ANALYSIS OF STRUCTURES

RED BOX RULES ARE FOR PROOF STAGE ONLY. DELETE BEFORE FINAL PRINTING. ANALYSIS OF STRUCTURES Eisley Waas An Introduction Including Numerical Methods University of Michigan, USA --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Analysis of Structures: An Introduction Including Numerical Methods offers an original way of introducing engineering students to the subject of stress and deformation analysis of solid objects, and helps them become more familiar with how numerical methods such as the finite element method are used in industry. Eisley and Waas secure for the reader a thorough understanding of the basic numerical skills and insight into interpreting the results these methods can generate. Throughout the text, they include analytical development alongside the computational equivalent, providing the student with the understanding that is necessary to interpret and use the solutions that are obtained using software based on the finite element method. They then extend these methods to the analysis of solid and structural components that are used in modern aerospace, mechanical and civil engineering applications. Analysis of Structures: An Introduction Including Numerical Methods is accompanied by a website (www.wiley.com/go/waas) housing exercises and examples that use modern software which generates color contour plots of deformation and internal stress. It offers invaluable guidance and understanding to senior level and graduate students studying courses in stress and deformation analysis as part of aerospace, mechanical and civil engineering degrees as well as to practicing engineers who want to re-train or re-engineer their set of analysis tools for contemporary stress and deformation analysis of solids and structures. • Provides a fresh, practical perspective to the teaching of structural analysis using numerical methods for obtaining answers to real engineering applications • Proposes a new way of introducing students to the subject of stress and deformation analysis of solid objects that are used in a wide variety of contemporary engineering applications • Casts axial, torsional and bending deformations of thin walled objects in a framework that is closely amenable to the methods by which modern stress analysis software operates. Enjoyed this book? www.wiley.com/go/waas ANALYSIS OF STRUCTURES Joe G. Eisley and Anthony M. Waas ANALYSIS OF STRUCTURES An Introduction Including Numerical Methods Joe G. Eisley and Anthony M. Waas Why not tell others about it and write a review on your favourite online bookseller. Cover design by Dan Jubb Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come ANALYSIS OF STRUCTURES --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-FM Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come ANALYSIS OF STRUCTURES AN INTRODUCTION INCLUDING NUMERICAL METHODS --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Joe G. Eisley Anthony M. Waas College of Engineering University of Michigan, USA A John Wiley & Sons, Ltd., Publication Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come This edition first published 2011  C 2011 John Wiley & Sons, Ltd Registered office John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, United Kingdom For details of our global editorial offices, for customer services and for information about how to apply for permission to reuse the copyright material in this book please see our website at www.wiley.com. The right of the author to be identified as the author of this work has been asserted in accordance with the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by the UK Copyright, Designs and Patents Act 1988, without the prior permission of the publisher. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books. Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The publisher is not associated with any product or vendor mentioned in this book. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought. Library of Congress Cataloging-in-Publication Data --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Eisley, Joe G. Analysis of structures : an introduction including numerical methods / Joe G. Eisley, Anthony M. Waas. p. cm. Includes bibliographical references and index. ISBN 978-0-470-97762-0 (cloth) 1. Structural analysis (Engineering)–Mathematics. 2. Numerical analysis. I. Waas, Anthony M. II. Title. TA646.W33 2011 624.1′ 71–dc22 2011009723 A catalogue record for this book is available from the British Library. Print ISBN: 9780470977620 E-PDF ISBN: 9781119993285 O-book ISBN: 9781119993278 E-Pub ISBN: 9781119993544 Mobi ISBN: 9781119993551 Typeset in 9/11pt Times by Aptara Inc., New Delhi, India Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come We would like to dedicate this book to our families. To Marilyn, Paul and Susan —Joe To Dayamal, Dayani, Shehara and Michael —Tony --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come Contents About the Authors xiii Preface xv 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 Forces and Moments Introduction Units Forces in Mechanics of Materials Concentrated Forces Moment of a Concentrated Force Distributed Forces—Force and Moment Resultants Internal Forces and Stresses—Stress Resultants Restraint Forces and Restraint Force Resultants Summary and Conclusions 1 1 1 3 4 9 19 27 32 33 2 2.1 2.2 2.3 Static Equilibrium Introduction Free Body Diagrams Equilibrium—Concentrated Forces 2.3.1 Two Force Members and Pin Jointed Trusses 2.3.2 Slender Rigid Bars 2.3.3 Pulleys and Cables 2.3.4 Springs Equilibrium—Distributed Forces Equilibrium in Three Dimensions Equilibrium—Internal Forces and Stresses 2.6.1 Equilibrium of Internal Forces in Three Dimensions 2.6.2 Equilibrium in Two Dimensions—Plane Stress 2.6.3 Equilibrium in One Dimension—Uniaxial Stress Summary and Conclusions 35 35 35 38 38 44 49 52 55 59 62 65 69 70 70 Displacement, Strain, and Material Properties Introduction Displacement and Strain 3.2.1 Displacement 3.2.2 Strain Compatibility 71 71 71 72 72 76 2.4 2.5 2.6 2.7 3 3.1 3.2 3.3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come viii 3.4 Contents 3.9 3.10 Linear Material Properties 3.4.1 Hooke’s Law in One Dimension—Tension 3.4.2 Poisson’s Ratio 3.4.3 Hooke’s Law in One Dimension—Shear in Isotropic Materials 3.4.4 Hooke’s Law in Two Dimensions for Isotropic Materials 3.4.5 Generalized Hooke’s Law for Isotropic Materials Some Simple Solutions for Stress, Strain, and Displacement Thermal Strain Engineering Materials Fiber Reinforced Composite Laminates 3.8.1 Hooke’s Law in Two Dimensions for a FRP Lamina 3.8.2 Properties of Unidirectional Lamina Plan for the Following Chapters Summary and Conclusions 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11 4.12 Classical Analysis of the Axially Loaded Slender Bar Introduction Solutions from the Theory of Elasticity Derivation and Solution of the Governing Equations The Statically Determinate Case The Statically Indeterminate Case Variable Cross Sections Thermal Stress and Strain in an Axially Loaded Bar Shearing Stress in an Axially Loaded Bar Design of Axially Loaded Bars Analysis and Design of Pin Jointed Trusses Work and Energy—Castigliano’s Second Theorem Summary and Conclusions 99 99 99 109 116 129 136 142 143 145 149 153 162 5 5.1 5.2 5.3 5.4 5.5 5.6 5.7 A General Method for the Axially Loaded Slender Bar Introduction Nodes, Elements, Shape Functions, and the Element Stiffness Matrix The Assembled Global Equations and Their Solution A General Method—Distributed Applied Loads Variable Cross Sections Analysis and Design of Pin-jointed Trusses Summary and Conclusions 165 165 165 169 182 196 202 211 6 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 Torsion Introduction Torsional Displacement, Strain, and Stress Derivation and Solution of the Governing Equations Solutions from the Theory of Elasticity Torsional Stress in Thin Walled Cross Sections Work and Energy—Torsional Stiffness in a Thin Walled Tube Torsional Stress and Stiffness in Multicell Sections Torsional Stress and Displacement in Thin Walled Open Sections 213 213 213 216 225 229 231 239 242 3.5 3.6 3.7 3.8 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 77 77 81 82 83 84 85 89 90 90 91 94 96 98 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come ix Contents 6.9 6.10 6.11 A General (Finite Element) Method Continuously Variable Cross Sections Summary and Conclusions 245 254 255 7 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10 7.11 7.12 Classical Analysis of the Bending of Beams Introduction Area Properties—Sign Conventions 7.2.1 Area Properties 7.2.2 Sign Conventions Derivation and Solution of the Governing Equations The Statically Determinate Case Work and Energy—Castigliano’s Second Theorem The Statically Indeterminate Case Solutions from the Theory of Elasticity Variable Cross Sections Shear Stress in Non Rectangular Cross Sections—Thin Walled Cross Sections Design of Beams Large Displacements Summary and Conclusions 257 257 257 257 259 260 271 278 281 290 300 302 309 313 314 8 8.1 8.2 8.3 8.4 8.5 8.6 A General Method (FEM) for the Bending of Beams Introduction Nodes, Elements, Shape Functions, and the Element Stiffness Matrix The Global Equations and their Solution Distributed Loads in FEM Variable Cross Sections Summary and Conclusions 315 315 315 320 327 341 345 9 9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9 9.10 9.11 More about Stress and Strain, and Material Properties Introduction Transformation of Stress in Two Dimensions Principal Axes and Principal Stresses in Two Dimensions Transformation of Strain in Two Dimensions Strain Rosettes Stress Transformation and Principal Stresses in Three Dimensions Allowable and Ultimate Stress, and Factors of Safety Fatigue Creep Orthotropic Materials—Composites Summary and Conclusions 347 347 347 350 354 356 358 361 363 364 365 366 10 10.1 10.2 Combined Loadings on Slender Bars—Thin Walled Cross Sections Introduction Review and Summary of Slender Bar Equations 10.2.1 Axial Loading 10.2.2 Torsional Loading 10.2.3 Bending in One Plane Axial and Torsional Loads Axial and Bending Loads—2D Frames 367 367 367 367 369 370 372 375 10.3 10.4 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come x 10.5 Contents Bending in Two Planes 10.5.1 When Iyz is Equal to Zero 10.5.2 When Iyz is Not Equal to Zero Bending and Torsion in Thin Walled Open Sections—Shear Center Bending and Torsion in Thin Walled Closed Sections—Shear Center Stiffened Thin Walled Beams Summary and Conclusions 384 384 386 393 399 405 416 11.4 11.5 11.6 11.7 Work and Energy Methods—Virtual Work Introduction Introduction to the Principle of Virtual Work Static Analysis of Slender Bars by Virtual Work 11.3.1 Axially Loading 11.3.2 Torsional Loading 11.3.3 Beams in Bending 11.3.4 Combined Axial, Torsional, and Bending Behavior Static Analysis of 3D and 2D Solids by Virtual Work The Element Stiffness Matrix for Plane Stress The Element Stiffness Matrix for 3D Solids Summary and Conclusions 417 417 417 421 421 426 427 430 430 433 436 437 12 12.1 12.2 12.3 12.4 12.5 12.6 12.7 12.8 12.9 Structural Analysis in Two and Three Dimensions Introduction The Governing Equations in Two Dimensions—Plane Stress Finite Elements and the Stiffness Matrix for Plane Stress Thin Flat Plates—Classical Analysis Thin Flat Plates—FEM Analysis Shell Structures Stiffened Shell Structures Three Dimensional Structures—Classical and FEM Analysis Summary and Conclusions 439 439 440 445 452 455 459 466 470 477 13 13.1 13.2 13.3 13.4 13.5 13.6 13.7 13.8 Analysis of Thin Laminated Composite Material Structures Introduction to Classical Lamination Theory Strain Displacement Equations for Laminates Stress-Strain Relations for a Single Lamina Stress Resultants for Laminates CLT Constitutive Description Determining Laminae Stress/Strains Laminated Plates Subject to Transverse Loads Summary and Conclusion 479 479 480 482 486 489 492 493 498 14 14.1 14.2 14.3 14.4 14.5 14.6 Buckling Introduction The Equations for a Beam with Combined Lateral and Axial Loading Buckling of a Column The Beam Column The Finite Element Method for Bending and Buckling Buckling of Frames 499 499 499 504 512 515 524 10.6 10.7 10.8 10.9 11 11.1 11.2 11.3 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-FM P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come xi Contents 14.7 14.8 Buckling of Thin Plates and Other Structures Summary and Conclusions 524 527 15 15.1 15.2 Structural Dynamics Introduction Dynamics of Mass/Spring Systems 15.2.1 Free Motion 15.2.2 Forced Motion—Resonance 15.2.3 Forced Motion—Response Axial Vibration of a Slender Bar 15.3.1 Solutions Based on the Differential Equation 15.3.2 Solutions Based on FEM Torsional Vibration 15.4.1 Torsional Mass/Spring Systems 15.4.2 Distributed Torsional Systems Vibration of Beams in Bending 15.5.1 Solutions of the Differential Equation 15.5.2 Solutions Based on FEM The Finite Element Method for all Elastic Structures Addition of Damping Summary and Conclusions 529 529 529 529 540 547 548 548 560 567 567 568 569 569 574 577 577 582 16.3 16.4 16.5 Evolution in the (Intelligent) Design and Analysis of Structural Members Introduction Evolution of a Truss Member 16.2.1 Step 1. Slender Bar Analysis 16.2.2 Step 2. Rectangular Bar—Plane Stress FEM 16.2.3 Step 3. Rectangular Bar with Pin Holes—Plane Stress Analysis 16.2.4 Step 4. Rectangular Bar with Pin Holes—Solid Body Analysis 16.2.5 Step 5. Add Material Around the Hole—Solid Element Analysis 16.2.6 Step 6. Bosses Added—Solid Element Analysis 16.2.7 Step 7. Reducing the Weight—Solid Element Analysis 16.2.8 Step 8. Buckling Analysis Evolution of a Plate with a Hole—Plane Stress Materials in Design Summary and Conclusions 583 583 584 584 585 586 587 588 590 591 592 592 594 594 A A.1 A.2 A.3 A.4 A.5 A.6 Matrix Definitions and Operations Introduction Matrix Definitions Matrix Algebra Partitioned Matrices Differentiating and Integrating a Matrix Summary of Useful Matrix Relations 595 595 595 597 598 598 599 B B.1 B.2 B.3 B.4 Area Properties of Cross Sections Introduction Centroids of Cross Sections Area Moments and Product of Inertia Properties of Common Cross Sections 601 601 601 603 609 15.3 15.4 15.5 15.6 15.7 15.8 16 16.1 16.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-FM P2: ABC JWST071-Waas July 4, 2011 11:3 Printer Name: Yet to Come xii Contents C C.1 C.2 C.3 C.4 C.5 Solving Sets of Linear Algebraic Equations with Mathematica Introduction Systems of Linear Algebraic Equations Solving Numerical Equations in Mathematica Solving Symbolic Equations in Mathematica Matrix Multiplication 611 611 611 611 612 613 D D.1 D.2 D.3 Orthogonality of Normal Modes Introduction Proof of Orthogonality for Discrete Systems Proof of Orthogonality for Continuous Systems 615 615 615 616 References 617 Index 619 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-babout JWST071-Waas July 6, 2011 9:48 Printer Name: Yet to Come About the Authors Joe G. Eisley received degrees from St. Louis University, BS (1951), and the California Institute of Technology, MS (1952), PhD (1956), all in the field of aeronautical engineering. He served on the faculty of the Department of Aerospace Engineering from 1956 to 1998 and retired as Emeritus Professor of Aerospace Engineering in 1998. His primary field of teaching and research has been in structural analysis with an emphasis on the dynamics of structures. He also taught courses in space systems design and computer aided design. After retirement he has continued some part time work in teaching and consulting. Anthony M. Waas is the Felix Pawlowski Collegiate Professor of Aerospace Engineering and Professor of Mechanical Engineering, and Director, Composite Structures Laboratory at the University of Michigan. He received his degrees from Imperial College, Univ. of London, U.K., B.Sc. (first class honors, 1982), and the California Institute of Technology, MS (1983), PhD (1988) all in Aeronautics. He joined the University of Michigan in January 1988 as an Assistant Professor, and is currently the Felix Pawlowski Collegiate Professor. His current teaching and research interests are related to lightweight composite aerostructures, with a focus on manufacturability and damage tolerance, ceramic matrix composites for “hot” structures, nano-composites, and multi-material structures. Several of his projects have been funded by numerous US government agencies and industry. In addition, he has been a consultant to several industries in various capacities. At Michigan, he has served as the Aerospace Engineering Department Graduate Program Chair (1998–2002) and the Associate Chairperson of the Department (2003–2005). He is currently a member of the Executive Committee of the College of Engineering. He is author or co-author of more than 175 refereed journal papers, and numerous conference papers and presentations. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-babout JWST071-Waas July 6, 2011 9:48 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-Preface JWST071-Waas July 2, 2011 14:26 Printer Name: Yet to Come Preface This textbook is intended to be an introductory text on the mechanics of solids. The authors have targeted an audience that usually would go on to obtain undergraduate degrees in aerospace and mechanical engineering. As such, some specialized topics that are of importance to aerospace engineers are given more coverage. The material presented assumes only a background in introductory physics and calculus. The presentation departs from standard practice in a fundamental way. Most introductory texts on this subject take an approach not unlike that adopted by Timoshenko, in his 1930 Strength of Materials books, that is, by primarily formulating problems in terms of forces. This places an emphasis on statically determinate solid bodies, that is, those bodies for which the restraint forces and moments, and internal forces and moments, can be determined completely by the equations of static equilibrium. Displacements are then introduced in a specialized way, often only at a point, when necessary to solve the few statically indeterminate problems that are included. Only late in these texts are distributed displacements even mentioned. Here, we introduce and formulate the equations in terms of distributed displacements from the beginning. The question of whether the problems are statically determinate or indeterminate becomes less important. It will appear to some that more time is spent on the slender bar with axial loads than that particular structure deserves. The reason is that classical methods of solving the differential equations and the connection to the rational development of the finite element method can be easily shown with a minimum of explanation using the axially loaded slender bar. Subsequently, the development and solution of the equations for more advanced structures is facilitated in later chapters. Modern advanced analysis of the integrity of solid bodies under external loads is largely displacement based. Once displacements are known the strains, stresses, strain energies, and restraint reactions are easily found. Modern analysis solutions methods also are largely carried out using a computer. The direction of this presentation is first to provide an understanding of the behavior of solid bodies under load and second to prepare the student for modern advanced courses in which computer based methods are the norm. Analysis of Structures: An Introduction Including Numerical Methods is accompanied by a website (www.wiley.com/go/waas) housing exercises and examples that use modern software which generates color contour plots of deformation and internal stress. It offers invaluable guidance and understanding to senior level and graduate students studying courses in stress and deformation analysis as part of aerospace, mechanical and civil engineering degrees as well as to practicing engineers who want to re-train or re-engineer their set of analysis tools for contemporary stress and deformation analysis of solids and structures. We are grateful to Dianyun Zhang, Ph.D candidate in Aerospace Engineering, for her careful reading of the examples presented. Corrections, comments, and criticisms are welcomed. Joe G. Eisley Anthony M. Waas June 2011 Ann Arbor, Michigan --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-Preface JWST071-Waas July 2, 2011 14:26 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 1 Forces and Moments 1.1 Introduction Mechanics of solids is concerned with the analysis and design of solid bodies under the action of applied forces in order to ensure “acceptable” behavior. These solid bodies are the components and the assemblies of components that make up the structures of aircraft, automobiles, washing machines, golf clubs, roller blades, buildings, bridges, and so on, that is, of many manufactured and constructed products. If the solid body is suitably restrained to exclude “rigid body” motion it will deform when acted upon by applied forces, or loads, and internal forces will be generated in the body. For “acceptable” behavior: 1. Internal forces must not exceed values that the materials can withstand. 2. Deformations must not exceed certain limits. In later chapters of this text we shall identify, define, and examine the various quantities, such as internal forces, stresses, deformations, and material stress-strain relations, which determine acceptable behavior. We shall study methods for analyzing solid bodies and structures when loaded and briefly study ways to design solid bodies to achieve a desired behavior. All solid bodies are three dimensional objects and there is a general theory of mechanics of solids in three dimensions. Because understanding the behavior of three dimensional objects can be difficult and sometimes confusing we shall work primarily with objects that have simplified geometry, simplified applied forces, and simplified restraints. This enables us to concentrate on the process instead of the details. After we have a clear understanding of the process we shall consider ever increasing complexity in geometry, loading, and restraint. In this introductory chapter we examine three categories of force. First are applied forces which act on the surface or the mass of the body. Next are restraint forces, that is, forces on the surfaces where displacement is constricted (or restrained). Thirdly, internal forces generated by the resistance of the material to deformation as a result of applied and restraint forces. Forces can generate moments acting about some point. For the most part we carefully distinguish between forces and moments; however, it is common practice to include both forces and moments when referring in general terms to the forces acting on the body or the forces at the restraints. 1.2 Units Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The basic quantities in the study of solid mechanics are length (L), mass (M), force (F), and time (t). To these we must assign appropriate units. Because of their prominent use in every day life in the P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 2 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods United States, the so-called English system of units is still the most familiar to many of us. Some engineering is still done in English units; however, global markets insist upon a world standard and so a version of the International Standard or SI system (from the French Système International d’Unités) prevails. The standard in SI is the meter, m, for length, the Newton, N, for force, the kilogram, kg, for mass, and the second, s, for time. The Newton is defined in terms of mass and acceleration as 1N = 1 kg · 1 m s2 (1.2.1) For future reference the acceleration due to gravity on the earth’s surface, g, in metric units is g = 9.81 m s2 (1.2.2) The standard English units are the foot, ft, for length, the pound, lb, for force, the slug, slug, for mass, and the second, s, for time. The pound is defined in terms of mass and acceleration as 1 lb = 1 slug · 1 ft s2 (1.2.3) The acceleration due to gravity on the earth’s surface, g, in English units is g = 32.2 ft s2 (1.2.4) We shall use SI units as much as possible. Most of you are still thinking in English units and so for quick estimates you can note that a meter is approximately 39.37 inches; there are approximately 4.45 Newtons in a pound; and there are approximately 14.59 kilograms in a slug. But since you are not used to thinking in slugs it may help to note that a kilogram of mass weighs about 2.2 pounds on the earth’s surface. For those who must convert between units there are precise tables for conversion. In time you will begin to think in SI units. Often we obtain quantities that are either very large or very small and so units such as millimeter are defined. One millimeter is one thousandth of a meter, or 1 mm = 0.001 m, and, of course, one kilogram is one thousand grams, or 1 kg = 1000 g. The following table lists the prefixes for different multiples: Multiple Prefix Symbol 109 106 103 10−3 10−6 10−9 giga mega kilo milli micro nano G M k m μ n One modification of SI is that it is common practice in much of engineering to use the millimeter, mm, as the unit of length. Thus force per unit length is often, perhaps usually, given as Newtons per millimeter or N/mm. Force per unit area is given as Newtons per millimeter squared or N/mm2 . One N/m2 is called a Pascal or Pa, so the unit of 1 N/mm2 is called 1 mega Pascal or 1 MPa. Mass density has the units of kilograms per cubic millimeter or kg/mm3 . Throughout we shall use millimeter, Newton, and kilogram in all examples, discussions, and problems. As noted in the above table: Only multiples of powers of three are normally used; thus, we do not use, for example, centimeters, decimeters, or other multiples that are the power of one or two. These are conventions, of course, so in the workplace you will find a variety of practices. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 3 Forces and Moments 1.3 Forces in Mechanics of Materials There are several types of forces that act on solid bodies. These consist of forces applied to the mass of the body and to the surface of the body, forces at restraints, and internal forces. In Figure 1.3.1 we show a general three dimensional body with forces depicted acting on its surface and on its mass. line force N mm body force N mm3 surface forces concentrated force N N mm 2 Figure 1.3.1 Forces that are volume or mass related are called body forces. In the system of units we are using they have the units of Newtons per cubic millimeter (N/mm3 ). Gravity forces are a good example. Inertia forces generated by accelerations are another. Surface forces can be specified in terms of force per unit area distributed over a surface and have the units of Newtons per square millimeter (N/mm2 ). As noted one Newton per square millimeter is also called one mega Pascal (MPa). If a force is distributed along a narrow band it is specified as a line force, that is, a force per unit length or Newtons per millimeter (N/mm). If the force acts at a point it is a concentrated force and has the units of Newtons (N). Concentrated forces and line forces are usually idealizations or resultants of distributed surface forces. We can imagine an ice pick pushing on a surface creating a concentrated force. More likely the actual force acts on a small surface area where small means the size of the area is very small compared to other characteristic dimensions of the surface. Likewise a line force may be the resultant of a narrow band of surface forces. When a concentrated, line, surface, or body force acts on the solid body or is applied to the body by means of an external agent it is called an applied force. When the concentrated, line, or surface force is generated at a point or region where an external displacement is imposed it is called a restraint force. In addition, for any body that is loaded and restrained, a force per unit area can be found on any internal surface. This particular distributed force is referred to as internal or simply as stress. Generally, in the initial formulation of a problem for analysis, the geometry, applied forces, and physical restraints (displacements on specified surfaces) are known while the restraint forces and internal stresses are unknown. When the problem is formulated for design, the acceptable stress limits may be specified in advance and the final geometry, applied forces, and restraints may initially be unknown. For the most part the problems will be formulated for analysis but the subject of design will be introduced from time to time. The analysis of the interaction of these various forces is a major part of the following chapters. For the most part we shall use rectangular Cartesian coordinates and resolve forces into components with respect to these axes. An exception is made for the study of torsion in Chapter 6. There we use cylindrical coordinates. --`,,,,``,``,,` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 4 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In the sign convention adopted here, applied force components and restraint force components are positive if acting in the positive direction of the coordinate axes. Positive stresses and internal forces will be defined in different ways as needed. We start first with a discussion of concentrated forces. 1.4 Concentrated Forces As noted, concentrated forces are usually idealizations of distributed forces. Because of the wide utility of this idealization we shall first examine the behavior of concentrated forces. In all examples we shall use the Newton (N) as our unit of force. Force is a vector quantity, that is, it has both magnitude and direction. There are several ways of representing a concentrated force in text and in equations; however, the pervasive use of the digital computer in solving problems has standardized how forces are usually represented in formulating and solving problems in the behavior of solid bodies under load. First, we shall consider a force that can be oriented in a two dimensional right handed rectangular Cartesian coordinate system and we shall define positive unit vectors i and j in the x, and y directions, respectively, as shown in Figure 1.4.1. Using boldface has been a common practice in representing vectors in publications. y j x i Figure 1.4.1 A force is often shown in diagrams as a line that starts at the point of application and has an arrowhead to show its direction as shown in Figure 1.4.2. y F Fyj θ x Fxi Figure 1.4.2 The concentrated force, F, can be represented by its components in the x and y directions. F = Fx i + Fy j (1.4.1) In keeping with the notation most commonly used for later computation we represent this force vector by a column matrix {F} as shown in Equation 1.4.2.   Fx {F} = (1.4.2) Fy --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 5 Forces and Moments In matrix notation the unit vector directions are implied by the component subscripts. From the properties of a right triangle the magnitude of the vector is given by  (1.4.3) F = Fx2 + Fy2 The orientation of the force can be represented by the angle between the force and either axis. For example, with respect to the x axis tan θ = Fy Fx θ = tan−1 → Fy Fx (1.4.4) Quite often we must sum two or more forces such as those shown in Figure 1.4.3 as solid lines. y F3 = F1 + F2 F3y F1 F2 x F3x Figure 1.4.3 To add or subtract vectors is simply to add or subtract components. For example,         F3x F1x + F2x F2x F1x {F3 } = {F1 } + {F2 } = = = + F3y F1y + F2y F2y F1y (1.4.5) The sum is shown by the dashed line and its components in the two coordinate directions by the dotted lines. ########### Example 1.4.1 Problem: Two forces are acting at a point at the origin of the coordinate system as shown in Figure (a). Sum the two to find the resultant force and its direction. y F1 = 85 N F2 = 100 N 45º 30º x Figure (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas 6 July 2, 2011 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Solution: Resolve the forces into components and sum. Solve for the resultant force and its orientation. The components are             86.6 100 cos 30◦ −60.1 −85 sin 45◦ F2x F1x {F } {F1 } = N (a) = = = = = 2 100 sin 30◦ F2y 85 cos 45◦ F1y 50 60.1 The sum is {F3 } = {F1 } + {F2 } =       −60.1 86.6 26.5 + = N 60.1 50 110.1 The total magnitude of the force is   F3 = Fx2 + Fy2 = (26.5)2 + (110.1)2 = 113.2 N (b) (c) The resultant force vector makes an angle with respect to the x axis, θ = tan−1 Fy 110.1 = tan−1 4.15 = 76.5◦ = tan−1 Fx 26.5 (d) The resultant force is shown as a dashed line and its components as dotted lines in Figure (b). y F3 = 113.2 N F1 = 85 N 45o F2 = 100 N 30o x Figure (b) ########### This can be extended to three dimensions. We shall define positive unit vectors i, j, k in the x, y, z directions, respectively, as shown in Figure 1.4.4. y j k x i z Figure 1.4.4 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 7 Forces and Moments The concentrated force, F, can be represented by its components in the x, y, and z directions as in Equation 1.4.6. F = Fx i + Fy j + Fz k (1.4.6) This is shown graphically in Figure 1.4.5. y Fy j F Fx i x Fz k z Figure 1.4.5 The components in matrix form are ⎡ ⎤ Fx {F} = ⎣ Fy ⎦ Fz (1.4.7) The magnitude of the vector F is given by F=  Fx2 + Fy2 + Fz2 (1.4.8) The angular orientation of the force F with respect to each axis is given by cos α = Fx F cos β = Fy F cos γ = Fz F (1.4.9) The angle between the force, F, and the x axis is α, between the force F and the y axis is β, and between the force F and the z axis is γ. The quantities in Equation 1.4.9 are called the direction cosines. As noted in the two dimensional case, to add or subtract vectors is simply to add or subtract components. For example, given three forces acting at a point the force representing the sum is ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ F1x F2x F3x F1x + F2x − F3x F4x {F4 } = {F1 } + {F2 } − {F3 } = ⎣ F1y ⎦ + ⎣ F2y ⎦ − ⎣ F3y ⎦ = ⎣ F1y + F2y − F3y ⎦ = ⎣ F4y ⎦ F1z F2z F3z F1z + F2z − F3z F4z (1.4.10) ########### Example 1.4.2 Problem: Two forces act in perpendicular planes as shown in Figure (a). Sum the two to find the resultant force and its direction. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 8 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y F1 = 200 N F2 = 98 N 45o 60o x z Figure (a) Solution: Resolve the forces into components and sum. Solve for the value of the resultant force and its orientation. The components of the forces are ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ F1x 0 0 {F1 } = ⎣ F1y ⎦ = ⎣ 200 sin 60◦ ⎦ = ⎣ 173.2 ⎦ N F1z 200 cos 60◦ 100 ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ F2x 98 cos 45◦ 69.3 {F2 } = ⎣ F2y ⎦ = ⎣ 98 sin 45◦ ⎦ = ⎣ 69.3 ⎦ N F2z 0 0 (a) The sum of the forces is ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ 0 69.3 69.3 {F} = {F1 } + {F2 } = ⎣ 173.2 ⎦ + ⎣ 69.3 ⎦ = ⎣ 242.5 ⎦ N 100 0 100 (b) The magnitude of the total force is F=  Fx2 + Fy2 + Fz2 =  (69.3)2 + (242.5)2 + (100)2 = 271.3 N (c) The direction cosines are 69.3 Fx = = 0.255 F 271.3 242.5 Fy = = 0.662 cos β = F 271.3 100 Fz = = 0.369 cos γ = F 271.3 cos α = → α = 75.2◦ → β = 26.6◦ → γ = 68.4◦ The final result is shown in Figure (b). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 9 Forces and Moments y F = 271.3 N F1 = 200 N F2 = 98 N 45o 60o x z Figure (b) ########### Another property of a matrix that we shall use shortly is multiplication of a matrix by a scalar. It is simply ⎤ ⎡ ⎤ ⎡ a Fx Fx (1.4.11) a {F} = a ⎣ Fy ⎦ = ⎣ a Fy ⎦ Fz a Fz Additional matrix operations will be introduced as needed. They are summarized in Appendix A. 1.5 Moment of a Concentrated Force A concentrated force can produce a moment about any given axis. In all examples we shall use Newton millimeter (N · mm) as our unit for moments. Consider the force applied to the rigid bar at point B as shown in Figure 1.5.1. F y A B x L Figure 1.5.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 10 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods If we take moments about points A and B we get MA = FL MB = M − FL = FL − FL = 0 (1.5.1) Now consider the force has been moved to point A and a concentrated moment equal to FL is added at point A as shown in Figure 1.5.2. y F A M = FL x B L Figure 1.5.2 The moments about points A and B in this new configuration are the same as for the first configuration. Summing moments about each point we get MA = F · 0 + M = M = FL MB = M − FL = FL − FL = 0 (1.5.2) We can, in fact, take moments about any point in the xy plane and get the same result for both configurations. For example, take moments about the point C as shown in Figure 1.5.3 located at L 2 xC = L 4 C y A yC = (1.5.3) F L/4 L/2 B x L Figure 1.5.3 From the configuration in Figure 1.5.3 we get MC = F L 2 (1.5.4) From the configuration in Figure 1.5.2 we get MC = M − F L L L = FL − F = F 2 2 2 (1.5.5) The problem can be posed in another way: If you move a force, what moment must be added to achieve an equivalent balance of moments? --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 11 Forces and Moments Let us consider the rigid bar in Figure 1.5.4 with the force initially at point B. We shall call this configuration 1. It is then moved to point C (shown by a dashed line). This we call configuration 2. What moment must be added (and at what location) to provide equivalence? F y L/2 A C x B L Figure 1.5.4 Let us take moments about point A for configuration 1 and 2. MA1 = F · L MA2 = F · L 2 (1.5.6) For equivalence we must add a moment that is equal to the difference in the two values or M = MA1 − MA2 = F · L − F · L L =F· 2 2 (1.5.7) Now where should it be added? The answer is anywhere. In this case anywhere along the bar, for example, at point A, or point C, or point B, or any point in between. y --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- A F L/2 C x B M = FL/2 L Figure 1.5.5 Just to be sure let us place the new applied moment at point A and sum moments about points A, B, and C in Figure 1.5.5. MA = FL FL + = FL 2 2 MB = − FL FL + =0 2 2 MC = FL 2 (1.5.8) If you compare this with the original configuration in Figure 1.5.1 you will see that the moments about points A, B, and C agree. The use of the half circle symbol in Figures 1.5.2 and 1.5.5 is one way of representing a concentrated moment in diagrams. It is used when the moment is about an axis perpendicular to the plane of the page. A common practice is to use a vector with a double arrowhead shown here in an isometric view to represent a moment. The vector is parallel to the axis about which the moment acts. The right hand rule of the thumb pointed in the vector direction and the curve fingers of the right hand showing the direction of the moment is implied here. The moment of Figure 1.5.2 is repeated in Figure 1.5.6 using a double arrowhead notation. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 12 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y F M = FL z --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- x Figure 1.5.6 ########### Example 1.5.1 Problem: A force is applied to a rigid body at point A as shown in Figure (a). If the force is moved to point B what moment must be applied at point C (origin of coordinates) to produce the same net moment about all points in space? y C 500 mm F z A B 100 mm 250 mm x Figure (a) Solution: Find the moment components at point C due to the force at point B and add the necessary moments so the total is equivalent to the moment components generated by the force at point A. The force at point A produces the following moment components about the origin (point C) MCAx = 0 MCAy = 0 MCAz = 500 · F This can be written in matrix form as a column vector. ⎡ ⎤ 0 {MCA } = F ⎣ 0 ⎦ 500 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) (b) P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 13 Forces and Moments The same force at point B would produce the following moment components at point C. MCBx = −250 · F MCBy = 0 MCBz = 400 · F (c) or ⎡ ⎤ −250 {MCB } = F ⎣ 0 ⎦ 400 (d) The needed moment components would be ⎡ ⎡ ⎤ ⎤ ⎡ ⎤ ⎡ ⎤ 0 MCx −250 250 {MC } = ⎣ MCy ⎦ = {MCA } − {MCB } = F ⎣ 0 ⎦ − F ⎣ 0 ⎦ = F ⎣ 0 ⎦ MCz 500 400 100 (e) This is illustrated in Figure (b). MCx = 250F Nmm y C 500 mm MCz = 100F Nmm F z A B 100 mm --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 250 mm x Figure (b) Let us check our answer by summing moments about the origin. Using Figure (b) Mx = 250F − 250F = 0 My = 0 Mz = 400F + 100F = 500F In matrix form we get the same answer as Equation (a) ⎡ ⎤ 0 {M} = F ⎣ 0 ⎦ 500 (f) (g) Summing moments about any point in space will prove that the answer is always the same. ########### So far we have considered a single force parallel to one of the axes. Consider now a force with components in all three dimensions acting at a point in space. We select a point about which we wish to find the moment components of this force with respect to a set of rectangular Cartesian coordinate axes. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- y Fy Fx x yo Fz zo xo z Figure 1.5.7 In Figure 1.5.7 we show the components of a force {F} and the axes about which we wish to find the moment. If we take the moment at the origin of the coordinate system about each axis in turn we get Mx = Fz yo − Fy z o M y = Fx z o − Fz xo Mz = Fy xo − Fx yo The moment components in matrix form about all three axes are ⎡ ⎤ ⎡ ⎤ Mx Fz yo − Fy z o {M} = ⎣ M y ⎦ = ⎣ Fx z o − Fz xo ⎦ Mz Fy xo − Fx yo (1.5.9) (1.5.10) This same information is often presented in the form of vector notation. The cross product of two vectors is stated as C=A×B (1.5.11) C = AB cos θ (1.5.12) The magnitude of the cross product is where θ is the angle between the two vectors. The vector C has a direction that is perpendicular to the plane containing A and B and its direction is defined by the right hand rule. Since the angle between unit ◦ ◦ ◦ vectors is either 90 , −90 , or 0 the cross product of unit vectors is found to be i×j=k j×k=i k×i=j i × k = −j j × i = −k k × j = −i i×i=0 j×j=0 k×k=0 (1.5.13) With this definition in mind the moment of the force in Figure 1.5.7 is represented as the cross product of a position vector and the force. Thus M=d×F (1.5.14) If the moment is taken with respect to the origin of the coordinates in Figure 1.5.7 the position vector is d = x o i + yo j + z o k Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (1.5.15) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 15 Forces and Moments The moment is then M = d × F = (x o i + yo j + z o k) × (Fx i + Fy j + Fz k) = xo Fx (i × i) + xo Fy (i × j) + xo Fz (i × k) (1.5.16) + yo Fx (j × i) + yo Fy (j × j) + z o Fz (j × k) + z o Fx (k × i) + z o Fx (k × j) + z o Fx (k × k) By combining terms this reduces to M = d × F = (yo Fz − z o Fy )i + (xo Fz − z o Fx )j + (xo Fy − yo Fx )k (1.5.17) This is often presented in the form of a determinant. i M = xo Fx j yo Fy k zo Fz (1.5.18) Equations 1.5.17 and 1.5.18 convey exactly the same information as that contained in Equation 1.5.10. The matrix formulation of vectors very often has replaced the boldfaced vector representation more common in past treatises. We shall discontinue the use of boldface in representing vectors since we shall be using the matrix representation in all future work. The representation of any vector quantity will be clear from the context of its use. As noted before we denote a moment using a vector with a double arrowhead as shown in Figure 1.5.8. y Fy My Fx yo x Mz Mx Fz zo xo z Figure 1.5.8 The vector components can be combined to obtain the total value of the moment.  M = Mx2 + M y2 + Mz2 (1.5.19) and the orientation can be given by the direction cosines. cos α = Mx M cos β = My M cos γ = Mz M (1.5.20) In all our deliberations applied forces are positive if they are in the positive directions of the axes and applied moments resulting from applied forces are positive by the right hand rule. Right handed rectangular Cartesian coordinate systems are used for the most part. Cylindrical coordinates will be used when we study torsion in Chapter 6. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 16 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 1.5.2 Problem: Find the moment components of the force shown in Figure (a) about the origin of the coordinate axes and the total value of the moment. y  −20  {F } =  −80 N  34  300 mm --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 280 mm x 600 mm z Figure (a) Solution: Use Equations 1.5.18–20. The components are ⎡ ⎤ −20 {F} = ⎣ −80 ⎦ N 34 ⎤ ⎡ ⎤ 600 xo {d} = ⎣ yo ⎦ = ⎣ 300 ⎦ mm zo −280 ⎡ (a) The moment can be obtained from i j k M = 600 300 −280 −20 −80 34 → ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ Mx 34 · 300 − 80 · 280 −12200 {M} = ⎣ M y ⎦ = ⎣ −34 · 600 + 20 · 280 ⎦ = ⎣ −14800 ⎦ N · mm Mz −80 · 600 + 20 · 300 −42000 (b) The total value of the moment is M=  Mx2 + M y2 + Mz2 =  (12200)2 + (14800)2 + (42000)2 = 46172.3 N mm (c) The orientation of the resultant moment is given by the direction cosines. cos α = −12200 Mx = = −0.2642 → α = 105.32◦ M 46172.3 −14800 My = = −0.3205 → β = 108.7◦ cos β = M 46172.3 −42000 Mz = = −0.9096 → γ = 155.46◦ cos γ = M 46172.3 ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 17 Forces and Moments To find the moment about some point other than the origin of the coordinate system requires only defining a new position vector. For example suppose we wish to find the moment components about point A as shown in Figure 1.5.9. y Fy Fx x yo d Fz zo A xo z Figure 1.5.9 The components of the new position vector would be ⎡ ⎤ xo {d} = ⎣ yo ⎦ 0 (1.5.21) A special case of a moment caused by forces occurs when there are two parallel forces of equal and opposite direction separated by a distance a. This might occur, for example, with the loads applied to a member as shown in Figure 1.5.10. y B L/2 3a/2 F A x a F L Figure 1.5.10 The resultant force of the two forces is zero. The resultant moment about any point in the plane is M = Fa (1.5.22) To illustrate that the location of the point in the plane is of no effect take moments about point A which is at the origin of the coordinates and about point B which is at x = L/2 and y = 3d/2. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 2, 2011 18 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods MA = F a a + F = Fa 2 2 MB = − a 3a − 2 2 (1.5.23) F+ a 3a + 2 2 F = Fa Such a force combination as shown in Figure 1.5.10 is called a couple. When d is small it can often be represented as a concentrated moment as in Figure 1.5.11. M=Fa A x Figure 1.5.11 ########### Example 1.5.3 Problem: Find the moment components about the axes of the set of couples shown in Figure (a). y z 2F 500 mm 300 mm 600 mm F F 2F x Figure (a) Solution: Use the definition of a couple. The moment components about the axes are simply Mx = 500 · 2F M y = −500 · F Mz = 0 (a) Example 1.5.4 Problem: A force system consists of a couple and another force as shown in Figure (a). Find the moment about the z axis at point A. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-01 P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 19 Forces and Moments Solution: Sum moments about the z axis. y 3F L/ 2 M=Fa A x Figure (a) The sum of moments about point A is MA = 3F · L L + M = 3F · + Fa = 2 2 3L +a F 2 (a) ########### 1.6 Distributed Forces—Force and Moment Resultants Forces may be distributed along a line, on a surface, or throughout a volume. It is often necessary to find the total force resultant of the distributed force and also find through what point it is acting. Consider the area shown in Figure 1.6.1 and the force per unit area acting upon it. We have chosen a planar rectangular area and a particular distribution for ease of explanation. Real surfaces with loads will be found in many shapes and sizes and can be external or internal surfaces. y fsy ( x) z L a/2 a/2 x Figure 1.6.1 This particular surface loading acts in the y direction, varies in the x direction, and is uniform in the z direction. We label this force f sy (x, z) where the subscript s denotes a surface force and the y its --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 20 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods component direction. We note that its units are force per unit area (N/mm2 ) and since it is uniform in the z direction we can multiply the surface force by the width, a, of the planar surface and obtain an equivalent line force with units of force per unit length (N/mm) as shown in Figure 1.6.2. This line force acts in the plane of symmetry, that is, in the xy plane at z = 0. Our coordinate system was selected conveniently with this in mind. Bear in mind that this is the resultant of the distributed surface force and not the actual force acting on the surface. It may be used for establishing equilibrium of the body. y fly ( x) dfly ( x) z x --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- L dx a/2 a/2 x Figure 1.6.2 We label this force fly (x) where the subscript l denotes a line force and note that fly (x) = a f sy (x, z) (1.6.1) Now consider an infinitesimal length, dx, along the line force at location x, as shown in Figure 1.6.2. On the length, dx, the force in the y direction is dF y = dfly (x) (1.6.2) If we sum the forces on all such dx lengths, ranging for x = 0 to x = L, we obtain the total resultant of the distributed force L Fy = 0 L fly (x)d x = a f sy (x)d x (1.6.3) 0 We can find the location, or line of action, of the force resultant by equating the moments of the distributed force to the moment of the force resultant as follows. Again, the sum of all the moments of all the forces on all the infinitesimal elements dx is L x̄ Fy = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS x fly (x)d x (1.6.4) 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 21 Forces and Moments And the location of the resultant is L x̄ = 1 Fy x fly (x)d x L 0 0 x fly (x)d x = (1.6.5) L fly (x) d x 0 The use of symmetry to locate the line force and the resultant force at z = 0 can be confirmed by equating the moments of the distributed force to the moment of the resultant force about the z axis. a 2 L z̄ Fy = 0 − a2 z f sy (x)dzd x → z̄ = 1 Fy a 2 L zdz f sy (x) d x − a2 (1.6.6) 0 Clearly the integral a 2 − a2 zdz = 0 (1.6.7) and therefore z̄ = 0. The line force and the resultant force and its location are depicted in Figure 1.6.3. y Fy x fly ( x) L 0 x Figure 1.6.3 ########### Example 1.6.1 Problem: A distributed force per unit area is applied to the surface as shown in Figure (a). The force is uniform in the z direction. x L Find the total value and its line of action. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- f sy (x, z) = f 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 22 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y z fsy ( x, z ) = f 0 x L L x b Figure (a) Solution: Convert the surface force (N/mm2 ) to a line force (N/mm) and integrate to find the total resultant force (N). Then equate moments to find its line of action. The distributed line force and resultant force are fly (x) = b f 0 L x L Fy = b f 0 0 x2 x d x = b f0 L 2L L 0 = b f0 L 2 (b) Find the line of action by equating moments. L x̄ Fy = xb f 0 0 x dx L → x̄ = 1 Fy L xb f 0 0 1 b f0 x 3 x dx = L Fy 3L L 0 = 2 L 3 (c) The line force and the location of the force resultant are shown in Figure (b). y Fy f0 fly(x) x x= L 2 L 3 Figure (b) ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 0 P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 23 Forces and Moments Symmetry may be used to identify quickly the location of a resultant force. The two line forces shown in Figure 1.6.4 are symmetrical about their midpoints and so the location of the resultant is known instantly as shown. F F fl fl Figure 1.6.4 Distributed surface forces may be functions of two variables; for example, as shown in Figure 1.6.5. y Fy z fsy ( x, z ) x z L a/2 a/2 x Figure 1.6.5 Then the resultant force for the area shown, using the same procedure as before, is a 2 Fy = − a2 L f sy (x, z)d xdz (1.6.6) 0 and its line of action is found by a 2 x̄ Fy = − a2 L 0 a 2 1 x f sy (x, z)d xdz → x̄ = Fy a 2 − a2 L 0 x f sy (x, z)d xdz = L x f sy (x, z)d xdz − a2 0 a 2 L f sy (x, z)d xdz − a2 0 a 2 a 2 z̄ Fy = − a2 L 0 z f sy (x, z)d xdz → z̄ = 1 Fy a 2 − a2 L 0 z f sy (x, z)d xdz = − a2 L z f sy (x, z)d xdz 0 a 2 − a2 L f sy (x, z)d xdz 0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (1.6.7) P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 24 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The location of the resultant force is also depicted in Figure 1.6.5. Of course, surfaces are not always planar and rectangular and surface forces may have components in all coordinate directions. We shall be satisfied with simplified geometry and forces until and unless the need arises for more complicated cases. ########### Example 1.6.2 Problem: A distributed force per unit area is applied to the surface shown in Figure (a). The surface force is represented by the function in Equation (a). Find the resultant force and its line of action. f sy (x, z) = f 0 1 − x2 a2 z 1 + 2 2b (a) y z x z a b x Figure (a) Solution: Integrate to find the resultant force. Equate moments to find its line of action. The resultant force is given by Equation (b) a Fy = 0 b 0 a f sy (x, z)dzd x = f 0 = f0 0 b 0 x− x2 a2 1− x3 3a 2 a 0 z 1 + dzd x 2 2b b z2 z + 2 4b 0 = f0 (b) ab 2 Its location is given by Equations (c) and (d). b a x̄ Fy = 0 0 b a x f sy (x, z)d xdz → x̄ = 1 Fy 2 = ab f0 0 0 x 1− x4 x2 − 2 2 4a a 0 x2 a2 z 1 dzd x + 2 2b z2 z + 2 4b b 0 = 3a 8 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 25 Forces and Moments b a z̄ Fy = z f sy (x, z)d xdz 0 0 z̄ = → = 1 f0 Fy 2 ab b a 0 0 x− x2 a2 z 1− x3 3a 2 a z 1 + dzd x 2 2b z3 z2 + 4 6b 0 b 0 = (d) 5b 9 ########### We may also need to find the resultants of body forces and their location. Body forces have the units of force per unit volume (N/mm3 ) and may be distributed over the volume. Consider the general solid shown in Figure 1.6.6. y z x y dV x W Figure 1.6.6 The most common body force is that due to gravity, that is, weight. The weight (dW) of a infinitesimal volume dV is given by dW = ρ(x, y, z)gdV (1.6.8) where ρ(x, y, z) is the mass density (kg/mm3 ) and g is the acceleration due to gravity (mm/s2 ). (See Section 1.2, Equation 1.2.2.) If the y axis is oriented parallel to the gravity vector then the total weight is given by W =g ρ (x, y, z) dV (1.6.9) V and its line of action, or, in this case, the point through which it acts is given by yρ (x, y, z) dV xρ (x, y, z) dV x̄ = V ρ (x, y, z) dV V ȳ = V ρ (x, y, z) dV V zρ (x, y, z) dV z̄ = V (1.6.10) ρ (x, y, z) dV V The location of this resultant force is called the center of gravity and is commonly abbreviated as C.G. Notice that the acceleration due to gravity is a constant that cancels out of the integrals for finding the C.G. and the resulting equations depended only upon the mass density and the volume. This is also called the center of mass. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- z P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 26 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In many situations the solid body is homogenous, that is, the mass density is a constant. In such cases the mass density terms also cancel. Equations 1.6.10 then become xdV x̄ = V dV V ydV ȳ = zdV V z̄ = dV V V (1.6.11) dV V This locates the centroid of the volume. In such cases the centroid, the center of mass, and the center of gravity are the same point. When a homogenous body can be divided into sub volumes with simple geometry so that the centroids of the sub volumes are known we can find the centroid of the total using the following formulas.    ys Vs z s Vs xs Vs ȳ =  z̄ =  (1.6.12) x̄ =  Vs Vs Vs ########### Example 1.6.3 Problem: A cylindrical bar has a portion hollowed out as shown in Figure (a). It is made of aluminum which has a mass density of 2.72 E-06 kg/mm3 . Find its total weight and center of gravity. The y axis is aligned with the gravity vector. y 15 mm x 5 mm 50 mm 100 mm Figure (a) Solution: The total weight is the volume times the mass density times the acceleration due to gravity. From axial symmetry we know the center of gravity will lie on the centerline of the cylinder. We find the x location by summing moments about the z axis. To find the total weight we find the weight of the outer cylinder and subtract the weight of the inner cylinder. πr 2 L 2 L − inner Wtotal = Wouter − Winner = ρg πrouter 2  = 2.72 · 10−6 · 9.81 π (15)2 100 − π (5)2 50 = (1.886 − 0.105) N = 1.781N Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The quantities xs , ys , and zs represent the distances from the base axes to the centroids of the sub volumes Vs . For a body with uniform mass density you can replace the volume in Equation 1.6.12 with the mass or the weight to find the center of mass or the center of gravity. All are at the same location. P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 27 Forces and Moments Using symmetry the center of gravity location is located on the centerline of the cylinder and at an x position given by x̄ Wtotal = 50Wouter − 75Winner = 50 · 1.886 − 75 · 0.105 = 86.425 → x̄ = 86.425 = 48.526 mm 1.781 ȳ = z̄ = 0 (b) ########### 1.7 Internal Forces and Stresses—Stress Resultants As we have said, when forces are applied to a solid body that is suitably restrained to eliminate rigid body motions it will deform and internal forces will be generated. It has been found convenient on any internal surface to define stress as the force per unit area (N/mm2 ). The stress on any internal surface is usually divided into components normal to that surface and tangent to it. The stress and stress components are depicted in Figure 1.7.1. Consider a force, p, acting on an element of area, A, in the interior of the solid. If we resolve the force p into components normal and tangential to the surface, pn and pt , respectively, we can define a normal component of stress, σ , and a tangential or shearing (or shear) stress component, τ, as shown in Figure 1.7.1. ∆ pn =σ ∆A y ∆p ∆A x ∆ pt =τ ∆A z Figure 1.7.1 If we allow A to shrink to an infinitesimal size, then, lim A→0 pn =σ A lim A→0 pt =τ A (1.7.1) To illustrate internal stresses let us first consider a uniform slender bar in equilibrium with equal but opposite distributed loads on each end as shown in Figure 1.7.2. These distributed forces have units of force per unit area and act on the end surfaces. There are no force components in the y and z directions. The applied force resultants in the x direction are F0 = A f x (0, y, z)dA = −F FL = A f x (L , y, z)dA = F --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (1.7.2) P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 28 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y fx (0,y,z) fx (L,y,z) x Figure 1.7.2 Let us choose distributed forces for which the applied moment resultants are zero, that is M y0 = Mz0 = A f x (0, yz)zdA = 0 My L = A f x (0, yz)ydA = 0 Mz L = A f x (L , yz)zdA = 0 A f x (L , yz)ydA = 0 (1.7.3) This is illustrated in Figure 1.7.3. y F F x x0 Figure 1.7.3 Now let us ask what the distributed internal force or stress is on an interior surface. Let us choose a flat surface normal to the x axis at, say, x = x0 and examine the stress on that surface by considering the equilibrium of the segment of the bar between x = x0 and x = L, as shown in Figure 1.7.4. As noted above, in the usual notation for stress, we designate the normal stress component with the symbol σ and the tangential or shearing stress component with the symbol τ . To further specify the specific components of stress we use a double subscript notation. The first subscript refers to the direction of the normal to the surface and the second to the direction of the stress. Thus we show the two possible stress components on this particular surface and label them σ xx and τ xy . Since normal stress components always have the same subscript repeated we usually use only one and so we use σ x for σ xx in subsequent applications. τ xy F σ xx x Figure 1.7.4 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 29 Forces and Moments The sign convention for stresses is different from that for external and restraint forces; the normal and shear stress components are positive as shown in Figure 1.7.4. Sign conventions for stress will be discussed in greater detail shortly. At this time we do not yet know the distribution of these stresses on that surface; however, we can define stress resultants on the surface as σx dA P= V = A τx y dA (1.7.4) A This is shown in Figure 1.7.5. V F x P Figure 1.7.5 If we sum the forces acting on the segment in the x and y directions we get Fx = F − P = 0 → P=F Fy = V = 0 → V =0 (1.7.5) While we do not yet know the actual distribution of stress on this surface we can define an average stress as   F V (σx )ave = τx y ave = =0 (1.7.6) A A Now consider that the chosen internal surface is not normal to the axis but is defined by a local coordinate system, tn, at an angle α to the xy coordinate system as shown in Figure 1.7.6. y α n t F F x --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- x0 Figure 1.7.6 The stresses on this surface, that we have labeled Aα , are shown in Figure 1.7.7. σn τnt F x Figure 1.7.7 Once again we can define stress resultants on this surface as Pn = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS σn dA Aα Vt = τnt dA Aα Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (1.7.7) P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 30 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods This is shown in Figure 1.7.8. Pn Vt F x Figure 1.7.8 When we sum the forces acting on the segment in the x and y directions we get Fx = F + Vt cos α − Pn sin α = 0 Fy = Vt sin α + Pn cos α = 0 → Vt cos α − Pn sin α = −F (1.7.8) Vt sin α + Pn cos α = 0 → Solving we get Pn = F sin α → Vt = −F cos α (1.7.9) And the average stresses are Pn F = sin α Aα Aα (τnt )ave = Vt F =− cos α Aα Aα (1.7.10) Stresses are always defined by their magnitude, direction, and the surface upon which they act. The actual distribution of the stresses on the interior surfaces in this example is discussed in detail in Chapter 4. To continue our discussion of stress we consider a thin flat plate acted upon by forces in the plane of the plate. We orient this plate in the xy plane of the coordinate system. With forces only in the plane of the plate on the thin plate edge surfaces and no forces on the plate surfaces in the z direction we can assume that the only stress components are those acting in the plane of the plate. To consider the stress at a point in the plate we take a small rectangular element, dxdy, of the plate as shown in Figure 1.7.9 and note the stress components as the size of the element approaches zero. y dy dx x Figure 1.7.9 The stresses on this rectangular element are shown in Figure 1.7.10. This is called a two dimensional state of stress. Remember this is a uniform stress through the thickness at a point where the edges dx and dy approach zero. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- (σn )ave = P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 31 Forces and Moments σy y τyx τxy σx σx dy dx τxy τyx σy x Figure 1.7.10 Stresses are positive as shown. If the normal to the surface upon which the stress acts points in the positive x direction the normal stress is positive in the positive x direction. If the normal points in the negative x direction the normal stress is positive in the negative x direction. If the normal points in the positive x direction the shear stress is positive in the positive y direction. If the normal points in the negative x direction the shear stress is positive in the negative y direction. We see that the forces called stress resultants have a different sign convention than the applied forces. At this time one would conclude that there are four components of stress in a two dimensional solid σx , σ y , τx y , τ yx . We note in Figure 1.7.10 that the τx y and τ yx components each form a couple and as the sides of the rectangle shrink to a point by applying moment equilibrium let us conclude that (1.7.11) τx y = τ yx This 2D state of stress is often depicted in matrix form as ⎡ ⎤ σx {σ } = ⎣ σ y ⎦ τx y (1.7.12) To depict a complete state of stress at a point in a three dimensional solid with respect to rectangular Cartesian coordinates consider a rectangular element dxdydz oriented as shown in Figure 1.7.11. Stress components are shown on the three faces with positive normal directions and on the one face with a negative normal (x) direction. Similar components are on the negative y and z directions but are not shown to avoid confusion. On a face whose normal is in the positive x direction we have three stress components as follows: σx = lim A→0 Px A τx y = lim A→0 Py A τx z = lim A→0 Pz A (1.7.13) On a face whose normal is in the negative direction of the axis the positive directions of both normal and shearing stresses are in the negative direction of their respective axes. At this time one would conclude that there are nine components of stress in a three dimensional solid σx , σ y , σz , τx y , τ yx , τx z , τzx , τ yz , τzy . By the same argument used in the two dimensional case each of the three pairs of shear stresses form a couple and as the element shrinks to a point, we can conclude from --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 32 14:39 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods σy τ yx τ yz τ xz τ xy τ zy σx σx τ zx τ xy τ xz Figure 1.7.11 moment equilibrium that τx y = τ yx τ yz = τzy τzx = τx z (1.7.14) Thus, there are six independent components of stress at a material point within a solid. These are variously arranged in matrix form as a column or square matrix according to their use in equations. ⎡ ⎤ σx ⎢ σy ⎥ ⎥ ⎢ ⎢ σz ⎥ ⎥ {σ } = ⎢ ⎢ τx y ⎥ ⎥ ⎢ ⎣ τ yz ⎦ τzx or ⎡ σx [σ ] = ⎣τx y τzx τx y σy τ yz ⎤ τzx τ yz ⎦ σz (1.7.15) At the beginning of our deliberation the stresses, generally, are unknown. Finding stresses in particular situations is a major part of the task before us. The state of stress on surfaces at other orientations than normal to the xyz axis is discussed in great detail in Chapter 9. Acceptable magnitudes of internal forces are limited by the properties of the material of the solid body. We all know that if they are too great the body will fail either from undesirable permanent deformations or from fracture. Material properties are introduced in Chapter 3 and discussed in greater detail in Chapter 9. 1.8 Restraint Forces and Restraint Force Resultants Solid bodies can be restrained at a point, they can be restrained along a line, or over a surface. When the body is subjected to applied forces there must be forces at the restraints. In a common statement of the problem these forces are unknown. Part of the purpose of our study in the following chapters is to find the value of these forces. In some circumstances it is possible to predict the value of the resultants of these forces without finding the actual distribution of the forces. This, in fact, is an important lesson of Chapter 2. Just how the restraint forces are distributed will be considered in Chapters 4 and beyond. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- σz P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come 33 Forces and Moments 1.9 Summary and Conclusions We have introduced the definition and notation for forces and moments of forces. An important part of the study of the mechanics of material is the interaction among applied forces, restraint forces, and internal forces and the moments generated by them. These forces come in various forms such as concentrated forces with units of Newtons (N), forces per unit length or line forces with units of Newtons per millimeter (N/mm), surface forces with units of Newtons per millimeter squared (N/mm2 ), and body forces with units of Newtons per millimeter cubed. (N/mm3 ). We have shown how force resultants of distributed forces are found and how their location is determined. We have defined internal forces and stresses, and restraint forces. Now we shall put all this to good use in Chapter 2 in establishing static equilibrium, that is, in satisfying Newton’s laws as the various forces interact. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-01 P2: ABC JWST071-Waas July 2, 2011 14:39 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 2 Static Equilibrium 2.1 Introduction While all solid bodies deform to some extent under loads these deformations are often so small that the bodies may be considered to be rigid for certain purposes. A rigid body is an idealization that considers a body to have no deformation when subject to loads. When all forces, both known applied forces and unknown restraint forces, are identified and we attempt to sum the forces and moments we can have two possible situations: 1. The number of unknown forces is less than the number of independent equations of motion. Rigid body motion may result. This is the subject of rigid body dynamics. We shall not study this case here. 2. The number of unknown forces is equal to or greater than the number of independent equations of motion. There is no rigid body motion and we have static equilibrium. There are two subcases: a. The body is statically determinate when the number of unknown forces is equal to the number of independent equations of static equilibrium. All the unknown forces can be determined by the summation of forces and moments without regard to what the body is made of or to its deformation. b. The body is statically indeterminate when the number of unknown forces is greater than the number of independent equations of static equilibrium. The summation of forces and moments is not sufficient to find the unknown forces. We need to consider the deformation of the body and the physical properties that contribute to resisting deformation. In the case of a statically indeterminate body we must introduce additional equations to obtain a solution. In this chapter we shall be interested primarily in statically determinate bodies. The subsequent chapters will contend with both statically determinate and statically indeterminate problems. Free Body Diagrams We shall illustrate each subcase with a simple example in two dimensions using concentrated forces only. In these examples it is assumed that the applied forces initially are known and the restraint forces are unknown. Consider the rigid body shown in Figure 2.2.1. We introduce some common symbols to illustrate forces and restraints. For now we consider only concentrated loads and, in this case, assume they act entirely in the xy plane. We add concentrated restraints. The restraint on the lower left prevents movement in both the x and y directions but does allow rotation about the z axis. The restraint on the lower right Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 2.2 P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 36 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods prevents movement only in the y direction, since it is assumed to roll freely (without friction) in the x direction. Since it is assumed that there are no forces in the z direction there is no need to consider restraints in the z direction. F2 F1 y x Figure 2.2.1 By replacing the restraint symbols in Figure 2.2.1 with visual representations of the characteristics to which we have assigned them, we create what is called a free body diagram. We draw the free body in Figure 2.2.2 showing all the applied forces and restraint forces. Unless otherwise stated we shall use the symbol F for concentrated applied forces and R for concentrated restraint forces. Appropriate subscripts will be added to identify individual forces and components. It is common to represent both the applied forces and the restraint forces in terms of their components in the appropriate coordinate directions. For the applied forces these components are F1x = F1 cos α F1y = F1 sin α F2x = F2 cos β F2y = F2 sin β a2 (2.2.1) F2 F1 α β y a1 RAx B A a3 RAy x RBy Figure 2.2.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- According to standard sign convention, restraint force components generally are drawn directed along the positive x and y directions, as defined by the coordinate system. If a force component acts in the direction it is drawn, it has a positive value. If it acts in the opposite direction, it has a negative value. We have static equilibrium between the known applied forces and the unknown restraint forces. There are four known applied force components (F1x , F1y , F2x , and F2y ), three unknown restraint force components (R Ax , R Ay , and R By ) and three equations of static equilibrium (summation of forces in the x and y directions and summation of moments about the z axis). The summation of moments about the z axis may be taken about any point in the xy plane. In the following equations the summation of moments Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 37 Static Equilibrium is taken about point A.  Fx = RAx + F1x − F2x = 0 → RAx = F2x − F1x = F2 cos β − F1 cos α  F1 a1 + F2 a2 Mz = RBy d3 − F1 a1 − F2 a2 = 0 → RBy = a3  F1 a1 + F2 a2 Fy = RAy + RBy − F1y − F2y = 0 → RAy = F1y + F2y − a3 F1 a1 + F2 a2 = F1 sin α + F2 sin α − a3 (2.2.2) In this case, the unknown restraint forces can be found by solving the equations of static equilibrium. Bodies that satisfy this condition are called statically determinate. These bodies will play a prominent role in subsequent chapters. The statically determinate case is considered in some detail in this text. Many structures are naturally, or are purposely made to be, statically determinate; however, that will not be the case for many of our problems in later chapters. Indeterminate, also called redundant, structures have important positive characteristics and are widely used. Consider the body in Figure 2.2.3. Suppose we remove the rollers from the bottom right restraint and add a third restraint at the top of the body as shown. F2 F1 y x The free body diagram is shown in Figure 2.2.4. RCy a2 C F1 F2 RCx α β y a1 RAx x RBx A a3 RAy B RBy Figure 2.2.4 Just by removing the rollers on the bottom right support we add a fourth restraint force component RAx . This alone would make the number of unknowns greater than the number of equations of static equilibrium. Add the third support and we then have six unknown restraint force components. We can Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure 2.2.3 P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 38 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods write down the three equations of static equilibrium and solve them for three of the unknowns in terms of the remaining three unknowns, but, we shall need additional information in order to find all the restraint forces. You shall soon see that, for a deformable body, equations based on distributed displacements and material properties will supply the additional necessary information. Distributed displacements of the deformed body will play a prominent role in subsequent chapters. The focus of most of this text will be the subject of statics. The bodies that we study will be in static equilibrium. Accordingly, all forces and all moments acting on those bodies will always sum to zero. In dynamics, on the other hand, bodies often experience rigid body motion and/or nonzero net forces. Some dynamic considerations are introduced in Chapter 15. So far our examples have been two dimensional. In three dimensions, when presented in rectangular Cartesian coordinates, the equations of statics are   Fx = 0 Mx = 0   Fy = 0 My = 0   Fz = 0 (2.2.3) Mz = 0 where the moment summations are taken about each axis at any point in space. For a statically determinate system, these equations will be sufficient to find the restraint forces and internal forces. For a statically indeterminate system, additional equations based on deformations will be needed. 2.3 Equilibrium—Concentrated Forces Concentrated forces occur in the form of point loads, restraints at a point, resultants of distributed forces, and stress resultants. It is often very useful to consider a deformable, restrained body under the influence of applied loading to be rigid and to replace its actual applied loads and restraints with equivalent concentrated applied and restraint force components. Several such cases will be introduced here and will have significant applications in later chapters. 2.3.1 Two Force Members and Pin Jointed Trusses If a structural member has only equal and opposite point forces applied at its ends, such as those examples shown in Figure 2.3.1, it is called a two force member. The forces, which could be applied via pins inserted through holes at each end of the member, would, in reality, be distributed, that is, where the pins bear against the inside surfaces of the holes. For simplicity of analysis, however, the actual distributed forces will be represented by their concentrated force resultants acting through the centers of the holes. Exactly how the forces are applied is not now of concern as long as no moments are generated. Since this is the study of statics, the member must be in equilibrium, and thus the two forces must be equal and opposite and must act along the direction defined by the dotted lines below. F --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- F Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS F F Figure 2.3.1 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 39 Static Equilibrium Structures are often constructed by joining the ends of several two force members with pins. This is called a pin jointed truss. A 2D example is shown in Figure 2.3.2a. Both 2D and 3D truss structures are used as load carrying components. Several conventions are used in illustrating pin jointed trusses. Each of the two force members is shown as a solid line. The actual geometry of the cross section of the member is not of concern at this time. Only the cross sectional area is relevant. The pinned joints are sometimes shown as small circles (Fig. 2.3.2a). Other times the pinned joints are implied and no circles are shown (Fig. 2.3.2b). y x (a) (b) Figure 2.3.2 Pinned joints are free to translate in directions in which they are not restrained. In Figure 2.3.2 the lower left joint is restrained in both the x and y directions, and thus is fixed. These two restraints prevent rigid body motion of the truss as a whole. The lower right joint is restrained in the y direction but is free to move in the x direction. The upper right joint is free to move in both the x and y directions. Because two of the joints are able to translate, the individual members are able to deform and rotate. If the assumption of two force members is to remain valid, forces can be applied only at the joints of the truss. Accordingly, member weights, the resultants of which act through the members’ respective centers of gravity, must be able to be neglected. Since the applied forces are frequently much larger than the member weights, this is often the case. In some cases the member weights are divided up as equivalent forces at the joints. Exceptions will be identified explicitly. ########### Example 2.3.1 Problem: Find the restraint forces and the forces acting on each member of the pin jointed truss shown in Figure (a). Neglect the weight of the members. y L L C D F H A α B x Figure (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 40 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Step 1: Draw the global free body diagram (Figure (b). According to standard sign convention, the restraint force components are drawn to have positive values when acting in the positive directions of the coordinate axes. y L L D C F H RAx α A x B RBy RAy Figure (b) To find the restraint forces sum all forces in the x and y directions and sum moments about the z axis. Specifically, we shall take moments about point A.    Fx = F + RAx = 0 → Mz = RBy L − FH = 0 Fy = RAy + RBy = 0 RAx = −F → R = FH By → (a) L RAy = −RBy → RAy = − FH L Step 2: We can now draw free body diagrams for each member and each pinned joint (Figures (c) and (d)). FAC FCD FCD FBD FBC FAC FBD FAB FAB FBC Figure (c) Free body diagrams for each member are shown in Figure (c). According to standard sign convention, forces acting on each member are drawn to have positive values when putting the member in tension. Thus, compressive forces will have negative values. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Solution: We shall first draw a “global” free body diagram for the entire truss to identify the restraint forces generated at its supports. Those restraint forces will then be found using the equations of static equilibrium. Once the restraint forces are known, free body diagrams will be drawn for each member and each joint so that the force on each member can be determined. P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 41 Static Equilibrium Free body diagrams for each pinned joint are shown in Figure (d). According to standard sign convention, joint forces are drawn to have positive values when directed away from the joint. This is consistent with the sign convention used for member forces in that forces directed away from the joint are tensile forces. Joint D Joint C FCD α FAC FBC F FBD Joint B α F α FBD FBC FAC Joint A FCD FAB α FAB FH/L FH/L Figure (d) Because the entire truss is in equilibrium, each member and each joint is also in equilibrium. Accordingly, equations of static equilibrium must be satisfied. Since we are dealing with two force members, the moment equilibrium equation is identically satisfied for each member and each joint. In summing forces at each joint we assume positive forces are in the direction of the positive coordinate axes. At Joint D:   At Joint B:  At Joint A:  Fy = −FBD sin α = 0 → Fx = F − FCD − FBD = 0 FBD = 0 → FCD = F FH FH = 0 → FBC = − L L = 0 → FAB = 0 Fy = FBD sin α + FBC + Fx = FBD cos α − FAB   Fy = FAC sin α − FH =0 L → Fx = FAC cos α + FAB − F = 0 FAC = → FH sin αL FAC = F cos α (b) (c) (d) It would appear that at Joint A we have found two different values for FAC ; however, if we note that H/L = tan α then both values are the same. We have found the values for all the forces on the members and so it is not necessary to sum forces at Joint C; however, we can use Joint C to confirm that the above answers are correct. It is always desirable to have an independent check of any analysis. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 42 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods At Joint C:   Fx = FCD − FAC cos α = 0 Fy = −FAC sin α − FBC = 0 F cos α = 0 cos α F sin α − F tan α = 0 cos α F− → → (e) Example 2.3.2 Problem: Find the support reactions (restraint forces) and the forces acting on each member for the pin jointed truss in Figure (a). Neglect the weight of the members. y 500 mm B 500 mm A α = 30ο C D x 866 mm 1000 N Figure (a) Solution: This time we shall first set up a free body diagram of one joint, solve for the internal forces in the members connected by that joint, and then proceed from joint to joint without first finding the support reactions. Step 1: The free body diagram for Joint D is shown in Figure (b). FDB --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- FDC α D 1000 N Figure (b) From summation of forces we get   Fy = FDB sin α − 1000 = 0 Fx = −FDC − FDB cos α = 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS → → 1000 1000 = = 2000N sin α sin 30◦ = −2000 cos α = −2000 cos 30◦ = −1732N FDB = FDC Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 43 Static Equilibrium Step 2: The free body diagram for Joint B is shown in Figure (c). B FAB α FDB = 2000 N FBC Figure (c) From summation of forces we get   Fx = 2000 cos α − FAB = 0 FAB = 2000 cos α = 2000 cos 30◦ = 1732N → Fy = −FBC − FDB sin α = 0 FBC = −FDB sin α = −2000 sin 30◦ = −1000N → (b) As defined previously, a positive value indicates a tensile force and a negative value indicates a compressive force. Step 3: The free body diagram for Joint A is shown in Figure (d) A FAB RAx RAy Figure (d) From summation of forces we get   Fx = RAx + FAB = 0 Fy = R Ay = 0 → → RAx = −FAB = −1732N R Ay = 0 (c) The free body diagram for Joint C is shown in Figure (e). FBC B RCx FDC Ry Figure (e) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- From summation of forces we get  Fx = RCx + FDC = 0  Fy = RCy + FBC = 0 → → RCx = −FDC = 1732N RCy = −FBC = 1000N (d) The support reactions are Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS RAx = −1732N RAy = 0 RCx = 1732N RCy = 1000N ########### Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (e) P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 44 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Not all pin jointed trusses are statically determinate. The truss shown in Figure 2.3.3 has more unknown forces than available equations of static equilibrium and therefore is statically indeterminate. Note that the two interior members are not joined where they cross. Figure 2.3.3 The method used to find the additional equations required to solve a statically indeterminate truss problem is considered in Chapters 4 and 5. As stated earlier trusses, both determinate and indeterminate, can also be three dimensional. In such cases, we will have one more relevant force equation and two more relevant moment equations, raising our number of relevant static equilibrium equations from three to six. We will deal with these cases in later chapters. 2.3.2 Slender Rigid Bars --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Slender bars that can carry lateral and rotational loads as well as axial loads are one structural form that has received a lot of specialized attention. A slender bar is defined as a solid body having one dimension that is much larger than the other two. Two force members used in trusses are a special case of slender bars where the bar deformation is restricted to be along the axis, that is, a one dimensional axially deformable member. Using slender bars allows certain simplifications to be made. When forces and moments are distributed over sufficiently small areas, they can be approximated as concentrated forces and moments. For a slender bar, the surface areas of the ends of the bar are very small when compared to the surface area of the entire bar. Accordingly, restraint forces created by supports at the ends of the bar, which are actually distributed over the small surface area of the ends of the bar, can be approximated as concentrated restraint force and moment resultants. All bodies deform when subjected to loads and restraints; however, in many instances the amount of deformation is very small. In those cases the loaded bodies very nearly maintain their unloaded sizes and shapes; therefore, their dimensions and location of loads and restraints in the unloaded state can be used when establishing, for example, equilibrium in the loaded state. This is an important assumption made in the analysis of structures that undergo small deformations. We did this with two force members in trusses. Thus, we extend this assumption when we examine the equilibrium of deformable bars in this section. Initially we restrict discussion to two dimensional cases, in which all forces lie in the xy plane and only moments about the z axis can be non zero. This keeps the problems simple and allows us to solve them without becoming overwhelmed by complications in geometry. There are several conventions for showing restraints and depicting restraint forces in drawings and diagrams. The special case of Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 45 Static Equilibrium using concentrated fixed restraint force resultants for a slender, rigid bar is shown in Figure 2.3.4 for three different types of supports. Once again we shall use R to designate concentrated restraint forces and use Q to designate concentrated restraint moments, with subscripts used to show the coordinate direction. While the forces are, in fact, distributed over the surface area of the end of the bar, they are represented here as concentrated resultant forces. In later chapters we shall consider distributed force restraints as well. Rx Qz Fixed support Ry Rx Simple (pinned) support Ry Roller support Ry Figure 2.3.4 The forces and moments are shown in what is the convention used in this text for their positive directions. At the beginning of a problem the values of these forces and moments will not be known. To find out more about them is part of what we are about to study. ########### Example 2.3.3 Problem: Consider the rigid bar fixed at the left end and free at the right end as shown in Figure (a). This is known as a cantilever support. It has two concentrated applied forces as shown. Find the restraint forces and moments. Neglect the weight of the bar. y F1y L1 F2x x L2 Figure (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 46 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Solution: Draw a free body diagram and sum forces and moments. The free body diagram is shown in Figure (b). F1y L1 F2x R0x Q0z L2 R0y Figure (b) Assuming F 1y and F 2x are known, the three restraint forces and moments can be found from the three available equations of static equilibrium.  Fx = R0x + F2x = 0 → R0x = −F2x Fy = R0y + F1y = 0 → R0y = −F1y  Mz = Q 0z + F1y L 1 = 0 → (a) Q 0z = −F1y L 1 Example 2.3.4 Problem: Consider the rigid bar with pinned supports at both ends as shown in Figure (a). This is known as a simply supported beam. It has two concentrated applied forces as shown. Rollers are added at the right end to allow the force F2x to be transmitted to the bar. Find the restraint forces. Neglect the weight of the bar. L2 F1y L1 F2x Figure (a) Solution: Draw a free body diagram and sum forces and moments. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`---  P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 47 Static Equilibrium The free body diagram is shown in Figure (b). F1y L1 L2 R0y R2y Figure (b) Assuming F1y and F2x are known, the three restraint forces and moments can be found from the three available equations of static equilibrium. If we sum moments about the left end and then sum forces we have    Mz = F1y L 1 + R2y L 2 = 0 → Fx = R0x + F2x = 0 R0x = −F2x → Fy = R0y + F1y + R2y = 0 R2y = − → F1y L 1 L2 (a) R0y = −F1y − R2y F1y L 1 = −F1y + = F1y L2   L1 −1 L2 Example 2.3.5 Problem: Consider the rigid bar fixed at one end and pinned at the other as shown in Figure (a). It has two concentrated applied forces as shown. Find the restraint forces. Neglect the weight of the bar. L2 F1y L1 F1x Figure (a) Solution: Draw a free body diagram and sum forces and moments. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- F2x R0x P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 48 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The free body diagram is shown in Figure (b). F1y L1 F1x R0x R2x Q0z L2 R0y R2y Figure (b) Assuming F1y and F2x are known, there are now five unknown restraint forces but just three available equations of static equilibrium.    Fx = R0x + F1x + R2x = 0 Fy = R0y + F1y + R2y = 0 (a) Mz = Q 0z + F1y L 1 + R2y L 2 = 0 We can reduce the number of unknowns to two by solving the equations but need additional equations to obtain a full solution. This is a statically indeterminate problem. Where and how to find the additional equations is the subject of later chapters. Example 2.3.6 Problem: Find the support reactions for the built up bar structure shown in Figure (a). Neglect the weight of the structure. y 1000 mm 300 mm 1000 mm F1x = 67 N x M2z =12345 N ⋅ mm Figure (a) Solution: Set up the free body diagram and sum forces and moments. Note that we have a concentrated applied moment at the right end. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 49 Static Equilibrium The free body diagram is shown in Figure (b). y 1000 mm 1000 mm 67 N Q0z 300 mm x R0x 12345 N ⋅ mm R0y Figure (b) And the solution is    Fx = R0x + 67 = 0 → R0x = −67N Fy = R0y = 0 (a) Mz = Q 0z + 12345 − 67 · 300 = 0 → Q 0z = 7755N · mm ########### 2.3.3 Pulleys and Cables Flexible cables combined with pulleys enable convenient changes in the direction of tensile forces. Consider the case shown in Figure 2.3.5. Neglect the weight of the cable, pulley, and pin jointed truss support. Rpx r l y Rpy 0 45 W W (weight) x Figure 2.3.5 Using the pulley free body diagram in Figure 2.3.5 we can obtain the force in the cable, Fc , by summing moments about the center of the pulley and the reactions on the pulley hub by summing forces. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Fc P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 50 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Note that the radius of the pulley is r.  Mz = Fc r − W r = 0 → Fc = W  Fx = R px − Fc = 0 → R px = Fc = W  Fy = R py − W = 0 → R py = W (2.3.1) The free body diagram of the body supporting the pulley is shown is Figure 2.3.6. Note that the pulley and the truss supporting it apply equal and opposite forces on each other through the pin that connects them. W A W B l 450 C Figure 2.3.6 A free body diagram of the pinned joint supporting the pulley is shown in Figure 2.3.7 W W FAB FBC Figure 2.3.7 The two equations of static equilibrium give us the forces in the truss members.   Fy = −FBC cos 45o − W = 0 → Fx = −FAB − FBC sin 45o − W = 0 FBC = − → 2W W = −√ cos 45o 2 (2.3.2) FAB = −FBC sin 45o − W = W − W = 0 ########### Example 2.3.7 Problem: Find the applied forces and the support reactions for the structure shown in Figure (a). The radius of the pulley is 100 mm. Neglect the weights of the structure and the pulley. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 51 Static Equilibrium y 1000 mm 1000 mm 300 mm 11.2o x 400 N Figure (a) Solution: To find the forces transmitted to the bar by the pulley set up the free body diagram of the pulley and sum forces and moments. To find the support reactions set up a free body diagram of the bar. The free body diagram of the pulley is shown in Figure (b). 97.47 Fc 11.2o 392.38 Rpx Rpx Rpy 400 N Rpy 400 N Figure (b) From moments about the hub of the pulley we have  Mz = Fc · 100 − 400 · 100 = 0 → Fc = 400N (a) From summation of forces on the pulley we have   Fx = R px − 400 cos 11.2◦ = R px − 392.38 = 0 → Fy = R py + 400 sin 11.2◦ − 400 = R py − 322.31 = 0 R px = 392.38N → R py = 322.31N Then set up the free body diagram of the structure as shown in Figure (c) and solve. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- y 1000 mm 1000 mm 400 N 300 mm Q0z 392.38 N R0x 322.31 N R0y Figure (c) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 52 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Actually, it is simpler to use a global free body diagram to find the support reactions. Here it is in Figure (d). y 1000 mm Q0z R0x 1000 mm 300 mm 11.2o x R0y 400 N Figure (d) From Figure (d) we can quickly determine that  Fx = R0x = 0  Fy = R0y − 400 = 0 → R0y = 400N  Mz = Q 0z − 400 · 2100 = 0 → Q 0z = 840000 N · mm (c) As we shall see in detail later the applied forces as shown in Figure (c) are useful in finding internal forces and stresses for the structure. ########### We can, of course, have components of applied forces and restraints in the z direction and applied moments and restraints about the y axis as well. We shall consider such cases in Section 2.5. Moments about the x axis will be considered separately. 2.3.4 Springs Springs are often used to support solid bodies. Their behavior is expressed in terms of overall displacement, d, and force, F. Consider a spring acted upon by equal and opposite forces (it is a two force member actually) or restrained at one end and loaded by a force at the other as shown in Figure 2.3.8. F F F For a linearly elastic spring the force displacement curve is as shown in Figure 2.3.9. At some value of force and displacement the spring will yield and eventually fracture. We shall limit discussion to the linear elastic range. The slope of this curve is designated as, k, and is called the spring constant. Thus F =k d → F = kd (2.3.3) The spring constant k has units of Newtons per millimeter (N/mm). Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure 2.3.8 P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 53 Static Equilibrium F F =k d d Figure 2.3.9 ########### Example 2.3.8 Problem: A rigid bar of length, L, is supported at the ends by two springs, both having spring constant, k, as shown in Figure (a). Find the forces in and elongations of the springs due to the weight of the bar. Neglect the weight of the springs. k Undeformed position of the bar k dA dB A B L Figure (a) Solution: Draw a free body diagram. Then solve for the forces in the springs. Finally, use Equation 2.3.3 to find the spring elongations (dA and dB ). The free body diagram is shown in Figure (b). FB FA L/2 W Figure (b) By summing forces and taking the moment about the left end we get   Fy = FA + FB − W = 0 L Mz =FB · L − W = 0 2 → → FA + FB = W L FB · L = W 2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 54 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods and the solution is FA = FB = W 2 dA = dB = → W 2k (b) Example 2.3.9 Problem: A weight equal to the weight of the bar is added to the right side of the bar, one fourth of the length of the bar from its right end as shown in Figure (a). Given the value of the spring constant at the left end (kA ) what should be the value of the spring constant at the right end (kB ) for the bar to remain level? Include the weight of the bar but neglect the weight of the springs. kA Undeformed position of the bar kB dA dB A B 3L/4 L/4 W Figure (a) Solution: Draw a free body diagram. Then solve for the spring forces (FA and FB ). Finally, use Equation 2.3.3 to find the spring constant of the right spring (kB ). The free body diagram is shown in Figure (b). FB FA L L/2 L/4 W W Figure (b) First find the moment about the left end. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`---  Mz = FB · L − W 3L L −W =0 2 4 → FB = 5W 4 (a) Next sum forces in the y direction.  Fy = FA + FB − 2W = 0 → FA = 2W − FB = 3W 4 Let dB = dA , where dA and dB represent the elongations of the respective springs. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 55 Static Equilibrium Then dA = 3W 4kA → kB = 5 FB 5W 4kA · = kA = dA 4 3W 3 (c) ########### 2.4 Equilibrium—Distributed Forces In Section 1.6 we noted that distributed forces can be resolved into force resultants which act at a point with a particular line of action. It is often the case that the forces per unit area applied to the surface of a bar, such as those we considered in Section 1.6, are constant across the width of the bar (see Example 1.6.1). These can be resolved into forces per unit length along the bar. Consider a cantilever bar with a distributed load fy (x) as shown in Figure 2.4.1. Neglect the weight of the bar. y fy(x) x L Figure 2.4.1 The distributed force has units of force per unit length (N/mm). The force resultant can be found.  L FR = f y (x)dx (2.4.1) 0 The centroid, or location of the line of action, of the distributed force is  L x f y (x)dx x = 0 L f y (x)dx (2.4.2) 0 A free body diagram of the beam is shown in Figure 2.4.2 with the force resultant applied in place of the actual distributed force. FR Qz x Ry Figure 2.4.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 56 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods From summation of forces and moments  Fy = R y + FR = 0 → R y = −FR  Mz = Q z + FR x = 0 → Q z = −FR x (2.4.3) ########### Example 2.4.1 Problem: The bar in Figure 2.4.1 is given a uniform distributed load (N/mm) as shown in Figure (a). Find the support reactions. Neglect the weight of the bar. y x L Figure (a) Solution: Find the force resultant and line of action of the distributed load. Then set up a free body diagram and solve. The force resultant can be found to be FR =  L f y (x)dx = f 0 0  L 0 dx = f 0 x|0L = f 0 L (a) By inspection, the location of the centroid of the distributed load, and thus the location of the line of action of the distributed load, is at L/2. Or, if you wish to find it formally by integration. f0 x=  f0 L f0 xdx 0  L dx = 0 L x 2  L2 L 2 0 = = f0 L 2L 2 The free body diagram is shown in Figure (b) f0L Q0z L/2 R0y Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- fy(x) = f0 P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 57 Static Equilibrium From summation of forces and moments   Fy = R0y + f 0 L = 0 Mz = Q 0z + f 0 L · → L =0 2 R0y = − f 0 L → Q 0z = − (c) f0 L 2 2 Example 2.4.2 Problem: A cantilever bar is loaded with a triangularly distributed line load (N/mm) as shown in Figure (a). Find the support reactions. Neglect the weight of the bar. Solution: Find the force resultant and line of action of the distributed load. Then set up a free body diagram and sum forces and moments. f y ( x) = f 0 1 − x L x L Figure (a) The resultant of the distributed load is  L  f y (x)dx = Fy = 0 L 0  f0 L x dx = f0 1 − L 2 and its line of action passes through the centroid of the load distribution.  L  L  x dx x f y (x)dx x 1− f0 L L 0 x = 0 L = =  L 3 x dx 1− f y (x)dx f0 L 0 0 The free body diagram is shown in Figure (b). L/3 f0L/2 Q0z R0y Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) (b) P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 58 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The support reactions are f0 L f0 L = 0 → R0y = − 2 2  f0 L L f0 L 2 · = 0 → Q 0z = − Mz = Q 0z + 2 3 6  Fy = R0y + (c) Example 2.4.3 Problem: A rigid bar one meter long is loaded as shown in Figure (a). What is the force in the spring and what is the total displacement at the right end of the bar in terms of f0 ? Neglect the weights of the bar and the spring. f0 (N/mm) k 560 mm Figure (a) Solution: First, replace the load distribution with equivalent force resultants. Then draw a free body diagram and sum forces and moments. Finally, calculate the spring deflection (ds ) and use that to find the displacement of the right end (dL ). It is convenient to split the load distribution into a triangularly distributed load and a uniformly distributed load. The functions for the triangularly distributed load f 1 (x) and the uniformly distributed load f 2 (x) are f 1 (x) = f 0 f 2 (x) = f 0 x 560 0 ≤ x ≤ 560 (a) 560 ≤ x ≤ 1000 The resultant force, F 1, of the triangularly distributed region is found by integration. F1 =  0 560 f0 560 x 2  x dx = f 0 = 280 f 0 560 1120 0 (b) You could also find F 1 by calculating the area under the triangular region F1 = 560 f 0 = 280 f 0 2 (c) From the area under the rectangular uniformly distributed region F2 = 440 f 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 1000 mm P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 59 Static Equilibrium The line of action of F 1 is x1 = 1 F1  560 x f0 0 560 1 1120 x 3  x dx = mm f0 =  560 F1 3 · 560 0 3 (e) or, from the equation for the centroid of the triangle, two thirds the distance from the left end. The line of action of F 2 is at the midpoint of the region of uniformly distributed load. x 2 = 560 + 220 = 780 mm (f) The free body diagram is shown in Figure (b). F2 F1 R0y 1120/3 Fs 560 780 Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- By summing moments about the left end we find the force in the spring.    1 1120 1120 − F2 780 = 0 → Fs = 280 f 0 + 440 f 0 780 = 799.5 f 0 Mz = 560Fs − F1 3 560 3 (g) The deflection of the spring is ds = f0 Fs = 799.5 k k (h) The displacement at the right end (dL ) is found by the simple ratio ds dL = 1000 560 → dL = f0 f0 1000 1000 ds = 799.5 = 1428 560 560 k k The restraint force at the left end is found by summing forces in the y direction.  Fy = R0y + Fs − F1 − F2 = 0 → R0y + 799.5 f 0 − 280 f 0 − 440 f 0 = 0 (i) (j) → R0y = −79.5 f 0 ########### 2.5 Equilibrium in Three Dimensions All the explanations and examples up to now have been limited to two dimensions, that is, to static equilibrium in a plane. We have arbitrarily chosen that plane to be the xy plane and have presented all Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 60 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods cases in terms of rectangular Cartesian coordinates. In such cases there are three equations of static equilibrium.  Fx = 0  Fy = 0  Mz = 0 (2.5.1) We are fortunate that a number of real problems can be formulated and solved in this way. Many other real problems must be formulated and solved in terms of three dimensions. For these cases the equations of static equilibrium in rectangular Cartesian coordinates are   Fx = 0 Mx = 0  Fy = 0  My = 0  Fz = 0  Mz = 0 (2.5.2) We shall limit our three dimensional discussion to a few simple examples. The principles are the same but the application can get quite complicated. ########### Example 2.5.1 Problem: Find the internal forces in the members of the three dimensional truss shown in Figure (a) in terms of the applied weight, W. Neglect the weight of the truss members. Solution: Set up free body diagrams and sum forces. y y 1000 mm C 1000 mm --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- D z x B 1000 mm 1000 mm W A Isometric view Side view W x Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 61 Static Equilibrium The free body diagram at Joint A is shown in Figure (b). y FAC FAB FAD A W Figure (b) We sum forces in all three coordinate directions. √  Fy = FAC sin 45◦ − W = 0  Fx = FAD cos 26.565◦ + FAB cos 26.565◦ + FAC cos 45◦ = 0  → FAC = ◦ ◦ Fz = FAD sin 26.565 − FAB sin 26.565 = 0 → → 2W → FAD = FAB W FAD + FAB = − cos 26.565◦ W FAD = FAB = − = −0.559W 2 cos 26.565◦ (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The internal forces are the same as the forces at the joints. PAB = FAB = −0.559W PAC = FAC = √ 2W PAD = FAD = −0.559W (b) Example 2.5.2 Problem: Find the support reactions for the rigid bar shown in Figure (a). A free body diagram is shown in Figure (b). Restraint moments are indicated by double arrowheads. Neglect the weights of the bar and the flanges. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 62 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y Q0y Q0x R0y 2F R0x Q0z z L/4 R0z F F L/4 L/2 L/2 2F 3F 3F x Figure (a) Figure (b) Solution: Sum forces and moments in all six degrees of freedom. The sum of the forces and moments and the solutions are    Fx = R0x + 3F = 0 → R0x = −3F Fy = R0y − 2F = 0 → R0y = 2F Fz = R0z − F = 0 → R0z = F L FL L Mx = Q 0x + 2F · − F · = 0 → Q 0x = − 4 4 4  M y = Q 0y + F · L = 0 → Q 0y = −FL  L Mz = Q 0z − 2F · = 0 → Q 0z = FL 2 (a)  --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ########### 2.6 Equilibrium—Internal Forces and Stresses Up to now we have been concerned with the equilibrium of the entire body or the separate members under the actions of applied loads and restraints. Now let us consider what is happening in the interior of the body. Any portion of the interior must also be in equilibrium. A free body diagram can be drawn for a portion of the solid body so that the summation of forces may consist of both external (applied) forces and internal forces (stress resultants). In Section 2.3 we found the forces acting on a member of a determinate pin jointed truss. A simple free body diagram of a portion of that member shows that the internal force is the same as the applied force (Fig. 2.6.1). We label the internal force, P, and note that it must be equal to F.  Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Fx = F − P = 0 → P=F (2.6.1) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 63 Static Equilibrium y F F x F P Figure 2.6.1 P is the force resultant of the stress distribution over the cross section upon which it acts. For the given coordinate system, the normal stress on that particular surface would be labeled σ x . Thus  P= σx dA (2.6.2) A We know that the value of P is equal to the integral of the stress, σ x , acting on the cross section surface but we do not yet know how that stress is distributed. The actual distribution of that stress is considered in detail in Chapter 4. Similarly we can make a free body diagram of the portion of the bar in Figure 2.6.2 that is to the right of a coordinate x0 on the bar. y fy(x) x x0 L --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Fyx0 Vx0 M x0 x0 x L Figure 2.6.2 We show an internal transverse force, V x0 , and an internal moment, M x0 , on the bar at x = x0 . We choose the notation V and M to distinguish the internal quantities from the restraint quantities which are generally designated by R and Q, for quantities of force and moment, respectively. We also choose a sign convention that will be explained in detail in subsequent chapters. Shown are the directions of positive values for internal force resultants in a slender bar according to said sign convention. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 64 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Then we can calculate V and M at x 0 by summation of forces and moments.  Fy = Fyx0 − Vx0 = 0 → Vx0 = Fyx0  Mz = (x − x0 ) Fyx0 − Mx0 = 0 → Mx0 = (x − x0 ) Fyx0 (2.6.4) Once again we can find the internal stress resultants but we do not yet know how the stress is distributed. There would be both a shearing stress, τ xy , and a normal stress, σ x , acting on a surface normal to the x axis. Thus the stress resultants are   τx y dA Mx0 = yσx dA (2.6.5) Vx0 = A A The actual distribution of the stresses is considered in detail in Chapter 7. ########### Example 2.6.1 Problem: For the bar in Figure (a) find the internal forces at L/2. f y ( x) = f 0 1 − x L x L Figure (a) Solution: First set up a free body diagram of the bar from x = L/2 to L. Then find the force resultant of the applied distributed load and its location. Finally, sum forces and moments. The free body diagram is shown in Figure (b). y f y ( x) = f 0 1 − VL/2 ML/2 x L x L/2 Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- To find the internal force and moment we first find the resultant of the applied distributed load and its location for the portion of the bar to the right of x0 .  L x f y (x)  L x f y (x) dx → x = 0 (2.6.3) Fyx0 = Fyx0 x0 P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 65 Static Equilibrium The applied force resultant for the portion of the beam from x = L/2 to L is found by taking the area of the triangle. f0 L 1 f0 L · · = 2 2 2 8 FR = (a) The location of the resultant is at the centroid of the triangle which is one third the distance from its left side. As measured from the origin the line of action is x= 1L 2L L + = 2 32 3 (b) The free body diagram with force resultants for the portion of the beam outboard of x = L/2 is shown in Figure (c). f0L/8 y VL/2 ML/2 x x0 = L / 2 x = 2L / 3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (c) The internal force and moment are f0 L − VL/2 = 0 → 8  f0 L L · − M L/2 = 0 Mz = 8 6  Fy = VL/2 = → f0 L 8 M L/2 = f0 L 2 48 (c) ########### Now let us go completely into the interior of the body and look at equilibrium there. This time we deal directly with the stresses rather than with the stress resultants. 2.6.1 Equilibrium of Internal Forces in Three Dimensions Consider a small rectangular element of infinitesimal dimensions dx by dy by dz somewhere inside the boundaries of a three dimensional solid body as shown in Figure 2.6.3. The body is restrained and has applied loads. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 66 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure 2.6.3 We wish to place this internal element in equilibrium. Using a Taylor series expansion we can write the change in stress from one face of the element to another face dx away as shown in Figure 2.6.4. We show the stress vectors as solid lines on faces with normals in the positive coordinate directions and as dotted lines on faces with normals in the negative coordinate directions. y dx τ xy + dz ∂τ xy dx ∂x --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- dy σx + τ zx σx ∂σ x dx ∂x x τ xy τ zx + z ∂τ zx dx ∂x Figure 2.6.4 For example, on the left face of the cube we have σx (0, y, z) and on the right face we have σx (0 + dx, y, z). We do a Taylor’s series expansion and neglect higher order terms. σx (0 + dx, y, z) = σx (0, y, z) + ∂σx d x + higher order terms ∂x (2.6.6) Also on those same two faces we have τzx (0 + dx, y, z) = τzx (0, y, z) + ∂τzx dx ∂x τx y (0 + dx, y, z) = τx y (0, y, z) + ∂τx y dx ∂x (2.6.7) A body force, that is, a force per unit volume, may also be acting on the body. It is designated by f b (x, y, z). An example of a body force is the force due to gravity or an inertial force caused by Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 67 Static Equilibrium acceleration. In many, but not all, of our examples this body force is small enough compared to the other forces present that it may be ignored. Similar expansions can be written for the faces of constant y and z as well. In Figure 2.6.5 we show all of the components in the x direction. y τ xy + dx ∂τ xy dy ∂y τ zx dz dy σx + f bx σx τ zx + ∂σ x dx ∂x x ∂τ zx dz ∂z τ xy z Figure 2.6.5 τx y (x, 0 + dy, z) = τx y (x, 0, z) + ∂τx y dy ∂y (2.6.8) τzx (x, y, 0 + dz) = τzx (x, y, 0) + ∂τzx dz ∂z (2.6.9) and For equilibrium, summation of forces in the x direction results in      ∂τx y ∂σx dx dydz − σx dydz + τx y + dy dxdz − τx y dxdz Fx = σx + ∂x ∂y   ∂τzx dz dxdy − τzx dxdy + f bx dxdydz = 0 + τzx + ∂z (2.6.10) This simplifies to ∂τx y ∂τzx ∂σx + + + f bx = 0 ∂x ∂y ∂z (2.6.11) Similar expansions and summations in the y and z directions result in ∂σ y ∂τ yz ∂τx y + + + f by = 0 ∂x ∂y ∂z ∂τ yz ∂σz ∂τzx + + + f bz = 0 ∂x ∂y ∂z Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (2.6.12) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- These include the expansions P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas 68 July 2, 2011 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods This may be put in matrix form as shown in Equation 2.6.13. ⎡ ⎤ ⎡ ∂ ∂ ∂ ⎤ σx 0 0 0 ⎢ ⎥ ⎢ ∂x ∂y ∂z ⎥ σy ⎥ ⎥⎢ ⎢ ⎡ ⎤ ⎥ ⎥⎢ ⎢ f bx ⎥ ⎥⎢ ⎢ ⎢ ⎥ σ ∂ ∂ ∂ z ⎥ ⎢ ⎥ ⎥ = −⎢ 0 0 ⎥⎢ [E] {σ } = ⎢ 0 ⎣ f by ⎦ = − { f b } ⎥ ⎥⎢ ⎢ ∂y ∂x ∂z τ ⎢ ⎥ x y ⎥⎢ ⎢ ⎥ f bz ⎥⎢ ⎢ ⎥ ⎣ ∂ ∂ ∂ ⎦ ⎣ τ yz ⎦ 0 0 0 τzx ∂z ∂y ∂x (2.6.13) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- These are the three dimensional differential equations of equilibrium in terms of the stress components in the interior of the body. These equations must be satisfied to obtain what we call an exact solution for the stress distribution in a three dimensional solid body. ########### Example 2.6.2 Problem: A cube of material is subject on all sides to a uniform hydrostatic pressure as shown in Figure (a). The same pressures are on all six faces. What are the stresses in the cube? p p p Figure (a) Solution: A possible solution would be uniform normal stress in all directions equal to the applied pressure. Give it a try. Substitute this assumed solution in the three dimensional equations of equilibrium to see if it satisfies equilibrium. Let σ x = σ y = σz = − p τx y = τ yz = τzx = 0 Substitute the values in Equation (a) into Equation 2.6.13. ⎤ ⎡ ⎡ ∂ ∂ ∂ ⎤ −p 0 0 0 ⎥ ⎢ ⎢ ∂x ∂y ∂z ⎥ −p ⎥ ⎢ ⎥⎢ ⎡ ⎤ ⎥ ⎡0⎤ ⎢ ⎢ ⎥⎢ f bx ⎢ ⎥ ⎢ −p ⎥ ⎥ ∂ ∂ ∂ ⎢ ⎥ ⎢ ⎥ ⎥ ⎥=⎢ 0 0 ⎥⎢ ⎢ 0 ⎥ ⎣ 0 ⎦ = − ⎣ f by ⎦ ⎢ ⎥⎢ ∂y ∂x ∂z 0 ⎥ ⎢ ⎢ ⎥⎢ ⎥ 0 f bz ⎢ ⎥⎢ ⎥ ⎣ ∂ ∂ ∂ ⎦⎣ 0 ⎦ 0 0 0 0 ∂z ∂y ∂x Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) (b) P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas July 2, 2011 14:50 Printer Name: Yet to Come 69 Static Equilibrium Since the derivative of a constant is zero it is seen to be a solution if no body force is applied. And the stress state does match the applied loads at the boundary faces. Both equilibrium and boundary conditions are satisfied. ########### 2.6.2 Equilibrium in Two Dimensions—Plane Stress In many cases the applied forces and restraint forces all lie within the same plane. This often happens when a thin sheet is loaded by in-plane edge loads. This state of stress is called plane stress. The plane stress equations of equilibrium are obtained from Figure 2.6.2, in which a small element of dimensions dx by dy and thickness h is shown under a state of plane stress. σy+ ∂σ y ∂y dy τ xy + ∂ τ xy dx ∂x f by σx+ f bx σx y ∂σ x dx ∂x τ xy σy x Figure 2.6.2 Equilibrium, from summation of forces in the x and y directions, results in      ∂τx y ∂σx dx hdy − σx hdy + τx y + dy hdx − τx y hdx + f bx hdxdy = 0 Fx = σx + ∂x ∂y      ∂σ y ∂τx y dy hdx − σ y hdx + τx y + dx hdy − τx y hdy + f by hdxdy = 0 Fy = σ y + ∂y ∂x (2.6.14) This reduces to ∂τx y ∂σx + + f bx = 0 ∂x ∂y ∂σ y ∂τx y + + f by = 0 ∂x ∂y (2.6.15) ⎡ ∂ ⎢ ∂x [E] {σ } = ⎢ ⎣ 0 0 ∂ ∂y ∂ ∂y ∂ ∂x ⎤⎡ σx ⎤ ⎥⎢ ⎥ ⎣ σy ⎥ ⎦=− ⎦ τx y  f bx f by  = − { fb } (2.6.16) One may conclude that a uniform stress on the boundary would produce a uniform stress in the interior. We shall examine this in more detail in Chapter 4. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The plane stress equilibrium equations can be written in matrix form as P1: TIX/XYZ JWST071-02 P2: ABC JWST071-Waas 70 July 2, 2011 14:50 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 2.6.3 Equilibrium in One Dimension—Uniaxial Stress In an even more simple case, all of the applied forces and restraint forces lie along a single line and there is stress in only one direction. This stress state is called uniaxial stress and is shown in Figure 2.6.3. fbx σx σx + ∂σ x dx ∂x dx Figure 2.6.3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The one dimensional equation of equilibrium is found by summing forces.    ∂σx dx hdy − σx hdy + f bx hdxdy = 0 Fx = σx + ∂x (2.6.17) This reduces to ∂σx = − f bx ∂x 2.7 (2.6.18) Summary and Conclusions We have applied the equations of static equilibrium to a number of cases. We have limited the geometry to simplified or idealized bodies. Geometry simplification has come in the form of two forces members, cables, springs, and slender, rigid bars. These topics will be expanded in subsequent chapters. Finally, we have examined the equilibrium of internal forces in terms of the six components of stress. The task before us is to find a connection between the external forces and restraints acting on the body and the stress state within the body. Trying to satisfy the equations of equilibrium for the stresses within the body for different combinations of external forces and restraints can be quite difficult. We understand the relationship between external forces and force resultants, but we don’t yet understand the exact stress distribution itself. Fortunately, by simplifying geometry, loading, and restraints it is possible to work with one and two dimensional equations in many practical situations. Examples of this simplified approach make up most of the rest of this text; however, we shall touch base with some three dimensional examples to justify the simplifying assumptions. Before trying to solve any problems of deformable bodies we need some further definitions and equations. That will be the subject of the next chapter. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 3 3.1 Introduction We now provide a brief look at the representation of displacements and deformation of the solid body. We know from experience that solid bodies are not truly rigid. The deformation may be small and not visually perceptible, such as when we walk across a floor, or quite perceptible, such as when we jump from a diving board. Our use of rigid body equilibrium equations assumes that the displacements are so small they do not affect the equations, that is, we can calculate the effect of the loads using the undeformed geometry. This is a central assumption in problems of linear kinematics, which is relaxed when we consider buckling problems in later chapters. We shall soon learn that knowing information about displacements is essential to solving indeterminate structures. The remaining chapters treat this in great detail. In the process of finding displacements it is desirable to define a quantity called strain. Strain is essentially a measure of the rate of change of displacement with respect to position. At first you may not see why it is necessary to define strain but it soon becomes obvious when material properties are discussed. For a given geometry we are well aware that the size of the displacement is related to the size of the load. We are also aware that the size of the displacement for a given load and geometry will depend upon the material. Under a given tensile load a rubber band will stretch much more than a steel wire of the same cross section area. How we represent material properties is introduced later in this chapter. 