Principles of Gas-Solid Flows-Fan, Zhu PDF

Principles of Gas-Solid Flows
Gas-solid flows are involved in numerous industrial processes and occur in vari-
ous natural phenomena. This authoritative book addresses the fundamental princi-
ples that govern gas-solid flows and the application of these principles to various
gas-solid flow systems.
The book is arranged in two parts: Part I deals with basic relationships and
phenomena, including particle size and properties, collision mechanics of solids,
momentum transfer and charge transfer, heat and mass transfer, basic equations,
and intrinsic phenomena in gas-solid flows. Part II discusses the characteristics of
selected gas-solid flow systems such as gas-solid separators, hopper and standpipe
flows, dense-phasefluidizedbeds, circulatingfluidizedbeds, pneumatic conveying
systems, and heat and mass transfer in fluidization systems.
As a comprehensive information source on gas-solid flows, this text will be
useful to a broad range of engineers and applied scientists - chemical, mechanical,
agricultural, civil, environmental, aeronautical, and materials engineers, as well
as atmospheric and meteorological scientists.
Cambridge Series in Chemical Engineering
Editor
Arvind Varma, University of Notre Dame
Editorial Board
Alexis T. Bell, University of California, Berkeley
John Bridgwater, University of Cambridge
L. Gary Leal, University of California, Santa Barbara
Massimo Morbidelli, Swiss Federal Institute of Technology, Zurich
Stanley I. Sandier, University of Delaware
Michael L. Shuler, Cornell University
Arthur W. Westerberg, Carnegie-Mellon University
Principles of Gas-Solid Flows
LIANG-SHIH FAN
The Ohio State University
CHAO ZHU
The Ohio State University
CAMBRIDGE
UNIVERSITY PRESS
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo
Cambridge University Press
The Edinburgh Building, Cambridge CB2 2RU, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521581486
Liang-Shih Fan and Chao Zhu 1998
This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.
First published 1998
This digitally printed first paperback version 2005
A catalogue recordfor this publication is available from the British Library
Library of Congress Cataloguing in Publication data

Fan, Liang-Shih.
Principles of gassolid flows / Liang-Shih Fan, Chao Zhu.
p. cm.
Includes bibliographical references (p. ).
ISBN 0-521-58148-6
1. Two-phase flow. 2. Gas flow. 3. Bulk solids flow. I. Zhu,
Chao, 1961- . II. Title.
TA357.5.M84F36 1997
531'.163-dc20 96-41141
CIP
ISBN-13 978-0-521-58148-6 hardback
ISBN-10 0-521-58148-6 hardback
ISBN-13 978-0-521 -02116-6 paperback

ISBN-10 0-521-02116-2 paperback
To
Lois Liang-Yen, Liang-Fu, and Jeannine Liang-Chi
and
Zhichi Zhu, Zhu Ruan, Baolan Shi, and Victor
Contents
Preface page xv
Part I Basic Relationships

1 Size and Properties of Particles 3
1.1 Introduction 3
1.2 Particle Size and Sizing Methods 3
1.2.1 Equivalent Diameters of a Nonspherical Particle 4
1.2.2 Particle Sizing Methods 10
1.3 Particle Size Distributions and Averaged Diameters 17
1.3.1 Density Functions 18
1.3.2 Typical Distributions 19
1.3.3 Averaged Diameters of a Particulate System 23
1.4 Material Properties of Solids 24
1.4.1 Physical Adsorption 25
1.4.2 Deformation and Fracture 28
1.4.3 Thermal Properties 32
1.4.4 Electrical Properties 35
1.4.5 Magnetic Properties 37
1.4.6 Material Densities 38
1.4.7 Optical Properties 39
Nomenclature 40
References 42
Problems 43
2 Collision Mechanics of Solids 46

2.1 Introduction 46
2.2 Stereomechanical Impact 47
2.2.1 Collinear Impact of Spheres 47
2.2.2 Planar Impact of Spheres 48
2.3 Theory of Elastic Contact of Solids 49
2.3.1 General Relations of Stresses in a Solid Medium in Equilibrium 50
2.3.2 Concentrated Force at a Point in an Infinite Solid Medium 52
2.3.3 Force on the Boundary of a Semiinfinite Solid Medium 53
2.3.4 Hertzian Theory for Frictionless Spheres in Contact 59
2.3.5 Theories for Frictional Spheres in Contact 63
2.4 Collision of Elastic Spheres 72
2.4.1 Normal Collision of Elastic Spheres 72
2.4.2 Collision of Frictional Elastic Spheres 74
vii
viii Contents
2.5
Collision of Inelastic Spheres 78
2.5.1 Onset of Plastic Deformation 78
2.5.2 Restitution Coefficient 80
Nomenclature 83
References 85
Problems 85
3 Momentum Transfer and Charge Transfer 87

3.1 Introduction 87
3.2 Particle-Fluid Interactions 87
3.2.1 Drag Force 87
3.2.2 Basset Force 88
3.2.3 Saffman Force and Other Gradient-Related Forces 95
3.2.4 Magnus Effect and Force Due to Rotation of a Sphere 97
3.3 Interparticle Forces and Field Forces 101
3.3.1 Van der Waals Force 101
3.3.2 Electrostatic Force 103
3.3.3 Collisional Force 104
3.3.4 Field Forces 105
3.4 Motion of a Single Particle 107
3.4.1 Basset, Boussinesq, and Oseen (BBO) Equation 107
3.4.2 General Equation of Motion 108
3.5 Charge Generation and Charge Transfer 111
3.5.1 Static Electrification of Solids 111
3.5.2 Charge Transfer by Collision 119
Nomenclature 123
References 126
Problems 128
4 Basic Heat and Mass Transfer 130

4.1 Introduction 130
4.2 Heat Conduction 130
4.2.1 Heat Transfer of a Single Sphere in a Quiescent Fluid 131
4.2.2 Heat Conduction in a Collision of Elastic Spheres 133
4.3 Convective Heat Transfer 138
4.3.1 Dimensional Analysis of Forced Convection in a Single-Phase
Flow 138
4.3.2 Heat Transfer of a Single Sphere in a Uniform Flow 138
4.3.3 Thermal Convection in Pseudocontinuum One-Phase Flow 141
4.4 Thermal Radiation 142
4.4.1 Single-Particle Scattering 143
4.4.2 Radiant Heating of a Particle 148
4.4.3 General Considerations in Radiation with a Particle Cloud 150
4.4.4 Radiation Through an Isothermal and Diffuse Scattering Medium 154
4.5 Mass Transfer 156
4.5.1 Diffusion and Convection 156
4.5.2 Mass and Heat Transfer Analogy 157
Contents ix
Nomenclature 159
References 161
Problems 162
5 B asic Equations 164

5.1.1 Eulerian Continuum Approach 164
5.1.2 Lagrangian Traj ectory Approach 165
5.1.3 Kinetic Theory Modeling for Interparticle Collisions 166
5.1.4 Ergun Equation 166
5.1.5 Summary 167
5.2 Modeling of Single-Phase Flows 167
5.2.1 General Transport Theorem and General Conservation 167
5.2.2 Governing Equations 169
5.2.3 Kinetic Theory and Transport Coefficients 170
5.2.4 Modeling for Turbulent Flows 174
5.2.5 Boundary Conditions 179
5.3 Continuum Modeling of Multiphase Flows 182
5.3.1 Averages and Averaging Theorems 182
5.3.2 Volume-Averaged Equations 189
5.3.3 Volume-Time-Averaged Equations 193
5.3.4 Transport Coefficients and Turbulence Models 196
5.3.5 Boundary Conditions of Particle Phase 205
5.4 Trajectory Modeling of Multiphase Flows 205
5.4.1 Deterministic Trajectory Models 206
5.4.2 Stochastic Trajectory Models 208
5.5 Kinetic Theory Modeling for Collision-Dominated Dense Suspensions 210
5.5.1 Dense-Phase Transport Theorem 211
5.5.2 Hydrodynamic Equations 213
5.5.3 Collisional Pair Distribution Function 215
5.5.4 Constitutive Relations 217
5.6 Equations for Flows Through Packed Beds 222
5.6.1 Darcy's Law 223
5.6.2 Straight Capillaric Model 224
5.6.3 Ergun's Equation 225
5.7 Dimensional Analysis and Similarity 230
5.7.1 Scaling Relationships for Pneumatic Transport
of Dilute Suspensions 230
5.7.2 Scaling Relationships for Fluidized Beds 232
Nomenclature 236
References 239
Problems 242
6 Intrinsic Phenomena in a Gas-Solid Flow 244

6.2 Erosion and Attrition 244
6.2.1 Ductile Erosion and Brittle Erosion 245
Contents
6.2.2 Locations of Erosive Wear 247

6.2.3 Mechanisms of Attrition 252
6.3 Thermodynamic Properties of a Gas-Solid Mixture 254
6.3.1 Density, Pressure, and Equation of State 254
6.3.2 Internal Energy and Specific Heats 257
6.3.3 Isentropic Change of State 258
6.4 Pressure Waves Through a Gas-Solid Suspension 259
6.4.1 Acoustic Wave 259
6.4.2 Normal Shock Wave 265
6.5 Instability 270
6.5.1 Wave Motions in Stratified Pipe Flows 270
6.5.2 Continuity Wave and Dynamic Wave 280
6.6 Particle-Turbulence Interaction 285
Nomenclature 288
References 292
Problems 293
Part II System Characteristics

7 Gas-Solid Separation 297
7.2 Separation by Rotating Flow 297
7.2.1 Mechanism and Type of Rotary Row Dust Separators 297
7.2.2 Flow Field in a Cyclone 300
7.2.3 Collection Efficiency of Cyclones 303
7.3 Electrostatic Precipitation 309
7.3.1 Mechanism of an Electrostatic Precipitator 310
7.3.2 Migration Velocity and Electric Wind 311
7.3.3 Collection Efficiency of Electrostatic Precipitators 312
7.4 Filtration 314
7.4.1 Mechanisms of Filtration and Types of Filters 314
7.4.2 Pressure Drop in a Filter 315
7.4.3 Collection Efficiency of Fabric Filters 319
7.5 Gravity Settling and Wet Scrubbing 320
7.5.1 Gravity Settling Chambers 321
7.5.2 Mechanisms of Scrubbing and Types of Scrubbers 323
7.5.3 Modeling for Scrubbing and Collection Efficiency 324
Nomenclature 328
References 330
Problems 331
8 Hopper and Standpipe Flows 333

8.2 Powder Mechanics in Hopper Flows 333
8.2.1 Mohr Circle for Plane Stresses 334
8.2.2 Mohr-Coulomb Failure Criterion and Coulomb Powders 336
8.2.3 Static Stress Distributions in Standpipes and Hoppers 337
Contents xi
8.2.4 Stress Distribution in a Steady Hopper Flow 340

8.2.5 Flowability of Powders in Hopper Design 342
8.3 Hopper and Standpipe Flow Theories 346
8.3.1 Moving Bed Flows in a Feed Hopper 346
8.3.2 Standpipe Flows 350
8.3.3 Hopper-Standpipe-Discharger Flow 354
8.3.4 Multiplicity of Steady Standpipe Flows 357
8.3.5 Leakage Flow of Gas in a Standpipe 359
8.4 Types of Standpipe Systems 361
8.4.1 Overflow and Underflow Standpipes 361
8.4.2 Inclined Standpipe and Nonmechanical Valves 364
Nomenclature 366
References 368
Problems 369
9 Dense-Phase Fluidized Beds 371

9.2 Particle and Regime Classifications and Fluidized Bed Components 371
9.2.1 Classification of Fluidized Particles 372
9.2.2 Fluidization Regimes 374
9.2.3 Components in a Dense-Phase Fluidized Bed 376
9.3 Minimum Fluidization and Particulate Fluidization 378
9.3.1 Minimum Fluidization 378
9.3.2 Particulate Fluidization 380
9.4 Bubbling Fluidization 381
9.4.1 Onset of Bubbling 381
9.4.2 Single Bubble in a Fluidized Bed 382
9.4.3 Bubble/Jet Formation and Bubble Coalescence and Breakup 388
9.4.4 Bubble/Jet Size and Bubble Rise Velocity 389
9.4.5 Gas Row Division and Bed Expansion 392
9.5 Turbulent Fluidization 396
9.5.1 Regime Transition and Identification 396
9.5.2 Determination of Transition Velocity 398
9.5.3 Hydrodynamic Characteristics 399
9.6 Entrainment and Elutriation 400
9.6.1 Mechanisms of Solids Ejection into the Freeboard 401
9.6.2 Correlations and Modeling 402
9.7 Slugging 403
9.7.1 Shapes and Rise Velocities of Single Slugs 403
9.7.2 Continuous Slugging 404
9.8 Spouted Beds 406
9.8.1 Onset of Spouting 407
9.8.2 Maximum Spoutable Bed Depth and Spout Diameter 408
9.8.3 Fountain Height 408
9.8.4 Gas Flow Distribution 408
Nomenclature 409
References 411
Problems 415
xii Contents
10 Circulating Fluidized Beds 421

10.2 System Configuration 422
10.3 Flow Regimes and Transitions Between Regimes 423
10.3.1 Flow Regimes and Regime Diagrams 423
10.3.2 Regime Transition Determination 425
10.3.3 Operable Fluidization Regimes 429
10.4 Hydrodynamic Behavior in a Macroscale 438
10.4.1 Axial Profiles of Cross-Sectional Averaged Voidage 438
10.4.2 Radial Profiles of Voidage and Solids Flux 440
10.4.3 Overall Solids Holdup 442
10.5 Local Solids Flow Structure 443
10.5.1 Transient Nature of Solids Flow 444
10.5.2 Characterization of Intermittent Solids Flow 445
10.6 Mathematical Models of Fast Fluidization 447
10.6.1 Models Based on the Concept of Cluster 447
10.6.2 Models Based on the Core-Annular Flow Structure 448
10.6.3 Models Based on the Axial Profiles of Solids Holdup 451
10.6.4 Two-Phase Flow Models and Computational Fluid Dynamics 451
Nomenclature 453
References 455
Problems 459
11 Pneumatic Conveying of Solids 461

11.2 Classifications of Pneumatic Conveying Systems 461
11.2.1 Horizontal and Vertical Transport 461
11.2.2 Negative- and Positive-Pressure Pneumatic Conveyings 462
11.2.3 Dilute Flow Versus Dense Flow 464
11.2.4 Flow Regimes and Regime Transitions 464
11.3 Pressure Drop 466
11.3.1 General Pressure Drop in One-Dimensional Flow 467
11.3.2 Drag Reduction 469
11.3.3 Pressure Drop and Acceleration Length in Developing
Regions 472
11A Critical Transport Velocities 474
11.4.1 Minimum Transport Velocity 475
11.4.2 Pick-up Velocity 476
11.5 Rows in Bends 478
11.5.1 Single-Phase Flow in a Curved Pipe 478
11.5.2 Paniculate Flow in a Bend 481
11.6 Fully Developed Dilute Pipe Flows 482
1 L6.1 Basic Equations and Boundary Conditions 483
11.6.2 Characteristic Relations 487
11.6.3 Temperature Distributions of Phases 489
Nomenclature 494
References 496
Problems 498
Contents xiii
12 Heat and Mass Transfer Phenomena in Fluidization Systems 499

12.2 Suspension-to-Surface Heat Transfer 499
12.2.1 Heat Transfer Modes and Regimes 500
12.2.2 Film Model 502
12.2.3 Single-Particle Model 503
12.2.4 Emulsion Phase/Packet Model 506
12.3 Heat Transfer in Dense-Phase Fluidized Beds 512
12.3.1 Particle-to-Gas and Bed-to-Gas Heat Transfer 512
12.3.2 Bed-to-Surface Heat Transfer 513
12.3.3 Effect of Operating Conditions 518
12.4 Heat Transfer in Circulating Fluidized Beds 521
12.4.1 Mechanism and Modeling 521
12.4.2 Radial and Axial Distributions of Heat Transfer Coefficient 524
12.4.3 Effect of Operating Parameters 525
12.5 Heat Transfer in Spouted Beds 526
12.5.1 Gas-to-Particle Heat Transfer 526
12.5.2 Bed-to-Surface Heat Transfer 527
12.6 Mass Transfer in Multiparticle Gas-Solid Systems 527
12.6.1 Mass Transfer in Dense-Phase Fluidized Beds 527
12.6.2 Mass Transfer in Circulating Fluidized Beds 532
Nomenclature 532
References 535
Problems 537
Appendix: Summary of Scalar, Vector, and Tensor Notations 540
Index 545
Preface
Gas-solid flows are involved in numerous industrial processes and occur in various
natural phenomena. For example, in solid fuel combustion, gas-solid flows are involved
in pulverized coal combustion, solid waste incineration, and rocket propellant combustion.
Gas-solid flows are encountered in pneumatic conveying of particulates commonly used
in pharmaceutical, food, coal, and mineral powder processing. Fluidization is a common
gas-solid flow operation with numerous important applications such as catalytic cracking
of intermediate hydrocarbons, and Fischer-Tropsch synthesis for chemicals and liquid fuel
production. Gas-solid flows occur in gas-particle separations, as exemplified by cyclones,
electrostatic precipitators, gravity settling, and filtration operations. Fine powder-gas flows
are closely associated with material processing, as in chemical vapor deposition for ceramics
and silicon production, plasma coating, and xerography. In heat transfer applications, gas-
solid flows are involved in nuclear reactor cooling and solar energy transport using graphite
suspension flows. Solid dispersionflowsare common in pigment sprays, dust explosions and
settlement, and nozzle flows. The natural phenomena accompanied by gas-solid flows are
typified by sand storms, moving sand dunes, aerodynamic ablation, and cosmic dusts. The
optimum design of the industrial processes and accurate account of the natural phenomena
that involve gas-solid flows as exemplified previously require a thorough knowledge of the
principles governing these flows.
This book is intended to address basic principles and fundamental phenomena associated
with gas-solid flows, as well as characteristics of selected gas-solid flow systems. It
covers the typical range of particle sizes of interest to gas-solid flows, i.e., 1 ^m-10
cm, recognizing that flow characteristics for submicrometer particles are also of great
industrial importance. The book features a systematic account of important theories or
models concerning particle mechanics as well as fluid dynamics from their origins of the
development. The physical interpretation and limitations in application of these theories
or models are emphasized. Various intrinsic phenomena underlying the gas-solid flow
systems are also illustrated. The book is aimed as a textbook for seniors and graduate
students who are interested in general or specific topics of gas-solid flows. In addition,
it can be used as a reference for researchers and practitioners who are interested in the
general field of multiphase flow. It is written with multidisciplinary engineering readers in
mind. Specifically, it will be of benefit to chemical and mechanical engineering readers as
well as readers in other engineering disciplines, including agricultural, civil, environmental,
pharmaceutical, aeronautical, mining, and atmospheric and meteorological sciences.
The book contains two parts; each part comprises six chapters. Part I deals with basic
relationships and phenomena of gas-solid flows while Part II is concerned with the char-
acteristics of selected gas-solid flow systems. Specifically, the geometric features (size
and size distributions) and material properties of particles are presented in Chapter 1. Basic
particle sizing techniques associated with various definitions of equivalent diameters of
particles are also included in the chapter. In Chapter 2, the collisional mechanics of solids,
based primarily on elastic deformation theories, is introduced. The contact time, area, and
xv
xvi Preface
force of colliding particles are discussed using theories of elastic collision, which are im-
portant to the formulation of the momentum, heat, and charge transfer processes involving
collisions of solids. Chapter 3 is devoted to the momentum and charge transfer of gas-solid
flows. Various forces in gas-solid flows due to gas-particle interactions, particle-particle
interactions, and external fields are delineated. Equations for single-particle motion, based
on a force balance analysis, are derived. Basic mechanisms of charge generation in gas-
solid flows are also introduced in the chapter, along with a detailed discussion of charge
transfer mechanism by particle collisions. Chapter 4 deals with fundamental concepts and
theories of heat and mass transfer in gas-solid flows. Highlights include thermal radiation
of the particulate phase and heat conduction in collisions of elastic particles. Chapter 5
presents four basic modeling approaches of gas-solid flows, namely, continuum modeling
of multiphase flows or multifluid modeling, trajectory modeling, kinetic theory model-
ing for collision-dominated dense suspensions, and the Ergun equation for flow through a
packed bed of particles. In this chapter, the hydrodynamic equations of single-phase flows
are first discussed. Here, basic concepts of kinetic theory of gas and turbulence models are
introduced as a preamble to discussion of these basic modeling approaches. In contrast to
the ke turbulence model for single-phase flows, the kek p model is introduced with
the continuum approach of gas-solid flows to account for gas-solid turbulence interac-
tions. Chapter 6 focuses on the discussion of intrinsic phenomena in gas-solid flows, such
as erosion and attrition, acoustic wave and shock wave propagation through a gas-solid
suspension flow, thermodynamic properties of a gas-solid mixture, flow instability, and
gas-solid turbulence interactions.
Chapter 7 is concerned with gas-solid separations. The basic separation methods intro-
duced in this chapter include cyclone, filtration, electrostatic precipitation, gravity settling,
and wet scrubbing. Chapter 8 deals with hopper flows and standpipe flows, which are
commonly encountered in the bulk solids handling and transport processes. In order to
understand the fundamental hopper and standpipe flow characteristics, some basic con-
cepts of powder mechanics are illustrated. Chapter 9 introduces the general concept of gas
fluidization. Specifically, the chapter addresses dense-phase fluidization, which represents
a gas-particle operation of enormous industrial importance. Various operating regimes
including particulate fluidization, bubbling/slugging fluidization, and turbulent fluidization
are discussed along with spouting. The fundamental properties of bubble, cloud, and wake
and the intrinsic bubble coalescence and breakup and particle entrainment phenomena are
illustrated. Chapter 10 continues the discussion of fluidization under higher-velocity con-
ditions which are characterized by fast fluidization. Fast fluidization is conducted in a
riser of a circulating fluidized bed system where solid particles are circulating in a loop.
This chapter illustrates the interactive relationship of gas-solid flows in a loop situation
by considering the flow behavior of the individual loop components and their effects on
the overall gas-solid flow characteristics. Chapter 11 is concerned mainly with the dilute
transport or pipe flow of gas-solid suspensions. Some pertinent phenomena such as drag
reduction are discussed. Fully developed pipe flow and gas-solid flow in a bend are also
illustrated. Chapter 12 describes transport phenomena underlying heat and mass transfer
in fluidized systems. Transport models and empirical correlations are introduced to allow
heat and mass transfer properties in various fluidized systems to be quantified. An appendix
which summarizes the scalar, vector, and tensor notations presented in the text is provided.
Throughout the text, unless otherwise noted, the correlation equations presented are given
in SI units. Common notations used across the chapters such as superficial gas velocity
Preface xvii
and particle Reynolds number are unified. The solution manual for homework problems is
available. Interested instructors are urged to contact the publisher about it.
The book is intended to be used in various ways depending on the specific information
that the readers desire. While the material is described in a logical sequence chapter by
chapter, each chapter is presented with extensive cross-references and maintains reasonable
independence. Thus, readers who wish to have a quick grasp of a specific subject may go
directly to the relevant chapters. It is important to note that gas-solid flow is a rapidly devel-
oping field of research, and that the physical phenomena of gas-solid flows are so complex
that a comprehensive understanding of the phenomena is far from complete. Therefore, the
present text is also intended to provide readers with ample fundamental concepts to allow
them to follow through new developments in the field.
We would like to express our sincere thanks to the following colleagues who have re-
viewed the text and provided constructive suggestions and overviews: Professor R. S.
Brodkey, Professor R. Clift, Professor J. F. Davidson, Dr. R. Davis, Professor N. Epstein,
Professor J. R. Grace, Dr. K. Im, Professor B. G. Jones, Professor D. D. Joseph, Dr. C.-H.
Lin, Dr. P. Nelson, Dr. S. L. Passman, Professor R. Pfeffer, Professor M. C. Roco, Professor
S. L. Soo, Dr. B. L. Tarmy, Professor U. Tiiziin, and Professor L.-X. Zhou. We are grateful
to Dr. E. Abou-Zeida, Dr. P. Cai, Mr. S. Chauk, Dr. T. Hong, Dr. P.-J. Jiang, Professor
J. Kadambi, Dr. T. M. Knowlton, Dr. S. Kumar, Dr. R. J. Lee, and Dr. J. Zhang for their
valuable technical assistance in providing information which was incorporated in the text.
Special thanks are due to Mr. R. Agnihotri, Dr. D.-R. Bai, Dr. H.-T. Bi, Dr. A. Ghosh-
Dastidar, Mr. E.-S. Lee, Dr. S.-C. Liang, Mr. J. Lin, Mr. T. Lucht, Mr. X.-K. Luo, Dr. S.
Mahuli, Mr. J. Reese, Mr. S.-H. Wei, Dr. J. Zhang, Mr. T.-J. Zhang, and Mr. J.-P. Zhang,
who have read part of the text and have provided valuable comments. The outstanding edi-
torial assistance of Dr. T. Hong and Dr. K. M. Russ is gratefully acknowledged. Thanks are
also extended to Dr. E. Abou-Zeida and Mrs. Maysaa Barakat for their excellent drawing
of the figures. The inquisitive students in the Chemical Engineering 801 course, Gas-Solid
Flows, and the 815.15 course, Fluidization Engineering, taught by the senior author in the
Department of Chemical Engineering at the Ohio State University have provided important
feedback about the text. Their input is indeed extremely helpful. Financial assistance to this
writing project provided by the members of the Ohio State University/Industry Consortium
Program on Fluidization and Particulates Reaction Engineering, including Shell Develop-
ment Co., E. I. duPont de Nemours & Co., Hydrocarbon Research Inc., Exxon Research &
Engineering Co., Texaco Inc., and Mitsubishi Chemical Co., is deeply appreciated.
PART I
Basic Relationships
CHAPTER 1
Size and Properties of Particles
1.1 Introduction
The flow characteristics of solid particles in a gas-solid suspension vary signifi-
cantly with the geometric and material properties of the particle. The geometric properties
of particles include their size, size distribution, and shape. Particles in a gas-solid flow of
practical interest are usually of nonspherical or irregular shapes and poly dispersed sizes.
The geometric properties of particles affect the particle flow behavior through an interac-
tion with the gas medium as exhibited by the drag force, the distribution of the boundary
layer on the particle surface, and the generation and dissipation of wake vortices. The
material properties of particles include such characteristics as physical adsorption, elastic
and plastic deformation, ductile and brittle fracturing, solid electrification, magnetization,
heat conduction and thermal radiation, and optical transmission. The material properties
affect the long- and short-range interparticle forces, and particle attrition and erosion be-
havior in gas-solid flows. The geometric and material properties of particles also represent
the basic parameters affecting the flow regimes in gas-solid systems such as fluidized
beds.
In this chapter, the basic definitions of the equivalent diameter for an individual particle
of irregular shape and its corresponding particle sizing techniques are presented. Typical
density functions characterizing the particle size distribution for polydispersed particle
systems are introduced. Several formulae expressing the particle size averaging methods
are given. Basic characteristics of various material properties are illustrated.
1.2 Particle Size and Sizing Methods

The particle size affects the dynamic behavior of a gas-solid flow [Dallavalle,
1948]. An illustration of the relative magnitudes of particle sizes in various multiphase
systems is given in Fig. 1.1 [Soo, 1990]. It is seen in this figure that the typical range
of particle sizes of interest to gas-solid flows is roughly from 1 /xm to 10 cm. The par-
ticle shape affects the flowability of powders, their packing, and the covering power of
pigments. Qualitative definitions for particle shapes are given in Table 1.1. The shape of
particles is commonly expressed in terms of shape factors and shape coefficients [Allen,
1990].
Particles used in practice for gas-solid flows are usually nonspherical and polydispersed.
For a nonspherical particle, several equivalent diameters, which are usually based on equiv-
alences either in geometric parameters (e.g., volume) or in flow dynamic characteristics
(e.g., terminal velocity), are defined. Thus, for a given nonspherical particle, more than
one equivalent diameter can be defined, as exemplified by the particle shown in Fig. 1.2,
in which three different equivalent diameters are defined for the given nonspherical parti-
cle. The selection of a desired definition is often based on the specific process application
intended.
11 Size and Properties of Particles
Table 1.1. Definitions of Particle Shape
Acicular needle-shaped
Angular sharp-edged or having roughly polyhedral shape
Crystalline freely developed in a fluid medium of geometric shape
Dendritic having a branched crystalline shape
Fibrous regularly or irregularly thread-like
Flaky plate-like
Granular having approximately an equidimensional irregular shape
Irregular lacking any symmetry
Modular having rounded, irregular shape
Spherical global shape
Source: T. Allen's Particle Size Measurements, Chapman & Hall, 1990.
L Typical particle size range covered in this book J

United States Department
of Agriculture (USDA)
.Clay- . Silt >L Sand J* Gravel J classification
- Aerosol reactor Fluidizedbed J
. Air pollution- Applications

Metallized Pneumatic conveying
L- propellant
I rocket
Electric mobility _
-Centrifuge sedimentation.
Measurements
Microscope _
LElectronJ . Sieves.
Imicroscopel
1
(Unit: m)
10- 10" 10" 10" 10" 10 10" 10" l
Figure 1.1. Magnitudes of particle sizes in gas-solid systems (after Soo, 1990).
1.2.1 Equivalent Diameters of a Nonspherical Particle
An equivalent diameter of a particle is usually defined in relation to a specific

sizing method developed on the basis of a certain equivalency criterion. Several equiva-
lent diameters of a spherical particle commonly employed are discussed in the following
sections.
1.2.1.1 Sieve Diameter

A sieve diameter is defined as the width of the minimum square aperture through which
the particle will pass. A common sizing device associated with this definition is a series of
sieves with square woven meshes. Two sieve standards, i.e., Tyler Standard and American
1.2 I Particle Size and Sizing Methods
Minimum
dimension
Figure 1.2. Schematic illustration of multidimensions of a particle and its equivalent volume
diameter, surface diameter, and sieve diameter.
Martin's diameter
Feret's diameter
Projected area
diameter
Figure 1.3. Schematic illustration of different particle diameters based on 2-D projected image.
Society for Testing and Materials (ASTM) Standard, are widely used; they are introduced
in 1.2.2.1.
1.2.1.2 Martin's Diameter, Feret's Diameter, and Projected Area Diameter

Martin's diameter, Feret's diameter, and projected area diameter are three diameters defined
on the basis of the projected image of a single particle. Specifically, Martin's diameter is
defined as the averaged cord length of a particle which equally divides the projected area.
Feret's diameter is the averaged distance between pairs of parallel tangents to the projected
outline of the particle. The projected area diameter is the diameter of a sphere having
the same projected area as the particle. These diameters are schematically represented in
Fig. 1.3. The projected area diameter of a particle d\ can be related to the particle projected
area A by
) (...)
J
Martin's diameter and Feret's diameter of a particle depend on the particle orientation
under which the measurement is made. Thus, obtaining a statistically significant measure-
ment for these diameters requires a large number of randomly sampled particles which are
measured in an arbitrarily fixed orientation. Since Martin's diameter, Feret's diameter, and
projected area diameter are based on the two-dimensional image of the particles, they are
generally used in optical and electron microscopy. The principles of microscopy as a sizing
method are discussed in 1.2.2.2.
1.2.1.3 Surface Diameter, Volume Diameter, and Sauter's Diameter

The surface diameter, d$, volume diameter, dy, and Sauter's diameter, J32, are defined such
that each of them reflects a three-dimensional geometric characteristic of an individual
particle. A surface diameter is given as the diameter of a sphere having the same surface
area as the particle, which is expressed by
= (L2)
* Vx
where S is the particle surface area. A volume diameter is the diameter of a sphere having
the same volume as the particle, which is defined by
dv = ( ) (1.3)
where V is the particle volume. The Sauter's diameter or surface-volume diameter is

defined as the diameter of a sphere having the same ratio of external surface to volume as
the particle, which is given by
The concept of the surface diameter may be mostly used in the field of adsorption
and reaction engineering, where the equivalent surface exposure area is important. The
determination of the surface area depends on the method of measurements; for example,
permeametry can give a much lower area than does gas adsorption. The latter often includes
the contribution of pore surface area, which is accessible to the gas molecules. The deter-
mination of particle surface area by gas adsorption is given in 1.2.2.4. The fundamentals
of gas adsorption are further covered in 1.4.1.
The volume diameter of a particle may be useful in applications where equivalent volume
is of primary interest, such as in the estimation of solids holdup in a fluidized bed or in the
calculation of buoyancy forces of the particles. The volume of a particle can be determined
by using the weighing method. Sauter's diameter is widely used in the field of reacting
gas-solid flows such as in studies of pulverized coal combustion, where the specific surface
area is of most interest.
1.2.1 A Dynamic Diameter

The dynamic response of a particle in gas-solid flows may be characterized by the settling
or terminal velocity at which the drag force balances the gravitational force. The dynamic
diameter is thus defined as the diameter of a sphere having the same density and the same
terminal velocity as the particle in a fluid of the same density and viscosity. This definition
leads to a mathematical expression of the dynamic diameter of a particle in a Newtonian
fluid as
(1.5)
dt =
where Ret is the particle Reynolds number at the terminal velocity; Co is the drag coefficient,
which is a function of Ret; fi denotes the viscosity of the fluid; p and p p represent the
densities of the fluid and the particle, respectively; Upt is the particle terminal velocity; g is
the gravitational acceleration; and dt is the equivalent dynamic diameter.
The relationship between Co and Ret for a sphere is given by Fig. 1.4 [Schlichting,
1979]. Mathematically, it can be expressed by
24
CD = Re t < 2
Ket
18 5 (1.6)
C D = 06 2 < Re t < 500
Ret
C D = 0.44 500 < Ret < 2 x 105
The three correlations in Eq. (1.6), in order from top to bottom, are known as Stokes's,
Allen's, and Newton's equations, respectively. Combining these equations with Eq. (1.5),
O 10
10 10 10 10 10 10 10
Figure 1.4. Drag coefficient for spheres as a function of Ret (from Schlichting, 1979).
the terminal velocity of a sphere is related to its diameter by
RCt
2<Ret<500 (1.7)
C/2 = 3.o 3 d t i p v ~ P>* 500 < Re, < 2 x 105
It is noted that in the laminar flow region, the particle moves in random orientation;
however, outside this region it orients itself so as to give the maximum resistance to the
motion. Thus, the dynamic diameter of an irregular-shaped particle in the intermediate
region may be greater than that in the laminar flow region.
Example 1.1 One of the applications of using Stokes's law to determine the particle
size is the Sedigraph particle analyzer. Table El.l shows the relationship between the
cumulative weight percentage of particles and the corresponding particle terminal velocities
for a powder sample. The densities of the particle and the dispersing liquid are 2,200
and 745 kg/m3, respectively. The liquid viscosity is 1.156 x 10~3 kg/m-s. Find out the
relationship of the mass fraction distribution to the equivalent dynamic diameter.
Table El.l. Cumulative Weight Percentage Versus Terminal Velocity
Upt (m/s) Cumulative wt% Upt (m/s) Cumulative wt%
4.4 x 10"3 99.9 1.1 X 10-5 65.6

2.5 x 10~3 99.3 6.2 x 10"6 47.2
1.7 x 10"3 99.2 2.7 x 10"6 21.2
6.2 x 10"4 98.5 6.9 x 10"7 1.2
1.5 x 10"4 96.0 4.4 x 10"7 1.0
6.9 x 10-5 93.0 2.5 x 10"7 0.8
4.4 x 10-5 90.1 1.7 x 10"7 0.4
2.5 x 10-5 83.5 1.1 x 10"7 0.2
1.7 x 10-5 76.8 2.7 x 10"8 0.1
Solution Rearranging Eq. (1.7), the dynamic diameter for Re t < 2 is given as
Upt (El.l)
( P P - P)g
which yields the dynamic diameter from the given terminal velocity, as given in Table E l .2.
The weight fraction within the range of two neighboring dynamic diameters is also tabulated
in Table E l . 2 , from which the mass distribution versus the dynamic diameter is obtained,
as shown in Fig. E l . l .
Table El.2. Mass Fraction (wt%) Versus

Dynamic Diameter
f/pt (m/s) dt (fim) /M (wt%)

3
4.4 x 10~ 80
0.6
2.5 x 10 3 60
0.1
1.7 x 10 3 50
0.7
6.2 x 10"4 30
2.5
1.5 x 10 4 15
3.0
6.9 x 10"5 10
2.9
4.4 x 10~5 8
2.5 x 10"5 6 6.6
1.7 x 10 5 5 6.7
11.2
1.1 x 10"5 4
18.4
6.2 x 10"6 3
26.0
2.7 x 10~6 2
20.0
6.9 x 10 7 1
0.2
4.4 x 10~7 0.8
2.5 x 10"7 0.6 0.2
0.4
1.7 x 10~7 0.5
0.2
1.1 x 10~7 0.4
2.7 x 10"8 0.2 0.1
0 1 2 3 4 5 6 7 9 10 11 12 13 14 15 dt, \im
Figure El.l. Mass fraction distribution based on data in Table El.2.

10 II Size and Properties of Particles
Table 1.2. Some Methods of Particle Size

Measurement
Method Size range {iim)
Sieving
Woven wire 37-5660
Electroformed 5-120
Punched plate 50-125,000
Microscopy
Optical 0.8-150
Electron 0.001-5
Sedimentation
Gravitational 5-100
Centrifugal 0.001-1,000
Fraunhofer diffraction 0.1-1,000
Doppler phase shift 1-10,000
1.2.2 Particle Sizing Methods

The sizing methods involve both classical and modern instrumentations, based
on a broad spectrum of physical principles. The typical measuring systems may be classi-
fied according to their operation mechanisms, which include mechanical (sieving), optical
and electronic (microscopy, laser Doppler phase shift, Fraunhofer diffraction, transmission
electron miscroscopy [TEM], and scanning electron microscopy [SEM]), dynamic (sed-
imentation), and physical and chemical (gas adsorption) principles. The methods to be
introduced later are briefly summarized in Table 1.2. A more complete list of particle sizing
methods is given by Svarovsky (1990).
1.2.2.1 Sieving
Sieving is the simplest and most widely used technique for powder classification. This
method is based only on the size of the particles and is independent of other particle
properties {e.g., density, optical properties, and surface roughness).
The common sieves are made of woven wire cloth and have square apertures. The sizes
of the sieve openings have been standardized, and currently two different sets of standard
series, the Tyler Standard and the U.S. Series ASTM Standard, are used in the United States.
The mesh number of a sieve is normally defined as the number of apertures per unit area
(square inch). Thus, the higher the mesh number the smaller the aperture. Typical mesh
numbers, aperture sizes, and wire diameters are given for the Tyler sieves and the U.S.
ASTM sieves in Table 1.3. Sieve analysis covers the approximate size range of 37 /im to
5,660 jLtm using standard woven wire sieves. Electroformed micromesh sieves extend the
range down to 5 jj,m or less while punched plate sieves extend the upper limit.
It should be pointed out that longer sieving time can improve the recovery of a given
particle size for a distribution; however, excessive sieving can lead to particle degradation
due to attrition or mechanical wear. This effect can be especially pronounced for particles
near the tail end of the size distribution. Unfortunately, neither good theories nor reliable
empirical formulae are available to estimate the optimum sieving time under which a narrow
error margin of the resulting size distribution can be ensured for a given sample.
1.2 I Particle Size and Sizing Methods 11
Table 1.3. Tyler Standard and U.S. ASTM Sieve Series
Tyler standard U.S. series ASTM standard

Size Wire diameter Size Wire diameter
Mesh no. (jim) Mesh no. (/xm) Om)
H4 5,660
4,760
1,280-1,900
1,140-1,680
H4 5,613
4,699
1,650
1,650
5 4,000 1,000-1,470 5 3,962 1,120
6 3,360 870-1,320 6 3,327 914
7 2,830 800-1,200 7 2,794 833
8 2,380 740-1,100 8 2,362 813
10 2,000 680-1,000 9 1,981 838
12 1,680 620-900 10 1,651 889
14 1,410 560-800 12 1,397 711
16 1,190 500-700 14 1,168 635
18 1,000 430-620 16 991 597
20 840 380-550 20 833 437
25 710 330-480 24 701 358
30 590 290-420 28 589 318
35 500 260-370 32 495 300
40 420 230-330 35 417 310
45 350 200-290 42 351 254
50 297 170-253 48 295 234
60 250 149-220 60 246 179
70 210 130-187 65 208 183
80 177 114-154 80 175 142
100 149 96-125 100 147 107
120 125 79-103 115 124 97
140 105 63-87 150 104 66
170 88 54-73 170 88 61
200 74 45-61 200 74 53
230 62 39-52 250 61 41
270 53 35^6 270 53 41
325 44 31-40 325 43 36
400 37 23-35 400 38 25
1.2.2.2 Microscopy
Microscopy is often referred to as an absolute method for the determination of size and
size distribution of small particles because it allows direct visualization and measurements
of individual particles. Three commonly used types are optical microscopy, transmission
electron microscopy (TEM), and scanning electron microscopy (SEM).
The optical microscope is one of the most basic instruments for particle sizing and is
applicable to a typical size range of 0.8 /xm to 150 /xm. The lower limit is a result of the
diffraction effects on the particle image as observed in a microscope. The limit of resolution
of an optical microscope can be estimated by (Yamate and Stockham, 1977)
1.22A
8= (1.8)
where 8 is the limit of resolution; X is the wavelength of the light; and N A is the numerical
Table 1.4. The Maximum Useful Magnification and the Eyepiece Required
for Different Objectives
Objective
Depth
Focal length Numerical of focus Maximum useful Eyepiece
Magnification (mm) aperture (/xm) magnification required
2.5 56 0.08 50 80 30
10 16 0.25 8 250 25
20 8 0.50 2 500 25
43 4 0.66 1 660 15
97 2 1.25 0.4 1,250 10
Source: A. G. Guy's Essentials of Materials Science, McGraw-Hill, 1976.
aperture of the objective. As an example, for visible light of X = 4,500 A and with an
objective aperture having NA = 1.25, the limit of resolution of the optical microscope can
be calculated from Eq. (1.8) as 0.2 ixm.
An appropriate selection of the maximum useful magnification of an optical microscope
for a given sample is also important. The magnification of the microscope is the product of
the objective-eyepiece combination. As a rule of thumb, the maximum useful magnification
for the optical microscope is 1,000 times the numerical aperture. Table 1.4 summarizes the
maximum useful magnification and the eyepiece required for different objectives.
The TEM and SEM are two advanced techniques which use electron beams for direct
determination of the particle size and surface area. They are usually applied to measurement
of particles in a size range of 0.001 /xm to 5 /xm. The TEM generates an image of a particle
sample on a photographic plate by means of an electron beam, through the transmissibility
of the electron beam on the sample. The SEM uses a fine beam of electrons of medium
energy (5-50 keV) to scan across the sample in a series of parallel tracks. These scanning
electrons produce secondary electron emission, back scattered electrons, light, and X-rays
which can be detected. Both the TEM and SEM are extensively used in the determination
of the pore structure and surface shape and area of the particle. The SEM is considerably
faster and gives more three-dimensional information than the TEM. Details about the TEM
and SEM are given by Kay (1965) and Hay and Sandberg (1967), respectively.
1.2.2.3 Sedimentation
The sedimentation techniques utilize the dependence of the terminal velocities of particles
on their size either in a gravitational field or in a centrifugal field. The lower limit of the
particle sizing by the gravitational sedimentation method is about 5 /zm because of the
effects of convection, diffusion, and Brownian motion as well as the long settling time
involved. These effects can be overcome by centrifuging the suspension, which accelerates
the settling process. Centrifugal sedimentation is mostly applied to the particle size range
of 0.001 /xmto 1 mm.
The sedimentation methods are normally used to measure the size of particles in a liquid
medium because of the relatively high viscosity effects in liquids compared to gases. The
particles in a liquid may become solvated, yielding increased weight and volume of the
particle. Meanwhile, the buoyant effect on the solvated particle in the surrounding medium
1.2 I Particle Size and Sizing Methods 13
increases. In the determination of the overall driving force for sedimentation, these two
effects are noted to cancel each other. Therefore, solvation usually has little effect on the
particle sizing results when the sedimentation methods in liquids are used.
By analogy, the definition of dynamic diameter in a centrifugal field can be simply
extended from Eq. (1.5) with the replacement of the gravitational acceleration, g, by the
centrifugal acceleration, co2r, as
J^-Re t 2
(L9)
where co is angular frequency and r is the radial distance from the center of the centrifugal
field.
1.2.2.4 Gas Adsorption

As indicated earlier, the surface area of porous particles is an important variable in character-
izing physical or chemical processes involving these particles. Porous particles commonly
encountered in catalysis and reaction engineering include activated carbon, alumina, silica,
and zeolites. For a given porous particle, the effective surface area is defined on the basis
of the specific transport phenomenon of interest in a process system. For example, thermal
radiation may be affected predominantly by the external surface area of the particle and the
exposed surface area due to superficial cracks and fissures. On the other hand, for most
chemical reactions and adsorption processes, the internal surface area provided by the inte-
rior pores of the particle may determine the overall rate process. A convenient classification
of pores according to their width divides them into three categories: micropores, less than
20 angstrom (A); mesopores, between 20 and 500 A; and macropores, more than 500 A.
An exception of a large specific surface which is wholly external in nature is provided by
a dispersed aerosol composed of fine particles free of cracks and indentations [Gregg and
Sing, 1982].
The most common method used for the determination of surface area and pore size
distribution is physical gas adsorption (also see 1.4.1). Nitrogen, krypton, and argon are
some of the typically used adsorptives. The amount of gas adsorbed is generally determined
by a volumetric technique. A gravimetric technique may be used if changes in the mass
of the adsorbent itself need to be measured at the same time. The nature of the adsorption
process and the shape of the equilibrium adsorption isotherm depend on the nature of the
solid and its internal structure. The Brunauer-Emmett-Teller (BET) method is generally
used for the analysis of the surface area based on monolayer coverage, and the Kelvin
equation is used for calculation of pore size distribution.
It is noted that in the evaluation of the particle surface diameter and Sauter's diameter,
as discussed in 1.2.1.3, only the external surface area of the particle is considered.
1.2.2.5 Fraunhofer Diffraction

The particle sizing technique using light scattering and diffraction possesses some advan-
tages. It is nonintrusive and much faster than that using a mechanical means, requiring
neither a conducting medium nor a large shearing force. The implementation of Mie theory
Right angle
scattering
Incident light ^ , ^^^ _ ,

^ Backward + A :_ Forward
scattering / \\T^--^ scattering
Figure 1.5. Illustration of the angular light intensity distribution of light scattered from a
single particle.
with Fraunhofer diffraction and side scatter permits the measurement of particle sizes over
a range of 0.1-1000 ^m [Plantz, 1984].
From the Beer-Lambert law, the transmittance for a light beam through a sample of
particles is given by
!
j=cxp(-nAJ) (1.10)
where / t is the intensity of the transmitted beam; h is the intensity of the incident beam; n
is the particle number concentration; Ae is the integrated cross section for extinction, which
includes the effects of reflection, refraction, diffraction, and absorption; and / is the optical
path length. The extinction cross section can be calculated from the Lorenz-Mie theory. A
typical angular distribution of light scattered from a single particle is illustrated in Fig. 1.5.
It shows that the most scattering is in the forward direction.
Although the Lorenz-Mie theory is exact, it does not lead to simple and analytical
solutions relating the particle size to transmittance measurements. However, there are
limiting cases where much simpler theories have been established. These limiting cases
are the Rayleigh scattering for particles much smaller than the wavelength of light and the
Fraunhofer diffraction for particles much larger than the wavelength of light. A criterion
for discerning limiting cases is proposed by van de Hulst (1981) as
K < 0.3 Rayleigh scattering

Kw 1 Lorenz-Mie theory (1-11)
K ^> 30 Fraunhofer diffraction
The parameter K is defined as

-H (M2)
A.
where nx is the relative index of refraction of the particle.

In this book, particles larger than 1 /xm are of primary interest, and thus, only the
Fraunhofer diffraction method, which can account for particles larger than 2-3 /xm, is
discussed here. The Fraunhofer diffraction theory is derived from fundamental optical
principles that are not concerned with scattering. To obtain the Fraunhofer diffraction, two
basic requirements must be satisfied. First, the area of the particle or aperture must be
much smaller than the product of the wavelength of light and the distance from the light
source to the particle or aperture. Second, this area must also be smaller than the product
12 I Particle Size and Sizing Methods 15
Collimated
light
Plane with a circular Observation

aperture plane
(a)
Focus Detection
lens plane
Expanded laser Particle
beam field
(b)
Figure 1.6. Fraunhofer diffraction system for particle size analysis: (a) Diffraction by a
circular aperture; (b) Diffraction by a particle cloud.
of the wavelength and the distance from the particle or aperture to the observation plane.
Therefore, Fraunhofer diffraction is known as far-field diffraction. A schematic diagram
for the Fraunhofer diffraction of a single particle or aperture is illustrated in Fig. 1.6(a),
whereas an optical schematic of a Fraunhofer diffraction instrument for the analysis of
particle sizes in a gas-solid suspension system using a laser beam as the light source is
shown in Fig. 1.6(b).
The transmittance of Fraunhofer diffraction for a circular aperture or spherical particles
of diameter d can be expressed by
(1.13)
where Ji is the first-order spherical Bessel function and x is given by
ndr
x = (1.14)
where r is the radial distance in the observation plane as measured from the optical axis
and F is the focal length of the lens. Thus, the Fraunhofer diffraction pattern for a circular
aperture or spherical particles can be determined as shown in Fig. 1.7. Consequently, by
measuring and analyzing the intensity distributions of the light beam over a finite area of
the detector, the equivalent particle diameter can be obtained. More detailed information
about the Fraunhofer diffraction method is given by Weiner (1984).
16 11 Size and Properties of Particles
x-
Figure 1.7. Fraunhofer diffraction pattern for circular aperture or opaque disk (from Weiner,
1984).
1.2.2.6 Laser Doppler Phase Shift

When a spherical particle enters the crossing volume of two laser beams, a Doppler effect
occurs not only in frequency shift but also in phase shift of the scattered light. The frequency
shift yields the velocity of the sphere, whereas the phase shift gives the particle size. The
phase Doppler principle has been employed to measure the size and size distributions of
spheres in addition to the particle velocity. The phase Doppler principle was first reported
by Durst and Zare (1975) and became a viable measurement tool one decade later [Bachalo
andHouser, 1984].
The phase Doppler principle can be described as follows: When light is scattered by a
small spherical particle traveling through a laser measurement volume, it yields frequency
signals, which can be measured to obtain the particle velocity. This frequency is known as
the Doppler shift frequency, which is identical in all spatial directions. When viewed from
two separate spatial locations the scattered signals exhibit a phase shift whose magnitude
depends on factors including the angle at which light is scattered to each photodetector,
the index of refraction of the material of the spherical particle, and parameters such as
the light wavelength and the beam intersection angle. When reflection is the dominant
mode of scattering, the phase shift is independent of the index of refraction. The phase
shift measured in the Doppler signal obtained from the same particle using two closely
spaced photodetectors varies linearly with the particle diameter for spherical particles and
hence provides a useful means for determining the spherical particle size. Evaluation of
the relationship of the phase shifts from the signals received at each of the photodetector
locations is complex but can be determined on the basis of Mie scattering theory [Bachalo
13 I Particle Size Distributions and Averaged Diameters 17
and Houser, 1984]. In principle, the measurement of particle size requires that the particle
entering the measurement volume be spherical, and the diameters of amorphous particles
cannot be measured using the phase Doppler method.
Typically, the phase Doppler method is good for the measurement of particle sizes
ranging from 1 /xm to 10 mm with a variation by a factor of 40 at one instrument setting.
As a rule of thumb, the maximum measurable concentration is 1,000 particles per cubic
millimeter (mm3). Commercial instruments using this technique are available, e.g., the
phase Doppler particle analyzer (PDPA) (Aerometrics) and the Dantec particle dynamics
analyzer (DPDA) (Dantec Electronics).
1.2.2.7 Coulter Principle

The Coulter principle underlies a method for determining particle sizes and size distributions
using an electrical sensing technique. The instrument based on the Coulter principle is
known as the Coulter counter. In the Coulter counter, particles are first suspended in an
electrolyte and then passed through a small orifice. The particle concentration is usually
so low that particles traverse the orifice one at a time. The orifice has immersed electrodes.
When a particle passes through the orifice, it displaces electrolyte within the orifice, which
results in a change in impedance leading to a voltage pulse with an amplitude proportional
to the volume of the particle. By regulating, sizing, and number counting of the pulses, the
particle size and size distributions are obtained. The typical sizing range by the Coulter
counter is from 1 to 50 /xm.
1.2.2.8 Cascade Impactor

When particles are small enough, the sedimentation method becomes inefficient as a result
of the impractically long settling time. An important design using the inertial technique
is known as the cascade impactor, which samples and classifies particle sizes by their
inertia. A cascade impactor consists of a series of collecting plates of the particle-laden
gas flow, which is gradually increased in the form of a succession of jets. Thus, deflected
by inertia, the particles are collected and graded on the series collecting plates. The extent
of the particle deposition on each plate depends on the impact velocity of the gas stream.
The intake velocity should be low enough to prevent any damage on the collecting plates.
However, it should also be high enough to ensure sufficient inertia of the particles. The
most commonly used cascade impactor is the one developed by May (1945). The May
cascade impactor is capable of sampling airborne particles from 0.5 to 50 /xm by using four
or more collecting glass discs. The particle sizing range by cascade impactors is typically
from 0.1 to 100/xm.
1.3 Particle Size Distributions and Averaged Diameters

For a system of poly dispersed particles, various averaged diameters may be defined
according to the diversity of needs in industrial applications. An averaged diameter depends
not only on the type of particle size distribution but also on the selection of a weighing factor.
A particle size density function can be defined in terms of either the number of particles or the
mass of particles within a given size range. The number density function is interconvertible
with its corresponding mass density function. Different weighing factors with their distinct
physical significance may be imposed to yield various averaged diameters for particles in a
polydispersed system.
1.3.1 Density Functions

A number density function, / N (b), is defined so that / N (b) db represents the particle
number fraction in a size range from b to b + db. Thus,
No
where dN is the number of particles within the size range of b to b + db for a total number No
of the sample particles. Clearly, the preceding expression leads to a normalized condition
:1 (1.16)
o
Thus, over a range from d\ to d2, the fraction of the total sample No of this size is obtained by
- ^ = / fN(b)db (1.17)
No idi
A particle density function can also be defined in terms of the particle mass. A mass
density function, fu(b), represents the particle mass fraction in size by
,, j iviv-/ (1-18)
Mo
where dM is the mass of particles within the size range of b to b + db for a total mass
Mo of the sample particles. Thus, the normalized condition for a mass density function is
given by
/ fM(b)db=l (1.19)
Jo
and, over a range from d\ to d2, the fraction of the total sample M o of the mass is found
from
~ id!
It is noted that the mass of particles can be expressed in terms of the number of particles
of the same size, or
dM=mdN (1.21)
where m is the mass of a single particle of size b. For a spherical particle, m can be
expressed by
m = ^ppb3 (1.22)
6
From Eqs. (1.15), (1.18), and (1.21), the number density function is related to the mass
density function by
fu(b) = ^ h i b ) (1.23)
Mo
The number density function is usually obtained by using microscopy or other optical
means such as Fraunhofer diffraction. The mass density function can be acquired by use
of sieving or other methods which can easily weigh the sample of particles within a given
size range.
1.3 I Particle Size Distributions and Averaged Diameters 19
1.3.2 Typical Distributions

In the applications of gas-solid flows, there are three typical distributions in particle
size, namely, Gaussian distribution or normal distribution, log-normal distribution, and
Rosin-Rammler distribution. These three size distribution functions are mostly used in the
curve fitting of experimental data.
1.3.2.1 Gaussian Distribution

The Gaussian distribution, also known as the normal distribution, has the density function
(d-do)2'
fN(d) = A N exp - (1.24)
2*1
where AN is the normalizing constant; do is the arithmetic mean of d\ and crd is the standard
deviation of d. Therefore, as given in Fig. 1.8, 2\/2<rd is the width of the distribution curve
defined as the cord length of two points where
/NW) 1 ,-. oc.
JNWO) e
For a given sample, the particle size range is bounded by d\ and di as shown in Fig. 1.8.
Thus, Eq. (1.16) becomes
rA
AN exp ( 5 I db = 1 (1.26)
/ A N exp
from which AN is obtained as
-1
(1.27)
djd, d/d0
Figure 1.8. Gaussian distribution function.