3.2 Displacement and Strain When a restrained body is acted upon by external forces it deforms and internal forces and stresses are generated. When a body deforms the distance between any two particles within the body changes, that is, the positions of particles within the body, in general, will change by different amounts. If all points in the body change their location by the same amount the body displaces without deforming. This is referred to as rigid body displacement. To solve certain problems it is necessary to find relations between the displacements and the internal stresses. To do this directly is not easy. An analysis of the deformation of the body through the use of strain facilitates the connection between stress and displacement. Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Displacement, Strain, and Material Properties P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 72 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 3.2.1 Displacement First consider a three dimensional solid body at rest and unloaded but suitably restrained from having any rigid body motion. Then, as it is loaded, it will deform. Consider a point within the body with coordinates (x, y, z) in the undeformed state. After deformation this point will have a new position coordinate (x∗ ,y∗ ,z∗ ). If the displacement components u(x,y,z), v(x,y,z), and w(x,y,z), are introduced in the x, y, and z directions then x* = x + u y* = y + v z* = z + w (3.2.1) Throughout this text we shall assume that the displacements are so small that they will have negligible effect on the location of loads and restraints, that is, changes in dimensions due to displacements caused by external loads are ignored unless otherwise stated. This is typical of most, but not all, situations that we shall consider. At the beginning of our deliberations the displacements, generally, are unknown. Finding them in particular situations is a major part of the task before us. 3.2.2 Strain The amount of deformation for any given applied load and set of restraints depends upon the properties of the material of the solid body. Some materials resist deformation more than others, for example, steel compared to soft rubber. To discuss this resistance to deformation it is convenient, indeed necessary, to define a deformation related quantity called strain. So far the quantities we have discussed – force, force per unit area (stress or at least pressure), and displacement are familiar from studies in physics. Unless you have had previous study in the behavior of solid bodies you may find strain less intuitive. Its value will soon become clear. For an introductory and elementary look at strain we find it convenient to consider a geometrically simple solid body, namely, a slender bar with a uniform cross section shown unloaded and undeformed by the shaded image in Figure 3.2.1. Away for the ends of the bar we scribe two lines a distance L apart. Now we load it on the ends by equal and opposite forces and note that it elongates by the amount shown by the unshaded extensions at each end. Next we measure the new distance between the scribed lines and note that the lines are now an additional distance apart, d. F F L L+d Figure 3.2.1 We call the ratio of this change in length to the original length as the strain, ε, that is, ε= d L (3.2.2) In a more general case of loading we assume that the distributed displacement in the x direction can be represented by a single independent variable, the displacement u(x). Given a quantity u(x), we can define a quantity called the normal strain, ε x , at any point along the bar. Consider the bar in Figure 3.2.2 and the displacement at x and at a point x + x as shown in the figure. --`,,,,``,``,,``, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 73 Displacement, Strain, and Material Properties x ∆x L Figure 3.2.2 At any point along the bar the local value of strain can be defined as εx = lim x→0 du u(x + x) − u(x) u = lim = x→0 x x dx (3.2.3) If we take a more general example, namely, a three dimensional body under the action of normal stress in all three coordinate directions by a similar argument we obtain εx = lim x→0 ∂u u = x ∂x ε y = lim y→0 ∂v v = y ∂y εz = lim z→0 ∂w w = z ∂z (3.2.4) It is observed that when a body is subjected to shearing stresses it will distort in another way. Consider an infinitesimal cubical element imbedded in a solid body and subjected to shearing stress τ xy . Once again we start with a geometrically simple body. The imposed shear stresses are shown in Figure 3.2.3 only in the xy directions, that is, there are no z direction components of shear stress in this example. y z τxy x Such an element will distort under small displacement. Looking at it in the xy plane as shown in Figure 3.2.4 we see the body before and after strain. The shear strain or shearing strain is defined as the change in angular orientation of points on the body. In Figure 3.2.4 after deformation the line elements AC and AD have rotated through the angles γ1 and γ2 to the positions A’C’ and A’D’. Since the derivatives of the displacements are small, these angles are approximately equal to the sine of the angle. Thus γ1 = ∂u ∂y γ2 = ∂v ∂x (3.2.5) It may be seen that the decrease in the right angle formed by AC and AD when the body deforms to the position A’C’ and A’D’ is equal to the shearing strain. γx y = γ1 + γ2 = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ∂v ∂u + ∂y ∂x Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (3.2.6) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure 3.2.3 P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 74 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods u+ ∂u dy ∂y B’ C’ γ1 B C D’ A’ υ+ γ2 v A ∂υ dx ∂x D Figure 3.2.4 In a similar way the shearing strains in the yz and zx planes may be found to be γ yz = ∂v ∂w + ∂y ∂z γzx = ∂w ∂u + ∂z ∂x (3.2.7) Equations 3.2.4, 3.2.6, and 3.2.7 collectively are called the strain displacement relations. There are six independent components of strain since γx y = γ yx γ yz = γzy γzx = γx z (3.2.8) The strains are variously arranged in matrix form as a column or square matrix according to their use in equations. ⎤ ⎡ εx ⎢ εy ⎥ ⎡ ⎤ ⎥ ⎢ εx γx y γzx ⎢ εz ⎥ ⎥ ⎢ ⎣ ⎦ {ε} = ⎢ (3.2.9) ⎥ or [ε] = γx y ε y γ yz ⎢ γx y ⎥ γ γ ε zx yz z ⎣ γ yz ⎦ γzx The matrix form of the strain displacement relations for a three dimensional solid is ⎡ ⎤ ∂ 0 0 ⎢∂x ⎥ ⎢ ⎥ ⎢ ⎥ ∂ ⎥ ⎤ ⎢ 0 ⎡ 0 ⎢ ⎥ εx ∂ y ⎢ ⎥ ⎢ εy ⎥ ⎢ ∂ ⎥⎡ ⎤ ⎥ u ⎥ ⎢ 0 ⎢ 0 ⎢ εz ⎥ ⎢ ∂z ⎥ ⎢ ⎥ ⎣ v ⎦ = [D] {u} ⎥ ⎢ {ε} = ⎢ ⎥ ⎥=⎢ ∂ ∂ ⎥ w ⎢ γx y ⎥ ⎢ 0 ⎥ ⎣ γ yz ⎦ ⎢ ∂ y ∂ x ⎢ ⎥ ⎢ ∂ ∂ ⎥ γzx ⎢ 0 ⎥ ⎢ ∂z ∂ y ⎥ ⎢ ⎥ ⎣ ∂ ∂ ⎦ 0 ∂z ∂x Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (3.2.10) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- u P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 75 Displacement, Strain, and Material Properties When a thin sheet is acted upon only by in plane loads there are no stresses in, say, the z direction. As noted in Section 2.6 this is called plane stress. The corresponding strain displacement equations are ⎤ ⎡ ∂ 0 ⎥ ⎡ ⎤ ⎢∂x ⎥ ⎢ εx ∂ ⎥ u ⎢ ⎥ 0 {ε} = ⎣ ε y ⎦ = ⎢ (3.2.11) = [D] {u} ⎢ ∂y ⎥ ⎥ v ⎢ γx y ⎣ ∂ ∂ ⎦ ∂y ∂x For a slender bar loaded in one dimension the equations become εx = du dx (3.2.12) ########### Example 3.2.1 Problem: A 100 mm square thin steel plate is deformed from the shape ABCD to A’B’C’D’ indicated by the dotted lines as shown in Figure (a). Determine εx , ε y , γx y . y 0.1 mm C B C’ B’ x A, A’ D D’ Figure (a) Solution: The normal strain is the change in length divided by the length. The shearing strain is the change in angle between two lines at right angles. Comparing the geometry of the deformed part to the original we obtain εx = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 0.2 A′ D′ − AD = = 0.002 AD 100 0.1 A′ B′ − AB εy = =− = −0.001 AB 100 γx y = angle BAB′ + angle DAD′ = 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 0.2 mm P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 76 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Example 3.2.2 Problem: A 100 mm square thin steel plate is deformed from the shape ABCD to A’B’C’D’ indicated by the dotted lines as shown in Figure (a). Determine εx , ε y , γx y . y 0.4 mm 0.1 mm C’ B C B’ 0.1 mm D’ D A, A’ 0.1 mm x 0.3 mm Figure (a) Solution: The normal strain is the change in length divided by the length. The shearing strain is the change in angle between two lines at right angles. Comparing the geometry of the deformed part to the original we obtain εx = A′ D′ − AD = AD (99.9)2 + (0.1)2 − 100 = −0.001 100 0.1 0.1 + tan−1 = angle BAB′ + angle DAD′ = tan−1 99.9 100.3 εy = --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- γx y (100.3)2 + (0.1)2 − 100 = 0.003 100 A′ B′ − AB = AB (a) = 0.002 rad ########### 3.3 Compatibility There are six components of strain defined in term of three components of displacement. From that you may conclude that the strain components are not independent of each other. Consider the two dimensional case expressed in Equation 3.2.12. Expanded we have εx = ∂u ∂x εy = ∂v ∂y γx y = ∂v ∂u + ∂y ∂x (3.3.1) Take the second derivatives of each strain component in the following way. ∂ 3u ∂ 2 εx = 2 ∂y ∂ x∂ y 2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ∂ 3v ∂ 2 εy = 2 ∂x ∂ x 2∂ y ∂ 3u ∂ 3v ∂ 2 γx y = + 2 2 ∂ x∂ y ∂ x∂ y ∂x ∂y (3.3.2) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 77 Displacement, Strain, and Material Properties By summing the first two equations in Equation 3.3.2 we note from the third equation that the following relation exists among the strain components. ∂ 2 εy ∂ 2 γx y ∂ 2 εx + = ∂ y2 ∂x2 ∂ x∂ y (3.3.3) This is called the compatibility equation in two dimensions. When applied to the three dimensional case additional relationships are found. We shall not go through the steps but only record that in three dimensions there are five more relations that must be satisfied. (3.3.4) We note that in both two and three dimensions the compatibility equations consist of relations among the second derivatives of the strain components with respect to the coordinates. It follows that any state of strain involving constant values of strain components satisfies compatibility and so do cases of linearly varying strain components. Whatever the state of strain in a body, to be a valid solution of a given problem, compatibility must be satisfied. We shall use this fact to advantage in several instances in later chapters. 3.4 Linear Material Properties There are two physical laws that govern the behavior of solid bodies under load. One of these is static equilibrium, discussed in detail in Chapter 2. The other is based on the properties of materials. When a material is homogeneous it has properties that do not change from point to point. When it is isotropic it has the same properties in all possible coordinate directions and it often has the simple properties that are about to be defined. These properties apply to many, but not all, materials that solid bodies used in structures are made of. The three most commonly used aerospace metals, aluminum, steel, and titanium, are both homogenous and isotropic; however, modern materials like carbon reinforced composites have more involved properties and will be dealt with later in this chapter. 3.4.1 Hooke’s Law in One Dimension—Tension Material properties are determined initially by a simple experiment. Consider that same slender bar in Figure 3.2.1 used to define strain. The force is applied in a manner yet to be determined but assume, for the moment, that it acts through the centroid of the cross section. If we examine a cross section normal to the axis at, say, position 0 < x < L, there must be an internal force P acting on that cross section that satisfies Newton’s law of equilibrium, namely, that the summation of forces is zero. We illustrate this with a free body diagram, that is, we consider a portion of the bar in Figure 3.2.1 as shown in Figure 3.4.1. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ∂ 2 εz ∂ 2 γ yz ∂ 2 εy + = ∂z 2 ∂ y2 ∂ y∂z ∂ 2 εx ∂ 2 γx z ∂ 2 εz + = ∂x2 ∂z 2 ∂z∂ x 2 ∂ ∂γx z ∂γx y ∂γ yz ∂ εx = + + − 2 ∂ y∂z ∂x ∂x ∂y ∂z ∂ ∂γ yz ∂γx z ∂γx y ∂ 2εy = − + 2 ∂z∂ x ∂y ∂x ∂y ∂z ∂ ∂γ yz ∂γx z ∂γx y ∂ 2 εx = + − 2 ∂ x∂ y ∂z ∂ x ∂y ∂z P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 78 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods F P(x) x L Figure 3.4.1 As noted before, it is common to define applied force components and restraint force components as positive if acting in the positive directions of the axes. So F is shown as a positive force. Two different conventions are used for internal forces. The most common has been to define normal force components acting away from the surface as positive, as P is shown in Figure 3.4.1. By this convention positive stresses produce positive internal force resultants. The occasion will arise when we shall use another convention but for now we shall use the one just stated. Another convention relates to the direction of the y axis in the figures. Civil engineers, in particular, are inclined to show the axis pointing down. Since applied loads in civil engineering tend to be gravity loads they can then be positive pointing down in the same direction of the positive y axis. We have chosen to have it pointing up, perhaps because of our aerospace background. Airplanes are particularly pleased when some of the forces acting upon them are acting up. From Newton’s laws the summation of forces in the x-direction must be zero. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Fx = F − P(x) = 0 → P=F (3.4.1) In this case P(x) is a constant for all values of x. The force P, however, must be distributed over the cross sectional area. Let us assume, for now, that it is uniformly distributed over that area. This is justified by experimental evidence. We can denote the force per unit area or stress by σ x . The stress is σx = P A P = σx A → (3.4.2) Note that it makes no difference what the material of the bar is. For any applied force F there will be an internal force P and for any area A the stress will be given by Equation 3.4.2. Note again that this applies only if the force is acting through the centroid of the cross sectional area. We would expect, and we can verify by experiment, that the bar will deform. In this case the ends would displace some distance. Just what distance would depend upon the material. You can readily agree that if the bar was made of soft rubber it would displace much more than if made of steel for a given value of F and A. To find the displacement then we must find some way to describe the properties of the material being used. We might, for example, actually construct a bar of a material of interest and measure the displacement of two scribe marks just as we did in defining strain. For many materials, such as those commonly used in structures of many kinds (for example, steel, aluminum, or titanium) we would obtain a plot similar to that shown in Figure 3.4.2. There would be a straight portion, that is, a linear relation between force and displacement, followed by a region in which the displacement increases more rapidly with respect to an increase in force. For many materials used in real structures the deflection returns to zero if the load is reduced to zero provided the maximum load does not exceed a certain value. Quite often this value is very near where the curve deviates from a straight line. Materials that behave this way are called elastic, or more precisely linearly elastic, up to the point where the curve is no longer straight. The straight portion of this plot up to a value of the force F 0 has the slope k where k= Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS F d (3.4.3) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 79 Displacement, Strain, and Material Properties F F0 linear elastic d Figure 3.4.2 The quantity k is called the stiffness of the bar. Given that F ≤ F0 and a value for k we can say d= F k (3.4.4) The greater the stiffness the less is the displacement for a given force. The same concept is used to define the stiffness of springs where k is called the spring constant. For the moment we shall restrict our interest to loads that do not exceed F 0 . In a later chapter we shall discuss what happens when larger loads are applied. This defines the principal material property over the linearly elastic range for this particular structure; however, if we change the cross sectional area A or the length L of the bar between scribe marks the test results would provide a different value of k. So we need a more general way to define the properties of the material that is independent of the geometry of the solid bar. Fortunately, a way has been found. We define the quantity εx = d du = dx L (3.4.5) This is our old friend the normal strain. If the stress is uniform across the cross section we can define the normal stress as σx = F P = A A (3.4.6) We will show this to be true in Chapter 4. Now the same data used to plot the curve in Figure 3.4.2 can be represented as shown in Figure 3.4.3. σx σ0 = F0 A Yield stress Proportional limit εx Figure 3.4.3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas 80 July 2, 2011 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The slope of the straight portion is designated by E. It follows that σx εx or σx = Eεx or εx = σx E (3.4.7) In this case the plot is independent of A and L and is representative of the material and independent of the particular geometry of the slender bar. Thus it has general validity for that material for any value of the length between scribe marks, L, or of the cross section area, A. The straight portion is an example of Hooke’s law for elastic materials. The quantity E is called Young’s modulus. We will study this in more detail later. The quantity σ0 is called the proportional limit. Beyond this point the material may or may not be elastic. For many metals a short distance beyond the proportional limit the material starts to have permanent deformation. That point it called the yield point or yield stress. Often the two points are so close together that the material is assumed to be elastic up to the yield stress and it is taken as the value at which our assumption of linear elasticity no longer applies. If the material we pulled on was amorphous glass or carbon fiber the stress-strain curve would be a straight line up to the point where it fractures with no nonlinear portion as shown in Figure 3.4.4. This is characteristic of brittle materials while Figure 3.4.3 is characteristic of ductile materials. σx εx Figure 3.4.4 When the stress is positive it is called a tensile stress, when negative a compressive stress. For many materials of interest the slope of the curve, E, is the same in the negative stress and strain region as in the positive region. For future reference we note that the total displacement in our test specimen between scribe marks is d = εx L = PL FL σx L = = E AE AE (3.4.8) ########### Example 3.4.1 Problem: Consider that the bar in Figure 3.2.1 is made of aluminum. For the aluminum alloy 6010-T6 the proportional limit is approximately 250 N/mm2 (MPa) and E = 68950 N/mm2 . The dimension between scribe marks is L = 200 mm and the cross section area is A = 4 mm2 . The applied load is F = 500 N. Find the stress and the total displacement between marks. Solution: Use Equations 3.4.6 – 3.4.8. The stress is σx = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 500 P = = 125 N/mm 2 A 4 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- E= P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 81 Displacement, Strain, and Material Properties Since this value is below the proportional limit the displacement would be d= 500 · 200 FL = = 0.3626 mm AE 4 · 68950 (b) The fact that the displacement is a small quantity when compared to the length of the bar is important in certain assumptions to be made in the development of the equations of mechanics of materials. In fact, this simple example points out what we are usually looking for in an analysis – the size of the stress and of the displacement. Most structures are said to fail when either one or both of these exceed acceptable limits. ########### 3.4.2 Poisson’s Ratio A careful observer of the experimental determination of Young’s modulus will note that the slender bar’s cross section narrows as the bar elongates under load. This is shown in side view in Figure 3.4.5(a) and in end view in Figure 3.4.5b with the change in dimensions greatly exaggerated. The shaded bar shows the original dimensions. The bar has a width b and a height h in the unloaded state. y F F x L L+dx (a) y h h - dy z b b - dz --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- (b) Figure 3.4.5 So far we have neglected this effect. We designate the amount of elongation as dx and of lateral displacement d y in the y direction and dz in the z direction. The corresponding strains are εx = dx L εy = − dy h εz = − dz b (3.4.9) It may be noted that for a linear elastic, homogenous, and isotropic material experimental evidence will confirm that εz εy =− =ν (3.4.10) − εx εx Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 82 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods where ν is a constant called Poisson’s ratio, named after the person who defined it. The negative sign accounts for lateral contractions in the y and z directions for a positive elongation in the x direction when a positive tensile force is applied. From this we may conclude that if εx = σx E (3.4.11) then from Equation 3.4.10 ε y = −νεx = − νσx E εz = −νεx = − νσx E (3.4.12) If there is a contraction in y and z directions the area of the cross section must be smaller than the nominal value that we have been using to determine the stress. We turn to an example to explore this effect. ########### Example 3.4.2 Problem: Find the Poisson’s ratio effect on the magnitude of the stress for the bar in Example 3.4.1. In a typical material, such as aluminum, the value of Poisson’s ratio is approximately ν = 0.3. Width and height of the cross section is b = h = 2 mm. Solution: Find the change in cross section area and calculate the stress based on the new area. The new cross section dimensions are bnew = b + εz b = b − νP 0.3 · 500 νσx b=b− b =2− · 2 = 1.9989 = h new E AE 4 · 68950 (a) Thus the new area is Anew = bnew · h new = (1.9989)2 = 3.9956 mm 2 (b) and the stress based on this area is σxnew = 500 P = 125.138 N/mm 2 = Anew 3.9956 (c) This may be compared to the stress value of 125 N/mm 2 found in Example 3.4.1. Such a small change justifies the use of nominal dimensions in nearly all analysis using most common structural materials. ########### 3.4.3 Hooke’s Law in One Dimension—Shear in Isotropic Materials When an experiment is performed that produces a pure shear stress it produces a curve similar to the one in Figure 3.4.2. When the shearing stress, τ , is plotted versus the shearing strain, γ , a curve very similar to Figure 3.4.3 is produced. The slope of the stress strain curve is designated, G, and is called the shear modulus. The corresponding Hooke’s law for shear is then τ = Gγ (3.4.13) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 83 Displacement, Strain, and Material Properties We shall show in Chapter 9 that G is not independent of E and ν , in fact, G= E 2(1 + ν) (3.4.14) For now we shall accept the experimental evidence confirms both Equations 3.4.13 and 3.4.14. 3.4.4 Hooke’s Law in Two Dimensions for Isotropic Materials Some structural components consist of thin flat plates with loads and restraints applied in the plane of the plate. This is called a state of plane stress. The stress components at a point in the interior of the plate are as shown in Figure 3.4.6. y σy τxy σx σx x τxy σy Figure 3.4.6 In such cases all z component of stresses are zero. σz = τ yz = τzx = 0 (3.4.15) The σx normal stress produces the same strains as given in Equations 3.4.11 and 3.4.12, namely, εx = σx E εy = − νσx E εz = − νσx E (3.4.16) Similarly the σ y normal stress produces σy E εx = − νσ y E εz = − νσ y E (3.4.17) These two can be combined to provide  1  σx − νσ y E  1  σ y − νσx εy = E  −ν  εz = σ y + σx E εx = and to this we add the shearing strains. γx y = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 1 τx y G (3.4.18) (3.4.19) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- εy = P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 84 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The first two of Equations 3.4.18 and Equation 3.4.19 can be put in the matrix form ⎤ ⎡ 1 −ν 0 ⎥⎡ ⎤ ⎢ E ⎤ ⎡ E ⎥ σx ⎢ εx ⎥ ⎢ −ν 1 ⎣ ⎦ ⎣ εy ⎦ = ⎢ 0⎥ ⎥ σ y = [Ŵ] {σ } ⎢ E E ⎥ τx y ⎢ γx y ⎣ 1⎦ 0 0 G (3.4.20) The matrix [Ŵ] can be inverted to put these equations in a more useful form. [Ŵ]−1 1 ⎢ E ⎢ ⎢ −ν =⎢ ⎢ E ⎢ ⎣ 0 ⎡ −ν E 1 E 0 ⎤−1 ⎡ E 0⎥ ⎢ 1 − ν2 ⎥ ⎢ ⎥ ⎢ νE 0⎥ ⎥ =⎢ ⎢ 1 − ν2 ⎥ ⎣ 1⎦ 0 G Hooke’s law for plane stress can now be written as ⎡ E νE ⎡ ⎤ ⎢ 1 − ν2 1 − ν2 σx ⎢ ⎢ νE E {σ } = ⎣ σ y ⎦ = ⎢ ⎢ 1 − ν2 1 − ν2 τx y ⎣ 0 0 νE 1 − ν2 E 1 − ν2 0 ⎤ 0⎥ ⎥ ⎥ = [G] 0⎥ ⎥ ⎦ G ⎤ 0⎥⎡ ⎤ ⎥ εx ⎥⎣ ε y ⎦ = [G] {ε} 0⎥ ⎥ ⎦ γx y G (3.4.21) (3.4.22) The notation here can be confusing. The shear modulus is labeled G and is a scalar quantity. The matrix [G] relates all components of stress to strain. We note that εz is not zero; therefore the thickness of the plate at any point is dependent upon the state of stress at that point. εz = −  v  σx + σ y E (3.4.23) For most structural materials the effect is small so that the nominal thickness of the plate is used in all calculations with minimal effect, that is, the z component of strain is neglected. 3.4.5 Generalized Hooke’s Law for Isotropic Materials If a material is subjected at a point to triaxial stress components we add the z components of strain to the components listed in Equations 3.4.16-17. εz = σz E εy = − νσz E εx = − νσz E (3.4.24) 1 E 1 εy = E 1 εz = E εx = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS      σ x − ν σ y + σz  σ y − ν (σx + σz ) (3.4.25)   σz − ν σ y + σ x Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- When all three stress components are acting we can superimpose the above to obtain P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 85 Displacement, Strain, and Material Properties When a homogenous, isotropic, and linear elastic material is subjected at a point to all three shearing stress components we have γx y = This can be put in matrix form ⎡ 1 ⎢ E ⎢ ⎢ −ν ⎤ ⎢ ⎡ ⎢ E εx ⎢ ⎢ ε y ⎥ ⎢ −ν ⎥ ⎢ ⎢ ⎢ εz ⎥ ⎢ E ⎥ ⎢ {ε} = ⎢ ⎢ γx y ⎥ = ⎢ ⎥ ⎢ ⎢ ⎣ γ yz ⎦ ⎢ 0 ⎢ ⎢ γzx ⎢ 0 ⎢ ⎢ ⎣ 0 1 τx y G γ yz = −ν E 1 E −ν E −ν E −ν E 1 E 0 1 τ yz G 0 0 0 0 0 0 0 1 G 0 0 0 0 1 G 0 0 0 0 where G= γzx = 1 τzx G ⎤ 0⎥ ⎥ ⎥ ⎡ ⎤ 0⎥ ⎥ σx ⎥ ⎥ ⎢ σy ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢ σz ⎥ ⎥⎢ ⎥ ⎥ ⎢ τx y ⎥ = [Ŵ] {σ } ⎢ ⎥ 0⎥ ⎥ ⎣ τ yz ⎦ ⎥ ⎥ τzx 0⎥ ⎥ ⎥ 1⎦ G E 2 (1 + ν) (3.4.26) (3.4.27) (3.4.28) The matrix [Ŵ] can be inverted to put these equations in a more useful form. [G] = [Ŵ]−1 0 0 0 G 0 0 0 0 0 0 G 0 where λ= 3.5 ⎤ ⎤⎡ 0 εx ⎢ ⎥ 0⎥ ⎥ ⎢ εy ⎥ ⎢ εz ⎥ 0⎥ ⎥⎢ ⎥ = [G] {ε} ⎢ ⎥ 0⎥ ⎥ ⎢ γx y ⎥ 0 ⎦ ⎣ γ yz ⎦ γzx G νE (1 + ν) (1 − 2ν) (3.4.30) (3.4.31) Some Simple Solutions for Stress, Strain, and Displacement To find a valid solution for the stress strain and displacement in a solid body in the linear elastic range we must satisfy 1. 2. 3. 4. The equations of equilibrium. The strain displacement equations consistent with compatibility. Hooke’s law for material properties. Boundary conditions on loading and restraint. We shall examine some simple examples here. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Hooke’s law for a 3D solid can now be written as ⎤ ⎡ ⎡ λ + 2G λ λ σx ⎢ σy ⎥ ⎢ λ λ + 2G λ ⎥ ⎢ ⎢ ⎢ σz ⎥ ⎢ λ λ λ + 2G ⎥ ⎢ {σ } = ⎢ ⎢ τx y ⎥ = ⎢ 0 0 0 ⎥ ⎢ ⎢ ⎣ τ yz ⎦ ⎣ 0 0 0 τzx 0 0 0 (3.4.29) P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 86 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 3.5.1 Problem: A thin plate has two sets of uniform edge loads as shown in Figures (a) and (b). The coordinate system is attached to the center of the plate. Find the stresses and displacements. fb0 y fa0 b x fa0 a fb0 Figure (a) Figure (b) The loading in Figure (a) is  a  f ax ± , y = ± f a0 N/mm 2 2 (a) The loading in Figure (b) is f by x, ± b 2 = ± f b0 N/mm 2 (b) Solution: Assume the stresses are the same constant values as the edge loads and check to see if the stress components satisfy the equilibrium equations, match the loads at the boundaries, satisfy Hooke’s law, satisfy the compatibility equations, and then integrate the strains to find the displacements. Since this is a thin plate and all forces lie in the xy plane this may be treated as a plane stress problem. It is suggested that the stress components in the interior of the plate are the same as on the edges where the load is applied in each case. For the loading in Figure (a) try this for the stress. σx (x, y) = f a0 σ y (x, y) = τx y (x, y) = 0 (c) Substitute these values in the equilibrium equations. ⎡ ∂ ⎢∂x ⎢ ⎣ 0 0 ∂ ∂y ∂ ⎤⎡ ⎤  f a0 ∂y ⎥ ⎥⎣ 0 ⎦ = 0 ∂ ⎦ 0 0 ∂x --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 87 Displacement, Strain, and Material Properties In the absence of body forces they are satisfied. The stresses at the boundaries match the applied load. From Hooke’s law the strains are f a0 σx = E E εx = ε y = −ν f a0 σx = −ν E E γx y = 0 (e) The compatibility equation is satisfied as shown in Equation (f). ∂ 2εy ∂ 2 γx y ∂ 2 εx + = 2 2 ∂y ∂x ∂ x∂ y → ∂2 ∂ y2 f a0 E + ∂2 ∂x2 −ν f a0 E = (f) ∂2 (0) ∂ x∂ y → 0+0=0 To find the displacements integrate the strains. εx = f a0 du = dx E → f a0 E u (x, y) = → f a0 dv = −ν dy E u (x, y) = v (x, y) = −ν → → dx = f a0 x +a E f a0 0+a =0 E u (0, 0) = → εy =  f a0 E  v (0, 0) = −ν f a0 x E dy = −ν (g) f a0 y+b E f a0 0+b =0 E v (x, y) = −ν → a=0 → → b=0 f a0 y E For the loading in Figure (b), following the same steps, we find that σ y (x, y) = f b0 σx (x, y) = τx y (x, y) = 0 (h) satisfies equilibrium and the stresses match the applied loads at the edges. The strains are εy = f b0 σy = E E εx = −ν f b0 σy = −ν E E γx y = 0 (i) and they satisfy compatibility. The displacements are found to be --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- v (x, y) = f b0 y E u (x, y) = −ν f b0 x E (j) If both loads are applied at the same time the stresses and displacements are simply the sum of those found, that is, the principle of superposition applies. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 88 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Example 3.5.2 Problem: The same thin plate has varying edge loads as shown in Figure (a). y a f0 x2/a2 b x –f0 x2/a2 Figure (a) These loads are f y x, ± b 2 = ± f0 x2 N/mm 2 a2 (a) Once again it is suggested that the stresses in the interior be the same as those on the boundaries, that is, σ y (x, y) = f 0 x2 a2 σx = τx y = 0 (b) Are these correct values for the stress components? Solution: As noted, to be valid the stress components must satisfy the equilibrium equations, match the loads at the boundaries, satisfy Hooke’s law, and the compatibility equations. Insert the suggested stress components into the equilibrium equations and evaluate. ⎤ ⎡ ⎡ ∂ 0 ∂ ⎤ 0 x2 ⎥  0 ⎢∂x ∂y ⎥ ⎢ f0 2 ⎥ ⎥⎢ ⎢ (c) = ⎢ ⎣ ∂ ∂ ⎦⎣ b ⎥ 0 ⎦ 0 0 ∂y ∂x The left hand side is identically zero in the absence of body forces. The equilibrium equations are satisfied. Great! And the stresses at the boundaries do match the applied loads. Now find the strain components. εy = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS f0 x 2 σy = E E a2 εx = −ν f0 x 2 σy = −ν E E a2 γx y = 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 89 Displacement, Strain, and Material Properties Insert these in the compatibility equation and evaluate. ∂2 ∂ y2 f0 x 2 E a2 + ∂2 ∂x2 −ν f0 x 2 E a2 = ∂2 (0) ∂ x∂ y → 0− 2ν = 0 Ea 2 (e) Compatibility is not satisfied. This cannot be the stress in the plate. The actual solution is quite difficult to obtain and is beyond our scope at this time, but we shall find a way to solve this problem later on. ########### 3.6 Thermal Strain A change in temperature from an initial state can cause a material to expand or contract. An increase in temperature will cause an expansion and for many materials this change is linear with respect to temperature. For such materials we denote a coefficient of thermal expansion with the symbol α. It has units of strain per degree of temperature. In one dimension for a change in temperature T the thermal strain is (3.6.1) εx T = αT When there is a temperature change in addition to a strain caused by a mechanical load Equation 3.2.12 for the one dimensional case can be written as the superposition of the mechanical and thermal strains. εx = εxmech + αT = du dx (3.6.2) Since the mechanical strain in one dimension is given by Equation 3.4.7, that is, εxmech = σx E (3.6.3) we have σx du = + αT dx E (3.6.4) In two dimensions the mechanical strain in terms of stress is given by Equation 3.4.20. The effect of adding thermal strain results in ∂ ⎢∂x ⎢ ⎢ ⎢ [D] {u} = ⎢ 0 ⎢ ⎢ ∂ ⎣ ∂y ⎡ ⎤ 0 ⎥ ⎥ ∂ ⎥ ⎥ u ∂y ⎥ ⎥ v ∂ ⎥ ⎦ ∂x 1 ⎢ E ⎢ ⎢ −ν ⎢ =⎢ ⎢ E ⎢ ⎣ 0 ⎡ −ν E 1 E 0 ⎤ 0⎥ ⎡ ⎤ ⎤ ⎥⎡ 1 ⎥ σx ⎥ 0 ⎥ ⎣ σ y ⎦ + αT ⎣ 1 ⎦ ⎥ τx y 0 1⎥ ⎦ G --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (3.6.5) P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas 90 July 2, 2011 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In three dimensions the mechanical strain in terms of stress is given by Equation 3.4.27. The effect of adding thermal strain results in ⎡ ⎡ ⎤ ⎤ 1 −ν −ν ∂ 0 0 0 0 0 ⎢ E ⎢∂x ⎥ ⎥ E E ⎢ ⎢ ⎥ ⎥ ⎢ −ν ⎢ ⎥ ⎥ 1 −ν ∂ ⎢ ⎢ 0 ⎥ 0 0 0 ⎥ 0 ⎥ ⎢ E ⎢ ⎥⎡ ⎡ ⎤ ⎤ E E ∂ y ⎢ ⎢ ⎥ σx ⎥ 1 ⎢ ⎢ ⎥ ⎥ ∂ ⎥ ⎡ ⎤ ⎢ −ν −ν 1 ⎢ ⎥ ⎢ σy ⎥ ⎢1⎥ 0 0 0 ⎥⎢ 0 ⎢ ⎢ 0 ⎥ ⎢ ⎥ ⎥ ⎢ E ⎢ ⎥ ⎢ σz ⎥ ⎢ ⎥ ∂z ⎥ u E E ⎥⎣ v ⎦ = ⎢ ⎥⎢ ⎥ + αT ⎢ 1 ⎥ [D] {u} = ⎢ ⎢ ⎢ ∂ ⎥ ⎢ τx y ⎥ ⎥ ⎢0⎥ ∂ 1 ⎢ 0 ⎢ ⎢ ⎢ ⎥ ⎥ 0 ⎥ 0 0 ⎥ 0 0 ⎢ ⎢ ∂y ∂x ⎥ ⎣ τ yz ⎦ ⎥ w ⎣0⎦ G ⎢ ⎢ ⎥ ⎥ ⎢ ⎢ ⎥ τzx ⎥ ∂ ∂ 0 1 ⎢ 0 ⎢ 0 ⎥ 0 ⎥ 0 0 0 ⎢ ⎢ ⎥ ⎥ ∂z ∂ y ⎥ G ⎢ ⎢ ⎥ ⎢ ⎢ ⎥ ⎥ ∂ ⎦ 1 ⎦ ⎣ ⎣ ∂ 0 0 0 0 0 0 ∂z ∂x G (3.6.6) Some examples of the use of these equations that have been extended to include changes in temperature will be given in later chapters. 3.7 Engineering Materials Steel and aluminum are metallic materials often used in structures. Titanium is used in high performance structures where high temperatures are important. The properties depend upon the particular alloy selected. For exact values for these and other materials we must refer to handbooks and manufacturer’s data that list properties for particular alloys. For use in example problems in the following chapters we list properties in Table 3.7.1 for a generic steel, aluminum, and titanium that are representative of these three metals. Table 3.7.1 Material Density kg/mm3 Young’s Modulus MPa Shear Modulus MPa Yield Stress MPa Poisson’s Ratio Thermal Expansion (10−6 )/◦ F Steel Aluminum Titanium 0.0078 0.0028 0.0044 206800 68950 120000 79500 26500 46000 240 360 900 0.3 0.3 0.3 16 24 9 In particular non metallic and composite materials are coming into increasing use in many types of structures. A brief introduction to composite material properties is given in the next section. 3.8 Fiber Reinforced Composite Laminates Modern aerospace structures use a significant amount of fiber reinforced polymer (FRP) matrix composite materials. There are many constructions of FRP laminates that are made by stacking together several lamina and then curing them in an autoclave under specified pressure and temperature histories. Each lamina of a FRP may be woven (like a textile), braided (a more complex weave pattern), or simply consist of a collection straight fibers held within a matrix, such as shown in Figure 3.8.1. We designate the direction of the fiber as 1 and the two perpendicular directions as 2 and 3. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 91 Displacement, Strain, and Material Properties 1 3 2 matrix fiber Figure 3.8.1 Using straight fiber stacked lamina to make a laminate is popular in modern aerospace construction. In particular the Airbus 380, Boeing 777 and 787, and most military aircraft use a substantial amount of laminate materials. Typically, the lamina shown in Figure 3.8.1 is about 0.13 mm thick. Carbon fibers, which are most popular in the aerospace industry, are about 5-7 μm in diameter, and within each lamina fibers occupy about 50 percent of the volume. We will focus here on understanding the stress-strain behavior of a single lamina. Later, when we analyze plate problems we shall derive the equations to characterize the stress-strain behavior of a collection of stacked laminae, or a laminate. 3.8.1 Hooke’s Law in Two Dimensions for a FRP Lamina Consider a FRP lamina with loads and restraints applied in the plane of the lamina as shown in Figure 3.8.2. 2 σ2 τ12 σ1 σ1 1 τ12 σ2 Figure 3.8.2 This is a state of plane stress as we noted before. We shall treat the lamina as a homogeneous but orthotropic layer, with “1” denoting the fiber direction and “2” the transverse direction. The assumption --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 92 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods about homogeneity implies that we are treating the lamina as a single material with regard to the discrete fibers and matrix. The reason is that the dimension of the lamina we are interested in is several times greater than the fiber diameter or the spacing between fibers. The fiber diameter is usually about 5-7 μm for carbon fibers and about 20 μm for glass fibers. The spacing between fibers is usually about 5-7μm for the carbon fibers and about 15-20 μm for the glass fibers. If one has equal amounts of fiber and matrix by volume in a lamina we say the lamina has a 50 percent volume fraction (Vf ) of fiber. For plane stress σ3 = τ23 = τ31 = 0 (3.8.1) Consider what happens when we have σ1 acting alone. Then, unlike isotropic materials, the amounts by which the lamina will strain in the 2 and 3 directions, that is, the Poison’s ratio effect, will be different from the 1 direction. This is because the properties of the lamina in the fiber direction will be different than those perpendicular to the fiber direction. So we can write ε1 = σ1 E1 ε2 = −ν12 σ1 E1 ε3 = −ν13 σ1 E1 (3.8.2) However, for a unidirectional lamina the 2 and 3 directions are identical in constitution and so ν12 = ν13 . Equation 3.8.2 becomes ε1 = σ1 E1 ε2 = −ν12 σ1 E1 ε3 = −ν12 σ1 E1 (3.8.3) The notation for ν12 signifies that the loading is in the “1” direction and the contraction occurs in the “2” direction. Young’s modulus E1 is in the fiber or 1 direction. We can generalize Equation 3.8.3 to include other stress components, like we did before for isotropic materials, to obtain ε1 = σ1 σ2 − ν21 E1 E2 ε2 = −ν12 σ2 σ1 + E1 E2 γ12 = τ12 G 12 (3.8.4) Equations 3.8.4 are the plane stress-strain relations for an orthotropic lamina. Notice that we have used 5 material constants, E 1 , E 2 , G 12 , ν12 and ν21 to describe the plane stress –strain relation for an orthotropic material. It turns out that only four of these are independent, because, E2 E1 = ν12 ν21 (3.8.5) which is called the reciprocity relation. In general, with i, j ranging from 1 to 3 Ej Ei = νi j ν ji (3.8.6) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 93 Displacement, Strain, and Material Properties When we generalize Equation 3.8.4 to 3D without the plane stress assumption but still with fibers in only the 1 direction we get ⎡ ⎤ 1 ν21 ν21 − − 0 0 0 ⎢ E1 ⎥ E2 E2 ⎢ ⎥ ⎢ ν12 ⎥ ν (= ν ) 1 32 23 ⎥⎡ ⎤ ⎤ ⎢− ⎡ − 0 0 0 ⎢ E ⎥ σ1 ε1 E E 1 2 2 ⎢ ⎥ ⎥ ⎢ σ2 ⎥ ⎢ ε2 ⎥ ⎢ ν ν 1 23 ⎥⎢ ⎥ ⎥ ⎢ 12 ⎢ − 0 0 0 ⎥ ⎢ σ3 ⎥ ⎢ ε3 ⎥ ⎢− ⎥⎢ ⎥ ⎥ = ⎢ E1 ⎢ E2 E2 (3.8.7) ⎥ ⎢ τ23 ⎥ ⎢ γ23 ⎥ ⎢ ⎥⎢ ⎥ ⎥ ⎢ ⎢ 2(1 + ν23 ) ⎥ ⎣ ⎦ ⎣ γ31 ⎦ ⎢ 0 0 0 ⎥ τ31 0 0 ⎢ E2 ⎢ ⎥ τ12 γ12 1 ⎢ ⎥ ⎢ 0 0 0 0 0 ⎥ ⎢ ⎥ G 12 ⎣ 1 ⎦ 0 0 0 0 0 G 12 There are five independent constants. These are E1, E2 , ν 12 , G12 , and ν 23 . This is referred to as a transversely isotopic material. ########### Example 3.8.1 Problem: A thin unidirectional lamina has uniform edge loads (along the fiber direction) as shown in Figure (a). 2 f0 f0 1 b a Figure (a) These loads are f 1 = f 0 N/mm 2 (a) Since this is a thin sheet and all forces lie in the 1,2 plane this may be treated as a plane stress problem. It is suggested that the stress components in the interior of the plate are --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS σ1 (1, 2) = f 0 σ2 = τ12 = 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) P2: ABC JWST071-Waas July 2, 2011 94 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Are these correct values for the stress components? Solution: Check to see if the stress components satisfy the equilibrium equations, match the loads at the boundaries, satisfy Hooke’s law, and satisfy the compatibility equations. Notice that we are using 1 and 2 as coordinate directions instead of x and y. The equilibrium equations in two dimensions are given in Section 2.6, Equation 2.6.16. Insert the suggested stress components and evaluate. ⎤ ⎡ ⎡ ⎤ ∂ ∂  0 0 ⎥ ⎢ ∂ x1 f b1 ∂ x 2⎥⎣ ⎢ ⎦ (c) ⎣ ∂ ∂ ⎦ f 0 = − f b2 0 0 ∂ x2 ∂ x1 The left hand side is identically zero. In the absence of body forces the equilibrium equations are satisfied. The stresses at the boundaries do match the applied loads. From Hooke’s law the strain components are ε1 = f0 E1 ε2 = −ν12 f0 E1 γ12 = 0 (d) Insert these in the compatibility equation and evaluate. ∂ 2 ε2 ∂ 2 γ12 ∂ 2 ε1 + = 2 2 ∂ x1 ∂ x2 ∂ x2 ∂ x1 → ∂2 ∂ x22 f0 E1 + ∂2 ∂ x12 −ν12 f0 E1 = ∂2 (0) ∂ x1 ∂ x2 → 0+0=0 (e) All necessary equations are satisfied. These stress values are correct. ########### 3.8.2 Properties of Unidirectional Lamina The most popular unidirectional material in the aerospace industry is graphite/epoxy. There are several suppliers of this material, which consists of graphite filaments (fibers) that are held in a preimpregnated (pre-preg) polymer matrix. The lamina are supplied as rolls. The manufacturer purchases these lamina and lays them up (stacks them one on top of another) to form a laminate by applying pressure and temperature to the stack in a specified manner. After curing, a laminate is made. AS4 fiber (a commercial grade of graphite fiber) in a 3501-6 thermoset polymer is widely used. The designation is AS4/3501-6. Another popular graphite/epoxy is T300/5208. The auto industry and some in the civil engineering sector use glass fiber/epoxy. A popular glass fiber pre-preg is S2 glass/epoxy. The lamina elastic properties for these three materials are listed below (from Herakovich, 1998)1 in Table 3.8.1 In a later chapter we will examine these and other materials in much greater detail including additional properties. 1 Herakovich, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS CT. Mechanics of Fibrous Composites, John Wiley & Sons, Ltd, 1998. Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-03 P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 95 Displacement, Strain, and Material Properties Table 3.8.1 Material Density, g/cm2 Axial Modulus, E11 , GPa Transverse Modulus, E22, GPa Poissons Ratio, ν 12 Poissons Ratio, ν 23 Shear Modulus, G12 Axial Coeff. Of thermal expansion, α 1 , (m/◦ C) Transverse Coeff. of Thermal Expansion, α 2 , (m/◦ C) Fiber Volume Fraction, Vf Lamina Thickness, mm AS4/3501 T300/5208 S2 Glass/Epoxy 1.52 148 10.50 0.30 0.59 5.61 −0.8 29 0.62 0.127 1.54 132 10.8 0.24 0.59 5.65 −0.77 25 0.62 0.127 2.0 43.5 11.5 0.27 0.40 3.45 6.84 29 0.60 varies ########### Example 3.8.2 Problem: Problem: For the plate in Example 3.8.1 assign these initial dimensions before the load is applied. b = 200 mm h = 2 mm (a) where h is the thickness. It is made of graphite/epoxy (AS4/3501-6) with E11 = 148 × 103 MPa, E22 = 10.5 × 103 MPa, G12 = 5.61 × 103 MPa and ν12 = 0.3. If f0 = 200 N/mm2 . What are its dimensions after it is loaded? Solution: Find the strains and multiply by the original lengths to find the change in dimensions. From, 3.8.4, we have σ1 σ2 − ν21 ε1 = E1 E2 σ2 σ1 + ε2 = −ν12 E1 E2 τ12 γ12 = G 12 (b) Thus ε1 = 200 σ1 = 0.0014, = E1 148000 ε2 = −ν12 200 σ1 = −0.00041, = −0.3 E1 148000 γ12 = 0 (c) To these, we must add the three components of the original dimensions. From Equation 3.8.3, ε3 = ε2 . All shear strains are zero. Since the strains are constant the new dimensions can be found as follows: anew = 300 (1 + ε1 ) = 300 (1 + 0.0014) = 300.42 mm bnew = 200 (1 + ε2 ) = 200 (1 − 0.00042) = 199.92 mm h new = 2 (1 + ε3 ) = 2 (1 − 0.00042) = 1.9992 mm ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- a = 300 mm P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 96 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 3.9 Plan for the Following Chapters Our goal in the following chapters is to use the physical laws of equilibrium (Equations 2.6.13, 2.6.16, 2.6.18) and material properties (Equations 3.4.7, 3.4.22, 3.4.30) and the strain displacement relations (Equations 3.2.10, 3.2.11, 3.2.12) to derive equations connecting the known quantities such as geometry, applied loads, restraints, and material properties with the unknown quantities of displacement, strain, stress, internal forces, and restraint forces. Symbolically, for homogenous and isotropic materials {F} → { fb } {σ } {ε} {u} {ε} [D] {u} ← {σ } = [G] {ε} { f b } = [E] {σ } {ρ} (3.9.1) The quantities in the first row of Equation 3.9.1 are the matrix components of the unknown quantities for which we shall seek answers {σ } stresses {ε} strains {u} displacements and the known quantities that are given in the statement of the problem {F} applied concentrated and surface forces ({F} = {Fc } + { f s }) { fb } body forces {ρ} geometric restraints {ε} = [D] {u} Strain displacement {σ } = [G] {ε} Hooke’s law { f b } = [E] {σ } Equilibrium (E is the equilibrium operator matrix) The arrows indicate that {F} and {ρ} appear in the boundary conditions. The particular formulation of these equations depends on the geometry of the solid body. There are three classes of bodies that are usually considered: r One dimensional (function of 1 independent variable): Slender bars – axial, torsional, bending r Two dimensional (function of 2 independent variables): Plane stress, plates, and shells r Three dimensional (function of 3 independent variables): General solid bodies These are considered in order in the following chapters. In the one dimensional case the equations of equilibrium (Equation 2.6.18), material properties (Equation 3.4.7), and strain displacement Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- These quantities are connected by the following equations where the matrices in the second, third, and fourth rows are the differential, integral, and algebraic operators which define the following equations. P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas July 2, 2011 14:52 Printer Name: Yet to Come 97 Displacement, Strain, and Material Properties Equation (3.2.12) in order are summarized here as Equation 3.9.2. ∂σx = − f bx ∂x σx = Eεx εx = du dx (3.9.2) In two dimensions we have for equilibrium (Equation 2.6.16) ⎡ ∂ ⎢∂x [E] {σ } = ⎢ ⎣ 0 ∂ ⎤⎡ ⎤  ∂ y ⎥ σx ⎥ ⎣ σ y ⎦ = − f bx ∂ ⎦ f by τx y ∂x 0 ∂ ∂y = − { fb } (3.9.3) νE 1 − ν2 E 1 − ν2 0 E ⎢ 1 − ν2 σx ⎢ νE {σ } = ⎣ σ y ⎦ = ⎢ ⎢ ⎣ 1 − ν2 τx y 0 ⎡ ⎤ ⎡ 0 ⎤ ⎡ ⎤ ⎥ εx ⎥ ⎥ ⎣ ε y ⎦ = [G] {ε} 0⎥ ⎦ γx y (3.9.4) G and for strain displacement (Equation 3.2.12) ⎡ ∂ ⎡ ⎤ ⎢∂x ⎢ εx ⎢ {ε} = ⎣ ε y ⎦ = ⎢ ⎢ 0 ⎢ γx y ⎣ ∂ 0 ⎤ ⎥ ⎥ ∂ ⎥ ⎥ u ∂y ⎥ ⎥ v ∂ ⎦ ∂x ∂y = [D] {u} (3.9.5) In three dimensions for equilibrium (Equation 2.6.13) ∂ ⎢∂x ⎢ ⎢ [E] {σ } = ⎢ ⎢ 0 ⎢ ⎣ 0 ⎡ 0 0 ∂ ∂y 0 0 ∂ ∂z ∂ ∂y ∂ ∂x 0 0 ∂ ∂z ∂ ∂y ⎤⎡ ⎤ ∂ σx ⎢ ⎥ ⎡ ⎤ ∂z ⎥ ⎥ ⎢ σy ⎥ f bx ⎥ ⎢ σz ⎥ ⎢ ⎥ ⎣ ⎦ 0 ⎥ ⎥ ⎢ τx y ⎥ = − f by = − { f b } ⎥⎢ ⎥ f bz ∂ ⎦ ⎣ τ yz ⎦ ∂x τzx (3.9.6) For material properties (Equation 3.4.30) ⎡ λ + 2G ⎤ ⎢ ⎡ ⎢ λ σx ⎢ ⎢ σy ⎥ ⎢ ⎥ ⎢ λ ⎢ ⎢ σz ⎥ ⎢ ⎥ ⎢ {σ } = ⎢ ⎢ τx y ⎥ = ⎢ 0 ⎥ ⎢ ⎢ ⎣ τ yz ⎦ ⎢ ⎢ 0 ⎢ τzx ⎣ 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS λ λ 0 0 0 λ + 2G λ 0 0 λ λ + 2G 0 0 0 0 G 0 0 0 0 G 0 0 0 0 ⎥⎡ ⎤ 0 ⎥ εx ⎥ ⎥ ⎢ εy ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢ εz ⎥ ⎥⎢ ⎥ = [G] {ε} ⎢ ⎥ 0⎥ ⎥ ⎢ γx y ⎥ ⎥ ⎣ γ yz ⎦ 0⎥ ⎥ γzx ⎦ G ⎤ Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (3.9.7) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- for material properties (Equation 3.4.22) P1: TIX/XYZ JWST071-03 P2: ABC JWST071-Waas 98 July 2, 2011 14:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods and for strain displacement (Equation 3.2.10) ⎡ ∂ --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎢∂x ⎢ ⎢ ⎤ ⎢ 0 ⎡ ⎢ εx ⎢ ⎢ εy ⎥ ⎢ ⎥ ⎢ 0 ⎢ ⎢ εz ⎥ ⎢ ⎥ ⎢ {ε} = ⎢ ⎢ γx y ⎥ = ⎢ ∂ ⎥ ⎢ ⎢ ⎣ γ yz ⎦ ⎢ ∂ y ⎢ ⎢ γzx ⎢ 0 ⎢ ⎢ ⎣ ∂ ∂z ⎤ 0 0 ∂ ∂y ⎥ ⎥ ⎥ 0 ⎥ ⎥ ⎥ ∂ ⎥⎡ ⎤ ⎥ u ∂z ⎥ ⎥ ⎣ v ⎦ = [D] {u} ⎥ 0 ⎥ w ⎥ ⎥ ∂ ⎥ ⎥ ∂y ⎥ ⎥ ∂ ⎦ 0 ∂ ∂x ∂ ∂z 0 (3.9.8) ∂x The material properties for laminate materials will also be introduced at an appropriate point. The study of the three dimensional Equations 3.9.6, 3.9.7, and 3.9.8 and their two and one dimensional cousins is called the theory of elasticity. Satisfying these equations consistent with their boundary conditions of applied load and restraints for a given geometry has been proved to be a very accurate description of the behavior of many load carrying solid bodies. Finding exact analytical solutions, however, has proved quite difficult except in certain cases of simplified geometry, loading, and restraint. Fortunately, simplifying assumptions can provide equations that can be solved for a wide variety of cases with great practical value. The rest of this text deals with the equations and their solutions made possible by these simplifying assumptions. On occasion we even find exact solutions to the complete equations of elasticity. 3.10 Summary and Conclusions The various quantities that make up the study of mechanics of solids are identified and defined. In the remaining chapters we apply external forces on solid bodies which are restrained and find the internal forces or stresses and the displacements. To derive suitable equations for analysis we invoke two physical laws that are defined here – equilibrium and materials properties. To properly define material properties it is necessary to introduce the quantity called strain and connect it to the stresses via Hooke’s law. Strain is defined in terms of the displacements. In the past many solution methods were based on finding a state of stress that first satisfied equilibrium. Then the strains were found using Hooke’s law. Finally the displacements were found from the strain displacement equations. Because the six components of strain are defined in terms of three components of displacement the strain components are not independent of each other. Thus the compatibility equations also must be satisfied for the displacements to be valid. In more recent times equilibrium equations usually are formulated in terms of displacements. From the displacements we determine the strains, then the stresses, and compatibility is satisfied implicitly. This introduction is deliberately brief. It is believed that the things studied so far will become familiar and more meaningful in the process of solving a variety of problems that are introduced in the subsequent chapters. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 4 Classical Analysis of the Axially Loaded Slender Bar --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 4.1 Introduction In Chapters 1–3 we learned that in a typical problem in the analysis of solid bodies the known quantities are the geometry, applied forces or loads, restraints, and material properties. We note that the unknown quantities usually are restraint forces, internal forces and stresses, displacements, and strains. To have a successful structure the stresses and displacements must fall within acceptable limits. To define these quantities we have presented the two physical laws, equilibrium and material properties, and one geometric relationship, strain-displacement, that must be satisfied. Now we must put them together in a form that obtains solutions for the displacements, strains, stresses, and internal forces. To illustrate this in the clearest way we shall use some examples that have very simple geometry, loading, and restraint. In that way we can concentrate on the process and not be diverted from understanding by the details of more complicated cases. In this chapter we consider the axially loaded slender bar. We shall assume it is made of a homogenous and isotropic material and that its strain and displacement follow Hooke’s law within the load limits imposed. 4.2 Solutions from the Theory of Elasticity To set the stage for developing the equations for an axially loaded slender bar consider a rectangular solid body that is loaded by a uniform force per unit area, f 0 , in the x direction on each end as shown in Figure 4.2.1. There are no loads on any other surface, no body forces, and no restraints. The coordinate axes originate at the centroid of the body and are aligned as principal axes. The body is in equilibrium as a rigid body. We wish to find solutions for the stress and deformation that satisfy equilibrium in the interior, match the loads on the boundaries, and satisfy the compatibility equations. It is intuitive that the normal stress, σ x , shall be constant throughout the body and all other stress components be zero, that is, σx (x, y, z) = f 0 σ y = σz = τx y = τ yz = τzx = 0 (4.2.1) In fact we showed this to be true for the plane stress case in Example 3.5.1. Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 100 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y y L fx(-L/2,y,z) = f0 z h L/2 b z fx(L/2,y,z) = f0 x Figure 4.2.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Let us see if these values do satisfy the three dimensional equations of equilibrium in Section 2.6, Equations 2.6.13. In fact, upon substitution all equations are satisfied exactly when there are no body forces. ⎡ ∂ ⎢ ∂x ⎢ ⎢ ⎢ 0 ⎢ ⎢ ⎢ ⎣ 0 0 0 ∂ ∂y 0 0 ∂ ∂z ∂ ∂y ∂ ∂x 0 0 ∂ ∂z ∂ ∂y ∂ ⎤⎡ f ⎤ 0 ∂z ⎥ ⎢ ⎥ ⎡ ⎤ ⎥⎢ 0 ⎥ 0 ⎥⎢ ⎥ 0 ⎥ = ⎣0⎦ 0 ⎥ ⎥⎢ ⎢ ⎥ ⎥⎢ 0 ⎥ 0 ⎥ ∂ ⎦⎣ 0 ⎦ ∂x 0 (4.2.2) The stresses match the applied loads at all boundaries.  L σx ± , y, z 2 = f0  h σ y x, ± , z 2 =0  b σz x, y, ± 2 =0 (4.2.3) and all shear stresses on the boundaries are zero. The state of strain represented by this state of stress is obtained from Hooke’s law. εx = f0 σx = E E εy = − ν f0 νσx =− E E εz = − ν f0 νσx =− E E γx y = γ yz = γzx = 0 (4.2.4) It is noted in Section 3.3 that the strain components are not independent of each other. The conditions that must be met are called conditions of compatibility. It is noted there that compatibility is satisfied Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 101 Classical Analysis of the Axially Loaded Slender Bar if the strain components are constant. Let us check Equation 3.3.3 which is repeated here as Equation 4.2.5. ∂ 2 εy ∂ 2 γx y ∂ 2 εx + = 2 2 ∂y ∂x ∂ x∂ y (4.2.5) Inserting the strain components we see that compatibility is satisfied. ∂2 ∂ y2  f0 E + ∂2 ∂x2  ν f0 − E = ∂2 (0) ∂ x∂ y → 0+0=0 (4.2.6) The displacements are found by integrating the strain components. ∂u f0 f0 f0 → = → u(x, y, z) = x + a(y, z) E ∂x E E ∂v ν f0 ν f0 ν f0 εy = − → =− → v(x, y, z) = − y + b(x, z) E ∂y E E ∂w ν f0 ν f0 ν f0 εz = − → ∂ =− → w(x, y, z) = − z + c(x, y) E dz E E εx = (4.2.7) If we assign zero displacement at the origin of the axes we have the following boundary conditions f0 · 0 + a(y, z) = 0 → a(y, z) = 0 E ν f0 v(x, 0, z) = − · 0 + b(x, z) = 0 → b(x, z) = 0 E ν f0 w(x, y, 0) = − · 0 + c(x, y) = 0 → c(x, y) = 0 E u(0, y, z) = (4.2.8) The displacements are u(x, y, z) = f0 x E v(x, y, z) = − ν f0 y E w(x, y, z) = − ν f0 z E (4.2.9) Note that the displacement in the x direction is independent of y and z and linear in x. All points lying on any yz plane move an equal amount in the x direction. This is referred to as plane sections remain plane. There is also a displacement in the y and z directions even though there are no forces in the y and z directions. This was noted in Section 3.4 as the Poisson’s ratio effect. Since for many materials the value of Poisson’s ratio is approximately ν = 0.3 we can see that the displacements in the y and z directions can be quite sizeable compared to the displacement in the x direction for a three dimensional body. If the body is much longer in the x direction compared to its height and thickness the displacements in the y and z directions are relatively small in comparison. We noted in Section 3.4 that for a slender bar these displacements can often be ignored. Example 3.4.2 also confirms that for a typical structural material, in that case aluminum, all displacements will be quite small. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 102 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The solution for this combination of geometry, loading, and restraint is one of the few exact solutions available in the theory of elasticity. If we modify the end loading to be non uniform an exact solution can become quite difficult to obtain. Refer again to Chapter 3, Section 3.5. To illustrate consider a parabolic end load such as that shown in Figure 4.2.2. Let the loading at each end be  4y 2 f x (±L/2, y, z) = ± f 0 1 − 2 h (4.2.10) y f x ( L / 2, y , z ) f x ( L / 2, y, z ) x Figure 4.2.2 Let us assume that the stress in the interior has the same form as the loads on the boundaries, that is, σ y = σz = τx y = τ yz = τzx = 0 (4.2.11) Let us check to see if Equation 4.2.11 is a possible solution. First let us check to see if this state of stress satisfies equilibrium. ∂ ⎢ ∂x ⎢ ⎢ ⎢ ⎢ 0 ⎢ ⎢ ⎢ ⎣ 0 ⎡ ∂ ∂y ∂ ∂x 0 0 ∂ ∂y 0 0 ∂ ∂z 0 0 ∂ ∂z ∂ ∂y ⎤⎡  4y 2 ∂ f0 1 − 2 ⎢ h ∂z ⎥ ⎥⎢ ⎥⎢ ⎢ 0 ⎥ 0 ⎥⎢ 0 ⎥⎢ ⎢ ⎥⎢ 0 ⎥ ∂ ⎦⎢ ⎣ 0 ∂x 0 ⎤ ⎥ ⎥ ⎡ ⎤ ⎥ 0 ⎥ ⎥ ⎣ ⎦ ⎥= 0 ⎥ 0 ⎥ ⎥ ⎦ (4.2.12) Sure enough it does in the absence of body forces. And by inspection we see that it satisfies the boundary conditions on applied loads at the boundaries. On the ends σx  L ± , y, z 2  4y 2 = f0 1 − 2 h (4.2.13) and the stresses are zero on all other faces. So far so good, but now we must check compatibility. The components of strain are εx = f0 σx = E E  1− 4y 2 h2 εy = − ν f0 νσx =− E E  1− 4y 2 h2 (4.2.14) ν f0 νσx =− εz = − E E Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  4y 2 1− 2 h γx y = γ yz = γzx = 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`---  4y 2 σx (x, y, z) = f 0 1 − 2 h P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 103 Classical Analysis of the Axially Loaded Slender Bar Insert the components of strain into Equation 4.2.5. ∂ 2εy ∂ 2 γx y ∂ 2 εx + = 2 2 ∂y ∂x ∂ x∂ y   4y 2 ∂2 f0 1− 2 → 2 ∂y E h + ∂2 ∂x2  − ν f0 E 1− 4y 2 h2 → −  (4.2.15) =0 8 f0 = 0 Eh 2 However, this equation is not satisfied. The second derivative of the normal strain in the x direction is a non zero constant. An exact analytical solution has not been found; however, approximate numerical, but highly accurate, solutions for this and other boundary conditions have been found. The validity of these solutions is supported by experimental evidence. We shall now present one of these numerical solutions without proof; however, as this text proceeds we shall develop the methods used here. In particular a numerical solution has been found for a non uniform end force per unit area as shown in Figure 4.2.2. The Unigraphics I-DEAS software was used in the following example. ########### Problem: A rectangular bar is loaded with a parabolic end load as shown in Figure (a). It has dimensions of L = 250 mm, h = 50 mm, b = 10 mm. It is made of aluminum with E = 68950 N/mm2 . The distributed end loading is  L f x ± , y, z 2  4y 2 = ± f0 1 − 2 h =±  4y 2 3 1− 2 2 50 N mm 2 (a) Find the stresses and displacements. y x 50 mm 250 mm Figure (a) Solution: Use the I-DEAS software to find a numerical solution based on 3D elasticity equations. Note that we can determine from symmetry that the resultant forces at the two ends act through the centroids of the end faces. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Example 4.2.1 P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 104 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The value of the resultant force at each end is h 2 F= − h2  4y 2 f0 1 − 2 h 25 bdy = −25   4y 3 4y 2 3 3 1 − 2 10dy = · 10 y − 2 50 2 3 · 502 25 = 500 N (b) −25 A contour plot of the σ x component of stress from the numerical analysis is shown in Figure (b). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (b) The stress contours are shown on a deformed model which greatly exaggerates the deformation so that it is visible. The deformed model shows the magnitude of the displacement, that is, the vector sum of the x and y displacements. The outline of the undeformed bar is visible behind the contour plot. The Poisson’s ratio effect is clearly visible. The maximum displacement is of the order of 10−6 so you can see it is quite small. The lines between colors are lines of constant stress. (For those looking at a black and white print the lines between shades of gray are lines of constant stress.) Over most of the bar the stress is a uniform 1 N/mm2 . This is the same value you would get if a uniform distributed load was placed on the ends as explained in Equations 4.2.1-3 with a value of 1 N/mm2 (500 N total force). The surprising fact is that over most of the length of the bar the stress is uniformly distributed no matter how the load is Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come Classical Analysis of the Axially Loaded Slender Bar 105 distributed over the ends. Only the value of the resultant applied force and the cross section area are important. Near the ends the stress distribution is affected. There is a transition from the uniform value to the value of the distributed applied load. We can see better what is happening near the ends if we zoom in as shown in Figure (c). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (c) The black and white printed version, which shows the colors in gray scale, is not so clear but in any case you can see that a portion of the bar is affected significantly at each end. The stress transitions from the value of the applied surface load on the end faces to a uniform value over a distance about equal to the height of the bar. Over the middle 60 percent the stress is very close to uniform as expressed by a single color. The maximum stress at the ends is approximately 50 percent greater than the average uniform stress over the middle portion. A contour plot of the x displacement is shown in Figure (d). As noted the displacement displayed is greatly exaggerated in order to make it visible. As we noted the actual displacement is quite small. Straight vertical edges of color zones are evidence of constant values of displacement in the bar. Of particular importance is that the displacement away from the ends is uniform for all values of y and z. The material particles that lie in any yz plane before the load is applied continue to lie on an yz plane after the load is applied except near the ends. This is referred to as plane sections remain plane. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas 106 July 4, 2011 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- This phenomenon of the localized effect of loading, that is, over most of the length of a slender bar the internal stresses and displacements depend only on the value of the total force and not on its particular distribution, was first discovered by St. Venant who enunciated his now famous St. Venant’s principle. According to this principle the strains that are produced in any body by an application, to a small part of its surface, of a system of forces statically equivalent to zero force and zero moment, are of negligible magnitude at distances which are large compared to the linear dimensions of the body. This principle will aid in developing a simplified theory of analysis for axially loaded slender bars. Figure (d) Now let us repeat the solution using a more slender bar. This has a length to height ratio of 10 to 1 as shown in Figure (e). The cross section and the loading are the same. y x Figure (e) In Figure (f) we have the contour plot of the stress. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come Classical Analysis of the Axially Loaded Slender Bar Figure (f) In Figure (g) we have the contour plot of the x displacement. Figure (g) --`,,,,``,``,,``, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST 107 P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas 108 July 4, 2011 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In this case the same effect is noted; however, only about 10 percent of the region at each end is affected, that is, a length roughly equal to the height. Thus as the bar becomes more slender the end effect is a less significant part of the total. Before continuing with the development of this particular theory of structures we note that constraints can also affect the stress and displacement distribution. Consider a slender bar built into a wall, or has what is called a fixed end. The fixed end boundary condition consists of a total restraint in all degrees of freedom. In particular we solve the same bar as in Figure (a) with the same load on the left end but with the right end fixed. In Figure (h) we show the results of the stress contour plot. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (h) This is a bit hard to see so we enlarge the image of the right end in Figure (i). The high stress levels at the corners are due to restraint against the Poisson’s ratio effect. There is a decrease in the height of the bar over most of the length due to this effect but at the right end it is restrained by the fixed condition. Once again the effect is localized in the vicinity of the restraint and does not affect the rest of the bar. The actual displacements in the y and z directions are quite small for the slender member. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come Classical Analysis of the Axially Loaded Slender Bar 109 Figure (i) ########### In the meantime we shall use the assumption that plane sections remain plane to learn what we can about the behavior of slender bars by neglecting the end effects. Many of the great achievements in structural design and analysis have been made possible by the validity of this assumption in appropriate circumstances. 4.3 Derivation and Solution of the Governing Equations --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- In the examples in Section 4.2 the loads and restraints are applied only at the ends of the body. We wish now to allow loads to be applied along the bar in the x direction as well as at the ends. Consider a slender bar of arbitrary cross section as shown in Figure 4.3.1. It may be restrained at one or both ends and it has some combination of distributed loads, fx (x), acting along the bar and concentrated loads, Fc , at points along the bar and at its ends. We pause for a moment to ask how the applied forces are, in fact, applied. A load distributed over the end faces can be resolved into a concentrated resultant force. It must satisfy the requirement of a line of action through the centroid. How distributed and concentrated forces are applied between the ends requires some interpretation. Let us for now assume that some device or substance is used to attach the forces to the cross section or to the surfaces, but always with the condition that the resultant acts through the centroid. Gravity and inertia loads are one way to apply distributed axial loads. Point loads can be applied by drilling a hole and applying loads to a pin or a lug or collar can be fastened to the bar Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 110 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y z fx(x) Fc x Figure 4.3.1 for purposes of applying loads as illustrated in Figure 4.3.2. We would, of course, rely on St. Venant’s principle to be able to ignore local effects. Figure 4.3.2 In whatever manner the forces are applied we assume that their effect on modifying the uniform stress and the plane section displacements is ignored. If the bar is slender enough this localized effect can be ignored by the grace and goodness of St. Venant’s principle. We shall assume that plane sections remain plane over the entire length of the bar. It is assumed that the x axis lies on the loci of centroids of the cross sections in the yz plane. All resultants of distributed forces and all concentrated forces act through the centroids of the cross sections. Finding centroids of areas is discussed in detail in Appendix B. In most of our examples the cross sections will be simple shapes with axes of symmetry; such as, rectangular or circular. The centroids are where the axes of symmetry cross. For more complicated cross sectional shapes refer to Appendix B. From experimental evidence and our discussion in Section 4.2 we assume that plane sections remain plane and that the stress, σ x , on a plane cross section surface normal to the x axis is constant. No shearing stresses occur on that surface. Furthermore the displacement is defined by a single component, u, which is constant across the plane and at most is only a function of x, that is, u(x). It follows that there is one component of strain, εx (x). While it is true that the axial stress will generate lateral strain from the Poisson’s ratio effect, as we have noted in Section 3.4, this effect is small and will be neglected in the slender bar. Let us consider that the slender bar in Figure 4.3.1 has constant cross section, A, is fastened at one end, and is loaded with a distributed force, fx (x), with units of force per unit length (N/mm), and one or more concentrated loads all acting in the direction of and coincident with the x axis. If a body force --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 111 Classical Analysis of the Axially Loaded Slender Bar is present it must be restricted to the same conditions and is resolved into a force per unit length and combined with any distributed surface forces. We shall denote the distributed displacement as u(x). From our discussion of stress and strain in Chapter 3 the strain displacement equation is εx (x) = du dx (4.3.1) From Hooke’s law we have σx (x) = Eεx = E du dx (4.3.2) Since the x axis is located at the centroid of the cross section of the bar we can assume that only an axial internal force is generated, that is, there are no internal moments about any axis and no internal lateral forces. We denote the internal stress resultant force with the letter P. The internal force in terms of the stress, strain, and displacement is P = σx A = EAεx = EA du dx (4.3.3) Now let us seek the equilibrium of a slice of the bar at x of length dx as shown in Figure 4.3.3. fx(x) P(x+dx) P(x) dx Figure 4.3.3 We are faced with defining a convention for the positive direction of a force. It is common to define applied force components and restraint force components as positive if acting in the positive directions of the axes. So F and fx (x) in Figure 4.3.3 are shown as positive forces. Two different conventions are used for internal forces. The most common has been to define internal force components acting away from the surface as positive, as P is shown in Figure 4.3.3. Internal force resultants have the same sign conventions as stresses, thus a positive stress component will generate a positive stress resultant. We shall use this convention until further notice. In some cases later on we shall use the same convention for internal and external forces, that is, positive if acting in the positive direction of the axes, for reasons that will be made clear when we encounter the need. If the internal force at x is P(x) then at x + dx it is P(x + dx). Using a Taylor’s series expansion and neglecting higher order terms, we have P(x + d x) = P(x) + dP d x + higher order terms dx (4.3.4) From Newton’s laws the summation of forces in the x-direction must be zero. Fx = P(x) + dP d x − P(x) + f x (x)d x = 0 dx --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (4.3.5) P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 112 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods which reduces to dP = − f x (x) dx (4.3.6) Combining Equations 4.3.3 and 4.3.6 we obtain d du dP = EA = − f x (x) dx dx dx (4.3.7) The quantities and equations governing the behavior of an axially loaded bar are summarized in Equations 4.3.8. As a result of the assumption of plane sections remaining plane throughout the bar the first row differs from Equation 3.9.1 in that internal force resultants are added explicitly, distributed applied forces are explicit in the equations and any body forces are absorbed into the applied forces. All concentrated applied forces and restraints are introduced in the boundary conditions. Since each quantity is the function of a single variable, the matrices reduce to single components. In summary we have → f x (x) P(x) σx (x) εx (x) dP = − f x(x) dx u(x) εx (x) = σx (x) = Eεx ← ρ du dx (4.3.8) du P(x) = σx A = EAεx = EA dx d du dP = EA = − f x (x) dx dx dx In the first row of Equation 4.3.8 we have the typical internal unknowns in the formulation of a problem. u(x) εx (x) σx (x) P(x) - distributed displacement normal strain normal stress internal force resultant and the known external quantities f x (x) - distributed applied force Fc - concentrated applied forces ρ - displacement restraints The arrows are there to indicate that both Fc and ρ are external boundary conditions to be satisfied in the process of solving the equations. The equations connecting these quantities are in the second and third rows. dP = − f x (x) - static equilibrium dx du - strain displacement εx = dx σx = Eεx - Hooke’s law for material properties Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Fc P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 113 Classical Analysis of the Axially Loaded Slender Bar In the fourth row we have combined the three expressions for the internal quantities to get the internal force resultant in terms of the displacement (Equation 4.3.3). P = σx A = EAεx = EA du dx (4.3.9) and in the fifth row equilibrium is expressed in terms of the single unknown displacement (Equation 4.3.7). d du dP = EA = − f x (x) dx dx dx (4.3.10) In the special case of a bar with a constant cross sectional area and constant Young’s modulus this equation becomes EA d 2u = − f x (x) dx2 (4.3.11) The solution of Equations 4.3.10 and 4.3.11 involve straightforward direct integration. From Equation 4.3.11 we have 1 du =− dx EA f x (x)dx + a → u(x) = − 1 EA f x (x)dx + ax + b (4.3.12) The constants of integration, a and b, are found by assigning boundary conditions. The boundary conditions shall consist of one displacement constraint and one applied force or two displacement constraints. We must have at least one displacement restraint to obtain static equilibrium. This solution works for both statically determinate and indeterminate cases. Having found the displacement we find the stress by differentiating the displacement. σx = Eεx = E du dx (4.3.13) In the determinate case (one displacement restraint) the internal force, P, can be determined directly from the equations of static equilibrium. Once we know the internal force we know immediately that the stress is P = σx A → σx = P A (4.3.14) If we need the displacement in this case we get it from the first order equation, Equation 4.3.9. P = EA du dx → du P = dx EA → u(x) = 1 EA P(x)d x + c (4.3.15) where c is the constant of integration to be determined from a displacement boundary condition. We shall illustrate both approaches to a solution in the next simple example. Note that since u is a function of x only compatibility is always satisfied ########### Example 4.3.1 Problem: Consider the simplest problem of them all – a uniform bar with equal and opposite forces at the ends as shown in Figure (a). Find the stresses and displacements. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 114 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods This is the same problem examined in Section 4.2 but now with the assumption of plane sections remaining plane regardless of how the forces are applied as long as their resultant acts through the centroid of the cross sections. y F F x L Figure (a) Solution: First, integrate the second order displacement equation and use two boundary conditions to find the constants of integration. From the displacement find the strain, stress, and internal force. Second, find the internal force and stress using the equations of static equilibrium and then solve the first order displacement equation. First method: Since there is no distributed applied load the second order equation and its solution are EA d 2u =0 dx2 → d 2u =0 → dx2 du =a dx → u(x) = ax + b (a) The known boundary condition at the right end of the bar yields P(L) = EA du(L) = EAa = F dx → a= F EA (b) For our other boundary condition we attach the origin of the coordinate system to the left end of the bar. The displacement there will then be zero. u(0) = a · 0 + b = 0 → b=0 (c) The solution is u(x) = F x EA (d) To find the internal force, P, we use Equation 4.3.9.  du F = EA P = EA dx EA =F (e) Note well that we find the displacement by directly integrating the second order displacement equation and then find the internal force by differentiating the displacement equation. The stress throughout the bar is found to be σx = F P = A A --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (f) P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 115 Classical Analysis of the Axially Loaded Slender Bar Plots of the displacement and stress are shown in Figure (b). FL/EA Displacement, u x Stress, σx x Figure (b) Second method: For an alternative approach, since this is a statically determinate case, we can first draw a free body diagram of a portion of the bar from some point x to the right end as shown in Figure (c). P F x Figure (c) The internal force may be found immediately from static equilibrium Fx = F − P = 0 → P=F (g) The stress is found directly to be σx = F P = A A (h) The governing equation for the displacement and its solution is from Equation 4.3.9 P = EA du dx → du F = dx EA → u(x) = F x +c EA (i) where c is a constant of integration. To find the value of the constant of integration we invoke the condition of zero displacement at x = 0 or u(0) = 0. u(0) = F 0 + c = 0 thus c = 0 EA (j) F x EA (k) The displacement u(x) is u(x) = Note well that in this second case the internal force is found first using the equations of static equilibrium and from it the stress is found and then the displacement is found by integrating the first order equation. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- L F/A P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 116 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods We note that the displacement at x = L is u(L) = FL EA (l) This alternative approach is possible only if the bar is statically determinate. The first method is good in all cases, both determinate and indeterminate. Clearly the first approach using the second order differential equation is more general since it applies to all cases but in many statically determinate cases it is faster and easier first to find the internal force and then to solve the first order equation. When the stress is all you are looking for, of course, this is the way to go. We must note that this is exactly the same problem with exactly the same solutions as that for a bar with the left end fixed as shown in Figure (d). y F x L Figure (d) A free body diagram of the entire bar as shown in Figure (e) quickly confirms that it is the same problem. y Rx F x L Figure (e) From static equilibrium Fx = Rx + F = 0 → Rx = −F (m) and the boundary conditions are u(0) = 0 P(L) = EA du(L) =F dx (n) ########### 4.4 The Statically Determinate Case We shall provide several more examples of the statically determinate case. Simply put, the number of restraints may not exceed the number of applicable equations of static equilibrium. For an axially loaded --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 117 Classical Analysis of the Axially Loaded Slender Bar slender bar that number is one. Since we have only one constant of integration the displacement restraint is necessarily the boundary condition of record. Let us summarize the various determinate cases that arise. With fx (x) and EA as continuous functions of x, with concentrated loads at one or both ends, or with a restraint at one end and a concentrated load at the other we have three possible configurations. One possibility is for no specified displacement restraint but for all axial forces to be in equilibrium. This is illustrated in Figure 4.4.1. Example 4.4.1 is a special case of this. fx(x) F0 FL Figure 4.4.1 To be in equilibrium the summation of all concentrated forces and the resultants of all distributed forces must be zero. L f x (x)dx = 0 Fx = F0 + FL + (4.4.1) 0 In such cases we assign the origin of the coordinate system to be the point from which displacements are measured. In Section 4.2 the origin was placed at the centroid of the bar and the axes were oriented as principal axes. The origin may be placed at any convenient point along the centroidal axis of the bar including the end points. In Example 4.3.1 the origin was placed at the left end. This is in effect a displacement restraint. With a restraint at one end, fx (x) a continuous function of x, and a concentrated load at the other end, the possible configurations are shown in Figure 4.4.2. fx(x) fx(x) or FL F0 Figure 4.4.2 In all three cases the internal forces P(x) can be found by using static equilibrium. To find the stress σx = P(x) A (4.4.2) To find displacements we integrate the first order equation. EA du = P(x) dx → u(x) = P(x) dx + c EA (4.4.3) The constant of integration is found by imposing a displacement restraint, either u(0) = 0 or --`,,,,``,``,,`` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS u(L) = 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (4.4.4) P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 118 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 4.4.1 Problem: The bar with the left end restrained is given a distributed load fx (x) = f0 , a constant value, as shown in Figure (a). Find the displacement and stress. f x (x ) = f 0 y x L f0 f x (x ) 0 L Figure (a) Solution: Find the internal force and stress using the equations of static equilibrium and then the displacement by integrating the first order equation. You can set up a free body diagram of a portion of the bar to help find the internal force as we did in Example 4.3.1. f0 P x L Figure (b) From static equilibrium the axial internal force is Fx = f 0 (L − x) − P(x) = 0 → P(x) = f 0 (L − x) (a) and we can immediately say that σx (x) = f0 P(x) = (L − x) A A --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 119 Classical Analysis of the Axially Loaded Slender Bar To find the deflection we use P = EA du dx (c) to get f0 du = (L − x) → dx EA u(x) = f0 EA  Lx − x2 2 (d) +c The displacement boundary condition determines that u(0) = c = 0 (e) The final answer is u(x) = f0 EA  Lx − x2 2 u(L) = f0 L 2 2EA (f) Plots of the displacement and stress are given in Figure (c). f0L2/2EA Displacement, u L f0L/A --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Stress, x L Figure (c) Example 4.4.2 Problem: Our next example is a bar with a triangular loading and a displacement restraint on the right end as shown in Figure (a). Find the displacement and stress. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 120 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y x L f x (x ) = f 0 1 x L f0 f x (x ) 0 Figure (a) Solution: As the functional equation for the loading gets more complicated it is convenient to use the equilibrium equation, Equation 4.3.6, to find the internal axial force in place of a free body diagram. Then solve the first order equation for the displacement. For this case x dP = − f x (x) = − f 0 1 − (a) dx L Integrate once. P(x) = − f 0 1−  x2 x dx = − f 0 x − L 2L (b) +a The boundary condition at x = 0 is that there is no concentrated load, so P(0) = a = 0 (c) We have for the stress σx (x) = f0 P(x) =− A A The differential equation for the displacement is  du x2 EA = P(x) = − f 0 x − → dx 2L  x− f0 EA The boundary condition is the restraint at x = L, that is, u(L) = 0.  f0 L 2 L3 f0 L 2 u(L) = − − +c =0 +c =− EA 2 6L 3EA Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (d) x2 dx x− 2L  2 x3 f0 x − +c u(x) = − EA 2 6L u(x) = − → x2 2L  → c= f0 L 2 3EA Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (e) (f) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- L P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 121 Classical Analysis of the Axially Loaded Slender Bar And so the displacement is u(x) = f0 f0 L 2 − 3EA EA  x3 x2 − 2 6L = f0 EA  x2 x3 L2 − + 3 2 6L (g) Plots of the displacement and stress are given in Figure (b). f0 L 2/3EA Displacement, u Stress, σx -f0 L/2A Figure (b) ########### When distributed loading is discontinuous, and/or concentrated loads are at midpoints, or EA is discontinuous, the equations must be written for each segment. Examples of possible configurations are shown in Figure 4.4.3. There are, of course, an infinite variety of loads and EA values, but just the three possible end loads or restraints. Figure 4.4.3 Let two adjacent segments be identified by the subscripts i and i + 1 and the coordinate where they meet by xj . All loads including concentrated loads at a boundary are satisfied by using static equilibrium Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- L P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 122 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods to find P(x) in each region. To find the stresses σxi = Pi (x) Ai (4.4.5) To find displacements: E i Ai du i = Pi (x) dx u i (x) = → Pi (x) dx + ci E i Ai (4.4.6) For boundary conditions apply whatever restraints and concentrated applied forces occur at the ends of the bar. Then apply continuity of displacement where segments meet. Then     u i x j = u i+1 x j (4.4.7) Where necessary apply equilibrium of internal and external forces between segments.     du i x j du i+1 x j + F j + E i+1 Ai+1 =0 − E i Ai dx dx (4.4.8) For the top example in Figure 4.4.3 to be in equilibrium the summation of all concentrated forces and the resultants of all distributed forces must be zero. n Fx = m i=1 Fi + j=1 f j (x)d x = 0 (4.4.9) The bar is attached to the origin of the coordinate system. This serves as the restraint boundary condition to prevent rigid body motion. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ########### Example 4.4.3 Problem: Consider the bar with two concentrated loads as shown in Figure (a). Find the displacements and stresses. y 2F F x L/2 L Figure (a) Solution: Find the internal forces and stresses using the equations of static equilibrium and then find the displacements by integrating the first order equation. Satisfy boundary restraints and forces at the ends and continuity at the mid point. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 123 Classical Analysis of the Axially Loaded Slender Bar Region 1 ............. Region 2 ............. P1(x) 2F P2(x) F F Figure (b) From static equilibrium  Region1:  Region2: The stresses are Fx = F − 2F − P1 (x) = 0 → Fx = F − P2 (x) = 0 σx1 = − F A 0≤x ≤ σx2 = 0≤x ≤ L 2 L <x≤L 2 P2 (x) = F → L 2 P1 (x) = −F F A L <x≤L 2 (a) (b) The deflections are found by integrating and applying boundary conditions. F F L du1 =− → u 1 (x) = − x + c1 0≤x ≤ dx EA EA 2 F F L du2 = → u 2 (x) = x + c2 ≤x≤L dx EA EA 2 (c) The boundary condition at the left end is u 1 (0) = c1 = 0 → u 1 (x) = − Now we use u 2 (L/2) = u 1 (L/2) to get   FL FL L L = + c2 = u 1 =− u2 2 2EA 2 2EA F x EA → u1 → u 2 (x) =  L 2 =− FL 2EA F (x − L) EA → u 2 (L) = 0 Plots of the displacement and stress are shown in Figure (c). Displacement, u L -FL/2EA F/A Stress, x -F/A Figure (c) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) (e) P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 124 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Example 4.4.4 Problem: Suppose the same bar has both a distributed load over its whole length as given in Example 4.4.1 and two concentrated forces as given in Example 4.4.3 as shown in Figure (a). Find the displacements and stresses. y f0 2F F x L/2 L Figure (a) Solution: We have two choices here. First we could proceed exactly as in the two previous examples and find the axial internal force. From it we can find the stress and solve the differential equation for the displacement. Second we could use superposition of known solutions. First Method: Region 1:  Fx = f 0 (L − x) + F − 2F − P1 (x) = 0  Fx = f 0 (L − x) + F − P2 (x) = 0 P1 (x) = f 0 (L − x) − F → 0≤x ≤ (a) L 2 Region 2: → P2 (x) = f 0 (L − x) + F (b) L <x≤L 2 The stress is f0 L F P1 (L − x) − = 0≤x ≤ A A A 2 (c) f0 F P2 L (L − x) + = <x≤L σx2 = A A A 2 The deflection is found by integrating the following two equations and applying the boundary restraint. σx1 = P1 f0 F du 1 = = (L − x) − dx EA EA EA  x2 f0 → u 1 (x) = Lx − EA 2 − F x + c1 EA → u 1 (0) = c1 = 0 0≤x ≤  P2 f0 F x2 F du 2 f0 = = (L − x) + → u 2 (x) = x + c2 Lx − + dx EA EA EA EA 2 EA     L2 L2 f0 L 2 F L f0 L 2 L L − + c2 = − = u2 = + → u2 2 2 EA 2 8 EA 2 EA 2 8 → c2 = − FL EA L 2 − (d) F L EA 2 L <x≤L 2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 125 Classical Analysis of the Axially Loaded Slender Bar The resulting displacements are  x2 f0 Lx − EA 2  x2 f0 Lx − u 2 (x) = EA 2 F x EA F (x − L) + EA u 1 (x) = − 0≤x ≤ L 2 (e) L <x≤L 2 Second Method: Actually, there is an easier way. From the principle of superposition we simply add the displacement from previous examples. If we label Example 4.4.1 as case a, Example 4.4.3 as case b, and this example as case c, then EA du a = Pa dx and EA du b = Pb dx (f) and since the boundary conditions are the same for both, then for EA du c = Pa + Pb dx (g) → uc = ua + ub Thus u c2 (x) = u a2 + u b2 x2 2 f0 EA  Lx − f0 = EA  x2 Lx − 2 u c1 (x) = u a1 + u b1 = − F x EA F (x − L) + EA 0≤x ≤ L 2 (h) L ≤x≤L 2 You will note that the stresses can also be found by adding the two solutions. Example 4.4.5 Problem: The slender bar is given three concentrated loads at equal intervals along the length as shown in Figure a. Let l = L/3. Find the displacements and stresses. y L --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- F 2F 3F x l l l Figure (a) Solution: Find the internal force and stress using the equations of static equilibrium and then the displacements by integrating the first order equations. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 126 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The bar is divided into three regions with separate equations covering each region. Since this is statically determinate we can find the constant axial internal force in each region to be  0≤x ≤l Region1: Fx = 3F − 2F + F − P1 = 0 → P1 = 3F − 2F + F = 2F  l ≤ x ≤ 2l Region2: Fx = 3F − 2F − P2 = 0 → P2 = 3F − 2F = F (a)  2l ≤ x ≤ L Region3: Fx = 3F − P3 = 0 → P3 = 3F The stresses follow immediately. σx1 = 2F A σx2 = F A σx3 = 3F A (b) The equations governing the displacement in each region are EA du 1 = P1 = 2F dx EA du 2 = P2 = F dx (c) EA du 3 = P3 = 3F dx The solutions are u1 = 2F x + c1 EA u2 = F x + c2 EA (d) u3 = 3F x + c3 EA The boundary conditions are u 1 (0) = 0 u 2 (l) = u 1 (l) (e) u 3 (2l) = u 2 (2l) Applying these boundary conditions: 2Fl 2F x → u 1 (l) = EA EA Fl 2Fl Fl + c2 = → c2 = u 2 (l) = u 1 (l) → EA EA EA 3Fl F (x + l) → u 2 (2l) = → u 2 (x) = EA EA 6Fl 3Fl 3Fl + c3 = → c3 = − u 3 (2l) = u 2 (2l) → EA EA EA 3F 6Fl (x − l) → u 3 (3l) = → u 3 (x) = EA EA u 1 (0) = c1 = 0 → u 1 (x) = (f) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- We can now say that u(0) = u 0 = 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS u(l) = u l = 2 Fl EA u(2l) = u 2l = 3 Fl EA u(3l) = u 3l = 6 Fl EA Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (g) P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 127 Classical Analysis of the Axially Loaded Slender Bar We plot the displacements and internal forces in Figure (b) noting that the displacements are linear between points of load application and the internal forces are constant between points of load application. 6 FL/EA 4 Displacement, u 2 0 0 l 2l s1 3l s3 s2 3F Internal force, P 2F F 0 Figure (b) We have labeled the displacements at the discrete points at which loads are applied as ul , u2l , and u3l . It has been found useful at times to represent the distributed displacement u(x)in terms of these discrete displacements and the local coordinates s1 , s2 , and s3 , as shown in Figure (b). For example, consider region 2. s2  s2 s2 u l + u 2l (h) u(s2 ) = u l +(u 2l − u l ) = 1 − l l l This can be put in a standard matrix form.    s2 s2  u l u(s2 ) = 1 − u 2l l l (i) As a check let us evaluate u(s2 ) at s2 = 0 and s2 = l. s2 u(s2 ) = 1 − l     s2  u l s2 1− = l l u 2l → ⎡ 2Fl ⎤  ⎥ s2 ⎢ ⎢ EA ⎥ l ⎣ 3Fl ⎦ EA u(0) = 2Fl EA → (j) u(l) = 3Fl EA In fact this applies to any and all regions, say, region m, when the appropriate local coordinate, sm , and the discrete displacements at each end of the region are used.  sm s m   u  n (k) u(sm ) = 1 − l lm m u n+1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- where u n is the displacement at sm = 0 and u n+1 is the displacement at sm = lm . Try it on regions 1 and 3 to satisfy yourself that it works. We shall have an important use for this way of representing the displacements later in the next chapter. ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 128 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods As we have noted determinate problems also can be solved using the second order differential equation. With fx (x) a continuous function of x and a concentrated load at one end only and a restraint at the other end as shown in Figure 4.4.1: EA d 2u = − f x (x) → dx2 1 du =− dx EA f x (x)d x + a → u(x) = − (4.4.10) 1 EA f (x)d xdx + ax + b Distributed loads are accounted for in the differential equation but concentrated loads must enter through the boundary conditions. The constants of integration are found by one boundary displacement restraint and one concentrated load condition. u(0) = 0 and EA du(L) = FL dx or EA du(0) = F0 dx and u(L) = 0 (4.4.11) To find the stress differentiate the displacement. EA du =P dx → σx = P A (4.4.12) When distributed loading is discontinuous, and/or concentrated loads are at midpoints, or EA is discontinuous the equations must be written for each segment. E 1 A1 d 2u1 = − f 1 (x) dx2 → 1 du 1 =− dx E 1 A1 → E 2 A2 d 2u2 = − f 2 (x) dx2 → u 1 (x) = − 1 du 2 =− dx E 2 A2 → ... d 2u1 E n An 2 = − f n (x) dx f 1 (x)d x + a1 → f 1 (x)d xd x + a1 x + b1 f 2 (x)d x + a2 u 2 (x) = − 1 du n =− dx E n An → 1 E 1 A1 1 E 2 A2 f 2 (x)d xd x + a2 x + b2 (4.4.13) f n (x)d x + an u n (x) = − 1 E n An f n (x)d xd x + an x + bn The constants of integration are found by imposing a displacement restraint, either u 1 (0) = 0 or u n (L) = 0 (4.4.14) and continuity of displacement between regions of integration. u 1 (x1 ) = u 2 (x1 ) u 2 (x2 ) = u 3 (x2 ) . . . u n−1 (xn−1 ) = u n (xn−1 ) (4.4.15) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 129 Classical Analysis of the Axially Loaded Slender Bar and equilibrium of internal and external forces between regions E 1 A1 du 1 (x1 ) du 2 (x1 ) − E 2 A2 − Fx1 = 0 dx dx du 3 (x2 ) du 2 (x2 ) − E 3 A3 − Fx2 = 0 E 2 A2 dx dx ... E n−1 An−1 du n−1 (xn−1 ) du n (xn−1 ) − E n An − Fxn = 0 dx dx (4.4.16) To find stresses differentiate the displacements. du 1 = P1 (x) dx du 2 = P2 (x) E 2 A2 dx E 1 A1 P1 (x) A1 P2 (x) = A2 → σx1 = → σx2 (4.4.17) ... E n An du n = Pn (x) dx → σxn = Pn (x) An When the second order equation is used the displacements are found first by integration. The constants of integration are found from a combination of displacement boundary conditions and force boundary conditions for statically determinate cases. The stresses are then found by differentiating the displacements to find the internal axial forces and dividing by the areas. If the problem is indeterminate you must use the second order equations. If the problem is determinate you may use either the first order or the second order equations. 4.5 The Statically Indeterminate Case y fx (x) x L Figure 4.5.1 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- We shall provide several examples of the statically indeterminate case. Simply put, the number of restraints exceeds the number of applicable equations of static equilibrium, that is, there are two displacement restraints and only one equilibrium equation. Let us summarize the various indeterminate cases that arise. First consider that E, A, and f x (x) are continuous functions of x and there are displacement restraints at both ends as shown in Figure 4.5.1. P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 130 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods A free body diagram of the whole bar gives us Figure 4.5.2. y fx(x) Rx2 Rx1 x Figure 4.5.2 From the summation of forces L f x dx + Rx1 + Rx2 = 0 Fx = (4.5.1) 0 Of course, this is our only equilibrium equation and it has two unknown restraint forces in it. This is a statically indeterminate problem. In such circumstances we cannot find the internal force P by means of equilibrium alone. In this case we must use the second order equation, which combined the equation of equilibrium with the bar strain displacement and material properties relations. d du dP = EA = − f x (x) dx dx dx (4.5.2) or when EA is constant EA d 2u = − f x (x) dx2 (4.5.3) ########### Example 4.5.1 Problem: Consider the special case represented by Figure 4.5.1 when fx (x) = f0 is a constant and EA is a constant. Find the displacement and stress. y fx (x) = f0 x L Figure (a) Solution: Solve the second order equation for the displacement and then differentiate to find the stress. The second order equation and its solution are EA d 2u = − f0 dx2 → f0 du =− x +a dx EA → u(x) = − f0 2 x + ax + b 2EA --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 131 Classical Analysis of the Axially Loaded Slender Bar where a and b are constants of integration. In this case we use two displacement boundary conditions to find the two constants of integration. The bar is restrained on both ends so u(0) = b = 0 u(L) = − f0 L 2 + aL + b = 0 2EA (b) From these we obtain b=0 a= f0 L 2EA (c) and so u(x) = f0 x (L − x) 2EA (d) Now, the equations for strain displacement and Hooke’s law are the same as we used before, and so is the stress resultant P in terms of the stress, thus εx = f0 du = dx EA  L −x 2 σx = Eεx =  f0 A L −x 2 P = σx A = f 0  L −x 2 (e) Plots of the displacement and stress are shown in Figure (b). f0L2/8EA Displacement, u f0L/2A Stress, L x - f0L/2A Figure (b) The support restraint forces are Rx1 = −P(0) = − f 0 L 2 Rx2 = P(L) = − f 0 L 2 (f) Note how the sign conventions for restraint forces and internal forces are accommodated in Equation (f). ########### When distributed loading is discontinuous, and/or concentrated loads are at midpoints, or EA is discontinuous, the equations must be written for each segment. An example of a possible configuration is shown in Figure 4.5.3. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 4, 2011 132 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods f1(x) Fx2 x1 x2 Figure 4.5.3 In this case we must use the second order equations. The constants of integration are found by imposing displacement restraints at both ends. u 1 (0) = 0 and u n (L) = 0 (4.5.4) and continuity of displacement between regions of integration. u 1 (x1 ) = u 2 (x1 ) u 2 (x2 ) = u 3 (x2 ) . . . u n−1 (xn−1 ) = u n (xn−1 ) (4.5.5) and equilibrium of internal and external forces between regions E 1 A1 du 1 (x1 ) du 2 (x1 ) − E 2 A2 = Fx1 dx dx du 2 (x2 ) du 3 (x2 ) − E 3 A3 = Fx2 E 2 A2 dx dx (4.5.6) ... E n−1 An−1 du n (xn−1 ) du n−1 (xn−1 ) − E n An = Fxn dx dx To find the stresses differentiate the displacements. du1 = P1 (x) dx du2 = P2 (x) E 2 A2 dx E 1 A1 ... dun = Pn (x) E n An dx P1 (x) A1 P2 (x) = A2 → σx1 = → σx2 → σxn = (4.5.7) Pn (x) An ########### Example 4.5.2 Problem: A uniform bar has fixed ends and a concentrated load F at x = x0 as shown in Figure (a). Find the displacement and stress. y x0 F x L Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-04 P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 133 Classical Analysis of the Axially Loaded Slender Bar Solution: Solve the second order equation for the displacement and then find the stress. We will distinguish the two regions with subscripts where u 1 (x) is from 0 ≤ x ≤ x0 and u2 (x) is from x0 ≤ x ≤ L. The solution to the first equation is EA d 2u 1 = 0 → u 1 (x) = a1 x + b1 dx2 0 ≤ x ≤ x0 (a) x0 ≤ x ≤ L (b) The solution to the second equation is EA d 2u2 =0 dx2 → u 2 (x) = a2 x + b2 We must find four boundary conditions to solve for the constants of integration. We have a displacement boundary condition at each end or u 1 (0) = a1 0 + b1 = 0 u 2 (L) = a2 L + b2 = 0 (c) u 1 (x0 ) = u 2 (x0 ) → a1 x0 + b1 = a2 x0 + b2 (d) Second, apply equilibrium at x = x0 . See the free body diagram in Figure (b). F P1 P2 Figure (b) From summation of forces Fx = P2 + F − P1 = 0 → P2 − P1 = EA du1 du2 − EA = −F dx dx (e) or du1 du2 F − = a2 − a1 = − dx dx EA (f) We have then four equations to solve for the four constants of integration. As the number of unknowns increase to four and more it becomes more tedious to solve by the usual substitution methods that work so well for two or three unknowns. We can cast the boundary condition equations in the following format a1 0 + b1 = 0 a1 x0 + b1 − a2 x0 − b2 = 0 a1 − a2 = F EA a2 L + b2 = 0 (g) and form matrix equations ⎡ 0 ⎢ x0 ⎢ ⎣1 0 1 0 1 −x0 0 −1 0 L ⎤ ⎡ ⎤⎡ ⎤ 0 0 a1 ⎢ 0 ⎥ ⎥ ⎢ ⎥ ⎢ −1 ⎥ ⎥ ⎢ b1 ⎥ = ⎢ F ⎥ ⎥ 0 ⎦ ⎣ a2 ⎦ ⎢ ⎣ EA ⎦ b2 1 0 (h) This can be solved with one of the symbolic equation solving software packages, such as, Mathematica, Maple, and others. Appendix C contains some Mathematica instructions for solving sets of equations symbolically and numerically. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Now it follows that we must have continuity at x0 . First, apply continuity of displacements. P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 134 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods By any method the solution is a1 = The displacement is x0  F 1− EA L u 1 (x) = Also b1 = 0 a2 = − x0  F 1− x EA L u 2 (x) = F x0 EA L b2 = F x0 EA x F 1− x0 EA L F x0  x0  F x0  du1 = 1− σ1x = Eε1x = 1− P1 = σ1x A = F 1 − dx EA L A L L F x0 du2 F x0 x0 =− σ2x = Eεx = − P2 = σ2x A = −F = dx EA L A L L (i) (j) ε1x = ε2x (k) The plots of the displacement and stress are shown in Figure (c) u(x0) Displacement, u x0 L x F 1− 0 L A Stress, σx − F x0 A L Figure (c) Note that if we label the nodal displacements u 1 (0) = u 0 = 0 u 1 (x0 ) = u 2 (x0 ) = u x0 u 2 (L) = u L (l) the distributed displacements can be written in terms of local coordinates and the nodal displacements.  s1 s1   0  u 1 (s1 ) = 1 − x x 0 0 u x0 (m)    s2 s2 u x0 u 2 (s2 ) = 1 − L − x0 L − x0 uL This form for the displacements applies to all axial cases with concentrated loads. ########### This solution can be used to solve for multiple concentrated loads for a given set of displacement boundary conditions. From the principle of superposition, if multiple concentrated loads are applied at --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 135 Classical Analysis of the Axially Loaded Slender Bar various values of x0 , you can substitute the appropriate force and location values and simply add them up to obtain the displacements, strains, stresses, and axial stress resultants. In all cases so far the displacement constraints have all been zero. It is possible to induce stress and displacement by imposing a non zero displacement constraint. The next example demonstrates this. ########### Example 4.5.3 Problem: The bar in Example 4.5.1 is modified by imposing a non zero restraint at the right end. This might be accomplished with a screw jack. In this case we create a negative displacement δ as shown in Figure (a). (a) u(L) = −δ y f0 x δ Figure (a) Solution: This is an indeterminate problem (two displacement restraints) so solve the second order equation. From Example 4.5.1 we have for the displacement u(x) = − f0 2 x + ax + b 2EA (b) The boundary conditions are now u(0) = b = 0 u(L) = − f0 L 2 + a L + b = −δ 2EA (c) f0 L δ + L 2EA (d) From these we obtain b=0 a=− and so u(x) = δ f0 x (L − x) − x 2EA L (e) Now, the equations for strain displacement and Hooke’s law are the same as we used before, and so is the stress resultant P in terms of the stress, thus   EAδ f0 L Eδ P du L = f0 −x − σx = = −x − (f) P(x) = EA dx 2 L A A 2 L --`,,,,``,``,,``,,`,`,,,,,``,,``-` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 136 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The support restraint forces are Rx1 = −P(0) = − f 0 E Aδ L + 2 L Rx2 = P(L) = − f 0 E Aδ L − 2 L (g) Plots of the displacement and stress are shown in Figure (b). f 0 L2 8 EA 2 Displacement, u f0 L 2A Stress, E L L x f0 L 2A E L Figure (b) ########### 4.6 Variable Cross Sections Consider an axially loaded slender bar with a cross section that varies with x, that is, A(x). Once again we shall assume that the stress is uniform across the section if the axial load acts through the centroid of the section. The same rules apply for statically determinate and indeterminate cases. Given the internal force we can find the stresses. σx = P(x) A(x) (4.6.1) Finding the deflection requires integrating the first order equation for the determinate case. EA(x) du = P(x) dx → u= P(x) dx + c EA(x) (4.6.2) or the second order equation for the indeterminate case d du EA(x) = − f x (x) dx dx → u(x) = − 1 EA(x)  f x (x)dx + a dx + b (4.6.3) An example will help. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 137 Classical Analysis of the Axially Loaded Slender Bar ########### Example 4.6.1 Problem: An axially loaded slender bar with linear taper, restrained as shown in Figure (a), has a uniform distributed load. Find the stress and displacement. The cross section area is x  (a) A(x) = A0 1 − 2L where A0 is the area at x = 0. f0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- L Figure (a) Solution: Find the internal force and stress and then solve the first order equation for the displacement. The internal force is Fx = f 0 (L − x) −P(x) = 0 → P(x) = f 0 (L − x) (b) The stress is σx = The deflection is 1 u(x) = E 1 P(x) dx = A(x) E It is convenient to introduce f 0 (L − x) P(x) = x  A(x) A0 1 − 2L f0 L 2 f 0 (L − x) x  dx = EA0 A0 1 − 2L ξ= (c) 1 x x d L − 1− 2L x L d x x L 1− 2L x L then the integral may be written and looked up in the integral tables. ⎫ ⎧ ⎪ ⎪ ⎪ ⎪   ⎬ 2 ⎨ f0 L 2 1 ξ f0 L ξ   dξ − dξ = u(ξ L) = 2 ln 1 − ⎪ ξ ξ EA0 ⎪ EA 2 0 ⎪ ⎪ ⎭ ⎩ 1− 1− 2 2 (d) (e) + 2ξ + c (f) Applying the boundary condition u(0) = 0 we get c = 0. Thus u(x) = x" x  f0 L 2 ! +2 2 ln 1 − EA0 2L L (g) It is always good practice to find a way to check your answer. Let us compare the displacement at x = L with the uniform bar with the same loading in Example 4.4.1. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 138 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods If the bar has a constant cross section equal to A0 , the value at x = 0, the deflection would be u(L) = 0.5 f0 L 2 EA0 (h) On the other# hand if the cross section was constant and equal to that at the end of the tapered bar (A(L) = A0 2), the end displacement would be u(L) = f0 L 2 EA0 (i) For this tapered case we have u(L) = 0.6137 f0 L 2 EA0 (j) It is logical that the displacement would fall between the two. Plots of the displacement, internal force, and stress are given in Figure (b). 0.6137f0L2/EA0 0.5f0L2/EA0 Displacement, u 0 L f0L 0.5f0L Axial force, P 0 f0L/A0 Stress, x 0 Figure (b) ########### If the restraints are such that the bar is statically indeterminate the second order equation must be used. du d EA = − f x (x) dx dx (4.6.4) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 139 Classical Analysis of the Axially Loaded Slender Bar Once we find the displacement u(x) by integrating twice and applying constraints to find the constants of integration we can find the axial internal force P(x) = EA(x) du dx (4.6.5) σx (x) = P(x) A(x) (4.6.6) ########### Example 4.6.2 Problem: Consider the same slender bar and loading as in Example 4.6.1 but with both ends fixed as shown in Figure (a). Find the displacement and stress. f0 L Figure (a) Solution: Solve the second order equation for the displacement and then find the stress. Integrate Equation 4.6.4 once du d EA = − f0 dx dx → EA du = − f0 x + a dx → where a is the constant of integration. The second integration provides ⎧ ⎪ 1 ⎨ u(x) = − f0 EA0 ⎪ ⎩ x 1− ⎫ ⎧ ⎪ ⎪ ⎬ ⎨ f0 x a 1 du   − + = x x ⎪ dx EA0 ⎪ ⎭ ⎩ 1− 1− 2L 2L x  dx + a 2L where b is the second constant of integration. x For convenience let ξ = , then L ⎧ ⎪ ⎪ 1 ⎨ ξ  − f0 L 2 u(ξ L) = dξ + a L ξ EA0 ⎪ ⎪ ⎩ 1− 2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 1 1− ⎫ ⎪ ⎬ (b) x  d x + b⎪ ⎭ 2L 1  1− ξ 2 dξ + b (a) ⎫ ⎪ ⎪ ⎬ ⎪ ⎪ ⎭ Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- and the stress P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 140 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods and from a table of integrals      ξ ξ 1 − 2a L ln 1 − − f 0 L 2 −2ξ − 4 ln 1 − EA0 2 2 +b (d) Now apply the boundary restraints. % 1 $ − f 0 L 2 [0 − 4 ln (1)] − 2a L ln (1) + b = 0 → b = 0 EA0 % 1 $ u(L) = − f 0 L 2 [−2 − 4 ln (0.5)] − 2a L ln (0.5) + 0 = 0 → a = 0.5573 f 0 L EA0 u(0) = (e) Finally, the displacement is u(x) = " f0 L ! x  + 2x 2.8854L ln 1 − EA0 2L (f) The internal force in the bar is P(x) = EA(x) du = − f 0 x + a = − f 0 x + 0.5573 f 0 L dx (g) Let us check to see if the two internal forces at the restraints are in equilibrium with the applied load. P(0) = 0.5573 f 0 L → P(L) = f 0 L(−1 + 0.5573) = −0.4427 f 0 L (h) Recognizing the definition of positive P we have the values shown in Figure (b). P(0) = 0.5573 f 0 L P( L) = 0.4427 f 0 L f0 Figure (b) Summing all forces we get f 0 L + P(L) − P(0) = f 0 L − 0.4427 f 0 L − 0.5573 f 0 L = 0 We see that the bar is in equilibrium. This is a good check on the answer. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (i) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- u(ξ L) = P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 141 Classical Analysis of the Axially Loaded Slender Bar The stress is σx (x) = du P(x) =E = E (− f 0 x + a) = E f 0 (0.5573L − x) A(x) dx (j) As a check we note the values of the stress at x = 0 and x = L. σ (0) = 0.5573 f0 L A0 σ (L) = −0.8854 f0 L A0 (k) Plots of the displacements, internal forces, and stresses are given in Figure (b). 0.2f0L2/EA0 Displacement, u 0 L 0.5f0L Axial force, P 0 -0.5f0L 0.5f0L/A0 Stress, x -0.5f0L/A0 -0.8854f0L/A0 -1.0f0L/A0 Figure (b) ########### In real life we may not find that the cross sectional area is so easily represented by a simple functional relationship; therefore, in today’s world it is more common to adopt approximate methods. For example, the tapered bar can be divided into a series of uniform segments as shown in Figure 4.6.1. We use an average value of the area for each segment. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 142 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods f0 L Figure 4.6.1 We shall demonstrate solutions in the next chapter in Section 5.5 and verify the validity of this type of approximation. 4.7 Thermal Stress and Strain in an Axially Loaded Bar --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Thermal strain occurs when a bar is heated. Defining the change in temperature from some base value as T we have for the thermal strain (See Chapter 3, Section 3.6) (4.7.1) εx T = α T The strain from applied loads we designate as εxmech then the total strain is εxTotal = εxmech + εx T = du σx +α T = E dx (4.7.2) If a statically determinate bar is heated a uniform amount it expands thus increasing the displacement. No stresses are induced. When the bar is restrained on both ends, that is, is indeterminate, stresses are induced in the absence of any applied loads but no displacement takes place. These effects can be added to the effects of applied loads by the principle of superposition. In the presence of a change in temperature the summary of equations in Equation 4.3.8 are modified as follows. Fc → f x (x) dP = − f x (x) dx P(x) σx (x) εxmech (x) T (x) εxTotal = εxmech + εx T = du − Eα T σx (x) = E dx  du P(x) = σx A = EA −α T dx   d dP du = −α T = − f x (x) EA dx dx dx u(x) ← ρ du σx +α T = E dx (4.7.3) ########### Example 4.7.1 Problem: The two bars shown in Figure (a) are initially at room temperature. Their temperature is then increased uniformly an amount T. What is the displacement and stress in each case? Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 143 Classical Analysis of the Axially Loaded Slender Bar y x y x --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (a) Solution: Use Equation 4.7.3. The fixed-free case is statically determinate so there is no internal force P and σx = 0 For the displacement  du EA −α T =0 dx → u(x) = α T x + c For the fixed-fixed case if α T is a constant   d 2u d du EA −α T = EA 2 = 0 → dx dx dx → (a) → u(0) = c = 0 → u(x) = α T x u(x) = cx + d u(0) = d = 0 → u(L) = cL = 0 (b) (c) → u(x) = 0 From Equation 4.7.3 σx = Eεx = E  du −α T dx = −Eα T (d) These solutions can be superimposed with the other solutions for stress and displacement under loading for a bar that is both loaded and heated. ########### 4.8 Shearing Stress in an Axially Loaded Bar From what we have studied so far it would appear that there is one component of normal stress,σx , and all other stress components are zero. In terms of internal surfaces on which stress components are defined, that is, planes normal to the rectangular Cartesian coordinates to which we have oriented the bar, this is true. At other orientations of the internal surfaces, however, other stress components can have non zero values. As we noted in Chapter 1 let us look at an internal surface of the bar which is not normal to the x axis as shown in Figure 4.8.1. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 144 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y t n F F x x0 Figure 4.8.1 The normal to the surface is at an angle θ to the x axis. A coordinate system nt is normal and tangent to the inclined surface. The forces and stresses on this surface, that we have labeled Aθ , are shown in Figure 4.8.2. P n, n F P, Pt, x nt Figure 4.8.2 The normal and shearing stresses are σn = Pn Aθ τnt = − Pt Aθ (4.8.1) Since Aθ = A cos θ Pn = P cos θ Pt = P sin θ (4.8.2) where A is the area normal to the x axis, the stresses are σn = P cos2 θ = σx cos2 θ A τnt = − P sin θ cos θ = σx sin θ cos θ A (4.8.3) Note that the normal stress is a maximum at θ = 0◦ and a minimum at θ = 90◦ . The shear stress is zero at θ = 0◦ and 90◦ and reaches a maximum value at θ = 45◦ . (τnt )max = −σx sin 45◦ cos 45◦ = − σx 2 (4.8.4) At that angle the normal stress is σt = σx cos2 45◦ = σx 2 (4.8.5) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 145 Classical Analysis of the Axially Loaded Slender Bar In Figure 4.8.3 we plot the variation in the normal stress and shearing stress as the angle changes. The upper plot is σn = cos2 θ σx (4.8.6) τnt = − sin θ cos θ σx (4.8.7) and the lower plot is 1.0 n 0.5 x 0 45 90 135 180 0.5 nt 0 x -0.5 Figure 4.8.3 This emphasizes that the value of the stress component depends not only on the size of the force but also on the orientation of the surface upon which it acts. A comprehensive discussion of the stress transformation that occurs when axes are rotated is presented in Chapter 9. 4.9 Design of Axially Loaded Bars In analysis we specify the geometry, materials, loads, and restraints and ask for the displacements, strains, and stresses. We wish to verify that the structure meets minimum requirements on load carrying ability. In design we are more likely to be given the loads and some of the restraints but are asked what geometry and materials should be used to achieve a certain acceptable value of stress and displacement. We wish to create a structure that will meet minimum requirements on load carrying ability. Design, however, is a much broader subject than that just stated. In addition to the geometric restraints we may have restraints on weight, cost, and ease of manufacture. Safety and margins of safety are ever present. Aesthetics are an additional consideration in many cases, such as the body panels on an automobile and the interior of an airliner. Because of the broad nature of design it is largely a team effort. We cannot begin to study all of these concerns so we shall limit ourselves to geometry, geometric restraints, materials, and weight for a given state of loading. These are the immediate concerns of the structural analyst member of the team. For an axially loaded slender bar the geometry consists of the length and the cross sectional area. The exact shape of the cross section might be determined by considerations of manufacturing, fasteners and --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 146 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods restraints, points or regions of load application, or other considerations. We shall restrict our discussion to the role of modifying the geometry and material to improve the design. ########### Example 4.9.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Problem: The bar in Example 4.5.2, repeated here in Figure (a), is made of aluminum for which the L yield stress is 255 MPa. If the load is applied at x0 = , what should the area of the cross section of the 2 bar be to limit the stress to two thirds of the yield stress, that is, limit the stress to 170 MPa? Repeat for 3L . x0 = 4 y F x0 x L Figure (a) Solution: Insert values in the equations obtained in Example 4.5.2 and calculate the areas. In Example 4.5.2 we found the stresses to be x0  F F x0 1− σ2x = − σ1x = A L A L In this case x0 = L/2. To the left of the load at x = L/2 the required area would be  F L F x0  F F = 170 MPa → A = 1− = 1− = σ1x = A L A 2L 2A 340 (a) (b) To the right of the load the required area would be σ2x = − F L F F x0 =− =− = −170 MPa → A L A 2L 2A A= F 340 (c) Both values define the same area. 3L what would be the minimum area. Note that this is a Now let us ask if the load is located at x0 = 4 uniform bar throughout its length. To the left of the load the required area would be  F 3L F x0  F F = 170 MPa → A = (d) 1− = 1− = σ1x = A L A 4L 4A 680 To the right of the load the required area would be σ1x = − F 3L 3F F x0 =− =− = −170 MPa → A L A 4L 4A A= 3F 680 For a uniform bar the larger of these two areas must prevail. ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (e) P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 147 Classical Analysis of the Axially Loaded Slender Bar Consider now that there are conditions on both stress and displacement. In the simplest of cases of a two force member of given length with a known axial force we select a material, stress level, and/or displacement level and calculate the cross section area to satisfy those conditions. Thus, for an internal force P = F, a material with Young’s modulus, E, a maximum allowable stress, σ max , and a maximum total displacement, δ, we obtain σx = P A → P σmax A= or δ= FL EA → A= FL Eδ (4.9.1) Then choose the larger area of the two. This would be result in the minimum weight for these conditions and this particular material. A change to a stiffer material would decrease the area based on displacement but might also require a change in the choice of maximum allowable stress. ########### Example 4.9.2 Problem: A slender bar 1000 mm long has an axial load of 8500 N as shown in Figure (a). The material choice is between aluminum and steel with the condition that the maximum stress may not exceed two thirds of the yield stress. The minimum weight design is sought. Case 1: No restriction on displacement Case 2: The total elongation is restricted to 2 mm 8500 N 8500 N 1000 mm Figure (a) Solution: Apply Equations 4.8.1, calculate weights, and compare. Materials with the following properties were chosen where ρ is the mass density: Aluminum: E = 68950 Mpa σyield = 255 MPa −6 ρ = 2.72·10 kg/mm Steel: E = 206800 MPa σyield = 342 MPa 3 ρ = 7.86·10−6 kg/mm 3 Case 1: The areas based on stress only and the corresponding weights are Wt = 50 · 1000 · 2.72 · 10−6 = 0.136 kg Steel: A= P 8500 3 · = 37.28 mm 2 = 2 342 2 σy 3 Wt = 37.28 · 1000 · 7.86 · 10−6 = 0.293 kg The aluminum bar is lighter. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Aluminum: 8500 3 P · = 50 mm 2 = A= 2 255 2 σy 3 P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 148 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Case 2: The areas based on the stress and the corresponding weights are the same as Case 1. For the aluminum bar the displacement would be δ= 8500 · 1000 PL = = 2.47 mm AE 50 · 68950 (c) A= PL 8500 · 1000 = = 61.64 mm 2 δE 2 · 68950 (d) to meet both requirements. Its weight would be W t = 61.64 · 1000 · 2.72 · 10−6 = 0.168 kg (e) For the steel bar the displacement would be δ= 8500 · 1000 PL = = 1.10 mm AE 37.28 · 206800 (f) The area needed to meet the stress requirement is more than is needed to meet the displacement criteria, but it cannot be reduced because of the stress criteria. Thus the weight of the steel bar necessary to meet both requirements is Wt = 37.28 · 1000 · 7.86 · 10−6 = 0.293 kg (g) The aluminum bar wins. Example 4.9.3 Problem: Redesign the cross sectional area of the beam in Example 4.4.1 for minimum weight. Assume a maximum stress but no displacement criteria; however, knowledge of the displacement is desired. y f x (x ) = f 0 x L Figure (a) Solution: Find the area for a constant maximum stress level and the resulting displacement using the equations in Section 4.6. The internal force is Fx = f 0 (L − x) − P(x) = 0 → P(x) = f 0 (L − x) (a) For a constant stress, σx , the area is A= Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS f0 P (L − x) = σx σx Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- therefore the area would have to be increased to P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 149 Classical Analysis of the Axially Loaded Slender Bar For a bar cross section of constant width the side view would be a triangular shape as shown in Figure (b). Figure (b) EA du =P dx → u(x) = 1 E σx f 0 (L − x) σx x +c = x dx = f0 E E (L − x) σx 1 P dx = A E (c) where c = 0 from the fixed boundary condition at the left end. For reasons other than stress values the actual choice would probably not be a bar that converges to a sharp point; however, some taper would be desirable in many cases. ########### As the loading gets more complicated the design process becomes more difficult. For statically determinate problems the areas can be calculated directly to meet stress requirements and then displacements can be analyzed to see if a displacement criterion has been met. For indeterminate cases the internal forces depend upon the restraints as well as the loads and so the second order displacement equations must be solved before stress levels can be determined. ########### 4.10 Analysis and Design of Pin Jointed Trusses We have just developed methods of analysis for axially loaded slender bars. Often several bars are joined together to form a structure called a pin jointed truss. The joints where bars meet are necessarily pinned to allow rotation so that no moments are applied to the bars, thus all members are axially loaded. It follows also that loads and restraints can only be applied at the joints. Such structures are easy to construct and easy to analyze and have important but limited use in practice. Nevertheless they are an interesting study that provides additional insight into the design, analysis, and construction of structures. Consider the simplest of trusses – one made of just two members as shown in Figure 4.10.1. A B Fx C Fy Figure 4.10.1 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The displacement is P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 150 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods If all you want is the stress it is easy to find. This is a statically determinate structure and a summation of forces at the point of load application will show you the internal axial load in each member. In fact in Section 2.3 we studied the forces in determinate trusses and found values in Examples 2.3.1 and 2.3.2. Now we must find stresses and displacements. The stress is easily determined since we know the internal forces in the members. The deflection requires us to consider joint compatibility in conjunction with knowledge of internal bar stresses. This will be dealt with next, through an example problem. It is common practice to show only line drawings such as in Figure 4.10.2 with the understanding that the joints are pinned. y x Figure 4.10.2 In this two dimensional structure you can find the support reactions by summation of forces in the x and y directions and the moment about the z axis. Starting at either support you can use summation of forces to find the internal force in each member that meets at that joint. As you work your way along from joint to joint in proper order there are never more than two unknowns at any joint. Thus quite large truss structures can be designed and analyzed in this way. Years ago many highway bridges appeared to be constructed this way and while the joints were not truly pinned the analysis was effective considering the large margins of safety allowed. They still are called truss style bridges. It should be noted that the failure of a single pin or a single member ensures the failure of the structure – so statically determinate pin jointed trusses are not recommended when safety is important. ########### Example 4.10.1 Problem: Consider the simple truss shown in Figure (a). Find the displacements and stresses. Let A1 = A2 = 400 mm2 F = 10000 N E = 206840 MPa. (a) Solution: Find the force in each member by static equilibrium. From the forces find the stresses. Find the displacements from the elongations of the members. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 151 Classical Analysis of the Axially Loaded Slender Bar 1000 mm 1000 mm B A A1 = 400 mm2 A2 = 400 mm2 1000 mm 30 deg C F = 10000 N Figure (a) We draw a free body diagram of the joint at C, where the load is applied as shown in Figure (b). P2 P1 F = 10000 N Figure (b) From summation of forces, where P1 and P2 are the internal forces in the left and right truss members respectively, we have  Fx = 10000 cos 30◦ − P1 cos 45◦ + P2 cos 45◦ = 0 (b)  Fy = −10000 sin 30◦ + P1 sin 45◦ + P2 sin 45◦ = 0 We can insert the values and put that in matrix form.      P1 −1 1 −8660 0.707 = P2 1 1 5000 (c) Solving we get P1 = 9660.5 N P2 = −2588.4 N P1 9660.5 = 24.15 MPa = A1 400 P2 −2588.4 = − 6.471 MPa → σ2 = = A2 400 → σ1 = (d) While finding the internal forces and hence the stresses can be easy, finding the displacement is a bit more difficult. We do know that members 1 and 2 will elongate an amount given by (see Example 4.3.1, Equation (l)) δ1 = P1 L 1 = 0.1651 mm EA1 δ2 = P2 L 2 = −0.0442 mm EA2 (e) In Figure (c) the dashed lines represent the original unloaded position and the solid lines represent the final loaded position – greatly exaggerated of course. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 152 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods u υ Figure (c) Let us suppose that the joint C undergoes a displacement of magnitude u in the horizontal direction and a displacement of magnitude v in the vertical direction as indicated. First let us look at member 1 and the possible u and v displacements due to its elongation. Since the actual displacements are very small we note that the change in angle of the position is also very small. Assuming that the angle has not changed we can say that v sin 45◦ + u cos 45◦ = δ1 = 0.1651 (f) v sin 45◦ − u cos 45◦ = δ2 = −0.0442 (g) and in a similar way that Solving for u and v u= (0.1651 + 0.0442) = 0.1480 mm 2 cos 45◦ (h) (0.1651 − 0.0442) = 0.0855 mm v= 2 sin 45◦ ########### --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- We must pause to note that there are several approximations made in arriving at this final analysis. As we have noted we did not recalculate the stress based on the change in area due to the Poisson’s ratio effect. We calculated the forces in the truss members based on the position of the forces in the undeformed structure, that is, we did not account for the position change of the applied forces due to the displacement. In finding the displacements we use the angular position of the truss members in their undeformed positions. Does this make our values invalid or, at least suspect? Not if the displacements are small enough. More detailed analysis confirms that these are valid assumptions for the slender bars of the kind we are considering here. Many pinned jointed trusses are indeterminate. For example, just add another member to the one in Figure 4.10.1 as shown in Figure 4.10.3. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 153 Classical Analysis of the Axially Loaded Slender Bar Fx Fy Figure 4.10.3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- In two dimensions you now have three unknown internal forces but just two equilibrium equations. The necessary third equation requires information about displacement. This will be taken up in Chapter 5. 4.11 Work and Energy—Castigliano’s Second Theorem When a force moves through a displacement work is done. Consider the experiment we used to define Young’s modulus. The force displacement curve from Figure 3.4.2 is reproduced here as Figure 4.11.1. Within the elastic range the work done as the bar displaces an amount under load F is shown by the shaded triangle. We call this the external work, W e , and note that its value is 1 F 2 We = (4.11.1) F F0 d Figure 4.11.1 As the body deforms the external work will produce stresses and strains which store energy in the body equal to the work done. To see this we use the stress-strain curve from Figure 3.4.3 repeated here as Figure 4.11.2. Within the elastic range this energy, called strain energy, U, is shown by the shaded triangle. We note that its value is U= 1 2 σx εx dV (4.11.2) V Assuming the stress is constant across the cross section of the bar we have εx = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS σx E (4.11.3) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 154 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods σx σ0 εx Figure 4.11.2 and so U= 1 2 σx εx dV = V 1 2 V σx2 dV E (4.11.4) We define the internal work, W i , done as the negative of the strain energy, that is, W i = −U (4.11.5) and note the sum of the external and internal work is zero We + Wi = We − U = 0 (4.11.6) Thus we can say that the external work done is equal to the strain energy stored. We = U (4.11.7) This can be used to solve problems, that is, to find the deflection in the direction of the load. 1 F 2 = 1 2 V σx2 dV E → = 1 F V σx2 dV E (4.11.8) ########### Example 4.11.1 Problem: Consider the axially loaded bar shown in Figure (a). Find the displacement at the end where the load is applied. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- F L Figure (a) Solution: Equate external work and internal strain energy and solve for the displacement in terms of the applied force. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 155 Classical Analysis of the Axially Loaded Slender Bar The deflection at the right end is u(L) = δ. The external work done as the load is applied and reaches its final value is We = 1 Fδ 2 (a) U= 1 2 σx εx dV = V 1 2 1 P2 σx2 dV = E 2 EA2 V L dx 0 dA = A 1 P2 L 2 EA (b) By equating the external work to the internal strain energy we can find the displacement at the point of application of the force and in the direction of the force. Since from static equilibrium P = F We = U → 1 P2 L 1 F2L 1 Fδ = = 2 2 EA 2 EA → δ= FL EA (c) ########### External work and strain energy can be useful in solving problems where springs are involved. When a spring is displaced by a force work is done and strain energy is stored in the spring. From Figure 2.3.8 and Equation 2.3.3 we note that We = 1 1 Fd = U = kd 2 2 2 (4.11.9) This is used in the next example. ########### Example 4.11.2 Problem: Find the displacement δ at the point where the load is applied for the rigid bar and spring shown in Figure (a). The bar is 1 meter long. F k 560 mm 1000 mm Figure (a) Solution: Equate external work and strain energy and solve for the displacement. For small displacements the displacement at the spring, d, is related to the displacement at the end, δ, by d = 0.56δ Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Given that σx = P/A the strain energy is P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 156 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Equating external work and strain energy gives us We = U → 1 1 Fδ = kd 2 2 2 → 1 1 Fδ = k(0.56δ)2 2 2 → δ= F 0.3136 · k (b) Check using static equilibrium by summing moments about the left end.  1000 · F 560 · k 1000 · F F d = = → δ= 0.56 0.56 · 560 · k 0.3136 · k Mz = 560 · kd − 1000 · F = 0 → d= (c) Static equilibrium and work and energy provided the same answer. ########### We can also find the displacement in trusses at the location of the force and in the direction of the force using work and energy. We equate the work done by the applied load to the strain energy stored in the truss members. The work done is We = 1 Fδ 2 (4.11.10) The strain energy in a two force member was found in Example 4.11.1 to be U= 1 P2 L 2 EA (4.11.11) and so in a truss we just sum all the members. U= Pn2 L n En A 1 2 (4.11.12) ########### Example 4.11.3 Problem: Find the displacement of the truss in Example 4.10.1 by work and energy. 1000 mm 1000 mm B A A1 = 400 mm2 A2 = 400 mm2 30 deg C F = 10000 N Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 1000 mm P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 157 Classical Analysis of the Axially Loaded Slender Bar Solution: Equate external work and strain energy and solve for the displacement. The strain energy is U = = 1 P22 L 2 1 P12 L 1 + 2 EA 2 EA (a) 1 (9660.5)2 1414 1 (−2588.4)2 1414 + = 854.74 N · mm 2 206840 · 400 2 206840 · 400 The external work and displacement are 1 Fδ = 854.74 2 → δ= 2 · 854.46 2 · 854.46 = = 0.1709 mm F 10000 (b) Note that this gives the displacement in the direction of the force. This checks with the values in Example 4.10.1, Equation (g), as follows: ' & (c) δ = u 2 + v 2 = (0.1480)2 +(0.0855)2 = 0.1709 mm ########### As the trusses get more complicated finding the stresses may be possible with reasonable effort; however, finding the displacements by the geometric resolution of the elongation of each member becomes tedious and prone to error. The simple solution by equating the work done to the strain energy is helpful but limited only to the displacement in the direction of a single load. If there are multiple loads or there is a desire for a displacement at a joint with no load the method does not work. Fortunately there is a work and energy method that can be extended to those cases. For any statically determinate structure we can find the strain energy in terms of discrete applied forces. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- U = U (F1 , F2 , . . . , Fi , . . . , Fn ) Now let us increase the ith force an amount order terms. U = U (F1 , F2 , . . . , Fi + (4.11.13) Fi , do a Taylor’s series expansion, and neglect the higher Fi , . . . , Fn ) = U (F1 , F2 , . . . , Fi , . . . , Fn ) + ∂U ∂ Fi Fi + terms of order ( Fi2 ) and higher Now let us look at the same event in another way. If we apply the force applied the external work done by it is We = 1 2 Fi u i (4.11.14) Fi before the other forces are (4.11.15) Since the external work done is equal to the strain energy stored the strain energy would now be U= 1 2 Fi u i (4.11.16) If now the other forces are applied the force Fi undergoes an additional displacement u i and does more work in the amount Fi u i and the strain energy changes a like amount. Both u i and ui are Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 158 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods displacements in the direction of the force. Since the order of load application does not affect the final value of the strain energy we can say that the total strain energy is now 1 2 U = (F1 , F2 , . . . , Fi , . . . , Fn ) + Fi u i + Fi u i (4.11.17) The strain energy represented in Equation 4.11.14 is the same as the strain energy represented in Equation 4.11.17 and so 1 Fi u i + 2 ∂U = u(F1 , F2 , . . . , Fi , . . . , Fn ) + Fi ∂ Fi Fi , . . . , Fn ) + u(F1 , F2 , . . . , Fi + And now we neglect the higher order terms including Fi u i + U = U + ∂U ∂ Fi 1 2 Fi Fi u i (4.11.18) Fi u i to obtain Fi u i = → ∂U ∂ Fi Fi (4.11.19) From which we can conclude that ui = ∂U ∂ Fi (4.11.20) This is known as Castigliano’s second theorem. Remember that this works only if the internal strain energy can be presented in terms of external concentrated loads, that is, it is determinate. The resulting displacement is in the direction of the force. Let us put this to the test with an example. ########### Example 4.11.4 Problem: Let us reexamine Example 4.10.1 using Castigliano’s theorem. 1000 1000 A1 A2 1000 F1 F F2 Figure (a) Solution: Find the strain energy in terms of the applied load and differentiate. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 159 Classical Analysis of the Axially Loaded Slender Bar We shall divide the force into components so that we can find the displacement in both the x and y directions. A free body diagram of the lower joint is shown in Figure (b) P2 P1 F1 F F2 Figure (b) Summing forces:   Fx = F1 − P1 cos 45◦ + P2 cos 45◦ = 0 Fy = −F2 + P1 sin 45◦ + P2 sin 45◦ = 0 (a) Solve for the internal forces in terms of the applied forces. P1 = F1 + F2 √ 2 P2 = F2 − F1 √ 2 (b) The strain energy is U= 1 P22 L 2 1 (F1 + F2 )2 L 1 1 (F2 − F1 )2 L 2 1 P12 L 1 + = + 2 EA 2 EA 4 EA 4 EA (c) We note that F1 and F2 have the values F1 = F cos 30◦ = 10000 · 0.866 = 8660N F2 = F sin 30◦ = 10000 · 0.5 = 5000N (d) but we must differentiate first to find the displacements and then insert the values. Since L 1 = L 2 = L --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- u= 1414 L L ∂U {(F1 + F2 ) −(F2 − F1 )} = F1 = 8660 = 0.1480 = ∂ F1 2EA EA 206840 · 400 v= L 1414 L ∂U {(F1 + F2 ) +(F2 − F1 )} = F2 = 5000 = 0.0855 = ∂ F2 2EA EA 206840 · 400 (e) This agrees with the solution found in Example 4.10.1. Note that the sign difference for the vertical displacement follows from the fact that the displacement here is in the direction of the load. ########### In cases where we want the displacement where there is no load, place a load there, find the strain energy in terms of all the loads, differentiate with respect to that load, then set the value of that load to zero. Here is a simple example. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 160 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 4.11.5 Problem: Find the displacement of joint B in the following truss in both the x and y directions. The truss members are made of steel with E = 206,800 MPa. The cross section area of member AB is 400 mm2 and that of member CB is 800 mm2 . A 1 meter B C F 1 meter Figure (a) Solution: Use Castigliano’s theorem. To find the horizontal displacement we must add a horizontal force in order to be able to differentiate with respect to it and then let it go to zero. The horizontal force is added in Figure (b). A 1 meter B C F1 F2 1 meter Figure (b) From summation of forces at point B:   Fy = FBA sin 45◦ − F2 = 0 → FBA = √ F2 = 2F2 ◦ sin 45 Fx = FBA cos 45◦ + FBC − F1 = 0 → FBC = F1 − FBA cos 45◦ = F1 − (a) ◦ cos 45 F2 = F1 − F2 sin 45◦ --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 161 Classical Analysis of the Axially Loaded Slender Bar The strain energy is 1 (FBA )2 L BA 1 (FBC )2 L BC 1 U= + = 2 EABA 2 EABC 2 2 √ 2F2 L BA EABA + 1 (F1 − F2 )2 L BC 2 EABC (b) Differentiate first to find the displacements and then insert F1 = 0. (0 − F2 ) 1000 (F1 − F2 ) L BC 2F2 L BA 2F2 1414 ∂U − = 4.032 · 10−5 F2 = − = ∂ F2 EABA EABC 206800 · 400 206800 · 800 (F1 − F2 ) L BC (0 − F2 ) 1000 ∂U = −0.6044 · 10−5 F2 u= = = ∂ F1 EABC 206800 · 800 v= (c) The displacement in the x direction is positive in the direction of the force F 1 which explains its negative value. Example 4.11.6 Problem: Extend Example 2.5.1 to find displacement of the three dimensional truss shown in Figure (a) in the direction of the applied weight. All three members have the same cross sectional area A and have the same Young’s modulus E. Solution: Given the forces in the members from Example 2.5.1 sum their contributions to the strain energy. Use Castigliano’s theorem to find the displacement. In Example 2.5.1 we found the forces in the members to be √ FAD = FAB = −0.559W (a) FAC = 2W y y --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 1000 mm C 1000 mm D z x B 1000 mm 1000 mm W A Isometric view Side view W x Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 162 6:57 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The strain energy is 1 2EA 1 = 2EA U = $ % (FAB )2 L AB +(FAC )2 L AC +(FAD )2 L AD ! (b) " √ 2(−0.559W ) L AB + ( 2W )2 L AC 2 The lengths of the members are L AB = L AD = 1118 mm L AC = 1414 mm (c) The displacement is δy A =  √ 2 1 W ∂U = 4 (−0.559)2 W · 1118 + 2 W · 1414 = 2112.7 ∂W 2EA EA (d) ########### 4.12 Summary and Conclusions The axially loaded bar is studied at some length because the simplified geometry, loadings, and restraints allow us to concentrate on the formulation of the problem and the solution process. In this chapter we have taken a sharp departure from many other traditional introductory texts on the subject of mechanics of solids. We have chosen to derive and then emphasize solving the differential equations in terms of the distributed displacements. In traditional introductory texts emphasis is placed on the force formulation of statically determinate problems. Stresses are then found from the internal forces determined from the equations of static equilibrium. When statically indeterminate problems are encountered a limited introduction of displacements at the points of restraint application are introduced to provide the additional equations necessary for a solution. Only after extensive use of this method are distributed displacements introduced, if at all. In our displacement formulation we do consider determinate problems; however, we solve for the distributed displacements, not just the discrete displacement at ends of the bar, using either the first order equations or the second order equations. We also solve both determinate and indeterminate cases using the second order equation for displacements. In this latter case the internal forces and stresses are found by differentiating the displacements. The classical analysis equations of the displacement method are summarized in Equations 4.3.8, repeated here as Equation 4.12.1. Fc → f x (x) P(x) σx (x) εx (x) dP = − f x (x) dx u(x) εx (x) = σx (x) = Eεx ← ρ du dx (4.12.1) du P(x) = σx A = EAεx = EA dx d du dP = EA = − f x (x) dx dx dx Several examples in Sections 4.4 and 4.5 show how these equations are used to find displacements, strains, stresses, internal forces, and restraint forces when EA has a constant value. The first order equation can --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come 163 Classical Analysis of the Axially Loaded Slender Bar be used for determinate cases only. EA du = P(x) dx → u(x) = P(x) dx + c EA (4.12.2) The second order equation can be used for determinate or indeterminate cases. EA d 2u = − f x (x) dx2 → u(x) = − 1 EA f n (x)dx + ax + b (4.12.3) In Section 4.6 we examine the solution for variable cross sections. The first order equation can be used for determinate cases only. EA(x) du = P(x) dx → u= P(x) dx + c EA(x) (4.12.4) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The second order equation can be used for determinate or indeterminate cases.  du d 1 f x (x)d x + a d x + b EA(x) = − f x (x) → u(x) = − dx dx EA(x) (4.12.5) In Section 4.7 we add the thermal strain component to the governing equations. The changes in the equations are summarized here. Fc → f x (x) P(x) dP = − f x (x) dx σx (x) εxmech (x) T (x) εxT otal = εxmech + εx T = u(x) ← ρ du σx +α T = E dx du − Eα T σx (x) = E dx  du −α T P(x) = σx A = EA dx   d dP du = −α T = − f x (x) EA dx dx dx (4.12.6) In Section 4.8 the presence of shearing stress on internal surfaces is examined. The design of axially loaded bars is discussed briefly in Section 4.9. In analysis the known quantities are geometry, material properties, external forces, and restraints. The unknowns are internal forces, stress, strain, and displacement. In design we often invert the process of selecting some of these quantities. For example, we might ask what geometry and material would be best suited to achieve some known value of stress and displacement. The analysis and design of pin jointed trusses is examined in Sections 4.10 and 4.11. We equate the external work to the internal strain energy to solve simple problems then extend work and energy methods to include Castigliano’s second theorem ui = ∂U ∂ Fi (4.12.7) to find displacements in statically determinate axially loaded bars and trusses. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-04 P2: ABC JWST071-Waas July 4, 2011 6:57 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 5 A General Method for the Axially Loaded Slender Bar 5.1 Introduction We have in Equations 4.3.8 the complete set for solving for stress, strain, and displacement in axially loaded bars that are either statically determinate or indeterminate and which are loaded with either one or both of concentrated and distributed loads. As the geometry, loading, and restraint get more complicated the analytical solutions become more time consuming and tedious. In this day of computers there must be a way to organize the solution process more efficiently than is suggested by the example problems considered so far. Note that those examples have the simplest of geometry, restraints, and loads. In practice you rarely find problems that are that simple. The analytical solution methods examined in Chapter 4 require integrating the differential equations. In the process constants of integration are generated. These constants are found by inserting the solutions into a set of boundary conditions. This results in a set of linear algebraic equations in terms of the unknown constants of integration. The problem is thus reduced to solving sets of linear algebraic equations to find the constants of integration. In the years B.C. (Before Computers) every effort was made to keep the number of simultaneous linear algebraic equations that needed to be solved to a minimum, perhaps not more than a single digit number. Thus, dividing bars into segments and solving the differential equations by matching boundary conditions where segments meet was limited to just a few segments. Today, with modern digital computers, there is virtually no limit to the number of simultaneous linear algebraic equations that can be solved quickly and efficiently. With the advent of the digital computer a new way of organizing the solution of problems in solid mechanics has evolved and continues to improve with time. For now, we call this the general method because it can be adapted to work with all combinations of material, load, restraint, and geometry. As you shall see distinction between determinate and indeterminate is no longer a consideration in this method. 5.2 Nodes, Elements, Shape Functions, and the Element Stiffness Matrix Consider an axially loaded slender bar with only concentrated loads. We have seen in Examples 4.4.5 and 4.5.2 that we can express the distributed displacement for each segment of the bar between load locations in terms of the discrete displacements at points of load application and a local coordinate s. Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. --`,,,,``,``,,``,,`,`,,,,,``,,``- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 166 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods See Example 4.4.5, Equation (i), which is repeated here.    s m sm un u(sm ) = 1 − lm lm u n+1 (5.2.1) The local coordinate sm has its origin at one end of the segment which has been numbered segment m. Let us generalize this a bit by considering a bar divided into segments as shown in Figure 5.2.1. We shall call each segment an element and number it. The points at the ends of elements are called nodes and also are numbered. Two elements have a common node where they meet, thus node n is common to elements m-1 and m. element m-1 element m element m+1 x node n-1 node n node n+1 node n+2 s pm,n, un pm,n+1, un+1 lm Figure 5.2.1 A particular element, say element m, has a length lm with the left end labeled node n and the right end labeled node n + 1. We define the discrete displacement at node n as u n and at node n + 1 as u n+1 . The origin of the coordinate sm is at node n. We also define internal forces for element m, also called stress resultants, at node n as pm,n and at node n + 1 as pm,n+1 . These stress resultants are the same as P used in the classical analysis except for the sign convention. Note the direction of these forces in Figure 5.2.1. These forces are chosen to be positive in the same direction as the positive axis direction at both ends because it is easier to keep track of them in the derivation that follows. We used the definition for P (normal away from the face as positive, same as positive normal stress) in the differential equation derivations because it is convenient and is most common in the literature of classical analysis. This new definition is most common in the literature of the numerical analysis of solids. Applying Equation 5.2.1 to element m gives us the distributed displacement u (sm ).    sm s m un = [n] {rm } (5.2.2) u(sm ) = 1 − lm lm u n+1 where we have designated  sm [n] = 1 − lm sm lm  {rm } =  un u n+1  (5.2.3) The functions in the matrix [n] are called shape functions. We introduce the notation {rm } for the discrete nodal displacement to avoid confusion. To call it {u m } might confuse it with the displacement at node m. The strain displacement equation is    un 1  du m d −1 1 (5.2.4) = εxm = [n] {rm } = u n+1 dsm dsm lm --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 167 A General Method for the Axially Loaded Slender Bar    un Em  d −1 1 [n] {rm } = u n+1 dsm lm At nodes n and n + 1 we define the internal forces    un E m Am  d 1 −1 pm,n = −Am σxm = −E m Am [n] {rm } = u n+1 dsm lm σxm = E m εxm = E m pm,n+1 = Am σxm E m Am  d −1 = E m Am [n] {rm } = dsm lm 1   un u n+1 (5.2.5) (5.2.6)  The negative value at node n results from defining the positive internal force opposite to the positive direction of the stress, while at n + 1 they are in the same direction. The nodal internal forces can be expressed in matrix form.      E m Am un 1 −1 pm,n { pm } = = [km ] {rm } = (5.2.7) u n+1 pm,n+1 −1 1 lm Now we shall add these new quantities, {rm } and { pm }, to the summary of equations shown in Equation 4.3.8 and use the shape functions in Equation 5.2.2 to represent the displacement, strain, stress, and internal forces for an element. In summary for element m {rm } u (sm )    sm sm un u (sm ) = [n] {rm } = 1 − lm lm u n+1    un 1  d du m −1 1 = εxm = [n] {rm } = u n+1 dsm dsm lm     Em d un −1 1 σxm = E m εxm = E m [n] {rm } = u n+1 dsm lm    un E m Am  d 1 −1 pm,n = −Am σxm = −E m Am [n] {rm } = u n+1 dsm lm    un E m Am  d −1 1 pm,n+1 = Am σxm = E m Am [n] {rm } = u n+1 dsm lm      E A un 1 −1 pm,n m m { pm } = = = [km ] {rm } pm,n+1 u n+1 −1 1 lm { pm } σxm εxm (5.2.8) In the first row of Equation 5.2.8 we have the typical unknowns in the formulation of a problem as applied to a single element m. {rm } - nodal displacements u(sm ) - distributed displacement εxm - normal strain σxm - normal stress { pm } - internal forces Among the known quantities in any given problem are the geometry, as expressed by A and lm , and material properties, as expressed by E, of the slender bar. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- and Hooke’s law is P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 168 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Starting in the second row where we define the displacement in terms of shape functions and nodal displacements and we move down and to the left to define other quantities in terms of the shape functions. In the third row we have the strain displacement equation, and in the fourth row Hooke’s law. Note that since the distributed displacement is a linear function of sm the distributed strain and stress are constants over the whole length of the element. In the next two rows we express the internal forces at each node and then combine them in a single matrix.      E m Am un 1 −1 pm,n { pm } = = [km ] {rm } = (5.2.9) u n+1 pm,n+1 −1 1 lm where the matrix [km ] = E m Am lm  1 −1 −1 1  (5.2.10) is called the element stiffness matrix. In the next example we shall show that the element stiffness matrix contains exactly the same information as the differential equations in Chapter 4. ########### Example 5.2.1 Problem: Demonstrate that the element stiffness matrix contains the same information as the equations in Chapter 4 by applying it to the bar in Example 4.3.1, Figure (d). The figure is repeated here as Figure (a). y F x L Figure (a) Solution: Take the entire bar as a single element and adapt the element matrix equations to this configuration. Call the bar element 1 and number the nodes as shown in Figure (b). e1 n1 n2 Figure (b) The element matrix equations, Equation 5.2.11, are then      E 1 A1 u1 1 −1 p1,1 = p1,2 u2 −1 1 L1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 169 A General Method for the Axially Loaded Slender Bar From Figure (a) we note that the boundary conditions are u1 = 0 p1,2 = F (b) Also we note that u 2 and p1,1 are unknown. Since p1,1 is an unknown restraint force let us assign it the symbol p1,1 = R1 (c) Note that p1,1 and R1 have the same sign convention. The matrix equations now become      EA 0 1 −1 R1 = F u2 1 L −1 (d) If we expand the matrix equations and solve for the unknowns we get F= EA EA (−1 · 0 + 1 · u 2 ) = u2 L L → u2 = FL EA EA EA FL EA (1 · 0 − 1 · u 2 ) = − u 2 = − = −F R1 = L L L EA (e) To find the stress we use Equation 5.2.5. E −1 σx = L 1   0 u2  ⎡ ⎤ 0  F E E FL −1 1 ⎣ FL ⎦ = = = L L EA A EA (f) To find the distributed displacement use Equation 5.2.2. Since s1 and x are the same coordinate we have   x x F x x FL 0 = x (g) = u2 = u(x) = 1 − u2 L L L L EA EA This is the same as the solution obtained in Example 4.3.1. This almost trivial example is presented to emphasize that the element matrix equations contain exactly the same information as the classical equations presented in Chapter 4 for bars with concentrated loads. ########### 5.3 The Assembled Global Equations and Their Solution A bar with several concentrated loads must be divided into several elements. The shape functions of Equation 5.2.2 apply to any and all segments between the concentrated loads. Consider that a bar with multiple concentrated loads has been divided into a number of segments and the nodes and elements are numbered as indicated in Figure 5.3.1. We make sure that there is a node at each location of an applied force; however, we may have nodes at other locations where the loads are zero. The elements may be different lengths. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 170 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y e1 e2 em … … eN-1 x n1 n2 ... nn nn+1 … nN Figure 5.3.1 To formulate the problem we must write a stiffness matrix for each element and then combine all the elements to represent the whole bar. Then we must add the applied loads and restraints to define the complete set of equations governing the behavior of the bar. The process of combining all the elements is called the assembly of the various element stiffness matrices to form the global stiffness matrix. For element 1 we have just nodes 1 and 2. We can say      E 1 A1 u1 1 −1 p1,1 { p1 } = = [k1 ] {r1 } = (5.3.1) u2 p1,2 −1 1 l1 For any intermediate element m we have nodes n and n+1. We have      E m Am un 1 −1 pm,n { pm } = = = [km ] {rm } pm,n+1 u n+1 −1 1 lm For the last element N-1 we have nodes N-1 and N. We have      E N−1 AN−1 u N−1 1 −1 pN−1,N−1 { pN−1 } = = [kN−1 ] {rN−1 } = uN pN−1,N −1 1 lN−1 (5.3.2) (5.3.3) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Note that we always have one more node than we have elements. We immediately encounter a problem when we try to assemble these equations. The assembly process requires a summation of the element stiffness matrices but by the rules of matrix addition. Two sets of equations in matrix form {X } = [A] {x} {Y } = [B] {y} (5.3.4) can be assembled, or summed, only if {x} is exactly equal to {y} and [A] and [B] have the same number of rows and columns, or the same order. If so, we obtain {Z } = {X } + {Y } = [A] {x} + [B] {y} = ( [A] + [B] ) {x} (5.3.5) Element matrices in Equations 5.3.1-3 cannot be added directly because each one refers to a different pair of nodes. To overcome this problem we imbed each element matrix in a global matrix format. For example, for element 1 the element stiffness matrix is reformatted as follows. ⎡ ⎤ E 1 A1 E 1 A1 − 0 0 ... 0 ⎢ l1 ⎥⎡ ⎤ l1 ⎡ ⎤ ⎢ ⎥ u1 p1,1 ⎢ ⎥ ⎢ E 1 A1 ⎥ ⎢ u2 ⎥ E 1 A1 ⎢ p1,2 ⎥ ⎢ ⎥⎢ ⎥ 0 0 . . . 0 − ⎢ ⎥ ⎢ ⎥ ⎢ u3 ⎥ l l ⎢ 0 ⎥ 1 1 ⎥⎢ ⎥ { p1 } = ⎢ (5.3.6) ⎥ = [k1 ] {u} = ⎢ ⎢ ⎥ ⎢ u4 ⎥ ⎢ 0 ⎥ ⎢ 0 ⎥⎢ ⎥ 0 0 0 . . . 0 ⎣ ... ⎦ ⎢ ⎥⎣...⎦ ⎢ 0 0 0 0 ... 0 ⎥ ⎢ ⎥ uN 0 ⎣ ... ... . . . . . . . . . . . .⎦ 0 0 0 0 ... 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 171 A General Method for the Axially Loaded Slender Bar This has exactly the same information as the shorter form. The stiffness matrix for element 2 becomes ⎡ 0 ⎢ ⎤ ⎢ 0 0 ⎢ ⎢ ⎢ p2,2 ⎥ ⎢ ⎥ ⎢ ⎢ ⎥ ⎢ { p2 } = ⎢ p2,3 ⎥ = [k2 ] {u} = ⎢ ⎢ 0 ⎢ 0 ⎥ ⎢ ⎣ ... ⎦ ⎢ ⎢ 0 ⎢ 0 ⎣... ⎡ 0 0 E 2 A2 l2 − 0 E 2 A2 − l2 E 2 A2 l2 0 ... 0 ... E 2 A2 l2 0 ... 0 ... 0 0 ... 0 ... ... ... 0 ... 0 0 ⎤ ⎥⎡ ⎤ 0 ⎥ u1 ⎥ ⎥ ⎢ u2 ⎥ ⎥ ⎥⎢ ⎥ ⎢ u3 ⎥ ⎢ ⎥ 0 ⎥ ⎥ ⎢ u4 ⎥ ⎥ ⎥⎢ ⎥⎣...⎦ 0 ⎥ ⎥ uN ...⎦ 0 (5.3.7) and so on until all the elements are reformatted. Now when all the element matrices are added together we get the global equation set shown in Equation 5.3.8. ⎡ ⎤ p1,1 ⎢ ⎥ p1,2 + p2,2 ⎢ ⎥ ⎢ ⎥ p2,3 + p3,3 ⎢ ⎥ ⎢ ⎥ p + p 3,4 4,4 ⎢ ⎥ ⎣ ⎦ ... pN−1,N−1 + pN−1,N ⎡ E A E 1 A1 1 1 − ⎢ l1 l1 ⎢ ⎢ E 1 A1 E 1 A1 E 2 A2 ⎢− + ⎢ l1 l1 l2 ⎢ ⎢ ⎢ E A 2 2 ⎢ − =⎢ 0 l2 ⎢ ⎢ ⎢ ⎢ 0 0 ⎢ ⎢ ... ⎢ ... ⎣ 0 0 0 0 E 2 A2 − l2 E 3 A3 E 2 A2 + l2 l3 E 3 A3 − l3 ... E 3 A3 l3 E 4 A4 E 3 A3 + l3 l4 ... 0 0 0 − 0 ... ... 0 ... 0 ... 0 ... ... E N−1 AN−1 lN−1 ... ⎤ ⎥ ⎥ ⎥⎡ ⎤ ⎥ u ⎥ 1 ⎥⎢ ⎥ ⎥ ⎢ u2 ⎥ ⎥⎢ ⎥ ⎥ ⎢ u3 ⎥ ⎥⎢ ⎥ ⎥ ⎢ u4 ⎥ ⎥⎣ ⎦ ⎥ ... ⎥ ⎥ uN ⎥ ⎥ ⎦ (5.3.8) Now we must define the forces acting at the nodes. The forces acting at node n+1 are shown in Figure 5.3.2. The applied load Fn+1 is shown above the node but is acting on the node. Fn+1 pm,n+1 pm+1,n+1 node n+1 Figure 5.3.2 From equilibrium of the forces at the node the external load Fn+1 is equal to the sum of the internal nodal forces, so Fn+1 − pm,n+1 − pm+1,n+1 = 0 → pm,n+1 + pm+1,n+1 = Fn+1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (5.3.9) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 172 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods It follows that ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ ⎤ ⎤ ⎡ F1 p1,1 ⎢ ⎥ ⎥ p1,2 + p2,2 ⎥ ⎢ F2 ⎥ ⎢ ⎥ ⎥ p2,3 + p3,3 ⎥ = ⎢ F3 ⎥ = {F} ⎥ ⎢ F4 ⎥ p3,4 + p4,4 ⎥ ⎥ ⎢ ⎦ ⎣...⎦ ... FN pN−1,N−1 + pN−1,N (5.3.10) We now represent Equation 5.3.8 as {F} = [K ] {r } (5.3.11) where [K] is called the global stiffness matrix. Since the applied forces {F} are known it would appear that {F} = [K ] {r } is an acceptable set of equations to solve. If we try we run into a problem. It so happens that [K] is a singular matrix, that is, the determinant of [K] is equal to zero. |K | = 0 (5.3.12) ⎡ E A 1 1 ⎢ l1 ⎢ ⎡ ⎤ ⎢ ⎢− E 1 A1 R1 ⎢ l1 ⎢ F2 ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ ⎢ F3 ⎥ ⎢ 0 ⎢ ⎥ ⎢ F4 ⎥ = ⎢ ⎢ ⎥ ⎢ ⎢ ⎣...⎦ ⎢ ⎢ 0 ⎢ FN ⎢ ⎢ ... ⎣ 0 E 1 A1 l1 E 2 A2 E 1 A1 + l1 l2 E 2 A2 − l2 0 − 0 ... E 2 A2 l2 E 3 A3 E 2 A2 + l2 l3 E 3 A3 − l3 ... E 3 A3 l3 E 4 A4 E 3 A3 + l3 l4 ... 0 0 0 0 0 − − ... 0 ... 0 ... 0 ... 0 ... ... E N−1 AN−1 lN−1 ... ⎤ ⎥ ⎥ ⎥⎡ ⎤ ⎥ 0 ⎥ ⎥⎢ ⎥ ⎥ ⎢ u2 ⎥ ⎥⎢ ⎥ ⎥ ⎢ u3 ⎥ ⎥⎢ ⎥ ⎥ ⎢ u4 ⎥ ⎥⎣ ⎦ ⎥ ... ⎥ ⎥ uN ⎥ ⎥ ⎦ (5.3.13) Of course [K] is still a singular matrix; however, we now have one less unknown nodal displacement and so there is one less equation to solve as a result of applying the restraint. We now do what is called partitioning the matrix equations. Let us separate the first equation from the rest as indicated by the Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- This means that the equations have no unique solution. Why is that? We have not yet applied boundary conditions. To complete the problem we must add boundary restraints and note that restraint forces exist at nodes where displacements are restrained. At least one of the nodal displacements must have a known value to satisfy static equilibrium. At any restrained node we must add a restraint force to the force vector {F}. For example, if the displacement at node 1 is zero then u1 = 0 and there is a restraint force R1 added as follows: P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 173 A General Method for the Axially Loaded Slender Bar dashed lines in Equation 5.3.14 and write the remaining equations separately. ⎡ E 1 A1 E 1 A1 − 0 0 ⎢ l1 l1 ⎢ ⎢ ⎡ ⎤ ⎢ ⎢ R1 ⎢ E 1 A1 E 2 A2 E 2 A2 E 1 A1 + − 0 − ⎢ ⎥ ⎢ l l l l2 ⎢F ⎥ ⎢ 1 1 2 ⎢ 2⎥ ⎢ ⎢F ⎥ ⎢ E 3 A3 E 3 A3 E 2 A2 E 2 A2 ⎢ 3⎥ = ⎢ + − 0 − ⎢F ⎥ ⎢ ⎢ l l l l3 2 2 3 4 ⎢ ⎥ ⎢ ⎣. . . ⎦ ⎢ E 4 A4 E A E A 3 3 3 3 ⎢ 0 + 0 − FN ⎢ l l l4 ⎢ 3 3 ⎢ ... ... ... ... ⎢ ⎣ 0 0 0 0 ... ... ... ... ... ... 0 ⎤ ⎥ ⎥ ⎥ ⎥⎡ ⎤ ⎥ 0 ⎥ ⎥⎢ ⎥ 0 ⎥⎢ ⎥ ⎥⎢ u 2 ⎥ ⎥⎢ ⎥ ⎥⎢ u 3 ⎥ 0 ⎥⎢ ⎥ ⎥⎢ u 4 ⎥ ⎥⎣ ⎦ ⎥ ... ⎥ 0 ⎥ uN ⎥ ⎥ ... ⎥ E N−1 AN−1 ⎦ lN−1 (5.3.14) We get the following set of equations with known forces and unknown displacements. This set is not singular. ⎡ ⎤ E 1 A1 E 2 A2 E 2 A2 + − 0 . . . 0 ⎢ l1 ⎥ l2 l2 ⎢ ⎥ ⎢ ⎥⎡ ⎡ ⎤ ⎢ E 3 A3 E 3 A3 E 2 A2 E 2 A2 ⎥ u ⎤ F2 ⎢ − ⎥ + − ... 0 2 ⎢ ⎥⎢ l l l l ⎢ F3 ⎥ ⎢ 2 2 3 3 ⎥ ⎢ u3 ⎥ ⎢ ⎥ ⎢ ⎥ ⎥⎢ ⎢ F4 ⎥ = ⎢ E 4 A4 E 3 A3 E 3 A3 ⎥ ⎢ u4 ⎥ ⎢ ⎥ ⎢ ⎥ + . . . 0 0 − ⎥ ⎣...⎦ ⎢ ⎥⎣...⎦ l3 l3 l4 ⎢ ⎥ ... ... ... ... ... FN ⎢ ⎥ uN ⎢ ⎥ ⎢ ⎥ ⎣ E N−1 AN−1 ⎦ 0 0 0 ... lN−1 (5.3.15) These equations can be solved for the nodal displacements. And once they are solved we can find the unknown restraint force with the other partitioned equation. ⎡ ⎤ u2  ⎢ u3 ⎥  ⎢ ⎥ E 1 A1 u4 ⎥ (5.3.16) 0 0 ... 0 ⎢ [R1 ] = − ⎢ ⎥ l1 ⎣...⎦ uN To find the stresses we apply Equation 5.2.5 for each element. σxm = E m εxm = E m    un Em  d −1 1 [n] {rm } = u n+1 ds lm (5.3.17) This relatively simple act of defining shape functions turns the problem from one of solving differential equations to that of directly solving linear algebraic equations. The differential equations are solved when creating the shape functions once and for all for each class of elements we examine. Note that the strain displacement, stress-strain, and equilibrium equations are satisfied without compromise. For a bar with truly concentrated loads this is not an approximation. All these relations are --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 174 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods found directly from integration of the governing equation for an axially loaded slender bar from classical theory. This is exactly what you have been studying in Chapter 4. It is just presented in a different way. The restraints can be applied at either end or at both ends, thus possible formulations of a problem can include ⎡E A 1 1 ⎢ l1 ⎢ ⎢ E A ⎤ ⎢ ⎡ ⎢− 1 1 F1 ⎢ l1 ⎢ F2 ⎥ ⎢ ⎥ ⎢ ⎢ ⎢ F3 ⎥ ⎢ ⎥ ⎢ 0 ⎢ ⎢ F4 ⎥ = ⎢ ⎥ ⎢ ⎢ ⎢ ... ⎥ ⎢ ⎦ ⎢ ⎣ ⎢ 0 ⎢ ⎢ RN ⎢ ... ⎢ ⎢ ⎣ 0 − E 1 A1 l1 0 E 2 A2 E 1 A1 + l1 l2 − E 2 A2 l2 − 0 E 2 A2 l2 0 E 3 A3 E 2 A2 + l2 l3 0 − ... 0 E 3 A3 l3 ... − ... E 3 A3 l3 ... E 4 A4 E 3 A3 + l3 l4 ... ... 0 ... 0 0 ... ... ⎤ ⎥ ⎥ ⎥ ⎥⎡ ⎤ ⎥ u 0 ⎥ 1 ⎥⎢ u ⎥ ⎥⎢ 2 ⎥ ⎥⎢ ⎥ ⎥⎢ u 3 ⎥ 0 ⎥⎢ ⎥ ⎥⎢ u 4 ⎥ ⎥⎢ ⎥ ⎥⎣. . .⎦ ⎥ 0 ⎥ ⎥ 0 ... ⎥ ⎥ ⎥ E N−1 AN−1⎦ lN−1 (5.3.18) and E 1 A1 ⎢ l1 ⎢ ⎢ ⎡ ⎤ ⎢ ⎢ E A R1 ⎢ 1 1 ⎢ ⎥ ⎢− l 1 ⎢F ⎥ ⎢ ⎢ 2⎥ ⎢ ⎢F ⎥ ⎢ ⎢ ⎢ 3⎥ ⎢ 0 ⎢ ⎥= ⎢ F4 ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ ... ⎥ ⎢ ⎣ ⎦ ⎢ ⎢ 0 ⎢ RN ⎢ ⎢ ... ⎢ ⎢ ⎣ 0 ⎡ − E 1 A1 l1 0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- E 2 A2 E 1 A1 + l1 l2 − E 2 A2 l2 − 0 E 2 A2 l2 0 E 3 A3 E 2 A2 + l2 l3 0 − ... 0 E 3 A3 l3 ... − ... E 3 A3 l3 ... E 4 A4 E 3 A3 + l3 l4 ... ... 0 ... 0 0 ... ... ⎤ ⎥ ⎥ ⎥ ⎥ ⎥⎡ ⎤ ⎥ 0 0 ⎥ ⎥⎢ ⎥ ⎥⎢ u 2 ⎥ ⎥⎢ ⎥ ⎥⎢ u 3 ⎥ ⎥⎢ ⎥ 0 ⎥⎢ u 4 ⎥ ⎥⎢ ⎥ ⎥⎢. . .⎥ ⎥⎣ ⎦ ⎥ 0 ⎥ 0 ⎥ ... ⎥ ⎥ ⎥ E N−1 AN−1⎦ lN−1 (5.3.19) We would normally call a problem defined by the constraints in Equations 5.3.13 and 5.3.18 statically determinate and a problem defined by Equations 5.3.19 statically indeterminate and use different approaches in solving them. By this method we need make no such distinctions. In summary, we assign nodes where known concentrated loads and where known concentrated restraints are applied and at other points if we wish. We then divide the bar into elements and number the nodes and elements. Next we find the element matrix for each element { pm } = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  pm,n pm,n+1  =  −Am σm,n Am σm,n+1  = E m Am lm  1 −1 −1 1  un u n+1  = [km ] {rm } (5.3.20) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 175 A General Method for the Axially Loaded Slender Bar Next the known nodal restraints are inserted in {r}; the corresponding unknown restraint forces in {F}; and the known applied loads in {F}. This leaves us with unknown nodal displacements in {r} and unknown restraint forces in {F} to be determined. We then partition the global equations into on set containing the unknown nodal displacements and the known applied forces that we solve for the nodal displacements. In partitioning another set is found for solving for the restraint forces in terms of the now known nodal displacements Once we know the nodal displacements we can find the distributed displacement, the strain, the stresses, and the internal forces in each element from the equations presented in Equation 5.2.6. This whole process is best understood by examples. ########### Example 5.3.1 Problem: A bar with both ends fixed has a concentrated force at the midpoint as shown in Figure (a). Find the displacements, internal forces, and stresses. Use the general method with two elements. F L/2 L Figure (a) Solution: Divide into elements, number the nodes and elements, assemble the element matrices, add the applied forces and restraints, partition, and solve. With nodes at the two ends and at the location of the applied load we choose two elements of equal length. The nodes and elements are numbered in Figure (b). e1 n1 e2 n2 n3 Figure (b) In order to assemble the element matrices we rewrite the stiffness matrix for each element by imbedding it in the global matrix format. For element 1 ⎡ ⎤ ⎡ ⎤⎡ ⎤      p1,1 1 −1 0 u1 2EA 2EA 1 −1 u 1 p1,1 ⎣−1 1 0⎦ ⎣ u 2 ⎦ → ⎣ p1,2 ⎦ = = (a) p1,2 1 u2 L −1 L 0 u3 0 0 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- and imbed each element stiffness matrix in the global format. The element matrices can then be assembled (summed) to form the global matrix equations.   { pm } = {F} = (5.3.21) [km ] {r } = [K ] {r } P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 176 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods For element 2  p2,2 p2,3    −1 u 2 1 u3  2EA 1 = L −1 → ⎡ ⎤ ⎡ 0 0 2EA ⎣ p2,2 ⎦ = ⎣0 L p2,3 0 0 1 −1 ⎤⎡ ⎤ 0 u1 −1⎦ ⎣ u 2 ⎦ u3 1 (b) Now assemble these two. ⎤ ⎡ 1 p1,1 2EA ⎣−1 ⎣ p1,2 + p2,2 ⎦ = L p2,3 0 ⎡ ⎤⎡ ⎤ −1 0 u1 2 −1⎦ ⎣ u 2 ⎦ u3 −1 1 (c) Looking closely we note that the nodal displacements are known at the restraints and the rest are unknown. Likewise the restraint forces are unknown at the restraints and applied forces are known at all other nodes. The applied loads and restraints for this case are quite simple and have been inserted in Equation (d). Symbols for the unknown restraint forces have been added. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎤ ⎡ 1 R1 2EA ⎣−1 ⎣ F ⎦= L R3 0 ⎡ ⎤⎡ ⎤ −1 0 0 2 −1⎦ ⎣ u 2 ⎦ 0 −1 1 (d) As noted, the global stiffness matrix is a singular matrix, meaning its determinant is zero. Thus this set of equations does not have a unique solution. The next step is to partition the matrix equation into one set for finding the nodal displacements and another for finding the restraint forces. The dashed lines show the partitions. ⎡ R1 ⎤ ⎡ 1 ⎥ ⎢ ⎢ F ⎥ = 2EA ⎢ ⎣−1 ⎦ ⎣ L 0 R3 ⎤⎡ ⎤ 0 0 ⎥ ⎥⎢ ⎢ u −1⎦ ⎣ 2 ⎥ ⎦ 1 0 −1 2 −1 (e) The equation for the displacement and its solution is F= 2EA · 2u 2 L → u2 = FL 4EA (f) The equations for the restraint forces and their solution are  R1 R3  = 2EA L   −1 u −1 2 → R1 = − F 2EA FL · = − = R3 L 4EA 2 (g) The internal stresses from Equation 5.2.5 for element 1 are σx1 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS E1  −1 = l1 ⎡ ⎤   0   u1 2E  F −1 1 ⎣ FL ⎦ = 1 = u2 L 2A 4EA Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (h) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 177 A General Method for the Axially Loaded Slender Bar and for element 2 σx2 ⎡ ⎤ FL   ⎢  u2 2E  F ⎥ −1 1 ⎣ 4EA ⎦ = − 1 = u3 L 2A 0 E2  −1 = l2 (i) To find the distributed displacements we note that s1 = x and s2 = x − L/2. Then for each element from Equation 5.2.2 we have  2s u 1 (s1 ) = 1 − 1 L 2s1 L  u1 u2   2s = 1− 1 L 2s1 L  ⎤ 0 ⎣ FL ⎦ = F s1 2EA 4EA ⎡ F x 2EA ⎤ ⎡    FL 2s2 FL 2s2 ⎣ 4EA ⎦ = 1− 4EA L L 0 x F  1− → u 2 (x) = 2EA L → u 1 (x) =  2s u 2 (s2 ) = 1 − 2 L  2s2 L u2 u3   2s = 1− 2 L (j) (k) Plots of displacements and stresses are given in Figure (c). The displacements between nodes are always linear so we may simply draw straight lines between nodal displacements. The internal forces and stresses are always constant between nodes. FL/4EA Displacement, u 0 L F/2A Stress, σx 0 -F/2A Figure (c) Compare with Example 4.5.2. Example 5.3.2 Problem: Repeat Example 4.4.5 using the general method. The bar is shown in Figure (a). Find the displacements and stresses. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 178 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y L F 2F 3F x l l l Figure (a) Solution: Assemble the global stiffness matrix equations, add the forces and restraints, partition the equations, and solve. Divide the bar into three elements and number the nodes and elements as shown in Figure (b). e1 n1 e2 n2 e3 n3 n4 Figure (b) Putting all the elements together into a set of equations describing the entire bar requires the assembly of the element matrices. We note that E, A, and l are the same in all three elements where l = L/3. ⎡ ⎤ ⎡ ⎤⎡ ⎤ p1,1 1 −1 0 0 u1 ⎢ p1,2 ⎥ EA ⎢−1 ⎥ ⎢ u2 ⎥ 1 0 0 ⎥ ⎢ ⎥ ⎢ ⎥ = [k1 ] {r } { p1 } = ⎢ (a) ⎣ 0 ⎦= l ⎣ 0 0 0 0⎦ ⎣ u 3 ⎦ u4 0 0 0 0 0 This conveys exactly the same information as Equation 5.2.10 but is specialized to element 1. For element 2 ⎤ ⎡ ⎡ ⎤⎡ ⎤ 0 0 0 0 0 u1 ⎢ p2,2 ⎥ EA ⎢0 ⎥ ⎢ u2 ⎥ 1 −1 0 ⎥ ⎢ ⎥ ⎢ ⎥ = [k2 ] {r } { p2 } = ⎢ ⎣ p2,3 ⎦ = l ⎣0 −1 1 0⎦ ⎣ u 3 ⎦ u4 0 0 0 0 0 (b) For element 3 ⎡ ⎤ ⎡ 0 0 0 0 ⎢ 0 ⎥ EA ⎢0 0 0 ⎢ ⎥ ⎢ { p3 } = ⎣ = p3,3 ⎦ 1 l ⎣0 0 p3,4 0 0 −1 ⎤⎡ ⎤ 0 u1 ⎢ ⎥ 0⎥ ⎥ ⎢ u 2 ⎥ = [k3 ] {r } −1⎦ ⎣ u 3 ⎦ u4 1 (c) The assembled global matrix equations for the entire bar are found by adding Equations (a), (b), and (c). ⎤ ⎡ ⎡ ⎤ ⎡ ⎤⎡ ⎤ p1,1 1 −1 0 0 F1 u1 ⎢ F2 ⎥ ⎢ p1,2 + p2,2 ⎥ EA ⎢−1 ⎥ ⎢ u2 ⎥ 2 −1 0 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ = [K ] {r } {F} = ⎢ (d) ⎣ F3 ⎦ = ⎣ p2,3 + p3,3 ⎦ = l ⎣ 0 −1 2 −1⎦ ⎣ u 3 ⎦ u4 F4 p3,4 0 0 −1 1 where [K] is the global stiffness matrix for the entire structure. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 179 A General Method for the Axially Loaded Slender Bar Let us now introduce the applied loads, restraints, and restraint forces. The restraint is zero displacement at node 1 and applied loads are concentrated forces at nodes 2, 3, and 4. The global equations become ⎡ ⎤ ⎡ ⎤⎡ 0 ⎤ R1 1 −1 0 0 ⎢ ⎢ ⎥ ⎥ ⎥ ⎢ F ⎥ EA ⎢ 2 −1 0⎥ ⎢−1 ⎥⎢ (e) ⎢ ⎢ u2 ⎥ ⎥= ⎣ −2F ⎦ l ⎣ 0 −1 2 −1⎦ ⎣ u 3 ⎦ 0 0 −1 1 u4 3F Partitioning leaves us with three equations for finding the three unknown displacements. ⎡ ⎡ ⎤ ⎤⎡ ⎤ F 2 −1 0 u2 EA ⎣ −2F ⎦ = ⎣−1 2 −1⎦ ⎣ u 3 ⎦ l u4 3F 0 −1 1 (f) Once this set of equation is solved we can then find the restraint force from the other partitioned equation which can now be written as ⎡ ⎤  u2 EA  −1 0 0 ⎣ u 3 ⎦ (g) [R1 ] = l u4 Solving Equation (f) by the usual substitution methods of elementary algebra is not difficult but this might be a good time to get familiar with one or more of the numerical and symbolic equation solving software packages. As our matrices grow larger through the use of many more elements it will be necessary to turn to these tools. Instructions for using Mathematica are given in Appendix C. By factoring the F from the loading matrix we can rewrite Equation (f) as ⎡ ⎤ ⎡ ⎤⎡ ⎤ 1 2 −1 0 u2 Fl ⎣ ⎦ ⎣ −2 = −1 2 −1⎦ ⎣ u 3 ⎦ (h) EA u4 3 0 −1 1 Using Mathematica we obtain the solutions ⎡ ⎤ ⎡ ⎤ 2 u2 EA  Fl ⎣3⎦ ⎣ u3 ⎦ = −1 R1 = EA l u4 6 0 ⎡ ⎤  u2 0 ⎣ u 3 ⎦ = −2F u4 (i) As noted before we can simply draw straight lines between nodal displacements in a plot. This is shown in Figure (c). Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Notice that a free body diagram of the entire bar would have provided R1 = −2F. So this checks out. Given the nodal displacements the distributed displacements are linear functions found from   s1 s1 u1 u 1 (s1 ) = 1 − u2 l l   s2 s2 u2 u 2 (s2 ) = 1 − (j) u3 l l   s3 s3 u3 u 3 (s3 ) = 1 − u4 l l P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 180 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Given the nodal displacements the stresses can be found for each element from Equations 5.2.5. These have the same sign convention as the classical analysis. σx1 E −1 = l σx2 = σx3 = E −1 l E −1 l ⎡ ⎤   0  u1  E 2F 1 −1 1 ⎣ 2Fl ⎦ = = u2 l A EA ⎡ ⎤ 2Fl    u2  ⎢ EA ⎥ E F ⎥ 1 −1 1 ⎢ = ⎣ 3Fl ⎦ = A u3 l EA ⎡ 3Fl    u3  ⎢ EA E 1 −1 1 ⎢ = ⎣ 6Fl u4 l EA (k) ⎤ ⎥ 3F ⎥= ⎦ A And if you want the internal forces they are simply the stresses times the area. These are exactly the same values obtained in Example 4.4.5. The stresses are plotted in Figure (c). FL/EA 6 4 Displacement, u 2 0 0 l s1 2l s2 3l s3 3F/A Stress, σx 2F/A F/A Figure (c) ########### As we noted three equations in three unknowns can be solved analytically; however, as the number of concentrated loads and hence the number of elements increases an analytical solution becomes tedious, although there are symbolic solver programs available that will ease the burden. It is normal to convert to numerical values even before assembling the equations and to use numerical linear equation solving software to provide answers. In fact, numerical solutions are common even for small problems. The global stiffness matrix depends only on the geometry and material properties of the bar. It is independent of a particular loading or restraints. The next example defines different loads and restraints but uses the same global stiffness matrix [K] that was used in Example 5.3.2. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 181 A General Method for the Axially Loaded Slender Bar ########### Example 5.3.3 Problem: A uniform slender bar has loading and restraints as shown in Figure (a). Find the displacements and stresses. y L 4F F x l l l Solution: Modify the global set of matrix equations in Example 5.3.2 and solve. The assembled global stiffness matrix is the same as in Example 5.3.2. When you add the loads, restraints and restraint forces for this problem the global set of equations becomes ⎤ ⎡ ⎡ ⎤⎡ ⎤ 1 −1 0 0 R1 0 ⎢ 4F ⎥ EA ⎢−1 ⎢ ⎥ 2 −1 0⎥ ⎥= ⎢ ⎢ ⎥ ⎢ u2 ⎥ (a) ⎣ F ⎦ 2 −1⎦ ⎣ u 3 ⎦ l ⎣ 0 −1 0 R4 0 0 −1 1 The partitioned equations are    EA 4F 2 = F l −1   −1 u 2 2 u3  R1 R4  =   3 2  R1 R2  =F  EA −1 0 l 0 −1  u2 u3  (b) and the solutions are  u2 u3  = Fl EA  −3 −2  (c) The stresses are σx,1 σx,2 σx,3 ⎡ ⎤ 0  E 3F −1 1 ⎣ 3Fl ⎦ = = l A EA ⎡ ⎤ 3Fl    u2  ⎢ EA ⎥ −F E E ⎥ −1 1 −1 1 ⎢ = = ⎣ 2Fl ⎦ = A u3 l l EA ⎡ ⎤ 2Fl    u3 ⎢ E E ⎥ −2F −1 1 −1 1 ⎣ EA ⎦ = = = 0 l l A 0 E −1 = l  1  0 u2  (d) Given the nodal displacements the distributed displacements are linear functions between nodes. The stresses are constants between nodes. These results are plotted in Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (a) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 182 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Fl/EA 3 Displacement, u 2 1 0 0 l 2l s2 s1 3l s3 3F/A Stress, σx -F/A -2F/A Figure (b) ########### 5.4 A General Method—Distributed Applied Loads To allow the method just described to work with distributed applied loads we must replace them with a set of concentrated loads that are in some sense statically equivalent. Replacing distributed loads with equivalent concentrated loads has been used in the past with the analytical methods discussed in Chapter 4 but has been restricted to a small number of concentrated loads because of the number of linear algebraic equations generated in the process of matching boundary conditions. With the advent of the digital computer this restriction is lifted. If a sufficiently large number of equivalent concentrated loads are used the answer closely approximates the exact solution based on distributed loads. One method of creating equivalent concentrated loads can be explained by the process shown in the next three figures. Consider the axially loaded bar in Figure 5.4.1 with a plot of the distributed axial load shown in Figure 5.4.2. We divide the bar into elements and the distributed load into segments as shown in Figure 5.4.2. fx(x) Figure 5.4.1 The area of each rectangular block can be the actual area under the corresponding segment of the curve or can be approximated by taking the value of the distributed force at the mid point of the segment times the length of the segment. We take that area under each block, which is the total force for that segment, and put half of it at each side of the segment as shown in Figure 5.4.3. These form a set of concentrated loads that is closely statically equivalent to the distributed load. The double arrows indicate --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 183 A General Method for the Axially Loaded Slender Bar fx(x) x Figure 5.4.2 Fe Figure 5.4.3 that the particular concentrated load comes from half the block on either side. The blocks need not be of equal width. If we designate the equivalent concentrated forces for each element as {Fem } =  Fem,n Fem,n+1  (5.4.1) Then the total equivalent forces are {Fe } =  {Fem } (5.4.2) The element equivalent forces must be embedded in a global format for summing. In the early days the nodal loads were calculated by the method described in the above figures. Later a formal method of calculating them was used that is presented in Equation 5.4.3 for, say, element m. {Fem } =  lm [n]T f x (sm ) dsm (5.4.3) 0 In Equation 5.4.3 the matrix [n]T is the transpose of [n]. Since [n] is a row matrix, [n]T is a column matrix. The coordinate sm is a local coordinate with origin at node n. These are called kinematically equivalent nodal loads and are identified with the subscript e. We shall derive Equation 5.4.3 in Chapter 11 using the principle of virtual work. The kinematically equivalent nodal loads are based on the principal that there is work equivalence between real applied loads acting over the real displacements and the nodal equivalent loads acting over the nodal displacements. Virtual work is an alternative statement of equilibrium that is equivalent to Newton’s laws in a “weak” or distributed sense for continuous structures. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 184 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods For the axial case for element m {Fem } =  lm lm  [n]T f x (sm ) dsm = 0 0 ⎡ sm ⎤   lm ⎥ ⎥ f x (sm ) dsm = Fem,n Fem,n+1 sm ⎦ 1− ⎢ ⎢ ⎣ (5.4.4) lm Our gallery of quantities used in the analysis of slender bars now reads as shown in Equation 5.4.5. Everything from { pm } to {rm } is exactly the same as in Equation 5.2.8. The distributed load is converted into equivalent nodal loads which are added to any concentrated loads to obtain the total loading. The restraints ρ are also added. In summary f x (s) → {Fc } {F} {Fem } =  lm { pm } [n]T f x (sm ) dsm σxm {Fe } = 0  {rm } ← {ρ} u (sm ) = [n] {rm } {Fem } {F} = {Fc } + {Fe } d du m = [n] {rm } dsm dsm d = E m εxm = E m [n] {rm } dsm εxm = σxm pm,n = −Am σxm u (sm ) εxm d = −E m Am [n] {rm } dsm { pm } = → d [n] {rm } dsm   { pm } = {F} = [km ] {r } = [K ] {r }  pm,n pm,n+1  (5.4.5) = [km ] {rm } pm,n+1 = Am σxm = E m Am We shall illustrate the two methods for obtaining equivalent nodal loads in Example 5.4.1. ########### Example 5.4.1 Problem: An axially loaded bar has a uniform distributed applied load f 0 with units of force per unit length (N/mm). The bar is divided into six equal elements as shown in Figure (a). Find the equivalent nodal loads. n1 e2 n2 e3 n3 e4 n4 e5 n5 e6 n6 n7 l L Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- e1 P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 185 A General Method for the Axially Loaded Slender Bar Solution: Find the equivalent nodal loads by both methods. Using the first method the load on each element is f 0 lm = f0 L 6 (a) and so the equivalent concentrated load on each element is found by putting half of this load on each node. ⎡ f L⎤ 0 ⎢ 12 ⎥ ⎥ {Fem } = ⎢ ⎣ ⎦ f0 L 12 The total equivalent load in this case is ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ {Fe } = ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ f0 L 12 f0 L f0 L + 12 12 f0 L f0 L + 12 12 f0 L f0 L + 12 12 f0 L f0 L + 12 12 f0 L f0 L + 12 12 f0 L 12 (b) ⎤ f0 L ⎥ ⎢ 12 ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ f0 L ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ 6 ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ f0 L ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ 6 ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ f0 L ⎥ ⎥=⎢ ⎥ ⎥ ⎢ 6 ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ f0 L ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ 6 ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ f L⎥ ⎥ ⎢ 0 ⎥ ⎥ ⎢ ⎥ ⎥ ⎢ 6 ⎥ ⎥ ⎢ ⎥ ⎦ ⎣ f L⎦ 0 12 ⎤ ⎡ (c) Using the second method for each element {Fem } =  lm 0  ⎤ ⎡  sm ⎡ lm sm ⎤ ds 1 − m 1− ⎥ ⎢ 0 l m ⎥ ⎢ ⎢ ⎢ lm ⎥ ⎥=⎢ ⎢ ⎥ ⎢   ⎥ ⎣ ⎣ sm ⎦ f 0 dsm = f 0 ⎢  lm sm ⎦ ⎣ ds m 0 lm lm ⎡ f 0 lm ⎤ ⎡ ⎢ 2 ⎥ ⎥=⎢ ⎦ ⎣ f 0 lm 2 f0 L ⎤ 12 ⎥ ⎥ ⎦ f0 L 12 (d) and so the total is the same as Equation (c). Note that this set of equivalent nodal loads applies to all boundary conditions. For more complex distributions the two methods may differ a bit with the second one the more accurate. As the elements increase in number and are, therefore, shorter in length the accuracy of the displacements and stresses improves. We shall demonstrate just how accurate in examples coming up. ########### Now let us test the accuracy of using equivalent nodal loads with a numerical example that we can compare to an exact analytical solution using a distributed load. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 186 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 5.4.2 Problem: Consider the bar in Example 4.4.1 as shown in Figure (a). Find the displacement and stress. Compare with the traditional analytical method. f0 Figure (a) Solution: Find the equivalent nodal loads; assemble the global matrix equations; introduce the loads, restraints, and restraint forces and solve. Compare with the solution in Example 4.4.1. Suppose we start with a very crude approximation of only two elements as in Figure (b). F1 e1 F2 e2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- l1 F3 l2 n1 n2 n3 Figure (b) Let us assume the l1 = l2 = L/2 and the E and A are the same for both elements. The element stiffness matrix for element 1 is  p1,1 p1,2  =  2EA 1 L −1   −1 u 1 = [k1 ] {r1 } 1 u2 (a)  p2,2 p2,3   2EA 1 = L −1   −1 u 2 = [k2 ] {r2 } 1 u3 (b) and for element 2 When these are imbedded in the global format the assembled global matrix equations are ⎡ 1 2EA ⎣ {F} = −1 L 0 ⎤⎡ ⎤ −1 0 u1 2 −1⎦ ⎣ u 2 ⎦ = [K ] {r } u3 −1 1 (c) Let us assign values for the equivalent nodal loads. By either method the total load on each element is f 0 L/2. Place half of this at each end, that is, f 0 L/4. At node 1 there is also a reaction force R1 that we Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 187 A General Method for the Axially Loaded Slender Bar must account for. Also at node 1 the displacement is zero. Our matrix equations become ⎫ ⎧ f0 L ⎪ ⎪ ⎪ ⎪ R + ⎪ ⎪ 1 ⎪ 4 ⎪ ⎪ ⎪ ⎡ ⎤⎡ ⎤ ⎪ ⎪ ⎪ ⎪ 1 −1 0 0 ⎬ 2EA ⎨ f0 L f0 L ⎣−1 2 −1⎦ ⎣ u 2 ⎦ = [K ] {r } = + ⎪ L 4 ⎪ ⎪ ⎪ 4 u3 0 −1 1 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ f0 L ⎪ ⎪ ⎩ ⎭ 4 Note that at node 1 there is both an applied load and a restraint force. We partition the equations as shown in Equation (e). ⎧ ⎫ f0 L ⎪ ⎪ ⎪ ⎪ R + 1 ⎪ ⎪ ⎪ ⎡ ⎤⎡ ⎤ 4 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ −1 0 1 0 ⎨ ⎬ 2EA f0 L f0 L ⎢ ⎥⎢ ⎥ = ⎣−1 + 2 −1⎦ ⎣ u 2 ⎦ = [K ] {r } ⎪ 4 ⎪ L 4 ⎪ ⎪ ⎪ ⎪ u3 0 −1 1 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ f L 0 ⎪ ⎪ ⎭ ⎩ 4 We obtain two matrix equations ⎡ ⎤ f0 L  ⎢ 2 ⎥ 2EA 2 ⎥= ⎢ ⎣ f L⎦ −1 L 0 4 and R1 +   −1 u 2 1 u3 (d) (e) (f)    u2 2EA  2EA f0 L −1 0 = u2 =− u3 4 L L (g) The solutions are --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`---  u2 u3  ⎤ 3 f0 L 2 ⎢ 8EA ⎥ ⎥ =⎢ ⎣ ⎦ f0 L 2 2EA ⎡ R1 = − f 0 L (h) How well does this compare with the exact answer? From Example 4.4.1, Equation (f) we get     f0 L 2 x2 3 f0 L 2 f0 L u(L) = Lx − → u = u(x) = EA 2 2 8EA 2EA (i) At 0, L/2, and L the answers are the same. From a plot of the displacements in Figure (c) we see the difference. f 0 L2 2 EA equivalent nodal loads Displacement, u(x) distributed load 0 L Figure (c) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 188 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The displacement between points of concentrated load application is always a linear function. It is a quadratic function in the exact case; however, the approximate curve is not far off. For the stresses from Example 4.4.1, Equation (b) σx (x) = f0 (L − x) A σx (0) = → f0 L A σx   f0 L L = 2 2A σx (L) = 0 (j) The stresses from equivalent nodal loads are σx1 E −1 = l σx2 E −1 = l ⎡ 0  2E  −1 1 ⎣ 3 f 0 L 2 = L 8EA ⎡ 3 f0 L 2    u2  ⎢ 8EA 2E  1 −1 1 ⎢ = ⎣ f L2 u3 L 0 2EA  1  0 u2  ⎤ ⎦ = 3 f0 L 4A ⎤ (k) ⎥ ⎥ = f0 L ⎦ 4A When using equivalent nodal loads the stress is constant in each element as represented by the dotted lines in Figure (d) while the exact answer for a distributed load shows a linearly varying stress represented by the solid line. f0 L A equivalent nodal loads distributed load Stress 0 L The first reaction is that this is not a very good way to solve for the stresses. But note that the average stress at node 2 using equivalent nodal loads is surprisingly accurate. Compare that average to the exact value       1 f0 L f0 L f0 L f0 L 1 3 f0 L L + σx (l) = = L− = σave,2 = 2 4 A 4 A 2A 2 A 2 2A and they agree exactly. Also let us use the stress at the support based on the support reaction (see Equation (h)) to get P(0) = −R1 = f 0 L → σx (0) = f0 L P (0) = A A (m) Likewise use the known value at the right end where there is no load to obtain P(L) = 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS → σx (L) = 0 P(L) = =0 A A Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (n) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (d) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 189 A General Method for the Axially Loaded Slender Bar and place these values on the graph repesented by small circles. A curve fitted through these values at the ends and the nodal average value at node 2 is the same as the exact answer. Clearly the stress between nodes is not accurate. But the stress is accurate at the nodes if you use the loads at the boundary to find the stress there and average the stress at interior nodes. A curve fitted to these values coincides with the analytical answer in this case and is very close in more complicated cases of distributed applied loads. This is common practice. Now let us double the number of elements to four and compare the answers. The process of assembling the matrix equations is just the same. You get ⎡ f0 L 8 f0 L 4 f0 L 4 f0 L 4 f0 L 8 R1 + ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ {F} = ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ ⎤ ⎥ ⎥ ⎥ ⎡ ⎥ 1 ⎥ ⎥ −1 ⎥ 4EA ⎢ ⎢ ⎥ ⎢ 0 ⎥= ⎥ L ⎢ ⎣ 0 ⎥ ⎥ 0 ⎥ ⎥ ⎥ ⎦ −1 0 2 −1 −1 2 0 −1 0 0 ⎤⎡ ⎤ 0 0 u1 ⎢ u2 ⎥ 0 0⎥ ⎥⎢ ⎥ ⎢ ⎥ −1 0⎥ ⎥ ⎢ u3 ⎥ ⎦ 2 −1 ⎣ u 4 ⎦ u5 −1 1 (o) Let us assign some numerical values. Let the material be aluminum for which E = 68950 N/mm2 . The length is L = 200 and the cross section area is 4 mm2 . The load is f 0 = 2.5 N/mm. Numerical values are inserted and the partitions are selected. ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ R1 + 62.5 125 125 125 62.5 ⎤ 1 −1 ⎥ ⎢−1 ⎥ ⎢ ⎥ ⎥ = 5516 ⎢ ⎢ 0 ⎥ ⎣ 0 ⎦ 0 2 −1 0 0 ⎡ 0 0 −1 0 2 −1 −1 2 0 −1 ⎤⎡ 0 ⎤ 0 ⎢ ⎥ ⎥ 0⎥ ⎥⎢ ⎢ u2 ⎥ ⎥ ⎥ 0⎥ ⎢ u ⎢ 3⎥ −1⎦ ⎣ u 4 ⎦ 1 u5 (p) The partitoned matrix equations are ⎡ ⎤ ⎡ ⎤⎡ ⎤ 125 2 −1 0 0 u2 ⎢ 125 ⎥ ⎢ ⎢ ⎥ 2 −1 0⎥ ⎢ ⎥ = 5516 ⎢−1 ⎥ ⎢ u3 ⎥ ⎣ 125 ⎦ ⎣ 0 −1 2 −1⎦ ⎣ u 4 ⎦ u5 62.5 0 0 −1 1 R1 + 62.5 = −5516u 2 (q) This is easily solved for the unknown nodal displacements and the restraint force using one of the software solvers or with your calculator. ⎤ ⎡ ⎤ 0.0793 u2 ⎢ u 3 ⎥ ⎢ 0.1360 ⎥ ⎥ ⎢ ⎥ {r } = ⎢ ⎣ u 4 ⎦ = ⎣ 0.1700 ⎦ u5 0.1813 ⎡ R1 = −499.92 This compares to the actual restraint force of −500. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (r) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 190 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Once again use the restraint force at node 1 and the actual applied load at node 5 to find the stresses there. 500 = 125 N/mm 2 4 σx1 = σx5 = 0 (s) The stresses at node 2 are 4•68950  −1 200 4•68950  −1 = 200 (σx )1,2 = (σx )2,3  0.0 = 109.35 N/mm 2 0.0793    0.0793 1 = 78.19 N/mm 2 0.1360  1  (t) The average stress at node 2 is σave,2 = 109.35 + 78.19 = 93.77 N/mm 2 2 (u) The analytical value is 93.75. The displacement and stress values at all the nodes are given in Table 5.4.1. Just for comparison we added the results from an 8 element equivalent nodal load solution. The displacement values are given in Table 5.4.1 for 2, 4, and 8 elements and for the exact analytical solution at the node locations. Table 5.4.1 Value of x Displ. 2 elem. Displ. 4 elem. Displ. 8 elem. Displ. analytic 0 0.0000 0.0000 0.0000 0.0000 Stress 2 elem. Stress 4 elem. Stress 8 elem. Stress analytic 125.022 124.980 125.000 125.000 25 50 75 0.0425 0.0425 0.0793 0.0793 0.0793 0.1105 0.1105 100 0.1360 0.1360 0.1360 0.1360 109.355 109.375 93.770 93.772 93.750 78.189 78.125 62.503 62.538 62.469 62.500 125 150 175 0.1558 0.1558 0.1700 0.1700 0.1700 0.1785 0.1785 200 0.1813 0.1813 0.1813 0.1813 46.886 46.875 31.234 31.303 31.250 15.583 15.625 0.000 0.000 0.000 0.000 It is customary to draw a smooth curve through the nodal values for displacement and for average stress in presenting this data. Within plotting accuracy the curves fall right on top of each other when several elements are used. Problem: Repeat Example 4.5.3 by the general method. u(L) = −δ Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Example 5.4.3 P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 191 A General Method for the Axially Loaded Slender Bar f0 δ Figure (a) Solution: Borrow the equivalent nodal loads and global stiffness matrix from Example 5.4.2. Insert the new displacement constraints and restraint forces, partition, and solve. Using just two elements the partitioned global equations are ⎧ f0 L ⎪ R1 + ⎪ ⎪ ⎪ 4 ⎪ ⎪ ⎪ ⎨ f0 L f0 L + ⎪ 4 4 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ R + f0 L 3 4 ⎫ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎬ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎭ = ⎡ ⎤⎡ ⎤ 0 0 ⎥⎢ 2 −1⎥ ⎣ u 2 ⎥ ⎦ ⎦ −δ −1 1 1 −1 2EA ⎢ ⎢−1 L ⎣ 0 (a) The partitioned equation for displacement and the solution is given in Equation (b). 2EA f0 L (2u 2 + δ) = 2 L → u2 = δ f0 L 2 − 8EA 2 (b) This agrees with the analytical solution in Example 4.5.3, Equation (e) at x = L/2. The partitioned equations for the restraints are ⎤ f0 L  + R ⎢ 1 4 ⎥ ⎥ = 2EA −1 ⎢ ⎣ L −1 f0 L ⎦ R3 + 4 ⎡   u2 0 1 −δ (c) and the values of the restraint forces are R1 = − δ f0 L + EA 2 L R3 = − δ f0 L − EA 2 L --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- These also agree with the values found in Example 4.5.3. The stresses are ⎤ ⎡   0  0  δ 2E  2E  f0 L −1 1 −1 1 ⎣ f 0 L 2 σx1 = −E = δ⎦= u2 L L 4A L − 8EA 2 ⎤ ⎡ 2   δ f0 L  u2  2E  δ 2E  f0 L − ⎦ −1 1 −1 1 ⎣ 8EA = −E σx2 = 2 =− −δ L L 4A L −δ Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) (e) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 192 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In the following plots the answers from Example 4.5.3 are shown as solid lines. The values found here are dotted lines. Circles are shown for the stresses based on restraint forces and the average value at node 2. Once again the curve fitted to the nodal values for stresses agrees closely with the analytical solution. We should use more elements to get a better approximation for displacements. f 0 L2 δ − 8EA 2 Displacement δ f 0 L Eδ − 2A L L Stress − f 0 L Eδ − L 2A ########### Example 5.4.4 Problem: Find the displacements, internal axial forces, and stresses by the general method using equivalent nodal loads for the bar shown in Figure (a). Use two elements. Compare with the exact analytical solution. y fx(x) = f0x/L x L Figure (a) Solution: The bar is divided into two elements and the nodes and elements are numbered exactly as in the previous three examples. The global stiffness matrix will be the same. The equivalent nodal loads Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (b) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 193 A General Method for the Axially Loaded Slender Bar and restraints will differ. Find the equivalent nodal loads, complete the global equations, partition, and solve. First find the equivalent nodal loads for element 1. ⎡ ⎤ 2s1   l   L 1− 2 ⎢ L ⎥ Fe1,1 ⎢ ⎥ f 0 s1 ds1 = [n]T f x (s1 ) ds1 = ⎣ Fe1,2 2s1 ⎦ L 0 0 ⎡ ⎢ ⎢ = f0 ⎢ ⎢ ⎣ L ⎤  2s12 s1 − 2 ds1 ⎥ L L ⎥ 0 ⎥ = f0 ⎥  L  2 ⎦ 2 2s1 ds1 2 L 0 L 2  2s 3 s12 − 12 ⎢ 2L 3L ⎢ ⎣ 3 2s1 3L 2 ⎡  ⎤ L2 ⎡ f L  0  ⎢ 24 ⎥ ⎥ = ⎢ ⎣ ⎦ f0 L   12 0 ⎤ ⎥ ⎥ ⎦ (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- For element 2 you must make a coordinate transformation. If you use the new coordinate s2 you must note that x = s2 + L/2 and replace x in the loading as shown in Equation (b). ⎤ ⎡ 2s2     l   L 1− 2 ⎢ s2 + L2 L ⎥ Fe2,2 T ⎢ ⎥ ds2 = [n] f x (s2 ) ds2 = ⎦ f0 ⎣ 2s Fe2,3 L 0 0 2 L ⎡ ⎡ ⎤ ⎤ 2s2 2s2  L 1−  L 1− 2 ⎢ 2 ⎢ L ⎥ L ⎥ ⎢ ⎢ ⎥ f 0 s2 ds2 + ⎥ f 0 ds2 = (b) ⎣ ⎣ ⎦ L 2s2 2s2 ⎦ 2 0 0 L L Note that the first of these integrals is the same as the one in Equation (a). The second is one half that found in Example 5.4.2 since the value of the load is f0 /2. ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ f0 L f0 L f0 L   ⎢ 24 ⎥ ⎢ 8 ⎥ ⎢ 6 ⎥ Fe2,2 ⎥ ⎢ ⎥ ⎢ ⎥ (c) =⎢ ⎣ f L ⎦+⎣ f L ⎦=⎣ 5f L ⎦ Fe2,3 0 0 0 8 12 24 You could have replaced s2 with s2 = x −L/2 in the integral in Equation (b) and integrated from L/2 to L. That would be a bit more difficult. Or you can arrive at the values for element 2 another way. The loading is triangular as plotted here. f0 f0/2 Figure (b) We have just found the equivalent nodal loads for the triangular load in element 1. Element 2 has an equal triangular load and a uniform load as shown by the darker shaded area. The uniform portion of the Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 194 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods load has two equivalent nodal loads each equal to the one half the total load represented by that area. The triangular load contribution to element 2 is the same as for element 1. The total load from the uniform part is F= f0 L · 2 2 (d) So the equivalent nodal load contribution of the uniform part is Fe2,2 Fe2,3  ⎤ f0 L ⎢ 8 ⎥ ⎥ =⎢ ⎣ f L⎦ 0 8 ⎡ (e) Add to it the triangular contribution.  Fe2,2 Fe2,3  ⎤ ⎡ f0 L f0 L ⎢ 8 ⎥ ⎢ 24 ⎥ ⎢ =⎢ ⎣ f L ⎦+⎣ f L 0 0 8 12 ⎡ ⎤ f0 L ⎥ ⎢ 6 ⎥ ⎥=⎢ ⎥ ⎦ ⎣ 5f L ⎦ 0 24 ⎤ ⎡ (f) The combination of all the equivalent nodal loads is in the assembled equations, Equations (g). The assembled equations are ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ f0 L 24 ⎫ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎬ ⎡ 1 −1 2EA ⎣ f0 L f0 L −1 2 = + ⎪ 12 L 6 ⎪ ⎪ ⎪ 0 −1 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 5f L ⎪ ⎪ ⎭ ⎩ R3 + 0 ⎪ 24 ⎤⎡ ⎤ 0 u1 −1⎦ ⎣ u 2 ⎦ 0 1 (g) The partitioned equations and solutions are ⎤ f0 L  ⎢ 24 ⎥ 2EA 1 ⎥= ⎢ ⎣ f L⎦ L −1 0 4 ⎡   −1 u 1 2 u2 →  u1 u2  ⎤ f0 L 2 ⎢ 6EA ⎥ ⎥ =⎢ ⎣ ⎦ 7 f0 L 2 48EA ⎡ ⎤ f0 L 2  u1  ⎢ 6EA ⎥ 2EA  2EA  5 f0 L ⎥ = − 7 f0 L 0 −1 0 −1 ⎢ R3 + = = ⎣ 2 ⎦ u2 24 L L 24 7 f0 L 48EA 5 f0 L 12 f 0 L f0 L 7 f0 L − =− =− → R3 = − 24 24 24 2  Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  (h) ⎡ Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (i) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`---  P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 195 A General Method for the Axially Loaded Slender Bar The internal forces and stresses are p1,2 p2,3 ⎤ f0 L 2  u1  ⎢ 6EA ⎥ 2EA  2EA  ⎥ = − f 0 L → σx1 = − f 0 L −1 1 −1 1 ⎢ = = ⎣ ⎦ 2 u L L 24 24A 2 7 f0 L 48EA ⎤ ⎡   7 f0 L 2  u2  2EA  2EA  7 f0 L 7 f0 L −1 1 −1 1 ⎣ 48EA ⎦ = − = → σx2 = − = 0 L L 24 24A 0  ⎡  (j) These displacements and internal forces are plotted in Figure (c) as dotted lines. The average internal force at node 2 is P2ave =   f0 L 7 f0 L f0 L 1 − − =− 2 24 24 6 (k) and the forces at each end are P (0) = 0 P(L) = R3 = − f0 L 2 (l) These values of internal forces at the nodes are plotted in Figure (c) as circles and a dashed curve is fitted through them. The analytical expressions for the internal force and the stress can be found from static equilibrium. x dP = − f x (x) = − f 0 dx L P (x) = − → f0 x 2 2L f0 x 2 P (x) =− A 2L A σx (x) = → (m) Particular values of the analytical solution for the internal forces at the nodal points are P (0) = 0 P   f0 L L =− 2 8 P(L) = − f0 L 2 (n) The displacements are found using the first order differential equation. EA du f0 x 2 = P(x) = − dx 2L --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- → u(L) = − → u (x) = → u (x) = − f0 L 2 +c =0 6EA f0 x 3 +c 6E AL → c= f0 L 2 6EA (o)  f0  3 L − x3 6E AL Particular values at the nodes are Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS u (0) = f0 L 2 6EA u   7 f0 L 2 L = 2 48 u(L) = 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (p) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas 196 July 11, 2011 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The analytical solutions are plotted in Figure (c) as solid lines. L 0 Axial force, P(x) f0L2/6EA Displacement, u(x) 0 Figure (c) Using a few more elements would greatly improve the solution. ########### We have called this a general method because it works for all loadings and all restraints. There need be no distinction between determinate and indeterminate bars. Now we must admit that it has another much better known name. It is called the finite element method (FEM) and the process is finite element analysis (FEA). ########### 5.5 Variable Cross Sections To use the elements stiffness matrices that we have developped we must have a uniform EA in each element. We shall approximate each element with a variable cross section with one that has a constant average value. An example will demonstrate the validity of this approach. ########### Example 5.5.1 Problem: Do both Examples 4.6.1 and 4.6.2 using the general method (FEM). The figures for these two problems are repeated here as Figure (a) and Figure (b). f0 L Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- -f0L/2 P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 197 A General Method for the Axially Loaded Slender Bar L Figure (b) Solution: Divide the bar into segments of constant cross section. Assemble the global stiffness matrix, find the equivalent nodal loads, assign boundary restraints, partition the equations, and solve. First we shall show the bar divided into only two elements, as in Figure (c), to show the process. In reality more elements should be used to more closely approximate the exact answer; however, even this crude approximation does quite well for most purposes. 3L/4 L/4 f0 e1 n1 e2 n2 n3 Figure (c) The support at the right end is free in Example 4.6.1 and fixed in 4.6.2, hence the ghostly image at the right end. Remember that we use the same global stiffness matrix in each case. The cross sectional area of the tapered beam is  x  (a) A (x) = A0 1 − 2L Each element has a constant cross section taking on the value of A at the midpoint of the variable cross section. ⎛ ⎛ 3L ⎞ L ⎞ ⎜ ⎜ ⎟ 7 ⎟ 5 A2 = A0 ⎝1 − 4 ⎠ = A0 (b) A1 = A0 ⎝1 − 4 ⎠ = A0 2L 8 2L 8 The loading is divided into concentrated loads for each element at each node. The total load at each node is f0 L f0 L f0 L f0 L Node 2: + Node 3: (c) Node 1: 4 4 4 4 We must pause here and point out that two approximations have been introduced: (1) the distributed load is replaced with equivalent nodal loads, and (2) the tapered cross section is replaced with uniform cross sections in each element. The element stiffness matrix in general is   E n An 1 −1 (d) [kn ] = −1 1 ln Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- f0 P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 198 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Let us imbed this in the global stiffness matrix for element 1 recognizing that A1 = 87 A0 and l1 = L2 and also for element 2 recognizing that A2 = 58 A0 and l2 = L2 . ⎤ ⎡ ⎤ ⎡ 7EA 0 0 0 7EA0 0 − 0 ⎥ ⎢ ⎢ 4L ⎥ 4L ⎥ ⎢ ⎢ ⎥ 5EA0 ⎥ 5EA0 ⎢ ⎢ 7EA ⎥ ⎥ ⎢ 7EA − 0 0 0 (e) [k2 ] = ⎢ [k1 ] = ⎢ 0⎥ 4L 4L ⎥ ⎢− ⎥ ⎥ ⎢ 4L 4L ⎣ ⎦ ⎦ ⎣ 5EA0 5EA0 0 − 0 0 0 4L 4L --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The complete assembled global stiffness matrix is ⎡ 7EA0 7EA0 − ⎢ 4L ⎡ ⎤ ⎢ 4L F1 ⎢ 5EA0 7EA 7EA ⎢ 0 0 {F} = ⎣ F2 ⎦ = [K ] {r } = ⎢ − + ⎢ 4L 4L 4L F3 ⎢ ⎣ 5EA0 0 − 4L 0 ⎤ ⎥⎡ ⎤ ⎥ ⎥ u 5EA0 ⎥ ⎣ 1 ⎦ ⎥ u2 − 4L ⎥ ⎥ u3 5EA0 ⎦ 4L (f) First let us apply the boundary conditions for Figure (b), where both ends are fixed. ⎡ ⎤ ⎤ ⎡ 7EA0 7EA0 f0 L − 0 ⎢ 4L ⎥ 4L ⎢ R1 + 4 ⎥ ⎢ ⎥⎡ ⎤ ⎥ ⎢ ⎢ ⎥ 0 ⎥ ⎢ ⎢ ⎥ ⎥ ⎢ f0 L ⎢ 7EA0 ⎥⎢ ⎥ 5EA 5EA 7EA ⎥ ⎢ 0 0 0 ⎢− ⎥ ⎢ u2 ⎥ {r } {F} = ⎢ = = [k] ⎥ + − ⎢ ⎣ ⎦ 2 ⎥ ⎢ 4L 4L 4L 4L ⎥ ⎢ ⎥ ⎥ ⎢ ⎢ ⎥ 0 ⎥ ⎢ ⎢ ⎥ ⎣ R3 + f 0 L ⎦ ⎣ 5EA0 5EA0 ⎦ 4 0 − 4L 4L The partitioned equation to obtain the displacement is   5EA0 7EA0 f0 L + u2 = → 4L 4L 2 u2 = 1 f0 L 2 6 EA0 (g) (h) and the equations for the restraint forces are 7EA0 f0 L =− u2 4 4L 5EA0 f0 L =− u2 R3 + 4 4L R1 + → R1 = −0.5417 f 0 L → R3 = −0.4583 f 0 L We can compare these to the exact answers found in Example 4.6.2. The deflection at x = (i) L is 2   f0 L 2 L = 0.1698 u 2 EA0 (j) The approximate value from the FE method is u 2 = 0.1667 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS f0 L 2 EA0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (k) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 199 A General Method for the Axially Loaded Slender Bar The support reactions are Exact: P(0) = 0.5572 f 0 L FEM: R1 = −0.5417 f 0 L P(L) = −0.4428 f 0 L R3 = −0.4583 f 0 L (l) The differences in sign result from the different definition of positive forces between internal forces and support restraints. The high accuracy for the displacement at the midpoint and the forces at the support when using such a crude model is amazing. The axial forces in the two elements are ⎤ ⎡          0 7A0 2E 7A0 2E 7 0 1 −1 1 −1 ⎣ −1 p1,1 2 ⎦ 1 f0 L f0 L = = = p1,2 u2 1 1 1 8 L −1 8 L −1 24 6 EA0 (m) ⎤ ⎡    1 f0 L 2       5A0 2E 5A0 2E 5 u2 1 −1 1 −1 ⎣ 1 p2,2 = f0 L = 6 EA0 ⎦ = p2,3 0 1 1 −1 8 L −1 8 L −1 24 0 The average value of P at node 2 is    17−5 1 1 L p1,2 + p2,3 = f0 L = f 0 L = 0.0417 f 0 L = P 2 2 2 24 24 (n) The stresses in each element are based on the area of the element and are uniform throughout the element. Of course the stresses will not be so accurate; however, with a little creative interpretation they are not so bad. If you use the actual area at the supports then the stresses there are reasonably accurate. f0 L A0 f0 L FEM: σx (0) = 0.5417 A0 Exact: σx (0) = 0.5572 f0 L A0 f0 L σx (L) = −0.9166 A0 σx (L) = −0.8856 (o) By the FEM at x = L2 we have two different stress components, one for element one and one for element two. These values are σx1 = f0 L f0 L 8 0.2917 = 0.3333 7 A0 A0 f0 L 8 f0 L σx2 = − 0.2083 = −0.3333 5 A0 A0 (p) This is terrible; we can’t have both a positive and negative stress at the same point with a distributed load. Of course, it is the result of concentrating forces at that point. The exact stress is   f0 L L = 0.0572 (q) σx 2 A0 Once again we find the average between the stresses in elements one and two. σ12ave = 0.0 (r) That is not so bad. Since this is a region of low stress the inaccuracy is not much of a problem. Remember that this is a very crude model. Using a few more elements improves accuracy tremendously. Plots are presented in Figure (d) for comparison. The solid lines are from the exact solutions obtained in --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 200 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Section 4.6, Example 4.6.2. The dotted lines are from the values obtained here. The dashed curves are fitted through the nodal values for axial force and stress. 0.2f0L2/EA0 Displacement, u 0 L L/2 0.5f0L Axial force, P 0 -0.5f0L --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 0.5f0L/A0 Stress, σx -0.5f0L/A0 -1.0f0L/A0 Figure (d) Because we use equivalent nodal loads the displacements are linear functions for each element. The internal forces and stresses are constant in each element. As we learned in previous examples we can improve the approximations by using the restraint forces to find the internal forces and stresses at the two ends and average the values at interior nodes. Both the constant values and the curves connecting the interpretive values are shown as dotted lines in the plots. Now let us also redo Example 4.6.1. Our matrix equations are found quickly from Equations (f). ⎡ ⎢ ⎢ ⎢ ⎢ {F} = ⎢ ⎢ ⎢ ⎢ ⎣ ⎡ 7EA ⎤ f0 L 0 ⎢ 4L 4 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ f0 L ⎥ = [k] {r } = ⎢ 7EA0 ⎢− ⎥ 2 ⎢ ⎥ 4L ⎢ ⎥ f0 L ⎣ ⎦ 4 0 − R1 + 7EA0 4L 5EA0 7EA0 + 4L 4L 5EA0 − 4L 0 ⎤ ⎥⎡ ⎤ ⎥ 0 ⎥ ⎥ ⎥ 5EA0 ⎥ ⎢ ⎥ ⎣ u2 ⎦ − ⎥ 4L ⎥ u 3 ⎦ 5EA0 4L (s) The partitioned equations are ⎤ ⎡ 7EA0 5EA0 f0 L ⎢ 2 ⎥ ⎢ 4L + 4L ⎥ ⎢ ⎢ ⎣ f L ⎦=⎣ 5EA0 0 − 4 4L ⎡ Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ⎤ 5EA0   4L ⎥ ⎥ u2 5EA0 ⎦ u 3 4L − Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (t) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 201 A General Method for the Axially Loaded Slender Bar and R1 = − f0 L 7EA0 u2 − 4L 4 (u) Solving for the displacements we get u2 = 3 f0 L 2 f0 L 2 = 0.4286 7 EA0 EA0 u3 = 22 f 0 L 2 f0 L 2 = 0.6286 35 EA0 EA0 (v) The exact values are u   f0 L 2 L = 0.4246 2 EA0 u(L) = 0.6137 f0 L 2 EA0 (w) The approximate restraint force is R1 = − 7EA0 3 f 0 L 2 7EA0 f0 L f0 L u2 − =− = − f0 L − 4L 4 4L 7 EA0 4 (x) which, of course, is the exact answer. The internal forces are   p1,1 p1,2  p2,2 p2,3  7A0 2E = 8 L  5A0 2E 8 L  = 1 −1 1 −1 −1 1  −1 1  0 u2  u2 u3  7A0 2E = 8 L  5A0 2E 8 L  = 1 −1 1 −1 ⎤ ⎡    0 3 −1 ⎣ −1 2 ⎦ 3 f0 L = f0 L 1 1 4 7 EA0 ⎡ ⎤ 3 f0 L 2 ⎢   ⎥ −1 ⎢ 7 EA0 ⎥ 1 −1 = f0 L ⎢ ⎥ 1 ⎣ 22 f 0 L 2 ⎦ 4 1 (y) 35 EA0 The internal forces and stresses must be interpreted once again. Using the correct area and the restraint force you get the correct stress at x = 0. The stress at x = L is known to be zero. Using the constant values of stress in each element we get         f0 L f0 L 3 7 5 1 f0 L ÷ A0 = 0.8571 f0 L ÷ A0 = 0.4 σ2 = (z) σ1 = 4 8 A0 4 8 A0 The average value at node 2 is σ12ave = 0.6286 f0 L A0 (α) The exact value is   f0 L L = 0.6667 2 A0 (β) The approximate values are shown in Figure (e) as dotted lines. The solid lines are the exact values from Section 4.6, Example 4.6.1. The moral of our story is that the values of displacement and stress at the nodes are quite accurate even for crude models such as just two elements. Let us assure you that with a few more nodes and elements Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- σ P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 202 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods the accuracy is excellent. By fitting a curve through the nodal values you can get a sufficiently accurate value between nodes. 0.6137f0L2/EA0 0.5f0L2/EA0 Displacement, u 0 L f0L 0.5f0L Axial force, P 0 f0L/A0 Stress, σx 0 Figure (e) ########### As you might have expected by now the general method is quite ripe for being programmed for computer analysis. In such cases the number of elements can be increased with little additional effort and answers can more closely approximate the exact answers. 5.6 Analysis and Design of Pin-jointed Trusses We shall now apply FEA to pin jointed trusses. The first thing that we must recognize is that the element stiffness matrix that we derived earlier for an axially loaded member is aligned with the x axis while those in a truss structure can be at various angles to the x axis. Since all external forces and displacements are measured by their components in the x and y directions (for 2D trusses) we must define each element stiffness matrix in terms of the displacement components in the x and y directions. Consider an element m at an angle α to the x axis as shown in Figure 5.6.1. What we need to do is find the displacement vectors uα and vα at each node in terms of its components in the xy axes directions. From Figure 5.6.1 we see that u αn = u n cos α + vn sin α u αn+1 = u n+1 cos α + vn+1 sin α vαn = −u n sin α + vn cos α vαn+1 = −u n+1 sin α + vn+1 cos α (5.6.1) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 203 A General Method for the Axially Loaded Slender Bar y u αn+1 υαn+1 un+1 l υn+1 υαn α un u αn υn x Figure 5.6.1 This can be put in matrix form      u αn un cos α sin α = vn vαn −sin α cos α  u αn+1 vαn+1  =  cos α −sin α sin α cos α  u n+1 vn+1  (5.6.2) For the two noded axial element there is no vα component, that is, the loads and displacements can only lie along the axis of the member, so we can set vαn = vαn+1 = 0 Let us define the nodal displacements as ⎡ ⎤ un ⎢ vn ⎥ ⎥ {rm } = ⎢ ⎣ u n+1 ⎦ vn+1 (5.6.3) ⎤ u αn ⎢ 0 ⎥ ⎥ {rαm } = ⎢ ⎣ u αn+1 ⎦ 0 ⎡ We can combine Equations 5.6.2 into one matrix set of equations. ⎤ ⎡ ⎡ ⎤ ⎤⎡ cos α sin α 0 0 u αn un ⎢ 0 ⎥ ⎢ −sin α cos α ⎥ ⎢ 0 0 ⎥ ⎥ ⎢ ⎥ ⎢ vn ⎥ = [T ] {rm } {rαm } = ⎢ ⎣ u αn+1 ⎦ = ⎣ 0 0 cos α sin α ⎦ ⎣ u n+1 ⎦ vn+1 0 0 0 −sin α cos α (5.6.4) (5.6.5) It just happens that this transformation matrix [T] has the interesting property that its inverse is equal to its transpose, or [T ]−1 = [T ]T . You can verify that [T ]T [T ] = [1] where [1] is the unit matrix ⎡ ⎤ 1 0 0 0 ⎢0 1 0 0⎥ ⎥ (5.6.6) [1] = ⎢ ⎣0 0 1 0⎦ 0 0 0 1 This allows us to state that {rαm } = [T ] {rm } {rm } = [T ]T {rαm } (5.6.7) { pαm } = [T ] { pm } { pm } = [T ]T { pαm } (5.6.8) and also that --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 204 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Since { pαm } = [kαm ] {rαm } [T ] { pm } = [kαm ] [T ] {rm } (5.6.9) [T ]T [T ] { pm } = { pm } = [T ]T [kαm ] [T ] {rm } = [km ] {rm } It follows that [km ] = [T ]T [kαm ] [T ] (5.6.10) Since [T] is a 4 × 4 matrix [kαm ] must be formatted as a 4 × 4 matrix as well. Note that [kαm ] is the element stiffness matrix for the two force member, that is, for α = 0. This element stiffness matrix in the standard format is    EA u αn 1 −1 (5.6.11) [kαm ] {rαm } = u αn+1 −1 1 l can be presented in the format ⎡ 1 EA ⎢ 0 ⎢ [kαm ] {rαm } = l ⎣ −1 0 0 −1 0 0 0 1 0 0 ⎤ ⎤⎡ 0 u αn ⎥ ⎢ 0⎥ ⎥⎢ 0 ⎥ 0 ⎦ ⎣ u αn+1 ⎦ 0 0 (5.6.12) so that the operation in Equation 5.6.10 can be carried out. This is best understood in an example. ########### Example 5.6.1 Problem: Find the element stiffness matrix for typical elements at 45◦ and 135o , 60o and 120o , and 90o and 180o . These will be used in the following examples. Solution: Find the matrix [T] and then find [km ] for each case. For 45o the matrix [T] is ⎡ ⎡ ⎤ cos α sin α 0 0 1 √ ⎢−sin α cos α 2⎢ 0 0 ⎥ −1 ⎢ ⎢ ⎥ = [T45 ] = ⎣ 0 0 cos α sin α ⎦ 2 ⎣ 0 0 0 −sin α cos α 0 The stiffness matrix for α = 45o is ⎤ ⎡ ⎡ 1 −1 0 0 1 √ EA ⎢ 2⎢ 1 1 0 0⎥ 0 T ⎥ ⎢ ⎢ [k45 ] = [T45 ] [k0 ] [T45 ] = 0 1 −1⎦ l ⎣−1 2 ⎣0 0 0 1 1 0 ⎡ ⎤ 1 1 −1 −1 EA ⎢ 1 1 −1 −1⎥ ⎢ ⎥ = 1 1⎦ 2l ⎣−1 −1 −1 −1 1 1 0 0 0 0 −1 0 1 0 1 0 1 0 0 1 0 −1 ⎤ 0 0⎥ ⎥ 1⎦ 1 ⎤ ⎡ 0 √ 1 1 ⎢ 0⎥ ⎥ 2 ⎢−1 1 0⎦ 2 ⎣ 0 0 0 0 0 (a) 0 0 1 −1 ⎤ 0 0⎥ ⎥ 1⎦ 1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 205 A General Method for the Axially Loaded Slender Bar For α = 135◦ the matrix [T] is ⎡ cos α ⎢−sin α ⎢ [T135 ] = ⎣ 0 0 sin α cos α 0 0 0 0 cos α −sin α ⎡ ⎤ 0 −1 √ ⎢ 0 ⎥ ⎥ = 2 ⎢−1 sin α ⎦ 2 ⎣ 0 cos α 0 1 0 −1 0 0 −1 0 −1 ⎤ 0 0⎥ ⎥ 1⎦ −1 (c) The stiffness matrix for α = 135◦ is [k135 ] = [T135 ]T [k0 ] [T135 ] ⎡ ⎡ ⎤ −1 −1 0 0 1 √ ⎢ ⎥ 2 ⎢ 1 −1 0 0⎥ EA ⎢ 0 ⎢ = 0 −1 −1⎦ l ⎣−1 2 ⎣ 0 0 0 1 −1 0 ⎡ ⎤ 1 −1 −1 1 ⎥ EA ⎢ −1 1 1 −1 ⎢ ⎥ = ⎣ 1 1 −1⎦ 2l −1 1 −1 −1 1 0 0 0 0 −1 0 1 0 ⎡ ⎤ ⎤ 0 √ −1 1 0 0 ⎢ ⎥ 0⎥ 2 ⎢−1 −1 0 0⎥ ⎥ ⎣ ⎦ 0 2 0 0 −1 1⎦ 0 0 0 −1 −1 (d) For α = 60◦ the matrix [T] is ⎡ cos α ⎢−sin α ⎢ [T60 ] = ⎣ 0 0 sin α cos α 0 0 0 0 cos α −sin α ⎡ ⎤ 1 0 √ ⎢ 0 ⎥ ⎥ = 1 ⎢− 3 sin α ⎦ 2 ⎣ 0 cos α 0 √ 3 1 0 0 ⎤ 0 ⎥ √0 ⎥ 3⎦ − 3 1 0 0 1 √ (e) The stiffness matrix for α = 60◦ is [k60 ] = [T60 ]T [k0 ] [T60 ] √ ⎤ ⎡ ⎡ 0 1 √1 − 3 0 ⎥ EA ⎢ 0 1⎢ 3 1 0 0 ⎢ √ ⎥ = ⎢ ⎦ ⎣−1 0 2⎣ 0 √1 − 3 l 0 3 1 0 0 √ √ ⎤ ⎡ 3 −1 √ − 3 √1 EA ⎢ 3 − 3 √ −3 ⎥ ⎥ ⎢ 3 √ = 3 1 3⎦ − 4l ⎣ −1 √ √ 3 3 − 3 −3 0 0 0 0 −1 0 1 0 ⎤ ⎡ 1 0 √ ⎢− 3 1 0⎥ ⎥ ⎢ 0⎦ 2 ⎣ 0 0 0 √ 3 0 1 0 0 1 √ 0 − 3 ⎤ 0 ⎥ √0 ⎥ 3⎦ 1 (f) For α = 120◦ the matrix [T] is ⎡ cos α ⎢−sin α [T120 ] = ⎢ ⎣ 0 0 --`,,,,``,``,,``,,`,`,,,,,``,, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS sin α cos α 0 0 0 0 cos α − sin α ⎡ ⎤ −1 0 √ ⎢− 3 1 0 ⎥ ⎥= ⎢ sin α ⎦ 2 ⎣ 0 cos α 0 √ 3 −1 0 0 ⎤ 0 0 0 √ 0⎥ ⎥ 3⎦ −1 √ − 3 −1 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (g) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 206 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The stiffness matrix for α = 120◦ is [k120 ] = [T120 ]T [k0 ] [T120 ] √ ⎤ ⎡ ⎡ −1 − 3 0 0 1 √ ⎥ EA ⎢ 0 1⎢ 3 −1 0 0 ⎥ ⎢ ⎢ √ = ⎣ 0 0 √ −1 − 3⎦ l ⎣−1 2 0 3 −1 0 0 √ √ ⎤ ⎡ 1 − 3 √ −1 3 √ ⎥ EA ⎢ 3 3 3 −3 − ⎢ √ ⎥ √ = ⎦ 3 1 − 3 −1 4l ⎣ √ √ 3 −3 − 3 3 0 0 0 0 −1 0 1 0 √ ⎤ ⎡ −1 3 0 0 √ ⎢ 0 0⎥ ⎥ 1 ⎢− 3 −1 0⎦ 2 ⎣ 0 0 −1 √ 0 0 0 − 3 ⎤ 0 ⎥ √0 ⎥ 3⎦ −1 (h) For α = 90◦ the matrix [T] is ⎡ cos α ⎢− sin α ⎢ [T90 ] = ⎣ 0 0 sin α cos α 0 0 0 0 cos α − sin α ⎤ ⎡ 0 0 ⎢ 0 ⎥ ⎥ = ⎢−1 sin α ⎦ ⎣ 0 cos α 0 1 0 0 0 0 0 0 −1 ⎤ 0 0⎥ ⎥ 1⎦ 0 (i) The stiffness matrix for α = 90◦ is ⎡ ⎤ −1 0 0 1 ⎢ 0 EA 0 0 0⎥ ⎢ ⎥ 0 0 −1⎦ l ⎣−1 0 1 0 0 ⎡ ⎤ 0 0 0 0 EA ⎢ 1 0 −1⎥ ⎢0 ⎥ = 0 0 0⎦ l ⎣0 0 −1 0 1 0 ⎢ 1 [k90 ] = [T90 ]T [k0 ] [T90 ] = ⎢ ⎣0 0 0 0 0 0 −1 0 1 0 ⎤⎡ 0 0 ⎢−1 0⎥ ⎥⎢ 0⎦ ⎣ 0 0 0 1 0 0 0 0 0 0 −1 ⎤ 0 0⎥ ⎥ 1⎦ 0 (j) For α = 180◦ the matrix [T] is ⎡ cos α ⎢− sin α [T180 ] = ⎢ ⎣ 0 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS sin α cos α 0 0 0 0 cos α − sin α ⎤ ⎤ ⎡ −1 0 0 0 0 ⎢ 0 0⎥ 0 ⎥ ⎥ ⎥ = ⎢ 0 −1 0 −1 0⎦ sin α ⎦ ⎣ 0 0 0 0 −1 cos α Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (k) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎡ P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 207 A General Method for the Axially Loaded Slender Bar The stiffness matrix for α = 180◦ is ⎡ −1 0 ⎢ 0 −1 T ⎢ [k180 ] = [T180 ] [k0 ] [T180 ] = ⎣ 0 0 0 0 ⎡ 1 EA ⎢ 0 ⎢ = l ⎣−1 0 ⎡ ⎤ 0 1 ⎢ 0⎥ ⎥ EA ⎢ 0 0⎦ l ⎣−1 −1 0 ⎤ −1 0 0 0⎥ ⎥ 1 0⎦ 0 0 0 0 −1 0 0 0 0 0 0 0 0 0 −1 0 1 0 ⎤⎡ ⎤ 0 −1 0 0 0 ⎢ 0⎥ 0 0⎥ ⎥ ⎢ 0 −1 ⎥ 0⎦ ⎣ 0 0 −1 0⎦ 0 0 0 0 −1 (l) Example 5.6.2 Problem: Find the displacements and stresses for the pin jointed truss in Figure (a). All members are made of the same material and have the same cross section area. We have numbered the nodes and elements on the figure. Let FH = 400 N FV = 1000 N 1000 mm n2 1000 mm n3 e2 e1 (a) n4 e3 60o 45o FH n1 FV Figure (a) Solution: Assemble the global equations, partition, and solve. The method for matrix multiplication and the equation solving which follow are presented for Mathematica in Appendix C. Note that this is an indeterminate truss. The length of the members are l1 = √ 2 · 1000 mm l2 = 1000 mm 2 l3 = √ · 1000 mm 3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 208 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Now the assembly process is a bit more complicated because the element matrices do not just descend down the diagonal as in the case of a single axially loaded member. In this case nodes 1 and 2 are associated with element 1, nodes 1 and 3 are associated with element 2, and nodes 1 and 4 are associated with element 3. Each element matrix must be imbeded in the larger global matrix as shown here. The stiffness matrix for element 1 (α = 135◦ )is given in Equation (c). √ √ √ ⎤ ⎡ √ 2 2 2 2 − − ⎢ 4 ⎥ 4 4 4 ⎢ ⎥ ⎢ ⎥⎡ ⎤ √ √ √ √ ⎡ ⎡ ⎤ ⎤⎡ ⎤ ⎢ ⎥ 2 2 2 2 p1,1x u1 1 −1 −1 1 ⎢− ⎥ u1 − ⎢ ⎥ ⎢ p1,1y ⎥ EA ⎢−1 ⎢ ⎥ 4 4 4 4 ⎥⎢ v1 ⎥ 1 1 −1⎥ ⎢ ⎢ ⎥= ⎥ ⎥ ⎢ v1 ⎥ = EA ⎢ ⎢ ⎥⎢ √ √ √ √ ⎣u 2 ⎦ (c) ⎣ p1,2x ⎦ 2l1 ⎣−1 1 1 −1⎦ ⎣u 2 ⎦ 1000 ⎢ ⎥ 2 2 2 2 ⎢ ⎥ − v2 p1,2y 1 −1 −1 1 ⎢− ⎥ v2 ⎢ 4 4 4 4 ⎥ ⎢ √ ⎥ √ √ √ ⎣ 2 2 2 2 ⎦ − − 4 4 4 4 It is embedded in the global matrix format in Equation (d). √ √ √ ⎡ √ 2 2 2 2 − − ⎢ 4 4 4 4 ⎢ √ √ √ ⎢ √ ⎢ ⎡ ⎤ 2 2 2 ⎢− 2 p1,1x − ⎢ 4 4 4 4 ⎢ ⎢ p1,1y ⎥ ⎢ √ ⎢ ⎥ √ √ √ ⎢ ⎢ p1,2x ⎥ 2 2 2 ⎢− 2 ⎢ ⎥ − ⎢ ⎢ p1,2y ⎥ 4 4 4 ⎢ ⎥ = EA ⎢ 4 ⎢ 0 ⎥ 1000 ⎢ √ √ √ √ ⎢ 2 ⎢ ⎥ 2 2 2 ⎢ ⎢ 0 ⎥ − − ⎢ ⎢ ⎥ ⎢ 4 4 4 4 ⎣ 0 ⎦ ⎢ ⎢ 0 0 0 0 0 ⎢ ⎢ 0 0 0 0 ⎢ ⎣ 0 0 0 0 0 0 0 0 ⎤ 0 0 0 0⎥ ⎥ ⎥ ⎥⎡ ⎤ 0 0 0 0⎥ ⎥ u1 ⎥ ⎢ v1 ⎥ ⎥⎢ ⎥ ⎥ ⎢u 2 ⎥ ⎢ ⎥ 0 0 0 0⎥ ⎥ ⎢ v2 ⎥ ⎥⎢ ⎥ ⎥ ⎢u 3 ⎥ ⎥⎢ ⎥ ⎥⎢ ⎥ 0 0 0 0⎥ ⎢ v3 ⎥ ⎥ ⎣u 4 ⎦ ⎥ 0 0 0 0⎥ ⎥ v4 0 0 0 0⎥ ⎥ 0 0 0 0⎦ 0 0 0 0 (d) Element 2 (α = 90◦ ) imbedded in the global matrix format is given in Equation (e). ⎤ ⎡ ⎡ ⎤⎡ ⎤ p2,1x 0 0 0 0 u1 ⎢ p2,1y ⎥ ⎢0 ⎥ ⎢ v1 ⎥ EA 1 0 −1 ⎥ ⎢ ⎢ ⎥⎢ ⎥ ⎣ p2,3x ⎦ = 1000 ⎣0 0 0 0⎦ ⎣u 3 ⎦ p2,3y v3 0 −1 0 1 ⎡ p2,1x ⎢ p2,y ⎢ ⎢ 0 ⎢ ⎢ 0 ⎢ ⎢ p2,3x ⎢ ⎢ p2,3y ⎢ ⎣ 0 0 ⎤ ⎡ 0 0 ⎥ ⎢0 1 ⎥ ⎢ ⎥ ⎢0 0 ⎥ ⎢ ⎥ ⎢0 EA 0 ⎥= ⎢ ⎥ 1000 ⎢0 0 ⎥ ⎢ ⎥ ⎢0 −1 ⎥ ⎢ ⎦ ⎣0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 −1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 ⎤⎡ ⎤ u1 0 ⎢ ⎥ 0⎥ ⎥ ⎢ v1 ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢u 2 ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢ v2 ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢u 3 ⎥ ⎥ ⎥ 0⎥ ⎢ ⎢ v3 ⎥ 0⎦ ⎣u 4 ⎦ v4 0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (e) P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come 209 A General Method for the Axially Loaded Slender Bar Element 3 (α = 60◦ ) imbedded in the global matrix format is given in Equation (f). √ √ ⎤⎡ ⎤ 3 −1 − 3 u1 √ √1 ⎥ ⎢ v1 ⎥ ⎢ 3 3 − 3 −3 ⎥ ⎥ ⎢ ⎢ √ √ ⎦ ⎣u 4 ⎦ ⎣ −1 − 3 1 3 √ √ v4 3 3 − 3 −3 √ √ ⎡ ⎤ 3 3 3 3 − − ⎢ 8 8 8 8 ⎥ ⎢ ⎥ ⎢ ⎥⎡ ⎤ √ √ 3 ⎥ u1 3 3 ⎢ 3 ⎢ ⎥ − 3 − 3 ⎢ ⎥ EA ⎢ 8 8 8 8 ⎥ ⎢ √ ⎥ ⎢ v1 ⎥ = √ ⎢ ⎥ ⎣u 4 ⎦ 1000 ⎢ 3 3 3 ⎥ ⎢− 3 ⎥ v4 − ⎢ 8 8 8 8 ⎥ ⎢ ⎥ ⎣ ⎦ 3√ 3 3 3√ − − 3 3 8 8 8 8 ⎡ √ ⎤ √ 3 3 3 3 0 0 0 0 − − ⎥ ⎢ 8 8 8 ⎥ ⎢ 8 ⎢ ⎥ √ √ ⎢ 3 ⎡ ⎤ ⎤ ⎡ 3 ⎥ 3 3 ⎢ ⎥ u1 p3,1x − 3 0 0 0 0 − 3 ⎢ 8 ⎥ 8 8 8 ⎥ ⎢v ⎥ ⎢ ⎢ p3,1y ⎥ ⎢ ⎥ ⎢ 1⎥ ⎥ ⎢ ⎢ ⎥ ⎢ 0 ⎢ 0 ⎥ 0 0 0 0 0 0 0 ⎥ ⎢ ⎥ ⎢u 2 ⎥ ⎥ ⎢ ⎢ 0 ⎥ ⎢ v2 ⎥ ⎢ 0 ⎥ EA 0 0 0 0 0 0 0 ⎢ ⎥⎢ ⎥ ⎥ ⎢ ⎢ ⎥ ⎢ 0 ⎥ = 1000 ⎢ 0 0 0 0 0 0 0 0 ⎥ ⎢ ⎥ ⎢u 3 ⎥ ⎥ ⎢ ⎢ 0 ⎥ ⎥ ⎢ 0 ⎥ 0 0 0 0 0 0 ⎥⎢ ⎢ v3 ⎥ ⎢ √ ⎥ ⎢ √0 ⎢ ⎥ ⎣ ⎣ p3,4x ⎦ u 3 3 3 ⎥ 4⎦ ⎢− 3 − 0 0 0 0 ⎢ ⎥ p3,4y 8 8 8 ⎥ v4 ⎢ 8 ⎢ ⎥ ⎢ 3 3√ 3 3√ ⎥ ⎣ − − 3 0 0 0 0 3⎦ 8 8 8 8 ⎤ p3,1x ⎢ p3,1y ⎥ EA ⎥ ⎢ ⎣ p3,4x ⎦ = 4l p3,4y ⎡ ⎡ (f) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The only nonzero displacements are at node one. To find the displacements we need only those terms in the assembled matrices that are part of the partitioned matrix equations. The assembled matrices are given in Equation (g) with just the necessary terms for finding the displacements. The symbol c is just to identify the location of the other terms in the global stiffness matrix. ⎤ ⎡ ⎡ ⎤ p1,1x + p2,1x + p3,1x 0.57006 0.02145 FH ⎢0.02145 2.00307 ⎢−FV ⎥ ⎢ p1,1y + p2,1y + p3,1y ⎥ ⎥ ⎢ ⎢ ⎢ ⎥ ⎥ ⎢ ⎢ ⎢ ⎥ ⎢ c ⎢ R2x ⎥ ⎢ ⎥ c c ⎥ ⎢ ⎢ ⎥ EA ⎢ ⎢ c ⎢ R2y ⎥ ⎢ ⎥ c c = = ⎥ ⎢ ⎢ ⎢ ⎥ ⎢ R3x ⎥ ⎢ ⎥ 1000 ⎢ c c c ⎥ ⎢ ⎢ ⎢ ⎥ ⎢ c ⎢ R3y ⎥ ⎢ ⎥ c c ⎥ ⎢ ⎢ ⎢ ⎥ ⎣ c ⎣ R4x ⎦ ⎣ ⎦ c c R4y c c c ⎡ Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c ⎤⎡ ⎤ u1 c ⎢ v1 ⎥ c⎥ ⎥⎢ ⎥ ⎥⎢ ⎥ ⎢ ⎥ c⎥ ⎥⎢ 0 ⎥ ⎢ ⎥ c⎥ ⎥ ⎢ 0 ⎥ (g) ⎥ ⎥ c⎥ ⎢ ⎢0⎥ ⎢0⎥ c⎥ ⎥⎢ ⎥ c⎦ ⎣ 0 ⎦ c 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas 210 July 11, 2011 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Partition and solve for the displacements.      EA 0.57006 0.02145 u 1 400 = −1000 1000 0.02145 2.00307 v1     1000 720.76 u → 1 = v1 EA −506.95 (h) Now you could complete the global stiffness matrix and partition and solve for the restraint forces and then resolve those forces to find the internal axial forces in each member. It is easier to solve for the internal forces directly since we already have those matrix equations formed. The internal force in element 1 can be found from the following equations. ⎤ p1,1x ⎢ p1,1y ⎥ ⎥ ⎢ ⎣ p1,2x ⎦ p1,2y ⎡ ⎡ √ 2 ⎢ 4 ⎢ ⎢ √ ⎢ ⎢− 2 EA ⎢ ⎢ 4 = ⎢ 1000 ⎢ √2 ⎢ ⎢− ⎢ 4 ⎢ √ ⎣ 2 √ √ √ ⎤ 2 2 2 − − 4 4 4 ⎥ ⎥ √ ⎥⎡ ⎤ √ √ 2 2 2⎥ ⎥ u1 − ⎥ ⎥ 4 4 4 ⎥⎢ ⎢ v1 ⎥ √ ⎥ √ √ ⎣0⎦ ⎥ 2 2 2⎥ − ⎥ 0 4 4 4 ⎥ √ √ √ ⎥ 2 2 2 ⎦ − − 4 4 4 4 ⎡ √ √ ⎤ 2 2   − ⎥  EA ⎢ 4 ⎥ u1 p1,2x ⎢ 4 → = ⎢ √ ⎥ p1,2y 1000 ⎣ √2 2 ⎦ v1 − 4 4      −434.06 720.76 −0.35355 0.35355 N = = 434.06 0.35355 −0.35355 −506.95 The internal axial force in member 1 is % P1 = (−434.06)2 + (434.06)2 = 613.85 N The internal force is element 2 is found from ⎡ ⎤ ⎡ ⎤⎡ ⎤ p2,1x 0 0 0 0 u1 ⎢ p2,1y ⎥ ⎢ ⎥ ⎢ v1 ⎥ EA 0 1 0 −1 ⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎣ p2,3x ⎦ = 1000 ⎣0 0 0 0⎦ ⎣ 0 ⎦ p2,3y 0 0 −1 0 1           EA 0 0 u1 0 720.76 0 0 p2,3x = N = = → p2,3y 506.95 0 −1 −506.95 1000 0 −1 v1 (i) (j) (k) The internal axial force in member 2 is P2 = 506.95 N Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (l) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 10:52 Printer Name: Yet to Come A General Method for the Axially Loaded Slender Bar The internal force is element 3 is found from √ ⎡ √ ⎤ 3 3 3 3 − − ⎢ 8 8 8 8 ⎥ ⎢ ⎥ ⎢ ⎥⎡ ⎤ ⎡ ⎤ √ √ 3 ⎥ u1 3 3 ⎢ 3 p3,1x ⎢ ⎥ − 3 − 3 ⎢ ⎥ ⎢ 8 ⎢ p3,1y ⎥ 8 8 8 ⎥ ⎢ ⎥ ⎢ v1 ⎥ ⎥ = EA ⎢ √ √ ⎢ ⎥ ⎣0⎦ ⎣ p3,4x ⎦ 1000 ⎢ 3 3 3 ⎥ ⎢− 3 ⎥ 0 − p3,4y ⎢ 8 8 8 8 ⎥ ⎢ ⎥ ⎣ ⎦ 3√ 3 3 3√ − − 3 3 8 8 8 8 ⎤ ⎡ √ 3 3     − − ⎥ u1 EA ⎢ p3,4x 8 8 ⎥ ⎢ → = ⎦ v1 p3,4y 1000 ⎣ 3 3√ − − 3 8 8      EA −0.21651 34.06 720.76 −0.375 N = = 58.99 −0.64952 −506.95 1000 −0.375 The internal axial force in member 3 is % P3 = (34.06)2 + (58.99)2 = 68.12 N 211 (m) (n) (o) As a check see if the truss member force components and the applied loads are in equilibrium at node 1.  Fy = (434.06 + 506.95 + 58.99 − 1000) = 0 (p)  Fx = (−434.06 + 0 + 34.06 + 400) = 0 They check. ########### The solution of even such a simple truss problem like this by either the classical differential equation method or the modern finite element method is quite a chore. Fortunately, the finite element method has been programmed for graphical input of the geometry, automatic creation of nodes and elements, assembly of the matrices, and direct application of loads and restraints to the assembled equations. Partitioning of the equation and their solution is also automated. And additional bonus is the graphical as well as numerical presentation of the results. 5.7 Summary and Conclusions The finite element method is especially useful for formulating problems for numerical analysis. The equations for this method are summarized in Equations 5.4.5 and repeated here in Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-05 P1: TIX/XYZ JWST071-05 P2: ABC JWST071-Waas July 11, 2011 212 10:52 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Equation 5.7.1. {F} f x (s) {Fc } {Fem } =  lm { pm } → [n]T f x (sm ) dsm 0 {Fe } = σxm  εxm du m d = = [n] {rm } dsm dsm σxm = E m εxm = E m d [n] {rm } dsm → {rm } ← {ρ} u (sm ) = [n] {rm } {Fem } {F} = {Fc } + {Fe } pm,n = −Am σxm = −E m Am u (sm ) εxm d [n] {rm } dsm { pm } =  d pm,n+1 = Am σxm = E m Am [n] {rm } ds & m & {F} = { pm } = [km ] {r } = [K ] {r } (5.7.1)  pm,n = [km ] {rm } pm,n+1 The critical factor here is the discovery and recognition of shape functions [n].    s m sm un u(sm ) = [n] {rm } = 1 − lm lm u n+1 (5.7.2) In Section 5.6 we extend the analysis to pin jointed trusses and introduce the transformation for axial element stiffness matrices oriented at positions other than along the x axis. The transformation matrix is ⎡ ⎤ cos α sin α 0 0 ⎢−sin α cos α 0 0 ⎥ ⎥ (5.7.4) [T ] = ⎢ ⎣ 0 0 cos α sin α ⎦ 0 0 −sin α cos α Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Note that for axial bars with concentrated loads the equations for the classical method are exactly satisfied. To use this method distributed loads are replaced with equivalent concentrated nodal loads. This is done using Equation 5.4.4 repeated here as Equation 5.7.3. The derivation of this relation is given in a later chapter. ⎡ sm ⎤    lm  lm 1 − Fem,n lm ⎥ ⎢ {Fem } = (5.7.3) [n]T f x (sm ) dsm = ⎣ sm ⎦ f x (sm ) dsm = Fem,n+1 0 0 lm P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 6 Torsion 6.1 Introduction In this chapter we consider the special case of a slender bar or shaft that is loaded in pure torque about its centroidal axis and is restrained from any rigid body motion. This has come to be known as the subject of torsion. We shall assume that any local effects of load application and restraint are neglected. There are, of course, local effects but we depend upon St. Venant’s principle to allow us to gain useful information about the behavior if the shaft is long and slender. The equations and solution methods found here for torsion are very similar in form to those in Chapters 4 and 5 that govern axial deformation. In fact, we will depend heavily on what we learned there to understand what is happening here. Torsional Displacement, Strain, and Stress To gain a preliminary understanding of torsion we conduct an experiment. Consider a slender bar with a circular cross section restrained at x = 0 and with a concentrated applied moment or torque ML at x = L as shown in Figure 6.2.1. A double arrowhead is used to indicate a moment. The moment is positive by the right hand rule. Just how the moment is applied is not yet defined. The body deforms in rotation and the total angle of twist at x = L as a result of the load is designated as φ. A γ B φ C ML x Figure 6.2.1 Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 6.2 P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 214 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Experimental evidence confirms that the distributed angle of twist β(x) is proportional to the distance along the bar. Then β(x) = φ x L (6.2.1) Let us look at a free body portion of the bar as shown in Figure 6.2.2 and invoke static equilibrium. T ML x L Figure 6.2.2 From summation of moments the internal torque T is a constant equal to the applied moment ML . Mx = M L − T = 0 → T = ML (6.2.2) We now adopt cylindrical coordinates with the origin at the center of the circular cross section and with r any point along the radius of the cross section. In Section 3.2 we define the shearing strain as a change in angular displacement. Such an angular change is noted in Figure 6.2.1 and is the angle between lines AB and AC at the outer radius. For small displacements the sine of the angle is equal to the angle and shearing strain at the outer surface is seen to be γmax = φ rmax L (6.2.3) It varies linearly to zero at the origin of the radius; therefore, the shearing strain at any point in the body is φ r L γ (x, r ) = (6.2.4) Given that Hooke’s law applies we have from Equation 3.4.13 τ = Gγ (6.2.5) where G is the shear modulus. Thus τ (x, r ) = Gφ r L (6.2.6) We note that the torque on any cross section is T = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  A τ rdA = G  A γ rdA = G φ L  r 2 dA (6.2.7) A Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`---  P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 215 Torsion We define  r 2 dA = J (6.2.8) A and we note that this is the polar moment of inertia of the circular cross section and is known as the torsional constant. Thus we obtain T = GJφ L → φ= TL ML L = GJ GJ (6.2.9) The stress in terms of the applied load and the properties of the bar is found by combining Equations 6.2.6 and 6.2.9.   Gφ G TL Tr τ= r= (6.2.10) r= L L GJ J Note, this is true only for a circular cross section. Non circular cross sections require a more advanced application of the theory of elasticity to define the stress distribution and the torsional constant. That is beyond the scope of this work for the most part; however, in Sections 6.5–6.8 we shall extend the analysis to consider the torsional stress in a class of cross sections of the so called thin walled type. The shear stress distribution given in Equation 6.2.10 is a linear function in the radial direction as shown in Figure 6.2.3. τ --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- r Figure 6.2.3 The maximum shear stress is found at the outer radius. τmax = T rmax J (6.2.11) It may be noted that the polar moment of inertia of a solid circular section is (see Appendix B) J= 4 πrmax 2 (6.2.12) An actual experiment can be performed to find the value of the shear modulus for a particular material. We can take a circular shaft of that particular material, fasten it at one end, and apply a moment at the other, as in Figure 6.2.1, and gradually increase the moment M L as we measure the rotation φ and plot Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 216 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods the results. We will get a plot very like the one in Figure 3.4.2, shown here with appropriate symbols for torsion in Figure 6.2.4. ML ML0 linear elastic φ Figure 6.2.4 The slope of the strait portion we shall call kt then kt = ML φ φ= ML kt (6.2.13) where kt is called the torsional stiffness or torsional spring constant. The proportional limit is reached at M L0 . This is the point at which the curve in no longer linear. These values apply to the particular bar tested and are dependent upon the length of the bar and the radius of the cross section as well as the material. To obtain the shear modulus independent of particular dimensions of the bar and thus related only to the material of the bar we replot the data in terms of stress and strain as shown is Figure 6.2.5. We then find G as the slope of the straight portion of the curve. ML r J max slope G γ max = φ r L max Figure 6.2.5 6.3 Derivation and Solution of the Governing Equations Now consider the more general case of a circular shaft loaded with some combination of distributed torsional loads, t(x), and concentrate torsional loads, M c , as shown, for example, in Figure 6.3.1. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- τ max = P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 217 Torsion y z t(x) Mc x Figure 6.3.1 We shall denote the distributed displacement as β(x). We may surmise as a generalization of Equation 6.2.4 that the local value of strain is proportional to the local rate of change of angular displacement. The strain displacement equation is then γ (x, r ) = dβ r dx (6.3.1) Tests verify that this is a valid assumption for long slender bars with circular cross sections. Combining Equations 6.2.5 and 6.3.1 we get for Hooke’s law τ (x, r ) = Gγ = G dβ r dx (6.3.2) The internal torque T (x) is T (x) =  τ rdA = G A  γ rdA = G A dβ dx  r 2 dA = GJ A dβ dx (6.3.3) Consider the equilibrium of a short segment of the bar of length dx with a distributed moment tx (x) applied as shown in Figure 6.3.2. We include the values of the internal torque at x and x + dx from a Taylor series expansion neglecting higher order terms. t(x) T (x + dx) = T ( x) + T(x) dT dx dx dx Figure 6.3.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 218 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Summing the torques we satisfy static equilibrium.  which reduces to Mx = T (x) + dT dx − T (x) + t(x)dx = 0 dx dT = −t(x) dx (6.3.4) (6.3.5) Combining Equations 6.3.3 and 6.3.5 we obtain dβ d GJ = −t(x) dx dx (6.3.6) From Equations 6.3.2 and 6.3.3 we get τ (x, r ) = Gγ (x, r ) = G T (x) dβ r= r dx J (6.3.7) In summary Mc → tx (x) T (x) dT = −tx (x) dx τ (x, r ) γ (x, r ) β(x) ← ρ dβ r γ (x, r ) = dx dβ τ (x, r ) = Gγ = G r dx    T (x) dβ dβ → τ (x, r ) = r T (x) = τ rdA =G γ rdA = G r 2 dA = GJ dx dx J A A (6.3.8) A d dβ dT = GJ = −t(x) dx dx dx In the first row of Equation 6.3.8 we have the typical unknowns in the formulation of a torsion problem. β(x) γ (x, r ) τ (x, r ) T (x) − − − − distributed rotational displacement shearing strain in cylindrical coordinates shearing stress in cylindrical coordinates internal torque about the x axis tx (x) − distributed applied torque Mc − concentrated applied torques ρ − displacement restraints The arrows are there to indicate that both Mc and ρ are boundary conditions to be satisfied in the process of solving the problem. The equations connecting these quantities are in rows two and three through four. dT = −tx (x) dx dβ r γ = dx dβ τ = Gγ = G r dx  dβ T = τ rdA = GJ dx − static equilibrium − strain displacement − Hooke’s law for shear − internal torque definition A Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- and the known quantities P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 219 Torsion and in the fifth row equilibrium is given in terms of the single unknown displacement. dβ d GJ = −t(x) dx dx (6.3.9) And, of course, when GJ is constant GJ d 2β = −t(x) dx2 (6.3.10) Now everything we learned in Chapter 4 about solving a second order differential equation by direct integration applies here. From Equation 6.3.10 we get 1 dβ =− dx GJ  tx (x)dx + a → β(x) = − 1 GJ    tx (x)dx dx + ax + b (6.3.11) where the constants of integration, a and b, are found by assigning boundary conditions. Having found the rotational displacement β(x) we then find the internal torque T from Equation 6.3.3 and then the stress τ from Equation 6.3.7. That is, in order we find β(x) → T = GJ dβ dx → τ= T r J (6.3.12) This solution works for both statically determinate and indeterminate cases. In the determinate case the internal torque, T, can be found directly from the equations of static equilibrium and so we can go directly to the stress in Equation 6.3.7. If the displacement is desired we use β(x) = 1 GJ  T (x)dx + c (6.3.13) where c is the constant of integration to be determined from a displacement boundary condition. We shall illustrate both approaches to a solution in the next simple example. Here we pause to remind you that the set of equations and flow chart for the torsion problem is identical to the set of equations and flow chart for axial deformation if we make the following substitutions shown in Table 6.3.1. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Axial Torsional u(x) EA f (x) dP/dx = − f (x) . . . β(x) GJ t(x) dT/dx = −t(x) . . . Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Table 6.3.1 P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 220 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 6.3.1 Problem: A shaft with a circular cross section is fixed at one end and has a concentrated applied torque at the other end as shown in Figure (a). Find the stresses and displacements of the shaft. ML L Figure (a) Note: This is the problem we used as an experiment in Section 6.2 to aid in the derivation of the torsional equations. We shall now use it to illustrate how the differential equations and boundary conditions are used to solve the displacements and stresses. Solution: First find the internal torque, the stress, and then solve the first order equation for the displacement. Then repeat the solution using the second order displacement equation. As a first step, use equilibrium to find the internal torque in terms of the applied moment. A simple free body diagram of a portion of the bar is shown in Figure 6.2.2, and repeated here as Figure (b) ML T x L Figure (b) → T = ML (a) From Equation 6.3.7 the stress is τ= ML T r= r J J (b) The displacement is obtained by integrating the first order equation.  ML dβ ML = M L → β(x) = dx + c = x +c GJ dx GJ GJ (c) From the boundary condition on the displacement at the left end β(0) = ML ·0+c =0 GJ → c=0 → β(x) = ML x GJ (d) Now let us repeat the solution by starting with the second order equation for the displacement, Equation 6.3.10, and integrating twice. We get GJ Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS d 2β =0 dx2 → dβ =a dx → β(x) = ax + b Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (e) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- From summation of moments the internal torque is  Mx =M L − T = 0 P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 221 Torsion Now apply the boundary conditions for displacement on the left end and for applied load on the right end to find the constants of integration. β(0) = a · 0 + b = 0 → b = 0 ML dβ(L) = GJa = M L → a = GJ dx GJ → ML x GJ β(x) = (f) Now that we know β(x) we can find T T = GJ dβ ML = GJ = ML dx GJ (g) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- and then find τ τ= ML T r= r J J (h) Take a quick look at Example 4.3.1 for the axially loaded bar. This is exactly the same problem but with different symbols. Mathematically the torsional and axial cases for finding the displacements and internal forces are identical. This will apply to many other problems in this chapter. Example 6.3.2 Problem: A shaft with a circular cross section is fixed at one end and has two concentrated applied torques as shown in Figure (a). Find the stresses and displacements of the shaft. ML/2 ML L/ 2 L Figure (a) Solution: Divide the shaft into two regions and integrate each equation. Apply boundary conditions to determine the values of the constants of integration. The stresses are easy to find because this is a statically determinate case. The internal torque is Region 1: Region 2:   Mz = M L/2 + M L − T1 = 0 Mz = M L − T2 = 0 → → T1 = M L/2 + M L T2 = M L 0≤x ≤ L 2 L <x≤L 2 (a) The stresses are found by inserting the internal torque values in Equation 6.3.7. L M L/2 + M L r 0≤x ≤ J 2 L ML r ≤x≤L τ2 = J 2 τ1 = (b) The displacements require a bit more work. Let us call the displacements β1 from 0 ≤ x ≤ Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS L 2 and β2 from L ≤x≤L 2 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 222 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods then dβ1 M L/2 + M L = M L/2 + M L → β1 = x + a1 dx GJ dβ2 ML = M L → β2 = x + a2 GJ dx GJ GJ (d) The boundary conditions are on the displacement at the left end and continuity of displacement at the point where the two displacement functions meet.     L L = β2 (e) β1 (0) = 0 β1 2 2 Therefore M L/2 + M L β1 (0) = · 0 + a1 = 0 → a1 = 0 GJ     ML L M L/2 + M L L L L = + a2 = β2 = β1 2 2 GJ 2 GJ 2 → a2 = M L/2 L 2GJ (f) The displacements are β1 = M L/2 + M L x GJ β2 = M L/2 L ML x+ GJ 2GJ (g) The same results can be obtained by integrating the second order equation twice for each region and applying a boundary condition at each end and continuity of displacement and torque at x = L/2. The equations and their solution are d 2 β1 =0 dx2 d 2 β2 GJ 2 = 0 dx GJ → β1 = a1 x + b1 → β2 = a2 x + b2 (h) The boundary conditions are the displacement at the left end, the applied moment at the right end, equilibrium of external and internal moments and continuity of displacement where the two regions meet. β1 (0) = 0 dβ2 (L) = ML GJ dx     L L dβ1 dβ2 2 2 − GJ + M L/2 = 0 GJ dx dx     L L = β1 β2 2 2 (i) Apply the boundary condition at the left end. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- β(0)1 = a1 · 0 + b1 = 0 → b1 = 0 → a2 = (j) Next satisfy the applied torque at the right end. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS GJ dβ2 (L) = GJa2 = M L dx ML GJ Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (k) P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 223 Torsion Now satisfy equilibrium at the point where the two regions meet.     L L dβ2 dβ1 2 2 GJ − GJ + M L/2 = GJa2 − GJa1 + M L/2 = M L − GJa1 + M L/2 = 0 dx dx M L + M L/2 → a1 = GJ Finally, satisfy continuity of displacement at the midpoint.     L ML L M L + M L/2 L L L L + b2 = = β1 = a2 · + b2 = a1 · + b1 = β2 2 2 2 2 GJ 2 GJ 2 → b2 = M L/2 L GJ 2 (l) (m) The displacements are β1 = M L/2 + M L x GJ β2 = M L/2 L ML x+ GJ 2GJ (n) Now that the displacements are known you can find the internal torques and the stresses as shown in Equation 6.3.12, repeated here as Equation (o). → T = GJ dβ dx → τ= T r J (o) Example 6.3.3 Problem: Consider other loadings and restraints, for example, a uniform distributed applied moment t0 with the left end fixed, as shown in Figure (a), or a shaft with both ends fixed and a concentrated applied moment at x0 as shown in Figure (b). t0 L Figure (a) M0 x0 L Figure (b) Solution: Integrate the second order equation and apply boundary conditions to determine the values of the constants of integration. Since the second order equation and restraints for torsion are the same form as the second order equation and restraints for an axially loaded bar the solution methods are identical. Just the symbols and the quantities they represent are different. Therefore the solution procedure is exactly the same for these two problems as found in Examples 4.4.1 and 4.4.2. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- β(x) P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 224 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods We shall just summarize the results here. For the shaft in Figure (a)   x2 t0 Lx − β(x) = GJ 2 dβ T = GJ = t0 (L − x) dx t0 dβ τ = G r = (L − x)r dx J (a) For the shaft in Figure (b) x0  Mx0  1− x GJ L  x0  dβ = Mx0 1 − T1 = GJ dx L dβ x0 = −Mx0 T2 = GJ dx L x M x0  1− x0 GJ L Mx0  x0  dβ τ1 = G r = 1− r dx J L Mx x0 dβ τ2 = G r = − 0 r dx J L β1 (x) = β2 (x) = (b) ########### In summary there are four possible sets of boundary restraint. For one the shaft can be in static equilibrium under a set of applied loads as shown in Figure 6.3.3. tx(x) ML Figure 6.3.3 In such cases we assign the origin of the coordinate system to be the point from which the displacements are measured. For all practical purposes it amounts to the kinds of restraints in the next figure. The one in Figure 6.3.3 and the two shown in Figure 6.3.4 are statically determinate. tx(x) tx(x) or ML M0 Figure 6.3.4 The fourth case shown in Figure 6.3.5 with both ends fixed is indeterminate. tx(x) Figure 6.3.5 Either the first order or the second order equation for displacement may be used in the first three cases. The second order equation must be used in the fourth case. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- M0 P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 225 Torsion If t(x) is discontinuous, or E is discontinuous, or A is discontinuous, or midpoint concentrated loads are present the shaft must be divided into regions with separate equations for each region. Continuity boundary conditions between regions are then imposed. This process presented in detail in Chapter 4 for the axial case applies here as well. 6.4 Solutions from the Theory of Elasticity How do the solutions in Section 6.3 hold up as solutions to the three dimensional equations of elasticity? We note that the torsional stress component, τ , in Equation 6.2.10 can be resolved into two components in rectangular Cartesian coordinates as shown in Figure 6.4.1. y τ τ xy τ zx r z Figure 6.4.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- It follows that τ= T r J → τxy = T z J τzx = T y J (6.4.1) In the absence of body forces these together with the other stresses σx = σ y = σz = τ yz = 0 (6.4.2) exactly satisfy the three dimensional equations of equilibrium, Equations 2.6.13. ⎤ ⎡ 0 ⎤ ⎡ ∂ ∂ ∂ ⎢ 0 ⎥ 0 0 0 ⎥ ⎡ ⎤ ⎢ ⎢ ∂x ∂y ∂z ⎥ 0 ⎥ ⎢ ⎥⎢ 0 ⎥ ⎢ ⎢ ⎥ ∂ ∂ ∂ ⎢T ⎥ ⎣ ⎦ ⎢ 0 0 0 ⎥ ⎢ z⎥= 0 ⎢ ⎥ ⎥ ∂y ∂ x ∂z ⎥⎢ ⎢ 0 ⎢ J ⎥ ⎣ ∂ ∂ ∂ ⎦⎢ 0 ⎥ ⎦ ⎣ 0 0 0 T ∂z ∂y ∂x y J (6.4.3) They also satisfy the conditions of applied load on the boundaries provided that the cross section is circular with the end loads distributed the same as the stresses in the interior. Compatibility is also satisfied. In Section 3.3 it is noted that compatibility requires that certain equations involving the second derivatives of the strain components must be satisfied. In this case the strain components are γxy = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS T τxy = z G GJ γ yz = 0 γzx = T τzx =− y G GJ Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (6.4.4) P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas 226 July 4, 2011 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The second derivatives of linear functions are zero and therefore the conditions of compatibility are satisfied. For the one equation   ∂2 ∂ 2εy ∂ 2 γxy ∂2 ∂2 T ∂ 2 εx → z =0 (6.4.5) + = (0) + (0) = ∂ y2 ∂x2 ∂ x∂ y ∂ y2 ∂x2 ∂ x∂ y GJ Similar results are obtained for the other compatibility equations in three dimensions. If you change the geometry, load distribution, or restraints you no longer have an exact solution with the torsional theory presented here. Non circular sections, except as noted in Section 6.5 for thin wall cross sections, require a more advanced theory. Since the circular section usually is the most efficient it is frequently used in structures where torsion is dominant if other requirements allow. Additional stress components are generated for a circular section if the end loads are anything other than linear with radius, or if the end is restrained in a particular way. In such cases we depend upon St. Venant’s principle to save us from our sins. It can be shown that the effects are localized near the restraint or load application. This leads us to ask just how loads are applied to a shaft and precisely what effect do they have both locally and globally? In Figure 6.4.2 we represent a shaft with the left end fixed and with a distributed load applied tangentially to a portion of the surface at the right end. This provides a pure torque at the right end but with a different distribution of local stress from the linear with radius values of Equation 6.2.10. Figure 6.4.2 A computer based numerical solution using the full three dimensional equations of elasticity was obtained for a specific case. The Unigraphics I-DEAS software is used. The basis for these equations and the validity of this solution are presented in Chapters 11 and 12. Contour plots of the shear stress values are shown in Figure 6.4.3. The stress over most of the length is equal to that obtained by Equation 6.2.10. What is shown on the cylindrical surface is the stress on that surface or τmax . For this particular geometry and loading the maximum shear stress at the surface obtained from Equation 6.2.11 is 0.8 MPa. At the loaded end there are higher stresses near the edge between the loaded surface and the free surface. They reach a maximum of 1.5 MPa in the region shown in red (darkest in gray scale) in Figure 6.4.3. This is another example of stress concentration that occurs at regions of abrupt change in geometry, restraint, or loading. Note that the region of transition is very short, thus confirming St. Venant once --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 227 Torsion Figure 6.4.3 again. At the fixed end it is essentially the same as for the rest of the shaft, that is, we can confirm a linear distribution with radius. Thus we conclude that in this case the restraint does not impose conditions that would modify the results from the simplified theory any significant amount. In practice torsional loadings are often provided by pulleys or gears attached to the shaft so that the shaft acts as an axle. We add a disk to the shaft in Figure 6.4.4 and apply a tangential load as before. The load magnitude is chosen to provide the same total torque as in the previous example. The left end is fixed as before. Figure 6.4.4 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-06 P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas 228 July 4, 2011 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Contour plots of the shear stress values are shown in Figure 6.4.5. The solution over most of the shaft is the same. Once again we get a stress concentration as the diameter changes abruptly where the shaft and disk join. The highest value of shear stress is 1.47 MPa which is only slightly different from the previous case. The color over the main portion is slightly different due to the way the contours are divided between the highest and lowest stress value. The stress values on the shaft are the same. One way to lessen the maximum stresses is to avoid abrupt changes in geometry, loading, or restraint. In Figure 6.4.6 we have added a fillet between the shaft and the disk. Figure 6.4.6 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure 6.4.5 P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 229 Torsion The contour plot is shown in Figure 6.4.7. Figure 6.4.7 The maximum stress is now just 0.904 MPa. This is just a bit over the 0.8 MPa for the shaft. Once again the color changes due to the way the contours are divided. Stress concentrations are discussed in more detail in later chapters. In any case the simplified theory works over most of the length of the shaft and therefore is an efficient way to get a first approximation to the behavior of the shaft. 6.5 Torsional Stress in Thin Walled Cross Sections The equations derived and solved in Section 6.2 apply for hollow circular cross sections as shown in Figure 6.5.1 as well. The shear stress obtained from Equation 6.2.10 has the distribution as shown. The inner radius is ri and the outer radius is ro . τ ro ri Figure 6.5.1 In this case the polar moment of inertia or torsional constant is (See Appendix B)  π 4 r − ri4 (6.5.1) J= 2 o This torsional constant may be used in all the torsional equations for hollow circular sections. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas 230 July 4, 2011 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In cases where the inner radius is only slightly smaller than the outer radius the values of the shear stress are only slightly different along the inner and outer radii. It is common practice to find the shear stress at the mid radius between the inner and outer radii and multiply it by the difference in two radii, or the wall thickness of the cylinder, which we shall designate as b. This is called the shear flow and has the units of force per unit length along the circumference. This is illustrated in Figure 6.5.2 where the mid radius is designated rmid and the shear flow q where q = τmid b (6.5.2) The shear stress is assumed to be constant across the wall thickness. q rmid b Figure 6.5.2 We can write the torsional constant in terms of the mid radius and the wall thickness.      π b 4 b 4 π  4 4 r − ri = rmid + − rmid − J= 2 o 2 2 2 (6.5.3) If we expand the right side and neglect all terms of order b2 and higher, we obtain the following approximation for the torsional constant when b is small compared to rmid . 3 J = 2πrmid b (6.5.4) Noting that q = τb = Trmid b Trmid b T T = = = 2 3 J 2A T 2πrmid 2πrmid b (6.5.5) 2 where A T = πrmid is the area enclosed by the circle of radius rmid . The subscript is added so that this area will not be confused with the area of the cross section material. This can be extended to non circular thin walled closed cross sections. Because there is no shear stress on the inner and outer surfaces of the tube and because the wall is so thin no significant shear stress can exist except tangent to the wall of the tube. Thus we assume a constant shear flow all around the cross section. This assumption is verified by experiment. Consider a tube with the cross section shown in Figure 6.5.3 and a constant wall thickness. ds rmid dAT Figure 6.5.3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 231 Torsion Assuming a constant shear flow the torque is   T = qrmid ds = q rmid ds (6.5.6) where the integral is around the closed boundary. This integral can be replaced by an area integral when we note that the triangular area is dAT = 1 rmid ds 2 (6.5.7) Thus T = 2q  dAT = 2q A T (6.5.8) where A T is the area enclosed by the thin walled tube. This applies to any closed thin walled section. Given the torque at a cross section the shear flow and shear stress are q= T 2A T τ= T 2A T b (6.5.9) ########### Example 6.5.1 Problem: Find the stress in a tube with a rectangular thin walled cross section with dimensions shown in Figure (a) in terms of the torque T. The wall thickness is a constant value b. b a c Figure (a) Solution: Apply Equation 6.5.9. If the wall is quite thin it makes little difference if the outside or if the inside or if the median dimensions are used in defining the enclosed area. q= T T = 2A T 2ac → τ= T 2acb (a) ########### 6.6 Work and Energy—Torsional Stiffness in a Thin Walled Tube Work and strain energy can be used to find displacements at points where concentrated moments are applied. The applied moments move through rotations to provide the external work. The strain energy --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 232 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods in torsion for a circular shaft of length L is UT = 1 2  γ τ dV = 1 2  1 τ2 dV = G 2  1 G  Tr J 2 dV = 1 2  L 0 T2 dx GJ 2  r 2 dA = A 1 2  0 L T2 dx GJ (6.6.1) In Example 6.6.1 we revisit Example 6.2.1. ########### Example 6.6.1 Problem: Find the displacement at the right end. The cross section is circular. ML L --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (a) Solution: Equate the external work to the internal strain energy and solve for the displacement. The external work done by the applied moment is We = 1 M L β(L) 2 (a) Equate this to the strain energy and solve. 1 1 M L β(L) = 2 2  L 0 1 T 2L T2 dx = GJ 2 GJ → β(L) = TL GJ (b) This agrees with the result we obtained in Example 6.3.1, Equation (f). ########### When there are multiple loads on a determinate shaft, including distributed and concentrated loads, for example, as in Figure 6.6.1,we can use Castigliano’s second theorem to find the displacements at the points of concentrated load application. Castigliano’s theorem was first presented in Chapter 4, Section 4.11. M1 tx M2 … Mn L Figure 6.6.1 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 233 Torsion First find the internal torques in terms of the applied moments, that is, find Ti (M1 , M2 , . . . , Mn , tx ) for each region. The total strain energy is then the sum of the strain energies for each region. UT (M1 , M2 , . . . , Mn , tx ) = n  1 Ti2 L 2 GJ i=1 (6.6.2) The displacements at the points of concentrated load application are ∂UT = βi ∂ Mi (6.6.3) ########### Example 6.6.2 Problem: A shaft is fixed at the left end. It is loaded with a concentrated applied moment at x = 400 mm and a distributed moment t over half the length as shown in Figure (a). Find the displacement at the right end where the concentrated moment is applied. The distributed moment is t(x) = −210 N · mm mm t(x) = 0 0 ≤ x ≤ 200 (a) 200 ≤ x ≤ 400 The material is steel with a shear modulus of 80,000 MPa. The polar moment of inertia is J = 20000 mm4 . t = –210 N ⋅ mm / mm l1 = 200 mm M400 = 16000 N ⋅ mm l 3 = 200 mm x Figure (a) Solution: Use Castigliano’s second theorem. We must carry the concentrated moment at the tip as a symbolic M 400 in order to be able to take the derivative. We can enter the numerical value after that. T1 = M400 − 210 · x N · mm T2 = M400 N · mm 0 ≤ x ≤ 200 200 ≤ x ≤ 400 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 234 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods UT = 1 2GJ 1 = 2GJ 1 = 2GJ 1 = 2GJ 1 = 2GJ    200 T12 dx + 0 200  400 200  T22 dx+ (M400 − 210 · x)2 dx + 0  400 (M400 )2 dx 200 200 2 (M400 ) dx − 0  200  (2M400 · 210 · x)dx + 0  200 2 (210 · x) dx + 0  400 2 (M400 ) dx 200    (M400 )2 · 200 − M400 · 210 · 2002 + 70 · 210 · 2003 + (M400 )2 · 200   (M400 )2 · 400 − M400 · 210 · 2002 + 70 · 210 · 2003 (c) The displacement is  ∂UT 1  2 · 400 · M400 − 210 · 2002 = ∂ M400 2GJ 2 · 400 · 16000 − 210 · 2002 = 0.001375 rad = 0.0788◦ = 2 · 80000 · 20000 β (400) = (d) If the displacement is desired at some other point place a symbolic concentrated applied load at that point, find the strain energy, differentiate with respect to that load, and then set the value of the concentrated load to zero. ########### The torsional constant for the thin walled tube of a non circular cross section is not the polar moment of inertia. To use the displacement equations requires a valid torsional constant. We can use the work and energy principles to obtain and effective torsional constant which we shall label Jeff . Suppose the shaft in Example 6.6.1 has a thin walled non circular cross section. The internal torque is  Mx = M L − T = 0 → T = ML (6.6.4) If we define a coordinate s that follows the mid radius of the tube as shown in Figure 6.5.3 the strain energy in the thin walled tube is U= 1 2  V τ γ dV = 1 2  V 1 τ2 dV = G 2G  V T 2L T2 dxdA = 8A2T G (2A T b)2  s ds b (6.6.5) The integral is along the path of the local mid radius. The external work done by the applied moment is We = 1 1 M L φ = M L β(L) 2 2 (6.6.6) where φ is the total angle of twist at the loaded end (see Figure 6.2.1). Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The strain energy is P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 235 Torsion Equating the external work to the strain energy gives us We = T 2L 1 M L β(L) = U = 2 8A2T G  s ds b → TL 4A2T G β(L) =  s ds b (6.6.7) For a circular section we note that ML L GJ (6.6.8) So let us define an effective torsion constant for the non circular section β(L) = TL 4A2T G  s ML L ds = b GJ eff → Jeff = 1 4A2T 1  s 4A2 =  T ds ds s b b (6.6.9) If you use this formula to calculate the Jeff of a thin walled circular section you get  2 2 4 πrmid b 4A2 3 = 2πrmid b=J Jeff =  T = ds 2πrmid s b (6.6.10) This new formula agrees with the value obtained before for the circular section in Equation 6.5.4. This effective torsional stiffness can be used to find the displacement using Equation 6.3.10 GJ eff d 2β = −t(x) dx2 (6.6.11) and to find the internal torque using Equation 6.3.3 GJ eff dβ =T dx (6.6.12) ########### Example 6.6.3 Problem: The shaft in Example 6.6.1 has the thin walled cross section in Example 6.5.1. What is its displacement? b a c Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- β(L) = P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 236 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Solution: Calculate Jeff and insert it in the displacement formula. For this cross section 2(ac)2 b 4A2 4(ac)2 b = Jeff =  T = ds 2a + 2c a+c s b (a) From Example 6.3.1 β(x) = ML ML a + c x x= GJ eff G 2(ac)2 b (b) β(L) = ML ML a + c L L= GJ eff G 2(ac)2 b (c) and the total twist angle is ########### Let us confirm the assumption that the stress is constant in a non circular thin walled tube by a numerical solution that accounts for the local effects of loading and restraint. This is the same analysis method used for the illustrations in Section 6.4 that will be explained in more detail in Chapters 11 and 12. The first example is one with an oval cross section as shown in Figure 6.6.2. Figure 6.6.2 Once again the left end is fixed and a torque is applied to the right end by tangent surface loads. The shear stress distribution is shown in Figure 6.6.3. --`,,,,``,``,,`` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 237 Torsion Figure 6.6.3 The dark color does not show well in gray scale. The point is that the whole surface is one color indicating a constant stress except for a very short region near the load application. Next we solve for a square cross section as shown in Figure 6.6.4. Figure 6.6.4 --`,,,,``,``,,``,,`, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas 238 July 4, 2011 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The stress distribution is shown in Figure 6.5.5. Figure 6.6.5 This shows a small difference at the sharp corners where the stress components make a 90o degree change in direction. On the flat surfaces the stress is constant. Let us round off the corners as shown in Figure 6.6.6 and try again. Figure 6.6.6 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 239 Torsion The stress contours are shown in Figure 6.6.7. Figure 6.6.7 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The stresses are constant except for the locale of the applied forces. 6.7 Torsional Stress and Stiffness in Multicell Sections We can extend our results to the torsion in slender bars with multi celled thin walled cross sections. The main assumption is that all cells of the cross section experience the same amount of twist. Experimental evidence confirms that this is a reasonable assumption. Consider the multi celled section in Figure 6.7.1. q1 cell 1 cell 2 q2 Figure 6.7.1 From Equation 6.5.8 we know that T = 2q A T (6.7.1) T = 2q1 A T 1 + 2q2 A T 2 (6.7.2) therefore, for two cells Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 240 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods From Equations 6.6.12 and 6.6.9 we have GJ eff 4A2 Jeff =  T ds s b dβ =T dx (6.7.3) Combining we obtain 1 dβ (2q1 A T 1 + 2q2 A T 2 ) = dx GJ eff (6.7.4) We also know that q ds Gb (6.7.5) 1 dβ = dx 2GAT 1     ds ds − q2 q1 s12 b s1 b (6.7.6) 1 dβ = dx 2GAT 2   q2  (6.7.7) 1 dβ = dx 2A T  and so for cell 1 and for cell 2 s12 ds − q1 b  s12 ds b where s12 is the length common to both cells. In Equations 6.7.4 and 6.7.6-7 we have three equations in three unknowns, q1 , q2 , and dβ/dx. Let us do an example. ########### Problem: A two celled cross section has the dimensions shown in Figure (a). The wall thickness is a constant value of b. Find the shear flows and the rate of twist for a pure internal torque T. 2a a a q2 q1 cell 1 cell 2 Figure (a) Solution: Solve for the three unknowns using the equations noted above. We can calculate the following:   a 4a ds ds = = A T 2 = 2a 2 AT 1 = a 2 b b b b s12 s1 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  s2 6a ds = b b Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Example 6.7.1 P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 241 Torsion It follows that T = 2q1 A T 1 + 2q2 A T 2 = 2a 2 q1 + 4a 2 q2       1 1 dβ ds a 4a ds = = − q2 − q q q1 2 1 dx 2GAT 1 2Ga 2 b b s12 b s b     1  1 1 dβ ds a ds 6a = − q1 − q = q q2 1 2 dx 2GAT 2 4Ga 2 b b s12 b s12 b (b) (c) (d) Equations (c) and (d) may be rewritten and solved for the shear flows in terms of the rate of twist. dβ dx dβ −q1 + 6q2 = 4Gab dx dβ 16 Gab 23 dx dβ 18 q2 = Gab 23 dx q1 = 4q1 − q2 = 2Gab → The torque can now be written in terms of the rate of twist.     dβ dβ 104 3 dβ 16 18 Gab Gab Ga b + 4a 2 = T = 2a 2 q1 + 4a 2 q2 = 2a 2 23 dx 23 dx 23 dx (e) (f) Since T = GJ eff 104 3 dβ dβ = Ga b dx 23 dx Jeff = → 104 3 a b = 4.52a 3 b 23 (g) T a2 (h) and q1 = 0.154 T a2 q2 = 0.173 The shear flow in the interior web is q2 − q1 = (0.173 − 0.154) T T = 0.019 2 a2 a (i) Let us check to see how the interior web affects the result by removing it and comparing the answers. If the interior web is not present, then  1 1 4 q dβ q q = ds = 8a = dx 2GAT s b 2G(3a 2 ) b 3 abG (j) 2 T = 2A T q = 6a q Therefore dβ dβ 3 dβ = 4.50a 3 bG = GJ eff T = 6a 2 abG 4 dx dx dx (k) Thus Jeff = 4.50a 3 b q = 0.17 T a2 (l) The internal web has a small effect. While the interior web has a small effect on the torsional stress we shall find out later that the effect on shear stress in bending is much greater. ########### --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 242 6.8 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Torsional Stress and Displacement in Thin Walled Open Sections It has been observed that the torsional stress in a narrow rectangular section has a stress distribution that is essentially parallel to the outer surface as illustrated in Figure 6.8.1. Figure 6.8.1 When the rectangle is very narrow and the section is subjected to pure torque the shear stress can be approximated as a linear distribution through the thickness as shown in Figure 6.8.2. The stress does have to stay parallel to the surface at the top and bottom; however, over most of the section the stress can be approximated. The stress at the narrow end is found to be small and can be neglected as far as finding maximum values of stress is concerned. y h z τ xy b Figure 6.8.2 The stress is found to be τxy = T z Jeff (6.8.1) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 243 Torsion where the maximum stress is found at the two surfaces, that is, at z = ±b/2, Without proof the torsional constant for the thin walled section is found to be hb3 (6.8.2) Jeff = 3 The displacement is found from dβ GJ eff =T (6.8.3) dx This can be extended to thin walled cross sections of other shapes. For the cross section in Figure 6.8.3a the effective torsional constant is given by Equation 6.8.2 where h is the length of the arc midway between the inner and outer edges. It can also be applied to multiple sections as, for example, in Figure 6.8.3b. Jeff = h 2 b23 h 1 b13 + 3 3 (6.8.4) h1 h b1 h2 b b2 (a) (b) Figure 6.8.3 ########### Example 6.8.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Problem: A cantilever bar with a cross section as defined in Figure (a) has a pure torque applied at the free end. Find the stress and displacement. s R ϕ b Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 244 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Solution: Use Equations 6.8.1–6.8.3. The torsional constant is π Rb3 3 (a) T 3T δ= δ Jeff π Rb3 (b) Jeff = The torsional stress is τ= where δ is measured normal to the mid radius and has the maximum values of ±b/2. The torsional displacement is GJ eff dβ =T dx → dβ T = dx GJ eff → β(x) = β(x) = T x GJ eff T x +a GJ eff (c) This is a cantilever bar so β(0) = a = 0 → (d) ########### This is a good time to compare the torsional stiffness and maximum stresses in open and closed thin walled sections. In Section 6.5, Equation 6.5.9 we note that the shear stress in a closed section under a pure toque load is q= T 2A T → τ= T 2A T b (6.8.5) The torsion stiffness is given in Section 6.6, Equation 6.6.9, to be Jeff = 1 4A2T 1  s 4A2 =  T ds ds b s b (6.8.6) For a circular thin walled section this reduces to the polar moment of inertia as shown in Equation 6.6.10. 4A2 3 b=J Jeff =  T = 2πrmid ds s b (6.8.7) To compare the stresses in an open section use Equations 6.8.1 and 6.8.2 and note that (τxy )max = 3 T 3T b T z= 3 = Jeff 2 hb2 hb 2 (6.8.8) The torsional rigidity of the open section is given by Equation 6.8.2. Let us compare a specific case of a circular cross section. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 245 Torsion Example 6.8.2 Problem: A circular thin walled shaft in pure torsion with a closed section is compare to one with a slit running the full length producing an open section. rmid b Figure (a) Solution: Compare the stresses and effective torsional stiffness. The maximum stress in the closed section is T T = 2 2A T b b 2πrmid (a) 3 T 3 T = 2 hb2 4 πrmid b2 (b) τc = The maximum stress in the open section is τo = The ratio of the stresses of the open to closed is 3 T τo = τc 4 πrmid b2  3r T = 2 2b 2πrmid b For example if the radius is 10 mm and the wall thickness is 1 mm the stress would be 15 times higher in the open section. The ratio of torsional rigidities is  2πrmid b3 1 b2 Jo 3 2πrmid = (c) b= Jc 3 3 r2 For an open section with the same dimensions the ratio would be 1/300, that is the open section would have a twist displacement of 300 times that of the closed section for the same applied torque. ########### 6.9 A General (Finite Element) Method Let us consider a shaft loaded only with concentrated moments. We divide the shaft into segments and number the nodes and elements just as we did in the axial case. Then label the nodal rotations and internal torques as in Figure 6.9.1. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ########### P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 246 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods element m-1 element m element m+1 x node n-1 node n node n+1 node n+2 s ψ m,n , β n ψ m ,n +1 , β n +1 lm Figure 6.9.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- We shall use {φm } to represent the nodal rotations and {ψm } to represent the internal nodal torques.     ψm,n βn {φm } = (6.9.1) {ψm } = ψm,n+1 βn+1 We define shape functions for element m that present the distributed rotational displacement in terms of the nodal values of rotation. β(sm ) = [n]{φm } (6.9.2) By analogy with the axial case the shape functions for element m are    sm sm βn β(sm ) = [n]{φm } = 1 − lm lm βn+1 Using Equation 6.3.1 the strain in terms of the nodal displacements is    βn dβ 1  −1 1 γm (r ) = r r= βn+1 dsm lm (6.9.3) (6.9.4) and from Hooke’s law the stress is τm (r ) = Gγm = G    βn G dβ −1 1 r r= βn+1 dsm lm For element m, the internal moments at the ends of the elements are     βn GJ  dβ 1 −1 ψm,n = − τm rdA = GJ = βn+1 dsm lm A     βn GJ  dβ −1 1 ψm,n+1 = = τm rdA = GJ βn+1 dsm lm A Equations 6.9.6 can be put in matrix form      G m Jm βn 1 −1 ψm,n {ψm } = = [km ]{φm } = βn+1 ψm,n+1 −1 1 lm (6.9.5) (6.9.6) (6.9.7) and so the element stiffness matrix is [km ] = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS G m Jm lm  1 −1 −1 1  (6.9.8) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 247 Torsion The element stiffness matrices are summed to provide the global stiffness matrix.   {M} = {ψm } = [km ]{φm } = [K ]{φ} (6.9.9) Does this look familiar? What we learned in Chapter 5 applies with only a change in symbols and coefficients. The equivalent nodal loads for element m are found from  lm [n]T tx (s) ds (6.9.10) {Mem } = 0 and the total equivalent nodal load matrix is the sum of the element matrices.  {Me } = {Mem } (6.9.11) These are added to any actual concentrated applied moments. (6.9.12) {M} = {Mc } + {Menl } By analogy with the axial case (compare with Equation 5.4.5) we can summarize the torsional case as follows: {Mc } tx (sm ) →  lm [n]T tx (sm )dsm {Mem } = {M} {ψm } τm (r ) {Me } = 0 γm (r )  β(sm ) {Mem } {φm } ← β(sm ) = [n]{φm } d dβ r= [n]{φm }r dsm dsm d dβ r=G [n]{φm }r τm (r ) = G dsm dsm dβ d = −GJ [n]{φm } ψm,n = −GJ   dsm dsm ψm,n → {ψm } = = [km ]{φm } ψm,n+1 dβ d ψm,n+1 = GJ = GJ [n]{φm } dsm dsm {M} = {Mc } + {Menl } {M} = {ρ} γm (r ) =  {ψm } =  (6.9.13) [km ]{φm } = [K ]{φ} {φm } β(sm ) γm (r ) τm (r ) {ψm } − − − − − nodal rotational displacements distributed rotational displacement shear strain shear stress internal moments and the knowns {ρ} − nodal displacement restraints {Mc } − concentrated moments at nodes {Me } − equivalent nodal moments The assembly process is exactly the same as the one we just learned in Chapter 5. The result is a set of global equations.   [km ]{φm } = [K ]{φ} (6.9.14) {M} = {ψm } = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- In the first row we have the typical unknowns in the formulation of the equations for a single element P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 248 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The boundary restraints are added to the global nodal matrix and the restrain moments to the global moment matrix, the equations are partitioned and solved. ########### Example 6.9.1 Problem: Find the FEM equations for the shaft in Example 6.2.2. Solution: Divide into elements, find the element stiffness matrices, assemble the global matrix equations, add the boundary conditions, and partition. Let us use just two elements as shown in Figure (a). e1 ML / 2 e2 ML L/ 2 L n1 n2 n3 Figure (a) By analogy with the axial case the assembled global equations are ⎡ ⎤ ⎡ ⎤⎡ ⎤ Q1 1 −1 0 0 2GJ ⎣ −1 2 −1 ⎦ ⎣ β2 ⎦ {M} = [K ]{φ} = ⎣ M L/2 ⎦ = L ML β3 0 −1 1 This may be partitioned to find the nodal displacements      2GJ β2 2 −1 M L/2 = ML β3 −1 1 L (a) (b) Generally numerical values are introduced and the equations solved by a linear algebraic equation solver rather than finding an analytical solution. Use Equation 6.9.5 to find the stresses and Equation 6.9.3 to find the distributed displacements. The restraint moment is found from the other partitioned equation.    β2 2GJ  −1 0 (c) Q1 = β3 L ########### At this point it is not so clear why the finite element method is better than the classical one we used. On simple problems it is not. As problems get more complicated it comes into its own because it lends itself to numerical solutions using existing software. The next example problem is complicated enough that it would be more difficult to solve the traditional analytical way. Again, we remind you of the similarity of the “general method” for the axial problem and the “general method” for the torsional problem just presented. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 249 Torsion ########### Example 6.9.2 Problem: A shaft is fixed on both ends, has changes in cross section, as shown in Figure (a) and has three applied concentrated moments, say, at x = L /4, L /2, and 3L /4. Find the displacements and stresses. e1 e2 e3 M2 M3 e4 M4 L n1 n2 n3 n4 n5 Figure (a) Solution: Divide into elements, find the element stiffness matrices, assemble the global matrix equations, add the boundary conditions, and solve. The element and nodal numbering is shown on Figure (a). To do this analytically by the traditional method would require solving the second order equation four times and satisfying eight boundary conditions. Let us set up some conditions. For example, assume that the shaft is divided into four elements of equal length, the value of GJ for elements e1 and e2 is twice that for elements e3 and e4, that is, G 3 J3 = G 4 J4 = The matrices for the four elements are    4G 1 J1 1 ψ1,1 = ψ1,2 −1 L    4G 2 J2 1 ψ2,2 = ψ2,3 −1 L    4G 3 J3 1 ψ3,3 = ψ3,4 −1 L    2G 4 J4 1 ψ4,4 = ψ4,5 −1 L G 1 J1 2 −1 1 −1 1 −1 1 −1 1     β1 β2 β2 β3 β3 β4 β4 β5 and     = G 2 J2 = G 1 J1 4G 1 J1 L 2G 1 J1 = L = 2G 1 J1 L    1 −1 −1 1 1 −1 −1 1 1 −1 −1 1 (a)    β2 β3 β3 β4 β4 β5   (b)  These must be imbedded in the global format and summed. The assembled equations are ⎡ ⎤ ⎡ ⎤⎡ ⎤ 0 2 −2 0 0 0 Q1 ⎢ ⎥ ⎢ ⎢ M2 ⎥ 4 −2 0 0⎥ ⎥ 2G 1 J1 ⎢ −2 ⎢ ⎥ ⎢ β2 ⎥ ⎢ ⎥ ⎢ 0 −2 ⎢ M3 ⎥ = 3 −1 0⎥ ⎥ ⎢ ⎥ ⎢ β3 ⎥ L ⎢ ⎣ 0 ⎣ M4 ⎦ 0 −1 2 −1 ⎦ ⎣ β4 ⎦ Q5 0 0 0 0 −1 1 (c) After partitioning ⎡ ⎤ ⎡ M2 4 −2 2G 1 J1 ⎣ M3 ⎦ = ⎣ −2 3 L M4 0 −1 ⎤⎡ ⎤ 0 β2 −1 ⎦ ⎣ β3 ⎦ β4 2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 250 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods and  Q1 Q5  2G 1 J1 = L  ⎡ ⎤  β −2 0 0 ⎣ 2⎦ β3 0 0 −1 β4 (e) Let us introduce some numerical values. Consider a steel shaft with G = 75,000 MPa. Its diameter at the fixed end is 20 mm; its length is 500 mm. It follows that π 4 π r = 104 = 15708 mm 4 = J2 2 1 2 π J1 = 7854 mm 4 = r34 = J4 → J3 = 2 2 J1 = (f) r3 = 8.409 mm = r4 Case 1 Case 2 M2 = 50000 M2 = 10000 M3 = 34000 M3 = −20000 M4 = 25679 M4 = 30000 (g) Case 1: Substituting the values for Case 1 into Equation (c) we have ⎡ ⎤⎡ ⎤ ⎤ 4 −2 0 β2 50000 2 · 75000 · 15708 ⎣ −2 ⎣ 34000 ⎦ = 3 −1 ⎦ ⎣ β3 ⎦ 500 β4 0 −1 2 25679 ⎡ (h) Using an equation solver for Case 1 we get ⎤ ⎡ ⎤ ⎡ ⎤ 7.734 · 10−3 0.443 β2 −3 ⎣ β3 ⎦ = ⎣ 10.136 · 10 ⎦ rad = ⎣ 0.582 ⎦ deg β4 7.806 · 10−3 0.447 ⎡ (i) For Case 1 the maximum stress in each element, that is, the stress at the surface, is 4G  −1 L 4G  −1 τ2 = L 4G  −1 τ3 = L 4G  −1 τ4 = L τ1 =    β1 = β2    β2 1 = β3    β3 1 = β4    β4 1 = β5 1 4 · 75000  −1 500 4 · 75000  −1 500 4 · 75000  −1 500 4 · 75000  −1 500  0 10 = 46.404 MPa 0.007734    0.007734 1 10 = 14.574 MPa 0.010163    0.010163 1 8.409 = −11.892 MPa 0.007806    0.007806 1 8.409 = −39.3844 MPa 0 1   (j) Case 2: Substituting the values for Case 2 into Equation (c) we have ⎡ ⎡ ⎤ ⎤⎡ ⎤ 10000 4 −2 0 β2 2 · 75000 · 15708 ⎣ −20000 ⎦ = ⎣ −2 3 −1 ⎦ ⎣ β3 ⎦ 500 β4 30000 0 −1 2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (k) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Consider two cases of applied moments (N . mm). P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 251 Torsion Using an equation solver for Case 2 we get ⎤ ⎡ ⎡ ⎤ ⎡ ⎤ 5.30515 · 10−4 0.0304 β2 ⎦ rad = ⎣ 0 ⎦ deg ⎣ β3 ⎦ = ⎣ 0 β4 3.18309 · 10−3 0.1824 (l) The stresses may be found using the same equations as in Equation (i). Note that once the main stiffness matrix is assembled a variety of loading conditions can be solved using the one entry for the stiffness matrix. Other restraints can also be used but the global matrix must then be repartitioned. Example 6.9.3 Problem: A shaft is fixed at both ends as shown in Figure (a). The three sections each have a different diameter. The center section has a uniform distributed applied moment applied over all its length. The material is the same in all parts. Write down the equations for finding the displacements and stresses. t(x) = t0 l1 = 200 l2 = 100 x l3 = 200 Figure (a) Solution: Divide into elements, find the element stiffness matrices, assemble the global matrix equations, add the boundary conditions, and partition. Three elements will do. e1 n1 e2 n2 e3 n3 n4 Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The equivalent nodal loads on element 2 are ⎡ ⎤ t0 100     ⎢ 2 ⎥ M2 ⎥ = 50t0 =⎢ ⎣ t 100 ⎦ 50t0 M3 0 2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 252 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The assembled equations are ⎡ GJ 1 ⎢ 200 ⎡ ⎤ ⎢ ⎢ GJ 1 F1 ⎢− ⎢ F2 ⎥ ⎢ ⎢ 200 ⎢ ⎥ {F} = [K {r }] = ⎣ ⎦ = ⎢ F3 ⎢ ⎢ 0 F4 ⎢ ⎢ ⎣ 0 GJ 1 200 GJ 2 GJ 1 + 200 100 GJ 2 − 100 0 − 0 GJ 2 − 100 GJ 3 GJ 2 + 100 200 GJ 3 − 200 0 ⎤ ⎥ ⎥⎡ ⎤ ⎥ β1 0 ⎥ ⎥⎢ ⎥ ⎥ ⎢ β2 ⎥ ⎥⎣ ⎦ β3 GJ 3 ⎥ ⎥ − ⎥ 200 ⎥ β4 ⎦ GJ 3 200 (b) When boundary conditions, restraints, and loads are inserted ⎡ GJ 1 ⎢ 200 ⎤ ⎢ ⎡ ⎢ GJ 1 Q1 ⎢− ⎢ 50t0 ⎥ ⎢ ⎥ ⎢ 200 ⎢ ⎣ 50t0 ⎦ = ⎢ ⎢ ⎢ 0 Q4 ⎢ ⎢ ⎣ 0 GJ 1 200 GJ 2 GJ 1 + 200 100 GJ 2 − 100 0 − 0 ⎥ ⎥⎡ ⎤ ⎥ 0 0 ⎥ ⎥⎢ ⎥ ⎥ ⎢ β2 ⎥ ⎥⎣ ⎦ β3 GJ 3 ⎥ ⎥ − 0 200 ⎥ ⎥ ⎦ GJ 3 200 GJ 2 100 GJ 3 GJ 2 + 100 200 GJ 3 − 200 − 0 ⎤ (c) Partition the equations for finding the displacements.  50t0 50t0  GJ 2 GJ 1 + ⎢ 200 100 =⎢ ⎣ GJ 2 − 100 ⎤ GJ 2   ⎥ β2 100 ⎥ GJ 3 ⎦ β3 GJ 2 + 100 200 ⎡ − (d) The equations for the restraint forces are  Q1 Q4  ⎡ ⎢ =⎢ ⎣ − GJ 1 200 ⎤ 0   ⎥ β2 ⎥ GJ 3 ⎦ β3 − 200 0 (e) Finally the internal torques are found from these equations    ψ1,1 ψ1,2 ψ2,2 ψ2,3 ψ3,3 ψ3,4    = GJ 1 = GJ 2 = GJ 3    1 −1 −1 1 1 −1 −1 1 1 −1 −1 1    0 β2 β2 β3 β3 0   (f)  And the stresses are found from τ1 (x) = ψ1 r J1 τ2 (x) = ψ2 r J2 τ3 (x) = ψ3 r J3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (g) P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 253 Torsion Example 6.9.4 Problem: A shaft is fixed at the left end. It is loaded with concentrated applied moments at x = 200 mm and 300 mm as shown in Figure (a). In addition it is subjected to an applied rotation, β L = 0.002 rad, at the right end. Write down the appropriate equations and boundary conditions for finding the displacements, internal torques, and stresses. The material is steel with a shear modulus of 80,000 MPa. The polar moments of inertia are J1 = 20000 mm 4 = J3 and J2 = 40000 mm 4 . M200 = 800 N-mm M300 = 4200 N-mm l1 = 200 mm β L = 0.002 rad l3 = 200 mm x Figure (a) Solution: Divide into elements, find the element stiffness matrices, assemble the global matrix equations, partition, and solve. The values for GJ are GJ 1 = 16 · 108 = GJ 3 GJ 2 = 32 · 108 (a) Three elements will do. e1 e2 n1 n2 e3 n3 n4 Figure (b) The assembled equations are ⎡ GJ 1 ⎢ 200 ⎡ ⎤ ⎢ ⎢ GJ 1 M1 ⎢− ⎢ M2 ⎥ ⎢ ⎢ 200 ⎢ ⎥ =⎢ {M} = [K {φ}] = ⎣ M3 ⎦ ⎢ ⎢ 0 M4 ⎢ ⎢ ⎣ 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS GJ 1 200 GJ 2 GJ 1 + 200 100 GJ 2 − 100 − 0 0 GJ 2 − 100 GJ 3 GJ 2 + 100 200 GJ 3 − 200 0 ⎤ ⎥ ⎥⎡ ⎤ ⎥ β1 0 ⎥ ⎥⎢ ⎥ ⎥ ⎢ β2 ⎥ ⎥⎣ ⎦ β3 GJ 3 ⎥ ⎥ − ⎥ 200 ⎥ β4 ⎦ GJ 3 200 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- l2 = 100 mm P2: ABC JWST071-Waas July 4, 2011 254 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods When displacement boundary conditions, restraint forces, GJ values, and loads are inserted ⎡ ⎤ 1 1 0 0 − ⎢ 200 ⎥ 200 ⎢ ⎥ ⎢ ⎥⎡ ⎤ ⎡ ⎤ ⎢ ⎥ Q1 2 2 1 1 0 ⎢− + − 0 ⎥ ⎢ ⎥⎢ ⎥ ⎢ 200 100 100 ⎢ 200 ⎥ ⎢ β2 ⎥ ⎢ 800 ⎥ ⎥ ⎥⎣ ⎥ = 16 · 108 ⎢ ⎢ ⎢ ⎥ β3 ⎦ ⎣ −4200 ⎦ 1 1 2 2 ⎢ 0 ⎥ + − − ⎢ 0.02 Q4 100 100 200 200 ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ 1 1 ⎦ 0 0 − 200 200 (c) Partition the equations for find the displacements. This time the fourth column is multiplied by a known value and so must be included in the partition. ⎤ ⎡     β 16 · 108 800 5 −4 0 ⎣ 2 ⎦ β3 = (d) −4200 −4 5 −1 200 0.02 The equations for the restraint forces are  Q1 Q4  16 · 108 = 200  −1 0 0 −1 ⎤ ⎡  β 0 ⎣ 2 ⎦ β3 1 0.02 (e) Finally the internal torques are found from these equations      0 1 −1 ψ1,1 = GJ 1 ψ1,2 β2 −1 1      β2 ψ2,2 1 −1 = GJ 2 β3 ψ2,3 −1 1      ψ3,3 β3 1 −1 = GJ 3 0.02 ψ3,4 −1 1 (f) And the stresses are found from τ1 (x) = ψ1 r J1 τ2 (x) = ψ2 r J2 τ3 (x) = ψ3 r J3 (g) ########### 6.10 Continuously Variable Cross Sections If J varies with x we can adapt the equations as we demonstrated for the axial case. If the shaft is statically determinate we find the internal torque and from it the stress. τ (x, r ) = T (x) r J (x) (6.10.1) Finding the displacement requires integrating T (x) dβ = dx GJ(x) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS → β(x) = 1 G  T (x) dx + c J (x) (6.10.2) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-06 P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 7:2 Printer Name: Yet to Come 255 Torsion or dβ d GJ(x) = −tx (x) dx dx → β(x) = −  1 GJ(x)   tx (x)dx + a dx + b (6.10.3) If the shaft is statically indeterminate we must use the second order equations. Thus dβ d GJ(x) = −tx (x) dx dx (6.10.4) Once we find the displacement β(x) by integrating twice and applying constraints to find the constants of integration we can find the internal torque. dβ dx (6.10.5) T (x) r J (x) (6.10.6) T (x) = GJ(x) and the stress is τ (x, r ) = We learned in the axial case that except for the simplest functions representing J (x) direct integration is an unwieldy process. It is more common to approximate the variation by a series of constant cross sections, such as is represented in Figure 6.10.1. In practice several sections might be used, not just the three shown. The distributed load would be replaced by equivalent nodal loads and the general method used. tx(x) L {Me} Figure 6.10.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- In Section 5.5 we learned how careful interpretation of the results can achieve a high degree of accuracy. We would use the restraint torques at the ends to find the stresses there and use average values at the nodes. 6.11 Summary and Conclusions Since the equations for torsion differ from the axial case only in the symbols used for the coefficients everything we learned in the axial case applies here. For this reason we do not do a lot of examples. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-06 P2: ABC JWST071-Waas July 4, 2011 256 7:2 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The equations for the classical analytical solution of torsional problems for slender shafts with circular or thin walled closed cross sections are summarized in Equations 6.3.8 and are repeated here as Equation 6.11.1. Mc tx (x) T (x) dT = −tx (x) dx → τ (x, r ) γ (x, r ) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- β(x) dβ r γ (x, r ) = dx dβ τ (x, r ) = Gγ = G r dx    T dβ dβ 2 → τ= r T (x) = τ rdA =G γ rdA = G r dA = GJ dx dx J A ← ρ (6.11.1) A A dT d dβ = GJ = −t(x) dx dx dx The equations for the general method (FEM) are summarized in Equations 6.9.2 and are repeated here as Equation 6.11.2. {Mc } tx (sm ) →  lm [n]T tx (sm )dsm {Mem } = {M} {ψm } τm (r ) {Me } = 0  γm (r ) {Mem } β(sm ) {φm } ← β(sm ) = [n]{φm } d dβ γm (r ) = r= [n]{φm }r {M} = {Mc } + {Menl } dsm dsm dβ d r=G [n]{φm }r τm (r ) = G dsm dsm d dβ = −GJ [n]{φm } ψm,n = −GJ   dsm dsm ψm,n → {ψm } = = [km ]{φm } ψm,n+1 dβ d ψm,n+1 = GJ = GJ [n]{φm } dsm dsm {M} =  {ψm } =  {ρ} (6.11.2) [km ]{φm } = [K ]{φ} We must emphasize that while the axial findings are true for any cross section shape the findings for torsion apply only to circular or thin walled cross sections. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 7 Classical Analysis of the Bending of Beams 7.1 Introduction We now consider slender bars with transverse loads and moments. A slender member that can carry such loads is referred to as a beam. If in addition the member also has axial loads that are compressive we call it a beam column. The effect of axial loading on a beam column is treated in detail in Chapter 13. Just as in the previous cases we have included in earlier chapters we place the x axis at the centroid of the cross section but we must be careful with the orientation of the y and z axes. To simplify the problem we consider that the y and z axes are principal axes of inertia of the cross section area. Just what this means is explained in Section 7.2. We must also deal with sign conventions that change according to the type of analysis. The so-called traditional sign convention, or the “elasticity” sign convention, that is used throughout this chapter is defined in the next section. 7.2 Area Properties—Sign Conventions 7.2.1 Area Properties Consider a cross section with a general shape such as is shown in Figure 7.2.1 with the x axis normal to the cross section. y z Figure 7.2.1 Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas 258 July 4, 2011 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The x axis is a centroidal axis if   ydA = 0 A zdA = 0 (7.2.1) A The area moments of inertia are Izz =  y 2 dA I yy = A  z 2 dA (7.2.2) A and the area product of inertia is Iyz =  (7.2.3) yzdA A y z h b Figure 7.2.2 The centroid of this section is easily determined by symmetry and its moments and product of inertia are h/2   h/2 b3 h by 3  bh 3 I yy = Iyz = 0 (7.2.4) y 2 dA = b y 2 dy = = Izz =  3 −h/2 12 12 −h/2 A This and several other cross sections and their properties are found and illustrated in Appendix B. The extension to more general cross section shapes, including cases where Iyz = 0, is presented in Chapter 10. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- In many cases Iyz = 0. This occurs when either the xy or the xz axes plane is a plane of symmetry. For cross sections without symmetry it is possible to orient the yz axes so that Iyz = 0. When Iyz = 0 the axes are called principal axes of inertia. While it is possible to derive the beam equations for any orientation of the axes, that is, when Iyz may not be zero, we shall limit discussion in this chapter to beams with the x axis centroidal and the yz axes aligned as principal axes. In such cases the equations governing displacement in the two principal axis directions are uncoupled. For the most part we shall derive equations and give examples only for applied forces in the y direction and applied moments about the z axis. Very similar equations are used in the xz plane. Since the equations in the two planes are uncoupled they can be solved independently. Information on finding the centroids and the moments and product of inertia of cross section areas, including finding the orientation of the yz axes so that Iyz = 0, may be found in Appendix B. In many of our examples we shall use a rectangular cross section, such as that illustrated in Figure 7.2.2, for illustrating the distribution of the shear and normal stresses. P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 259 Classical Analysis of the Bending of Beams 7.2.2 Sign Conventions Sign conventions can be confusing and the different standards used in different books and publications do not help; however, nearly everyone uses the same conventions for stress and strain. For a reminder, positive values for stress are shown in Figure 7.2.3 in two dimensions. (See Chapter 1, Section 1.7.) σy τxy σx σx τxy σy Figure 7.2.3 The sign convention for positive internal forces and moments in beams is usually as shown in Figure 7.2.4 in two dimensions in the xy plane when explaining traditional analytical methods. M(x) is the internal moment about the z axis and V(x) is the internal shear force in the y direction. The direction of the normal to the face of interest determines the positive direction. y V(x) M(x) M(x) x V(x) Figure 7.2.4 The shear force V(x) and the moment M(x) are stress resultants, that is, V (x) =  τxy dA M(x) = − A  yσx d A (7.2.5) A The minus sign in the moment definition is needed because of the convention that a positive stress at a positive y produces a negative moment. At this time we do not yet know just how the stress is distributed on the cross section. That will be coming right up. The sign convention for applied loads usually is to have all applied forces as positive in the positive direction of the axes and all applied moments as positive according to the right hand rule when a right hand rectangular Cartesian coordinate system is used. The same convention applies for forces and moments at the restraints. Note that while the right hand rules apply to internal moments on a face whose normal is in the positive x direction it does not apply to internal moments on a face whose normal is in the negative x direction. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 260 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 7.3 Derivation and Solution of the Governing Equations First consider a slender bar acted upon by pure moments at each end as shown in Figure 7.3.1. For convenience the origin of the coordinates is placed at the mid point or centroid of the volume of the beam. y Mz Mz x Figure 7.3.1 It is observed in experiments that, as the beam deforms laterally in the y direction an amount v(x), plane cross sections normal to the initially straight centroidal axis remain essentially plane and normal to the deformed shape of the curve passing through the centroids of the cross sections. This is illustrated in Figure 7.3.2. y x Figure 7.3.2 As the plane cross section rotates any point on it may also undergo a displacement u in the x direction. Consider a segment of the beam as shown in Figure 7.3.3 in both the undeformed and deformed position. The displacement is greatly exaggerated. dυ dx y y B dυ dx υ u A y x Figure 7.3.3 We are particularly interested in the movement of a point on the beam from its undeformed position A to its deformed position B. For small deflections it can be assumed that the plane cross section moves up an amount v and rotates an amount dv/dx. For a slender beam the Poisson’s ratio effect can be ignored and so the point remains a distance y from the centroid of the cross section. This assumes that all points on a cross section at a given value of x displace the same amount in the y direction. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 261 Classical Analysis of the Bending of Beams For small displacements there is a relation between the u and v displacements as expressed in Equation 7.3.1 u(x, y) = −y dv (x) dx (7.3.1) While u is a function of both x and y, the displacement v is a function of x only. From the strain displacement equation the normal strain on the cross section is then   dv d 2 v(x) d du(x, y) −y = −y = εx (x, y) = dx dx dx dx2 (7.3.2) and from Hooke’s law the stress is σx (x, y) = Eεx (x, y) = −y E d 2 v(x) dx2 The moment on an internal surface is   d 2v d 2 v(x) M(x) = − σx yd A = E 2 y 2 d A = EIzz dx dx2 A (7.3.3) (7.3.4) A In cases where the moment can be found from static equilibrium, that is, the beam is statically determinate, you can find the normal stress immediately by combining Equations 7.3.3 and 7.3.4. σx (x, y) = − M(x)y Izz (7.3.5) You integrate Equation 7.3.4 twice to find the displacement. When the beam is not statically determinate we must find some other way to satisfy equilibrium. Consider a slender bar that is restrained to remove rigid body motion and is acted upon by a lateral distributed load as shown in Figure 7.3.4. We shall restrict the case to actions taking place in the xy plane and assume that the yz axes are principal axes of inertia. In Figure 7.3.4 it appears that we have applied the lateral load to the upper surface. Actually it is assumed to be applied along the centroidal axes. Eventually, we shall show that it has little effect whether applied to the upper or lower surface, the centroid axis, or as a body force. fy(x) Figure 7.3.4 If lateral loads are present there must be both internal shear forces and stresses and internal moments and normal stresses. Consider a thin slice of length dx of this beam, as shown in Figure 7.3.5, in the initial undeformed configuration. Later when we treat buckling problems due to destabilizing axial loads we need to consider the forces on the thin slice in the deformed configuration. For now, since the deformations are small writing the equilibrium equations in the undeformed configuration incurs no significant error. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 262 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods fy(x) y V ( x + dx ) M(x) x V(x) M (x + dx) dx Figure 7.3.5 Using a Taylor’s series expansion and neglecting higher order terms we recognize that dM dx dx dV V (x + d x) = V (x) + dx dx M(x + d x) = M(x) + + higher order terms (7.3.6) + higher order terms Summation of forces in the y direction and summation of moments about the z axis produce these two equilibrium equations. dV dV d x − V (x) + f y (x)d x = 0 → = − f y (x) dx dx  dM dM dx d x − M(x) + V d x − f y d x =0 → = −V (x) Mz = M(x) + dx 2 dx  Fy = V (x) + (7.3.7) Since we are still assuming plane sections remain plane we can continue to use Equation 7.3.4 and combine it with Equations 7.3.7. d d 2v dM = EIzz 2 = −V (x) dx dx dx − d2 M dV d2 d 2v = = EI = f y (x) zz dx dx2 dx2 dx2 (7.3.8) (7.3.9) And, of course, if EIzz is constant --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- EIzz d 3v = −V (x) dx3 EIzz d 4v = f y (x) dx4 (7.3.10) Can we still accurately assume plane sections remain plane? Yes and no. The shear force does cause a shear deformation that warps the cross section; however, if the beam is slender enough that effect is negligible. At some value, as the beam is less and less slender, the shear deformation becomes important and the assumption of plane sections no longer holds. For now we shall proceed with slender beams for which we can neglect shear deformation, but that does not mean we can neglect the shear stress. So far we have equations for finding the displacement, internal moment and normal stress, and internal shear force, but not the shear stress. To find the shear stress consider a slice of the beam and the normal stresses acting on it as shown in Figure 7.3.6a. For the time being consider this to be a rectangular cross section. In Figure 7.3.6b we show a portion of the section above the coordinate value of y. The unbalanced normal force caused by the normal stresses in the x direction must be balanced by the horizontal shear force resulting in the shear stress τ xy . Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 263 Classical Analysis of the Bending of Beams σx+ σx ∂σ x dx ∂x σx+ σx τ xy ∂σ x dx ∂x y x dx --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- (a) (b) Figure 7.3.6 If we assume that the shear stress is constant across the width, b, of the cross section we get  Fx =  h 2 y    h 2 ∂σx d x bdy − σx + σx bdy − τxy bd x = 0 ∂x y (7.3.11) or  h 2 y ∂σx d xbdy − τxy bd x = 0 ∂x → τxy =  h 2 y ∂σx dy ∂x (7.3.12) The integration is carried out over the cross section area above the coordinate y. This area is defined as (7.3.13) d A y = bdy and is given the subscript y to identify it from the total cross sectional area. When Equations 7.3.5 and 7.3.7 are noted we get  1 ∂σx dM y d Ay = − d Ay ∂ x b Ay A y dx Izz   1 dM V (x) =− ydA y = ydA y Izz b dx A y Izz b A y τxy (x, y) = 1 b  (7.3.14) For a rectangular cross section of height h and width b we get τxy (x, y) = V (x) Izz b  ydA y = Ay V (x) Izz b  h 2 ybdy = y V (x) 2Izz  h2 − y2 4  (7.3.15) For a rectangular section the magnitude of the shear stress is parabolic as illustrated in Figure 7.3.7. The maximum value is at y = 0. τmax (x, 0) = V (x)h 2 3V (x) = 8Izz 2bh (7.3.16) The direction of the shear stress is parallel to the surface. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 264 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 3V(x) 2bh Figure 7.3.7 This formula is reasonably accurate for beams with rectangular cross sections; however, is it not adequate for cross sections that have abrupt changes in width. In such cases the more advanced theory of elasticity may necessary. We shall consider shear in other cross section shapes in Section 7.9 and in Chapter 10. In particular we can adapt beam theory to thin walled cross sections. To summarize, we have listed the quantities and equations used to solve the beam bending problem. Mc , Fc → f y (x) dV = − fy dx V (x) M (x) σx (x, y) εx (x, y) dM = −V (x) dx u (x, y) v(x) ← ρ u(x, y) = −y d 2v du = −y 2 dx dx d 2v σx (x, y) = Eεx = −E y 2 dx  2 2  (x) y M d v d v y 2 d A = EIzz 2 → σx = − M(x) = − σx yd A =E 2 dx dx Izz A A  d V (x) d 2v dM = EIzz 2 = −V (x) → τxy (x, y) = ydA y dx dx dx Izz b A y 2 2 2 d d v d M = EIzz 2 = f y (x) dx2 dx2 dx dv dx εx (x, y) = (7.3.17) Listed in the first row are the unknown quantities in a typical problem v(x) – distributed displacement in the y direction u(x, y) – displacement in the x direction εx (x, y) – normal strain σx (x, y) – normal stress M(x) – internal moment V (x) – internal shear force and also the known quantities f y (x) – distributed applied loads – force per unit length Mc , Fc – concentrated applied moments and forces ρ– displacement restraints --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 265 Classical Analysis of the Bending of Beams The arrows are there to indicate that Mc , Fc , and ρ are boundary conditions to be satisfied in the process of solving for constants of integration. Note that the shear displacement, strain, and stress are not used to derive the displacement equations and therefore do not appear in the first row. Shear displacement is neglected and the shear stress is derived from the normal stress by an equilibrium argument as noted above. In the second row we have the equilibrium equations on the left and the plane sections remain plane relation on the right. dM = −V (x) − static equilibrium of moments dx dV = − f y (x) − static equilibrium of forces dx dv − plane sections remain plane assumption u(x, y) = −y dx In the third row we have the strain displacement relation and in the fourth row Hooke’s law. du − strain displacement dx σx (x, y) = Eεx (x, y) − Hooke’s law for material properties εx (x, y) = In row five we define the moment in terms of displacement by combining the above terms and then show the stress in terms of the moment. M(x) = −  σx ydA = E d 2v dx2 A  y 2 dA = EIzz d 2v dx2 → σx (x, y) = − M (x) y Izz (7.3.18) A In row six we define the shear force in terms of the displacement and the shear stress in terms of the shear force. d 2v d EIzz 2 = −V (x) dx dx → Z τxy (x, y) = V (x) Izz b  ydA (7.3.19) Ay Finally, in the last row we have the equilibrium equation relating the displacement and the applied load. d 2v d2 EIzz 2 = f y (x) 2 dx dx (7.3.20) Of course, when EIzz is constant we get EIzz d 3v = −V (x) dx3 EIzz d 4v = f y (x) dx4 (7.3.21) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 266 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The solution to Equations 7.3.20 and 7.3.21 involve straight forward integration.For Equations 7.3.21 we have f y (x) d 4v = dx4 EIzz  1 d 3v f y (x)d x + a = → dx3 EIzz   2 1 d v (7.3.22) f y (x)d x + ax + b = → dx2 EIzz   1 dv a = f y (x)dx + x 2 + bx + c → dx EIzz     2 1 a b f x (x)d x+ x 3 + x 2 + cx + d → v(x) = EIzz 6 2 where a, b, c, and d are the constants of integration. The four boundary conditions needed to find the constants of integration are from the restraints and/or applied loads at the ends of the beam segments, that is, some combination of the values of the displacement, slope, moment, or shear force at the ends of the beam segments. Equations 7.3.22 apply to both statically determinate and indeterminate beams. In the determinate case you can find the internal moment directly from the equations of static equilibrium and use the second order equation from Equation 7.3.4 to find the displacement. EIzz d 2v = M (x) dx2 (7.3.23) In this case the solution is M d 2v = 2 dx EIzz → →  1 dv = M(x)d x + e dx EIzz   1 M(x)d x + ex + f v(x) = EIzz (7.3.24) where e and f are the constants of integration. The constants of integration are found by applying the boundary conditions on displacement and slope. With all these equations swirling before your eyes it is time for some examples to help clarify what is going on. ########### Example 7.3.1 Problem: Find the stresses and displacements in a beam loaded by pure concentrated moments of equal value at the ends as shown in Figure 7.3.1 and reproduced here in Figure (a). The coordinate system origin is placed at the midpoint as shown. M0 y M0 x Figure (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 267 Classical Analysis of the Bending of Beams Solution: Since this beam is statically determinate we can find the internal moment and shear force, find the stresses, and then use the second order equation for the displacement. From static equilibrium the internal moment has a constant value  Mz = M0 − M(x) = 0 → (a) M(x) = M0 and so the stress is a constant in x and linear in y, that is σx (x, y) = − M(x)y M0 y =− Izz Izz (b) There is no shear force and therefore no shear stresses and so EIzz dM d 3v = −V (x) = 0 = dx3 dx (c) We can use the second order equation to find the displacement. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- EIzz d 2v = M0 dx2 → M0 x dv = +e dx EIzz → v(x) = M0 x 2 + ex + f 2EIzz (d) There are no physical restraints; however, the beam is in static equilibrium so we can assign displacements to orient the beam with respect to the coordinate system. Let us attach the coordinate system at the origin so that dv(0) =0 dx v(0) = 0 (e) By inserting the displacement solution into these boundary conditions we obtain v(0) = f = 0 dv(0) =e=0 dx → (f) The deflection is v(x) = M0 x 2 2EIzz (g) The plot of the displacement is given in Figure (b). y M 0 L2 8 EIzz − L 2 L 2 Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 268 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Example 7.3.2 Problem: Find the stresses and the displacement of a cantilever beam fixed at the left end and with a single concentrated load at the right end as shown in Figure (a). The load is applied on the centroidal x axis. The cross section is rectangular with height h and width b. F y x L Figure (a) Solution: Let us first solve this using the fourth order equation. Then we shall repeat the solution using the second order equation. Using the fourth order equation: Since there is no distributed applied load we have --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- EIzz d 4v d 3v d 2v d 4v = f y (x) = 0 → =0 → =a → = ax + b 4 4 3 dx dx dx dx2 a a b dv = x 2 + bx + c → v(x) = x 3 + x 2 + cx + d → dx 2 6 2 (a) The boundary conditions are dv (0) = 0 dx v(0) = 0 EIzz d 3v (L) = −F dx3 EIzz d 2v (L) = 0 dx2 (b) Thus v(0) = d = 0 EIzz dv (0) = c = 0 dx d 2v (L) = EIzz (a L + b) = 0 dx2 → d 3v (L) = EIzz a = −F dx3 FL b= EIzz EIzz → a=− F EIzz (c) The displacement is v(x) = FL 2 F F 3 a 3 b 2 x + x + cx + d = − x + x +0·x +0= 6 2 6EIzz 2EIzz EIzz  x3 Lx2 − 2 6  (d) We differentiate the displacement to find the bending moment and shear force.   x2 dv = F Lx − dx 2 d 2v EIzz 2 = F (L − x) = M → M = F (L − x) dx d 3v EIzz 3 = −F = −V → V = F dx EIzz Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (e) P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 269 Classical Analysis of the Bending of Beams From this we can obtain the stresses. σx (x, y) = − F(L − x) M y=− y Izz Izz τxy (x, y) = F Izz b  ydA y (f) Ay If the cross section is rectangular with a height h and a width b the shear stress is F τxy (x, y) = Izz b  Ay F ydA y = Izz b  h 2 y F ybdy = 2Izz  h2 − y2 4  (g) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- It is a common practice to plot the shear force and bending moment to have a visual picture of their distribution. This makes it easy to spot where these quantities are the largest and hence where the stresses are the maximum. In this case it is very simple as shown in Figure (b). F Shear, V(x) x 0 L FL Moment, M(x) x 0 L Figure (b) Clearly the maximum normal stress is at x = 0 and at the largest value of y on the cross section. If the cross section is rectangular of height h and width b then the maximum value is   FL h h =± σmax = σx 0, ± 2 Izz 2 (h) The shear force is a constant for all values of x. The maximum shear stress occurs at y = 0. τmax = τxy (x, 0) = 3F F h2 = 2Izz 4 2bh (i) Using the second order equation: For an alternative solution first recognize that this is a statically determinate beam. The internal moment and the shear force can be obtained directly from a free body diagram. F V(x) M ( x) x L Figure (c) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 270 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods From static equilibrium  and the stresses are  Mz = F(L − x) − M(x) = 0 Fy = F − V (x) = 0 σx (x, y) = − → M(x) = F(L − x) (j) V (x) = F → M F(L − x) y=− y Izz Izz τxy (x, y) = F Izz b  (k) ydA y Ay The normal stress and shear stress are found directly as in Equation (f). Since we can determine the internal moment directly from static equilibrium the second order equation suffices for finding the displacement. M(x) F d 2v = = (L − x) 2 dx EIzz EIzz (l) Integrate twice F dv = dx EIzz  Lx − x2 2  +e → v(x) = F EIzz  x3 Lx2 − 2 6  (m) + ex + f where e and f are constants of integration. Apply the boundary conditions and obtain the solution. When the second order equation is used the two boundary conditions are both displacement restraints. That is because a minimum of two displacement restraints are needed on a beam to prevent rigid body motion. v(0) = f = 0 dv (0) = e = 0 dx → v(x) = F EIzz  x3 Lx2 − 2 6  (n) Note that the displacement is a cubic polynomial, the same as Equation (d). The maximum displacement is at x = L. vmax = v(l) = FL3 3EIzz (o) A plot of the displacement is given in Figure (c). FL3 3EI zz Displacement, υ(x) L 0 x Figure (d) For a statically determinate beam either the fourth order or the second order equation is suitable. ########### --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 271 Classical Analysis of the Bending of Beams 7.4 The Statically Determinate Case With EIzz and fy (x) continuous functions of x and concentrated loads at ends the possible statically determinate cases include a beam with all external forces in equilibrium for which Example 7.3.1 is a special case and beams with the restraints as shown in Figure 7.4.1. All loads including concentrated forces and moments at a boundary are satisfied by using static equilibrium to find both V(x) and M(x). fy(x) fy(x ML M0 or FL F0 or fy(x) M0 ML Figure 7.4.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Once the internal shear forces and bending moments are found you can find the stresses.  M(x) V (x) y τxy (x, y) = ydA σx (x, y) = − Izz Izz b A y (7.4.1) To find displacements EIzz d 2v = M(x) dx2 → dv(x) = dx  M(x) d x + e → v(x) = EIzz  M(x) d x + ex + f EIzz (7.4.2) The constants of integration are found by imposing two displacement restraints, either v(0) = 0 and dv(0) =0 dx or v(L) = 0 and dv(L) =0 dx or v(0) = 0 and v(L) = 0 (7.4.3) When distributed loading is discontinuous, and/or concentrated loads are at midpoints, or EIzz is discontinuous, the equations must be written for each segment. Several possibilities are represented in Figure 7.4.2. There are, of course, an infinite number of variations. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 272 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods f1 (x) F2 x1 FL x2 or f1 (x) F2 f1 (x) or F2 or f1 (x) F2 Figure 7.4.2 All loads including concentrated loads at a boundary are satisfied by using static equilibrium to find Mn (x) and Vn (x) in each region. To find the stresses in each region  Vn (x) Mn (x) ydA (7.4.4) y τx yn (x, y) = σxn (x, y) = − In In bn A y To find displacements: dv 1 d 2 v1 = = M1 (x) → dx2 dx E 2 I2 d 2 v2 = M2 (x) dx2 E n In d 2 vn = Mn (x) dx2 →  dv2 = dx  dvn = dx  M1 (x) d x + e1 → v1 (x) = E 1 I1 M2 (x) d x + e2 → E 2 I2  M1 (x) d xd x + e1 x + f 1 E 1 I1  M2 (x) v2 (x) = d xd x + e2 x + f 2 E 2 I2 ··· Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS → Mn (x) d x + en → E n In vn (x) =  Mn (x) d xd x + en x + f n E n In (7.4.5) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- E 1 I1 P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 273 Classical Analysis of the Bending of Beams The constants of integration are found in part by imposing displacement restraints as given in Equation 7.4.3 in the first three cases respectively. The last example in Figure 7.4.2 is a special case of a simple support at a location other than the ends. As long as there are only two simple supports they may be any place along the beam and it will still be statically determinate. In addition we must enforce continuity of displacement and slope between regions of integration. v1 (x1 ) = v2 (x1 ) dv2 (x1 ) dv1 (x1 ) = dx dx v2 (x2 ) = v3 (x2 ) vn−1 (xn−1 ) = vn (xn−1 ) ... dv3 (x2 ) dv2 (x2 ) = dx dx ... dvn−1 (xn−1 ) dvn (xn ) = dx dx (7.4.6) When the second order equations are used the shear forces, V, and the internal moments, M, are found first and the stresses are calculated. Then the displacement is found by integrating twice. The constants of integration are evaluated by using two displacement boundary conditions. Here are some more examples of statically determinate cases. ########### Example 7.4.1 Problem: Find the stress and displacement in a cantilever beam with a uniform distributed load as shown in Figure (a). fy(x) = f0 N/mm x Figure (a) Solution: First find the internal shear force and moment to find the stresses and then solve the second order equation for the displacement. A free body diagram of a portion of the beam showing the internal force and moment is shown in Figure (b). f0 V M x L Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 274 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods From Figure (b) the internal shear and moment are  Fy = f 0 (L − x) − V (x) = 0 →  Mz = f 0 (L − x) V (x) = f 0 (L − x) (L − x) − M(x) = 0 2 → M(x) = (a) f0 (L − x)2 2 The shear and bending moment diagrams are shown in Figure (c). f0L Shear, V(x) x 0 L f 0 L2 2 Moment, M(x) x 0 Figure (c) The normal stress is σx (x, y) = − M(x)y f0 =− (L − x)2 y Izz 2Izz (b) The shear stress is τxy (x, y) = V Izz b  ydA y = Ay f 0 (L − x) Izz b  ydA y (c) Ay If the cross section is rectangular with height h and width b the shear stress is   f 0 (L − x) h 2 − y2 τxy (x, y) = 2Izz 4 (d) In such a case the maximum stresses are at x = 0 and at y = ±h/2 for normal stress and y = 0 for shear stress.   f0 L 2 h f0 L 2 =± h τmax = τxy (0, 0) = h (e) σmax = σx 0, ± 2 4Izz 8Izz Integrate twice to find the displacement. (f) The boundary conditions are v(0) = f = 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS dv (0) = e = 0 dx Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (g) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- M f0 d 2v = = (L − x)2 dx2 EIzz 2EIzz   f0 x3 dv 2 2 = +e L x − Lx + → dx 2EIzz 3   2 2 3 4 Lx x f0 L x − + + ex + f → v(x) = 2EIzz 2 3 12 P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 275 Classical Analysis of the Bending of Beams The displacement is v(x) = f0 2EIzz  Lx3 x4 L2x2 − + 2 3 12  (h) Note that the displacement is a fourth order polynomial. The maximum displacement is at x = L. vmax = v(L) = f0 L 4 8EIzz (i) Example 7.4.2 Problem: Find the stresses and displacements in a beam with two transverse forces as shown in Figure (a). The cross section is a rectangle of height h and width b. 2F F l l L Figure (a) Solution: We can divide the beam into two regions with separate equations covering each region. This is statically determinate and so we can find the moments and shear forces in each region to be  Mz = F (2l − x) − 2F (l − x) − M1 = 0  Region 2: 0≤x ≤l F − 2F − V1 = 0 → V1 = −F  Mz =F (2l − x) − M2 = 0 → M2 = F (2l − x)  Fy = F − V2 = 0 Fy = (a) l ≤ x ≤ 2l → V2 = F The shear and bending moment diagrams are shown in Figure (b). F Shear, V(x) 0 x l 2l -F FL 2 Moment, M(x) x 0 Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- → M1 = F (2l − x) − 2F (l − x) = F x Region 1: P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 276 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods One can quickly see where the moments and shear forces are a maximum and consequently where the largest stresses occur. The maximum shear stress occurs at y = 0 all along the beam; however its value is negative along the left half and positive along the right half.   2 V F 2 h − y2 → τmax = ± h (b) τxy (x, y) = 2Izz 4 8Izz The maximum bending stress occurs at x = σx (x, y) = − L and at y = ± h2 . 2 FL y 2Izz → σmax = ± FL h 4Izz (c) The equations governing the displacement in each region are EIzz d 2 v1 = Fx dx2 EIzz d 2 v2 = F (2l − x) dx2 (d) The boundary conditions are dv 1 (0) = 0 dx v1 (l) = v2 (l) dv2 dv 1 (l) = (l) dx dx (e) The solutions are found by integrating twice in each region.  2  3 F F dv 1 x x + e1 + e1 x + f 1 (x) = v1 (x) = dx EIzz 2 EIzz 6     F x2 F x3 dv2 (x) = + e2 + e2 x + f 2 v2 (x) = 2lx − lx 2 − dx EIzz 2 EIzz 6 (f) Applying boundary conditions, the displacement in the first region is v1 (0) = f 1 = 0 dv 1 (0) = e1 = 0 dx → v1 (x) = F x3 EIzz 6 0≤x ≤ L 2 (g) The boundary conditions for the second region are the continuity of slope and displacement where the two regions meet.   F l2 dv2 F l2 dv 1 2 (l) = = (l) = 2l − + e2 → e2 = −l 2 dx EIzz 2 dx EIzz 2 (h)   F l3 l3 F l3 = v2 (l) = l 3 − − l 3 + f2 → f2 = v1 (l) = EIzz 6 EIzz 6 3 The displacement in the second region is   F l3 x3 v2 (x) = − l2x + lx 2 − EIzz 6 3 l ≤ x ≤ 2l (i) Note that in each region the displacement is a cubic polynomial. It is sometimes useful to define the displacements in term of the discrete values of displacement at the points of load application and restraint and in terms of local coordinates. See Example 4.4.5 in the axial case. In the bending case the discrete displacements must include both the lateral displacement and the slope in order to represent the cubic polynomial, that is, we need four quantities to define the four coefficients of the cubic polynomial. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- v1 (0) = 0 P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 277 Classical Analysis of the Bending of Beams We use the following notation for the slopes θ= dv dx (j) The plot of the displacements based on Equations (g) and (i) is shown in the Figure (c). υ2l ,θ 2l υ(x) υl ,θ l υ0 ,θ 0 l 2l s1 s2 Figure (c) To this plot we have added two local coordinates, s1 and s2 and the lateral displacements and slopes at the discrete points of load application and restraint. Let us define the cubic polynomial for region 2 where the displacements are vl , θl at s2 = 0 and v2l , θ2l at s2 = l. The distributed displacement and slope are v(s2 ) = as23 + bs22 + cs2 + d (k) dv (s2 ) = 3as22 + 2bs2 + c dx Evaluate the distributed displacements at the discrete points s2 = 0, l. θ (s2 ) = --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- v(0) = d = vl v(l) = al 3 + bl 2 + θl l + vl = v2l θ(0) = c = θl θ(l) = 3al 2 + 2bl + θl = θ2l (l) Solve for a and b in terms of the discrete values of displacement and slope. 1 (2vl + θl l − 2v2l + θ2l l) l3 The displacement in region 2 may be written a= b= 1 (−3vl − 2θl l + 3v2l − θ2l l) l2 (m) 1 1 (2vl + θl l − 2v2l + θ2l l) s23 + 2 (−3vl − 2θl l + 3v2l − θ2l l) s22 + θl s2 + vl l3 l This may be rearranged into the following matrix form ⎡ ⎤ vl ⎢ 1  3 θl ⎥ ⎥ −l 2 s22 + ls23 ⎢ 3ls22 − 2s23 l 3 s2 − 2l 2 s22 + ls23 v(s2 ) = 3 l − 3ls22 + 2s23 ⎣ v2l ⎦ l θ2l v(s2 ) = (n) (o) In a similar fashion we can write the displacement in region one to be v(s1 ) = 1  3 l − 3ls12 + 2s13 l3 l 3 s1 − 2l 2 s12 + ls13 3ls12 − 2s13 −l 2 s12 + ls13 ⎡ ⎤ v0 ⎢ θ0 ⎥ ⎢ ⎥ ⎣ vl ⎦ θl (p) These are the shape functions for a beam element which we shall utilize in Chapter 8 when we introduce the general (finite element) method for the beam. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 278 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In fact this format given in Equations (o) and (p) in Example 7.4.2 applies to all regions between concentrated loads in any uniform beam segment, or, in general for element m ⎡ ⎤ vn ⎢ θn ⎥ 1  ⎥ v(sm ) = 3 lm3 − 3lm sm + 2sm3 lm3 sm − 2lm2 sm2 + lm sm3 3lm sm2 − 2sm3 −lm2 sm2 + lm sm3 ⎢ ⎣ vn+1 ⎦ lm θn+1 (7.4.7) Now why did we go to all this fuss when the displacements given in Equation (g) look simpler and easier to use? It is because Equation 7.4.7 can be applied to all beams with concentrated loads – determinate or indeterminate. And it will make the solution of more complicated cases of loading and restraint much easier. We shall explore the use of these functions in great detail in Chapter 8. Much of what has been written on the subject of beams, especially in the introductory texts, dwells at great length on determinate beams. Because of the importance of bending moments and shear forces a visual representation is sought. So these texts spend much time on bending moment and shear diagrams. From this visual representation you quickly see where the bending moments and shear forces are a maximum and thus where the normal stresses and shear stresses are a maximum. However, in real life many beams are statically indeterminate and the bending moments and shear forces cannot be found from static equilibrium alone. In such cases you must integrate the fourth order equations to find the displacements, apply boundary conditions to find the constants of integration and then differentiate twice to find the moments and three times to find the shear forces. We shall demonstrate this in Section 7.6. First let us take a look at the use of work and energy in the statically determinate case. 7.5 Work and Energy—Castigliano’s Second Theorem The strain energy in bending for a beam of length L consists of a contribution from the normal stress and another from the shear stress.    2  2 τxy 1 σx 1 1 1 dV + dV (7.5.1) σx εx dV + γxy τxy dV = U B = Un + Us = 2 2 2 E 2 G where σx (x, y) = − M(x)y Izz τxy (x, y) = V (x) Izz b  (7.5.2) ydA y Ay The contribution to the strain energy of the normal bending stress is     2  1 My 2 σx 1 1 1 dV = − σx εx dV = dV Un = 2 2 E 2 E Izz  L   1 1 L M2 M2 = dx y2d A = dx 2 2 0 EIzz 2 0 EIzz A (7.5.3) The contribution to the strain energy of the shear stress in bending is 1 Us = 2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  1 γxy τxy dV = 2  2 τxy 1 dV = G 2G   V Izz b  ydA y Ay 2 dV (7.5.4) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ########### P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 279 Classical Analysis of the Bending of Beams With Castigliano’s second theorem (see Chapter 4, Section 4.11) we can find the deflection at the point of application of a concentrated load in the direction of that load. vi = ∂U ∂ Fi (7.5.5) In the next two examples we shall learn that the contribution of the shear strain energy is very small and usually can be neglected. ########### Example 7.5.1 Problem: Consider the common case of a beam of uniform material and cross section and with one end fixed and a concentrated force at the other end as shown in Figure (a). The cross section is rectangular with a height h and width b. Find the displacement at the right end. See Example 7.3.2 for comparison. F L Solution: Use Castigliano’s second theorem. From equilibrium the internal moment is  Mz = F (L − x) − M(x) = 0 M(x) = F (L − x) → (a) The contribution to the strain energy of the normal bending stress in this case is  L    1 L F 2 (L − x)2 F2 x 3  F2 L3 1 L M2 2 2 L x − Lx + dx = dx = = Un =  2 0 EIzz 2 0 EIzz 2EIzz 3 0 6EIzz The shear force is constant, that is,  Vy = F − V (x) = 0 V (x) = F → (b) (c) For the rectangular cross section the shear stress is from Equation 7.3.14     2 F V (x) h 2 h 2 2 −y = −y τxy (x, y) = 2Izz 4 2Izz 4 (d) The strain energy of shear is Us = 1 2  γxy τxy dV = 1 2  2 τxy G dV = F 2b 8GI2zz  h 2 − h2  L 0  h2 − y2 4 2 d xd y = F 2 bh 5 L 240GI2zz Using Castigliano’s second theorem we have the displacement at the point of load application.    2 3 FL3 ∂Un ∂Us ∂ ∂ Fbh 5 L ∂U B F L F 2 bh 5 L = = + = + v(L) = + 2 ∂F ∂F ∂F ∂ F 6EIzz ∂F 3EIzz 96GIzz 120GI2zz Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (e) (f) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (a) P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 280 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The value found in Example 7.3.2, Equation (m), by solving the differential equation is v(L) = FL3 3EIzz (g) ########### This very suspiciously looks as if we found only the normal stress contribution to the displacement when we solved the differential equation. In fact, that is what happened because our assumption in developing the beam theory neglects shear deformation. That the shear contribution can be neglected is examined in the next example. ########### Example 7.5.2 Problem: Examine the contribution of shear energy to the displacement of the beam in Example 7.5.1. Solution: Compare the displacements with and without the strain energy of shear stress. Let us take a specific case to compare. Use an aluminum beam with E = 68950 MPa and G = 26520 MPa. It has the dimensions L = 1000 mm h = 100 mm b = 40 mm (a) Its moment of inertia is 40 · 1003 bh 3 = 12 12 The strain energy of normal stresses in bending is Izz = Un = F 2 · 10003 · 12 1 F2 L3 1 · = · = F 2 · 0.0007252 2 3EIzz 2 3 · 68950 · 40 · 1003 (b) (c) The strain energy of shear stresses in bending is Us = 3.1000 F 2 bh 5 L 122 F 2 bh 5 L 3F 2 L · 2 6 = = F 2 = 5.656 × 10−6 F 2 = 2 240G b h 5bhG 5 · 40 · 100 · 26520 240GIzz (d) The ratio of the strain energy of normal stresses to the strain energy of shear stresses is The displacement under the load including shear energy is   v(L) = F · 2 · 7.252 × 10−4 + 5.626 × 10−6 = 0.0014617F (e) (f) From bending energy alone the displacement is v(L) = F · 2 · (0.0007252) = F · 0.0014504 (g) This is a 1.94 % change. The contribution of shear to displacement is usually neglected for slender beams. ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 7.252 × 10−4 Un = = 128.22 Us 5.656 × 10−6 P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 281 Classical Analysis of the Bending of Beams When there are multiple loads on a determinate beam, including distributed and concentrated forces and concentrated moments, for example, as shown in Figure 7.5.1, we can use Castigliano’s theorem to find the displacements at the points of concentrated load application. fy(x) F1 F2 … Fn M1 L Figure 7.5.1 First find the strain energy as a function of all the loading.   U B F1 , F2 , . . . , Fn , f y1 , . . . , f yn , M1 , . . . Mn (7.5.5) Do not insert numerical values for the loads until after the differentiation has been done. The displacements at the points of concentrated load application are vi = ∂U B ∂ Fi θi = ∂U B ∂ Mi (7.5.6) where θi = dvi dx (7.5.7) Note that differentiation with respect to a force finds the lateral displacement and differentiation with respect to a moment finds the slope. If you wish to find the displacement or slope at a point where no load is applied you introduce a force or moment at that point, differentiate to obtain the expression for that displacement and then set the load to zero. ########### 7.6 The Statically Indeterminate Case With EIzz and fx (x) continuous functions of x and three or more displacement restraints the fourth order equation must be used. d 2v d2 EI = f y (x) zz dx2 dx2 (7.6.1) and, of course, when EIzz is constant EIzz d 4v = f y (x) dx4 (7.6.2) When EIzz is constant we integrate to find the displacement as explained in Equation 7.3.22.     1 a b v(x) = f y (x)d x + x 3 + x 2 + cx + d (7.6.3) EIzz 6 2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 282 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Distributed loads are accounted for but concentrated loads must enter through the boundary conditions. When the beam is indeterminate the constants of integration are found by three or more boundary displacements and one boundary force or four boundary displacements. Some of the possible configurations of restraints are shown in Figure 7.6.1. fx(x) or fx(x) M0 or fy(x) ML Figure 7.6.1 For the beam with both ends fixed the boundary conditions are v(0) = 0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- dv(0) =0 dx dv (L) =0 dx v(l) = 0 (7.6.4) For the two beams with a simple support at one end and a fixed support at the other we have v(0) = 0 and EI d 2 v(0) = M0 dx2 dv (L) =0 dx (7.6.5) d 2 v (L) = ML dx2 (7.6.6) and v(l) = or v(0) = dv(0) = 0 and dx v(l) = 0 and EI To find the stresses differentiate the displacement. M(x) d 2v y = M(x) → σx (x, y) = − I dx2  3 d v V (x) EI 3 = −V (x) → τxy (x, y) = ydA dx I b Ay EI Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (7.6.7) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 283 Classical Analysis of the Bending of Beams When distributed loading is discontinuous, and/or concentrated loads are at midpoints, or EI is discontinuous but constant for each segment the equations.     a1 b1 f y1 (x) 1 d 4 v1 f y1 (x)d x + x 3 + x 2 + c1 x + d1 = → v (x) = 2 dx4 E 1 I1 E 1 I2 6 2     a 2 3 b2 2 f y2 (x) 1 d 4 v2 f y2 (x)d x + x + x + c2 x + d2 = → v2 (x) = (7.6.8) dx4 E 2 I2 E 2 I2 6 2 ···     ai bi f yi (x) 1 d 4 vi f yi (x)d x + x 3 + x 2 + ci x + di = → v (x) = i dx4 E i Ii E i Ii 6 2 The constants of integration are found by imposing boundary conditions. Here we can have a great variety of restraints because we can have intermediate supports as well as end supports. There must be a minimum of two displacement restraints but there may be many more. For example the beam shown in Figure 7.6.2 has five displacement restraints. It would have to be divided into at least seven segments in order to solve. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- fy(x) F ML Figure 7.6.2 Apply whatever displacement restraints are necessary at the ends of each segment of the beam. Then apply continuity of displacement and slope where segments meet. Let two adjacent segments be identified by the subscripts i and i + 1 and the coordinate where they meet by x j .         dvi+1 x j dvi x j = (7.6.9) vi x j = vi+1 x j dx dx and where necessary apply equilibrium of internal and external forces between segments. When concentrated moments M j or forces F j are present this means d 2 vi (x j ) d 2 vi+1 (x j ) − M j − E i+1 Ii+1 =0 2 dx dx2 3 3 d vi (x j ) d vi (x j ) + F j + E i+1 Ii+1 =0 −E i Ii dx3 dx3 E i Ii (7.6.10) Once the displacement are known differentiate to find the stresses. Mi (x) d 2 vi (x) y = Mi (x) → σxi = − Ii dx2  3 Vi (x) d vi (x) = −Vi (x) → τx yi = ydA y E i Ii dx3 I b Ay E i Ii (7.6.11) There are too many possibilities to list them all. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas 284 July 4, 2011 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods When the fourth order equation is used the displacements are found first by integration. The constants of integration are found from a combination of displacement boundary conditions and force boundary conditions. The stresses are then found by differentiating the displacements to find the internal shear forces and moments and inserting these values in the appropriate formulas. If the problem is indeterminate you must use the fourth order equations. If the problem is determinate you may use either the second order or the fourth order equations ########### Problem: Find the stresses and displacement of a uniform beam with both ends fixed and loaded with a uniform distributed applied load as shown in Figure (a). The cross section is rectangular with a height of h and width b. fy(x) = f0 x Figure (a) Solution: Solve the fourth order beam equation for the displacements and then differentiate to find the internal shear force and moment and hence the stresses. A free body diagram of a portion of the beam in Figure (b) shows that there are four unknowns – the internal force and moment and the restraint force and moment. We have just two equations of static equilibrium; therefore, this is a statically indeterminate problem. f0 V(x) M(x) Q x L R Figure (b) We turn to the fourth order equation in such cases. For this problem we have EIzz Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS d 4v = f y (x) = f 0 dx4 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Example 7.6.1 P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 285 Classical Analysis of the Bending of Beams Integrating four times we obtain four constants of integration. f0 d 4v = dx4 EIzz → → → → d 3v f0 = x +a dx3 EIzz f0 2 d 2v = x + ax + b 2EIzz dx2 dv f0 3 a 2 = x + x + bx + c dx 6EIzz 2 bx 2 ax 3 f0 4 + + cx + d x + v(x) = 24EIzz 6 2 (b) Boundary conditions are v(0) = d = 0 dv (0) = c = 0 dx bL 2 f0 a L3 + =0 v(L) = L4 + 24EIzz 6 2 f0 3 a 2 dv (L) = L + L + bL = 0 dx 6EIzz 2 (c) and so a=− f0 L 2EIzz b= f0 L 2 12EIzz c=0 d=0 (d) Our final answer for the displacement is v(x) = f0 L 3 f0 L 2 2 f0 f0 x4 − x + x = (x 4 − 2L x 3 + L 2 x 2 ) 24EIzz 12EIzz 24EIzz 24EIzz (e) The shear force is V (x) = −EIzz f0 L d 3v = f0 = − f0 x + dx3 2  L −x 2  (f) and the bending moment is M(x) = EIzz d 2v = f0 dx2  Lx L2 x2 − + 2 2 12  --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (g) P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 286 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The shear and bending moment diagrams are shown in Figure (c). f0 L 2 Shear, V(x) x 0 f0 L 2 − f 0 L2 12 Moment, M(x) x 0 − L f 0 L2 12 Figure (c) The shear and normal stresses are   2     y2 f0 L f0 V h L −x −x − τxy (x, y) = ydA y = ydA y = Izz b A y Izz b 2 Izz 2 8 2 Ay   2 Lx L2 M(x) f0 x σx (x, y) = − − + y y=− Izz Izz 2 2 12 (h) The maximum shear stresses and the maximum bending stresses are at the fixed ends. τmax = ± f0 L 2 h 16Izz σmax = ± f0 L 2 h 24Izz (i) ########### Example 7.6.2 Problem: Find and plot the displacements, shear forces, and bending moments for the beam shown in Figure (a). Mc Figure (a) Solution: Use the fourth order equations. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- L L/2 P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 287 Classical Analysis of the Bending of Beams The beam is indeterminate so we must divide it into two regions and use the fourth order equations and find the displacements first. d 4 v1 =0 → dx4 dv 1 x2 = a1 → dx 2 d 4 v2 EIzz 4 = 0 → dx dv2 x2 = a2 → dx 2 EIzz d 3 v1 = a1 dx3 → + b1 x + c1 → 3 d v2 = a2 dx3 → + b2 x + c2 → d 2 v1 = a1 x + b1 dx2 x3 x2 v1 = a1 + b1 + c1 x + d1 6 2 d 2 v2 = a2 x + b2 dx2 x3 x2 v2 = a2 + b2 + c2 x + d2 6 2 (a) The boundary conditions at the ends are v1 (0) = 0 dv 1 (0) =0 dx dv2 (L) =0 dx v2 (l) = 0 → d1 = 0 → c1 = 0 → → L2 + b2 L + c2 = 0 2 L3 L2 + b2 + c2 L + d2 = 0 a2 6 2 (b) a2 The boundary conditions for continuity between the two regions at L/2 are     L2 L3 L2 L L3 L L v1 + b1 = a2 + b2 + c2 + d2 = v2 → a1 2  2   48 8 48 8 2 L L dv2 dv 1 L2 L L2 L 2 2 = → a1 + b1 = a2 + b2 + c2 dx   dx 8 2 8 2   L L 2 2 d v1 d v2 Mc L L 2 2 − a2 − b2 = 0 EIzz − M − EI = 0 → a1 + b1 − c zz 2 EIzz 2 dx2 dx2     L L d 3 v1 d 3 v2 2 2 − = 0 → a1 = a2 dx3 dx3 (c) We can use a1 = a2 to simplify this set of linear algebraic equations. a2 L2 + b2 L + c2 = 0 2 L2 L3 + b2 + c2 L + d2 = 0 a2 6 2 2 L L2 L − b2 − c2 − d2 = 0 b1 8 8 2 L L b1 − b2 − c2 = 0 2 2 Mc b1 − b2 = EIzz --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 288 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The five algebraic equations for the constants can be put in matrix form. ⎡ 0 ⎢ ⎢ ⎢ ⎢ 0 ⎢ ⎢ ⎢ L2 ⎢ ⎢ 8 ⎢ ⎢L ⎢ ⎣ 2 1 L2 2 L3 6 0 0 0 L L2 2 L2 − 8 L − 2 −1 1 L − L 2 −1 0 0 ⎤ ⎥ ⎡ ⎥⎡ ⎤ 0 ⎥ b1 ⎢ 0 1⎥ ⎥ ⎢ a2 ⎥ ⎢ ⎥⎢ ⎥ ⎢ 0 ⎥ ⎢ b2 ⎥ = ⎢ ⎢ ⎥ ⎢ −1⎥ ⎥ ⎣ c2 ⎦ ⎢ 0 ⎥ ⎣ Mc ⎥ d2 ⎥ 0⎦ EIzz 0 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (e) Using a symbolic solver (see Appendix C) the constants are found to be ⎡ 3 ⎢ 2L ⎢ ⎢ 1 ⎢ − ⎡ ⎤ ⎢ 4 a1 ⎢ ⎢ b1 ⎥ ⎢ 3 ⎢ ⎥ ⎢ ⎢ a2 ⎥ ⎢ M ⎢ ⎥ = c ⎢ 2L ⎢ b2 ⎥ EI ⎢ 5 zz ⎢ ⎢ ⎥ ⎢ −4 ⎣ c2 ⎦ ⎢ ⎢ L d2 ⎢ ⎢ ⎢ 2 ⎣ L2 − 8 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (f) The displacements are Mc Mc 2 x2 x3 + b1 = x3 − x 6 2 4EIzz L 8EIzz Mc 5Mc 2 Mc L Mc L 2 x2 x3 x3 − x + x− v2 = a2 + b2 + c2 x + d2 = 6 2 4EIzz L 8EIzz L 2EIzz 8EIzz v1 = a1 (g) The bending moments are d 2 v1 = EIzz (a1 x + b1 ) = dx2 2 d v2 M2 = EIzz 2 = EIzz (a2 x + b2 ) = dx M1 = EIzz Mc 3Mc x− 2L 4 5Mc 3Mc x− 2L 4 (h) The shear forces are 3Mc d 3 v1 = −EIzz a1 = − dx3 2L 3Mc d 3 v2 V2 = −EIzz 3 = −EIzz a2 = − dx 2L V1 = −EIzz (i) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 289 Classical Analysis of the Bending of Beams Plots of shear and bending moment are given in Figure (b) 0 x Shear force, V(x) L 3M c − 2L Mc 2 Moment, M(x) 0 − x Mc 2 Figure (b) The stresses are     Mc 5Mc M2 M1 y y 3Mc 3Mc x− x− y=− σ2 = − y=− Izz 2L 4 Izz Izz 2L 4 Izz    3Mc h 2 V − y2 ydA y = − τ1 = τ2 = Izz b A y 4L Izz 4 σ1 = − (j) h The bending stress is maximum at x = L/2 and y = ± . The shear stress has a constant value 2 throughout the length and is maximum at y = 0. Plots of the displacements are given in Figure (c). McL2/4EI x L/2 L -0.025 Figure (c) In the indeterminate case we solve the fourth order equation for the displacement and then find the bending moments and shear forces in order to find the stresses. ########### In the determinate case the bending moments and shear forces are inputs to the solution for the stresses and the displacements. In the indeterminate case the bending moments and shear forces are output from the displacement solution. It is important to note that the fourth order equation is equally successful in solving determinate and indeterminate cases. In the past, because of the limitations in analysis capability, many structures were designed to be determinate. Now with more powerful methods this is no longer the norm. In fact, the Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 0.025 P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 290 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods concept of fail safe structures demands indeterminacy, that is, you must have multiple internal load paths in the structure so that a local failure does not cause a global failure. This inherently implies an indeterminate structure. So we shall not dwell on methods created to find bending moments and shear forces as inputs to the problem. In our favored methods these are outputs from the displacement solution. 7.7 Solutions from the Theory of Elasticity Let us pause and ask whether the beam equations and their solutions based on plane sections remaining plane are, if fact, valid. We have, so far, invoked experimental evidence. The fact that these equations have been used successfully for so many decades adds credence. Still we must ask under what conditions they might fail. Consider a beam with a rectangular cross section with moments applied at each end via a linearly varying stress as shown in Figure 7.7.1. The load distribution at the ends of the beam exactly duplicates the internal stress obtained from the beam equations. Let the cross section have a height h and width b. The length of the beam is L and the origin of the coordinates is at the centroid of the cross section as shown. Let the end loads be expressed by a force per unit area of value   2 f0 L y f x ± , y, z = ∓ 2 h (7.7.1) The loads on all other surfaces are zero and there are no body forces. Let us assume that the state of stress throughout the beam is σx = − 2 f0 y h σ y = σz = τxy = τyz = τzx = 0 (7.7.2) and ask if this stress distribution satisfies the complete equations of elasticity. y y L  L  2f f x  − , y, z  = 0 y h  2  z L/2 h b z --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 2f L  f x  , y, z  = − 0 y h 2  x Figure 7.7.1 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 291 Classical Analysis of the Bending of Beams Upon substituting these values into the three dimensional equations of equilibrium (see Chapter 2, Section 2.6, Equations 2.6.13) we find that they are satisfied identically if the body forces are zero. ⎡ ∂ ⎢∂x ⎢ ⎢ ⎢ 0 ⎢ ⎢ ⎣ 0 0 0 ∂ ∂y 0 0 ∂ ∂z ∂ ∂y ∂ ∂x 0 ∂ ∂z ∂ ∂y 0 ⎡ − ∂ ⎤⎢ ⎢ ⎢ ∂z ⎥ ⎥⎢ ⎥⎢ ⎢ 0 ⎥ ⎥⎢ ⎥⎢ ⎢ ∂ ⎦⎢ ⎢ ∂x ⎣ ⎤ 2 f0 y⎥ h ⎥ 0 ⎥ ⎥ ⎡ ⎤ 0 ⎥ 0 ⎥ ⎥ = ⎣0⎦ ⎥ 0 0 ⎥ ⎥ ⎥ 0 ⎦ 0 (7.7.3) We note that the internal stresses match the surface forces at the boundaries. The strain components are εx = 2 f0 σx =− y E Eh ε y = −v 2v f 0 σx = y E Eh εz = −v 2v f 0 σx = y E Eh γxy = γyz = γzx = 0 (7.7.4) Again we note that linear functions of strain satisfy compatibility (see Chapter 3, Section 3.3), that is, ∂ 2 γxy ∂ 2 εy ∂ 2 εx + = 2 2 ∂y ∂x ∂ x∂ y → ∂2 ∂ y2  −    ∂ 2 2v f 0 ∂2 2 f0 y + 2 y = (0) = 0 Eh ∂x Eh ∂ x∂ y (7.7.5) Now let us find the displacements associated with this loading. If we ignore the Poisson’s ratio effect and examine the displacement in the x direction we have εx = 2 f0 ∂u =− y ∂x Eh → u(x, y) = − 2 f0 yx + f (y, z) Eh (7.7.6) Since u(0, y, z) = 0 → f (y, z) = 0 → u(x, y) = − 2 f0 yx Eh (7.7.7) it is confirmed that u(x, y, z) is linear in y, or that plane sections remain plane. There will be displacements in the y and z directions from the Poisson’s ratio effect, but for a long slender beam these effects are small. Numerical analysis methods allow us to solve the more exact equations of material behavior which do not make assumptions about what will happen to planar sections after loads are applied. Computer graphics allow us to present visual confirmation of this behavior. The results of the numerical analysis for a beam with the loading given in Equation 7.7.1 are shown in Figure 7.7.2. The displacement is greatly exaggerated to make it visible. In this particular software program the origin of the axes is at the centroid of the deformed shape. Note that the shape compares favorably to Figure (b) in Example 7.3.1 if you move the axes. Note also that the end planes appear still to be plane. The contour lines between changes in color are lines of constant stress. The uniform parallel lines of color confirm that the stress distribution in the interior is the same as that applied to the ends. This infers that the plane sections do remain plane. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas 292 July 4, 2011 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure 7.7.2 Once again a change in geometry, loading, or restraint will produce a problem for which it usually is impossible to find an exact analytical solution. The next demonstration applies the end stresses as shown in Figure 7.7.3. There is a parabolic distribution of f x (x, y, z) in the y direction that is compressive for positive y and tensile for negative y.     L 2y For positive y : f x ± , y, z = ∓ f 0 y 1 − 2 h (7.7.8)     L 2y For negative y : f x ± , y, z = ∓ f 0 y 1 + 2 h y fx(x) fx(x) x Figure 7.7.3 We cannot expect that the stress in the interior will be the same as the stress on the ends, but let us check just in case. Assume   2y For positive y : σx (x, y, z) = − f 0 y 1 − h (7.7.9)   2y For negative y : σx (x, y, z) = − f 0 y 1 + h Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-07 P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 293 Classical Analysis of the Bending of Beams Equilibrium is satisfied in the absence of body forces for positive y as noted in the next equation.   ⎡ 2y ⎤ − f y 1 − ⎡ ∂ ∂ ∂ ⎤⎢ 0 h ⎥ ⎢ ⎥ 0 0 0 ⎥ ⎡ ⎤ ⎥⎢ ⎢∂x ∂ y ∂z 0 ⎥ ⎥⎢ ⎢ 0 ⎥ ⎥⎢ ⎢ ∂ ∂ ∂ ⎥ = ⎣0⎦ ⎥⎢ ⎢ 0 0 0 0 (7.7.10) ⎥ ⎥⎢ ⎢ ∂y ∂ x ∂z ⎥ ⎥⎢ ⎢ 0 0 ⎥ ⎣ ∂ ∂ ∂ ⎦⎢ ⎢ ⎥ 0 0 0 ⎣ ⎦ 0 ∂z ∂y ∂x 0 It is obvious that it is also satisfied for negative y. The interior forces match the surface forces at the boundaries. So equilibrium and boundary conditions are satisfied. Continuing, the strain components would be     f0 y v f0 y 2y 2y σx σx =− = 1− ε y = −v 1− εx = E E E E h  h  (7.7.11) v f0 y 2y σx = 1− γxy = γyz = γzx = 0 εz = −v E E h Next we check compatibility. ∂ 2 γxy ∂ 2 εy ∂ 2 εx + = ∂ y2 ∂x2 ∂ x∂y       2y ∂ 2 ν f0 y 2y ∂2 ∂2 f0 y (0) = 0 1 − + 1 − − → − ∂ y2 E h ∂x2 E h ∂ x∂ y (7.7.12) Figure 7.7.4 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- It fails compatibility – this cannot be a solution. So we turn to a numerical analysis. The results of a numerical analysis are shown in Figure 7.7.4. As you can see the parabolic stresses on the ends rapidly transform to the linear distribution of beam theory away from the ends. The end loading indeed has a localized effect. What is important to the interior region is the magnitude of the moment, not its distribution on the ends. P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 294 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure 7.7.5 We may infer that in the interior away from the ends this warping effect is extremely small. In the above examples there are no shear forces. Our next demonstration is for the beam in Example 7.3.2. The shear stress in Equation (g) of Example 7.3.2 that was obtained by beam theory is   2 F h 2 −y (7.7.13) τxy (x, y, z) = 2Izz 4 The bending stress is σx (x, y, z) = − F(L − x) y Izz (7.7.14) The forces at the boundaries are as shown in Figure 7.7.6. At the right end we have a distributed shear force.   2 F h − y2 (7.7.15) f y (L , y, z) = 2Izz 4 At the left end there is a distributed shear force f y (0, y, z) = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS F 2Izz  h2 − y2 4  (7.7.16) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- You can notice a small warping at the former plane end section. You can see this better in the enlarged image in Figure 7.7.5. P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 295 Classical Analysis of the Bending of Beams and a distributed normal force f x (0, y, z) = − FL y Izz (7.7.17) y f x ( 0, y, z ) = − FL y I zz x f y ( 0, y , z ) = F h2 ( − y2) 2 I zz 4 fy (L y z) = F h2 − y2 2 I zz 4 Figure 7.7.6 Assume that the stresses in the interior are exactly the same as in Equations 7.7.13 and 7.7.14 as found from beam theory. Do these stresses satisfy the three dimensional equations of equilibrium? ⎡ ∂ ⎢∂x ⎢ ⎢ ⎢ 0 ⎢ ⎢ ⎣ 0 0 0 ∂ ∂y 0 0 ∂ ∂z ∂ ∂y ∂ ∂x 0 0 ∂ ∂z ∂ ∂y ⎤ F(L − x) − y ⎥ Izz ∂ ⎤⎢ ⎢ ⎥ ⎢ ⎥ 0 ⎢ ⎥ ⎡ ⎤ ∂z ⎥ ⎥ ⎥⎢ ? ⎥ ⎥⎢ 0 ⎢ ⎣ ⎦ 0 ⎥ ⎥ ⎥= ? ⎥ ⎢ F  h2 ⎥ ⎥⎢ ? − y2 ⎥ ∂ ⎦⎢ ⎢ 2Izz 4 ⎥ ⎥ ∂x ⎢ 0 ⎣ ⎦ ⎡ (7.7.18) 0 Expanding the equations we find     2  ∂ F(L − x) F ∂ F F h − y2 − = y + y− y=0 ∂x Izz ∂ y 2Izz 4 Izz Izz   2  ∂ F h − y2 =0 ∂ x 2Izz 4 (7.7.19) and so the equilibrium equations are satisfied exactly in the absence of body forces. These stresses match the surface forces at the boundaries. Given these stresses the strains will be F(L − x) vF(L − x) σx σx =− = y ε y = −v y E EIzz E EIzz   2 τxy F h = − y2 γyz = γzx = 0 = G 2GIzz 4 εx = γxy εz = −v vF(L − x) σx = y E EIzz (7.7.20) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 296 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Check for compatibility. ∂ 2 γxy ∂ 2 εy ∂ 2 εx + = ∂ y2 ∂x2 ∂ x∂ y      2   ∂2 F(L − x) vF(L − x) ∂2 ∂2 F h 2 −y → − y + 2 − y − ∂ y2 EIzz ∂x EIzz ∂ x∂ y 2GIzz 4 =0+0−0=0 (7.7.21) The other compatibility equations are satisfied – trust me. Contour plots follow. The contour plot for σ x bending stresses is shown in Figure 7.7.7. Figure 7.7.7 The contour plot for τ xy shearing stresses is shown in Figure 7.7.8. This confirms the accuracy of the elementary bending theory. Now let us find the displacements associated with this loading. If we ignore the Poisson’s ratio effect and examine the displacement in the x direction we have   x2 F Lx − F(L − x) ∂u 2 =− y → u(x, y) = − y + f (y, z) (7.7.22) εx = ∂x EIzz EIzz Apply the boundary condition u(0, y, z) = 0 → f (y, z) = 0 →   x2 F Lx − 2 u(x, y) = − y EIzz (7.7.23) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 297 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Classical Analysis of the Bending of Beams Figure 7.7.8 This confirms that u(x, y, z) is linear in y, or that plane sections remain plane. For one more example to establish St. Venant’s principle consider a cantilever beam, that is, one with the left end fixed and with a constant shear force on the right end, as shown in Figure 7.7.9. Both the fixed condition and the applied load violate the beam theory stress distributions at the boundary. y fy = f0 x Figure 7.7.9 The contour plot for σ x bending stresses is shown in Figure 7.7.10. The contour plot for τ xy shearing stresses is shown in Figure 7.7.11. You will note a small distortion in the bending and shear stresses at the fixed end and at the free end for the shear stress. Over the rest of the beam the approximations are quite valid. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas 298 July 4, 2011 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure 7.7.10 Figure 7.7.11 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 299 Classical Analysis of the Bending of Beams Finally we explore one more approximation in beam theory – the application of a distributed lateral load as shown in Figure 7.7.12 to the upper surface. y f0 fy = f0 x Figure 7.7.12 The bending stress contours are shown in Figure 7.7.13. Figure 7.7.13 The shearing stress contours are shown in Figure 7.7.14. Again St. Venant’s principle dashes to the rescue. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 300 July 4, 2011 Figure 7.7.14 The σ y component of stress is so small compared to the σ x and τ xy components that it is ignored for a slender beam. 7.8 Variable Cross Sections If Izz varies with x we can adapt the equations as we demonstrated for the axial and torsional cases. If the beam is statically determinate we find the internal shear forces and moments and from it the stresses. σx (x, y) = − M(x) y Izz (x) τxy (x, y) = V (x) Izz (x) b  ydA y (7.8.1) Ay Finding the displacement requires integrating with EIzz inside the integral. M(x) d 2v = EIzz (x)  dx2 M(x) dv = dx + e → dx EIzz (x)   M(x) → v(x) = d x + ex + f EI zz (x) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (7.8.2) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 301 Classical Analysis of the Bending of Beams d 2v d2 EIzz (x) 2 = f y (x) 2 dx dx  d 2v d EIzz (x) 2 = f y (x) d x + a → dx dx  d 2v f y (x) d x + ax + b → EIzz (x) 2 = dx     x2 dv 1 f y (x) d x d x + a + bx + c = → dx EI (x) 2     zz x2 x3 1 f y (x) d x d x + a + b + cx + d → v= EIzz (x) 6 2 (7.8.3) Once we find the displacement v(x) by integrating four times and applying boundary conditions to find the constants of integration we can find the internal shear forces and moments. V (x) = − d 2v d EIzz (x) 2 dx dx M(x) = EIzz (x) d 2v dx2 (7.8.4) and the stresses are given by Equation 7.8.1. fy(x) L {Fe) Figure 7.8.1 As we learned in the axial case, except for the simplest functions for the variable cross section properties direct integration is an unwieldy process. It would be even more inconvenient when the multiple integrations for bending would be necessary. It is more common to approximate the variation by a series of constant cross sections, such as is represented in Figure 7.8.1. In practice several sections might be used, not just the three shown. The distributed load would be replaced by equivalent nodal loads and the finite element method used. In Section 5.5 for the axial case we learned how careful interpretation of the results can achieve a high degree of accuracy. We would use the restraint moments at the ends to find the stresses there and use Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- If the beam is statically indeterminate we must use the fourth order equations with EIzz inside the integral. P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas 302 July 4, 2011 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods average values at the nodes. This will be demonstrated in Chapter 8 where the finite element method for beams is presented. 7.9 Shear Stress in Non Rectangular Cross Sections—Thin Walled Cross Sections --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The formula for finding the shear stress in beams that is presented in Section 7.3 assumes that the shear stress is uniformly distributed over the width of the cross section. This is quite accurate for rectangular cross sections. For other sections that do not have large or abrupt changes in width, such as those shown in Figure 7.9.1, it may serve as a reasonable first approximation but not as a final analysis. Figure 7.9.1 The assumption of a uniform shear stress across the width would produce vector components as shown in the figure on the left. Since there are no shear stresses on the side faces of the beam, the stress cannot have components normal to those surfaces at each side. A more reasonable assumption is for the shear stress at the side surfaces to be parallel to the surfaces as shown in the figure on the right. For sections such as the thick walled I beam shown in Figure 7.8.2 the formula is inadequate except as noted below. y z Figure 7.9.2 Using the formula from Equation 7.3.14, repeated here as Equation 7.9.1  V τxy (x, y) = ydA y Izz h A y (7.9.1) we would obtain the shear stress profile shown in Figure 7.9.3. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 303 Classical Analysis of the Bending of Beams y x Figure 7.9.3 This shows a significant shear stress on the horizontal inside faces of the flanges. Clearly the shear stress must be zero on those faces since there is no load on those faces. The values in the vertical web of the I beam, in particular at the midpoint where the shear stress is largest, are accurate. They are inaccurate at the junction of the vertical web and the horizontal flanges and in the horizontal flanges. The more advanced theory of elasticity is necessary for an accurate analysis. There is one other class of cross sections, however, for which a simple theory of shear stress distribution is possible. These are the so called thin walled cross sections such as the examples shown in Figure 7.9.4. Figure 7.9.4 If the surfaces of the beam are free of shear stress then any significant component of the shear stress must be tangent to the surface. For practical purposes any shear in the thin walled cross section can be approximated as uniform across the width of the section. Once again we can define shear flow (7.9.2) q = τb where b is the wall thickness. We can use the method in Section 7.3 to find the shear stress by equating the shear force necessary to place a segment of length dx in equilibrium with the normal forces acting on that segment as we have done in Figure 7.3.6 and Equations 7.3.11-13. σx τ zx σ x+ (a) ∂σ x dx ∂x (b) Figure 7.9.5 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 4, 2011 304 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Consider an I beam as shown in Figure 7.9.5(a) and a small segment of the upper flange in Figure 7.9.5(b). The shear at the cut is found from the differences in the normal stresses at sections x and x + dx. This can be extended to a closed section by considering an element with dimensions ds by dx as shown in Figure 7.9.6(a) with shear flow and stresses as shown in Figure 7.9.6(b). σx q dx ds q+ ∂q ds ∂s σx + (a) ∂σ x ∂x dx (b) Figure 7.9.6 From equilibrium of the element      ∂σx ∂q d x bds − σx bds − qd x + q + ds d x = 0 Fx = σx + ∂x ∂s (7.9.3) From this we obtain ∂σx ∂q = −b ∂s ∂x (7.9.4) The bending stresses are given by σx (x, y) = − My Izz (7.9.5) When Equation 7.9.5 is substituted into Equation 7.9.4 we get b ∂M bV ∂q = y=− y ∂s I ∂x Izz This may be integrated along the path s to obtain the shear flow.  s V q = q0 − byds 0 Izz (7.9.6) (7.9.7) where q0 is the value of the shear flow at the origin of the coordinate s. A simple example will confirm the validity of this equation. ########### Example 7.9.1 Problem: Solve for the shear stress in the simplest of cases: a thin rectangular cross section of height h shown in Figure (a). The width is b. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-07 P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 305 Classical Analysis of the Bending of Beams y z h s Vy Figure (a) Solution: Apply Equation 7.9.7. The shear flow is τxy b = q = q0 −  y − h2 y Vb V V bydy = 0 − b y 2 − h = 2 Izz Izz 2Izz  h2 − y2 4  (a) where we note that q0 = 0 since there are no shear forces on the upper or lower surfaces. This is, of course, exactly what we obtained for a rectangular section in Section 7.3, Example 7.3.2, Equation (g). ########### When the section is not symmetrical about the xy axes plane we encounter a problem with the load application. We have assumed so far that the lateral loads are along the centroidal axis of the beam. If the cross section is symmetrical about the y axis and the load is applied in the xy plane only bending occurs. If the load is off the centroidal axis both bending and torsion may occur. When the cross section is not symmetrical about the y axis this generalization is no longer true. A load in the xy plane may cause both bending and torsion. This situation is presented in detail in Chapter 10. For now we restrict all cross sections, thin walled or not, to ones with symmetry about the y axis and lateral loading only in the xy plane. ########### Example 7.9.2 Problem: Find the shear flow in a circular thin walled cross section shown in Figure (a). The shear force is applied through the centroid of the cross section. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 306 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods s R ϕ Vy --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (a) Solution: Use Equation 7.9.7. The shear flow is q = q0 −  0 s V b R2 V byds = q0 − Izz Izz  ϕ cos ϕdϕ = q0 − 0 V b R2 sin ϕ Izz (a) We can argue from symmetry that the shear flow will be zero at s = 0 and π . Thus q0 is zero. Example 7.9.3 Problem: Consider the thin walled I beam shown in Figure (a). A shear force V is acting at the centroid. The vertical web and the flanges are the same thickness b. Find the shear flows in the cross section. y a a s5 s4 h z b h s3 s2 s1 V Figure (a) Solution: Use Equation 7.9.7. The origins and directions of the local coordinates, s1 to s5 , are shown on Figure (a). Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 307 Classical Analysis of the Bending of Beams Starting with the flange on the lower left Equation 7.9.7 becomes q = q0 − s  0 V byds Izz → q1 (s1 ) = 0 − Vb Izz s1  (−h) ds1 = 0 V bh s1 Izz → q1 (a) = V bha Izz (a) This is plotted in Figure (b). The direction of the shear flow is shown by the arrows. For the flange on the lower right we have q = q0 −  0 s V byds Izz → q2 (s2 ) = 0 − Vb Izz  s1 (−h) ds2 = 0 V bh s2 Izz → q2 (a) = V bha Izz (b) It also is plotted in Figure (b). The positive signs indicate that the flows are in the same direction as the local coordinates. y a a s5 s4 h x h s3 V --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (b) At the juncture where s1 , s2 , and s3 meet we must establish equilibrium, that is, from Figure (c) we note that q3 (0) = q1 (a) + q2 (a) = 2V bha Izz (c) Note that this represents the state of stress at the point where the two flanges meet the vertical web. q3(0) q1(a) q2(a) Figure (c) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 308 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods For the vertical web it is convenient to stick with the y coordinate rather than s3 . The shear flow in the vertical web is then   s Vb y 2V bha V byds → q3 (y) = − ydy q = q0 − Izz Izz −h 0 Izz  2V bha 2V bha V b 2  y Vb  2 (d) = y − h2 − y −h = − Izz 2Izz Izz 2Izz  Vb  = 4ha − y 2 + h 2 2Izz Note that when y = h q3 (h) = 2V bha Izz (e) and when y = 0 q3 (0) =  Vb  4ha + h 2 2Izz (f) A plot of this shear flow is added in Figure (d). It is placed off to the side to avoid confusion. y --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- h h V Figure (d) Finally, the shear flows in the two upper flanges are  s1  s V V bh V byds → q4 (s4 ) = 0 − s4 bhds4 = − q = q0 − I I Izz zz zz 0 0  s1  s V V bh V byds → q5 (s5 ) = 0 − bhds1 = − s5 q = q0 − I I Izz zz zz 0 0 V bha Izz V bha → q5 (a) = − Izz (g) These plots have also been added to Figure (d). The minus signs indicate that the positive flow is in the opposite direction of the local coordinates. → q4 (a) = − ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 309 Classical Analysis of the Bending of Beams 7.10 Design of Beams As noted earlier in analysis we specify the geometry, materials, loads, and restraints and ask for the displacements, strains, and stresses. In design we are more likely to be given the loads and some of the restraints but are asked what geometry and materials should be used to achieve certain acceptable values of stress and displacement. In addition to the geometric restraints we may have restraints on weight, cost, and ease of manufacture among other considerations. The design of a beam in bending is more involved than the axial case because the shape of the cross section, in particular, the area moment of inertia, in addition to the area, has a significant influence on the design. Once again consider one of the simplest of problems, the cantilever beam with a concentrated force on the end that was analyzed in Example 7.3.2. It is shown here in Figure 7.10.1. F L Figure 7.10.1 From static equilibrium we found the bending moment and shear force to be   Mz = F(L − x) − M = 0 Fy = F − V = 0 → M = F(L − x) → V =F (7.10.1) σx = − F (L − x)y Izz τxy (x, y) = F 2Izz  h2 − y2 4  (7.10.2) The displacement is v(x) = F EIzz  x2 Lx3 − 2 6  (7.10.3) Given a particular load and length we note that increasing the moment of inertia Izz decreases both stress components and the displacement. Choosing a material with greater E decreases the displacement but has no effect on the stresses. For a rectangular cross section increasing the height h would appear to increase the stresses; however, this also affects the moment of inertia and so the effect is not entirely clear. Let us examine the effect of different shaped cross sections on a specific beam with a constant cross section and a load as shown in Figure 7.10.1. ########### Example 7.10.1 Problem: The beam in Figure 7.10.1 has a length of 1000 mm and a load of 1600 N. It is made of steel with E = 206800 MPa and a yield stress σ yield = 240 MPa. A maximum normal stress of 160 MPa is Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The normal and shear stresses for a rectangular cross section of height h are P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 310 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods chosen to allow a margin of safety. A rectangular cross section is chosen initially of dimensions as shown in Figure (a). h = 40 mm b = 30 mm Figure (a) Initial calculations show that the normal stress exceeds the maximum allowable value of 160 MPa. The moment of inertia is Izz = 30 · 403 bh 3 = = 160000 mm 4 . 12 12 (a) The maximum stresses are   1600 · 1000 · 20 FL h h =± = ±200 MPa =± σmax = σx 0, 2 Izz 2 160000 τmax F h2 1600 (40)2 = 2.0 MPa = τxy (x, 0) = = 2Izz 4 8 · 160000 (b) and the maximum displacement is vmax = 1600 · (1000)3 FL3 = 16.12 mm = 3EIzz 3 · 206800 · 160000 (c) Refine the cross section size and shape to obtain an acceptable level of stress. Solution: Given the material and the stress find a new cross section shape and size. First let us reduce the normal stress without increasing the amount of material used by changing the dimensions of the rectangular cross section. Let us try a section with the same area where b = 20 mm and h = 60 mm as shown in Figure (b). h = 60 mm --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- b = 20 mm Figure (b) The moment of inertia is now Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Izz = 20 · 603 bh 3 = = 360000 mm 4 12 12 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 311 Classical Analysis of the Bending of Beams The maximum normal stress is now   1600 · 1000 · 30 FL h h = = 133.33 MPa =± σmax = σx 0, 2 Izz 2 360000 (e) This is below the limit we imposed. The displacement is also reduced. vmax = 1600 · (1000)3 FL3 = 7.16 mm = 3EIzz 3 · 206800 · 360000 Let us say that the shear stress is so small that it does not concern us and the displacement is acceptable. We can continue to modify the dimensions of the rectangular section until the right stress levels are achieved. Now let us see if we can actually reduce the amount of material used by changing the shape of the cross section. Let us try an I beam. Let us remove some material to produce the cross section shown in Figure (c). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- c=5 a = 40 h = 60 mm b = 20 mm Figure (c) The moment of inertia now is Izz = (b − c) a 3 20 · 603 15 · 403 bh 3 − = − = 280000 mm 4 12 12 12 12 (f) The material removed amounts to a reduction in the moment of inertia of 80000 mm4 . The stress increases to   1600 · 1000 · 30 FL h h =± = = 171.4 MPa (g) σmax = σx 0, 2 Izz 2 280000 This is close to our target value. In the process we removed half of the material of the beam and saved weight and material cost. Just to make sure let us check the shear stress to see if it is still small.   20  30 V V V ydA y = y5dy + y20dy τmax = τxy (x, 0) = I c I 5 Izz 5 20  zz20 A y  zz 0  (h) 30 1600 1600 y 2  y 2  (200 = + 1000) = 6.86 MPa = + 4 280000 2 0 2 20 280000 It is still small. The tip displacement increases to 9.21 mm. ########### Noting the bending moment distribution for a cantilever beam one might suggest a tapered beam since the normal stress varies linearly from a maximum at the fixed end to zero at the free end. The shear stress, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 312 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods however, is constant over the whole length so we cannot use the bending stress as the sole criterion. Nevertheless, a tapered beam is called for in many cases. The tapered wing of an airplane is a good example. For beams supported at both ends the maximum bending moment is somewhere in the middle and so a beam with greater moment of inertia in the mid section is called for. You see this often in bridge construction. The topic of beams with varying moments of inertia will be considered in the next chapter. ########### Example 7.10.2 Problem: The uniform beam in Example 7.6.1, shown here as Figure (a), has a solid square cross section 42 mm on a side. The length is 1 meter. It is made of steel with a Young’s modulus of 206800 MPa. It may carry a load that produces a normal bending stress of up to 400 MPa. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- (1) Find the maximum acceptable load. (2) Change the rectangular cross section dimensions to cut the maximum normal bending stress in half without increasing the amount of material used. Retain the uniform beam condition and the load found in (1). fy(x) = f0 x Figure (a) Solution: Use equations from Example 7.6.1 which determine the relation between stress and load and between stress and the area moment of inertia. (1) From Example 7.6.1, Equation (g), the maximum moment at x = L/2 is f0 L 2 12 (a) Mmax ymax f0 L 2 h =± Izz 12Izz 2 (b) Mmax = The maximum stress is given by σmax = − where h/2 is half the height of the section. The maximum y has both a plus and minus value. The area moment of inertia is Izz = 42 · 423 bh 3 = = 259308 mm 4 12 12 (c) Solve for the applied load f 0 in Equation (b). f0 = 12Izz σmax 12 · 424 · 400 = 59.27 N/mm = h 12 · 10002 · 21 L2 2 (d) (2) We strive for a value of σmax = 200 MPa. The cross section geometry figures in both the moment of inertia, I zz , and in the height of the section, ymax . To decrease the maximum stress to 200 MPa the we Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 7:5 Printer Name: Yet to Come 313 Classical Analysis of the Bending of Beams must double the value of Izz /ymax . It must then equal 259308 Izz = 24696 mm 3 =2· ymax 21 (e) Consider a rectangular section shown in Figure (b) with the same total area. h Figure (b) To maintain the original cross section area A = bh = 1764 → b= 1764 h (f) The moment of inertia would then be Izz = 1764h 2 bh 3 = = 147h 2 12 12 (g) Thus we find b and h 147h 2 Izz = 24696 = = 294h h ymax 2 → h= 24696 = 84 mm 294 → b= 1764 = 21 mm 84 (h) We can also achieve this value of Izz /ymax with a smaller total area at the same maximum stress level and thus reduce the weight by going to, say, an I beam cross section. ########### 7.11 Large Displacements In all the work up to now we have assumed small displacements. Displacements so small, in fact, that we have assumed that the internal forces and moments can be calculated based upon the location of the applied loads in the undeformed position. We have also assumed that stresses are small enough to stay within the linearly elastic range of the material and the strains are small enough to neglect higher order displacement terms in the strain displacement equations. It is possible to have displacements exceed the values consistent with the applied loads in the undeformed position and still not violate the conditions on stress and strain. Consider the displacement curve of a cantilever beam with a load at the end shown in both an undeflected and a deflected position in Figure 7.11.1. Clearly the bending moment in the beam is less in the deflected position compared to what would be calculated in the undeflected position due to the leftward movement of the load by a distance . There would also be an axial load induced by the Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- b P1: TIX/XYZ JWST071-07 P2: ABC JWST071-Waas July 4, 2011 314 7:5 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods component of the applied load tangent to the beam axis that would induce and axial internal force and displacement. L ∆ Figure 7.11.1 One solution in the literature assumes the axial load is insignificant and determines that the length of the curved beam is the same as the undeflected beam. More recently computer based analysis provides a solution based upon a series of incremental linear solutions. A small increment of the final load is applied and a linear analysis is performed. The deflected shape is used to redefine the new position of the load and the shape of the beam. The load is then increased another increment and the new deflected position is found. In this step both the lateral and axial deformations are calculated to define the location of loads and the current shape of the beam. This is repeated until the final total load is reached. With the development of efficient computer codes this is both rapid and economical. In the past analysis methods for large displacement were inadequate or too expensive for general use. When large displacements were necessary experimental methods were used and they were often time consuming and expensive. The point being made is that it is no longer necessary to depend only upon linear analysis or testing. Beams with large displacement can be used if the situation warrants and the analysis is neither inefficient nor expensive. 7.12 Summary and Conclusions --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- We continue our emphasis on the displacement method. In our displacement method the classical analytical equations are summarized in Equations 7.3.16, repeated here. Mc , Fc → f y (x) dV = − fy dx V (x) M(x) σx (x, y) εx (x, y) dM = −V (x) dx u (x, y) u(x, y) = −y εx (x, y) = v(x) ← ρ dv dx d 2v du = −y 2 dx dx d 2v σx (x, y) = Eεx = −E y 2 dx   2 M(x)y d 2v d v y 2 d A = EIzz 2 → σx = − M(x) = − σx yd A = E 2 Izz dx dx A A  2 d V (x) d v dM = EIzz 2 = −V (x) → τxy (x, y) = ydA y dx dx Izz b A y dx d2 d2 M d 2v = 2 EIzz 2 = f y (x) 2 dx dx dx (7.12.1) Several examples are provided to illustrate the classical analytical method of solution. Once again the beams can be classified as statically determinate or indeterminate; however, both can be solved using the fourth order equation. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 8 A General Method (FEM) for the Bending of Beams --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 8.1 Introduction As we noted in the axial and torsional cases, a new way of organizing the equations of solid mechanics evolved with the advent of the digital computer. We called this the general method because it is easily adapted to all possible materials, geometries, applied loads, and restraints. We have admitted that it is more commonly known as the finite element method or FEM for short. As noted in the axial and torsional cases the method depends upon dividing the structure into segments, called elements, and developing displacement functions for each element in terms of the nodal displacements at the ends of each element. In the process of satisfying Hooke’s law for material properties, equilibrium, and the strain displacement equations we arrive at an element stiffness matrix which relates the internal forces to the nodal displacements. The elements are then assembled into a global stiffness matrix that relates the external applied loads and restraint forces to the nodal displacements. The result is a set of linear algebraic equations that can be solved for the nodal displacements. From these nodal displacements we can work back to the distributed displacements, strains, internal forces, and stresses. We now adapt this method to beams. 8.2 Nodes, Elements, Shape Functions, and the Element Stiffness Matrix At the heart of the finite element method is the representation of the displacement of the element in terms of the nodal displacements and a local coordinate. These displacement functions are called shape functions. The shape functions for a beam were introduced in Example 7.4.2 and presented in Equation 7.4.7 for an element m based on the solution of the differential equations of beam theory applied to a segment of a beam between concentrated loads. Consider the deformed shape of a beam that has been divided into elements as shown in Figure 8.2.1. Let us select one element that we identify as element m between nodes n and n+1. Applied loads, if any, are concentrated loads located at nodes. From the solution of the differential equation for the displacement of such a beam we found that the shape of an element is a cubic polynomial. To represent this cubic polynomial we define a lateral displacement and a rotational displacement at each node. This provides four quantities needed to define the four coefficients of the cubic polynomial. Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 316 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods element m υ(x) node n node n+1 x Figure 8.2.1 Let us isolate this element and define the lateral and rotational displacements at each node and the local coordinate sm as shown in Figure 8.2.2. lm υ(x) υn , θn υn+1, θn+1 sm x Figure 8.2.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- When v (sm ) is the lateral displacement we define the rotation at each node as dvn+1 dvn θn = θn+1 = dsm dsm In matrix form the displacements at each node are     vn vn+1 θn+1 θn (8.2.1) (8.2.2) Thus the nodal displacements for element m are a combination of lateral displacements and rotations at both nodes. ⎡ ⎤ vn ⎢ θn ⎥ ⎥ {rm } = ⎢ (8.2.3) ⎣ vn+1 ⎦ θn+1 We now write the distributed displacement in terms of the nodal displacements where [n] is the matrix of shape functions. v (sm ) = [n] {rm } (8.2.4) The process of obtaining the shape functions is presented in detail in Example 7.4.2. In Equation 7.4.7 we presented ⎡ ⎤ vn ⎢ θn ⎥ 1 ⎥ v(sm ) = 3 lm3 − 3lm sm + 2sm3 lm3 sm − 2lm2 sm2 + lm sm3 3lm sm2 − 2sm3 −lm2 sm2 + lm sm3 ⎢ ⎣ vn+1 ⎦ lm θn+1 (8.2.5) Thus the shape function matrix [n} is [n] = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 1 3 l − 3lm sm + 2sm3 lm3 m lm3 sm − 2lm2 sm2 + lm sm3 3lm sm2 − 2sm3 −lm2 sm2 + lm sm3 (8.2.6) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 317 A General Method (FEM) for the Bending of Beams As before, sm is a local coordinate in the x direction and lm is the length of the element. Each term in the shape function matrix is a cubic polynomial. This is the exact solution to the differential equation for a beam segment between concentrated loads. There are internal forces in the form of bending moments and shear forces at each end of the element. In matrix form at each node for element m we have     Vm,n+1 Vm,n (8.2.7) Mm,n+1 Mm,n Thus the internal force matrix for the element m is ⎡ ⎤ Vm,n ⎢ Mm,n ⎥ ⎥ { pm } = ⎢ ⎣ Vm,n+1 ⎦ Mm,n+1 (8.2.8) These are represented in Figure 8.2.3 for element m. Vm ,n , υn V m , n +1 , υn +1 element m M m,n+1 , θn+1 M m ,n , θ n node n node n+1 Figure 8.2.3 We must note the sign convention used here. In the traditional analysis in Chapter 7 we used the sign convention presented in Figure 8.2.4 for positive forces and moments. See Figure 7.3.5. y V(x+dx) M(x) x M(x+dx) V(x) Figure 8.2.4 At node n+1 the two agree. At node n they are of opposite sign. In Figure 8.2.3 all internal forces are positive in the same direction as the positive coordinate axes and all internal moments are positive by the right hand rule. The purpose is to make the assembly of elements easier. Use the values at the higher node when comparing the FEM results with the traditional analytical solutions. We now note that given Equation 8.2.4, repeated here as Equation 8.2.9, v(sm ) = [n] {rm } (8.2.9) we can find the displacement u (sm , y), assuming plane sections remain plane, in terms of the shape functions and the nodal coordinates. Thus u(sm , y) = −y dv d = −y [n] {rm } dsm dsm (8.2.10) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 318 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods and from the strain displacement equation the strain is εxm (sm , y) = d2 ∂u = −y 2 [n] {rm } ∂sm dsm (8.2.11) Finally from Hooke’s law the stress is σxm (sm , y) = Eεx = −y E d2 [n] {rm } ds2m (8.2.12) From our knowledge of beam theory detailed in Chapter 7 (see Equations 7.3.4 and 7.3.8), it follows that the internal forces at the node n+1, where the sign conventions are the same in both the classical and FEM analysis, are yσx (sm ) dA = E m Im Mm,n+1 = − A d 2 vm d2 = E m Im 2 [n] {rm } 2 dsm dsm (8.2.13) Vm,n+1 dMn+1 d 3 vm d3 = −E m Im 3 = −E m Im 3 [n] {rm } =− dsm dsm dsm At the node n the signs are reversed from the classical values and so the internal forces are Mm,n = yσx (sm ) dA = −E m Im A d2 d 2 vm = −E m Im 2 [n] {rm } 2 dsm dsm (8.2.14) Vm,n dMn d 3 vm d3 = E m Im 3 = E m Im 3 [n] {rm } =− dsm dsm dsm This relatively simple act of defining shape functions turns the problem from one of solving differential equations to that of solving linear algebraic equations. The differential equations are solved when creating the shape functions once and for all for each class of elements we examine. The genius of this approach is that all the differentiation indicated in Equations 8.2.10-14 can be performed explicitly since the shape functions [n]are known. Please note that the strain displacement, Hooke’s law, and equilibrium equations are satisfied without compromise. Given that shape functions in Equation 8.2.6 we note that the distributed displacement is u(sm ) = −y =− d dvm = −y [n] {rm } dsm dsm y −6lm sm + 6sm2 lm3 (8.2.15) lm3 − 4lm2 sm + 3lm sm2 6lm sm − 6sm2 −2lm2 sm + 3lm sm2 {rm } The strain and the normal stress in that element are εxm = d2 ∂u m = −y 2 [n] {rm } ∂sm dsm =− y −6lm + 12sm lm3 −4lm2 + 6lm sm 6lm − 12sm −2lm2 + 6lm sm {rm } (8.2.16) σxm --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ∂u m d2 = E = −y E 2 [n] {rm } ∂sm dsm =− yE −6lm + 12sm lm3 −4lm2 + 6lm sm 6lm − 12sm −2lm2 + 6lm sm {rm } Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 319 A General Method (FEM) for the Bending of Beams As noted the internal forces are evaluated at sm = 0 for node n and at sm = lm for node n+1. Vm,n = E m Im E m Im d3 [n] {r } = 3 [12 lm ds3m Mm,n = −E m Im Vm,n+1 − 12 6lm ] {rm } 6lm E m Im d2 6lm [n] {r } = 3 lm ds2m 4lm2 E m Im d2 6lm [n] {r } = 3 2 lm dsm From Equations 8.2.17 we obtain ⎡ ⎤ Vm,n ⎢ Mm,n ⎥ E m Im ⎥ { pm } = [km ] {rm } = ⎢ ⎣ Vm,n+1 ⎦ = l 3 m Mm,n+1 2l2m {rm } (8.2.17) E m Im d3 = −E m Im 3 [n] {r } = − 3 [12 lm dsm Mm,n+1 = E m Im − 6lm 6lm 2lm2 ⎡ 12 6lm 2 ⎢ 6lm 4l m ⎢ ⎣−12 −6lm 6lm 2lm2 − 12 − 6lm −12 −6lm 12 −6lm 6lm ] {rm } 4lm2 {rm } ⎤⎡ ⎤ 6lm vn ⎥ ⎢ 2lm2 ⎥ ⎥ ⎢ θn ⎥ ⎦ ⎣ −6lm vn+1 ⎦ 4lm2 θn+1 (8.2.18) Thus the element stiffness matrix is ⎡ 12 6lm E m Im ⎢ 6lm 4lm2 ⎢ [km ] = 3 ⎣ −12 −6lm lm 6lm 2lm2 −12 −6lm 12 −6lm ⎤ 6lm 2lm2 ⎥ ⎥ −6lm ⎦ 4lm2 (8.2.19) This element stiffness matrix can be used in all subsequent beam problems that lie in the xy plane. The work we just went through need never be done again. In summary, the path to the element stiffness matrix is { pm } σxm (sm , y) εxm (sm , y) u(sm , y) v(sm ) {rm } v(sm ) = [n]{rm } dvm d = −y [n]{rm } u (sm , y) = −y dsm dsm d2 ∂u m = −y 2 [n]{rm } ∂sm dsm d2 σxm (sm , y) = Eεxm = −y E 2 [n]{rm } dsm dMm,n d3 d2 Vm,n = − = E m Im 3 [n]{r } = yσxm (sm )dA = −E m Im 2 [n]{r } dsm dsm dsm A εxm (sm , y) = Mm,n Mm,n+1 = − yσxm (sm )dA = E m Im A 1 3 l − 3lm sm + 2sm3 lm3 m ⎡ ⎤ Vm,n ⎢ Mm,n ⎥ E m Im ⎥ { pm } = ⎢ ⎣ Vm,n+1 ⎦ = l 3 m Mm,n+1 Vm,n+1 = − lm3 sm − 2lm2 sm2 + lm sm3 [n] = --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS d2 [n]{r } ds2m ⎡ 12 ⎢ 6lm ⎢ ⎣−12 6lm 6lm 4lm2 −6lm 2lm2 −12 −6lm 12 −6lm dMm,n+1 d3 = −E m Im 3 [n]{r } dsm dsm 3lm sm2 − 2sm3 −lm2 sm2 + lm sm3 ⎤ ⎤⎡ 6lm vm,n ⎥ ⎢ 2lm2 ⎥ ⎥ ⎢ θm,n ⎥ = [km ]{rm } (8.2.20) ⎣ ⎦ vm,n+1 ⎦ −6lm 2 θm,n+1 4lm Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas 320 July 11, 2011 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 8.3 The Global Equations and their Solution Refer back to the process of assembly of the element equations to form the global equations for the axial case in Chapter 5, Section 5.3. We summarize the same process for the beam equations. We seek the global equations in the form {F} = { pm } = [km ]{rm } = [K ]{r } (8.3.1) where [K ]is the global stiffness matrix and {r }is the total set of nodal displacements including known restraints and unknown values to be determined. The matrix {F} contains both applied and restraint forces and moments. From equilibrium of forces at each node we can define {F} as ⎡ --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- F1 ⎢ M1 ⎢ ⎢ F2 ⎢ {F} = ⎢ ⎢ M2 ⎢ ... ⎢ ⎣ FN MN ⎤ ⎡ ⎤ V1,1 ⎥ ⎢ ⎥ M1,1 ⎥ ⎢ ⎥ ⎥ ⎢ V1,2 + V2,2 ⎥ ⎥ ⎢ ⎥ ⎥ = ⎢ M1,2 + M2,2 ⎥ = [K ] {r } ⎥ ⎢ ⎥ ⎥ ⎢ ⎥ ... ⎥ ⎢ ⎥ ⎦ ⎣ ⎦ VN MN (8.3.2) As in the axial case we cannot sum the element matrices directly because they do not refer to the same nodal matrix. Thus we reformat each of the element matrices by imbedding them in a format that includes the complete set of nodal displacements. The assembly of the element matrices into a global matrix was described in detail for the axial and torsional cases and is the same process here. It is best shown by example. ########### Example 8.3.1 Problem: Find the stresses and displacements for the beam in Example 7.4.2 by the FE method. The beam is fixed at the left end and has two transverse forces as shown in Figure (a). Let l = L/2. The cross section is a rectangle of height h and width b. 2F F l l L Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 321 A General Method (FEM) for the Bending of Beams Solution: Determine nodes and elements, assign values to the element stiffness matrices, and assemble the global stiffness matrix. Apply boundary conditions, partition, and solve the resulting set of linear algebraic equations. Let us divide the beam into two elements and number the nodes and elements as shown in Figure (b). e1 e2 n1 n2 n3 Figure (b) The process of assembling the global stiffness matrix from the element matrices is to first imbed each element matrix into the global matrix format. You may wish to review the process first outlined in Section 5.3 for the axial case. The size of the global stiffness matrix is determined by the total number of nodal displacements. For this case ⎡ ⎤ v1 ⎢ θ1 ⎥ ⎢ ⎥ ⎢ v2 ⎥ ⎥ {r } = ⎢ (a) ⎢ θ2 ⎥ ⎢ ⎥ ⎣ v3 ⎦ θ3 thus the global stiffness matrix [K ] is of order 6×6. For element 1 ⎡ 12 6l 2 ⎢ 6l 4l ⎢ EIzz ⎢ −12 −6l [k1 ] = 3 ⎢ 2l 2 l ⎢ ⎢ 6l ⎣ 0 0 0 0 −12 −6l 12 −6l 0 0 6l 2l 2 −6l 4l 2 0 0 0 0 0 0 0 0 ⎤ 0 0⎥ ⎥ 0⎥ ⎥ 0⎥ ⎥ 0⎦ 0 (b) For element 2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎡ 0 ⎢0 ⎢ EIzz ⎢0 [k2 ] = 3 ⎢ l ⎢ ⎢0 ⎣0 0 0 0 0 0 0 12 0 6l 0 −12 0 6l 0 0 6l 4l 2 −6l 2l 2 ⎤ 0 0 0 0 ⎥ ⎥ −12 6l ⎥ ⎥ −6l 2l 2 ⎥ ⎥ 12 −6l ⎦ 2 −6l 4l (c) Summing the two, that is, [K ] = [k1 ] + [k2 ], give us the global stiffness matrix. ⎡ 12 6l 2 ⎢ 6l 4l ⎢ ⎢ EIzz −12 −6l [K ] = 3 ⎢ 2l 2 l ⎢ ⎢ 6l ⎣ 0 0 0 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS −12 6l −6l 2l 2 24 0 0 8l 2 −12 −6l 6l 2l 2 ⎤ 0 0 0 0 ⎥ ⎥ −12 6l ⎥ ⎥ −6l 2l 2 ⎥ ⎥ 12 −6l ⎦ −6l 4l 2 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas 322 July 11, 2011 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods We can, of course, solve the assembled algebraic equations analytically; however, it is common to insert values for the various quantities and then to solve numerically using a linear equation solving software program. This is the practical approach when a large number of elements are used. Let us assume that the material is steel with E = 206,800 N/mm2 , the length is L=1000 mm, and the cross section is rectangular with a width b = 20 mm and a height h = 30 mm. The force is F = 250 N. Equation (e) becomes ⎡ ⎡ ⎤ ⎤⎡ ⎤ 12 3000 −12 3000 0 0 0 R1 ⎢3000 1000000 −3000 500000 ⎢ Q1 ⎥ 0 0 ⎥⎢ 0 ⎥ ⎢ ⎢ ⎥ ⎥⎢ ⎥ ⎢ ⎢ ⎥ ⎢ −500 ⎥ 24 0 −12 3000 ⎥ ⎢ ⎥ = 74.448 ⎢ −12 −3000 ⎥ ⎢ v2 ⎥ (f) ⎥ ⎢3000 500000 ⎢ 0 ⎥ ⎥ 0 2000000 −3000 500000 ⎥ ⎢ ⎢ ⎢ θ2 ⎥ ⎢ ⎥ ⎣ 0 ⎣ 250 ⎦ 0 −12 −3000 12 −3000 ⎦ ⎣ v3 ⎦ 0 0 0 3000 We partition Equation (f) to find the displacements. ⎡ ⎡ ⎤ 24 0 −500 ⎢ 0 ⎢ 0 ⎥ 2000000 ⎢ ⎢ ⎥ ⎣ 250 ⎦ = 74.448 ⎣ −12 −3000 3000 500000 0 500000 −12 −3000 12 −3000 Solving with a linear equation solver we get ⎡ ⎤ ⎡ ⎤ v2 0.5597 ⎢ θ2 ⎥ ⎢ 0.003358 ⎥ ⎢ ⎥=⎢ ⎥ ⎣ v3 ⎦ ⎣ 3.3580 ⎦ θ3 0.006716 −3000 1000000 θ3 ⎤⎡ ⎤ 3000 v2 ⎢ θ2 ⎥ 500000 ⎥ ⎥⎢ ⎥ −3000 ⎦ ⎣ v3 ⎦ θ3 1000000 mm rad mm rad (g) (h) Let us compare these displacements with the analytical results found in Example 7.4.2, Equations (g) and (i).   F L3 F L3 L (0) = = 0.560 mm v (L) = = 3.358 mm v =0 v 2 EIzz 48 EIzz 8 θ (0) = dv (0) =0 dx   L dv F L2 2 = = 0.003358 rad dx EIzz 8 (i) 2 dv (L) FL = = 0.006716 rad dx EIzz 4 In matrix format we have from Example 7.4.2 the same values as those from FEM. ⎡ ⎤ ⎡ ⎤ ⎤ ⎡ ⎤ ⎡ v1 0 0 θ1 {v} = ⎣ v2 ⎦ = ⎣ 0.560 ⎦ mm {θ} = ⎣ θ2 ⎦ = ⎣ 0.003358 ⎦ rad θ3 v3 0.006716 3.358 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (j) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- At node 1 there are two known support restraints, v1 = 0 and θ1 = 0, and two unknown reactions, F1 = R1 and M1 = Q 1 . At nodes 2 and 3 there are four known loads, F2 = −2F, M2 = 0, F3 = F, and M3 = 0. The assembled equations are then ⎤⎡ ⎤ ⎡ ⎡ ⎤ 0 12 6l −12 6l 0 0 R1 ⎢ ⎥ ⎢ 6l ⎢ Q1 ⎥ 4l 2 −6l 2l 2 0 0 ⎥ ⎥⎢ 0 ⎥ ⎢ ⎢ ⎥ ⎢ ⎥ ⎢ −2F ⎥ EIzz ⎢−12 −6l 24 0 −12 6l ⎥ ⎥ ⎢ v2 ⎥ ⎢ ⎥ {F} = [K ] {r } = ⎢ (e) 2 2 2 ⎥⎢ ⎥ ⎢ 0 ⎥ = l 3 ⎢ 6l 0 8l −6l 2l 2l ⎥ ⎢ θ2 ⎥ ⎢ ⎢ ⎥ ⎣ ⎦ ⎦ ⎣ 0 ⎣ F ⎦ 0 −12 −6l 12 −6l v3 θ3 0 0 6l 2l 2 −6l 4l 2 0 P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 323 A General Method (FEM) for the Bending of Beams This is as it should be. The method used here is exactly the same since the shape functions are cubic polynomials and the analytical solution of the differential equations for the displacements are also cubic polynomials. The other partitioned equation set for the support reactions is  R1 Q1  = 74.448  −12 −3000 3000 500000 ⎤ v2 ⎥ 0 0 ⎢ ⎢ θ2 ⎥ ⎣ v3 ⎦ 0 0 θ3  ⎡ (k) From which we obtain  R1 Q1  =  250.000 −0.037  N N · mm (l) The exact values from static equilibrium as found in Example 7.4.2 are  R1 Q1  =  250 0  N N · mm (m) We obtained a very small round off error. The value for Q1 = −0.037 is really a very small error for the moments. The shear forces and bending moments are found from Equation 8.2.8 which is repeated here. ⎡ ⎤ Vm,n ⎢ Mm,n ⎥ E m Im ⎥ { pm } = [km ] {rm } = ⎢ ⎣ Vm,n+1 ⎦ = l 3 m Mm,n+1 ⎡ 12 6lm 2 ⎢ 6lm 4l m ⎢ ⎣−12 −6lm 6lm 2lm2 −12 −6lm 12 −6lm ⎤ ⎤⎡ 6lm vn 2 ⎥⎢ 2lm ⎥ ⎢ θn ⎥ ⎥ −6lm ⎦ ⎣ vn+1 ⎦ θn+1 4lm2 (m) Since all the elements are the same length the [km ] matrix is the same for all elements. For element 1: ⎡ ⎤ 12 3000 −12 V1,1 ⎢ ⎢ M1,1 ⎥ ⎥ = 74.448 ⎢3000 1000000 −3000 ⎢ ⎣ −12 −3000 ⎣ V1,2 ⎦ 12 M1,2 3000 500000 −3000 ⎡ ⎤⎡ ⎤ ⎡ ⎤ 3000 0 250.000 N ⎢ ⎥ ⎢ −0.037 ⎥ N · mm 500000 ⎥ 0 ⎥⎢ ⎥=⎢ ⎥ −3000 ⎦ ⎣ 0.5597 ⎦ ⎣ −250.000 ⎦ N 1000000 0.003358 125000 N · mm (n) For element 2 ⎤⎡ ⎤ ⎡ ⎤ 3000 0.5597 −250.000 N ⎢ ⎥ ⎢ ⎥ 500000 ⎥ ⎢ 0.003358 ⎥ ⎢ −125000 ⎥ N · mm ⎥ = −3000 ⎦ ⎣ 3.358 ⎦ ⎣ 250.000 ⎦ N 1000000 0.006716 0.037 N · mm (o) We may also compare the shear forces and bending moments by the two methods. There may be some confusion over signs since the definition of positive shear forces and moments is different for the analytical equations and the FE method. Refer back to Figure 8.2.1 and 8.2.2. In plotting we use the convention for the analytical equations. Thus the signs at node n must be reversed for comparison. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎡ ⎤ 12 3000 −12 V2,2 ⎢3000 1000000 −3000 ⎢ M2,2 ⎥ ⎢ ⎥ ⎢ ⎣ V2,3 ⎦ = 74.448 ⎣ −12 −3000 12 M2,3 3000 500000 −3000 ⎡ P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 324 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods For creating the shear diagram, then we use V1,2 (or – V1,1 ) for the shear force in element 1 and V2,3 (or – V2,2 ) for element 2. For bending moment we use the values -M1,1 M1,2 from element 1 and M2,3 from element 2. This is plotted in Figure (c). 250 Shear, V (N) 0 L -250 125000 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Moment, M (N . mm) 0 Figure (c) These are exactly the same values as for the analytical method. See Example 7.4.2. Note that since the shape functions are cubic polynomials the moments will always be second derivatives or linear polynomials and the shear force will always be third derivatives or constants for each element. This is exact when only concentrated forces are present but will be an approximation when distributed forces are applied as we shall show in the next section. One check on the results is that at each node the sum of the internal forces must be equal to the applied force and the sum of the internal moments must be equal the applied moments. At node 2 −2F = V1,2 + V2,2 0 = M1,2 + M2,2 F = V2,3 At node 3 0 = M2,3 → −500 = −250 − 250 = −500 → 0 = 125000 − 125000 = 0 (p) → 250 = 250 . → 0 = 0.037 = 0 The maximum normal stress is at x = L/2 and y = ±h/2. σx = − Mz y Izz → σmax = ± 125000 · 15 = ±41.67N/MPa 45000 (q) The maximum shear stress is y = 0 and at all values of x. τx y (x, y) = V 2Izz  h2 − y2 4  → τmax = ± 250 (15)2 = ±0.625 MPa 2.45000 (r) ########### At this point the FE method appears to be a lot more work than the analytical method. The next example is easier by FEM. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 325 A General Method (FEM) for the Bending of Beams ########### Example 8.3.2 Problem: Repeat Example 7.6.2 by the FEM. Plot the displacements, shear forces, and bending moments for the beam shown in Figure (a). L Mc Figure (a) Solution: Divide into two elements, assembled the global equations, partition, and solve. The global stiffness matrix is the same as in the previous example. Only the restraints and applied forces have changed. Inserting boundary conditions we get ⎤ ⎡ 12 R1 ⎢ 6l ⎢ Q1 ⎥ ⎥ ⎢ ⎢ ⎢ 0 ⎥ EIzz ⎢−12 ⎥= ⎢ {F} = [K ] {r } = ⎢ ⎢ Mc ⎥ l3 ⎢ ⎥ ⎢ 6l ⎢ ⎣ 0 ⎣ R3 ⎦ Q3 0 ⎡ 6l 4l 2 −6l 2l 2 0 0 6l 2l 2 0 8l 2 −6l 2l 2 −12 −6l 24 0 −12 6l ⎤⎡ ⎤ 0 0 0 ⎢0⎥ 0 0 ⎥ ⎥⎢ ⎥ ⎢ ⎥ −12 6l ⎥ ⎥ ⎢ v2 ⎥ 2 ⎥⎢ −6l 2l ⎥ ⎢ θ2 ⎥ ⎥ 12 −6l ⎦ ⎣ 0 ⎦ −6l 4l 2 0 (a) The partitioned equations and their solution for the displacements are  0 Mc  EIzz = 3 l  24 0 0 8l 2  v2 θ2  →  v2 θ2  ⎡ 0 ⎤ (b) ⎢ ⎥ = ⎣ Mc l ⎦ 8EIzz The nodal shear forces and bending moments are ⎡ ⎡ ⎤ V1,1 12 ⎢ M1,1 ⎥ EIzz ⎢ 6l ⎢ ⎢ ⎥ ⎣ V1,2 ⎦ = l 3 ⎣−12 M1,2 6l Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 6l 4l 2 −6l 2l 2 −12 −6l 12 −6l ⎤⎡ 6l ⎢ 2l 2 ⎥ ⎥⎢ ⎢ ⎦ −6l ⎣ 2 4l 3Mc l 2 ⎢ 4E Izz ⎢ ⎤ ⎢ 0 ⎢ Mc l 3 0 ⎥ EI ⎢ zz ⎢ 4E I zz ⎥ 0 ⎥= 3 ⎢ l ⎢ ⎢ 3Mc l 2 Mc l ⎦ ⎢− ⎢ 4E Izz 8EIzz ⎢ ⎣ Mc l 3 2E Izz ⎡ ⎤ ⎡ 3M c ⎥ ⎥ ⎢ 4l ⎥ ⎢ ⎥ ⎢ Mc ⎥ ⎢ ⎥ ⎢ 4 ⎥=⎢ ⎥ ⎢ 3Mc ⎥ ⎢− ⎥ ⎢ 4l ⎥ ⎣ ⎥ Mc ⎦ 2 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- L/2 P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 326 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 3Mc l 2 ⎢ 4E Izz ⎢ ⎢ ⎤⎡ 0 ⎤ ⎢ Mc l 3 6l ⎢ l M c ⎥ 2 ⎥⎢ 2l ⎥ ⎢ ⎥ EIzz ⎢ ⎢ 2E Izz = ⎥ ⎢ 8EI zz 3 ⎢ −6l ⎦ ⎣ ⎦ l ⎢ 3Mc l 2 0 ⎢− 4l 2 ⎢ 4E Izz 0 ⎢ ⎣ Mc l 3 ⎡ ⎤ ⎡ V2,2 12 6l 2 ⎢ M2,2 ⎥ EIzz ⎢ 6l 4l ⎥ ⎢ ⎢ ⎣ V2,3 ⎦ = l 3 ⎣−12 −6l M2,3 6l 2l 2 ⎡ −12 −6l 12 −6l 4E Izz ⎤ ⎡ 3M c ⎥ ⎥ ⎢ 4l ⎥ ⎢ ⎥ ⎢ Mc ⎥ ⎢ ⎥ ⎢ 2 ⎥=⎢ ⎥ ⎢ 3Mc ⎥ ⎢− ⎥ ⎢ 4l ⎥ ⎣ ⎥ Mc ⎦ 4 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (d) Plots of shear and moment are based on our knowledge that shear forces are constant between nodes and bending moments are linear. This is exactly the same results we obtained in Example 7.6.2 by solving the differential equations. 0 Shear force − L 3M c 4l Mc 2 Moment 0 − Mc 2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (b) The displacements are ⎡ ⎤ 0 v1 ⎢ ⎢ θ1 ⎥ ⎢ 0 ⎢ ⎥ ⎢ 0 ⎢ v2 ⎥ ⎢ ⎢ ⎥ = ⎢ Mc l ⎢ θ2 ⎥ ⎢ ⎢ ⎥ ⎢ 8EIzz ⎣ v3 ⎦ ⎢ ⎣ 0 θ3 0 ⎡ ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (e) To plot the distributed displacement we invoke the shape functions. We have dropped the subscripts from s and l since they are the same values for each element. For element 1 ⎡ ⎤ 0 ⎢ 0 ⎥ 1 ⎢ ⎥ v1 (s) = 3 l 3 − 3ls 2 + 2s 3 l 3 s − 2l 2 s 2 + ls 3 3ls 2 − 2s 3 −l 2 s 2 + ls 3 ⎢ 0 ⎥ ⎣ Ml ⎦ l (f) c 8EI zz    Mc l 1  s3 Mc = 3 −l 2 s 2 + ls 3 = −s 2 + l 8EIzz l 8EIzz Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 327 A General Method (FEM) for the Bending of Beams For element 2 v2 (s) = = 1 3 l − 3ls 2 + 2s 3 l3 l 3 s − 2l 2 s 2 + ls 3 3ls 2 − 2s 3 −l 2 s 2 + ls 3    Mc l 1 3 s3 Mc 2 2 3 2 l s − 2l s + ls = ls − 2s + l3 8EIzz l 8EIzz ⎡ 0 ⎢ Mc l ⎢ ⎢ 8EIzz ⎢ ⎣ 0 0 ⎤ ⎥ ⎥ ⎥ ⎥ ⎦ (g) Plot of displacements (Note the anti symmetry): 0.025 2 Mcl /EIzz -0.025 Figure (c) ########### 8.4 Distributed Loads in FEM Next we will continue with an example only slightly more complicated but where a distributed load is replaced with equivalent nodal loads. The equivalent nodal loads are obtained for each element from lm [n]T f y (sm ) dsm {Fem } = (8.4.1) 0 and the total loading from equivalent nodal loads is {Fe } = {Fem } (8.4.2) As noted before, this expression in Equation 8.4.1 will be derived in Chapter 11. The applied loads consist of both concentrated loads and equivalent nodal loads as expressed by {F} = {Fc } + {Fe } (8.4.3) Note that applied concentrated moments are common but distributed applied moments in bending are rarely encountered and so are not included in this discussion. In any case they all would be included in equivalent nodal loads. The shape functions consist of four cubic polynomials. It follows that there will be four equivalent nodal loads obtained from a distributed lateral load for each element. These will consist of two forces and two moments as follows. ⎤ ⎡ 3 ⎡ ⎤ lm − 3lm sm2 + 2sm3 Fem,n lm ⎢ 3 lm 2 2 3 ⎥ ⎢ Mem,n ⎥ ⎢ lm s − 2lm sm + lm sm ⎥ f y (sm ) ⎥ {Fem } = [n]T f y (sm ) dsm = (8.4.4) ⎥ 3 dsm = ⎢ ⎢ 2 3 ⎣ Fem,n+1 ⎦ ⎦ lm ⎣ 3lm sm − 2sm 0 0 Mem,n+1 −lm2 sm2 + lm sm3 It may seem strange that a lateral distributed load would produce both equivalent lateral concentrated loads and concentrated moments but this is the best possible equivalent load system. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 328 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In summary the FEM equations are {F} f y (sm ) → { pm } {Fc } σxm (sm , y) εxm (sm , y) u (sm , y) l {Fem } = [n]T f y (sm ) dsm {Fe } = 0 {Fem } εxm (sm , y) = −y σxm (sm , y) = −y E d [n]{rm } dsm d2 [n]{rm } ds2m d2 [n]{rm } ds2m d2 [n] {r } ds2m d2 = E m Im 2 [n] {r } dsm d3 [n] {r } ds3m d3 = −E m Im 3 [n] {r } dsm Vm,n = E m Im Mm,n = −E m Im Mm,n+1 v(sm ) = [n]{rm } u (sm , y) = −y {F} = {Fc } + {Fe } v (sm ) {rm } ← {ρ} Vm,n+1 v(sm ) = [n]{rm } = 1 3 l − 3lm sm + 2sm3 lm3 m lm3 sm − 2lm2 sm2 + lm sm3 ⎡ ⎤ Vm,n ⎢ Mm,n ⎥ ⎥ { pm } = [km ] {rm } = ⎢ ⎣ Vm,n+1 ⎦ = Mm,n+1 {F} = { pm } = E m Im 3 lm 3lm sm2 − 2sm3 ⎡ 12 6lm ⎢ 6lm 4lm2 ⎢ ⎣−12 −6lm 6lm 2lm2 −lm2 sm2 + lm sm3 ⎡ ⎤ vm,n ⎢ θm,n ⎥ ⎢ ⎥ ⎣ vm,n+1 ⎦ θm,n+1 ⎤ ⎤⎡ 6lm vm,n ⎥ ⎢ 2lm2 ⎥ ⎥ ⎢ θm,n ⎥ −6lm ⎦ ⎣ vm,n+1 ⎦ θm,n+1 4lm2 −12 −6lm 12 −6lm (8.4.5) [km ]{rm } = [K ] {r } ########### Example 8.4.1 Problem: Repeat Example 7.6.1 as shown in Figure (a) using the FE method. fy (x) = f0 x Figure (a) Solution: Determine nodes and elements, assign values to the element stiffness matrices, and assemble the global stiffness matrix. Find equivalent nodal loads, apply boundary conditions, and solve the resulting set of linear algebraic equations. Let us use four elements. The nodes and elements are shown in Figure (b). e1 n1 e1 n2 e3 n3 e4 n4 n5 Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 329 A General Method (FEM) for the Bending of Beams Let us say that the length of each element is lm = l for a total length of 4l. In this case the equivalent nodal loads on each element are f0l ⎢ ⎡ 3 ⎤ ⎢ 22 lm − 3lm sm2 + 2sm3 ⎢ f0l ⎢ l ⎢ l 3 s − 2l 2 s 2 + l s 3 ⎥ ⎢ 12 m m⎥ ⎢ m m m ⎢ ds = ⎢ ⎥ m ⎢ f0l 2 3 ⎦ 3lm sm − 2sm 0 ⎣ ⎢ ⎢ ⎢ 2 −lm2 sm2 + lm sm3 ⎣ f 2 0l − 12 ⎡ {Fem } = f0 l3 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (a) Note again that while the applied load is only a transverse load the equivalent nodal loads include moments as well as forces. When we imbed just the first element stiffness matrix into the global stiffness matrix we get f0l ⎢ 2 ⎢ ⎢ f0l 2 ⎢ ⎢ 12 ⎢ ⎢ f0l ⎢ ⎢ ⎢ 2 ⎢ f l2 ⎢ 0 ⎢− ⎢ 12 ⎢ ⎢ 0 ⎢ ⎢ 0 ⎢ ⎢ 0 ⎢ ⎢ 0 ⎢ ⎣ 0 ⎡ 0 ⎤ ⎥ ⎥ ⎡ ⎥ 12 6l ⎥ ⎥ 2 ⎢ 6l 4l ⎥ ⎢ ⎥ ⎢ ⎥ ⎢−12 −6l ⎥ ⎢ 6l 2l 2 ⎥ ⎥ EI ⎢ 0 0 zz ⎢ ⎥ ⎥= 3 ⎢ 0 0 ⎥ l ⎢ ⎢ ⎥ ⎢ 0 0 ⎥ ⎢ ⎥ ⎢ 0 ⎥ 0 ⎢ ⎥ ⎣ 0 ⎥ 0 ⎥ ⎥ 0 0 ⎥ ⎦ −12 −6l 12 −6l 0 0 0 0 0 0 6l 2l 2 −6l 4l 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎤⎡ ⎤ v1 0 ⎢ θ1 ⎥ 0⎥ ⎥⎢ ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢ v2 ⎥ ⎥ ⎥ 0⎥ ⎢ ⎢ θ2 ⎥ ⎢ v3 ⎥ 0⎥ ⎥⎢ ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢ θ3 ⎥ ⎥ ⎥ 0⎥ ⎢ ⎢ v4 ⎥ ⎢ ⎥ θ 0⎥ ⎢ 4 ⎥ ⎥ 0⎦ ⎣ v5 ⎦ 0 (b) θ5 When we add the other three element stiffness matrices and enter the rest of the forces and the restraints we get f0l ⎢ 2 ⎢ 2 ⎢ f ⎢ Q 1 + 0l ⎢ 12 ⎢ ⎢ ⎢ f0l ⎢ ⎢ 0 ⎢ ⎢ f0l ⎢ ⎢ 0 ⎢ ⎢ f 0l ⎢ ⎢ 0 ⎢ ⎢ ⎢ ⎢ R + f0l ⎢ 5 2 ⎢ ⎣ f0l 2 Q5 − 12 ⎡ R1 + ⎤ ⎥ ⎥ ⎡ ⎥ 12 6l ⎥ 2 ⎥ ⎢ 6l 4l ⎥ ⎢ ⎥ ⎢−12 −6l ⎥ ⎢ ⎥ ⎢ ⎥ 2l 2 ⎢ 6l ⎥ ⎥ EIzz ⎢ 0 0 ⎢ ⎥= ⎢ ⎥ 3 ⎢ 0 0 l ⎥ ⎢ ⎥ 0 0 ⎢ ⎥ ⎢ ⎥ ⎢ 0 0 ⎥ ⎢ ⎥ ⎣ 0 ⎥ 0 ⎥ ⎥ 0 0 ⎥ ⎦ −12 −6l 6l 2l 2 24 0 0 8l 2 −12 −6l 6l 2l 2 0 0 0 0 0 0 0 0 0 0 0 0 −12 6l −6l 2l 2 24 0 0 8l 2 −12 −6l 6l 2l 2 0 0 0 0 0 0 0 0 0 0 −12 −6l 24 0 0 0 6l 2l 2 0 8l 2 −12 6l −6l 2l 2 ⎤⎡ ⎤ 0 0 0 ⎥⎢ 0 ⎥ ⎥⎢ ⎥ ⎢ ⎥ 0 0 ⎥ ⎥ ⎢ v2 ⎥ ⎥⎢ ⎥ 0 0 ⎥ ⎢ θ2 ⎥ ⎥⎢ ⎥ 0 0 ⎥ ⎢ v3 ⎥ ⎥⎢ ⎥ 0 0 ⎥ ⎢ θ3 ⎥ ⎥⎢ ⎥ −12 6l ⎥ ⎢ v4 ⎥ ⎥⎢ ⎥ −6l 2l 2 ⎥ ⎢ θ4 ⎥ ⎥⎢ ⎥ 12 −6l ⎦ ⎣ 0 ⎦ −6l 4l 2 0 0 0 (c) Notice that the equivalent nodal moments cancel at all interior nodes in this case. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 330 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Partition the global matrix equations to find the nodal displacements. ⎡ ⎤ ⎡ ⎤⎡ ⎤ v2 f0l 24 0 −12 6l 0 0 2 2 ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ θ2 ⎥ 8l −6l 2l 0 0 ⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎢ f 0 l ⎥ EIzz ⎢−12 −6l 24 0 −12 6l ⎥ ⎢ ⎥= ⎢ ⎥ ⎢ v3 ⎥ 2 2 2 ⎢ ⎥ ⎢ 0 ⎥ ⎢ ⎥ 3 0 8l −6l 2l 6l 2l l ⎢ ⎢ ⎥ ⎥ ⎢ θ3 ⎥ ⎣ f0l ⎦ ⎣ 0 0 −12 −6l 24 0 ⎦ ⎣ v4 ⎦ 0 8l 2 θ4 0 0 0 6l 2l 2 The equations for the support reactions are ⎡ ⎤ f0l R + 1 ⎢ 2 ⎥ ⎢ ⎥ ⎡ ⎢ 2 ⎥ f −12 0l ⎥ ⎢ ⎢ Q1 + ⎥ EI ⎢ ⎢ zz ⎢ −6l 12 ⎥ ⎢ ⎥= 3 ⎣ 0 ⎢ f l ⎥ l ⎢ R5 + 0 ⎥ 0 ⎢ ⎥ 2 ⎢ ⎥ ⎣ f0l 2 ⎦ Q5 − 12 6l 2l 2 0 0 0 0 0 0 0 0 0 0 −12 0 0 6l ⎡ ⎤ v2 ⎥ 0 ⎢ ⎢ θ2 ⎥ ⎢ v3 ⎥ 0 ⎥ ⎥⎢ ⎥ ⎥ −6l ⎦ ⎢ ⎢ θ3 ⎥ 2 ⎣ v4 ⎦ 2l θ4 ⎤ (d) (e) Let us try a numerical solution. Let the cross section be a rectangle 20 mm high by 10 mm wide. The element length is l = 200 mm and the material is aluminum with E = 68950 MPa. The loading is f 0 = 1 N/mm. Then with EIzz N = 57.458 3 l mm the equations for the displacement are ⎡ ⎡ ⎤ 24 0 200 ⎢ 0 ⎢ 0 ⎥ 320000 ⎢ ⎢ ⎥ ⎢ ⎢ 200 ⎥ ⎢ ⎥ = 57.458 ⎢ −12 −1200 ⎢1200 80000 ⎢ 0 ⎥ ⎢ ⎢ ⎥ ⎣ 0 ⎣ 200 ⎦ 0 0 0 0 −12 −1200 24 0 −12 1200 (f) 1200 0 80000 0 0 −12 320000 −1200 −1200 24 80000 0 From a linear solver: ⎡ ⎤ ⎡ ⎤ v2 1.3053 ⎢ θ2 ⎥ ⎢ 0.008702 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ v3 ⎥ ⎢ 2.3205 ⎥ ⎢ ⎥=⎢ ⎥ ⎢ θ3 ⎥ ⎢ ⎥ 0 ⎢ ⎥ ⎢ ⎥ ⎣ v4 ⎦ ⎣ 1.3053 ⎦ θ4 −0.008702 The equations for the support restraints are ⎤ R1 + 100 ⎡ ⎥ ⎢ −12 1200 40000 ⎥ ⎢ ⎥ ⎢ Q1 + ⎢ −1200 80000 ⎥ ⎢ 12 ⎥ = 57.458 ⎢ ⎢ ⎣ 0 0 ⎢ R5 + 100 ⎥ ⎥ ⎢ 0 0 ⎦ ⎣ 40000 Q5 − 12 ⎡ 0 0 0 0 0 0 0 0 ⎤⎡ ⎤ 0 v2 ⎢ θ2 ⎥ 0 ⎥ ⎥⎢ ⎥ ⎢ ⎥ 1200 ⎥ ⎥ ⎢ v3 ⎥ ⎥ ⎥ 80000 ⎥ ⎢ ⎢ θ3 ⎥ 0 ⎦ ⎣ v4 ⎦ θ4 320000 mm rad mm rad mm rad (h) ⎡ ⎤ v2 ⎡ ⎤ ⎢ θ2 ⎥ 0 0 −300 N ⎢ ⎥ ⎢ v3 ⎥ ⎢ −50000 ⎥ N · mm 0 0 ⎥ ⎥⎢ ⎥ = ⎢ ⎥ ⎥ ⎣ −300 ⎦ −12 −1200⎦ ⎢ N ⎢ θ3 ⎥ 1200 80000 ⎣ v4 ⎦ 50000 N · mm θ4 ⎤ (i) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (g) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 331 A General Method (FEM) for the Bending of Beams Thus the restraint forces are ⎤ ⎡ ⎤ −400 R1 N ⎢ Q 1 ⎥ ⎢ −53333 ⎥ N · mm ⎥=⎢ ⎢ ⎥ ⎣ R5 ⎦ ⎣ −400 ⎦ N Q5 53333 N · mm ⎡ (j) Let us check these values with the analytical solution in Example 7.5.1. From Equation (e), where L =800 mm, E = 68950 N/mm2 andI = 10 · (20)3 /12 mm 4 . v(x) = f0 (x 4 − 2L x 3 + L 2 x 2 ) 24E Izz (k) Evaluated at the nodes ⎤ ⎡ ⎤ 1.305 v2 ⎢ θ2 ⎥ ⎢ 0.008702 ⎥ ⎥ ⎢ ⎥ ⎢ ⎢ v3 ⎥ ⎢ 2.321 ⎥ ⎥ ⎢ ⎥=⎢ ⎥ ⎢ θ3 ⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎢ ⎣ v4 ⎦ ⎣ 1.305 ⎦ θ4 −0.008072 ⎡ mm rad mm rad mm rad (l) Compared with Equation (h) we see that the displacements at the interior nodes have excellent accuracy for such a limited number of elements. Remember that the displacement which is a fourth order polynomial in the analytical solution (Example 7.6.1, Equation (e)) is replaced by segmented cubic polynomials in each element in the approximate solution. Now let us continue and find the moments and shear forces from which the stresses may be found. Equation 8.2.18 repeated here may be used. ⎤ ⎡ ⎡ ⎤ 12 6lm −12 6lm ⎡ vn ⎤ Vm,n ⎢ Mm,n ⎥ 4lm2 −6lm 2lm2 ⎥ E m Im ⎢ θn ⎥ ⎥⎢ ⎢ 6lm ⎥ ⎥ { pm } = [km ] {rm } = ⎢ (m) ⎥⎢ ⎣ Vm,n+1 ⎦ = l 3 ⎢ ⎣ ⎦ v ⎦ ⎣ −12 −6l 12 −6l n+1 m m m θn+1 Mm,n+1 6lm 2l 2 −6lm 4l 2 m m Since E m , Im , andlm are the same for all the elements the [km ] matrix is the same for all elements. For element 1: ⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤ 12 1200 −12 1200 V1,1 0 −300 N ⎢1200 160000 −1200 80000 ⎥ ⎢ ⎢ M1,1 ⎥ ⎥ ⎢ −50000 ⎥ N · mm 0 ⎢ ⎥ ⎢ ⎥⎢ ⎥=⎢ ⎥ (n) ⎣ V1,2 ⎦ = 57.458 ⎣ −12 −1200 12 −1200 ⎦ ⎣ 1.3053 ⎦ ⎣ 300 ⎦ N M1,2 1200 80000 −1200 160000 0.008702 −10000 N · mm For element 2: ⎤ ⎡ ⎡ 12 V2,2 ⎢ ⎢ M2,2 ⎥ ⎥ = 57.458 ⎢1200 ⎢ ⎣ −12 ⎣ V2,3 ⎦ M2,3 1200 For element 3: ⎡ ⎤ ⎡ 12 V3,3 ⎢1200 ⎢ M3,3 ⎥ ⎢ ⎥ ⎢ ⎣ V3,4 ⎦ = 57.458 ⎣ −12 M3,4 1200 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ⎤⎡ ⎤ ⎡ ⎤ 1200 −12 1200 1.305 −100 N ⎢ ⎥ ⎢ ⎥ 160000 −1200 80000 ⎥ ⎥ ⎢ 0.008702 ⎥ = ⎢ 10000 ⎥ N · mm (o) −1200 12 −1200 ⎦ ⎣ 2.321 ⎦ ⎣ 100 ⎦ N 80000 −1200 160000 0 −30000 N · mm ⎤⎡ ⎤ ⎤ ⎡ 1200 −12 1200 100 2.321 N ⎢ ⎥ ⎢ 30000 ⎥ N · mm 160000 −1200 80000 ⎥ 0 ⎥⎢ ⎥ ⎥=⎢ −1200 12 −1200 ⎦ ⎣ 1.305 ⎦ ⎣ −100 ⎦ N 80000 −1200 160000 −10000 N · mm −0.008702 (p) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-08 P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 332 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods For element 4: ⎡ ⎤ 12 V4,4 ⎢ ⎢ M4,4 ⎥ ⎥ = 57.458 ⎢1200 ⎢ ⎣ −12 ⎣ V4,5 ⎦ M4,5 1200 ⎡ ⎤⎡ ⎤ ⎡ ⎤ 1200 −12 1200 1.305 300 N ⎢ ⎥ ⎢ ⎥ 160000 −1200 80000 ⎥ ⎥ ⎢ −0.008702 ⎥ = ⎢ 10000 ⎥ N · mm (q) ⎦ ⎣ −300 ⎦ −1200 12 −1200 ⎦ ⎣ 0 N 80000 −1200 160000 0 50000 N · mm Remember that, since the shape function in each element is a cubic polynomial and the shear force is the third derivative of the displacements, the shear forces are necessarily constant in each element. These are plotted as dotted horizontal lines in Figure (c) using the values on the right end of each element where the sign convention for the analytic solution agrees with the FEM solution. 400 200 Shear, V 0 (N ) L -200 -400 Figure (c) But we know that the shear force at each restraint is 400 N at the left end and −400 N at the right end, so let us put a point there at each end. Now we know that in the real case with a distributed load there is no jump in the shear stress at any interior node, so at each interior node average the two shear forces and put a point there. Now draw a line, shown as a dashed line, connecting the points. This is the same shear force diagram as obtained in the analytical solution as noted in Equations (r) and (s). From the analytical equations the shear forces at the node locations are ⎤ ⎤ ⎡ 400 V1 ⎢ V2 ⎥ ⎢ 200 ⎥ ⎥ ⎢ ⎥ ⎢ ⎢ V3 ⎥ = ⎢ 0 ⎥ N ⎥ ⎢ ⎥ ⎢ ⎣ V4 ⎦ ⎣ −200 ⎦ −400 V5 ⎡ (r) ⎤ ⎤ ⎡ 400 V1 = −R1 ⎥ ⎢ 200 ⎥ ⎢ V2 ⎥ ⎥ ⎢ ⎢ ⎥=⎢ 0 ⎥N ⎢ V3 ⎥ ⎥ ⎢ ⎢ ⎦ ⎣ −200 ⎦ ⎣ V4 −400 V5 = R5 ⎡ (s) Since the shape functions are cubic polynomials the bending moments, which are the second derivatives of the displacement are necessarily linear polynomials. These also require some interpretation when Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- From the FE equations adjusted by averaging at the nodes and applying support reactions the shear forces are P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 333 A General Method (FEM) for the Bending of Beams compared with the analytical solution. In Equation (t) we list the moment at each node as follows ⎤ ⎡ ⎤ ⎡ ⎤ −M1,1 50000 M1 ⎢ M2 ⎥ ⎢ M1,2 ⎥ ⎢ −10000 ⎥ ⎥ ⎢ ⎥ ⎢ ⎢ ⎥ ⎢ M3 ⎥ = ⎢ M2,3 ⎥ = ⎢ −30000 ⎥ N · mm ⎥ ⎢ ⎢ ⎥ ⎥ ⎢ ⎣ M4 ⎦ ⎣ M3,4 ⎦ ⎣ −10000 ⎦ M5 M4,5 50000 ⎡ (t) At the left end the sign convention for M 11 is the opposite of that for the analytical solution and the sign must be reversed. All the others are calculated at the right end of the element and have the same sign convention as the analytical solutions. These moments are plotted as dotted lines in Figure (d). From the analytical solution the moments at the nodes are given in Equation (u). These are plotted with a smooth curve passing through the points in Figure (d) ⎤ ⎡ ⎤ 53333 M1 ⎢ M2 ⎥ ⎢ −6667 ⎥ ⎥ ⎢ ⎥ ⎢ ⎢ M3 ⎥ = ⎢ −26667 ⎥ N · mm ⎥ ⎢ ⎥ ⎢ ⎣ M4 ⎦ ⎣ −6667 ⎦ 53333 M5 ⎡ (u) What accounts for the differences? In finding the moments from Equation (n) the moments on the elements generated by Equation (a) are not taken into account. These moments are 40000 f0l 2 = = 3333 N · mm 12 12 (v) If these are added to the values for the moments in Equations (n) through (q) the nodal values from the approximate solution are identical to the analytical values. The dashed lines in Figure (d) then move up to coincide with the solid curve at the nodes. FEM codes are programmed to make these adjustments. 5000 0 -3000 Figure (d) We see that the values at the nodes of displacement, shear forces, and moments are remarkably accurate even when a small number of nodes are used. For values between nodes linear interpolation is usually quite satisfactory. h The maximum normal stress is at x = 0 and L, and at y = ± . 2 σx = − Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Mz y 53333 · 10 = ±80 MPa → σmax = ± Izz 6667 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (w) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- L P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 334 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The maximum shear stress is at x = 0 and L, and at y = 0. V τx y (x, y) = 2Izz  h2 − y2 4  → τmax = ± 400 (10)2 = ±3 MPa 2.6667 (x) ########### Now let us do a problem with enough complexity to show off the advantage in using finite elements. ########### Example 8.4.2 Problem: Find the stresses and displacements for a beam that is restrained and loaded as shown. Elements and nodes are numbered as shown in Figure (a). F f0 e1 n1 e2 n2 e3 n3 e4 n4 n5 Figure (a) Solution: Use the FE method. The loads and restraints were conveniently located so that we can divide the beam into four elements of equal length. Let us say that the length of each element is lm = l for a total length of 4l. The equivalent nodal loads for the first element are f0l ⎢− 2 ⎤ ⎡ 3 ⎢ l − 3ls12 + 2s13 ⎢ f0l 2 ⎢− l ⎢ l 3 s − 2l 2 s 2 + ls 3 ⎥ ⎢ 12 ⎢ 1 ⎥ 1 ⎢ ds = ⎥ ⎢ 1 ⎢ f0l 3 2 ⎦ ⎣ 3ls1 − 2s1 0 ⎢− ⎢ 3 2 2 2 ⎢ −l s1 + ls1 ⎣ f0l 2 12 ⎡ {Fe1 } = − f0 l3 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 335 A General Method (FEM) for the Bending of Beams When we imbed just the first element stiffness matrix into the global stiffness matrix we get ⎡ ⎤ f0l − ⎢ 2 ⎥ ⎢ ⎥ ⎢ f l2 ⎥ ⎡ ⎤⎡ ⎤ 0 ⎢− ⎥ v1 12 6l −12 6l 0 0 0 0 0 0 ⎢ ⎥ 2 2 ⎢ 12 ⎥ ⎢ 6l ⎥ ⎢ θ1 ⎥ 4l −6l 2l 0 0 0 0 0 0 ⎢ ⎢ ⎥ ⎥⎢ ⎥ ⎢ f0l ⎥ ⎢−12 −6l 12 −6l 0 0 0 0 0 0⎥ ⎢ v2 ⎥ ⎢− ⎢ ⎥ ⎥⎢ ⎥ 2 2 ⎢ ⎢ ⎥ ⎢ 6l −6l 4l 0 0 0 0 0 0⎥ 2l 2 ⎥ ⎢ ⎢ ⎥ ⎥ ⎢ θ2 ⎥ ⎢ f l 2 ⎥ EIzz ⎢ 0 ⎥ ⎢ v3 ⎥ 0 0 0 0 0 0 0 0 0 0 ⎢ ⎢ ⎥= ⎥⎢ ⎥ ⎢ ⎢ ⎥ ⎥ 0 0 0 0 0 0 0 0 0⎥ l3 ⎢ ⎢ 12 ⎥ ⎢ 0 ⎥ ⎢ θ3 ⎥ ⎢ ⎢ 0 ⎥ ⎥ ⎢ v4 ⎥ 0 0 0 0 0 0 0 0 0 ⎢ 0 ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎢ ⎥ ⎢ 0 ⎥ 0 0 0 0 0 0 0 0 0⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎢ θ4 ⎥ ⎢ ⎥ ⎣ 0 ⎦ 0 0 0 0 0 0 0 0 0 ⎣ v5 ⎦ ⎢ 0 ⎥ ⎢ ⎥ θ5 0 0 0 0 0 0 0 0 0 0 ⎢ 0 ⎥ ⎢ ⎥ ⎣ 0 ⎦ 0 (b) When we add the other three element stiffness matrices and enter the rest of the forces and the restraints we get ⎡ f l ⎤ 0 − ⎢ ⎥ 2 ⎢ ⎥ ⎡ ⎤⎡ ⎤ ⎢ f0l 2 ⎥ v1 12 6l −12 6l 0 0 0 0 0 0 ⎢− ⎥ ⎢ 12 ⎥ 2 2 ⎢ 6l 4l −6l 2l 0 0 0 0 0 0 ⎥⎢ θ ⎥ ⎢ ⎥ ⎢ ⎥⎢ 1 ⎥ ⎢ ⎥ ⎢ ⎢ ⎥ ⎥ v −12 −6l 24 0 −12 6l 0 0 0 0 ⎢ f0l ⎥ ⎢ ⎥⎢ 2 ⎥ ⎢− ⎥ 2 2 2 ⎢ ⎢ ⎥ ⎥ 0 8l θ −6l 2l 0 0 0 0 6l 2l 2 ⎢ ⎥ 2 ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ f 0 l 2 ⎥ EI ⎢ ⎢ 0 0 −12 −6l 24 0 −12 6l 0 0 ⎥⎢ 0 ⎥ ⎢ ⎥ ⎥ ⎢ ⎥ (c) ⎢ 12 ⎥ = 3 ⎢ ⎥⎢θ ⎥ 2 2 2 ⎢ ⎥ l ⎢ 0 0 6l 2l 0 8l −6l 2l 0 0 ⎢ ⎥⎢ 3 ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ R3 ⎥ 0 0 0 0 −12 −6l 24 0 −12 6l ⎢ ⎥ ⎢ v4 ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥ 2 2 2 ⎢ ⎥ 0 8l −6l 2l 0 0 0 0 6l 2l ⎢ ⎥ ⎢ θ4 ⎥ ⎢ 0 ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎣ ⎢ −F ⎥ 0 0 0 0 0 0 −12 −6l 12 −6l ⎦ ⎣ 0 ⎦ ⎢ ⎥ ⎢ 0 ⎥ 0 0 0 0 0 0 6l 2l 2 −6l 4l 2 0 ⎢ ⎥ ⎣ R5 ⎦ Q5 Note that the global stiffness matrix is exactly the same as in Example 8.4.1. This is one of the strengths of the method. The stiffness matrix is independent of the loads and the restraints: thus, a variety of problems can be solved using the same global stiffness matrix as we shall demonstrate. This can be partitioned – first, to find the nodal displacements. ⎡ f l ⎤ 0 − ⎢ 2 ⎥ ⎢ ⎥ ⎡ ⎢ f0l 2 ⎥ 12 6l −12 6l 0 ⎢− ⎥ 2 2 ⎢ 12 ⎥ ⎢ 6l 4l −6l 2l 0 ⎢ ⎥ ⎢ ⎢ f0l ⎥ ⎢ 24 0 6l ⎢− ⎥ EIzz ⎢−12 −6l ⎢ ⎥= ⎢ 6l 0 8l 2 2l 2 2l 2 2 ⎢ ⎥ ⎢ 3 l ⎢ ⎢ f0l 2 ⎥ 0 6l 2l 2 8l 2 ⎢ ⎥ ⎢ 0 ⎢ 12 ⎥ ⎣ 0 0 0 0 −6l ⎢ ⎥ ⎢ 0 ⎥ 0 0 0 0 2l 2 ⎢ ⎥ ⎣ −F ⎦ 0 0 0 0 −6l 24 0 ⎤⎡ ⎤ v1 0 ⎢ θ1 ⎥ 0⎥ ⎥⎢ ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢ v2 ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢ θ2 ⎥ ⎢ ⎥ 2l 2 ⎥ ⎥ ⎢ θ3 ⎥ ⎦ 0 ⎣ v4 ⎦ 8l 2 θ4 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-08 P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 336 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods And then find the support reactions. ⎤ ⎡ 0 R3 EI zz ⎣0 ⎣ R5 ⎦ = 3 l Q5 0 ⎡ 0 −12 0 0 0 0 −6l 0 0 ⎡ ⎤ v1 ⎢ θ1 ⎥ ⎤⎢ ⎥ ⎥ 0 −12 6l ⎢ ⎢ v2 ⎥ ⎥ θ 0 −12 −6l ⎦ ⎢ ⎢ 2⎥ ⎥ θ 0 6l 2l 2 ⎢ 3 ⎢ ⎥ ⎣ v4 ⎦ θ4 (e) Let us try a numerical solution. Let the cross section be a rectangle 20 mm high by 10 mm width. The length is l = 200 mm and the material is aluminum with E = 68950 MPa. The loading is f 0 = −1.2 N/mm and F = −150 N. Then the global stiffness matrix is ⎤ ⎡ 12 1200 1200 −12 0 0 0 0 0 0 0 0 0 0 0 0 ⎥ ⎢1200 160000 −1200 80000 ⎢ −12 −1200 24 0 −12 1200 0 0 0 0 ⎥ ⎥ ⎢ ⎢1200 80000 0 320000 −1200 80000 0 0 0 0 ⎥ ⎥ ⎢ ⎢ 0 0 −12 −1200 24 0 −12 1200 0 0 ⎥ ⎥ [K ] = 57.458 ⎢ ⎢ 0 0 1200 80000 0 320000 −1200 80000 0 0 ⎥ ⎥ ⎢ ⎢ 0 0 0 0 −12 −1200 24 0 −12 1200 ⎥ ⎥ ⎢ ⎢ 0 0 0 0 1200 80000 0 320000 −1200 80000 ⎥ ⎦ ⎣ 0 0 0 0 0 0 −12 −1200 12 −1200 0 0 0 0 0 0 1200 80000 −1200 160000 (c) The nodal forces are ⎤ ⎤ ⎡ ⎤ ⎡ −120 −120 0 ⎢ 0 ⎥ ⎢ −4000 ⎥ ⎢ −4000 ⎥ ⎥ ⎥ ⎢ ⎥ ⎢ ⎢ ⎢ 0 ⎥ ⎢ −120 ⎥ ⎢ −120 ⎥ ⎥ ⎥ ⎢ ⎥ ⎢ ⎢ ⎢ 0 ⎥ ⎢ 4000 ⎥ ⎢ 4000 ⎥ ⎥ ⎥ ⎢ ⎥ ⎢ ⎢ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ ⎥N ⎥ ⎢ ⎥ ⎢ {F} = {Fc } + {Fe1 } = ⎢ ⎥ ⎥=⎢ ⎥+⎢ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ −150 ⎥ ⎢ 0 ⎥ ⎢ −150 ⎥ ⎥ ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎥ ⎥ ⎢ ⎥ ⎢ ⎢ ⎣ 0 ⎦ ⎣ 0 ⎦ ⎣ 0 ⎦ ⎡ --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 0 0 0 (d) The displacement restraints are v3 = v5 = θ5 = 0 (e) The complete global equations with applied forces, restraint forces, and restraints are ⎡ −120 ⎤ ⎡ 12 ⎢ −4000 ⎥ ⎢1200 ⎢ −120 ⎥ ⎢ −12 ⎢ 4000 ⎥ ⎢1200 ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ 0 ⎢ R3 ⎥ ⎢ ⎥ ⎢ ⎢ 0 ⎥ = 57.458 ⎢ ⎢ 0 ⎥ ⎢ ⎢ 0 ⎢ −150 ⎥ ⎢ ⎥ ⎢ ⎢ 0 ⎢ 0 ⎥ ⎣ ⎦ ⎣ 0 R5 Q5 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 0 1200 160000 −1200 80000 −12 −1200 24 0 1200 80000 0 320000 0 0 −12 −1200 0 0 1200 80000 0 −12 −1200 24 0 0 0 0 1200 0 0 80000 0 0 0 −12 1200 320000 −1200 80000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 −12 1200 0 −1200 80000 24 0 0 320000 −12 1200 −1200 80000 0 −12 −1200 12 −1200 ⎤ ⎡ v1 ⎤ θ1 ⎥ ⎥⎢ v2 ⎥ ⎥⎢ ⎢ ⎥ ⎢ θ2 ⎥ ⎥⎢ ⎥ ⎥ ⎥ 0 ⎥⎢ 0 ⎥ ⎥ ⎥⎢ ⎥ 0 ⎥⎢ ⎢ θ3 ⎥ ⎥ 1200 ⎥ ⎢ v4 ⎥ ⎢ ⎥ 80000 ⎥ ⎢ θ4 ⎥ ⎦⎣ ⎦ −1200 0 0 0 0 160000 0 0 (f) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 337 A General Method (FEM) for the Bending of Beams This can be partitioned – first, to find the nodal displacements. ⎡ ⎤ −120 12 1200 −12 ⎢1200 160000 −1200 ⎢ −4000 ⎥ ⎢ ⎢ ⎥ ⎢ −12 −1200 ⎢ −120 ⎥ 24 ⎢ ⎢ ⎥ ⎢ 4000 ⎥ = 57.458 ⎢1200 80000 0 ⎢ ⎢ ⎥ ⎢ 0 ⎥ ⎢ 0 0 1200 ⎢ ⎢ ⎥ ⎣ −150 ⎦ ⎣ 0 0 0 0 0 0 0 ⎡ 1200 80000 0 320000 80000 0 0 0 0 0 0 1200 0 80000 0 320000 −1200 −1200 24 80000 0 ⎤⎡ ⎤ v1 0 ⎢ θ1 ⎥ 0 ⎥ ⎥⎢ ⎥ ⎢ ⎥ 0 ⎥ ⎥ ⎢ v2 ⎥ ⎢ ⎥ 0 ⎥ ⎥ ⎢ θ2 ⎥ (g) ⎢ ⎥ 80000 ⎥ ⎥ ⎢ θ3 ⎥ 0 ⎦ ⎣ v4 ⎦ θ4 320000 Put this in one of the linear solvers and you get ⎡ ⎤ ⎤ ⎡ v1 −12.748442 ⎢ θ1 ⎥ ⎢ 0.038398 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ v2 ⎥ ⎢ −5.24296 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ θ2 ⎥ = ⎢ 0.034917 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ θ3 ⎥ ⎢ 0.014032 ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ v4 ⎦ ⎣ 0.592824 ⎦ −0.003507997 θ4 mm rad mm rad rad mm rad (h) ⎤ ⎡ 0 0 −12 −1200 0 R3 ⎣ R5 ⎦ = 57.458 ⎣0 0 0 0 0 Q5 0 0 0 0 0 ⎡ ⎤ v1 ⎢ ⎥ ⎤ ⎤ ⎢ θ1 ⎥ ⎡ ⎥ 556.875 N −12 1200 ⎢ ⎢ v2 ⎥ ⎥ = ⎣ −166.875 ⎦ N θ −12 −1200⎦ ⎢ 2 ⎢ ⎥ ⎥ 24750 N · mm 1200 80000 ⎢ ⎢ θ3 ⎥ ⎣ v4 ⎦ θ4 ⎡ (i) Now that you know the support reactions and the applied loads you can find the bending moment and shear force diagrams by traditional methods, that is, by static analysis. Or you can find them by using Equations 8.2.18. If you call within the next 5 minutes you can try another loading at virtually no increase in cost. Your equation solver is all fired up and ready to go. Just put in a new load matrix and solve again using the same stiffness matrix. As an example let us extend the distributed load over the first two elements as shown in Figure (b). F f0 Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The support reactions are found from P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 338 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The new load matrix is ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ {F} = {Fc } + {Fe1 } + {Fe2 } = ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ 0 0 0 0 0 0 −150 0 0 0 ⎤ ⎤ ⎡ ⎤ ⎡ −120 0 −120 ⎥ ⎢ −4000 ⎥ ⎢ 0 ⎥ ⎢ −4000 ⎥ ⎥ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ −120 ⎥ ⎢ −120 ⎥ ⎢ −240 ⎥ ⎥ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 4000 ⎥ ⎢ −4000 ⎥ ⎢ 0 ⎥ ⎥ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 0 ⎥ ⎢ −120 ⎥ ⎢ −120 ⎥ ⎥ ⎥=⎢ ⎥+⎢ ⎥+⎢ ⎥ ⎢ 0 ⎥ ⎢ 4000 ⎥ ⎢ 4000 ⎥ ⎥ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ −150 ⎥ ⎥ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎥ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎦ ⎦ ⎣ 0 ⎦ ⎣ 0 ⎦ ⎣ 0 ⎡ ⎤ 0 0 0 (j) The global equations are now −120 ⎤ ⎢ −4000 ⎥ ⎢ −240 ⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎡ 12 ⎢1200 ⎢ −12 ⎢1200 ⎢ ⎡ ⎢ ⎥ ⎢ ⎢R3 − 120⎥ ⎢ ⎥ = 57.458 ⎢ ⎢ ⎢ 4000 ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ −150 ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ 0 ⎥ ⎣ ⎣ ⎦ R5 Q5 1200 160000 −1200 80000 −12 −1200 24 0 1200 80000 0 320000 0 0 −12 −1200 0 0 1200 80000 0 0 −12 −1200 24 0 0 0 0 0 0 0 1200 0 0 80000 0 0 0 −12 1200 320000 −1200 80000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 −12 1200 −1200 80000 24 0 0 320000 −12 1200 −1200 80000 0 0 0 0 0 0 −12 −1200 12 −1200 ⎤⎡ v1 ⎤ θ1 ⎥ ⎥⎢ v2 ⎥ ⎥⎢ ⎢ ⎥⎢ θ2 ⎥ ⎥⎢ ⎥ ⎥ ⎥ 0 ⎥⎢ 0 ⎥ ⎥ ⎥⎢ 0 ⎥⎢ θ3 ⎥ ⎢ ⎥ 1200 ⎥ v4 ⎥ ⎥⎢ ⎢ ⎥ 80000 ⎥⎢ θ4 ⎥ ⎦⎣ ⎦ −1200 0 0 0 0 160000 0 0 (k) Note that the equivalent nodal load at node 3 must be added to the restraint there. The partitioned equations for the displacement are ⎡ ⎡ ⎤ 12 −120 ⎢ −4000 ⎥ ⎢1200 ⎢ ⎢ ⎥ ⎢ −240 ⎥ ⎢ −12 ⎢ ⎢ ⎥ ⎢ 0 ⎥ = 57.458 ⎢1200 ⎢ ⎢ ⎥ ⎢ 0 ⎢ 4000 ⎥ ⎢ ⎢ ⎥ ⎣ 0 ⎣ −150 ⎦ 0 0 1200 −12 160000 −1200 −1200 24 80000 0 0 1200 0 0 0 0 1200 80000 0 320000 80000 0 0 0 0 0 0 1200 0 80000 0 320000 −1200 −1200 24 80000 0 ⎤⎡ ⎤ v1 0 ⎢ ⎥ 0 ⎥ ⎥ ⎢ θ1 ⎥ ⎢ ⎥ 0 ⎥ ⎥ ⎢ v2 ⎥ ⎢ ⎥ 0 ⎥ ⎥ ⎢ θ2 ⎥ (l) ⎥ ⎥ 80000 ⎥ ⎢ ⎢ θ3 ⎥ 0 ⎦ ⎣ v4 ⎦ θ4 320000 Run the solver again. ⎡ ⎤ ⎡ ⎤ v1 −16.055206 ⎢ θ1 ⎥ ⎢ ⎥ 0.0471 ⎢ ⎥ ⎢ ⎥ ⎢ v2 ⎥ ⎢ −6.809322 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ θ2 ⎥ = ⎢ 0.043619 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ θ3 ⎥ ⎢ 0.019253 ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ v4 ⎦ ⎣ 0.853885 ⎦ −0.004813298 θ4 mm rad mm rad rad mm rad ⎤ ⎡ ⎤ 766.875 N R3 − 120 ⎦ = ⎣ −256.875 ⎦ ⎣ R5 N Q5 0.0003675 N · mm ⎡ (m) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- That displacement at the left end may be a little too large. For a slight increase in shipping and handling you can put a restraint there. If we keep the distributed load on the first two elements we have the loads Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 339 A General Method (FEM) for the Bending of Beams and restraints shown here in Figure (c). F f0 Figure (c) The global equations are now 1 − 120 ⎤ ⎡ 12 ⎢ −4000 ⎥ ⎢1200 ⎥ ⎢ ⎢ −12 ⎢ −240 ⎥ ⎢ ⎢ 0 ⎥ ⎢1200 ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ 0 ⎢ R3 − 120⎥ ⎥ =57.458 ⎢ ⎢ ⎢ 0 ⎢ 4000 ⎥ ⎢ ⎥ ⎢ ⎢ 0 ⎢ −150 ⎥ ⎢ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ ⎣ ⎦ ⎣ 0 0 R5 Q5 1200 −12 1200 0 0 0 0 0 160000 −1200 80000 −1200 24 0 80000 0 320000 0 −12 −1200 0 1200 80000 0 0 0 0 0 0 0 0 0 0 −12 −1200 24 0 −12 1200 0 0 0 0 1200 0 0 80000 0 0 0 −12 1200 320000 −1200 80000 −1200 24 0 80000 0 320000 0 −12 −1200 0 0 0 0 0 0 0 0 0 0 −12 1200 −1200 80000 12 −1200 ⎤⎡ 0 ⎤ θ1 ⎥ 0 ⎥⎢ ⎥⎢ ⎥ 0 ⎥⎢v2 ⎥ ⎢ θ2 ⎥ 0 ⎥ ⎥ ⎥⎢ ⎢ ⎥ ⎥ 0 ⎥⎢ 0 ⎥ ⎢ ⎥ ⎥ 0 ⎥ ⎥⎢ ⎢ θ3 ⎥ 1200 ⎥⎢v4 ⎥ ⎥⎢ ⎥ 80000 ⎥⎢θ4 ⎥ ⎦ −1200 ⎣ ⎦ 0 160000 0 0 (n) From the previous example you must take out the first row of all matrices and also the first column of the stiffness matrix. The partitioned equations for the displacements are ⎡ ⎡ ⎤⎡ ⎤ ⎤ θ1 160000 −1200 80000 0 0 0 −4000 ⎢ ⎥ ⎢ −1200 ⎢ −240 ⎥ 24 0 1200 0 0 ⎥ ⎢ ⎢ ⎥ ⎢ v2 ⎥ ⎥ ⎢ ⎥ ⎢ 80000 ⎢ 0 ⎥ 0 320000 80000 0 0 ⎥ ⎢ ⎢ ⎥ ⎢ θ2 ⎥ ⎥ (o) ⎢ 4000 ⎥ = 57.458 ⎢ 0 ⎥ ⎢ θ3 ⎥ 1200 80000 320000 −1200 80000 ⎢ ⎢ ⎥⎢ ⎥ ⎥ ⎣ ⎣ 0 ⎣ −150 ⎦ ⎦ ⎦ 0 0 −1200 24 0 v4 0 0 0 80000 0 320000 0 θ4 Solving, the displacements are ⎡ The support reactions are ⎤ ⎡ ⎤ θ1 −0.004506397 ⎢ v2 ⎥ ⎢ −0.501919 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ θ2 ⎥ ⎢ 0.0006138024 ⎥ ⎢ ⎥=⎢ ⎥ ⎢ θ3 ⎥ ⎢ 0.002051188 ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ v4 ⎦ ⎣ −0.00621572 ⎦ θ4 −0.0005127969 rad mm rad rad mm rad ⎡ ⎤ −0.0045064 ⎡ ⎤ ⎤ ⎢ −0.50192 ⎥ 77.6786 1200 −12 1200 0 0 0 R1 − 120 ⎢ ⎥ −12 −1200 0 −12 1200 ⎥⎢ 0.0006138 ⎥ ⎢ 272.6786 ⎥ ⎢ 0 ⎢ R3 − 120 ⎥ ⎢ ⎥= ⎦ = 57.458 ⎣ 0 ⎣ R 0 0 0 −12 −1200⎦⎢ 0.0020512 ⎥ ⎣ 39.64286 ⎦ N 5 ⎣ −0.0062157 ⎦ −2785, 71 N 0 0 0 0 1200 80000 Q5 ⎡ ⎤ ⎡ (p) N N · mm · mm −0.0005128 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (q) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎡R P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 340 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Given the displacements we find the moments and shear forces using Equation 8.2.18, repeated here. ⎡ ⎤ ⎡ ⎤⎡ ⎤ Vm,n −12 6lm 12 6lm vn ⎢ Mm,n ⎥ ⎥ ⎢ E m Im ⎢ 4lm2 −6lm 2lm2 ⎥ ⎢ 6lm ⎥ ⎢ θn ⎥ ⎥ { pm } = [km ] {rm } = ⎢ (r) ⎣ Vm,n+1 ⎦ = l 3 ⎣−12 −6lm ⎣ ⎦ vn+1 ⎦ 12 −6lm m θn+1 6lm 2lm2 −6lm 4lm2 Mm,n+1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Let us use the values from the first case considered, that is, the distributed load on the first element. Note that in the following symbols for the shear forces and moment we use two subscripts – the first for the element and the second for the node. Remember there are two values of shear force and bending moment at each interior node – one for each element at that node. ⎡ ⎤⎡ ⎤ ⎡ ⎤ ⎤ ⎡ 12 1200 −12 1200 −119.96 −12.748442 V1,1 N ⎢1200 160000 −1200 80000 ⎥ ⎢ 0.038398 ⎥ ⎢ −3995.537 ⎥ N · mm ⎢ M1,1 ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎥=⎢ ⎣ V1,2 ⎦ = 57.458 ⎣ −12 −1200 12 −1200 ⎦ ⎣ −5.24296 ⎦ ⎣ 119.96 ⎦ N M1,2 1200 80000 −1200 160000 −19996.44 N · mm 0.034917 (s) We are picking up a little calculation round off error here. The exact answers based on this model would be ⎤ ⎡ ⎤ ⎡ −120 N V1 ⎢ M1 ⎥ ⎢ −4000 ⎥ N · mm ⎥ ⎢ ⎥ ⎢ (t) ⎣ V2 ⎦ = ⎣ 120 ⎦ N M2 −20000 N · mm Using the same matrix equation adapted to the second, third, and fourth elements we get ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ −239.986 −240 N V2,2 ⎢ M2,2 ⎥ ⎢ 24001.82 ⎥ . ⎢ 24000 ⎥ N · mm ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎣ V2,3 ⎦ = ⎣ 239.986 ⎦ = ⎣ 240 ⎦ N M2,3 −71999.01 −72000 N · mm (u) ⎤ ⎡ ⎤ ⎡ ⎤ 316.876 317 N V3,3 ⎢ M3,3 ⎥ ⎢ 72000.13 ⎥ . ⎢ 72000 ⎥ N · mm ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎣ V3,4 ⎦ = ⎣ −316.876 ⎦ = ⎣ −317 ⎦ N M3,4 8624.979 8625 N · mm ⎡ ⎤ ⎡ ⎤ ⎤ ⎡ 167 166.875 N V4,4 ⎢ M4,4 ⎥ ⎢ 8624.979 ⎥ . ⎢ 8625 ⎥ N · mm ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎣ V4,5 ⎦ = ⎣ −166.875 ⎦ = ⎣ −167 ⎦ N M4,5 24750 N · mm 24749.98 ⎡ (v) (w) Now let us check these out. Compare the values at the nodes with the support reactions. You see that they agree, that is, the sum of the shear force at node 3 from elements 2 and 3 is equal to the restraint force. Also the shear and moment at node five from the element agree with the restraint values. The maximum moment is at node 3 and so the maximum normal stress is σx,max = − M ymax 72000 · 10 · 12 = ±108 MPa =± Izz 80000 (x) and the maximum shear force is in element 3 and so the maximum shear stress is τmax = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS V h2 317 · 100 · 12 = −2.3775 MPa =− I 8 80000 · 2 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (y) P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 341 A General Method (FEM) for the Bending of Beams This may look like quite a large amount of work. Actually, in the equation solving software you have to enter only a few matrices. Many of these are used several times, so, it really is not that bad. Try solving this problem by dividing the beam into segments, solving the differential equation in each segment, and then satisfying the boundary conditions. You will soon become a convert to this method. We should note that this method, when it requires replacing distributed loads with equivalent nodal loads, does introduce a measure of approximation into the solution. For example, we appear to have a non zero shear force at node one when, in fact, the shear force is zero there. This requires careful interpretation of the results as we have noted in Example 8.2.2. In practice this often is handled by using a large number of elements, thereby approaching the exact answer at each node with less interpretation needed. Remember that there are FEM computer codes that do most of this work for you. We just wanted you to see what is going on inside those codes. ########### 8.5 Variable Cross Sections It is common practice with tapered beams to break them down into a set of elements with uniform cross sections similarly to what we did in Chapter 5, Section 5.5, for the axially loaded case. An example will help to explain. ########### Example 8.5.1 Problem: A beam has a rectangular cross section with a linear taper in width as shown in Figure (a) and a uniform height h. It has a uniform load fy (x) = f 0 as shown in Figure (b). Find the displacement and stresses. b(x) = b0(1-x/2L) b0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- x L z Figure (a) y f0 h x L Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 342 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Solution: Model the beam with two elements of constant cross section each. Assemble the global stiffness matrix, find the equivalent nodal loads, assign boundary restraints, partition the equations, and solve. First we shall show the bar divided into only two elements, as in Figure (c). In reality more elements should be used to more closely approximate the exact answer but this will be sufficient to demonstrate the process. The value of Izz at x = L/ 4 will be assigned to element 1 and that at x = 3L/ 4 to element 2. 3L/4 L/4 e1 e2 n1 n2 n3 Figure (c) The moments of inertia are based on the cross sections at x=L/ 4 and x=3L/ 4.         7 5 L 3L 3L L = b0 1 − = b0 b = b0 1 − = b0 b 4 4 · 2L 8 4 4 · 2L 8 (a) The element moments of inertia are then 7 I0 8 I2 = 5 I0 8 The stiffness matrix for element 1 imbedded in global format is ⎡ 84 21L −84 21L 2 2 ⎢ 21L 7L −21L 3.5L ⎢ EI0 ⎢ −84 −21L 84 −21L [k1 ] = 3 ⎢ 2 −21L 7L 2 L ⎢ ⎢ 21L 3.5L ⎣ 0 0 0 0 0 0 0 0 The stiffness matrix for element 2 imbedded in global format is ⎡ 0 0 0 0 0 ⎢0 0 0 0 0 ⎢ EI0 ⎢ 0 0 60 15L −60 [k2 ] = 3 ⎢ −15L 5L 2 L ⎢ ⎢0 0 15L ⎣0 0 −60 −15L 60 0 0 15L 2.5L 2 −15L The equivalent nodal loads for each element are Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS f0 l3 0 0 0 0 0 0 ⎤ 0 0⎥ ⎥ 0⎥ ⎥ 0⎥ ⎥ 0⎦ 0 (c) ⎤ 0 0 ⎥ ⎥ 15L ⎥ ⎥ 2.5L 2 ⎥ ⎥ −15L ⎦ 5L 2 ⎤ ⎡ f0 L f0l ⎢ 2 ⎥ ⎢ 4 ⎢ ⎥ ⎢ ⎤ ⎡ 3 ⎢ f0l 2 ⎥ ⎢ f0 L 2 l − 3ls 2 + 2s 3 ⎢ ⎥ ⎢ l ⎢ 3 2 2 ⎢ ⎥ ⎢ s + ls 3 ⎥ ⎥ds = ⎢ 12 ⎥ = ⎢ 48 ⎢ l s − 2l 2 3 ⎢ f0l ⎥ ⎢ f0 L ⎦ ⎣ 3ls − 2s 0 ⎢ ⎥ ⎢ ⎢ 2 ⎥ ⎢ 4 −l 2 s 2 + ls 3 ⎢ ⎥ ⎢ ⎣ f0l 2 ⎦ ⎣ f0 L 2 − − 48 12 ⎡ {Fem } = (b) (d) ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (e) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- I1 = P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 343 A General Method (FEM) for the Bending of Beams The assembled global equations including restraints are ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢− ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ f0 L 4 f0 L 2 Q1 + 48 R1 + ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎡ ⎥ 84 ⎥ ⎥ ⎢ 21L ⎢ f0 L ⎥ f0 L ⎥ + ⎥ EI0 ⎢ ⎢−84 4 4 ⎥= ⎢ ⎥ L 3 ⎢ 21L 2 ⎥ 2 f0 L ⎥ f0 L ⎢ ⎣ 0 + 48 48 ⎥ ⎥ 0 ⎥ ⎥ f0 L ⎥ ⎥ 4 ⎥ ⎦ 2 f0 L − 48 21L 7L 2 84 −21L 21L 3.5L 2 0 0 −21L 3.5L 2 0 0 84 + 60 −21L + 15L −60 15L −21L + 15L 7L 2 + 5L 2 −15L 2.5L 2 −60 −15L 60 −15L ⎤⎡ ⎤ 0 0 ⎢ ⎥ 0 ⎥ ⎥⎢ 0 ⎥ ⎥⎢ ⎥ ⎢ ⎥ 15L ⎥ ⎥ ⎢ v2 ⎥ 2⎥ ⎢ 2.5L ⎥ ⎢ θ2 ⎥ ⎥ −15L ⎦ ⎣ v3 ⎦ 5L 2 θ3 (f) Partition and solve as we have been doing. ########### As we noted before, a cantilever beam tapered in depth is suggested since the bending moment varies linearly from a maximum at the fixed end to zero at the free end. A tapered beam with half the height at the free end and constant width was considered as shown in Figure 8.5.1 h(x) = h0(1-x/2L) h0 x L F Figure 8.5.1 This, of course, removes some material compared to a uniform cross section and thus reduces the weight. The shear force is constant; however, the shear stress is highest at the free end and decreases as the cross section moment of inertia increases toward the fixed end. This has been solved with a numerical FE code and in Figure 8.5.2 we show the stress contours for the bending stress, σ x . The contour plots of shear stress in Figure 8.5.3 show more variation. This shows a more efficient use of the material as far as bending goes. It has been noted that while the shear force is constant the shear stress is much higher at the free end than at the fixed end because the moment of inertia is smaller. In fact, the peak value at y = 0 is double the value on the tapered beam compared to the uniform beam. Although the contour plots do not make this clear the distribution is nearly parabolic all along the length. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas 344 July 11, 2011 17:48 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure 8.5.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Figure 8.5.3 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come 345 A General Method (FEM) for the Bending of Beams 8.6 Summary and Conclusions As the problems become more complex the classical method becomes cumbersome. A general (finite element) method based on the definition of shape functions is then developed. The equations for this method are summarized in Equation 8.4.5 and repeated here. {F} f y (sm ) → { pm } {Fc } σxm (sm , y) εxm (sm , y) u (sm , y) v (sm ) {rm } ← {ρ} l {Fem } = [n]T f y (sm ) dsm {Fe } = 0 {Fem } v(sm ) = [n]{rm } u (sm , y) = −y {F} = {Fc } + {Fe } εxm (sm , y) = −y σxm (sm , y) = −y E d2 [n] {r } ds2m d2 = E m Im 2 [n] {r } dsm d [n]{rm } dsm d2 [n]{rm } ds2m d2 [n]{rm } ds2m d3 [n] {r } ds3m d3 = −E m Im 3 [n] {r } dsm Mm,n = −E m Im Vm,n = E m Im Mm,n+1 Vm,n+1 ν(sm ) = [n]{rm } = 1 3 l − 3lm sm + 2sm3 lm3 m ⎡ lm3 sm − 2lm2 sm2 + lm sm3 ⎤ Vm,n ⎢ Mm,n ⎥ E m Im ⎥ { pm } = [km ] {rm } = ⎢ ⎣ Vm,n+1 ⎦ = l 3 m Mm,n+1 {F} = ⎡ 12 6lm 2 ⎢ 6lm 4l m ⎢ ⎣−12 −6lm 6lm 2lm2 { pm } = 3lm sm2 − 2sm3 −12 −6lm 12 −6lm −lm2 sm2 + lm sm3 ⎤ ⎤⎡ 6lm vm,n ⎥ ⎢ 2lm2 ⎥ ⎥ ⎢ θm,n ⎥ −6lm ⎦ ⎣ vm,n+1 ⎦ θm,n+1 4lm2 ⎡ ⎤ νm,n ⎢ θm,n ⎥ ⎢ ⎥ ⎣νm,n+1 ⎦ θm,n+1 [km ]{rm } = [K ] {r } --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (8.6.1) P1: TIX/XYZ JWST071-08 P2: ABC JWST071-Waas July 11, 2011 17:48 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 9 More about Stress and Strain, and Material Properties 9.1 Introduction --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- For the most part we have presented the results for stress and strain with respect to a fixed rectangular Cartesian coordinate system. One exception is torsion where we have used cylindrical coordinates. For the axially loaded slender bar the state of stress has only one non zero component, σ x , with respect to the chosen coordinate system. Briefly we did note that shear stress is present in an axially load bar if the axes of reference are rotated with respect to the original axes. We must ask whether changes in orientation of the axes will produce different states of stress in other cases. States of stress that we have examined so far include a rectangular solid loaded by uniform hydrostatic pressure in Chapter 2, Example 2.6.2, resulting in a three dimensional state of stress. We have examined briefly rectangular plates with uniform and linear planar edge loads in Chapter 3, Example 3.5.1, and in Chapter 4, Section 4.2, and concluded that the stresses in the interior matched the edge loads. In Chapter 6 we examined the state of shear stress in torsion and in Chapter 7 the state of both normal and shear stress in bending. If we put these together in various combinations, and we can by the principle of superposition, we can solve for some quite complicated states of stress. Several cases of this type are examined briefly in Chapter 10. It is time to extend our discussion to a more general state of stress and strain and examine what happens when different orientations of the axes are used. A restriction imposed until this chapter is to consider that all stresses and strains are below the proportional limit of a homogenous, isotropic, linearly elastic solid except for Chapter 3, Section 3.8, where fiber reinforced composite laminates were very briefly considered. Might it be that the stresses with respect to the axes used so far are not the ones that must be used to satisfy our restrictions on material properties? And are there other material properties that must be considered in the analysis and design of solid bodies? All this and more will be coming right up. 9.2 Transformation of Stress in Two Dimensions The stress components we find with respect to the chosen rectangular Cartesian coordinates in any case tell only part of the story. Let us reexamine what happens in an axially loaded bar when the axes are rotated as shown in Figures 4.8.1 and 4.8.2, repeated here in Figure 9.2.1. Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 348 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y n t F θ x Pn , σ n P, σ x Pt ,τ nt Figure 9.2.1 We saw briefly in Chapter 4, Section 4.8, that if we consider a coordinate system nt with rotation θ about the z axis the stress components normal and tangential to the face normal to the n axis are P P cos2 θ = σx cos2 θ (9.2.1) τnt = − sin θ cos θ = −σx sin θ cos θ A A In Figure 9.2.2 (repeat of Figure 4.8.3) we plot the variation in the normal stress and shearing stress as the angle changes. σn = 1.0 σn σ 0.5 0 45 90 135 180 0.5 τ nt τ 0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- -0.5 Figure 9.2.2 The upper plot is σn = cos2 θ σx (9.2.2) τnt = sin θ cos θ τx y (9.2.3) σn = σx = σmax (9.2.4) and the lower plot is We note that when θ = 0◦ then Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 349 More about Stress and Strain, and Material Properties and as θ increases the normal stress decreases until it reaches zero at θ = 90◦ . The shear stress is zero at θ = 0◦ and θ = ±90◦ and reaches a maximum at θ = ±45◦ or τmax = ± P 2A (9.2.5) Now let us consider the general case of a two dimensional stress at a point where there are non zero components ⎡ ⎤ σx {σ } = ⎣ σ y ⎦ (9.2.6) τx y and zero components σz = τzx = τ yz = 0 (9.2.7) In Figure 9.2.3(a) we show symbolically the stresses at a point with respect to the xy axes and in Figure 9.2.3(b) we show them with respect to the sr axes which are rotated an amount θ. σy σr y τsr σs τxy σx σx r s σs τxy τsr σr σy θ x --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- (a) (b) Figure 9.2.3 We wish to express the stress components with respect to the rotated sr axes in terms of the components of stress with respect to the xy axes. Let us find the equilibrium of the stresses at a point as shown in Figure 9.2.4. τsr σs σx τxy σy Figure 9.2.4 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 350 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Let the area of the face on which σs acts be A. Then on the face normal to the x axis the area is A cos θ and on the y normal face the area is A sin θ. Equilibrium of forces in the r and s direction provides  Fs = σs A − τx y A sin θ cos θ − σ y A sin θ sin θ − τx y A cos θ sin θ − σx A cos θ cos θ = 0 → σs = σx cos2 θ + σ y sin2 θ + τx y 2 sin θ cos θ  (9.2.8) Fr = τsr A + τx y A sin θ sin θ − σ y A sin θ cos θ − τx y A cos θ cos θ + σx A cos θ sin θ = 0 → τsr = (σ y − σx ) sin θ cos θ + τx y (cos2 θ − sin2 θ) (9.2.9) Using the trigonometric identities sin2 θ = (1 − cos 2θ )/2 and sin 2θ = 2 sin θ cos θ, cos2 θ = (1 + cos 2θ )/2 (9.2.10) we can write σx − σ y σx + σ y + cos 2θ + τx y sin 2θ 2 2 σx − σ y sin 2θ + τx y cos 2θ =− 2 σs = (9.2.11) τr s (9.2.12) We can get the normal stress in the r direction by substituting θ + 90◦ for θ in the expression for σs . Thus σx − σ y σx + σ y − cos 2θ + τx y sin 2θ 2 2 (9.2.13) Now that we can find the state of stress for any orientation of the axes of particular interest are rotations of axes that find the maximum values of normal and shear stress components. That will be treated in the next section. 9.3 Principal Axes and Principal Stresses in Two Dimensions If we differentiate Equation 9.2.11 with respect to θ and set the result to zero we can find the angle for maximum and minimum normal stresses. This angle identifies axes called the principal axes and the stresses with respect to those axes are called the principal stresses. σx − σ y dσs =− 2 sin 2θ + τx y 2 cos 2θ = 0 dθ 2 (9.3.1) Note that we get the same result by setting the shear stress to zero, that is, from Equation 9.2.12 τr s = − σx − σ y sin 2θ + τx y cos 2θ = 0 2 (9.3.2) From either Equation 9.3.1 or 9.3.2 we obtain tan 2θ p = τx y (σx − σ y )/2 (9.3.3) We have added the subscript p to indicate that this is the rotation necessary to define a principal axis. To find the rotation for which the normal stresses are a maximum or a minimum the following Figure 9.3.1 is helpful. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- σr = P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 351 More about Stress and Strain, and Material Properties 2  σ x −σ y  2   + τ xy  2  τ σ x −σ y 2 2θp2 τxy 2θp1 σ σ x −σ y τxy 2 Figure 9.3.1 There are two different rotations that are 90o apart that define the axes of principal stress. Given tan2θ p from the figure you can identify the sine and cosine of 2θ p. sin 2θ =  τx y σx − σ y 2 cos 2θ =  2 + τx2y If we substitue these into Equation 9.2.11 we obtain  σx + σ y σx − σ y σ p max = + 2 2 2 for the rotation that produces the maximum principal stress and  σx + σ y σx − σ y σ p min = − 2 2 2 σx − σ y 2 σx − σ y 2 (9.3.4) 2 + τx2y + τx2y (9.3.5) + τx2y (9.3.6) for the rotation that produces the minimum principal stress. The maximum and minimum principal stresses are found at the angle of rotation for which the shear stresses are zero, thus we can say τ p1 = τ p2 = 0 (9.3.7) If we differentiate Equation 9.2.12 with respect to θ and set the result to zero we can find the angle for maximum shearing stresses. σx − σ y dτr s = 2 cos 2θ + τx y 2 sin 2θ = 0 dθ 2 σx − σ y tan 2θsh = − 2τx y (9.3.8) The maximum shear is then τmax =  σx − σ y 2 2 + τx2y (9.3.9) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 352 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The normal stress at τmax is the same in both directions σ = σx + σ y 2 (9.3.10) We see that the maximum and minimum stresses may not be the values obtained with respect to the original axes chosen for representing the equations and their solutions. One must find those values from the above equations. There is a widely used visual representation of the stress components that is helpful in interpreting the results of an analysis. Let us rewrite Equations 9.2.11 and 12 as follows σs − σx − σ y σx + σ y = cos 2θ + τx y sin 2θ 2 2 σx − σ y sin 2θ + τx y cos 2θ τr s = − 2 (9.3.11) We can eliminate θ by squaring both Equations 9.3.11 and adding the results. Then   σx + σ y 2 σx − σ y 2 2 σs − + τr s = + τx2y = R 2 2 2 (9.3.12) where we have set the right hand side equal to R2 . This equation can be recognized as the equation of a circle. Such a circle is plotted in Figure 9.3.2 and is known as Mohr’s circle. τ σx +σ y σ x −σ y 2 2 σy σs ,τsr τxy 2θp σ 2θ −τxy σPmin θ=0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- σx σPmax Figure 9.3.2 This allows us to see the values of stress at any rotation θ at a glance. ########### Example 9.3.1 Problem: The state of plane stress at a point is found to be as shown in Figure (a). Find the principal stresses and the angle of the principal axes. Then find the maximum shear stress and the values of the normal stress at maximum shear. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 353 More about Stress and Strain, and Material Properties 100 MPa 60 MPa 40 MPa Figure (a) Solution: Substitute the values of stress into the appropriate formulas. For the principal stresses σ p max, p min σx + σ y ± = 2  σx − σ y 2 σ p max = 137.082 MPa 2 + τx2y 100 + 40 ± = 2  40 − 100 2 2 + 602 (a) σ p min = 2.918 MPa The rotation angle for the principal axes is tan 2θ p = 2τx y 2 · 60 = −2 = σx − σ y 40 − 100 → θp = − 63.435◦ = −31.718◦ 2 (b) The maximum shear stress is τmax =  σx − σ y 2 2 + τx2y =  40 − 100 2 2 + 602 = 67.082 MPa (c) The rotation angle for maximum shear is tan 2θsh = − σx − σ y 40 − 100 = 0.5 =− 2τx y 2 · 60 → 2θsh = 26.57◦ → θsh = 13.28◦ (d) The normal stress at the angle of maximum shear is the same in both directions. σ = 40 + 100 σx + σ y = = 70 2 2 Note that the rotation angle between maximum principal stress and maximum shear is 45 degrees. This is represented on a Mohr’s circle in Figure (b). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (e) P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 354 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods τ 70 2θp = 63.4º 67.082 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- σ −60 σPmin 2.918 40 137.08 Figure (b) ########### 9.4 Transformation of Strain in Two Dimensions Plane strain has a similar transformation. Let us relate the strain at a point with respect to the sr axes to the strain with respect to the xy axes. The sr axes are rotated by an angle θ from the xy axes as shown in Figure 9.4.1. y P r s θ x Figure 9.4.1 With respect to the xy axes we have the strain displacement relations εx = ∂u ∂x εy = ∂v ∂y γx y = ∂v ∂u + ∂y ∂x (9.4.1) ∂vs ∂u s + ∂r ∂s (9.4.2) With respect to the sr axes the strain displacement relations are εs = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ∂u s ∂s εr = ∂vs ∂r γsr = Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 355 More about Stress and Strain, and Material Properties From Figure 9.4.1 we note the coordinates of a point P are related in the two coordinate systems by x = s cos θ − r sin θ y = s sin θ + r cos θ (9.4.3) vr = v cos θ − u sin θ (9.4.4) and displacements of the point P are related by u s = u cos θ + v sin θ Using Equation 9.4.3 we can write εs = ∂u s ∂ x ∂u s ∂ y ∂u s ∂u s ∂u s = + = cos θ + sin θ ∂s ∂ x ∂s ∂ y ∂s ∂x ∂y (9.4.5) Then with Equation 9.4.4 we can obtain εs = ∂u cos2 θ + ∂x  ∂u ∂v + ∂x ∂y cos θ sin θ + ∂v sin2 θ ∂y (9.4.6) Finally using Equation 9.4.1 we can write εs = εx cos2 θ + ε y sin2 θ + γx y cos θ sin θ (9.4.7) or εs = εx − ε y γx y εx + ε y + cos 2θ + sin 2θ 2 2 2 (9.4.8) By the same process we can also write εx − ε y γx y εx + ε y − cos 2θ + sin 2θ 2 2 2 γr s εx − ε y γx y =− sin 2θ + cos 2θ 2 2 2 (9.4.9) εr = Just as we did in Equation 9.3.12 for stress we can do for strain.   εx + ε y 2 γx y 2 γx y εx − ε y 2 εx − + + = 2 2 2 2 (9.4.10) 2 = R2 (9.4.11) A Mohr’s circle for strain can be drawn. We can use these relations to establish a relationship between the shear modulus and Young’s modulus by considering an element of material in a state of pure shear, that is σx = σ y = 0 τx y = 0 (9.4.12) σ p min = −τx y (9.4.13) The maximum and minimum principal stresses are then σ p max = τx y The corresponding maximum principal normal strain is ε p max = 1 1+ν (σ p max − νσ p min ) = τx y E E (9.4.14) From Equation 9.4.10 and Hooke’s law we get ε p max = τx y γx y = 2 2G (9.4.15) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 356 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods By equating Equation 9.4.14 and 9.4.15 we can conclude that G= E 2 (1 + ν) (9.4.16) We used this back in Chapter 3, Section 3.4, to simplify writing the stress-strain relations in two and three dimensions. Now we have evidence that it is true. 9.5 Strain Rosettes In experimental mechanics stress cannot normally be measured directly; however, strain can be measured using a strain guage based on the electrical resistance of a wire or foil bonded to the surface of the member under strain. When the wire is stretched its length and cross sectional area change and the resitance to electrical current changes. This effect can be calibrated. These guages are usually arranged in a pattern in a group of three. This group is called a strain rosette. They may be arranged in a general pattern such as shown in Figure 9.5.1. 1 2 3 Figure 9.5.1 More commonly, they are arranged in a group at 45◦ intervals as shown in Figure 9.5.2(a) or at 60◦ intervals as shown in Figure 9.5.2(b). y y 2 2 3 1 3 1 x (a) x (b) Figure 9.5.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 357 More about Stress and Strain, and Material Properties In the general case we can relate the strain measured by each guage to the strain in xy coordinates by Equation 9.4.7. ε1 = εx cos2 θ1 + ε y sin2 θ1 + γx y cos θ1 sin θ1 ε2 = εx cos2 θ2 + ε y sin2 θ2 + γx y cos θ2 sin θ2 (9.5.1) ε3 = εx cos2 θ3 + ε y sin2 θ3 + γx y cos θ3 sin θ3 or in matrix form ⎡ cos2 θ1 ⎣ cos2 θ2 cos2 θ2 sin2 θ1 sin2 θ2 sin2 θ3 ⎤ ⎡ ⎤ ⎤⎡ sin θ1 cos θ1 ε1 εx sin θ2 cos θ2 ⎦ ⎣ ε y ⎦ = ⎣ ε2 ⎦ ε3 γx y sin θ3 cos θ3 (9.5.2) Solve these equations for the strains in the xy coordinates and from those strains find the stresses {σ } = [G] {ε} (9.5.3) In the case of a 45◦ rosette θ1 = 0, θ2 = 45◦ , and θ3 = 90◦ and so relations in Equation 9.5.2 reduce to ⎡ ⎤ ⎤ ⎡ εx ε1 ⎣ εy ⎦ = ⎣ ⎦ ε3 (9.5.4) 2ε2 − (ε1 + ε3 ) γx y In the case of a 60◦ rosette θ1 = 0, θ2 = 60◦ , and θ3 = 120◦ and so the relations reduce to ⎤ ⎡ ε1 ⎤ ⎡ εx ⎥ ⎢1 (2ε2 + 2ε3 − ε1 ) ⎥ ⎣ εy ⎦ = ⎢ ⎥ ⎢3 ⎦ ⎣ 2 γx y √ (ε2 − ε3 ) 3 (9.5.5) ########### Example 9.5.1 Problem: A 45◦ strain rosette on the surface of a thin flat plate measures the following strains. The material is aluminum with E = 68950 MPa and ν = 0.3. ⎡ ⎤ ⎡ ⎤ ε1 0.0012 ⎣ ε2 ⎦ = ⎣ 0.0002 ⎦ (a) ε3 0.0005 The rosette is oriented with the xy axes as shown in Figure 9.5.2(a). Find the stresses in the xy coordinates, the maximum principal stress, and the rotation angle of the principal axes. Solution: Use Equation 9.5.4 to find the strains with respect to the xy axes. Use Hooke’s law to find the stresses. Then find the principal stresses and axes using the formulas from Section 9.3. From Equation 9.5.4 ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ 0.0012 ε1 εx ⎦ = ⎣ 0.0005 ⎦ ⎣ εy ⎦ = ⎣ ε3 (b) 2ε2 − (ε1 + ε3 ) γx y −0.0013 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 358 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The stresses from Equation 3.4.22 are E ⎡ ⎤ ⎢ 1 − ν2 σx ⎢ ⎣ σy ⎦ = ⎢ ⎢ νE ⎢ τx y ⎣ 1 − ν2 νE 1 − ν2 E 1 − ν2 ⎤ 0 ⎥⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤ 75769 22731 0 0.0012 102.29 ⎥ εx ⎥⎣ ε y ⎦ = ⎣ 22731 75769 0 ⎦ ⎣ 0.0005 ⎦ = ⎣ 65.16 ⎦ 0⎥ ⎥ 0 0 26519 −0.0013 −34.47 ⎦ γx y 0 0 G (c) The maximum principal stress from Equation 9.3.5 is σ p max σx + σ y + = 2  σx − σ y 2 2 + τx2y = 122.877 MPa (d) The rotation angle for the principal axes from Equation 9.3.3 is tan 2θ p = 2τx y = −1.8567 σx − σ y → θ p = −30.85◦ (e) ########### 9.6 Stress Transformation and Principal Stresses in Three Dimensions Transformation of stress and strain in three dimensions and the principal axes and stresses and strains will be presented here without proof. The full development may be found in books on the theory of elasticity. We show two sets of axes – one rotated with respect to the other – in Figure 9.6.1. The components of the vector F are shown with respect to both the axes xyz and x′ y′ z′ . y y′ Fy x′ Fy′ F Fx′ Fx x Fz Fz ′ z z′ Figure 9.6.1 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎡ P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 359 More about Stress and Strain, and Material Properties To simplify the notation we define a set of direction cosines in the following table. x′ y′ z′ x y z l1 l2 l3 m1 m2 m3 n1 n2 n3 where, for example, l1 = cos xx′ , where xx′ signifies the angle between the x axis and the x′ axis. Similarly, m3 = cos yz′ , where yz′ signifies the angle between the y axis and the z′ axis. We define a transformation matrix ⎡ ⎤ l1 m 1 n 1 [T ] = ⎣ l2 m 2 n 2 ⎦ (9.6.1) l3 m 3 n 3 You can verify that the components of the force F are given by the matrix equation {F ′ } = [T ]{F} (9.6.2) Thus given the force components with respect to a particular orientation of the axes we can find the components with respect to any rotated axes. For the transformation of stress we must recall the definition of stress must specify both the direction of the force and the orientation of the surface to which it is referred. It can be shown that the transformation of stress is given by [σ ′ ] = [T ][σ ][T ]T (9.6.3) where [σ ] is written in the form ⎡ σx [σ ] = ⎣ τx y τzx τx y σy τ yz ⎤ τzx τ yz ⎦ σz (9.6.4) It can also be shown that there are three orthogonal planes on each of which the stress at a point consists of a normal stress and no shearing stress. These are called the principal stresses and the rotated coordinate axes that are normal to each of these planes are called principal axes. Let us look at the stresses at a point on a small tetrahedron as shown in Figure 9.6.2. The face of the tetrahedon is oriented so that the normal stress σ p is the only stress acting on that face, thus it is a principal stress. It follows from Equation 9.6.3 that [σ p ] = [T p ][σ ][T p ]T (9.6.5) where [T p ] is the appropriate set of direction cosines for the pincipal axes and [σ p ] are the principal stresses. The orientation of the principal axes for the normal stress in Figure 9.6.2 is unknown. Let us identify them as ⎡ ⎤ l ⎣m ⎦ (9.6.6) n It can be shown that ⎡ σx ⎣ τx y τzx τx y σy τ yz ⎡ ⎤ ⎤⎡ ⎤ τzx l l τ yz ⎦ ⎣ m ⎦ = σ p ⎣ m ⎦ σz n n --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (9.6.7) P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 360 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y σz τzx σp σx τyz x τxy σy z Figure 9.6.2 This can be rewritten as a classical eigenvalue problem. ⎤⎡ ⎤ ⎡ ⎤ ⎡ τx y τzx l σx − σ p 0 ⎣ τx y σy − σ p τ yz ⎦ ⎣ m ⎦ = ⎣ 0 ⎦ τzx τ yz σz − σ p n 0 (9.6.8) One possible solutions is ⎡ ⎤ ⎡ ⎤ l 0 ⎣m ⎦ = ⎣0⎦ n 0 (9.6.9) however, equations of this form have solutions other than zero for discrete values of s determined by    σx − σ τx y τzx    τx y σy − σ τ yz  = 0 (9.6.10)   τzx τ yz σz − σ  When expanded it yields the following cubic equation σ 3 − S1 σ 2 + S2 σ − S3 = 0 (9.6.11) where S1 = σx + σ y + σz 2 S2 = σx σ y + σx σ y + σx σ y − τx2y − τ yz − τzx2 S3 = σx σ y σz + 2τx y τ yz τzx − 2 σx τ yz − σ y τzx2 − (9.6.12) σz τx2y The three roots of Equation 9.6.11 are the three principal stresses. The direction of the principal stresses are found by substituting one of the principle stresses into Equation 9.6.13. ⎤⎡ ⎤ ⎡ ⎤ ⎡ τx y τzx l σx − σ p 0 ⎣ τx y σy − σ p τ yz ⎦ ⎣ m ⎦ = ⎣ 0 ⎦ (9.6.13) τzx τ yz σz − σ p n 0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 361 More about Stress and Strain, and Material Properties Since this is a set of homogenous equations we can only solve for two of the direction cosines in terms of the third. The third one is found from the property of direction cosines that l 2 + m 2 + n2 = 1 (9.6.14) The software products, such as Mathematica, Maple, Matlab, and Mathcad, among others have eigenvalue solvers for exactly this form of the equations. ########### Example 9.6.1 Problem: Find the principal normal stresses and the maximum shear stress at a point in the three dimensional solid where the stresses are σx = 20 MPa τx y = −40 MPa σ y = −40 MPa τ yz = 60 MPa σz = 100 MPa τzx = 0 MPa Solution: Use Equation 9.6.13 With the stress values inserted the eigenvalue equations become ⎤⎡ ⎤ ⎡ ⎤ ⎡ −40 0 l 20 − σ p 0 ⎦⎣m ⎦ = ⎣0⎦ ⎣ −40 60 −40 − σ p 0 60 100 − σ p n 0 The three eigenvalues, or principal stresses, are ⎤ ⎡ ⎤ ⎡ 124.2 σ p1 ⎣ σ p2 ⎦ = ⎣ −76.9 ⎦ MPa σ p3 32.7 The eigenvector, or orientation of the principal stresses, is found stresses into Equation (a) and solving for ⎤ ⎡ ⎡ l2 l3 −0.1421 0.3701 l1 ⎣ m 1 m 2 m 3 ⎦ = ⎣ −0.3642 −0.8820 n1 n2 n3 0.9204 −0.2918 (a) (b) (c) by inserting one of the principal ⎤ 0.9181 0.2992 ⎦ 0.2601 (d) Each column presents the eigenvectors or direction cosines for the corresponding principal stress. ########### 9.7 Allowable and Ultimate Stress, and Factors of Safety Up to now we have limited the discussion to levels of stress and strain that do not exceed the linearly elastic range of the material. Most load bearing structures are designed so that the levels of stress and strain always remain in this linear range. The point at which higher values of stress causes the stress-strain curve to deviate from a straight line is called the proportional limit. We need to examine what happens when stress values exceed this limit. A generic stress-strain curve is shown as obtained from a uniaxial test in Figure 9.7.1. The point at which the curve is no longer elastic is called the yield stress. In many typical structural metals the yield stress is very near the proportional limit and so we do not have to take into account any small region of non linear elasticity. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 362 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods σ Ultimate stress A Yield stress Fracture point Proportional limit Unloading ε B Figure 9.7.1 In metals, when stresses are below the yield value and when loads are reduced the stresses and strains follow the curve down on the same path that was followed when increasing the load. If, in fact the stress does exceed the yield value, say at point A, then unloading follows the path AB as shown and a permanent strain results. The details of the stress-strain curve for such loading paths for other materials like fiber reinforced composites can be different, influenced by the polymer stress-strain curve. These details are beyond the scope of this text that provides an introduction. Going back to the typical response for a metal as shown in Figure 9.7.1, at some point the stress reaches a maximum value called the ultimate stress and there is continued strain without an increase in stress, in fact there is a relaxation in which strain increases even with a decrease in stress. Finally, the material fractures. In design, a certain value of stress is determined as permissible under normal conditions of loading. This is called the allowable stress. In most cases in the design of reusable structures the allowable stress is well below the yield point. After all, we do not want the structure to have permanent deformation under normal use. If we define failure as the value of stress where permanent deformation occurs the ratio of that value to the allowable value is called the factor of safety. F.S. = σfail σallow (9.7.1) A suitable value for F.S. depends on many factors, such as, degree of accuracy of loads and restraints, approximations used in analysis, deteriorizations over lifetime of use, repetitive and dynamic loading, accuracy of material properties, and many more consideration. Just what value to use in design requires careful consideration. Structures such as bridges and building must have fairly high factors of safety because they have extended lives and must survive often under severe environmental conditions and human safety is a paramount concern. Aircraft must have much lower values in order to keep the weight down to be economically viable; however, they have shorter lives, frequent maintenance, and carefully trained operators. But always human safety is a paramount concern. Unmanned space craft can have very low values of F.S. because they are most carefully operated and human life is not at risk. In the case of uniaxial stress it is quite clear that when the stress excedes the yield point the material will achieve permanent deformation. In the case of plane stress and a state of three dimensional stress it might seem logical that when the maximum principal stress exceeds that yield point the material will deform permanently. Actually it is not as simple as that. Apparently the mechanism for yield is more --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 363 More about Stress and Strain, and Material Properties complicated than that. There are in fact several criteria offered for when a material will yield under multiaxial stress. We shall present only one called the von Mises yield stress, which is widely accepted for designing with metals. Without further explanation, the von Mises criteria is based on the maximum strain energy of distortion. In the calculation of strain energy, part of that energy is a result of change in volume; another part is a result of distortion or change in shape. The distortion energy per unit volume of an isotropic material in plane stress is Ud =  1  2 σ − σ p1 σ p2 + σ p22 6G p1 (9.7.2) In the case of a uniaxial test specimen, such as used to find the stress strain curve in Chapter 3, Section 3.4, we note that at yield σ p2 = 0 σ p1 = σY (9.7.3) Thus (Ud )Y = 1 2 σ 6G Y (9.7.4) For plane stress the von Mises criterion stated that the general 2D state of stress is safe as long as, Ud represented by (9.7.2) is less than Ud that one finds in a uniaxial test, i.e. given by (9.7.4). Thus, it follows that, σ p21 − σ p1 σ p2 + σ p22 ≤ σY2 Based on (9.7.5), we define the von Mises stress in two dimensions as  σVM = σ p21 − σ p1 σ p2 + σ p22 (9.7.5) (9.7.6) and compare it to the yield stress based on a one dimensional test of the axially loaded slender bar, as stated earlier. The distortion energy for a state of three dimensional stress is Ud = 1 [(σ p1 − σ p2 )2 + (σ p2 − σ p3 )2 + (σ p3 − σ p1 )2 ] 12G (9.7.7) For three dimensional stress, using a similar reasoning to that followed for the 2D case, the von Mises criterion states that the state of stress is safe as long as (σ p1 − σ p2 )2 + (σ p2 − σ p3 )2 + (σ p3 − σ p1 )2 ≤ 2σY2 , (9.7.8) which is obtained from (9.7.8) and (9.7.4). We define the von Mises stress for three dimensions as  (σ p1 − σ p2 )2 + (σ p2 − σ p3 )2 + (σ p3 − σ p1 )2 (9.7.9) σVM = 2 and compare it to the yield stress based on a one dimensional test of the axially loaded slender bar. 9.8 Fatigue Before we move onto examples, we must note a few things regarding designing against yield using a factor of safety for those structures like wings and bridges which experience repeated loading or cyclic loads. In those instances, a structure can experience failure at a level of load that is below the load which causes permanent set. This is due to fatigue. In fatigue, a material can fail at a load level that is below --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas 364 July 4, 2011 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods that which is needed to cause yield. Failure due to repeated loading can be seen when one bends a paper clip back and forth until failure. Consider a structural component that is subjected to sinusoidal cylic loading. We can describe such loading with an amplitude and a period (or frequency). Even though the stress amplitude may not be as large to cause failure in one cycle, with repeated loading, it is found that the structure undergoes fatigue failure by the growth of small flaws (cracks) that can accumulate damage over each cycle and reach a critical state of damage that can lead to large and abrupt crack growth. In general, the number of cylces until failure increases with a diminishing stress amplitude and there exists a stress amplitude below which the structure has infinite fatigue life, the latter being what is ideally desirable. For each material, the resistance to fatigue can be measured by doing tests at cyclic loading of constant amplitude and frequency. In those cases two quantities are needed to characterize the loading, σmax + σmin σmax − σmin , and the mean stress, defined as, σave. = . the average stress, defined as σave. = 2 2 To find the fatigue stress level for loading with zero mean value, the number of cycles to failure, N, is recorded as a function of stress amplitude. A plot of stress amplitude vs. N produces a S-N diagram which is used as a material property, the fatigue resistance curve. In general, the S-N curve is a monotonically decreasing curve and below a certain stress amplitude the material shows no limit in number of cycles to failure – referred to as the fatgue limit. The notion of “safe life design” requires that the part sees service stresses below this fatigue limit, ensuring an infinite life! Fatiuge life is also affected by mean stress and usually, it is found that the larger the mean stress, the more diminished is the S-N diagram. For many aerospace structural materials S-N diagrams are available in handbooks. Designing against fatigue is an involved subject that goes beyond the scope of this introductory text and the reader is referred to Bathias and Pineau (2010) for more details. ########### Example 9.8.1 Problem: Find the von Mises stress for the state of stress in Example 9.3.1. Solution: Use Equation 9.7.5. The principal stresses are from Example 9.3.1. σ p1 = 2.918 MPa σ p2 = 137.082 MPa Substituting the values   σVM = σ p21 − σ p1 σ p2 + σ p22 = (2.918)2 − 2.918 · 137.082 + (137.082)2 = 135.65 MPa (a) (b) Note that in this case it is slightly smaller than the maximum principle stress given in Equation (a). ########### These relations are programmed and automatically calculated in the FEM computer codes. Factors of safety are often based on the von Mises stresses. 9.9 Creep In some material and environmental conditions the displacement may increase slowly with time although the load is constant and the stress is below the normal yield stress. In particular some materials behave this way when the temperature is increased beyond some appropriate level. This phenomenon is known --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas July 4, 2011 7:10 Printer Name: Yet to Come 365 More about Stress and Strain, and Material Properties as creep. For most aerospace metals, creep is significant only at high temperatures and in those situations, such as in engine structures, special metal alloys have been developed that exhibit high creep resistance. Creep is a specialized subject that is beyond the scope of this introductory text. 9.10 Orthotropic Materials—Composites There is growing use of materials for structures that are neither homogenous nor istropic. Plywood, for example has been around for a long time. Built up layers of cloth glued together is another. For some time these were used primarily in low load situations. Today there are a number of laminated and filament reinforced materials used in very high tech structures, for example, in spacecraft. A new generation of aircraft make substantial use of composites. In many cases these can be modeled as orthotropic materials, that is, the materials have specific and different properties in each of the three coordinate directions. Without proof the strain stress relations for an orthotropic material are 1 ⎢ Ex ⎢ ⎢ −vx y ⎢ ⎤ ⎢ E ⎡ ⎢ x εx ⎢ ⎢ ε y ⎥ ⎢ −vx z ⎥ ⎢ ⎢ ⎢ εz ⎥ ⎢ E x ⎥ ⎢ ⎢ ⎢ γx y ⎥ = ⎢ ⎥ ⎢ ⎢ ⎣ γ yz ⎦ ⎢ 0 ⎢ ⎢ γzx ⎢ ⎢ 0 ⎢ ⎢ ⎣ 0 ⎡ −v yx Ey 1 Ey −v yz Ey −vzx Ez −vzy Ez 1 Ez 0 0 0 ⎤ 0 0 0 0 0 0 0 0 1 Gxy 0 0 0 0 0 1 G yz 0 0 0 0 0 1 G zx ⎥ ⎥ ⎥ ⎥ ⎥⎡ ⎤ ⎥ σx ⎥ ⎥ ⎢ σy ⎥ ⎥⎢ ⎥ ⎥ ⎢ σz ⎥ ⎥⎢ ⎥ ⎥ ⎢ τx y ⎥ ⎥⎢ ⎥ ⎥ ⎣ τ yz ⎦ ⎥ ⎥ τzx ⎥ ⎥ ⎥ ⎥ ⎦ (9.10.1) It can be shown that vx y v yx = Ex Ey vzx vx z = Ex Ez v yz vzy = Ez Ey (9.10.2) This set of equations can be inverted to provide {σ } = [Gortho ] {ε} or ⎡ ⎡ ⎤ E x0 (1 − vzy v yz ) σx ⎢ E y (vx y + vx z vzy ) ⎢ σy ⎥ ⎢ ⎥ ⎢ ⎢ σz ⎥ 1 ⎢ ⎢ E z (vx z + vx y v yz ) ⎥ ⎢ ⎢ ⎢ τx y ⎥ = 0 ⎥ ⎢ ⎢ ⎣ ⎣ τ yz ⎦ 0 τzx 0 E y (vx y + vx z vzy ) E y (1 − vzx vx z ) E z (v yz + v yx vx z ) 0 0 0 E z (vx z + vx y v yz ) E z (v yz + v yx vx z ) E z (1 − vx y v yx ) 0 0 0 (9.10.3) 0 0 0 Gxy 0 0 0 0 0 0 G yz 0 ⎤ ⎤⎡ εx 0 ⎢ ⎥ 0 ⎥ ⎥ ⎢ εy ⎥ ⎢ εz ⎥ 0 ⎥ ⎥ ⎥⎢ ⎢ ⎥ 0 ⎥ ⎥ ⎢ γx y ⎥ ⎣ ⎦ γ yz ⎦ 0 γzx G zx (9.10.4) where = 1 − vx y v yx − vx z vzx − v yz vzy − 2vx z vzy v yx (9.10.5) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-09 P2: ABC JWST071-Waas 366 July 4, 2011 7:10 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods This expression for [Gortho ] can be used in place of [G] wherever it is found in any of the previous equations provided the xyz axes are alligned with the material axes in an orthotropic material. 9.11 Summary and Conclusions In Chapter 1 a brief introduction was provided for the concepts of displacement, stress, strain, equilibrium, and material properties. Then in Chapters 2 through 6 methods for finding the values of displacement, stress, and strain consistent with the laws of equilibrium and material properties. Now, in this chapter we have examined these quantities in more detail. --`,,,,``,``,,``,,`,`,,,, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come Combined Loadings on Slender Bars—Thin Walled Cross Sections 10.1 Introduction The geometry of the slender bar, its material properties, the applied loads, and the restraints determine the displacements and stresses in the bar. In previous chapters we have limited the geometry, loading, and restraints to allow uncoupled axial, torsional, and bending displacements and stresses. Each case is considered independently. In the case of bending we have further restricted the analysis to displacements in the xy plane. We now remove some of these restrictions to consider simultaneous multiple application of loading. We also remove both the restriction of bending in one plane and the requirement that I yz = 0. Of particular interest is the extension of the theory of thin walled cross sections to include simultaneous bending and torsion and the addition of concentrated stiffeners to the sections. We start with a brief review of slender bar theory as presented in previous chapters and then embark on the extensions of the analysis just noted. 10.2 Review and Summary of Slender Bar Equations It is convenient to summarize the equations used in the preceding chapters to analyze slender bars. In the following sections of this chapter we shall apply these equations to combined loadings and in some cases extend them to more complicated geometry. 10.2.1 Axial Loading We found in Chapter 4 that the differential equation for the displacement of an axially loaded slender bar is du d EA = − f x (x) dx dx (10.2.1) and when EA is constant EA d 2u = − f x (x) dx2 → u(x) = − 1 EA  f x (x)dx + ex + f (10.2.2) Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 10 P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas 368 July 2, 2011 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The equation is integrated to find u(x) and boundary conditions are used to find the two constants of integration. Once the displacements are known the internal force is found from P = EA du dx (10.2.3) and the stress is σx = P A (10.2.4) We note that when the bar is statically determinate we can find P directly from equilibrium and immediately solve for the stress. To find the displacement we then solve the first order equation, that is, Equation 10.2.3. An alternative is to use the finite element method as presented in Chapter 5. The first step is to divide the bar into elements and number the elements and the nodes. Shape functions [n] that represent the distributed displacement u(sm ) in terms of the nodal displacements {rm } of each element are shown in Equation 10.2.5.    sm sm un (10.2.5) u(sm ) = [n]{rm } = 1 − lm lm u n+1 From these shape functions we find the element stiffness matrix that represents the internal forces { pm } in terms of the nodal displacements.      E m Am 1 −1 un pm,n = [km ]{rm } = { pm } = (10.2.6) u n+1 pm,n+1 −1 1 lm The element stiffness matrices [km ] are assembled to form the global stiffness matrix [K ].  [K ] = [km ] For each element the distributed applied loads are converted to equivalent nodal loads by ⎡ sm ⎤    lm  lm 1 − lm ⎥ Fem,n ⎢ [n]T f x (s)dsm = {Fem } = ⎦ f x (s)dsm = ⎣ s m Fem,n+1 0 0 lm (10.2.7) (10.2.8) These are assembled to form the total equivalent nodal loads {Fe } and are added to the actual applied concentrated loads {Fc } to form global load matrix {F}.  {Fem } = {Fc } + {Fe } (10.2.9) {F} = {Fc } + The global equations for the nodal displacements are {F} = {Fc } + {Fe } = [K ]{r } (10.2.10) The known nodal restraints and the unknown restraint forces are inserted in the global equations and then the equations are partitioned into one set for solving the unknown nodal displacements and another set for solving the unknown restraint forces. Once the nodal displacements are known the distributed displacements are found for each element by Equation 10.2.5 repeated here as Equation 10.2.11.    sm sm un (10.2.11) u(sm ) = [n]{rm } = 1 − lm lm u n+1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 369 Combined Loadings on Slender Bars—Thin Walled Cross Sections The internal forces are found by Equations 10.2.6 and the stresses are found for each element by   Em d un −1 1 (10.2.12) [n]{rm } = σxm = E m u n+1 dsm lm As explained in Chapter 5 the axial forces and stresses are exact values for bars with concentrated loads only, but must be interpreted when equivalent nodal loads are used (such as when you have distributed axial loads). In particular forces are averaged at the nodes and curves are fitted through the resulting values. Chapter 5, Section 5.4, has an example demonstrating this. 10.2.2 Torsional Loading We found in Chapter 6 that the differential equation for the displacement of a torsionally loaded slender bar is dβ d GJ = −tx (x) dx dx (10.2.13) and when GJ is constant GJ d 2β = −tx (x) dx2 → β(x) = − 1 GJ  t(x)dx + ex + f (10.2.14) The equation is integrated to find β(x) and boundary conditions are used to find the two constants of integration. Once the rotational displacement is known the internal torque is found from T = GJ dβ dx (10.2.15) and the stress is τ= T r J (10.2.16) where τ is expressed in polar coordinates. These equations are restricted to circular cross sections and J is the area polar moment of inertia of the cross section. In fact notice that these equations for torsion are identical in form to the equations that govern the axial deformation of slender bars, Equations 10.2.1 – 10.2.4. We note that when the bar is statically determinate we can find T directly from equilibrium and immediately solve for the stress. To find the displacement we then solve the first order equation, that is, Equation 10.2.15. An alternative is to use the finite element method as presented in Chapter 6. The first step is to divide the bar into elements and number the elements and the nodes. Shape functions [n] that represent the distributed rotational displacement βm (s) in terms of the nodal displacements {φm } of each element are shown in Equation 10.2.17 (see similarity between Equation 10.2.17 and 10.2.5).    sm sm βn (10.2.17) β(sm ) = [n]{φm } = 1 − lm lm βn+1 From these shape functions we find the element stiffness matrix that represents the internal torques {ψm } in terms of the nodal displacements.      G m Jm 1 −1 βn ψm,n = [km ]{φm } = (10.2.18) {ψm } = βn+1 ψm,n+1 −1 1 lm Again Equation 10.2.18 is similar to Equation 10.2.6. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-10 P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 370 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The element stiffness matrices [km ] are assembled to form the global stiffness matrix [K ].  [K ] = [km ] (10.2.19) For each element the distributed applied moments, tx (sm ), are converted to equivalent nodal moments by [n]T tx dsm =  0 lm ⎡ sm ⎤   lm ⎥ Mm,n ds = t ⎦ x m sm Mm,n+1 lm 1− ⎢ ⎣ (10.2.20) These are assembled to form the total equivalent nodal moments {Me } and are added to the actual applied concentrated moments {Mc } for form global moment matrix {M}.  {M} = {Mc } + {Mem } = {Mc } + {Me } (10.2.21) The global equations for the nodal displacements are {M} = {Mc } + {Me } = [K ] {φ} (10.2.22) The known nodal restraints and the unknown restraint forces are inserted in the global equations and then the equations are partitioned into one set for solving the unknown nodal displacements and another set for solving the unknown restraint forces. Once the nodal displacements are known the distributed displacements are found for each element by    sm sm βn (10.2.23) β(sm ) = [n]{φm } = 1 − lm lm βn+1 For each element the internal moments are found by Equation 10.2.18 and the only unknown shear stresses by   Gm βn −1 1 r (10.2.24) τm = βn+1 lm As explained in Chapter 6 the internal torques and stresses are exact values for bars loaded by concentrated torques but must be interpreted when the applied loads are distributed and equivalent nodal loads are used. In particular torques are averaged at the nodes and curves are fitted through the resulting values. Chapter 6, Section 6.9, has examples and details of this stress recovery process. 10.2.3 Bending in One Plane We found in Chapter 7 that the differential equation for the displacement of a laterally loaded slender bar when I yz = 0, that is, when at least one of the cross sectional axes, y or z, is a principal axis (symmetrical cross sections are an example) and the load entirely in the xy plane is d 2v d2 EI = f y (x) zz dx2 dx2 (10.2.25) and when EIzz is constant the solution is EIzz d 4v = f y (x) dx4 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS → v(x) = 1 EIzz  f y (x)d x + a 3 b 2 x + x + cx + d 6 2 (10.2.26) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- {Mem } =  P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 371 Combined Loadings on Slender Bars—Thin Walled Cross Sections The boundary conditions are used to find the four constants of integration. Once the displacement is known the internal bending moment and shear force are found from M = EIzz d 2v dx2 V = −EIzz d 3v dx3 (10.2.27) The normal stresses are σx = − My Izz (10.2.28) and for a rectangle cross section the shear stresses are V τx y (x, y) = Izz b  ydA (10.2.29) Ay where b is the width of the cross section. We note that when the beam is statically determinate we can find V and M directly from equilibrium and immediately solve for the stresses. To find the displacement we then solve the second order equation, that is, Equation 10.2.27. An alternative is to use the finite element method as presented in Chapter 8. The first step is to divide the bar into elements and number the elements and the nodes. Shape functions [n] that represent the distributed displacement v(sm ) in terms of the nodal displacements {rm } of each element are shown in Equation 10.2.30. v(sm ) = [n]{rm } 1 = 3 lm3 − 3lm sm + 2sm3 lm lm3 sm − 2lm2 sm2 + lm sm3 3lm sm2 − 2sm3 −lm2 sm2 + lm sm3 {rm } (10.2.30) From these shape functions we find the element stiffness matrix that represents the internal shear forces and moments, { pm }, in terms of the nodal displacements, {rm }. ⎡ ⎤⎡ ⎡ ⎤ ⎤ 12 6lm −12 6lm Vm,n vm,n ⎢ ⎥ 2 2 ⎢ Mm,n ⎥ 4lm −6lm 2lm ⎥ ⎢ θm,n ⎥ E m Im ⎢ 6lm ⎥ ⎥ (10.2.31) { pm } = [km ]{rm } = ⎢ ⎥⎢ ⎣ Vm,n+1 ⎦ = l 3 ⎢ ⎣ −12 −6lm 12 −6lm ⎦ ⎣ vm,n+1 ⎦ m Mm,n+1 θm,n+1 6l 2l 2 −6l 4l 2 m m m m The element stiffness matrices [km ] are assembled to form the global stiffness matrix [K ].  [K ] = [km ] (10.2.32) For each element the distributed applied loads are converted to equivalent nodal loads by ⎤ ⎡ 3 ⎡ ⎤ lm − 3lm sm2 + 2sm3 Fem,n  lm  lm ⎢ 3 2 2 3 ⎥ ⎢ Mem,n ⎥ ⎢ lm s − 2lm sm + lm sm ⎥ f y (sm ) ⎥ {Fem } = [n]T f y (sm ) dsm = ⎥ 3 dsm = ⎢ ⎢ ⎣ Fem,n+1 ⎦ 2 3 ⎦ lm ⎣ 3lm sm − 2sm 0 0 Mem,n+1 −l 2 s 2 + l s 3 m m (10.2.33) m m These are assembled to form the total equivalent nodal loads {Fe } and are added to the actual applied concentrated loads {Fc } for form the global load matrix {F}.  {Fem } = {Fc } + {Fe } {F} = {Fc } + (10.2.34) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas 372 July 2, 2011 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The global equations for the nodal displacements are {F} = {Fc } + {Fe } = [K ]{r } (10.2.35) The known nodal restraints and the unknown restraint forces are inserted in the global equations and then the equations are partitioned into one set for solving the unknown nodal displacements and another set for solving the unknown restraint forces. Once the nodal displacements are known the distributed displacements are found for each element by Equation 10.2.30 repeated here as Equations 10.2.36. v(sm ) = [n]{rm } 1 = 3 lm3 − 3lm sm + 2sm3 lm lm3 sm − 2lm2 sm2 + lm sm3 3lm sm2 − 2sm3 −lm2 sm2 + lm sm3 {rm } (10.2.36) The internal moments and shear forces are found by Equation 10.2.31. The normal and shear stresses are found at the nodes for each element using Equations 10.2.28 and 10.2.29 repeated here as Equation 10.2.37  V My τx y (x, y) = ydA (10.2.37) σx = − Izz Izz b A y The normal stresses are linear functions between nodes and the distributed values are found for each element by ⎡ ⎤ vm,n ⎢ θm,n ⎥ yE ⎥ σxm = − 3 −6lm + 12sm −4lm2 + 6lm sm 6lm − 12sm −2lm2 + 6lm sm ⎢ ⎣ vm,n+1 ⎦ (10.2.38) lm θm,n+1 The shear force in each element is a constant and consequently the shear stress is the same at all values of s. They are equal to the values calculated at either node. As explained in Chapter 8 the shear forces and stresses and the bending moments and stresses are exact values for beams with only concentrated loads but must be interpreted when equivalent nodal loads are used. In particular, forces and moments are averaged at the nodes and curves are fitted through the resulting values. Chapter 8, Section 8.4, has examples and details for this process, but we will give examples to clarify this. 10.3 Axial and Torsional Loads For bars with certain restrictions on the cross sections, combined axial and torsional loads can produce both axial and torsional displacements that are uncoupled In these instances each case can be solved separately and superposition can be used to find answers for the case of combined loading. In the FEM method the equations can be combined into one set. A propeller shaft (circular) is a good example. The drag on the propeller blades produces a torque while the thrust of the propeller produces an axial force. Here is a simple example. ########### Example 10.3.1 Problem: A shaft with a circular cross section is loaded as shown in Figure (a). This simulates the loads on a propeller shaft. Find the displacement and stress in the shaft. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 373 Combined Loadings on Slender Bars—Thin Walled Cross Sections F α d α F Figure (a) Solution: The axial and torsional loads can be applied separately and the results for displacement and stress simply added. The net effect is an axial load (thrust) which we shall designate FT and a moment (from drag on the propeller blades) which we shall designate MD . M D = 2dF sin α FT = 2F cos α (a) From Example 4.3.1 the displacement and stress for the axial case are u a (x) = FT x EA σxa = FT A (b) From Example 6.3.1 the displacement and stress for the torsional case are β(x) = MD x GJ τ= MD r J (c) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ########### The finite element equations can be solved separately for axial and torsional effects also. Often they are combined into one set. In the case where an element lies along the x axis we simply place the axial and torsion nodal displacements in the same matrix. Let the nodal displacements and internal forces be ⎡ ⎡ ⎤ ⎤ pm,n un ⎢ ψm,n ⎥ ⎢ βn ⎥ ⎥ ⎥ (10.3.1) { pm } = ⎢ {rm } = ⎢ ⎣ pm,n+1 ⎦ ⎣ u n+1 ⎦ ψm,n+1 βn+1 Combining Equations 10.2.6 and 10.2.18 the element matrix equations become ⎡ E m Am E m Am 0 − 0 lm lm ⎡ ⎤ ⎢ ⎢ pm,n G m Jm G m Jm ⎢ 0 0 − ⎢ ψm,n ⎥ ⎢ ⎢ l lm m ⎢ ⎥ =⎢ { pm } = [km ]{rm } = ⎣ E m Am pm,n+1 ⎦ ⎢ E m Am 0 0 ⎢− ψm,n+1 lm lm ⎢ ⎣ G m Jm G m Jm 0 − 0 lm lm Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ⎤ ⎥⎡ ⎤ ⎥ ⎥ u m,n ⎥⎢ ⎥ ⎢ βm,n ⎥ ⎥ ⎥⎣ ⎥ u m,n+1 ⎦ ⎥ ⎥ βm,n+1 ⎦ (10.3.2) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 374 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 10.3.2 Problem: Solve Example 10.3.1 by finite elements. Solution: Use Equation 10.3.2. We can use just one element. Insert restraints, restraint forces, and applied loads. ⎡ EA ⎤ EA 0 − 0 ⎥⎡ ⎤ L ⎡ ⎤ ⎢ ⎢ L ⎥ Rx1 GJ GJ ⎥ 0 ⎢ ⎢ ⎥⎢ ⎥ 0 0 − ⎢ Q x1 ⎥ ⎢ 0⎥ L L ⎥ ⎢ ⎥ ⎥⎢ ⎣ FT ⎦ = ⎢ EA ⎢ EA ⎥ ⎣ u2 ⎦ ⎢− 0 0 ⎥ MD ⎢ L ⎥ β2 L ⎣ GJ GJ ⎦ 0 − 0 L L The partitioned equations for the displacement are ⎡ EA   FT ⎢ =⎣ L MD 0 This is easily solved.  u2 β2  0 ⎤   ⎥ u2 ⎦ GJ β2 L ⎡ F L ⎤ T ⎢ EA ⎥ =⎣ MD L ⎦ GJ (a) (b) (c) The internal forces are  p1,2 ψ1,2  ⎡ EA ⎢ =⎣ L 0 0 ⎤     FT ⎥ u2 = ⎦ GJ MD β2 L (d) and the stresses are σx = FT A τa = MD r J (e) The distributed displacements are  x u(x) = 1 − L  x β(x) = 1 − L   x  0 x = u2 L L   x  0 x = β2 L L FT FT L = x EA EA MD MD L = x GJ GJ You can compare these results with those in Example 10.3.1. ########### --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (f) P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 375 Combined Loadings on Slender Bars—Thin Walled Cross Sections 10.4 Axial and Bending Loads—2D Frames When loads produce both bending and axial component forces in a bar, the problems may be treated as uncoupled and each case solved separately when the axial force is sufficiently small. The results can then be found using the principle of superposition. There are circumstances, however, particularly when the axial forces are large and compressive, when the bending and axial effects are coupled. This is considered in detail, later, in Chapter 14. In this section we limit consideration to cross sections which are symmetrical about the xy plane, for which I yz = 0, where the loads are in the xy plane, and when the axial loads are sufficiently small. An example will explain what must be done. ########### Example 10.4.1 Problem: Consider the case when the load on a slender bar is at an angle to the x axis other than 0 or 90 degrees as shown in Figure (a). The cross section is rectangular with a height h and a width b. Find the displacements and stresses. Solution: The applied force can be broken into components as shown and the axial and bending problems solved separately. Fortunately the two equations are uncoupled. The displacements and stresses can be found by adding the two results. Superposition holds. Fy y F α Fx From Example 4.3.1 the displacement and stress for the axial case are u a (x) = F cos α Fx x= x EA EA σxa = F cos α Fx = A A (a) where A = hb. From Example 7.3.2 the displacement and stress for the bending case are vb (x) = Fy EIzz   x3 F sin α L x 2 − EIzz 2 6 Fy (L − x) F sin α(L − x) y=− y σxb (x, y) = − I I  zz zz F sin α Fy ydAy = ydAy τx yb (x, y) = Izz b A y Izz b Ay  x3 Lx2 − 2 6  = (b) where Izz = bh 3 /12. Since in bending the axial displacement is u b (x, y) = −y Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS dv dx Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (a) P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 376 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The total axial displacement at any point in the bar is given by u(x, y) = u a + u b = F sin α F cos α x− EA EIzz  Lx − x2 2  y (d) The combined stresses are F sin α(L − x) F cos α − y A Izz  F sin α ydA y τ (x, y) = τx y (x, y) = Izz b Ay σx (x, y) = σa + σxb = (e) ########### We can solve finite element equations independently and add the results or we can adapt them to work with combined loads. In the case where an element lays along the x axis we simply place the axial and bending nodal displacements in the same matrix. The element nodal displacement matrix and the internal force matrix are ⎤ ⎤ ⎡ ⎡ un pm,n ⎢ vn ⎥ ⎢ Vm,n ⎥ ⎥ ⎥ ⎢ ⎢ ⎢ θn ⎥ ⎢ Mm,n ⎥ ⎥ ⎥ ⎢ ⎢ { pm } = ⎢ (10.4.1) {rm } = ⎢ ⎥ ⎥ ⎢ u n+1 ⎥ ⎢ pm,n+1 ⎥ ⎣ vn+1 ⎦ ⎣ Vm,n+1 ⎦ θn+1 Mm,n+1 Combining Equations 10.2.6 and 10.2.31 the element matrix equations become ⎡ E A E m Am m m 0 0 − 0 0 ⎢ lm lm ⎢ ⎢ 12E m Im 12E m Im 6E m Im 6E m Im ⎢ 0 0 − ⎡ ⎤ ⎢ 3 2 3 l l l lm2 ⎢ m m m pm,n ⎢ 6E m Im 6E m Im 4E m Im 2E m Im ⎢ Vm,n ⎥ ⎢ ⎢ ⎥ ⎢ 0 0 − 2 ⎢ Mm,n ⎥ ⎢ lm2 lm lm lm ⎢ ⎥=⎢ ⎢ pm,n+1 ⎥ ⎢ E m Am E A m m ⎢ ⎥ ⎢ 0 0 0 0 ⎣ Vm,n+1 ⎦ ⎢ − l lm ⎢ m ⎢ Mm,n+1 ⎢ 12E m Im 6E m Im 12E m Im 6E m Im ⎢ 0 − − 2 0 − 2 3 3 ⎢ lm lm lm lm ⎢ ⎣ 6E m Im 6E m Im 2E m Im 4E m Im 0 0 − 2 lm2 lm lm lm ⎤ ⎥ ⎥ ⎥ ⎥ ⎥⎡ ⎤ ⎥ un ⎥ ⎥ ⎢ vn ⎥ ⎥⎢ ⎥ ⎥ ⎢ θn ⎥ ⎥⎢ ⎥ ⎥ ⎢ u n+1 ⎥ ⎥⎢ ⎥ ⎥ ⎣ vn+1 ⎦ ⎥ ⎥ θn+1 ⎥ ⎥ ⎥ ⎥ ⎦ (10.4.2) ########### Example 10.4.2 Problem: Solve Example 10.4.1 by finite elements. Solution: Use Equation 10.3.2. We can use just one element as numbered in Figure (a). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 377 Combined Loadings on Slender Bars—Thin Walled Cross Sections node 1 node 2 element 1 Figure (a) The restraints, restraint forces, and loads have been inserted. We have added nodal coordinate subscripts to better keep track of the forces. ⎡ EA ⎤ EA 0 0 0 0 − ⎢ L ⎥ L ⎢ ⎥ ⎢ ⎥ 12EI 12EI 6EI 6EI ⎢ ⎥⎡ ⎤ 0 0 − ⎡ ⎤ ⎢ 3 2 3 2 R1x L L L L ⎥ ⎢ ⎥ 0 ⎥⎢ 0 ⎥ ⎢ R1y ⎥ ⎢ 6EI 6EI 4EI 2EI ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎥⎢ 0 ⎥ 0 − 0 ⎢ Q1 ⎥ ⎢ 2 2 ⎢ ⎥⎢ ⎥ L L L L ⎢ ⎥ ⎢ ⎥⎢ ⎥ (a) ⎢ ⎥= ⎥⎢ ⎥ ⎢F ⎥ ⎢ EA EA ⎢ ⎥ ⎢ u2 ⎥ 2x ⎢ ⎥ ⎢− 0 0 0 0 ⎥ ⎣F ⎦ ⎢ L ⎥ ⎣ v2 ⎦ L 2y ⎢ ⎥ ⎢ 0 12EI 6EI 12EI 6EI ⎥ ⎢ 0 ⎥ θ2 − − 0 − ⎢ L3 L2 L3 L2 ⎥ ⎢ ⎥ ⎣ ⎦ 6EI 6EI 2EI 4EI 0 − 2 0 2 L L L L Partition and solve for the displacements. ⎡ EA ⎤ ⎢ L ⎢ F2x ⎢ ⎣ F2y ⎦ = ⎢ 0 ⎢ ⎢ 0 ⎣ ⎡ 0 0 12EI L3 6EI − 2 L 0 ⎤ --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎥⎡ ⎤ ⎥ u2 6EI ⎥ ⎣ v2 ⎦ − 2 ⎥ L ⎥ ⎥ θ2 ⎦ 4EI L (b) The displacements are ⎡ ⎡ ⎤ ⎢ ⎢ u2 ⎢ ⎣ v2 ⎦ = ⎢ ⎢ ⎢ θ2 ⎢ ⎣ F2x L EA F2y L 3 3EI F2y L 2 2EI ⎡ F cos αL ⎥ ⎢ EA ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ F sin αL 3 ⎥=⎢ ⎥ ⎢ 3EI ⎥ ⎢ ⎦ ⎣ F sin αL 2 2EI ⎤ ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (c) Once the nodal displacements are known the distributed displacements can be found using the shape functions. The restraint forces are found from the other partitioned equation ⎤ ⎤ ⎡ F cos αL ⎤ ⎡ EA ⎡ EA − − 0 0 0 0 ⎥ EA ⎥⎡ ⎤ ⎢ L ⎥⎢ ⎡ ⎤ ⎢ L ⎥ ⎥ u2 ⎥⎢ ⎢ ⎢ R1x ⎢ ⎥ ⎢ ⎢ 12EI 6EI ⎥ 12EI 6EI ⎥ ⎢ F sin αL 3 ⎥ ⎥ ⎥ ⎢ ⎣ R1y ⎦ = ⎢ 0 ⎣ ⎦ v (d) = ⎢ ⎥ − 0 − 2 ⎢ ⎢ ⎥ L3 L2 ⎥ L3 L2 ⎥ 3EI ⎥ θ2 ⎥⎢ ⎢ ⎢ Q1 ⎥ ⎦ ⎦⎢ ⎣ ⎣ 6EI 6EI 2EI 2EI ⎣ F sin αL 2 ⎦ 0 − 2 0 − 2 L L L L 2EI Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 378 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The stresses are obtained from the nodal forces using the usual formulas. In this case of using just one element the element nodal forces are just the applied loads and the restraint forces, that is, ⎤ ⎤ ⎡ R1x p1,1 ⎢ V1,1 ⎥ ⎢ R1y ⎥ ⎥ ⎥ ⎢ ⎢ ⎢ M1,1 ⎥ ⎢ Q 1 ⎥ ⎥ ⎥=⎢ ⎢ ⎢ p1,2 ⎥ ⎢ F2x ⎥ ⎥ ⎥ ⎢ ⎢ ⎣ V1,2 ⎦ ⎣ F2y ⎦ 0 M1,2 ⎡ (e) ########### For a simple frame structure such as the one shown in Figure 10.4.1 we are concerned about both axial and bending displacements and stresses in the same member. The differential equation solution can be found by dividing the frame into two members, the vertical part and the horizontal part, and matching boundary conditions where the two parts meet. It becomes a bit of a chore, especially when the frames become more complicated. We shall go directly to the FEM formulation and find that it is straight forward to obtain a solution. Figure 10.4.1 For a simple frame, like the one in Figure 10.4.1, we must set up a transformation matrix to find the element stiffness matrices for different orientations of the elements. In Section 5.6 we developed the transformation matrix for axial displacements only for use in pin jointed truss problems. It becomes more complicated when both axial and bending displacements are combined. We show an element oriented at an angle α to the x axis in Figure 10.4.2. υα n +1 y υn uα n +1 θ n +1 un +1 υn υα n α un x uα n θn Figure 10.4.