Given the number density function of Eq. (1.24), the corresponding mass density function
becomes
Md) = AM^ppd3exp (-^'^ ) d-28)
The normalizing constant AM can be calculated from
* = l (1.29)
There is no simple, exact, and explicit expression for AM- However, for the case of a very
narrow size distribution where cr^/do <3C 1, AN and AM are given by
AN L - (1.30)
V2
and
For particle sizes following the Gaussian distribution described by Eqs. (1.24) and (1.30),
95 percent of the particles are of sizes between (2cr d + do) and (2ad + do).
1.3.2.2 Log-Normal Distribution

Most systems of fine particles have the log-normal type of particle size distribution. That
is, with the logarithm of the particle size, the particle size distribution follows the normal or
Gaussian distribution in semilog scales. Therefore, the density function for the log-normal
distribution can be expressed by
N2
1 [ 1 flnd-lnd(oi
Md) = j= exp - - i - (1.32a)
L
or
fsQnd) = -7= exp
Here, do\ and ad\ are parameters defining the log-normal distribution. do\ is the median
diameter, and && is the natural log of the ratio of the diameter for which the cumulative-
distribution curve has the value of 0.841 to the median diameter. In doi and ad\ are not
equivalent to the arithmetic mean and the standard deviation of In d, respectively, for the
log-normal distribution (Problem 1.3). Note that, for the log-normal distribution, the particle
number fraction in a size range of b to b + db is expressed by fy(b) db; alternatively, the
particle number fraction in a parametric range of In b to In b + d(ln b) is expressed by
1.3.2.3 Rosin-Rammler Distribution

For broken coal, moon dust, and many irregular particles, the mass distribution is found to
follow a form known as the Rosin-Rammler distribution. A Rosin-Rammler distribution
has the density function
Md) = aMa~X expC-^") (1.33)
131 Particle Size Distributions and Averaged Diameters 21
Table 1.5. a and P for Some Materials
Material a P x lOVm)- 1
(a) Fine grinding

Marlstone 0.675 33
Marlslate 0.839 33
Brown coal (lignite) 0.900 63
Feldspar 0.900 71
Cement clinker 1.000 29
Glass powder 1.111 25
Coal 1.192 21
(b) Coarse grinding
Fullers clay 0.727 0.40
Coal, type 1 0.781 0.067
Coal, type 2 0.781 0.15
Limestone with 7% bitumen 0.781 0.13
Limestone, medium hardness 0.933 0.083
Limestone, hard 1.000 0.40
Clinker 1.036 0.50
Feldspar 1.111 0.50
Source: G. Herdan's Small Particle Statistics, Butterworths,

London, 1960.
where a and ft are constants. Integrating Eq. (1.33) yields the cumulative distribution
function, F , as
fd
= / fu(b) db=l- exp(-pda) (1.34)

Jo
However, the Rosin-Rammler distribution is often expressed in terms of R defined by
f
R = Pd a) (1.35a)
d
Jd
Then, we have
1 \
(1.35b)
Equation (1.35b) shows that a linear relationship exists when ln[ln(l//?)] is plotted against
Ind. From the slope and intercept of this straight line, a and p can be determined, a and
P are typically obtained from the particle size distribution data based on sieve analyses.
Table 1.5 provides a list of typical values of a and P for some materials for the Rosin-
Rammler density function with d in the function having the unit micrometers (/xm).
Example 1.2 A coarsely ground sample of corn kernel is analyzed for size distribution,
as given in Table El.3. Plot the density function curves for (1) normal or Gaussian distri-
bution, (2) log-normal distribution, and (3) Rosin-Rammler distribution. Compare these
distributions with the frequency distribution histogram based on the data and identify the
distribution which best fits the data.
Table El.3. Data of Size Distribution
Size range Number of Size range Number of

(mm) particles (mm) particles
0.05-0.10 1 0.50-0.55 3
0.10-0.15 5 0.55-0.60 1
0.15-0.20 6 0.60-0.65 2
0.20-0.25 7 0.65-0.70 0
0.25-0.30 8 0.70-0.75 1
0.30-0.35 6 0.75-0.80 0
0.35-0.40 4 0.80-0.85 1
0.40-0.45 4 0.85-0.90 0
0.45-0.50 4 0.90-0.95 1
Solution The data on numbers of particles in each particle range given in

Table El.3 can be converted to relative frequencies per unit of particle size as given in
Table El .4. The histogram for the relative frequency per unit of particle size for the data is
plotted in Fig. El.2; the histogram yields a total area of bars equal to unity. Superimposed
on the histogram is the density function for the normal distribution based on Eqs. (1.24)
and (1.30). For this distribution, the values for d0 and crd are evaluated as 0.342 and 0.181,
respectively. Also included in the figure is the density function for the log-normal distribu-
tion based on Eq. (1.32a). For this distribution, the values for In doi and crd\ are evaluated
as -1.209 and 0.531, respectively.
Table El.4. Relative Frequency per Unit of Particle Size Data Given in Table El.3
Particle size Relative frequency per

(averaged, mm) Number of particles Relative frequency unit of particle size
0.075 1 0.019 0.370

0.125 5 0.093 1.852
0.175 6 0.111 2.222
0.225 7 0.130 2.593
0.275 8 0.148 2.963
0.325 6 0.111 2.222
0.375 4 0.074 1.481
0.425 4 0.074 1.481
0.475 4 0.074 1.481
0.525 3 0.056 1.111
0.575 1 0.019 0.370
0.625 2 0.037 0.741
0.675 0 0.000 0.000
0.725 1 0.019 0.370
0.775 0 0.000 0.000
0.825 1 0.019 0.370
0.875 0 0.000 0.000
0.925 1 0.019 0.370
Sum 54 1.000 20.00
For the Rosin-Rammler distribution, the distribution constants (a and /?) are obtained
from the particle mass distribution data. To obtain the mass density distribution, the data on
1.3 / Particle Size Distributions and Averaged Diameters 23
the number density distribution given in Table El .3 need to be converted by using Eq. (1.23).
From the converted data and with the least-square fitting based on Eq. (1.35b), a and ft can
be obtained as a = 3.71 and fi = 4.88 mm" 1 . Note that the unit for d in Eq. (1.33) would
be millimeters when these a and f$ values are used. Thus, the mass density function of the
Rosin-Rammler distribution can be calculated. Converting the mass density function to
the number density function, the results for the Rosin-Rammler distribution are plotted as
shown in Fig. El.2.
7.0
6.0 Given data
Normal distribution
5.0
Log-normal distribution
4.0 Rosin-Rammler distribution
0.075 0.275 0.475 0.675 0.875

Diameter, mm
Figure E1.2. Comparisons of the relative frequency distribution based on the data with three
density functions.
A graphical comparison of the three distributions with the given data shown in the figure
reveals that the log-normal distribution best approximates the data.
1.3.3 Averaged Diameters of a Paniculate System

For a given size distribution, various averaged diameters can be calculated, depend-
ing on the forms of weighing factors. The selection of an appropriate averaged diameter
of a particle system depends on the specific needs of the application. For instance, in a
pulverized coal combustion process, the surface area per unit volume may be important. In
this case, Sauter's averaged diameter should be chosen.
1.3.3.1 Arithmetic Mean Diameter
The arithmetic mean diameter d\ is the averaged diameter based on the number density
function of the sample; d\ is defined by
f
Jo 3
bfN(b)db
(1.36)
f
Jo
Mb) 6b
24 1 / Size and Properties of Particles
1.3.3.2 Surface Mean Diameter

The surface mean diameter J s is the diameter of a hypothetical particle having the same
averaged surface area as that of the given sample; ds is given by
4 =^ (1-37)
Jo
1.3.3.3 Volume Mean Diameter
The volume mean diameter dy is the diameter of a hypothetical particle having the same
averaged volume as that of the given sample; dy is determined by
3
[bMb)db
dY = ^ 5 5 (1.38)
Mb)db
1.3.3.4 Sauter's Mean Diameter

Sauter's mean diameter d& is the diameter of a hypothetical particle having the same
averaged specific surface area per unit volume as that of the given sample; J32 is defined by
/ b2Mb)db
Jo
Note that Sauter's mean diameter in Eq. (1.39) is defined for a range of particle size, which
is different from Sauter's diameter in Eq. (1.4), defined for a single particle size.
1.3.3.5 DeBroucker's Mean Diameter

DeBroucker's mean diameter J43 is the averaged diameter based on the mass density function
of the sample; ^43 is evaluated by
p oo poo
b44fN(b)db // bfM(b)db
I
= 44oo d-40)
b3Mb)db / fM(b)db
Jo
1.4 Material Properties of Solids

The material properties of solids are affected by a number of complex factors. In
a gas-solid flow, the particles are subjected to adsorption, electrification, various types of
deformation (elastic, plastic, elastoplastic, or fracture), thermal conduction and radiation,
and stresses induced by gas-solid interactions and solid-solid collisions. In addition, the
particles may also be subjected to various field forces such as magnetic, electrostatic, and
gravitational forces, as well as short-range forces such as van der Waals forces, which may
affect the motion of particles.
In this section, we briefly discuss several aspects of the material properties of solids that
are of interest to gas-solidflowapplications. They include physical adsorption, deformation
1.4/ Material Properties of Solids 25
and fracture, thermal properties, electrical properties, magnetic properties, and optical
properties.
1.4.1 Physical Adsorption

The physical adsorption phenomenon (also see 1.2.2.4) was first described by de
Saussure in 1814 [Bikerman, 1970]. In his experiment, activated carbon particles were put
into a flask containing chlorine. The green color of the gas disappeared. With the heating
of the flask, the greenish chlorine gas reappeared. This phenomenon, known as adsorption,
is the accumulation of gas on a solid surface.
The outermost layer of the solid molecules on the surface is bound on only one side
to the inner layer of molecules by atomic and molecular forces. To compensate for this
imbalance of the binding force, an attraction force from the solid surface captures the
surrounding gas, vapor, or liquid. This attraction can be either physical or chemical, de-
pending on the temperature and the interacting forces between the solid and the surrounding
fluid.
Physical adsorption, or van der Waals adsorption, results from a relatively weak interac-
tion between the solid and the gas. The forces responsible for adsorption are "dispersion"
forces (characterized by London; see 3.3.1) and/or electrostatic forces (Coulombic; see
3.3.2) if either the gas or the solid is polar in nature. Physical adsorption is reversible;
hence all the gas adsorbed by physical adsorption can be desorbed by evacuation at the same
temperature. Chemical adsorption is a result of a more energetic interaction between the
solid and the gas than that of physical adsorption. Reversal of chemical adsorption using
a vacuum requires elevated temperature, and even that may not be sufficient. Physical
adsorption, being of more interest in gas-solid flows, is the focus of the following sections.
1.4.1.1 Freundlich Isotherm

The most common way to present adsorption data is an adsorption isotherm. The amount
adsorbed is plotted as a function of the partial pressure of the adsorbate, p, at a constant
temperature. The first empirical isotherm was proposed by Freundlich in 1906 [Freundlich,
1926], and since then, a number of empirical adsorption isotherms have been reported. The
Freundlich isotherm is usually written in the form
VR = apfi (1.41)
where Va is the volume of gas adsorbed while a and ft are empirical constants derived from
linear graphs of log( Va) against log(p).
1.4.1.2 Langmuir Isotherm

Langmuir (1918) laid down a systematic and theoretically grounded adsorption isotherm,
the well-known Langmuir isotherm, which has served as a basis for a number of future
isotherms, such as the BET isotherm, described in 1.4.1.3. The basic assumptions of the
Langmuir adsorption model include monolayer coverage, localized adsorption (i.e., the
specific adsorption sites exist and the interactions are between a specific molecule and the
site), and homogeneous materials. In addition, the heat of adsorption is independent of
coverage. He proposed that the rate of adsorption is proportional to the dimensionless
pressure of the adsorbate, p/po (= /?*), where po is the saturated vapor pressure, and the
percentage of unoccupied adsorption sites (1 ), where is the ratio of the number of

occupied sites to the total number of adsorption sites on the solid surface. At equilibrium,
the rate of adsorption balances the rate of desorption so that
V d - O = Kdt (1.42)
where Ka is the adsorption rate constant and K& is the desorption rate constant. When the
entropy effects are neglected, the equilibrium constant, Ke(= KJK&), can be defined as
a 4 3 )
where AH is theheat of adsorption per mole at the standard temperature and pressure
(STP) condition. R^ is the universal gas constant, which is equal to 82.06 x 10~3 m3 atm/
kgmole K. The Langmuir isotherm is rearranged to the linearized form
Therefore, the plot of ~1 against p*~l yields Ke from the linearized data.
1.4.1.3 Brunauer-Emmett-Teller (BET) Isotherm

The assumption of monolayer adsorption in the Langmuir isotherm model is unrealistic in
most cases, and a modification to multilayer adsorption should be considered. In 1938,
Brunauer, Emmett, and Teller modified the Langmuir approach of balancing the rates of
adsorption and desorption for the various molecular layers [Brunauer et ai, 1938]. This
approach is known as the BET method. The BET isotherm assumes that the adsorption of
the first layer has a characteristic heat of adsorption AHa and the adsorption and desorption
on subsequent layers are controlled by the heat of condensation of the vapor, AHC. The
derivation of the BET equation is beyond the scope of this book; however, a common form
of the BET equation is given as
' ^ ^ U.45)
Va(/>o - p) V mC VmCp0
where Vm is the monolayer capacity at STP and C is given as
Most commercial instruments using gas adsorption for surface area and porosity deter-
mination are based on the BET isotherm. In Eq. (1.45), the monolayer capacity Vm can be
used to calculate the surface area on the basis of the area occupied by each adsorbed gas
molecule. According to Eq. (1.45), a plot of p/[V&(po p)] versus p/po is linear. From
the slope and the intercept, Vm can be obtained. Thus, the specific surface area Sy can be
obtained as
Sy = ^ ^ (1.47)
Vgmm
where A is Avogadro's number (6.02 x 1023); sm is the area per molecule of the adsorbed
gas; Vgm is the molar volume of the adsorbed gas; and m is the mass of the solid.
1.4 /Material Properties of Solids 27
Example 1.3 From the data given in Table El.5, estimate the specific surface area of
22.5 mg of Ca(OH)2 powder. The data are for adsorption of nitrogen at a temperature of
195.8C, with the specific surface area of the solid adsorbent S w = 4.35 Vm/m (m2/g).
Table E1.5. BET Data for
Specific Surface Area
Calculation ofCa(OH)2
P/Po V a (cm 3 )
0.1 0.1855
0.2 0.2113
0.3 0.2472
0.35 0.2652
Solution The data given are plotted in the form as required by BET equation,
i.e., Eq. (1.45), as shown in Fig. E1.3, from which the intercept on the ordinate / = 0.0122
(cm" 3) and the slope of the curve s = 5.77 (cm" 3 ). The values of s and / are used to obtain
Vmby
1
Vm = = 0.173 cm3 (El.2)
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Figure E1.3. Plot of BET equation for data given in Table E1.5.
28 1 /Size and Properties of Particles
The specific surface area of Ca(OH) 2 can be obtained by substituting Vm and the weight of
the solid in the expression for Sy.
Sy = 4.35 x ^ ^ - = 33.4m 2/g (E1.3)
1.4.2 Deformation and Fracture

All materials undergo a deformation, either noticeable or unnoticeable, whenever
they are subjected to an external force. The deformation can be elastic, plastic, elastomeric,
or viscoelastic. When an external force is applied on a body, the displacement of points of
the body relative to neighboring points is measured as strain and the strength of the force
applied to the local point is measured as stress. Elastic deformation such as rubber band
stretch is recovered when the stress, or external force, is removed. Plastic deformation such
as a dent on a metal car body is permanent and is not recoverable with the removal of the
deforming stresses.
1.4.2.1 Stress and Strain
Deformation is caused by stress from either an external force or an imbalance of internal
forces. Quantitatively, a stress a on an area of a specimen is equal to the force applied per
unit area. Since a force is a vector with three components, the stress component from the
normal component of the force is called normal stress; it causes elongation or contraction
of the material depending on the direction of the force. The stress components from the
two tangential components of the force are called shear stresses; they are responsible for
the shear deformation.
Imagine a small block cut from a material with its faces perpendicular to the three axes
of an orthogonal coordinate system {e.g., Cartesian coordinates). Across any face, a force
is exerted on the block by the adjoining material. Thus, a total of nine stress components
can be obtained from the three faces perpendicular to the three coordinates. Once the small
block shrinks to a point, the group of nine stresses, usually in the form of a stress tensor, can
give information about the internal force at a local point within the material. For a symmetric
stress tensor (such as might exist for a homogeneous material), the number of independent
stress components reduces to six, three normal components and three shear components. It
should be pointed out that a stress tensor, similarly to a vector, is independent of the choice
of the orthogonal coordinate systems.
Although stress and strain always occur together, the two are distinguishable. Basically,
a stress is associated with the forces applied to a local area, whereas a strain is related to
the displacements which give rise to changes in shape and size of the material in question.
Therefore, in rigid-body motion where no relative displacement among the points in the
body occurs, the strains are always zero. Analogously to the nine stress components at a
local point, the local total displacements lead to nine independent strains, i.e., three normal
strains and six shear strains.
A strain is equal to the ratio of the displacement AL caused by a stress to the total length
L. Two common methods for strain calculation are used, depending on the magnitude of the
strain. When a strain is small, the reference axes remain virtually unchanged so that L essen-
tially maintains its initial length L. Hence the strain, known as the nominal strain, is simply
e= ^ (1.48)
In the cases where L changes appreciably from its initial value, the method for so-called true
strain should be used instead of the nominal strain method. The true strain is calculated by
1.4.2.2 Hooke 's Law and Elastic Constants

The mathematical relationship between the stress and the strain depends on material prop-
erties, temperature, and the rate of deformation. Many materials such as metals, ceramics,
crystalline polymers, and wood behave elastically at small stresses. For tensile elastic de-
formation, the linear relation between the stress, <r, and strain, 6, is described by Hooke's
law as
= % (1.50)
where E is a material property known as Young's modulus. A material is anisotropic if the
value of a given property depends on the direction in which testing is performed. Typical
examples of anisotropic materials include sapphire crystal, whose atomic distribution varies
with the testing directions, and wood, whose specific stiffness (defined as the ratio of Young's
modulus to the material density) is much larger in the longitudinal direction (with the grain)
than in the radial or tangential direction (against the grain). Poly crystalline materials such
as copper are generally isotropic as a result of the random orientations of their crystals.
For isotropic materials, a general relationship correlating the three normal components
of the stress tensor and strain is expressed as
ei = ^--(<r 2 + a 3 : 4 (1.51)
tL tL
where \ is the tensile strain in the direction of a\, and cr2 and a 3 are the transverse stresses.
This strain increases in proportion to the longitudinal stress &\ and decreases with the
transverse stresses a2 and 03 as a result of the Poisson contraction. The degree of the
Poisson contraction is determined by Poisson's ratio, v, i.e., the ratio of the contraction to
the extension in the direction of a given applied stress. Young's modulus and Poisson's
ratio are among the most important elastic properties of materials. Typical values of E and
v of some materials are listed in Table 1.6.
1.4.2.3 Plastic Deformation

Most solids are subjected to permanent deformation or breakup once the applied stresses
exceed a certain limit. Hence, most solid particles may be classified into two categories:
elastoplastic particles and elastic-brittle particles. Typical elastoplastic materials include
metals and polymers, while typical elastic-brittle materials include coal, activated carbon,
and ceramics. Materials that are elastoplastic at room temperature may become brittle at
low temperatures and those that are brittle at room temperature may become plastic at high
temperatures.
Some plastic materials have different tensile and compressive characteristics. For ex-
ample, polystyrene is tough under compressive load but very brittle in tension. However,
for most elastoplastic materials, the stress-strain curves in compression are the same as
in tension. Hence, the deformation properties of these materials in tension may also be
applied to those in compression, which is of great interest to gas-solid flows.
For elastoplastic materials, two stresses representing the yield strength and the tensile
strength are important to consider. The yield strength is the stress at which appreciable
Table 1.6. Elastic Properties of Some Solid Materials
Young's modulus Poisson's ratio

Material E, 1010N/m2 v
Graphite 100
AI2O3 crystals
[1010] 230
[1120] 125
[0001] 48
Boron 45 0.21
Sintered carbide (WC) 65 0.20
Glass-ceramic 10 0.25
Silica glass 8 0.24
Aluminum alloys 7 0.33
Steel 20 0.28
Tungsten 41 0.28
Wood (typical)
Longitudinal 1 0.04
Radial 0.07 0.3
Tangential 0.06 0.5
Copper alloys 12 0.35
Nylon 0.3 0.48
Polyethylene 0.04 0.3
Source: A. G. Guy's Essentials of Materials Science, McGraw-

Hill, 1976.
plastic deformation begins, whereas the tensile strength is the maximum nominal stress that
a specimen can support before fracture. The yield strength is generally defined as the 0.2
percent offset yield stress, i.e., the point at which the stress-strain curve crosses a line offset
0.2 percent along the strain axis parallel to the linear part of the stress-strain curve (see
Fig. 1.9). Yield strength values of some typical materials are given in Table 1.7.
In contrast to the simplicity of elastic deformation, plastic deformation occurs in diverse
ways. Figure 1.9 illustrates the stress-strain curves for two typical elastoplastic materials
(hardened metal and polymer). Both materials show similar linear relationships between
stress and strain for the elastic deformation {i.e., before yield strength) but quite different
correlations in the yielding processes before fracture.
1.4.2.4 Fracture and Fatigue

Besides deformation, fracture is the other response of materials to a stress. Fracture is the
stress-induced breakup of a material. Two types of fracture are commonly defined. A brittle
fracture is breakup which occurs abruptly without localized reduction in area. A ductile
fracture is the failure of the material which is preceded by appreciable plastic deformation
and localized reduction in area (necked region). The brittle fracture and ductile fracture are
schematically illustrated in Fig. 1.10.
For brittle materials, the compressive strength is usually an order of magnitude greater
than the tensile strength, and the fracture occurs by a shearing process [Guy, 1976]. For
Table 1.7. Yield and Tensile Strength of Some Solid Materials
Material Yield strength (MPa) Tensile strength (MPa)
(a) Metals and alloys

Carbon steels 177-571 320-830
ASTM steels 204-340 306-680
Stainless steels 204-1,800 544-1,900
Cast aluminum alloys 55-435 35-70
Wrought aluminum alloys 27-618 68-666
Bronze 68-410 230-660
Brass 68-440 265-680
(b) Polymer
Nylons 58-86 65-86
PVC-acrylic alloy 37-44 38-74
Polyacetals 60-68 60-68
Polystyrene 34-69 34-69
(c) Ceramics (bending strength)
Alumina 272-320 136-272
Electric porcelain 34-82 14-48
Polycrystalline glass 102-170
Mica 82-88 34-41
Graphite and carbon 4-27 5-108
Source: E. R. Parker's Material Data Book, McGraw-Hill, 1967 and J. R

Shackelford and W. Alexander's CRC Materials Science and Engineering Hand-
book, CRC Press, 1992.
Tensile strength
Fracture
Yield strength
0. .2 .4 .6 .8 1.
Strain, %
10.0 20.0 30.0 40.0
Nominal strain, %
Figure 1.9. Deformation of typical elastoplastic materials (hardened metal and polymer) in the
stress-strain diagram (after Guy, 1976).
Brittle fracture Ductile fracture

Figure 1.10. Brittle fracture and ductile fracture.
ductile materials, fracture requires a large amount of energy compared to that necessary for
brittle fracture.
Fractures usually refer to a single cycle of either compressive or tensile loading. A
more common failure of a material is caused by repeated stressing. This type of failure is
known as fatigue. The fatigue process in a clean ductile material involves crack initiation,
crack growth, and tensile failure at which the crack attains a critical size. Crack initiation
is due to the characteristic that plastic deformation is not completely reversible during a
cyclic loading. In many engineering materials, fatigue failures can occur well below the
theoretical predictions because impurities and heterogeneities within the material structure
permit plastic deformations to occur in areas of local weakness.
1.4.3 Thermal Properties

When a solid is exposed to a thermal environment, it will either absorb or release
heat. This thermal energy is delivered via heat transfer mechanisms of conduction, con-
vection, and radiation. Typical thermal properties of a solid include specific heat, thermal
conductivity, thermal expansion, and thermal radiation properties.
1.4.3.1 Specific Heat and Coefficient of Thermal Expansion

The specific heat at constant pressure, cp, is defined as the amount of heat needed to increase
the temperature of 1 kg of material by 1 K. Thus, cp can be expressed as
Cp (1.52)
~ mAT
where AQ is the heat absorbed over a temperature increase of AT and m is the mass of
the testing material. The heat absorbed is mainly used as an energy source for increased
vibration of the crystal lattice. Since most solid materials exhibit similar lattice vibrations
1.4 / Material Properties of Solids 33
at room temperature, they have similar values of c p . This value of c p can be estimated as
(1.53)
where RM is the gas constant (see Eq. (6.37)). At reduced temperatures, c p initially decreases
gradually but rapidly approaches zero as the temperature approaches 0 K.
When an object is heated or cooled, the characteristic length of the body is expanded or
contracted by thermal expansion. Thermal expansion also changes stresses in the material
and, in some cases, makes stresses large enough to cause fracture. The simplest form for a
linear thermal expansion is given by
^ (1.54)
LJ
where a^ is the coefficient of thermal expansion and AL is the change in length.
I A3.2 Thermal Conductivity

When a temperature gradient exists in a material, energy in the form of heat is conducted
from a high-temperature region to a low-temperature region through intermolecular and
atomic impacts, lattice vibrations, and transport of electrons. This type of thermal energy
transfer is called conductive heat transfer. The relation between heat flux induced by thermal
conduction and temperature can be described by Fourier's law as
Jq = -KVT (1.55)
where Jq is the heat flux vector driven by the temperature gradient V77, and the proportion-
ality coefficient K is defined as thermal conductivity.
The thermal conductivity K mainly varies with temperature and type of the material. For
solid materials, the quantum of lattice vibrational energy is known as a phonon. Heat con-
duction in solid material can be explained by the movement of phonons and electrons. The
electrons acquire energy from a higher-energy phonon and carry it to a lower-temperature re-
gion, and then transfer the energy to a lower-energy phonon. For good metallic conductors,
conduction of heat by electrons produces distinctly higher values of thermal conductivity
than that from lattice vibration or phonon mechanism. This explains why a good thermal
conductor, such as copper or silver, is usually a good electric conductor. For poor metallic
conductors, such as a highly alloyed metal, phonons and electrons contribute comparably
to the thermal conductivity. At ambient or higher temperatures, the thermal and elec-
tric conductivities of good metallic conductors can be described by the Wiedemann-Franz
law, which has the following expression when the energy transfer by means of phonons is
neglected
K = LooT (1.56)
8 2
where Lo is the Lorentz constant with a value of 2.45 x 10~ W 2/K and a is the electrical
conductivity, discussed in 1.4.4.1.
1.4.3.3 Thermal Radiation Properties

Radiative heat transfer of a solid involves both receiving and emitting radiant energy. Most
solids in gas-solid flows can be regarded as thermally opaque. When a radiant heat flux
strikes a solid surface, the incident energy is both reflected and absorbed with little heat
transmitted. The fraction of absorbed radiant energy defines the thermal absorptivity of the
solid, whereas the fraction reflected defines the thermal reflectivity of the material. Thus,
we have
aR + PR = 1 (1.57)
where aR is the absorptivity and PR is the reflectivity.
The emissive power of a solid ER is defined as the energy emitted by the body per
unit area and per unit time. The body having a theoretical maximum emissive power at a
given temperature is called a blackbody. The actual emissive power of a solid at a given
temperature is less than or equal to that of the blackbody for the same temperature. Hence,
we define the emissivity of a solid as the ratio of the emissive power of the solid to the
emissive power of a blackbody at the same temperature, which is
(1.58)
For a blackbody, ERh is given by Stefan-Boltzmann's law

^ R b Gbl U-J y )
where ob is the Stefan-Boltzmann constant, which has a value of 5.67 x 10~ W/m K4.
8 2
From Kirchoff's law, we have

aR = R (1.60)
In general, the emissivity of a solid is affected by the temperature as well as the wave-
length of the radiation. The concept of monochromatic emissivity is related to the radiant
emission by a solid at a specific wavelength. The monochromatic emissivity x is defined
as the ratio of the monochromatic-emissive power of a solid Ex to the monochromatic-
emissive power of a blackbody Ebk at the same temperature and wavelength, i.e.,
Moreover, a solid is termed a gray body if its monochromatic emissivity e^ is indepen-

dent of the wavelength. The monochromatic-emissive power of a blackbody at a specific
temperature and wavelength is given by Planck's formula
(L62)
T
where X is wavelength (fim); T is absolute temperature (K); C\ = 3.743 x 108 (W ^m 4 /m 2 );
andC2 = 1.4387xlO 4/zm-K. In addition, the wavelength Am at which the monochromatic-
emissive power of a blackbody is maximum for a given temperature can be derived from
Eq. (1.62) as
XmT = C3 (1.63)
where C3 = 2898 /xm K. Equation (1.63) is known as Wien's displacement law.
The total emissivity of the solid is related to the monochromatic emissivity by
The thermal properties of some solid materials are given in Table 1.8.
1.4 / Material Properties of Solids 35
Table 1.8. Thermal Properties of Solids
Thermal conductivity Coefficient of thermal

Material K, W / m K expansion a T , 10~6 K"1
Copper 390 18
Beryllia (BeO) 250 10
Aluminum 200 25
Graphite 150 2
Sintered carbide (WC) 80 7
Steel 50 12
Boron nitride (BN) 28 8
A12O3 ceramics 17 7
Pyrex glass 1.2 3
Kovar 15 5
Soda-lime glass 0.8 9
Polyethylene 0.3 300
Rubber 0.1 670
Urethane foam 0.05 90
Invar 11 1
1.4.4 Electrical Properties

Electrons are driven to move by electric force when an electrical potential differ-
ence exists in a material. The resistance R to this force depends on the cross-sectional area
A, the length /, and the conductivity a of an object as
R = -t- (1.65)
aA
1.4.4.1 Electrical Conductivity

The electrical conductivity a is defined as the electrical current density or the amount of
charges passing through a unit cross-sectional area per second in an electrical field with
strength E of 1 V/m. The electrical conductivity can be determined from the particle
number concentration n, the charge on a particle q, and the mobility of a particle JJL by
a = nqiJL (1.66)
The mobility here is the average velocity of a particle that accelerates between successive
collisions under unit electric driving force {i.e., electric field strength E of 1 V/m).
For metallic materials, charges are conducted by electrons or holes because most of
the energy bands of metals are partially filled. A common equation for the electrical
conductivity of metals can be derived from the following simple model. Applying Newton's
second law to an electron in a crystal, it yields
dve
-eE = me-^ (1.67)
at
where me is the effective mass of an electron and ve is the electron velocity. The electron
will accelerate continuously under the driving force eE until it collides with impurities,
lattice imperfections, etc. Assume that the velocity of the electron drops to zero at each
collision and the average time between collisions is r so that
eEr=mevem (1.68)
where vem is the maximum velocity gained just before the next collision. From the definition
of mobility, which equals the average velocity per unit electrical field strength,
âv vem ex
M = = - - = - (1.69)
E IE 2me
Therefore, the electrical conductivity of metals is obtained from Eqs. (1.66) and (1.69)
as
ne z
a=nefi = (1.70)
2me
For nonmetallic substances, the electrons cannot move as freely as in the case of met-
als because their energy bands are essentially completely full or empty. The electrical
conductivity in nonmetallic materials is dominated by another mechanism, i.e., the defect
mechanism, instead of electron conduction. In ionic crystals such as salts (e.g., sodium
chloride), two types of ions, cations and anions, are driven to move by the electrical force
qE once an electrical field is applied. The ions can move only by the defect mechanism;
that is, they exchange position with a vacancy of the same type. At the room temperature,
the fraction of vacancies for salt is very small (of the order of 10~17) with low exchange
frequency (of the order of 1 Hz) so that electrical conductivity is extremely low. Although
impurities and high temperature can affect electrical conductivity by a large factor, non-
metallic materials generally have very low electrical conductivity and these substances are
widely used as electrical insulators.
1.4.4.2 Permittivity
Permittivity e is the most basic property of a dielectric material. To understand the concept
of permittivity, consider first a parallel-plate capacitor with two metallic plates. When a
voltage is applied across the metallic plates, charges are generated on the surface of each
plate. The charge Q on the plate surface is given by
Q = eAj = CV (1.71)
where A is the surface area of the plate; V is the voltage; / is the distance between the
plates; is the permittivity of the medium; and C is the capacitance, which depends on the
geometry of the capacitor and the medium with which it is filled.
In the case of a vacuum between the plates, the permittivity (e 0) has a value of 8.854 x
10~12 F/m. When a dielectric material fills the space between the plates rather than a
vacuum, the charge on the plate surface can be increased because of the nature of polarization
of the dielectric as shown in Fig. 1.11. The ratio of e to o is the relative permittivity or the
dielectric coefficient of the dielectric.
In general, permittivity may be expressed in the complex form
= ' -ie" (1.72)
where the imaginary part e" accounts for the electrical power loss or heating in the dielectric
and the real part e' is equivalent to the permittivity. Both ef and e" vary with the frequency,
which is a function of several mechanisms of polarization in a solid dielectric material.
Table 1.9. Dielectric and Ferroelectric Materials
Dielectric coefficient Curie temperature

Material r, at 106 Hz and 25C C
(a) Dielectrics
Polyethylene 2.3
Polyvinyl chloride 2.8
Nylon 3.6
Porcelain 5
Mica 7
Alumina ceramic 9.6
(b) Ferroelectrics
BaTiO3 (ceramic) 1700 120
PZT-5A (ceramic) 1700 365
PZT-5H (ceramic) 3400 193
Region of no
net charge
Area of A
Figure 1.11. Illustration of induced charge by a dielectric medium (from Guy, 1976).
In comparison to ordinary dielectrics, the permittivities of the so-called ferroelectric

materials are about 103 times larger. The ferroelectric material can be transformed into
a new type of material called piezoelectric material by heating the ferroelectric above its
Curie temperature and then cooling it in a powerful electric field. A piezoelectric crystal
changes its polarization once subjected to a mechanical strain. As a result, it can deform
mechanically under an electric field or produce electric impulses as a result of mechanical
impulses. Currently, piezoelectric materials are widely used as force or pressure transducers
with fast response times and very sensitive output. Permittivities of common dielectric and
ferroelectric materials are given in Table 1.9.
1.4.5 Magnetic Properties

A few types of solids can be magnetized and therefore are responsive to the mag-
netic force of a magnetic field. Magnetic solids include ferromagnetic and ferrimagnetic
materials. Most materials, however, are nonmagnetic. These nonmagnetic solids can be
classified as either diamagnetic or paramagnetic.
Magnetic effects in solids originate in the tiny electrical currents associated either with
electrons in atomic orbits or with the spin of electrons. In diamagnetic materials, the effects
of those currents cancel out each other within an atom so that no net magnetic moment
is generated. In paramagnetic materials, although each individual atom has net magnetic
moments, the random orientations for a group of atoms known as the domain still result in
zero net magnetization.
In ferromagnetic materials, the magnetic moments of all types are aligned within a single
domain. In ferrimagnetic materials, the magnetic moments of each type are aligned within
a domain. In the absence of an external magnetic field, the orientation of the net magnetic
moment of a domain may be random so that the solid is unmagnetized. With a strong
external magnetic field, both ferromagnetic and ferrimagnetic solids become magnetized
because the domains in the material are aligned or nearly aligned. Once the external
magnetic field is removed, thermal motion may cause most domains to return to a random
arrangement. However, some domains retain a residual alignment which keeps the material
partially magnetized. Figure 1.12 schematically illustrates the magnetic behavior of a
typical material in each of the classes.
Analogously to the electrostatic force in an electric field, the magnetic force, F m, is
expressed in the form
Fm = mixrBo (1.73)
where BQ is the magnetic flux density in a vacuum; m is the number of north magnetic
poles; and /xr is the relative permeability of the material, defined as
Mr = - (1.74)
Mo
where IXQ is the permeability in a vacuum and [i is the permeability of the material.
For both diamagnetic materials and paramagnetic materials, /xr is about unity. For
ferromagnetic materials, /JLT ranges from 104 to 105. Typical ferromagnetic materials include
iron, nickel, cobalt, and six rare-earth elements, and their alloys.
1.4.6 Material Densities

One of the most important properties of a material is its density, for which there
are several expressions, namely, bulk, particle, and skeletal densities. The bulk density
of solids is the overall density of the material including the interparticle distance of sep-
aration. It is defined as the overall mass of the material per unit volume, which can be
determined by simply pouring a preweighed sample of particles into a graduated cylinder
and measuring the volume occupied. The material can become denser with time and settling,
and its bulk density reaches a certain limiting value, known as the tapped or packed bulk
density.
Particle density is the density of a particle including the pores or voids within the individ-
ual solids. It is defined as the weight of the particle divided by the volume occupied by the
entire particle. Particle density is sometimes referred to as the material's apparent density.
Direct measurement of particle density can be made by immersing a known quantity of the
material in a nonwetting fluid, such as mercury, which does not penetrate into the pores.
The volume of the particle is the volume change of the fluid.
L4 /Material Properties of Solids 39
Zero atomic
moment
ooo6
oooo
oooo
ooo Cu atom
(a)
Domain Domain
1 2
0
Featom
(c) (d)
Figure 1.12. Schematic illustrations of the atomic origins of four types of magnetic
behavior without external magnetic field (from Guy, 1976): (a) Diamagnetism (copper);
(b) Paramagnetism (chromium); (c) Ferrimagnetism (iron); (d) Ferromagnetism
The skeletal density, also called the true density, is defined as the density of a single
particle excluding the pores. That is, it is the density of the skeleton of the particle if the
particle is porous. For nonporous materials, skeletal and particle densities are equivalent.
For porous particles, skeletal densities are higher than the particle density. Measurements
of the skeletal density can be made by liquid or gas pycnometers.
1.4.7 Optical Properties

The most important optical property of a material is the ability to transmit light.
The transmittance of a material is defined as the percentage of the light intensity of an initial
light beam remaining after passing through 10 mm of the material. Thus, the transmittance
of a material also depends on the wavelength of the light. Transmittances of some typical
materials across the ultraviolet, visible, and infrared light spectra are illustrated in Fig. 1.13.
tttravioletl visible Infrared

spectrum
100
Optical glass*NMgF2, polycrystalline
Qrtranl)
Plastic
80 (Potyme&yl-
methacrylate)
60
40
I
3
20
0
0.2 0.5 1 2 5 10
Wavelength of light, urn
Figure 1.13. Light transmittance for three typical optical materials (from Guy, 1976).
Another important optical property of a material is the index of refraction. It is defined

as the ratio of the speed of light in a vacuum, Uw, to the speed of light in the material Um
i = 7 T d-75)
ri[ is about unity in air. For various glass and acrylic (Plexiglas) materials, nx varies from
1.3 to 1.5.
Nomenclature
A Cross-sectional area CD Drag coefficient
A Projected area Cp Specific heat at constant pressure
Ae Integrated cross section for d Particle diameter
extinction do Arithmetic mean diameter in the
AM Normalized constant in the normal distribution
particle mass distribution do\ Parameter in the log-normal
AN Normalized constant in the distribution
particle number distribution d32 Sauter's diameter
B Magnetic flux density J 43 DeBroucker's diameter
Bo Magnetic flux density in a vacuum dA Projected area diameter
b Variable for particle diameter dx Arithmetic mean diameter
C Capacitance dM Martin's diameter
C Factor, defined by Eq. (1.46) ds Surface diameter
Nomenclature 41
dt Dynamic diameter m Mass of the testing material

dy Volume diameter m Number of north magnetic poles
E Young's modulus me Mass of an electron
E Electric field strength N Number of particles
E Total emissive power of radiation No Total number of particles in a
Eb Total emissive power of sample
blackbody radiation N& Numerical aperture
ER Energy flux due to thermal n Number density, or number of
radiation, or emissive power of particles per unit volume
radiation nA Avogadro's number
ERb Energy flux due to thermal nx Relative index of refraction
radiation by a blackbody P Polarization
Ex Monochromatic-emissive power p Partial pressure of the adsorbate
of radiation p* Dimensionless pressure of
E bx Monochromatic-emissive power adsorbate
of blackbody radiation po
Saturated vapor pressure
e Charge carried by an electron Q
Total charge
F Force AQ
Heat absorbed by solids
F Focal length q
Charge carried by a particle
F Cumulative distribution function R
Resistance
Fm Magnetic force R
Rosin-Rammler distribution
/M Mass density function function
/N Number density function Re t Particle Reynolds number at the
g Gravitational acceleration particle terminal velocity
condition
I Intercept
RM Gas constant
Ix Intensity of incident light beam
R^ Universal gas constant
7t Intensity of transmitted beam
r Radial distance
J\ First-order spherical Bessel
S Surface area
function
Sy Specific surface area of particles
Jq Heat flux vector
s Slope
K Thermal conductivity
sm Area per molecule of the absorbed
K& Adsorption rate constant
gas
^d Desorption rate constant
T Absolute temperature
KQ Equilibrium constant
Um Speed of light in a material
L Length
Particle terminal velocity
LQ Initial length /pt
Speed of light in a vacuum
Lo Lorentz constant Uy
Particle volume
I Optical path length V
Volume of gas adsorbed at relative
M Mass of particles Va
pressure p/po of adsorbate
Mo Total mass of particles in a sample
Molar volume of the adsorbed gas
m Mass of a particle ^gm
Monolayer capacity to total number of sites on solid

Velocity of an electron surface adsorption
Maximum velocity of electron K Parameter, defined by Eq. (1.12)
between collisions X Wavelength
Parameter, defined by Eq. (1.14) Wavelength at which the
monochromatic-emissive power
Greek Symbols of a blackbody is maximum
a Parameter, defined by Eqs. (1.33) Mobility of a particle
and (1.41) Permeability
OCR Absorptivity due to thermal Dynamic viscosity of fluid
radiation Mr Relative permeability
Coefficient of thermal expansion V Poisson's ratio
P Parameter, defined by Eqs. (1.33) Average electron velocity
and (1.41) VQ
Electron velocity
AH Heat of adsorption Ver Maximum velocity gained by an
AHa Characteristic heat of adsorption electron just before the next
AHC Characteristic heat of collision
condensation of vapor P Density of fluid
8 Limit of resolution of the optical Density of particles
microscope PR Reflectivity due to thermal
Normal strain radiation
6 Permittivity, or dielectric constant <7
Electrical conductivity
Total emissivity a Normal stress
Permittivity in a vacuum Ob Stefan-Boltzmann's constant
Emissivity due to thermal &d Root of mean square deviation in
radiation the normal distribution
r Dielectric coefficient (Tdl Parameter in the log-normal
Monochromatic emissivity due to distribution
thermal radiation r Average time between collisions
r Ratio of number of occupied sites CO Angular frequency
References
Allen, T. (1990). Particle Size Measurement, 4th ed. New York: Chapman & Hall.
Bachalo, W. D. and Houser, M. J. (1984). Phase/Doppler Spray Analyzer for Simultaneous
Measurements of Drop Size and Velocity Distributions. Optical Engineering, 23, 583.
Bikerman, J. J. (1970). Physical Surfaces. New York: Academic Press.
Brunauer, S., Emmett, P. H. and Teller, E. (1938). Adsorption of Gases in Multimolecular Layers. J.
Am. Chem. Soc, 60, 309.
Cadle, R. D. (1965). Particle Size. New York: Reinhold.
Dallavalle, J. M. (1948). Micrometrics: The Technology of Fine Particles, 2nd ed. New York:
Pitman.
Problems 43
Durst, F. and Zare, M. (1975). Laser Doppler Measurements in Two-Phase Rows. Proceedings of
LDA Symposium, Copenhagen.
Freundlich, H. (1926). Colloid and Capillary Chemistry. Ed. Trans. H. S. Hatfleld. London:
Methuen.
Ghadiri, M., Farhadpour, F. A., Clift, R. and Seville, J. P. K. (1991). Particle Characterization: Size
and Morphology. In Powder Metallurgy: An Overview. Ed. Jenkins and Wood. London:
Institute of Metals.
Gregg, S. J. and Sing, K. S. W. (1982). Adsorption, Surface Area and Porosity, 2nd ed. London:
Academic Press.
Guy, A. G. (1976). Essentials of Materials Science. New York: McGraw-Hill.
Hay, W. and Sandberg, P. (1967). The Scanning Electron Microscope, a Major Breakthrough for
Micropaleontology. Micropaleontology, 13, 407.
Herdan, G. (1960). Small Particle Statistics. London: Butterworths.
Kay, D. H. (1965). Techniques for Electron Microscopy, 2nd ed. Oxford: Blackwell Scientific.
Langmuir, I. (1918). The Adsorption of Gases on Plane Surfaces of Glass, Mica, and Platinum. J.
Am. Chem. Soc, 40, 1361.
May, K. R. (1945). The Cascade Impactor: An Instrument for Sampling Coarse Aerosols. J. Sci.
Instrum., 22, 187.
Parker, E. R. (1967). Material Data Book. New York: McGraw-Hill.
Plantz, P. E. (1984). Particle Size Measurements from 0.1 to 1000 [im, Based on Light Scattering
and Diffraction. In Modern Methods of Particle Size Analysis. Ed. H. G. Barth. New York:
John Wiley & Sons.
Schlichting, H. (1979). Boundary Layer Theory, 7th ed. New York: McGraw-Hill.
Shackelford, J. F. and Alexander, W. (1992). The CRC Materials Science and Engineering
Handbook. Boca Raton, Fla.: CRC Press.
Soo, S. L. (1990). Multiphase Fluid Dynamics. Beijing: Science Press; Brookfield USA: Gower
Technical.
Svarovsky, L. (1990). Characterization of Powders. In Principles of Powder Technology. Ed. M.
Rhodes. New York: John Wiley & Sons.
van de Hulst, H. C. (1981). Light Scattering by Small Particles. New York: Dover.
Weiner, B. B. (1984). Particle and Droplet Sizing Using Fraunhofer Diffraction. In Modem Methods
of Particle Size Analysis. Ed. H. G. Barth. New York: John Wiley & Sons.
Yamate, G. and Stockham, J. D. (1977). Sizing Particles Using the Microscope. In Particle Size
Analysis. Ed. Stockham and Fochtman. Ann Arbor, Mich.: Ann Arbor Science Publishers.
Problems
1.1 Two particles with an identical density of 2,000 kg/m3 undergo free fall in an 80 percent
glycerin/water solution at 20C. Their terminal velocities, /pt, are measured to be 0.007
and 0.114 m/s. What are the equivalent dynamic diameters of these particles? The viscosity
and density of the solution are 62 x 10~3 kg/m s and 1,208.5 kg/m3, respectively.
1.2 A fly ash sample from a fluidized bed coal combustor is analyzed to obtain particle size data.
Table PI .1 shows the distributions of the projected area equivalent diameter of the particle
dA obtained by the image analysis and the volume diameter of the particle dy obtained by
the electrozone technique [Ghadiri etal, 1991].
Specify the type of density functions that can best represent the d& and dy distributions.
How do they differ from the dt distribution?
Table P. 1.1. Data for Cumulative

Distribution of dA and dy for Fly Ash
Volume % dA dy dt (experimental)
undersize /xm /xm /xm
2 0.8 0.50
6 1.2 0.75
10 1.6 0.90
20 2.0 1.05 1.00
30 2.2 1.40
40 2.8 1.80
50 3.0 2.00 2.10
60 3.5 2.20 3.00
70 4.0 2.80
80 4.2 3.20 4.05
90 5.0 4.05 5.20
1.3 The sizes of a powder sample are found to follow a log-normal distribution with <7di = 0.411
and doi = 0.445 mm. Calculate (a) the arithmetic mean diameter, (b) the surface mean di-
ameter, (c) the volume mean diameter, (d) the Sauter mean diameter, and (e) the DeBroucker
mean diameter.
1.4 From the data given in Table PI.2, determine the specific surface area of 11.32 mg of
CaO powder. The data are based on adsorption of nitrogen at 195.8C, with Sy =
4.35V m /m(m 2 /g).
Table P. 1.2. BET Data for
Specific Surface Area
Calculation of CaO
P/Po Va(cm3)
0.1 0.080
0.2 0.086
0.3 0.105
0.35 0.115
Vapors in equilibrium with liquid in fine capillaries or pores will have depressed vapor
pressure as a result of the Kelvin effect. In fact, if the pores are adequately small, the
vapor will condense at pressures far below normal. By measuring the volume of nitrogen
adsorbed at a relative pressure, i.e., p/po, of 0.99 and with prior knowledge of the surface
area, the average pore diameter can be calculated.
In an experiment with the given CaO powder it was found that at a relative pressure
of 0.99, 1.13 cm3 of nitrogen gas was adsorbed. Assuming cylindrical pores for CaO,
determine the average pore diameter of the given powder. To convert volume of gaseous
nitrogen adsorbed to liquid nitrogen, Vnq, held inside the pores, use Eq. (Pl.l),
R
M
where pa is the ambient pressure (1 atm); Va is the volume of gaseous nitrogen adsorbed
per gram of adsorbent; and Vmoi is the molar volume of liquid nitrogen, 34.7 cm3/mole.
Problems 45
1.5 In a simple tensile test, a specimen of an aluminum alloy is elongated elastically by 0.18 mm.
If the original length of the specimen was 25 cm, determine the axial stress in the specimen.
1.6 Prove that the Stefan-Boltzmann law for thermal radiation given in Eq. (1.59) can be derived
by using Planck's formula given in Eq. (1.62). Also show that
(J ==
(k)'c, (PI.2)
Hint
I exp(x) - 1 = (Pl-3)
CHAPTER 2
Collision Mechanics of Solids
2.1 Introduction
In particulate multiphase flows, collision occurs between the particles or between
the particle and the wall. In gas-solid flows, the collision frequency mainly depends on the
particle concentration and particle size, although it is also influenced by the flow pattern
and system configurations. Kinetic energy loss of the particles due to collisions may occur
and can be in the form of frictional heat generation, wall surface erosion, particle breakage
(attrition), particle deformation, particle agglomeration, or solid electrification. The total
momentum of all colliding bodies obeys the law of conservation of momentum, namely,
the total momentum before collision equals the total momentum after collision. However,
as a result of the collision, the momentum of each individual particle may change.
A collision without permanent deformation or heat generation is called an elastic col-
lision. Otherwise, the collision is inelastic and energy loss occurs mainly in the form of
permanent deformation such as particle breakage and frictional heat loss. Generally, a force
is exerted on the particle by its colliding counterpart during the collision. This force is
expressed as an impulse, which is defined as the product of the force and the duration of the
collision. From Newton's law, the impulse is equal to the momentum change of the particle
during the collision. Thus, for the same change in momentum, a shorter duration requires
a larger impact force. It is this impact force that causes the momentum transfer from one
particle to other particles, particle deformation or breakage, and frictional heat loss.
The simplest theory of impact, known as stereomechanics, deals with the impact between
rigid bodies using the impulse-momentum law. This approach yields a quick estimation
of the velocity after collision and the corresponding kinetic energy loss. However, it does
not yield transient stresses, collisional forces, impact duration, or collisional deformation
of the colliding objects. Because of its simplicity, the stereomechanical impact theory
has been extensively used in the treatment of collisional contributions in the particle mo-
mentum equations and in the particle velocity boundary conditions in connection with the
computation of gas-solid flows.
In this chapter, two simple cases of stereomechanical collision of spheres are analyzed.
The fundamentals of contact mechanics of solids are introduced to illustrate the interrela-
tionship between the collisional forces and deformations of solids. Specifically, the general
theories of stresses and strains inside a solid medium under the application of an external
force are described. The intrinsic relations between the contact force and the corresponding
elastic deformations of contacting bodies are discussed. In this connection, it is assumed
that the deformations are processed at an infinitely small impact velocity and for an infinitely
long period of contact. The normal impact of elastic bodies is modeled by the Hertzian
theory [Hertz, 1881], and the oblique impact is delineated by Mindlin's theory [Mindlin,
1949]. In order to link the contact theories to collisional mechanics, it is assumed that the
process of a dynamic impact of two solids can be regarded as quasi-static. This quasi-static
approach is valid when the impact velocity is small compared to the speed of the elastic
46
2.2 I Stereomechanical Impact 41
wave. Finally, collision of inelastic solids is addressed. The concepts, criteria, and mod-
eling of plastic deformation and the coefficient of restitution for an inelastic collision are
introduced and the rebounding velocities of colliding particles are illustrated. It should be
noted that the effects of gas-solid interactions and field forces, such as gravitational and
electrostatic forces, are usually insignificant during a collision process and are excluded in
this chapter.
2.2 Stereomechanical Impact

In this section, collisions among rigid spherical particles are studied. Two simple
cases, collinear and planar collisions, are described. For the general theory of stereo-
mechanical impact of irregular-shaped rigid bodies in arbitrary motion, readers may refer
to Goldsmith (1960).
2.2.1 Collinear Impact of Spheres

Consider a collinear collision of two rigid, frictionless, and nonrotating spheres,
as shown in Fig. 2.1. Neither sphere has a tangential momentum component in this system.
Therefore, conservation of the normal momentum component of the two-ball system yields
m\U\ + m2U2 = mxU[ + m1U'1 = U(mx + m2) (2.1)
where the superscript' represents the state after impact and U is the normal velocity of the
center of mass of the system. The normal impulses / and Jf during the compression and
rebounding processes of the collision, respectively, are given by
J= ml(Ui-U)=m2(U-U2)
J1 = mx(U-U[) =m2{U!l-U)
Now, define the coefficient of restitution, e, as
Substituting Eqs. (2.1) and (2.2) into Eq. (2.3) shows
J
e= - (2.4)
J
Figure 2.1. Normal collision of rigid and smooth spheres.

48 2 / Collision Mechanics of Solids
The rebounding velocities of the colliding spheres can be expressed in terms of the coefficient
of restitution as
Ul= U\ {U\ -
+ m2 (2.5)
x
U2 = U2-\ ;
m IT
(U\ - U2)
+m2 energy of the system is given by
Therefore, the loss of the kinetic
~ . v- ~ /v-i - U2)2 (2.6)

2/i + m 2
It is clear that e = 1 for the normal impact of perfect elastic spheres, while e = 0 for the
normal impact of perfect plastic spheres.
Equation (2.3) also provides a basis for the experimental determination of the coefficient
of restitution. Consider the case where a ball at rest is dropped from a height h to a horizontal
stationary massive rigid surface, rebounding back to a height of h''. If we label the ball with
the subscript 1 and the massive plane with 2, Eq. (2.3) can be rearranged to
(2.7)
The coefficient of restitution in an impact depends not only on the material properties of
the colliding objects but also on their relative impact velocity. More discussion of the
coefficient of restitution is given in 2.5.2.
2.2.2 Planar Impact of Spheres

Consider a planar collision of a sphere of mass m i and radius a\ against an initially
stationary sphere of mass m2 and radius a2. Select the Cartesian coordinates so that the
x- and v-axes are normal and tangential to the contact surfaces, respectively, as shown in
Fig. 2.2. The conservation of linear momentum yields
mi (JV TlA -A- m^JV 0
(2.8)
Figure 2.2. Planar collision of rigid spheres.

2 3 I Theory of Elastic Contact of Solids 49
and the conservation of angular momentum with respect to the contact point gives
k2M-co1)-al(v;-vl)=o
0
where k is the radius of gyration. For a sphere, k is evaluated by
k1=I- = la2 (2.10)

m 5
where / is the moment of inertia.
In order to determine the unknown velocities after collision uniquely, two more rela-
tionships for impact velocities need to be specified. For certain values of the coefficient
of friction and the coefficient of restitution, simple expressions of impact velocities can be
obtained. In a collision of two completely rough and inelastic spheres, where the coefficient
of friction / reaches infinity and the coefficient of restitution e is equal to zero, the relative
velocities must vanish. Therefore, we have
U[ = U^
(2.11)
V[ + cJxax = V2' + co'2a2
When two perfectly smooth and yet incompletely inelastic spheres collide, i.e., f = 0 and
0 < e < 1, no change in the tangential velocity of each colliding sphere occurs. Thus, it
gives
V[ = V\
_ t// _ y/ (2.12)
The condition V2 = V2 = 0 is automatically satisfied from Eq. (2.12) and Eq. (2.8). For an
impact of two perfectly smooth and elastic spheres, we have
V{ = Vi
It can be easily shown that Eq. (2.12) and Eq. (2.13) are equivalent when e = 1.
The collisional impulse can be obtained by using the impulse-momentum equations
[Goldsmith, I960]. However, this stereomechanical approach does not yield the transient
stresses, collisional forces, impact duration, or collisional deformation of the colliding
objects. These quantities have to be obtained from the analysis of stresses and strains of
the solids due to the impact, which is introduced later.
2.3 Theory of Elastic Contact of Solids

During a collision, the colliding solids undergo both elastic and inelastic (or plastic)
deformations. These deformations are caused by the changes of stresses and strains, which
depend on the material properties of the solids and the applied external forces. Theories on
the elastic deformations of two elastic bodies in contact are introduced in the literature uti-
lizing Hertzian theory for frictionless contact and Mindlin's approach for frictional contact.
As for inelastic deformations, few theories have been developed and the available ones are
usually based on elastic contact theories. Hence, an introduction to the theories on elastic
contact of solids is essential.
In this section, an introduction of the general relations of stresses in equilibrium in

an infinitely large solid medium is presented, followed by a special application where a
concentrated force is acting on a point inside the solid. Also presented is the case of forces
on the boundary of a semiinfinite solid medium, which is of importance to the contact of
two solid objects. As consequences of the boundary compression, displacements due to the
changes of stresses and strains in the region of contact can be linked to the contact force by
the Hertzian theory for frictionless contacts and by Mindlin's theory for frictional contacts.
For more details on the Hertzian theory for contact, interested readers may refer to books on
elasticity [Goldsmith, 1960; Timoshenko and Goodier, 1970; Landau and Lifshitz, 1970].
2.3.1 General Relations of Stresses in a Solid Medium in Equilibrium

For an element in equilibrium with no body forces, the equations of equilibrium
were obtained by Lame and Clapeyron (1831). Consider the stresses in a cubic element in
equilibrium as shown in Fig. 2.3. Denote T^ as a component of the stress tensor T acting
on a plane whose normal is in the direction of e\ and the resulting force is in the direction
of ey In the Cartesian coordinates in Fig. 2.3, the total force on the pair of element surfaces
whose normal vectors are in the direction of ex can be given by
aixz
Fx = ex + dx (2.14)
A A A
dx + dx Ax Ay Az
Similarly, the total forces on the other two pairs of element surfaces are given by
I VX O -*VV ^ -*VZ
(2.15)
dy dy
Figure 2.3. Stress relationship in a differential cube.