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 379 Combined Loadings on Slender Bars—Thin Walled Cross Sections ⎡ Am lm2 Im 0 ⎢ ⎢ ⎢ ⎢ ⎢ 0 E m Im ⎢ ⎢ { pαm } = [kαm ]{rαm } = 3 ⎢ lm ⎢ Am lm2 ⎢− ⎢ Im ⎢ ⎢ 0 ⎣ 0 0 0 12 6lm 6lm 4lm2 0 0 −12 6lm − Am lm2 Im 0 0 0 −12 −6lm Am lm2 0 −6lm Im 0 12 2lm2 0 −6lm ⎤ 0 ⎥ ⎥⎡ u ⎤ αn 6lm ⎥ ⎥⎢ ⎥ v ⎥ αn ⎢ ⎥ 2lm2 ⎥ ⎢ θ ⎥ ⎢ αn ⎥ ⎥ ⎥⎢ ⎥ ⎢ u αn+1 ⎥ ⎥ 0 ⎥ ⎥ ⎣ vαn+1 ⎦ ⎥ θαn+1 −6lm ⎥ ⎦ 2 4lm (10.4.3) Following the steps in the axial case (see Chapter 5, Section 5.6, Equations 5.6.1-5) we can see that the transformation relations at node n are  u αn vαn  =  cos α − sin α sin a cos α  un vn  θαn = θn (10.4.4) The full set of transformations for both nodes is presented in Equation 10.4.5. ⎤ ⎡ cos α u αn ⎢ vαn ⎥ ⎢ − sin α ⎥ ⎢ ⎢ ⎢ θαn ⎥ ⎢ 0 ⎥ ⎢ [rαm ] = ⎢ ⎢ u αn+1 ⎥ = ⎢ 0 ⎥ ⎢ ⎢ ⎣ vαn+1 ⎦ ⎣ 0 θαn+1 0 ⎡ sin α cos α 0 0 0 0 0 0 1 0 0 0 0 0 0 cos α − sin α 0 0 0 0 sin α cos α 0 ⎤⎡ ⎤ 0 un ⎢ ⎥ 0⎥ ⎥ ⎢ vn ⎥ ⎢ θn ⎥ 0⎥ ⎥⎢ ⎥ = [T ]{rm } ⎢ ⎥ 0⎥ ⎥ ⎢ u n+1 ⎥ 0 ⎦ ⎣ vn+1 ⎦ θn+1 1 (10.4.5) The same transformation applies to the internal force matrix, thus { pαm } = [T ]{ pm } (10.4.6) As explained in Chapter 5 a transformation matrix has the interesting property that its inverse is equal to its transpose, that is, [T ]−1 = [T ]T . You can verify that [T ]T [T ] = [1]. Following the same steps used in Section 5.6, Equations 5.6.7 -10, we can state that { pm } = [T ]T { pαm } (10.4.7) and since { pαm } = [kαm ]{rαm } → [T ]{ pm } = [kαm ][T ]{rm } (10.4.8) → [T ]T [T ]{ pm } = { pm } = [T ]T [kαm ][T ]{rm } = [km ]{rm } Thus [km ] = [T ]T [kαm ][T ] Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (10.4.9) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Here we have both u and v components at each node and have added the rotational nodal displacements. From the notation in Figure 10.4.2 we note that P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas 380 July 2, 2011 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 10.4.3 Problem: The frame in Figure 10.4.1 has the dimensions and loads shown in Figure (a). The cross section is 24 mm by 24 mm. It is made of aluminum, E = 68950 MPa. Solve for the displacements and stresses. Solution: Choose three elements as shown in Figure (a). Number the nodes and elements and insert values in the appropriate equations. Note that since the loads are concentrated this will obtain exactly the same answer as the differential equation solution. y 1000 6000 N n3 1000 mm 6000 N e3 n4 e2 n2 e1 1000 mm n1 x Figure (a) ◦ --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- For elements 1 and 2 the angle α is 90 these two elements is ⎡ 0 −1 0 ⎢1 0 0 ⎢ ⎢0 0 1 [k1 ] = [k2 ] = ⎢ ⎢0 0 0 ⎢ ⎣0 0 0 0 0 0 , thus, cos 90◦ = 0 and sin 90◦ = 1. The stiffness matrix for 0 0 0 0 1 0 0 0 0 −1 0 0 where ⎡ Am lm2 Im 0 ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ 0 E m Im ⎢ [kαm ] = 3 ⎢ ⎢ lm ⎢ Am lm2 ⎢− ⎢ Im ⎢ ⎢ 0 ⎣ 0 ⎡ ⎤ 0 0 ⎢ −1 0⎥ ⎢ ⎥ ⎢ 0⎥ ⎥ [kαm ] ⎢ 0 ⎢ 0 0⎥ ⎢ ⎥ ⎣ 0 ⎦ 0 0 1 0 0 12 6lm Am lm2 − Im 0 6lm 4lm2 0 0 0 −12 −6lm 6lm 2lm2 Am lm2 Im 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 −1 0 0 0 −12 −6lm 0 12 −6lm ⎤ 0 0 0 0⎥ ⎥ 0 0⎥ ⎥ 1 0⎥ ⎥ 0 0⎦ 0 1 (a) ⎤ 0 ⎥ ⎥ 6lm ⎥ ⎥ ⎥ 2 ⎥ 2lm ⎥ ⎥ ⎥ ⎥ 0 ⎥ ⎥ ⎥ −6lm ⎥ ⎦ 4lm2 (b) Since all elements are the same length and have the same cross sections we will drop the subscripts m. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 381 Combined Loadings on Slender Bars—Thin Walled Cross Sections When Equation (b) is inserted in Equation (a) and the multiplications are carried out the first two element stiffness matrices are ⎤ ⎡ 12 0 −6l −12 0 −6l ⎥ ⎢ ⎥ ⎢ Al 2 Al 2 ⎥ ⎢ 0 0 0 − 0 ⎥ ⎢ I I ⎥ ⎢ ⎢ 2 2 ⎥ −6l 0 4l 6l 0 2l ⎥ EI ⎢ ⎥ (c) [k1 ] = [k2 ] = 3 ⎢ ⎥ l ⎢ 0 6l 12 0 6l ⎥ ⎢ −12 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ Al 2 Al 2 ⎢ 0 0 0 0 ⎥ − ⎦ ⎣ I I 2 2 −6l 0 2l 6l 0 4l Now we must be careful in the interpretation of the internal nodal forces in the equation for each element. For element 1 { p1 } = [k1 ]{r1 } (d) For element 1, the x direction displacements and forces are lateral forces associated with bending and the y direction displacements and forces are axial. We need a suitable notation to identify the direction of the internal and external forces at each node. Note in Equation (e) we have associated V1,1 with u 1 and P1,1 with v1 , that is, shear forces with the x direction displacement and axial forces with the y direction displacement. Note also that the matrix transformation exhibited in Equation (a) produces the axial stiffness components is the proper row and column. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎡ ⎢ ⎢ ⎢ { p1 } = ⎢ ⎢ ⎢ ⎣ V1,1 P1,1 M1,1 V1,2 P1,2 M1,2 ⎡ 12 ⎢ ⎢ ⎢ 0 ⎢ ⎢ ⎥ ⎢ ⎥ ⎥ EI ⎢ −6l ⎢ ⎥= ⎥ l3 ⎢ ⎢ −12 ⎥ ⎢ ⎦ ⎢ ⎢ ⎢ 0 ⎣ −6l ⎤ 0 −6l −12 −6l ⎥ ⎥⎡ ⎤ 0 ⎥ ⎥ u1 ⎥ ⎢ v1 ⎥ ⎥⎢ 2l 2 ⎥ ⎢ θ1 ⎥ ⎥ ⎥⎢ ⎥ ⎥ ⎢ u2 ⎥ 6l ⎥ ⎢ ⎥ ⎥ ⎣ v2 ⎦ ⎥ ⎥ 0 ⎥ θ2 ⎦ 2 4l Al 2 I 0 0 0 4l 2 6l Al 2 I 0 0 6l 12 0 0 0 2l 2 6l − Al 2 I 0 − ⎤ 0 Al 2 I 0 (e) The same transformation occurs for element 2; however, element 3 needs no transformation and the usual displacements and forces are associated. We have for element 3 with α = 0 ⎡ Al 2 I 0 ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ EI ⎢ 0 [k3 ] = 3 ⎢ l ⎢ ⎢ Al 2 ⎢− ⎢ I ⎢ ⎢ 0 ⎣ 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 0 0 12 6l 6l 0 4l 2 0 −12 −6l 6l 2l 2 − Al 2 I 0 −12 0 −6l Al I 0 0 0 2 0 12 −6l ⎤ 0 ⎥ ⎥ 6l ⎥ ⎥ ⎥ 2 ⎥ 2l ⎥ ⎥ ⎥ ⎥ 0 ⎥ ⎥ ⎥ −6l ⎥ ⎦ 4l 2 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (f) P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 382 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods We must extend this notation to the global matrices. The applied force matrix with restraint forces and applied load added and the nodal displacement matrix with the restraints added become ⎤ ⎡ ⎤ 0 u1 ⎢ v1 ⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ θ1 ⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ u2 ⎥ ⎢ u2 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ v2 ⎥ ⎢ v2 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ θ2 ⎥ ⎢ θ2 ⎥ ⎥ ⎢ ⎥ {r } = ⎢ ⎢ u3 ⎥ = ⎢ u3 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ v3 ⎥ ⎢ v3 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ θ3 ⎥ ⎢ θ3 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ u4 ⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎢ ⎥ ⎣ v4 ⎦ ⎣ 0 ⎦ θ4 0 ⎡ ⎤ ⎤ ⎡ R1x F1x ⎢ F1y ⎥ ⎢ R1y ⎥ ⎢ ⎥ ⎥ ⎢ ⎢ M1 ⎥ ⎢ Q 1 ⎥ ⎢ ⎥ ⎥ ⎢ ⎢ F2x ⎥ ⎢ 6000 ⎥ ⎢ ⎥ ⎥ ⎢ ⎢ F2y ⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎥ ⎢ ⎢ M2 ⎥ ⎢ 0 ⎥ ⎢ ⎢ ⎥ ⎥ {F} = ⎢ ⎥ ⎥=⎢ ⎢ F3x ⎥ ⎢ 0 ⎥ ⎢ F3y ⎥ ⎢ −6000 ⎥ ⎢ ⎥ ⎥ ⎢ ⎢ M3 ⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎥ ⎢ ⎢ F4x ⎥ ⎢ R4x ⎥ ⎢ ⎥ ⎥ ⎢ ⎣ F4y ⎦ ⎣ R4y ⎦ Q4 M4 ⎡ (g) The complete global matrix equation is shown in Equation (h). ⎡ R1x R1y Q1 ⎤ --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ 6000 ⎥ ⎢ ⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎢ 0 ⎥ [F] = ⎢ ⎥ = [K ]{r } ⎢ 0 ⎥ ⎢ ⎥ ⎢ −6000 ⎥ ⎢ ⎥ ⎢ 0 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ R4x ⎥ ⎢ ⎥ ⎣ R4y ⎦ Q4 ⎡ 12 0 −6l −12 0 ⎢ Al 2 Al 2 ⎢ ⎢ 0 0 0 − ⎢ I I ⎢ 6l 0 0 4l 2 ⎢ −6l ⎢ ⎢ ⎢ −12 0 6l 24 0 ⎢ ⎢ 2 Al 2Al 2 ⎢ ⎢ 0 − 0 0 ⎢ I I ⎢ 0 0 −6l 0 2l 2 EI ⎢ ⎢ = 3 ⎢ l ⎢ 0 0 0 −12 0 ⎢ ⎢ ⎢ Al 2 ⎢ 0 0 0 − ⎢ 0 ⎢ I ⎢ 0 0 0 −6l 0 ⎢ ⎢ ⎢ ⎢ 0 0 0 0 ⎢ 0 ⎢ ⎢ 0 0 0 0 0 ⎣ Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 0 0 0 0 0 −6l 0 0 0 0 0 0 0 0 0 0 0 2l 2 0 0 0 0 0 0 −12 0 −6l 0 0 0 0 0 0 0 Al 2 − I 0 0 −12 0 −6l 0 0 8l 2 0 − 6l Al 2 6l 12 + I 0 2l 2 0 6l Al I 0 2 0 2l 2 6l Al 2 + 12 6l I 6l 8l 2 0 0 0 Al 2 − I 0 −12 −6l Al 2 I 0 0 0 6l 2l 2 0 0 12 0 ⎤ ⎥ ⎥ 0 ⎥ ⎥⎡ ⎤ ⎥ 0 ⎥ 0 ⎥⎢ 0 ⎥ ⎥⎢ ⎥ 0 ⎥⎢ 0 ⎥ ⎥⎢ ⎥ ⎥⎢ ⎥ ⎥⎢ ⎥ 0 ⎥ ⎢ u2 ⎥ ⎥⎢v ⎥ ⎥⎢ 2 ⎥ 0 ⎥⎢ θ ⎥ ⎥⎢ 2 ⎥ ⎥⎢u ⎥ ⎥⎢ 3 ⎥ 0 ⎥⎢ ⎥ ⎥ ⎢ v3 ⎥ ⎥⎢ ⎥ ⎥⎢ θ ⎥ 6l ⎥ ⎢ 3 ⎥ ⎥⎢ ⎥ ⎢ ⎥ 2l 2 ⎥ ⎥⎢ 0 ⎥ ⎥⎣ 0 ⎦ ⎥ ⎥ 0 ⎥ 0 ⎥ −6l ⎥ ⎦ −6l 4l 2 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (h) P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come Combined Loadings on Slender Bars—Thin Walled Cross Sections 383 Insert l = 1000 mm, A = 576 mm2 , E = 68950 Mpa and the partitioned equations for the deflections are ⎡ ⎡ ⎤⎡ ⎤ ⎤ 24 0 0 −12 0 −6000 u2 6000 ⎢ 0 ⎢ 0 ⎥ ⎥ ⎢ v2 ⎥ 41667 0 0 −20833 0 ⎢ ⎢ ⎥⎢ ⎥ ⎥ ⎢ ⎥ ⎢ 0 ⎢ 0 ⎥ 0 8000000 6000 0 2000000 ⎥ ⎢ ⎢ ⎥ ⎢ θ2 ⎥ ⎥ (i) ⎢ 0 ⎥ = 1.9063 ⎢ −12 ⎥ ⎢ u3 ⎥ 0 6000 20845 0 6000 ⎢ ⎢ ⎥⎢ ⎥ ⎥ ⎣ 0 ⎣ −6000 ⎦ −20833 0 0 20845 6000 ⎦ ⎣ v3 ⎦ θ3 −6000 0 2000000 6000 6000 8000000 0 Using one of the linear equation solvers (see, for example, Appendix C) we get ⎡ ⎤ ⎡ ⎤ u2 164.05 mm ⎢ v2 ⎥ ⎢ −0.1887 ⎥ mm ⎢ ⎥ ⎢ ⎥ ⎢ θ2 ⎥ ⎢ −0.0329 ⎥ rad ⎢ ⎥=⎢ ⎥ ⎢ u 3 ⎥ ⎢ 0.0661 ⎥ mm ⎢ ⎥ ⎢ ⎥ ⎣ v3 ⎦ ⎣ −0.3775 ⎦ mm θ3 0.1315 rad (j) The internal forces are ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤ V1,1 12 0 −6000 −12 0 −6000 0 −3376.2 N ⎢ P1,1 ⎥ ⎢ 0 ⎥⎢ ⎥ ⎢ 7495.5 ⎥ 20833 0 0 −20833 0 0 N ⎥ ⎢ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎢ M1,1 ⎥ ⎢ ⎢ ⎥ ⎢ 1750877 ⎥ N · mm 0 4000000 6000 0 2000000 ⎥ 0 ⎥ = 1.9063 ⎢ −6000 ⎢ ⎥⎢ ⎥=⎢ ⎥ ⎢ V1,2 ⎥ ⎢ −12 ⎢ ⎥ ⎢ ⎥ 0 6000 12 0 6000 ⎥ N ⎥ ⎢ ⎢ ⎥ ⎢ 164.05 ⎥ ⎢ 3376.2 ⎥ ⎣ P1,2 ⎦ ⎣ 0 ⎦ ⎣ −0.1887 ⎦ ⎣ −7495.5 ⎦ −20833 0 0 20833 0 N M1,2 −6000 0 2000000 6000 0 4000000 −0.0329 1625344 N · mm ⎡ (k) ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤ 12 0 −6000 −12 0 −6000 164.05 2623.8 V2,2 N ⎢ 0 ⎥ ⎢ −0.1887 ⎥ ⎢ 7495.5 ⎥ ⎢ P2,2 ⎥ 20833 0 0 −20833 0 N ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ M2,2 ⎥ ⎥ 0 4000000 6000 0 2000000 ⎥ ⎥ = 1.9063 ⎢ −6000 ⎥ ⎢ −0.0329 ⎥ = ⎢ −1625344 ⎥ N · mm ⎢ ⎢ −12 ⎢ ⎥ ⎢ ⎢ V2,3 ⎥ ⎥ 0 6000 12 0 6000 ⎥ N ⎥ ⎢ ⎥ ⎢ 0.0661 ⎥ ⎢ −2623.8 ⎥ ⎢ ⎣ 0 ⎦ ⎣ −0.3775 ⎦ ⎣ −7495.5 ⎦ ⎣ P2,3 ⎦ −20833 0 0 20833 0 N M2,3 −6000 0 2000000 6000 0 4000000 0.1315 −998435 N · mm ⎡ (l) ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤ 20833 0 0 −20833 0 0 0.0661 2623.8 N P3,3 ⎢ ⎥ ⎢ −0.3775 ⎥ ⎢ 1495.5 ⎥ ⎢ V3,3 ⎥ 0 12 6000 0 −12 6000 N ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎢ ⎢ ⎢ ⎥ ⎢ ⎥ ⎢ M3,3 ⎥ 0 6000 4000000 0 −6000 2000000 ⎥ ⎥ = 1.9063 ⎢ ⎥ ⎢ 0.1315 ⎥ = ⎢ 998435 ⎥ N · mm ⎢ ⎢ −20833 0 ⎥⎢ ⎥ ⎢ −2623.8 ⎥ ⎢ P3,4 ⎥ 0 20833 0 0 0 N ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎢ ⎣ ⎦ ⎣ −1495.5 ⎦ ⎣ V3,4 ⎦ 0 −12 −6000 0 12 −6000 ⎦ ⎣ 0 N M3,4 0 6000 2000000 0 −6000 4000000 0 497059 N · mm ⎡ To find the stresses you must combine the bending and axial terms, that is   My P V 122 2 σ =− + τ= −y I A 2I 4 (m) (n) To find the distributed displacements use the shape functions for each element. The final shape (not to scale) would be approximately as shown in Figure (b) with the bending displacements dominating the axial ones. For example, the vertical displacement of node 2 (−0.1887 mm) due to the axial displacement of element 1 is so small that it can reasonably be neglected; however, that lateral displacement due to bending is quite large. At node 3 both the vertical and horizontal displacement resulting from the axial displacements of the elements also is very small; however, the rotation at node 3 (0.1315 rad) is large enough to produce a significant bending displacement in the member. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas 384 July 2, 2011 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure (b) ########### The solution of even such a simple frame problem like this by either the classical differential equation method or the modern finite element method is quite a chore. Fortunately, the finite element method has been programmed for graphical input of the geometry, automatic creation of nodes and elements, assembly of the matrices, and direct application of loads and restraints to the assembled equations. Partitioning of the equation and their solution is also automated. An additional bonus is the graphical as well as numerical presentation of the results. A tutorial on using a finite element code for solving this and more complicated frame problems is provided separately. 10.5 Bending in Two Planes 10.5.1 When I yz is Equal to Zero So far we have only considered beams that have the yz axes oriented as principal axes of inertia (I yz = 0). This is true for all beams with cross sections symmetrical about at least one of the axes, y or z. We also have limited the discussion to loads only in the xy plane. The simple act of placing a load at an angle to the y axis will induce displacements in both the y and z directions. Consider a beam with a distributed loading represented by both f y (x) and f z (x). If I yz = 0 the problem in each component direction is uncoupled from the other. In addition to the equations we have been using (xy plane) we must solve those in the xz plane. The displacement in the z direction is denoted by w(x). The width of the cross section in the y direction is h. To summarize, these equations for bending in the xz plane when I yz = 0 are Mcy , Fcz → f z (x) Vz (x) M y (x) dM y d Vz = − f z (x) = −Vz (x) dx dx --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS σx (x, z) εx (x, z) u (x, z) w(x) ← ρ dw u (x, z) = −z dx d 2w du = −y 2 εx (x, z) = dx dx d 2w σx (x, z) = Eεx = −E y 2 dx   2 My z d 2w d w z 2 dA = EI yy 2 → σx = − (10.5.1) M y (x) = − σx zdA = E 2 I yy dx dx A A  d Vz d 2w dM y = EI yy 2 = −Vz (x) τzx (x, z) = zdAz dx dx I yy h Az dx d2 d 2 My d 2w = 2 EI yy 2 = f z (x) 2 dx dx dx Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 385 Combined Loadings on Slender Bars—Thin Walled Cross Sections Since we know how to deal with similar equations in the xy plane it should present no particular difficulty to solve these equations for specific cases. The results for the two sets are simply added together, that is, the total magnitude of the displacement is the vector sum of v(w) and w(x) and the normal stresses are σx (x, y, z) = − My Mz y− z Izz I yy (10.5.2) In a similar way we write the finite element equations for the two planes. There is a small problem with sign conventions. In the classical differential equations the sign convention in the xz plane is as shown in Figure 10.5.1. Subscripts have been added to be sure which components are in effect. z Vz My My x Vz Figure 10.5.1 Note that the shear forces are positive by the same rule as shear stresses and the moments are positive if they produce compressive normal stresses in the positive z space (this is the convention we have chosen). This is the same rule as we used in the xy plane and so the equations come out with the same signs. In finite elements the sign convention is to have all internal forces positive according to the right hand rule as shown in Figure 10.5.2. This is desirable for the addition, that is, the assembly, of the element matrices to form the global matrix equations. z Vm, Vm,n+1 element m Mm,n+1 Mm,n node n x node n+1 Note that the moments are in the opposite direction from those in Chapter 8, Figure 8.2.3. This is because in Figure 8.2.3 the z axis is positive in the direction toward the observer and in Figure 10.5.2 the y axis is positive away from the observer. The result is that the element stiffness matrix in the xz plane has similar components but some different signs from those in the xy plane. ⎡ ⎤ Vm,n ⎢ Mm,n ⎥ E m Im ⎥ { pm } = [km ]{rm } = ⎢ ⎣ Vm,n+1 ⎦ = l 3 m Mm,n+1 ⎡ 12 ⎢ −6lm ⎢ ⎣ −12 −6lm −6lm 4lm2 −6lm 2lm2 −12 6lm 12 6lm ⎤ ⎤⎡ −6lm vm,n 2 ⎥⎢ 2lm ⎥ ⎢ θm,n ⎥ ⎥ 6lm ⎦ ⎣ vm,n+1 ⎦ θm,n+1 −4lm2 (10.5.3) In any case the problem can be solved in each plane and the results added for the total effect. The two stiffness matrices can be combined into one as we did for combined axial and bending. In fact the axial, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure 10.5.2 P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 386 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods torsional, and two bending matrices can be combined into one. This gets to be a lot of writing. The FEM computer codes do this nicely for us and there is no need to explicitly write the lengthy equations. 10.5.2 When I yz is Not Equal to Zero When I yz = 0 we must modify the governing equations. We can still assume that plane sections remain plane but now consider the rotation of the cross section about both the y and z axes at the same time. By neglecting the Poisson’s ratio effect and shear deformation we can assume v(x) and w(x) are the lateral displacements. The displacement in the x direction becomes u(x, y, z) = −y dw dv −z dx dx (10.5.4) The strain component of interest is (remember we neglect shear deformation) εx (x, y, z) = d 2v d 2w ∂u = −y 2 − z 2 ∂x dx dx (10.5.5) The axial stress in the interior of the beam is σx (x, y, z) = Eεx = −y E d 2w d 2v − zE 2 2 dx dx (10.5.6) The stress resultant moments are defined by     2 d v d 2w y 2 + z 2 zdA M y (x) = − σx (x, y, z)zdA =E dx dx A A  2    2  d 2w d v d v 2 = E yzdA + z dA = E I yz + 2 2 dx A dx A dx2     2 d v d 2w y 2 + z 2 ydA Mz (x) = − σx (x, y, z)ydA =E dx dx A A  2    2  2 d w d v d v 2 = E y dA + yzdA = E Izz + 2 2 dx A dx A dx2 d 2w I yy dx2  (10.5.7) d 2w I yz dx2  (10.5.8) For a statically determinate beam the moments can be found from static equilibrium and therefore are known quantities. The area moments of inertia and the product of inertia are known for any given cross section. Thus the two equations, Equations 10.5.7 and 10.5.8, can be solved for the second derivatives of the displacements in terms of the moments. 1 I yy Mz − I yz M y d 2v = 2 E I yy Izz − I yz dx2 1 I yz Mz − Izz M y d 2w =− 2 E I yy Izz − I yz dx2 (10.5.9) The stress is found by substituting Equations 10.5.9 into Equation 10.5.6.     d 2w d 2v I yz Mz − Izz M y I yy Mz − I yz M y y+ z σx (x, y, z) = Eεx = −y E 2 − z E 2 = − 2 2 I yy Izz − I yz I yy Izz − I yz dx dx (10.5.10) These equations can be made to look less intimidating by the following notation, let 2 I = I yy Izz − I yz (10.5.11) and Izz = 2 I yy Izz − I yz I yy = I I yy I yy = 2 I yy Izz − I yz Izz = I Izz I yz = 2 I yy Izz − I yz I yz = I I yz (10.5.12) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 387 Combined Loadings on Slender Bars—Thin Walled Cross Sections Equations 10.5.9 and 10.5.10 may then be written 1 I yy Mz − I yz M y d 2v = 2 E I yy Izz − I yz dx2 d 2v = I yy Mz − I yz M y dx2 1 I yz Mz − Izz M y d 2w =− 2 2 E I yy Izz − I yz dx → EI d 2v Mz My = − EIzz EI yz dx2 → d 2w d 2w Mz My → EI 2 = I yz Mz − Izz M y → =− + 2 EI yz EI yy dx dx     I yz Mz − Izz M y I yy Mz − I yz M y y+ z σx (x, y, z) = − 2 2 I yy Izz − I yz I yy Izz − I yz     My My Mz Mz y+ z → σx (x, y, z) = − − − Izz I yz I yz I yy (10.5.13) (10.5.14) Just as a check, if Iyz = 0, these equations reduce to d 2w = dx2 Mz y− σx (x, y, z) = − Izz Mz d 2v = EIzz dx2 My EI yy My z I yy (10.5.15) These are the same as the ones we derived before for Iyz = 0 in Chapter 7 and presented in the first part of this section. ########### Example 10.5.1 Problem: A cantilever beam has a cross section as shown in Figure (a). It is oriented with respect to rectangular Cartesian coordinates as shown. The x axis is the centroidal axis. y y’ d 25 mm 25 mm 100 mm z c z’ 100 mm Figure (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 388 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods A concentrated moment about the z axis is applied at the free end as shown in Figure (b). Find the normal stresses in and the displacement of the beam. y Mz = M x 1m Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Solution: First find the centroid of the cross section. Then we have two choices. One is to find the principal axes of inertia of the cross section with respect to centroidal axes and use Equations 10.5.15. The other is to find the moments and product of inertia with respect to the centroidal axes as shown in Figure (a) and use Equations 10.5.9 and 10.5.10. We will do the latter. First we must find the centroid of the cross section and the centroidal moments and product of inertia. The centroid is found by dividing the cross section into two rectangular portions as shown in Figure (c) and use the transfer formula from Appendix B, Equation B.2.3. The centroid is measured from the y′ z′ axes and is located at  12.5 · 75 · 25 + 50 · 100 · 25 z s As = 33.93 mm (a) = d=  75 · 25 + 100 · 25 As y y’ d = 33.93 mm centroid c = 33.93 mm z z’ Figure (c) From symmetry c = 33.93 mm (b) Now we must find the moments of inertia with respect to the yz centroidal axes. See Appendix B, Equations B.3.5-7. The dimensions from the centroid of the angle to the centroids of the two segments are shown in Figure (d). The centroidal moments of inertia are  75 · 253 25 · 1003 I yc yc + Az c2 = + 75 · 25 · (21.43)2 + + 100 · 25 · (16.07)2 12 12 = 3.688 · 106 mm4 I yy =  (c) From symmetry Izz = 3.688 · 106 mm4 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 389 Combined Loadings on Slender Bars—Thin Walled Cross Sections y 21.43 21.43 28.57 z 16.07 Figure (d) To find the product of inertia we note that, for each rectangular area, one of the offset distances in negative and one is positive. The product of inertia is --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- I yz =  (I yc zc + Ayc z c ) = −75 · 25 · (28.57)(21.43) − 100 · 25 · (21.43)(16.07) (e) = −2.100 · 106 mm4 Next we find the following quantities. 2 = (3.688 · 106 )2 − (−2.100 · 106 )2 = 9.565 · 1012 I = I yy Izz − I yz (f) and noting that the new equivalent quantities are non-italics, Izz , Iyy and Iyz Izz = I yy = I yz = 2 I yy Izz − I yz I yy I yy Izz − 2 I yz Izz 2 I yy Izz − I yz I yz = = = I I yy = 9.565 · 1012 = 2.594 · 106 3.688 · 106 I 9.565 · 1012 = = 2.594 · 106 Izz 3.688 · 106 I I yz = (g) 9.565 · 1012 = −4.761 · 106 −2.100 · 106 The normal stress is    My My Mz Mz y+ z − − Izz I yz I yz I yy     M M y− z =− 2.594 · 106 4.761 · 106 σx (x, y, z) = −  (h) The displacements are found by integrating the following two equations. My M Mz M d 2v − = = 0.3855 · 10−6 = 2 6 EI EI E · 2.594 · 10 E dx zz yz 2 My M Mz M d w = 0.2100 · 10−6 + = =− EI yz EI yy E · 4.761 · 106 E dx2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (i) P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 390 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Integrate each equation twice and apply the boundary conditions. 0.3855 · 10−6 M 2 x + a1 x + b1 2E −6 0.2100 · 10 M 2 x + a2 x + b2 w(x) = 2E v(x) = (j) The boundary conditions are dv(0) =0 dx v(0) = 0 w(0) = 0 dw(0) =0 dx (k) You can quickly see that a1 = b1 = a2 = b2 = 0 (l) The displacements are v(x) = 0.3855 · 10−6 M 2 x 2E w(x) = 0.2100 · 10−6 M 2 x 2E (m) The consequence of a non zero product or inertia is that a moment about the z axis produces displacements in both the y and z directions. In fact the displacement in the z direction is quite large. Example 10.5.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Problem: Repeat Example 10.5.1 by finding the principal axes of inertia and using the uncoupled equations for bending in two planes. Solution: Use Equations 10.2.26-28 and Equations 10.5.15. From Appendix B, the principal axes are rotated by an amount tan 2θ = I yz I yz = Izz − I yy 0 → 2θ = 90◦ → θ = 45◦ This might have been anticipated from the symmetry as shown in Figure (a). y yp 45o z zp Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 391 Combined Loadings on Slender Bars—Thin Walled Cross Sections The principal moments of inertia are   I yy + Izz I yy − Izz 2 2 I max ± = + I yz min 2 2   0 2 6 = 3.688 · 10 ± + (2.100 · 106 )2 = 3.688 · 106 ± 2.100 · 106 2 (b) Thus I yy p = 3.688 · 106 + 2.100 · 106 = 5.788 · 106 mm 4 Izz p = 3.688 · 106 − 2.100 · 106 = 1.588 · 106 mm 4 (c) The moments with respect to the principal axes are M y p = −M cos 45◦ = −0.7071 · M Mz p = M sin 45◦ = 0.7071 · M (d) The normal stress is σx (x, y p , z p ) = − Myp ′ Mz ′ ′ 0.7071 · M 0.7071 · M z −− y = zp − yp 6 I yy p Izz p 5.788 · 10 1.588 · 106 = (0.1222 · z p − 0.4453 · y p )M · 10−6 (e) The displacements are d 2v p = Mz p = 0.7071M dx2 Mz p d 2v p 0.7071M M → = = = 0.4453 · 10−6 2 6 EIz p E · 1.588 · 10 E dx dv p M = 0.420 · 10−6 x + a1 → dx E M x2 + a1 x + b1 → v p = 0.420 · 10−6 E 2 2 d wp EI y p = M y p = −0.7071M dx2 Myp d 2w p 0.7071M M → = = −0.1222 · 10−6 =− EI y p E · 5.788 · 106 E dx2 EIz p (f) dw p M x2 = −0.124 · 10−6 + a2 dx E 2 M x2 + a2 x + b2 → w p = −0.124 · 10−6 E 2 → From the boundary conditions we can see immediately that a1 = b1 = a2 = b2 = 0 (g) And so the displacements are v p = 0.4453 · 10−6 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS M x2 E 2 w p = −0.1222 · 10−6 M x2 E 2 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (h) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-10 P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas 392 July 2, 2011 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods For a quick check let us calculate the stress and displacement at the lower right corner of the angle. From Example 10.5.1 the yz coordinates at the corner are y = −33.9 z = −33.9 (i) From Example 10.5.1, Equation (f), the stress is σx (x, y, z) = (−0.3846y − 0.2092z)M · 106 → σx (x, −33.9, −33.9) = (0.3846 · 33.9 + 0.2092 · 33.9)M · 10−6 (j) = 20.13 · M · 10−6 N/mm 2 The yp zp coordinates at the corner are y p = −47.98 zp = 0 (k) The stress from Equation (e) in this example is σx (x, y p , z p ) = σx (x, −47.98, 0) = (0.4211 · 47.98)M · 10−6 (l) = 20.20 · M · 10−6 N/mm 2 A little round off error has crept in. You can check and find that the displacements also agree. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ########### The above equations may be adapted to solve statically indeterminate cases for I yz = 0. We can use equilibrium to obtain the fourth order equations necessary to solve indeterminate problems that may also be used with determinate cases. We have already noted in Equations 7.3.17 and 10.5.1 that d Vy = − fy dx dMz = −Vy dx d Vz = − fz dx dM y = −Vz dx (10.5.16) By differentiating Equations 10.5.13 and inserting Equations 10.5.16 we obtain Vy Vz d 3v =− + 3 dx EIzz EI yz Vz Vy d 3w =− + 3 dx EI yy EI yz (10.5.17) fz fy d 4w = − dx4 EI yy EI yz (10.5.18) and fy fz d 4v = − dx4 EIzz EI yz Once Equations 10.5.18 are solved for the displacements the bending moments are obtained from Equations 10.5.7 and 10.5.8. To find the shear forces  2   3  d dMz d 2w d 3w d v d v =− E = −E I + I I + I Vy = − zz yz zz yz dx dx dx3 dx3 dx2 dx2 (10.5.19)  2    d d 2w d 3w dM y d v d 3v =− E = −E Vz = − I + I I + I yz yy yz yy dx dx dx3 dx3 dx2 dx2 To find the shear stresses is not so easy. The method we used in Chapter 7, Section 7.3, is accurate, strictly speaking, for rectangular cross sections and can obtain reasonable results for some other sections as discussed in Section 7.9. A true three dimensional elasticity solution may be needed in many cases. We will offer some methods in Chapter 12. Because of the ability to solve fully 3D problems has been Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 393 Combined Loadings on Slender Bars—Thin Walled Cross Sections made easy with the finite element methods, we will see in due course that we can get accurate solutions for fairly complicated cross section shapes. When lateral loads are applied to beams with I yz = 0 it is common to get displacements in both the y and z directions. An additional problem often arises. A lateral load may cause twist as well as bending in two planes. The shear stress is then a combination of that found in bending and that found in torsion. For sections or arbitrary cross section this may become a problem for 3D elasticity theory and not amenable to solving with the equations we have developed so far. There is hope, however, to obtain solutions in the analysis of thin walled cross sections using slender bar theory. That will be considered in Sections 10.6 and 10.7. The finite element formulation for I yz = 0 is available and has been captured in FEM software. The same shape functions are used together with the equations we develop in this section. We shall not write it down here but trust that it has been appropriately captured in the finite element computer codes 10.6 Bending and Torsion in Thin Walled Open Sections—Shear Center In Chapter 7, Section 7.9, we considered the shear flow due to transverse shear forces in thin walled closed and open sections. We determined that the shear flow can be calculated by Equation 7.9.7 repeated here as Equation 10.6.1.  s V byds (10.6.1) q = q0 − 0 I zz The analysis was restricted to sections that were symmetrical about the xy axes plane. Let us extend the theory of thin walled sections to those with I yz = 0 and with loading in the y direction but without symmetry about the y axis. Two examples of such cross sections are shown in Figure 10.6.1. y z Vy Vy Figure 10.6.1 Depending upon where the lateral load is applied, twist as well as bending may occur. There is a position along the z axes on each cross section through which the load must be applied to produce only bending but no torsion. This is defined as the point about which the moment of the applied load is equal to the moment of the shear flow. This point is called the shear center. The locus of such points along the axis is called the elastic axis. The shear center location on the cross section is defined by Equation 10.6.2.  (10.6.2) Vy ez = rqds where ez is the moment arm of the Vy force about some point of reference. This is best understood by an example. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas 394 July 2, 2011 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 10.6.1 Problem: A thin walled cantilever beam has a C shaped cross section. A rigid plate is fastened at the free end so that a concentrated load can be applied at different locations along the z axis as shown in Figure (a). The origin of the yz axes is at the centroid of the cross section. The force is shown applied at the shear center, that is, at a point where the beam will bend without twisting. The problem here is to find the shear flow and the shear center, that is, the location of the force along the z axis. F F y x y z L Figure (a) Solution: Find the shear flow using Equation 10.6.1 and then find the shear center using Equation 10.6.2. The details of the cross section are shown in Figure (b). The uniform wall thickness is b. y’ ec = 2R/π z’ R e ϕ s F ez Figure (b) We define a coordinate s that starts at the bottom edge and follows the centerline of the curved section. It is convenient to use the y′ z′ coordinates to find the centroid and the shear flows. In defining beam bending behavior we switch to centroidal axes. In terms of the radius R and the angle ϕ as shown in Figure (b) we have y ′ = −R cos ϕ z ′ = R sin ϕ ds = Rdϕ The centroid is found to be π  ′ (R sin ϕ) b Rdϕ z dA −R cos ϕ|π0 2R  = = = 0 ec = π Rb π π dA --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) (b) P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 395 Combined Loadings on Slender Bars—Thin Walled Cross Sections The moment of inertia of the cross section is Izz =  A y ′2 dA =  0 π  ϕ  sin ϕ cos ϕ π π R3b + R 2 cos2 ϕ Rbdϕ = R 3 b  = 2 2 2 0 (c) The shear flow is defined as positive in the positive direction of the coordinate s. The value of the shear flow at s = 0 is zero, that is, q0 = 0. This can be stated because there is no shear on the edge at 90o to the cross section. The shear flow in the curved section is q = q0 −  s 0 F Vy ′ y bds = 0 + Izz Izz  ϕ 0 (R cos ϕ) Rbdϕ = FbR2 2F sin ϕ sin ϕ = Izz πR (d) Vy ez = Fez =   rqds → π 0 qRRdϕ = 2F πR  0 π (sin ϕ) R 2 dϕ = − 4FR 2FR cos ϕ|π0 = π π (e) Therefore the shear center is located at ez = 4R π (f) or from the centroid of the section. e = ez − ec = 2R 2R 4R − = π π π (g) The shear center is to the left of the section, approximately as shown in Figure (b). With the moment of inertia for this cross section added the displacement in bending only, as was found in Example 7.3.2, Equation (d), is repeated here. v(x) = F EIzz  x3 Lx2 − 2 6  = F π R 3 bE  Lx2 − x3 3  → v (L) = 2FL3 3π R 3 bE (h) The formula for the normal bending stresses is the same as that found in Example 7.3.2, Equation (f). The normal stresses are σx (x, y) = − F(L − x) 2F(L − x) M y y=− y= Izz Izz π R3b → σx (0, R) = 2FL π R2b (i) Let us add some values to the quantities to get a better feel of the actual behavior of a cantilever beam. Let L = 1000 mm R = 50 mm b = 2 mm E = 206800 N/mm2 F = 1000 N (j) The centroid and the shear center are located at ec = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS 2R = 0.6366R = 31.831 mm π ez = 4R = 1.273R = 63.66 mm π Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (k) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- To find the shear center equate the moment of the applied force to the moment of the shear flow (principle of equivalence). Using Equation 10.6.2 we get P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 396 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The shear flow and the maximum values of the shear flow and shear stress at ϕ = π/2 are q(ϕ) = 2000 2F sin ϕ = sin ϕ = 12.732 sin ϕ N/mm πR π 50 π  = 12.732 N/mm → q 2 π  q 12.732 → τ = 6.366 N/mm2 = = 2 b 2 (l) The bending stress and its maximum value at x = 0, y = 50 are σx (x, y) = 2 · 1000 2F(L − x) y= (1000 − x)y = 0.0025465(1000 − x)y π R3b π · 503 · 2 → σ (0, 50) = 0.0025465(1000)50 = 127.324 N/mm2 (m) The tip displacement is v(L) = 2 · 1000 · 10003 2FL3 = = 4.105 mm 3 3π R bE 3 · π · 503 · 2 · 206800 (n) For the load given these appear to be quite modest values. We shall find some interesting comparisons with the results of a similar problem in Example 10.6.2 coming right up. ########### Having found the shear center and the shear flow associated with a force through the shear center let us ask what happens when the shear force is acting through some point other than the shear center. Clearly some twist will occur. To solve this problem we replace the actual applied force with a statically equivalent applied force of equal value through the shear center and a moment about the shear center. The effect of the moment is a shear stress and a twist as determined from torsional theory for open thin walled sections. In Chapter 6, Section 6.8, we extended the torsional analysis to an open thin walled cross section. We found the stress to be τ= T δ Jeff (10.6.3) where δ can be interpreted to be a coordinate normal to the line midway between the two outer surfaces. If the wall thickness is b then the maximum shear stress is on the two outer edges, that is, τmax = ± T b Jeff 2 (10.6.4) The effective torsional stiffness is given as hb3 3 (10.6.5) where h is the length of the midline and b is the wall thickness. This effective stiffness can be used in the differential equation for the displacement. GJeff dβ =T dx (10.6.6) This equation can be integrated to find the displacement. Once again an example will make this all very clear. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Jeff = P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 397 Combined Loadings on Slender Bars—Thin Walled Cross Sections ########### Example 10.6.2 Problem: The beam in Example 10.6.1 now has a force F applied at the centroid of the cross section at the free end as shown in Figure (a). Find the stresses and the displacements of the beam. y y x L F z F Figure (a) Solution: Resolve the applied load at the centroid into a lateral force applied through the shear center and a torque about the shear center. Add the torsional stress using Equation 10.6.3 to the shear flow found in Example 10.6.1. The equivalent loading for purposes of analysis is to have the force F at the shear center for bending shear analysis and the moment M about the shear center for torsional analysis. When the load is applied through centroid a moment about the shear center is   2R 2R 4R − =F· (a) M=F π π π These loads are shown in Figure (b). The results for lateral displacement and normal stress for this beam are the same as presented in Example 10.6.1.     2 x3 F x3 Lx F 2 − = L x (b) − v(x) = EIzz 2 6 π R 3 bE 3 M F(L − x) 2F(L − x) σx (x, y) = − y = − y (c) y= Izz Izz π R3b y s R ϕ M z 2R π F Location of centroid π Location of shear center 4R b π Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 2R P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 398 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The shear stress in bending was found in Example 10.6.1 to be  s Fb R 2 2F Vy ′ sin ϕ y bds = sin ϕ = q = q0 − I I π R zz zz 0 (d) To this must be added the shear stress of torsion. From Equation 10.6.5 the effective torsional constant is π Rb3 hb3 = 3 3 (e) 3 T 2R F 6F 3 · δ = 2 3δ δ = M 3δ = Jeff hb π π Rb3 π b (f) Jeff = From Equation 10.6.3 the torsional stress is τ= where δ is measured normal to the mid radius and has the maximum values at ±b/2. The torsional displacement for a cantilever beam with a moment at the free end was found in Example 6.3.1. to be M x GJeff (g) 3 6F 2R F M · x = 2 3x x= GJeff π Gπ Rb3 π Gb (h) β(x) = For this case we have β(x) = Let us put in the same numbers, add the shear modulus, and compare. Let L = 1000 mm R = 50 mm G = 79538 N/mm2 b = 2 mm F = 1000 N (i) The bending stress and the lateral tip displacement will be the same as in Example 10.6.1. To the shear stress we must add the effect of the torsional moment and now we will have torsional displacement as well. The distributed torsional shear stress and its maximum value at δ = ±1 mm will be τ (δ) = 6 · 1000 6F δ = 2 3 δ = 75.991δ π 2 b3 π ·2 → τ (±1) = ±75.991 N/mm2 (j) This must be added to the shear stress of bending (see Example 10.6.1, Equation(1)). The maximum value will be at ϕ = π/2, δ = ±1.    π b , ±1 = (6.366 ± 75.991) N/mm2 =τ (k) τ ϕ, ± 2 2 Now let us look at the twist. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- β(x) = 6 · 1000 6F x= 2 x = 0.0009554x rad π 2 Gb3 π · 79538 · 23 → β(L) = 0.0009554 · L = 0.9554 rad = 54.744◦ (l) This is quite a large rotational displacement and may very well be unacceptable. In fact, using the undeformed geometry to define the moment is no longer valid. Open sections are very weak in torsion and are generally avoided as load carrying members. ########### We can extend this to thin walled beams with I yz = 0. From Equation 7.9.4 we have ∂σx ∂q = −b ∂s ∂x Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (10.6.8) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 399 Combined Loadings on Slender Bars—Thin Walled Cross Sections When we insert the value of the normal stress in Equation 10.5.14 into Equation 10.6.8 we get          d Vz Vz My My Vy Vy ∂q Mz Mz =b y− z = − + by − − bz (10.6.9) − − + ∂s dx Izz I yz I yz I yy Izz I yz I yz I yy The shear forces to be used in Equation 10.6.9 are found using Equations 10.5.19. ########### 10.7 Bending and Torsion in Thin Walled Closed Sections—Shear Center Let us extend the theory of thin walled closed sections to those with I yz = 0 and with loading in the y direction but without symmetry about the y axis. An example of such a cross section is shown in Figure 10.7.1. y z Vy Figure 10.7.1 Depending upon where the lateral load is applied, twist as well as bending may occur. As in the case of the open section we shall first find the shear for pure bending and the location of the shear center. When the transverse force is not through the shear center we will resolve it into a force at the shear center and a moment about the shear center. The same equations for the shear flow and shear center used in Section 10.6 apply here. q = q0 −  0 s Vy byds Izz V y ez =  rqds (10.7.1) In this case of bending only we shall use the condition of no twist to find the value of q0 in Equation 10.7.1. Consider the element under strain as shown in Figure 10.7.2. q γ dx ds q T x Figure 10.7.2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-10 P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 400 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Since we are using the shear flow q we assume the stress is constant through the thickness and we can identify the geometry of the element by its mid plane surface as shown in Figure 10.7.3. γ υt (s) dx dβ r Figure 10.7.3 The displacement vt (s) is in the direction of the coordinate s. From Hooke’s law γ = τ q = G Gb (10.7.2) From the Figure 10.7.2 vt = γ d x = r dβ γ =r → q dβ = dx Gb (10.7.3) If we integrate around the section  q ds = Gb  r dβ ds dx and assume that dβ/dx is a constant around the section then from Equations 6.5.6-8   dβ dβ rds = 2A T · 2A T rds = → dx dx where AT is the enclosed area of the section. Thus we find the rate of twist is  1 q dβ = ds dx 2A T Gb (10.7.4) (10.7.5) (10.7.6) This applies to any value of q. When q is constant, as in the case of pure torque, from Equation 6.5.9   T 1 T q 1 dβ q= = ds = ds (10.7.7) → 2A T dx 2A T Gb Gb 4A2T Equation 10.7.7 is the result we got in Chapter 6, Section 6.6. From Equation 10.7.7 we defined GJeff dβ =T dx (10.7.8) where --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 4A2 Jeff =  T ds b (10.7.9) Equation 10.7.6 will be a valuable tool in finding the shear flow and shear center for a beam with a closed thin walled cross section. The next example illustrates this. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come Combined Loadings on Slender Bars—Thin Walled Cross Sections 401 ########### Example 10.7.1 Problem: Find the shear flow and shear center in the closed single cell thin walled cross section shown in Figure (a). The shear force is applied at the shear center of the cross section in the y direction. y’ C sCA 1 z’ B R ϕ sAB A ez F Figure (a) y ′ = −R cos ϕ z ′ = R sin ϕ dsAB = Rdϕ just as we did in Example 10.6.1. The shear flow in the curved part is   s FbR2 ϕ FbR2 F ′ by dsAB = qA + sin ϕ cos ϕdϕ = qA + qABC = q0 − Izz 0 Izz 0 I zz (a) (b) Note that qABC at ϕ = π , that is at point C, equals qA . Then, along the vertical web qC = qA and dsCA = −dy ′ .  y′ Fb ′2 F ′ ′ by dy = qA + (y − R 2 ) (c) qCA = qC + 2Izz −R I zz To find qA assume the section does not twist. Use Equation 10.7.6.   1 q dβ = ds = 0 → qds = 0 dx 2A T Gb      π  −R  Fb ′2 FbR2 2 qA + qA + qds = sin ϕ Rdϕ + (y − R ) (−dy ′ ) Izz 2Izz R 0  ′3 −R  FbR3 Fb y 2 ′  − R − cos ϕ|π0 − qA y ′ |−R y = RqA ϕ|π0 − R  Izz 2Izz 3 R   2R 3 FbR3 Fb 3 + 2R − = RqA π − (−2) − qA (−2R) − Izz Izz 3   FbR3 4 4 FbR2 = 0 → qA = − = (Rπ + 2R)qA + Izz 3 3(π + 2) Izz Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Solution: Use Equations 10.7.1 and 10.7.6. For the curved portion use P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 402 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The total shear flow is qABC = qA + FbR2 4FbR2 FbR2 FbR2 sin ϕ = − + sin ϕ = Izz 3(π + 2)Izz Izz Izz Fb 2 4FbR2 Fb 2 qCA = qA + (y − R 2 ) = − + (y − R 2 ) 2Izz 3(π + 2)Izz 2Izz   Fb 1 4R 2 = + (y 2 − R 2 ) − Izz 3(π + 2) 2  − 4 + sin ϕ 3(π + 2)  (e) (f) To find the shear center, let   FbR2 qA + sin ϕ R 2 dϕ Izz 0 s    π 4 FbR2 + sin ϕ R 2 dϕ − = Izz 3(π + 2) 0 π 4   4ϕ FbR FbR4 2(π + 6) − cos ϕ  = − = Izz 3(π + 2) Izz 3(π + 2) 0 Fez = or  qrds =  π ez = (g) b R 4 2(π + 6) Izz 3(π + 2) (h) Let us assign some numerical values. L = 1000 mm R = 50 mm b = 2 mm E = 206800 N/mm2 F = 1000 N (i) The moment of inertia is b(2R)3 π · 503 · 2 2 · 1003 π R3b + = + 2 12 2 12 = 392699.08 + 166666.67 = 559365.75 mm 4 Izz = (j) The shear center is located at ez = 2 · 504 · 2(π + 6) b R 4 2(π + 6) = = 26.488 mm Izz 3(π + 2) 559365.75 · 3(π + 2) (k) The shear flows are   4 FbR2 + sin φ = 8.939 (−0.2593 + sin ϕ) − Izz 3(π + 2) = −2.3180 + 8.9387 sin ϕ      1 2 1 y2 4R 2 4 FbR2 Fb 2 + (y − R ) = + − − −1 = Izz 3(π + 2) 2 Izz 3(π + 2) 2 R 2    2 y −1 = −6.787 + 0.0017878y 2 = 8.939 −0.2593 + 0.5 2500 qABC = qCA (l) In the above analysis the positive direction of the shear flow in the vertical web was in the negative y direction, that is, from point C to A. Let us redefine it as positive in the positive y direction and label it qAC , then qAC = 6.787 − 0.0017878y 2 (m) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 403 Combined Loadings on Slender Bars—Thin Walled Cross Sections The shear flows qABC and qAC are plotted in Figure (b). Note that the shear flow qABC follows the curved web and the vertical component of the shear flow in the curved web and that in the vertical web add up to a total force equal to the applied load. 180 160 140 ϕ (Degree) 120 100 80 60 40 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 20 0 –3 –2 –1 0 1 2 3 qABC (N/mm) 4 5 6 7 2.5 3 3.5 4 4.5 5 qAC (N/mm) 5.5 6 6.5 7 50 40 30 20 y (mm) 10 0 –10 –20 –30 –40 –50 2 Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 404 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Example 10.7.2 Problem: The shear force for the beam with the D shaped cross section in Example 10.7.1 is applied along the vertical web as shown in Figure (a). Find the shear flow in the section. y’ C sCA z’ B R ϕ sAB A F Figure (a) Solution: Resolve the applied force into a force through the shear center and a moment about the shear center. Add the shear flow from that moment to the shear flow found in Example 10.7.1. The equivalent shear force and moment are shown in Figure (b). y’ C T z’ B sCA R ϕ sAB F Figure (b) The moment causing twist is T = Fez = FbR4 2(π + 6) Izz 3(π + 2) (a) The shear flow from this moment is qT = − Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS FbR2 (π + 6) T FbR4 2(π + 6) =− =− 2 2A T 2π R Izz 3(π + 2) π Izz 3(π + 2) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- A P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 405 Combined Loadings on Slender Bars—Thin Walled Cross Sections The minus sign in needed because the shear flow in Example 10.7.1 was positive in the counter clockwise direction. Add this to the values found in Example 10.7.1, Equations (e) and (f). ########### 10.8 Stiffened Thin Walled Beams --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Thin walled sections are often combined with stiffeners such as shown in Figure 10.8.1a. Consider that this is a section of a beam with a lateral load acting in the y direction through the centroid of the section. Assume the bending moment Mz and the shear force Vy have been found. y A A A = A + bh/ 2 b z h s (a) (b) (c) Figure 10.8.1 The discussion of the solution for shear flow in the web of an I beam in Section 7.9 applies here. There are two addition assumptions to simplify the analysis in common use. In Figure 10.8.1 (b) we depict the stiffener area as acting at a point but maintain the web thickness. The moment of inertia contributed by the two stiffeners is then approximated as  2 h2 h A (10.8.1) A= (Izz )st = 2 2 2 where A is the area of the stiffener. This neglects the moment of inertia of the stiffeners about their own centroids. This is a common approximation when h is sufficiently large. The total moment of inertia in case (b) is then bh 3 h2 A+ 2 12 Izz = (10.8.2) In Figure 10.8.1 (c) if the moment of inertia contributed by the web is small compared to the contribution of the stiffener we can neglect it entirely or we can lump the web area in with the stiffener area by adding half of the web area to each stiffener. This new area that we label A is then used in the various equations. A= A+ bh 2 (10.8.3) The corresponding moment of inertia is  Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS I zz  st =2  2 h2 h A= A 2 2 (10.8.4) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 406 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The normal stress on the stiffener is σx = − M y Izz → (σx )st = − Mh M M h = y=− Izz Izz 2 A (10.8.5) The shear flow in then a constant q= Vy h (10.8.6) Another way to approach the solution is to consider the equilibrium of a free body of the segment of the top stiffener as shown in Figure 10.8.2. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P dx qc P+ dP dx dx Figure 10.8.2 Summing forces we get  Fx = P + dP d x − P − qdx = 0 dx → qc = d d dP = (σx A) = − dx dx dx  dM Vy h h M A y A = − dx A = Izz Izz 2 Izz 2 (10.8.7)  We can use either A or A in Equation 10.8.6. When we insert the moment of inertia in Equation 10.8.6 we get qc = 2Vy h Vy Vy h A= 2 A= Izz 2 h A2 h (10.8.8) We can replace A with A and get the same result for the shear flow. Using A the normal stress will be lower owing to the larger value of the moment of inertia. Actually since the shear flow is a constant in the web we can go directly to qc = Vy h (10.8.9) This is depicted in Figure 10.8.3 (c). If instead we use the approximation in Figure (b) the shear flow in the web is   Vy s Vy y Vy h Vy h A− A− ybdy ybds = qb = Izz 2 Izz 0 Izz 2 Izz − h2 (10.8.10)    2 Vy h Vy b 2  y Vy h = A− − y2 hA + b y −h = 2 Izz 2 2Izz 2Izz 4 where the moment of inertia is given by Equation 10.8.2. This distribution of shear flow is depicted in Figure 10.8.3 (b). Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 407 Combined Loadings on Slender Bars—Thin Walled Cross Sections y A A+bh/2 A b z h (a) (b) (c) Figure 10.8.3 A numerical example will provide a measure of the validity of these approximations. ########### Example 10.8.1 Problem: The cross sections in Figure 10.8.1 are given the following numerical parameters: The stiffeners are 10 mm by 10 mm. The height is 100 mm and the web thickness is 2 mm. Compare the two approximations in Figures 10.8.3 (b) and (c). Solution: Calculate the normal and shear stresses by the two approximations. First let us check the accuracy of neglecting the moment of inertia of the stiffeners about their own centroids. Including the terms neglected  2 1002 104 a4 h + 100 A=2 (Izz )st = 2 + 2 12 2 12 2 (a) = 1,666.67 + 500,000 = 501,666.67 mm 4 Including the moments of inertia about the stiffeners’ centroids increases the total by only a factor of 0.0033. It is reasonable to neglect them. In this case if we add the shear web to the moment of inertia we have bh 3 1002 2 · 1003 h2 A+ = 100 + = 500,000 + 166,666.67 = 666666.67 mm 4 (b) Izz = 2 12 2 12 The shear stress in Figure 10.8.2 (b) is    2 Vy Vy h − y2 [15000 − 2y 2 ] = hA + b (c) qb = 2Izz 4 1333333 At the stiffeners (y = ±50 mm) the value is Vy [15000 − 2 · 502 ] = 0.0075 · Vy qb = (d) 1333333 At the centroid of the whole section (y = 0) the value is Vy [15000 − 2 · 02 ] = 0.01125 · Vy qb = (e) 1333333 Now let us try the approximation of Figure 10.8.1 (c). We know that the shear flow is constant and therefore Vy = 0.01 · Vy (f) qc = h --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 408 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Example 10.8.2 Problem: Compare the shear flow magnitudes and distributions in a thin walled rectangular cross section without and with stringers. Compare the cross section in Example 7.9.1 with two solutions in Example 10.8.1. From Example 7.9.2 the cross section is shown in Figure (a). y h z s Vy Figure (a) In Example 7.9.1 we found the shear flow to be q= Vb 2Izz  h2 − y2 4  (a) Given the same dimension and properties as in Example 10.8.1 and Vy = 1000 N we have q = 0.006(2500 − y 2 ) (b) In Example 10.8.1 we found the shear flow to be qb = Vy [15000 − 2y 2 ] = 0.0015(7500 − y 2 ) 1333333 These two values along with a uniform shear flow are plotted in Figure (b). y 50 q 0 8 16 -50 Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 409 Combined Loadings on Slender Bars—Thin Walled Cross Sections The solid line refers to the section with no stiffeners, Equation (b); the dashed line refers to the section with stiffeners, Equation (c); and the dotted line refers to the case where a uniform shear flow is assumed. A section with a larger stiffener area would approach the constant case. Note that adding stiffeners decreases the normal stress and decreases the maximum value of the shear flow; however, the integral of the shear flow will always add up to the applied load. Thus, the area under each of those curves in Figure (b) will be the same. ########### This approximation may be extended to more complex shapes and is especially helpful for closed sections including multi cell sections. The next example considers a single cell thin walled section with stiffeners. ########### Example 10.8.3 Problem: Find the stresses in the single cell thin walled section shown in Figure (a). The dimensions are the same as in Example 10.8.1. The load is in the y direction and is acting through the shear center. Assume the bending moment Mz and the shear force Vy have been found. Compare approximations in Figures (b) and (c). y A A A or A C bb z B h A Vy Figure (a) Figure (b) Figure (c) Solution: Find the shear flows using the same equations as in Example 10.7.1. Use equilibrium of the shear flows at the stiffener and the condition of no twist to find the unknown shear flows at the start of each web. First consider the case in Figure (b) as shown in Figure (d). y’ C sCA z’ B --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS R ϕ sAB A F Figure (d) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 410 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The shear flows are qABC = qA + FbR2 sin ϕ Izz qCA = qC + Fb 2 (y − R 2 ) 2Izz (a) To find qA and qC , place the upper stiffener in equilibrium as shown in Figure (e). P qA dP dx dx P+ qC Figure (e) Summing forces we get  Fx = P + dP d x − P + (qC − qA )d x = 0 dx → qA − qC = d dP d = (σx A) = − dx dx dx  dM F M y A = − dx R A = RA Izz Izz Izz (b)  This is one equation in two unknowns, qA and qC . To find another equation use the condition of no twist.    −R    π Fb 2 FbR2 qC + qA + sin ϕ Rdϕ + (y − R 2 ) (−dy ′ ) qds = Izz 2Izz R 0  3 −R 3  FbR Fb y 2  − R − cos ϕ|π0 − qC y|−R y = RqA ϕ|π0 − R  Izz 2Izz 3 R (c)   3 3 2R Fb FbR 3 + 2R − (−2) − qC (−2R) − = RqA π − Izz 2Izz 3   FbR3 4 4FbR2 = 0 → πqA + 2qC = − = RπqA + 2RqC + Izz 3 3Izz Solving the two equations provides qA = F (π + 2)Izz  hA − 4b R 2 3  qC = qA − F h A Izz 2 (d) Let us put some numbers in since all these symbolic quantities are difficult to interpret. Using the same values from Example 10.7.1 and with stiffener areas of 100 mm2 we add the stiffeners to the moment of inertia. Izz = 559365.75 + 2R 2 A = 559365.75 + 2 · 502 · 100 = 1059365.75 mm 4 (e) Given these values F 1000 · 50 · 100 = 4.7198 N/mm RA = Izz 1059365.77 4FbR2 4 · 1000 · 2 · 2500 = −6.2930 N/mm πqA + 2qC = − =− 3Izz 3 · 1059365.75 qA − qC = --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (f) P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 411 Combined Loadings on Slender Bars—Thin Walled Cross Sections Solving for qA and qC we get qA = 0.6120 N/mm qC = −4.1078 N/mm (g) The shear flows in the webs are 1000 · 2 · 2500 FbR2 sin ϕ = 0.6120 + 4.7198 sin ϕ sin ϕ = 0.6120 + Izz 1059365.75 1000 · 2 Fb 2 (y 2 − 2500) (y − R 2 ) = −4.1078 + = qC + 2Izz 2 · 1059365.75 = −4.1078 + 0.000944(y 2 − 2500) → qAC = 4.1078 − 0.000944(y 2 − 2500) qABC = qA + qCA (h) Fez =  π 0 qABC R 2 dϕ =  π 0 (0.6120 + 4.7198 sin ϕ)502 dϕ = (0.6120ϕ − 4.7198 cos ϕ)|π0 502 = (0.6120 · π − 4.7198(−2))502 = 1000ez (i) → ez = 28.4057 mm The shear flows are plotted as dashed lines in Figure (f). The solid lines are the values obtained for the section with no stiffeners from Example 10.7.2. 180 No Siffeners With Stiffeners 160 140 ϕ (Degree) 120 100 80 60 40 20 0 –3 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS –2 –1 0 1 2 3 qABC (N/mm) 4 5 6 7 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Now find the shear center. P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 412 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 50 No Siffeners With Stiffeners 40 30 20 y (mm) 10 0 –10 –20 –30 –40 –50 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 qAC (N/mm) Figure (f) The idealization in Figure (c) assumes that the bending stresses are carried entirely by the stiffeners. Depending on the relative size of the stiffeners and the thickness of the webs we can either ignore the web thickness or apportion it to the stiffener areas. The unknown shear flows are labeled in Figure (g). Since all the bending stress is carried by the stiffeners the shear flows in the webs are constant values. y C B z qAC qABC A F Figure (g) In either case we can place a segment of each stiffener in equilibrium. The upper stiffener is shown in Figure (h). We have represented the shear flow q AC in the vertical web in the direction we anticipate is positive. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 413 Combined Loadings on Slender Bars—Thin Walled Cross Sections P qABC dP dx dx P+ q AC Figure (h) Summing forces --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P+ dP d x − P − qABC d x − qAC d x = 0 dx → qABC + qAC dM d F dP = (σx A) = − dx y A = RA = dx dx Izz Izz (j) Note that Izz = 2R 2 A (k) and so qABC + qAC = F FRA F = RA = Izz 2R 2 A 2R (qABC + qAC )2R = F → (l) This is one equation containing the two unknown shear flow components. We need another equation. Setting twist to zero and finding the shear center is one way.  qds = qABC π R − qAC 2R = 0 πqABC − 2qAC = 0 → (m) Solving the two simultaneous Equations (d) and (e) we get qABC = F (π + 2)R qAC = πF 2(π + 2)R (n) Equate the applied moment to the moment of the shear flow to find the shear center. Fez = qABC π R · R = Fπ R (π + 2) → ez = πR (π + 2) (o) Now let us put some numbers in. 1000 F = = 3.890 N/mm (π + 2)R (π + 2)50 π 50 πR ez = = = 30.55 mm (π + 2) (π + 2) qABC = qAC = π 1000 πF = = 6.110 N/mm 2(π + 2)R 2(π + 2)50 (p) These shear flow values are added to the previous figures as shown in Figure (i). Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 414 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 180 160 140 ϕ (Degree) 120 No Siffeners With Stiffeners Uniform Shear Flow 100 80 60 40 20 0 –3 –2 –1 0 1 2 3 qABC (N/mm) 4 5 6 7 50 40 30 20 y (mm) 10 No Siffeners With Stiffeners Uniform Shear Flow 0 –10 –20 –30 –40 –50 2 2.5 3 3.5 4 4.5 5 qAC (N/mm) 5.5 6 6.5 7 Figure (i) Note that when the assumption of constant shear flows in the web is invoked the size of the stringers has no effect on the shear. Normal stress will be quite different, of course. ########### --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 14:35 Printer Name: Yet to Come 415 Combined Loadings on Slender Bars—Thin Walled Cross Sections The analysis of multi cell sections proceeds in the same manner. It gets to be quite complicated and so we will just outline the procedure in the next example for the case where the shear flows are constant in each web. In Chapter 12 we will provide a computer based method of analysis for solving for shear flow in thin walled sections that takes some of the pain out of the effort. ########### Example 10.8.4 Problem: Consider the multi cell section shown in Figure (a). Given the shear force find the shear flow in the section and the shear center. y h A bb h z Vy Figure (a) Solution: Use the approximation that all bending stresses are carried in the stiffeners and the shear flow is constant in each web between stiffeners. The section is idealized as in Figure (b). q23 3 A or A 4 2 q12 q53 q543 5 1 q51 Figure (b) Set up free body diagrams of each stiffener as shown in Figure (c). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-10 P2: ABC JWST071-Waas July 2, 2011 416 14:35 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods P2 P3 q23 q23 q543 q53 P3 + q53 P5 dP3 dx dx q12 q53 P2 + dP2 dx P1 q1 2 q51 q51 q543 P5 + dP5 dx dx P1 + dP1 dx Summing forces on each of the stiffeners will provide four equations in five unknowns. To get the fifth equation equate the moment of the applied load to the moment of the shear flows. 10.9 Summary and Conclusions In this chapter we have been exploring ever increasing complexity of structural problems. In Section 10.3 we examined combined torsional and axial loading which, fortunately, are uncoupled and can be combined by simple superposition. The combined axial and bending problems in Section 10.4 are a bit more complicated because of the variety of geometrical arrangement of the frame members. And bending in two planes in Section 10.5, especially when Ix y = 0, presents some challenges. In all three cases the FEM provides advantages in solving problems. The analysis of thin walled cross sections of Sections 10.6-8 extends our under standing of beam bending and torsional theory. It soon becomes evident that this analysis becomes tedious and cumbersome beyond the simple examples given there. Clearly we need to develop ever more powerful methods. We have formulated problems that use the capabilities of linear equation solvers in Chapters 5, 8, and this chapter. Now we must go on to still more powerful methods. Actually we have been using them without formulation to illustrate 3D solutions that justify our 1D approximate methods. These will be introduced in more detail in the next two chapters. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (c) P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 11 Work and Energy Methods—Virtual Work 11.1 Introduction There is an alternative way of arriving at Newton’s laws of static equilibrium different from the method of summation of forces and moments using a free body diagram. These are methods based on work and energy. Equating work done directly to strain energy stored has limited usefulness as described in Chapter 4, Section 4.11. The displacement of an axially loaded bar or pin jointed truss can be found at the point of application of a single force in the direction of that force if the body is statically determinate. Castigliano’s second theorem, introduced in the same section (Chapter 4), extends the finding of the displacement at the point of application of a concentrated force to include multiple forces but the body must still be statically determinate. In Chapter 6, Section 6.6, we used the argument that work done equals strain energy stored to define the effective torsional stiffness of a thin wall section. In Chapter 7, Section 7.5 we extended these arguments and Castigliano’s second theorem to beams in bending. Now we extend consideration of work and strain energy principles to provide general solutions to indeterminate as well as determinate bodies. There are several different formulations of the work and energy methods that may be found in the various references that extend these methods. We shall present here only the principle of virtual work. This is sufficient for our immediate needs. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 11.2 Introduction to the Principle of Virtual Work Setting the summation of forces and moments to zero on a free body diagram as a statement of equilibrium is intuitively believable and simple to understand. The principle of virtual work is an equivalent statement for bodies that are in equilibrium. To cast it is its simplest possible form (in the hope that this will gain your confidence in believing it in more complicated cases) we consider a single particle in equilibrium. Quite obviously the summation of forces acting on the particle must be zero if it is in equilibrium. The principle of virtual work (PVW) states the very same thing in the following way: A particle is in equilibrium if and only if the virtual work of all the forces acting on the particle is zero during an arbitrary virtual displacement. Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 418 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Let us first examine what a virtual displacement is. We are not speaking of an actual displacement caused by the action of forces but of an imagined (virtual) displacement with no change in the forces that are acting. Instead we ask what work (virtual) would be done if the particle was displaced (virtually) with no changes in the values of the forces. If the virtual work is zero then that body is in equilibrium. Consider a single particle acted upon by several forces as shown in Figure 11.2.1. F1 y F2 δa F3 … x Fn z Figure 11.2.1 If the particle is in equilibrium the imagined (virtual) work δW done by the imagined (virtual) displacement δa (with components δu, δv, δw in the three coordinate directions) would be δW =  Fi x δu +  Fi y δv +  Fi z δw = 0 (11.2.1) and the PVW states that δW = 0. Thus, for arbitrary values of δu, δv, δw it follows that for δW = 0 it must be that    Fi y = 0 Fi z = 0 (11.2.2) Fi x = 0 So, in this case at least, the principle of virtual work and Newton’s laws are equivalent. Will it be harder to believe in more complicated cases? Will you just accept it on faith for a while until a number of illustrations of the use of it have succeeded? After all you are taking Newton’s laws on faith in much the same way. You believe them because you have had so many illustrations of their truth; that is, they are verified in practice. Now let us consider a slightly more complicated case. Let there be a linear spring with spring constant k attached to the particle and fixed at the other end as shown in Figure 11.2.2. For convenience of illustration let us restrict all forces to lie in the xy plane and have the x-direction coincide with the direction of the spring force. Now imagine the particle to have achieved a state of equilibrium under a spring extension u, that is, under the action of the actual forces the particle is displaced until the spring force exactly equals the sum of the applied forces, and at that state the spring has stretched and amount u. F1 y F2 δa k x F3 … Fn z Figure 11.2.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 419 Work and Energy Methods—Virtual Work Now, the total virtual work is defined as the sum of the external virtual work δW e and the internal virtual work δW i where the internal virtual work is equal to the negative of the virtual strain energy δU , that is, (11.2.3) Now, give the particle a virtual displacement with components δu and δv. We know that, at equilibrium, the linear spring has a force (11.2.4) F = ku The spring will store strain energy by an amount shown in the crosshatched region in Figure 11.2.3 as a result of the initial displacement. F k δu u Figure 11.2.3 That strain energy stored is U= 1 1 Fu = ku 2 2 2 (11.2.5) When a virtual displacement is imposed, the virtual strain energy stored is shown by the shaded rectangular area in Figure 11.2.3, that is, by neglecting higher order terms δW i = −δU = − [U (u + δu) − U (u)]   1 1 = − k (u + δu)2 − ku 2 2 2    1 1  2 = − k u + 2uδu + δu 2 − ku 2 2 2 (11.2.6) = −(ku)δu Remember that the forces including the spring force do not change under a virtual displacement therefore the external virtual work is  (Fi x δu + Fi y δv) (11.2.7) δW e = Now according to the principle of virtual work δW is zero if the body is in equilibrium or  (Fi x δu + Fi y δv) − (ku)δu = 0 δW = δW e + δW i = δW e − δU = (11.2.8) By regrouping terms we have δW = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS     Fi x − ku δu + Fi y δv = 0 (11.2.9) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- δW = δW e + δW i = δW e − δU P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 420 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods To obtain equilibrium for arbitrary virtual displacements δu and δv it must be that   Fi y = 0 Fi x − ku = 0 (11.2.10) These are the familiar equations of equilibrium obtained from Newton’s laws; however, this time obtained from virtual work. Let us now extend the PVW to an elastic body such as a slender bar. The principle of virtual work, in this case, is stated as A deformable solid body is in equilibrium if and only if the total virtual work is zero for any virtual displacement that does not violate the geometric constraints. Such a virtual displacement is called kinematically admissible. Now consider the principle of virtual work applied to the slender bar in Figure 11.2.4. We wish to find the displacement at the end where the load is applied. F L Figure 11.2.4 Since the stress and strain in the bar are constant we can define the strain in the bar to be uL εx = L (11.2.11) and from Hooke’s law the stress is (11.2.12) σx = Eεx Thus the strain energy in the bar in terms of displacement u L is UA = 1 2 εx σx d V = 1 2 V L εx2 E Ad x = 0 1 2 L 0 EA 2 u 2L u E Ad x = L2 2L L (11.2.13) Impose a virtual displacement δu L , that is, the virtual displacement of the bar at u(L). The external virtual work would be δW e = Fδu L (11.2.14) The virtual strain energy would be (neglecting higher order terms) δU A = U A (u L + δu L ) − U A (u L ) = EA EA EA (u L + δu L )2 − (u L )2 = u L δu L 2L 2L L (11.2.15) The body is in equilibrium if δW = δW e − δU A = Fδu L − EA u L δu L = δu L L F− EA uL L =0 (11.2.16) Since δu L is an arbitrary virtual displacement the above equation is true only if F− EA uL = 0 L → uL = FL EA (11.2.17) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 421 Work and Energy Methods—Virtual Work This is exactly what we get using Newton’s laws for equilibrium. It is also exactly what we got equating work to strain energy in Chapter 4, Section 4.11. It is also what we get using Castigliano’s second theorem. By now you should be a little more convinced that the PVW gets the same results as Newton’s laws for equilibrium for our purposes. This was done for a statically determinate bar with a special case of axial loading. In the next section we shall extend the results of the principle of virtual work to the general case of a slender bar in axial, torsional, and bending displacement with distributed and concentrated loadings. 11.3 Static Analysis of Slender Bars by Virtual Work We shall repeat the derivation of the general (FEM) method for slender bars using the principle of virtual work. This is an excellent use of virtual work and will lead us to an extension of this approach to more complicated problems. 11.3.1 Axially Loading Consider a slender bar with axial concentrated loads, a constant cross section A, and Young’s modulus E. The bar is divided into nodes and elements as shown in Figure 11.3.1. There must be a node at the location of each concentrated load and at other points where you wish to define a nodal displacement or have the stress reported. element m x node n node n+1 s pm,n, un pm,n+1, un+1 Figure 11.3.1 We define the nodal displacements and internal nodal forces for element m as shown in Figure 11.3.1. {rm } =  un u n+1  { pm } =    pm,n pm,n+1  (11.3.1) The virtual displacements are δu n {δrm } = δu n+1 (11.3.2) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Now, we treat the element shown in Figure 11.3.1 as the “system” with the external loads being the force resultants { pm }. Then, the external virtual work on element m is Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS δWme = {δrm }T { pm } = δu n δu n+1  pm,n pm,n+1  (11.3.3) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 422 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Let us assume that the distributed displacement for element m can be written in terms of the nodal displacements{rm }, that is, we can define the shape functions [n] so that u m (s) = [n]{rm } (11.3.4) Given the shape functions the strain displacement and Hooke’s law equations can be written in terms of the nodal displacements. εsm = d du m = [n] {rm } ds ds σsm = Eεsm = E d du m =E [n] {rm } ds ds (11.3.5) The strain energy is Um = = 1 2 εsm σsm d V = V 0 lm 1 = 2 0 EA 2 =  lm 1 2 1 2 V Eεs2m d V  du ds 2 du EA ds 2 lm d [n]T ds E d A ds A {rm }T 0 lm 1 ds = 2 0 d EA [n] {rm } ds (11.3.6) 2 ds d [n] {rm } ds ds Note that to square a matrix you must take the transpose of a matrix times the matrix (See Appendix A), that is, 2 du ds d [n] {rm } ds = 2 d [n]T ds = {rm }T d [n] {rm } ds (11.3.7) The distributed virtual displacements are (11.3.8) δu = [n]{δrm } The virtual strain energy for element m is (neglecting higher order terms) δUm = Um (u + δu) − Um (u) = E 2 lm 0 lm = EA 0 d (u + δu) ds 2 Ads − E 2 dδu du ds = {δrm }T E A ds ds lm 0 lm 0 du ds 2 Ads (11.3.9) d d [n]T [n] ds {rm } ds ds Now set the virtual work to zero to establish equilibrium. δWm = δWme − δUm lm d d [n]T [n] ds {rm } ds ds 0 d d [n]T [n] ds {rm } = 0 ds ds = {δrm }T { pm } − {δrm }T E A lm = {δrm }T { pm } − E A 0 (11.3.10) T With {δrm } an arbitrary virtual displacement the only way this can be true is if lm { pm } − E A 0 d d [n]T [n] ds {rm } = 0 ds ds (11.3.11) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 423 Work and Energy Methods—Virtual Work or { pm } = [km ] {rm } (11.3.12) where lm d d [n]T [n] ds ds ds [km ] = E A 0 (11.3.13) We remind you that the arbitrary virtual displacement for an elastic body must be kinematically admissible. This means that the shape functions must be chosen to satisfy any geometric constraints whether internal or at the boundaries. In this case it means that the axial displacement at a node where two elements are joined must be the same for both elements and any realizable displacement constraints at the boundaries must be satisfied. Fortunately the shape functions used in our previous work, namely, Equation 5.2.2, meet these requirements.    s s un (11.3.14) u m (s) = [n] {rm } = 1 − lm lm u n+1 Substituting [n] into Equation 11.3.13 we get lm [km ] = E A 0 lm d d [n]T [n] ds = E m Am ds ds 0 which reduces to lm [km ] = E m Am 0 Thus ⎤ 1  ⎢ lm ⎥ ⎥ −1 ⎢ ⎣ 1 ⎦ lm lm ⎡ − { pm } = [km ] {rm } =  ⎡ d ⎢ ⎣ ds s ⎤  s lm ⎥ d 1− ⎦ s l ds m lm 1−   E m Am 1 1 ds = −1 lm lm pm,n pm,n+1  = E m Am lm  1 −1 −1 1  −1 1 s lm  un u n+1  ds (11.3.15) (11.3.16)  (11.3.17) This is exactly the same result that we obtained in Chapter 5, Equation 5.2.9, and also summarized in Chapter 10. There are differences, however. We used the PVW and not Newton’s laws for equilibrium and the polynomials used for the shape functions were chosen without knowledge of the exact answer (or could be chosen so). For all we know (if we had done it this way first) this is an approximate relation for the distributed displacement and so for the stiffness. Later on you will learn that this method will work to obtain approximate equations in cases where the exact shape function has never been found. Fortunately, it can be shown that you can approach the exact answer as closely as you wish with this method by using smaller and smaller elements even if the shape function itself is not the exact shape of the element displacement. Having found the element stiffness matrix we can proceed to assemble the global stiffness matrix as before.   { pm } = {F} = (11.3.18) [km ] {rm } = [K ] {r } Of course the element matrices must be imbedded in the global format before summing. As in Chapter 5 concentrated loads at nodes are added directly. To these we add distributed loads converted to equivalent nodal loads. In Chapter 5 (and in Chapter 10) we presented the formula for --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 424 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods converting distributed loads into equivalent concentrated loads without proof. Now let us derive the formula we introduced there (Equation 5.4.4). The distributed external load f s (s) contributes to the external virtual work term for element m as follows lm ln (δu m )T f s (s)ds = {δrm }T δWme = 0 [n]T f s (s)ds 0 lm = {δrm }T ⎡ s ⎤ lm ⎥ T s ⎦ f s (s)ds = {δrm } {Fem } ⎢ ⎣ 0 (11.3.19) 1− lm Thus the equivalent nodal loads contributed by a distributed load on element m are found from lm lm [n]T f s (s)ds = {Fem } = 0 0 ⎡ ⎢ ⎢ ⎣ s ⎤ lm ⎥ ⎥ f (s)ds s ⎦ s 1− (11.3.20) lm This is our justification for finding the equivalent nodal loads using Equation 5.4.4. For example, a uniform load f0 would contribute {Fem } =  Fem,n Fem,n+1  lm = ⎡ ⎡ s ⎤ ⎢ lm ⎥ ⎢ s ⎦ f 0 ds = ⎣ lm 1− ⎢ ⎣ 0 f 0 lm ⎤ 2 ⎥ ⎥ ⎦ f 0 lm 2 (11.3.21) while a triangular load f s (s) = f 0 lms would contribute {Fem } =  Fem,n Fem,n+1  lm = 0 ⎡ ⎢ ⎣ ⎡ s ⎤ f 0 lm ⎢ 6 lm ⎥ s ⎢ s ⎦ f 0 lm ds = ⎣ f 0 lm lm 3 1− ⎤ ⎥ ⎥ ⎦ (11.3.22) We now have derived a formal way for defining equivalent nodal loads that ensures the best possible distribution for satisfying equilibrium. For the global equivalent nodal loads we have {Fe } =  {Fem } (11.3.23) Putting this all together, the global equations would be {F} = {Fc } + {Fe } = [K ] {r } (11.3.24) where the global stiffness [K ] is assembled from the element stiffness matrices as described in Chapter 5. To complete the solution the restraints are added to the nodal displacement matrix and the restraint forces are added the nodal force matrix, the equations are partitioned and solved. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 425 Work and Energy Methods—Virtual Work In summary {F} {Fc } f s (s) → { pm } Um lm {Fem } = [n]T f s (s)ds  {Fe } = 0 σsm u m (s) εsm {Fem } u m (s) = [n (s)] {rm } {F} = {Fc } + {Fe } εsm = d du = [n (s)] {rm } ds ds σsm = Eεsm = E Um = 1 2 εsm σsm d V = 1 2 V 2 Eεsm dV = lm EA 2 0 V = EA 2 lm {rm }T 0 lm δWm = δWme − δUm = {δrm }T { pm } − E A lm { pm } = E A --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`---  s [n] = 1 − lm { pm } =  pm,n pm,n+1  E m Am = Lm [K ] =   0 s lm 1 −1 {rm } ← ρ 0 du ds d [n (s)] {rm } ds (11.3.25) 2 ds d [n]T ds d [n] {rm } ds ds d d [n]T [n] ds {rm } = 0 ds ds d d [n]T [n] ds {rm } = [km ] {rm } ds ds  −1 1  un u n+1  = [km ] {rm } [km ] {F} = {Fc } + {Fe } = [K ] {r } In the first row of Equation 11.3.25 we have the typical unknowns in the formulation of a problem as applied to a single element m. {rm } - nodal displacements u m (s) - distributed displacement εsm - normal strain σsm - normal stress Um - strain Energy { pm } - internal forces Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 426 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods and the known quantities f s (s) - distributed force {Fc } - concentrated forces {F} - total applied forces ρ - restraints In the final global set of equations the nodal restraints ρ are added to the nodal displacement matrix{r }, the corresponding nodal restraint forces are added to the global applied load matrix {F}, the resulting equations are partitioned and solved. So far we have found what we already knew; however, in the very next chapter we shall put the principle of virtual work to good use in deriving FEM equations for 2D and 3D structures for which the methods we used in Chapters 5, 6, and 8 are inadequate. 11.3.2 Torsional Loading Just as in Chapter 6, torsional equations can be derived by direct analogy to the axial case. As a reminder the nodal rotational displacements and internal nodal torques are shown in Figure 11.3.2 for element m. element m x node n node n+1 s ψm,n, βn ψm,n+1, βn+1 Figure 11.3.2 In matrix form the nodal displacements and torques are     ψn βn {ψm } = {φm } = ψn+1 βn+1 (11.3.26) The rotational displacement βm (s) is given in terms of the nodal displacements {φm } and shape functions [n]. βm (s) = [n (s)] {φm } (11.3.27) In this case the shape functions must satisfy compatible rotations where two elements are joined and any realizable displacement rotational constraints at the boundaries to be kinematically admissible. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 427 Work and Energy Methods—Virtual Work We shall just summarize the equations here. {ψm } → Um lm {Mem } = [n]T ts (s)ds {Me } = 0 {M} = {Mc } + {Me } Um = 1 2 V γ τdV = GJ = 2 lm 0 1 2 τm  γm βm (s) {Mem } {φm } ← ρ βm (s) = [n (s)] {φm } d dβ r= [n (s)] {φm } r ds ds d τm = Gγm = G [n (s)] {φm } r ds   1 lm dβ 2 2 2 G r d A ds Gγ d V = 2 0 ds A γm = V dβ ds 2 ds = lm GJ 2 {φm }T 0 lm δWm = δWme − δUm = {δφm } {ψm } − G J 0 d [n]T ds d [n] {φm } ds ds (11.3.28) d d [n]T [n] ds {φm } = 0 ds ds lm d d [n]T [n] ds {φm } = [km ] {φm } ds ds   s s [n] = 1 − lm lm      G m Jm 1 −1 βn ψm,n {ψm } = = [km ] {φm } = βn+1 ψm,n+1 −1 1 Lm  [K ] = [km ] {ψm } = G J 0 {M} = {Mc } + {Me } = [K ] {φ} The point of all this is that the PVW provides the same set of FEM equations that we obtained in Chapter 6, Section 6.9, by enforcing equilibrium in an alternative manner. The solution of the equations proceeds in exactly the same way as before. 11.3.3 Beams in Bending The displacements and internal forces on a beam element m are shown in Figure 11.3.3. y element m x node n node n+1 Mm,n+1, θn+1 s Mm,n, θn Vm,n+1, υn+1 Vm,n, υn Figure 11.3.3 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- {M} {Mc } ts (s) P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 428 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In matrix form the nodal displacements and internal forces are ⎡ ⎡ ⎤ ⎤ vn Vm,n ⎢ θn ⎥ ⎢ Mm,n ⎥ ⎥ ⎥ { pm } = ⎢ {rm } = ⎢ ⎣ vn+1 ⎦ ⎣ Vm,n+1 ⎦ θn+1 Mm,n+1 (11.3.29) The strain energy for beams in bending consists of a part based on normal stresses and a part based on shear stresses. 1 2 σx εx d V + 1 2 (11.3.30) γx y τx y d V It was noted, however, that the strain energy contribution of the shear stress for a slender bar is quite small and so can be neglected. In Example 7.5.2 the rationale for neglecting it is demonstrated. So we shall proceed using only the normal stress part. Remember that we did not include shear deformation in the previous development of the FEM equations in Chapter 8. The distributed displacements for an element m are now presented in terms of the nodal displacements and a shape function matrix. vm (s) = [n] {rm } u m (s, y) = −y d dv = −y [n] {rm } ds ds (11.3.31) The strain displacement and Hooke’s law equations are εsm (s, y) = d2 du = −y 2 [n] {rm } ds ds σxm (s, y) = Eεsm = −y E d2 [n] {rm } ds 2 (11.3.32) The strain energy of bending for a beam element m in terms of displacement is found to be Um = 1 2 1 = 2 = (εsm σsm )d V = V lm 0 E Im 2  E d 2v ds 2 lm {rm }T 0 1 2 E(εsm )2 d V = V  2 2 y d A ds = A E Im 2 1 2 E −y V lm 0 d 2v ds 2 d 2v ds 2 2 dV 2 (11.3.33) ds d2 [n] {rm } ds ds 2 d2 [n]T ds 2 The virtual strain energy of bending for a beam element with a uniform cross section is then (again neglecting higher order terms) δUm = Um (v + δv) − Um (v) = E Im 2 lm 0 d 2 (v + δv) ds 2 lm 2 {δrm }T = E Im 0 2 ds − d [n]T ds 2 lm E Im 2 0 d 2v ds 2 2 lm ds = E Im 0 d 2 δv d 2 v ds ds 2 ds 2 (11.3.34) 2 d [n] {rm } ds ds 2 The nodal matrices can be taken outside the integral. lm δUm = {δrm }T E Im 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS d2 d2 [n]T 2 [n] ds {rm } 2 ds ds (11.3.35) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- UB = P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 429 Work and Energy Methods—Virtual Work For element m we apply the principle of virtual work. lm δWm = δWme − δUm = {δrm }T { pm } − {δrm }T E Im 0 d2 d2 [n]T 2 [n] ds {rm } = 0 2 ds ds (11.3.36) Since {δrm }T is an arbitrary virtual displacement it follows that lm { pm } = E Im 0 2 d2 T d [n] [n] ds {rm } = [km ] {rm } ds 2 ds 2 (11.3.37) In this case to be kinematically admissible the shape functions must satisfy both displacement and slope compatibility at all the nodes. For an element m, by the same arguments used in introducing Equation 8.2.6, we select the following shape functions. [n] = 1 3 l − 3lm s 2 + 2s 3 lm3 m lm3 s − 2lm2 s 2 + lm s 3 3lm s 2 − 2s 3 −lm2 s 2 + lm s 3 (11.3.38) These are cubic polynomials that represent the true shape of a uniform beam segment between concentrated loads. The second derivative is 1 d2 [n] = 3 −6lm + 12s ds 2 lm −4lm2 + 6lm s −2lm2 + 6lm s 6lm − 12s (11.3.39) Upon substituting Equation 11.3.39 into Equation 11.3.37 and simplifying the result we get ⎡ Vm,n ⎢ M ⎢ m,n { pm } = [km ] {rm } = ⎢ ⎣ Vm,n+1 Mm,n+1 ⎤ ⎥ E m Im ⎥ ⎥= 3 ⎦ lm ⎡ 12 6lm −12 ⎢ 6l 4lm2 ⎢ m ⎢ ⎣−12 −6lm 6lm −6lm 2lm2 12 −6l 6lm ⎤⎡ vn ⎤ ⎢ ⎥ 2lm2 ⎥ ⎥ ⎢ θn ⎥ ⎥ ⎥⎢ −6lm ⎦ ⎣ vn+1 ⎦ 4lm2 (11.3.40) θn+1 Thus the element stiffness matrix is [km ] = E m Im lm3 ⎡ 12 6lm --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎢ 6l 4lm2 ⎢ m ⎢ ⎣−12 −6lm 2lm2 6lm −12 −6lm 12 −6lm 6lm ⎤ 2lm2 ⎥ ⎥ ⎥ −6lm ⎦ (11.3.41) 4lm2 This is exactly what we got in Chapter 8, Equations 8.2.19. Furthermore, from the external virtual work, we obtain Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS lm [n]T f y (s)ds {Fe } = (11.3.42) 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 430 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods In summary f y (s) → { pm } Um σsm (s, y) lm {Fem } = [n]T f y (s)ds {Fe } = 0  εsm (s, y) {Fem } {F} = {Fc } + {Fenl } 1 2 vm (s) {rm } ← ρ vm (s) = [n (s)] {rm } d dv = −y [n (s)] {rm } ds 2 ds d du εsm (s, y) = = −y 2 [n (s)] {rm } ds ds d2 (11.3.43) = −y E 2 [n (s)] {rm } ds u m (s, y) = −y σsm (s, y) = Eεsm Um = u m (s, y) (εsm σsm )d V = V 1 2 E(εsm )2 d V = V = E Im 2 lm 0 d 2v ds 2 E 2 −y 2 ds = E Im 2 lm {rm }T 0 d2 [n]T ds 2 d2 [n] {rm } ds ds 2 d2 d2 [n]T 2 [n]ds {rm } = 0 2 ds ds 2 d2 T d [n] [n]ds {rm } = [km ] {rm } ds 2 ds 2 1 3 l − 3lm s 2 + 2s 3 lm3 m ⎡ ⎤ Vm,n ⎢ Mm,n ⎥ E m Im ⎥ { pm } = ⎢ ⎣ Vm,n+1 ⎦ = l 3 m Mm,n+1 lm3 s − 2lm2 s 2 + lm s 3 [n] = [K ] = dV V δWm = δWme − δUm = {δrm } { pm } = E Im { pm } = E Im 2 d 2v ds 2 ⎡ 12 6lm ⎢ 6l 4lm2 ⎢ m ⎢ ⎣−12 −6lm  6lm 2lm2 −12 −6lm 12 −6lm [km ] {rm } 3lm s 2 − 2s 3 6lm −lm2 s 2 + lm s 3 ⎤⎡ ⎤ vn ⎥⎢ ⎥ ⎢ θn ⎥ ⎥ = [km ] {rm } ⎥ −6lm ⎦ ⎣ vn+1 ⎦ θn+1 4l 2 2lm2 m {F} = {Fc } + {Fenl } = [K ] {r } The solution proceeds in exactly the same way as in Chapter 8. 11.3.4 Combined Axial, Torsional, and Bending Behavior The principal of virtual work can be applied to bending in the xz plane, to beams with I yz = 0, and to combined axial, torsional, and bending behavior. The results are exactly the same as described in Chapter 10. 11.4 Static Analysis of 3D and 2D Solids by Virtual Work Let us extend virtual work to the case for a general state of stress, that is, for 3D solid bodies without restrictions on geometry and to 2D thin plates within plane loads. We need the strain displacement equations and the Hooke’s law equations for the 3D and 2D cases. These are taken from Chapter 3 and repeated here for convenience. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- {F} {Fc } P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 431 Work and Energy Methods—Virtual Work The 3D strain displacement equations are ⎡ ∂ ⎢ ∂x ⎢ ⎢ ⎢ ⎡ ⎤ ⎢ 0 ⎢ εx ⎢ ⎢ εy ⎥ ⎢ ⎢ ⎥ ⎢ 0 ⎢ εz ⎥ ⎢ ⎥ ⎢ {ε} = ⎢ ⎢ γx y ⎥ = ⎢ ∂ ⎢ ⎥ ⎢ ⎣ γ yz ⎦ ⎢ ⎢ ∂y ⎢ γzx ⎢ ⎢ 0 ⎢ ⎢ ⎣ ∂ 0 0 ∂ ∂y 0 0 ∂ ∂z ∂ ∂x ∂ ∂z 0 ∂ ∂y ∂ ∂x 0 ∂z ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥⎡ ⎤ ⎥ u ⎥ ⎥ ⎣ v ⎦ = [D] {u} ⎥ ⎥ w ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (11.4.1) and in 2D plane stress are ⎤ 0 ⎥ ⎥  ∂ ⎥ ⎥ u = [D] {u} ⎥ ∂y ⎥ v ⎥ ∂ ⎦ ⎡ ∂ ∂ ⎡ ⎤ ⎢ ⎢ x εx ⎢ ⎢ {ε} = ⎣ ε y ⎦ = ⎢ 0 ⎢ γx y ⎢ ⎣ ∂ ∂y Hooke’s law in 3D is ⎡ where ⎤ ⎡ σx ⎢ ⎢ σy ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ σz ⎥ ⎢ ⎢ ⎥=⎢ {σ } = ⎢ ⎥ ⎢ ⎢ τx y ⎥ ⎢ ⎣ τ yz ⎦ ⎢ ⎢ ⎣ τzx λ + 2G λ ∂x λ 0 0 0 λ λ + 2G λ 0 0 0 λ λ λ + 2G 0 0 0 0 0 0 G 0 0 0 0 0 0 G 0 0 0 0 0 0 G λ= (11.4.2) ⎤ ⎡ ⎤ ⎥ εx ⎥⎢ ⎥ ⎢ εy ⎥ ⎥ ⎥⎢ ⎥ ⎢ εz ⎥ ⎥ = [G] {ε} ⎥⎢ ⎥ ⎢ γx y ⎥ ⎥ ⎥⎣ ⎥ γ yz ⎦ ⎦ γzx νE (1 + ν) (1 − 2ν) (11.4.3) (11.4.4) And, in 2D it is E ⎢ 1 − ν2 σx ⎢ ⎢ ⎣ ⎦ {σ } = σ y = ⎢ ν E ⎣ 1 − ν2 τx y 0 ⎡ ⎤ ⎡ νE 1 − ν2 E 1 − ν2 0 0 ⎤ ⎤ ⎡ ⎥ εx ⎥ ⎥ ⎣ ε y ⎦ = [G] {ε} 0⎥ ⎦ γx y (11.4.5) G For an element m we define nodal displacements {rm } and if we can find some shape functions [n] such that {u m } = [n] {rm } (11.4.6) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 432 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods where {u} is a matrix of the distributed displacement components then {εm } = [D] {u m } = [D] [n] {rm } → {σm } = [G] {εm } = [G] [D] {u m } = [G] [D] [n] {rm } (11.4.7) The strain energy is Um = V {ε}T {σ }d V = V ([D] [n] {rm })T ([G] [D] [n] {rm }) d V (11.4.8) Symbolically the virtual work is the sum of the external and internal virtual work or δWm = δWme + δWmi = δWme − δUm = 0 (11.4.9) The external work is done by forces acting on the surfaces of the body whatever its shape. These surfaces forces are represented symbolically by {Fs } and will be given specific values in example problems in later chapters. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- δWme = {δu}T {Fs } d A (11.4.10) A δUm = {δε}T {σ }d V (11.4.11) V The virtual internal strain energy is Setting the virtual work equal to zero for equilibrium obtains δWm = δWme − δUm = A {δu}T {Fs } d A − V {δε}T {σ }d V = 0 (11.4.12) Following our lead with the slender bar, perhaps it is possible to divide the solid body or the thin plate into elements, define shape functions, and turn this integral equation into a set of linear algebraic equations. We designate the applied forces on the surfaces of a 3D solid or on the edges of a thin plate element m as f sm . They may be forces per unit length in plane stress cases or forces per unit area in 3D solids. The virtual work becomes δWm = {δr }T ([D] [n])T [G] [D] [n] {rm } d V = 0 [δrm ]T [n]T { f sm } d A − (11.4.13) Since the virtual nodal displacements {δrm } and the real nodal displacements {rm } are constants they can be brought outside the integrals to obtain δWm = [δrm ]T ([D] [n])T [G] [D] [n] d V {rm } = 0 [n]T { f sm } d A − (11.4.14) For arbitrary values of the virtual nodal displacements {δr } the quantity in the parenthesis must be zero. [n]T { f sm } d A − ([D] [n])T [G] [D] [n] d V {rm } = 0 (11.4.15) We define for each element { pm } = [n]T { f sm } d A (11.4.16) which we recognize as the equivalent nodal forces. And we define the element stiffness matrix for the element m. [km ] = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ([D] [n])T [G] [D] [n] d V (11.4.17) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 433 Work and Energy Methods—Virtual Work The internal element forces are { pm } = [km ] {rm } (11.4.18) These are assembled into our final set of algebraic equations, where [K ] is the global stiffness matrix.  { pm } = [K ] {r } {F} = (11.4.19) In summary {F} {Fc } {Fem } = {Fs } → { pm } Am [n]T {Fs } d Am {F} = {Fc } + {Fe } {σm } {εm }  {Fe } = {Fem } {u m } {rm } ← ρ {u m } = [n] {rm } {εm } = [D] {u m } = [D] [n] {rm } --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- {σm } = [G] {εm } = [G] [D] [n] {rm } Um = V {εm }T {σm }d V = V ([D] [n] {rm })T ([G] [D] [n] {rm }) d V δWm = δWme − δUm = {δrm } { pm } − ([D] [n])T [G] [D] [n] d V {rm } = 0 ([D] [n])T [G] [D] [n] d V {rm } = [km ] {rm }  {km } [K ] = { pm } = {F} = [K ] {r } 11.5 (11.4.20) The Element Stiffness Matrix for Plane Stress There is no simple polynomial expression for the plane stress shape functions for any but the simplest geometry and loading. We cannot conveniently turn to analytical solutions for suggestions to what the shape functions should be as we did for axial, torsional, and bending cases. The displacement functions are continuous, however, and so approximate polynomial shape functions can be derived that, when used in a large number of small elements, closely approximate the exact solution. The earliest attempt to implement the finite element method for plane stress was to divide the plate into triangular segments, or elements, resulting in three nodes for each element. A triangular element is shown in Figure 11.5.1 with the notations for element displacements and forces. For plane stress there are two displacement components and two force components at each node. Ym,n+2,υm,n+2 y Xm,n+2,um,n+2 Ym,n+1,υm,n+1 Ym,n,υm,n Xm,n+1,um,n+1 x Xm,n,um,n Figure 11.5.1 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 434 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The nodal displacement and internal force matrices are ⎤ ⎤ ⎡ ⎡ un X m,n ⎢ vn ⎥ ⎢ Ym,n ⎥ ⎥ ⎥ ⎢ ⎢ ⎢ u n+1 ⎥ ⎢ X m,n+1 ⎥ ⎥ ⎥ ⎢ ⎢ {rm } = ⎢ ⎥ { pm } = ⎢ Ym,n+1 ⎥ ⎥ ⎢ vn+1 ⎥ ⎢ ⎣ u n+2 ⎦ ⎣ X m,n+2 ⎦ vn+2 Ym,n+2 (11.5.1) In an early development simple linear polynomial displacements were assumed of the form u(x, y) = a1 + a2 x + a3 y v(x, y) = a4 + a5 x + a6 y (11.5.2) Since the strains are first derivatives of the displacements the resulting strains are constants, thus, this element is called the constant strain triangle. Remember that the axial bar element is a constant strain element. A difference is that the shape functions are exact for the axial elements with truly concentrated loads but are only approximate for the plane stress element under nearly all conditions. However, the PVW assures us the best possible result for equilibrium within the limits of the approximate shape function. The development of shape functions from the polynomial form of the displacements is straight forward but a bit of a mess. In fact, the process is described in detail in many books but the actual shape functions usually are not even written down explicitly. Briefly, the process is given in the next few equations. Equation 11.5.2 can be rewritten in this form. ⎡ ⎤ a1 ⎢ ⎥  ⎢ a2 ⎥    ⎥ 1 x y 0 0 0 ⎢ u ⎢ a3 ⎥ = [] {a} {u} = (11.5.3) = ⎥ a 0 0 0 1 x y ⎢ v 4 ⎢ ⎥ ⎣ a5 ⎦ a6 ⎤ ⎡ 1 u1 ⎢ v1 ⎥ ⎢0 ⎢ ⎥ ⎢ ⎢ u 2 ⎥ ⎢1 ⎥ ⎢ {rm } = ⎢ ⎢ v2 ⎥ = ⎢0 ⎢ ⎥ ⎢ ⎣ u 3 ⎦ ⎣1 v3 0 ⎡ x1 0 x2 0 x3 0 y1 0 y2 0 y3 0 0 1 0 1 0 1 0 x1 0 x2 0 x3 ⎤⎡ ⎤ a1 0 ⎢ ⎥ y1 ⎥ ⎥ ⎢ a2 ⎥ ⎢ ⎥ 0⎥ ⎥ ⎢ a3 ⎥ = [h] {a} ⎢ ⎥ y2 ⎥ ⎥ ⎢ a4 ⎥ ⎦ 0 ⎣ a5 ⎦ a6 y3 (11.5.4) Invert [h] to obtain {a} in terms of the nodal displacements. {a} = [h]−1 {rm } (11.5.5) Insert Equation 11.5.5 into Equation 11.5.3. {u} = [] [h]−1 {r } = [n] {rm } (11.5.6) [n] = [] [h]−1 (11.5.7) Thus Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The nodal displacements {rm } can be written in terms of {a} as shown in Equation 11.5.4 P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 435 Work and Energy Methods—Virtual Work The shape functions are of the form    u (x, y) n 11 {u} = = n 21 v (x, y) n 12 n 22 n 13 n 23 n 14 n 24 n 15 n 25 ⎤ un ⎥ ⎢  ⎢ vn ⎥ ⎥ u n 16 ⎢ ⎢ n+1 ⎥ = [n] {rm } ⎥ v n 26 ⎢ n+1 ⎥ ⎢ ⎣ u n+2 ⎦ vn+2 ⎡ (11.5.8) We can find [h]−1 and the shape functions [n] explicitly and write them down; however, they are rarely used in explicit format. What is actually done is the above process is programmed and only the coordinates of the nodes are actually entered and the program then generates the shape functions for each element for use in the next steps. Once you have defined the shape functions you can insert them in the expression for virtual work. The result is the element stiffness matrix [km ] = ([D] [n])T [G] [D] [n] d V (11.5.9) and the formula for equivalent nodal loads, namely. {Fem } = [n]T { f s } d A (11.5.10) Note that strain displacement, Hooke’s law, and equilibrium are all satisfied in the process. Since we are using kinematically admissible shape functions compatibility is always satisfied. The element stiffness matrix is of the form (where we entered just the order of the [km ] matrix) ⎡ ⎤ ⎤ un X m,n ⎢ vn ⎥ ⎢ Ym,n ⎥ ⎢ ⎥ ⎥ ⎢ ⎢ u n+1 ⎥ ⎢ X m,n+1 ⎥ ⎢ ⎥ ⎥ ⎢ { pm } = ⎢ ⎥ = [6x6] ⎢ vn+1 ⎥ = [km ] {rm } ⎢ ⎥ ⎢ Ym,n+1 ⎥ ⎣ u n+2 ⎦ ⎣ X m,n+2 ⎦ vn+2 Ym,n+2 ⎡ (11.5.11) The element matrices are then assembled to find the global equations. The process of assembly requires imbedding the element matrices in the global matrix format and summing. This has quite naturally led to commercial computer based programs that generate the shape functions from input of the nodal coordinates, add material properties and continue on to create the element stiffness matrices, and assemble them for solution. These programs also create the equivalent nodal loads from specific distributed loads on the edges and, of course, create nodal restraints based on distributed restraints. The constant strain triangle was never very successful because the restriction to constant strain was too severe and so it required too many elements to converge to an accurate answer. Eventually additional elements including quadrilateral configurations were derived using higher order polynomials. These will be discussed in the next chapter. The above explanation does give you some idea of how the problem of creating the element stiffness may be approached. We should mention that among the conditions imposed on the element is that two adjacent elements using two nodes in common must be compatible in that they share a common edge, that is, no gaps or overlaps. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 436 11.6 14:37 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The Element Stiffness Matrix for 3D Solids Deriving the element stiffness matrix for 3D solids encounters problems similar to those for plane stress. One of the earliest attempts was for a four noded tetrahedron as shown in Figure 11.6.1. At node n ⎡ ⎤ un {rm }n = ⎣ vn ⎦ (11.6.1) wn y Yn , υn n X n , un Z n , wn x n+1 n+3 z n+2 Figure 11.6.1 There are four nodes for element m with three nodal displacements at each node. In an early development simple linear polynomial displacements were assumed of the form u(x, y) = a1 + a2 x + a3 y + a4 z v(x, y) = a5 + a6 x + a7 y + a8 z w(x, y) = a9 + a10 x + a11 y + a12 z --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Steps similar to those described for plane stress eventually lead to a shape function matrix of the form ⎡ ⎤ u (x, y, z) ⎣ v (x, y, z) ⎦ = [3x12] [12x1] = [n] {rm } (11.6.2) w (x, y, z) The element stiffness matrix is formed from the same virtual work equations as summarized in Equation 11.4.20. We enter the order of the matrices rather than type them in full. [ pm ] = [12x1] = [12x12] [12x1] = [km ] {rm } (11.6.3) The assembly process to produce the global matrices proceeds are described before. This element was never very successful. Eventually additional elements including hexahedral configurations were derived using higher order polynomials. Currently a ten noded tetrahedral element with mid edge nodes is the favorite. These will be discussed in the next chapter. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come 437 Work and Energy Methods—Virtual Work 11.7 Summary and Conclusions The PVW supplies a straightforward way to extend the general method to other structural forms. It also provides the optimum method for converting distributed applied loads into equivalent nodal loads. This is summarized in Equation 11.3.25 for axially loaded slender bars, in Equation 11.3.28 for torsionally loaded shafts, in Equation 11.3.43 for beams in bending, and in Equation 11.4.20 for 3D and 2D solids. In the case of slender bars the shape functions for axial, torsional, and bending can be found from the exact analytical equations when the bar is restricted to concentrated loads; however, this is not possible for many other structural forms of interest as we shall see in the remaining chapters. The extension to 2D and 3D solids is presented in the next chapter. In the case of slender bars with truly concentrated loads the FEM is an exact analytical method, that is, it satisfies the differential equation for slender bars exactly. In succeeding chapters the PVW and FEM will be used to find approximate solutions to problems for which exact analytical solutions are more difficult or impossible to find. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-11 P2: ABC JWST071-Waas July 2, 2011 14:37 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 12 Structural Analysis in Two and Three Dimensions 12.1 Introduction We have studied in great detail the stress, strain, and displacement of slender bars (1D) with axial, torsional, and bending loads. 1. In the axial case the problem is formulated in rectangular Cartesian coordinates with just one component of displacement, u, normal stress, σx , and normal strain, εx . All other displacement, stress, and strain components are zero or are ignored. This is possible because of the assumption, supported by St. Venant’s principle, that the stress is uniform across the cross section and the plane cross section remains plane when displaced. The x axis is the loci of centroids of the cross section area. A simple relation defines strain as a first derivative of displacement and another simple relation exists between the stress and strain represented by Young’s modulus E. Applying equilibrium provides a second order differential equation in terms of the displacement that can be solved by direct integration. From this solution for the displacement we can find the stresses and strains. 2. In the torsional case with respect to cylindrical coordinates there is just one component of displacement, β, shearing stress, τ , and shearing strain, γ . All other displacement, stress, and strain components are zero or are ignored. This is possible because of the assumption, supported by St. Venant’s principle, that for solid circular cross sections the stress varies linearly with the radius of the cross section and the plane cross section rotates as a plane when displaced. The x axis is the loci of centroids of the cross section area. The shearing strain is defined in terms of the first derivative of the displacement. A simple relation exists between the stress and strain as represented by the shear modulus G. Applying equilibrium provides a second order differential equation in terms of the displacement that can be solved by direct integration. From this solution for the displacement we can find the stresses and strains. A similar formulation is provided for thin walled open and closed sections. 3. In bending in the xy plane, with I yz = 0 and with respect to rectangular Cartesian coordinates, we have two displacement components, u and v, two stress components, σx and τx y , and two strain components, εx and γx y . The x axis is the loci of centroids of the cross section area and the yz axes are principal axes of the cross sectional area. A simplified relation exists between the u and v displacements and the normal and shear components based on the assumption, supported by St. Venant’s principle, that Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 440 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods plane sections remain plane. This formulation considers bending displacement but neglects shear displacement. Applying equilibrium provides a fourth order differential equation in terms of the displacement that can be solved by direct integration. From this solution for the displacement we can find the stresses and strains. Similar equations can be found for bending in xz plane. Since the governing displacement is a function of a single independent variable in each case, that is, u(x) for axial, β(x) for torsional, and v(x) for bending in the xy plane, and w(x) for bending in the xz plane, these are often called one dimensional structures. These are very effective structures as far as they go, but load carrying bodies are not such simple shapes for the most part. At one time many structures were designed as slender bars or as assemblies of slender bars, such as trusses and frames, because they could be analyzed easily while more complicated shapes were avoided because they were difficult to analyze. As more powerful methods of analysis have been developed the geometry of structures has become more complex. This has resulted in lower costs for manufacture and assembly, reduced material costs, lighter weight and greater efficiency, and other benefits. We shall now consider more general cases of stress and strain in two and three dimensions for which simplifying assumptions are much harder to find. Actually we have considered some cases for simple geometry and loading. We introduced a simple three dimensional problem in Chapter 2, Example 2.6.2, a cube of material under hydrostatic pressure. In Chapter 3, Example 3.5.2, we examined a thin plate with edge loads in the plane of the plate, and in Chapter 4, Section 4.2, we examined a slender bar loaded with uniform and varying end loads. In all these cases solutions were based on the three dimensional equations of elasticity. Also in Chapter 3, Example 3.5.2, we examined a particular two dimensional state of stress that satisfied equilibrium but concluded that it could not be a solution. In Chapter 6, Section 6.4, we found that the linearly radial distribution of torsional stress in a circular cross section did, indeed, satisfy the three dimensional equations of elasticity under certain circumstances. Finally in Chapter 7, Section 7.7, we examined the circumstances under which the stresses in beams did satisfy the three dimensional equations. The results of these previous examples suggest that some exact solutions are available for certain simple geometrical shapes if the surface loading is simple enough, such as, uniform or linearly varying, and the boundary restraints are idealized. The attempt at solutions becomes more difficult with minor changes in geometry, loading, and restraint. We need to extend our methods to handle these more difficult cases. A more complete study of such problems is the subject of the theory of elasticity in more advanced texts and is much beyond the scope of this effort; however, we shall introduce the finite element method as applied to two and three dimensional solid bodies. This is the principal way such problems are solved today. 12.2 The Governing Equations in Two Dimensions—Plane Stress Consider a thin flat plate of arbitrary shape and uniform thickness h restrained to prohibit rigid body motion and loaded on unrestrained edges in the plane of the plate, for example, as shown in Figure 12.2.1. The plate is made of homogeneous and isotropic material. A small element in the interior of the body of dimensions dx by dy is shown in Figure 12.2.2 under a state of stress. The equilibrium of this element in the x-y plane was considered previously in Chapter 2, Section 2.6. We shall assume that the loads and restraints are such that --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS σz = τzx = τ yz = 0, (12.2.1) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 441 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Structural Analysis in Two and Three Dimensions dx dy y x Figure 12.2.1 σy + ∂σ y ∂y dy τ xy + ∂τ xy ∂x dx fby y fbx σx σx + ∂σ x dx ∂x τ xy σy x Figure 12.2.2 which corresponds to state of plane stress. Equilibrium was presented in matrix form in Chapter 2, Equation 2.6.16 and is repeated here as Equation 12.2.2. ⎡ ⎤ ∂ ∂ ⎡ ⎤  0 ⎢ ∂x ⎥ σx ∂ y ⎥ ⎣ σ y ⎦ = − f bx = − { f b } (12.2.2) [E] {σ } = ⎢ ⎣ ∂ ∂ ⎦ f by τx y 0 ∂y ∂x The strain displacement equations were presented in Chapter 3, Equation 3.2.11, and are repeated here as Equation 12.2.3. ⎤ ⎡ ∂ 0 ⎡ ⎤ ⎢ ∂x ⎥ εx ⎢ ∂ ⎥ ⎥ u ⎢ 0 ⎣ ⎦ {ε} = ε y = ⎢ (12.2.3) = [D] {u} ∂y ⎥ ⎥ v ⎢ γx y ⎣ ∂ ∂ ⎦ ∂y Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ∂x Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 442 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods εz = − ν σx + σ y E (12.2.4) The result of the εz component of strain is a change in the thickness of the plate depending upon the local value of the stresses; however, the change has little effect on the behavior of the plate if the plate is sufficiently thin compared to its other two dimensions and so it is generally neglected. In Chapter 3, Section 3.3, it was noted that the strains, εx , ε y , and γx y are not independent. They satisfy the compatibility equation (Equation 3.3.3), ∂ 2 εy ∂ 2 γx y ∂ 2 εx + = 2 2 ∂y ∂x ∂ x∂ y (12.2.5) From Chapter 3, Equation 3.4.20, the plane stress, strain-stress relations for a linear, homogeneous, and isotropic material are, ⎤ ⎡ 1 −ν 0 ⎥⎡ ⎡ ⎤ ⎢ E ⎤ E ⎥ σx ⎢ εx ⎥ ⎢ −ν 1 ⎥ ⎢ ⎣ εy ⎦ = (12.2.6) 0 ⎥ ⎣ σy ⎦ ⎢ E E ⎥ τx y ⎢ γx y ⎣ 1 ⎦ 0 0 G It is convenient to invert Equation 12.2.6 to obtain ⎡ E νE ⎤ ⎡ ⎢ 1 − ν2 1 − ν2 σx ⎢ ⎣ σy ⎦ = ⎢ ν E E ⎢ ⎣ 1 − ν2 1 − ν2 τx y 0 0 0 ⎤ ⎡ ⎤ ⎥ εx ⎥ ⎥ ⎣ ε y ⎦ = [G] {ε} 0⎥ ⎦ γx y (12.2.7) G In summary the set of equations to be satisfied are {F} → { fb } {σ } {ε} {σ } = [G] {ε} [E] {σ } = − { f b } [E] [G] [D] {u} = − { f b } {ε} = [D] {u} {u} ← {ρ} (12.2.8) In the first row we have the typical internal unknowns in the formulation of a problem. {u} − distributed displacements {ε} − strains {σ } − stresses and the known external quantities { f b } − body forces {F} − surface forces {ρ} − restraints The arrows are there to indicate that both {F} and {ρ} are external boundary conditions to be satisfied in the process of solving the equations. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- As shown in Chapter 3, Equations 3.4.20–3.4.23, it is true that there is a non zero component of strain in the z direction given by P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 443 Structural Analysis in Two and Three Dimensions The equations connecting these quantities are in the second, third, and fourth rows. {ε} = [D] {u} − strain displacement {σ } = [G] {ε} − Hooke’s law [E] {σ } = − { f b } − static equilibrium In the fifth row we have combined these equations to form a set of equation in terms of the displacements. [E] [G] [D] {u} = − { f b } (12.2.9) There are many books written on the subject of finding analytical solutions to Equations 12.2.8 for various combinations of geometry, applied loads, and restraints. We can find some solutions to the equations for very simple geometry, loading, and restraint. The number of exact analytical solutions is very limited. Most often stresses are found that satisfy equilibrium, compatibility, and boundary conditions then the displacements are calculated. The displacement based set of equations, Equations 12.2.9, is seldom used directly; however, one such solution is found in the next example. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ########### Example 12.2.1 Problem: A thin flat rectangular plate has uniform loads on edges as shown in Figure (a). The plate has a thickness h, has a length in the x direction of a and a width in the y direction of b. The origin of the coordinates is in the center of the plate. The edge loads are constant and have the units of N/mm2 . Find the stresses, strains, and displacements of the plate. y fy fx fx x fy Figure (a) Solution: Uniform loads are likely to provide a uniform interior stress which would suggest a linear displacement. So let us assume u = cx v = dy where c and d are constants to be determined if, in fact, this is a valid assumption. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 444 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Substituting into the strain displacement relations we get ⎡ ∂ ⎤ ⎡ ∂ 0 ⎤ ⎢ ∂x ⎡ ⎢ ∂x ⎥ εx ⎢ ⎢ ∂ ⎥ ⎢ 0 ⎥ u 0 ⎣ εy ⎦ = ⎢ = ⎢ ⎥ ⎢ ∂y ⎥ v ⎢ ⎢ γx y ⎣ ∂ ⎣ ∂ ∂ ⎦ ∂y ∂x 0 ∂ ∂y ∂ ∂x ∂y ⎤ ⎡ ⎤ ⎥ c ⎥ ⎥ cx = ⎣d ⎦ ⎥ ⎥ dy 0 ⎦ (b) These values of constant strain can then be entered in the Hooke’s law equations, Equations 12.2.7. ⎤ ⎤ ⎡ ⎡ E vE vE E 0 ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ 0 ⎥ c ⎥ εx ⎢ 1 − v2 1 − v2 ⎢ 1 − v2 1 − v2 σx ⎥ ⎥ ⎢ ⎢ ⎥⎣d ⎦ ⎥ ⎣ ε y ⎦ = ⎢ vE ⎣ σ y ⎦ = ⎢ vE E E ⎥ ⎥ ⎢ ⎢ ⎣ 1 − v2 1 − v2 0 ⎦ 0 ⎣ 1 − v2 1 − v2 0 ⎦ γx y τx y (c) 0 0 G 0 0 G ⎡ ⎤ c + vd E ⎣ vc + d ⎦ = 1 − v2 0 One can quickly note that constant values of stress satisfy the equilibrium equations, Equation 12.2.2, identically if the body forces are zero. ⎤ ⎤ ⎡ ⎡ ∂ ∂ ∂ ∂ ⎡ ⎡ ⎤ ⎤ ⎡ ⎤ 0 0 σ c + vd 0 x ⎥ ⎥ E ⎢ ∂x ⎢ ∂x ∂ y ∂ y ⎥ ⎣ σy ⎦ = ⎢ ⎥ ⎣ vc + d ⎦ = ⎣ 0 ⎦ (d) [E] {σ } = ⎢ ⎦ ⎦ ⎣ ⎣ 2 ∂ ∂ ∂ ∂ 1−v τx y 0 0 0 0 ∂y ∂x ∂y ∂x What remains is to satisfy the boundary conditions. E a (c + vd) = f x = σx ± 2 1 − v2   E b (vc + d) = f y = σy ± 2 1 − v2 (e) Solving for c and d c= 1 E fx − v f y d= 1 E f y − v fx (f) The displacement is therefore u (x, y) = 1 E fx − v f y x v (x, y) = 1 E f y − v fx y (g) Now we learned earlier (Chapter 3, Section 3.3) that satisfying equilibrium is not enough. Let us check to see if compatibility (Equation 3.3.3) is satisfied. ∂ 2 εy ∂ 2 γx y ∂ 2 εx + = ∂ y2 ∂x2 ∂ x∂ y → ∂ 2c ∂ 2d ∂ 20 + = ∂ y2 ∂x2 ∂ x∂ y → 0+0=0 (h) So the assumed form is a valid solution. It satisfies strain displacement, Hooke’s law, equilibrium, and all boundary conditions. Note that by starting with displacements compatibility is satisfied. ########### --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 445 Structural Analysis in Two and Three Dimensions Solutions are also possible for linearly varying edge loads; however, as soon as we restrain an edge, have more complicated edge loads, or change the geometry of the plate the search for an analytic solution becomes very difficult. Known useful results are limited. What we really need is a method for all combinations of geometry, loading, and restraint. That method is based on virtual work as was first introduced in Chapter 11. It is called the finite element method. It will be extended to plane stress problems in the next section. 12.3 Finite Elements and the Stiffness Matrix for Plane Stress The equations for the finite element method are developped in Chapter 11, Section 4, and are summarized here in Equations 12.3.1. These are repeated here for convenience. { pm } {Fs }  {Fem } = [n]T {Fs } dAm {F} Am {σm } {εm }  {Fem } {F} = {Fe } = {u m } {rm } ← ρ {u m } = [n] {rm } {εm } = [D] {u m } = [D] [n] {rm } {σm } = [G] {εm } = [G] [D] [n] {rm }  {εm }T {σm }dV =  ([D] [n] {rm })T ([G] [D] [n] {rm }) dV   δWm = δWme − δUm = {δrm } { pm } − ([D] [n])T [G] [D] [n] dV {rm } = 0  { pm } = ([D] [n])T [G] [D] [n] dV {rm } = [km ] {rm }  [K ] = {km } Um = V (12.3.1) V {F} = [K ] {r } There are no exact polynomial expressions for the plane stress shape functions for any but the simplest geometry and loading. We cannot conveniently turn to analytical solutions for suggestions to what the shape functions should be as we did for axial, torsional, and bending cases. The displacement functions are continuous, however, and so approximate polynomial shape functions can be derived that, when used in a large number of elements, closely approximate the exact solution. We have briefly described in Chapter 11, Section 11.5, the quest for shape functions for the constant strain triangle element. Even for this simplest of plane stress elements it is impossible to actually write down the shape functions and the resulting stiffness matrices for each element and assemble them by hand as we have illustrated for slender bars. We must depend on computer programs that do this quickly and neatly upon given our instructions. As noted in Chapter 11 this has quite naturally led to commercial programs that generate the shape functions from input of the nodal coordinates, add material properties, and continue on to create the element stiffness matrices and assemble them for solution. These programs also create the equivalent nodal loads from specific distributed loads on the edges and, of course, create nodal restraints based on distributed restraints. We also noted that the constant strain triangle was never very successful, so great effort was expended to develop more advanced element stiffness matrices using higher order polynomials for the shape functions. The actual process was to take a square element and map it into a general quadrilateral as shown in Figure 12.3.1. Still later two other popular elements were developped – the six noded triangle and the eight noded quadrilateral each with curved edges as illustrated in Figure 12.3.2. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 446 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y x Figure 12.3.1 y Figure 12.3.2 The derivation of these elements is well beyond the scope of this work, but you get some idea of what is involved by the example given in Chapter 11, Section 11.5. Plane stress elements were used in the solutions presented in Chapter 4, Section 4.2, and in Chapter 7, Section 7.7. They were also adapted to use on thin walled shell structures such as the hollow tubes in Chapter 6, Section 6.6. Because of the size and complexity of the element stiffness matrices we do not attempt to write them out for each element. Because a large number of elements are needed we do not assemble the elements, find equivalent nodal loads, partition the matrices, and insert the resulting matrices into a linear equation a linear equation solver. Instead all this has been programmed and offered in a number of commercial FE codes. In addition there are a number of pre and post processors that work directly with surface and solid models to create automatically the nodes and elements and to present the results in the form of color contour plots. We shall use one such set of commercial codes to create solid models, attach boundary restraints and loads to the geometry of the models, to create nodes and elements (called a mesh), to solve for the displacements and stresses, and finally to present the results in color contour plots. One such set of codes was already used in Chapters 4, 6, and 7 to demonstrate the validity of the simplifying assumptions of slender bar theory. Additional examples will be offered of ever increasing complexity. Steps in the process of solution will be presented. We shall now offer a plane stress solution based on FEM for one of the classic problems of structural analysis – the plate with a hole. The analysis starts with a surface created by solid modeling or computer aided design (CAD) software. The applied loads and restraints are applied directly to the CAD model. The nodes and elements are created by a modern automatic mesh generator. The elements used are Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- x P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 447 Structural Analysis in Two and Three Dimensions sophisticated quadrilaterals. The software converts the loads and restraints to equivalent nodal loads and restraints. A powerful linear equation solver completes the solution. Finally, the results are presented in the form of contour plots. ########### Example 12.3.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Problem: A square plate with a hole is loaded with uniform edge loads as shown in Figure (a). Find the stresses and displacements. The plate has a dimension of 100 mm on a side and is 2 mm thick. The hole is centered and has a diameter of 30 mm. The material is aluminum with E = 68950 N/mm2 . A uniform load of 2 N/mm is applied to the two opposite edges as shown. This provides an edge stress of 1 N/mm2 . f0 = 2 N/mm f0 = 2 N/mm Figure (a) Solution: Create a CAD model; add loads to the edges; use the automatic mesh generator to create nodes and elements; invoke the solver; and present contour plots of stresses and displacements. A solid model of the plate was created in the I-DEAS software as shown in Figure (b). Figure (b) The plane stress elements are based on surfaces and the thickness is added as a scalar quantity in a property table. A surface was selected and loads were added to the edges. These are shown in Figure (c). Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 448 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure (c) Nodes and elements were then created by the automatic mesh generator. We have the choice of 3 or 6 noded triangular, or 4 or 8 noded quadrilateral elements. We have chosen the 4 noded quadrilateral for this example. We have the choice of specifying the density of the mesh, that is, the number of nodes and elements, in several ways. One way is to specify the nominal edge length of each element. An edge length of 10 mm was chosen. In anticipation of high stress in the vicinity of the hole we have used a feature of the mesh generator that allows us to increase the number of nodes on the edge of the hole. This provides smaller element edge lengths there. Sixteen equally spaced nodes were placed on the edge of the hole. For visual reasons the element size was kept fairly large for the first solution attempt. In creating the elements a thickness of 2 mm was specified. The nodes and elements are shown in Figure (d). Nodes are at the corners of each element. Figure (d) You will note that while the geometry and loading are symmetrical about both the x and y axes the elements are not, that is, the automatic mesh generator does not maintain perfect symmetry. There are mesh generator functions that will enforce symmetry but they require a little more work. We will address this problem shortly. The statement of the problem is now complete. The solver was invoked. Tables of the values of the displacements and stresses at the nodes are available; however, more useful are contour plots of these quantities throughout the plate. We have the option of displaying contour plots of the three Cartesian components of stress, the maximum and minimum principle stresses, the maximum shear stresses, and the von Mises stresses. In Figure (e) we show a contour plot of the σx stress on an undeformed plate. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come Structural Analysis in Two and Three Dimensions 449 We note the large increase in stress at the top and bottom of the hole. The stress at the boundary is 1 N/mm2 while the maximum stress is 3.54 N/mm2 . This phenomenon is known as stress concentration. Figure (e) In Figure (f) we show a contour plot of the displacements placed on a deformed model. What is plotted is the maximum displacement, that is, the vector sum of the x and y displacements. Figure (f) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 450 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The displacements are greatly exaggerated for visual reasons. The actual displacements are quite small, 0.000373 mm is the maximum. In Figure (g) we show a contour plot of the stresses on a deformed model. This time we show the von Mises stresses. Figure (g) Increased accuracy is obtained by using a finer and finer mesh until convergence is observed. Not long ago the cost of computation was so high that various means were used to keep the number of nodes and elements to a minimum. Computation has become so fast and cheap that for simple problems like this it is not a problem to increase the density substantially. We now repeat the whole process using a much finer mesh density to examine the accuracy of the method. The nodes and elements are shown in Figure (h). There are 4100 nodes and 3960 elements in this FE model. The mesh in the vicinity of the hole was made much finer deliberately. Figure (h) --`,,,,``,``,,``,,`,`,,, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come Structural Analysis in Two and Three Dimensions 451 A contour plot of the von Mises stresses are shown in Figure (i) on a deformed model to compare with those in Figure (g). Note the symmetry of the solution – it must certainly have converged. The maximum von Mises stress has changed from 3.41 to 3.84. Nowhere near this number of elements are needed for an accurate answer. Figure (i) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The displacement contour plot is shown in Figure (j). The maximum displacement has changed from 0.000373 mm to 0.000383 mm. Figure (j) ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 452 12.4 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Thin Flat Plates—Classical Analysis When a thin plate is subject to lateral loads and is restrained at the edges to prevent rigid body motion, bending of the plate occurs. We shall orient the plate in the xy plane as shown in Figure 12.4.1. A rectangular plate of uniform thickness is shown but the plate can be any shape. y b z Figure 12.4.1 The in plane displacements are u and v in the x and y directions respectively. The lateral displacement w is in the z direction. Distributed lateral loads are normal to the undeformed plate surface; are designated fz (x,y); and have units of N/mm2 . There are no in plane loads. With the x and y axes in the plane of the plate and the z axis normal to the plate we assume the following simplification to the in-plane displacements, due to bending, u(x, y, z) = −z ∂w (x, y) ∂x v(x, y, z) = −z ∂w (x, y) ∂y (12.4.1) Note that w (x, y) is a function of only x and y while u(x, y, z) and v(x, y, z) are functions of x, y, and z. This is equivalent to the “plane sections remain plane” assumption of beam theory and is verified in practice. In fact, this can be thought of as the 2D extension of this assumption. The strain components corresponding to (12.4.1) become, εx (x, y, z) = −z ∂ 2w ∂x2 ε y (x, y, z) = −z ∂ 2w ∂ y2 γx y (x, y, z) = −2z ∂ 2w ∂ x∂ y (12.4.2) The strain components in matrix form are ∂2 ⎡ ⎤ ⎢ ∂x2 ⎢ εx ⎢ ∂2 {ε} = ⎣ ε y ⎦ = ⎢ ⎢ −z ∂ y 2 ⎢ γx y ⎣ ∂2 −2z ∂ x∂ y ⎡ −z ⎤ ⎥ ⎥ ⎥ ⎥ {w} = [D] {w} ⎥ ⎥ ⎦ where {w} has only one row and one column. The stress-strain equations in matrix form are given in Equation 12.2.7, as ⎤ ⎡ νE E ⎡ ⎡ ⎤ ⎤ 0 2 2 ⎥ εx ⎢ 1−ν σx 1−ν ⎥ ⎢ ⎥ ⎣ ε y ⎦ = [G] {ε} ⎣ σy ⎦ = ⎢ ν E E ⎢ 0⎥ 2 2 ⎦ γx y ⎣ τx y 1−ν 1−ν 0 0 G Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (12.4.3) (12.4.4) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- x a P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 453 Structural Analysis in Two and Three Dimensions The stresses in terms of the displacements are, σx (x, y, z) = E E (εx + vε y ) = −z 1 − v2 1 − v2  σ y (x, y, z) = E E (ε y + vεx ) = −z 1 − v2 1 − v2  τx y (x, y, z) = Gγx y = −z E ∂ 2w 1 + v ∂ x∂ y ∂ 2w ∂ 2w +ν 2 2 ∂x ∂y ∂ 2w ∂ 2w +ν 2 ∂ y2 ∂x   (12.4.5) When dealing with equilibrium it is convenient to define stress resultants in terms of moments per unit length. With reference to Figure 12.4.2, and in the sense as indicated there (using the double arrow convention and the right hand cork-screw rule), Mx (x, y) = −  h 2 − h2 σx zdz = h 2 E 1 − v2 h 2  − h2 h 2  ∂ 2w ∂ 2w +ν 2 2 ∂x ∂y  z 2 dz = Eh 3 1 − v2  ∂ 2w ∂ 2w +ν 2 2 ∂x ∂y    2  ∂ 2w 2 ∂ 2w Eh 3 ∂ w ∂ 2w z + ν + ν dz = ∂ y2 ∂x2 1 − v2 ∂ y 2 ∂x2 − h2 − h2      h  h 2 2 Eh 3 (1 − v) ∂ 2 w E ∂ 2w z 2 dz = Mx y (x, y) = − τx y zdz = 1 + v − h2 ∂ x∂ y 1 − v2 ∂ x∂ y − h2 (12.4.6) and the transverse shear forces per unit length are, M y (x, y) = −  σ y zdz = E 1 − v2 Q x (x, y) =    h 2 − h2 τzx dz Q y (x, y) =  h 2 − h2 τ yz dz (12.4.7) The forces and moments on a small element of the plate are shown in Figure 12.4.2. My + ∂M y ∂y dy Qy + ∂Qy ∂y dy M xy + ∂M xy ∂y dy Qx Mx + M xy z ∂M x dx ∂x p ( x, y ) y M xy + x M xy Qy My Qx + ∂M xy ∂x dx ∂Qx dx ∂x Figure 12.4.2 From summation of forces in the z direction we get Fz = ∂ Qy ∂ Qx + + p (x, y) = 0 ∂x ∂y (12.4.8) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 454 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods From the summation of moments about the x and y axes we get  Mx = ∂ My ∂ Mx y + + Qy = 0 ∂x ∂y  My = ∂ Mx ∂ Mx y + + Qx = 0 ∂y ∂x (12.4.9) To simplify the notation let Dh = Eh 3 12(1 − ν 2 ) (12.4.10) where, h is the uniform plate thickness. Now find the forces in terms of the displacements by combining Equations 12.4.6 and 12.4.9.   2   ∂ Mx y ∂ Mx ∂ 2w ∂ ∂ ∂ 2w ∂ w (1 − = −Dh − D Qx = − + ν − ν) h ∂x ∂y ∂x ∂x2 ∂ y2 ∂ y ∂ x∂ y   ∂ 2w ∂ ∂ 2w + = −Dh ∂x ∂x2 ∂ y2 (12.4.11)   2   ∂ Mx y ∂ ∂ 2w ∂ My ∂ ∂ 2w ∂ w − = −Dh + ν 2 − Dh (1 − ν) Qy = − ∂y ∂x ∂ y ∂ y2 ∂x ∂ x ∂ x∂ y  2  2 ∂ w ∂ ∂ w + = −Dh ∂ y ∂ y2 ∂x2 Finally Equation 12.4.8 in terms of displacement becomes    4  4 ∂ Qy ∂ 4w ∂ 4w ∂ Qx ∂ w ∂ w + + p (x, y) = −Dh − D + p (x, y) = 0 + ν + ν h ∂x ∂y ∂x4 ∂ x 2∂ y2 ∂ y4 ∂ x 2∂ y2 (12.4.12) This may be written in the standard form ∇4w = ∂ 4w ∂ 4w ∂ 4w p +2 2 2 + = 4 ∂x ∂x ∂y ∂ y4 Dh (12.4.13) where the operator ∇ 4 is ∇4 = ∂4 ∂4 ∂4 +2 2 2 + 4 4 ∂x ∂x ∂y ∂y (12.4.14) This partial differential equation has been the subject of much study in mathematics as well as in applied engineering. There are whole books written on the subject of exact and approximate analytical solutions. Several are presented in the classic text by Timoshenko and Woinowsky-Krieger (1987). Exact solutions have been found for only a very few special cases of geometry, boundary conditions, and loading. One set of conditions for which there is an exact solution is a rectangular plate with simple support on all four edges and a sinusoidal loading. The boundary conditions for simple support are w(0, y) = 0 w(x, 0) = 0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ∂ 2 w (0, y) =0 ∂2x ∂ 2 w (x, 0) =0 ∂2 y w(a, y) = 0 w(x, b) = 0 ∂ 2 w (a, y) =0 ∂2x ∂ 2 w (x, b) =0 ∂2 y (12.4.15) The sinusoidal loading is Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS p (x, y) = cij sin iπ y iπ x sin a b (12.4.16) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 455 Structural Analysis in Two and Three Dimensions The displacement equation becomes cij iπ y ∂ 4w ∂ 4w iπ x ∂ 4w sin +2 2 2 + = sin 4 4 ∂x ∂x ∂y ∂y Dh a b (12.4.17) A solution is suggested of the form w (x, y) = Cij sin iπ y iπ x sin a b (12.4.18) One can quickly verify by substitution that this satisfies the boundary conditions. Upon substitution the trial solution into Equation (12.4.16) we obtain   4 4 cij i 2 j 2π 4 j 4π 4 i π = + 2 + (12.4.19) C a4 a 2 b2 b4 Dh The trouble is that plates usually are not rectangular, seldom, if ever, are simply supported, and do not have such a convenient form of loading. There are other cases for which approximate analytical solutions have been found; however, the idealizations in geometry, loading, and restraint necessary to obtain analytical solutions have largely given way to the methods used in the next section which do not require these idealizations. The main usefulness of the few exact solutions that have been obtained is to verify the validity of the approximate methods that have been devised. Before we dive into attacking problems using the finite element method, one point regarding boundary conditions for plate analysis is warranted. Suppose, in the previous problem, imagine that three edges are simply supported and the edge x = a is a free edge. Then the boundary conditions on the free edge, in terms of the forces and moments denoted in Figure 12.4.2 would become, Mx (a, y) = Mx y (a, y) = Q x (a, y) = 0. But, when these are expressed in terms of displacements, there is one extra condition on this edge, since for the fourth order symmetric governing equation, one only needs two boundary conditions on each edge. This apparent paradox was solved by Kirchhoff who combined two of the conditions in defining the effective Kirchhoff transverse shear resultant on an edge. Thus, introduce the Kirchhoff shear resultants, Vx , Vy , which are operational on edges with shear resultants, Q x , Q y , respectively, and are defined as, Vx = Q x + ∂ Mx y ; ∂y Vy = Q y + ∂ Mx y ∂x (12.4.20) Now, the boundary conditions on the edge x = a for a free edge are, Vx (a, y) = Q x (a, y) + ∂ Mx y (a, y) = 0 and ∂y Mx (a, y) = 0 (12.4.21) M y (x, b) = 0 (12.4.22) and, if in a problem, the edge, y = b is a free edge, then, Vy (x, b) = Q y (x, b) + ∂ Mx y (x, b) = 0 and ∂x Equations 12.4.20 and 12.4.21 are the Kirchoff plate boundary conditions at a free edge. 12.5 Thin Flat Plates—FEM Analysis The plate shape functions have been defined for the same configurations as for plane stress, namely, three and six noded triangles, and four and eight noded quadrilaterals. We shall illustrate just the four noded quadrilateral. We show the internal forces and displacements at just one of the nodes in Figure 12.5.1. The nodal force is designated Z and the nodal moments Mx and My . --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 456 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods z x Z,w x Mx,φ My,θ Figure 12.5.1 The displacements at one node consist of the lateral displacement w and two rotations θ= ∂w ∂x φ= ∂w ∂y (12.5.1) At node n ⎡ ⎤ wn {rm }n = ⎣ θn ⎦ φn (12.5.2) There are four nodes for element m with three nodal displacements at each node. The shape function matrix is of the form wm (x, y) = [n] {rm } = [1x12] [12x1] The element stiffness matrix is formed from the same virtual work equation as before.  { pm } = [km ] {rm } = ([D] [n])T [G] [D] [n] dV {rm } The matrix [D] is given in Equation 12.4.3 and [G] is given in 12.4.4. The formula for equivalent nodal loads is  {F}enl = [n]T { f s } dA (12.5.3) (12.5.4) (12.5.5) The order of the matrices are shown in Equation 12.5.6 [ pm ] = [12x1] = [12x12] [12x1] = [km ] {rm } (12.5.6) The assembly process to produce the global matrices proceeds are described above. The steps for a solution are the same as those described for Example 12.3.1. To emphasize the fact that ease of solution does not depend upon simplified geometry, loading, and restraints we shall give an example that would be very difficult to solve by any of the exact or approximate analytical methods that have preceded the FEM. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 457 Structural Analysis in Two and Three Dimensions ########### Example 12.5.1 Problem: A trapezoidal shaped plate has a fixed restraint on three edges and is free on the fourth edge as shown in Figure (a). The coordinates of the four corners are given in millimeters in the parentheses. The plate is 10 mm thick. The material is steel with E = 206800 N/mm2 . The load consists of a uniform distributed load of 0.2 N/mm2 in the positive z direction. Find the stresses and displacements. y D(200,800) fixed edge C(900,700) free edge fixed edge B(1000,200) fixed edge A(0,0) x Figure (a) Solution: Create a CAD model; add restraints to the edges; add loads to the surface; use the automatic mesh generator to create nodes and elements; invoke the solver; and present contour plots of stresses and displacements. A solid model of the plate was created in the I-DEAS software. Restraints were added to the edges and loads to the surface. These are shown in Figure (b). Figure (b) Nodes and elements were then created by the automatic mesh generator. The plate elements are based on surfaces and the thickness is added as a scalar quantity in a property table. Once again we have the --`,,,,``,``,,` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 458 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods choice of specifying the density of the mesh, that is, the number of nodes and elements, in several ways. We have specified nominal edge length of 25 mm. For visual reasons the element size was kept fairly large. The nodes and elements are shown in Figure (c). Figure (c) The solver was invoked. In Figure (e) we show a contour plot of the von Mises stresses on a deformed plate. Figure (e) In Figure (f) we show a contour plot of the displacements. The displacements are greatly exaggerated for visual reasons. --`,,,,``,``,,``,,`,`,,,,,` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 459 Structural Analysis in Two and Three Dimensions Figure (f) ########### 12.6 Shell Structures The plate bending and plane stress elements are combined into a single element for dealing with both bending and in plane loads, stresses, and displacements. Such is often the case for curved plates and shell structures. These elements are called shell elements and have five components of displacement and force at each node as shown in Figure 12.6.1. z y Z,w x M,x φ; Y,υ M,y θ; X,u Figure 12.6.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 460 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods This element is shown in the xy plane. Appropriate coordinate transformations are required to place this in position in 3D space. We shall show in an example how such elements can be used to solve a thin walled beam like structure. ########### Example 12.6.1 Problem: A cantilever beam with a D shaped cross section such as the one examined in Example 10.7.2 is one meter long and has the cross section dimensions shown in Figure (a). It is loaded by a distributed load of 1 N/mm along the vertical web at the free end. The wall thickness is 2 mm. The material is steel with E = 206800 N/mm2 . fy(x) = 1 N/mm R = 500 mm Figure (a) Solution: Create a CAD model; add restraints to the edges; add loads to the vertical edge; use the automatic mesh generator to create nodes and elements; invoke the solver; and present contour plots of stresses and displacements. A solid model of the beam was created in the I-DEAS software as shown in Figure (b). The shell elements will be created on the surfaces of the model and the thickness entered as a scalar quantity in an element properties table. Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 461 Structural Analysis in Two and Three Dimensions Restraints and loads are now added to the appropriate edges. These are shown in Figure (c). Y Z X Figure (c) The nodes and elements are created next. A four noded quadrilateral element is chosen. A nominal size for an element is a 10 mm edge length. This is shown in Figure (d). Figure (d) The solver was invoked next. In Figure (e) we show a contour plot of the von Mises stresses. This is shown in what is called the step shaded format. Of particular interest is the shear stress. From our study in Chapter 10 we know that the shear stress is tangent to the thin walls. This software only presents stresses in the coordinate directions in contour plots. You can go to the report data and resolve the stress components given in rectangular Cartesian coordinate into components in cylindrical components if necessary. We shall discuss only general conclusions here. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 462 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure (e) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- In Figure (f) we show the τ yz stress component. This may be compared with the results obtained in Chapter 10, Example 10.7.2. The gray scale contour plots are difficult to interpret. Looking at the color contour plot shows that the shear stress in the vertical web agrees closely to the value obtain in Chapter 10. A more detailed analysis of the shear stresses in the curved web is needed to confirm that the results agree with those in Chapter 10; however, a look at the τ yz stress in the curved web as shown in Figure (g) confirms that the largest stresses are in the vertical web and it also shows the reversal, that is, negative direction of the shear stresses at the top and bottom of the curved web. Figure (f) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 463 Structural Analysis in Two and Three Dimensions Figure (g) ########### The method is well adapted to multi celled structures as shown in the next example. ########### Example 12.6.2 Problem: A two celled thin walled beam with a cross section is first analyzed with out stiffeners and then stiffeners are added in the next section. The beam CAD model is shown in Figure (a). It has the same loads and restraints as Example 12.6.1 as shown in Figure (b). Solution: Create the CAD model, FE model, etc., using exactly the same steps as in Example 12.6.1. Y Z X Figure (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 464 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The loads and restraints are shown in Figure (b). Figure (b) The mesh is shown in Figure (c). Figure (c) A contour plot of von Mises stresses is shown in Figure (d). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come Structural Analysis in Two and Three Dimensions Figure (d) In Figure (e) we show a contour plot of the τ yz stress component. Figure (e) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST 465 P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 466 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods A comparison of the FEM results with the analytical results from thin walled beam theory establishes good agreement with the exception of the usual local effects at the loadings and the restraints. These local effects are ignored in the thin walled beam theory with the blessing of St. Venant. 12.7 Stiffened Shell Structures We shall now apply the FEM to stiffened shell structures. ########### Example 12.7.1 Problem: Solve the cantilever beam with the D shaped cross section and stiffeners examined in Example 10.8.3, that is, repeat Example 12.6.1 with stiffeners added. Let the stiffeners be 10 mm by 10 mm and the wall thickness is 2 mm. Let R = 50 mm Solution: Beam elements are added to the model and the problem is resolved. In Figure (a) we show the FE model with beam elements added. The restraints and loads are the same as in Example 12.6.1 Figure (a) The von Mises stresses are shown in Figure (b). There is a substantial lowering of the stress by adding the stiffeners as might be expected. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- E = 206800 MPa P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 467 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Structural Analysis in Two and Three Dimensions Figure (b) The τ yz shear stresses are shown in Figure (c). Figure (c) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 468 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The results compare with those obtained in Chapter 10 for the vertical web. A comparison with Example 12.6.1 shows a more nearly uniform value of the shear stress in the vertical web. This supports the assumption of a constant shear stress in the web as explained in Chapter 10 to obtain a first approximation. A look at the curved web in Figure (d) shows the much lower shear stresses and the stress reversal obtained in Chapter 10. Figure (d) Example 12.7.2 Problem: Add stiffeners to Example 12.6.2 and solve. Solution: Beam elements are added to the model and the problem is resolved. Now let us add stiffeners and repeat the analysis. The beam with stiffeners is shown in Figure (f). The loading and restraints are the same as in Example 12.7.2. Figure (f) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come Structural Analysis in Two and Three Dimensions 469 A contour plot of the von Mises stresses is shown in Figure (g) Figure (g) A contour plot of the τ yz stress component is shown in Figure (h) Figure (h) Again the von Mises stresses are much lower. The shear stress in the vertical web is more closely uniform and we have the usual local effects where the loading and restraints are applied. Comparison with a thin walled beam theory solution would show good agreement. ########### --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 470 12.8 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Three Dimensional Structures—Classical and FEM Analysis The classical equations were developed in previous chapters. The strain displacement equations are given in Equation 3.2.10 and are repeated here as Equation 12.8.1. ∂ ⎢ ∂x ⎢ ⎢ ⎤ ⎢ 0 ⎡ ⎢ εx ⎢ ⎢ εy ⎥ ⎢ ⎥ ⎢ 0 ⎢ ⎢ εz ⎥ ⎢ ⎥ ⎢ {ε} = ⎢ ⎢ γx y ⎥ = ⎢ ∂ ⎥ ⎢ ⎢ ⎣ γ yz ⎦ ⎢ ∂ y ⎢ ⎢ γzx ⎢ 0 ⎢ ⎢ ⎣ ∂ ∂z ⎡ 0 0 ∂ ∂y 0 0 ∂ ∂z ∂ ∂x ∂ ∂z 0 0 ∂ ∂y ∂ ∂x ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥⎡ ⎤ ⎥ u ⎥ ⎥ ⎣ v ⎦ = [D]{u} ⎥ ⎥ w ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (12.8.1) The stress strain equations are given in Equation 3.4.30 and are repeated here as Equation 12.8.2. ⎡ ⎤ ⎡ σx ⎢ ⎢ σy ⎥ ⎢ ⎢ ⎥ ⎢ ⎢ σz ⎥ ⎢ ⎢ ⎥=⎢ {σ } = ⎢ ⎥ ⎢ ⎢ τx y ⎥ ⎢ ⎣ τ yz ⎦ ⎢ ⎢ ⎣ τzx λ + 2G λ λ 0 0 λ λ + 2G λ 0 0 λ λ λ + 2G 0 0 0 0 0 0 0 0 0 0 0 G 0 0 0 G 0 0 ⎤ ⎡ ⎤ ⎥ εx 0⎥ ⎢ ⎥ ⎢ εy ⎥ ⎥ ⎥⎢ εz ⎥ 0⎥ ⎥ = [G] {ε} ⎥⎢ γx y ⎥ ⎥⎢ ⎢ ⎥ 0⎥ ⎥ ⎣ γ yz ⎦ 0 ⎦ γzx G (12.8.2) where λ= νE (1 + ν) (1 − 2ν) The equilibrium equations are given in Equation 2.6.13 and are repeated here as Equation 12.8.3. ⎤⎡ ⎡ ⎤ ∂ ∂ ∂ σx 0 0 0 ⎥ ⎢ ∂x ∂y ∂z ⎥ ⎢ σ y ⎥ ⎡ ⎤ ⎢ ⎥ f bx ⎥⎢ ⎢ ∂ ∂ ∂ ⎢ σz ⎥ ⎥ ⎢ 0 ⎥ = − ⎣ f by ⎦ = − { f b } 0 0 ⎥⎢ (12.8.3) [E] {σ } = ⎢ ⎥ ∂y ∂ x ∂z ⎥⎢ ⎢ ⎢ τx y ⎥ f bz ⎥ ⎢ ⎣ ⎦ ∂ ∂ ∂ ⎦ τ yz ⎣ 0 0 0 τzx ∂z ∂y ∂x Exact analytical solutions to these equations have been found only in instances of very simple geometry, loading, and material properties. In Chapter 2, Example 2.6.2, we solved them for the stress in a cube under uniform hydrostatic pressure. In Chapter 3, Example 3.5.1 we solved them for a thin rectangular plate with a uniform edge load, but in Example 3.5.2, we show that the solution becomes much more difficult when the loading is not uniform. Continuing in Chapter 4, Section 4.2, we show that the equations are solved exactly for a slender bar with a rectangular cross section as long as the axial force in uniformly distributed over the cross section. In Chapter 6, Section 6.4, we show that they are solved for a slender shaft with a circular cross section if the loading is a radially linear force per unit area applied at the ends. And in Chapter 7, Section 7.7, --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come 471 Structural Analysis in Two and Three Dimensions we show that they are solved for a beam with a rectangular cross section and a linearly varying normal force per unit area and/or a parabolic shear force per unit area on the ends. In all cases the material is homogenous and isotropic. And in all cases the message is that any change from the simple geometry assumed or any variation from the constant or linearly varying loading makes an exact analytical solution at best difficult and most likely impossible. There are so few analytical solutions to these equations we turn directly to virtual work as expressed in Equations 12.3.1 and derive the FEM equations. The shape functions for four elements, the four and ten noded tetrahedrons and the eight and twenty noded hexahedrons, are the ones primarily in use. The four noded tetrahedron is shown in Figure 12.8.1. y Y,υ X,u Z,w x z Figure 12.8.1 At node n ⎡ ⎤ un {rm }n = ⎣ vn ⎦ wn (12.8.4) There are four nodes for element m with three nodal displacements at each node. The shape function matrix is of the form ⎡ ⎤ u (x, y, z) ⎣ v (x, y, z) ⎦ = [3x12] [12x1] = [n] {rm } (12.8.5) w (x, y, z) The element stiffness matrix is formed from the same virtual work equations as Equations 11.4.20. We enter the order of the matrices rather than type them in full. [ pm ] = [12x1] = [12x12] [12x1] = [km ] {rm } (12.8.6) The assembly process to produce the global matrices proceeds are described before. The ten noded tetrahedral element is shown in Figure 12.8.2. This currently is the most used element because of its utility in automatic mesh generation. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 472 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y Y,υ X,u Z,w x z Figure 12.8.2 The hexahedral elements are six sided blocks. Each face is a quadrilateral. Its irregular shape allows it to match solid bodies of various shapes. Solid elements must be used when the simplifying assumptions of slender bar and shell theory do not apply. Here is an example. ########### Example 12.8.1 Problem: The bracket shown in Figure (a) is loaded with a vertical force on a pin placed in the hole. The head of the pin transmits the load to the bracket. The bracket is attached to what is assumed to be a rigid wall by an adhesive. Find the stresses and displacements and modify the design to reduce the highest stresses. The material is steel with E = 206899 MPa. The load is 1000 N. Solution: Use a commercial FEM code to find the stresses and displacements. Figure (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come Structural Analysis in Two and Three Dimensions 473 First we create a surface on which to apply the load as shown in Figure (b). Figure (b) Now apply the boundary condition restraints and loads as shown in Figure (c). Figure (c) We now invoke the automatic mesh generator and use ten noded tetrahedral elements as shown in Figure (d). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 474 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Next the solver. As might be expected the highest stresses are at the sharp corner as shown in the contour plot of the von Mises stresses in Figure (e). Figure (e) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure (d) P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come Structural Analysis in Two and Three Dimensions 475 The maximum displacement is 0.645 mm. If the maximum von Mises stress of 225 MPa is higher than desired it can be reduced by design changes in the geometry. A common change is to put a fillet in the corner. This has been done as is shown in Figure (f). Figure (f) The boundary conditions are added; the new model is meshed; the solver is run again and the results are displayed in Figure (g). Figure (g) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas 476 July 4, 2011 7:16 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The maximum stress is reduced but perhaps not enough so let us consider another design change. A stiffener is added as shown in Figure (h). Figure (h) The results of this analysis are shown in Figure (i). Figure (i) There is a considerable reduction in both the maximum stress and the maximum displacement. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come Structural Analysis in Two and Three Dimensions 477 The purpose of this exercise is not to create an optimum design but rather to illustrate that changes in geometry no longer restrict the ability to get accurate solutions. All three of these brackets would be very difficult to analyze, even with approximate analytic methods, before the arrival of FEM. 12.9 Summary and Conclusions --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Exact analytical solutions of the classical differential equations of two and three dimension structures are limited to only a few very simple cases of geometry, loading, and restraint. Several books on the theory of elasticity grapple with this problem and for the most part are forced to go to approximate methods. The finite element method is by far the most productive. Therefore we emphasize this approach and largely skip all others. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-12 P2: ABC JWST071-Waas July 4, 2011 7:16 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 13 Analysis of Thin Laminated Composite Material Structures 13.1 Introduction to Classical Lamination Theory Fiber-reinforced composite laminates are increasingly being used in the aerospace industry and are now being used in a wide variety of aerospace structural applications. The Boeing 787 commercial airplane represents a revolutionary step in the application of thin-walled laminated composites in primary (load bearing) aircraft structural components such as the fuselage, tail and wing. Ongoing developments and advancements in the fiber and polymer manufacturing industries will undoubtedly enhance the performance of future laminated structural components with stronger fibers and tougher resin systems. MATRIX FIBER --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Laminate Typical Lamina Figure 13.1.1 Schematic of a continuous fiber laminated composite A laminated structure contains multiple layers that are laminated together with each layer oriented in a specified direction selected by the design engineer, as schematically shown in Figure 13.1.1. Since the Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 480 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods fiber orientation and the manner by which the layers are stacked (usually called the stacking sequence) can be changed as desired, the overall structural properties of a fiber reinforced laminate can be tailored as needed. This type of flexibility is one of the main advantages of using a laminate in a structural application as will be evident in this chapter. Because one can change structural properties, a design engineer can “optimize” the deformation response and dynamic behavior of structural components as desired. Although there are a wide variety of fiber reinforced laminates in the market today, in this chapter we will focus on analyzing the deformation of continuous, straight fiber reinforced laminates, that are made by stacking laminae, the latter being the basic building block of laminates. When a laminate is fabricated, they act as one single layer material. The bond between any two laminae in a laminate is assumed to be perfect, that is, infinitesimally thin and not shear deformable. Thus, laminae cannot slip over each other, and the displacements therefore must remain continuous across the bond line (or the interface). 13.2 Strain Displacement Equations for Laminates Let us consider the deformation of a section of a laminate in the xz plane when we place the coordinate axes such that the xy plane coincides with the geometrical midplane of the laminate, as shown in Figure 13.2.1 xy plane u0 A C x A′ D z C′ D′ B Undeformed w0 α B′ zα Deformed Figure 13.2.1 1. Assume that a line ACDB, originally straight and perpendicular to the x axis, also remains straight and perpendicular to the deformed geometric midplane of the laminate. 2. Further assume that point C undergoes displacements u 0 (x, y) , v0 (x, y) and w0 (x, y) along the x, y and z directions respectively. Consider now the deformation of point D, which was located at a distance Z below the midplane in the undeformed configuration. Upon deformation to point D’ we can write the following displacement field for point D, assuming small (compared to the total laminate thickness) displacements; u (x, y) = u 0 (x, y) − zα (x, y) (13.2.1) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 481 Analysis of Thin Laminated Composite Material Structures where α(x, y) is the slope of the laminate’s midplane is the x-direction. Notice that we are assuming α(x, y) to be small in the sense that tan α ≈ α and cos α ≈ 1, that is, small slopes. Then α= ∂w0 ∂x → u = u0 − z ∂w0 ∂x (13.2.2) By a similar consideration v = v0 − z ∂w0 ∂y (13.2.3) The displacement w (x, y, z) of any point on ACDB is the displacement w0 (x, y) of the midplane (z = 0), plus any stretching of ACDB. It is assumed that the stretch of ACDB is insignificant compared to w0 itself and thus the “w” displacement of any point on ACDB is w0 of the midplane (point C) itself. Let us write the infinitesimal strains εx , ε y and εz for the displacement field, given by Equation 13.2.4 as shown in Equation 13.2.5 ∂w0 (x, y) ∂x ∂w0 (x, y) v (x, y, z) = v0 (x, y) − z ∂y u (x, y, z) = u 0 (x, y) − z (13.2.4) w (x, y, z) = w0 (x, y) The strains are therefore, ∂u 0 (x, y) ∂ 2 w0 (x, y) ∂u = −z ∂x ∂x ∂x2 2 ∂v0 (x, y) ∂ w0 (x, y) ∂v = −z εy = ∂y ∂y ∂ y2 ∂v ∂u 0 (x, y) ∂v0 (x, y) ∂ 2 w0 (x, y) ∂u + = + − 2z 2εxy = ∂y ∂z ∂y ∂x ∂ x∂ y εx = (13.2.5) Notice that γ yz and γzx are zero and that εz is zero on account of the assumption that w (x, y, z) = w0 (x, y). We can write Equation 13.2.5 as ⎡ 0⎤ ⎡ ⎤ ⎡ ⎤ εx χx εx ⎢ 0⎥ ⎣ εy ⎦ = ⎢ εy ⎥ + z ⎣ χy ⎦ (13.2.6) ⎣ ⎦ χxy γxy 0 γxy where εx0 ⎡ ⎤ ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ∂v0 ⎦ + ∂y ∂x ⎢ ⎢ ⎢ ⎢ 0⎥ ⎢ ⎣ εy ⎦ = ⎢ ⎢ ⎢ γxy0 ⎣ ∂u 0 ⎡ ∂u 0 ∂x ∂v0 ∂y ⎤ ∂ 2 w0 (x, y) − ⎥ ∂x2 ⎡ ⎤ ⎢ ⎥ ⎢ χx ⎢ ∂ 2 w (x, y) ⎥ ⎥ ⎢ 0 ⎣ χy ⎦ = ⎢ − ⎥ 2 ⎥ ⎢ ∂ y χxy ⎥ ⎢ ⎣ ∂ 2 w0 (x, y) ⎦ −2 ∂ x∂ y ⎡ (13.2.7) To obtain the stress distribution across the laminate we need to use the strain distribution, Equation 13.2.6, and the lamina constitutive description. This will require using the stress-strain relations for laminae, as presented next. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 482 13.3 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Stress-Strain Relations for a Single Lamina For a transversely isotropic fiber reinforced lamina as shown in Figure 13.3.1, we have, ⎡ ⎤ 1 v21 v21 − − 0 0 0 ⎢ E1 ⎥ E2 E2 ⎢ ⎥ ⎢ v12 ⎥ v32 (= v23 ) 1 ⎢− ⎥ − 0 0 0 ⎢ ⎥⎡ ⎤ ⎡ ⎤ ⎢ E1 E2 E2 ⎥ σ ε1 ⎢ ⎥ 1 ⎥⎢ ⎥ v12 v23 1 ⎢ ε2 ⎥ ⎢ ⎢ ⎥ ⎢ σ2 ⎥ − − 0 0 0 ⎢ ⎥ ⎢ ⎥ E2 E2 ⎢ ε3 ⎥ ⎢ E 1 σ3 ⎥ ⎥⎢ ⎢ ⎥ ⎥ ⎥⎢ ⎢ τ23 ⎥ ⎢ γ23 ⎥ = ⎢ 2(1 + v23 ) ⎢ ⎥ ⎢ ⎢ ⎥ ⎢ 0 0 0 0 0 ⎥⎣ ⎥ ⎣ γ31 ⎦ ⎢ E2 ⎥ τ31 ⎦ ⎢ ⎥ ⎢ ⎥ τ12 γ12 1 ⎢ 0 ⎥ 0 0 0 0 ⎢ ⎥ G 12 ⎢ ⎥ ⎢ ⎥ 1 ⎣ ⎦ 0 0 0 0 0 G 12 (13.3.1) 3,z 2 y x y 1 D C 1 2 θ A Figure 13.3.1 x B A single lamina of a fiber reinforced laminate. For a condition of plane stress (in the 1-2 plane), we can set the out-of-plane stress components to be negligible (owing to the relatively small cross-sectional area of planes with normals in the “1” and “2” directions, because of the small lamina thickness) compared to the in-plane stress components; thus, set σ3 = τ23 = τ31 = 0 (13.3.2) so that ⎡ ⎤ ⎡ ε1 S11 ⎣ ε2 ⎦ = ⎣ S21 γ12 0 S12 S22 0 ⎤ ⎡ ⎤ 0 σ1 0 ⎦ = ⎣ σ2 ⎦ τ12 S66 (13.3.3) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 483 Analysis of Thin Laminated Composite Material Structures where S11 = 1 E1 S22 = 1 E2 S12 = S21 = − ν21 ν12 =− E2 E1 S66 = 1 G 12 (13.3.4) Therfore, stress in terms of strain is ⎡ ⎤ ⎡ σ1 Q 11 ⎣ σ2 ⎦ = ⎣ Q 21 τ12 0 Q 12 Q 22 0 ⎡ ⎤ ⎡ ε1 ⎤ 0 ⎢ ⎥ ⎢ 0 ⎦ ⎣ ε2 ⎦ = [Q] ⎣ γ12 2Q 66 2 ⎤ ε1 ε2 ⎥ ⎦ γ12 2 (13.3.5) where E1 S22 = 2 1 − ν21 ν12 S11 S22 − S12 ν12 E 2 S12 = = = Q 21 2 1 − ν21 ν12 S11 S22 − S12 E2 S11 = = 2 1 − ν21 ν12 S11 S22 − S12 1 = = G 12 S66 Q 11 = Q 12 Q 22 Q 66 (13.3.6) Consider a lamina, whose fiber direction (the “1” direction) is at some angle, θ, to the “x” coordinate direction. Using the rules for stress tranformations, the components of stress in the x-y plane can be found from, ⎡ ⎤ ⎤⎡ ⎤ ⎡ ⎤ ⎡ 2 s2 −2cs σ1 σ1 c σx −1 2 2 ⎣ σy ⎦ = ⎣ s c 2cs ⎦ ⎣ σ2 ⎦ = [T ] ⎣ σ2 ⎦ (13.3.7) τ12 τ12 τxy cs −cs c2 − s 2 where c = cos θ s = sin θ (13.3.8) It follows that --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎤ ⎡ ⎤ σx σ1 ⎣ σ2 ⎦ = [T ] ⎣ σ y ⎦ τ12 τxy ⎡ (13.3.9) where [T ] −1 ⎡ c2 = ⎣ s2 cs s2 c2 −cs ⎤ −2cs 2cs ⎦ c2 − s 2 The tensorial strain transformation is similar. ⎤ ⎤ ⎡ ⎡ εx ε1 ⎢ ε ⎥ ⎢ ε2 ⎥ ⎦ = [T ] ⎣ y ⎦ ⎣ γ12 γxy 2 2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (13.3.10) (13.3.11) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 6, 2011 484 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Note that we have used half of the engineering shear strain, γ12 , in the above tensor transformation relations. Therefore ⎡ ⎤ ⎡ ⎤ εx σx ε ⎥ ⎣ σ y ⎦ = [T ]−1 [Q] [T ] ⎢ (13.3.12) ⎣ y ⎦ γxy τxy 2 which ultimately provides (now, using engineering shear strain in the final result) ⎤ ⎡ ⎤ ⎡Q Q 12 Q 16 ⎡ ε ⎤ 11 σx x ⎥ ⎣ σy ⎦ = ⎢ (13.3.13) ⎣ Q 12 Q 22 Q 26 ⎦ ⎣ ε y ⎦ γxy τxy Q Q Q 16 26 66 where Q 11 = Q 11 c4 + Q 22 s 4 + 2 (Q 12 + 2Q 66 ) s 2 c2 Q 12 = (Q 11 + Q 22 − 4Q 66 ) c2 s 2 + Q 12 (s 4 + c4 ) Q 22 = Q 11 s 4 + Q 22 c4 + 2 (Q 12 + 2Q 66 ) s 2 c2 Q 16 = (Q 11 − Q 12 − 2Q 66 ) c3 s − (Q 22 − Q 12 − 2Q 66 ) cs 3 (13.3.14) Q 26 = (Q 11 − Q 12 − 2Q 66 ) cs 3 − (Q 22 − Q 12 − 2Q 66 ) c3 s Q 66 = (Q 11 + Q 22 − 2Q 12 − 2Q 66 ) c2 s 2 + Q 66 (s 4 + c4 ) There are only four independent constants in the Equation 13.3.12, we obtain ⎡ ⎤ ⎡ S 11 εx ⎣ ε y ⎦ = ⎣ S 12 γxy S 16 where set of equations given above in 13.3.13. Inverting S 12 S 22 S 26 ⎤⎡ ⎤ S 16 σx S 26 ⎦ ⎣ σ y ⎦ τxy S 66 (13.3.15) S 11 = S11 c4 + (2S12 + S66 ) s 2 c2 + S22 s 4 S 12 = S12 (s 4 + c4 ) + (S11 + S22 − S66 ) s 2 c2 S 22 = S11 s 4 + (2S12 + S66 ) s 2 c2 + S22 c4 S 16 = (2S11 − 2S12 − S66 ) sc3 − (2S22 − 2S12 − S66 ) s 3 c (13.3.16) S 26 = (2S11 − 2S12 − S66 ) s 3 c − (2S22 − 2S12 − S66 ) sc3 S 66 = 2 (2S11 + 2S22 − 4S12 − S66 ) s 2 c2 + S66 (c4 + s 4 ) For example, suppose we want the modulus of elasticity associated with uni-axial loading along the x-direction. Then, in Equation 13.3.14, we set, σx = σ σ y = τxy = 0, (13.3.17) so that, Ex = = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS σx σ 1 1 = = = 4 + (2S + S ) s 2 c2 + S s 4 εx S c S 11 σ S 11 11 12 66 22 1  ν12 1 1 4 c + −2 + E1 E1 G 12 s 2 c2 + 1 4 s E2 . (13.3.18) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-13 P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 485 Analysis of Thin Laminated Composite Material Structures S12 s 4 + c4 + (S11 + S22 − S66 ) s 2 c2 εy S 12 =− =− εx S11 c4 + (2S12 + S66 ) s 2 c2 + S22 s 4 S 11  1 1 ν12 4 1 s 2 c2 s + c4 + + − E1 E1 E2 G 12  =− ν12 1 1 4 1 4 s 2 c2 + c + −2 + s E1 E1 G 12 E2 υxy = − (13.3.19) We can also define shear coupling coefficients (similar to Poisson’s ratios) now. For example, ηx,x y = γxy S 16 = εx S 11 (13.3.20) which is a measure of the amount of shear strain generated in the xy plane per unit normal strain along the direction of applied normal stress, σx . Clearly ηx,x y = 0 for θ = 0◦ and θ = 90◦ . Another shear coupling constant exists for different types of load states. For example, when τxy = 0 and σx = σ y = 0, then ηx y,y = εy S 26 = γxy S 66 (13.3.21) This characterizes the normal strain response along the y-direction due to a shear stress in the xy plane. We can also write Q ij in the following convenient form; Q 11 = u 1 + u 2 cos 2θ + u 3 cos 4θ Q 12 = u 4 − u 3 cos 4θ Q 22 = u 1 − u 2 cos 2θ + u 3 cos 4θ u2 Q 16 = − sin 2θ − u 3 sin 4θ 2 u2 Q 26 = − sin 2θ + u 3 sin 4θ 2 1 Q 66 = (u 1 − u 4 ) − u 3 cos 4θ = u 5 − u 3 cos 4θ 2 (13.3.22) where 1 (3Q 11 + 3Q 22 + 2Q 12 + 4Q 66 ) 8 1 u 2 = (Q 11 − Q 22 ) 2 1 (13.3.23) u 3 = (Q 11 + Q 22 − 2Q 12 − 4Q 66 ) 8 1 u 4 = (Q 11 + Q 22 + 6Q 12 − 4Q 66 ) 8 1 u 5 = (u 1 − u 4 ) 2 Now, u 1 , u 2 , u 3 , u 4 , u 5 are invariants and therefore do not change with respect to rotation in the 1-2 plane. All the rotational effects are captured in the trigonometric functions appearing in Equation 13.3.21. Thus these equations can be useful for optimization problems where optimum lay-ups (or stacking sequences of the laminae) are sought for a given application. Having obtained the lamina stress-strain relations in the x-y coordinate frame, we can now obtain the stresses in the kth lamina. Let any point on the kth lamina be situated between z = z k and z = z k−1 . Then, u1 = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Furthermore, the major Poisson’s ratio follows as, P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 486 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods from the constitutive equations, Equations 13.3.12, and Equations 13.2.6 ⎡ ⎤ Q 11 σx ⎣ σy ⎦ = ⎢ ⎣ Q 12 τxy k th Q 16 ⎤ Q 12 Q 16 Q 22 Q 26 ⎥ Q 26 ⎦ Q 11 Q 12 Q 16 ⎢ = ⎣ Q 12 Q 22 ⎥ Q 26 ⎦ ⎡ ⎡ Q 16 Q 26 Q 66 k th ⎤ Q 66 ⎤ εx ⎣ εy ⎦ γxy ⎡ ⎡ k th εx0 (13.3.24) ⎤ ⎡ ⎤ Q 11 Q 12 Q 16 ⎢ ⎢ 0⎥ ⎣ ε y ⎦ + z ⎣ Q 12 Q 22 ⎥ Q 26 ⎦ γxy0 Q 16 Q 26 Q 66 k th ⎡ ⎤ χx ⎣ χy ⎦ χxy where z k−1 ≤ z ≤ z k 13.4 (13.3.25) Stress Resultants for Laminates The stresses in a laminate vary from layer to layer, as sketched in Figure 13.4.1. 1 2 x 3 4 z Linear varitation of strain through the thickness Laminate zk−1 z Variation of stress is linear across each lamina but is discontinuous across the laminate thickness (see later eqn.7) zk Kth lamina Figure 13.4.1 Thus, it is convenient to introduce the notion of stress resultants and moment resultants, which are thickness integrated quantities, as defined in Equations 13.4.1 and 13.4.2, and acting as shown in Figure 13.4.2 A resultant is defined as one that is obtained by integrating a particular component over the entire thickness of the laminate. Thus they are also referred to as ‘thickness average’ resultants. Introduce, the in-plane force resultants, N x , N y and Nxy , as, h 2 Nx = − h2 h 2 σx dz Ny = − h2 h 2 σ y dz Nxy = − h2 τxy dz (13.4.1) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 487 Analysis of Thin Laminated Composite Material Structures Mx δF = (σx dz) δMx = (σx dz)z x dz h/2 h/2 δF Mx = σx zdz δMx = −h/2 −h/2 zk−1 z zk zk+1 Kth lamina Nomenclature for integration x z x Mx Nx Myx y Ny Mxy My y Moment-Axial force conventions Figure 13.4.2 Stress and Moment Resultants Similarly the resultant moments can be defined as follows; h 2 Mx = − h2 h 2 σx zdz My = − h2 h 2 σ y zdz Mxy = − h2 τxy zdz (13.4.2) Positive forces and moments are as indicated in Figure 13.4.2. Since the stresses in each layer are different, the integrals defined in Equations 13.4.1 and 13.4.2 have to be converted to sums over each layer. For example h 2 Nx = − h2 −h 1 σx dz = hn −h 2 σx dz + −h 0 σx dz . . . . . . . . . .. + σx dz (13.4.3) h n−1 −h 1 The thickness coordinates used in the integration are as shown in Figure 13.4.3. The force resultants can be written as, ⎡ ⎤ ⎡ ⎤ ⎤ ⎡ h n σx σx Nx hk 2 ⎣ σ y ⎦dz, ⎣ σ y ⎦dz = ⎣ Ny ⎦ = (13.4.4) − h2 k=1 h k−1 τxy τxy Nxy and the moment resultants, as, ⎡ ⎤ Mx ⎣ My ⎦ = Mxy h 2 − h2 ⎤ σx ⎣ σ y ⎦zdz = τxy ⎡ n k=1 ⎤ σx ⎣ σ y ⎦zdz h k−1 τxy hk ⎡ (13.4.5) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 6, 2011 488 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods h0 1 2 3 h1 h2 h h3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-13 Geometric midplane x hn-3 hn-2 hn hn-1 k n Figure 13.4.3 Nomenclature for thickness integration By substituting Equation 13.3.23 into Equations 13.4.4 and 13.4.5 we get ⎤ ⎡ 0⎤ ⎡ ⎤ ⎡ Q 11 Q 12 Q 16 εx n Nx hk ⎥ ⎢ 0⎥ ⎢ ⎣ Ny ⎦ = ε ⎣ Q 12 Q 22 Q 26 ⎦ ⎣ y ⎦dz k=1 h k−1 Nxy Q 16 Q 26 Q 66 k th γxy0 n hk k=1 h k−1 + and ⎤ Mx ⎣ My ⎦ = Mxy ⎡ n hk k=1 h k−1 ⎡ ⎡ Q 12 Q 16 ⎢ ⎣ Q 12 Q 22 ⎥ Q 26 ⎦ Q 16 Q 26 k th ⎤ ⎡ Q 12 Q 16 ⎢ ⎣ Q 12 Q 22 ⎥ Q 26 ⎦ Q 16 hk k=1 h k−1 + ⎡ Q 26 Q 66 ⎤ Q 12 Q 16 ⎢ ⎣ Q 12 Q 22 ⎥ Q 26 ⎦ Q 16 Q 26 εx0 ⎤ ⎢ 0⎥ ⎣ ε y ⎦ zdz γxy0 k th Q 11 Q 66 (13.4.6) ⎤ χx ⎣ χ y ⎦zdz χxy ⎡ Q 66 Q 11 n ⎤ Q 11 (13.4.7) ⎡ k th ⎤ χx ⎣ χ y ⎦z 2 dz χxy Noting that the midplane strains and plate curvatures remain constant not only within a lamina but for all laminae in the laminate, we obtain from Equations 13.4.6 and 13.4.7 ⎤ ⎡ ⎤ ⎡ ε0 ⎤ ⎡ ⎤⎡ ⎤ ⎡ x B11 B12 B16 A11 A12 A16 χx Nx ⎥ ⎢ ⎣ N y ⎦ = ⎣ A12 A22 A26 ⎦ ⎣ ε0y ⎦ + ⎣ B12 B22 B26 ⎦ ⎣ χ y ⎦ (13.4.8) 0 Nxy A16 A26 A66 B B B χ 16 26 66 xy γxy and ⎤ ⎡ B11 Mx ⎣ M y ⎦ = ⎣ B12 Mxy B16 ⎡ Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS B12 B22 B26 ⎤ ⎡ ε0 ⎤ ⎡ x D11 B16 ⎢ ⎥ B26 ⎦ ⎣ ε0y ⎦ + ⎣ D12 D16 B66 γ0 xy D12 D22 D26 ⎤⎡ ⎤ D16 χx D26 ⎦ ⎣ χ y ⎦ D66 χxy (13.4.9) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 489 Analysis of Thin Laminated Composite Material Structures where n Aij = k=1 Bij = Dij = 13.5 1 2 1 3  Q ij n k=1 n k=1   Q ij  (h k − h k−1 ) k  Q ij k  k h 2k − h 2k−1 (13.4.10) h 3k − h 3k−1 CLT Constitutive Description Equations 13.4.8 and 13.4.9 can be represented by    {N } [A] = {M} [B] [B] [D]    0  ε {χ } , (13.5.1) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- where each of the symbols A, B and D represent a 3 × 3 matrix. [A] is an extensional stiffness matrix, [B] is a coupling (extensional-bending) stiffness matrix, and [D] is a bending stiffness matrix. Note that in-plane loads are determined by mid-plane strains and plate curvatures (stretch-bending coupling), that is, both stretching (and compressing) and bending induce mid-plane stress resultants. The bending effects in this case are related to the B matrix and it is apparent that there would be no in-plane/bending coupling if B = 0. The Bij are even functions of the h k and thus will be zero for laminates which are symmetric about the midplane, that is, for each layer above the midplane there is an identical layer (in properties, thickness, and orientation) located at the same distance below the midplane. Examination of the Aij shows another form of coupling: in-plane shear and normal forces as represented in Equation 13.5.2. ⎤ ⎡ ⎤ ⎡ ε0 ⎤ ⎡ x A11 A12 A16 Nx ⎥ ⎣ N y ⎦ = ⎣ A12 A22 A26 ⎦ ⎢ ε (13.5.2) ⎣ 0y ⎦ 0 Nxy A16 A26 A66 γxy This particular coupling can be removed if A16 = A26 = 0. From the definition of Aij we see that Aij can be zero only if all the Q ij are zero or if some Q ij are positive and some negative. The Q 11 , Q 12 , Q 22 and Q 66 are always positive; and so are the corresponding Aij . The Q 16 and Q 26 however are zero for 00 and 900 orientation and can be either positive or negative, since odd parameters c and s appear in the defining expression. So, if for every lamina with positive +θ orientation, there is another lamina with the identical properties and thickness at negative θ, then A16 = A26 = 0. The z position of the lamina is not important, that is, it need not be midplane symmetric. Of course, if it were, then additionaly Bij = 0. From Equation 13.5.1 we have   {M} = [B] ε 0 + [D] {χ } (13.5.3) We see that [B]also relates in-plane and bending effects. Furthermore, note that the moments Mx and M y are due, in part, to midplane shearing and plate twisting while Mxy is partly due to midplane normal strain and normal plate curvature. The coupling between plate twisting and normal bending is eliminated if D16 = D26 = 0. This will occur if Q 16 = Q 26 = 0, that is, for 00 or 900 orientation or if for every layer at +θ at a given distance above the midplane there is an identical layer, that is, in properties and thickness, at the same distance below the midplane but at negative θ. The h 3k − h 3k−1 terms are the same for both the layers but the Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 490 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Q terms are of opposite sign. Unfortunately, such a laminate would not possess midplane symmetry; however, the D16 and D26 → 0 for thick laminates since the ±θ layers tend to cancel each other. Knowledge of the [A], [B], and [D] matrices allows the stresses within the laminate to be found. The stress resultants will be known from an analysis of the structure, hence the midplane strains and plate curvatures may be found by inverting the constitutive equation. Then, using the transformation matrix [T ], for each layer, the lamina stresses in the principal material directions, 1-2, are obtained. Usually a limiting strain would be of interest. For example, it is of interest to know the fiber direction tensile strain, ε1 , to examine if it exceeds the fiber direction maximum allowable tensile strain. This can be obtained from coupon level material tensile tests or from a handbook of materials strength data. In practice, the situation is slightly more complex than indicated since a thermal strain must be included in the formulation of the constitutive equation to account for stresses that arise during the curing of the laminate. These stresses may be high enough to crack the laminate. The above equations are commonly known as ‘Classical Lamination Theory’ and are adequate for predicting laminate stiffness behavior and laminae stresses away from an edge. In the region of an edge a three dimensional analysis is needed to calculate the through the thickness stresses. ########### Example 13.5.1 Problem: A two layer laminate consists of layer (sometimes, the word “ply” is also used to describe a layer) orientation of 00 and 450 with respect to the x-y axes as shown in Figure (a). x y 3 mm 5 mm Figure (a) Two layer laminate --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The 00 lamina is 5 mm thick and the 450 lamina is 3 mm thick. The [Q]matrix for the ply material is given in Equation 13.5.1. ⎡ ⎤ 20 0.7 0 0 ⎦ × 109 N/mm2 (a) [Q] = ⎣ 0.7 20 0 0 0.7 The thickness coordinates are (refer to Figure 13.4.3 for the nomenclature) h 0 = −4 mm h 1 = −1 mm h 2 = 4 mm (b) Obtain the laminate constitutive description, that is, the A, B and D matrices. Solution: The [Q] matrix is defined in Equations 13.3.5 and 13.3.6 and its values are given inthe   statement of the problem. The Q matrix is defined in Equation 13.3.12 and 13.3.13. First find the Q matrices for 00 and 450 and then use Equations 13.4.10 to find [ A], [B], and [D]. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 491 Analysis of Thin Laminated Composite Material Structures     Step 1: Find Q 00 and Q 45◦ .   The Q 00 matrix is the same as [Q] since cos θ = 1and sin θ = 0, that is, ⎡ ⎤ 20 0.7 0   0 ⎦ GPa Q 00 = ⎣ 0.7 20 0 0 0.7   For the Q 45◦ matrix we have (c) Q 11 = 20 (cos 45◦ )4 + 2 (sin 45◦ )4 + 2 (0.7 + 2 × 0.7) (sin 45◦ )2 (cos 45◦ )2 = 6.55 GPa (d) Similarly, Q 22 = 6.55 GPa Q 12 = 5.15 GPa Q 66 = 5.15 GPa Q 16 = 4.50 GPa (e) and so ⎡ Q 11 Q 12   ⎢ Q 45◦ = ⎣ Q 12 Q 22 Q 16 Q 16 ⎤ ⎡ 6.55 ⎥ ⎣ Q 26 ⎦ = 5.15 4.50 Q Q 26 66 ⎤ 5.15 4.50 6.55 4.50 ⎦ GPa 4.50 5.15 (f) Step 2: Find [A], [B], and [D]. Note that the interface does not coincide with the midplane. n Bij = 1 2 k=1  k     (h k − h k−1 ) = Q ij 45◦ ((−1) − (−4)) + Q ij 00 ((4) − (−1)) n k=1  Q ij = −7.5 Q ij 1 3 Q ij     = 3 Q ij 45◦ + 5 Q ij 0◦ GPa mm ⎡ ⎤ 119.65 18.95 13.50 [A] = ⎣ 18.95 29.65 13.50 ⎦ GPa mm 13.50 13.50 18.95  Dij =   k  45◦ n k=1   = 21 Q ij Q ij  45◦  k h 2k − h 2k−1 =   1  1  Q ij 45◦ (−1)2 − (−4)2 + Q ij 0◦ (4)2 − (−1)2 2 2 + 7.5 Q ij 0◦ GPa ⎡ 100.9 [B] = ⎣ −33.4 −33.75 h 3k − h 3k−1 =  + 21.67 Q ij  mm (h) (i) 2 −33.4 −34.1 −33.75 ⎤ −33.75 −33.75 ⎦ GPa mm2 −33.4 1  1  Q ij 45◦ (−1)3 − (−4)3 + Q ij 0◦ (4)3 − (−1)3 3 3 (j) (k) 3 GPa mm ⎡ 571 123 [D] = ⎣ 123 181 94.5 94.5 0◦ ⎤ 94.5 94.5 ⎦ GPa mm3 123 ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (g) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (l) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Aij = P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas 492 13.6 July 6, 2011 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Determining Laminae Stress/Strains We know from Equation 13.5.1, repeated here as Equation 13.6.1, that       0  {N } ε [A] [B] = {M} {χ } [B] [D] (13.6.1) Expanding we have   {N } = [A] ε0 + [B] {χ }   {M} = [B] ε 0 + [D] {χ }   Solving for ε0 from the first equation of 13.6.2, we have,  0 ε = [A]−1 ({N} − [B] {χ }) (13.6.2) (13.6.3) From the second of Equations 13.6.2, we get {M} = [B] [ A]−1 ({N } − [B] {χ }) + [D] {χ } Thus we get a partially inverted system of equations given by   0   ∗   {N } ε [A ] [B ∗ ] = ∗ ∗ {χ } {M} [C ] [D ] (13.6.4) (13.6.5) where [A∗ ] = [A]−1 [B ∗ ] = − [A]−1 [B] [C ∗ ] = [B] [ A]−1 = − [B ∗ ]T (13.6.6) [D ∗ ] = [D] − [B] [ A]−1 [B] From the second of Equations 13.6.5 we get  −1  ∗   −1 {M} − D ∗ {χ } = D ∗ C {N } Using this in the first of Equations 13.6.5 we get  0   ∗   ∗   ∗ −1  ∗     −1 {M} {N } + B ∗ D ∗ ε = A − B D C From Equations 13.6.7 and 13.6.8 we get the fully inverted form   0    ′  ′    B A {N } ε =  ′  ′ {χ } {M} D C (13.6.7) (13.6.8) (13.6.9) where  ′ A = [A∗ ] − [B ∗ ] [D ∗ ]−1 [C ∗ ]  ′ B = [B ∗ ] [D ∗ ]−1  ′  T C = − [D ∗ ]−1 [C ∗ ] = B ′  ′ D = [D ∗ ]−1 (13.6.10) This inverted form, Equation 13.6.9, is also very useful for solving certain types of boundary value problems. --`,,,,``,``,,``,,` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 493 Analysis of Thin Laminated Composite Material Structures 13.7 Laminated Plates Subject to Transverse Loads The theory that was developed (CLT) assumes that each lamina and the laminate as a whole is in a state of plane stress and further it is assumed that plane sections remain-plane and normal to the deformed centerline, as in classical plate theory. Furthermore, with the restriction to linear elastic material response, it is assumed that strains are small. In Figure 13.7.1 the resultant forces acting on an element of a plate with transverse loading is shown. We can now derive the equilibrium equations for laminated plates, using a direct approach, in which we use summation of forces and summation of moments. Then, with respect to Figure 13.7.1, summation of forces in the x-direction provides N x dy + ∂ Nxy ∂ Nx dydx + Nxy dx + dydx − N x dy − Nxy dx = 0 ∂x ∂y which gives us x --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Nyx x dy dx z dy Nx Nx + dx ∂Nx ∂x dx ∂Nyx dy y Nyx + ∂y y Only “x” forces shown My Qy x q (x, y) Qx + Qx ∂Qx dy Mxy Mxy + dx y My + ∂My ∂y Figure 13.7.1 ∂x dx ∂Mxy ∂x dx ∂Qy dy Qy + ∂y dy Force and moment resultants and transverse pressure loads Force and moment resultants and transverse pressure loads q(x, y), on an infinitesimal “chunk” of a laminate in the x-y plane ∂ Nxy ∂ Nx + =0 ∂x ∂y (13.7.1) Similarly in the y-direction N y dx + ∂ Ny ∂ Nxy dydx + Nxy dy + dydx − N y dx − Nxy dy = 0 ∂y ∂x ∂ Nxy ∂ Ny + =0 ∂y ∂x Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (13.7.2) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 494 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods For force equilibrium in the z- direction (out of plane), the edge shears equilibrate the transverse pressure loading q (x, y). Q x dy + ∂ Qx ∂ Qy dydx + Q y dx + dydx − Q x dy − Q y dx + q (x, y) dxdy = 0 ∂x ∂y ∂ Qy ∂ Qx + + q (x, y) = 0 ∂x ∂y (13.7.3) Moment equilibrium about the x axis, gives, −M y dx − ∂ My ∂ Mxy ∂ Qy dydx − Mxy dy − dydx + Q y dydx + dydxdy ∂y ∂x ∂y + qdxdy dy dy ∂ Q x dy dy + Q x dy + dydx + M y dx + Mxy dy − Q x dy =0 2 2 ∂x 2 2 (13.7.4) ∂ Mxy ∂ My + − Qy = 0 ∂y ∂x Similarly, moment equilibrium about the y axis, provides, ∂ Mxy ∂ Mx + − Qx = 0 ∂x ∂y (13.7.5) Using the last two equations in the third force equation (z-direction) of equilibrium, we obtain, ∂ ∂x  ∂ Mxy ∂ Mx + ∂x ∂y ∂ + ∂y  ∂ Mxy ∂ My + ∂y ∂x +q =0 ∂ 2 My ∂ 2 Mx ∂ 2 Mxy + + +q =0 ∂x2 ∂ x∂ y ∂ y2 (13.7.6) This last equation and the first two equations are in a form that is convenient for solving laminated plate bending problems. These three equations can be cast in terms of centerline displacements, u 0 , v0 , and w, by using the strain-displacement equations, Equation 13.2.7, and the laminated plate constitutive equations, Equation 13.5.1, to arrive at, A11 ∂ 2u0 ∂ 2u0 ∂ 2 v0 ∂ 2 v0 ∂ 2u0 + 2A + A + A + A 16 66 12 16 ∂x2 ∂ x∂ y ∂ x∂ y ∂ y2 ∂x2 + A66 ∂ 2 v0 ∂ 2 v0 ∂ 3w ∂ 3w + A26 2 − B11 3 − 3B16 2 ∂ x∂ y ∂y ∂x ∂x ∂y --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- − (B12 + 2B66 ) A16 ∂ 3w ∂ 3w − B26 3 = 0 2 ∂ x∂ y ∂y ∂ 2u0 ∂ 2u0 ∂ 2u0 ∂ 2 v0 ∂ 2 v0 (A ) + A + + A + A + 2A 26 12 66 66 26 ∂x2 ∂ x∂ y ∂ y2 ∂x2 ∂ x∂ y Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (13.7.7) ∂ 3w ∂ 3w ∂ 3w ∂ 3w ∂ 2 v0 + A22 2 + B16 3 − (B12 + 2B66 ) 2 − 3B26 − B22 3 = 0 2 ∂y ∂x ∂x ∂y ∂ x∂ y ∂y (13.7.8) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 495 Analysis of Thin Laminated Composite Material Structures D11 ∂ 4w ∂ 4w ∂ 4w ∂ 4w ∂ 4w (D ) + 2 + 4D + 2D + 4D + D 16 12 66 26 22 ∂x4 ∂ x 3∂ y ∂ x 2∂ y2 ∂ x∂ y 3 ∂ y4 − B11 ∂ 3u0 ∂ 3u0 ∂ 3u0 ∂ 3u0 ∂ 3 v0 (B ) − + 2B − 3B − B − B 12 66 16 26 16 ∂x3 ∂ x 2∂ y ∂ x∂ y 2 ∂ y3 ∂x3 − (B12 + 2B66 ) (13.7.9) ∂ 3 v0 ∂ 3 v0 ∂ 3 v0 − 3B26 − B22 3 = q (x, y) 2 2 ∂x ∂y ∂ x∂ y ∂y We will now demonstrate how these equations can be used to solve laminates subjected to transverse loads. ########### Example 13.7.1 Consider a rectangular, symmetric, cross-ply laminate that is free from any in-plane forces and moments on its edges and, in fact, is free to move in the x and y directions on all edges, but is prevented from out of plane motions at the edges. Suppose such a laminate is subjected to uniform transverse pressure loads. Assume the origin of the xyz coordinate system is at the geometric center of the plate. a q0 = const b x y Figure (a) z Cross-ply laminate under transverse pressure loads Since the plate is symmetric and of cross-ply construction Cross − ply → Symmetric → A16 = A26 = D16 = D26 = 0 Bij = 0 for all i and j As a result, the governing equations simplify to, ∂ 2 v0 ∂ 2u0 ∂ 2u0 =0 + A66 2 + (A11 + A66 ) 2 ∂x ∂y ∂ x∂ y ∂ 2u0 ∂ 2 v0 ∂ 2 v0 (A12 + A66 ) + A66 2 + A22 2 = 0 ∂ x∂ y ∂x ∂y ∂ 4w ∂ 4w ∂ 4w D11 4 + 2 (D12 + 2D66 ) 2 2 + D22 4 = q0 ∂x ∂x ∂y ∂y A11 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas 496 July 6, 2011 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods with q0 being the magnitude of the uniform load. The boundary conditions are as follows: a → 2 ∂v0 ∂u 0 + A12 =0 N x = A11 ∂x ∂y  ∂v0 ∂u 0 Nxy = A66 + =0 ∂y ∂x ∂ 2w ∂ 2w Mx = −D11 2 − D12 2 = 0 ∂x ∂y w= 0 (b) b → 2 ∂u 0 ∂v0 + A22 =0 N y = A12 ∂x ∂y  ∂v0 ∂u 0 + =0 Nxy = A66 ∂y ∂x ∂ 2w ∂ 2w M y = −D12 2 − D22 2 = 0 ∂x ∂y w= 0 (c) At x = ± At y = ± --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Sometimes the form of the solution can be determined by inspection. For this problem a solution of the form u 0 (x, y) = 0 v0 (x, y) = 0 ∞ ∞ Wmn cos w (x, y) = m=1,3,5,... n=1,3,5,...  mπ x  a cos  nπ y  (d) b satisfies governing equations, Equation (a), and all the boundary conditions at each edge, Equations (b) and (c). The Wmn can be determined by substituting the form for w (x, y) into the third differential equation, expanding the uniform load in the same double cosine series and equating coefficients. For any cross-ply laminate with these same boundary conditions, the following procedure is valid if the load can be expanded in terms of a double cosine Fourier series. The double cosine Fourier series representation of the uniform load is ∞ ∞ Q mn cos q0 (x, y) = m=1,3,5,... n=1,3,5,...  mπ x  a cos  nπ y  b (e) with Q mn = 16q0 m+n (−1) 2 −1 mnπ 2 (f) Substituting the series expressions for w (x, y) and q0 (x, y) into the third differential equation, and equating the like terms in m and n leads to   mπ 4  nπ 4  mπ 2  nπ 2 D11 Wmn = Q mn + 2 (D12 + 2D66 ) + D22 (g) a a b b Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 13:22 Printer Name: Yet to Come 497 Analysis of Thin Laminated Composite Material Structures This results in or Wmn =  Wmn D11  mπ 4 a Q mn  nπ 4  mπ 2  nπ 2 + 2 (D12 + 2D66 ) + D22 a b b (h) 16q0 m+n (−1) 2 −1 2 mnπ =   nπ 4  mπ 4  mπ 2  nπ 2 D11 + 2 (D12 + 2D66 ) + D22 a a b b (i) With Wmn known, the curvatures are ∞ ∞ ∞ ∞ χx = −  nπ y   mπ 2  mπ x  ∂ 2w cos Wmn = cos 2 ∂x a a b m=1,3,5,... n=1,3,5,... χy = −  nπ y   nπ 2  mπ x  ∂ 2w cos Wmn = cos 2 ∂y b a b m=1,3,5,... n=1,3,5,... ∞ ∞ χxy = −2 (j)  mπ x   nπ y   mπ   nπ  ∂ 2w = −2 sin sin Wmn ∂ x∂ y a b a b m=1,3,5,... n=1,3,5,... The moment resultants are Mx = D11 χx + D12 χ y ∞ ∞ = m=1,3,5,... n=1,3,5,...  D11  mπ 2 + D12  nπ 2 Wmn cos  mπ x  cos  nπ y   D12  mπ 2 + D22  nπ 2 Wmn cos  mπ x  cos  nπ y   mπ x  cos M y = D12 χx + D22 χ y ∞ ∞ = m=1,3,5,... n=1,3,5,... Mxy = D66 χxy ∞ a ∞ = −2D66 m=1,3,5,... n=1,3,5,... a b  mπ   nπ  a b b Wmn cos a a a b  nπ y  b The transverse shear force resultants are ∂ Mxy ∂ Mx + Qx = ∂x ∂y  ∞ ∞  mπ 3  nπ 2  mπ   nπ 2  mπ  + 2D66 D11 + D12 =− a b a b a m=1,3,5,... n=1,3,5,...  nπ y   mπ x  cos ×Wmn sin a b ∂ Mxy ∂ My Qy = + ∂y ∂x  ∞ ∞  nπ   mπ 2  nπ   mπ 2  nπ 3 D22 + D12 + 2D66 =− b b a b a m=1,3,5,... n=1,3,5,...  nπ y   mπ x  sin ×Wmn cos a b --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (k) b Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (l) P1: TIX/XYZ JWST071-13 P2: ABC JWST071-Waas July 6, 2011 498 13:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods However, these are not the effective transverse shear force resultants discussed earlier, in Chapter 12 (note that Vx and Vy are the effective or Kirchhoff shear resultants). The effective shear force resultants consist of a combination of the transverse shear resultants, Q x and Q y , and the gradients along each plate edge, of the twisting moments, Mxy and M yx , respectively. That is, with respect to Figure 13.7.1, along the edge ∂M where Q x , acts, the effective transverse Kirchhoff shear force resultant, Vx , is Vx = Q x + ∂ yxy , and along ∂M the edge where Q y , acts, the effective transverse Kirchhoff shear force resultant, Vy , is Vy = Q y + ∂ xyx . More details of the derivation of Vx and Vy are given by Timoshenko and Woinowsky-Krieger (1987). The strains at any point x, y, z within the laminate are ∞ ∞  nπ y   mπ 2  mπ x  cos Wmn cos εx = zχx = z a a b m=1,3,5,... n=1,3,5,... ∞ ∞ Wmn ε y = zχ y = z m=1,3,5,... n=1,3,5,... ∞ ∞  nπ 2 b Wmn γxy = zχxy = −2z m=1,3,5,... n=1,3,5,... cos  mπ x  a  mπ   nπ  a b cos sin  nπ y  (m) b  mπ x  a sin  nπ y  b Since the plate deflection, the internal stress and moment resultants and the strains within the laminate are related to the external load, one can find the value of the external load that will cause any of these quantities to be excessive. More examples of composite laminates under different loading conditions are provided in the textbook by Kollar and Springer (2003). The textbooks by Hyer (1996) and Herakovich (1998) are recommended for further reading. Summary and Conclusion Our purpose in this chapter is to provide an introduction to analyze thin fiber reinforced laminates that are increasingly used in the Aerospace industry. The simplest theory to analyze panels, classical lamination thin plate theory (CLT), has been introduced and some example problems have been done to illustrate the solution process of the displacement equations of equilibrium. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 13.8 P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 14 Buckling 14.1 Introduction We have assumed up to now that the differential equations for the bending displacement of slender bars subjected to lateral loads are uncoupled from the effects of axial loads. We have made a similar assumption for the effects of lateral and in plane loads in thin plates. This is not always the case. When slender bar and thin plate structures are subjected to low levels of axial and in plane stresses they behave as described by the various equations considered in the previous chapters. When the axial or in plane compressive loads and therefore the compressive stresses become large enough the bending displacements are substantially greater for any given set of lateral loads. What this means is that in an infinitesimal element of the structure, considered in the deformed configuration, the in-plane loads can produce out-of-plane load resultants that are significant and of a magnitude that is non-negligible. This leads to another phenomenon referred to as geometric instability or buckling. This can cause a catastrophic failure at levels of stress well below the yield stress of the material. We shall study buckling of bars and thin walled structures in this chapter. 14.2 The Equations for a Beam with Combined Lateral and Axial Loading Consider a beam with both lateral and axial loads as shown in Figure 14.2.1. The loading may include concentrated forces and concentrated moments as well. You will note that in the previous chapters the distributed forces appear in the governing equations while any concentrated forces are included via boundary conditions. In the following development the effect of axial concentrated forces as well as axial distributed forces will appear in the differential equation for the lateral displacement in bending. For this example the loads act only in the xy plane. The cross sectional area is oriented so that Ix y = 0. y x Figure 14.2.1 Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 500 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Referring back to the derivation of the displacement equations for the axially loaded bar in Chapter 4 and the bending of beams in Chapter 7 we can combine the two effects into one set of equations for bending. Let us label the axial displacement of the beam centroidal axis as u A (x) to distinguish it from the total displacement in the x direction which includes the bending component. Keeping v (x) as the label for the lateral displacement in bending under combined loading we can say that the total displacement in the x direction is u (x, y) = u A − y dv dx (14.2.1) where it is understood that u A = u A (x) and v = v (x). From the strain displacement equations we get for moderately large rotations and small strains,     du A d 2v 1 dv 2 du du A d 2v 1 dv 2 = −y 2 + + = −y 2 (14.2.2) εx (x, y) = dx dx dx 2 dx dx 2 dx dx Notice that we have now included extra terms (quadratic terms) in Equation 14.2.2 that were neglected before. This is because when the bending deflections are significant, the magnitude of the quadratic   1 dv 2 term, , is comparable to the other terms in the expression for εx (x, y) in Equation 14.2.2. From 2 dx Hooke’s Law we have    d 2v 1 dv 2 du A − Ey 2 + σx (x, y) = Eεx = E (14.2.3) dx 2 dx dx Now we must establish equilibrium. Here we show the circumstances under which the axial forces produce significant out-of-plane force resultants. The bar will displace in both the x and and y directions as shown in Figure 14.2.2. fy(x) y F dx fx (x) x M Figure 14.2.2 If we look at the equilibrium of a thin slice of this beam of width dx, that was originally in the undeformed state, then, due to the deformation, the slice gets rotated in the deflected position, as shown in Figure 14.2.3. We have defined the internal axial force by the symbol N. N receives contributions from both the distributed and concentrated axial applied loads. In previous chapters we used the symbol P but now we want to reserve the symbol P for a compressive axial force to be consistent with the most common notation used in other texts. From equilibrium in the x direction we get,  Fx = 0 → N (x + d x) cos(θ + dθ) − N (x) cos (θ) + V (x) sin (θ) (14.2.4) − V (x + d x) sin (θ + dθ ) + f x (x) d x = 0 Where, we have used the Taylor expansion to express θ (x + d x) ≈ θ(x) + θ ′ (x)d x = θ + dθ. Let θ be small, so that cosθ ≈ 1, and sinθ ≈ θ. These approximations specify the smallness of θ . Note that θ (x) = dv (x) dx (14.2.5) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 501 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Buckling fy(x) y V(x+dx) M(x) M(x+dx) N(x+dx) fx(x) θ (x+dx) N(x) θ (x) V(x) x Figure 14.2.3 With these approximations, Equation 14.2.4 reduces to N (x + dx) − N (x) + V (x) θ − V (x + dx) (θ + dθ) = f (x) = 0 (14.2.6) This further simplifies to d (V θ) dN + f x (x) − =0 dx dx (14.2.7) Now, compared to N (x), V (x) θ is a second order quantity. Therefore, (14.2.7) finally simplifies to, dN = − f x (x) dx (14.2.8) This is the same as Equations 4.3.6 in Chapter 4 where only axial loads in the x direction are applied. From equilibrium in the y direction we get  Fy = 0 → V (x + dx) cos(θ + dθ) − V (x) cos (θ) − N (x) sin (θ) + N (x + dx) sin (θ + dθ) + f y (x) dx = 0 Using the same simplifications as in the axial case we get dV d (N θ ) + + f y (x) = 0 dx dx (14.2.9) This is similar to the first of Equations 7.3.7 in Chapter 7, where only lateral loads in the y direction are applied, but now notice that because we satisfy equilibrium in the deformed configuration, the axial load produces a vertical resultant (the second term in Equation 14.2.9), which would not be there if we considered equilibrium in the undeformed configuration. Finally, from summation of moments about the z axis and at the left hand face we get  Mz = 0 → [V (x + dx) cos(θ + dθ)] dx + [N (x + dx) sin (θ + dθ)] dx dv + [V (x + dx) sin (θ + dθ ) − N (x + dx) cos (θ + dθ)] dx (14.2.10) dx   dx =0 + M (x + dx) − M (x) + f y dx 2 When simplified this becomes dM = −V (x) dx (14.2.11) Unlike in the beam theory presented in Chapter 7 we will now need a new expression for the vertical force resultant on a cross section because the internal resultant forces N and V are now rotated. Such Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 502 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods an expression can be easily obtained by considering the forces marked in Figure 14.2.3. If we call the vertical (in the y direction) force resultant V̂ (x), then V̂ = V (x) cos θ + N (x) sin θ = V (x) + N (x) θ = V + N dv dx (14.2.12) This term did not appear in the derivations in Chapter 7 because V̂ and V are identical in the case where we consider equilibrium in the undeformed configuration. Equations 14.2.8-9 and 14.2.11 are the three equations of equilibrium that we will deal with in this chapter. Since the moment is a resultant of the stress, integrated over the beam cross-section area, it can be given in terms of the lateral displacement. Using Equation 14.2.3 we get      d 2v 1 dv 2 du A − E y 2 ydA + E M (x) = − σx ydA = − dx 2 dx dx A A (14.2.13) 2 2 d v d v 2 = E 2 y dA = EIzz 2 dx A dx and the axial force resultant in terms of displacement is      d 2v 1 dv 2 du A − E y 2 dA + E N (x) = σx dA = dx 2 dx dx A A   2  1 dv du A = EA + dx 2 dx (14.2.14) The uniform axial stress across the cross section does not contribute to the moment, and the linear stress through the thickness does not contribute to the axial force resultant.       1 dv 2 1 dv 2 du A du A ydA = E + + ydA = 0 → ydA = 0 (14.2.15) E dx 2 dx dx 2 dx A A A Combining Equations 14.2.11-13 we get V (x) = − d d 2v dM = − EIzz 2 dx dx dx → V̂ (x) = − dv d d 2v EIzz 2 + N dx dx dx By combining Equation 14.2.9 with Equation 14.2.11 we have   dv d2 M d N + f y (x) = 0 − 2 + dx dx dx     dv d2 d d 2v N = f y (x) → 2 EIzz 2 − dx dx dx dx (14.2.16) (14.2.17) This shows that the internal axial force does affect the lateral displacement. In the statically determinate case N is found directly by applying equilibrium. For statically indeterminate cases the second order equation for axial displacement must be solved simultaneously with Equation 14.2.17 since the equations for u A (x) and v (x) are coupled. That is, solve       1 dv 2 1 dv 2 du A d du A = − f x (x) → N (x) = EA (14.2.18) EA + + dx dx 2 dx dx 2 dx --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- When EIzz is constant Equation 14.2.17 may be written as,     dv dv f y (x) d d 4v 1 d d 4v N = f y (x) → N = EIzz 4 − − 4 dx dx EI dx dx EIzz dx dx zz Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (14.2.19) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 503 Buckling Further, when the axial distributed load, f x (x) = 0, then, from the axial equilibrium equation, it follows that N (x) = constant = N , say, which further reduces (14.2.19) to EIzz d 2v d 4v − N 2 = f y (x) 4 dx dx → d 4v N d 2v f y (x) − = 4 EIzz dx2 EIzz dx (14.2.20) In summary Mc , Fc → f y (x) N (x) V (x) M(x) σx (x, y)   d dv dM dv + = −V (x) N + f y (x) = 0 dx dx dx dx dN = − f x (x) dx εx (x, y) u(x, y) u (x, y) = u A − y εx (x, y) = 1 du A + dx 2   d 2v 1 dv 2 du A − Ey 2 + σx (x, y) = Eεx = E dx 2 dx dx   2  1 dv du A + N (x) = σx dA = EA dx 2 dx A  σx ydA = E M(x) = − d 2v dx2 y 2 dA = EIzz d 2v dx2 v(x) ← ρ  dv dx 2 −y dv dx d 2v dx2 (14.2.21) → σx = M(x)y N (x) − A Izz A A 2 d dM d v = EIzz 2 ; dx dx dx d dv d 2v EIzz 2 + N → dx dx dx   dv d d2 d 2v N = f y (x) EI − zz dx dx dx2 dx2 V̂ (x) = − τx y (x, y) = V̂ Izz h ydA Ay The summary in Equation 14.2.21 looks much the same as Equation 7.3.17 in Chapter 7 but with the changes in force equilibrium. This results in an additional term in the definition of shear force (vertical force resultant), V̂ (x), Equation 14.2.16, and in the final equation for displacements, Equation 14.2.17. Note that the distributed applied lateral load, f y (x), is explicit in the displacement equation but the distributed applied axial loads, f x (x) are implicit in the value of N. Furthermore, concentrated applied lateral loads appear in the boundary conditions while concentrated applied axial loads are also implicit in the value of N. For the next few examples we will limit discussion to applied axial forces that are not functions of x, that is, f x (x) = 0; however, concentrated axial forces do exist. Distributed applied axial forces would result in distributed internal axial forces that are function of x, that is, N (x). This would create differential equations with non constant coefficients and this presents exceptional difficulty in attempting solutions. For the same reason we shall consider only cases where EIzz is constant. We deal with situations in which N(x) and/or EIzz (x) by other means later in this chapter. In those cases where distributed axial forces are absent (but concentrated forces can still be present) and when EIzz is constant, we find that the equations of equilibrium can be uncoupled and the solution becomes much easier, because the internal axial force, N, is constant. This is easily seen if we put f x (x) = 0, in Equation 14.2.8, then the axial equilibrium equation admits the solution, N = constant, and using this fact in Equation 14.2.19, with EIzz constant, we have an ordinary differential equation with constant coefficients. N d 2v f y (x) d 4v − = 4 2 EIzz dx EIzz dx (14.2.22) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 504 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods This has a well known solution. Of special interest is the case where the lateral load, f y (x), is zero and the axial load is compressive. Such a structure is commonly called a column. We consider this in the next section. 14.3 Buckling of a Column Consider the case of a uniform beam with no lateral load but a constant axial compressive load that produces a constant internal force N = −P. Equation 14.2.22 becomes P d 2v d 4v + =0 4 EIzz dx2 dx (14.3.1) For convenience let λ2 = P EIzz (14.3.2) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- then d 4v d 2v + λ2 2 = 0 4 dx dx (14.3.3) This equation has a very interesting solution. Either v (x) = 0 or λ takes on a series of discrete non zero values for which a solution other than v(x) = 0 is possible. In the study of differential equations this is called an eigenvalue equation. The discrete values of λ imply discrete values of P. The lowest non zero value of P is called the buckling load. Furthermore for each discrete value of λ there is a non zero solution for v(x) which is called the eigenvector or mode shape. We shall now develop this solution. Ordinary differential equations with constant coefficients have solutions of the form v = ceαx (14.3.4) If we substitute this into Equation 14.3.3 we get α 4 + λ2 α 2 ceαx = 0 → α 4 + λ2 α 2 = 0 (14.3.5) This equation has four roots as follows α1 = iλ √ α3 = 0 α2 = −iλ α4 = 0 (14.3.6) where i = −1. The solution to Equation 14.3.3 is v(x) = aeiλx + be−iλx + cx + d (14.3.7) or since eiλx = cos λx + i sin λx e−iλx = cos λx − i sin λx this may be rewritten as v(x) = aeiλx + be−iλx + cx + d = a (cos λx + i sin λx) + b (cos λx − i sin λx) + cx + d = (a + b) cos λx + i (a − b) sin λx + cx + d (14.3.8) = A sin λx + B cos λx + Cx + D The values of these constants of integration are obtained from the boundary conditions. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 505 Buckling For a typical column with two boundary conditions at each end the result is a set of four homogeneous algebraic equations containing the four constants of integration. These can be written in matrix form where the coefficients of the constants of integrations are just shown as a 4 × 4 matrix. ⎡ ⎤ ⎡ ⎤ A 0 ⎢ B ⎥ ⎢0⎥ ⎥ ⎢ ⎥ [4x4] ⎢ ⎣C ⎦ = ⎣0⎦ 0 D (14.3.9) One possible solution is that ⎡ ⎤ ⎡ ⎤ A 0 ⎢ B ⎥ ⎢0⎥ ⎢ ⎥=⎢ ⎥ ⎣C ⎦ ⎣0⎦ D 0 (14.3.10) but this means there is no lateral displacement. It so happens that there is another possibility under certain circumstances. If the determinant of the coefficient matrix is zero then it is possible that the constants of integration are not all zero and a lateral displacement occurs, that is, if |[4x4]| = 0 (14.3.11) If we set the determinant to zero we can expand it to obtain an equation in terms of the unknown λ. The roots of this equation have interesting meaning, namely, the values of the applied axial compressive load for which a solution other than no displacement is possible. As noted above problems of this type are called eigenvalue problems. The roots are called eigenvalues and the corresponding loads are called critical buckling loads. We shall find that three of the four constants of integration can be found but a fourth one is always indeterminate. Thus we can define the shape of the displacement but not its amplitude. This is best explored by examples. ########### Example 14.3.1 Problem: A uniform column has both ends pinned, or simply supported, and a constant axial compressive load as shown in Figure (a). Find the buckling load and mode. F L Figure (a) Solution: Insert Equation 14.3.8 into the four boundary conditions to obtain the four homogenius equations. Solve for the lowest buckling load and its mode shape. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 506 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The four required boundary conditions are v(0) = 0 M(0) = EIzz d 2 v(0) =0 dx2 v(L) = 0 M(L) = EIzz (a) d 2 v(L) =0 dx2 Substitute the solution, Equation 14.3.8, into each of the boundary conditions and you get v(0) = A sin λ0 + B cos λ0 + C · 0 + D = B + D = 0 (b) 2 d v(0) = −Aλ2 sin λ0 − Bλ2 cos λ0 = −Bλ2 = 0 dx2 v(L) = A sin λL + B cos λL + C L + D = 0 (c) (d) 2 d v(L) = −Aλ2 sin λL − Bλ2 cos λL = 0 dx2 (e) In matrix form ⎡ 0 ⎢ 0 ⎢ ⎣ sin λL −λ2 sin λL 1 −λ2 cos λL −λ2 cos λL 0 0 L 0 ⎤⎡ ⎤ ⎡ ⎤ A 0 1 ⎢ ⎥ ⎢ ⎥ 0⎥ ⎥⎢ B ⎥ = ⎢0⎥ 1⎦⎣C ⎦ ⎣0⎦ 0 D 0 (f) so we examine under what conditions a non zero displacement is possible by setting the determinant of the coefficient matrix equal to zero.    0 1 0 1    0 −λ2 0 0   =0 (g)  sin λL cos λL L 1    −λ2 sin λL −λ2 cos λL 0 0  A sin λL + CL = 0 A sin λL = 0 (h) If A = 0 then C = 0 and there would be no displacement; but there is another possibility, namely, that sin λL = 0. If this is true then C = 0; however, A is indeterminate. Under what circumstances is sin λL = 0? When  nπ 2 EI (i) λL = nπ → P = L where n is an integer. This suggests that there are an infinite number of discrete values of P, one for each integer value of n, for which a solution other than A = B = C = D = 0 exists. We did determine that B = C = D = 0 therefore the general solution of the equation for these boundary conditions is v(x) = A sin nπ x L (j) We know the shape of the deflection but the amplitude A is indeterminate. This shape is called the mode shape. We simply do not have enough information to determine the amplitude of the displacement. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- We can expand the determinant; however, in this case you can simplify it by noting that from Equation (c) you get B = 0; then from the Equation (b) you get D = 0. The Equations (d) and (e) reduce to P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 507 Buckling In practice the lowest positive non zero value, that is, n = 1, is of most interest. As a load is slowly applied once it reaches the lowest value of the buckling load the damage has been done. So we usually just say that the critical buckling load and the corresponding critical buckling stress are Pcr = π2 EI L2 σcr = − π2 P =− EI A AL 2 (k) Experiments bear this out. Real columns will deflect catastrophically when this load is reached. The mode shape is a simple half sine wave. v(x) = A sin π x L (l) and is plotted in Figure (b). We arbitrarily assign the value A = 1 for plotting purposes. 1.0 0.8 0.6 A 0.4 0.2 0.2 0.4 0.6 0.8 1.0 x/L Figure (b) Example 14.3.2 Problem: Pin jointed trusses are made up of an assembly of slender bars simply supported at the ends. Those members in compression may be candidates for buckling. Consider the truss in Example 4.10.1. What value of F would cause one of the members of the truss to buckle? The truss is shown here in Figure (a). The cross section is a 20 mm × 20 mm square. A1 = A2 = 400 mm 2 F = 10000 N 1000 mm E = 206840 MPa. 1000 mm A1 A2 1000 mm 30 deg F Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-14 P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 508 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Solution: Increase the load until a member buckles. We know from Example 4.10.1 that member 2 is in compression. The critical buckling load in member 2 would be  π 2 π2 204 = 13611N (b) · 206800 · Pcr = 2 EI = L 1414 12 The corresponding compressive stress, called the critical stress, σcr , is Pcr = −34.03 MPa A The actual stress found in Example 4.10.1 is (c) σcr = − (d) σ2 = −6.471 MPa This is well below the critical stress. This is a linear problem, that is, the internal forces are proportional to the external loads. If F is increased by a factor of 34.03 = 5.259 or 6.471 F = 52588N (e) the member will buckle. Example 14.3.3 Problem: Consider a column that is fixed on both ends and loaded with an axial force as shown in Figure (a). We have to be careful here and allow the axial displacement. The right end is fixed in rotation and lateral deflection but axial displacement is permitted. Find the buckling load and mode. y F x Figure (a) Solution: Insert Equation 14.3.8 into the four boundary conditions to obtain the four homogenius equations. Set the determinant of the coefficient matrix to zero and solve for the critical buckling load and mode. The boundary conditions are dv(L) dv(0) =0 v(L) = 0 =0 dx dx Substitute the solution into each of the boundary condtions and you get v(0) = 0 v(0) = A sin λ0 + B cos λ0 + C · 0 + D = B + D = 0 dv(0) = Aλ cos λ0 − Bλ sin λ0 + C = Aλ + C = 0 dx v(L) = A sin λL + B cos λL + CL + D = 0 dv(0) = Aλ cos λL − Bλ sin λL + C = 0 dx --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) (b) (c) (d) (e) P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 509 Buckling This can be presented as a set of four equations in four unknowns in matrix equation form. ⎡ 0 ⎢ λ ⎢ ⎣ sin λL λ cos λL 1 0 cos λL −λ sin λL 0 1 L 1 ⎤⎡ ⎤ ⎡ ⎤ 0 A 1 ⎢ ⎥ ⎢ ⎥ 0⎥ ⎥⎢ B ⎥ = ⎢0⎥ 1⎦⎣C ⎦ ⎣0⎦ 0 D 0 (f) Again an obvious solution is ⎡ ⎤ ⎡ ⎤ A 0 ⎢ B ⎥ ⎢0⎥ ⎢ ⎥=⎢ ⎥ ⎣C ⎦ ⎣0⎦ D 0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Another possible solution is that   0   λ   sin λL   λ cos λL 1 0 cos λL −λ sin λL 0 1 L 1 (g)  1  0  =0 1  0 (h) We can simplify the equations and reduce this to a 2 × 2 determinant by noting the D = −B and C = −Aλ from Equations (b) and (c). Equations (d) and (e) can then be written A (sin λL − λL) + B (cos λL − 1) = 0 (i) Aλ (cos λL − 1) − Bλ sin λL = 0 (j) and the determinant becomes   sin λL − λL   cos λL − 1 If we expand the determinant we get  cos λL − 1  =0 sin λL  (k) λL sin λL + 2(cos λL − 1) = 0 (l) By inspection the lowest positive non zero root is found to be λL = 2π → Pcr =  2π L 2 (m) EI Substituting λL = 2π back into the equations we get A (sin 2π − 2π ) + B (cos 2π − 1) = 0 → −2π A + B · 0 = 0 Aλ (cos 2π − 1) − Bλ sin 2π = 0 → 0 = 0 → A=0 (n) If A = 0 then from Equation (c) C = 0 and since D = −B from Equation (b) the mode shape is found to be   2π x −1 (o) v(x) = B cos L Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 510 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The mode shape is shown in Figure (b) with a value B = −0.5 assigned for plotting purposes. This plots the maximum amplitude as 1.0. Remember that the actual amplitude is indeterminate. 1.0 0.8 B 0.6 0.4 0.2 0.2 0.4 0.6 0.8 1.0 x/L Figure (b) Example 14.3.4 Problem: Now consider a column fixed on one end and free on the other as shown in Figure (a). Find the buckling load and mode. y F x Figure (a) Solution: Insert Equation 14.3.8 into the four boundary conditions to obtain the four homogenius equations. Expand the determinant of the coefficients and solve for the lowest buckling load and mode. With one exception the boundary conditions are the same as discussed in Chapter 7 for a cantilever beam. We specify the displacement and slope on the left end and the moment and shear force on the right end. The exception is that the shear force applied at the right end is now given by EIzz dv d 3v = −V +P dx dx3 (a) The boundary conditions are v(0) = 0 dv(0) =0 dx (b) d 2 v(L) =0 dx2 dv(L) d 3 v(L) =0 +P −V (L) = EIzz dx dx3 M(L) = EIzz --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 511 Buckling Substituting we get v(0) = A sin λ0 + B cos λ0 + C · 0 + D = B + D = 0 (c) dv(0) = Aλ cos λ0 − Bλ sin λ0 + C = Aλ + C = 0 dx d 2 v(L) = −Aλ2 sin λL − Bλ2 cos λL = 0 dx2 dv(L) d 3 v(L) + λ2 dx dx3 = −Aλ3 cos λL + Bλ3 sin λL + λ2 (Aλ cos λL − Bλ sin λL + C) (d) (e) (f) = Cλ2 = 0 We can set up the determinant of the coefficients or in this case note that C = 0 from Equation (f) and therefore A = 0 from Equation (d). We are left with Bλ2 cos λL = 0 B+D=0 (g) from Equations (c) and (e). Either B = 0 and thus D = 0 or cos λL = 0. The condition for cos λL = 0 to be true is for λL = (2n + 1)π 2 n = 0, 1, 2, . . . (h) In this case the lowest critical buckling load is for n = 0, or Pcr = The mode shape is v(x) = B cos  π 2 EIzz 2L (i)   πx πx + D = B cos −1 2L 2L (j) 1.0 0.8 0.6 0.4 0.2 0.2 0.4 0.6 0.8 1.0 Figure (b) ########### We can go on giving examples, including multiple compressive loads at different points along the beam and variable cross sections. As the loading gets more complicated there is a better way. This will be shown in Section 14.5. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The mode shape is shown in Figure (b) with a value B = −1 assigned for plotting purposes. P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 512 14.4 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The Beam Column When both axial and lateral loads are present and the compressive axial load is well below the buckling load the axial and bending effects are nearly uncoupled. As the axial load approaches the buckling load we can expect the behavior to be quite different. This structure is sometimes called a beam column. The differential equation for the case with both a compressive axial load and a lateral load is f y (x) d 2v d 4v + λ2 2 = 4 EIzz dx dx (14.4.1) Let us examine what happens with a simple example. ########### Example 14.4.1 Problem: Find the displacement and stresses in the beam in Figure (a). Let the axial load vary from a small to a large value approaching the buckling load. FB y x FA L Figure (a) Solution: Solve Equation 14.2.22 for a range of values of the axial load. Adapt Equations 14.2.20 to this problem. N d 2v f y (x) d 4v − = 4 EI dx2 EI dx → d 4v d 2v + λ2 2 = 0 4 dx dx (a) The solution is given in Section 14.3, Equation 14.3.8. v(x) = A sin λx + B cos λx + Cx + D (b) The boundary conditions are v(0) = 0 dv(0) =0 dx M(L) = EI d 2 v(L) =0 dx2 FB d 3 v(L) dv(L) =− + λ2 −V (L) = dx EI dx3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 513 Buckling Substituting we get v(0) = A sin λ0 + B cos λ0 + C · 0 + D = B + D = 0 (d) dv(0) = Aλ cos λ0 − Bλ sin λ0 + C = Aλ + C = 0 dx d 2 v(L) = −Aλ2 sin λL − Bλ2 cos λL = 0 dx2 V (L) dv(L) d 3 v(L) =− + λ2 dx EI dx3 (e) = −Aλ3 cos λL + Bλ3 sin λL + λ2 (Aλ cos λL − Bλ sin λL + C) = Cλ2 = − (g) FB EI From Equation (g) we have C =− FB EI FB FB =− =− 2 λ EI PEI P (h) From the Equation (e) we have Aλ + C = Aλ − FB =0 P A= → FB λP (i) From the Equation (f ) we have A sin λL + B cos λL = B cos λL + FB sin λL = 0 λP → B=− FB tan λL λP (j) Finally from Equation (d) we have B+D=0 → D = −B = FB tan λL λP (k) And so the solution is v(x) = A sin λx + B cos λx + Cx + D = FB FB FB FB sin λx − tan λL cos λx − x+ tan λL λP λP P λP = FB FB FB sin λx + tan λL (1 − cos λx) − x λP λP P (l) The lateral tip deflection is v (L) = FB FB FB sin λL + tan λL (1 − cos λL) − L λP λP P FB FB FB FB sin λL + tan λL − sin λL − L λP λP λP P   FB 1 = tan λL − L P λ = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (m) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- (f) P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 514 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The value of λ and the buckling load when FB is zero are (see Example 14.3.4) λ= π 2L → Pcr =  π 2 EIzz 2L (n) At P = 0, that is, when there is no axial load, the value of the tip deflection as presented in Equation (m) is indeterminate. Through a limiting process you can find the value, but, of course, from Example 7.3.2, Equation (d) we know the deflection is v (x) = FB EI  x3 Lx2 − 2 6  → v (L) = FB L 3 3EI (o) So now let us plot the value of the tip deflection versus the value of λL from zero to the buckling load. For plotting purposes we format the tip deflection as follows. v (L) = FB P  1 tan λL − L λ  = FB FB L 3 (tan λL − λL) (tan λL − λL) = λP EI (λL)3 (p) Plot v(L) versus λL in Figure (b). It is quite clear that as the axial load approaches the buckling load the lateral displacement grows rapidly. Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- To complete the plot we ask what happens when the axial force is tension. Then the equation is d 2v d 4v − λ2 2 = 0 4 dx dx (q) The solution is straight forward following the same steps that we just presented. We shall not go through them here. Suffice it to say that the effect is a small reduction in the amplitude of the lateral displacement caused by the tensile axial force. The moment generated by the axial force times the lateral displacement is very small compared to the moment generated by the lateral force and its moment arm. The plot in Figure (b) has been extended downward to show this. ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 515 Buckling We can continue to consider other loading and boundary conditions as we have done for beams in Chapter 7. We can also extend the analysis to beams with different internal forces in different regions by writing equations for each region and matching boundary conditions where regions meet. This becomes quite time consuming and tedious so we will defer these problems to the next section. We can continue examples of beam column analysis to include more complicated cases including frames; however, the solutions become quite laborious. There is a better way which we consider in the next section. 14.5 The Finite Element Method for Bending and Buckling We shall extend the FEM for cases that include the effect of internal axial forces on the lateral bending of beams. First we show the beam column form of these equations and then we simplify for buckling analysis. When both axial and lateral loads are present the virtual work consists of the external work done by the loads minus the strain energy of bending. Again we look at the bar in the deflected position. Then δW = δWae + δWle + δW i = δW e − δU (14.5.1) where δWae is the external virtual work of the axial loads and δWle is the external virtual work of the lateral loads. Consider an element m with an axial compressive load and a set of lateral loads. The lateral bending displacement is represented by nodal displacements and a shape function. v(sm ) = [n] {rm } (14.5.2) When these loads are applied to an element m the strain energy of bending is precisely what we found before in Chapter 11, Equation 11.3.33. 2  EIm lm d 2 v 1 ds (εxm σxm )d V = Um = 2 V 2 0 ds2     EIm lm d2 d2 {rm }T 2 [n]T } {r ds (14.5.3) = [n] m 2 0 ds ds2 and the virtual strain energy is given in Equation 11.3.34.  2  lm lm  2 d 2 δv d 2 v d T d T {δr } } {r ds ds = EI δUm = EIm [n] [n] m m m ds2 ds2 ds2 ds2 0 0 (14.5.4) Note that we do not include the strain energy of the axial stress. It is negligible when compared to that of bending and so has little effect on the solution. Now consider a beam in the deflected position with an axial load as shown in Figure 14.5.1. n1 node n+1 node n F element m Figure 14.5.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 516 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The applied axial force F provides an internal axial force P on element m. To find the external virtual work consider an axial compressive load on element m. The work done by the axial force P is equal to the force times the distance it moves as a result of the lateral displacement. For example in Figure 14.5.2 we can identify the shortening of the distance between the two ends of an element due to the lateral displacement. The undeflected element is shown with a dashed outline. ξ s Figure 14.5.2 We isolate the element m and assign a coordinate ξ which follows the centerline of the curved beam and we retain coordinate s of the undeflected element. We express the shortening of the distance on a small slice of the element as du = ds − dξ (14.5.5)   2  dv 2 dξ = ds + dv 2 = ds 1 + ds (14.5.6) where For small values of the slope this can be approximated by expanding in a series and neglecting higher order terms.    1 dv 2 + higher order terms (14.5.7) dξ = ds 1 + 2 ds Thus  1 du = ds − dξ = ds − ds 1 + 2  dv ds 2  =− 1 2  dv ds 2 ds (14.5.8) The total axial displacement due to the lateral displacement for the element m is 1 u(lm ) = − 2 lm 0  dv ds 2 ds (14.5.9) The virtual displacement δu that results from a virtual displacement δv is then lm δu(lm ) = − 0 dδv dv ds ds ds (14.5.10) Since the axial force P is in the same direction as the displacement the external virtual work done by the axial force is lm e δWam = Pδu = P 0 dδv dv ds = {δrm }T P ds ds lm 0 d [n]T d [n] ds {rm } ds ds (14.5.11) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 517 Buckling The external virtual work done by the lateral forces is e = {δrm }T { pm } δWlm (14.5.12) The total virtual work is lm δW = {δrm }T { pm } + {δrm }T P 0 lm T − {δrm } EIm 0 d [n]T d [n] ds {rm } ds ds (14.5.13) T d 2 [n] d 2 [n] ds {rm } = 0 ds2 ds2 or lm { pm } = −P 0 d [n]T d [n] ds {rm } ds ds lm +EIm 0 (14.5.14) T d 2 [n] d 2 [n] ds {rm } = 0 ds2 ds2 Note again that we can neglect the strain energy contributed by the axial stress and strain because its small compared to the bending strain energy, and we can neglect the work done by compressive axial displacement, but we include the external work done by the axial force due to lateral displacement. This leaves us with   [km ] − P kmd {rm } = { pm } where lm [km ] = EIm 0 T d 2 [n] d 2 [n] {rm } ds ds2 ds2 lm   kmd = 0 (14.5.15) d [n]T d [n] ds ds ds (14.5.16) For a finite element solution we could consider shape functions based on the general homogenous solution of the fourth order differential equation. It has been found that a simpler approach, namely, using the same shape functions that were used for the beam with lateral loading only, is more convenient. It is easier to obtain a desired level of accuracy by using more elements than it is to fuss over the accuracy of each element. The shape functions for bending developed in Chapter 8 and summarized in Equation 10.2.30, repeated here, are v(sm ) = [n] {rm } 1 = 3 lm3 − 3lm sm2 + 2sm3 lm lm3 sm − 2lm2 sm2 + lm sm3 3lm sm2 − 2sm3 ⎡ ⎤ wn ⎢ θ ⎥ n ⎥ −lm2 sm2 + lm sm3 ⎢ ⎣ wn+1 ⎦ θn+1 (14.5.17) These are not the solution to the differential equation for buckling but they meet the criteria for kinematically admissible shape functions. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 518 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods As demonstrated in Chapter 11, Equations 11.3.38-41 we get for [km ] for element m lm [km ] = EI 0 ⎡ 12 6lm T 4lm2 E m Im ⎢ d 2 [n] d 2 [n] ⎢ 6lm ds = 3 ⎢ lm ⎣ −12 −6lm ds2 ds2 6lm 2lm2 −12 −6lm 12 −6lm ⎤ 6lm 2lm2 ⎥ ⎥ ⎥ −6lm ⎦ 2 4lm (14.5.18)   The matrix kmd has various names. It is called the differential stiffness matrix, or the geometric stiffness matrix, or the initial stress matrix, among others. Using the same shape functions we get lm   kmd = 0 ⎡ 36 ⎢ −3l 1 d d m ⎢ [n]T [n] ds = ⎢ ds ds 30lm ⎣ −36 −3lm −3lm 4lm2 3lm −lm2 ⎤ −36 −3lm 3lm −lm2 ⎥ ⎥ ⎥ 36 3lm ⎦ 3lm 4lm2 (14.5.19) For the element m we get 12 6lm 2 ⎜ E I ⎢ 6lm 4l  d m ⎜ m m⎢ [km ] − P km {rm } = ⎜ 3 ⎢ ⎝ lm ⎣ −12 −6lm 6lm 2lm2 ⎛ ⎡ ⎡ 36 −3l P ⎢ m ⎢ − ⎢ 30L m ⎣ −36 −3lm −3lm 4lm2 3lm −lm2 −12 −6lm 12 −6lm −36 3lm 36 3lm ⎤ 6lm 2lm2 ⎥ ⎥ ⎥ −6lm ⎦ 4lm2 (14.5.20) ⎤⎞ ⎡ ⎤ −3lm vn 2 ⎥⎟ ⎢ −lm ⎥⎟ ⎢ θn ⎥ ⎥ ⎥⎟ 3lm ⎦⎠ ⎣ vn+1 ⎦ θn+1 4lm2 Next we must assemble the matrices. [K ] = ! [km ]   ! d km Kd = (14.5.21) The global equations are   [K ] − P K d {rm } = {F} (14.5.22) So using the same shape functions from bending theory and assuming a constant EI we can summarize the process of finding the global matrix equations. The summary provided in Equation 14.5.22 is   here very similar to that in Chapter 8, Equation 8.4.5 with the addition of the kmd element matrix. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 519 Buckling In summary Vm (sm ) Pm Mzm (sm ) σsm (sm , y) lm Fem = f y (sm ) ds ! → 0 εsm (sm , y) u(sm , y) v (sm ) = [n (sm )] {rm } Fem u m (sm ) = −y F = Fc + Fe εsm (sm ) = −y σsm (sm ) = −y E lm δW = {δrm }T P lm 0 v(sm ) {rm } ← ρ d [n] {rm } ds d2 [n] {rm } ds2 d2 [n] {rm } ds2 d [n]T d [n] ds {rm } − {δrm }T EIm ds ds (14.5.23) lm 0 T d 2 [n] d 2 [n] ds {rm } = 0 ds2 ds2 T d [n]T d [n] ds {rm } ds ds 0 # −lm2 sm2 + lm sm3 ⎤ lm d 2 [n] d 2 [n] ds ds2 ds2  [km ] = EIm 0 " 1 l 3 − 3lm sm2 + 2sm3 lm3 m  kmd = P lm3 sm − 2lm2 sm2 + lm sm3 3lm sm2 − 2sm3 ⎛ ⎡ −12 6lm 12 6lm ⎜ E m Im ⎢ 6lm  d 4lm2 −6lm 2lm2 ⎥ ⎜ ⎢ ⎥ [km ] − P km {rm } = ⎝ 3 ⎣ −12 −6lm 12 −6lm ⎦ lm 6lm 2lm2 −6lm 4lm2 ⎤ ⎤⎞ ⎡ ⎡ vn 36 −3lm −36 −3lm 2 ⎥ ⎟⎢ P ⎢ 3lm −lm2 ⎥ ⎥⎟ ⎢ θn ⎥ ⎢ −3lm 4lm − ⎦ ⎣ ⎠ ⎦ ⎣ v −36 3l 36 3l 30L m n+1 m m θn+1 −3lm −lm2 3lm 4lm2 [K ] = ! [km ]  ! d km Kd =  d {r } = {F} [K ] − P K  We shall show how the element stiffness matrices are assembled to solve a buckling problem in an example. In such cases the applied lateral loads are zero. ########### Example 14.5.1 Problem: Reconsider Example 14.3.3. This time do it by FEA. y F x Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- {F} → f ym (sm ) P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 520 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Solution: Divide the column into elements and label the nodes and elements. Then assemble the elements stiffness and differential stiffness matrices, add boundary conditions, partition, and solve. Let us divide the column into two elements and let L 1 = L 2 = L2 = l as shown in Figure (b). e1 e2 n1 n2 n3 Figure (b) We assemble the element equations to obtain the global equations, Equation 14.5.22, add the boundary restraints, and partition. ⎡ ⎡ ⎤ ⎤⎞ ⎡ ⎤ ⎤ ⎛ ⎡ 0 12 −6l −12 −6l 0 36 −3l −36 −3l 0 R1 0 0 ⎢ −6l 4l 2 6l 2l 2 0 ⎢ −3l 4l 2 3l −l 2 0 ⎥ ⎥⎟ ⎢ 0 ⎥ ⎢ Q1 ⎥ ⎜ 0 0 ⎢ ⎢ ⎥ ⎥⎟ ⎢ ⎥ ⎥ ⎜ ⎢ ⎢ ⎥ ⎥⎟ ⎢ ⎥ ⎥ ⎜ ⎢ P ⎢ ⎢ −36 3l 72 0 −36 −3l ⎥⎟ ⎢ v2 ⎥ ⎢ 0 ⎥ ⎜ EIzz ⎢ −12 6l 24 0 −12 −6l ⎥ − = ⎥ ⎥ ⎜ 3 ⎢ ⎢ ⎢ 2 2 2 ⎥ 2 2 2 ⎥⎟ ⎢ 0 8l 6l 2l ⎥ 30l ⎢ −3l −l 0 8l 3l −l ⎥⎟ ⎢ θ2 ⎥ ⎢ 0 ⎥ ⎜ l ⎢ −6l 2l ⎢ ⎥ ⎥⎟ ⎢ ⎥ ⎥ ⎜ ⎢ ⎢ ⎣ 0 ⎣ R3 ⎦ ⎝ ⎣ 0 0 −12 6l 12 6l ⎦ 0 −36 3l 36 3l ⎦⎠ ⎣ 0 ⎦ Q3 0 0 −6l 2l 2 6l 4l 2 0 0 −3l −l 2 3l 4l 2 0 (a) This quickly reduces the problem to just two equations in two unknowns. " # " # " # # "   d P 72 0 EIzz 24 0 0 v2 {r } = = − [K ] − P K θ2 0 0 8l 2 l3 30l 0 8l 2 (b) Either v2 = θ2 = 0 or the determinant of the coefficients is zero. To simplify the notation let β= Pl 2 30EIzz (c) Then the determinant of the coefficients reduces to    24 − 72β  0  =0  0 8l 2 (1 − β)  (d) There are two possibilities: 24 − 72β = 0 1−β =0 → → β1 = β2 = 1 = Pl 2 1 = 3 30EIzz Pl 2 30EIzz → Pcr = 40EIzz L2 → P1 = P2 = 10EIzz 40EIzz = l2 L2 30EIzz 120EIzz = l2 L2 (e) The lowest value is the critical value. We can compare this with the exact answer in Example 14.3.3.  2 39.478EIzz 2π EIzz = Pcr = L L2 The error is 1.32%. This is quite a good approximation for such a small number of elements. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (f) (g) P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 521 Buckling The mode shape is obtained by substituting the critical value back into the equations.   1 (24 − 72β1 ) v2 = 24 − 72 v2 = 0 · v2 = 0 3  1−  1 2 θ 2 = θ2 = 0 3 3 (h) It follows that θ2 = 0 and v2 is indeterminate. The mode shape is --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- v(x) = [n 1 ] {r1 } = v(x) = [n 2 ] {r2 } =  1  2 8 3ls − 2s 3 v2 = 3 3 l L  1 3 8 l − 3ls 2 + 2s 3 v2 = 3 l3 L  " # 3L 2 x − 2x 3 v2 2 3L L3 − 8 2  x− L 2 2 0≤x ≤ L 2   L 3 v2 +2 x − 2 L ≤x≤L 2 (i) The mode shape from the analytical solution in Example 14.3.3 is   2π x −1 v(x) = B cos L (j) What we have done is approximate a mode shape that is exactly a cosine function with two piecewise cubic polynomials. Let us compare the exact and approximate shapes. Remember that mode shapes do not have a defined amplitude. The coefficients v2 in Equation (i) and B in Equation (j) are not determined. In order to compare we assign a value to each coefficient so that the maximum displacements, in this case at L/2, are the same amplitude. This is called normalizing. In this case we assign B=− 1 2 v2 = 1 (k) We shall evaluate the mode shapes at intervals of L/8 and present the results in Table (a) Table (a) x 0 L/8 L/4 3L/8 L/2 5L/8 3L/4 7L/8 L Equation (j) Equation (i) 0.000 0.146 0.500 0.854 1.000 0.854 0.500 0.146 0.000 0.000 0.156 0.500 0.844 1.000 0.844 0.500 0.156 0.000 This shows that even a crude approximation of just two elements provides excellent correlation. Maybe there is something to this finite element stuff after all. Both the exact analytical and the FEM approximate modes shapes are shown in Figure (c). They are so close a single line is shown. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 522 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 1.0 0.8 0.6 0.4 0.2 0.2 0.4 0.6 0.8 1.0 Figure (c) ########### In more complicated problems involving three or more equations it is helpful to use one of the numerical equation solvers. Some of the solvers can deal with the format we have here, that is   [K ] − P K d {r } = 0 or   [K ] {r } = P K d {r } (14.5.24) Other solvers often require that the set of equations be put in standard form for solving for the  −1   eigenvalues and eigenfunctions. That form requires the inversion of K d , that is, K d . Then  Kd −1   [K ] − P K d {r } = 0 →  Kd −1  [K ] − P [1] {r } = 0 (14.5.25) and the eigenvalue determinant becomes     d −1  [K ] − P [1] = 0  K (14.5.26) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- where [K ]−1 [K ] = [1], the unit matrix that has ones on the diagonal and zeros elsewhere. Since for large determinants the solver may use an iterative procedure that finds the highest eigenvalue first it is sometimes desirable to invert the matrix [K ] instead and format the problem as follows.   [K ]−1 [K ] − P K d {r } = 0 →     1 [K ]−1 K d − [1] {r } = 0 P (14.5.27) This way we get the lowest critical load first and can stop the iteration. ########### Example 14.5.2 Problem: Here is a problem that lends itself to solution by one of the equation solvers. Consider the column with multiple loads. Let L 1 = L 2 = L 3 = L3 = l as shown in Figure (a). Set up the global equations for buckling analysis. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 523 Buckling y L 2F F 3F x l l l Figure (a) Solution: Use three elements. Please note how tedious this would be to solve by direct analytical means. We number the nodes and elements in Figure (b). n2 e1 n3 n4 e2 e3 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- n1 Figure (b) The assembled stiffness matrix is ⎡ 12 ⎢ −6l ⎢ ⎢ −12 ⎢ EIzz ⎢ −6l [K ] = 3 ⎢ l ⎢ ⎢ 0 ⎢ 0 ⎢ ⎣ 0 0 −6l 4l 2 6l 2l 2 0 0 0 0 −12 −6l 6l 2l 2 24 0 0 8l 2 −12 6l −6l 2l 2 0 0 0 0 0 0 0 0 −12 −6l 6l 2l 2 24 0 0 8l 2 −12 6l −6l 2l 2 ⎤ 0 0 0 0 ⎥ ⎥ 0 0 ⎥ ⎥ 0 0 ⎥ ⎥ −12 −6l ⎥ ⎥ 6l 2l 2 ⎥ ⎥ 12 6l ⎦ 6l 4l 2 (a) Next we must find the axial load in each element. P1 = 4F P2 = 2F P3 = 3F (b) The assembled differential stiffness or initial stress matrix is ⎡ 144 ⎢ −12l ⎢ ⎢ −144 ⎢  d F ⎢ ⎢ −12l K = 30l ⎢ ⎢ 0 ⎢ 0 ⎢ ⎣ 0 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS −12l 16l 2 12l −4l 2 0 0 0 0 −144 12l 144 + 72 12l − 6l −72 −6l 0 0 −12l −4l 2 12l − 6l 16l 2 + 8l 2 6l −2l 2 0 0 0 0 −72 6l 72 + 108 6l − 9l −108 −9l 0 0 −6l −2l 2 6l − 9l 8l 2 + 12l 2 9l −3l 2 0 0 0 0 −108 9l 108 9l ⎤ 0 0 ⎥ ⎥ 0 ⎥ ⎥ 0 ⎥ ⎥ −9l ⎥ ⎥ 2⎥ −3l ⎥ 9l ⎦ 12l 2 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas 524 July 4, 2011 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The boundary conditions are v1 = θ1 = v4 = 0. By partitioning we get ⎡ ⎛⎡ ⎤⎞ ⎡ ⎤ ⎡ ⎤ ⎤ v2 24 0 −12 −6l 0 0 216 6l −72 −6l 0 2 2 2 2 ⎢ 6l ⎜⎢ 0 ⎥⎟ ⎢ θ2 ⎥ ⎢ 0 ⎥ ⎥ 8l 24l 6l −2l 0 6l 2l 0 ⎢ ⎜⎢ ⎥⎟ ⎢ ⎥ ⎢ ⎥ ⎥ ⎟⎢ ⎥ ⎢ ⎥ ⎢ ⎜⎢ −12 6l 6l 180 −3l −9l ⎥ 24 0 −6l ⎥ ⎜⎢ ⎥⎟ ⎢ v3 ⎥ = ⎢ 0 ⎥ ⎥ − β ⎢ −72 ⎣ −6l −2l 2 −3l 20l 2 −3l 2 ⎦⎠ ⎣ θ3 ⎦ ⎣ 0 ⎦ ⎝⎣ −6l 2l 2 0 8l 2 2l 2 ⎦ 0 θ4 0 0 −9l −3l 2 12l 2 0 0 −6l 2l 2 4l 2 (d) where Fl 2 (e) 30EIzz To proceed numerically we need only define the length of the elements. Let it be 100 mm; we shall add numbers for the other quantities later. Equation (d) becomes ⎤ ⎛⎡ 24 0 −12 −600 0 ⎜⎢ 0 80000 600 20000 0 ⎥ ⎥ ⎜⎢ ⎥ ⎜⎢ −12 600 24 0 −600 ⎥ ⎜⎢ ⎝⎣ −600 20000 0 80000 20000 ⎦ 0 0 −600 20000 40000 ⎡ ⎤⎞ ⎡ ⎤ ⎡ ⎤ v2 0 216 600 −72 −600 0 ⎢ 600 ⎥⎟ ⎢ θ2 ⎥ ⎢ 0 ⎥ 240000 600 −20000 0 ⎢ ⎥⎟ ⎢ ⎥ ⎢ ⎥ ⎟⎢ ⎥ ⎢ ⎥ 600 180 −300 −900 ⎥ (f) −β ⎢ ⎢ −72 ⎥⎟ ⎢ v3 ⎥ = ⎢ 0 ⎥ ⎣ −600 −20000 −300 200000 −30000 ⎦⎠ ⎣ θ3 ⎦ ⎣ 0 ⎦ θ4 0 0 0 −900 −30000 120000 β= This can be inserted into an equation solver for eigenvalue extraction. By this example we have established the validity of the FEM for buckling. As the problems become more complicated the commercial codes are used. ########### The FEM is especially good when the column is tapered and/or the axial force is not a constant. Then the differential equation has non constant coefficients and is very difficult to solve. The FEM equations can be divided into constant cross section elements and we can use concentrated equivalent nodal loads. 14.6 Buckling of Frames To extend the FEM analysis to frames   requires applying the transformation matrix presented in Section 10.4, Equations 10.4.3-9, to the kmd matrix as well as to the [km ] matrix. As you can imagine when there are several elements requiring the transformation this can become time consuming and tedious. It is best to go directly to commercial FEM codes that have this process nicely programmed. We shall not pursue this further here. 14.7 Buckling of Thin Plates and Other Structures Thin flat plates are normally the next component to consider for being susceptible to buckling. By an argument similar to that used for columns the plate equations can be extended to include the effect of both lateral and in plane loads on the deflected plate. Consider a flat plate with in plane loads on its edges. Prior to buckling this is a plane stress problem and the equations of Chapter 12, Section 12.2, apply. Assume we have solved the problem and the internal --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 525 Buckling stresses have been found. If we multiply these stresses by the plate thickness h we have the internal forces per unit length. N x (x, y) = σx (x, y) h N y (x, y) = σ y (x, y) h N x y (x, y) = τx y (x, y) h (14.7.1) When we apply equilibrium in the deflected position the plate equations with both lateral and in plane loads become   4 ∂ 4w ∂ 4w ∂ 2w ∂ 2w ∂ 2w ∂ w + Ny 2 = p(x, y) + N x 2 + 2N x y + 2 + (14.7.2) Dh 4 2 2 4 ∂x ∂x ∂y ∂y ∂x ∂ x∂ y ∂y where Dh = Eh 3 12(1 − ν 2 ) (14.7.3) For those who wish to examine the derivation of this equation we refer you to books on the subject. We will restrict our discussion to a brief look at the subject. When p(x,y) = 0 and there are compressive loads present the plate is a candidate for buckling. Exact analytical solutions for buckling of plates have been found for a few rare cases of geometry, restraint, and loading. One such case is a rectangular plate with uniform edge loads and simply supported on all four edges. In practice plates are not often rectangular, edge loads are seldom uniform, and truly simply supported edges do not exist. Even in tightly controlled laboratory experiments it is virtually impossible to build a rectangular plate with truly simply supported edges. Nevertheless the solution of this problem provides great insight into the nature of plate buckling and exists as a measure for accuracy of approximate solutions. In Section 12.2 we found the internal plane stress for a plate with uniform edge loads to be uniform internal stress and hence uniform internal forces per unit length. Let us use the solution from Example 12.2.1 to examine the buckling of a rectangular plate. ########### Example 14.7.1 Problem: A rectangular plate with uniform compressive edge loads is shown in Figure (a). It is simply supported on all four edges. The edge loads have the units of force per unit area. There are no lateral loads. Find the buckling load. y f0 f0 b x a Figure (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 526 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Solution: An exact analytical solution has been found that we shall adopt. The plane stress solution from Example 12.2.1 is N x (x, y) = σx (x, y) h = f 0 h = N0 N y (x, y) = 0 N x y (x, y) = 0 (a) Equation 14.7.2 becomes ∂ 4w ∂ 4w N0 ∂ 2 w ∂ 4w + 2 + = − ∂x4 ∂ x 2∂ y2 ∂ y4 DH ∂ x 2 (b) It may be noted that the simply supported boundary conditions require that ∂ 2w (0, y) = ∂2x ∂ 2w w (x, 0) = w (x, b) = 2 (x, 0) = ∂ y w (0, y) = w (a, y) = ∂ 2w (a, y) = 0 ∂2x ∂ 2w (x, b) = 0 ∂2 y (c) These are the conditions for no lateral displacement and no moments on all four edges. There is no restraint of in plane displacement in order for the in plane load to be applied. Try a solution of the form nπ y mπ x sin (d) w (x, y) = Cmn sin a b This satisfies the boundary conditions, so let us see if it satisfies the governing equation. Upon substitution we obtain the value of N0 for the critical buckling load. 2 2 #2 " 2  n2 a n2 m 2 N0 b π DH m + + (e) = → N = m 0 a2 b2 a2 π 2 DH a m b b2 Take, for example, a square plate for which a = b. Then the lowest buckling load is for m = n = 1 or N0 = 4 π 2 Dh b2 (f) and the mode shape is w (x, y) = C sin πx πy sin b b (g) This is plotted in Figure (b). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 1.0 1.0 0.5 0.0 0.0 0.5 0.5 1.0 0.0 Figure (b) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 7:19 Printer Name: Yet to Come 527 Buckling Other combinations of m and n can be found for other ratios of a/b. For example, if a/b = 2 the lowest buckling load is for m = 2 and n = 1, or N0 = 4 π 2 Dh b2 (h) and the mode is w (x, y) = C sin πy 2π x sin a b (i) This is plotted in Figure (c). 1.0 0.5 0.0 –0.5 –1.0 0.0 1.0 0.5 0.5 1.0 1.5 2.0 0.0 Figure (c) Unfortunately, the slightest changes in plate shape, edge loading, or edge support make it impossible to find an exact analytical solution and quite difficult to find an approximate analytical solution. But fortunately, there is help on hand—FEM to the rescue. Please accept that appropriate shape function has been found and element stiffness and differential stiffness matrices have been derived. It is beyond our efforts in this introductory text to carry this further. ########### 14.8 Summary and Conclusions Compressive loads can cause catastrophic failure at loads below those which achieve yield stress when the bars are slender or plates are thin. For slender bars in summary we have the following equations for --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-14 P2: ABC JWST071-Waas July 4, 2011 528 7:19 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods classical buckling analysis. dN = − f x (x) dx εx (x, y) u (x, y) u (x, y) = u A − y εx (x, y) = 1 du A + dx 2   d 2v 1 dv 2 du A − Ey 2 + σx (x, y) = Eεx = E dx 2 dx dx   2  1 dv du A + σx dA = EA N (x) = dx 2 dx A  M(x) = − σx ydA = E d 2v dx2 A y 2 dA = EIzz v (x) ← ρ  dv dx 2 −y dv dx d 2v dx2 (14.8.1) d 2v dx2 → σx = M (x) y N (x) − A Izz A d dM d 2v = EIzz 2 ; dx dx dx d dv d 2v → V̂ (x) = − EIzz 2 + N dx dx dx   dv d d 2v d2 N = f y (x) EI − zz dx dx dx2 dx2 τx y (x, y) = V̂ Izz h ydA Ay The equivalent finite element equations are summarized as follows. {F} → f ym (s) Pm Vm (s) Mzm (s) σxm (s, y) εxm (s, y) u m (s, y) vm (s) {rm } ← ρ vm (s) = [n (s)] {rm } d u m (s) = −y [n] {rm } ds d2 εxm (s) = −y 2 [n] {rm } ds d2 σsm (s) = −y E 2 [n] {rm } ds T T lm lm d 2 [n] d 2 [n] d d [n] [n] ds {rm } − {δrm }T EIm ds {rm } = 0 δW = {δrm }T P 2 ds ds ds ds2 0 0 T lm lm  d d [n]T d [n] d 2 [n] d 2 [n] ds {rm } ds km P [km ] = EIm 2 2 ds ds (14.8.2) ds ds 0 0  1  3 [n] = 3 lm − 3lm s 2 + 2s 3 lm3 s − 2lm2 s 2 + lm s 3 3lm s 2 − 2s 3 −lm2 s 2 + lm s 3 lm ⎛ ⎡ ⎤ −12 6lm 12 6lm 2 2 ⎜ E m Im ⎢ 6lm   4lm −6lm 2lm ⎥ ⎢ ⎥ [km ] − P kmd {rm } = ⎜ ⎝ l 3 ⎣ −12 −6lm 12 −6lm ⎦ m 6lm 2lm2 −6lm 4lm2 ⎡ ⎤⎞ ⎡ ⎤ 36 −3lm −36 −3lm vn 2 ⎟⎢ ⎥ P ⎢ 3lm −lm2 ⎥ ⎢ −3lm 4lm ⎥⎟ ⎢ θn ⎥ − 3lm 36 3lm ⎦⎠ ⎣ vn+1 ⎦ 30L m ⎣ −36 −3lm −lm2 3lm 4lm2 θn+1 [K ] = ! [km ]   ! d Kd = km Similar equations can be derived for plates and other structures. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Mc , Fc → f y (x) N (x) V (x) M(x) σx (x, y)   d dv dM dv + = −V (x) N + f y (x) = 0 dx dx dx dx P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 15 Structural Dynamics 15.1 Introduction In the analysis of structures up to this point we have assumed that loads are applied slowly, so slowly, in fact, that all the loads and the resulting stresses and displacements are independent of time. We are all aware, however, that loads are often applied rapidly and time dependent motion of the structure occurs. We are also aware that time dependent motion can be induced by imposing a displacement with a force or a restraint and then releasing the force or restraint. We speak here of time dependent displacement with no net rigid body motion. The study of such motion is called structural dynamics or vibration of structures. When structures are subjected to dynamic loads inertia terms must be added to the equations. To set the stage for the study of structural dynamics we introduce the inertia term in a simple case of mass/spring systems in the next section. In this way the nature of vibration, the form of the governing equations, and the methods of analysis can be introduced in their simplest forms. We then examine the axial motion of a slender bar. This is the simplest of the structural forms. We follow this with the torsional motion of shafts, the dynamic bending of beams, plates and shells, and dynamic motion of general three dimensions structures. In each case we shall consider first the exact analytical solution based on the differential equations of motion and then we shall solve the same problems by the finite element method. It just so happens that the equations for the dynamics of continuous structures such as slender bars, plates and shells, and general solid bodies are reduced by the finite element method to equations of the same form as mass/spring systems. For each structural form we shall first consider free motion, that is, motion induced by an initial displacement, or an initial velocity, or both. The next major concern is a phenomenon called resonance, that is, a displacement driven by an oscillatory load. The final concern is the dynamic response due to a time dependent but non oscillating load. 15.2 Dynamics of Mass/Spring Systems 15.2.1 Free Motion --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Insight into the nature of structural dynamics or vibration is quickly gained by looking at a one degree of freedom mass/spring system as shown in Figure 15.2.1 where k is the spring stiffness with units of Newtons per millimeter (N/mm), M is the value of the mass in kilograms (kg), and F(t) the applied load in Newtons (N). The displacement is v(t) in millimeters (mm). We allow the mass to stretch the Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 530 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods spring under the action of gravity until it is in static equilibrium and measure the displacement from that position. The spring is assumed to be massless. This is a reasonable assumption when the mass of the spring is small compared to the attached mass. y k υ(t) M x F(t) Figure 15.2.1 From D’Alembert’s principle the magnitude of the inertia force, FI , is equal to the mass times the acceleration, and is directed opposite to the direction of motion (y-direction). We can indicate this force in the positive y-direction as FI = −M d 2v = −M v̈ dt 2 (15.2.1) This is often called the reversed effective force. We have introduced the common practice of using dots over the symbol to indicate differentiation with respect to time. A free body diagram of the mass is shown in Figure 15.2.2. kυ(t) FI F(t) Figure 15.2.2 The equation of motion is then the sum of the applied load, the inertia force, and the spring force set equal to zero. F(t) + FI − kv = 0 → F(t) − M v̈ − kv = 0 → v̈ + ω2 v = F(t) M (15.2.2) where ω2 = Mk for convenience of notation. When there is no applied load, that is, F(t) = 0, we have what is called free motion or free vibration. Motion is induced by an initial displacement or velocity of both. The equation for free vibration is v̈ + ω2 v = 0 (15.2.3) v(t) ≈ Ceαt (15.2.4) If we substitute this into Equation 14.2.3 we get Cα 2 eαt + Cω2 eαt = (α 2 + ω2 )Ceαt = 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (15.2.5) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- From the methods for the solution of ordinary differential equations with constant coefficients the homogeneous equation has a solution of the form P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 531 Structural Dynamics If C = 0 there is no motion. Since eαt is not zero it is possible that α 2 + ω2 = 0 → (15.2.6) α = ±iω −1. There are two possible solutions to Equation 14.2.3 v(t) = C1 eiωt and v(t) = C2 e−iωt (15.2.7) For a linear equation the principle of superposition holds and so the sum of these two solutions is also a solution. It is usually more convenient to convert the solution from its exponential form to its trigonometric form. v(t) = C1 eiωt + C2 e−iωt = C1 (cos ωx + i sin ωx) + C2 (cos ωx − i sin ωx) (15.2.8) = A sin ωt + B cos ωt To find the motion in a specific case we must assign values to the constants A and B. This is done by providing initial conditions. The most general statement of initial conditions is to assign values to an initial displacement and an initial velocity. (15.2.9) v̇(0) = v̇0 v(0) = v0 When the initial conditions are applied to find the constants of integration we get v(t) = A sin ωt + B cos ωt v(0) = B = v0 → v̇(t) = Aω cos ωt − Bω sin ωt → v̇(0) = Aω = v̇0 → A= v̇0 ω (15.2.10) or v(t) = v̇0 sin ωt + v0 cos ωt ω (15.2.11) where ω=  k rad/s M (15.2.12) is known as the natural frequency of the system. The frequency has units of radians per second, rad/s. The standard unit of frequency is cycles per second or Hertz, Hz. We shall use the symbol f for frequency measured in Hertz, thus, f = ω Hz 2π (15.2.13) The natural period, T, that is, the length in time of one cycle of the motion is T = 2π ω (15.2.14) The motion can be excited by any initial displacement, initial velocity, or combination of initial √ displacement and velocity. In any case the frequency of oscillation will always be ω = k/M and the motion will always be sinusoidal. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- where i = √ α 2 = −ω2 → P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 532 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ########### Example 15.2.1 Problem: A single degree of freedom mass/spring system has the following properties. M = 2 kg k = 32 N /mm (a) From its static equilibrium position it is given an initial displacement of 4 mm, no initial velocity, and released. Find the natural frequency and resulting motion. Solution: Use Equations 15.2.12 and 15.2.13r. The natural frequency is ω=  k = M  32 = 4 rad/s 2 → f = ω 4 = = 0.6366 Hz 2π 2π (b) The resulting motion is v(t) = v̇0 sin ωt + v0 cos ωt = 0 · sin 4t + 4 cos 4t = 4 cos 4t ω (c) A plot of the first few cycles of the motion is given in Figure (a). 4 υ(t) 2 1 2 3 4 t –2 Figure (a) ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- –4 P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 533 Structural Dynamics Now consider a system with two masses and two springs as shown in Figure 15.2.3. k1 υ1(t) M1 F1(t) kF2(t) υ2(t) M2 F2(t) Figure 15.2.3 To derive the equations of motion we set up a free body diagram for each mass as shown in Figure 15.2.4. k2(υ2-υ1) k1υ1 FI 1 = − M 1υ1 F F1(t) FI 2 = − M 2υ2 F2(t) k2(υ2-υ1) Figure 15.2.4 From summation of forces F1 (t) − k1 v1 + k2 (v2 − v1 ) − M1 v̈1 = 0 F2 (t) − k2 (v2 − v1 ) − M2 v̈2 = 0 → M1 v̈1 + k1 v1 − k2 (v2 − v1 ) = F1 (t) (15.2.15) M2 v̈2 + k2 (v2 − v1 ) = F2 (t) → Collecting terms we get M1 v̈1 + (k1 + k2 )v1 − k2 v2 = F1 (t) These may be put in the matrix form     k + k2 v̈1 M1 0 + 1 −k2 v̈2 0 M2 −k2 k2  v1 v2  =  F1 (t) F2 (t)  (15.2.17) This may be written as [M]{v̈} + [K ]{v} = {F(t)} Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (15.2.18) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- (15.2.16) M2 v̈2 − k2 v1 + k2 v2 = F2 (t) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 534 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods where {v} =  v1 v2  [M] =  M1 0 0 M2  [K ] =  k1 + k2 −k2 −k2 k2  {F(t)} =  F1 (t) F2 (t)  (15.2.19) For free motion, that is, there is no applied force but motion is excited by initial conditions, we have         M1 0 0 v1 k + k2 −k2 v̈1 (15.2.20) = + 1 v2 −k2 k2 v̈2 0 M2 0 From our knowledge of the solution for the one degree of freedom system let us try a solution of the form     C1 iωt v1 e ≈ (15.2.21) C2 v2 By substituting Equation 15.2.21 into Equation 15.2.20 we obtain         C1 iωt M1 0 0 k1 + k2 −k2 e = − ω2 C2 −k2 k2 0 M2 0 (15.2.22) Since eiωt = 0  k1 + k2 −k2   −k2 M1 − ω2 k2 0 0 M2   C1 C2  =   0 0 (15.2.23) One possibility is that {C} = 0, but then there would be no motion. From the properties of homogeneous linear algebraic equations another possibility is that the determinant of the coefficient matrix is equal to zero, that is,     (k1 + k2 ) − ω2 M1 −k2 =0  (15.2.24)  −k2 k 2 − ω 2 M2  Once again we have the classic eigenvalue problem. When this is expanded we get a quadratic equation in terms of ω2 , that is, an equation of the form α1 (ω2 )2 + α2 ω2 + α3 = 0 (15.2.25) This equation has two roots ω12 and ω22 . These are squares of the two natural frequencies of the system. To find the values for {C} we substitute each natural frequency in turn back into Equation 15.2.23. For each natural frequency there will be non zero values for {C}.         