23 I Theory of Elastic Contact of Solids 51
and
3TZ7
F7 = Ax Ay Az (2.16)
dz -e-x +' dz *y +' dz
From the condition of equilibrium, EFi = 0, the equation of equilibrium is given as
V.J = 0 (2.17)
For convenience in the following discussion, Eq. (2.17) is expressed in cylindrical coordi-
nates. The symmetric stress tensor T is given as
T = (2.18)
where a's are the normal stresses and r's are the shear stresses. Thus, the equation of
equilibrium takes the form
1 %TXQ drTZ
_
3r r 30 3z
daz rrz _ (2.19)
dr r 30 3z r
\<ê_ <êz _
dr r 30 dz r
The general relations between strains and stresses are represented by Hooke's law as
_ 1 _ 1
r E r 0 z , 6 g 0 r z
1 2(1 + v)
Yrz = *> (2.20)
2(1 + v) 2(1+ v)
/ 11/ j-, w
ll/ 7 YOz
/UL. = j-, ~VL
where 6 's are the normal strains; y 's are the shear strains; E is Young's modulus of elasticity
in tension and compression; and v is Poisson's ratio. For most homogeneous solids, both
E and v are material properties whose values can be obtained from materials handbooks.
Moreover, the displacements are related to the strains by
<9/r
e =
lT 1 dk
1 *z = a/
z
dr' r r do
(2.21)
Jz~
Yrz =
a/r
dlz ia/ r die k
YrS == r"dO~K a 7 ' r '
die 1 dlz
Yez = Jz~+ r do
dz dr'
where /'s are the displacements in the direction of the corresponding coordinates.
Most collisions with spherical particles can be regarded as axisymmetric, torsionless,
and quasi-static. Therefore, Eq. (2.19) is simplified to
a<Tr arrz
3r 3z
(2.22)
3xrz 3o^ rrz _
3r 3z r
52 21 Collision Mechanics of Solids
and the displacement-strain relationships become

9/r lT dlz 9/r 9/z
In order to solve Eq. (2.22), we introduce Love's stress function, yjr, which is defined
[Love, 1944] by
(2.24)
9 / , 32i/r \ 9 / d2i\
It can be shown that, with the substitution of Eq. (2.24), Eq. (2.22) becomes
VV = 0 (2.25)
which is the equation of equilibrium of solids in terms of Love's stress function \/r in an
axisymmetric and torsionless contact.
Hence, in principle, once V" is found, all the stresses can be directly obtained from
Eq. (2.24). Consequently, the corresponding strains and displacements can be evaluated
from Eq. (2.20) and Eq. (2.23), respectively.
2.3.2 Concentrated Force at a Point in an Infinite Solid Medium

Consider the case of a concentrated force applied to a point in an infinite solid
medium. Tofindthe relationships between the point force and the resulting stresses, Love's
stress function may be selected, from sets of solutions of Eq. (2.25), as [Timoshenko and
Goodier, 1970]
V^ = A(r 2 + z2)5 (2.26)
where A is a constant to be determined. By substituting this stress function into Eq. (2.24),
the corresponding stresses are obtained as
or = A [(1 - 2v)z(r 2 + z 2 )"i - 3r 2 z(r 2 + z 2 )"]
<7, = A ( l - 2 v ) z ( r 2 + z 2 )-
az = -A [(1 - 2v)z(r 2 + z 2 )" + 3z 3 (r 2 + z 2 )"]
rrz = - A [(1 - 2v)r(r2 + z 2 )~i + 3rz 2 (r 2 + z 2 )"i]
Note that all the stresses in Eq. (2.27) become singular at the origin where the point
force is applied. To avoid this singularity, consider a small spherical cavity whose center is
located at the origin of the coordinates as shown in Fig. 2.4. The coordinates are arranged
such that the force is in the z-direction and is applied at the origin of the coordinates. Thus,
the summation of the surface forces from the stresses in the direction of the z-axis balances
the point force inside the solid medium.
Consider the ring-shaped element shown in Fig. 2.4. The z-component of the summation
of surface stresses on the element, TZi is given by
Z
Tz = -r r z sin0 - <rzcos0 = A[-f--^ + ) (2.28)
2 2
\r + z (r + z2)2J
2
Infinite solid
medium
Spherical
cavity
Ring-shaped
element
Figure 2.4. Surface stresses on a spherical cavity in an infinite solid medium.
Thus, the total surface force Fz is found by integrating the surface stresses over the entire
surface as
- v) (2.29)
Jo
Since Fz is equal to the applied point force, the constant A is obtained from the preceding
equation as
Fz
A= (2.30)
- v)
8TT(1
The shear stress rrz on the plane z = 0 is then related to the point force Fz by Eq. (2.27)
and Eq. (2.30) as
- 2v)
Trz = - (2.31)
8JT(1 - v)r2
2.3.3 Force on the Boundary of a Semiinfinite Solid Medium

In this section, the case of a semiinfinite solid with a concentrated force acting on
the boundary is introduced. This case was originally solved by Boussinesq (1885). It should
be noted that the only difference between this case and the case of a point force in an infinite
solid medium is the boundary conditions. Shear stresses vanish on the boundary of the
semiinfinite solid. In the following, the concept of a center of compression is introduced. The
stress field in a semiinfinite solid with a boundary force can be obtained by superimposing
the stress fields from a point force and a series of centers of compression. A center of
compression is defined as the combination of three perpendicular pair forces.
2.3.3.1 A Center of Compression

Consider a case of two equal but opposite forces (in the z-direction) positioned a small
distance 8 apart in an infinite solid body, as shown in Fig. 2.5. At any point M, the stresses
54 2 I Collision Mechanics of Solids
Figure 2.5. Stresses produced by the pair forces.
produced by force F at the origin may be given by Eq. (2.27). To arrive at an expression for
the stresses produced by force F\ (F\ = F) applied at Oi in Fig. 2.5, noting that it is acting
in the opposite direction and considering that <5 is infinitely small, any term in Eq. (2.27)
expressed in a general function 4>(r, z) can be replaced by [O The stresses
produced by the pair forces can thus be given as
[(1 - 2v)z(r 2 - 3r 2 z(r 2
'8z (2.32)
az = B^- [(1 - 2v)z(r 2 + z2) 3zV
rrz = B [(1 - 2v)r(r2 + z2)^ + 3rz2(r2 + z 2 )"i]
where B = AS. Once in equilibrium, the relations of stresses on a triangular element at

point M (shown in Fig. 2.5) can be expressed as
= <rrAzsin0 + (Tz + r r z Azcos0 -f r rz Ar si
(2.33)
A/ = crrAz cos(j) G r rz Az sin</> + r rz Ar cos
or
CTR = ox sin2 0 + crz cos2 0 4- 2rrz cos 0 sin 0
(2.34)
rR0 = (<jr a z ) sin 0 cos 0 rrz(sin2 0 cos2 0)
Substituting Eq. (2.32) into Eq. (2.34) yields
<jR
2(1 4- v),

=
(2.35)
sin 0 cos 0
2.3 I Theory of Elastic Contact of Solids 55
Thus, the stresses at any point due to the pair forces along the z-direction are given by
Eq. (2.35). If the pair forces are relocated along the r-direction but still applied near the
origin, the resulting stresses at any point M can be obtained by replacing 0 by (0 n/2)
in Eq. (2.35) as
2(1 + v)B ( 2 2 ( 2 - v ) . z-
<x R = ( -cos 0 + i i sm 0
(2.36)
sin 0 cos 0
Similarly, if the pair forces are relocated along the #-direction and still applied near the
origin, the stresses at M are obtained by setting 0 = it/2 in Eq. (2.35) as
(7R = =0 (2.37)
Hence, by combining Eqs. (2.35), (2.36), and (2.37), the total stresses at M due to the three
perpendicular pair forces are
4(2v - \)B
; (2.38)
which indicates that the compression stress in the radial direction depends only on the
distance from the center of compression and is inversely proportional to the cube of that
distance.
To obtain an expression for the normal stress 0$ in the tangential direction, consider the
equilibrium of an element cut from a sphere by two concentric spherical surfaces of radii R
and R + &R and by a circular cone with a small angle d0, as shown in Fig. 2.6. The force
balance along the radial direction gives
~(d0) 2 /? &R = ( crR + ^ d J ? J(R (2.39)

4 \ d/c /
Figure 2.6. Equilibrium of an element in spherical coordinates.

zy
(a) (b)
Figure 2.7. A point force on the boundary of a semiinfinite medium: (a) r z coordinates;
(b) R (j> coordinates.
which yields
RdcrR
(2.40)
2.3.3.2 Stresses Induced by a Boundary Point Force

Consider the case where a point force is acting on the boundary of a semiinfinite body, as
shown in Fig. 2.7. Let the boundary be in the plane z = 0. Then, the shear stress on the
boundary would be given by Eq. (2.27) as
rrz = -A(1~2v) (2.41)

r2
where A is a constant to be determined. Since shear stresses on the boundary cannot be
produced by the concentrated force acting on the boundary, the shearing force in Eq. (2.41)
has to be balanced by the stress resulting from the compression in the upper semiinfinite
body.
The stress distribution corresponding to the center of compression is given by
(2.42)
where C is a constant to be determined. From Eq. (2.40), we have

daRR \C
(2.43)
From the equilibrium of the element in Fig. 2.7, using Eqs. (2.42) and (2.43), it can be
obtained that
(2.44)
2(
Assume that the centers of pressure are uniformly distributed along the z-axis from z = 0
to z = oo. The stresses in an infinite solid medium can be obtained by integration from
z to oo as
<yT = o [l ~~ z^"2 ~^~ z 2 ) ~ 5 ~~ zr2(r2 + z 2 ) " ] > G

z = Z(r 2 + z 2 )~ 5
2 (2 45)
r c c -
__ V(Y I <7 ^ 2 /Tî 1 7 (V I "7 ^ 2
Hence, on the plane z = 0, stresses are given by
^z = 0, rra - ^ (2.46)
To ensure that the boundary surface is free from stresses, the constants A and C in
Eq. (2.41) and Eq. (2.46) can be selected such that the shearing forces on the boundary are
eliminated. Thus, we have
C = 2A(1 - 2v) (2.47)
The stresses in a semiinfinite body can be obtained by combining Eq. (2.27) with Eq. (2.45)
as
oe = A]^ [z(r2 + z 2 )"^ - 1 + zr\r2 + z 2 )"i] (2.48)
az = - 3 V ) , rz V )
To determine A, consider the equilibrium of the ring-shaped element shown in Fig. 2.4.
The force on the element in the z-direction is
Tz = - ( r r z s i n 0 + crz cos 0) = 3Az 2 (r 2 + z 2 )" 2 (2.49)
The balance between the point force and the integration of Tz over the hemispherical surface
yields
A=^ (2.50)
In
Therefore, with the substitution of Eq. (2.50) into Eq. (2.48), the stresses due to a normal
force on the boundary of a semiinfinite solid are expressed in cylindrical coordinates by
a, = ^l-^[2(r2 + z2y\ - 1 + zr\r2 + z2)^} (2.51)

L J
2T r
z V +z)2, Va =
ZTT In
2.3.3.3 Displacements Induced by a Boundary Point Force

As a result of applying a point force on the boundary, displacements in the semiinfi-
nite solid are produced. These displacements can be calculated from Hooke's law and
the displacement-strain relationships. The displacement in the r-direction, / r, is given by
Eq. (2.20) and Eq. (2.23) as
lT = eer = [cr e- v(aT + a z )]
(2.52)
2nEr
Similarly, the displacement in the z-direction, lz, is given by
- T - = <?z = [<7z - V(ff r + Og)]

9z
a 53)
( J
dh_ dU _ 2(1 + v) _<M,
which yields
[(1 + v)z\r 2 + z2)~i + 2(1 - v 2 )(r 2 + z 2 )"^] (2.54)
where the integration constant is related to an axial rigid-body translation and is set to zero
here for simplicity. Thus, the displacements on the boundary z = 0 are expressed as
( l - 2 v ) ( l + v)F z (1 - v2)Fz
/JO = , / z0 = (2.55)
2nEr nE
The displacements given in Eq. (2.55) for a point force on a semiinfinite solid indicate that
the displacements would be infinite when r approaches zero at the origin where the point
force is applied. This singularity may be avoided by using a load distribution over a finite
area to replace the point force.
Consider a pressure distribution p on a circular area A of radius rc such that
pdA = Fz (2.56)
A
The influence of this pressure distribution on the displacements in a neighboring area on

the same boundary surface can be analyzed as shown in Fig. 2.8. Let M be an arbitrary
point outside the circular area (Fig. 2.8(a)) or inside the circular area (Fig. 2.8(b)) with a
distance r from the center of the circular area. A small element within the loaded zone is
chosen as sdfids, where s is the distance between this element and M. From the solution
of lz for a point force on a semiinfinite solid, the increment of the vertical displacement at
location M under the influence of this element force would be
(1 - v2)
d/ z 0 = _ psdfids (2.57)
TTES
Thus, the total vertical displacement on the boundary surface can be obtained by integrating
over the whole circular area, which gives
V}
JJ
2.31 Theory of Elastic Contact of Solids 59
(a) (b)
Figure 2.8. Displacement under compression: (a) M is outside the compressional circle;
(b) M is inside the compressional circle.
Similarly, the total radial displacement on the boundary surface is given by
(2.59)
Assuming uniform distribution of the pressure p, it can be shown that the maximum
displacement at the center of the circle is TC/2 times the displacement at the edge of the
circle.
2.3.4 Hertzian Theory for Frictionless Spheres in Contact

When two elastic and frictionless spheres are brought into contact under compres-
sional forces or pressures, deformation occurs. The maximum displacement and contact
area depend not only on the compressional force but also on the elastic material properties
and radii of the spheres. The contact between two elastic and frictionless spherical bodies
under compression was first investigated by Hertz (1881) and is known as the Hertzian
contact.
Consider two frictionless spheres in contact as shown in Fig. 2.9. M is a point on the
surface of sphere 1 with a distance r from the Zi-axis. N is the opposite point on the surface
of sphere 2 with the same distance r from the Z2-axis. O is the contact point and also serves
as the origin of both Zi-axis and Z2-axis. N, M, and O are in the same plane. The distances
from M or N to the tangential plane which is normal to the Zi-axis and Z 2-axis are denoted
by z\ and Z2, respectively. In Fig. 2.9, the triangle O2AO is similar to the triangle NBO.
Hence, the theorem of similarity of triangles gives
2z2
(2.60)
a2
Figure 2.9. Hertzian contact.
which yields
(2.61)
When r/a\ and r/a2 are small enough, it can be obtained that
a\ + a2 2
(2.62)
During contact, location M tends to be moved in the direction of Zi by a vertical

displacement /zOi. Similarly, N is going to be moved by a distance /z02 in the direction of
Z 2 . When the center of sphere 1 approaches the center of sphere 2 by a distance a (note
that a here is an approaching distance rather than the distance between the two centers of
the spheres), the change of distance between M and N is then given by a (/zoi + lzo2)-
Consider the case of deformation in which M and N are on the edge of the surface of contact
as a result of local compression. We have
(2.63)
where the vertical displacements /zOi and /zo2 are given by Eq. (2.58) as
(1 - v\
lzO2 = pdsdfi (2.64)
7tE2
Hence, it gives
p ds dfi = a - - (2.65)
2a*
p(f>
Figure 2.10. Hertz pressure distribution for frictionless contact.
where E* is the contact modulus and a* is the relative radius defined as

1 1 - v? 1 _ 1 1
(2.66)
E* E\ E2 ' a** a\ _
It is necessary tofindan appropriate form for the pressure distribution p so that Eq. (2.65)
can be satisfied for any r provided that r is small compared to the radii of spheres. One
such pressure distribution is the Hertz pressure distribution in which the pressure at any
location within the contact area is represented, as illustrated in Fig. 2.10, by
p(r)=C(r2c-r2y (2.67)
It is noted that the total loading force Fz is equal to the pressure integrated over the contact
area so that the constant C can be expressed by
(2.68)
Hence, the pressure distribution becomes
(2.69)
It is shown in Eq. (2.69) that the maximum pressure p0 occurs at the center of the contact
area and is given by
3 Fz
(2.70)
which is about 1.5 times the averaged pressure on the contact surface.
Using the Hertz pressure distribution, an integration along the cord mn in Fig. 2.8(b)
gives
2-r2sin2) (2.71)
(2.72)
Thus, we have
3F, 2 2 r2
^3(2r c -r )=a-^- (2.73)
Since the preceding equation is valid for any small r, the corresponding terms on both
sides of the equation should be equal. This argument suggests
3F Z 3FZ = 1
By rearranging the preceding two relations, the radius of the surface of contact rc and the
centers approaching distance a can be related to the contact force F z , the radius, and the
material properties of the spheres as
3F
it%
' - (2.75)
The corresponding maximum pressure p0 at the center of the contact is thus obtained by
substituting Eq. (2.75) into Eq. (2.70) as
Therefore, for Hertzian contact, using Eq. (2.75) and Eq. (2.76), the maximum radius
of contact r c , the maximum approaching distance a, and the corresponding maximum
pressure can be calculated on the basis of the contact force, the elastic material properties
of the spheres, and the radii of the spheres.
Example 2.1 Figure E2.1 (a) shows a packed bed of height 1 m with monodispersed glass
beads (silica glass) in a large container. Estimate the elastic displacement and contact area
of a particle at the bottom of the bed. The density of the glass beads is 2,500 kg/m 3. The
particle diameter is 1 cm.
Solution For a packed bed with monodispersed spheres in a large container, the
volume fraction for solids is about 0.45. Thus, for particles under the bed height of 1 m,
the average pressure caused by gravity is
p = avpvgH = 0.45 x 2,500 x 9.8 x 1 = 1.1 x 104N/m2 (E2.1)
As shown in Fig. E2.1(b), the effective area of a single particle in a horizontal layer is
equivalent to that of an equal-sized triangle with side length of one particle diameter d v.
Thus, the vertical loading on a single sphere under the bed height of 1 m is given by
4
F = d*p = x 10" x 1.1 x 104 = 0.48 N (E2.2)
Assume that the particles in consideration are not located in the bottom layer so that the
contact arrangement among spheres may be represented as in Fig. E2.1(c). Hence, the
compressional force between two neighboring spheres is given by
F 0.48
Fz = = = 0 . 2 0 N (E2.3)
V6 V6
(a)
(b) (c)
Figure E2.1. Particle contact arrangement in a packed bed: (a) A packed bed with
monodispersed spheres; (b) Top view of a horizontal layer of monodispersed spheres; (c) Local
contact mode of spheres.
The values for Young's modulus and Poisson's ratio for silica glass are found from
Table 1.6 as
E = 8 x 1010 N/m2; v = 0.24 (E2.4)
Hence, from Eq. (2.75), the radius of the contact circle of two contacting glass beads is
3 F7dL 3 0.2 x 0.01 2

1 0 -(1-0.24
rc = -
KS 8 x l O
VA
*""' )" =
" 2x 'm(E2.5)
Note that the elastic displacement of a single sphere equals half of the total displacement
of two contacting spheres. Therefore, from Eq. (2.74), we have
a
a 3 Fz(l - v2) 3 0.2x(l-0.24 2 )
= 0.0 x 10 m (E2.6)
h = = Erc 4 8 x 1010 x 2 x 10"5
2.3.5 Theories for Frictional Spheres in Contact

In the previous sections, only the normal contact of two elastic spheres with per-
fect smooth surfaces (i.e., no tangential force) is considered. However, for oblique con-
tact between two frictional spheres, tangential forces are encountered, and, consequently,
tangential displacements have to be considered. The relations among tangential traction,

normal traction or pressure, and tangential displacements for the contact of two frictional
elastic bodies were first developed by Mindlin (1949) and have been widely used ever since.
Therefore, in this section, oblique contact based on Mindlin's approach is introduced.
In an oblique contact, tangential forces are caused by both the normal compression
and either the tangential sliding or rotation on the frictional contact interface. Thus, these
forces and the resulting displacements should be related to the nature of friction on the
contact surface. Once the two contact objects are in a tangentially relative motion, the
frictional force on the contact interface obeys Amontons's law of sliding friction. That says
the frictional force is proportional to the loading force normal to the contact area when a
steady sliding motion between the two bodies occurs. The proportional constant is called
the coefficient of kinetic friction; its value is determined by the materials and the physical
conditions of the contact surface {e.g., surface roughness). If there is no tangentially relative
motion between the two contact objects, the tangential force is then called the static frictional
force; its magnitude is less than the force of limiting friction. The limiting tangential force
of static friction on the contact surface is equal to the normal loading multiplied by the
coefficient of static friction when sliding motion between the two bodies is just about to
occur. The kinetic friction coefficient is usually smaller than the static friction coefficient.
It should be pointed out that, although the static frictional force does not give rise to a
relative sliding motion, it does lead to frictional traction or microsliding on the contact area
[Johnson, 1985].
2.3.5.1 Sliding Contact of Spheres

Assume that Amontons's law of sliding friction can be applied at each elementary area of
the interface, such that
^ 4 =A (2.77)
P(x, y)
where q is the tangential traction; p is the normal pressure; and /k is the coefficient
of kinetic friction, whose value depends on the materials and physical conditions of the
contact interface only. Thus, the total tangential force F t becomes
^t = AFZ (2.78)
Assume that the normal pressure and displacements are not affected by the existence
of the tangential traction and resulting displacements. Hence, the normal pressure and
contact area can be determined by the Hertzian theory. For the sliding contact of spheres,
substituting Eq. (2.69) into Eq. (2.77) gives rise to the tangential traction as
<2.79)
With the tangential traction parallel to the x-axis, the general displacements on the contact
surface due to the tangential traction can be expressed by
h
= ^ T ^ / / liH, n) 3 d di;
2 n G j h S
(2.80)
-y)
Figure 2.11. Coordinates conversion between (, rj) and (s, p).
where s2 = ( x)2 + (rj v) 2 , and G is the shear modulus, defined by G = E/2(l + v).
The corresponding surface shear stress rxy is
- y)
^xv = i (2.81)
The detailed derivation of Eq. (2.80) and Eq. (2.81) is omitted here. Interested readers may
refer to the work of Johnson (1985).
In order to perform the surface integration of Eq. (2.80) and Eq. (2.81), it is convenient
to change the coordinates from (, rf) to (s, ft), as shown in Fig. 2.11. Consider the dis-
placements and stresses at a point M(JC, y) when an element of tangential traction is applied
at C(, rf). From the conversion of the coordinates, we have
= x + s cos j3
(2.82)
rj = y + s sin &
By considering only the surface displacement within the loading circle of radius r c , the limit
of s should be on the edge of the circle. For a fixed /?, this limiting point is represented by
N in Fig. 2.11. Denote s\ as the limiting length of s from M to N and g as the distance
between P and M. Thus, we have
g = x cos P + y sin /3 (2.83)
and
?)2=. (2.84)
which yields
(2.85)
66 2/ Collision Mechanics of Solids
where
a = ^Jrl - (JC2 + y2) (2.86)
Thus, Eq. (2.80) and Eq. (2.81) become
j f>2n rs\
k = / / q(s, j8)(l - v sin2 P) dp ds
2nG JO JO
(2.87)
/ v /f
Z7T /î
V ~~ I
2nG Jo
and i pin ps
z
= 1 f2n /f" q(s, (2.88)
2TT JO JO
For sliding contact of spheres, the displacement in the x -direction is obtained by substi-
tuting Eq. (2.79) into Eq. (2.87). First, we have
r*\ 3 f p rsi
/ q(s,f$)ds = - ^ J (a2- 2gs-s2)>ds
Jo 2 jrrc Jo
Note that, from Eq. (2.83), g(P) = -g(P + n). Substituting Eq. (2.89) into Eq. (2.87)
gives rise to
( 4
" "v)y2>) (2 90)
'
Similarly, the displacement in the y-direction is

3 fvF7
y
~ 32"
It should be noted that the displacement in the z-direction due to tangential traction
is usually not zero at the interface. Therefore, the actual interface with tangential force
mismatches the flat surface caused by the normal loading [Johnson, 1985]. In this text, it
is assumed that this mismatch of normal displacements can be neglected; this turns out to
be an acceptable approximation for most cases.
2.3.5.2 Nonsliding Contact of Spheres

When a tangential force is less than the limiting tangential force of "static friction," no slid-
ing motion is produced. However, tangential traction will still occur in this case. Consider
the general case illustrated in Fig. 2.12. Ai and A2 represent two points on the contact
surface which were coincident before a tangential force Ft is applied. With the action of Fu
A\ and A2 are shifted by tangential displacements lx\ and /X2 relative to the distant points
B\ and B2 in the undeformed region of each sphere, and B\ and Bi are moved by rigid
tangential displacements Sx\ and <5X2, respectively. Denoting sx\ and sX2 as the absolute
2.3 / Theory of Elastic Contact of Solids 67
Figure 2.12. Tangential tractions in nonsliding contact.
tangential displacements of A\ and A2, respectively, the relative tangential slip sx between
A\ and A 2 can be expressed by
= Sj\ - 5 x2 = (/Xl - kl) - (2.92)
A similar relation governs the tangential displacements in the y-direction. For the no-slip
region, sx becomes zero so that
1x2 = 5 x i 5X2 = 5 X
(2.93)
^y2 = 5 y i 5 y 2 = 5 y
where 8X and 8y denote the relative but rigid tangential displacements between the two
spheres so that both 8X and 8y are independent of the position of A\ and A2. Therefore,
Eq. (2.93) leads to a conclusion that all surface points within the no-slip region undergo the
same tangential displacement.
Consider the distribution of traction to be in the form of
_i
- r2) (2.94)
This distribution leads to a uniform tangential displacement of the contact surface. qo is the
maximum tangential traction at the center of the contact surface. By integrating Eq. (2.94)
over the contact surface, q0 is found to be
(2.95)
Substituting the distribution of traction in Eq. (2.94) into Eq. (2.87) gives
2 1
/ q(s,p)ds =qorc / (a2 - 2gs - s )-i ds = qorc | 'j - tan" U)\ (2.96)
Jo Jo
Using the symmetric condition g(fi) = g(P + 7r) yields
'* = I F A 1 ~v s i n 2 ^ = ^ ^ (2 97)
-
Similarly, we have
ly = 0 (2.98)
Therefore, a constant tangential displacement induced by the no-slip contact between two
spheres is given by
which indicates that Eq. (2.94) is the distribution of tangential traction of no-slip contact
between two spheres. However, according to Eq. (2.94), the tangential traction at the edge
of the contact circle would be infinitely large. Hence, near the edge of contact, a small zone
inevitably exists wherein slip occurs. This slip is relatively small compared to normal slip,
and thus can be termed microslip. This argument also suggests that the contact area consists
of a core region (r < c) for no-slip motion and an annulus region (c < r < rc) for microslip.
The distribution of traction for the microslip is similar to that of sliding contact. Thus,
q'(r) = i*Jrl-r*
r V
(2.100)
c
where / s is the coefficient of static friction. To match the no-slip condition in the core
region, another distribution of traction should be obtained so that the superposition of this
distribution with Eq. (2.100) can yield constant displacements in the nonslip region. To do
so, consider a distribution of traction
(2.101)
Similarly to Eq. (2.90) and Eq. (2.91), the tangential displacements due to the distributions
of tangential traction given by Eq. (2.100) and Eq. (2.101) are expressed, respectively, as
( 4 3 v ) x 2 ( 4 v ) y 2 (2 102)
" " " " " ^ '
(2.103)
and
K = - J r ^ [ 4 ( 2 - v)c2 - (4 - 3v)x2 - (4 - v)y2] (2.104)
Therefore, the tangential displacements in the core region under joint actions portrayed by
Eq. (2.100) and Eq. (2.101) can be obtained as
2-v)(rc2-c2) (2.106)
(2.107)
2.3/ Theory of Elastic Contact of Solids 69
The displacements satisfy the condition for no-slip within the core region with the result
that
3fsFz(2-vl t 2-v2 - c2
1, ~ ~r ~~z (2.108)
16
The radius of the core region c can be determined from the given tangential force F t by
Ft = P qr2nr dr + f\q' + q")2nr dr = fsFz (l - C

-) (2.109)
or
(2.110)
Thus, the area of the core region decreases with increasing tangential force while the
normal force is maintained constant. Once the tangential force reaches the limiting static
friction force (F t = / S F Z ), the core region has shrunk to a single point so that the sliding
motion is about to begin.
2.3.5.3 Torsion of Elastic Spheres in Contact

Besides the oblique contact, tangential displacements may also be produced in the contact
of two elastic spheres under the actions of a compressional twist, as shown in Fig. 2.13.
Since the torsional couple does not give rise to a displacement in the z-direction, the pressure
distribution is not influenced by the twist and is thus given by the Hertzian contact theory.
If a torsional couple or twist moment Mz were applied to the two spheres in contact,
the contact surface would undergo a rigid rotation relative to distant points in each body if
there were no-slip motion at the interface. In this case, the tangential displacements could
be expressed by
hi =
hi = (2.111)
where ft is the angle of twist in each body. To yield a uniform angle of twist throughout the
whole contact area, the distribution of tangential traction may be in the form of [Johnson,
Figure 2.13. Contact by normal compression and twist.

1985]
W
~ r 'r*-r2y* (2.112)
which leads to an infinite value of tangential traction at the edge of the contact circle. Hence,
similarly to the argument for no-slip oblique contact, microslip has to occur in an annular
region (c < r < rc) while no-slip remains in the core region (r < c).
The microslip motion is in the circumferential direction and the surface distribution of
traction in this annular region takes the limiting value from Amontons's law as
(2.113)
Several methods were presented by Lubkin (1951) for the determination of the radius
of the no-slip core and the distribution of tangential traction in this region. Only the final
results of Lubkin's work are introduced here.
The relation between the angle of twist and the penetration of microslip in sphere 1 is
given by
<hL
JF
L
An
(2.114)
S Z
where x *s defined by
2
x = l-(c/rc)2 (2.115)
and K(x) and E(x) are the complete elliptic integrals of the first and second kinds of
modulus of x, respectively, defined by
I ' ^ l2-sin20nd0
K(x)== I'^l-x (2.116)
JJo
E(X)= / (l-x 2 sin 2 ^)2d^ (2.117)
Jo
Thus, the total angle of twist is related to c by
- J - V ( x ) - E(x)] (2.118)
G2J
The distribution of tangential traction is given by
9 = -ZTT\l- ~ U + [*(*) - E(x)]K(k) - K(X)E(k) ) (2.119)
where A is defined by
(2.120)
It can be seen that the boundary value of Eq. (2.120) matches that given by Eq. (2.112).
The twisting moment is given by
M Z = 2TT [C qr2dr (2.121)

Jo
2.3 / Theory of Elastic Contact of Solids 71
where q(r) is obtained by Eq. (2.112) for c < r < rc and Eq. (2.119) for r < c. Thus, the
relation between Mz and c is given by
l
Jh_ = 3TT 3X2K(X) I _ /*2 sin (X sin a) da
fsFzrc 16 4TT I Jo ( 1 - A 2 sin 2 a ) i
r -3,-3/ / 2 2
(2.122)
4TT
Jo V 1-A sin a
The relation between the twist moment Mz and the total angle of twist ft is shown
in Fig. 2.14(a), which is based on the numerical integration of Eq. (2All). A typical
distribution of the tangential traction q(r) is illustrated in Fig. 2.14(b), which is obtained
fromEqs. (2.112) and (2.119).
(b)
Figure 2.14. Twist angle and tangential traction by torsion (from Lubkin, 1951): (a) Twist
moment Mz versus total twist angle /J; (b) Typical distribution of tangential traction.
2.4 Collision of Elastic Spheres

The basic theories of elastic deformations associated with various contact forces
under static contact conditions have been introduced in the last section. Assuming that
an impact process of two solids can be regarded as quasi-static, the theories given in 2.3
are used directly to link the dynamic deformations of the colliding solids with the impact
forces. In this section, the collisions of elastic spheres are described.
2.4.1 Normal Collision of Elastic Spheres

Collisions between particles with smooth surfaces may be reasonably approxi-
mated as elastic impact of frictionless spheres. Assume that the deformation process during
a collision is quasi-static so that the Hertzian contact theory can be applied to establish the
relations among impact velocities, material properties, impact duration, elastic deformation,
and impact force.
Consider a collision between two frictionless elastic spheres so that only normal force
and normal velocities are involved; i.e., no tangential forces or tangential velocities need
to be accounted for in this case. A general case is shown in Fig. 2.15, where two spheres
with different sizes, velocities, and material properties collide with each other. Only the
collisional force is considered.
From Newton's law, it gives
dU2
(2.123)
where m\ and ni2 are masses of particle 1 and particle 2, respectively; U\ and U2 denote
the corresponding collision velocities; and /12 is the collisional force. Defining a as the
approaching distance between the centers of the particles Oi and O2 during the collision, it
yields
(2.124)
da (2.125)
dt2
al ,EX a 2 , E2
Figure 2.15. Normal collision of elastic spheres.

2.4 / Collision of Elastic Spheres 73
Note that fn can be related to a by Eq. (2.75) as
fa = ^ E * v W (2.126)
Hence, we have
= V a a~i2 (2.127)
dt2 3 m*
where m* is the relative mass, defined by
= + (2.128)
m* m\ rri2
The initial conditions are given by
da\
= J7io - U 2Q = I/12; r=0 = 0 (2.129)
d
'A
Multiplying both sides of Eq. (2.127) by cto/df yields
Since the maximum deformation occurs at da/dt = 0, we have
From Eq. (2.126), the maximum collisional force fm can be expressed by
(2.32,
From Eq. (2.75), the corresponding maximum radius of collisional contact area can be given
by
The maximum pressure in a collision can also be given based on Eq. (2.70) as
1 F*
(2.134)
The collisional contact duration can be obtained by rearranging Eq. (2.130) to the form
where f is defined as a/a m . Hence, the total collisional time duration tc can be obtained as
(2.136,
1.00
0.75
0.50
\
0.25
\
4- compression
! i \
1.47 2.943
0 1.0 2.0 3
X
Figure 2.16. Change of contact area with time.
The dependence of the contact area Ac on the time of the compression process can be
obtained from Eq. (2.130) and Eq. (2.75) as
dAc 16
(2.137)
~dF
Since the rebound process is identical to the compression process, the change of the contact
area with time in a rebound process follows a path similar to that of the compression process,
as illustrated in Fig. 2.16. In this figure, the dimensionless area A* and dimensionless time
r are defined by
A r
' cm
2
(2.138)
As an approximation, the change of A* with r in a collision may be fitted into a sinusoidal

form by
7t
A* = sin (2.139)
2.943
2.4.2 Collision of Frictional Elastic Spheres

As noted before, the collision or dynamic contact from normal compression with-
out any tangential forces involved is based on the Hertzian contact theory. Similarly,
collisions with frictional forces can be approached using Mindlin's contact theory. To sim-
plify the analysis, the effect of torsion is neglected. Consider the impact of two balls in a
coplanar motion, as shown in Fig. 2.17. / t is the frictional force, V\ and V2 are the tangential
velocities, and co\ and a>2 represent the angular velocities. The momentum equation in the
tangential direction is given by
d
\ (Vi +<o\a\) = ft; J = -ft (2.140)
at d7'
Figure 2.17. Two-dimensional collision with tangential friction and twist.
For each ball, the angular momentum about the center of contact O is conserved provided
that the contact area is small and no external torque exists. Hence, we have
j - [mi Vxai + mxeoi {a\ + k\)} = 0

(2.141)
d r
1m 2 V2a2 + mm2a)2a)
2{a\
2v[ul 7
dt
where k\ and k2 are the radii of gyration of the balls about their centers of mass, which gives
d(D\ a\ dV\ dco2 a2 dV2
(2.142)
dt ~ a\ + k\ dt dt ~~ a\ + k\ dt
Thus, substituting the preceding equation into Eq. (2.140) yields
mi dVi m2 dV2 (2.143)
/t =
d/
Let
m2 1 _ 1 1
mi = m2 = (2.144)
m* mi m2
We have
4
m*
(2.145)
or
ft
(2.146)
dt2 m*
where 8 is the relative tangential displacement on the contact surface. The tangential de-
formation is affected by the existence of microslip, the magnitude of normal compressional
force /12 (no matter whether / i 2 is constant or not), the coefficient of friction, and the
loading history (i.e., loading, unloading, and reloading) [Johnson, 1985].
For the case in Fig. 2.17, the tangential velocities of the colliding spheres are not equal
to each other before the impact, which would most likely lead to a sliding on the contact
interface. Consider the simple case of two colliding spheres without relative rotation. To
analyze this case, a constant coefficient of friction is assumed. It is further assumed that
(1) the elastic deformation is small compared to the radii of the spheres so that the relative
tangential shift of the two spheres has no effect on the change of normal velocity components
throughout the whole course of collision; (2) the tangential velocity difference is not too
high nor too low so that the maximum compression by the normal velocity component can
be achieved while sliding is maintained in the collision. From Eq. (2.140) and Amontons's
law, we have
dVn = _hfn
dt m*
where Vn is the relative tangential velocity (V2 Vi). Using the Hertzian relations for the
normal deformation and substituting Eq. (2.126) into Eq. (2.147) yield
dVn 4*/ k Va* 3
ir = T T 5 ( 2 - 148 )
dt 3m*
The relation between a and t is obtained from Eq. (2.135) (where is a/am) as
' = 7T A 1 - * ' ) " * < * (2 149)

-
U\2 JO
for the compression process and
1
' c - ' = ~ f (l-
(^r (2.150)
^ 1 2 JO
for the rebound process. Thus, the change of relative tangential velocity after the collision
can be calculated by
a # - m , f*df (2.151)
3m* Jo
Substitution of Eq. (2.131), Eq. (2.149), and Eq. (2.150) into Eq. (2.151) yields
(2.152)
which shows that the change of relative tangential velocity is related only to the surface
friction coefficient and normal velocity component and is independent of other material
properties. Equation (2.152) also indicates that, to ensure the sliding assumption, the relative
incident angle should be larger than tan~ 1 (2/k).
It is clear that the collision between two elastic but frictional spheres is inelastic due
to the inevitable sliding at contact which yields the kinetic energy loss by frictional work.
Furthermore, the preceding analyses of both Hertzian collision and frictional collision can
also be applied to the particle-wall collision, where the radius of the wall is simply set to
be infinitely large.
Example 2.2 Consider an impact between a polyethylene particle (dp = 1 cm) and a
copper wall. The incident velocity is 2 m/s, and the incident angle is 30. The friction
coefficient of the interface is 0.2. The densities of polyethylene and copper are 950 and
8,900 kg/m3, respectively. What is the contact time duration for the collision? Estimate the
rebound velocity of the particle. Repeat the problem for a copper particle colliding with a
polyethylene wall.
Figure E2.2. Collision of a sphere with a wall.
Solution As shown in Fig. E2.2, the normal velocity Un is given by
Uu = U cos 0 = 2 cos 30 = 1.73 m/s (E2.7)
The wall can be regarded as a large sphere with an infinite radius. Thus, we have
a* = ^-; m*=m = ^d3pPp (E2.8)
The following data are available from Table 1.6
Ei = 0.04 x 1010 N/m2; vi = 0.3 (E2.9)
for polyethylene, and
E2 = 12 x 1010 N/m2; v2 = 0.35 (E2.10)

for copper.
Thus, it gives
(E2.ll)
For a polyethylene particle hitting a copper wall, we have
a* = ^ =0.005 m; m* = - d 3y p = - x 10"6 x 0.95 x 103 = 0 . 5 x 10" 3 kg (E2.12)

2 6 6
Hence, the collision time is calculated by Eq. (2.136)
2.94 / 1 5 m*U}2 \ * 2.94 / 1 5 0.5 x 10"3 x 1.732 \ 5

^ __ I L_ ] ^ [ j _ j 2 x 10 S
c
U \16 E*Vd*J 1 . 7 3 V 1 6 4 . 38
38 x l 8 0 V8 //
n
(E2.13)
and the change of tangential velocity is obtained by Eq. (2.152)
A V12 = 2/ k t/i2 = 2 x 0.2 x 1.75 = 0.7 m/s (E2.14)

Thus, the rebound velocity is
U' = [U^ + iUsinO- AV 12 ) 2 ]* = [1.732 + (2sin30 - 0.7)2 p = 1.76m/s (E2.15)
Similarly, for a copper particle hitting a polyethylene wall, we have
a* = - = 0.005 m; m* = -dip? = - x 10" 6 x 8.9 x 103 = 4.66 x 10"3 kg

2 6 p 6
(E2.16)
Hence, it yields
2 9 15 m
t = ' V *^ V = 2 - 9 V 1 5 466 x 1Q 3 x L732
" =3 V Q ~ iQ-4 s
tfi2\16E*vW 1.73 V16 4.38 x 108 x V a 0 0 5 /
(E2.17)
while the rebound velocity remains the same as the previous case since the change in
tangential velocity difference depends only on the friction coefficient and normal velocity.
2.5 Collision of Inelastic Spheres

Any solid material has its own upper limit of elastic deformation under either
normal or tangential stresses. Once the stresses exceed this limit, plastic deformation will
occur. In this section, collisions of inelastic spheres are presented. The degree of inelastic
deformation is characterized by the restitution coefficient.
2.5.1 Onset of Plastic Deformation

In order to estimate the onset yield stress of the material, three common criteria
are introduced. The Tresca criterion is based on maximum shear stress and is given as
max(|ori - or2|, \cr2 - <r3|, \a3 - ax\) = In = Y (2.153)
where cr\, cr2, a n d 03 are the principal stresses; n is the yield stress in simple shear; and Y
is the yield stress in simple compression or tension. The von Mises criterion considers the
shear strain-energy, which is calculated by
(cri - a 2 ) 2 + (a2 - a3)2 + (a3 - ax)2 = 6n2 = 2F 2 (2.154)
The third criterion is obtained from the maximum reduced stress and is expressed as
2
max(|a! - a|, \a2 - cr\, \cr3 - a\) = n = -Y (2.155)
where a = (a\ + a 2 + ^3)73.

No significant difference is found for the predictions of the yield stresses from these
three criteria. However, Tresca's criterion is more widely used than the other two because
of its simplicity. When two solid spherical particles are in contact, the principal stresses
along the normal axis through the contact point can be obtained from the Hertzian elastic
2.5 / Collision of Inelastic Spheres 79
contact theory as [Johnson, 1985]
(2.156)
where z denotes the depth inside the sphere along the axis of symmetry; rc is the radius of
contact area; and p0 is the maximum pressure on the contact surface.
It can be proved from Eq. (2.156) that, for materials with Poisson's ratio of 0.3 (which
is true for most solids), the maximum shear stress \az crT\ occurs at z/rc = 0.48. Con-
sequently, according to Tresca's criterion, the yield stress Y in a simple compression is
0.62 p0. Therefore, when the hardness or the yield stress Y of the particle material is less
than 0.62 times the maximum contact pressure, the sphere will, most likely, undergo plastic
deformation. From the elastic collision of two solid spheres, the maximum contact pressure
is given by Eq. (2.134). Thus, the relation between the critical normal collision velocity,
/i2Y> and the yield stress is given by
(2.157)
Moreover, it can be shown from Eq. (2.156) that the maximum shear stress \oz or\ is
insensitive to the variation of Poisson's ratio for solids, only deviating 5 percent when
Poisson's ratio varies from 0.27 to 0.36 (see Problem 2.4). Thus, Eq. (2.157) can also be
applied for solids with Poisson's ratio in the range of 0.27 to 0.36.
Example 2.3 For a normal impact between a polyethylene particle (d p = 1 cm) and a
copper wall, estimate the critical normal collision velocity of the particle, above which
plastic deformation would occur. The yield strength for polyethylene is 2 x 107 N/m2 and
the yield strength for copper is 2.5 x 108 N/m2. What would be the critical normal collision
velocity for a copper sphere colliding with the copper wall?
Solution Since the yield strength for polyethylene is lower than that for copper,
the polyethylene particle will undergo plastic deformation before the copper wall does in
the collision. As obtained in Example 2.2, we have
m* = 0.5 x 10" 3 kg; a* = 0.005 m; * = 4.38 x 108 N/m2 (E2.18)
Thus, from Eq. (2.157), the critical normal collision velocity is
If the particle is copper, we have
m* = 4.66 x 10"3 kg; a* = 0.005 m (E2.20)

Hence, it yields
0.0053 x (2.5 x 108)5

4.66 x 10- 3 x (6.8 x
This example illustrates that even small impact velocities in collisions between metallic
materials (or with hard solids) may lead to plastic deformation or breakage of colliding
parts.
2.5.2 Restitution Coefficient

Once the relative impact velocity between two colliding spheres is higher than the
critical yield velocity, plastic deformation must occur. Heat loss is another phenomenon
often coupled with such collisions. Collisions with plastic deformation are referred to as
inelastic collisions. All the energy transfer in the form of plastic deformation and heat loss
in an inelastic collision is considered as a kinetic energy loss.
The recoverability or restitution of the kinetic energy during a normal collision between
two objects can be represented by the coefficient of restitution defined by Eq. (2.3). Note
that the coefficient of restitution cannot be used as a criterion to judge whether a collision
is elastic or not unless the collision is solely considered as a normal collision. For example,
the sliding at contact for the collision between two elastic spheres will make the collision
inelastic while the value of the coefficient of restitution in this case is equal to 1.
A complete course of collision consists of the compression and rebound processes. The
total work done during the compression process without heat loss can be calculated as
m
^m*U?2 = [ fa da (2.158)
2 Jo
and the total work done during the rebound process without heat loss is
\m*U?2= I" f{2da' (2.159)

^ Jo
where a'm is approximately equal to am when the ratio of the plastic deformation to the
elastic deformation is negligibly small.
Prediction of the restitution coefficient has been a challenging research topic for decades.
Unfortunately, no reliable and accurate prediction method has been found so far. However,
some useful simplified models with certain limits have been developed. One of them is
the elastic-plastic impact model in which the compression process is assumed to be plastic
with part of the kinetic energy stored for later elastic rebounding, with the rebound process
considered to be completely elastic [Johnson, 1985]. In this model, it is postulated that
(1) during the plastic compression process, a = r^/2^*; (2) during the compression pro-
cess, the averaged contact pressure pm is constant and is equal to 37; and (3) the elastic re-
bound process starts when maximum deformation is reached. Therefore, the compressional
force is
/i2 = nrlpn (2.160)

2.5 / Collision of Inelastic Spheres 81
so that the total work during the compression process is given by
Contrary to the rather straightforward calculation of the total work during the com-
pression process, the rebound work is hard to estimate because of the difficulties in the
formulation of the rebound force. In the recovery process, the repelling force is related
to the radius of contact area by the elastic deformation. However, for the same degree of
compression (i.e., for the same a), the contact area in a fully plastic deformation is twice
as big as that in an elastic case. Since the actual repelling force starts from the contact area
by plastic deformation, the variation between the repelling force and the radius of the con-
tact area in the rebounding process follows neither fully elastic nor fully plastic relations.
Hence, an approximation has to be made to express the repelling force in terms of the radius
of the contact area rc or the approaching distance a.
First, for both elastic and plastic deformation, we have
S
f =4 (
Furthermore, assume that Eq. (2.74), obtained for Hertzian contact, is valid with r c being
the actual radius of contact area so that
/12 = -^-ct (2.163)
which, combined with Eq. (2.162), yields the relation between a and r c for the rebounding
process as
/m (2 164)
= d^ -
Therefore, the rebounding velocity can be obtained by Eq. (2.159) as
The coefficient of restitution is then obtained by eliminating r cm from Eq. (2.161) as
e = (2.166)
The elastic-plastic model reveals that the restitution coefficient depends not only on the
material properties but also on the relative impact velocity. Equation (2.166) also indicates
that the restitution coefficient decreases with increasing impact velocity by an exponent of
1/4, which is supported by experimental findings, as shown in Fig. 2.18. For high relative
impact velocities, the model prediction is reasonably good. However, for low relative
impact velocities, the prediction may be poor because the deformation may not be in a fully
plastic range as presumed.
Eliminating the yield stress in Eq. (2.166) by substitution of the critical yield velocity
from Eq. (2.157), the restitution coefficient becomes
i
(2.167)
j. hard bronze
o brass
lead
slope of lines = -1/4
0.1
0.1 1.0 10.
Velocity of impact V, m/s
Figure 2.18. Measurements of the restitution coefficient of a steel ball on blocks of various
materials (from Goldsmith, 1960).
The preceding equation is only valid for a certain range of impact velocities. The lower
limit comes from the fact that the coefficient of restitution has to be less than unity so that
Un (2.168)
The upper limit of the impact velocity can be deduced from the fact that r cm < min(i,
For a collision between two spheres of the same diameter dp, Eq. (2.161) yields
(2.169)
Similarly, for the collision of a sphere with a solid wall, we have
Un <0.53W (2.170)
where Y = min(7i, Y2) and p p is density of the sphere.
Example 2.4 A copper ball of 1 cm diameter normally collides with a stainless steel wall
with an impact velocity of 0.5 m/s. Estimate the restitution coefficient using the elastic-
plastic model. What is the rebound velocity of the ball? The yield strength of copper is
2.5 x 108 N/m2. It can be assumed that the yield strength of the stainles steel is higher than
that of copper.
Solution First, let us calculate the yield collisional velocity between the copper
ball and the stainless steel wall. Assume the radius of the wall is infinitely large so that
a* = - = 0.005 m; m* = -dlpp = - x 10"6 x 8.9 x 103 = 4.66 x 10"3kg

2 6 6
(E2.23)
Nomenclature 83
The following data are available
Ei = 12 x 1010 N/m2; vi = 0.35 (E2.24)

for copper and
E2 = 20 x 1010 N/m2; v2 = 0.28 (E2.25)

for stainless steel. Thus, we have
1-0.352 l-O^^" 1
E*='1-Vt iU
= 8.4 x 10IOXT /2
N/m
Ei E2 20 x 1010
(E2.26)
The critical normal collision velocity is
1
a*3Y5 _ / O.OO53 x (2.5 x 108)5
= 7.5 x 10~3 m/s
m**4 ~ ' V 4.66 x 10~3 x (8.4 x 1010)4
(E2.27)
To examine the velocity range suitable for the elastic-plastic model, the lower limit is
given by
UX2 > 27.5UUY = 27.5 x 0.0075 = 0.21 m/s (E2.28)
and the upper limit is
T _ 12.5 x 108 _
(E2.29)
Thus, U\2 = 0.5 m/s is reasonable for the model. From Eq. (2.167), the restitution coeffi-
cient is obtained as
0.0075
(E2.30)
Hence, the rebounding velocity is found from Eq. (2.166) as

U[2 = eUn = 0.80 x 0.5 = 0.40 m/s (E2.31)
Nomenclature
Constant, defined by Eq. (2.26) C Constant, defined by Eq. (2.42)
Contact area C Constant, defined by Eq. (2.67)
a Particle radius c Radius of contact circle without
a* Relative radius, defined by Eq. microslip
(2.66) dp Particle diameter
B Constant, defined by Eq. (2.32) E Total kinetic energy
E Young's modulus t Time

E Complete elliptic integral of the tc Contact time duration
second kind, defined by Eq. U Normal velocity component in a
(2.117) biparticle collision
Contact modulus, defined by Eq. U\ Critical normal collision velocity
(2.66) Ui Normal velocity component for
e Coefficient of restitution particle 1 in a biparticle collision
F Force u[ Normal rebounding velocity
Ft Tangential force component for particle 1 in a
Fz Normal impact force (or force biparticle collision
component) u2 Normal velocity component for
f Coefficient of friction particle 2 in a biparticle collision
hi Collisional force between Normal rebounding velocity
particles 1 and 2 component for particle 2 in a
biparticle collision
h Coefficient of kinetic friction
Maximum collisional force in a V Tangential velocity component in
/m
normal impact a biparticle collision
/s Coefficient of static friction X Cartesian coordinate
G Shear modulus Y Yield stress in simple
h Height compression or tension
I Moment of inertia y Cartesian coordinate
J Normal impulse during the z Cylindrical coordinate
compression process
Greek Symbols
r Normal impulse during the
rebounding process a Center approaching distance
K Complete elliptic integral of the Center approaching distance
first kind, defined by Eq. (2.116) under maximum deformation
k Radius of gyration Arbitrary angle, defined by Fig.
i Displacement 2.8
Mz Twist moment ft Twist angle
m Mass of a particle y Shear strain
m* Relative mass Normal strain
n Yield stress in simple shear 0 Cylindrical coordinate
P Pressure 0 Spherical coordinate
Pm Averaged contact pressure 8 Distance between the points of
q Tangential traction application of pair forces
R Spherical coordinate Rigid displacement
r Cylindrical coordinate f Dimensionless center approaching
rc Radius of contact circle distance
s Arbitrary coordinate Parameter for penetration of
T Total stress tensor microslip, defined by Eq. (2.120)
Poisson's ratio
Problems 85
Pv Density of particles X Parameter for penetration of

a Normal stress microslip, defined by
X Shear stress Eq. (2.115)
Length, defined by Eq. (2.83) Love's stress function
Spherical coordinate Angular velocity
References
Boussinesq, J. (1885). Application des Potentiels a VEtude de VEqulibre et du Mouvement des
Solids Elastiques. Paris: Gauthier-Villars.
Goldsmith, W. (1960). Impact: The Theory and Physical Behavior of Colliding Solids. London:
Edward Arnold.
Hertz, H. (1881). Uber die Beriihrung fester elastischer Korper. J. Reine Angew. Math. (Crelle), 92,
155.
Johnson, K. L. (1985). Contact Mechanics. Cambridge: Cambridge University Press.
Lame, G. and Clapeyron, B. P. E. (1831). Memore sur Iequilibre interieur des corps solides
homogenes. /. E Math. (Crelle), 1.
Landau, L. D. and Lifshitz, E. M. (1970). Theory of Elasticity, 2nd ed. New York: Pergamon Press.
Love, A. E. H. (1944). A Treatise on the Mathematical Theory of Elasticity, 4th ed. New York:
Dover.
Lubkin, J. L. (1951). The Torsion of Elastic Sphere in Contact. Trans. ASME, J. ofAppl. Meek, 18,
183.
Mindlin, R. D. (1949). Compliance of Elastic Bodies in Contact. Trans. ASME, J. ofAppl Meek,
16, 259.
Timoshenko, S. P. and Goodier, J. N. (1970). Theory of Elasticity. New York: McGraw-Hill.
Problems
2.1 Show that the selected set of stresses expressed by Love's stress function satisfy the equation
of equilibrium for axisymmetric and torsionless deformation in a solid.
2.2 A spherical solid ball is in normal contact with a spherical seat, as shown in Fig. P2.1. Derive
the expressions for the radius of the contact area, rc, and the distance change between the
centers of the contact bodies, a. The radii of the spherical seat and the solid ball are R\ and
R2, respectively.
Figure P2.1. Normal contact of a spherical solid ball with a spherical seat.
Figure P2.2. Collision between a solid ball and a spherical seat.
2.3 Derive the expressions for the maximum collision force and the collision time when a
spherical solid ball collides with a spherical seat, as shown in Fig. P2.2, with a contact
velocity U. This seat can be assumed unmovable.
2.4 Using the stresses given in Eq. (2.156), show that the yield stress in simple compression
varies from 0.59 to 0.63 times the maximum pressure for materials with Poisson's ratio
varying from 0.27 to 0.36.
2.5 Calculate the yield impact velocity for the collision between a stainless steel ball bearing
and a brass ball. The diameter for each ball is 3 mm.
2.6 Verify that, for a collinear collision of two frictionless, nonspinning, rigid spherical particles,
the total kinetic energy loss can be expressed by Eq. (2.6).
CHAPTER 3
Momentum Transfer and Charge Transfer
3.1 Introduction
In gas-solid flows, flow patterns of both phases depend not only on the initial
conditions and physical boundaries of the system but also on the mechanisms of momentum
transfer or the interacting forces between the phases. The forces controlling the motions of
particles may be classified into three groups: (1) forces through the interface between fluid
and particles, (2) forces due to the interactions between particles, and (3) forces imposed
by external fields. Although interparticle forces and field forces do not directly change
the course of the fluid motion, they may indirectly influence the motion via particle-fluid
interactions.
Among the significant characteristics in a gas-solid flow are the charge transfer and
charge accumulation among particles due to the interparticle collisions and/or particle-wall
collisions. The charges carried by the particles may directly affect the particle motion and
particle concentration distribution in the flow by the electrostatic forces, especially when
an external electric field is applied. Even in the absence of an external electric field, the
accumulation of particle charges may yield adverse effects such as arc discharge and dust
explosion.
In this chapter, various forces affecting the motion of a gas-solid flow along with the
momentum equations governing the motions of particles are delineated. The mechanisms
of charge generation and charge transfer by collisions of particles are discussed.
3.2 Particle-Fluid Interactions

The motion of a single particle without short-range interactions such as the van der
Waals force and collisional forces in the absence of external force fields may be generally
described by a nonspherical particle accelerating and rotating in a nonuniform flow field.
The complex movement of the particle may be logically perceived to comprise four types
of simple motion of a sphere or an equivalent sphere assuming that they are noninteractive:
(1) a particle moving with a constant velocity in a uniform flow field, (2) a particle accele-
rating in a uniform flow field, (3) a particle moving with a constant velocity in a nonuniform
flowfield,and (4) a particle rotating with a constant angular velocity in a uniform flow field.
Four different forces, each representing one of the types of simple motion given, are the
drag force, Basset force, Saffman force, and Magnus force, respectively.
3.2.1 Drag Force

In a particle-fluid flow, the particle velocity Uv generally differs from the fluid
velocity U. The slip velocity, U f/p, leads to unbalanced pressure distribution as well
as viscous stresses on the particle surface, which yield a resulting force known as the
drag force. The drag force of a single particle in a uniform flow field can be generally
87
3 I Momentum Transfer and Charge Transfer
expressed by
^\U - UV\(U - tfp) (3.1)
where A is the exposed frontal area of the particle to the direction of the incoming flow and
CD is the drag coefficient, which is a function of the particle Reynolds number, Re p , and
the local turbulent intensity of the fluid. Re p is defined by
At the two extremes, Rep ^> 1 but not high enough to exceed that for the transition
regime and Rep ^ 1 for the creeping flow regime, CD is well expressed in light of either
experimental or analytical work. In the high Reynolds number regime, C D = 0.44, originally
determined from Newton's experiments in 1710. Newton's relation covers a range of Re p
from 700 to 105 over which the inertia effect is predominant. In the creeping flow regime
where the viscous effect dictates the flow and the inertia is negligible, Stokes solved the
simplified governing equations and revealed that CD = 24/Re p in 1850.
An enormous amount of experimental data on the drag coefficient for a single sphere
at various Rep has been compiled into a standard curve [Schlichting, 1979], as shown in
Fig. 1.4. As the figure illustrates, the sharp reduction in the drag coefficient at high Re p
(around 3 x 105) corresponds to the transition from a laminar to a turbulent boundary
layer over the particle. This transition is due to the change of surface pressure distribution
around the particle caused by the change of the wake structure behind the particle in the
turbulent regime. It should be pointed out that the standard curve was obtained under
uniform and undisturbed flow conditions. However, in most gas-solid flows, the fluid itself
is usually turbulent even when the relative velocity and the resulting particle Reynolds
number are small. The local turbulent intensity of the fluid causes deviations in CD from
the standard curve. Unfortunately, no general conclusive relations between CD and the
local turbulent intensity of the fluid are available at this stage. The drag coefficient of a
particle in a multiphase system can be further affected by the existence and the movement of
neighboring particles. Studies on the drag forces of interacting particles in multiphase flows
[Rowe and Henwood, 1961; Lee, 1979; Tsuji et aL, 1982; Zhu et al., 1994] suggest that the
drag coefficient of an individual particle under the influence of a neighboring particle may
be expressed by
= i (i A )expfM (3.3)
CDo V dvJ
where CDO is the drag coefficient from the standard curve in Fig. 1.4; / is the distance
between the two interacting particles; and A and B are empirical coefficients which may
be a function of Rep and the deflecting angle between the direction of the relative velocity
and the line connecting the centers of the two interactive particles.
3.2.2 Basset Force

Once a particle is accelerating or decelerating in a fluid, a force known as the
Basset force becomes important. The Basset force can be rigorously derived from the
motion of a single accelerating sphere in the Stokes regime in a quiescent fluid. For particle
Reynolds numbers beyond the Stokes regime, the original Basset term may be modified
3.2 I Particle-Fluid Interactions 89
by introducing a correction factor [Odar and Hamilton, 1964; Clift et a/., 1978]. In this
section, the original form of the Basset force is derived [Basset, 1888], along with a brief
discussion of the importance of this force relative to the Stokes drag.
3.2.2.1 Stream Function for Linear Motion of a Sphere

For simplicity, we start with the derivation in cylindrical coordinates. The resultant equa-
tions are then transformed to polar coordinates in which the final form of the Basset force
is obtained.
Consider a sphere of radius a moving along a straight line in an infinite stagnant fluid
medium at a velocity V. The cylindrical coordinates are selected such that the z-axis co-
incides with the path of the sphere, and the origin is an arbitrarily fixed point on the path.
Assuming that the fluid is incompressible and its motion is axisymmetric, the stream func-
tion \j/ may be defined such that
uT = , u z = - (3.4)
r dz r or
where uT and uz are the radial and axial velocity components of the fluid, respectively.
Assume that the sphere moves in the Stokes regime where the inertia effects can be neglected.
Thus, the Navier-Stokes equations reduce to
(3.5)
z
= - - + vV wz (3.6)
where v is the kinematic viscosity. Defining an operator D as
D= 1 (3.7)
Equations (3.4) through (3.6) yield
1 dp _ 1 d (df \
V
~~pVz--rYr\JF~ )
1 dp 1 d (dxlf \
= vDxIf I
p dr r dz \ dt )
Eliminating the pressure term from the preceding equations leads to
The solution of this equation is
where ^r[ and xfr^ respectively, satisfy
= 0 (3.11)
Consequently, the pressure field is determined by
dp = - I dr - !- dz (3.12)
r dt \ dz dr J
90 3 I Momentum Transfer and Charge Transfer
0 r
Figure 3.1. Transformation from (r, z) to (R, 6).
The preceding equations are obtained with the coordinate defined relative to a fixed
origin. Now, consider a coordinate defined with the center of the sphere as the origin and
denote as the distance from afixedpoint. We may express the axial component of velocity
uz as
(3.13)
so that
(3.14)
dt dt 3%
The second term on the right-side of Eq. (3.14) is of the same order as the square of the
velocity and so represents an inertia. Hence, this term can be omitted in Stokes flows.
Therefore, it can be stated that both Eq. (3.5) and Eq. (3.6) are valid whether the origin is
fixed or in motion.
We now transform the governing equations in cylindrical coordinates into polar coor-
dinates. Since the motion is axisymmetric, the transformation from (r, z) to (R,0), as
shown in Fig. 3.1, is analogous to the transformation from Cartesian coordinates (x, y) to
cylindrical coordinates (r, 0) in a two-dimensional domain. The stream function is related
to the velocity components in polar coordinates by
1 d\ff
(3.15)
0 RsinOdR
The operator D, by analogy, is changed directly from Eq. (3.7) to
a2 1a I a
2
1 /a
2 (3.16)
dR RdR R2d02 RsinOKdr
Noting that
do cos# a
(3.17)
I
\drJzdR
Equation (3.16) becomes
a2 I a2 cot<9 a
D = (3.18)
dR2 R2d02 R2 80
Similarly, Eq. (3.12) can be expressed in polar coordinates as
(3.19)
The boundary conditions at the surface of the sphere are

1 d
UR\R=G = VcosO = 2
a sinO 30 R:=a
(3.20)
1 3^
ue\R=a = -V sin# = rr
asmO 3R R=a
3.2.2.2 Solution of Equations in Terms of Stream Function

Equation (3.9) in polar coordinates can be satisfied if ^ is in a form
where yj/\ and \j/2 are functions of R and t, which satisfy
= 0 (3.22)
3R2 R2
32\j/2 2\j/2 1 3x^2
2 = (3.23)
~3R ~ ~W v~JT
Solutions of Eq. (3.22) and Eq. (3.23) can be expressed by
i f^ r ,., ( c2 (3
2R\ vtjo - - ' ''-* -24)
and
R nr a r </>() / (R-a + &2\
2 \j vt dR Jo R \ Avt )
where x (O and 0() are functions to be determined by the boundary conditions given by
Eq. (3.20).
Using Eq. (3.20) and assuming 0(0) = 0r(O) = 0 and 0() and 0 / () are finite as
-> oo, x and 0 can be determined as [Basset, 1888]
Va (3 2 2\
X(f) = - f + ^ aK + a (3.26)
n \2 )
and
^ (3.27)
Lit
Hence, \j/\ and fo are obtained as
and ^ = l ! f3vt +
3a V 2
f2 = - ^ i ( (2Wvt - R +a) + 2Wvt - R +a)e-^ drj = -Vf(t)
(3.29)
92 31 Momentum Transfer and Charge Transfer
The preceding solution is for the case involving a particle moving at a constant velocity
V. For the case involving an accelerating particle, in order to obtain the stream function,
we introduce the following theorem first. Let vs(t, R) be a solution of the equation
where O is an operator. Then, for the initial condition v s(0, R) = 0, w is also a solution of
Eq. (3.30) if
nt
w= G(f-T)v s (r,/?)dr (3.31)
Jo
where G(r) is an arbitrary function which is independent of R and t, and does not become
infinite between the limits. Thus, replacing G(r) in Eq. (3.31) with F'(r) (=dF/dr) and
noting that /(0) = 0 and f{t) is a solution of Eq. (3.23), we may rewrite T/^ as
ir2 = - f F\t - r)/(r)dr (3.32)

Jo
where F\x) is a function to be determined. The new \j/2 is also a solution of Eq. (3.23).
Following the same procedure, a new \jr may be deduced from Eqs. (3.28) and (3.29) by
changing t into r and V into Fr{t r) dr, and integrating the resulting equations from t to
Oas
^ + a2) - / ( r ) l d r (3.33)
7T / J
Equation (3.33) is also a solution of Eq. (3.9). If F(0) = 0, it can be found by substituting
the new xj/ into Eq. (3.20) that F(t) is the velocity of the sphere from rest. Hence, Eq. (3.33)
represents the stream function for the case of the accelerating motion of a particle.
3.2.2.3 Basset Force and Carried Mass

In evaluating the particle resistance due to relative motion to the fluid, it is convenient to
begin with the constant velocity case and work to the time-dependent case. Consider the
case of a sphere moving in a gravitational field and assume that the gravitational acceleration
is in the same direction as the sphere motion. Denote Fz as the resistance. Since the total
resistance is due to the surface shear and surface pressure, and the flow is axisymmetric, F z
can be postulated as
Fz = 2na r ((pa
cos 0 - p sin 2 0) sinfl dO (3.34)
^0 (V t )a
From Eq. (3.19) and taking the gravitational force into consideration, we have
^0 V
dp aVi
- =psin0-^- - gpasinO (3.35)
ou ot oK
Substituting Eqs. (3.28), (3.29), and (3.35) into Eq. (3.34) yields
M (336)
^
3 2 I Particle-Fluid Interactions 93
where Ma is the mass of the fluid displaced. Changing t into r and V into F'(t r) dr and
then integrating the term in square brackets in Eq. (3.36) from t to 0, Eq. (3.36) becomes
(Problem 3.3)
(3.37,
where v is equal to F(t) with F(0) = 0. The first and last terms on the right-hand side of
Eq. (3.37) are the force due to the acceleration of the carried mass and the Basset force,
respectively. The second term is the buoyancy force. The third term is the Stokes drag
force. The carried mass is also termed the added mass or the virtual mass; it contributes
to the resistance because acceleration of the particle is also accompanied by acceleration
of the fluid; the mass of the carried fluid that has the same acceleration as the particle is
noted as the carried mass. The volume of the carried mass, which generally depends on the
particle geometry, is equal to one-half the particle volume when the particle is spherical.
The Basset force or the Basset history integral accounts for the effect of past acceleration
on the resistance. Here (t r) represents the time elapsed since past acceleration from 0
tot.
The Basset force can be substantial when the particle is accelerated at a high rate. The
total force on a particle in acceleration can be many times that in a steady state [Hughes and
Gilliland, 1952]. In a simple model with constant acceleration, the ratio of the Basset force
to the Stokes drag, /?BS> was derived [Wallis, 1969] and rearranged to [Rudinger, 1980]
(3.38)
where rs is the Stokes relaxation time defined as
rs = ^ (3.39)
18
The Basset force may be negligible when the fluid-particle density ratio is small, e.g., in
most gas-solid suspensions, and the time change is much longer than the Stokes relaxation
time or the acceleration rate is low.
Example 3.1 Consider the collision of two particles due to a wake attraction, as shown
in Fig. E3.1. Some assumptions for the collision process are as follows:
(1) The motion of the leading particle is not affected by the approach of the trailing
particle.
(2) Particles are equal-sized, rigid, and spherical.
(3) Initially, the velocities of the two particles differ only slightly and are close to their
respective particle terminal velocities. The particles are separated by a character-
istic distance /o which is within the wake-influencing distance.
(4) Particle acceleration is proportional to the relative velocity.
Develop a model to estimate the relative particle collision velocity due to the wake
attraction.
Figure E3.1. Schematic flow pattern of two interactive spheres due to wake attraction.
Solution A general momentum equation for the trailing particle takes the form
1 \ dU
2P Vdt = (/ p

where V is the volume of the particle; U is the relative velocity of the trailing particle to
the leading particle; and FB is the Basset force. It is assumed that the expression for the
Basset force given in the following can be extended beyond the Stokes regime
dU
L o y/t - X
On the basis of Assumption (3), the drag force of a single noninteracting particle FDo is
(E3.2)
written as
= (PP - p)Vg (E3.3)
The drag coefficient of the trailing particle can be given by Eq. (3.3). Substituting Eqs. (E3.2),
(E3.3), (3.1), and (3.3) into Eq. (E3.1) yields
dU
dU _ 2(p p - p)
g{\ - A)exp
B rtc Udt\- dr (E3.4)
dt Lpv -\- p {adolt
v j t ) yz,pv -t- p)7iav j 0
where tc is the elapsed time for particles to collide. From Assumption (4), we have
=CU (E3.5)
dt
32 I Particle-Fluid Interactions 95
which leads to
/ = / c exp[-C(f c -O] (E3.6)

where Uc is the collision velocity and C is a constant. On the basis of Eq. (E3.6), the initial
velocity may be expressed by
Uo = Uc exp(-Ctc) (E3.7)
Substituting Eqs. (E3.5) and (E3.6) into Eq. (E3.4) gives rise to
ISU
(E3.8)
where erf(x) is the error function of x. At the moment of contact, Eq. (E3.8) reduces to
(UC-UO)TT 2(pp-p) _ 18C/C h ~ Uo)

erf
/n (2pp + p)dv y /0
(E3.9)
Equation (E3.9) expresses the collision velocity implicitly as a function of the given flow
conditions and the properties of the fluid and particles.
3.2.3 Saffman Force and Other Gradient-Related Forces

When a sphere moves in a flow where a velocity gradient, pressure gradient, or
temperature gradient exists, additional forces related to these gradients can be as important
as the drag force.
3.2.3.1 Saffman Force

In a region where a velocity gradient exists, for instance, near a wall or in a high shear
region, a sphere moving at a constant velocity is subjected to a lift force caused by the
velocity gradient. This lift force is termed the Saffman force. The Saffman force was
originally derived for the motion of a sphere at a constant velocity in a simple shear flow
at low Rep [Saffman, 1965, 1968]. Using asymptotic expansions and Fourier transforms of
the velocity field, Saffman was able to deduce an expression for the lift force
(3.40)
dy
where the constant K was determined to be 6.46 on the basis of a numerical integration for
creeping flows at low shear rates. The ratio of the Saffman force to the Stokes drag, Rss, is
given as
Kd 1 d(U-Up)
(3.41)
I2n \ v dy
In a shear flow with a constant shear rate, Rss can be estimated by
(3.42)
dp/dz
Figure 3.2. Force on a sphere due to pressure gradient.
which indicates that the Saffman force is negligible at very small shear rate or very low
Rep.
3.2.3.2 Force Due to Pressure Gradient

Besides the drag force, Basset force, and Saffman force, another force may act on the
particle as a result of the existence of a pressure gradient in the fluid. Using the axisymmetric
condition, the force on a differential element of a sphere in a pressure gradient field shown
in Fig. 3.2 can be expressed by
dfp = -lira3 sin 0 cos2 0 &0 (3.43)

dz
Equation (3.43) gives
Fv = -lira3 r ^ sin 0 cos2 0 dO = - ^ ^ (3.44)

Jo dz dz 6
The minus sign indicates that the force is in the opposite direction of the pressure gradient
[Tchen, 1947]. The significance of this force is evident, for example, when a shock wave
propagates through a gas-solid suspension.
3.2.3.3 Radiometric Force

Forces acting on a particle as a result of a temperature gradient in a gas (thermophoresis)
or nonuniform radiation (photophoresis) are known as radiometric forces. For a particle
diameter much larger than the mean free path of a gas, the force due to the temperature
gradient is given by [Hettner, 1926]
r
(3.45)
2 ' p
where RM is the gas constant, and the temperature gradient inside the sphere, V7 p , is given
by [Epstein, 1929]
vrp = vr (3.46)
where K and Kv are the thermal conductivities of the gas and the solid, respectively.
Equation (3.45) shows that the force is directed toward the lower-temperature region.
At room conditions and in air, the radiometric forces are important only for submicrom-
eter particles. However, under conditions of high temperatures and/or large temperature
gradients such as found in plasma coating, the thermophoresis effect may also be significant
for larger particles.
3.2.4 Magnus Effect and Force Due to Rotation of a Sphere

Particle rotation may result from collisions of the particle with a solid wall or other
particles, or from a velocity gradient in a nonuniform flow region. In low Reynolds number
flows, particle rotation leads to fluid entrainment, resulting in an increase in the velocity
on one side of the particle and a decrease in the velocity on the other side. Thus, a lift
force is established which moves the particle toward the region of higher velocity. This
phenomenon is known as the Magnus effect [Magnus, 1852] and the lift force is called the
Magnus force.
For a sphere rotating in a uniform flow at low Reynolds number, the lift force can be
obtained by solving the Navier-Stokes equations using an asymptotic expansion and a
matching technique [Rubinow and Keller, 1961]. The solution procedure is noteworthy
because of its uniqueness in yielding not only the lift force but also the Stokes drag and the
Oseen drag of a spinning sphere. The Stokes stream function method cannot be applied to
this situation because of the asymmetric nature of the flow around the rotating sphere.
Consider a steady motion of a sphere of radius a, velocity Up, and angular velocity ft in
an incompressible fluid at rest, as shown in Fig. 3.3. By selecting the Cartesian coordinates
with the origin attached to the center of the sphere, with the Jt-axis pointing to the opposite
direction of Up, and with the x-y plane containing 17, we have
= 0 (3.47)
U- W = - - (3.48)
P
U\r>=a = X U\r>=00 = (tfp, 0, 0), p\r,=oo = 0 (3.49)
where r' is the radial distance in the spherical coordinate. It is convenient to define dimen-
sionless quantities u,~p,u>, and r by
U r1
u = r= (3.50)
u
Figure 3.3. Forces on a rotating and moving sphere without wake separation.
so that Eqs. (3.47) through (3.49) become

V M= 0 (3.51)
M-Vi Vp+ V2M) (3.52)
*=R^(~
U\r=x U\r =oo = (1,0,0), p\r=oo = 0 (3.53)
where Rea is defined by
Rea = (3.54)
V
Assume that the solutions of u and ~p can be expressed by the Stokes expansion in the form
M = M0 + ReaMi + 6>(Rea); "p = ~p0 + Re a ^ t + 6>(Rea) (3.55)
The zero-order Stokes approximation thus satisfies
V M0 = 0 (3.56)
V 2 M 0 - Vp 0 = 0 (3.57)
5o|r=i=wxr (3.58)
and the first-order Stokes approximation satisfies

V i = 0 (3.59)
2
V MI - VpY = MO VM0 (3.60)
ui\r=i=0 (3.61)
It is noted that the Stokes expansion is not uniformly valid in the neighborhood of infin-
ity. Therefore, another expansion, i.e., the Oseen expansion, is introduced to satisfy the
boundary conditions at infinity. By using the matching technique, the final results can be
obtained.
The Oseen approximation satisfies Eqs. (3.51) through (3.53) in stretched coordinates
V M* = 0 (3.62)
w* . VM* = - V p * + V2M* (3.63)
M*U Rea = ^ - u> x s; u*\5=O0 = (1, 0, 0), p*\s=oo = 0 (3.64)

Re a
where
M* = M, p* = J-i s = Rea r (3.65)

Rea
Similar to the Stokes expansion, the Oseen expansion is of the form
u* = ul + Rea u\ + 0(Rea), p* = p* + Rea p* + o(Rea) (3.66)
so that the zero-order Oseen approximation satisfies
V ul = 0 (3.67)
V2M* - V/7* = ul Vul (3.68)
The first-order Oseen approximation satisfies
V M* = 0 (3.70)
V2u\ - Vp\ = ul Vw* + u\ VM* (3.71)
=0 (3.72)
The matching principle requires that the Stokes expansion and the Oseen expansion must
be asymptotically equal in their common domain of validity. The Stokes expansion is valid
from the sphere out to some large distance, whereas the Oseen expansion is valid from
infinity to some small radii in the stretched variables. Hence, the common domain is a
spherical shell within which the unstretched radius is large but the stretched radius is small.
In this domain, the matching principle yields
olr=oo = Ul + 0(1), P0\r=oc = 0(1) (3.73)
Choose a particular solution for WQ and p$ as
S = (1,0,0), rf=0 (3.74)

On the basis of Eq. (3.74), u\ and p* can be solved as [Rubinow and Keller, 1961]
(3 75)
-TT -
where X is the x-component of s, and i is the unit vector in the .x-direction.

Using Eq. (3.73), o and ~p0 axe obtained by
1 \. 3x ( 1 1\ 1 _ 3x
and Mi and p t are given by
Ul
~ 32 [_ r r3 \r ^2 ^4 ^/J' 4 \r3 r4
3 T2 3 1 1 1
+ +
32[r~r2 r3~H: + ^
wsinyg {6 6 1 1
&z sin^ / 2 3 7 2\ wx(wxr)/ 1 1

16 \ r 3 r 4
r 5
r (
and
1 [ 9 7 3 1 9x 2/18 21 18 3 \1
Pi = T T
16 T + -T 7 - ; 6 =- +JC 2 ^ + r- 1
2r
coz sin
(3.78)
where x and z are the two components of the Cartesian coordinate, ft is the angle between
fl and MQ. The force per unit area exerted by the fluid on the surface of the sphere can be
evaluated from ~p and u as [Lamb, 1945]
r d
(3.79)
r dr
Thus, by integrating Eq. (3.79) over the entire surface of the sphere, the total force is
obtained as
F = -67ta/iUv 11 + -Re a j Rea u x Up + o Rea) (3.80)
The first term on the right-hand side of Eq. (3.80) is the total drag force in the opposite
direction of /p, including both Stokes drag and Oseen drag. The second term represents
a lift force in the direction perpendicular to Up. Thus, the lift force or Magnus force for a
spinning sphere in a uniform flow at low Reynolds numbers is obtained as
n
x uv (3.81)
It is noted that F M is independent of the viscosity \i and is similar to the Kutta-Joukowsky
lift formula for two-dimensional potential flow over an airfoil.
The ratio of the Magnus force to the Stokes drag can be expressed by
d
\ P
(3.82)
24 IJL
Therefore, the lift force due to the particle spin is negligibly small compared to the drag
force when the particle size is small or the spin velocity is low.
Figure 3.4. Forces on a rotating and moving sphere with wake separation.
3 3 I Interparticle Forces and Field Forces 101
At high Reynolds numbers, the rotation of the sphere yields an asymmetric wake, as
shown in Fig. 3.4. In this case, the theoretical analysis of the Magnus force and the drag
force becomes rather complex because of the difficulties in obtaining the expressions for the
pressure and velocity distributions around the surface of the sphere. Thus, the determination
of the lift force as well as the drag force relies mainly on the empirical approach.
3.3 Interparticle Forces and Field Forces

The motion of particles is affected by the short-range interparticle forces such as
the van der Waals force, electrostatic force, and collision force. It is also affected by the
long-range field forces such as the gravitational force, electric force, and magnetic force.
This section discusses the basic relationships which quantify these interparticle and field
forces.
3.3.1 Van der Waals Force

In 1873 van der Waals reported the nonideality of gases and attributed it to the
existence of molecular or atomic interactions due to interacting dipoles. The van der Waals
force is related to the dispersion effect, which is the interaction between the instantaneous
dipoles formed in the atoms by their orbiting electrons [London, 1937]. The very rapidly
changing dipole of one atom produces an electric field that acts upon the polarizability of
a neighboring atom. The induced dipole of the neighboring atom tends to move in phase
with the original dipole, producing an attractive atomic interaction known as the van der
Waals force. The van der Waals force, F v , can be expressed by
d
/\ = 4~ (3-83)
ds
where E is the energy of interaction and s is the separation distance between two atoms.
3.3.1.1 Van der Waals Forces: Microscopic Approach
For atoms in a vacuum, the induced dipole of the neighboring atom moves in phase with
the original dipole. Thus, the energy of interaction between two atoms i and j separated by
a distance s is obtained by [London, 1937]
E =- ^ (3.84)
sb
where Ay is London's constant, which is a function of the interacting atoms. This energy of
interaction is attractive provided the medium is not too optically dense such as in a vacuum
or gas. Equation (3.84) is only valid at a distance less than the wavelength corresponding
to the transition between the ground and the excited states of the atoms, i.e., the absorption
wavelength. At separations greater than the absorption wavelength, retardation effects
become important and the attractive interaction energy is inversely proportional to s 1.
Retardation effects result when the electromagnetic field has to travel farther at greater
separations. By the time the field reaches the neighboring atom, the original atomic dipole
has already changed its orientation. This effect leads to the interaction's being slightly out
of phase. The interaction energy is reduced but still attractive.
The van der Waals force exists not only between individual atoms and molecules but
also between solids. Equations for the van der Waals force between solids were deduced
by use of the additivity concept, which allowed the force to be calculated in terms of the
interaction between individual atoms in the solids [Hamaker, 1937]. The nonretarded energy
of interaction between two materials, 1 and 2, of volumes V\ and V2 and atom density q\
and q2 is given by
- dV2 dVi (3.85)

JV\ Jvv2 Ju
Thus, substituting Eq. (3.85) into Eq. (3.83) yields the van der Waals force between two
solids. For two spheres of radii a\ and a2, FY can be obtained as
(3.86)
OS"
Here s refers to the distance of separation between the two surfaces of the interacting
materials. In Eq. (3.86), a* is the relative radius defined by Eq. (2.66), and Ax2 is Hamaker's
constant (or coefficient), given by
A\2 = n qiq2k\2 (3.87)
Thus, for a sphere of radius ax and a plane surface where a2 approaches oo, Eq. (3.86)
reduces to
Fv = ^ 1 (3.88)
For two plane surfaces, the solutions of Eqs. (3.83) and (3.85) give a van der Waals
pressure, Py, or the van der Waals force per unit area of contact by
(3.89)
Equations (3.86) and (3.88) give some examples of the nonretarded van der Waals forces
for ideal contact geometries. For retarded interactions, the exponent for the distance of
separation increases by 1 with the change of the corresponding numerical coefficients.
The preceding theory, assuming complete additivity of forces between individual atoms, is
known as the microscopic approach to the van der Waals forces.
3.3.1.2 Van der Waals Forces: Macroscopic Approach

The concept of additivity is unsatisfactory when applied to closely packed atoms in a
condensed body. In this case, a new approach to the energy of interaction needs to be
developed or a modification of Hamaker's constant is desirable.
According to Maxwell's equations, spontaneous electromagnetic fluctuations change the
fluctuation field in the surrounding region, known as screening. A macroscopic theory of
the van der Waals forces relating the nonadditivity to the thermodynamic fluctuations in
the interior of a medium was developed by Lifshitz (1956). The macroscopic theory can
eliminate the additivity problem by regarding solids as a continuum and deriving the forces
in terms of bulk properties such as the dielectric constants and the refractive indices of the
materials.
For a material, Hamaker's constant is expressed by
A = -hm (3.90)
4TT
3.3 I Interparticle Forces and Field Forces 103
where hm is the Lifshitz-van der Waals constant, which depends only on the material and
is independent of geometry. Therefore, Eqs. (3.86), (3.88), and (3.89) can be applied, in
conjunction with Eq. (3.90), for various geometries in the evaluation of the van der Waals
force or pressure.
The Lifshitz-van der Waals constant for the interaction between two bodies of materials
1 and 2 separated by a medium 3 is given by
y (3.9D
where h is Planck's constant and 8[ is the dielectric constant of material i along the imaginary
frequency axis . This integral is a real function and decreases monotonously from the value
60, the static dielectric constant, when f = 0, to 1 when f = oo. Equation (3.91) holds for
two materials in a vacuum or gas by simply replacing s3 with 1.
In order to obtain the Lifshitz-van der Waals constant from Eq. (3.91), an expression for
(), related to the dielectric response, must be determined. The dielectric response of a
material is represented by the complete dielectric spectra or the complex dielectric perme-
ability, i.e., s (co) versus co. The real portion of the complex dielectric permeability measures
the transmission of electromagnetic radiation of frequency co through a body. The imagi-
nary portion measures the absorption of electromagnetic radiation of frequency co through
the same body. This absorption determines the dissipation of energy in an electromagnetic
wave propagated in the medium. Therefore, it is the macroscopic property of absorption
that determines the strength of the van der Waals interaction between solid bodies.
Solving Eq. (3.91) for the Lifshitz-van der Waals constant or Hamaker's constant usually
is not trivial. Many approximations for its calculation have been proposed for various cases
[Landau and Lifshitz, 1960; Hough and White, 1980; Visser, 1989; Israelachvili, 1992].
For materials 1 and 2 interacting through material 3, Hamaker's constant becomes
Al32 = c(An + A33 - Al3 - A23) (3.92)
where c is a function of the materials interaction with an average value of about 1.7.
3.3.1.3 Van der Waals Forces: Surface Effects

Surface deformation increases the van der Waals forces as a result of the increasing area of
contact [Krupp, 1967; Dahneke, 1972]. Thus, for deformable materials, the equations of
the van der Waals force, which are valid for rigid materials, should be modified to account
for the effect of the increased area of contact. Surface absorption also affects the van der
Waals forces. The thickness of the adsorbed layer may increase the distance of separation
between the interacting materials, thereby decreasing the van der Waals force [Krupp, 1967].
Moreover, if the thickness of the adsorbed layer is greater than the separation distance, the
dielectric properties of the adsorbed layer will dominate the base material in the van der
Waals interaction [Langbein, 1969].
3.3.2 Electrostatic Force

One of the important characteristics of gas-solid multiphase flows is concerned
with the electrostatic effect. Particles can be charged by surface contact in a collision,
by corona charging and scattering in an ionized gas, by thermionic emission in a high-
temperature environment, and by other charging mechanisms such as colloidal propulsion
for dielectric materials in an intense electric field [Soo, 1990]. The motion of a charged
particle in a gas-solid flow is affected by the electrostatic force imposed by nearby charged
particles.
Although the origin or basic mechanism of static electrification has not been fully un-
derstood, the well-known Coulomb's law, which quantifies the electrostatic force, was
unraveled as early as two centuries ago. Coulomb's law states that for two charged objects
that are much smaller than the distance between them, the force between them is propor-
tional to the product of the charges and inversely proportional to the square of the distance
of separation. The force is acting along a straight line from one charged object to the other.
Thus, the electrostatic force can be expressed by
(3.93)
where qx is the charge carried by object i\r is the distance between the two objects; and e
is the permittivity of the surrounding medium.
In a gas-solid suspension system, let nq be the charge density and V be the effective
domain. The electrostatic force of a particle carrying a charge q in the direction of n can
be expressed by
(r n) dV (3.94)
4TT6 JV r3
The origin of solid electrification and the mechanism of charge transfer are elaborated
on in 3.5.
3.3.3 Collisional Force

The effect of the collisional force due to the impact of particles should be included
when accounting for the motion of a particle except in a very dilute gas-solid flow situation.
Basic mechanisms of collision between two particles or between a particle and a solid
wall are discussed in Chapter 2. The collisional force between a particle and a group of
neighboring particles in a shear suspension is discussed in 5.3.4.3. In a very dense system
where particle collisions dominate the flow behavior, collisional forces can be described
by using kinetic theory, as detailed in 5.5. The key equations derived in other chapters
pertaining to the collisional forces can be summarized in the following.
For elastic spheres, the maximum collisional force in a collinear impact between two
particles, F c (= fm in Eq. (2.132)), is given by
(2.132)
where U (= Un in Eq. (2.132)) is the relative velocity, m* is the relative mass defined by
Eq. (2.128), a* is the relative radius, and E* is the contact modulus defined by Eq. (2.66).
The averaged collisional stress between a particle and a group of neighboring particles
of the same diameter in a shear flow can be expressed by (/x p A p ), where fip is given by
^ ^ ( A p : A p )i (5.220)
and Ap is evaluated by Eq. (5.217).

3.3 I Interparticle Forces and Field Forces 105
For a collision-dominated dense suspension, the collisional stress tensor Pc is given by

6k
(5.310)
where k is the thermal conductivity of collisional flux of fluctuation energy and is given by
Eq. (5.309); D is a tensor defined by Eq. (5.311); Tc is the granular temperature; and the
symbol tr denotes the trace.
3.3.4 Field Forces

The field forces, also known as the body forces, are long-range forces exerted by
various fields outside the flow system. Typical field forces in a gas-solid system include
the gravitational force, the electric force, and the magnetic force.
When an external electric field is applied to a gas-solid flow, the charged particles are
subjected to an electric force, F E , which is expressed by
FE = qE (3.95)
where q is the charge carried by a particle and E is the electric field intensity. Driven by this
force, particles carrying positive charges tend to move to the cathodic side while negatively
charged particles move to the anodic side. Electrostatic precipitation is an example of the
effect of the electric force.
If particles are magnetism-sensitive, they are subjected to a magnetic force once they are
exposed to an electromagnetic field in a gas-solid flow. The magnetic force, F m , is given
by
Fm = mfJLTBo (1.73)
where B$ is the magnetic flux density in a vacuum; m is the number of north magnetic
poles; and \xx is the relative permeability of the material. Most magnetic materials are highly
conductive. Thus, the electrostatic effect in magnetic materials can usually be neglected.
One example involving the magnetic force is magnetic fluidization (Problem 9.2).
In general, the forces and moments acting on a charged particle in an electromagnetic
field comprise those due to the net charge, electric dipole (permanent or induced dipoles) in
the electric field, and magnetic dipole in the induced magnetic field [Soo, 1964]. Neglecting
the effects of magnetic dipoles, the force, known as Lorentz force, on a charged particle
moving in an electromagnetic field is expressed by
Fem = q(E + UpxB) + V(pd E) (3.96)
where pd is the dipole moment, given for dielectric materials, by
where is permittivity of the material and er is relative permittivity or dielectric coefficient

of the dielectric material. Application of Eq. (3.97) to a dilute suspension of charged
particles is illustrated in the following example [Soo, 1964].
Example 3.2 Consider a large number of uniformly charged solid particles initially kept
in a spherical barrier of radius R$ with a symmetric density distribution. When the barrier
is suddenly removed, the particles start to emerge from that spherical domain. The viscous
drag in the gas is assumed to be negligible. Find the ratio of the force due to dipole to that
due to electrostatic repulsion and show that for dilute suspensions, the dipole effect due
to self-field is negligible. Also discuss the spreading of the solid particles in this simple
symmetric system.
Solution This is a one-dimensional problem. The equation of motion of a single

particle in the /^-direction is obtained from Eq. (3.96) as
at dR
where m is the mass of the particle. From Eq. (3.94), E can be given by
1 f 2 ($ I Wl)
E = / 47ZT nQq dr = MR (E3.11)
AneR2 JQ AneR2
where
MR= / 4nr2ppavdr (E3.12)
Jo
Expressing Uv by
V,-f (E3.I3)
dt
and substituting Eq. (3.97) into Eq. (E3.10) give rise to
p (E3.14)
dR \mj ATTR2 pve{ex + 2)\m) (4n)2R5
where Mo = MR(R0) at t = 0, where Uv = 0. It is seen that the dipole effect is due to the
gradient of electric field intensity, which, in this case, produces a deceleration. The ratio of
the force due to dipole F di to that due to electrostatic repulsion F er is thus obtained by
î 12fe-l)J^L
Hence, for uniform distribution of solids within R = RQ, Eq. (E3.15) reduces to
= 4 (E 3.l6)
Kr fe + 2) ap
Thus, for a dilute suspension of charged particles, the dipole effect due to self-field is
negligible compared to that due to the electrostatic repulsion.
With the dipole force neglected, the direct integration of Eq. (E3.14) yields the particle
velocity distribution as
AM (E3,7)
M0(q/m)2
3 A I Motion of a Single Particle 107
Using Eq. (E3.13), Eq. (E3.17) can be integrated with the initial condition of R = RQ at
t = 0 to yield
/",(/")' = *_(*_ A + J *_ + JL-A (E3 .18)
Thus, Eqs. (E3.17) and (E3.18) determine both the position and the velocity at a given
time of each particle initially at radius RQ. In addition, each particle reaches an asymptotic
velocity as R approaches infinity.
3.4 Motion of a Single Particle

The most convenient mathematical description of the motion of a particle in a
flow field can be given using the Lagrangian coordinate which has its origin attached to
the center of the moving particle. For the linear motion of a particle at very low particle
Reynolds numbers, the particle motion is governed by the BBO equation, originated by
Basset (1888), Boussinesq (1903), and Oseen (1927). To describe the general motion
of particles subjected to various forces, the governing equation may be formulated by
superposition of all applicable forces. This general formulation for particle motion forms
the basis of the Lagrangian trajectory modeling for gas-solid flows, which is presented in
this section. Examples of the application of the Lagrangian trajectory model in describing
the particle motion in rotating and oscillating flow fields are also presented.
3.4.1 Basset, Boussinesq, and Oseen (BBO) Equation

The BBO equation delineates a sphere in linear motion at low Reynolds numbers
[Basset, 1888; Boussinesq, 1903; Oseen, 1927]. For a spherical particle, the BBO equation,
based upon Eq. (3.37) with the replacement of the buoyancy force with the pressure gradient
force, can be expressed as [Soo, 1990]
^pÛ - Up)
(3.98)
where d/dt is the substantial derivative following the particle flow. The five terms on the
right-hand side of Eq. (3.98) are, in order from left to right, the forces due to the Stokes
drag, pressure gradient, added mass, Basset historic integral, and other external forces such
as gravitational force and electrostatic force.
It is suggested that the pressure gradient may be estimated from the Navier-Stokes
equation of a single-phase fluid by [Corrsin and Lumley, 1956]
-Vp = p- fjiV2U (3.99)
where D/Dt is the substantial derivative following the gasflow(see Eq. (5.7)). Substitution
of Eq. (3.99) into Eq. (3.98) yields

dUR = _3p
dt
(3.100)
In gas-solid flows well beyond the Stokes regime, the effect of convective acceleration
of the gas surrounding the particle is important. To incorporate this effect into the preceding
formulation, modifications of the expressions for the Stokes drag, carried mass, and Basset
force in the BBO equation are necessary [Odar and Hamilton, 1964]. The modified BBO
equation takes the form [Hansell et al, 1992]
S_^| P _ 0 ,| ( 0 _ v + c I
l
At 4dp pp 2 p p At
where CD, CI, and CB are the drag, carried mass, and Basset coefficients, respectively. In
general, Ci and CB are functions of the particle Reynolds number, Re p, defined by Eq. (3.2)
and the acceleration number, An, defined by [Faeth, 1983]
d(V-Vv)
(3.102)
For a simple harmonic motion at Re p < 62, correlations of C\ and CB are suggested as
[Odar and Hamilton, 1964]
0.132 An* O52An3
(1+0.12 An2) (1+An)3
3.4.2 General Equation of Motion

For general particle motion formulation, additional forces such as the Saffman
force, Magnus force, and electrostatic force should be included. Assuming that all forces
applied on the moving particle are additive, the equation of motion of a particle in an
arbitrary flow can be expressed by
m-^- = F D + F A + F B + F S + * M + ^ C + (3.104)
d
where Fn is the drag force; F& is the added (or carried) mass force; FQ is the Basset force; Fs
is the Saffman force; FM is the Magnus force; F c is the collisional force; and " " denotes
other forces such as the electrostatic force, the van der Waals force, the gravitational force,
the magnetic force, and forces associated with various field gradients due to shock wave
propagation, electrophoresis, thermophoresis, and photophoresis phenomena.
It should be noted that the forces in Eq. (3.104) are not generally linearly additive. The
drag force, Basset force, Saffman force, and Magnus force all depend on the same flow
3.4 / Motion of a Single Particle 109
field. The inertia term in the Navier-Stokes equation, V (UU), is not generally negligible,
so the equation is highly nonlinear. Thus, linear superposition of theflowfieldfor each flow
mode becomes undesirable because of the nonlinearity of the equation. However, when
the governing equations of fluids are linear, the forces on the solids may be additive; this
relation has been partially proved by Passman (1986), using one type of the mixture theory
which allows for separated phases.
Example 3.3 Consider a rotating gas flow in a cylindrical chamber with a small particle
injected into the flow. Assume that the gas rotates as a rigid body with a constant angular
velocity co and the only driving force is the Stokes drag [Kriebel, 1961]. Initially, the relative
particle velocity is normal to the flow. Develop the equations for the particle trajectory in
this rotating flow and discuss the effect of particle sizes on the trajectory.
Solution Select a Cartesian coordinate system with the origin on the axis of
rotation and the JC-axis coinciding with the line connecting the origin and the initial location
of the particle, as shown in Fig. E3.2. The velocity of gas is given by
U = cor sin 0 = coy
(E3.19)
V = cor cos cp = cox
y *
x
cN Initial location of x
particle and inlet
port
Figure E3.2. Coordinate system for particle motion in a rotating gas.
From Eq. (3.98), we have

dUp d 2x 1 fdx
(E3.20)
d2v _ 1 / dv
dT d^ ~ A d7
where rs is the Stokes relaxation time given by Eq. (3.39). Replacing d/dt by co(d/dcp),
110 3 /Momentum Transfer and Charge Transfer
Eq. (E3.20) becomes

dAx 1 dx
= 0
(E3.2!)
d2y 1 dv X
d 0 2 "f A d0 ""A
where A is defined by
A = COTS (E3.22)
The initial conditions can be expressed by
dx (E3.23)
= x0
0=0 co 0=0
Using the Laplace transformation to solve Eq. (E3.21) with the initial conditions given by
Eq. (E3.23), the particle trajectory can be obtained by
a(\ + 2AB)-
x = cosh(aV0 sinh(aVO )
(E3.24)
y=
J
b \
x sin 1
si (E3.25)
2? )
where
f =
(E326)
2 1 2 2
c = a + fc = VT+16A
The trajectories for different values of A and B are illustrated in Fig. E3.3. As expected,
these trajectories are spirals which start from the injection point and end at the cylinder
wall. The angular position of the particle lags behind the rotating gas by an angle 0, which
is given by
l
0 = cot tan (E3.27)
For very small particles, A <^C 1. Equations (E3.24) and (E3.25) are reduced to
x ^ xo(l + AB)eA* cos(l - 2A 2 )0
y JCO(1 + AB)eA* sin(l - 2A 2 )0
For large particles, A 1. The particle trajectory is given by
y ^ xo<t>
3.5/Charge Generation and Charge Transfer 111
A = oo
A
A = 0.25
A B= 2
O B= 1
5=
G) = 7,200 rpm
pp = 3,000 kg/m 3
ji=1.81xlO' 5 kg/m-s
Figure E3.3. Calculated particles' trajectories (from Kriebel, 1961).
3.5 Charge Generation and Charge Transfer

Charge transfer may occur during the contact between particles or between particles
and the solid surface. A gas-solidflowfieldcan be significantly affected by the electrostatic
forces induced by the charges carried by the particles. In this section, the origins of particle
charges due to electrification involving surface contact andfieldcharging are discussed. The
mechanism of charge transfer during the collision of two particles is described. In terms of
this mechanism, the electrostatic ball probe theory, which illustrates the interrelationship
between the particle mass flux and the charging current on the ball probe in a gas-solid
flow, is presented.
3.5.1 Static Electrification of Solids

The phenomena of the electrification of solids are complex. In gas-solid flows,
surface contact by collisions, ion collection, and thermionic emission are known to be the
major modes of particle electrification. Details of these three charging modes are introduced
in the discussion that follows.
3.5.1.1 Charging by Surface Contact

The charge equilibrium between two objects in contact requires the Fermi energy level of the
objects to be equalized. When two surfaces of dissimilar materials are brought into contact,
charge transfer may occur through the contact surface as a result of the difference in the
initial Fermi energy levels of these materials. Thus, the particles become electrostatically
charged after their separation from the contact. One particle is positively charged because
of the loss of electrons while the other is negatively charged because of the gain of electrons.
+ Silk _
Figure 3.5. Circular-charging relationship of glass, zinc, and silk.
This electrostatic charging is commonly known as triboelectric charging. The word tribo,
meaning "to rub," is from Greek. In fact, the triboelectric charging does not require any
rubbing processes other than surface contact. Normally, rubbing could result in increased
charge transfer due to an increased contact area.
Most solid materials can be arranged into a triboelectric series based on charge polarity in
a simple contact and separation process. Several triboelectric series are given in Table 3.1.
Although a triboelectric series can predict the charge polarity with reasonable accuracy,
exceptions are not uncommon because of many influential factors, such as surface finish,
preconditioning, and contamination. An overlapping circular polar relationship shown in
Fig. 3.5 gives an example of these exceptions [McAteer, 1990].
It is interesting to estimate the percentage of atoms on a solid surface involved in a
charge transfer process during a simple contact and separation [Cross, 1987]. A maximum
accumulated charge on a surface is limited by the breakdown electric field strength of the
surrounding medium. At atmospheric pressure with air as the medium, this breakdown
strength is about 3 x 106 V/m. From Gauss's law, the electric field strength due to an
assembly of charges can be estimated.
Gauss's law states that if an imaginary enclosed surface of area S is drawn around a
uniform density of charge, the component of the electric field at the surface, which is
perpendicular to the surface, is proportional to the total charge enclosed. Hence, Gauss's
law is expressed as
E= ^ ^ (3.105)
where ^2 Q{ is the sum of the charges enclosed by the surface. At the maximum or
breakdown electric field strength in air, the maximum charge density on the surface is
tfmax = ^ ^ = max60 = 2.6 X 1 0 " 5 - ^ (3.106)

o mz
Since the charge carried by a single electron is 1.6 x 10~19 C, the maximum number density
of electrons on the charged surface becomes 1.6 x 1014 m~2. The atomic density on a solid
surface is approximately 2 x 1019m~2. It can therefore be estimated that fewer than lOatoms
per million atoms on the surface are involved in the charge transfer process. Such a small
percentage of surface atoms participating in the charge transfer indicates the sensitivity of
the effect of the surface impurity and contamination on the amount of charges transferred.
3.5/Charge Generation and Charge Transfer 113
Table 3.1. Triboelectric Series
Material Polymer type Source
(a) Montgomery (1959;)

+ Charge Wool
t Nylon
Viscose Cellulose
Cotton
Silk
Acetate rayon Cellulose
Lucite or Perspex Polymethyl methacrylate
Polyvinyl alcohol
Dacro Copolyester of ethylene
glycol and
terephthalic acid
Orion Polyacrylonitrile
PVC
Dynel Copolymer acrylonitrile/
vinyl chloride
Velon Copolymer vinylidene
chloride/vinyl chloride
Polyethylene
Charge Teflon PTFE
(b) Webers (1963)
+ Charge Polyox Polyethylene oxide Union Carbide
t Polyethylene amine Chemirad
Gelatin
Vinac Polyninyl acetate Colton Chemical
Lucite 44 Polybutyl methacrylate Du Pont
Lucite 42 Polymethyl methacrylate Du Pont
AcryloidAlOl Polymethyl methacrylate Rohm and Haas
Zelec DX Polycation Du Pont
Polyacrylamide Cyanamid
Cellulose Eastman
Acysol Polyacrylic acid Rohm and Haas
Carbopol Polyacid BF Goodrich
Polyethylene
Polyvinyl butyral
Charge Polyethylene Du Pont
(c) Williams (1976)
+ Charge Lucite 2041 Methyl methacrylate Du Pont
t Dapon Daillyl phthalate
Lexan 105 Poly-bisphenol-A-carbonate GE
Formvar Polyvinylformal Monsanto
Estane Polyurethane Goodrich
Du Pont 49000 Polyester Du Pont
Durez Phenol formaldehyde Durez
Ethocel 10 Ethyl cellulose Hercules
Polystyrene 8X Polystyrene Kopper
Epolene C Polyethylene Eastman
Polysulphone P-3500 A diphenyl sulphone Union Carbide
Hypalon 30 Chorosulphonated PE Du Pont
CyclolacH-1000 Acrylonitrile-butadiene- Borg Warner
styrene terpolymer
{continued on p. 114)
Table 3.1 (cont.)
Material Polymer type Source
Uncoated iron
Cellulose acetate butyrate
Epon 828/V125 Epoxy amine curing agent Shell/General
Polysulphone P-1700 Union Carbide
| Cellulose nitrate
Charge Kynar Polyvinyldene fluoride Penwalt
Source: J. A. Cross' Electrostatics: Principles, Problems and Applications, AdamHilger, 1987.
To understand the basic mechanism of charge transfer, a brief introduction of various

electron energy bands in a solid is in order. Each electron in a single isolated atom moves
along a given orbit surrounding the nucleus of the atom. Each orbit represents a specific
electron energy level. An electron can jump across certain energy levels by gaining or losing
energy. In a solid particle, atoms interact with each other and, thus, the energy levels of
atoms are influenced by neighboring atoms. In light of this interaction, the energy levels can
be defined in terms of energy bands. The inner electrons are in narrow bands, the so-called
core bands. All the valence electrons stay in the valence bands. The free electrons are in the
conduction bands. An area of forbidden energy can lie between energy bands, the forbidden
energy band. The Fermi energy level, Ef, is defined as the highest filled energy level at an
absolute temperature of zero degree. At an absolute temperature above zero degree, the
thermal energy allows some electrons to stay at the energy level above the Fermi energy.
The probability of an electron state of energy in a metal can be described by the Fermi-Dirac
distribution function as
f(E)= [l+exp^^J (3.107)

where E is the energy level, k is the Boltzmann constant, and T is the absolute tempera-
ture. Unless the temperature is very high, the Fermi energy can be used as a reasonable
approximation of the energy of electrons to be freed for most ambient conditions.
The work function, W, is defined as the work required to remove an electron out of a
metal material from the top of the energy distribution. The work function is related to the
potential function 0, known as the thermionic work function, by
W = ecj) (3.108)
where e is the charge of an electron.

An intrinsic potential difference exists between dissimilar dry metals in contact. This
potential difference, known as the Volta potential or the contact potential, can be related to
the work functions of the two dissimilar materials A and B as
VA-VB= WA WB
~ = 0A _ 0B (3.109)
e
The contact potential reflects the work needed to transfer an electron from one solid to
the other. Thus, when two solids are brought into contact, the charge transfer results from
3.5 / Charge Generation and Charge Transfer 115
the potential difference, which is related to the difference in the Fermi energy level. Upon
separation of the two bodies, although some electrons may escape back as a result of the
tunnel effect, a net charge results.
The actual mechanism for charging by surface contact is complicated, and many causes
remain unresolved. However, it is clear that electron and/or ion transfers are principal factors
responsible for the contact charging. More information is available in Montgomery (1959),
Harper (1967), and Cross (1987).
3.5.1.2 Charging by Ion Collection

Various electric effects in multiphase flows have been summarized by Soo (1990). Typically,
particles can be electrified in an electric field via three distinct charging mechanisms: field
charging, diffusion charging, and corona charging. Field charging becomes predominant
for particles larger than 1 /xm where random motion of ions is insignificant and, thus, the
diffusion charging mechanism can be ignored. For particles smaller than 0.2 /xm, diffusion
charging must be accounted for and the contribution of an external electric field becomes
insignificant. When the electric field strength exceeds a certain limit, ionization of the
surrounding gaseous medium occurs. By the bombardments of the ions and electrons, the
solids are charged. Such a charging process is generally known as corona charging.
In the following discussion on these charging mechanisms, three basic assumptions are
imposed: i.e., all particles are spherical; particles of the same size are charged to the same
degree; and interacting effects between particles are negligible.
A. FIELD CHARGING
When a conducting particle of diameter dv is exposed to a uniform electric field E in an
air medium or in a vacuum with a charge density distribution qo, the electric potential V is
governed by the Poisson equation given by
V2y = - ^ (3.110)
where o is the particle permittivity in a vacuum. The relation between the electric field
strength, E, and the electric potential is given as
E = -WV (3.111)
With the particle surface charge of Qs, it can be derived that the electric field strength at
any point on the sphere, s , satisfies
% (3.112)
where 0 is the angle between a point on the sphere and the electric field. The second term on
the right-hand side of Eq. (3.112) represents the contribution of the particle surface charge
based on Gauss's law. Thus, the total charge current, /, entering the particle can be obtained
by [Oglesby and Nichols, 1978]
i = -=^ = / q0KEs27tazsin0 d<9 (3.113)

dt J
where K is the mobility of ions and 0m is the limiting angle for the charge flow to the
particle. 0m is determined from Eq. (3.112) by setting Es equal to zero, i.e.,
6s
cosOm= (3.114)
37td*E
Thus, from Eqs. (3.112) through (3.114), the equation for the charging rate of a single
particle in an electric field with unipolar ion density can be written as
_Q1
where g s m is the saturation charge, which is given by

Gsm = 37T6o'o^p (3.116)
The total charge acquired by an initially neutral particle within the charging time t is given
by
/ t \
(3.117)
where r is the time constant of the saturation charge, expressed by
^ l (3.118)
For dielectric particles, the saturation charge can be given as
where er is the relative permittivity of the dielectric materials.
B. DIFFUSION CHARGING
When a particle is placed in a field of unipolar ions with a charge density of qo and an
electric field strength of EQ, the particle will be charged by ion striking as a result of the
random motion of ions. The rate of diffusion charging of particles can be derived in terms
of the kinetic theory of gases. Assuming that every ion striking the particle is captured, the
charging rate of a particle in the absence of an external electric field is expressed by [White,
1963]
An IT / n * \
(3.120)
where k is the Boltzmann constant; T is the absolute temperature; and v is the mean thermal
speed of the ions, which is expressed by
v=\/ (3.121)
V nm
where m is the mass of an ion. Hence, the charges acquired by the initially neutral particle
within the charging time t can be obtained from Eq. (3.120) as
(3,22)
where x& is the time constant of the diffusion charging, expressed by
(3.123)
The diffusion charging of particles in the presence of an external electric field is available
in Murphy (1956) and Liu and Yeh (1968).
C. CHARGING IN AN IONIZED GAS

Consider an ideal case where a group of particles are uniformly suspended in an ionized
gas. Assume that all the charges carried by ions and electrons arriving at the particle
surface are completely attached to the particle surface. As the electrons have higher random
velocities than the ions do because of the smaller mass of the electron, the particles would be
negatively charged. Because of the complexity involved in treating the electric interaction
among particles, we only consider the cases where the interparticle spacing is larger than
the Debye shielding distance, A,D, given by
(3.124)
where #eo is the electron charge density. This case limits the analysis given late, making it
to be applicable only when the particle number density n is less than (2A,D)~3.
From the kinetic theory, the charging rate by electrons can be estimated by [Soo, 1990]
dt
where a is the electrothermal number, defined as the ratio of the particle electrostatic energy
to the thermal energy, and is expressed as
a = (3.126)
27t0dpkT
a is negative for electron charging and positive for ion charging. Similarly, the charging
rate by ions is
(3127)
dt 4
At equilibrium, the charging rate by ions equals that by electrons. Thus, we have
(3 128)
m
qi0 V
For any volume element, we have the charge balance as
Qsn = qio-qeo (3.129)
Thus, the saturated surface charge of each particle is given as
(3.130)
Example 3.4 Particles of 10 /xm size flowing at a velocity of 20 m/s collect charges from
unipolar ions in a corona discharge over a distance of 5 cm. Assume that the ions can be
regarded as electrons. The average ion number density, ne, in the corona discharge field
is 1016 m~3. The particle density is 1,400 kg/m3. Estimate the charge-to-mass ratio ac-
quired. Assume that the interparticle effects can be neglected. In addition, the simultaneous
charging of positive ions in the corona is negligible. Initially, the particles are neutral. The
environment is at 300 K.
Solution The charging rate of a single particle by electrons can be expressed by

Eqs. (3.125) and (3.126). In this case, the particles are only charged by electrons, which can
be regarded as negative ions. Thus, Qs = Q es. From Eqs. (3.125) and (3.126), we have
= Ae-BQ* (E3.30)
dt
where
7td n e / RkT
A = -^J (E3.31)
4 y 7tme
and
B= * (E3.32)
2n0dpkT
With the initial condition of
2eslr=0=0 (E3.33)
Equation (E3.30) yields
(E3.34)
Thus, for a particle passing through a corona over a distance of / at a velocity of Up, the
charge-to-mass ratio can be expressed by
^--4-^ln(l + ABl) (E3.35)

Substituting all the given data into Eq. (E3.35) gives
dpnee2l I 2
m edZPo \ 4eot/0 V nmekT
12 x 8.85 x 10- 12 x 1.38 x 10"23 x 300
1.6 x 10"19 x (10" 5 ) 2 x 1400
10"5 x 1016 x (1.6 x 10" 19 ) 2 x 0.05
x In 1 +
4 x 8.85 x 10"12 x 20
= 0.00029 (E3.36)
n x 9.1 x 10"31 x 1.38 x 10"23 x 300/ ' kg
3.5.1.3 Charging by Thermionic Emission

When solid particles are exposed to a high-temperature environment, typically for T >
1,000 K, charging by thermal electrification becomes important. The electrons inside the
solid can acquire the energy from the high-temperature field and be freed by overcoming
the energy barrier or the work function. By losing electrons in such a thermionic emission
process, the particles are thermally electrified.
From the Richardson-Dushman equation, the current density / e by thermionic emission
can be obtained as
J (3.131)
where (p is the work function of the solid and Ao is an experimentally determined coefficient.
It is noted that the rate of electrification is not constant. Once the tendency is established
for an electron to escape from the solid particle by thermionic emission, the charge buildup
occurs on the particle, which then attempts to recapture the to-be-freed electron by the
attracting Coulomb force. Therefore, the equilibrium of thermal electrification of solid
particles in afinitespace is possible. Details on the equilibrium and the rate of electrification
concerning the thermionic emission are available in Soo (1990).
3.5.2 Charge Transfer by Collision

Charge transfer occurs when particles collide with each other or with a solid wall.
For monodispersed dilute suspensions of gas-solid flows, Cheng and Soo (1970) presented
a simple model for the charge transfer in a single scattering collision between two elastic
particles. They developed an electrostatic theory based on this mechanism, to illustrate the
interrelationship between the charging current on a ball probe and the particle mass flux in
a dilute gas-solid suspension. This electrostatic ball probe theory was modified to account
for the multiple scattering effect in a dense particle suspension [Zhu and Soo, 1992].
3.5.2.1 Charge Transfer by Single Impact

Consider the charge transfer between two spherical particles of diameters d v\ and dp2. A
direct analogy between the charge transfer by collisions and the heat transfer by convection
appears to be in order. Thus, the current density through the contact area of these two
particles, / e 2 i, is expressed as
Jc2\ = hM(V2 - Vi) = -^-(Vc ~ Vi) = -p-(V 2 - Vc) (3.132)
dpi dp2
where Aei2 is the charge transfer coefficient through the contact area A12; V\, V2, and V c
are the electric potentials of particles 1 and 2 and the contact surface, respectively; and 01
and 02 are the electrical conductivities of the materials. By eliminating the contact potential
yc from the preceding equation, the charge transfer coefficient can be obtained as
axGl
hel2 = (3.133)
o\ dp2 + d
For each isolated particle, the charging current can be related to the particle capacity. Thus,
we have
^ ^ V l) (3.134)
where C\ and C2 are the capacitances of the particles.