0 C11 M1 0 k1 + k2 −k2 − ω12 = −k2 k2 C12 0 M2 0 (15.2.26)         0 0 C M k1 + k2 −k2 21 1 = − ω22 C22 −k2 k2 0 M2 0 We learned in Chapter 14 that for homogenous equations we cannot solve for all the constants of integration, but we can solve for all but one of the constants in terms of that one. We know the shape of the displacement but not its magnitude. These solutions are of the form ⎤ ⎤ ⎡ ⎡         1 1 C21 1 1 C11 = C21 ⎣ C22 ⎦ = C21 {ϕ}2 = C21 = C11 ⎣ C12 ⎦ = C11 {ϕ}1 = C11 C22 C12 ϕ1 ϕ2 C11 C21 (15.2.27) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 535 Structural Dynamics The resulting values are called the modes or mode shapes. The quantities {ϕ}1 and {ϕ}2 are called the normal modes. Note that we could have chosen to factor C12 or C22 and normalize that way. The actual amplitude of any normal mode is arbitrary. Equations 15.2.26 also can be written in terms of the normal modes since the factored constant can be divided out of each equation. [K ] − ω12 [M] C11 {ϕ}1 = 0 [K ] − ω12 [M] {ϕ}1 = 0 → --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- [K ] − ω22 [M] C21 {ϕ}2 = 0 (15.2.28) [K ] − ω22 [M] {ϕ}2 = 0 → Since each mode is a solution to its respective equation in Equation 15.2.28, these modes can be used to find the complete solution to the original equations, Equation 15.2.20. We take our cue from the one degree of system solution and note that there is a solution to the equations based on the first frequency and mode {v(t)}1 =  v1 (t) v2 (t)  1 = (a1 sin ω1 t + b1 cos ω1 t) C11  1 ϕ1  = (A1 sin ω1 t + B1 cos ω1 t)  1 ϕ1  1 ϕ2  (15.2.29) A second solution is based on the second frequency and mode. {v(t)}2 =  v1 (t) v2 (t)  2 = (a2 sin ω2 t + b2 cos ω2 t) C21  1 ϕ2  = (A2 sin ω2 t + B2 cos ω2 t)  (15.2.30) If each is a solution, then from the principle of superposition for linear systems, the sum of the two is a solution  v (t) {v(t)} = {v(t)}1 + {v(t)}2 = 1 v2 (t) =    v (t) + 1 v 2 (t) 1  2   (15.2.31) 1 1 v1 (t) = (A1 sin ω1 t + B1 cos ω1 t) + (A2 sin ω2 t + B2 cos ω2 t) v2 (t) ϕ1 ϕ2   Initial conditions are used to find the values of the constants A1 , B1 , A2 , B2 . Just how this is done will be demonstrated with specific examples. In Example 15.2.2 we shall find the frequencies and modes. In Example 15.2.3 we shall find the complete solution for the motion for specific initial conditions. ########### Example 15.2.2 Problem: The mass spring system shown in Figure (a) has the following values assigned to its properties. M1 = 2M M2 = M k1 = 2k k2 = k Find the natural frequencies and modes. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas 536 July 4, 2011 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 2k υ1(t) 2M k υ2(t) M Figure (a) Solution: Use Equations 15.2.17-28. The coefficient matrices are   2 0 [M] = M 0 1  k + k2 [K ] = 1 −k2 The equations of motion are     3 v̈1 (t) 2 0 + k M −1 v̈2 (t) 0 1 −k2 k2  3 −1 =k −1 1  v1 v2  −1 1  =   0 0 The frequencies are obtained from       2      [K ] − ω2 [M] = k 3 −1 − ω2 M 2 0  =  3k − 2Mω    −1 1 −k 0 1 Expanding the determinant we get 5 (ω2 M)2 − k(ω2 M) + k 2 = 0 2 → 5 k 2 (ω ) + 2M (ω2 )2 −   (b) (c)  −k  =0 k − Mω2  k M 2 =0 (d) (e) This is a quadratic equation in terms of ω2 whose two roots are ω12 = k 2M ω22 = 2k M (f)  k 2M ω2 =  (g) or ω1 = 2k M The equations for the first normal mode are then          k 2 0 1 0 3 −1 = − [K ] − ω12 [M] {ϕ}1 = k ϕ1 0 −1 1 2 0 1      0 1 2 −1 = → 0 ϕ1 −1 0.5 (h) We can solve for ϕ1 . Use either one of the two equations; they give the same result. 2 − ϕ1 = 0 − 1 + 0.5ϕ1 = 0 → ϕ1 = 2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (i) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 537 Structural Dynamics The first normal mode can be written   1 2 (j) Follow the same procedure for the second mode.          0 2 0 3 −1 1 = − 2k [K ] − ω22 [M] {ϕ}2 = k 0 −1 1 ϕ2 0 1      0 1 −1 −1 = → 0 ϕ2 −1 −1 (k) We can solve for ϕ2 . − 1 − ϕ2 = 0 − 1 − ϕ2 = 0 → ϕ2 = −1 (l) The second normal mode can be written {ϕ}2 =  1 −1  (m) We pause here to note an interesting property of normal modes that we will use to advantage later on. Note that          2 1 2 0 =0 = 1 2 M {ϕ}1T [M]{ϕ}2 = 1 2 M −1 −1 0 1 (n)          4 1 3 −1 {ϕ}1T [k] {ϕ}2 = 1 2 k =0 = 1 2 k −2 −1 −1 1 In fact         2 1 2 0 −1 M =0 = 1 −1 M 2 2 0 1          1 1 3 −1 {ϕ}2T [k] {ϕ}1 = 1 −1 k =0 = 1 −1 k 1 2 −1 1  {ϕ}2T [M]{ϕ}1 = 1 (o) We also note that  {ϕ}1T [M]{ϕ}1 = 1 {ϕ}2T [M]{ϕ}2  = 1  {ϕ}1T [k] {ϕ}1 = 1  {ϕ}2T [k] {ϕ}2 = 1         4 1 2 0 2 M = 8M = 0 = 1 2 M 2 2 0 1         2 1 2 0 −1 M = 3M = 0 = 1 −1 M −1 −1 0 1         1 1 3 −1 2 k = 3k = 0 = 1 2 k 1 2 −1 1         4 1 3 −1 −1 k = 6k = 0 = 1 −1 k −2 −1 −1 1 (p) (q) This property is known as orthogonality. It can be shown that in general {ϕ}iT [M]{ϕ} j = 0 i = j = 0 i = j {ϕ}iT [k] {ϕ} j = 0 i = j = 0 i = j We shall make good use of orthogonality when we study forced motion. ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (r) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- {ϕ}1 = P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 538 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods We have found the natural frequencies and modes but now we must find the actual motion. Let us first find the motion for a set of initial conditions. For two degrees of freedom there are four initial conditions. v1 (0) = v10 v̇1 (0) = v̇10 v2 (0) = v20 v̇2 (0) = v̇20 (15.2.32) The solution given in Equation 15.2.31, repeated here as Equation 15.2.33, has four constants to be determined.       1 1 v (t) = (A1 sin ω1 t + B1 cos ω1 t) + (A2 sin ω2 t + B2 cos ω2 t) (15.2.33) v(t) = 1 v2 (t) ϕ1 ϕ2 The initial conditions provide us with the necessary information to find the four constants of integration. ########### Example 15.2.3 Problem: The system in Example 15.2.2 has the following initial conditions. v1 (0) = v10 v̇1 (0) = 0 v2 (0) = 0 v̇2 (0) = 0 (a) Find the resulting motion. (b) Substituting Equation (b) into the four initial conditions we get v1 (0) = (A1 sin ω1 0 + B1 cos ω1 0) 1 + (A2 sin ω2 0 + B2 cos ω2 0) 1 = B1 + B2 = v10 v̇1 (0) = (A1 ω1 cos ω1 0 − B1 ω1 sin ω1 0) 1 + (A2 ω2 cos ω2 0 − B2 ω2 sin ω2 0) 1 = A1 ω1 + A2 ω2 = 0 v2 (0) = (A1 sin ω1 0 + B1 cos ω1 0) 2 (c) − (A2 sin ω2 0 + B2 cos ω2 0) 1 = 2B1 − B2 = 0 v̇2 (0) = (A1 ω1 cos ω1 0 − B1 ω1 sin ω1 0) 2 − (A2 ω2 cos ω2 0 − B2 ω2 sin ω2 0) 1 = 2A1 ω1 − A2 ω2 = 0 From the second and fourth equations you find that A1 = 0 A2 = 0 (d) From the first and third equations we have B1 = v10 3 B2 = 2v10 3 (e) The resulting motion is {v(t)} = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS        1   2v v10 v1 (t) 1 10 = cos ω1 t cos ω2 t + v2 (t) −1 2 3 3 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (f) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Solution: Find a solution based on normal modes. Equation 15.2.33 for this case is       1 1 v (t) = (A1 sin ω1 t + B1 cos ω1 t) + (A2 sin ω2 t + B2 cos ω2 t) v(t) = 1 v2 (t) −1 2 P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 539 Structural Dynamics Let us add some numerical values and plot the displacements. Let M = 2 kg v10 = 3 mm k = 64 N /mm (g) Then ω1 =  k = 2M  64 = 4 rad/s 2·2 ω2 =  2k = M  2 · 64 = 8 rad/s 2 (h) The displacements of the two masses are v1 (t) = cos 4t + 2 cos 8t v2 (t) = 2 cos 4t − 2 cos 8t (i) These are plotted in Figure (a). 3 υ1(t) 2 1 1 2 3 4 5 1 2 3 4 5 t –1 –2 2 υ2(t) 1 t –1 –2 –4 Figure (a) ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- –3 P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas 540 July 4, 2011 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods When there are more masses and springs and hence more degrees of freedom the equations for the free motion are of the form [M]{v̈} + [K ]{v} = {0} (15.2.34) where the number of equations is equal to the number of degrees of freedom. The solution is found by first finding the natural frequencies. These are determined by assuming a solution of the form {v(t)} ≈ {C}eiωt (15.2.35) which results in [K ] − ωi2 [M] {C}eiωt = 0 → The natural modes are then found from   [K ] − ω2 [M] = 0 (15.2.36) ([K ] − ω12 [M]){C}1 = 0 ([K ] − ω22 [M]){C}2 = 0 .. . (15.2.37) ([K ] − ωn2 [M]){C}n = 0 The normal modes are found in shape but their amplitude is not determined. As noted it is customary to factor out one of the constants so the remaining matrix {ϕ}i is known, for example, as shown in Equation 15.2.27. {C}i = Ci1 {ϕ}i (15.2.38) It follows that [K ] − ωi2 [M] {ϕ}i = 0 (15.2.39) The constant Ci1 is often assigned an arbitrary value, say, Ci1 = 1. The resulting matrix {ϕ}i is then called a normal mode. To satisfy the initial conditions the normal modes {ϕ}i are summed. {v(t)} = n  1 (Ai sin ωi t + Bi cos ωi t){ϕ}i (15.2.40) The constants are evaluated using the initial conditions. 15.2.2 Forced Motion—Resonance When an applied force is present this is known as forced motion. For a single degree of freedom mass/spring system the equation of motion is given by Equation 15.2.2 repeated here as Equation 15.2.41. v̈ + ω2 v = F(t) M (15.2.41) Its solution is made up of a homogenous and a particular part v(t) = A sin ωt + B cos ωt + vparticular (15.2.42) Of major interest is the solution for a periodic applied force, for example, F(t) = F0 sin t (15.2.43) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 541 Structural Dynamics We try a particular solution of the form v p = P sin t (15.2.44) Substitute this into Equation 15.2.41. v̈ + ω2 v = F0 sin t M d2 F0 (P sin t) + ω2 P sin t = sin t → dt 2 M F0 → (−2 + ω2 )P sin t = sin t M F0 → P= 2 (ω − 2 )M (15.2.45) The complete solution is then v(t) = A sin ωt + B cos ωt + F0 sin t (ω2 − 2 )M (15.2.46) Initial conditions are used to find the values of the constants A and B. For initial conditions v(0) = v0 (15.2.47) v̇(0) = v̇0 we get F0 sin 0 = v0 → B = v0 − 2 )M F0  cos 0 v̇(0) = Aω cos ω0 − Bω sin ω0 + 2 (ω − 2 )M F0 F0 v̇0  = Aω + 2  = v̇0 → A = − 2 (ω − 2 )M ω (ω − 2 )M ω v(0) = A sin ω0 + B cos ω0 + (ω2 (15.2.48) When these values of A and B are inserted into Equation 15.2.46 the complete solution becomes F0 sin t (ω2 − 2 )M v̇0 F0 F0  = sin ωt − 2 sin ωt + v0 cos ωt + 2 sin t ω (ω − 2 )M ω (ω − 2 )M   F0  v̇0 = v0 cos ωt + sin ωt + 2 sin t − sin ωt ω (ω − 2 )M ω v(t) = A sin ωt + B cos ωt + (15.2.49) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- We note that when the applied frequency approaches the natural frequency the particular solution grows rapidly and when they are equal the displacement is infinite for any and all initial conditions. Note that the amplitude of the motion is determined by the coefficient of the particular solution which can be reformulated. P= (ω2 F0 F0 = 2  2 −  )M ω M 1 1−  ω2 2 F =   2 1− 2 ω → P =  F 1 1− 2 ω2  (15.2.50) where F= Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS F0 ω2 M (15.2.51) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 542 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods   P  in Figure 15.2.5 shows the dependence of amplitude The plot of the absolute value of   versus F ω on the relation between the applied frequency and the natural frequency. When the applied frequency  equals the natural frequency, that is, when = 1, the amplitude is infinite. ω --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 4 Ρ F 3 2 1 Ω 0.5 1.0 1.5 2.0 2.5 3.0 ω Figure 15.2.5 ########### Example 15.2.4 Problem: Find the forced motion for a single degree of freedom mass/spring system when F(t) = F0 sin t (a) and the initial conditions are v(0) = v0 = 0 v̇(0) = v̇0 = 0 Solution: Use initial conditions to find the constants in Equation 15.2.49. Applying the initial conditions we get   F0  v̇0 sin ωt + 2 sin t − sin ωt v(t) = v0 cos ωt + ω (ω − 2 )M ω    F0 sin t − sin ωt = 2 (ω − 2 )M ω (b) (c) If we insert numbers from Example 15.2.1 ω = 4 rad/s M = 2 kg (d)  = 2 rad/s (e) and assign numerical values to the others F0 = 10 N Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 543 Structural Dynamics the motion is v(t) =    2 10  sin 2t − sin 4t sin ωt = (16 − 4) 2 ω 4 5 5 sin 2t − sin 4t = 0.4167 sin 2t − 0.2083 sin 4t = 12 24 F0 (ω2 − 2 )M  sin t − (f) If we change the applied frequency to twice the value of the natural frequency, that is, to  = 8 rad/s, we see a decrease in the amplitude of the motion.     8 10  F0 sin 8t − sin ωt = sin 4t sin t − v(t) = 2 (16 − 64) 2 (ω − 2 )M ω 4 (g) 5 5 sin 4t = −0.1042 sin 8t + 0.2083 sin 4t = − sin 8t + 48 24 If we change the applied frequency to a value much closer to the natural frequency, say,  = 3.6 rad/s, we see a large increase in the amplitude of the motion.     3.6 10  F0 sin ωt = sin 4t sin t − sin 3.6t − v(t) = 2 (16 − 12.96) 2 (ω − 2 )M ω 4 (h) = 1.645 sin 3.6t − 1.480 sin 4t Ω = 3.6 4 Ρ F 3 2 Ω =8 1 Ω=2 0.5 1.0 1.5 2.0 2.5 3.0 Ω ω Figure (a) For this mass/spring system with a natural frequency of 4 rad/s the amplitudes on Figure 15.2.5 would be as shown in Figure (a) for each of the applied frequencies given in Equations (f), (g), and (h). ########### We see that when the applied frequency  approaches the natural frequency ω the motion grows quite large. This is a phenomenon called resonance that can be quite dangerous and, at times, destructive. Thus an important design consideration is to tune the values of M and k to avoid any natural frequencies --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 544 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods ω that might be near values of applied frequencies . Alternatively, we can place operating restrictions so that there is no applied frequency at or near the natural frequency. Let us extend this to two or more degrees of freedom. The equations of motion are of the form [M]{v̈} + [K ]{v} = {F(t)} (15.2.52) The homogeneous equations can be solved for the frequencies and modes as just described. Once we have found the frequencies and modes we can note that [K ] − ωi2 [M] {ϕ}i = 0 [K ]{ϕ}i = ωi2 [M]{ϕ}i → (15.2.53) We noted in a two degree of freedom system the complete solution is the sum of the solution based on the two normal modes as expressed in Equation 15.2.33 and this can be extended to as many degrees of freedom as there are shown in Equation 15.2.40. This can be extended to forced motion and an advantage taken by the orthogonality property of normal modes mentioned briefly for the special case of two degrees of freedom in Example 15.2.2. This property allows us to simplify the solution for a large number of degrees of freedom. For many degrees of freedom a solution is found by assuming a series of the normal modes. {v} = n  ξi (t){ϕ}i (15.2.54) i=1 where ξi (t) is to be determined. Insert this into Equation 15.2.51 [M] n  i=1 ξ̈i (t){ϕ}i + [K ] n  i=1 ξi (t){ϕ}i = {F(t)} (15.2.55) and premultiply by {ϕ}Tj {ϕ}Tj [M] n  i=1 ξ̈i (t){ϕ}i + {ϕ}Tj [K ] n  ξi (t){ϕ}i = {ϕ}Tj {F(t)} (15.2.56) ξi (t){ϕ}Tj [K ]{ϕ}i = {ϕ}Tj {F(t)} (15.2.57) i=1 Now rearrange the terms n  i=1 ξ̈i (t) {ϕ}Tj [M]{ϕ}i + n  i=1 The orthogonality relation applies to all normal modes. {ϕ}Tj [M]{ϕ}i = 0 = Mj i = j (15.2.58) i= j From Equations 15.2.53 and 15.2.58 we can say {ϕ}Tj [K ]{ϕ}i = {ϕ}Tj ωi2 [M]{ϕ}i = 0 i = j = ω2j M j i= j (15.2.59) We can rewrite Equation 15.2.57 as follows M j ξ̈ j + ω2j M j ξ j = f j (t) → ξ̈ j + ω2j ξ j = f j (t) Mj (15.2.60) where f j (t) = {ϕ j }T {F(t)} (15.2.61) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 545 Structural Dynamics Each equation contains only one dependent variable. This is the same equation as for the one degree of freedom system and the solution is the same. By expanding the displacement in a series of normal modes and taking advantage of the orthogonality of modes we have transformed the set of coupled equations, Equations 15.2.52, into a set of uncoupled equations, Equations 15.2.60. It is time for an example. ########### Example 15.2.5 Problem: The mass/spring system in Example 15.2.2 is initially at rest and is then excited by a sinusoidal force applied to the first mass equal to F1 (t) = F10 sin t (a) The initial conditions are v(0) = v0 = 0 v̇(0) = v̇0 = 0 (b) Find the resulting motion. Solution: Use orthogonality of modes to find the uncoupled equations. Solve the uncoupled equations and sum according to Equations 15.2.54. 2k υ1(t) 2M F10 sin Ωt k υ2(t) M Figure (a) The equations of motion are M  2 0 0 1  v̈1 (t) v̈2 (t)  + k  3 −1 −1 1  v1 v2  =  F0 sin t 0  (c) We use the natural frequencies and normal modes found in Example 15.2.2 to convert the equations to the uncoupled form. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ JWST071-15 P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 546 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods To convert to the form of Equations 15.2.60 we find     1 2 0 2 M = 6M 2 0 1      1 2 0 = 3M M2 = {ϕ}2T [M]{ϕ}2 = 1 −1 M −1 0 1     F10 sin t f 1 (t) = {ϕ}1T {F(t)} = 1 2 = F10 sin t 0     F10 sin t f 2 (t) = {ϕ}2T {F(t)} = 1 −1 = F10 sin t 0  M1 = {ϕ}1T [M]{ϕ}1 = 1 (d) The first uncoupled equation is ξ̈1 + ω12 ξ1 = f 1 (t) M1 ξ̈1 + ω12 ξ1 = → F10 sin t 6M (e) This equation if the same for as the single mass/spring system and so the particular solution is ξ1 particular = F0 sin t ω12 − 2 6M (f) The complete solution for the first uncoupled equation is ξ1 (t) = a1 sin ω1 t + b1 cos ω1 t + ω12 F0 sin t − 2 6M (g) By the same procedure the second uncoupled equation is ξ̈2 + ω22 ξ2 = f 2 (t) M2 ξ̈2 + ω22 ξ2 = → F10 sin t 3M (h) and the complete solution for the second equation is ξ2 (t) = a2 sin ω2 t + b2 cos ω2 t + F0 sin t ω22 − 2 3M (i) The initial conditions are {v(0)} = n  i=1 [ξi (0)] {ϕ}i = 0 →     1 1 =0 + ξ2 (0) ξ1 (0) −1 2 → ξ1 (0) = 0 ξ2 (0) = 0 {v̇(0)} = n   i=1  ξ̇i (0) {ϕ}i = 0 → ξ̇1 (0) = 0 → (j)     1 1 =0 + ξ̇2 (0) ξ̇1 (0) −1 2 ξ̇2 (0) = 0 From the solution for the single degree of freedom system given in Equation 15.2.49 for initial conditions at rest the solutions for the uncoupled equations are    F10 sin t − sin ω t ξ1 (t) = 1 ω1 ω12 − 2 6M (k)    F10 sin t − ξ2 (t) = sin ω t 2 ω2 ω22 − 2 3M --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 547 Structural Dynamics It follows that        n  F10 1 v1 (t) = sin t − ξi (t){ϕ}i = sin ω1 t {v} = 2 2 6M v2 (t) 2 ω ω −  1 1 i=1       F10 1 sin t − sin ω2 t + −1 ω2 ω22 − 2 3M  (l) Note that each time the applied frequency approaches a natural frequency the motion grows quite large in amplitude. ########### For multi degree of freedom systems with the same applied load on the first mass and the same initial conditions the complete solution is   n   F10 sin t − sin ω t {ϕ}i (15.2.62) v(t) = i ωi ωi2 − 2 Mi i=1 15.2.3 Forced Motion—Response For multi degree of freedom systems when f j (t) is not periodic we get ξ j = a j sin ω j t + b j cos ω j t + ξ j particular (15.2.63) The particular solutions are not periodic and so resonance does not occur. In this case finding the maximum amplitude is often a determining design factor. Again let us see an example. ########### Example 15.2.6 Problem: The one degree of freedom mass/spring system is at rest at t = 0, that is, v(0) = v̇(0) = 0 (a) and then a constant force is applied to the mass. F(t) = 0 F(t) = F0 t <0 t ≥0 (b) y k υ(t) x M F(t) = F0 Figure (a) Find the resulting motion. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 548 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Solution: Find the particular solution for this case. The governing equation is v̈ + ω2 v = F0 M (c) Try a particular solution of the form vp = P (d) Then ω2 P = F0 M (e) The complete solution is v(t) = a sin ωt + b cos ωt + F0 ω2 M (f) Apply initial conditions. F0 =0 ω2 M v̇(0) = aω cos ω0 − bω sin ω0 = 0 → v(0) = a sin ω0 + b cos ω0 + → b=− F0 ω2 M (g) a=0 The complete solution is F0 (1 − cos ωt) ω2 M F0 This is a steady state oscillation of amplitude 2 . ω M (h) v(t) = ########### The extension to multiple degree of freedom systems is a matter of finding the appropriate particular solutions. We shall explore this further later on in this chapter. Next let us consider the vibration of various structural forms starting with the axial motion of a slender bar 15.3 Axial Vibration of a Slender Bar 15.3.1 Solutions Based on the Differential Equation We shall see that the axial bar can have a very similar behavior to that of a mass/spring system. The equation for the static axial displacement of a bar was given in Chapter 4, Section 4.3, Equation 4.3.7. du d EA = − f x (x) (15.3.1) dx dx The governing equation for dynamic behavior is found by considering the inertia force as an additional applied load using D’Alembert’s principle. We note that u(x, t), that is, u is a function of both space and time and so partial derivatives are needed. The governing equation is ∂u ∂ EA = − f x (x, t) − FI = − f (x, t) + m ü ∂x ∂x where m is the mass per unit length of the bar. (15.3.2) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 549 Structural Dynamics For a uniform bar this becomes EA ∂ 2u − m ü = − f x (x, t) ∂x2 (15.3.3) First we consider what is called the free vibration, that is, vibration induced by initial conditions but no applied load. EA ∂ 2u − m ü = 0 ∂x2 (15.3.4) To find a solution we use a method called separation of variables, that is, we assume --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- u(x, t) = U (x)T (t) (15.3.5) and substitute it into Equation 15.3.4. The result can be separated into two distinct equations. ∂ 2U EA 2 T − mU T̈ = 0 ∂x d 2U 2 d x 2 = 1 d T = constant = −ω2 mU T dt 2 EA → (15.3.6) On the left side of the equation all the quantities are functions of x only, while those to the right are functions of t only. This can be true only if both are equal to the same constant. The choice of the constant as −ω2 is done by hindsight. It turns out that ω is the natural frequency of the system. Thus we obtain two separate equations. T̈ + ω2 T = 0 d 2U d 2U mω2 U = + + α2U = 0 dx2 EA dx2 (15.3.7) where α2 = mω2 EA (15.3.8) The solutions are T (t) = a sin ωt + b cos ωt U (x) = c sin αx + d cos αx (15.3.9) The natural frequencies and modes are found by applying boundary conditions to the second of equations 15.3.9. We shall demonstrate this with a specific example. ########### Example 15.3.1 Problem: Consider the bar in Figure (a) with a cross sectional area A (mm2 ), Young’s modulus E (N/mm2 ), length L (mm), and mass per unit length m (kg/mm). It is restrained on the left end as shown in Figure (a). Find the natural frequencies and modes. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 550 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y E, A, m, L x L Figure (a) Solution: Separate variables and apply boundary conditions to the second of Equations 15.3.9. Given U (x) = c sin αx + d cos αx (a) we apply boundary conditions of no displacement at x = 0 and no load at x = L. u(0, t) = T (t)U (0) = 0 EA → du(L , t) dU (L) = EA T (t) = 0 dx dx U (0) = c sin α · 0 + d cos α · 0 = 0 → dU (L) = cα cos αL = 0 dx → d=0 (b) It must be true that either c = 0, or α = 0, or cos αL = 0. If either c or α is zero then nothing happens, that is, there is no displacement; however, if   (2i − 1) π EA (2i − 1) mω2 cos αL = 0 → αi L = π=L → ωi = (c) 2 EA 2 L m where i is any positive integer, we can have displacement. The first four frequencies, for example, are     1 π EA 3 π EA 5 π EA 7 π EA ω2 = ω3 = ω4 = (d) ω1 = 2L m 2L m 2L m 2L m From the solution to the time dependent part of Equation 15.3.9 T (t) = a sin ωt + b cos ωt (e) we can recognize that the ω′ s are the natural frequencies and that there are an infinite number of them. As noted if c = 0 then Ui = ci ϕi (x) = ci sin αi x (f) Since the ci are indeterminate we assign them the value of 1 and the normal modes are ϕi (x) = sin αi x = sin (2i − 1) πx 2L There are an infinite number of them. The first four normal modes are π x ϕ1 (x) = sin α1 x = sin 2 L 3π x ϕ2 (x) = sin α2 x = sin 2 L 5π x ϕ3 (x) = sin α3 x = sin 2 L 7π x ϕ4 (x) = sin α4 x = sin 2 L These modes are plotted in Figure (b). --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (g) (h) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 551 Structural Dynamics Mode 1 1.0 ϕ(x) 0.5 Mode 3 0.2 –0.5 0.4 0.6 0.8 1.0 x L Mode 4 Mode 2 –1.0 Figure (b) ########### Taking our cue from the mass/spring systems it is suggested that the motion can be described by a sum of the normal modes as follows u(x, t) =  ( Ai sin ωi t + Bi cos ωi t) ϕi (x) (15.3.10) The initial conditions would then be satisfied by u(x, 0) = u̇(x, 0) =   (Ai sin ωi 0 + Bi cos ωi 0) ϕi (x) =  Bi ϕi (x)  (Ai ωi cos ωi 0 − Bi ωi sin ωi 0) ϕi (x) = Ai ωi ϕi (x) (15.3.11) The actual initial conditions may not be the same functional shape as the normal modes but an interesting property of the normal modes saves the day for us. Similar to what happens in the mass/spring systems it just so happens that --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`---  L ϕi (x)mϕ j (x)d x(x) = 0, j = i = Mi j =i 0 (15.3.12) This is the property called orthogonality as it applies in this case. In this case it is weighted by the distributed mass. If we multiply both sides of Equations 15.3.7 by mϕ j (x) and integrate over the length of the bar as follows:   Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS L u(x, 0)ϕ j (x)md x = 0 L 0 u̇(x, 0)ϕ j (x)md x =   L 0 L 0   Bi ϕi (x)ϕ j (x)md x =  Ai ωi ϕi (x)ϕ j (x)md x = Bi   L ϕi (x)ϕ j (x)md x 0 Ai ωi  L (15.3.13) ϕi (x)ϕ j (x)md x 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 552 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods We note that all the terms in the summation on the right side are zero except when i = j. We then can solve for Ai and Bi .  L 1 u(x, 0)ϕi (x)md x Mi 0  L 1 u̇(x, 0)ϕi (x)md x Ai = Mi ωi 0  L (ϕi (x))2 md x Mi = Bi = (15.3.14) (15.3.15) 0 Now there are an infinite number of frequencies and modes and therefore and infinite number of terms in the series solution given in Equation 15.3.10. Fortunately this is a highly convergent series and only a few terms need be evaluated to achieve an accurate representation of the ensuing motion in most cases. Let us look at an example. ########### Example 15.3.2 Problem: The bar in Example 15.3.1, shown here in Figure (a), is given an initial displacement and no initial velocity. u(x, 0) = u 0 x L u̇(x, 0) = 0 (a) From Chapter 4 we know that u(L) = FL = u0 EA → F= u 0 EA L (b) The force is applied to provide an initial displacement and then at t = 0 is suddenly released. Find the resulting motion. y E, A, m, L x --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- u0 L Figure (a) Solution: Find a series solution using Equation 15.3.10. The dynamic displacement based on the normal modes is u(x, t) = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  (Ai sin ωi t + Bi cos ωi t) sin αi x Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 553 Structural Dynamics Use the initial conditions to find the values of the constants.  L  L 1 1 x Bi = u(x, 0)ϕi (x)md x = u 0 (sin αi x) md x Mi 0 Mi 0 L  L u0m u0m L  (sin αi x − αi x cos αi x) x (sin αi x) d x = 2 = 0 Mi L 0 αi Mi L u0m (sin αi L − sin αi 0 − αi L cos αi L + αi 0 cos αi 0) = 2 αi Mi L u0m (sin αi L) = 2 αi Mi L  L  L 1 1 Ai = u̇(x, 0)ϕi (x)md x = 0 (sin αi x) md x = 0 Mi ωi 0 Mi ωi 0 (d) where Mi =  L (sin αi x)2 md x = 0 L m  (αi x + sin αi x cos αi x) 0 2αi mL m (αi L − αi · 0 + sin αi L cos αi L − sin αi · 0 cos αi · 0) = = 2αi 2 (e) The displacement is  Bi sin αi x cos ωi t = Remember from Example 15.3.1 we found that cos αi L = 0  u0m (sin αi L) sin αi x cos ωi t αi2 Mi L → αi L = (2i − 1) π 2 (f) (g) and therefore sin αi L = 1 = −1 i = odd i = even (h) The first few values of Bi are u0m 2u 0 m 2u 0 (sin αi L) = 2 (sin αi L) (sin αi L) = αi2 Mi L αi m L 2 (αi L)2 8u 0 8u 0 8u 0 8u 0 → B1 = 2 , B2 = − 2 , B3 = , B4 = − π 9π 25π 2 49π 2 The displacement can now be written as  u(x, t) = Bi sin αi x cos ωi t Bi = = π 8u 0 x cos ω1 t sin π2 2L 3π 8u 0 x cos ω2 t sin − 9π 2 2L 8u 0 5π + x cos ω3 t sin 25π 2 2L 7π 8u 0 x cos ω4 t sin − 49π 2 2L + ··· ########### Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (i) (j) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- u(x, t) = P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 554 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Let us demonstrate the rate of convergence of the series solution given by Equation 15.3.10. A numerical example will help. ########### Example 15.3.3 Problem: The bar in Example 15.3.2 is given the following properties. A = 100 mm 2 L = 300 mm E = 68950 N /mm 2 m = 0.271 kg/mm Show how many terms are needed to demonstrate convergence of the series solution for the motion given by Equation (j) in Example 15.3.2. Solution: Evaluate Equation (j) for the first four modes in the series solution and compare one term, two term, three term, and four term solutions. A solution based on the first four terms, that is, summing the first four terms of the infinite series would be  u(x, t) = Bi sin αi x cos ωi t = π 8u 0 x cos ω1 t sin π2 2L 3π 8u 0 x cos ω2 t sin − 9π 2 2L 5π 8u 0 x cos ω3 t sin + 2 25π 2L 7π 8u 0 x cos ω4 t sin − 49π 2 2L The natural frequencies found in Example 15.3.1 are   1 π 1 π EA 68950 · 100 = = 26.411 rad/sec ω1 = 2L m 2 300 0.271  3 π EA = 3ω1 = 79.232 rad/sec ω2 = 2L m  5 π EA = 5ω1 = 132.054 rad/sec ω3 = 2L m  7 π EA = 7ω1 = 184.875 rad/sec ω4 = 2L m (a) (b) For plotting purposes let u 0 = 1 and let us evaluate the motion at x = L. u(L , t) = 8 cos 26.411t π2 + 8 cos 79.232t 9π 2 + 8 cos 132.054t 25π 2 8 cos 184.875t + 49π 2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 555 Structural Dynamics Note that the amplitude of the second term is 1/9 of the first term and the frequency is three times as great. The amplitudes get smaller and the frequencies higher for each successive term. One term, two term, three term, and four term solutions are plotted in order in Figures (a), (b), (c), and (d). u (L,t) 0.5 t (sec) 0.2 0.4 0.6 0.8 1.0 –0.5 Figure (a) u (L,t) 0.5 0.2 0.4 0.6 0.8 1.0 t (sec) –0.5 Figure (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 556 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods u (L,t) 0.5 0.2 0.4 0.6 0.8 1.0 t (sec) –0.5 Figure (c) u (L,t) 0.5 t (sec) 0.2 0.4 0.6 0.8 1.0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- –0.5 Figure (d) The addition of the higher frequency terms has a noticeable but small effect on the overall motion. For this case adding still higher terms would have little effect. ########### When a dynamic axial force is applied to a uniform slender bar the equation of motion becomes EA Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ∂ 2u − m ü = − f (x, t) ∂x2 (15.3.16) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 557 Structural Dynamics In pursuit of a solution we first find the natural frequencies and modes as just presented. Having found the frequencies and modes we assume a solution of the form u(x, t) = ∞  ξi (t)ϕi (x) (15.3.17) i=1 We insert this form into Equation 14.3.12 EA ∞ ∞  ∂2  ξ (t)ϕ (x) − m ξ̈i (t)ϕi (x) = − f (x, t) i i ∂ x 2 i=1 i=1 (15.3.18) We can rearrange the terms. EA ∞  ∞ ξi i=1  ∂ 2 ϕi −m ξ̈ ϕi = − f (x, t) 2 ∂x i=1 Now multiply all terms by ϕ j (x) and integrate all terms over the length of the bar.   L   ∞ ∞  ∂ 2 ϕi = − f (x, t) dx ξi − m ϕ j EA ξ̈ ϕ i ∂x2 0 i=1 i=1  L  L  L ∞ ∞   ∂ 2 ϕi ξi EAϕ j ϕ mϕ d x = − d x − ϕ j f (x, t)d x = ξ̈ i j i ∂x2 0 0 0 i=1 i=1 (15.3.19) (15.3.20) The normal modes are solutions to the separated equation EA d 2U + mω2 U = 0 dx2 (15.3.21) EA d 2 ϕi + mωi2 ϕi = 0 dx2 (15.3.22) d 2 ϕi = −mωi2 ϕi dx2 (15.3.23) It follows that Thus we can set EA d 2 ϕi with −mωi2 ϕi Equation 15.3.18 becomes dx2  L  L  ∞ ∞   2 ϕ j mϕi d x − ξi ωi ϕ j mϕi d x = − ξ̈i − By replacing EA i=1 0 i=1 0 L ϕ j f (x, t)d x (15.3.24) 0 We designate f j (t) = −  L ϕ j f (x, t)d x (15.3.25) 0 We note that from orthogonality  L ϕ j mϕi d x = 0 0 = Mj i = j (15.3.26) i= j --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 558 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Finally the equation for forced motion may be written ξ̈ j + ω2j ξ = f j (t) Mj (15.3.27) We now solve for ξ j (t) and insert it in the series of Equation 15.3.17. Once again an example will help. ########### Example 15.3.4 Problem: A periodic force is applied to the right end of the bar as shown in Figure (a). y F(t) = F0 sin Ωt E, A, m, L x L Figure (a) Solution: Solve Equations 15.3.27, repeated here as Equation (a). f j (t) ξ̈ j + ω2j ξ = Mj (a) Then sum the results as given by Equation 15.3.17, repeated here as Equation (b). u(x, t) = ∞  ξi (t)ϕi (x) (b) i=1 From Example 15.3.1, Equation (c), we have the natural frequencies and the normal modes   (2i − 1) (2i − 1) π EA mω2 π =L → ωi = cos αL = 0 → αi L = 2 EA 2 L m ϕi (x) = sin αi x (c) First evaluate  0 L ϕ j f (x, t)d x = − L  L F0 sin t sin α j xd x 0 F0 sin t −cos αi x|0L αi 0 F0 F0 F0 = sin t(cos αi L − cos αi 0) = sin t(0 − 1) = − sin t αi αi αi  L  L ϕ j mϕ j d x = m(sin α j x)2 d x Mj = = −F0 sin t 0  sin α j xd x = − 0 m = (α j x − sin α j x cos α j x)|0L 2α j m (α j L − sin α j L cos α j L − α j 0 − sin α j 0 cos α j 0) = 2α j mL m αj L = = 2α j 2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- f j (t) = − P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 559 Structural Dynamics Insert these values in Equation (a). 2F0 f j (t) sin t =− Mj αjmL ξ̈ j + ω2j ξ j = (e) When you compare this equation with the single degree of freedom mass/spring equation in Equation 15.2.41 and its solution in Equation 15.2.46 you see immediately that ξ j (t) = A j sin ω j t + B j cos ω j t − 2F0 sin t α j ω2j − 2 m L (f) Apply initial conditions to solve for A j and B j . The complete solution is given by Equation (b). Resonance will occur when an applied frequency approaches any one of the infinite number of natural frequencies. ########### The response to a non periodic applied force is reduced to finding the particular solutions to Equations 15.3.27. ########### Example 15.3.5 Problem: Instead of a periodic force a step function force is applied at the right end of the bar in Example 15.3.4. F(t) = 0 = F0 t <0 t ≥0 (a) y E, A, m, L F(t) x L Figure (a) Solution: Find fj (t) and then the particular solutions. The applied force becomes  L  L F0 F0 ϕ j f (x, t)d x = − F0 sin αi xd x = − (cos αi L − 1) = − f j (t) = − αi αi 0 0 (b) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The uncoupled equations of motion are then ξ̈ j + ω2j ξ j = − F0 Mjαj (c) This has a solution of the same form as that obtained in Example 15.2.6. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS ξ j (t) = A j sin ω j t + B j cos ω j t − F0 ω2j M j α j Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 560 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The same steps as in Example 15.2.6 are followed and the final motion is given by Equation 15.3.17. u(x, t) = ∞  ξi (t)ϕi (x) (e) i=1 ########### --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 15.3.2 Solutions Based on FEM From Chapter 5, Section 5.3, we note that the stiffness matrix of an axial member for element m is { pm } = [km ]{rm } = E m Am lm  1 −1 −1 1  rm,n rm,n+1  (15.3.28) When inertia forces are assigned the equivalent nodal loads resulting from the inertia forces for element m are    l  l r̈n {FI }enl = = − [m] {r̈m } [n]T (−m ü) d A = − (15.3.29) [n]T [n]mds r̈n+1 0 0 We have just defined a mass matrix [m m ] =  l [n]T [n]mds (15.3.30) 0 If m is a constant value the element mass matrix is ⎡ s⎤    l 1−  s s m m lm 2 1 l ⎥ ⎢ mds = [m m ] = ⎣ s ⎦ 1− 1 2 l l 6 0 l (15.3.31) There is confusing notation here. The subscript m refers to element m. The other m is the mass per unit length. These are assembled to get the global stiffness matrix [K ] and the global mass matrix [M] in the same way as the static case. The assembled equations for free vibration are [M]{r̈} + [K ]{r } = {0} (15.3.32) This is best seen by example. ########### Example 15.3.6 Problem: Solve for the free vibration of the same problem posed in Example 15.3.2 by the FEM. The initial conditions are u(x, 0) = u 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS x L u̇(x, 0) = 0 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 561 Structural Dynamics y E, A, m, L x L Figure (a) Solution: Assemble the FE equations and solve using the same methods as those used for the mass/spring system. Let us use just one element. The node and element numbering is shown in Figure (b). e1 n1 n2 Figure (b) Then the equation of motion is [M]{r̈} + [K ]{r } = mL 6  2 1 1 2    EA 1 0 + ü 2 L −1 −1 1  0 u2  =  R1 0  (b) After partitioning the remaining equation for the displacement is mL EA u2 = 0 ü 2 + 3 L ü 2 + → 3EA u 2 = ü 2 + ω2 u 2 = 0 m L2 (c) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- This is exactly the same form as Equation 15.2.3 for the mass/spring system and has a solution in exactly the same form. The natural frequency is  3EA 3EA ω2 = → ω = (d) m L2 m L2 Assigning initial conditions u(x, 0) = u 0 u̇(x, 0) = 0 x L → u(L , 0) = u 2 (0) = u 20 (e) → u̇(L , 0) = u̇ 2 (0) = 0 we obtain u(t) = a sin ωt + b cos ωt → u(0) = b = u 20 u̇(0) = aω = 0 (f) The final solution for the motion of node 2 is u 2 (t) = u 20 cos ωt It follows that the complete motion for the bar is approximated by    s s  u 1 (t) u(s, t) = 1 − u 2 (t) l l    x x  x 0 = (u 20 cos ωt) = u(x, t) = 1 − u 2 (t) L L L Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (g) (h) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 562 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Only one frequency results for the one element approximation and it may be compared with the exact value for the lowest frequency in Example 15.3.1, Equation (d).    √  1.732 EA 1.571 EA π EA 3 EA = Exact → ω1 = = (i) FE → ω1 = L m L m 2L m L m This is not bad for such a crude approximation. The normal mode shape is ϕ1 = x L (j) where we have arbitrarily assigned u 20 = 1. We can compare this mode with the exact value. The two are plotted in Figure (c). Note that this very crude approximation for the mode shape, namely, a linear function, compared to the exact shape, a sine function, still gives a ball park value for the frequency. 1.0 0.8 Exact 0.6 1 element FEM 0.4 0.2 0.2 0.4 0.6 0.8 1.0 Figure (c) We shall improve on the FEM solution by using more elements. Let us do free vibration with two elements, each of length L/2. The nodal and element numbers are shown in Figure (d) e1 n1 e2 n2 n3 Figure (d) The assemble global FE equations for free vibration with restraints inserted are ⎡ ⎡ ⎤⎡ ⎤ ⎡ ⎤ ⎤⎡ ⎤ R1 2 1 0 1 −1 0 0 0 mL ⎣ 2EA ⎣ −1 2 −1 ⎦ ⎣ u 2 ⎦ = ⎣ 0 ⎦ 1 4 1 ⎦ ⎣ ü 2 ⎦ + 12 L u3 ü 3 0 0 −1 1 0 1 2 (k) The effect of the partitioned equations is to obtain          2EA 2 −1 u 2 iωt 4 1 0 2 mL −ω e = u3 −1 1 0 L 12 1 2 The resulting equations have a solution other than     u2 0 = u3 0 (l) (m) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 563 Structural Dynamics only if the determinant of the coefficients is equal to zero, that is,       2EA 2 −1 m L 4 1   =0 − ω2  L −1 1 12 1 2  (n) Expanding the determinant we get a polynomial in ω2 . To simplify things a bit let λ= ω2 m L 2 24EA (o) Then the determinant becomes    2 − 4λ −1 − λ     −1 − λ 1 − 2λ  = 0 Expanded we get 7λ2 − 10λ + 1 = 0 (p) (q) The two roots of this polynomial and the natural frequencies are λ1 = 0.1082 λ2 = 1.320 λ1 24EA EA ω12 = = 2.5968 2 mL m L2 ω22 = --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The frequencies compared to exact values are  1.611 EA Exact FE → ω1 = L m  5.629 EA Exact FE → ω2 = L m λ2 24EA EA = 31.68 2 mL m L2   1.571 EA π EA ω1 = = 2L m L m   4.712 EA 3π EA ω2 = = 2L m L m → → (r) (s) As we increase the number of elements we get more frequencies and improve the accuracy of the lower frequencies. Note that this time we shall normalize by factoring c13 from the equation. This will provide a unit value for the normal mode displacement at the free end and will simplifying comparing the FEM modes with the exact analytical modes which all have unit values at the free end. Remember that the amplitude of the normal modes can be arbitrarily assigned. The first mode is found from these equations.         0 c12 4 1 2 −1 = − λ1 c13 0 1 2 −1 1       c12 /c13 4 1 2 −1 (t) c13 − 0.1082 → 1 1 2 −1 1      0 ϕ12 1.5672 −1.1082 = → 0 1 −1.1082 0.7836 From the first equation in Equation (s) we obtain 1.5672ϕ12 − 1.1082 = 0 → ϕ12 = 0.707 (u) The first normal mode is then {ϕ1 } = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  0.707 1  Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (v) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 564 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The second normal mode is found from these equations.         0 4 1 c22 2 −1 = − λ2 c23 0 1 2 −1 1         0 c22 /c23 4 1 2 −1 = − 1.320 → c23 0 1 1 2 −1 1      0 ϕ22 −3.28 −2.32 = → 0 1 −2.32 −1.64 (w) From the first equation in Equation (v) we obtain − 3.28ϕ22 − 2.32 = 0 → (x) ϕ22 = −0.707 The second normal mode is {ϕ2 } =  −0.707 1  (y) We compare the first two exact analytical normal modes with those obtained by the two element FEM solution. In Figure (e) we compare the first modes. 1.0 Exact 1st Mode 0.8 0.6 2 element FEM 0.4 0.2 0.2 0.4 0.6 0.8 1.0 0.8 1.0 Figure (e) In Figure (f) we compare the second modes. 1.0 Exact 2nd Mode 0.5 0.2 0.4 0.6 –0.5 2 element FEM mode –1.0 Figure (f) The complete solution is found by normal modes exactly as in the analytical solution. We just have approximate values for the modes and frequencies and a finite number of modes according to the number of nodes and elements. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 565 Structural Dynamics The solution in terms of the two normal modes is       u 2 (t) 1 1 {u(t)} = = (A1 sin ω1 t + B1 cos ω1 t) (z) + (A2 sin ω2 t + B2 cos ω2 t) u 3 (t) −1.414 1.414 The initial conditions are --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- u 2 (0) = u0 2 u̇ 2 (0) = 0 u̇ 3 (0) = 0 u 3 (0) = u 0 (aa) Lo and behold, we now have four constants to satisfy the four initial conditions. u 2 (0) = (A1 sin ω1 0 + B1 cos ω1 0)1 + (A2 sin ω2 0 + B2 cos ω2 0)1 = B1 + B2 = u0 2 u̇ 2 (0) = (A1 ω1 cos ω1 0 − B1 ω1 sin ω1 0)1 + (A2 ω2 cos ω2 0 − B2 ω2 sin ω2 0)1 = A1 ω1 + A2 ω2 = 0 u 3 (0) = (A1 sin ω1 0 + B1 cos ω1 0)1.414 (bb) − (A2 sin ω2 0 + B2 cos ω2 0)1.414 = 1.414B1 − 1.414B2 = u 0 u̇ 3 (0) = (A1 ω1 cos ω1 0 − B1 ω1 sin ω1 0)1.414 − (A2 ω2 cos ω2 0 − B2 ω2 sin ω2 0)1.414 = 1.414A1 ω1 − 1.414A2 ω2 = 0 Solving we get A1 = 0 A2 = 0 √ 1+ 2 u0 B1 = 4 √ 1− 2 u0 B2 = 4 (cc) The resulting motion at the nodes is √ √      1+ 2 1− 2 1 1 u 2 (t) √ √ = + u 0 cos ω1 t u 0 cos ω2 t {u(t)} = u 3 (t) 2 − 2 4 4  (dd) The distributed displacement is found by using the shape functions for each element. Increasing the number of elements increases the accuracy of both frequencies and modes. For free motion only a few modes are needed. ########### The finite element equations for free motion, Equation 15.3.32, are modified to add the applied load. [M]{r̈} + [K ]{r } = {F(t)} (15.3.33) The coefficient matrices are assembled from the element matrices. For element m the stiffness matrix is   EA 1 −1 (15.3.34) [km ] = L −1 1 and the mass matrix is {m m } = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS mL 6  2 1 1 2  (15.3.35) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 566 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Concentrated loads at nodes can be entered directly. Distributed applied loads must be converted to equivalent nodal loads. From Section 4.4 ⎡ s⎤   1− l ⎥ ⎢ {Fenl } = [n]T F(t)ds = ⎣ F(t)ds (15.3.36) s ⎦ l ########### Example 15.3.7 Problem: Find the response of the bar in Example 15.3.5 to a sinusoidal load in time. Solution: Use just one element. For just one element our FE equations become     EA 1 mL 2 1 0 + [M]{r̈} + [K ]{r } = ü 2 6 1 2 L −1 −1 1  0 u2  =  R1 F0 sin t  (a) EA mL u 2 = F0 sin t ü 2 + 3 L → ü 2 + 3EA 3F0 sin t u2 = 2 mL mL → ü 2 + ω2 u 2 = 3F0 sin t mL (b) Compare Equation (b) with Equation 15.2.41 with the periodic force applied as shown here F0 d 2w sin t (c) + ω2 w = dt 2 m With the same applied load and the same initial conditions it has the same solution as given in Equation 15.2.49, or    3F0 sin ωt (d) sin t − u 2 (t) = 2 (ω − 2 )m L ω ########### Clearly the same phenomenon of resonance is observed. Anytime an applied frequency is at or near any one of the natural frequencies of a structure there is danger of catastrophic resonance failure. This can be alleviated by adding damping to the system, but for now that is another subject. In our analytic solution we noted an infinite number of frequencies. In the FE solution the number of frequencies you obtain is equal to the number of unrestrained nodal displacements. The lowest frequency is found and the others are in ascending order. Those needing evidence of the validity of this approach can turn to the mathematical and engineering literature on the classical eigenvalue problem. A majority of problems in vibration are solved by finding the natural frequencies and then adjusting the mass and stiffness of the structure to avoid likely applied frequencies. Alternatively, operating limits can be enforced to ensure that no applied frequency is allowed that is at or near any natural frequency. Of course, non periodic applied forces also are important, for example, when an airplane enters a gust, performs a maneuver or, say, on landing. Then the mass and stiffness must be adjusted to provide acceptable displacements and stresses or operating constraints must be imposed to provide the acceptable responses. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- and so the partitioned equation for the motion is P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 567 Structural Dynamics 15.4 Torsional Vibration 15.4.1 Torsional Mass/Spring Systems Torsional motion can also be described in terms of the rotation of masses and the resistance of springs. Consider the shaft with masses as shown in Figure 15.4.1. The rotation is denoted by β(τ ) with units of radians or degrees, the shaft stiffness is denoted by a torsion spring constant kβ with units of moment/radian or degrees (Nmm/rad Nmm/deg,), and the polar mass moment of inertia I M of the disk with units of kilograms-millimeters squared (kg-mm2 ). kβ IM Τ(t) β(t) x Figure 15.4.1 From summation of moments about the x axis the equation of motion is I M β̈ + kβ β = M(t) → β̈ + ω2 β = T(t) IM (15.4.1) where ω2 = kβ IM (15.4.2) This equation is exactly the same form as Equation 15.2.3 for the linear mass/spring system. The solutions for free and forced motion are easily obtained by analogy. For free motion, that is, when T(t) = 0, the solution is β(t) = A sin ωt + B cos ωt (15.4.3) and initial conditions are applied in the same way. Multiple degrees of freedom torsional systems are also represented by the matrix equations of the form [I M ]{β̈} + [K β ]{β} = {0} (15.4.4) The natural frequencies are found from |[K β ] − ω2 [I M ]| = 0 (15.4.5) [K β ] − ωi2 [I M ] {C}i = 0 (15.4.6) and the modes from The complete solution is β(t) = n  i=1 (Ai sin ωi t + Bi cos ωi t)ϕi (15.4.7) where ϕi are the normal modes and the constants Ai and Bi are evaluated using the initial conditions. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 568 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Forced motion, that is, when {M(t)} = {0}, is defined by the equations [I M ]{β̈} + [K β ]{β} = {T(t)} (15.4.8) To find the solution the frequencies and normal modes are found as just discussed for the axial case and a solution is assumed of the form {β} = n  i=1 ξi (t){ϕi } (15.4.9) As in the case of axial vibration [K β ] − ωi2 [I M ] {ϕi } = 0 [K β ]{ϕi } = ωi2 [I M ]{ϕi } → (15.4.10) and the modes are orthogonal {ϕi }T [I M ]{ϕ j } = 0 i = j = IM j (15.4.11) i= j If we substitute Equation 15.4.9 into Equation 15.4.8 we get [I M ] n  i=1 ξ̈i (t){ϕi } + [K β ] n  i=1 ξi (t){ϕi } = {T(t)} Now use Equation 15.4.10 and pre multiply by ϕ j   n n   2 T {ϕ j } [I M ] ξi (t){ϕi } = {T(t)} ξ̈i (t){ϕi } + ωi [I M ] (15.4.12) (15.4.13) i=1 i=1 We then rearrange terms and take advantage of orthogonality to obtain I M j ξ̈ j + ω2j I M j ξ j = T j (t) ξ̈ j + ω2j ξ j = → T j (t) IM j (15.4.14) where T j (t) = {ϕ j }T {T(t)} (15.4.15) The solutions to these equations are obtained in the same way as for the linear mass spring systems. 15.4.2 Distributed Torsional Systems The differential equation for the torsional vibration of a uniform bar with a circular cross section is obtained from Equation 6.3.10 by including the rotational inertia as part of the applied load. GJ ∂ 2β ∂ 2β = −t(x) − I m ∂x2 ∂t 2 → GJ ∂ 2β ∂ 2β + I = −t(x) m ∂x2 ∂t 2 (15.4.16) where Im is the mass polar moment of inertia per unit length about the x axis. This is the same differential equation form as the axial case and all we learned there carries over with just a change in coefficients. As in the axial case the first step is to find the natural frequencies and this requires separation of variables. β(x, t) = B(x)T (t) (15.4.17) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 569 Structural Dynamics resulting in Im ω 2 d 2B B=0 + dx2 GJ T̈ + ω2 T = 0 (15.4.18) To this we add the mass matrix which by analogy to the axial case is   Im l m 2 1 [I m ] = 1 2 6 (15.4.20) The assembled matrix equations are of the form [I]{r̈} + [K]{r } = {F} (15.4.21) The element matrices are assembled exactly as presented in Section 15.3. Again it should be noted that the shape functions are not exact in this case since the inertia forces are a distributed, that is, not concentrated, load. 15.5 Vibration of Beams in Bending 15.5.1 Solutions of the Differential Equation For the free vibration of uniform beams we have EI ∂ 4v = f y (x, t) = −m v̈ ∂x4 EI → ∂ 4v + m v̈ = 0 ∂x4 (15.5.1) Separate variables v(x, t) = V (x)T (t) (15.5.2) to obtain EI d4V − mω2 V = 0 dx4 T̈ + ω2 T = 0 (15.5.3) d4V − α4 V = 0 dx4 (15.5.4) The first of these equations may be rewritten mω2 d4V V =0 − 4 dx EI → The general solution for V(x) is found by assuming an exponential form. V ≈ Ceax (15.5.5) Substituting this into Equation 15.5.4 gives us a 4 Ceax − α 4 Ceax = 0 → a4 = α4 (15.5.6) There are two pairs of roots. One pair ±α is real and the other ±iα is imaginary. Thus V (x) = C1 eαx + C2 e−αx + C3 eiαx C4 e−1αx Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (15.5.7) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The solution proceeds exactly as presented in Section 15.3 for the axial case. In a finite element solution the shape functions and element stiffness matrix are the same as used in Chapter 6, Section 6.6.     s s GJ 1 −1 [kβm ] = (15.4.19) [n] = 1 − lm lm lm −1 1 P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 570 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods which can be put in the more convenient form of V (x) = c sinh αx + d cosh αx + e sin αx + f cos αx (15.5.8) To find the natural frequencies substitute Equation 15.5.8 into the boundary conditions. An example will help. ########### Example 15.5.1 Problem: Find the natural frequencies and modes of the uniform simply supported beam shown here in Figure (a). L Figure (a) Solution: Insert Equation 15.5.8 into the boundary conditions to obtain the eigenvalue matrix. Find the frequencies and mode shapes. The boundary conditions are that the displacements and moments are zero at both ends. v(0, t) = V (0)T (t) = 0 EI d 2 V (0) ∂ 2 v(0, t) = EI T (t) = 0 ∂x2 dx2 v(L , t) = V (L)T (t) = 0 EI → → ∂ 2 v(L , t) d 2 V (L) = EI T (t) = 0 2 ∂x dx2 V (0) = 0 → d 2 V (0) =0 dx2 V (L) = 0 → (a) d 2 V (L) =0 dx2 Upon substitution V (0) = c sinh α0 + d cosh α0 + e sin α0 + f cos α0 = d + f = 0 d 2 V (0) = cα 2 sinh α0 + dα 2 cosh α0 − eα 2 sin α0 − f α 2 cos α0 = α 2 (d − f ) = 0 dx2 V (L) = c sinh αL + d cosh αL + e sin αL + f cos αL = 0 (b) d 2 V (L) = cα 2 sinh αL + dα 2 cosh αL − eα 2 sin αL − f α 2 cos αL = 0 dx2 When we have a set of homogeneous linear algebraic equations of the form [Ŵ]{C} = 0 (c) either {C} = {0} or |Ŵ| = 0 that is, the unknowns are zero or the determinant of the coefficients is zero. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (d) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 571 Structural Dynamics When we have an eigenvalue problem the trivial solution {C} = {0} would proclaim that we had no motion. But our experience says this is not so; thus, we look for the conditions where |Ŵ| = 0. We can immediately reduce the size of the set of equations by noting that from the first two equations d= f =0 (e) The remaining equations are c sinh αL + e sin αL = 0 (f) c sinh αL − e sin αL = 0 Adding the two we get c sinh αL = 0 c=0 → (g) which leaves us with e sin αL = 0 (h) and either e = 0, the trivial solution, or sin αL = 0 → αL = iπ → α= iπ L (i) where i is an integer. And there we have our natural frequencies.  2   4 mω2 EI iπ iπ → ω= α4 = = L EI L m (j) The mode shapes and the normal modes are V (x) = e sin iπ x L → ϕi = sin iπ x L (k) We know the shape of the mode but the amplitude is unknown so we arbitrarily assign e = 1 for plotting purposes. ########### For forced motion the equation of motion is EI ∂ 4v + m v̈ = f y (x, t) ∂x4 (15.5.9) Just as we did in the axial and torsional cases we assume a solution of the form v(x, t) = ∞  ξi (t)ϕi (x) (15.5.10) i=1 We insert this form in Equation 15.7.9 EI ∞ ∞  ∂4  ξi (t)ϕi (x) − m ξ̈i (t)ϕi (x) = − f y (x, t) 4 ∂ x i=1 i=1 (15.5.11) We can rearrange the terms. EI ∞  i=1 ξi ∞  ∂ 4 ϕi −m ξ̈ ϕi = − f y (x, t) 4 ∂x i=1 (15.5.12) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 572 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Now multiply all terms by ϕ j (x) and integrate all terms over the length of the bar.    ∞  d 4 ϕi −m ξi ϕ j EI ξ̈ ϕi = − f y (x, t) d x dx4 0 i=1 i=1  L  L  L ∞ ∞   d 4 ϕi d x − ϕ mϕ d x = − ξi ϕ j f y (x, t)d x EIϕ j = ξ̈ i j i dx4 0 0 0 i=1 i=1 L ∞  (15.5.13) The normal modes are solutions to the separated equation EI d4V + mω2 V = 0 dx4 (15.5.14) It follows that EI d 4 ϕi + mωi2 ϕi = 0 dx4 EI → d 4 ϕi = −mωi2 ϕi dx4 (15.5.15) 4 By replacing EI dd xϕ4i with −mωi2 ϕi Equation 15.5.13 becomes − ∞  i=1 ξi ωi2  L 0 ϕ j mϕi d x − ∞  i=1 ξ̈i  L ϕ j mϕi d x = 0  L ϕ j f y (x, t)d x (15.5.16) 0 We designate f y j (t) = −  L ϕ j f y (x, t)d x (15.5.17) 0 We note that from orthogonality --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`---  L 0 ϕ j mϕi d x = 0 i = j = Mj (15.5.18) i= j Finally the equation for forced motion may be written ξ̈ j + ω2j ξ = f y j (t) Mj (15.5.19) We now solve for ξ j (t) and insert it in the series of Equation 15.7.10. Once again an example will help. ########### Example 15.5.2 Problem: The simple supported beam in Example 15.5.1 is initially at rest and then is subjected to a dynamic lateral load as shown in Figure (a). f y (t) = f 0 sin t Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 573 Structural Dynamics f y (t ) L Figure (a) Solution: Find a solution based on a series of normal modes. The governing equation is ξ̈ j + ω2j ξ = f y j (t) Mj (b) where f y j (t) =  L  ϕ j f y (t)d x = 0 L sin 0 jπ x L (1 − cos jπ ) sin t f 0 sin td x = f 0 L jπ (c) When j is an even integer f y j (t) = 0. This is logical since there is symmetry of geometry, restraint, and loading about the mid length of the beam. All modes for j even are anti symmetric about the mid length. Thus f y j (t) = 2 f 0 L sin t jπ j = odd (d) j = even =0 We also note that Mj =  L ϕ 2j md x = 0  L m sin2 0 mL jπ xd x = L 2 (e) The solution to Equation (b) is ξ j (t) = A j sin ω j t + B j cos ω j t + ξjparticular (f) The particular solution for an equation of this form was found in Section 15.2, Equation 15.2.46. We just need to use the correct coefficients. ξjparticular = f y j (t) 4 f0 sin t sin t = ω2j − 2 M j jπ ω2j − 2 m j = odd (g) From Equation 15.5.10 v(x, t) = ∞  ξ j (t)ϕ j (x) = j=1 ∞  j=1   jπ 4 f0 x A j sin ω j t + B j cos ω j t + sin t sin L jπ (ω2j − 2 )m (h) Apply initial conditions v(0) = 0 v̇(0) = 0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (i) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 574 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods And things proceed just as they have done in other cases. ########### 15.5.2 Solutions Based on FEM For the finite element formulation of the beam vibration problem we use the same shape functions developed in Chapter 8, Section 8.2, Equation 8.2.5. The shape function matrix for element m is 1  3 l − 3lm s 2 + 2s 3 lm3 m lm3 s − 2lm2 s 2 + lm s 3 3lm s 2 − 2s 3 The stiffness matrix for element m was found to be ⎡ 12 6lm 2 E m Im ⎢ 6l 4l m m [km ] = 3 ⎢ ⎣ −12 −6lm lm 6lm 2lm2 −lm2 s 2 + lm s 3 ⎤ 6lm 2lm2 ⎥ ⎥ −6lm ⎦ 4lm2 −12 −6L 12 −6L If we use D’Alembert’s principle the equivalent nodal load for the mass term is ⎡ ⎤ v̈1  lm   lm ⎢ θ̈ ⎥ 1⎥ ⎢ [n]T m v̈ds = − [n]T [n]mds ⎢ ⎥ [n]T {FI }ds = − {F} = ⎣ v̈ 2⎦ 0 0 θ̈2  (15.5.20) (15.5.21) (15.5.22) We define a mass matrix for element m. [m m ] =  lm [n]T [n]mds (15.5.23) 0 Then the equation of motion for an element becomes [m m ]{r̈m } + [km ]{rm } = {0} (15.5.24) ⎡ (15.5.25) where ⎤ v̈n ⎢ θ̈ ⎥ ⎢ n ⎥ {r̈m } = ⎢ ⎥ ⎣ v̈n+1 ⎦ θ̈n+1 When evaluated the mass matrix becomes [m m ] =  lm 0 ⎡ 156 mlm ⎢ 22lm T ⎢ [n] [n]mds = 420 ⎣ 54 −13lm 22lm 4lm2 13lm −3lm2 54 13lm 156 −22lm ⎤ −13lm 2 ⎥ −3lm ⎥ −22lm ⎦ 4lm2 (15.5.26) These are assembled to get the global stiffness matrix [K ] and the global mass matrix [M] in the same way as the static case to form the global equations. [M]{r̈} + [K ]{r } = {0} Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS (15.5.27) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- [n] = P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 575 Structural Dynamics ########### Example 15.5.3 Problem: Let us find the first two natural frequencies of a uniform beam with fixed supports on both ends as shown in Figure (a) suing the FEM. x Figure (a) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Solution: Use two elements. We divide the beam into two elements with three nodes and assemble the global mass and stiffness matrices. The mass matrix is     ⎡ ⎤ L L 156 22 54 −13 0 0 ⎢ ⎥ 2 2 ⎢ ⎥    2  2   ⎢ ⎥ ⎢ ⎥ L L L L ⎢ 22 ⎥ 4 −3 13 0 0 ⎢ ⎥ 2 2 2 2 ⎢  ⎥   ⎢ L ⎥ L ⎢ ⎥ 312 0 54 −13 54 13 ⎥ ⎢ 2 2 ⎥ mL ⎢ ⎢ ⎥ (a) [M] =           840 ⎢ L L 2 L 2⎥ L 2 L ⎢ ⎥ −13 −3 −3 0 8 13 ⎢ ⎥ ⎢ ⎥ 2 2 2 2 2 ⎢ ⎥     ⎢ ⎥ L L ⎢ ⎥ 156 −22 0 0 54 13 ⎢ ⎥ 2 2 ⎢ ⎥ ⎢ ⎥         ⎣ L L 2 L 2 ⎦ L 0 0 −13 −3 4 −22 2 2 2 2 and the stiffness matrix is ⎡ 12 ⎢ ⎢ ⎢   ⎢ L ⎢6 ⎢ 2 ⎢ ⎢ ⎢ ⎢ −12 8EI ⎢ [K ] = 3 ⎢   L ⎢ L ⎢ ⎢6 ⎢ 2 ⎢ ⎢ ⎢ 0 ⎢ ⎢ ⎢ ⎣ 0   L −12 2  2   L L 4 −6 2 2   L 24 −6 2  2 L 2 0 2 6 0 0 −12 6   L 2   L 2  2 L 2 2 6 0 0 0 −12  2   L L −6 2 2   L 12 −6 2  2   L L 2 −6 2 2 8 The boundary conditions are 0 ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ 0 ⎥   ⎥ L ⎥ ⎥ 6 ⎥ 2 ⎥  2 ⎥ ⎥ L ⎥ 2 ⎥ ⎥ 2  ⎥ L ⎥ ⎥ −6 2 ⎥ ⎥  2 ⎥ ⎦ L 4 2 v1 = θ1 = v3 = θ3 = 0 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (b) (c) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 576 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods The partitioned equations reduce to just two. ⎤ ⎡ ⎡   24 312 0 8EI ⎣ mL ⎣  2 ⎦ v̈2 + 3 [M]{r̈} + [K ]{r } = L θ̈2 0 0 8 840 L 2 ⎤     0  2 ⎦ v2 0 = L θ2 0 8 2 (d) Assuming {r } = {rc }eiωt (e) and substituting ⎛⎡ ⎝⎣ 24 0 ⎤ ⎤⎞ ⎡     0 312 0 ω2 m L 4 ⎣  2 ⎦  2 ⎦⎠ vc2 0 = − L L θc2 0 8 0 8 840 · 8 · EI 2 2 (f) Let γ = ω2 m L 4 840 · 8 · EI The determinant of the coefficients is ⎡ ⎤ ⎡  24 0 312    2⎦ ⎣ ⎣ − γ L  0 8 0  2 ⎤  0  2 ⎦ L =0 8  2 This expands to (24 − 312γ ) (1 − γ ) 8 (g) (h)  2 L =0 2 (i) For the lowest frequency we get (24 − 312γ ) = 0 → γ = 1 = 0.0769 13 → ω12 = 516.92 EI m L4 (j) The second frequency is (1 − γ ) = 0 → γ =1 → ω22 = 6720 EI m L4 (k) Thus  EI ω1 = 22.74 m L4 ω2 = 81.98  EI m L4 The exact values from a differential equation solution   EI EI ω = 61.67 ω1 = 22.37 2 m L4 m L4 (l) (m) Not bad for such a crude approximation. ########### --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 577 Structural Dynamics 15.6 The Finite Element Method for all Elastic Structures The element FE equations for the vibration of all elastic structures are of the form [m m ]{r̈m } + [km ]{rm } = {Fm } (15.6.1) These are assembled into global equations of the form [M]{r̈} + [K ]{r } = {F} (15.6.2) This is exactly the same form as the equations for multi degree of freedom mass/spring systems described in detail in Section 15.2. The natural frequencies ωi and the corresponding modes {ϕ}i are found from the homogeneous equations. [M]{r̈} + [K ]{r } = {0} (15.6.3) We assume harmonic motion (a fast way to separate variables) of the form {r (t)} = {rc }eiωt (15.6.4) ([K ] − ω2 [M]){rc }eiωt = {0} (15.6.5) and obtain The frequencies are found by setting the determinant of the coefficients to zero. |[K ] − ω2 [M]| = 0 (15.6.6) By expanding this determinant we get a polynomial in ω2 and the roots of this polynomial are the natural frequencies. The natural frequencies are reinserted into the homogeneous equations and the normal modes are found. All that is different for various structures, such as, beams, shafts, plates, shells, and 3D solid elastic bodies are the specific forms of the mass and stiffness matrices. Solutions for specific initial conditions and applied forces are found by assuming a series solution based on the normal modes (see Equation 15.2.54) {r } = n  ξi (t){ϕ}i (15.6.7) i=1 and proceeding as presented in Equations 15.2.55-61. In all cases resonance is encountered whenever an applied frequency is at or near a natural frequency. 15.7 Addition of Damping The motion of structures as described in the previous sections of this chapter continue indefinitely once started by imposing initial conditions and/or applied loads. In practice this motion often is modified or even dies out as a result of forces that resist the motion. These are called the damping forces. The actual physical phenomena that cause damping are complex and generally hard to model mathematically. The actual damping mechanisms are beyond the space available and the scope of this text; however, one idealization of the true damping forces that has a certain amount of validity and mathematical simplicity is called viscous damping. This type generates a force proportional to the velocity and in the opposite direction of the velocity. An additional device called a viscous damper is often added to mass spring systems. It is shown schematically in Figure 15.7.1. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 578 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods y k υ(t) x M F(t) Figure 15.7.1 This device exerts a force opposing the motion proportional to the velocity, that is, FD = −μv̇ (15.7.1) where μ is the damping coefficient and has the units of kilograms per second (kg/s) or Newton-seconds per millimeter (Ns/mm). Add this to the free body diagram as shown in Figure 15.7.2. k υ(t) FI F(t) FD Figure 15.7.2 The equation of motion becomes F(t) + FI + FD − kv = 0 → F(t) − M v̈ − μv̇ − kv = 0 (15.7.2) → M v̈ + μv̇ + kv = F(t) F(t) → v̈ + β v̇ + ω2 v = M where β= μ M ω2 = k M (15.7.3) To solve this equation we try v ≈ Ceλt (15.7.4) (λ2 + βλ + ω2 )Ceλt = 0 (15.7.5) λ2 + βλ + ω2 = 0 (15.7.6) The result is λt Using the quadratic formula we obtain  β β2 − ω2 λ1 = − + 2 4 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS λ2 = − β − 2  β2 − ω2 4 (15.7.7) Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Since Ce = 0 it follows that P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 579 Structural Dynamics When β 2 ≥ 4ω2 (15.7.8) the roots are real and negative. The solution to the homogenous equation is then v = Aeλ1 t + Beλ2 t Both the terms represent an exponential decay with time. When β 2 ≤ 4ω2 both roots are complex λ1 = − β + i 2 λ2 = − β + i 2 =  ω2 − β2 4 (15.7.10) and the homogenous solution for free motion may be written as β β v(t) = e− 2 t (aeit + be−it ) = e− 2 t (A sin t + B cos t) (15.7.11) This is an oscillatory motion of frequency  multiplied by an exponential decay. ########### Example 15.7.1 Problem: A single degree of freedom mass/spring system with viscous damping has the following properties. M = 2 kg k = 32 N mm μ = 0.8 kg- sec mm (a) From its static equilibrium position it is given an initial displacement of 4 mm, no initial velocity, and released. Find the natural frequency and resulting motion. Solution: Use Equations 15.7.11 to find the motion. The natural frequency for the undamped system, as found in Example 15.2.1, is ω=  k = M  32 = 4 rad/s 2 → f = ω 4 = = 0.6366 Hz 2π 2π (b) With damping it is = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  ω2 − √ √ β2 = 16 − 0.04 = 15.96= 3.995 rad/s 4 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (c) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- (15.7.9) P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 580 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Apply initial conditions to Equation 15.7.11 we get β v(t) = e− 2 t (A sin t + B cos t) → v(0) = e0 (A sin 0 + B cos 0) = 4 → B=4 β −βt e 2 =0 2 β → v̇(0) = e0 (A cos 0 − B sin 0) − (A sin 0 + B cos 0) e0 = 0 2 B 4 Bβ = 0 → A = β = 0.4 = 0.8 → A − 2 2 2 β v̇(t) = e− 2 t (A cos t − B sin t) − (A sin t + B cos t) (d) Thus β v(t) = e− 2 t (A sin t + B cos t) = e−0.2t (0.8 sin 3.995 t + 4 cos 3.995 t) (e) A plot of the first few cycles of the motion is given in Figure (a). 4 3 2 1 υ(t) t 1 2 3 4 5 -1 -2 -3 Figure (a) ########### Now let us consider an oscillatory applied force F(t) = F0 sin t (15.7.12) that is, v̈ + β v̇ + ω2 v = F0 sin t F(t) = M M (15.7.13) A particular solution has been found by assuming the following form for the trial solution v p = P(sin t − ψ) (15.7.14) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas July 4, 2011 7:22 Printer Name: Yet to Come 581 Structural Dynamics When this is substituted into Equation 15.7.13 and the results collected we find that F  2 2 P = "  1− 2 ω 2 2 β  + ω2 β 2 tan ψ =  ω 2   1− 2 ω F= F0 ω2 M (15.7.15) We note that the homogenous part of the solution dies out in time for all cases of positive damping leaving only the particular solution, which for this reason is called the steady-sate solution. It is instructive to plot the amplitude of the steady-state solution, or more precisely, the absolute value of FP versus ω , for several values of damping β. The effects of damping are very clear. The resonance peaks are now finite. 3.0 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 2.5 P F Increasing β 2.0 1.5 1.0 0.5 0.5 1.0 1.5 2.0 2.5 3.0 Ω ω Figure 15.7.3 This can be expanded to multi degree of freedom systems forming the equations [M]{v̈} + [C]{v̇} + [K ]{v} = {F(t)} (15.7.16) Again the natural frequencies and normal modes are found and a solution based on a series of normal modes is sought. As noted, the dynamics of structures can be formed into equations that mathematically are the same as mass/spring/damper systems via FEM and the same solution methods used. The finite element equations are [M]{r̈} + [C]{ṙ } + [K ]{r } = {F(t)} (15.7.17) where [C] is the damping matrix. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-15 P2: ABC JWST071-Waas 582 15.8 July 4, 2011 7:22 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Summary and Conclusions --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- A brief introduction to the subject of structural dynamics has been completed. It places much attention on discrete mass spring systems because this is the clearest and simplest way to explain and demonstrate the process of solution. The equations of motion of continuous structures are first solved for the natural frequencies and modes. Then a solution is sought by forming a series which is the sum of the natural modes. In both classical and FEM analysis this reduces the problem to solving equations which are the same mathematically as mass/spring and mass/spring/damper systems. The addition of inertia forces introduces time as a variable and requires a new process called separation of variables. Part of the problem reduces to solving an eigenvalue equation. This solution method is familiar from the study of buckling. Advanced treatment of structural dynamics can be found in many modern textbooks, and the reader is recommended to consult Craig (2006)) for a broader coverage of the subject. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas July 4, 2011 11:7 Printer Name: Yet to Come 16 Evolution in the (Intelligent) Design and Analysis of Structural Members Introduction In this chapter we shall follow a series of steps taken in designing and analyzing, then repeated redesigning and analyzing, until a final design and analysis of a part is achieved for two specific cases. In each case the goal will be to reduce high stresses and deformations as appropriate, always keeping in mind the reduction of weight or amount of material. A commercial FEM computer program will be used. The subject of design as distinct from analysis was introduced briefly in Chapter 4, Section 4.9, for an axially loaded bar, and in Section 4.10 for a pin jointed truss. Then in Chapter 7, Section 7.10, the design of beams was considered. As is noted in those sections, in analysis we specify exactly the geometry, materials, loads, and restraints and ask for the displacements, strains, and stresses. We wish to verify that the structure meets minimum requirements on load carrying ability. In design, we are given a task to perform by a structural member, and we inquire what geometry and materials should be used to achieve a certain acceptable value of stress and displacement. The loads and restraints are determined as suitable to the task. Any extensive discussion of design is well beyond the scope of this text, and the interested student can follow on by taking a modern course in structural design optimization that introduces the student to the ideas of finding solutions to problems with constraints. Since we have developed analysis tools, in particular the finite element method, to include two and three dimensional structures, we take this opportunity to extend our understanding of design methods. Again as previously noted, design is a much broader subject than that just stated. In addition to the geometric restraints we may have restraints on weight, cost, and ease of manufacture. Safety and margins of safety are ever present. Aesthetics are an additional consideration in many cases, such as the body panels on an automobile and the interior of an airliner. Because of the broad nature of design it is largely a team effort. We cannot begin to study all of these concerns so we shall limit ourselves to geometry, geometric restraints, materials, and weight for a given state of loading. These are the immediate concerns of the structural analyst member of the team. Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 16.1 P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas 584 July 4, 2011 11:7 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods For an axially loaded slender bar the geometry consists of the length and the cross sectional area. The exact shape of the cross section might be determined by considerations of manufacturing, fasteners and restraints, points or regions of load application, or other considerations. We shall restrict our discussion to the role of modifying the geometry and material to improve the design. The first example is a truss member. It will be analyzed first by 1D slender bar theory, then by 2D plane stress theory, and finally by 3D solid body theory. There will follow a plane stress example. These examples are limited to just a few of the cases considered in the previous chapters; however, they illustrate the process of improving design through analysis. The same process can be applied to many other examples. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- 16.2 Evolution of a Truss Member Suppose a pin jointed truss has several members. In the first analysis of a truss the axial load in each member is found; however, the geometry of the joints where the members are pinned together is not modeled. At first the truss will be analyzed by the methods used in Chapter 3, Section 3.10 and Chapter 4, Section 4.4. One member of the truss has a known tensile load on it and is isolated for further design changes and analysis. The member we choose is 1000 mm long and has an axial load of 100000 N. We shall now account for the way the truss member is pinned to the base or to other members and take the truss member through several design changes as shown in Figure 16.2.1. Figure 16.2.1 16.2.1 Step 1. Slender Bar Analysis At the top in Figure 16.2.1 we represent a slender bar with just a line. This line is the centroidal axis of the axially loaded member. In slender bar analysis the length and the cross sectional area are the only geometric quantities of interest. The actual shape of the cross section and any details on how this member is attached to a base or to other truss members are ignored. For the given load we shall assign a cross section area to achieve a desired stress level and a satisfactory displacement. We shall give it an initial cross sectional area of 1000 mm2 without specifying a shape for the cross section. The load is a total force of 100000 N. From slender bar theory, as explained in Chapter 4, the Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas July 4, 2011 11:7 Printer Name: Yet to Come 585 Evolution in the (Intelligent) Design and Analysis of Structural Members stress is, σx = 100000 = 100 N /mm 2 or Mpa 1000 (16.2.1) If the bar is made of steel with a Young’s modulus of 206800 N/mm2 the total elongation of the bar is u (L) = 100000 · 1000 FL = = 0.484 mm AE 1000 · 206800 (16.2.2) If we choose a steel alloy with a yield stress of approximately 350 Mpa the factor of safety based on the yield stress is 350 = 3.5 (16.2.3) 100 At this point the large factor of safety is to account in part for the fact that we don’t know how the part is attached to a base or to other truss members. We do not need to know the actual shape of the cross section, since the area of the cross section is all we need to know in this analysis. F.S. = 16.2.2 Step 2. Rectangular Bar—Plane Stress FEM The second one down in Figure 16.2.1 is a member with a specific cross section. In this case a rectangular section with a height of 100 mm and a depth of 10 mm. It is modeled as a surface; the thickness is entered as a scalar quantity in the plane stress equations. If a uniform stress is applied on the edges at each end, from the plane stress equations, we obtain a constant uniform stress throughout the member, and the same displacement as in Step 1. We also can solve this by FEA using shell elements as shown in Figure 16.2.2, but this only confirms what we already know, that is, it will return a uniform stress of 100 MPa Figure 16.2.2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas 586 July 4, 2011 11:7 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods and a total elongation of 0.484 mm. Note that in this analysis the displacement is measured from the centroid of the model on each end face and is 0.242 mm in each of the positive and negative directions for a total elongation of 0.484. 16.2.3 Step 3. Rectangular Bar with Pin Holes—Plane Stress Analysis The third one down in Figure 16.2.1 is more representative of an actual member that is attached to a base and to other members with pins, that is, it is a truly pin-jointed truss member. Material is added at the ends so that the distance between the centers of the holes is 1000 mm. This one has no neat analytical solution, so we need to look at FEA. We need to simulate the application of the load which in reality is the pin in contact with the edge of the hole. In this case it is done with a rigid element bearing against the side of the hole. This is an approximation since the pin in reality is a flexible body and will deform to some extent under contact with the edge of the hole thus modifying the distribution of load. The rigid element does give a reasonable first approximation. We show the results in Figure 16.2.3 with the stress contours shown on the displaced model. The displacements are greatly exaggerated to be made more visible. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure 16.2.3 In Figure 16.2.4 we present a close up of the right end with the stress contours shown on the undeformed model. The result is a large increase in the maximum stress due to the stress concentration effect at the Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas July 4, 2011 11:7 Printer Name: Yet to Come Evolution in the (Intelligent) Design and Analysis of Structural Members 587 Figure 16.2.4 hole. The stress of 657 MPa exceeds the yield stress by a large factor and is unacceptable. Away from the holes the stress is the same as in our simple models. This is an excellent example of St. Venant’s principle that was discussed earlier. The total displacement of 0.604 mm results from the elongation of the region in the vicinity of the holes. The elongation of the mid section is the same as that obtained from the simple analysis. 16.2.4 Step 4. Rectangular Bar with Pin Holes—Solid Body Analysis Now we shall repeat the solution in Step 3 using a solid body and solid elements. The model is meshed using a nominal element length of 2.5 mm and a solution is found. This is essentially a repeat of the plane stress solution with small differences in the maximum stress due to modeling differences. If a greater number of elements were used in both the plane stress and solid cases the results would converge. The results show that something dramatic is happening in the vicinity of the holes. And it confirms that over most of the length the stress is uniform at 100 MPa. This too is an excellent example of St. Venant’s principle. For a better understanding, let us take a closer look at a contour plot of the region around the hole as shown in Figure 16.2.5. In this case we show the stress contours on a deformed model. The deformation is greatly exaggerated of course. We note that in both the plane stress and solid cases, when loaded by the pins bearing on the hole surfaces, we get much higher stress levels in the vicinity of the hole but essentially the same stress as in --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas 588 July 4, 2011 11:7 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure 16.2.5 the first two cases in the middle portion. The values of the several stress components and displacement are Von Mises Maximum Principle σx τxy Maximum shear Displancement in x = = = = = = 706 Mpa 761 753 245 357 0.1882 (total elongation) (16.2.4) The maximum von Mises stress reported at the top and bottom of the hole is 706 MPa. This is much higher than the yield stress of 350 MPa and would produce a negative factor of safety. We need to modify the design. 16.2.5 Step 5. Add Material Around the Hole—Solid Element Analysis We note that the stresses are too high at the hole and we must change the design to reduce the highest stresses to prevent yielding. The first attempt is to place more material in the vicinity of the hole by enlarging the radius surrounding the hole as shown in the fourth figure down and in Figure 16.2.6. The outside diameter is increased to 30 mm. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas July 4, 2011 11:7 Printer Name: Yet to Come Evolution in the (Intelligent) Design and Analysis of Structural Members 589 Figure 16.2.6 This is solved with the same density of mesh and the same loading as in the previous case. The results are shown in Figure 16.2.7. Figure 16.2.7 Von Mises = 590 Mpa Maximum Principle = 625 = 619 σx (16.2.5) = 182 τxy Maximum shear = 310 Displancement in x = 0.1262 (total elongation) The maximum von Mises stress is reduced to 590 MPa. That is still too high. More changes are needed. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- The values of the several stress components and displacement are P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas 590 16.2.6 July 4, 2011 11:7 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Step 6. Bosses Added—Solid Element Analysis We shall add bosses to the material surrounding the hole as shown in the fifth figure and in Figure 16.2.8. The thickness in this region is now 12 mm. Figure 16.2.8 This is solved with the same density of mesh and the same loading as the previous case. The results are shown in Figure 16.2.9 on an undeformed model. Figure 16.2.9 The values of the several stress components and displacement are Von Mises Maximum Principle σx τxy Maximum shear Displancement in x = = = = = = 221 Mpa 246 244 78 119 0.1030 (total elongation) (16.2.6) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas July 4, 2011 11:7 Printer Name: Yet to Come 591 Evolution in the (Intelligent) Design and Analysis of Structural Members The maximum von Mises stress is now 221 MPa. This is well within the range providing a safety factor of 350 = 1.58 221 (16.2.7) But wait! This meets the design requirement on stress and displacement, but is it an efficient design? Much of the material is operating at a safety factor of 3.5. This may be unnecessarily high, producing a heavier member than desired. 16.2.7 Step 7. Reducing the Weight—Solid Element Analysis Finally, we can reduce the amount of material and reduce the weight by removing material in the regions of low stress as shown in the sixth figure and in Figure 16.2.10. We can remove some of the material in the mid region by narrowing the height of the mid cross section by half. Figure 16.2.10 This should provide a stress of 200 Mpa in the mid region and a safety factor of. F.S. = 350 = 1.75 200 (16.2.8) With that change a contour plot of the von Mises stresses is given in Figure 16.2.11. The region of highest stress has now shifted to where the rectangular middle bar joins the curved boss. The stress values are now Von Mises Maximum Principle σx τxy Maximum shear Displancement in x = = = = = = 240 Mpa 257 225 88 122 0.1882 (16.2.9) This provides a safety factor based on the von Mises stresses of F.S. = 350 = 1.45 240 (16.2.10) Further changes, such as a larger fillet between the mid bar and the boss, will refine the design further. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- F.S. = P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas July 4, 2011 592 11:7 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods Figure 16.2.11 16.2.8 Step 8. Buckling Analysis Under certain conditions the loading on the truss member may be reversed subjecting the member to a compressive load of 10000 N. Can this member withstand that load without buckling? First use slender column theory. In the yx plane it acts as a column simply supported on both ends; in the zx plane it approximates a column with both ends fixed. From Example 8.3.1, for the simply supported case, Pcr = π2 π2 5 · 103 = 21261 N E Izz = 206800 2 2 L 200 12 (16.2.11) From Example 8.3.3, for the fixed end case, Pcr = 4π 2 4π 2 10 · 53 = 21261 N E Iyy = 206800 2 2 L 200 12 (16.2.12) Coincidentally, both buckling loads turn out to have the same magnitude, and both are sufficiently high to not warrant a FEA for buckling. 16.3 Evolution of a Plate with a Hole—Plane Stress Consider the plate in Example 12.5.1. We found the maximum von Mises stress to be 354 N/mm2 at the upper and lower diameter of the hole. This is the classical example of stress concentration. It is 3.54 times the value of the maximum stress if no hole were present. A typical solution is to reinforce the plate around the hole. In Figure 11.2.1 we take the surface which represents the plate and split into a ring --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas July 4, 2011 11:7 Printer Name: Yet to Come Evolution in the (Intelligent) Design and Analysis of Structural Members 593 Y Y Z X X Figure 16.3.1 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- surface around the hole with a radius of 200 mm – the hole has a radius of 150 mm. For that surface we specify a thickness of 30 mm while the rest of the plate remains at 10 mm. We repeat the analysis with the new geometry. A contour plot of the results is shown in Figure 16.3.2. The maximum stress is now only 116 N/mm2 . Figure 16.3.2 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-16 P2: ABC JWST071-Waas 594 July 4, 2011 11:7 Printer Name: Yet to Come Analysis of Structures: An Introduction Including Numerical Methods 16.4 Materials in Design 16.5 Summary and Conclusions Our purpose in this chapter is to show how increased analysis capability inherent in the finite element method is used to refine the design of a structural member. We do not pretend that the steps we have taken achieve the most efficient design. Prior to the development of FEA sometimes extreme simplifications of the geometry, restraints, and loadings were necessary to obtain solutions for stress and displacement. Now such simplifications are important in selecting initials values of geometry but are superceded by the more exact methods of analysis provided by FEA and modern computing power. As shown here structural parts often start out with simple geometry, idealized restraints, and simple loading using approximate theories, such as σx = F A (16.5.1) was used to start the design of the truss member. Then more detailed geometry, restraints, and loading are added requiring more capable analysis equations, such as FEA. These same ideas can be used in the evolution of a design to meet buckling and vibration criteria. Start with the simple geometry and material of a structure that meets buckling or vibration criteria and make modification to improve the design. Often a new product is the derivative of a previous design and the evolutionary process starts with the previous product. The Boeing 787, mentioned in Chapter 13, Section 13.1, has a shape and form determined by aerodynamics that is very much like previous airliners, its engines are moderately improved versions of previous ones; however, it structure is truly a quantum jump in performance. These new materials are the future of design. Your world of opportunity as a structural analyst is beyond description. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- In the last century materials used for structures generally were homogeneous and isotropic. In high stress devices and machines, such as aircraft and automobiles, the materials were dominantly those listed in Chapter 3, Section 3.7, along with other metals. In less high stress areas, isotropic molded plastics were and are used. Orthotropic materials such as plywood, layered metal plates with honeycomb cores, and layered fiber glass were used in selected applications. Then toward the end of the twentieth century new materials became available. Foremost among these for high technical application are fiber reinforced composite laminates. We introduced fiber reinforced laminate material properties in Chapter 3, Section 3.8, and used them in structural applications in Chapter 13. The significance of such laminates in design is truly immense. In the past, we accepted certain material properties and built the design around certain geometry consistent with the necessary boundary conditions, that is, restraints and applied loads. We can now design a material that best meets the requirements. Thus in the examples given earlier in this chapter we can both modify geometry and materials to best meet the design requirements. No longer must the material be homogenous nor isotropic. The challenge to do this with classical analysis tools, that is direct solutions of the differential equations would be an insurmountable task, except for trivial configurations and trivial cases of external loading; however, the finite element method is clearly adaptable to the task and will be the analysis method of choice now and in the future. P1: TIX/XYZ P2: ABC JWST071-App01 JWST071-Waas July 4, 2011 9:8 Printer Name: Yet to Come Appendix A Matrix Definitions and Operations A.1 Introduction A brief review of basic matrix definitions and operation is given here. The notation is consistent with that in the main text. A.2 Matrix Definitions A rectangular matrix of order m × n is an array of quantities in m rows and n columns as follows: a11 ⎢ a21 ⎢ ⎢ . [a] = ⎢ ⎢ . ⎢ ⎣ . am1 a12 a22 . . . am2 ⎤ . . . a1n . . . a2n ⎥ ⎥ ... . ⎥ ⎥ ... . ⎥ ⎥ ... . ⎦ . . . amn (A.2.1) An element in the mth row and nth column is represented by the notation amn . A set of square brackets [ ] will denote a rectangular matrix of any order. For convenience we shall define some special cases and use some special notation as explained in the following paragraphs. A column matrix has m rows and 1 column, or order m × 1. Curved braces { } are used to designate a column matrix. The element in the mth row is designated am . For example, if a1 = 1, a2 = 3, a3 = 2, the matrix is of order 3 × 1 and given by ⎡ ⎤ ⎡ ⎤ a1 1 {a} = ⎣ a2 ⎦ = ⎣ 3 ⎦ a3 2 (A.2.2) A row matrix has 1 row and n columns, for example,  [a] = a1 a2  a3 = 1 3 2 (A.2.3) and the element of the nth column is designated an . Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ⎡ July 4, 2011 9:8 Printer Name: Yet to Come 596 Appendix A: Matrix Definitions and Operations A square matrix has the same number of rows and columns, or m = n. For example a square matrix of order 3 × 3 might look like this ⎡ ⎤ 1 3 2 (A.2.4) [a] = ⎣4 8 1⎦ 7 5 2 A symmetrical matrix is a square matrix that has exactly the same value when rows and columns are interchanged, or amn = anm (A.2.5) For example, ⎡ 1 [a] = ⎣3 2 3 8 1 ⎤ 2 1⎦ 2 (A.2.6) A diagonal matrix is a square matrix in which all elements are zero except those on the principal diagonal, or if m = n amn = 0, (A.2.7) For example, ⎡ 1 [a] = ⎣0 0 0 8 0 ⎤ 0 0⎦ 2 (A.2.8) A diagonal matrix is always symmetrical. The identity matrix is denoted by [I] and is a special case of the diagonal matrix for which amn = 1 when m = n (A.2.9) For example ⎡ 1 [I ] = ⎣0 0 0 1 0 ⎤ 0 0⎦ 1 (A.2.10) It is also frequently called the unit matrix. A single element matrix is a scalar and may be written with or without brackets. The determinant of [a] is formed from the elements of a square matrix and is designated |a|. The transpose of a matrix is formed by interchanging rows and columns and is denoted by the superscript T. The elements of the transpose of a matrix are found by setting (amn )T = anm (A.2.11) For example, ⎡ ⎤ 1 2 [a] = ⎣7 9⎦ 6 3 [a]T = 1 2 7 9 6 3 (A.2.12) The transpose of a row matrix is a column matrix and vice versa. A symmetrical matrix is identical to its transpose. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ P2: ABC JWST071-App01 JWST071-Waas P1: TIX/XYZ P2: ABC JWST071-App01 JWST071-Waas July 4, 2011 9:8 Printer Name: Yet to Come 597 Appendix A: Matrix Definitions and Operations A.3 Matrix Algebra We now define certain rules of matrix mathematics that make it especially useful. 1. Equality. Two matrices are equal if they are of the same order and all corresponding elements are equal. That is, amn = bmn if [a] = [b] (A.3.1) 2. Addition and subtraction. Matrices must be of the same order to add and subtract. Addition is performed by adding corresponding elements and subtraction by subtracting corresponding elements. That is, where [a] ± [b] = [c] cmn = amn ± bmn (A.3.2) For example, ⎡ 2 ⎣3 1 ⎤ ⎡ 1 7 1 6 9⎦ + ⎣1 −4 0 0 ⎤ ⎡ 6 7 3 7 −2 0⎦ = ⎣4 4 0 1 1 −4 ⎤ 14 9⎦ 1 (A.3.3) 3. Multiplication by a scalar. Any matrix may be multiplied by a scalar by multiplying each element by the scalar. That is, a [b] = [c] where cmn = abmn (A.3.4) 4. Multiplication. We define multiplication of two matrices, say, [a] and [b], provided certain conditions exist. The number of columns in [a] must equal the number of rows in [b]. Each element in the product [c] is obtained by multiplying the elements of the corresponding row in [a] by the elements of the corresponding column in [b] and adding the results according to the rule For example, ⎡ ⎤ 3 2 1 2 ⎣1 1⎦ 6 −3 7 −1 amn bnk ⎡ ⎤ ⎡ 3·1+2·6 3·2−2·3 3·1+2·1 15 1 = ⎣1 · 1 + 1 · 6 1 · 2 − 1 · 3 1 · 1 + 1 · 1⎦ = ⎣ 7 1 7·1−1·6 7·2+1·3 7·1−1·1 1 (A.3.5) ⎤ 0 5 −1 2⎦ 17 6 (A.3.6) likewise, 1 6 ⎡ 3 2 1 ⎣ 1 −3 1 7 ⎤ 2 12 3 1·3+2·1+1·7 1·2+2·1−1·1 1⎦ = = 22 8 6·3−3·1+1·7 6·2−3·1−1·1 −1 (A.3.7) Note the order of the product in each case. Matrix multiplication is associative, thus, [a] ([b] [c]) = ([a] [b]) [c] (A.3.8) [a] ([b] + [c]) = [a] [b] + [a] [c] (A.3.9) distributive, thus, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- cmk = P1: TIX/XYZ P2: ABC JWST071-App01 JWST071-Waas July 4, 2011 9:8 Printer Name: Yet to Come 598 Appendix A: Matrix Definitions and Operations but, in general, not commutative, thus, (A.3.10) [a] [b] = [b] [a] 5. Inversion. Division, as such, is not defined for matrices but is replaced by something called inversion. We denote the inverted matrix with the symbolic form [a]−1 , and it is defined so that [a]−1 [a] = [I ] (A.3.11) The elements of an inverse matrix can be obtained algebraically. In practice this is seldom done. Instead, the inverse is found numerically. Numerically, inverting matrices, by developing special algorithms, is a specialized subject which is becoming increasingly important with the speeding up of calculations. A.4 Partitioned Matrices --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- A useful operation is the partitioning into submatrices. These submatrices may be treated as elements of the parent matrix and manipulated by the rules just reviewed. For example, ⎡ 5 9 2 9 2 8 4 7 6 ⎤ 3 8⎦ = 5 A11 = 7 3 5 0 9 2  A21 = 1 2 8  7 [a] = ⎣ 3 1    A11  A21 A12 A22  (A.4.1) where   A.5  A12 = 4 3 7 8  A22 = 6 (A.4.2) 5 Differentiating and Integrating a Matrix Differentiate each element in the conventional manner. For example, if ⎡ x [a] = ⎣x 2 3x x2 x4 2x ⎤ 3x 2x ⎦ x3 then ⎡ 1 d [a] = ⎣2x dx 3 2x 4x 3 2 ⎤ 3 2 ⎦ 3x 2 (A.5.1) Integrate each element in the conventional manner. For example, if ⎡ 1 [a] = ⎣2x 3 2x 4x 3 2 ⎤ 3 2 ⎦ 3x 2 then  ⎡ x [a]d x = ⎣ x 2 3x x2 x4 2x ⎤ 3x 2x ⎦ + [C] x3 (A.5.2) where [C] are the constants of integration. For definite integrals each term is evaluated for the limits of integration present. Below is a summary of useful relations for following the main text, but the interested reader is alerted to specialized texts on the subject of matrix algebra for wider coverage. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-App01 JWST071-Waas July 4, 2011 9:8 Printer Name: Yet to Come 599 Appendix A: Matrix Definitions and Operations A.6 Summary of Useful Matrix Relations [a] [I ] = [I ] [a] = [a] [a] ([b] + [c]) = [a] [b] + [a] [c] a ([b] + [c]) = a [b] + a [c] [a] ([b] [c]) = ([a] [b]) [c] = [a] [b] [c] [a] + ([b] + [c]) = ([a] + [b]) + [c] = [a] + [b] + [c] [a] + [b] = [b] + [a] [a] [b] = [b] [a] ([a] [b])T = [b]T [a]T --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ([a] [b])−1 = [b]−1 [a]−1  T −1  −1 T = [a] [a] Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-App01 JWST071-Waas --`,,,,``,``,,`` Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS July 4, 2011 9:8 Printer Name: Yet to Come Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas July 4, 2011 9:9 Printer Name: Yet to Come Area Properties of Cross Sections B.1 Introduction The area, the centroid of area, and the area moments of inertia of the cross sections are needed in slender bar calculations for stress and deflection. To simplify the problem we place the x axis so that it coincides with the loci of centroids of all cross sections of the bar. In our examples the cross sections lie in the yz plane. Furthermore, for beam bending analysis in these chapters we orient the y and z axes so that they are principal axes of inertia of the cross section area. This simplifies the equations for stress and displacement. Just what this means is explained in the following sections. B.2 Centroids of Cross Sections Consider a cross section with a general shape such as shown in Figure B.2.1 with the x axis normal to the cross section. y z Figure B.2.1 The x axis is a centroidal axis if  ydA = 0 A  zdA = 0 (B.2.1) A Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Appendix B P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas July 4, 2011 9:9 Printer Name: Yet to Come 602 Appendix B: Area Properties of Cross Sections If the cross sectional is symmetrical the centroid is easily found since it will always lie on the axis of symmetry. For sections with double symmetry, that is, symmetry about both the y and z axes, such as those sections in Figure B.2.2, the location is obvious. y y y z z z (a) (b) (c) Figure B.2.2 For sections with symmetry about just one axis we know the centroid lies on that axis but we must locate just where on that axis. For the sections in Figure B.2.3 we use the formulae   ydA zdA A ȳ =  dA A A z̄ =  (B.2.2) dA A as appropriate. y y y z z z z y z (a) (b) (c) Figure B.2.3 When an area can be divided into sub areas with simple geometry so that the centroid of the sub area is easily identified the process of finding the centroid of the original area is simplified to   z s As ys As (B.2.3) z̄ =  ȳ =  As As --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas July 4, 2011 9:9 Printer Name: Yet to Come 603 Appendix B: Area Properties of Cross Sections where ys and zs represent the distances from base axes to the centroids of the sub areas and As represents the areas of the sub areas. An example will help. ########### Example B.2.1 Consider the cross section in Figure (a) to which some dimensions have be added. y 20 z 100 z 100 Figure (a) We place the y axis on the axis of symmetry where we know the centroid lies and the z axis conveniently at the right edge. The cross section is divided into three rectangular areas for which their centroids are known. To find z̄ z̄ = 50 · 100 · 20 + 10 · 60 · 20 + 50 · 100 · 20 z 1 A1 + z 2 A2 + z 3 A3 = 40.77 = A1 + A2 + A3 100 · 20 + 60 · 20 + 100 · 20 (a) For slender bar analysis the y axis is moved to the new location. Centroids of some common shapes are given in the last section of this appendix. B.3 Area Moments and Product of Inertia The area moments of inertia are Izz =  y 2 dA I yy = A  z 2 dA (B.3.1) A and the area product of inertia is Iyz = Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS  yzdA A Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (B.3.2) --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- ########### P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas July 4, 2011 9:9 Printer Name: Yet to Come 604 Appendix B: Area Properties of Cross Sections We are interested primarily in values referred to centroidal axes. In many cases Iyz = 0. This occurs when either the xy or the xz axes plane is a plane of symmetry or the yz axes are oriented so that Iyz = 0. Then the axes are called principal axes of inertia. For sections with double symmetry the integration is often straight forward. An example will help. ########### Example B.3.1 Consider first the rectangular cross section with double symmetry as shown in Figure (a). y z h b Figure (a) Given a height h and a width b we have Izz =  h 2 − h2 y 2 bdy = 1 3 bh 12 I yy =  b 2 z 2 hdz = − b2 1 3 b h 12 Iyz =  b 2 − b2  h 2 yzdydz = 0 (a) − h2 Moments of inertia of typical double and single symmetry sections are given in the last section of this appendix. ########### For sections made up of subsections with known moments of inertia about the centroids of the sub sections there is a transfer process. It is known as the parallel axis theorem. Let the yc zc axes be centroid axes for an area whose moments and product of inertia are known. We wish to find the moments of inertia of this area with respect to a yz set of axes. Let ȳc and z̄ c be the distances from the yz axes to the yc zc axes as shown in Figure B.3.1. Izz =  (yc + ȳc )2 dA = A  yc2 dA + 2 ȳc A  yc dA+yc2 A  dA (B.3.3) A Since  yc dA = 0 A  dA = A (B.3.4) A --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas July 4, 2011 9:9 Printer Name: Yet to Come 605 Appendix B: Area Properties of Cross Sections yc y zc yc this becomes Izz = Izc zc + A ȳc2 (B.3.5) I yy = I yc yc + Az̄ c2 (B.3.6) Iyz = I yc zc + A ȳc z̄ c (B.3.7) Likewise and ########### Example B.3.2 Find the area moments of inertia with respect to centroidal axes for the cross section in Example B.2.1. Units are millimeters. y 20 40.77 100 z 100 Figure (a) Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Figure B.3.1 P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas July 4, 2011 9:9 Printer Name: Yet to Come 606 Appendix B: Area Properties of Cross Sections Transferring the three sections from top to bottom: 1 1 1 · 100 · (20)3 + 100 · 20 · (40)2 + · 20 · (60)3 + · 100 · (20)3 + 100 · 20 · (40)2 12 12 12 Izz = = 6,893,333 mm I yy = (a) 4 1 1 · 20 · (100)3 + 100 · 20 · (9.23)2 + · 60 · (20)3 + 60 · 20 · (30.77)2 12 12 1 · 20 · (100)3 + 100 · 20 · (9.23)2 + 12 (b) = 4,850,256 mm 4 From symmetry the product of intertia Iyz = 0. ########### For a section with no symmetry the process requires also finding the product of inertia. We show an example next. ########### Example B.3.3 Consider the following section. Find the centroid and the area moments and product of inertial with respect to the centroidal axes. y yc 20 zc zc 100 yc z 50 100 Figure (a) First find the centroid: 100 · 20 · 50 + 60 · 20 · 10 + 50 · 20 · 25 z 1 A1 + z 2 A2 + z 3 A3 = 32.62 mm = A1 + A2 + A3 100 · 20 + 60 · 20 + 50 · 20 y1 A1 + y2 A2 + y3 A3 100 · 20 · 90 + 60 · 20 · 50 + 50 · 20 · 10 = 59.52 mm ȳc = = A1 + A2 + A3 100 · 20 + 60 · 20 + 50 · 20 z̄ c = --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas July 4, 2011 9:9 Printer Name: Yet to Come 607 Appendix B: Area Properties of Cross Sections Now the moments of inertia: Izz = 1 1 · 100 · (20)3 + 100 · 20 · (30.48)2 + · 20 · (60)3 + 60 · 20 · (9.52)2 12 12 1 · 50 · (20)3 + 50 · 20 · (49.52)2 + 12 = 4,879,048 mm 4 I yy = 1 · 20 · (100)3 + 100 · 20 · (17.38)2 12 1 1 · 60 · (20)3 + 60 · 20 · (22.62)2 + · 20 · (50)3 + 50 · 20 · (7.62)2 + 12 12 (b) = 3,191,190 mm 4 Iyz = 100 · 20 · 17.38 · 30.48 + 20 · 60 · 9.52 · 22.62 + 50 · 20 · 7.62 · 49.52 = 1,695,238 mm 4 ########### Since in the main text we do all analysis with respect to principle axes, that is, axes for which Iyz = 0, we must reorient the axes to apply those methods. Consider the rotated y′ z′ axes in Figure B.3.2. y y′ θ z z′ Figure B.3.2 A point with the coordinates y and z with respect to the yz axes have the coordinates y’ and z’ with respect to the y′ z′ axes. The transformation equations are y ′ = y cos θ + z sin θ z ′ = z cos θ − y sin θ (B.3.8) From this we obtain Iz ′ z ′ =  A  ′ 2 y dA =  (y cos θ + z sin θ)2 dA A = Izz cos2 θ + I yy sin2 θ + 2Iyz sin θ cos θ I yy − Izz I yy + Izz − cos 2θ + Iyz sin 2θ = 2 2 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (B.3.9) P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas July 4, 2011 9:9 Printer Name: Yet to Come 608 Appendix B: Area Properties of Cross Sections I y′ y′ =   ′ 2 z dA = A  (z cos θ − y sin θ)2 dA A = I yy cos2 θ + Izz sin2 θ − 2Iyz sin θ cos θ I yy − Izz I yy + Izz + cos 2θ − Iyz sin 2θ = 2   2 y ′ z ′ dA = I y′ z′ = A (B.3.10) (y cos θ + z sin θ ) (z cos θ − y sin θ) dA A = I yy sin θ cos θ Izz sin θ cos θ + Iyz sin2 θ − cos2 θ I yy − Izz sin 2θ + Iyz cos 2θ = 2 (B.3.11) There is a value of θ for which Ixy = 0. I may be found by setting 2Iyz I yy − Izz sin 2θ + Iyz cos 2θ = 0 → tan 2θ = 2 Izz − I yy (B.3.12) It may be noted that when Ixy = 0 then Iyy is either a maximum or a minimum and Izz is a corresponding minimum or a maximum. These values are I max min I yy + Izz ± = 2  I yy − Izz 2 2 2 + Iyz (B.3.13) ########### Example B.3.4 Find the rotation angle of the axes to obtain principal axes of inertia and the resulting values for the cross section in Example B.3.2. y′ yc y zc zc yc z′ z Figure (a) The angle of rotation is tan 2θ = --`,,,,``,``,,``,,`,`,,,,,``, Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Iyz 1,695,238 = 1.00437 → θ = 22.56 ◦ = Izz − I yy 4,879,048 − 3,191,190 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST (a) P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas July 4, 2011 9:9 Printer Name: Yet to Come 609 Appendix B: Area Properties of Cross Sections The principal moments of inertia are I min = 2,141,433 mm 4 I max = 5,928,805 mm 4 (b) ########### B.4 Properties of Common Cross Sections A = bh h z Iyy = bh 3/12 Izz = hb 3/12 Iyz = 0 b yc A = πR 2 Iyy = Izz = π R 4/4 R zc J = π R 4/2 Iyz = 0 yc A = bh/2 h Izz = bh3/36 Iyy = hb3/36 zc h/3 Iyz = b2h2/72 b/3 b Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- yc July 4, 2011 9:9 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- P1: TIX/XYZ P2: ABC JWST071-App02 JWST071-Waas Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-App03 JWST071-Waas July 4, 2011 9:11 Printer Name: Yet to Come Appendix C Solving Sets of Linear Algebraic Equations with Mathematica C.1 Introduction You will have the opportunity to solve sets of linear algebraic equations of increasing complexity as the text progresses. It will be advantageous for you to learn the use of one or more software packages as soon R , as possible. Among the software packages that may be available are Mathematica, Maple, MATLAB and Mathcad. There may be others. Any will do. For those who do not already know a package here is a very brief introduction to Mathematica. C.2 Systems of Linear Algebraic Equations Simply stated the problem is to solve the equations [A] {q} = {B} (C.2.1) where [A] and {B} are known and {q} is to be found. Several software packages can conveniently solve these equations either symbolically or numerically. Here are some simple instructions for using Mathematica. If you are already familiar with Maple or other software that does the job please feel free to use it instead. Our first interest is in numerical solutions, that is, where both [A] and {B} contain only numerical values. There are circumstances, however, when symbolic solutions may be desired so we will cover that as well, but first, numerical solutions. C.3 Solving Numerical Equations in Mathematica The following equations are used to illustrate the numerical solution. ⎡ 5.25 ⎢−4.1 ⎢ [A] {q} = ⎣ 1.0 0 −4.1 6.05 −4.1 1.0 1.0 −4.1 6.05 −4.1 ⎡ ⎤ ⎤⎡ ⎤ 0.01 0 q1 ⎢ ⎥ ⎢ ⎥ 1.0 ⎥ ⎥ ⎢ q2 ⎥ = {B} = ⎢ 1.1 ⎥ ⎣ 0.0 ⎦ −4.1 ⎦ ⎣ q3 ⎦ q4 0.1 5.25 (C.3.1) Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-App03 JWST071-Waas 612 July 4, 2011 9:11 Printer Name: Yet to Come Appendix C: Solving Sets of Linear Algebraic Equations with Mathematica The brackets and braces are not needed with the symbols A, q, and B. That they represent matrices is declared by the format you use when you enter the values. Open Mathematica. Enter the data on your Mathematica worksheet. What you enter is given in boldface. That not in boldface is supplied by the program. The symbol <cr> means press the enter key. A={{5.25,-4.1,1.0,0},{-4.1,6.05,-4.1,1.0},{1.0,-4.1,6.05,-4.1},{0,1.0,-4.1,5.25}} <cr> q={{q1},{q2},{q3},{q4}} <cr> B={{0.01},{1.1},{0},{0.1}} <cr> Solve[A.q==B] <shift>+<cr> <shift>+<cr>is the instruction to do the calculation. The software will respond with the following output. Out[1] = {{5.25,-4.1,1.0,0},{-4.1,6.05,-4.1,1.0},{1.0,-4.1,6.05,-4.1},{0,1.0,-4.1,5.25}} Out[2] = {{q1},{q2},{q3},{q4}} Out[3] = {{0.01},{1.1},{0},{0.1}} Out[4] = {{q1->2.97414,q2->4.97075,q3->4.77586,q4->2.80196}} There is also a command called LinearSolve. For Solve the format is [A.q = =B]; for LinearSolve the format is [A,B]. Try it. C.4 Solving Symbolic Equations in Mathematica Let us try a simple set of equations right out of introductory algebra. 1x + 2y + 3z = R 2x + 4y + 5z = S 3x + 5y + 6z = T or ⎡ 1 [A {q}] = ⎣2 3 2 4 5 ⎤⎡ ⎤ ⎡ ⎤ 3 x R 5⎦ ⎣ y ⎦ = {B} = ⎣ S ⎦ 6 z T (C.4.1) Enter in Mathematica: A = {{1,2,3},{2,4,5},{3,5,6}} q = {{p1},{p2},{p3}} B = {{R},{S},{T}} Solve[A.B==q] <cr> <cr> <cr> <shift>+<cr> Out[5]={{1,2,3},{2,4,5},{3,5,6}} --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Out[6]={{q1},{q2},{q3}} Out[7]={{R},{S},{T}} Out[8]={{q1->R-3S+2T,q2->-3R+3S-T,q3->2R-S}} Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-App03 JWST071-Waas July 4, 2011 9:11 Printer Name: Yet to Come Appendix C: Solving Sets of Linear Algebraic Equations with Mathematica 613 C.5 Matrix Multiplication You will often need to multiply matrices to complete a solution. Take a case which is similar to one that you will encounter. ⎡ ⎤   0.52 2 3 4.5 ⎣ R1 1.20 ⎦ = 12 (C.5.1) R2 7 1 −6 0.26 12 {{2,3,4.5},{7,1,-6}} . {0.52,1.20,0.26} <shift>+<cr> Out[1]={69.72, 39.36} This is the dot or inner product of two matrices. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Enter in Mathematica: P1: TIX/XYZ P2: ABC JWST071-App03 JWST071-Waas July 4, 2011 9:11 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ P2: ABC JWST071-App04 JWST071-Waas July 4, 2011 9:13 Printer Name: Yet to Come Appendix D Orthogonality of Normal Modes D.1 Introduction In Chapter 15, Equation 15.2.54, we note the matrix form of the equations for multi degree of freedom mass/spring systems. It is repeated here are as Equation D.1.1. [M] {v̈} + [K ] {v} = {F (t)} (D.1.1) In the preceding pages the homogenous form of these is equations are solved for the natural frequencies, ωi , and for the normal modes, {ϕ}i . In the ensuing paragraphs the equations for finding the forced motion of the system are developed. In the process, the orthogonality of normal modes, Equation 15.2.58 repeated here as Equation D.1.2, is used without proof. i = j i= j (D.1.2) We should note that the equations for FEM analysis are exactly the same form as Equation D.1. In the case of the mass/spring system the mass matrix, [M], contains the concentrated masses while for the FEM case the mass matrix contains equivalent nodal masses depending upon the particular elements used in the derivation. Similarly, the stiffness matrix contains the spring constants for the mass/spring system and the equivalent stiffnesses for the FEM case. The proof which follows in Section D.2, then, applies to all discrete systems such as mass/spring and FEM formulations. In Section D.3 we extend the proof to continuous systems. D.2 Proof of Orthogonality for Discrete Systems Consider two dissimilar frequencies and modes of Equations D.1     [K ] − ωi2 [M] {ϕ}i = 0 [K ] − ω2j [M] {ϕ} j = 0 (D.2.1) If we premultiply the first by {ϕ}Tj and the second by {ϕ}iT , we get   {ϕ}Tj [K ] − ωi2 [M] {ϕ}i = 0   {ϕ}iT [K ] − ω2j [M] {ϕ} j = 0 (D.2.2) Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- {ϕ}Tj [M] {ϕ}i = 0 = Mj P1: TIX/XYZ P2: ABC JWST071-App04 JWST071-Waas July 4, 2011 9:13 Printer Name: Yet to Come 616 Appendix D: Orthogonality of Normal Modes Since in matrix multiplications [a]T [b] [c] = [c]T [b] [a] when [b] is symmetric, subtracting the first of Equation D.2.2 from the second, we have    2 ω j − ωi2 {ϕ}Tj [M] {ϕ}i = {ϕ}Tj [K ] {ϕ}i − {ϕ}Tj [K ] {ϕ}i = 0 (D.2.3) (D.2.4) From this we conclude that {ϕ}Tj [M] {ϕ}i = 0 = Mj ωi = ω j ωi = ω j (D.2.5) Thus Equation 15.2.58 is justified. D.3 Proof of Orthogonality for Continuous Systems Consider first the differential equation for the forced motion of a uniform axial bar, Equation 15.3.3 repeated here as Equation D.3.1. EA ∂ 2u − m ü = − f x (x, t) ∂x2 (D.3.1) Once the natural frequencies, ωi , and normal modes, ϕi , are found we can consider two different normal modes as follows. EAϕi ′′ − ωi2 mϕi = 0 EAϕ j ′′ − ω2j mϕ j = 0 (D.3.2) If we premultiply the first by ϕ j and the second by ϕi and integrate over the length of the bar, we get   EAϕi ′′ ϕ j d x − ωi2 mϕi ϕ j d x = 0   (D.3.3) EAϕ j ′′ ϕi d x − ωi2 mϕ j ϕi d x = 0 when [b] is symmetric, subtracting the first of Equations D.2.2 from the second, we have       2 mϕi ϕ j d x = EAϕ j ′′ ϕi d x − EAϕi ′′ ϕ j d x ω j − ωi2 By integrating the right hand side of Equation D.3.4 by parts, we obtain From this we conclude that  mϕi ϕ j d x = 0 ωi = ω j = Mj (D.3.4) (D.3.5) ωi = ω j Thus Equation 15.3.8 is justified. This can be extended to any and all continuous elastic structures. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-REF P2: ABC JWST071-Waas July 11, 2011 16:3 Printer Name: Yet to Come References Bathias, C. and A. Pineau (2010) Fatigue of Materials and Structures, John Wiley & Sons Ltd. Craig, R. and A. Kurdila (2006) Fundamentals of Structural Dynamics, 2nd edition, John Wiley & Sons Ltd. Crandall, S.H., N.C. Dahl, and T.S. Lardner (1972) An Introduction to the Mechanics of Solids, 2nd edition, McGraw-Hill: New York. Grandt, A. (2004) Fundamentals of Structural Integrity, John Wiley & Sons Ltd. Herakovich, C.T. (1998) Mechanics of Fibrous Composites, John Wiley & Sons Ltd. Hyer, M.W. (1998) Stress Analysis of Fiber-Reinforced Composite Materials, McGraw-Hill: New York. Kollar, L. and G. Springer (2003) Mechanics of Composite Structures, Cambridge University Press. Matthews, F. (1981) Course Notes on Aircraft Structures, AY 1981–1982, Imperial College, London, UK. Timoshenko, S. and S. Woinowsky-Krieger (1987) Theory of Plates and Shells, McGraw-Hill: New York. Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-REF P2: ABC JWST071-Waas July 11, 2011 16:3 Printer Name: Yet to Come --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-IND P2: ABC JWST071-Waas July 8, 2011 14:35 Printer Name: Yet to Come Index Allowable stress, 361 Angle of twist, 213 Applied loads, 36 Area properties, 257 Assembly of element matrices, 170 Axially loaded bars Castigliano’s second theorem, 153 design of, 145 displacement equations, 190 elasticity solutions, 99 finite element method (FEM), 165 statically determinate case, 116 statically indeterminate case, 129 shear stress in, 143 thermal stress, 142 trusses, 38, 60, 149, 202, 507 two force members, 38 variable cross sections, 136 vibration of bars, 548, 560 Beams applied loads, 261 boundary conditions, 266, 282 Castiglinao’s theorem, 278 classical differential equations, 264 design, 309 finite element method, 315 large displacements, 313 elasticity solutions, 290 shear flow in thin walled, 304 statically determinate, 271 statically indeterminate, 281 thin walled cross sections, 302 variable cross sections, 300 vibration by classical methods, 569 vibration by FEM, 574 Beam columns, 512 Body forces, 25 Buckling beam columns, 512 columns, 504 combined axial and lateral loads, 499 critical buckling load, 507 critical stress, 507 eigenvalue, 504 eigenvector, 504 frames, 524 differential stiffness matrix, 518 FEM equations, 515 geometric stiffness matrix, 518 modes, 50 plates, 524 shape functions, 517 virtual work, 517 Castigliano’s second theorem, 153, 278 Creep, 364 Columns, 504 Combined loading, 367 axial and torsional, 372 axial and ending, 375 bending in two planes, 384 bending and torsion, 393 thin walled closed sections, 399 Compatibility, 76 Composites, 479 classical lamination theory, 490 fiber angle, 483 Fourier series, 496 invariants, 485 Kirchhoff shear, 498 lamina, 480 laminate, 479 major poissons ratio, 485 Analysis of Structures: An Introduction Including Numerical Methods, First Edition. Joe G. Eisley and Anthony M. Waas. © 2011 John Wiley & Sons, Ltd. Published 2011 by John Wiley & Sons, Ltd. --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-IND P2: ABC JWST071-Waas July 8, 2011 14:35 Printer Name: Yet to Come 620 Composites (Continued ) optimization, 485 shear coupling, 485 transverse loads, 493 Constant strain triangle, 433 Critical buckling load, 507 D’Alembert’s principle, 530 Damping forces, 577 Deformation slender bars, 111 beams, 260 plates, 452 shafts, 213 strain displacement, 71 solid bodies, 470 Differential stiffness, 518 Eigenvalues buckling, 504 vibration, 534 Eigenvectors buckling, 504 Elasticity, theory of compatibility, 76 displacement and strain, 71 equilibrium, 66 Hooke’s law, 77, 83, 84 Elements, 166 Element stiffness matrix slender bar, 168 beam, 319 plate, 433 shaft, 246 solid, 436 Equations of compatibility, 76 Equations of elasticity, 66, 71, 76, 84 Equilibrium static, 35 concentrated forces, 38 distributed forces, 55 internal forces, 62, 70 Equivalent nodal loads, 182 Factors of safety, 351 Fatigue, 363 Fiber reinforce laminates, 90–5 Hooke’s law for lamina, 90–4 Unidirectional lamina, 94–5 Finite element method axial bar, 165 beams, 315 boundary constraints, 172, 247 buckling, 512 Index concentrated loads, 165 distributed loads, 182 elements, 166, 246, 316 equivalent nodal loads, 182, 327 frames, 378 free vibrations, 569 matrix assembly, 170 nodes, 166, 246, 316 partitioning, 174 pin jointed trusses, 202 plane stress applications, 445 plate applications, 455 shape functions, 166, 246, 316 three dimensional solids, 470 torsion, 245 virtual work, 421 Forced motion, 540 Forces concentrated, 4 damping, 577 distributed, 19–27 inertia, 530 internal, 27–32 resultants, 19–21 restraint, 32 Frames static analysis, 378 buckling, 524 Free body diagrams, 35 Free motion, 530 Hooke’s law, 77 Inertia Forces, 530 Initial conditions, 531 Instability, see buckling Isotropic materials, 83, 84 Kirchhoff shear, 455, 498 Lamination theory, 479 Laminates Classical lamination theory (CLT), 489 Strain displacement equations, 480 Stress strain relations, 482 Stress resultants, 486 Transverse loads, 493 Kirchhoff shear, 498 Mass matrix, 574 Mass/spring systems linear, 529 torsional, 567 damping, 577 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-IND P2: ABC JWST071-Waas July 8, 2011 14:35 Printer Name: Yet to Come 621 Index --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Material properties orthotropic, 365 effect of temperature change, 89 isotropic, 77 laminates, 90–5 linear, 77–85 orthotropic, 365 stress strain curve, 79 shear modulus, 83, 356 Young’s modulus, 80 Matrices assembly process, 170 definitions, 597 element stiffness, 168, 246, 319, 433, 436 global stiffness, 172 mass, 574 partitioned, 174 Matrix Algebra, 597, 598 Mid edge nodes, 446 Natural frequency, 531 Nodal displacements, 134, 166–168, 172–173, 175–176, 315 Nodal loads, equivalent, 182 Nodes, 166 Normal modes, 535 Orthogonality, 537 , 617 Orthotropic, 365 Partioning, 174 Pin jointed trusses classical analysis, 149 analysis and design by FEM, 202 Plane sections, 105 Plane stress elasticity equations, 440 element stiffness matrix, 433, 445 Plates classical differential equations, 452 buckling, 524 finite element method, 455 Poisson’s ration, 81 Resonance, 540 Restraint forces, 36, 38 Reversed effective force, 530 Rigid body, 35 Rigid bars, 44 Safety factor, 351 St. Venant’s principle, 106 Separation of variables, 549 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Shape functions axial bars, 166 bending of beams, 316 plates, 435 torsional shafts, 246 Shear strain, 73 stress, 27 Shear center, 393, 399 Shear modulus, 83, 356 Sign conventions beams with classical analysis, 259 beams with FEA, 317 Slender bars applied loads, 110 axial stress, 112 axial vibration, 548, 560 boundary restraints, 113 classical analysis of, 99 finite element analysis of, 169 work and energy methods, 421 Solid elements, 436 Spring constant, 52 Stiffened thin walled beams, 405 Statically determinate, 37, 116, 271 Statically indeterminate, 129, 281 Stiffness matrix differential, 518 element, 168, 246, 319, 433, 436 geometric, 518 global, 172 Strain displacement and, 71 normal, 72 rosettes, 356 principal, 355 shearing, 73 thermal, 89 transformation in two dimensions, 354 Strain energy, 153 Stress allowable, 361 compressive, 80, 499 definition of, 27 normal and shearing, 27 plane, 440 principal axes, 350, 358 principal stress, 350, 358 resultants, 27 transformation in two dimensions, 347 transformation in three dimensions, 358 two dimensional, 31 three dimensional, 32 ultimate, 351 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST P1: TIX/XYZ JWST071-IND P2: ABC JWST071-Waas July 8, 2011 14:35 Printer Name: Yet to Come 622 Stress strain equations, 82 Structural dynamics axial vibration, 548, 560 bending vibration, 569 torsional vibration, 567, 569 Superposition, 531, 535 --`,,,,``,``,,``,,`,`,,,,,``,,``-`-`,,`,,`,`,,`--- Thermal stress,142 Thermal strain, 89 Thin walled cross sections shear and torsion in closed sections, 399 shear and torsion in open sections, 393 shear center, 393 shear stress in, 302 stiffened, 405 torsion, 229 Three dimensional solids, 470 Torsion axial and torsional loads, 372 coupled bending torsion, 393 displacement, 216–9 elasticity solution, 225–9 finite element method, 245–8 in multi celled beams, 239–41 strain, 215 stress, 215 shear in thin walled closed sections, 229–31 shear in thin walled open sections, 242–5 variable cross sections, 254 Torsional constant, 215 effective torsional constant, 235 Torsional stiffness, 216, 231 Torsional vibration, 567 Trusses, 38–39 Copyright Wiley Provided by IHS under license with WILEY No reproduction or networking permitted without license from IHS Index Units, 1 Unstable structures, see buckling Variable cross sections, 136, 254, 300, 341 Variables, separation of, 549 Vibration analysis axial by classical methods, 548 axial by FEA, 560 beams, 569 damping, 577 FEA for all structures, 577 free motion, 530 forced motion, 540 initial conditions, 531 mass matrix, 560 resonance, 540 slender bar, 548 separation of variables, 549 torsional by classical methods, 567 torsional by FEA, 569 Virtual displacements, 417 Virtual work principle of, 417 internal virtual work, 419 2D and 3D solids, 430 Work and energy, 153, 417. See also virtual work Castigliano’s second theorem, 157 internal work, 419 strain energy, 153 torsional stiffness, 231 effective torsional stiffness, 234 Young’s modulus, 80 Yield stress, 80, 361, 363, 499 Licensee=PDVSA - Los Teques site 6/9986712011, User=Sanchez, Carmelo Not for Resale, 03/09/2016 08:24:43 MST