For a homogeneous spherical particle, the capacitance is expressed by

C = 2ndp (3.135)
where e is the permittivity of the material. The electric potential of the particle is
V = $ + ? (3.136)
2d
where 0 is the work function and q is the charge carried by the particle. Therefore, the
equation for the charge transfer can be expressed as
V2_Vi) (3
( V 2 V i )
dt C*
where C* is the relative capacitance, defined by
<3138)
From Eq. (2.74), the relationship between the contact radius rc and the approaching
distance a for elastic collision of particles can be given by
rl = aa* (3.139)
Thus, the contact area A12 is given as
Au = izrl = naa* (3.140)
The charge transfer equation becomes
d[ln(V2 - VQ] na*
A t C
From Eq. (2.135), we obtain the relation of the contact time t and the approaching distance
a as
2
dt = [Ul - V ^ a H da (3.142)
\ 15m* /
Let x = a/am and define am as
The rearranged form of the charge transfer equation is obtained as
-d[ln(V 2 - V,)] = " ^ a * ? ; 1 2 J c ( l - s i ) - * dx (3.144)
Integrating the preceding equation for a complete elastic impact process including the
compression and the rebounding processes, we have
(V20 - V2) - (V10 " Vi) - (V20 - V10)(l - e~B) (3.145)

where subscript "0" denotes the initial state and B is a constant given by
{l_xl)-idx ( 3 . 1 4 6)
The transferred charge 212 relates to the loss or gain of the colliding particles as given by
Qu = Q20 - Qi = Qx - G10 (3.147)
Thus, gi2 can be evaluated from
02 - 0i + (1 - e ) (3.148)
c2 C\ j
3.5.2.2 Theory of Electrostatic Ball Probe
Consider a ball probe of a conductive sphere such as a stainless steel ball bearing which is
connected to an electrometer. When the ball probe is placed in a flow system, the electric
current induced by charge transfer between the ball and the flowing charged particles can be
measured. The amount of current measured will vary with the particle velocity and particle
mass flux.
Let us first consider the case where only single scattering is responsible for the charge
transfer. The electric current, ib, measured by the ball probe is proportional to the product
of the collision frequency, / c , and the averaged charge transfer per impact, Qbp, as given
by
h = /cC*(Vp0 - VM)(1 - e~B) (3.149)
where B is
g = 2.03aêbp (3.150)
Here, subscripts b and p refer to ball probe and particle, respectively. The collision frequency
is expressed by
fc = nUp^(dp + dh)2 (3.151)
where n is the particle number density and Uv is the particle velocity. For B <^ 1, from
Eqs. (3.149) and (3.151) and the Taylor expansion, ib becomes
( m* \ 5 3 8
J a^nU'5 (3.152)
Note that the preceding equation is for ideal cases, in which the particles are monodis-
persed, spherical, and totally elastic, and the contact surface is clean. In practice, the
particles are usually nonspherical and polydispersed; the collision could have involved
some heat loss, plastic deformation, or even breakup; and the contact surface may have
impurities or contaminants. In these cases, a correction factor rj is introduced to account
for the effects of these nonideal factors. The applicable form of the electric current through
the ball probe is, thus, given by
ib = AU^/5Jp (3.153)
where Jp (= a p p p Up) is the particle massfluxand A is a constant related to the characteristic
properties of the ball probe and particles. A can be expressed by
A = ^ ( 4 + 4>)2(VPo - Vbo)*ebp(^) V 1 (3.154)

122 3 / Momentum Transfer and Charge Transfer
The theory given is applicable to a very dilute gas-solid suspension where the charge
transfer is dominated by single scattering. When applied to concentrated gas-solid suspen-
sions, the effects of multiple scattering, sliding, and charge saturation become significant
and should be incorporated in the formulation of the theory.
Consider a monodispersed dense suspension. From the kinetic theory analogy, the
number of simultaneous impacts by particles on a ball probe is given by
^ (3.155)
where X& is the mean free path of particle-to-ball probe collisions. AB can be expressed as
*B = ,,4 , (3.156)
The flow conditions ranging from single scattering to "boundary layer flow" of particles
are shown in Fig. 3.6. It is noted that the single scattering is bounded by the condition of
db/Xft = 1. In terms of Eq. (3.156), the boundary of the single scattering is a function of
the particle size, ball probe size, and particle concentration.
The effect of multiple scattering, sliding, and charge saturation on the ball probe current
can be taken into account in a theory via the modification of the electric potential difference
between the ball and each single colliding particle. Assume that the initial potential of a
charged particle is the same for both single and multiple scattering. For the ball probe, we
have
(3.157)
0.354
100
100
O
.'3 io
10
1.0 1.0
0.1 ii 0.1
10 4
IO" 10" 10 10' 1.0
Volume fraction of solid particles
Figure 3.6. Particle flow regimes over an electrostatic probe (after Zhu and Soo, 1992).
Nomenclature 123
Integration of the preceding equation yields
yb-yp==(ybs_yps)exp (3.158)
where subscript s denotes the single scattering and t can be treated as a time scale for the
multiple scattering effect. It is desired to relate the time scale to the particle mass flux and
the particle velocity so that the potential difference in dense suspensions is expressed as
(3.159)
where Cc is a dimensional concentration factor with a unit of m4/kg-s, which is determined

experimentally. Thus, the ball probe current for dense suspensions takes the form
(3.160)
The preceding equation has been experimentally validated [Zhu and Soo, 1992]. By
defining y as k/U^6 and x as Jp/U^, the equation for the ball probe current can be
arranged in a simple form as
y = Axe~CcX (3.161)
Using the condition of dy/dx = 0, the upper limit of the particle mass flux that yields the
univalent correspondence between the particle mass flux and the ball probe current can be
obtained as given by:
(3.162)
C
Nomenclature
A Exposed frontal area of a particle B Parameter, defined by Eq. (E3.26)
to the direction of an incoming B Constant, defined by Eq. (3.146)
flow Magnetic flux density in a vacuum
Bo
A Hamaker's constant c Electric capacitance of a particle
A Empirical coefficient, defined by c Empirical constant, defined by
Eq. (3.3)
Eq. (E3.5)
A Parameter, defined by Eq. (E3.22)
cb Electric capacitance of a ball probe
A Parameter, defined by Eq. (3.154)
Cc Concentration factor, defined by
Ao Constant, defined by Eq. (3.131)
Eq. (3.159)
An Acceleration number Ci Carried mass coefficient
a Particle radius CB Basset coefficient
a* Relative radius, defined by CD Drag coefficient
Eq. (2.66) c* Relative capacitance, defined by
B Magnetic field intensity vector
Eq. (3.138)
B Empirical coefficient, defined by c
Eq. (3.3) Coefficient, defined by Eq. (3.92)
D Operator, defined by Eq. (3.7)
dp Particle diameter K Mobility of ions

4 Ball probe diameter K Thermal conductivity of gas
E Electric field strength Kv Thermal conductivity of particles
E Energy of interaction between atoms k Boltzmann constant
E Energy level k Collisional thermal conductivity
Eo Electric field strength in a vacuum I Interparticle distance
E{ Fermi energy level M3 Mass of displaced fluid
Es Electric field strength at surface m Number of north magnetic poles
E* Contact modulus m Mass of a particle
E Electric field intensity vector m Mass of an ion
e Charge of an electron m* Relative mass
Fdi Force due to dipole me Mass of an electron
Fe Electrostatic force m[ Mass of an ion
Fer Electrostatic repulsion force Nc Number of simultaneous impacts
Fm Magnetic force by particles on a ball probe
FP Force due to pressure gradient n Particle number density
Fv Van der Waals force nq Charge density
Fz Total resistance to particle motion Pv Van der Waals pressure
FA Carried mass force vector Pc Collisional stress tensor
FB Basset force vector p Pressure
Fc Collisional force vector p* Dimensionless pressure for the
FD Drag force vector Oseen expansion
Electric force vector p Dimensionless pressure for the
FE
Lorentz force vector Stokes expansion
* em
Magnus force vector Pf Pressure of fluid
FM
Pd Dipole moment vector
Fs Saffman force vector
FT Qb Averaged charge transfer per
Radiometric force vector
impact
/ General function
Qe Surface charge of a particle carried
/c Collision frequency by electrons
g Gravitational acceleration Qis Surface charge of a particle carried
h Planck's constant by ions
he Charge transfer coefficient Qs Surface charge of a particle
hm Lifshitz-van der Waals constant <2 S] Saturated surface charge of a
i Surface charge current particle
i Total charge current q Charge carried by a particle
k Electric current through a ball q Atom density in the material
probe qo Charge density in a vacuum
Electric current density 4eO Electron charge density
Mass flux of particles Ion charge density
K Constant, defined by Eq. (3.41) R Spherical coordinate
Nomenclature 125
/?BSRatio of Basset force to Stokes UQ Gas velocity in the #-direction,

drag defined byEq. (3.15)
RM Gas constant M* Dimensionless gas velocity vector
RMS Ratio of Magnus force to Stokes for the Oseen expansion
drag u Dimensionless gas velocity vector
Rss Ratio of Saffman force to Stokes for the Stokes expansion
drag V Electric potential
Rea Particle Reynolds number, based V Volume of a particle
on particle radius and relative V Relative velocity of particle to gas,
velocity defined byEq. (3.14)
Rep Particle Reynolds number, based V Gas velocity in the y -direction,
on particle diameter and relative defined by Eq. (E3.19)
velocity Vp Particle velocity in the y -direction,
r Cylindrical coordinate defined by Eq. (E3.20)
rc Contact radius v Mean thermal speed of ions
r Position vector v Time-dependent relative velocity
of particle to gas, defined by
S Surface area
Eq. (3.37)
s Separation distance between two
vs A solution of Eq. (3.30)
atoms
W Work function
s Separation distance between two
x Cartesian coordinate
surfaces of interacting materials
y Cartesian coordinate
s Stretched coordinate
z Cartesian coordinate
T Absolute temperature
z Cylindrical coordinate
Tc Granular temperature
Tp Particle temperature
t Time Greek Symbols
tc Elapsed time for particles to
a Approaching distance
collide
a Electrothermal number, defined by
U Gas velocity in the x-direction,
Eq. (3.126)
defined by Eq.(E3.19)
6 Medium permittivity
U Relative velocity of trailing
particle to leading particle, defined e0 Particle permittivity in a vacuum
byEq. (E3.1) r Relative permittivity of dielectric
Up Particle velocity in the x-direction, materials
defined by Eq. (E3.20) f Imaginary frequency axis
U Gas velocity vector rj Correction factor
Up Particle velocity vector 0 Cylindrical coordinate
wR Gas velocity in the /^-direction, 0 Spherical coordinate
defined byEq. (3.15) 0 Angle between a point on the
wr Gas velocity in the r-direction, sphere and the electric field
defined by Eq. (3.4) XQ Mean free path of
uz Gas velocity in the z-direction, particle-to-particle collisions
defined by Eq. (3.4)
Debye shielding distance Time constant of diffusion

London's constant charging
hi
M Dynamic viscosity TS Stokes relaxation time
Mr Relative permeability 0 Potential function
V Kinematic viscosity 0 Thermionic work function
Distance from a fixed point 0 Rotating angle
P Density of gas Stream function
Pp Density of particles ft Angular velocity vector

a Electrical conductivity CO Angular velocity
T Time constant of saturation u> Dimensionless angular velocity
charge vector
References
Basset, A. B. (1888). A Treatise on Hydrodynamics, 2. Cambridge: Deighton, Bell; (1961) New
York: Dover.
Boussinesq, J. (1903). Theorie Analytique de la Chaleur, 2. Paris: Gauthier-Villars.
Cheng, L. and Soo, S. L. (1970). Charging of Dust Particles by Impact. 7. AppL Phys., 41, 585.
Clift, R., Grace, J. R. and Weber, M. E. (1978). Bubbles, Drops, and Particles. New York:
Academic Press.
Corrsin, S. and Lumley, J. (1956). On the Equation of Motion for a Particle in Turbulent Fluid. AppL
Sci. Res., 6A, 114.
Cross, J. A. (1987). Electrostatics: Principles, Problems and Applications. Philadelphia: Adam
Hilger.
Dahneke, B. (1972). The Influence of Flattening on the Adhesion of Particles. J. Colloid Interface
5c/., 40, 1.
Epstein, P. S. (1929). Zur Theorie des Radiometers. ZS.f. Phys., 54, 537.
Faeth, G. M. (1983). Evaporation and Combustion of Sprays. Prog. Energy Combust. ScL, 9, 1.
Hamaker, H. C. (1937). The London-Van der Waals Attraction Between Spherical Particles.
PhysicalV, 10, 1058.
Hansell, D., Kennedy, I. M. and Kollmann, W. (1992). A Simulation of Particle Dispersion in a
Turbulent Jet. Int. J. Multiphase Flow, 18, 559.
Harper, W. R. (1967). Contact and Frictional Electrification. Oxford: Oxford University Press.
Hettner, G. (1926). Zur Theorie der Photophorese. ZS.f. Phys., 37, 179.
Hough, D. B. and White, L. R. (1980). The Calculation of Hamaker Constants from Lifshitz Theory
with Applications to Wetting Phenomena. Adv. Colloid Interface Sci., 14, 3.
Hughes, R. R. and Gilliland, E. R. (1952). The Mechanics of Drops. Chem. Eng. Prog, 48, 497.
Israelachvili, J. N. (1992). Intermolecular and Surface Forces, 2nd ed. New York: Academic Press.
Kriebel, A. R. (1961). Particle Trajectories in a Gas Centrifuge. Trans. ASME, J. Basic Eng., 83D,
333.
Krupp, H. (1967). Particle Adhesion Theory and Experiment. Adv. Colloid Interface Sci., 1, 111.
Lamb, H. (1945). Hydrodynamics, 6th ed. New York: Dover.
Landau, L. D. and Lifshitz, E. M. (1960). Electrodynamics of Continuous Media. Oxford:
Pergamon Press.
Langbein, D. (1969). Van der Waals Attraction Between Macroscopic Bodies. /. Adhesion, 1, 237.
References 127
Lee, K. C. (1979). Aerodynamic Interactions Between Two Spheres at Reynolds Number Around
104. Aerosp. g., 30, 371.
Lifshitz, E. M. (1956). The Theory of Molecular Attractive Force Between Solids. Soviet Phys.,
2,73.
Liu, B. Y. H. and Yeh, H. C. (1968). On the Theory of Charging of Aerosol Particles in an Electric
Field. J. Appl. Phys., 39, 1396.
London, F. (1937). The General Theory of Molecular Forces. Trans. Faraday Soc, 33, 8.
Magnus, G. (1852). Uber die Abweichung der geschosse, nebst einem Anhange: Ubereine
auffallende Erscheinung bei rotirenden Korpern. Berlin: F. Dummler.
Massimilla, L. and Donsi, G. (1976). Cohesive Forces Between Particles of Fluid-Bed Catalysis.
Powder Tech., 15, 253.
McAteer, O. J. (1990). Electrostatic Discharge Control. New York: McGraw-Hill.
Montgomery, D. J. (1959). Static Electrification of Solids. Solid State Phys., 9, 139.
Murphy, A. T. (1956). Charging of Fine Particles by Random Motion of Ions in an Electric Field.
Ph.D. Dissertation. Carnegie Institute of Technology.
Odar, F. and Hamilton, W. S. (1964). Forces on a Sphere Accelerating in a Viscous Fluid. J. Fluid
Meek, 18, 302.
Oglesby, S. and Nichols, G. B. (1978). Electrostatic Precipitation. New York: Marcel Dekker.
Oseen, C. W. (1927). Hydrodynamik. Leipzig: Akademische Verlagsgescellschafe.
Passman, S. L. (1986). Forces on the Solid Constituent in a Multiphase Flow. J. Rheology, 30, 1077.
Rowe, P. N. and Hen wood, G. A. (1961). Drag Forces in a Hydraulic Model of a Fluidised Bed-Part
I. Trans. Instn. Chem. Engrs., 39, 43.
Rubinow, S. I. and Keller, J. B. (1961). The Transverse Force on a Spinning Sphere Moving in a
Viscous Fluid. / Fluid Mech., 11, 447.
Rudinger, G. (1980) Fundamentals of Gas-Particle Flow. Amsterdam: Elsevier Scientific.
Saffman, P. G. (1965). The Lift on a Small Sphere in a Slow Shear Flow. / Fluid Meek, 22, 385.
Saffman, P. G. (1968). Corrigendum. J. FluidMech., 31, 624.
Soo, S. L. (1964). Effect of Electrification on Dynamics of a Particulate System. I&EC Fund., 3, 75.
Soo, S. L. (1990). Multiphase Fluid Dynamics. Beijing: Science Press; Brookfield, Vt.: Gower
Technical.
Tchen, C. M. (1947). Mean Value and Correlation Problems Connected with the Motion of Small
Particles in a Turbulent Field. Ph.D. Dissertation. Delft University, Netherlands.
Tsuji, Y, Morikawa, Y and Terashima, K. (1982). Fluid-Dynamic Interaction Between Two
Spheres. Int. J. Multiphase Flow, 8, 71.
Visser, J. (1989). Van der Waals and Other Cohesive Forces Affecting Powder Fluidization. Powder
Tech., 58,1.
Wallis, G. B. (1969). One-Dimensional Two-Phase Flow. New York: McGraw-Hill.
Webers, V J. (1963). Measurement of Triboelectric Position. J. Appl. Polymer Sci., 1, 1317.
White, H. J. (1963). Industrial Electrostatic Precipitation. Reading, Mass.: Addison-Wesley.
Williams, M. W. (1976). The Dependence of Triboelectric Charging of Polymers on Their Chemical
Composition. J. Macromolecular Sci. Rev. Macromol. Chem., C14, 251.
Zhu, C , Liang, S.-C. and Fan, L.-S. (1994). Particle Wake Effects on the Drag Force of an
Interactive Particle. Int. J. Multiphase Flow, 20, 117.
Zhu, C. and Soo, S. L. (1992). A Modified Theory for Electrostatic Probe Measurements of Particle
Mass Flows in Dense Gas-Solid Suspensions. J. Appl. Phys., 72, 2060.
Problems
3.1 Calculate the van der Waals force (newtons) between (a) two diamond spheres of radii
al=a2 = 10 /xm; (b) two graphite spheres of radii a\ = 1 /xm and a2 = 10 jLim; and (c) a
graphite plate and a diamond sphere of radius a\ = 5 /xm. The Hamaker coefficients for
diamond and graphite are 28.4 x 10~13 and 47.0 x 10~13 erg, respectively. For the van
der Waals interaction between two different solid materials, the Hamaker coefficient can be
approximated by the geometric mean of the individual coefficients [e.g., Visser, 1989] as
An = VAnA22 (P3.1)
Assume that the separation distance is 4 A.
3.2 Consider the collision of particles due to wake attraction, as shown in Fig. E3.1. It is
assumed that (a) the motion of the leading particle is not affected by the approach of the
trailing particle; (b) particles are equal-sized, rigid, and spherical; and (c) initially, the
particles move nearly at their terminal velocities with a very small velocity difference and
are separated by a characteristic distance / 0. An empirical relation can be used to describe
the effects of the interparticle distance / and particle Reynolds number Re p on the drag force
of the trailing particle as
^ - = 1 - (1 - A)exp( B-t- j (P3.2)
where FD0 is the drag force of a single noninteracting particle and the coefficients A and
B, for 20 < Rep < 150, are given by [Zhu et al, 1994]
A = 1 - exp(-0.483 + 3.45 x 10"3Rep - 1.07 x 10"5Re^) (P3.3)
and
B = -0.115 - 8.75 x 10"4Rep + lO^Rej; (P3.4)
(a) Derive an expression for the particle contact velocity without considering the Basset
effect, (b) Use the result of (a) and Eq. (E3.9), which considers the Basset effect (see
Example 3.1), to predict the contact velocity under the conditions Re p = 90; \i = 0.057
kg/m s; p p = 1,500 kg/m3; p = 1,200 kg/m3; dv = 0.019m; and/ 0 = 2dp. The measured
contact velocity is 0.13 m/s. Compare the calculated contact velocities with the measured
velocity to show that the Basset force is important and cannot be neglected in the formulation.
3.3 Prove that the resistance on a sphere moving in a gravitational field with constant velocity
V can be obtained in the form given in Eq. (3.36). If the sphere starts moving from rest
with an accelerating velocity, v, show that the resistance can be expressed by Eq. (3.37).
3.4 Determine the initial acceleration of a free-falling glass bead with the density of 2,300
kg/m3 in the air and that of a rising spherical air "bubble" in a fluidized bed. The fluidized
bed can be regarded as a pseudocontinuum medium with a density of 600 kg/m 3. Discuss
the significance of carried mass force for both cases.
3.5 For a fixed spherical particle in a fully developed laminar pipe flow, determine the Saffman
force on the particle at various radial positions. Identify the location of the maximum
Saffman force. Discuss the case if the flow is turbulent (using the 1/7 power law for the
velocity profile).
3.6 In the transport of fine particles, it is important to understand interparticle force effects
during the particle motion. On the basis of the data provided by Massimilla and Donsi
(1976), i.e., km = 2 eV; s = 50 A; p p = 1,000-2,000 kg/m3; dp = 20 - 1,000 /xm,
discuss the relative significance of the gravitational force to the van der Waals force for
disperse particles with these properties.
Problems 129
3.7 Consider a large number of uniformly charged solid particles initially kept in a spherical
barrier of radius RQ with a symmetric density distribution. When the barrier is suddenly
removed, the particles start to emerge from that spherical domain. The viscous drag coeffi-
cient is assumed to be constant. It is also assumed that the dipole effect due to self-field can
be negligible. Show that the relationship between the spreading velocity and the position
of a charged particle initially located at Ro can be given by
Uv=cxp(-^pj\ f êxp(A?)d?l (P3.5)

where
A= 3pCD
2 pvdv 2ne
(see Example 3.2 for notations).
3.8 An account of the behavior of acoustic wave propagation in a gas-solid suspension or
particle movement in a turbulent eddy requires comprehensive knowledge of the dynamics
of particle motion in an oscillating flow field. This oscillating flow can be analyzed in terms
of one-dimensional simple harmonic oscillation represented by
U = UQ sin(<uf) (P3.6)
Assume that the only force imposed on the particle in the oscillating flow field is the Stokes
drag. Derive expressions for the particle velocity and the phase lag between the particle
velocity and the gas velocity for the condition when the particle is placed in the flow field
for sufficiently long time (i.e., t ^> rs).
3.9 For an electron number density of 1015 m~3 and a mobility of 0.022 (m2/V-s), calculate the
time required to reach 99 percent of the saturation charge of an initially neutral particle by
field charging. If the length of the charging chamber is 5 cm, what is the maximum particle
velocity allowed to ensure 99 percent saturation charge?
3.10 For the ball-probe measurement of charge current in a dilute gas-solid suspension, prove
that, for a given type of particle and a given mass flow ratio, the current measured is
independent of the particle size for dh > dv. It is assumed that the charge transfer coefficient
is constant.
CHAPTER 4
Basic Heat and Mass Transfer
4.1 Introduction
Gas-solid flows involving heat and mass transfer are common in many engineer-
ing operations including petroleum refining, nuclear reactor cooling, solid fuel combustion,
rocket nozzle jetting, drying, and bulk material handling and transport. In the chemical
process industries, reactors for polymerization and hydrogenation commonly require cool-
ing or heating to maintain desired reaction temperatures. In nuclear reactor cooling or solar
energy transportation, graphite suspension flows are used in the heat exchanger. Graphite
suspension coolants can have high heat transfer coefficients in addition to other noted prop-
erties such as high thermal capacity, good temperature stability, and low pressurization
characteristics [Boothroyd, 1971]. Combustion of pulverized coal, gas adsorption on a cat-
alyst, and synthesis of chemicals are among many examples of mass transfer in gas-solid
flows.
There are three basic modes of heat transfer, namely, thermal diffusion (heat conduction),
convection, and radiation. All three modes may occur simultaneously, or one of them may
dominate under certain conditions. The first two modes are analogous to their counterparts
in momentum transfer in a fluid. Thermal radiation, a form of energy transfer through
electromagnetic waves, is governed by rather different laws and can occur even through a
perfect vacuum. It is also noted that an analogy can exist between mass transfer and heat
transfer (excluding radiation) because of the similarity of their governing equations.
This chapter describes the fundamental principles of heat and mass transfer in gas-
solid flows. For most gas-solid flow situations, the temperature inside the solid particle
can be approximated to be uniform. The theoretical basis and relevant restrictions of this
approximation are briefly presented. The conductive heat transfer due to an elastic collision
is introduced. A simple convective heat transfer model, based on the pseudocontinuum
assumption for the gas-solid mixture, as well as the limitations of the model applications
are discussed. The chapter also describes heat transfer due to radiation of the paniculate
phase. Specifically, thermal radiation from a single particle, radiation from a particle cloud
with multiple scattering effects, and the basic governing equation for general multiparticle
radiations are discussed. The discussion of gas phase radiation is, however, excluded because
of its complexity, as it is affected by the type of gas components, concentrations, and gas
temperatures. Interested readers may refer to Ozisik (1973) for the absorption (or emission)
of radiation by gases. The last part of this chapter presents the fundamental principles of
mass transfer in gas-solid flows.
4.2 Heat Conduction

Conductive heat transfer is the dominant mode of intraparticle heat transfer. Under
low Reynolds number flow situations, conductive heat transfer is also an important mode
for fluid heat transfer. This section analyzes the conductive heat transfer characteristics of a
130
42/ Heat Conduction 131
single spherical particle in a quiescent fluid and in an elastic collision with another spherical
particle.
4.2.1 Heat Transfer of a Single Sphere in a Quiescent Fluid

For simplicity, it is assumed that the temperature distribution throughout the sphere
is radially symmetric, no heat generation occurs inside the sphere, and no thermal radiation
takes place between the sphere and the surrounding environment. Thus, the transient heat
balance in the particle leads to
(41)
The boundary and initial conditions are
v ^P / ,r> T\ * D (4.2)
oK
Tp = Tpi(R)
Define dimensionless variables as
Kp '
r
ppca2
(43)
where Bi and Fo are the Biot number and Fourier number, respectively.
The Fo and Bi are two important dimensionless parameters in heat conduction. Fo can
be interpreted as
Kp 3
_ ~^2a rate of heat conduction across a2 in a3
rate of heat storage in a 3 (4.4)
t
Bi can be interpreted as
hv convective heat transfer coefficient at sphere surface
Bi = = (4.5)
Kp/a thermal conductance of solid across depth a
Equations (4.1) and (4.2) can be reduced to dimensionless forms as
i a2 , N dT*
-(rr;) = p- (4.6)
with
P. = 0 at R* = 0
a/?*
r*
<4J)
" J? . _ o r - i
3R
T* = T^(R*) forFo = 0
132 4 /Basic Heat and Mass Transfer
The solution for Eqs. (4.6) and (4.7) gives
where Xn's are the eigenvalues of
A.ncotA.n + B i - l = 0 (4.9)
Equation (4.8) indicates that the one-dimensional transient temperature distribution in-
side a solid sphere without internal heat generation varies with Fo and Bi. On the basis of
the equation, the temperature distribution in the solid can be considered uniform with an
error of less than 5 percent when Bi < 0.1, which is the condition for most gas-solid flow
systems. In transient heat transfer processes where the gas-solid contact time is very short,
it also requires Fo > 0.1 [Gel'Perin and Einstein, 1971] for the internal thermal resistance
within the particles to be neglected. In the following, unless otherwise noted, it is assumed
that the temperature inside a solid particle is uniform.
For a quasi-steady-state heat conduction between an isothermal sphere and an infinitely
large and quiescent fluid, the temperature distribution in the fluid phase is governed by
with boundary conditions

T = Tp atR = a
(4.11)
T = TOO at R -> oo
which leads to the solution
(4.12)
The corresponding heat flux at the surface of the sphere is

j = 2K(Tp - Top)
(4.13)
Note that the heat flux can also be expressed as

Jq = hv(Tv - Too) (4.14)
Thus, the limiting Nusselt number at Rep > 0 is expressed by
Nup = (4.15)
K
Equation (4.15) indicates that, for an isolated, isothermal solid sphere in an infinite fluid,
Nup approaches 2 in the limit of negligible forced or natural convection. This relationship
was confirmed experimentally by Ranz and Marshall (1952).
Consider the heat conduction through a spherical shell of an inner diameter d\ and an
outer diameter d^. It can be shown that
Nup = ] (4.16)
-d\/d2
4.2 /Heat Conduction 133
Equation (4.16) can be reduced to Eq. (4.15) as d2 becomes infinitely large. For a packed
bed of monodispersed spheres arranged in a cubic particle array, the mean thickness of the
"shell" of gas surrounding the particle can be approximated by
(4.17)
nd\
Thus, from Eqs. (4.16) and (4.17), Nup ^ 1 0 , indicating that the Nup of a single particle
due to thermal conduction in a gas-solid suspension roughly varies from 2 to 10. A more
detailed discussion on the variation of Nup with 8/dp is given by Zabrodsky (1966).
4.2.2 Heat Conduction in a Collision of Elastic Spheres

In a collision between two spheres of different temperatures, heat conduction
occurs at the interface. The contact area is usually negligibly small compared to the cross-
sectional area of the spheres. Since the duration of the impact is also very short, the temper-
ature change of the colliding particles is confined to a small region around the contact area.
Therefore, the heat conduction between the two particles can be treated as that between
two semiinfinite media. It is also assumed that there is no thermal resistance between the
contact surfaces. Hence, the temperature and heat flux distributions are continuous across
the contact area. The surfaces outside the contact area are assumed to be flat and insulated.
For general information on collision mechanisms of solids, readers may refer to Chapter 2.
For simplicity, it is assumed that the impact is a Hertzian collision. Thus, no kinetic
energy loss occurs during the impact. The problem of conductive heat transfer due to the
elastic collision of solid spheres was defined and solved by Sun and Chen (1988). In this
problem, considering the heat conduction through the contact surface as shown in Fig. 4.1,
the change of the contact area or radius of the circular area of contact with respect to time
is given by Eq. (2.139) or by Fig. 2.16. In cylindrical coordinates, the heat conduction
between the colliding solids can be written by
1 37] 1 3 / 3T{\ d2T{
r + l l2 (418)
{ ) ^
The boundary and initial conditions are
dTx dT2
Ti = T2; ATpi-r = Kp2 atz = Oandr < rc(t)
oz oz
r\ rri r\ rp
- = = 0 at z = 0 and r > rc(t)

dz dz (4.19)
- = -1=0 atr = 0
or or
Tx = T 10 ; T2 = T20 at t = 0
where D t 's are the thermal diffusivities of solids. Dt is defined by
A = ^ (4.20)
When the contact time is very short, the heat penetration by conduction from one medium
to the other would not be very deep from the contact surface. Therefore, one-dimensional
Contact Surface
i
1
~~~~ Isotherms
dz dz dz
Contact Surface
dz
Figure 4.1. Conductive heat transfer by two-ball collision (from Sun and Chen, 1988).
approximation can be made to account for the heat transfer behavior. This behavior can
usually be characterized by the Fourier number in terms of the total impact duration and
maximum contact radius, as defined by
Fo i = (4.21)
Thus, for a small Fourier number, the one-dimensional heat transfer yields the following
equation
(4.22)
Ai dt ~ dz1 ' l
with the boundary and initial conditions given by Eq. (4.19).

For constant properties, a general solution of the problem
dT d2T
D
(4.23)
71,=o = 0; T\z=0 =
is given by [Carslaw and Jaeger, 1959]
(4.24)
/Dtt
At z = 0, the preceding equation reduces to

0(0) (4.26)
~dz z=0 y/Dt7Tt
Now, we can use the solutions, i.e., Eqs. (4.24) and (4.26), for the problem imposed by
Eq. (4.22) and its boundary and initial conditions. Using the continuity condition for the
temperature across the interface, we have
<t>x=<h + (720 - 7i0) (4.27)
Since the heat flux across the interface is continuous, we have
Ai A2
^qw == 01 OD =
02 (0) (4.28)
where the minus sign is due to the opposite outward-directed normals of the two media at
the contact interface. The interface heat flux is thus obtained as
-1
, (4.29)
Here t is measured from the moment of initial contact and, thus, can be expressed as a
function of the radius of contact area r c, as shown in Fig. 2.16. The total heat exchange per
impact is
rrcm rtc-2t
Qc = 2n / Jqwdtrdr (4.30)
Jo Jo
with t related to r by either Eq. (2.137) or Eq. (2.139). The numerical integration of Eq.
(4.30) gives
Qc0 = 2.73(r2o - r lo )r c 2 m V^f ^ +^ ) (4.31)

where rcm and tc are given by Eqs. (2.133) and (2.136), respectively; and the subscript "0"
for Qc indicates that it is for the case in which the Fourier number approaches zero.
The preceding result of heat exchange is derived for heat conduction only in the z-
direction when Fo is negligibly small. When heat conduction in the r-direction is not
negligible, more heat would be transferred between the two colliding media. To quantify
it, a correction factor C may be introduced in such a way that the total heat exchanged per
impact becomes
Qc = CQc0 (4.32)
Since there is no general analytical solution for Eq. (4.18) with the appropriate boundary
and initial conditions, the factor C can only be evaluated by the numerical approach.
For most solid materials, the parameter pp\C\/pP2C2 varies from about 0.1 to 10 and the
parameter K\/K2 varies from about 0.001 to 1. The Fourier number can be considered
as a measure of the impact conditions. For example, a larger value of the Fourier number
represents a smaller impact velocity. Within the range of interest, the solutions for the
1.0 1.0
0.001 10.0 0.001 10.0
Figure 4.2. Relations of the correction factor with the Fourier number: (a) K1/K2 = 0.001;
(b) Ki/K2 = 1.0 (from Sun and Chen, 1988).
correction factor C are plotted in Figs. 4.2(a) and 4.2(b) for two limiting values of K1/K2.
Each plot gives the variations of C with Fo for different pv\C\/pV2C2 values. It can be seen
that, as Fo increases, the heat conduction in the radial direction becomes significant in the
total heat conduction process.
In gas-solid suspensions and fluidized beds, the heat transfer between particles and the
wall surface or between a particle at one temperature and a group of other particles at another
temperature is largely due to particle impacts. Thus, the average rate of heat transfer may
be expressed in terms of a collisional heat transfer coefficient hc, which is defined by
BAI/pQc
hc = (4.33)
^20 ~ T\Q
where AUV is the relative impact velocity; n is the particle number density; and the product
of n and AUP is the rate of particle surface impact per unit area. Thus, from Eqs. (4.31),
(4.32), and (4.33), the collisional heat transfer coefficient is given by
-1
(4.34)
Example 4.1 Determine the collisional heat transfer coefficients under each of the follow-
ing conditions: (1) collisions of a cloud of hot particles with a cold particle; (2) collisions
of a cloud of cold particles with a hot wall. Assume the particles are in random motion with
the average impact velocity of 0.1 m/s. All the particles are spherical and of the same diam-
eter of 100 jLtm. The particles and wall are made of steel with v = 0.3, E = 2 x 105 MPa,
p p = 7,000 kg/m3, Kv = 30 W/m K, and c = 500 J/kg K. The particle volume fraction
is 0.4.
Solution The maximum contact radius and contact time for an elastic sphere
collision can be estimated from Eqs. (2.133) and (2.136), respectively. For collisions among
particles, we have
1/5
(AU
2 V 16 E ^
2 V 16 2 x 1011 '
and
2.94x4 r 2 m = 11.76 x ( 0 . 6 3 x 1 0 - ^
0.1 x 10"4
The Fo is obtained from Eq. (4.21) as
3Q x 4 7 x
Fo=
F ^ = ' = 10 (E4 3)
6 2
pvcr%m 7,000 x 500 x (0.63 x 10" )
Therefore, from Fig. 4.2, the correction factor C is found to be 5. The collisional heat
transfer coefficient by interparticle collisions is thus given by Eq. (4.34) as
_ SA9KpapAUvCrcmtc
3
V
8.19 x 30 x 0.4 x 0.1 x 5 x 0.63 x 10~6 x 4.7 x 1(T7 _ g W
Similarly, for collisions between a group of particles and the wall, we have
_ dp /57T p p
"" 2 U 2X10 11
and
2.94 x 2r2m = 5.88 x (0.95 x 1Q-6)2 = 7
AUpdp 0.1 x 10- 4 ' X S
The Fo is obtained by
Anrc JU
x 1Q
F o = = = 5 (E4 7)
ppcrc2m 7,000 x 500 x (0.95 x 10" 6 ) 2 V
'J
From Fig. 4.2, the correction factor C is about 4. The collisional heat transfer coefficient
by interparticle collisions is given by
nc
8.19 x 30 x 0.4 x 0.1 x 4 x 0.95 x 10"6 x 5.3 x 10"7 W

= 2
.(10-4)3^5 fe
138 4 / Basic Heat and Mass Transfer
This example indicates that impact of particles among themselves or with the wall is in
general not an effective means of heat transfer in gas-solid flows.
4.3 Convective Heat Transfer

When a thermal surface is exposed to a flow field, the heat transfer from the
surface is enhanced by thermal convection in addition to heat conduction. Although thermal
convection in general consists of forced convection and natural convection, the effect due
to natural convection in a gas-solid system is usually insignificant. This section presents
the dimensional analysis which gives rise to the basic dimensionless groups accounting
for forced convection in a single-phase flow. The thermal convection involving a single
spherical particle in a uniform flow and a flow of gas and solids treated as a pseudocontinuum
medium is discussed.
4.3.1 Dimensional Analysis of Forced Convection in a Single-Phase Flow

The influential parameters in a thermal convection system may be represented
by the characteristic length /, velocity U, density p, viscosity /x, specific heat at constant
pressure cp, and thermal conductivity K. Thus, we have
h = / ( / , U, p, ii, c p, K) (4.35)
where h is the convective heat transfer coefficient. In terms of dimensional units, Eq. (4.35)
takes the form
2
[M] _ / [L] LPVI^J L[M]
J L [L]
J L [M][L]
JL J \
= f [Li !_i (4 35)
3 2
[T]3[0] V m ' [L] ' [L][T]' [T] [0]' [T]3[0];
where [M] denotes the mass unit; [L] represents the length unit; [T] stands for the time unit;
and [0] is the thermal unit. From the Pi theorem, it yields
Nu = f(Re, Pr) (4.37)
where Nu is the Nusselt number and Pr is the Prandtl number. The physical significance of
Nu and Pr may be interpreted as
hi hi2 AT convection
Nu = - = x r = (4.38)
K v,i conduction
T
and
fji/p momentum diffusivity
A thermal diffusivity
4.3.2 Heat Transfer of a Single Sphere in a Uniform Flow

For forced convective heat transfer over a sphere in a uniform flow, a frequently
used empirical relation was proposed by Ranz and Marshall (1952) as
Nup = 2.0 + 0.6(Re p ) 1/2 (Pr) 1/3 (4.40)
where the particle diameter is used as the characteristic length. In the equation, the first
term matches the theoretical requirement of Nup = 2 at Rep = 0, whereas the second term
4.3 / Convective Heat Transfer 139
may be regarded as the enhancement factor due to forced convection. Equation (4.40) is
applicable to a range of Rep up to 104 and Pr > 0.7, which covers most gas-solid flows. It
is noted that although Eq. (4.40) was originally obtained from the study of liquid droplets,
this relation was proved also to be valid for the case of rigid spheres [Hughmark, 1967].
For most gases, Pr is nearly constant (~0.7). Therefore, for convective heat transfer of an
individual solid sphere in a gaseous medium, Nup is only a function of Rep. An alternative
estimation of Nup for thermal convection of a sphere in a uniform flow was recommended
by Me Adams (1954) as
Nup = 0.37 Rep16 for 17 < Rep < 70,000 (4.41)
where all associated thermal properties are determined on the basis of the ambient temper-
ature of the gas.
The analytical solution for convective heat transfer from an isolated particle in a Stokes
flow can be obtained by using some unique perturbation methods, noting that the standard
perturbation technique of expanding the temperature field into a power series of the Peclet
number (Pe = RepPr) fails to solve the problem [Kronig and Bruijsten, 1951; Brenner,
1963]. The Nup for the thermal convection of a sphere in a uniform Stokes flow is given by
Nup = 2 + 1 (RepPr) +tf(RepPr) (4.42)
which is proved by the following example using the method developed by Kronig and
Bruijsten (1951).
Example 4.2 Show that the convective heat transfer from a sphere in aflowingmedium at
low Reynolds number can be expressed by Eq. (4.42). It is assumed that the sphere is kept
at a constant temperature Tp and the temperature inside the sphere is uniform. Theflowis
uniform with velocity of UQQ and temperature of T^ at infinity. All the thermal properties
are constant.
Solution The energy balance equation can be expressed by

2
v r - u vr = o (E4.9)
with the boundary conditions (in polar coordinates)

T = TV atR = a; T = T^ at R - > o o (E4.10)
For simplicity, take the dimensionless form of the equation as
v*2r* - eu* v*r* = o (E4.ii)
where
T - r^ U R 1
7* = , U* = , V* = flV, r = - , e = -RepPr (E4.12)
Tp - T^ UOQ a 2
and the boundary conditions become
; T*=0 atrôo (E4.13)
Upon taking the polar axis parallel to the streamlines at infinity, the velocity components
of the Stokes flow can be given by
u:
= 0 - h + 2^)cos*' u; =
"0 - 1 ~ i) s i n ( ? (E414)
-
Substituting Eq. (E4.14) into Eq. (E4.ll) and setting cos# = /z, Eq. (E4.ll) becomes
(V*2 + sF + eG)T* = 0 (E4.15)
with
The term F represents the contribution from the uniformflow,whereas the term G accounts
for the deviation from the parallel flow due to the presence of the sphere.
Now express T* by a new variable 0 with the relation
T* = 0 e x p f - | r ( l - JLO) (E4.17)
The governing equation for 0 is obtained by substituting Eq. (E4.17) into Eq. (E4.15) as
while Eq. (E4.13) gives
0 = expf-(l-/x)j forr = l (E4.19)
For 0, we use thefirst-orderapproximation

0 = 0O + eSi + o(e) (E4.20)
Equation (E4.18) will be satisfied, provided
V*20O = 0, V*20! = f A + 1 - G\0o (E4.21)
while Eq. (E4.19) leads to
0 O = 1, 0 i = ^(1 - ix), forr = 1 (E4.22)

Equations (E4.21) and (E4.22) are fulfilled by
(E4 23)
o =-, i = - + (~ + A - A V '
r 2r \ 4r Sr2 8r 3 /
Thus, from Eqs. (E4.17), (E4.20), and (E4.23), we have
1
)r=i (l +
+ ++ 0 ( (e ) ) (E4.24)
which yields Eq. (4.42).

4.3 / Convective Heat Transfer 141
4.3.3 Thermal Convection in Pseudocontinuum One-Phase Flow

General considerations of the heat transfer in a gas-solid flow require separate
energy equations for the two phases that are coupled through a local heat transfer coefficient
as well as local phase velocities. These approaches are complex and are introduced in
Chapter 5.
One of the simplified heat transfer models of two-phase flows is the pseudocontinuum
one-phase flow model, in which it is assumed that (1) local thermal equilibrium between
the two phases exists; (2) particles are evenly distributed; (3) flow is uniform; and (4) heat
conduction is dominant in the cross-stream direction. Therefore, the heat balance leads to a
single-phase energy equation which is based on effective gas-solid properties and averaged
temperatures and velocities. For an axisymmetric flow heated by a cylindrical heating
surface at Tw, the heat balance equation can be written as
dT Ke a / 8T\
() (4.43)
with
T = Tw at r = R
(4.44)
T = 7] at z = 0
where z and r are in the streamwise and cross-stream directions, respectively; K Q is the
effective thermal conductivity; and U is the pseudocontinuum one-phase velocity.
Using the method of separation of variables, the temperature distribution from Eqs. (4.43)
and (4.44) is obtained as
J A
T(r. z) = Tw + 2(7] - rw) f exp(-/8&) (4.45)
where Jo and Ji are the zeroth and first-order Bessel functions of the first kind, respectively,
and An's are the eigenvalues obtained from the eigenequation
J0(KR) = 0 (4.46)
and P is a system-related constant given by
P = JT^- (4.47)
The heat transfer coefficient is

IK
(4 48)
^jE^W) -
n=\
and the length-averaged Nusselt number is
The applicability of the preceding pseudocontinuum approach to convective heat transfer

of gas-solid systems without heat sources depends not only on the validity of the phase
continuum approximation but also on the appropriateness of the local thermal equilibrium
assumption. The local thermal equilibrium may be assumed only if the particle-heating
Packed and '

incipient
incipient Ii
fluidized beds I
I J
o 100 Fast fluidized Granular

I Pneumatic I beds flows
conveying I and dense
| 10 and dilute I suspensions
suspensions i
1
| .Bubbling and
slugging
0.1 'fluidized bedsl
0.01 0.1 1
Solids fraction ocp
Figure 4.3. Regime map for gas-solid flows (from Hunt, 1989).
surface residence time, r ps , is much greater than the particle thermal diffusion time, r pd.
Here, r ps and rpa may be estimated by
td=d[ L (4.50)
Furthermore, we may define a modified Fourier number as

t P sA
Fo' = ^ = (4.51)
A regime map of Fo' versus the solid volume fraction, ap, for various gas-solid flows
was presented by Hunt (1989), as shown in Fig. 4.3. Hunt (1989) suggested that except
when Fo' > 1 and ap > 0.1, use of the pseudocontinuum model is inappropriate. Thus,
from Fig. 4.3, it can be seen that the pseudocontinuum model is applicable to packed
beds, incipient fluidized beds, and granular flows, whereas it is not applicable to pneumatic
transport flows, dilute suspensions, bubbling beds, and slugging fluidized beds [Glicksman
and Decker, 1982; Hunt, 1989].
4.4 Thermal Radiation

An element in a thermally radiative environment absorbs, reflects, refracts, diffracts,
and transmits incoming radiative heat fluxes as well as emits its own radiative heat flux.
Most solid materials in gas-solid flows, including particles and pipe walls, can be reason-
ably approximated as gray bodies so that absorption and emission can be readily calculated
from Stefan-Boltzmann's law (Eq. (1.59)) for total thermal radiation or from Planck's
formula (Eq. (1.62)) for monochromatic radiation. Other means of transport of radiative
4.4 / Thermal Radiation 143
energy can be generally described by "scattering," which may encompass various effects
due to reflection, refraction, diffraction, and transmission. In this section, we focus on the
determination of scattering of radiative energy by solids in various gas-solid flows.
4.4.1 Single-Particle Scattering

When a beam of incident radiation strikes a particle, some of the radiation is ab-
sorbed by the surface while the remaining portion is scattered out of the surface. The reflec-
tion, refraction, diffraction, and transmission depend not only on the optical properties of the
particle but also on the particle size dv relative to the wavelength of the incident radiation k.
The radiant energy scattered and absorbed by a particle can be characterized by the
"cross section" of the particle for scattering and absorption, respectively. The cross section
for scattering, C s, is defined as the area on which the amount of incident energy equals the
total energy scattered in all directions, and the cross section for absorption, C a , is defined
as the area on which the amount of incident energy equals the total energy absorbed into
the particle. Similarly, we may also introduce the cross section for extinction, C e , defined
as the area on which the amount of incident energy equals the total energy removed from
the original beam. From energy conservation, we have
Ce = C s + C a (4.52)
It is convenient to introduce the efficiency factors for extinction, scattering, and absorp-
tion, which are defined by
Ge = ~T7> Gs = ~pT' Ga = ~~^T (4.53)

where C is the geometric cross section. From Eq. (4.52), we have
Ge = Gs + Ga (4.54)
In general, G' s a r e functions of the orientation of particles and the state of polarization of
incident beams. However, for spherical and homogeneous particles, G' s a r e independent
of both.
The most commonly used theory for single-particle scattering was originally developed
by Mie (1908), who applied electromagnetic theory to derive the properties of the electro-
magnetic field with a plane monochromatic wave incident upon a sphere. Although the
Mie scattering theory covers the entire range of particle diameters, general formulations
for cross sections or efficiency factors are very complicated. The Mie scattering theory can
be greatly simplified when dv <& k or dp > A., leading to Rayleigh scattering and geometric
scattering, respectively. From Wien's displacement law (Eq. (1.63)), for the temperature
range of most gas-solid flows, the wavelengths at maximum emissive power are less than
10 fim (see Example 4.3). Thus, particles for Rayleigh scattering are in the submicrometer
range, which is beyond the particle size range of interest in this book. Therefore, discus-
sions on Rayleigh scattering are excluded here. For Rayleigh scattering of submicrometer
particles, interested readers may refer to Bohren and Huffman (1983).
Example 4.3 For gas-solid flows, the typical temperature range where thermal radiation
should be considered is from 500 K to 2,000 K. Show that for Rayleigh scattering, the particle
sizes of relevance are in the submicrometer range. The thermal radiation is assumed to be
monochromatic, and the particles can be regarded as blackbodies.
Solution The wavelength at which the monochromatic-emissive power of a

blackbody is maximum for a given temperature is given by Eq. (1.63) as
XmTv = 2,898 jLtm-K (E4.25)
For 500 < Tp < 2,000 K, the range of Am is obtained as
1.4 Aim < Am < 5.8/zm (E4.26)
To fulfill the requirement of Rayleigh scattering (i.e., dv <^ Am), we may take the maximum
particle size dpm for Rayleigh scattering to be 10 times smaller than A.m. Therefore, we
have
0.14 fim < dpm < 0.58 /xm (E4.27)
which is in the submicrometer range.
4.4.1.1 Mie Scattering

A typical single-particle scattering consists of an electromagnetic plane wave traveling
through a medium and impinging upon a sphere. This problem was first solved by Mie in
1908 using Maxwell's equations, from which the efficiency factors for extinction, scattering,
and absorption can be obtained [Ozisik, 1973; Bohren and Huffman, 1983]. We have
no intention of presenting the detailed derivation of the Mie theory here because of its
complexity, but rather will focus on its applications to simple cases.
Considering an incident plane wave on a sphere of no net surface charge, the scattering
and extinction efficiency factor for the field far away from the particle can be approximated
by [Van de Hulst, 1957]
2
n=l
n=l
(4.55)
2 ln
Ge = ~2 Y^ + !) Re ( A n + Bn)
* n=\
where Re(z) denotes the real part of the complex z and = 2jta/k . The coefficients An
and Bn are determined by
where r\ 27rriia/X; a is the particle radius; n\ is the index of refraction; and functions
i/fniz) and l;n(z) are defined by
(4.57)
where J and H are the Bessel functions of the first and second kind of order n + 1/2,
respectively.
The directional distribution of the scattering intensity can be described by phase func-
tions. A phase function is defined as the ratio of scattering intensity in a direction to the
scattering intensity in the same direction if the scattering is isotropic. Thus, it is a normalized
function and is defined over all directions. Typical phase functions for small, intermediate,
and large values of and nx are illustrated in Fig. 4.4, with the spheres assumed to be
nonabsorbing [Tien and Drolen, 1987]. It is shown that the phase functions mainly vary
100
900
-100
(a) (b)
(c)
100
900
-100
(d) (e)
Figure 4.4. Scattering phase functions for small, intermediate, and large values of , n\ (from
Tien and Drolen, 1987): (a) = 0.1, m = 1.05; (b) = 30.0, n{ = 1.05; (c) = 1.0, m = 1.5;
(d) = 0.1, i = 3.0; (e) = 30.0, n { = 3.0.
with the change of particle size. With increasing particle size, the distribution becomes less
uniform.
4.4.1.2 Simple Scattering from Large Spheres

For some typical modes of scattering from large spherical particles ( > 5), simple formula-
tions of phase functions can be obtained. These modes include scattering from a specularly
reflecting sphere, scattering from a diffuse reflection sphere, and scattering by diffraction
from a sphere.
A. SPECULARLY REFLECTING SPHERES

Consider an unpolarized incident radiation on a dielectric sphere. The phase function <p(6)
can be expressed as [Siegel and Howell, 1981]
A sin2(<9 - cos2(6> + x)
1+ (4.58)
2 sin ((9 + x) V ' cos2(^ - x)
2
where 0 is the incident angle; A is a function of the index of refraction and is determined
from the relationship of an integration of 0 (0) over all 0 to be equal to 1; and x is given by
= sin# (4.59)
The phase function based on Eq. (4.58) is plotted for various indices of refraction nx, as
shown in Fig. 4.5(a). The maximum directional scattering is from the forward scattering
with 0 of A at 0 = n/2, whereas the minimum is from the backward scattering with 0 of
A(n[ l)2/(fti +1) 2 at 0 = 0. Thus, for example, the backward scattering is only 4 percent
of the forward scattering for specularly reflecting spheres of nx = 1.5.
120 90 60
150
330
240 300
Figure 4.5a. Scattering diagram for a specularly reflecting sphere that is large compared with
the wavelength of incident radiation (from Siegel and Howell, 1981).
120 90
150
330
240 270 300
Figure 4.5b. Scattering phase function for a diffuse reflecting sphere which is large compared
with the wavelength of incident radiation and with constant reflectivity (from Siegel and
Howell, 1981).
B. REFLECTION FROM A DIFFUSE SPHERE

For a diffuse sphere, each surface element that intercepts incident radiation will reflect the
energy into the entire In solid angle above that element. Thus, the radiation scattered into a
specified direction will arise from the entire region of the sphere that receives radiation and
is also visible from this specified direction. Consequently, the phase function for a diffuse
sphere can be obtained as [Siegel and Howell, 1981]
= (sinj8-0cosj8) (4.60)
3TT
where /? is the scattering angle between the observer and forward direction. From Eq.
(4.60), 0 (/3) is plotted against various /3, as shown in Fig. 4.5(b). In this case, the maximum
scattering occurs from backward scattering where j3 = n, and the minimum is from forward
scattering.
C. DIFFRACTION FROM A LARGE SPHERE

To account for the total directional scattering from a large sphere, effects of both diffraction
and reflection must be considered. When a spherical particle is in the path of incident
radiation, the diffracted intensity may be obtained from Babinet's principle, which states
that the diffracted intensity is the same as that for a hole of the same diameter. The phase
function for diffraction by a large sphere is given by [Van de Hulst, 1957]
*(/?) = (4.61)
sin 2 ^
.001
0 1 2 3 4 5 6 7 8 9 10
Figure 4.5c. Phase function for diffraction from a large sphere.
where Ji is the Bessel function of the first kind of order one. Equation (4.61) is plotted
against various /Ts, as shown in Fig. 4.5(c). Thus, for large , the diffracted radiation lies
within a narrow angular region in the forward scattering direction.
4.4.2 Radiant Heating of a Particle

Consider a situation in which a particle is suddenly exposed to a radiant flux from
a thermal environment such as a heated wall or surrounding objects. For simplicity, it is
assumed that the radiant flux is isotropic and the temperature distribution inside the particle
is uniform. Absorption of radiation by the gas is negligible. Moreover, all the physical
properties are constant.
The transient behavior of the particle temperature is defined by the energy balance
mc-^- = hSiToo - Tp) + SaRcrh(T* - Tp4) (4.62)
where S is surface area of the particle and h is the convective heat transfer coefficient.
Equation (4.62) can be simplified to yield a simple solution as given later.
When the temperature difference between the particle and its thermal environment is
small, we may linearize the radiation term by
w ~TV) (4.63)
Hence, Eq. (4.62) reduces to
,d7P (4.64)
me r w - Tp)
dt Tp)+4
which yields a solution of the form
= e-Bt -e~Bt (4.65)
Tv T + A ( 1
4 A / Thermal Radiation 149
where the coefficients A and B are given by
Jl) (4.66)
For particle heating with Tw ^> Tp, it is reasonable to neglect particle radiation to the
surroundings. Thus, Eq. (4.62) becomes
1 rp
^ hSiT - Tp) + 4SaRabT* (4.67)
which has a solution with the exact form of Eq. (4.65), but the coefficients A and B now
are given by
4 hS
A = TOO + -aRcrbT*, B = (4.68)
h me
Example 4.4 Consider a solid sphere in a free fall passing through a heated chamber.
The particle is a 500 /xm alloy with a density of 4,000 kg/m 3, a thermal capacity of
150 J/kg K, and an absorptivity of 0.6. The initial temperature of the particle is 300 K.
Both the temperature of gas inside the chamber and the temperature of the wall are 1,000 K.
Estimate the length of the chamber required for the particle to reach 700 K. The physical
properties of gas are given as /i = 4 x 10~5 kg/m-s, p = 0.4 kg/m3, Pr = 0.7, and
K = 0.067 W/m K. The particle falls at its terminal velocity.
Solution The particle terminal velocity Upt and the particle Reynolds number
Rep are obtained from Eq. (1.7) and Eq. (1.5), respectively, as
= f 0 . 0 7 2 ( 5 x l 0 ; 4 4 ) 1 6 ( 4 ' 0 0 0 - 0 0 r 8 ) " = 5 m/s (E4.28)

V 0.40-4 x (4 x 10- 5 ) 0 - 6 J '
and
R e g = a4X
^ ^ r 4 5 x X i o 34 x 5 = 2 5 (E4 29)
-
In terms of Eq. (4.40), the convective heat transfer coefficient is obtained as
h = (
'067-. (2 + 0.6 x 25^ x 0.75) = 625 W/m2 K (E4.30)

5 x 10
From Eq. (4.68), the coefficients A and B are calculated as
8
= TOO + âRcrbT* = 1,000+ ( 0 . 6 x 5.67 x 10" x l,0004) = 1,218K
dvpvc 5 x 10~4 x 4,000 x 150

Thus, the heating time required to reach 700 K can be estimated from Eq. (4.65) as
J, (E4.32)
uuy
Thus, the required length of chamber is estimated by
l = Uptt = 5x 0.046 = 0.23 m (E4.33)
4.4.3 General Considerations in Radiation with a Particle Cloud

The thermal radiation with a particle cloud may be conveniently studied by a simple
model in which the energy is radiated from one heating surface to the other surface through
a uniform particulate medium. The particulate medium is treated as a pseudocontinuum
where particles absorb, emit, and scatter the radiative heat fluxes. Here, the scattering
represents the combined effects of reflection, refraction, diffraction, and transmission of
the radiation by the particles. In gas-solid flows involving thermal radiation, it is often
assumed that the gas is transparent to the radiation. For radiation through a particle cloud,
it is convenient to introduce the monochromatic coefficients for extinction, scattering, and
absorption, which are defined by
3a p 3a p 3ofp
s
Aa Aa Aa
where ap is the solid volume fraction. Thus, on the basis of Eq. (4.54), cre equals the sum
of crs and a a .
4.4.3.1 Single Versus Multiple Scattering

If a single ray of radiative beam traversing a medium is scattered only once before leaving
the medium, the medium is characterized by single scattering. If more than one scattering
occurs per ray, the medium is characterized by multiple scattering.
In gas-solid flows, single scattering occurs in dilute suspensions where only a few parti-
cles exist in the domain of interest. In this case, the scattered intensity of the suspension is
equal to the scattered intensity from a single particle multiplied by the number of particles.
As the number of particles within the space increases, multiple scattering becomes increas-
ingly significant. The classification between single and multiple scattering in a particulate
medium may be based on the value of the characteristic optical path length / s. One typical
criterion for single scattering was suggested by Bayvel and Jones (1981) as
fL
ls= asdl <0.1 (4.70)
Jo
where / is the characteristic length of the system and as is the scattering coefficient.
4.4.3.2 Independent Versus Dependent Scattering

Independent scattering refers to the scattering from a single particle in a cloud which is not
affected by the neighboring particles, whereas dependent scattering refers to the scattering
which is affected by the neighboring particles. As suggested by Hottel et al. (1971), the
general form of the efficiency factor for dependent scattering is
Gs,D = (4.71)
where c is the interparticle clearance and a p = f(c/a). Similarly, the general form of the
efficiency factor for independent scattering is
Gs,i = / ( M i ) (4.72)
It is noted that the dependent effects are functions of av and c/X instead of just the particle
separation alone. Brewster and Tien (1982) suggested that the scattering is independent
if c/X > 0.3 for the range of 0.66 < < 74 and 0.01 < av < 0.7. A schematic diagram
of independent and dependent scattering regimes for some particulate systems is given in
Fig. 4.6. In this text, all scattering is assumed to be independent for simplicity.
4.4.3.3 Radiation Transport Equation

Consider the energy balance of radiant flux through a scattering medium of monodispersed
spherical particles, in which absorption, emission, and scattering by spherical particles
are included. As shown in Fig. 4.7, the equation for the change in monochromatic radiant
103
102
INDEPENDENT
1
Pulverized i SCATTERING
! coal I
I
10 I combustion'
I I
L4--i
I Colloidal suspensions,
8 1 paints, pigments, etc.
Microspherd
1
1
insulation
10 Soot in flames and md
smoke layers l I
conglomerated
I soot
I particles
io-2
io- 6 io- 5 10" 4 io- 3 io- 2 IO1 1
Particle volume fraction ccp
Figure 4.6. Independent and dependent scattering regime map: particle size parameter versus
volume fraction (from Tien and Drolen, 1987).
Figure 4.7. Coordinate system for scattering function.
intensity with respect to distance along a pencil of rays at s, 0, <p can be expressed by [Love,
1968]
dI(s,0,<p)
, 0,
ds
_ flit pTL
s,6',(p')S(6,(p,0f,<pf)sin6f( (4.73)
4TT JO JO
where / represents the local radiant intensity, p p is the density of the particles, S is the
scattering phase function, and 7^ is the Planck intensity function, expressed by Eq. (1.62).
On the right-hand side of the equation, the first term represents the extinction of the ray in
accordance with the Bourger-Beer relation; the second term is the increase in intensity due
to thermal-energy emission from the particles; and the last term is the energy scattered into
0, <p from rays traversing in all directions. On the basis of Kirchoff 's law (Eq. (1.60)), the
monochromatic emission coefficient is assumed to be equal to the absorption coefficient.
The scattering function S is defined in such a way that the expression
askl(s, 0\ (p')S(O, (p, 6>', <p')df dm dco (4.74)

An
is the rate at which radiant energy is scattered by the differential element of mass dm from
the pencil of ray enclosed in the solid angle dcof having an intensity I(s,6\(p') within the
frequency range / and f -\-df into the solid angle dco characterized by 0 and <p. Note that
the scattering function S satisfies the condition
/
f47T S(0,(p,O f
,(p')dco = (4.75)
0 '
In general, solving for the transmitted and reflected energy from a multiple scatter-
ing medium requires a numerical solution of the integral-differential transport equation,
Eq. (4.73). The techniques for solving the three-dimensional radiation transport equa-
tion for absorbing, emitting, and scattering media can be represented by the differential
approximation [Im and Ahluwalia, 1984], modified differential approximation [Modest,
x=0
Figure 4.8. Coordinate scheme for axially symmetric case.
1989], discrete-ordinate approximation [Fiveland and Jamaluddin, 1991], hybrid method

[Ahluwalia and Im, 1994], and exact Monte Carlo solution [Gupta et ah, 1983]. In the cases
of single scattering, the equation can be greatly simplified since the last term in Eq. (4.73)
vanishes.
A A3 A Approximation of Scattering Integral by Gaussian Quadrature

In order to simplify the integral-differential equation of radiant transfer, certain geomet-
ric and optical restrictions have to be imposed. Consider a system which contains many
uniformly sized spherical particles suspended in a transparent medium and bounded by
parallel plates which emit and reflect radiation in a diffuse manner [Love and Grosh, 1965].
As shown in Fig. 4.8, x represents the component of s normal to surface 1. Because of
symmetry, the radiant intensity varies only with respect to polar angle 0 and is independent
of the azimuthal angle <p. Express optical depth / as
/ = / pporeA dx (4.76)
Jo
Let
fi = cos 0 (4.77)
so that Eq. (4.73) becomes

d/(/,u)
2atk ;_!
where
i r2yT
, At') = / S(0,<p,O',<p')d<p- (4.79)
For an isotropic medium, by substituting the Gaussian quadrature formula for the integral
in Eq. (4.78), the integral-differential equation may be reduced into a system of ordinary lin-
ear differential equations. Specifically, the integral can be treated as [Chandrasekhar, 1960]
fJo
(4.80)
where n is the order of the quadrature and Cj is the corresponding weight factor specified by
the quadrature formula. Because of the discontinuity at \i = 0, the intensity function may
be further divided into a forward part / ( / , /x) for 0 < /x < 1 and a backward part / ( / , /x)
for 1 < /x < 0 [Sykes, 1951]. The resulting equations become
C / /
Mi-jjj = - / ( / , Mi) + hx{l) + 2^ J ( ' Mj)S(Mi, Mj)
(4.81)
for the forward radiant intensity, and

CT sA .
) j/(Z, -Mj) 5(-Mi, -/*j) (4.82)
for the backward radiant intensity.
4.4.4 Radiation Through an Isothermal and Diffuse Scattering Medium

Thermal radiation through an isothermal and diffuse scattering medium represents
one of the simplest cases of radiation in a particulate medium. This problem was treated
by Love and Grosh (1965).
For an isothermal medium at temperature Tp, the monochromatic radiant emission from
particles I^x is a function of Tp and A only and is independent of /. In addition, we may
assume that the solutions of Eqs. (4.81) and (4.82) are of the form
/(/, /Xi) = Xieyl, 1(1, -Mi) = *(i+n)eW (4.83)
Substitution of Eq. (4.83) into Eqs. (4.81) and (4.82) leads to 2n linear algebraic equations
in x\ and JC(i+n) with parameter y. The matrix of the coefficients yields 2n eigenvalues ) \
as well as 2n sets of eigenvectors X[ and X(i+n) for each eigenvalue. Here, k is an iterative
index. The solutions can be expressed as
2n
(4.84)
k=l
where Dk's are the 2n constants of integration.

The determination of ZVs depends on the boundary conditions defined. With the as-
sumption of diffuse walls, the intensities at the wall may be written as
7(0, MO = J2 W j
j !
n (4.85)
+ 2p2 J^
where e\9 e2, pi, and P2 represent the emissivity and reflectivity of wall surfaces 1 and 2,
respectively. Substituting Eq. (4.85) into Eq. (4.84) gives
2n
Dk *i,k-2pi
2n " 7 : (4.86)
-2p2
k=l
Therefore, Z\'s can be expressed in terms of Ibx(T\), Ibx(T2), and Ibx(Tp) by solving the
preceding equations.
Now, substituting these D^s from Eq. (4.86) into Eq. (4.84) leads to the solutions of
radiant intensity inside the scattering medium in a form
/(/, /xO = Ei/bxdi) + E2Ibx(T2) + EpIbx(Tp)
(4.87)
where "s and F's are parameters independent of temperature and dependent on the nature of
single scattering, emissivity of the bounding surfaces, optical spacing between the bounding
surfaces, and optical depth of interest. Consequently, the net monochromatic flux J Tx(l)
can be obtained by the integration of the intensity function as
- G2Ihx(T2) - GpIbk(Tv) (4.88)
where G's are parameters similar to 's and F's. As a result, the total radiant flux in the
scattering medium is given by
poo /oo
Jr(l) = 4d/= [GMTy) - G2Ibx(T2) - GpIbK(Tp)] d / (4.89)
Jo Jo
which generally requires numerical integration for the solution.
The significance of the scattering and emission by the particle cloud can be illustrated
by the following numerical example [Soo, 1990]. Consider a gas-solid flow through a
parallel-plate system. Wall 1 is at 1,111 K with reflectivity of 0.1, and wall 2 is at 278 K
with reflectivity of 0.9. The gap between the two plates is 61 mm. The moving particles
are 2 /xm iron particles, and the temperature of the particles is maintained at 556 K. The
paniculate suspension density is 0.16 kg/m3. It can be shown that the net heat flux from
wall 1 is about 69.22 kW/m2, whereas the net heat flux into wall 2 is about 5.17 kW/m 2.
The net heat input into the particle cloud is thus obtained as 64.05 kW/m 2. It is interesting
to note that without particles the net heat flux between the two plates is only 8.54 kW/m 2.
4.5 Mass Transfer

The transfer of mass from one point to the other may take place by two different
modes, namely, diffusion and convection. The basic mechanisms for these modes of mass
transfer are similar to those for heat transfer discussed in 4.2 and 4.3. Specifically, the
mechanism for mass convection is analogous to heat convection and that for mass diffusion
is analogous to heat conduction.
4.5.1 Diffusion and Convection

Mass or molar diffusion is characterized by small-scale "random-walk" motion
caused primarily by molecular collisions. The flux of diffusion follows the spatial gradients
of a specific local driving force. If the driving force is concentration, the mass diffusion
is termed ordinary diffusion. If the driving force is pressure (as in centrifugal separation)
or temperature (as in thermophoresis), the mass diffusion is called pressure diffusion or
thermal diffusion, respectively. In these cases, the direction of the flux is from the high to
the low driving force. For diffusion of a species through a pore of a particle, ordinary dif-
fusion occurs when the pore diameter is larger than the mean free path of the gas molecule.
However, when the pore diameter is smaller than the mean free path of the gas molecule,
molecules collide more frequently with the walls than with each other; this mode of molecu-
lar diffusion is known as Knudsen diffusion. In ordinary diffusion, the quantification of the
mass flux is more complicated than that of the heat flux as the former requires accounting
for the mixture properties such as mixture composition and individual species velocity in
the mixture. As a phenomenological model equation, similar to Fourier's equation in heat
conduction, Fick's equation expresses the flux of mass or molar diffusion for a species in
a mixture at steady state, i.e., constant temperature and pressure. For a binary mixture,
the diffusive mass flux of species A, j A , is linearly proportional to the local concentration
gradient and is directed toward lower concentration, i.e.,
JA = -DABPmVcoA (4.90)
where DAB is the binary diffusivity; p m is the density of the mixture; and coA is the mass
fraction of species A in the mixture. Fick's equation for the molar diffusive flux for species
A, JA, can be given by
JA = -DABCmVXA (4.91)
where Cm is the overall molar concentration and XA is the mole fraction of species A in the
mixture.
Mass convection may be due to the bulk motion of the carrier gas or may be associated
with the drift of the solute through the carrier gas as a result of net forces applied directly
to the solute (e.g., centrifugation due to centrifugal force). Here, we discuss only the
convection due to bulk motion of the carrier gas.
For a binary mixture, the convective molar flux, Jc, for species A can be expressed as
7cA = XA(NA + NB) (4.92)
4.5 /Mass Transfer 157
where NA and A^ are the molar flux for species A and B, respectively. Here, the molar flux
NA or A^B comprises both the diffusive flux and the convective flux. Thus, for species A,
NA can be expressed, from Eqs. (4.91) and (4.92), by
NA = JA + JcA = -DABCmVXA + XA(NA + AW (4.93)
For multicomponent mixtures, the overall molar flux for species i in the mixture can be
expressed by
i + Xi J2 Ni ( 4 - 94 )
where Dim is the molecular diffusivity of species i in the mixture. VX{ can be expressed
by the Stefan-Maxwell equation as [Taylor and Krishna, 1993]
i7V
J " xiNi> (4 95)
'
The primary interest in the mass transfer occurring in gas-solid flow concerns the intra-
particle diffusive mass transfer and interfacial convective mass transfer between the particle
and the gas medium. In both cases, Fick's equation, i.e., Eq. (4.91), or a more general form
of the flux equation, i.e., Eq. (4.93), can be applied. The basic formulation accounting for
the mass transfer phenomenon for a single particle resembles that for heat transfer. Hence,
the following discussion is focused on the analogy between the mass and heat transfer
concerning a single particle.
4.5.2 Mass and Heat Transfer Analogy

In an analysis of mass (or molar) diffusion and mass (or molar) convection of a
single particle, the resulting equations are in a form similar to those of heat conduction
and thermal convection in a single particle given in 4.2 and 4.3. More precisely, the
resulting equations for heat transfer between fluid and a particle given in 4.2 and 4.3
can be applied analogously to mass transfer by simply replacing the principal variables
including the thermal diffusivity (see Eq. (4.20)), Prandtl number, and Nusselt number
with the molecular diffusivity, Schmidt number, and Sherwood number, respectively. The
analogy between heat and mass transfer for a single particle, however, is generally valid
only under the following conditions:
(1) Constant physical properties
(2) Small rate of mass transfer
(3) No chemical reactions in the fluid
(4) No viscous dissipation
(5) No emission or absorption of radiant energy
(6) No pressure diffusion, thermal diffusion, or forced diffusion (e.g., electrophoresis)
In flow situations, empirical analogies between mass and heat transfer are usually em-
ployed. For single-particle mass transfer, the boundary layer analysis for mass transfer is
similar to that for heat transfer and thus is used for typical applications such as sublimation
of a solid (e.g., naphthalene ball) or evaporation of a liquid drop falling in air. For a single
sphere of diameter dv moving in a fluid, in terms of a boundary layer analysis analogous
to heat transfer, the convective mass transfer coefficient, kc, accounting for fluxes from the
surface of a single sphere to the bulk, or vice versa, is given by Froessling (1938) as
Sh = Ê = 2.0 + 0.6 Re^Sc 1 / 3 (4.96)
where Sc is the Schmidt number, defined as v/DAB. The equation provides a good cor-
relation of data at low Re p, but Sh calculated for high Rep is lower than the values found
experimentally. Equation (4.96) is practically the same as Eq. (4.40) when the heat and
mass transfer analogy is applied, i.e., Nup replaced with Sh and Pr replaced with Sc.
The conditions for negligible convection compared to diffusion for mass transfer analo-
gous to heat transfer are [Rosner, 1986]
(Re pSc) 1/2 1 (4.97)
and
(Ra m Sc) 1/4 1 (4.98)
where Ram is the Rayleigh number for mass transfer defined by
Ram = |s0w(tt>Afs ~ Â,OO) | | | = GrSc (4.99)
where L is the characteristic length; Gr is the Grashof number; coA,S and &>A,OO are the
mass fractions of species A on the solid surface and in the bulk, respectively; and /?w is
the parameter defining the variation of local gas density with mass composition at constant
temperature and pressure, i.e.,
fa = - (.?-) (4.100)
P \d(oj
Equation (4.97) is usually satisfied for mass transfer in the gas phase, such as transport
of oxygen or hydrogen to the outer surface of a small catalyst particle which is stagnant
relative to the surrounding gas. In this case, when Eq. (4.98) is also satisfied, Sh of 2 is a
good approximation, which is analogous to the heat transfer counterpart given in Eq. (4.15).
Equation (4.97) is especially important when there is simultaneous heat transfer and mass
transfer.
These two conditions (Eqs. (4.97) and (4.98)) are usually sufficient for assuming the
medium as quiescent in dilute systems in which both GJA,S and &>A,OO are small. However,
in nondilute or concentrated systems the mass transfer process can give rise to a convection
normal to the surface, which is known as the Stefan flow [Taylor and Krishna, 1993].
Consider a chemical species A which is transferred from the solid surface to the bulk
with a mass concentration &>A,OO- When the surface concentration <Â,S is high, and the
carrier gas B does not penetrate the surface, then there must be a diffusion-induced Stefan
convective outflux, which counterbalances the Fickian influx of species B. In such situations
the additional condition for neglecting convection in mass transport systems is [Rosner,
1986]
Km 1 (4.101)
Nomenclature 159
where
(4.102)
In very concentrated systems, the increased contribution of the Stefan flow leads the diffu-
sional mass transfer rate corresponding to the Sh for quiescent systems to be modified as
[Rosner, 1986]
Shs = 2 (4.103)
where Shs corresponds to the mass transfer coefficient with the Stefan flow. The Stefan
reduction factor for mass transfer, Spm* is given as
Shs Km)
SFm = = (4.104)
Sh Km
The Stefan convection, however, does not alter the analogy between heat and mass transfer
because the laws governing the change in the Nusselt number with the Stefan flow are
identical to those governing Shs.
Nomenclature
A Constant, defined by Eq. (4.66) in a binary mixture
and Eq. (4.68) A Thermal diffusivity
A Function of the index of Am Molecular diffusivity of species /
refraction in a mixture
An Constant, defined by Eq. (4.56) dp Particle diameter
a Particle radius Fo Fourier number
B Constant, defined by Eq. (4.66) Fo' Modified Fourier number, defined
and Eq. (4.68) byEq. (4.51)
Bi Biot number Fo; Fourier number, based on contact
Constant, defined by Eq. (4.56) time and maximum contact radius
C Geometric cross section Gr Grashof number for mass transfer
C Correction factor, defined by Eq. h Convective heat transfer
(4.32) coefficient
Ca Cross section for absorption h Length-averaged heat transfer
Ce Cross section for extinction coefficient
cm Overall molar concentration hc Collisional heat transfer
Cs Cross section for scattering coefficient
c Specific heat of particles hp Particle convective heat transfer
c Interparticle clearance coefficient
j
Specific heat of gas at constant I Radiant intensity

cp
pressure Planck intensity function
Cpe Effective thermal capacity of a JA Diffusive molar flux of species A
mixture JA Diffusive mass flux of species A
Molecular diffusivity of a species JcA Convective molar flux of species A
Particle Reynolds number, based

h Heat flux
on particle diameter and relative
Heat flux at the interface
Radiant heat flux velocity
Jr
K Thermal conductivity of gas r Cylindrical coordinate
Ke Effective thermal conductivity of rc Radius of contact area
a mixture rCm Maximum contact radius
Km Parameter, defined by Eq. S Scattering phase function,
(4.101) defined by Eq. (4.74)
KP Thermal conductivity of S Surface area of a particle
particles Spir Stefan reduction factor for mass
kc Convective mass transfer transfer
coefficient Sc Schmidt number
L Characteristic length Shs Sherwood number in presence of
I Optical depth Stefan flow
I Characteristic length T Absolute temperature
h Characteristic optical path length Ti
Temperature of colliding sphere 1
m Mass of a particle T2 Temperature of colliding sphere 2
N Number of components in the T
P Temperature of particle
mixture Tpo Reference particle temperature
NA Overall molar flux of species A Tvi Initial particle temperature
Nu Nusselt number distribution function
Nup Particle Nusselt number Too Temperature of the surrounding
n{ Index of refraction environment when R > oo
n Particle number density t Time
Pe Peclet number Contact time duration
Pr Prandtl number u Gas velocity
P Pressure U Pseudocontinuum one-phase
Ga Efficiency factor for absorption velocity
Qc Total heat exchange per impact j j
Particle velocity
Efficiency factor for extinction Upt Particle terminal velocity
2e
Qs Efficiency factor for scattering X Mole fraction
2s,D Efficiency factor for dependent
scattering Greek Symbols
Gs,i Efficiency factor for independent
scattering a Volume fraction of the gas phase
R Spherical coordinate Up Volume fraction of the particle
R Radius of a cylindrical tube phase
R* Dimensionless spherical Absorptivity for thermal radiation
coordinate P Scattering angle
Ram Rayleigh number for mass P Constant, defined by Eq. (4.47)
transfer Pw Parameter, defined by Eq. (4.100)
Re Reynolds number y Parameter, defined by Eq. (4.83)
References 161
8 Thickness of gas surrounding a absorption

particle Monochromatic coefficient for
Emissivity of the surface extinction
r) Constant, defined as Inarii/k Monochromatic coefficient for
0 Incident angle scattering
k Wavelength Tpd Particle thermal diffusion time,
km Wavelength at which the defined by Eq. (4.50)
monochromatic emissive power tps Particle-heating surface
of a blackbody is maximum residence time, defined by Eq.
kn Eigenvalue (4.50)
fi Dynamic viscosity of gas 0 Phase function
fi Parameter, defined by Eq. (4.77) 0 Temperature distribution

function, defined by Eq. (4.23)
v Kinematic viscosity
X Angle, defined by Eq. (4.59)
Constant, defined as 2na/k
O)A Mass fraction of species A
Parameter, defined by Eq. (4.24)
Mass fraction of species A at the
p Density of gas
solid surface
p Reflectivity of the surface
Mass fraction of species A in the
pm Density of a mixture bulk
pp Density of particles Parameter, defined by
<7b Stefan-Boltzmann constant Eq. (E4.12)
Gax Monochromatic coefficient for
References
Ahluwalia, R. K. and Im, K. H. (1994). Spectral Radiative Heat-Transfer in Coal Furnaces Using a
Hybrid Technique. J. Institute of Energy, 67, 23.
Bayvel, L. P. and Jones, A. R. (1981). Electromagnetic Scattering and Its Applications. London:
Applied Science.
Bohren, C. F. and Huffman, D. R. (1983). Absorption and Scattering of Light by Small Particles.
New York: John Wiley & Sons.
Boothroyd, R. G. (1971). Flowing Gas-Solids Suspensions. London: Chapman & Hall.
Brenner, H. (1963). Forced Convection Heat and Mass Transfer at Small Peclet Numbers from a
Particle of Arbitrary Shape. Chem. Eng. Sci., 18, 109.
Brewster, M. Q. and Tien, C. L. (1982). Radiative Transfer in Packed/Fluidized Beds: Dependent
Versus Independent Scattering. Trans. ASME, J. Heat Transfer, 104, 573.
Carslaw, H. S. and Jaeger, J. C. (1959). Conduction of Heat in Solids, 2nd ed. Oxford: Oxford
University Press.
Chandrasekhar, S. (1960). Radiative Transfer. New York: Dover.
Fiveland, W. A. and Jamaluddin, A. S. (1991). Three-Dimensional Spectral Radiative Heat Transfer
Solutions by the Discrete-Ordinate Method. J. Thermophysics, 5, 335.
Froessling, N. (1938). The Evaporation of Falling Drops. Gerlands Beitr. Geophys., 52, 170.
Gel'Perin, N. I. and Einstein, V. G. (1971). Heat Transfer in Fluidized Beds. In Fluidization. Ed.
Davidson and Harrison. New York: Academic Press.
Glicksman, L. R. and Decker, N. (1982). Heat Transfer from an Immersed Surface to Adjacent
Particles in a Fluidized Bed: The Role of Radiation and Particle Packing. Proceedings of
the Seventh International Heat Transfer Conference, Miinchen, Germany. Ed. Grigull,
Hahne, Stephan, and Straub, 2, 24.
Gupta, R. P., Wall, T. F. and Truelove, J. S. (1983). Radiative Scatter by Fly Ash in
Pulverized-Coal-Fired Furnaces: Application of the Monte Carlo Method to Anisotropic
Scatter. Int. J. Heat & Mass Transfer, 26, 1649.
Hottel, H. C , Saroflm, A. R, Dalzell, W. H. and Vasalos, I. A. (1971). Optical Properties of
Coatings: Effect of Pigment Concentration. AIAA J., 9, 129.
Hughmark, G. A. (1967). Mass and Heat Transfer from Rigid Spheres. AIChEJ., 13, 1219.
Hunt, M. L. (1989). Comparison of Convective Heat Transfer in Packed Beds and Granular Flows.
In Annual Review of Heat Transfer. Ed. C. L. Tien. Washington, D.C.: Hemisphere.
Im, K. H. and Ahluwalia, R. K. (1984). Combined Convection and Radiation in Rectangular Ducts.
Int. J. Heat & Mass Trans., 27, 221.
Kronig, R. and Bruijsten, J. (1951). On the Theory of the Heat and Mass Transfer from a Sphere in a
Flowing Medium at Low Values of Reynolds Number. Appl. Set Res., A2, 439.
Love, T. J. (1968). Radiative Heat Transfer. Columbus, Ohio: Merrill.
Love, T. J. and Grosh, R. J. (1965). Radiative Heat Transfer in Absorbing, Emitting, and Scattering
Media. Trans. ASME, J. Heat Transfer, 87c, 161.
McAdams, W. A. (1954). Heat Transmission, 3rd ed. New York: McGraw-Hill.
Mie, G. (1908). Optics of Turbid Media. Ann. Phys., 25, 377.
Modest, M. F. (1989). Modified Differential Approximation for Radiative Transfer in General
Three-Dimensional Media. J. Thermophysics, 3, 283.
Ozisik, M. N. (1973). Radiative Transfer. New York: John Wiley & Sons.
Ranz, W. E. and Marshall, W. R. (1952). Evaporation from Drops. Chem. Eng. Prog., 48, 141.
Rosner, D. E. (1986). Transport Processes in Chemically Reacting Flow Systems. Stoneham, Mass.:
Butterworths.
Siegel, R. and Howell, J. R. (1981). Radiative Heat Transfer, 2nd ed. New York: McGraw-Hill.
Soo, S. L. (1990). Multiphase Fluid Dynamics. Beijing: Science Press; Brookfield, Vt: Gower
Technical.
Sun, J. and Chen, M. M. (1988). A Theoretical Analysis of Heat Transfer due to Particle Impact. Int.
J. Heat & Mass Transfer, 31, 969.
Sykes, J. B. (1951). Approximate Integration of the Equation of Transfer. Monthly Notes Royal
Astronomical Society, 111, 377.
Taylor, R. and Krishna, R. (1993). Multicomponent Mass Transfer. New York: John Wiley & Sons.
Tien, C. L. and Drolen, B. L. (1987). Thermal Radiation in Particulate Media with Dependent and
Independent Scattering. In Annual Review of Numerical Fluid Mechanics and Heat
Transfer. Ed. T. C. Chawla. Washington, D.C.: Hemisphere.
Van de Hulst, H. C. (1957). Light Scattering by Small Particles. New York: John Wiley & Sons.
Zabrodsky, S. S. (1966). Hydrodynamics and Heat Transfer in Fluidized Beds. Cambridge, Mass:
MIT Press.
Problems
4.1 Solve the one-dimensional, unsteady conductive heat transfer equation for a homogeneous
solid sphere. (Hint: (1) Let 0 = R*T* so that
(P4.1)
3R*2 9Fo * ;
Problems 163
with
0 = 0 at R* = 0
+ (Bi - 1)9 = 0 at R* = 1 (P4.2)

9/v*
0 = i(R*) forFo = 0
(2) Use the separation of variables to obtain the solution.)
4.2 For a general forced convection system, prove that Nu = /(Re, Pr).
4.3 For a dilute suspension of particles in creeping flows (Rep <$C 1), compare the difference in
convective heat transfer coefficient in Eq. (4.40) and Eq. (4.42). It is assumed that Pr = 0.7.
4.4 Show that, using Eq. (4.58), the ratio of forward scattering to backward scattering is
(rii l)2/(rii + I) 2 for specularly reflecting spheres.
4.5 Consider the cases where very hot particles are suddenly exposed to a cold environment.
Derive an expression to account for the transient behavior of the particle temperature using
Eq. (4.62). Assume that Tp > Tw and the cooling is dominated by thermal radiation.
4.6 Consider a solid sphere in free fall passing through a heated chamber. The particle is
a 500 fim alloy of density 4,000 kg/m3, thermal capacity 150 J/kg- K, and absorptivity
0.6. The initial temperature of the particle is 300 K. The temperature of the gas inside the
chamber is 900 K, and the temperature of the walls is 1,000 K. Determine the temperature
of the particle at the chamber outlet if the length of the chamber is 0.3 m. Also, determine
the ratio of the heat acquired by the particle as a result of thermal radiation to that due to
heat convection. The physical properties of the gas are given as fi = 4 x 10~5 kg/m s,
p = 0.4 kg/m3, Pr = 0.7, and K = 0.067 W/m K. The particle falls at its terminal velocity.
4.7 A spherical benzoic acid particle with a diameter of 2 mm is dropped into a quiescent air duct
at ambient conditions (25C). During the fall, sublimation of the particle occurs. Taking
the molecular diffusivity of benzoic acid in air as 4 x 10~6 m2/s, calculate the convective
mass transfer coefficient when the particle is falling at its terminal velocity condition.
CHAPTER 5
Basic Equations
5.1 Introduction
Developing a mathematical model for general description of a particulate multi-
phase flow system has long been a challenging issue for researchers in the field. Although
at this stage comprehensive constitutive relationships for modeling have not been fully
established, several approaches which have proved useful are available for predicting the
gas-solid flow behavior, particularly for engineering application purposes. Among them
are the Eulerian continuum approach, Lagrangian trajectory approach, kinetic theory mod-
eling for interparticle collisions, and Darcy's law and the Ergun equation for flows through
packed beds. These approaches are the most commonly used and are introduced in this
chapter.
The Eulerian continuum approach, based on a continuum assumption of phases, provides
a "field" description of the dynamics of each phase. The Lagrangian trajectory approach,
from the study of motions of individual particles, is able to yield historical trajectories of
the particles. The kinetic theory modeling for interparticle collisions, extended from the
kinetic theory of gases, can be applied to dense suspension systems where the transport in
the particle phase is dominated by interparticle collisions. The Ergun equation provides
important flow relationships, which are useful not only for packed bed systems, but also for
some situations in fluidized bed systems.
5.1.1 Eulerian Continuum Approach

Evaluating the performance of a gas-solid transport system usually requires a
means of macroscopic "field" description of the distribution of basic flow properties such
as pressure, mass fluxes, concentrations, velocities, and temperatures of phases in the
system. To conduct such an evaluation, the Eulerian continuum or multifluid approach is
usually the best choice among the available approaches.
The Eulerian continuum approach is basically an extension of the mathematical for-
mulation of the fluid dynamics for a single phase to a multiphase. However, since neither
the fluid phase nor the particle phase is actually continuous throughout the system at any
moment, ways to construct a continuum of each phase have to be established. The transport
properties of each pseudocontinuous phase, or the turbulence models of each phase in the
case of turbulent gas-solid flows, need to be determined. In addition, the phase interactions
must be expressed in continuous forms.
The most fundamental step in constructing a continuum is to select a statistically valid
"fluid element" of the continuum. The fluid element needs to be selected in such a way that
it contains sufficient particles so that the fluid element possesses statistical thermodynamic
properties such as temperature and density, as well as velocity. At the same time, the fluid
element must be small enough compared to the characteristic length of the system so that,
164
from a macroscopic point of view, the concept of a continuous phase offluidcan be adopted.
Volume averaging is commonly applied to express the pseudocontinuum properties of each
phase in a multiphase flow. In order that the volume averaging is of statistical significance
and yet the control volume in the averaging is small enough to satisfy the pseudocontinuum
condition, there exists a minimum control volume for the volume averaging. To account for
the turbulence effects, time averaging needs to be adopted after the volume averaging [Soo,
1989]. Since the pseudocontinuum approach is based on the concept of volume averaging,
the pseudocontinuum phase in a gas-solid flow should be regarded as compressible because
of the nonuniform behavior of the volume fraction distribution of the phases even though
the gas or particle itself may be incompressible.
The transport properties of gases in a single phase, laminarflowcan be described in terms
of the kinetic theory of gases. Particles in a gas-solid flow may be treated as "amplified
molecules" so that a direct analogy of their behavior to gas molecules can be postulated.
It is noted that, in the kinetic theory, the transport coefficients of gases are derived from
the Maxwell-Boltzmann velocity distribution, which is based on Boltzmann statistics and
the possible states governed by Schrodinger's wave equation. Since the interparticle forces
in gas-solid systems can be quite different from the intermolecular forces in gas systems,
the validity of the kinetic theory in describing the transport coefficients of particles needs
to be further assured. For single-phase turbulent flows, the transport properties can be
determined by turbulence models; these include the widely adopted mixing length model
and k- model. In multiphase continuum modeling for gas-solid flows, these two models,
with some modifications accounting for the compressibility of the pseudocontinuum phase
and phase interactions, can be extended to illustrate the turbulence in the gas phase. In
this chapter, the particle-particle interaction is described by using particle collision models
while the particle-turbulence interaction is illustrated by using the Hinze-Tchen model and
the k--kv model.
For the Eulerian continuum modeling discussed in this chapter, it is assumed that the
basic form of the Navier-Stokes equation can be applied to all phases. For some dense
suspension cases, the particle phase may behave as a non-Newtonian fluid. In these cases,
a simple extension of the Navier-Stokes equation may not be appropriate.
5.1.2 Lagrangian Trajectory Approach

The Lagrangian approach, also known as trajectory modeling, provides a direct
description of particulate flows by tracking the motion of individual particles. In trajectory
modeling, the motion of a particle can be expressed by ordinary differential equations in
Lagrangian coordinates. Consequently, the particle velocity and the corresponding particle
trajectory can be solved by directly integrating the Lagrangian equation of the particle.
Therefore, the Lagrangian approach is convenient to reveal the discrete and transient nature
of the motion of particles.
Typical Lagrangian approaches include the deterministic trajectory method and the
stochastic trajectory method. The deterministic trajectory method neglects all the turbulent
transport processes of the particle phase, while the stochastic trajectory method takes into
account the effect of gas turbulence on the particle motion by considering the instantaneous
gas velocity in the formulation of the equation of motion of particles. To obtain the statistical
166 51 Basic Equations
average of the variables in the equations, statistical computations based on techniques such
as the Monte Carlo method need to be performed.
To solve the Lagrangian equation for a moving particle, the dynamic behavior of the gas
phase and other particles surrounding this moving particle should be predetermined. In this
regard, the behavior of the gas phase can be simulated by the Eulerian method. Since the gas
flow and the motion of all particles are coupled through gas-particle and particle-particle
interactions, a large number of iterations are required in order to obtain converged solutions
for all phases.
The development of the Lagrangian models has been limited mainly by the inherent need
for large computing capacity to carry out statistical averaging and computation of the phase
interactions. The Lagrangian models are particularly applicable to very dilute or discrete
flow situations for which multifluid models are not appropriate, or to situations in which
the historic tracking of particles is important (such as in pulverized coal combustion in a
furnace or the tracking of radioactive particles in gas-solid flows).
5.1.3 Kinetic Theory Modeling for Interparticle Collisions

When the gas-solid flow in a multiphase system is dominated by the interparticle
collisions, the stresses and other dynamic properties of the solid phase can be postulated
to be analogous to those of gas molecules. Thus, the kinetic theory of gases is adopted in
the modeling of dense gas-solid flows. In this model, it is assumed that collision among
particles is the only mechanism for the transport of mass, momentum, and energy of the
particles. The energy dissipation due to inelastic collisions is included in the model despite
the elastic collision condition dictated by the theory.
The model yields a set of hydrodynamic equations for the solid phase. For equation
closure, additional constitutive relations, which can be obtained by using the kinematic
argument of the collision and by assuming the Maxwellian velocity distribution of the
solids, are needed. Two examples are given to illustrate the applications of this model in
this chapter.
5.1.4 Ergun Equation

The Ergun equation relates the pressure drop in a packed bed to the flow rate
and the properties of particle and gas. However, the application of this equation has been
extended beyond the limits of fixed bed systems since it was first formulated in 1951. Thus,
a detailed account of the origin of this equation is necessary.
The general behavior of viscous flow through a homogeneous porous medium can be
described by Darcy 's law. In order to estimate the pressure drop across a porous medium, it
is assumed that the general porous medium can be represented by an assemblage of channels
of various cross sections with a fixed length. The viscous energy loss at low gas flow rates
in this case is given by the Kozeny theory. When the gas flow rate increases, the effect of
inertia becomes important. The kinetic energy loss at high gas flow rates can be described
by the theory of Burke and Plummer. The general expression of the pressure drop for flow
through a fixed bed, as suggested by Reynolds, may be presented in an algebraic form as
the sum of the viscous and inertial contributions. The Ergun equation incorporates these
two contributions in a semiempirical form. The equation may not be applicable to turbulent
cases without further accounting for the turbulence effects.
5.2 I Modeling of Single-Phase Flows 167
5.1.5 Summary
In most engineering applications, the desired model is one that is simple, capable
of providing a full flow field description, and yet reasonably accurate. The continuum
model has such a simplicity feature and is widely used in spite of its crude continuum
assumptions for all phases involved. When the particle concentrations are so high that the
interparticle collisions become the dominant transport mechanism, the dynamic behavior
of the particulate flow may be reasonably modeled on the basis of the kinetic theory.
The kinetic theory can also be used in the formulation of the momentum transfer due
to interparticle collisions in continuum modeling of the particle phase. For a very dilute
multiphase suspension, the gas phase can be treated by the continuum approach while the
motion of the particles may be described by the trajectory model. For gas flow through a
packed bed or an unsuspended portion of a solids transport system, the Ergun equation can
be applied. The Ergun equation is also able to describe the limiting situation of a dense
suspension.
In this chapter, the continuum modeling of a single-phase flow is presented. The Eulerian
approach of multiphase flows using the theories of volume averaging and time averaging
is described. The Lagrangian method is introduced and the kinetic theory modeling for
collision-dominated dense suspensions is delineated. Finally, the development of the Ergun
equation is given.
5.2 Modeling of Single-Phase Flows

In this section, the governing equations and transport coefficients of a single-
phase flow are presented. First, we derive the general transport theorem for conservation
equations in continuum mechanics. The governing equations for mass, momentum, and
energy transport are then obtained by applying the general transport theorem. The kinetic
theory of gases, which provides the basis for the determination of transport coefficients, is
described. In a turbulent flow, the transport coefficients are approximated by turbulence
models. Hence, two commonly used turbulence models, the mixing length model and the
k- model, are introduced. Various boundary conditions are also presented.
5.2.1 General Transport Theorem and General Conservation

All conservation equations in continuum mechanics can be derived from the gen-
eral transport theorem. Define a variable F (t) as a volume integral over an arbitrary volume
v(t) inanr-space
F(t)= [ f(r,t)dv (5.1)
where f(r,t) is an integrable function representing a parameter such as mass, momentum,

or energy; t is time; and r is a position vector. In an r-space, r = (n, r 2 , r 3 ). To calculate
the rate of change of F, it is necessary to introduce a -space where the volume is fixed with
respect to time so that an interchange of differentiation and integration is possible for the
derivative of F. A transformation of the volume from r -space to -space can be expressed
as
dv = dn dr2 dr3 = / di d2 d3 = / dv0 (5.2)
where / is the Jacobian determinant, defined as

3(ri,r 2 ,r 3 )
/ ( }
It can be proved that (Problem 5.1)
^ = (V U)J (5.4)
where the velocity U is defined as

dr
U=- (5.5)
Therefore, we have
<5 6)
-
Note that d//df is the substantial derivative, which can be expressed by
d = d +u vf (5J)
i i -
The equation describing the general transport theorem can be obtained by substituting
Eq. (5.7) into Eq. (5.6) as
/ f(r,t)dv= [ (y- + V.fu)dv (5.8)

& Jv(t) Jv{t) \dt )
The rate of change of F can be found from the net flow of / across the closed surface
ofv(t) and the generation of / inside v(t). Denote r/r as the flux vector of / and O as the
rate of generation of / per unit volume. Thus,
d f f f
Odi; (5.9)
fdv = - / w-^dAH- /
where n isd*aJv JA
unit normal vector directed JV
outwardly so that the first term on the right-hand
side represents the net inflow of / across the closed surface A. Note that, according to
the Gauss theorem, a surface integral over a closed area can be transformed into a volume
integral, i.e.,
(5.10)
A - / Jv'
Combining the general transport theorem, Eq. (5.8), with Eqs. (5.9) and (5.10) gives
+ V fU + V f - 4>I dv = 0 (5.11)
dt J
Since the volume v is arbitrarily selected, the general conservation equation in a single
phase of fluid can be written in the form
(
+ V-/t/ + V-V- = 0 (5.12)
dt
5.2.2 Governing Equations

The conservation equations of mass, momentum, and energy of a single-phase
flow can be obtained by using the general conservation equation derived previously.
5.2.2.1 Continuity Equation (Conservation of Mass)

For the total mass conservation of a single-phase fluid, / represents the fluid density p.
ty represents the diffusional flux of total mass, which is zero. For flow systems without
chemical reactions, <l> = 0. Therefore, from Eq. (5.12), we have the continuity equation as
^ + V {PU) = 0 (5.13)
ot
5.2.2.2 Momentum Equation (Newton's Law)
For the momentum conservation of a single-phase fluid, the momentum per unit volume /
is equal to the mass flux pU. The momentum flux is thus expressed by the stress tensor
ifr = (pi r ) . Here p is the static pressure or equilibrium pressure; / is a unit tensor;
and r is the shear stress tensor. Since <$> = pf where/ is the field force per unit mass,
Eq. (5.12) gives rise to the momentum equation as
dpu
+ V (pUU) = - V p + V . r + pf (5.14)
dt
where
In Eq. (5.15), JJL' is the second viscosity or bulk viscosity, which reflects the deviation
of the averaged pressure from the pressure at equilibrium due to nonuniformity in the local
velocity distribution. <$ij is the Kronecker delta, y! is given by
(5 16)
*' = - w -
where pm is the averaged pressure. Except for the strong nonequilibrium regions such as
the shock wave region, it is assumed that /z' = 0.
5.2.2.3 Energy Equation (Conservation of Energy)

For total energy conservation, / represents the total energy per unit volume pE. The flux
and the generation rate of the total energy per unit volume come from three sources: the
heat flux vector Jq; the rate of heat generation per unit volume / e , including Joule's heating
and thermal radiation; and the rate of work done by the surface force and the field force.
Thus, the energy equation can be given as
^ + V (pUE) = -V'Up + V'(U-r) + pU'f-V-Jq + Je (5.17)

ot
where E = e + 1/2 U -U and e is the internal energy per unit mass.
The energy equation expressed in terms of internal energy becomes
^ + V (pUe) = - p V U + 0 - V Jq + / e (5.18)
ot
where <p is the dissipation function defined by
4> = V-(U-T)-U-(V'T) (5.19)
which represents the dissipation rate of energy per unit volume due to viscous effects. The
energy is dissipated in the form of heat. 0 can be proved to be always positive.
The energy equation expressed in terms of temperature is convenient for evaluating
heat fluxes. Let e = cyT with cv the specific heat at constant volume and T the absolute
temperature of the fluid. Assuming the heat flux Jq obeys Fourier's law, Eq. (5.18) takes
the form
+ V (pUcyT) = -pV U + 0 + V (KVT) + / e (5.20)

ot
where K is the thermal conductivity of the fluid.
5.2.3 Kinetic Theory and Transport Coefficients

A brief discussion of the kinetic theory of gases is essential to the understanding of
the fundamental assumptions or conditions used in the derivation of the transport coefficients
of the particle phase in the continuum modeling of gas-solid multiphase flows. When the
conserved quantities such as the number of molecules, total momentum of molecules, and
total kinetic energy of molecules (assuming elastic collisions) are initially nonuniformly
distributed in an enclosed volume, a tendency for these quantities to be moving toward
an equilibrium state is established. The rate of change of a conserved quantity toward
the equilibrium state is characterized by the transport coefficient. Specifically, a transport
coefficient is defined as the proportionality constant of the transport flux to the density
gradient of the conserved quantity. The transport coefficients of a gas can be evaluated,
from the kinetic theory, in terms of the microscopic properties of the fluid. The rigorous
microscopic theory of transport processes can be represented by the Boltzmann transport
equation. However, because of the complexity of this equation, it is omitted here. Instead, we
adopt a phenomenological approach to derive the transport coefficients using the elementary
kinetic theory and probabilistic concept.
In the kinetic theory of gases, the molecules are assumed to be smooth, rigid, and elastic
spheres. The only kinetic energy considered is that from the translational motion of the
molecules. In addition, the gas is assumed to be in an equilibrium state in a container where
the gas molecules are uniformly distributed and all directions of the molecular motion
are equally probable. Furthermore, velocities of the molecules are assumed to obey the
Maxwell-Boltzmann distribution, which is described in the following section.
5.2.3.1 Maxwell-Boltzmann Velocity Distribution

Consider a system that comprises numerous identical, independent, and indistinguishable
particles in equilibrium in a container at an absolute temperature T. If there is no limit on the
number of particles occupying a given energy level, the probability of finding n\ particles
with energy level ex in such a system can be described by Bose-Einstein statistics [Reif,
1965]. Denote gx as the degeneracy of the energy level ex (i.e., the number of the quantum
states having the same or nearly the same energy). In Bose-Einstein statistics, the number
of possible arrangements for the \th level of energy for any nx, WÊ, is
The total number of possible arrangements for a certain set of nx in the system, WBE> is
therefore obtained as
(5-22)
In a special case where g-x ^ nx, Eq. (5.22) can be reduced to the corrected Maxwell-
Boltzmann statistics (note: the classical Maxwell-Boltzmann statistics is for distinguishable
particles) as
wMB = wBE = 1r 1r ^- (5.23)

n{\
When the gas in a container is in an equilibrium condition, it can be shown (Problem 5.2)
that Eq. (5.23) can be expressed as
n- g\e~kT
- = - (5.24)
n z
where n is the total number of particles in the system; k is the Boltzmann constant; and z
is the partition function, defined as
The possible energy levels are determined by Schrodinger's wave equation [Reif, 1965].
For translational motion of a particle, the wave equation takes the form
V2* = - ^ e * (5.26)
hz
where h is Planck's constant, m is the mass of the particle, and ^ is the wave function.
6 is the eigenvalue of Eq. (5.26), which represents the energy as expressed by
^2/^2 2 2 \
e= + + (5 27)
8^z| 4 4J -
where kx,ky, and kz are the integer quantum numbers; and Lx, Ly, and L z are the geometric
dimensions in the x-, y-, and z-directions, respectively. Moreover, the partition function z
can be obtained as
z = V e ~ * = ^-(2nmkT)32 (5.28)
^-^ h5
where V is the volume of the container, which is equal to LxLyLz.
If we consider a cubic system with a characteristic dimension L and assume that the
energy distribution is continuous, we have
h2r2
where r is the radius in a space of quantum numbers (note that only the first octant is
meaningful since all the quantum numbers should be positive). As a result, the degeneracy
dg between and + de can be expressed as
dg = -r2dr = 2n (- ) (2m) U de (5.30)

2 \h )
Substituting Eqs. (5.28) and (5.30) into Eq. (5.24) yields

2
= . e ~ M d (5.31)
With 6 = l/2mu 2 ,we have finally obtained the Maxwell-Boltzmann velocity distribution
as
(5.32)
where dP represents the probability of particles having velocity between v and (v + dv).
5.2.3.2 Collision Frequency

The molecular collision is the basic mechanism that governs the transport processes in
gases. Thus, it is necessary to examine the collision frequency and collision cross section
before quantifying the transport coefficients. First consider the probability that a collision
occurs between r and r + dr. Denote X as the mean free path and P(r) as the probability
of no collisions within r. Hence, we have
(5.33)
The Taylor series expansion of P(r + dr) yields
<">
With P(0) = 1, it is found that the probability of no collisions within the length of path r
follows the Poisson distribution as
P(r)=e-i (5.35)
Now, consider a gas with two different types of particles in a container. The radius of
the sphere of influence for collision between the two different particles is
dx+d2
rn = - (5.36)
where subscripts 1 and 2 denote the two different types of particles.

The average relative velocity between particles 1 and 2 can be evaluated by
(vn) = / dui / dv2FlF2\vl - v2\ (5.37)

- / * * /
where (t>i2) is the averaged relative velocity and F is the velocity distribution function.
Assume that the motion of each type of particle obeys the Maxwell-Boltzmann distribution.
Then, (i?i2) can be obtained as (Problem 5.3)
(5.38)
The collision frequency per unit volume between different types of particles is obtained as
/12 = nin27rr^2(vn) = nxn2(dx +d2f\f\ ^^ (5.39)

V 2mm
which can be easily extended to yield the collision frequency among the same type of
particles as
/n = \n\nr2n{vu) = 2n\dh / (5.40)

2 y m\
where the factor 1/2 is to account for the overestimation of the collisions of identical
particles.
5.2.3.3 Transport Coefficients

The coefficients of transport properties considered here include the viscosity, diffusivity, and
thermal conductivity of a gas. The transport coefficients vary with gas properties if the flow
is laminar. When the flow is turbulent, the transport coefficients become strongly dependent
on the turbulence structure. Here we only deal with the laminar transport coefficients; the
discussion of the turbulent transport coefficients is given in 5.2.4.
A. COEFFICIENT OF VISCOSITY
Consider a cubic unit volume containing n particles in a Cartesian coordinate system. On
average, about n/6 particles move in the + v-direction, and the same number of particles
move in the other five directions. Each particle stream has the same averaged velocity
(v). Since particle collision is responsible for the momentum transport, the averaged x-
component of the particle momentum transported in the y -direction may be reasonably
estimated by
1
xy = -n(v)[mUx(y - A) - mUx(y + A)] (5.41)
6
where rxy is shear stress in the x-y-plane; A is the mean free path; m is the mass of the
particle; and Ux(y-X) represents the averaged *-component of velocity at the position (y-A)
along the y-axis. Using the Taylor expansion, Eq. (5.41) becomes
1 3U
/ v i
rxy = --n(v)mX-^ (5.42)
3 dy
Thus, the transport coefficient of viscosity, denoted by /x, is
\i = -n(v)mX (5.43)
B. COEFFICIENT OF SELF-DIFFUSION (DIFFUSIVITY)

The diffusivity can be obtained by using the same method as that for the viscosity. The
transported quantity in this case is the number of particles instead of the momentum of
particles. The net flux of particles in the v-direction may be expressed as
Jy = \(v)[n(y - A) - n(y + A)] (5.44)

6
where n(y-X) represents the particle number density at the position {y-X) along the v-axis.
From the Taylor expansion, Eq. (5.44) has the form
Jy = -\(v)k^ (5.45)
3 d
which gives the coefficient of self-diffusion D as
D=l-{v)X (5.46)
C. THERMAL CONDUCTIVITY
On the basis of the same approach, we can obtain an expression for the thermal conductivity
based on the energyflux.The net energy transported in the y-direction can be estimated from
qy = l-n(v)[e{y - X) - e(y + X)] = ~\n{v)Xy (5.47)
where e is the averaged energy of a particle. Since e is related to the absolute temperature
T, we have
Be de dT dT
= = c (5.48)
dy dT dy dy
where c is the specific heat of the particle. Therefore, the thermal conductivity K is in the
form
K = -n(v)cX (5.49)
D. TRANSPORT COEFFICIENTS BASED

ON MAXWELL-BOLTZMANN DISTRIBUTION
The averaged velocity (v) can be expressed in terms of Eq. (5.32) as
i i I
f dn (2y(my f 3 _nUpL SkT
(v)= v = - / v 5e *T dv = \ (5.50)
Jo n \ir J \kT) Jo V nm
Using Eq. (5.40), the mean free path can be estimated as
(5.51)
Hence, the transport coefficients can be given as

2 V^f
<5 52)
'
D = * * 1J K (5 .53)
3 nnd V nm
(5.54,
m
5.2.4 Modeling for Turbulent Flows

The most widely adopted method for the turbulent flow analysis is based on time-
averaged equations using the Reynolds decomposition concept. In the following, we discuss
the Reynolds decomposition and time-averaging method. There are other methods such
as direct numerical simulation (DNS), large-eddy simulation (LES), and discrete-vortex
simulation (DVS) that are being developed and are not included here.
In using the Reynolds decomposition, closure of the time-averaged Navier-Stokes equa-
tions cannot readily be realized because of the unknown correlation terms such as turbulent
Reynolds stresses generated by time averaging. Thus, additional equations are needed to
correlate these terms with time-averaged quantities. These additional equations may come
from turbulence models. The two most commonly used turbulence models, the mixing
length model and the k-e model, are introduced.
5.2.4.1 Time-Averaged Navier-Stokes Equations

It is assumed that the instantaneous Navier-Stokes equations for turbulent flows have the
exact form of those for laminar flows. From the Reynolds decomposition, any instantaneous
variable, 0, can be divided into a time-averaged quantity and a fluctuating part as
0(O = * + 0'(O (5-55)
where
* = - / 0 ( 0 dt (5.56)
to Jo
The integral time period to should be short compared to the characteristic time scale of the
system. At the same time, to must also be long enough so that
i/V (0 dt = O (5.57)
*o Jo
Taking the time average over Eqs. (5.13) and (5.14), we can obtain the time-averaged
continuity equation and time-averaged momentum equation as
^7 + :(^J+^D=0 (5.58)
(U
<5 59)
-
where ~pu[u'} is defined as the turbulent Reynolds stress and the time-averaged stress tensor
tfij is
2 3J7 mo 2 3u' (du[ du\\

3 9xffl 3 dxm \dxi dxj
In Eq. (5.60), \i! denotes the viscosity fluctuation which may result from the temperature
fluctuation. In Eq. (5.60) the quantities with the same subscript m in the third and fourth
terms on the right-hand side of the equation indicate the Einstein summation.
For simplicity, consider the steady, incompressible, and isothermal turbulent flows where
d/dt = 0; pf = 0; and \x! 0. The continuity equation is thus simplified from Eq. (5.58)
and the momentum equation is reduced from Eq. (5.59) to

To render Eq. (5.62) solvable, it is necessary to provide an expression for the turbulent
Reynolds stress. For isotropic turbulent flows, similar to the transport processes in laminar
flows, a scalar turbulent viscosity /XT is defined using the Boussinesq formulation
It should be noted that the turbulent viscosity is a function of the turbulence structure
in addition to the physical properties of the gas. We may further introduce an effective
turbulent viscosity /xeff as
Meff = II + MT (5.64)
Thus, Eq. (5.62) can be expressed as
For nonisotropic turbulentflowssuch as strong swirlingflowsor buoyantflows,the turbulent

viscosity is a tensor instead of a scalar. Therefore, all six components of the Reynolds stress
need to be modeled individually.
To determine M-M] or /xT, various types of turbulence models have been proposed. A
turbulence model is characterized by a set of transport equations or constitutive relations to
simulate higher-order correlations using lower-order correlations or time-averaged quan-
tities. Currently, no general turbulence model is available since the physics of turbulence
is not fully understood. However, for nonswirling and nonbuoyant flows (isotropic turbu-
lence), the mixing length model and the k-e model can be used and are introduced in the
following. For nonisotropic turbulent flows, the second-order moment closure model in
the form of algebraic expressions or differential transport equations is the simplest [Zhang
et al., 1992; Zhou, 1993]. The model in the form of differential transport equations is,
however, too complex to be used for general engineering purposes. Specifically, it requires
solving six equations instead of two in the k-e model. In addition, eight empirical constants
need to be determined instead of the five in the k-e model. Furthermore, specifying the
boundary conditions for each Reynolds stress component is difficult. Hence, in this text,
we only discuss the isotropic turbulent flows.
5.2.4.2 Mixing Length Model
The mixing length model was originally proposed by Prandtl (1925). The model is based
on two analogies: One is between the laminar viscosity and the turbulent viscosity of gas;
the other is between the ratio of the averaged fluctuating velocity to the averaged fluctuating
length and the strain from the averaged velocity. Prandtl proposed that
(5.66)
+ (5.67)
Combining Eq. (5.67) with Eq. (5.66) yields
(5.68)
' = P'n +
where /m is the turbulent mixing length, which is determined empirically and depends on
the type of flow.
The mixing length model is simple with /XT defined by an algebraic expression. Hence,
for simple flows such as jet flows, flows of the boundary layer type, and pipe flows, / m
can be easily estimated. For flows with recirculation regions or with complex geometries,
the determination of /m may be impossible. The selection of lm for various simple flows
can be found in many references [e.g., Launder and Spalding, 1972]. However, the mixing
length model indicates that the effect of turbulence will vanish where the strain, based on
the averaged velocity, is zero (e.g., near the end of a recirculation region or near a symmetric
axis). This is not always true.
5.2.4.3 k-e Model

If we assume that the local state of the turbulent flow depends on some turbulent quan-
tities governed by the corresponding transport equations, how many independent scaling
quantities are required? This question may be answered from the dimensional analysis.
To simulate the turbulent Reynolds stress, first we examine the dimensions of the
Reynolds stress, the rate of strain, and the turbulent viscosity as follows
[M]
; M
(5.69)
where [M] denotes the mass unit; [L] represents the length unit; and [T] is the time unit.
In order to correlate the Reynolds stress with the rate of strain, a length scale and a time
scale or any two independent combinations of the length and time scales (with ~p as the
mass scale) are needed to characterize the local turbulence. That is, two more governing
equations are required to describe the transport of these two scaling quantities.
A common choice for the two scaling parameters is the kinetic energy of turbulence k
and its dissipation rate e, defined as
k= ^ 6= â
l x
2 p axj dx}
From the dimensional analysis, it can be shown that k and e are the two independent scaling
parameters of time and length scales
Thus, the turbulent viscosity /JLJ may be related to k and e in the form
k2
MT = C^p (5.72)
where CM is an empirical constant.

For steady-state, incompressible, and isothermal flows, the transport equation of the
turbulent kinetic energy can be derived as (Problem 5.4)
(5J3)
(Uk) TW1 ? fo ? < $ ^1
178 5 / Basic Equations
The third term on the right-hand side of the equation represents the transport of k due to
molecular and turbulent diffusion. Hence, with an analogy to the laminar transport, this
term may be expressed as
(5.74)
where ak is an empirical constant. Thus, Eq. (5.73) takes the form
(pZ7j) = ( ) + (/xeff - fi)- ( - H ) - pe (5.75)

3XJ 3XJ V crk 3XJ J 3XJ \ d*j ox{ J
Equation (5.75) is called the ^-equation.
The transport equation for the rate of turbulent dissipation can also be derived as (Problem
5.5)
_ 3 aw; awj du[ du\

dx\ dx\ 3x[ 3XJ
. du[ d2u{
(5.76)
Similarly to the approach used for Eq. (5.74), the first term on the right-hand side of
Eq. (5.76) may be expressed as
3 _36 _-n _dpf 3wj 3 Z/Xgff 36

(5.77)
uX\ OX\ uX] uX] 3XJ \ ore 3XJ
where cre is an empirical constant. The next three terms on the right-hand side of Eq. (5.76)
may be simplified to the form
du'idu'j du\du[\dUi _du'iBu'jdu[ _ , du[

-2U
dx \ dx
dx\ dx dx dX dx
dx\ dx
dx\ dx ' dx\
-W, "-^+-P ^ (5.78)

where C\ is an empirical constant. The last term on the right-hand side of Eq. (5.76) may
be related to k and 6 by
-w^'- (5.79)
where C2 is an empirical constant. The final form of Eq. (5.76) thus becomes
(5.80)
k
Equation (5.80) is called the 6-equation.
For thefiveempirical constants in the k-e model, it is suggested by Launder and Spalding
(1972) that CM = 0.09; C\ = 1.44; C2 = 1.92; ak = 1.0; and a = 1.22.
52 A A Closure Equations
In order for a model to be closured, the total number of independent equations has to match
the total number of independent variables. For a single-phase flow, the typical independent
equations include the continuity equation, momentum equation, energy equation, equation
of state for compressible flow, equations for turbulence characteristics in turbulent flows,
and relations for laminar transport coefficients (e.g., \JL = f(T)). The typical independent
variables may include density, pressure, velocity, temperature, turbulence characteristics,
and some laminar transport coefficients. Since the velocity of gas is a vector, the number of
independent variables associated with the velocity depends on the number of components
of the velocity in question. Similar consideration is also applied to the momentum equation,
which is normally written in a vectorial form.
As an example, for steady, incompressible, and isothermal turbulent flows using the
k-e model, the independent equations are (1) the continuity equation, Eq. (5.61); (2) the
momentum equation, Eq. (5.65); (3) the definition of the effective viscosity, /zeff (combina-
tion of Eq. (5.64) and Eq. (5.72)); (4) the equation of turbulent kinetic energy, Eq. (5.75);
and (5) the equation for the dissipation rate of turbulent kinetic energy, Eq. (5.80). Thus,
for a three-dimensional model, the total number of independent equations is seven. The
corresponding independent variables are (1) velocity (three components); (2) pressure; (3)
effective viscosity; (4) turbulent kinetic energy; and (5) dissipation rate of turbulent ki-
netic energy. Thus, the total number of independent variables is also seven, and the model
becomes solvable.
5.2.5 Boundary Conditions

The solutions of the governing equations are subjected to the boundary conditions
specified. The number of boundary conditions required depends on the types of the gov-
erning equations, e.g., elliptical, parabolic, or mixed. Because of the complexity of the
nonlinear partial differential equations involved, there is no general form of boundary con-
dition that can guarantee the solvability of the equations. For single-phase turbulent flows,
the typical boundaries may include impermeable solid wall (the wall may be movable),
free surface, axis of symmetry, and inlet and outlet conditions. The common independent
variables whose boundary conditions are needed to obtain the solutions include velocity,
temperature, turbulent kinetic energy, and dissipation rate of the turbulent kinetic energy
(assuming the k- model is used for turbulence simulation). Since usually only the relative
pressure field is of interest and it can be obtained by solving the momentum and conti-
nuity equations, the discussion of the boundary conditions concerning pressure is omitted
here.
5.2.5.1 Impermeable Solid Wall

Assume that a no-slip condition of velocity at the solid wall boundary is valid. We have
U=0 (5.81)
The boundary condition of temperature may be in one of the three following forms:
(1) Given temperature distribution
T = Tw(ro) (5.82)
where r0 is the position vector on the wall.
(2) Given heat flux distribution
-K= qv/(r0) (5.83)

on
where n is a coordinate normal to the boundary.
(3) General balance of heat fluxes
-K= hiTco -T)+oe(7 - T 4) (5.84)
where e is the wall emissivity.

For turbulent flows using the k-e model, both k and e near the wall must be determined.
The most straightforward method is to apply experimentally established boundary condi-
tions, such as the logarithmic law of the wall, at some point near the wall [Launder and
Spalding, 1972]. For the near-wall region, the convection and diffusion of the turbulent ki-
netic energy are negligible. In most turbulent flows, the turbulent properties in the near-wall
region are functions only of the normal coordinate. Thus, Eqs. (5.72) and (5.75) yield
= - ^ = (5-85)
where rw is the turbulent shear stress near the wall.

For the rate of turbulent energy dissipation near the wall, from Eq. (5.72) we have
(5.86)
In order to give an approximation of the velocity gradient near the wall, an empirical
equation, known as the wall function, is introduced. The wall function is expressed as
(5.87)
where E and K are empirical constants (K is known as von Karman's constant). Thus,
from Eq. (5.85) and Eq. (5.87), we obtain the boundary condition of e near the solid wall,
i.e.,y = <5w,as
<= # (5.88,
where <5W is the normal distance of the point from the wall.
It is noted that turbulence may also affect the boundary conditions of velocity and heat
flux near the wall. A detailed discussion using the k-e model for these boundary conditions
is given by Launder and Spalding (1974).
5.2.5.2 Free Surface

For the velocity at a free surface, we have
dU
= 0; Un = 0 (5.89)
dn
where Un is the normal component of the velocity at the surface and n denotes the coordinate
normal to the free surface.
For temperature, the same form of Eq. (5.84) may be applied. For both k and e, it is
assumed that
= 0 (5.90)
and
= 0 (5.91)
5.2.5.3 Axis of Symmetry

On an axis of symmetry, we have the symmetric condition
= 0 (5.92)
where 0 can be U, T, k,or e, and r is the radial coordinate. The radial component of the
velocity is equal to zero.
5.2.5.4 Inlet Conditions

The inlet velocity profile and temperature profile are given as
V = Ui (5.93)
T = I- (5.94)
The inlet profiles of k and e are generally assumed.
5.2.5.5 Outlet Conditions

Several types of boundary conditions at the outlet can be specified.
A. FULLY DEVELOPED FLOW
If the flow becomes fully developed at the outlet, the partial derivatives for U, T, k, and
along the outlet direction are zero at the outlet.
B. LARGE PECLET NUMBER NEAR THE OUTLET
The Peclet number is defined as the product of the Reynolds number with the Prandtl
number. For a large Peclet number near the outlet, there is no upstream influence because
the flow is dominated by the downstream convection. In this case, no boundary condition
information is needed for the outflow boundary [Patankar, 1980].
C. FREE STREAM APPROXIMATION
The boundary conditions for k and e at the outlet may be represented by the limiting cases
of the k and equations in a free stream [Launder and Spalding, 1972]. In a free stream, all
terms containing derivatives with respect to the coordinates perpendicular to the streamline
are zero. Therefore, from the ^-equation, we have
* - *- <5-95)
dx /Ti
where /xxi is the turbulent viscosity at the outlet, and x is the coordinate along the streamline.
Similarly, for e at the outlet, we have
U~ = -C2p (5.96)
dx /x T i
5.3 Continuum Modeling of Multiphase Flows

Since a multiphaseflowusually takes place in a confined volume, the desire to have
a mathematical description based on a fixed domain renders the Eulerian method an ideal
one to describe the flow field. The Eulerian approach requires that the transport quantities
of all phases be continuous throughout the computational domain. As mentioned before, in
reality, each phase is time-dependent and may be discretely distributed. Hence, averaging
theorems need to be applied to construct a continuum for each phase so that the existing
Eulerian description of a single-phase flow may be extended to a multiphase flow.
5.3.1 Averages and Averaging Theorems

To construct a continuum for each phase, volume averaging and time averaging
are essential to the analysis of a multiphase flow where each phase has discrete spatial and
temporal distributions. Although the continuum for each phase may be constructed either
by volume averaging [Slattery, 1967a, 1967b; Whitaker, 1969; Delhaye and Archard, 1976]
in terms of the volume fraction or by time averaging in terms of the fractional residence
time [Ishii, 1975], the dynamic and thermodynamic properties of a mixture are cumulative
with the volume fraction, but not with the fractional residence time [Soo, 1989]. Moreover,
a priori time averaging based on the fractional residence time may eliminate the identity of
different dynamic phases because of the different corresponding velocities. Hence, volume
averaging is preferred for constructing a continuum for each phase while time averaging
may be carried out after volume averaging to account for the high-frequency fluctuations
[Soo, 1989].
5.3.1.1 Phase Average and Intrinsic Average

Using the subscript k to denote the transport quantities of phase k, for a control volume V
as shown in Fig. 5.1, the volume average of any quantity of phase k can be defined as
/ fadV (5.97)
V Jvv
k
Figure 5.1. Concept of volume average.

531 Continuum Modeling of Multiphase Flows 183
where rfr^ is a quantity of phase k ( ^ can be a scalar, vector, or tensor) and Vk is the volume
occupied by phase k inside the control volume V. (ifa) is termed the phase average of ^rk.
Besides the phase average, it is also important to introduce the intrinsic average, which
is defined by
=-[ (5.98)
Hence, the phase average can be related to the intrinsic average by the volume fraction of
phase k, o?k, as
(^> = y I ' ( ^ ) = a k l " ( ^ > (5.99)

Usually, the intrinsic average reflects the real physical property or quantity such as
density and velocity, while the phase average gives a pseudoproperty or quantity based on
the selection of control volume. Phase averages are used to construct the continuum of each
phase to which Eulerian description can be applied.
Note that volume averaging can only be applied to quantities per unit volume, such as
density, momentum per unit volume, and energy per unit volume. The intrinsic average of
the velocity of phase k may thus be defined as
^ (5.100)
and the intrinsic averaged internal energy of phase k is
Vk) = JT^ 77 / PkekdV = (5.101)
5.3.1.2 Minimum Control Volume of Phase Average

As mentioned, the pseudocontinuum approach can be used for the multiphase flow analysis
only when the control volume contains enough particles that the volume average is of
statistical significance, and at the same time, the control volume is small enough to ensure
the validity of the continuum assumption. Thus, in order to have a statistically meaningful
phase average, there exists a minimum averaging volume.
The minimum averaging volume of a statistically meaningful phase average can be
obtained from the study of the effects of control volumes on the computed volume fractions
of a particulate phase [Celmins', 1988]. Consider a three-dimensional hexagonal model as
an example to illustrate the method of analysis. The spherical particles are arranged with
a regular repeating pattern, as shown in Fig. 5.2. For convenience, a series of cocentered
spherical control volumes are used for volume averaging. In principle, the center of these
spherical control volumes can be at any arbitrary location inside a typical three-dimensional
hexagonal cell. Two special locations of the center, however, can adequately represent the
limiting cases of the volume average, i.e., one in the center of a sphere and the other in the
symmetric center of four neighboring spheres whose centers form a pyramid, as shown in
Fig. 5.2.
The computed volume fraction of particles, a*, for a given control volume in general
includes the volumes of fully and partially enclosed solid spheres inside the enclosure,
as indicated by the shaded area in Fig. 5.3. With increasing control volumes, the volume
fractions of solids obtained from different centered control volumes are close to the same
Pyramid
a Q o o o
Q )( O )( P/ P \
+ \ v / y + x, *
Q ^j \ Q__\ ._/_ _^ / Q
+ + + +
o o o o o
(b)
Figure 5.2. Three-dimensional hexagonal structure of spherical particles:

(a) Three-dimensional view; (b) Top view.
oooo oo o
o o-o o
oogo
OOOO (a)
ODO (b)
Figure 5.3. Top view of volume averaging over a hexagonal arrangement of spheres:
(a) Cocentrated with a sphere; (b) Centered in a pyramid of spheres.
asymptotic value a p , as shown in Fig. 5.4(a). This is the true ambient volume fraction of
solids. Consequently, given an error margin of relative deviation of the calculated volume
fraction of particles, 8 (defined as |a - a p |/a p ), the minimum radius of averaging volume
for statistically valid phase volume averages can be defined, as illustrated in Fig. 5.4(b). It
can be realized that in the previous computation, the volume fraction of solids for a given
type of particle arrangement depends on two geometric characteristic lengths, e.g., diameter
5.3 I Continuum Modeling of Multiphase Flows 185
5.0
Cocentered
Pyramid-centered
4.0
3.0
2.0
1.0
0.0
0.0 2.0 8.0 10.0
10.0
1
III
1
1 1 .
8.0
6.0
1 ,
4.0
1 1
= 1.0%
2.0
16
*~ 0.0 -
o
!
1
-2.0
,
1
-4.0
,
1
-6.0
,
1
1
-8.0
,
i i 1
-10.0
1
0.0 2.0 4.0 6.0 8.0 10.0

R/l
(b)
Figure 5.4. Determination of minimum radius of control volume for two different control
volume centers: (a) Effect of control volume on calculated particle volume fraction; (b) Effect
of control volume on relative deviation of calculated particle volume fraction.
of a solid sphere dv and center-to-center distance of two neighboring particles /. The average
interparticle distance can be further estimated from the particle number density, n, by
(5.102)
6 ar
Moreover, this expression suggests that l/dv / ( a p ) . Hence, the relationship of R/1 =
g(ctp) can be written as
= -f g( p) = f(ap)g(av) (5.103)
which means that Rm[n/dp is only a function of av for a given 8 and intersphere structure.
Thus, for a particulate phase with a uniform distribution of monodispersed spheres (includ-
ing the three-dimensional hexagonal structure which represents the maximum packing of
monodispersed spheres), Rm{n/dp versus av is given by [Celmins, 1988]
(5.104)
dp
Equation (5.104) provides the quantitative criterion for minimum radii of averaging vol-
umes to demarcate the applicable range of the pseudocontinuum approach in the particulate
multiphase flow analysis. The minimum averaging volume is mainly a function of particle
volume fraction and particle size. The typical range of the minimum radius of an aver-
aging volume for particle sizes from 50 to 1,000 /xm with 8 = 1 percent is illustrated in
Fig. 5.5. Equation (5.104) may be used to estimate the physically meaningful resolution
for phase distributions of a multiphase flow using the pseudocontinuum approach or to
estimate the minimum sampling volume for instantaneous optical measurements of local
particle concentrations.
Example 5.1 Consider the cube-shaped resolution domains in the evaluation of solutions
for gas-solid flows using the pseudocontinuum approach. The domain, having a side length
of AL, can be estimated by equating the volume of the domain to the minimum averaging
volume for a given particle size and volume fraction. Particle volume fractions and particle
diameters in three gas-solid systems are given in Table E5.1, in which Case 1 represents a
typical condition for dilute pneumatic transport; Case 2 characterizes a typical condition for
dispersion of dusty cloud; and Case 3 delineates a typical dense phase transport or fluidized
bed. For each case, estimate AL by using the Eq. (5.104) with 8 = 1 percent.
Table E5. 1. a p and dp for Three
Typical Cases
Case P(%) dp(/xm)
1 1 100
2 0.1 10
3 40 300
Solution By equating the volume of the resolution domain to the minimum

averaging volume for a given particle size and volume fraction, AL is related to R m[n by
3/?min = L6Rmin (E5A)

AL = ( Y )
50-
Figure 5.5. Typical range of minimum radius of control volume.
If ^min is evaluated from Eq. (5.104), the side lengths of the resolution domain for the
numerical simulation of the cases given in Table E5.1 are calculated to be 5.2 mm, 1.1 mm,
and 1.7 mm, respectively.
5.3.1.3 Volume-Averaging Theorems

Volume-averaging theorems are derived to express the volume averages over integrals or
derivatives in terms of the integrals or derivatives of the volume averages. For simplicity,
let us consider a point on an arbitrary continuous curve s in a multiphase mixture, as shown
in Fig. 5.6. With each point on the curve as its center, we can have a sphere over which the
volume averaging may be applied. The portion occupied by phase k is denoted as V^. V^ is
bounded by an interfacial area A^ and a surface area Ake, which is on the sphere centered at
the point s and is interfaced by phase k. The closed surface Ac (=Ak 4- Ake) is considered
to be moving by translation (without rotation) along this arbitrary curve. From the general
transport theorem, i.e., Eq. (5.8), the change of the integral over V as a function of s may
Figure 5.6. Volume averaging in a multiphase mixture.
be expressed, by replacing t with s, as
k (5.105)
& Jvk Jvk os JA
where nk is the outwardly directed normal for the closed surface Ak and Ake. Note that
is a function of time and space coordinates and is independent of s; therefore, we have
which yields
/ ir kdV = / ^ k -^.
^ W k dA+ / i/.Jk J (5.107)
as JVk JAAk as
as JJAke
Ake ds
ds
k
The interfacial area Ak is assumed to be smooth and continuous when the center of the
integral sphere shifts from s to s + ds so that dr/ds is a tangential vector on Ak(s). Thus,
the surface integral over Ak in Eq. (5.107) becomes
dr
"k = 0 (5.108)
ds
For the surface integral over Ake in Eq. (5.107), it is noted that Ake is not continuous when
the center of the sphere shifts from s to s + ds; in other words, Ake(s + ds) can be a
completely different surface from Ake(s). Hence, dr/ds is not a tangential vector on A ke(s)
or dr/ds nke is not zero. However, if r(s) is divided into ro(s) and P(s), where r 0 is the
position vector of the center of the sphere and P is a position vector relative to the center s,
and since the closed surface is translated, we have
dP
77= (5.109)
So Eq. (5.107) can be written in the form
ds
(5.110)
Jv*
5.5 / Continuum Modeling of Multiphase Flows 189
The derivative d/ds is expressed in terms of ro(s) as [Whitaker, 1969]
j- = :TT- = ? v = TT' v (5-m>

1
ds ds 9xi ds ds
Equation (5.110) finally takes the form
dr0 f f f \
V / i/fkdV / Vk^kedA =0 (5.112)
13
\ JVk J Akfi J
\ /
Since r 0 is arbitrary, the preceding equation can be reduced to
V / \j/kdV = / \lrknkedA (5.113)
Jvk JAkt
From the Gauss theorem, we have
= / V ^ k d V - / ^kkdA (5.114)
Jvk JAk
Combining Eq. (5.114) with Eq. (5.113), we have an averaging theorem as
= V f fkdV + I ^knkdA (5.115)

Jvk JAk
or
Similarly, it can be proved that
(V </>k) = V <t/.k> + i f xj>k k dA (5.117)

/Ak
The averaging theorem for the time derivative can be derived directly from the general
transport theorem. Consider the phase k in Fig. 5.7 to be fixed in space while Ak varies as
a function of time as a result of phase change. Applying Eq. (5.8) to the system in Fig. 5.7,
with d/dt = d/dt, we have
Adt f fkdV= [ ^pdV+ [ fkU (5.118)

Jv
lvkk Jvkk oz
Jv $t jJAAkk
where Us is the speed of displacement of the interface. Let the control volume of the mixture
be constant and larger than V^. From the definition of the phase average, we obtain
1 f
V / V^k^s-%dA (5.119)
JAk
Equations (5.116), (5.117), and (5.119) characterize the volume-averaging theorems of
derivatives [Slattery, 1967b; Whitaker, 1969].
5.3.2 Volume-Averaged Equations

For a dispersed multiphase flow, each phase may be regarded as a single phase
flowing over other phases. Thus, the volume-averaged equations for phase k in a dispersed
multiphase flow system can be obtained by applying the averaging theorem to the governing
equations of single-phase flows.
Figure 5.7. Expansion of phase k in a fixed control volume.
The volume-averaged continuity equation for dispersed laminar multiphase flows is given
by applying the volume-averaging theorems to Eq. (5.13) as
V (5.120)
Ot
where Fk is given by
rk = ~ f - Us) n k dA (5.121)
Fk represents the rate of mass generation of phase k per unit volume. It is noted that the unit
"volume" in this context is that of the gas-solid system. If a phase is defined in a physical
sense such as the solid phase or the gas phase, F k may be caused by chemical reactions
or phase changes. On the other hand, when a phase is defined in a dynamic sense [Soo,
1965], Fk may result from the size change due to attrition or agglomeration in addition to
the chemical reaction or phase change. From the mass balance of the mixture, we have
5k =0 (5.122)
Applying the averaging theorems to Eq. (5.14), the volume-averaged momentum equa-
tion of phase k is obtained as
V (r k )
ot
(5.123)
where F^ is given as
r k ) nkcL4 (5.124)
which accounts for the transfer of pressure and viscous stresses across the interface per unit
volume. FAk includes the drag force, Saffman force, and Magnus force of the phase.
in Eq. (5.123) takes the form
Frk = - A / frUk(Uk - Us) - nk dA (5.125)
which represents the momentum transfer across the interface per volume due to the mass
generation of the phase. The field force per unit mass / is taken to be a constant. For gas-
solid suspensions, / may include the gravitational force, electromagnetic force, electrostatic
force, and other material-related forces such as the van der Waals force.
The volume-averaged equation of energy conservation in terms of internal energy is
given by averaging Eq. (5.18) as
- ^ T ^ + V (Pkt/^k) = -</7kV Uk) - V (/qk) + </ek) + (0k> + GAk + 2rk

ot
(5.126)
where Q^k accounts for the heat transfer across the interface as
2Ak = - ^ / Jqk-nkdA (5.127)
V
JAk
and Qrk is the internal energy transferred by the mass generation as a result of phase change
and is expressed by
(Qrk) = / Pkek{Uk Us)-nkdA (5.128)
The term (0k) is the volume-averaged dissipation function for the energy dissipated by the
viscous force, which is irreversible dissipation of mechanical work into thermal energy or
heat. For the solid-particle phase, the kinetic energy loss by attrition or inelastic collision
may be included in this term.
The volume-averaged products in the preceding volume-averaged equations can be fur-
ther expressed by the products of the volume averages. From Eq. (5.100), the volume-
averaged mass flux of phase k is given by
(pkUk) = (Ac) \Uk) = cxk W \Uk) (5.129)

Similarly, the momentum flux of phase k can be denoted by
(pkUkUk) = (pk) \UkUk) = a k W W k > (5.130)

Furthermore, the intrinsic average of velocity correlation may be approximated by a product
of intrinsic averages of velocities as
iUi) (5.131)
From Eqs. (5.99) and (5.101), the internal energy per unit volume can be expressed by
(Pkek> = (A) W = k '(A) W (5.132)

and
W WW (5.133)
Thus, the internal energy carried by the mass flux of phase k can be expressed by
(pkek) l(Uk) = ak W \ek) \Uk) (5.134)
The approximation in the preceding equation is to circumvent any conflict with Eq. (5.100),
which defines the intrinsic average of velocity. This approximate relation becomes an equiv-
alent one under the isothermal condition of phase k with the expansion work estimated by
t/k> (pk)V <<l/k> (5.135)
For cases without the phase change between gas and solid phases, I \ , Frk and
vanish. Thus, assuming constant material densities and using Eq. (5.129), the volume-
averaged continuity equation, Eq. (5.120), becomes
0 (5.136)
Similarly, the volume-averaged momentum equation, Eq. (5.123), reduces to
Moreover, the volume-averaged equation of energy conservation in terms of internal energy,

given in Eq. (5.126), becomes
(Pk) (^^ + Wk> V W ) = -<Pk)V l(Uk) - V </qk) + (/ek) + <0k> + QAk
(5.138)
For an isothermal gas-solid two-phase flow without the phase change, Eq. (5.136) and
Eq. (5.137) can be expressed by
da
+ V-(al(U})=0 (5.139)
at
da
- +V-((l-a) I <tf p = 0 (5.140)
+ i{U) v <t7> = v
^dT ' ' ) 'W
p p (l - a ) (^^- + l(Up) V '(Vp)) - V (J p ) + p p (l - a)f - FA (5.142)
Equations (5.139) to (5.142) are the basic equations for a gas-solid flow. More detailed
information on both the fluid-particle interacting force FA and the total stresses T and
J p must be specified before these equations can be solved. One approach to formulate
the fluid-particle interacting force FA is to decompose the total stress into a component E
representing the "macroscopic" variations in the fluid stress tensor on a scale that is large
compared to the particle spacing, and a component e representing the effect of detailed
variations of the point stress tensor as the fluid flows around the particle [Anderson and
Jackson, 1967]. Under this condition, we may express the total tensor T of the fluid phase
by
T =E+e (5.143)
and
\T) = '{> (5.144)
The interacting force imposed by the particles on the fluid phase, FA, is given by
~ [ T-npdA = ~ / (E + e)-npdA
v JAv V JA
1
-Fp (5.145)
~~vjVp
where F p is defined by
Fp = - [ e-nvdA (5.146)
Thus, Eq. (5.141) and Eq. (5.142), respectively, become
pa I ^ - ^ + \U) V \U)I = aV \E) + ap/ - F p (5.147)
V '(tf p>) = V '(rp> + (1 - a)V '<) + (1 - a ) p p / + F p
(5.148)
In the preceding equations, F p can be expressed as a combination of local averaged drag

force and virtual mass force [Anderson and Jackson, 1967].
5.3.3 Volume-Time-Averaged Equations

The purpose of the time averaging after the volume averaging is to express averages
of products in terms of products of averages and to account for turbulent fluctuations and
high-frequency fluctuations [Soo, 1989]. The volume-time averaging is presented here in
a similar way to that of the Reynolds analysis of single-phase turbulent flow.
The averaging time duration should be chosen in such a way that THF <^ T < rLF-
The characteristic time of the high-frequency component THF may be estimated from the
reciprocal of the characteristic spectral frequency of the fluctuation, while the characteristic
time of the low-frequency component TLF may be determined from the time required to travel
the characteristic dimension of the physical system at the local characteristic low-frequency
speed. Thus, a time averaging after volume averaging can be defined as
1 fT
= -1 / WOd/ (5.149)
Jo
An instantaneous volume-averaged quantity is therefore expressed by two parts
(5.150)
where the prime denotes the high-frequency component and

(K) = 0 (5.151)
Further, we have
dV = ,
(5.152)
which indicates that the time averaging over a volume-averaged quantity equals the volume
averaging over a time-averaged quantity. This equivalence is valid when the spatial integral
and temporal integral become interchangeable with a proper choice of T (i.e., THF <^ T <
?LF)- When the time averaging is applied to the mass flux of phase k, we obtain
= <Pk> KUk) = <pk> KUk) + (p k ) iK) (5.153)
The correlation of two fluctuating components may be best formulated using Boussinesq's
approach [Boussinesq, 1877] by introducing a transport coefficient. Thus, the last term in
Eq. (5.153) can be expressed by
(5.154)
where Dk is the eddy diffusivity of phase k in the mixture.
Assuming the material density or the intrinsic averaged density of phase k to be constant
(as is true for the solids and almost true for the fluid when the effect of temperature variation
on the material density can be neglected), we have
(pk) = oik l(Pk) = akPkm (5.155)
where Pkm is the material density of phase k.
The expression for the momentum flux of phase k in terms of the product of averages can
be derived from the definition of the volume-averaged momentum flux of phase k. From
Eq. (5.130), we obtain
k) = <Pk> Wktfk) = <Pk + P
= ?*> + (Pk))C<Pk~Pk~> + f(PkPk> + '"(PkPO + '"(
If the local low-frequency component of a phase quantity is assumed to be constant through-
out the volume and time averaging, we can further obtain
(5.157)
which leads to Eq. (5.156) in the form of
(/\UkUk) = Got) '(tfk)'(ffk) + (GdKVM +
'VJ TO (5.158)
where the second term on the right-hand side of the equation represents the Reynolds
stress and the last two terms account for the momentum flux arising from the eddy mass
diffusion.
A similar procedure can be applied to the energy transport terms. When the volume-
averaged energy equation is expressed in terms of the internal energy and time averaging
5.31 Continuum Modeling of Multiphase Flows 195
is applied to the equation, the local volume-averaged internal energy flux becomes
(PkUkek) = Got) t(Uk) l(ed + (ft) Vk^k) + '&)(PLUO + l<Vk)(pWk) + (PkUke'k)

(5.159)
where the second and last terms on the right-hand side represent the thermal transport due
to the fluctuations in velocity and internal energy (or temperature) and can be expressed
by
/ & ) + (pkU'kek) = -GSk)Z>rkV 'fed (5.160)
where Dxk is the eddy thermal diffusivity. The third term on the right-hand side of Eq. (5.159)
accounts for the energy transport by diffusion, which may be expressed by
(5.161)
The fourth term on the right-hand side of Eq. (5.159) is the energy transport by the fluc-
tuations in density and internal energy (or temperature), which is linked to the averaged
quantities by
V^) (5.162)
where Cek is an empirical constant associated with the energy transport by fluctuations in
density and temperature. In addition, the volume-time-averaged expansion work is esti-
mated by
Uk) + (p'kV U'k) <p k)V l(Uk) (5.163)
The volume-time-averaged equations can be summarized as follows:

(1) Continuity equation
^ + V . (at'<! = V (>kVc*D + ^ (5.164)

vt Plan
(2) Momentum equation
| - (k"'(l7k) - D kVc) + V ('<!) TO)

ot
= v [/^(VoD'TO + ^kTOvot] - v(p k)

Aon
+ V Tk> + T J ) + (k/) + + (5.165)

Pkm Pkm Pkm
where the Reynolds stress of phase k is defined as
rl = -((Pk) W +^'W) (5-166)

which needs to be modeled by the turbulence models of phase k.
(3) Energy conservation equation
(akAcmSk) + V (a k Pkm '{Vk)ek)
k3
+ V C(ek)DkV(akPkm)) + V ( C ek Âcm l (U k )Va k V'(ek)
- (pk)W l(Uk) + </ek) + (<Pk) + 2 A k + Qn (5.167)
For convenience, we remove the {} signs and time-averaging bars from the preceding
equations. Thus, the volume-time-averaged equations can be simplified to the form
^ + V (akUk) = V (D k V k ) + (5.168)
with
^ k = 1 (5.169)
Yt /k - DkVak) + V- (akUkUk)
= V [Dk(Vak)Uk ^- D k^kVofk]
1 1 T Âk ^"rk
( ) (5.170)
Pkm Pkm l Tk r
k) Pkm Pkm
d
i ke (oikUke- k)
= V (k^T kV^k) + V 7kVc Vet |
1
+ Pkm ? kV Uk.+ êk + 0k + C >Ak + Grk) (5.171)
Assuming that the effects of turbulence can be reasonably modeled such that the cor-
relations of turbulent fluctuations can be linked to the averaged quantities, the dependent
variables for a n-phase flow are ak, Uk, pk, and Tk, where k varies from 1 to n. Thus, the
total number of dependent variables is 6n. Although Eqs. (5.168) through (5.171) consti-
tute 5n + 1 equations, additional n 1 equations (pk = pk(ak, Tk)) are needed to reach a
closure. The development of these n \ equations requires the mechanistic understanding
of the pseudofluid nature of each phase.
5.3.4 Transport Coefficients and Turbulence Models

The fluxes of mass, momentum, and energy of phase k transported in a laminar or
turbulent multiphase flow can be expressed in terms of the local gradients and the transport
coefficients. In a gas-solid multiphase flow, the transport coefficients of the gas phase may
be reasonably represented by those in a single-phase flow although certain modifications
may be needed to account for the effects of the solid phase, especially when the solids
concentration is high.
In this section, we discuss the transport coefficients of the solid phase. Specifically, the
basic mechanisms for the transport processes of solids suspended in a gas medium may
be understood from the gas-particle interactions and the particle-particle collisions. In
turbulent gas-particle flows, the diffusion of particle clouds by turbulent eddies was modeled
by Tchen and Hinze [Tchen, 1947; Hinze, 1959]. The Hinze-Tchen model considers the
motion of a small single particle tracking a turbulent eddy. The model is simple and always
yields a particle turbulent intensity less than the gas turbulent intensity. A more advanced
turbulence model for gas-solid flows is known as the k-e-kv model [Zhou, 1993]. This
model accounts for the effect of gas-particle interactions on both the gas turbulence and
particle velocity fluctuation in terms of the particle turbulent kinetic energy kv. In addition,
the k--kv model uses a transport equation of kv instead of an algebraic relation as in Hinze-
Tchen's model for the simulation of particle phase turbulence. Both the Hinze-Tchen model
and the k--kp model are for dilute gas-solid flows where the effect of interparticle collision
can be neglected. When the solids concentration is high, the momentum and energy transfers
due to single scattering of the collisions among particles become significant. In this section,
an expression for the viscosity of solids from a simple shearflowmodel for particle collisions
with single scattering and specular reflection is derived [Soo, 1989]. Further increase in
the solids loading leads to a dense suspension in which the dynamics of the solid phase is
dominated by particle collisions. In this case, the viscosity of the solids may be simulated
by using the kinetic theory of gases [Savage, 1983; Jenkins and Savage, 1983; Gidaspow,
1993], which is discussed in 5.5.
5.3.4.1 Hinze-Tchen s Model

Here we present the derivation of the transport coefficients due to diffusion of discrete
particles in a homogeneous turbulent flow. Although the Hinze-Tchen model was developed
in a more general form for general particle-fluid multiphase flows, we introduce this model
only for the cases of gas-solid flows. Some assumptions for this model are the following:
(1) The turbulence of the fluid is homogeneous and steady.
(2) The domain of turbulence is infinite.
(3) The particle is spherical and follows Stokes's law.
(4) The particle is small compared to the smallest wavelength in turbulence.
(5) The particle is always entrapped inside the same turbulent eddy.
(6) For gas-solid suspensions, pv/ p ^> 1.
To simplify the derivation, we may consider only the one-dimensional case. Denote
s(to + t) as the distance traveled by a marked fluid element with starting time to during the
time interval t. Let u be the instantaneous velocity and u' be the fluctuating velocity. In the
following derivation, the time-averaged fluid velocity is taken as zero. Thus, we have
t)= [ u'(t0+ r)dr (5.172)

I
It can be shown that [Taylor, 1921]
pt ft
ft'
= 2ua dt' R{x)dx (5.173)

Jo Jo
where R(r) is a dimensionless autocorrelation function or the Lagrangian correlation co-

efficient given by
.,,
The transport rate of a scalar quantity y can be expressed in the form suggested by
Boussinesq (1877) as
where Dy is the coefficient of diffusion of y and F is the corresponding mean value of the
quantity. Assuming a linear variation of T with respect to x, we have
y\t) = s(f)^ (5.176)
Since
u\t) = ^ (5.177)
df
from Eq. (5.175), we obtain
dx T Jo dt 2 dr
Hence, from Eq. (5.172) and Eq. (5.173), the coefficient of diffusion of y is
DY = ~=lfi f R(T)dr (5.179)

2 dr Jo
The dynamic behavior of discrete particles in turbulence can be described by the BBO
equation [Basset, 1888; Boussinesq, 1903; and Oseen, 1927] for the slow motion of a
spherical particle in a fluid at rest. This equation was extended by Tchen (1947) to the case
of a fluid moving with time-dependent velocity as
71
J3 dp ft ,3 / dw dwj
dw dwt
dr dr
f dr (5.180)
From the Navier-Stokes equation, we have

dp \du du] d 2u
The substantial derivative in a Lagrangian coordinate moving with the discrete particle can
be expressed by
i = h+u>h (5182)
-
Thus, the modified BBO equation becomes

3p \du 2/x8 2 w] ( 36/x
du dwp
du 18 rpn r fa - -^ dr (5183)
p/
2pp + p ax (2pp
Assuming the effect of Basset force is negligible and noting that in gas-solid flows
Pp/p 1, Eq. (5.183) reduces to
* ^ ^ (5.184)
dr rs
where r s is the Stokes relaxation time of the particle diffusion, defined by Eq. (3.39).
Both u and wp may be represented by the Fourier integrals as
(a cos cot + ft sin cot) dco

Jo
(5.185)
/OO
wp= / (y cos&tf + 8 sin cot) dco

Jo
where co is the angular frequency and is equal to 2nf. Substituting Eq. (5.185) into
Eq. (5.184), we obtain the relations ofa,p,y, and 8 as
COTS8 a y
(5
X R -186)
Defining the one-sided autospectral density function G(f) by
f2
f
POO
G(f)=4u R(r)cos27tfrdr (5.187)

Jo
from the inverse Fourier transform, we have the corresponding expression of R(r) as
1 f
R(r) = = G(f) cos 2ixfxdf (5.188)
ut2 Jo
Note that R(0) = 1, which yields
/*OO
^ = / G(/)d/ (5.189)
Jo
Thus, on the basis of Eq. (5.189), G(f) df represents the contribution to the mean square
of the turbulence fluctuation of velocity between frequencies / and / + df. When u is
expressed by Eq. (5.185) u becomes zero, and thus u is equal to u'. It can be shown [Hinze,
1959] that
(az + pz)dco (5.190)
where T is the integral time. Thus, G(f) can be related to a and P by
G{f) = n-(a2 + p2) (5.191)

Hence, we have
G((o) a2
(5.192)
Gv(co)
We assume that the Lagrangian correlation coefficient for fluid motion may be repre-
sented by an exponential function
J?(f)=exp( - - (5.193)
where Xf is the characteristic time of the fluid motion. The corresponding one-sided
autospectral density function then becomes, from Eq. (5.187),
~~ (5.194)
Thus, substituting Eq. (5.186) and Eq. (5.194) into Eq. (5.192), Gp(co) is obtained as
a 4rf
Gv(co) = uu 2
(5.195)
+Q) Tf )
Consequently, we have
<?=T- r (5.196)
2n Jo
and
1
r i / / t\ ( t
RM) = = / Gp(Q))cos(otdco= rfexp -rsexp
2nu/2 Jo tf-rsV V *i) V r
s
(5.197)
from which we may conclude that the Lagrangian correlation coefficient for discrete particle
motion is no longer a single exponential function. In terms of the definitions of the transport
coefficients for turbulent diffusion given in Eq. (5.179), it can be shown that
D YV
= 1+ (5.198)
l-exp(--
For a short diffusion time (t < Tf and t <$C ts), we have

-l
(5.199)
Equation (5.199) is known as the Hinze-Tchen equation.

Using the analogy of the kinetic theory of gases, from Eqs. (5.43), (5.46), and (5.49), it
yields
-l
Vp _ D_YV
= i+ (5.200)
v ~~ Dv
where v is the kinematic viscosity. The corresponding thermal conductivity can be given
by
K
" ~"~v" WI
" - "" ' (5.201)
where C is the specific thermal capacity of the gas and C p is the specific thermal capacity
of the particle.
5.3.4.2 k-e-kp Turbulence Model

The gas-particle interaction may strongly enhance or attenuate the gas phase turbulence.
Thus, a turbulence model for gas-solid flows needs to include the effect of gas-particle
interactions. Here, we consider only isotropic turbulence and dilute gas-solid flows without
phase changes. The flow is assumed to be steady and isothermal.
Following the approaches used in the single-phase k-e model, the k and e equations for
dilute gas-solid flows can be expressed as [Huang and Zhou, 1991; Zhou, 1993]
(p(l-ap)t/j*) = I
dxj 3JCJ \
(5.202)
and
where Gk is the generation of k due to the gas-solid interaction. G k can be expressed by
Gk = 1 L a P ( C k y ^ - 2k) + ^ ^ ^ ^ p - } (5.204)
where Ck is an empirical parameter; kv is the particle turbulent kinetic energy defined as
*p = i ^ (5.205)
and /i p is the particle turbulent viscosity due to gas-particle interaction. fx v is given by
k2
MP = C M p P p a p ^ (5.206)
The particle turbulent kinetic energy is governed by its own transport equation. Similarly
to the derivation of the ^-equation, the &p-equation is given by
(5.207)
where ep is the rate of generation or destruction of kv due to the gas-solid interaction. ep

can be expressed by
(5.208)
The model given above is called the k--kv model, which can be used for dilute, non-
swirling, nonbuoyant gas-solid flows. For strongly anisotropic gas-solid flows, the unified
second-order moment closure model, which is an extension of the second-order moment
closure model for single-phase flows [Zhou, 1993], may be used.
5.3.4.3 Particle Viscosity Due to lnterparticle Collisions

When the particle concentration is high, the shear motion of particles leads to interparticle
collisions. The transfer of momentum between particles can be described in terms of a
pseudoshear stress and the viscosity of particle-particle interactions. Let usfirstexamine the
transfer of momentum in an elastic collision between two particles, as shown in Fig. 5.8(a).
Particle 1 isfixedin space while particle 2 collides with particle 1 with an initial momentum
in the x-direction. Assume that the contact surface is frictionless so that the rebound of the
particle is in a form of specular reflection in the rx-plane. The rate of change of the
x -component of the momentum between the two particles is given by
m2U2 - [m2U2 COS(TT - 20)] = 2m2U2 cos 2 0 (5.209)
where 0 is the contact angle in the r-x-plane.
Particle 1 Particle 2
(a)
A representative particle
Up(y) in a cloud of particles
Differential surface of
the fixed particle
(b)
Figure 5.8. Impaction due to shear motion: (a) Elastic collision between two particles;
(b) Elastic collision of a single sphere with a cloud of particles.
Consider a collision between two particles 1 and 2 with masses m \ and m2 and velocities
U\ and U2, respectively. From Newton's law, we have
J7i\ = f\2\ yn2 = f\ 2 (5.210)

6t 6t
where f\2 is the collisional force. Equation (5.210) can be reduced to the form
m * ^ = /l2 (5.211)
6t
where AU U\-U2, the relative velocity between the two particles; and m* is the relative
mass defined by Eq. (2.128). Thus, the collision between two moving particles is equivalent
to the case where a particle collides with a fixed particle with the same relative mass and
relative velocity.
Now consider a case of collision of a fixed single sphere with a cloud of particles as
shown in Fig. 5.8(b). The sphere of radius r is subjected to a shear flow of a particle cloud
with a velocity gradient of dUv/dy. The particle number density is denoted as n, and the
mass of a particle is m. Select the velocity Uv as zero in the center plane of the sphere. The
relative velocity is estimated by
dUv dUn
AUV = E v = l rsm0cos(p (5.212)
dy dy
where 0 is the azimuthal angle in the y-z plane. The differential area of impact on the
sphere is r 2 sinO 60 6(j). The projected area of impact normal to the incoming velocity Uv
is r2 sin (9 cos 0 60 d0 and the rate of change of the x-component of the momentum for a
particle is 2mAUpcos20 from Eqs. (5.209) and (5.212), assuming single scattering and
specular reflection. The force acting on the semisphere is given by
F=4 / n(AUp)2m r2 sinO cos3 0 60 6(p = nm(Mr 4 (5.213)

Jo Jo 12 V ^y )
Consider a mixture of two groups of particles with radii a\ and a2, and corresponding
number concentrations of n\ and n2. Assume that the group of particle 1 is subjected to a
simple shear flow of particle 2. The x-component of the collision force between particle 1
and particle 2 can be obtained, from Eq. (5.213), as
2 ( dll \ 2
F\2 = n\n 2 m* ( I (a\ + a2) (5.214)
18 \ dy )
where (a\ + a2) is the radius of impact.
By introducing an equivalent force, i.e., equivalent to the x-component of the collision
force, acting on a projected area normal to the v-axis, a shear stress of particle phase can
be defined as
2Fl2 n fdUv\2 ,
(Txy)i2 = : ~ = ^n {n2m* [ - ^ ) (ax + a2)5 (5.215)
9 V ^
Equation (5.215) can be extended to a generalized form as [Soo, 1989]
nln2mlm2/__ +, a_2)55(Ap2 _: AAp2)^
xi
Ap2 = -/Xpi2Ap2 (5.216)
rn2)
where Ap has the Cartesian components

_ Wpi
(5.217)
and
A p : A p = ApijApji (5.218)
The viscosity due to the collision between the group of particle 1 and the group of particle
2 is defined from Eq. (5.216) as
A p2 :A p2 )2 (5.219)
9 (wii+/fi 2 )
For the special case of identical particles, Eq. (5.216) yields
Atp = ia^^(Ap:Ap)i (5.220)

The thermal conductivity of the particle collisions, Kv, can be obtained on the basis of
the kinetic theory. From Eqs. (5.43) and (5.49), for the identical particles, we have
(5.221)
m 7tdp
So far, in this model, we assume that all the particles are elastic and the collision is
of specular reflection on a frictionless smooth surface. For inelastic particles, we may
introduce the restitution coefficient e, which is defined as the ratio of the rebound speed to
the incoming speed in a normal collision. Therefore, for a collision of an inelastic particle
with a frictionless surface as shown in Fig. 5.9, we have
. % . != (5.222)
so the reflection angle 0' is larger than the incident angle 0; i.e., the reflection is not specular.
In Eq. (5.222) U[ and U[ are, respectively, the tangential and normal reflecting velocity after
collision. The basic procedure in deriving the pseudostress and the resulting viscosity for
inelastic particles is similar to that for elastic particles.
Ur
Figure 5.9. Inelastic impaction on a frictionless surface.

5.4 I Trajectory Modeling of Multiphase Flows 205
5.3.5 Boundary Conditions of Particle Phase

The boundary conditions for the gas phase are assumed to be of the same form
as that for the single gas phase discussed in 5.2.5. For the particle phase, the boundary
conditions are given as follows.
For the normal components of particle velocities, the boundary conditions may be ex-
pressed by using the coefficient of restitution as
U;n = -eUpn (5.223)
where U^n is the normal component of the rebound velocity of the particle. Hence, for
elastic particles, U^n U pn. For the tangential velocities of particles at a solid wall, slip
conditions should be imposed; they may be expressed by
U;t = fUpt (5.224)
where / is a frictional factor related to the material properties of both wall and particles and
the incident angle of impact. For frictionless surfaces, / = 1, which leads to U^ t = Uvi.
The boundary conditions of particle temperature may be obtained from the heat transfer
due to the collision of two bodies. Sun and Chen (1988) formulated the heat transfer
per impact of two elastic particles by considering the collision of two elastic particles
with different temperatures and assuming the heat conduction occurs only in the normal
direction.
We may consider the solid wall as a very large particle with an infinite thermal capacity.
Thus, for the collision between a particle and a solid wall, we have
rp) (5.225)
where Ac is the maximum contact area given by Eq. (2.133) and tc is the contact time
duration given by Eq. (2.136).
For the boundary condition of particle concentration at the wall, a zero normal gradient
condition is frequently adopted; that is
=0 (5.226)
Furthermore, it is assumed that there is no particle fluctuation on the wall so that the
boundary condition of particle turbulent kinetic energy at the wall is given by
(*P)W = 0 (5.227)
5.4 Trajectory Modeling of Multiphase Flows

Trajectory models quantify the dynamic characteristics of particles in Lagrangian
coordinates. The trajectory model is useful when the particle phase is so dilute that the
description of particle behavior by continuum models may not be suitable.
Two basic trajectory models, i.e., the deterministic trajectory model [Crowe et ai, 1977]
and the stochastic trajectory model [Crowe, 1991], are introduced in this section. The de-
terministic approach, which neglects the turbulent fluctuation of particles, specifically, the
turbulent diffusion of the mass, momentum, and energy of particles, is considered the most
classical and simplest approach that uses the Lagrangian concept. However, the oversim-
plified assumptions for the transport behavior of the particles, especially on the energy
transport processes, may yield inaccurate predictions of the particle number density distri-
bution and the temperature distribution of the particles. The stochastic approach, with the
aid of the Monte Carlo method, can directly simulate the instantaneous dynamic behav-
ior of the particles on the basis of instantaneous momentum equations of the particles in
Lagrangian coordinates. Depending on the choice of turbulence models and phase inter-
action mechanisms, this approach, in principle, can simulate the turbulence effect on the
interactions between the gas and the particles. However, to obtain the statistical averages
with reasonable accuracy, calculation of tens of thousands of trajectories, which sometimes
requires enormous computing power, may be needed. It should be noted that in the mod-
eling of gas-solid flows, the trajectory model is applied to account for the particle phase
only, while the gas phase is usually described by the Eulerian model. Using the Lagrangian
approach, the governing equations of the particle phase are expressed in the form of ordinary
differential equations instead of partial differential equations.
To simplify the following analysis, we assume that (1) the particles are spherical and of
identical size; (2) for the momentum interaction between the gas and solid phases, only the
drag force in a locally uniform flow field is considered, i.e., all other forces such as Magnus
force, Saffman force, Basset force, and electrostatic force are negligible; and (3) the solids
concentration is low so that particle-particle interactions are excluded.
5.4.1 Deterministic Trajectory Models

Consider that the turbulent diffusion of the particle mass, momentum, and en-
ergy can be neglected; i.e., the particle viscosity /JLV, particle diffusivity D p , and thermal
conductivity of particle phase Kp are negligibly small. The governing equations of mass,
momentum, and energy of the particle phase in the Lagrangian form can be obtained from
the corresponding volume-time-averaged equations in the Eulerian form derived in 5.3.3,
by setting all the correlation terms on the right-hand side of the particle-phase equations
to zero. In this case, the time-averaged particle-phase equations in Lagrangian coordinates
are formally identical to the instantaneous equations in Lagrangian coordinates.
5.4.1.1 Mass Conservation Equation

K P P ) + V (apppt/p) = F p (5.228)
ot
If we express the density of the particle phase in terms of the product of the particle number
density n and the mass of a particle m, i.e., a p p p = nm, and noting that F p = n(dm/dt),
we have
+ V (nUv) = 0 (5.229)
ot
For a steady flow, Eq. (5.229) can be reduced to
V (nUv) = 0 (5.230)
Thus, integrating the preceding equation over any cross section of a stream tube of the
particle phase and using the Gauss theorem, we can obtain
nUp dA = const = Np (5.231)
where A is the vector cross sectional area of the stream tube and Np is the flow rate of
the number of particles passing through A. Equation (5.231) reflects that the total particle
number flux is conserved along the stream tube. Thus, it is concluded that the total particle
number flux is conserved along the particle trajectory.
5.4.1.2 Momentum Equation

a.
(cxpppUp) + V (apPpUpUp) = apPpf + FAp + FTp (5.232)
where F r p is the momentum transfer due to the change of mass and is given by Eq. (5.125) as
F
rP = ~ [ ctpPpUp(Up - Ut) p dA UpTp (5.233)
v
JAP
FA P is the drag force between the gas and the solids, which can be expressed as
p -Up) (5.234)
where r^ is the relaxation time of the particle. For a low relative motion, r^ may be obtained
from Eq. (3.39). In Eq. (5.234), the effect of interparticle collisions is neglected since most
applications of the trajectory model deal with very dilute multiphase flows.
Using Eq. (5.228), Eq. (5.232) becomes
^ ( t f - f f p ) (5.235)
where d/dtp is defined as

d a
(5 236)
d,p a, - ^ -
Moreover, the particle trajectory s can be obtained from
s-so= I Updt (5.237)

Jtn
5.4.1.3 Energy Equation

(a p ppC p r p ) + V . (apPpUpCvTv) = Jep + QAp + QrP (5.238)
Using Eq. (5.228) and assuming that <2rP can be approximated by
2r P = ~ [ Pvep(Up -U8)'dA*t CpTpTp (5.239)

Eq. (5.238) is simplified to

dT- 1
/EP + QA P ) (5.240)
Note that / e p in Eq. (5.238) is replaced with / E p for Eq. (5.240), where / E p is the heat
generated by thermal radiation per unit volume and QAP is the heat transferred through
the interface between gas and particles. Thus, once the gas velocity field is solved, the
particle velocity, particle trajectory, particle concentration, and particle temperature can all
be obtained directly by integrating Eqs. (5.235), (5.237), (5.231), and (5.240), respectively.
Since the equations for the gas phase are coupled with those for the solid phase, final
solutions of the governing equations may have to be obtained through iterations between
those for the gas and solid phases.
5.4.2 Stochastic Trajectory Models

In trajectory models, the particle turbulent diffusion can be considered by calculat-
ing the instantaneous motion of particles in the turbulent flow field. In order to simulate the
stochastic characteristics of the instantaneous gas velocity in a turbulent flow, it is required
to generate random numbers in the calculation process.
The instantaneous motion of a particle is governed by
^ l = (u-uv)+f (5.241)
(it T
v^-tp c^rp
where u and uv are the instantaneous velocities of the gas and the particle, respectively. /
is the body force acting on the unit mass of the particle. If we write Eq. (5.241) in a scalar
form in cylindrical coordinates, the axial component of the equation becomes
= (U + uf - p) + gz (5.242)
dtv Trp
while the radial component is
di; n 1 u>n
= (V + i/ - vp) + + & (5.243)
and the tangential component is
^ = (W + w'_Wp).M!l+u (5.244)
where U, V, and W are the axial, radial, and tangential components, respectively, of the
time-averaged gas velocity; uf, v\ and wf are the axial, radial, and tangential components,
respectively, of the fluctuating gas velocity; and gz, gY, and g^ are gravitational acceleration
in the axial, radial, and azimuthal directions, respectively.
Assuming that the turbulence is isotropic, we have
'u71 = ^ 2 = ^ 2 = -k (5.245)
where Ic is the turbulent kinetic energy, which may be determined from the k-e model.
Furthermore, it is assumed that the local velocity fluctuations of the gas phase obey a
Gaussian probability density distribution. Thus, when a particle passes through a turbulent
eddy, we have
v
' = ?y|*. ' = !)/fk' w
' = S\lfk (5-246)
where f is a random number.
Substituting Eq. (5.246) into Eqs. (5.242), (5.243), and (5.244), we have the integrable
instantaneous equations for the particle velocities
(5 247)
) -
dvo 1 / . . FT l
3- -,. . o. (5.248)
and
(5.249)
Thus, by coupling Eqs. (5.247), (5.248), and (5.249) with the governing equations for the
gas phase, the instantaneous particle velocities can be obtained. Moreover, the stochastic
trajectories of the particles can also be obtained by
z = zp0 + / Up dtp, r =rpO+ vv dtp, <9P = <9p0 + / ^ d;p (5.250)

Jo Jo Jo r
In the numerical integration of a particle trajectory, the selection of the integral time step
is important. A typical way of choosing the integral time step is based on the interacting
duration between the turbulent eddy and the particle. This interacting duration Z{ may be
determined by
^ = min[Te,TR] (5.251)
where Te is the eddy existence time, while TR is the flying time for the particle passing
through the eddy. Te may be estimated from the k- model as
e ,_ l
VW2 + v'2 + wa
where /e is the characteristic length of the turbulent eddy. Thus,
j* (5.253)
is determined by
(5254)
K^)
and Ti is taken to be Te when /e > xÛ Uv\.
-
In computation using the stochastic trajectory model, the Monte Carlo approach is com-
monly employed. It is necessary to calculate several thousands, or even tens of thousands, of
trajectories to simulate the particle flow field. The central issue in developing the stochastic
trajectory model is how to model the instantaneous turbulent gas flow field. The method
given uses the Gaussian distribution of velocityfluctuations.Others use the Langevin model,
which is a statistical model [Elghobashi, 1994]. Recognizing the large computational re-
quirement for the stochastic trajectory model, the alternative method is to calculate the most
probable trajectories combined with the presumed probability density distribution functions
of the particle velocity, concentration, and other quantities [Baxter, 1989].
5.5 Kinetic Theory Modeling for Collision-Dominated

Dense Suspensions
When the particle concentration is high, the transport mechanism is significantly
influenced or even dominated by interparticle collisions. For gravity-free suspensions of
solid particles in a uniform shear flow, three distinct regimes were proposed by Bagnold
(1954) on the basis of their rheological properties. These regimes are the macroviscous, the
transitional, and the grain inertia regime, which are characterized by the Bagnold number,
Ba, as
Ba = A ^ f ^ V n e r t i a l f o r C e (5.255)
M V dv / viscous force
where X is a parameter related to the solids fraction. The macroviscous regime is demarcated
by Ba < 40, where the fluid viscosity dominates the motion of solids, and the shear stress
is linearly related to the shear rate. The particle motion in this regime can be described by
using either the continuum modeling or the trajectory modeling. When Ba > 450, where
moving granular flows occur, the particle motion is largely governed by particle-particle
interactions, and the shear stress depends on the square of the shear rate. This regime is thus
defined as the grain inertia regime. The transition regime falls between the macroviscous
and the grain inertia regimes [Hunt, 1989]. The preceding regime classifications have been
validated both for shear flows [Savage, 1979; Savage and Sayed, 1984] and for hopper
and chute flows [Zeininger and Brennen, 1985]. The study of grain inertia flows may be
conducted in light of the kinetic theory modeling, where the interstitial gas plays little role
in the momentum transport of solids.
Culick (1964) was among the earliest researchers who suggested an analogy between
particle collisions in suspensions and molecular collisions in the kinetic theory of gases.
However, the approach which rigorously followed the kinetic theories of gases for solid
particles came to a halt as a result of the complexity surrounding the direct application of
the Boltzmann equation in accounting for interparticle collisions. An alternative approach
using simplified kinetic theories of gases based on mechanistically derived or intuitive
relationships in place of the Boltzmann equation has been viewed as viable. For example,
Savage and Jeffrey (1981), Jenkins and Savage (1983), and Lun et al. (1984) introduced
elaborate collision models to determine the solid stress and other behavior of solids due to
the interparticle collisions for collision-dominated dense suspensions. This approach has
been applied to many gas-solid flow systems including fluidization and pneumatic transport
[Gidaspow, 1993].
The physical condition of the kinetic theory of gases can be described by elastic collisions
of monodispersed spheres with the Maxwellian velocity distribution in an infinite vacuum
space. Therefore, for an analogy between particle-particle interactions and molecular in-
teractions to be directly applicable, the following phenomena in gas-solid flows should not
be regarded as significant in comparison to particle-particle interactions: the gas-particle
5.5 I Kinetic Theory Modeling for Collision-Dominated Dense Suspensions 211
interactions involving velocity-dependent forces and pressure-dependent forces, energy dis-

sipation due to inelasticity and friction in particle collisions, and particle-wall interactions.
In this section, the discussion is focused on the dense gas-solid suspensions, where colli-
sions among particles are the only mechanism for the transport of mass, momentum, and
energy.
5.5.1 Dense-Phase Transport Theorem

For a dense system of hard, smooth, and elastic spherical particles, a transport
theorem based on the analogy of the kinetic theory of dense gases [Reif, 1965] may be
derived. Define an ensemble average of any property \j/ of a particle as
r, v, t) dv (5.256)
- ; /
where n is the particle number density and / ( 1 ) (r, v, t) represents the single-particle velocity
distribution function. Let us focus on the change of \[r in a fixed volume element dr which
is located between r and r + dr and contains ndr particles. The rate of change of the
ensemble average (n\[r dr) of the quantity iff for all particles in dr is affected by three
factors: (1) the velocity v of each particle varies with time; (2) particles bearing \fr enter and
leave the volume element dr; and (3) particles collide with each other in dr [Jenkins and
Savage, 1983]. Accordingly, the rate of increase of the ensemble average (n\{r dr) can be
expressed by
| - (nir) dr = A{ + Af + Ac (5.257)
ot
where A{ represents the rate of variation in the ensemble average of x/r due to the change of
\jr for each particle inside dr; Af accounts for the change of \fr due to the net flux of particles
entering d r; and Ac describes the change of \[r due to particle collisions in d r.
In a time duration dt, each particle changes its velocity by dv = (F/m) dt, where F is
the total external force acting on an individual particle of mass m. The change in \jr can be
given by
d\l/ F{ F dilr
J^^dt = --^dt (5.258)
av\ m m dv
Hence, the rate of variation in the ensemble average of i/r in dr is equal to
A{ = {n drDxjr) = n dr(Df) (5.259)
where
Df = - . ^ - (5.260)
m dv
A change of \jr in dr is also possible through a net flux of particles entering the volume
element. This increase rate is simply given by
Af = - V . (nvxjfdr) (5.261)
The change of \\r due to collisions may be related to the changes of velocities of colliding
particles in the form
Ar/r = xjri (r, v[, t) + ^ 2 ( r , v' 2 , t) - x//i(r, v u t ) - fair, v 2 , 0 (5.262)
The number of collisions is determined on the basis of three factors: (1) the scattering proba-
bility a (vi, V2 -> v\, v'2) dvj dvf2\ (2) the relative flux | vi V21 / ( r , vi, ) dvi of group 1 parti-
cles incident upon group 2 particles; and (3) the number of group 2 particles / (r, V2, t) dr dv2
which can do such scattering. Thus, Ac can be expressed as
= -dr / Ai/rv 12/1 (5.263)
where f12 = |vi V2I; /1 = / ( r , vi, 0; h = /( r > V2, 0; and the factor of 1/2 accounts
for the double counting of the colliding pairs in the preceding integration. The transport
theorem, also known as Enskog's equation of change, can be obtained by substituting
Eqs. (5.259), (5.261), and (5.263) into Eq. (5.257) as
-(nf) = n - V + (5.264)
ot
where C(T/O is the collisional rate of increase of the mean of \jr per unit volume. From
Eq. (5.263), C(i/f) can be written as
A If (5.265)
= - = - A\lfvn/i/2crdvidv2dv[dv2
For a system of hard, smooth, but inelastic particles of uniform diameter dv , the collision
term can be more explicitly expressed in terms of a collision transfer contribution and a
source term [Lun et al.9 1984]. Consider a binary collision between the particles labeled
1 and 2 in Fig. 5.10. In time 8t prior to the collision, particle 1 moves through a distance
V\28t relative to particle 2, where V12 = Vi V2- Thus, for a collision to occur within 8t,
the center of particle 1 must lie within the volume d28k(v\2 k)8t. The probable number
of collisions such that the center of particle 2 lies within the volume element 8r and vi, V2,
and k lie within the ranges <5vi, 8V2, and 8k is
. k)/(2)(r - dk, vi; r, v2; t)8k8vl8v28r8t (5.266)
where / ( 2 H v i, ru V2, ri\ 0 is termed the complete pair distribution function and is defined
such that / ( 2 ) (vi, #*i; v2, r2; t)8vi8v28r\8r2 is the probability of finding a pair of particles in
the volume element 8r\, 8r2 centered on the points r\, r2 and having velocities within the
ranges vi and v\ + 5vi, and V2 and V2 + 8v2, respectively.
Figure 5.10. Binary collision of particles of diameter dv.

During the collision, particle 2 gains a quantity (xj/^ ^2) of ^ , where primed and
unprimed quantities refer to the values after and before the collision, respectively. Consid-
ering only particles about to collide (i.e., taking vw k > 0), we can express the collisional
rate of increase for the mean of ^ per unit volume as
CW) = d2 f ( ^ - if2)(vn - k)fi2\r - d*, vi; r, v2; 0 dfcdv! dv2 (5.267)

Jvl2k>0
Similarly, by interchanging the subscripts 1 and 2 and replacing k by k, we can also write
C(V) = d2 / (f[ - fi)(V12 *)/ ( 2 ) (r, v i ; r + d*, v2; 0 d^dvT dv2 (5.268)
Expanding the pair-distribution function in a Taylor series, we have
/ ( 2 ) (r, vi; r + dft, v2; t) = f(2)(r-dk, vx\ r, v2; 0

v (dk v)2+ (dA: v)3+
(2)
'" h' h ' *
x/ (r, vi; r + dt, v2; r) (5.269)
Substituting Eq. (5.269) into Eq. (5.268) and adding this to Eq. (5.267), we obtain an
expression for C ( ^ ) as
C(VO = - V - 0 + x (5.270)
where 6 is the collisional transfer contribution
2 Jvl2l2k>o
x / ( 2 ) (r, v i ; r + d*, v2; f) dftdv, dv2 (5.271)
and / is the source term

d2 f
X = -* / {f 2 + f[ ~ *i ~ to) (V12 *)/ ( 2 > (r - dJt, V,; r, v2; r) d*dv, dv2 (5.272)
1
Jvi2k>0
Finally, for collisional dense phase suspensions of hard, smooth, and inelastic spherical
particles, the transport theorem is expressed as
| - (nf) = n { D f ) - V (nvir) - V 0 + X (5.273)

ot
where Df, 6, and x are given by Eqs. (5.260), (5.271), and (5.272), respectively.
5.5.2 Hydrodynamic Equations

The fundamental quantities which are conserved in a collision are the mass of
a particle and each component of the total momentum of the colliding particles. Thus,
conservation equations can be obtained by the use of the transport theorem. Although the
energy equation can also be derived from the transport theorem, the total kinetic energy of
the colliding particles is not conserved because of energy dissipation in the form of heat
loss.
5.5.2.1 Conservation of Mass

Setting ifr = m, Eq. (5.273) leads to
a
(otvpp) + V (QfpPpf/p) = 0 (5.274)
ot
where appp = nm and Up = (v).
5.5.2.2 Conservation of Momentum

By taking \j/ = mv, we can obtain the local form of the balance equation of the linear
momentum as
fdUp \ F
\ dt J m
where the total stress tensor Pp is the sum of the kinetic part P k and the collisional part Pc.
Pk is given by
P k = (apPpuu) (5.276)
where u = v (v) = v Up. Pc is given by
P c = Oimv) = m d l \ (v\ - v^)(y n k)k 1 - - ( d t V) + (dt V) 2 +

2 Jvi2-k>o ^ ' L 2 3! J
x / ( 2 ) ( n , vi; r 2 , v2; t) dtdvi dv2 (5.277)
5.5.2.3 Equation of Energy

Taking i/r to be \/2mv2, it is noted that
(5.278)
where qc is the collisional contribution of the heat flux of fluctuation energy qp and is
expressed in terms of \/2mu2 as given in the following:
1 2
qc = o(-mu
2
= -\mdl f (W? - M )(v12 k)k\\ - i(d*. V) + i(dJk. V)2 + .1
x / ( 2 ) (ri, vi; r 2 , v2; 0 dA:dvi dv2 (5.279)
If we define Tc as the granular temperature to represent the specific kinetic energy of the
velocity fluctuations or the translational fluctuation energy in such a way that
(5.280)
we obtain the energy equation in the form
ôfppp (^ + Up Vr c j = - P p : Vf/p - V qv + y (5.281)
where y is the collisional rate of dissipation per unit volume and is given by
Y =
= \md2 f (vf + v'2 -v2- v2)(vn * ) / ( 2 ) ( n , v i ; r, v2; 0 d*dv! dv2 (5.282)
and qp is the heat flux of fluctuation energy, which consists of the kinetic contribution qk
and the collisional part qc. The kinetic part qk is given by
qk=-(apPpu2u) (5.283)
Consider the collision of two spherical particles that are identical, smooth, but inelastic.
The loss of kinetic energy in the collision is given by Eq. (2.6)
l
-m(vf2 + vf2 -v2- v2) = l-m{e2 - 1)(* v 1 2 ) 2 (5.284)
Hence, the collisional rate of dissipation and the collisional stress tensor are related to the
coefficient of restitution by Eq. (5.284) and Eq. (5.277).
5.5.3 Collisional Pair Distribution Function

In order to evaluate the collisional integrals Pc,qc, and y, explicitly, it is important
to know the specific form of the pair distribution function / ( 2 ) (î ? r\\ V2, r^\ t). The pair
distribution function / ( 2 ) may be related to the single-particle velocity distribution function
/ ( 1 ) by introducing a configurational pair-correlation function g(ri, r 2 ). In the following,
we first introduce the distribution functions and then derive the expression of / ( 2 ) in terms
of / ( 1 ) by assuming / ( 1 ) is Maxwellian and particles are nearly elastic (i.e., 1 e <, 1).
Define an L-particle configurational distribution function (L) (ri, r 2 , . . . , r\) such that
(L)
(r\, r 2 , . . . , nJSri " -8rL is the probability of finding a particle in each of the volume
elements 8r\, <5r 2 ,..., 8r^ centered on r\, r 2 , . . . , r^. The one-particle distribution function
(1) (r) is just the number density n of particles at r. For a homogeneous bulk phase, we
have
n( (r) = ^ = n (5.285)
where iVp is the averaged number of particles in the volume V.

For an amorphous mass of particles there is no correlation between particles that are far
apart. The joint probability of finding particles at r\ and r2 is simply the product of the
individual probabilities. Let us define a configurational pair-correlation function g (r\, r 2 ) as
n(2)(rur2) / c _
g(rur2) = 5 (5.286)
so that
(5.287)
For a gas at equilibrium, i.e., no mean deformation, there is a spatial homogeneity and
thus g(r\,r2) depends only on the separation distance r = \r\ r2\. Then g = go(r) is
termed the radial distribution function, which may be interpreted as the ratio of the local
number density at a distance r from the central particle to the bulk number density. For a
system of identical spheres, the radial distribution function go(r) at contact {i.e., r = dv)
can be expressed in terms of the volume fraction of solids av as
ap) = + 3ap
+ / /p (5.288)
(1 - a p ) 2 ( 1 - Qfp)2 2 ( 1 - Qfp)3
which is based on a semiempirical equation of state [Carnahan and Starling, 1969].
Assume that a complete pair distribution function can be expressed as the product of the
spatial pair distribution function and the two single-particle velocity distribution functions.
Thus, we have
/ (2) (vi,r,; V2,r2, t) = g(run, t)fw(vuru t)fm(v2,r2, t) (5.289)

The anisotropy in g (r\, r2) may be determined by the use of a kinematic argument. Consider
a bulk of particle subjected to a mean shear flow. The radial distribution function, which
is spherical in equilibrium, becomes distorted into an ellipsoidal distribution as a result
of the presence of the mean shear. Hence, in order for g(r\, r2) to exhibit an anisotropy,
g(r\, r2) should depend not only on a p , r\, and r2 but also on Tc, v\, and v2. For dimensional
homogeneity, g can only be a function of ap, k U2\/T^2, and U\x/Tc. For a small
deformation rate (or when the magnitude of U2\ is small relative to T^2), it is assumed that
g(r\, ri) takes the form [Jenkins and Savage, 1983]
,, r2)= so [ l - ^ = ] (5-290)
Assuming that the single-particle velocity distribution is Maxwellian and takes the form
we can express the complete pair distribution function at collision as
47r 2 r c l r c 2 /
x exp + (5.292)
2Tcl 2Tc2
Consider two spheres in contact at r, as shown in Fig. 5.11, such that r\ = r (dv/2)k
and r2 = r+ (dp/2)k. We may expand / ( 2 ) in terms of the Taylor series about the point of
contact r. This allows the evaluation of the mean fields and the corresponding derivatives at
the contact point. Discarding the terms with spatial derivatives higher than the first order,
we can obtain the approximation of this expansion [Chapman and Cowling, 1970] as
F X K V
2? ' 2 ; 5O7lO 720 / TTT * rs
/20
Figure 5.11. Schematic diagram of two spheres in contact at r.
where/ ) (i= 1,2) is
(5.294)
5.5.4 Constitutive Relations

Now, we proceed to the evaluation of the collisional integrals P c , qc, and y by
using Eq. (5.293) as the form of a collisional pair distribution function. Use the coordinates
in Fig. 5.12, in which ez is chosen to be parallel to the relative velocity v\2. 0 and 0 are
the polar angles of k with respect to ez and the plane of ez and ex, respectively. ex, ey, and
ez are the three mutually perpendicular unit vectors corresponding to each coordinate in
Fig. 5.12. k, as mentioned in the previous section, is the unit normal on the collision point
directed from the center of particle 1 to the center of particle 2. Thus, we have
k ex sin 6 cos 0 + ey sin 0 sin 0 + ez cos 0 (5.295)
The element of solid angle dk is given by
(5.296)
Note that the integrals are to be taken over all values of k for which v\ 2 k is positive. Since
v12 k = V\2 cosO, the limits of integration are, for 0, 0, and 7r/2, and for 0, 0, and 2n. All
terms of the integrand containing odd powers of sin 0 or cos 0 become zero.
It is noted that the bouncing velocities of the colliding particles may be expressed by
, _ 1
2
(5.297)
The first-order approximation of the collisional stress tensor Pc may be simplified from
Eq. (5.277) and Eq. (5.297) as
Pc ^ -md3(l + e) I k(y\2 - k)3f(2\ri, v\\ r2, v2; t) dkdv\ dv2 (5.298)

4 Jvl2k>0
k
A
Figure 5.12. The coordinate system for the evaluation of collisional pair function.
Similarly, from Eq. (5.284), qc can be expressed as
2, v 2 ; dv2 (5.299)
where uf u\ can be further expressed by Eq. (5.297) as
(5.300)
From Eq. (5.282) and Eq. (5.284), y becomes
^-^"-^l,,, (vI2 kff(2\ri, vur, v2; dv2 (5.301)
It is convenient to derive a set of basic integrals to aid in the integration of Eqs. (5.298),
(5.299), and (5.301). From Eq. (5.295) and Eq. (5.296), we have [Chapman and Cowling,
1970; Jenkins and Savage, 1983]
k(vu k) dk = v i2 (5.302)
kk(vl2 . k)2 dk=^ (2v12v12 + vj2l) (5.303)
where / is the unit tensor.

If a is any vector independent of 6 and 0, it can be obtained that
(5.304)
and
f(a k)2(vn kf dk = ^v12 [3(a v 1 2 ) 2 + v22a2] (5.305)

Note that in Eq. (5.293)
fiofio = - ^ exp \-^- [(V! - Uv)2 + (v2 - Up)2]\ (5.306)

c L c J
and
ln
ff^\ 1
n^T = ^ F [("i - tfp) ~ ^ ~ ^P> ] (5-307>
2rc
V/ioV
Hence, after some manipulation, the constitutive relations can be obtained [Jenkins and
Savage, 1983; Lun etal., 1984] as follows:
(1) Collisional flux of fluctuation energy qc
qc = -kVTc (5.308)
where
^ (5.309)
(2) Collisional stress tensor Pc
y/nTc- trD )l - D (5.310)

dpp 5 ) 5
d 5 ) 5
where D is the tensor defined as
where the superscript 71 denotes the transpose, and "tr" denotes the trace. Thus,
trD = Da = Dn + D22 + D33 (5.312)
where Da represents the Einstein summation.

(3) Rate of energy dissipation y
6d-e)*L ( . r t J ,-, t r D | (5-313)
So far, we are able to construct the constitutive equations for qc, P c , and y. For mod-
erate solids concentrations, we can neglect the kinetic contributions in comparison to the
collisional ones. Thus, we can assume / \ <C Pc and qk <^C qc. Substituting the constitutive
relations into Eqs. (5.274), (5.275), and (5.281), after neglecting the kinetic contributions,
yields five equations for the five unknowns a p , l/ p , and Tc (or (u2)). Hence, the closure
problem is resolved.
Example 5.2 On the basis of the kinetic theory, which is used to model collision-
dominated gas-solid flows, derive a general expression of solid stresses of elastic spheres
in a simple shear flow.
Solution For elastic particles, e = 1. For a simple shearflow,D can be expressed

as
1 0\
0 0 (E5.2)
2 dy
0 0 0
V
Therefore, trD = 0; and k is given by
(E5.3)
The collisional stress tensor Pc can be obtained from Eq. (5.310) as
1 0 0\ /0
1 oI- ^WâJW^^ I1 0 0 I (E5.4)
L o o i \o o o>
The first term gives rise to the particle phase pressure, which, at low shear rates, is
isotropic. The second term is the shear stresses, which indicate the effective viscosity in
the form of
12 IT
(E5.5)
Thus, in this case, the elastic granular flow is Newtonian.
Example 5.3 Consider the two-dimensional gravity flow of granular materials down an
inclined plane, as shown in Fig. E5.1 [Savage, 1983]. Assume that theflowis fully developed
and collision-dominated with a free surface.
Figure E5.1. Granular flow down an inclined chute.
(1) From the momentum equation and the constitutive relation of collisional stress tensor,
show that Tc is governed by
(E5.6)
where ft is
0 = tan 2 ? - 6(1 - e) (E5.7)
(2) Assume that there are no slip of solids and no flux of fluctuation energy at the wall.
Determine the distributions of TC,UP, and ap.
Solution (1) From the momentum equation Eq. (5.275) and the constitutive re-
lation of collisional stress tensor Eq. (5.310), we have
ry
/ apPpdy (E5.8)
Jo
and
P = - = g sin f / appp dy (E5.9)

y
5 dy Jo
Equations (E5.8) and (E5.9) lead to
(E5.10)
The fluctuation energy equation Eq. (5.281) can be expressed by
dTc dUn d I di c \
Pp-^-0=-PCxy-f + - U - f -K (E5.ll)
where the rate of dissipation is given by Eq. (5.313)
6(1 - e)
_ IK, A Q î_( J . 1 i<^
Equation (E5.11) gives the relationship among the shear work, the energy flux gradient,
and the rate of dissipation. Substituting Eq. (E5.9) and Eq. (E5.10) into Eq. (E5.ll) yields
Eq. (E5.6).
(2) From Eq. (E5.8), the following approximation holds
~^/TTT'C^ gcosi;apppy (E5.13)

dp
where the averaged value of ap is defined by ap as given in the following:
1 fh
ap = - apdy (E5.14)
n Jo
Substituting Eq. (E5.13) into Eq. (E5.6) and defining / as Tc1/2, we have
(E5 15)
d , ' - -
which is the Bessel equation of zero order. At the free surface, there is no flux of fluctuation
energy. Thus, it gives
y~) =0 (E5.16)
d
yJy=o
Note that for a fully developed flow, an energy flux from the granular materials to the
wall corresponds to ft > 0, whereas an energy flux from the wall to the granular materials
corresponds to /? < 0. For zero energy flux into the wall, /3 = 0, i.e.,
tan = tan Co = \ -- (E5.17)

V 5n
where 0 is the natural angle of repose.
Applying Eq. (E5.13) at the wall and using Eq. (5.309) yield the fluctuation-specific
kinetic energy or the "granular temperature" Tc as
where apw is the volume fraction of the particles at the wall. Thus, from the nonslip
condition at the wall, we can obtain the velocity profile from Eq. (E5.10) as
l ^ - y) (E5.19)
The distribution of volume fraction of the particles can also be obtained from Eq. (E5.13)
and Eq. (E5.9) as
- = - (E5.20)
Two interesting points can be noted from this example. First, for elastic materials, 0 = 0
and it is impossible to have an energy flux from the wall to the granular materials. Second,
for inelastic materials, if fluctuation energy is put into the materials by using a vibrating
device, the slopes required for the chute flow can be considerably less than their natural
angle of repose.
5.6 Equations for Flows Through Packed Beds

The physical understanding of gas flows through a packed bed is important for the
study of the gas-solid transport system as it represents the limiting case of the gas-solid
transport. A typical example is the minimumfluidizationcondition of gas-solid fluidization.
In a gas-solid flow system, the presence of unsuspended particles is also common, and,
thus, understanding the gas flow in a packed bed is essential.
Experiments were conducted by Darcy (1856) to study viscous flows through homo-
geneous porous media, which resulted in the well-known Darcy law. There are various
models from which Darcy's law may be derived to describe gas flow through a packed bed
[Scheidegger, 1960; Bear, 1972]. Among them, the capillary tube model is considered to
be the simplest. In this model, the porous medium is characterized by a bundle of straight
circular capillary tubes with permeability of the tube bundle equal to that of the actual
medium. Thus, the pressure drop through a single, straight circular tube can be described
by the Hagen-Poiseuille equation derived for the steady tubular flow.
The Darcy law does not consider the effects of inertia. For a more general description of
flows in a packed bed, particularly when the effects of inertia are important, Ergun (1952)
presented a semiempirical equation covering a wide range of flow conditions, which is
known as the Ergun equation [Ergun and Orning, 1949; Ergun, 1952].
5.6.1 Darcy's Law

To study the general behavior of viscous flow through homogeneous porous media,
experiments were designed and performed by Darcy (1856). A schematic drawing of Darcy's
experiment is shown in Fig. 5.13. A homogeneous filter bed, of length L in the test section
and uniform cross-sectional area A, is filled with an incompressible liquid. The pressure
difference for a given flow rate is measured by the difference of the manometer fluid level,
Ah. By varying the experimental variables, the original form of Darcy's law can be deduced
as
KAAh
(5.314)
Figure 5.13. Schematic diagram of Darcy's filtration experiment (from Darcy, 1856).
224 5 I Basic Equations
where Q is the volumetric flow rate and K is a constant depending on the properties of the
fluid and of the porous medium. Ah can be expressed in terms of the pressure difference
A P. Furthermore, to separate the influence of the porous medium from that of the fluid, the
constant K can be replaced by the ratio of the specific permeability of the porous medium
k to the viscosity of the fluid /z. Hence, Eq. (5.314) becomes
k AP
Q = A (5.315)
[i L
or in a differential form
(5.316)
dz
where U is the superficial velocity of the fluid.
Equations (5.315) and (5.316) are the commonly used forms of Darcy's law. It is noted
that the gravitational effect is considered negligible in the Darcy's law analysis.
5.6.2 Straight Capillaric Model

The simplest capillaric model is the one representing a porous medium by a bundle
of straight parallel capillaries of average diameter <5, as shown in Fig. 5.14. The equivalent
voidage a can be related to the averaged diameter by
a = - (5.317)
4
where n is the number of capillaries per unit area of cross section.
Let us first consider the case when aflowpasses through a single circular tube. The flow
is assumed to be steady, incompressible, highly viscous, and fully developed. Thus, the
radial velocity distribution can be obtained as (Problem 5.6)
(5.318)
The flow rate through the tube becomes
(5.319)
\
0 \
0 (r
0 0
a \ 1
1 f;
Nv
\
0 0
\s0 craa- ' ()
o-
!
1 \. - ':
N 0 i8 ^\\ i)
Figure 5.14. Straight capillaric model.

which is the Hagen-Poiseuille equation. Equation (5.319) indicates that the pressure drop
is proportional to the flow rate (or the cross-sectional averaged velocity) for highly vis-
cous pipe flows. The flow per unit area, or the superficial velocity in a porous medium,
becomes
n p) (5320)
î
By comparing the preceding equation with Darcy's law, the specific permeability k can be
given as
*= ^ (5.321)
It is noted that straight capillary tubes may not portray the complex structure of the
porous medium. Thus, in practice, the factor 32 in Eq. (5.321) is commonly replaced by an
empirical parameter known as tortuosity. The tortuosity accounts for the tortuous paths of
the porous medium.
5.6.3 Ergun 's Equation

The generalized relation for the pressure drop for flows through a packed bed was
formulated by Ergun (1952). The pressure loss was considered to be caused by simultaneous
kinetic and viscous energy losses. In Ergun's formulation, four factors contribute to the
pressure drop. They are (1) fluid flow rate, (2) properties of the fluid (such as viscosity and
density), (3) closeness (such as porosity) and orientation of packing, and (4) size, shape,
and surface of the solid particles.
5.6.3.1 Effect of Flow Rate

The effect of the flow rate on the pressure drop for a flow through a packed bed can be
analyzed using two different approaches. The first approach treats the flow as an assembly
of many internal channel flows between the particles so that the pressure drop is due to
the wall friction of the channel flows. In the second approach, the pressure drop of the
flow through the packed bed is considered to be caused by the drag forces of the particles.
The drag force of a single particle is proportional to the fluid velocity in the creeping flow
regime and is proportional to the square of the velocity in the inertia regime. Thus, the
pressure drop through a fixed bed is proportional to the fluid velocity at low flow rates, and
approximately to the square of the velocity at high flow rates. As suggested by Reynolds
(1900), for a general flow through a packed bed, the pressure drop can be expressed by the
sum of two terms, i.e., one proportional to the first power of the fluid velocity and the other
proportional to the product of the density of the fluid and the second power of the fluid
velocity
=aU + bpU2 (5.322)

Li
where a and b are factors to be determined.
5.6.3.2 Effect of Fluid Properties

For a limiting case where the velocity U approaches zero, Eq. (5.322) yields
a i5323)
!&h=
With the aid of the straight capillaric model and according to the Hagen-Poiseuille equa-
tion and Darcy's law, a is proportional to the viscosity of the fluid. Thus, Eq. (5.322)
becomes
= a'nU + bpU2 (5.324)

JLJ
5.6.3.3 Effect of Voidage of Packing

The voidage a may affect the pressure drop in a column containing solid particles through
the parameter (1 a)1 /ot\ where i and j are exponents to be determined. As suggested by
Ergun and Orning (1949), the flow rate of the fluid through the porous medium varies with
the voidage of the medium. The values of af and b in Eq. (5.324) vary with theflowregimes.
The dependency of the viscous energy loss and the kinetic energy loss on the voidage can
be illustrated by the Kozeny theory [Scheidegger, 1960] and the Burke and Plummer theory
[Burke and Plummer, 1928], respectively, as discussed in the following.
A. KOZENY THEORY
Kozeny (1927) proposed equations accounting for the relationship between the pressure
drop and the voidage of the porous medium under low flow rate conditions [Ergun and
Orning, 1949]. A detailed discussion of the Kozeny theory is given in Scheidegger (1960).
It is assumed that the porous medium can be represented by an ensemble of channels
of various cross sections, but of definite length. The channel flow is horizontal, lami-
nar, and fully developed. The inertial effect is negligibly small compared to the viscous
effect.
Denote Aa as the pore cross-sectional area and A as the total cross-sectional area of the
column. The voidage is
a = ^ (5.325)
A
Now, consider the flow through a single straight channel of an irregular cross section. If a
cross section of the tube is chosen in the x-y plane and the fluid flows in the z-direction,
the equation of motion can be expressed as
d2W d2W 1 dp
~^~T + V T = ~~T~ (5.326)
dx2 dy2 [i dz
Introducing new variables
Eq. (5.326) becomes
+
d^ dr,2
which has the solution
^Kl.ij) (5-329)
az
where the function \jr is given by
~ .OnCr + ry) (5.330)
where 4> denotes a harmonic function. Moreover, assuming a no-slip condition at the wall
for fluid flow, we have
^^ 5> (5.331)
where (x0, jo) is a point on the boundary.
Let C be the circumference of the cross section of the pores and R be the equivalent
radius of a circle having the same circumference, i.e.,
R=~ (5.332)
Then, we may express Eq. (5.330) in polar coordinates r and 0 as
where B is defined as
>n(re/0)^ - ^ (5.334)
The averaged interstitial velocity u can be obtained by
u = I [wdA = --^^fBm (5.335)

AJ Io7rz/x dz
where
Bm = Brdrdc/) (5.336)
If the expansion of the circumference is represented by )3 = C2/Aa, it is clear that B and

u become smaller ifftbecomes larger. When /3 approaches infinity, u approaches 0. In this
case, from the analysis based on Eq. (5.335), Bm must decrease faster than the expansion
of \6n2c/p. This requirement can be satisfied if Bm is expressed by
where r > 1 (5.337)
c in Eq. (5.337) is known as the Kozeny constant, which depends only on the shape of the
cross section.
Consider a packed bed whose fluid cross section consists of n single-pore cross sections
with each having the cross-sectional area Asa and circumference C s , i.e.,
Aa = nAsa, C = nCs (5.338)
Substituting Eq. (5.337) and Eq. (5.338) into Eq. (5.335) yields
c dp Aa c dp A[a 1
U= 7--^r = : oTZo-TT (5.339)
/JL dz pt~l \i dz ct n
Since the interstitial velocity u should not depend on the number of pores, must be
equal to 2. Equation (5.339) then reduces to
u = -L*L*k Lf* (5.340)

/xdzC2 fi dz S1
where S is the specific surface area of the pore and is defined as
S= ^ (5.341)
A
The superficial velocity U can be extended from Eq. (5.340) as
fi dz S2
In order to generalize the expression of S for various packing materials, S is redefined
as the total surface area per unit packed volume and is expressed by
S = S0(l-a) (5-343)
where So is the surface area exposed to thefluidper unit volume of solid materials. Equation
(5.342) thus becomes
Therefore, for viscous energy loss at low flow rates, we have

Ap (!-)
QA -
L cr
B. BURKE AND PLUMMER THEORY
For kinetic energy loss at high flow rates, the theory of Burke and Plummer [Ergun and
Orning, 1949] assumes that the total resistance of the packed bed can be treated as the sum
of the resistances of the individual particles. For the fully developed turbulent flow, the
drag force acting on an isolated spherical particle is
FD = fpdy (5.346)
where / is the friction factor and is a constant.
Consider a packed column filled with spheres of equal diameter dv. The number of
particles per unit of packed volume is given by
n = -^-(l-a) (5.347)
itdl
Assume the drag force, in the form of Eq. (5.346), is acting on each particle. The rate of
work done by the drag force, or the power, is equal to the product of the interstitial velocity
and the force acting on the particles. Hence, we have
P = nfpdlu2u = 6-u 3(l - a) (5.348)

p
n dv
where P is the power per unit volume of packed bed. On the other hand, it is clear that P
can also be related to the pressure gradient and superficial velocity by
P = U (5.349)
Combining Eq. (5.348) and Eq. (5.349) yields
L n dv a5
For spherical particles, the specific surface area, So, can be expressed in terms of the
particle diameter, dp, as
So = T (5.351)
dp
From Eqs. (5.350) and (5.351), it gives
2
^ 4 ^ (5.352)
L TC a3
which indicates
% ^ (5353)
Therefore, by substituting Eq. (5.345) and Eq. (5.353) into Eq. (5.324), we obtain
^=W<Z*+y.f)U2*Z*L (5.354)
L a5 a5
5.6.3.4 Effects of Size, Shape, and Surface Area of Particles
For viscous energy loss, from Kozeny's equation, the pressure drop is proportional to the
square of the specific surface area of solids SQ. For kinetic energy loss, from Burke and
Plummer's relation, the pressure drop is proportional to So. So is related to the particle
diameter by Eq. (5.351) for spherical particles; for nonspherical particles, the dynamic
diameter (see 1.2) may be used for the particle diameter. The general form of the pressure
drop can be expressed as
Ap (i )2tiU , (1(l-a)pU
- )
2
k*
L M
a3 K2
a3 dp (5.355)
where k\ and &2 are universal constants which are determined experimentally. Equation
(5.355) can be rearranged into a linear form
y = ki+k2x (5.356)
where
x - f>Udv ( 5 357)
y x
y
L (l-a)Vf/' n(l-a)
By plotting (y/x) versus x in a logarithmic scale based on experimental data, the values
of k\ and &2 given in Eq. (5.356) can be determined. The results shown in Fig. 5.15 indicate
200
100 \
Experimental
\
\
V
5
10 B
in
ft,Q
1
1 10 100 1,000 4,000
Re p /(l-a)
Figure 5.15. Pressure drop in fixed beds (from Ergun, 1952).
that k\ and &2 are 150 and 1.75, respectively. The Ergun equation is thus expressed by
[Ergun, 1952]
(5.358)
5.7 Dimensional Analysis and Similarity

Dimensional analysis is an analytical method wherein a number of experimental
variables that govern a given physical phenomenon reduce to form a smaller number of
dimensionless variables. A phenomenon which varies as a function of k independent pa-
rameters can be reduced to a relationship between (k m) dimensionless parameters, where
m is the number of dimensionally independent parameters.
The resulting dimensionless groups from the dimensional analysis can be used as the basis
to account for the similarity or scaling relationships or laws for different design and operating
situations. There are two common approaches in dimensional analysis to obtain the dimen-
sionless variables: application of the Buckingham pi theorem and nondimensionalization
of the governing equations and relevant boundary conditions. From these dimensionless
variables, completely independent dimensionless parameters can be selected as the basis
for the scaling law application. The validation of these scaling relationships depends on
the appropriateness of the governing equations used and their related boundary conditions.
5.7.1 Scaling Relationships for Pneumatic Transport of Dilute Suspensions

Scaling for a dilute gas-solid suspension may represent a simple case of using the
nondimensionalization method. The following example is a simplified version of the work
ofChao(1982).
For pneumatic transport of solids in a dilute suspension, the effects of apparent mass,
Basset force, diffusion, and electric charge of the particles may be ignored. Thus, the
dynamic equation of a small particle in a gas medium is given by
,- -,., (5.359)
Lit ^rp
If we denote the displacement vector of the particle by r p , we have
tfp = ^ (5.360)
By introducing a reference length / and a reference velocity V, we may define the nondi-
mensional quantities
(5361)
v ^^y
(5362)
where eg is a unit vector in the direction of the gravitational acceleration; St is the Stokes
number; and Fr is the Froude number. St and Fr are defined as
Vr V2
St=-^, Fr=f- (5.363)
/ Ig
The dimensionless variable St is not always useful in the analysis because the relaxation
time of the particle r^ in the expression depends on the particle Reynolds number Re p in a
general form
(5 .364)
where C is the non-Stokesian correction factor, which can be correlated as a function of

Rep. In the Stokes region, C = 1. The particle Reynolds number Rep can be expressed as
Rep = ^l\U-Uv\= Re;|IT - li;\ (5.365)
where Re* is the pseudoparticle Reynolds number, which is given by
Re; = ^ (5.366)
P
M
In view of the preceding discussion concerning Eqs. (5.363) and (5.364), a similarity of St
requires that, in addition to Re, another dimensionless parameter *I>, given in the following,
be similar
18/z/
where *I> is termed the inertia parameter. Thus, for geometrically similar systems with
kinematically similar boundary conditions, based on the model, the dynamic similarity can
be obtained if Re*, *I>, and Fr are the same.
5.7.2 Scaling Relationships for Fluidized Beds

The derivation of scaling relationships for a dense gas-solid suspension in flu-
idized beds has been attempted on the basis of scale effects on the fundamental variables
governing the dispersed phase and the interstitial gas dynamics [Horio et aL, 1986]. How-
ever, the scaling relationships for fluidized beds may be more generally established by
using the nondimensionalization method accounting for the system behavior, as discussed
later [Fitzgerald and Crane, 1980; Glicksman, 1984; 1988]. In the governing equations of
this method, it is assumed that the interparticle forces, such as the collisional forces and
the electrostatic forces, are negligible. In addition, the gas-solid suspension in a fluidized
bed is treated as if in a steady-state laminar flow condition without phase change. The
following derivation follows the approach used by Glicksman (1984). A limiting case for
low Reynolds numbers is also considered.
The equations of conservation of mass for gas and particle phases are given, respectively,
by
V (alJ) = 0 (5.368)
= 0 (5.369)
The equation of motion of the gas can be written as
+ (U - t/p) = 0 (5.370)
ap ap
where fi(U Up) represents the drag force between the gas and the particles.
The equation of motion of the particles can be expressed, in a similar way, as
/-t/ p )=O (5.371)
The relevant boundary conditions for this system are given as follows:
(1) For particle velocities at solid walls such as the side wall of the bed in a batch-solids
situation
= 0 (5.372)
where /PN is the component of the particle velocity normal to the surface.
(2) For gas velocities at the base of the bed as well as at a height above the expanded bed
surface when a screen or a fixed bed is used as the distributor
U = Uex (5.373)
where U is the superficial gas velocity and ex is the unit vector along the axial direction.
Along the bed walls
U=0 (5.374)
(3) For the pressure at the side walls
= 0 (5.375)
dn
Along the bottom surface, for a uniform distributor

. P Umf*-*mf /c T-IS\
p = Pi-\ (5.376)
where p f represents the pressure in the freeboard region; Umf is the minimum fluidization
velocity; Lmf is the characteristic length at minimum fluidization; and amf is the bed voidage
at minimum fluidization. Along the top surface, above the expanded bed surface
P = Pf (5377)
By introducing the nondimensional quantities given next
p
U' U' / ' pU2
the nondimensional equations, corresponding to Eqs. (5.368) through (5.371), become
V* (all*) = 0 (5.379)
V* [(1 - a)Ul\ = 0 (5.380)
U2 a a pU ^ p
^
The dimensionless boundary conditions corresponding to Eqs. (5.372) through (5.377) can
then be expressed as
U;N = 0 (5.383)
U* = 1 (5.384)
U*=0 (5.385)
= 0 (5.386)
dn*
The boundary condition for the pressure given in Eq. (5.376) can be rearranged as
(P ~ Pi) = PUmfLmf/amf = ppgapl (5.387)
Equations (5.377) and (5.387) can be nondimensionalized by the term pU2 as
(P ~ Pi)
pU2 pU2am{ P \U2J
Pi (5.389)
From Eqs. (5.379) through (5.389), the governing nondimensional parameters can be
identified as
PI gl D p p{
PU' ip- T 7P ' W1 (
The nondimensionalized pressure term, pf/pU2, can be ignored when the gas velocity is
low compared to the speed of sound or when the effects of pressure on the thermodynamic
properties are not significant. Note that U*, a, and ap are nondimensional dependent
variables which are identical for two similar fluidized beds.
It should be noted that the coefficient fi is not an independent parameter; it depends on
the bed properties. In a fluidized bed, particles are closely spaced; it is assumed that the
Ergun equation can be applied to account for the pressure drop in the bed. Thus, from
Eq. (5.358), we have
* E = p l U - Up] = 1 5 O ( i > ^ g p l + L 7 5 ( i ) P ( g g P ) (5.391)
/ a2 d2 a dv
Rearranging this equation yields
pU a2 pUd2 a dv U
The second term in the right-hand side of Eq. (5.392) represents the fluid inertia, whereas
the first term represents the viscous contribution to the pressure drop. At low Reynolds
numbers, the Ergun equation can be simplified by neglecting the inertial term. Under this
condition, Eq. (5.392) can be expressed as
l 5 0 ^ ( ^ ) ( ) (5,93)
2
pU \UdJ\dJ
Equation (5.393) shows that (/31/pU) is only a function of (pUdp//x) and (dp/l) for low
Reynolds number cases. In this limit, the independent nondimensional scale-up parameters
can be identified as
pUdp gl dp D P
(5.394)
ft I' Pv I'
Further discussion on scaling relationships for various fluid-particle systems is given in
Glicksman et al. (1994).
Example 5.4 Design a geometrically similar laboratory-scale cold model fluidized bed
to simulate the hydrodynamics of a large-scale fluidized bed combustor. Also specify the
operating conditions for the cold model. The combustor is a square cross section column
with a width of 1.0 m and a height of 6 m. The fluidized bed combustor is operated at a
temperature of 1,150 K, a superficial gas velocity of 1.01 m/s, and a bed height of 1.06 m.
Particles with a density of 2,630 kg/m3 and a diameter of 677/zm are used for the combustor.
The cold model is operated at a temperature of 300 K. Air is used for both the cold model
and hot model fluidized beds. The physical properties of air are
pc = 1.16 kg/m3 p h = 0.3075 kg/m3
Me = 1-846 x 10"5 kg/m s fih = 4.565 x 10~5 kg/m s
at 300 K and 1 atm at 1,150 K and 1 atm
Solution The similarity analysis of Eqs. (5.378) through (5.394) indicates that
for low Reynolds number (in this case, Reynolds number = 4.61), the dimensionless groups
which must be matched between the model and the prototype are
gl dv D p
2 (E5.21)
fi ' u' r r pp
Since the temperature influences only the physical properties of air, the design of the hot
model can begin with the matching of the Reynolds number. The similarity condition
Udpp Udvp
(E5.22)
h
requires that
Ucdpc _ ph/ic
= 0.1072 (E5.23)
The subscripts c and h refer to the cold and hot model, respectively. With the equality of
the Froud number, U2/gl, the scale factor for the superficial gas velocity can be expressed
as
The condition of geometrical similarity requires
^ = ^ (E5.25)
Solving Eqs. (E5.23) through (E5.25) simultaneously yields that Uc/Uh = 0.4750 and
dpc/dph = /c//h = 0.2257.
The similarity of D/l requires that
l
Rl = = 0 2257 (E5.26)
Similarly, we have
^ = El = 3.77 (E5.27)
In terms of the preceding scale factors, the operating conditions for both the hot and cold
models can be obtained as given in Table E5.2.
Table E5.2. The Operating Conditions for the Hot Model
and Cold Model Fluidized Beds
Operating conditions Cold model Hot model
Temperature (K) 300 1,150

Superficial gas velocity (m/s) 0.48 1.01
Bed height (m) 0.24 1.06
Particle density (kg/m3) 9,921 2,630
Particle size (mm) 0.153 0.677
Bed width (or length) (m) 0.2257 1.0
Nomenclature
A Cross-sectional area Momentum transfer vector across
Ac Maximum contact area the interface per unit volume due
Ak Interfacial area of phase k to the mass generation of phase k
a Particle radius Fo Fourier number
Ba Bagnold number Fr Froude number
Collisional rate of increase Pair distribution function
C
Frictional factor
C Circumference
Single-particle velocity
C Specific thermal capacity of gas
distribution function
C\ Empirical constant in the k-e
f(2) Complete pair distribution
model
function
C2 Empirical constant in the k-e
model f Body force vector acting on the
unit mass of the particle
Cek Empirical constant, defined by
Eq. (5.162) G One-sided autospectral density
function
Cp Specific thermal capacity of
particle Generation of phase k
Empirical constant in the k-e Configurational pair-correlation
model function
Specific heat of particles Degeneracy of energy level
c
Kozeny constant Gravitational acceleration
c
Diffusivity go Radial distribution function
D
Column diameter or pipe Degeneracy of energy level ex
D
diameter Gravitational acceleration in the
D Operator in the dense phase radial direction
transport theorem Gravitational acceleration in the
Eddy diffusivity of phase k axial direction
Eddy thermal diffusivity of Gravitational acceleration in the
phase k azimuthal direction
Collisional shear stress tensor h Planck's constant
D
dv Particle diameter h Vertical coordinate
E Total energy per unit mass I Unit tensor
E Empirical constant, defined by J Mass flux
Eq. (5.87) J Jacobian determinant
e Internal energy per unit mass h Rate of heat generation per unit
e Coefficient of restitution volume by thermal radiation
e Directive unit vector Rate of heat generation per unit
Momentum transfer vector across volume
the interface per unit volume due Flux vector of a physical quantity
to the pressure and viscous Heat flux vector
stresses of phase k K Thermal conductivity
Nomenclature 237
Turbulent thermal conductivity of Pk Kinetic stress tensor of particles

phase k D
P
Total stress tensor of particles
k Boltzmann constant P Pressure
k Collisional thermal conductivity Pf Pressure in freeboard region
k Quantum number Pm Averaged pressure
k Specific permeability Q Volumetric flow rate
k Turbulent kinetic energy of the GAk Heat transfer across the interface
fluid phase Grk Internal energy transferred by the
kP Turbulent kinetic energy of mass generation due to phase
particles change
k Collisional vector q Energy flux of particles
L Bed length 9c Collisional heat flux vector
L Geometric dimension qk Kinetic heat flux vector
Characteristic length at minimum 9V Total heat flux vector of
fluidization fluctuation energy
I Interparticle distance R Lagrangian correlation
I Reference length coefficient
u Characteristic length of turbulent

eddy
R
Rmin
Radius of averaging volume
Minimum radius of averaging
Mixing length volume
m Mass of a particle Re Reynolds number
m* Relative mass, defined by Rep Particle Reynolds number
Eq. (2.128) r Radius
N Number of particles r Coordinate in r-space
Nv Number flow rate of the particle S Specific surface area
n Number of capillaries per unit So Surface area exposed to fluid per
area of cross section unit volume of solids
n Total number of particles in the St Stokes number
system s Arbitrary curve path
n Configurational distribution s Distance
function T Absolute temperature
n Particle number density T Averaging time duration
n Coordinate normal to a boundary Tc Granular temperature
Hi Number of particles having T Total stress tensor
energy ^ t Time
P Probability tc Contact time duration
P Power per unit volume of the U Superficial gas velocity
packed bed U Gas velocity component
P Total pressure tensor U\ Velocity of particle 1
P Position vector u2 Velocity of particle 2
pc Collisional stress tensor of Ui Gas velocity component in
particles direction i
238 5 I Basic Equations
/j Gas velocity component in u Relative velocity vector of a

direction j particle
Umf Minimum fluidization V Gas velocity component
velocity V Particle volume
Un Gas velocity component normal V Reference velocity, defined by
to the wall Eq. (5.361)
Up Particle velocity component v Instantaneous velocity of a
Upi Particle velocity component in particle
direction i vf Fluctuating radial velocity
t/pj Particle velocity component in component of gas
direction j Vp Instantaneous radial velocity
UT Normal impact velocity component of particle
U'x Normal reflecting velocity after v Instantaneous velocity vector of a
collision particle
Us Speed of displacement of the W Gas velocity component
interface W Number of possible arrangements
Ut Tangential impact velocity WBE Total number of possible
U[ Tangential reflecting velocity arrangements for a certain set of
after collision nx in Bose-Einstein statistics
U Velocity vector of fluid Wi,BE Number of possible arrangements
for the \th level of energy for any
/* Dimensionless gas velocity
ri[ in Bose-Einstein statistics
vector
WMB Total number of possible
Uk Velocity vector of phase k
arrangements for a certain set of
U'k Fluctuating velocity vector of ri[ in the corrected
phase k Maxwell-Boltzmann statistics
Up Velocity vector of particles w' Fluctuating tangential velocity
/* Dimensionless particle velocity component of gas
vector Wp Instantaneous tangential velocity
u Interstitial gas velocity component of particle
u Instantaneous axial velocity x Cartesian coordinate
component of gas V Yield strength
uf Fluctuating axial velocity y Cartesian coordinate
component of gas y Normal distance from wall
u[ Gas fluctuating velocity z Cartesian coordinate
component in direction i z Cylindrical coordinate
wj Gas fluctuating velocity z Partition function
component in direction j
Greek Symbols
Up Instantaneous axial velocity
component of particle a Volume fraction of the gas phase
wp Fluctuating axial velocity Qfp Volume fraction of the particle
component of particle phase
wpi Particle fluctuating velocity
component in direction i
References 239
Mass generation rate of phase k /x Deformation viscosity of fluid

rk per unit volume Mek Effective viscosity of phase k
Y Collisional rate of dissipation per V Kinematic viscosity
unit volume p Density of fluid
Y Shear strain Density of particles
Yf Scalar quantity, defined by (J Total stress tensor of the fluid
Eq. (5.176) phase
8 Diameter of a capillary tube Empirical constant associated
êk
8 Relative deviation of calculated ap with the energy transport by
h Kronecker delta fluctuations in density and
Distance from wall temperature
6 Normal strain Ok Empirical constant in the k-e
6 Dissipation rate of turbulent model
kinetic energy Empirical constant in the k-e
Quantum energy model
6 Wall emissivity T Shear stress tensor
Averaged energy of a particle X Shear stress
Energy at quantum level i Eddy existence duration
tf Characteristic time of fluid
Generation or destruction rate of
particle turbulent kinetic energy motion
Interacting duration
f Random number
Flying time for a particle passing
e Cylindrical coordinate
through an eddy
0 Spherical coordinate Relaxation time of particle
rp
e Contact angle motion
e Collisional transfer contribution ts Stokes relaxation time
vector (|) Rate of generation of / per unit
K Von Karman's constant volume
X Mean free path (J) Averaged variable
X Parameter, defined by Eq. (5.255) (j) Dissipation function
Dynamic viscosity (j) Instantaneous variable
Transport coefficient of viscosity, 0 Azimuthal angle
defined by Eq. (5.43) X Collisional source term
Meff Effective viscosity Wave function
Turbulent viscosity Flux vector of /
// Fluctuation of fluid viscosity CO Angular frequency
References
Anderson, T. B. and Jackson, J. (1967). A Fluid Mechanical Description of Fluidized Beds. / & EC
Fund., 6, 527.
Bagnold, R. A. (1954). Experiments on a Gravity-Free Dispersion of Large Solid Spheres in a
Newtonian Fluid under Shear. Proc. R. Soc. London, A225, 49.
Basset, A. B. (1888). Hydrodynamics. Cambridge: Deighton, Bell; (1961) New York: Dover.
Baxter, L. L. (1989). A Statistical-Trajectory Particle Dispersion Model. Ph.D. Dissertation.
Brigham Young University.
Bear, J. (1972). Dynamics of Fluids in Porous Media. New York: American Elsevier.
Boussinesq, J. (1877). Theory de L'ecoulement Tourbillant. Mem. Pres. Par Div. Savants a L'acad.
Sci., Paris. 23, 46.
Boussinesq, J. (1903). Theorie Analytique de la Chaleur, 2. Paris: Gauthier-Villars.
Burke, S. P. and Plummer, W. B. (1928). Suspension of Macroscopic Particles in a Turbulent Gas
Stream. / & EC, 20, 1200.
Carnahan, N. F. and Starling, K. E. (1969). Equations of State for Nonattracting Rigid Spheres. /.
Chem. Phys., 51, 635.
Celmins, A. (1988). Representation of Two-Phase Flows by Volume Averaging. Int. J. Multiphase
Flow, 14, 81.
Chao, B. T. (1982). Scaling and Modeling. In Handbook of Multiphase Systems. Ed. G. Hetsroni.
New York: Hemisphere; McGraw-Hill.
Chapman, S. and Cowling, T. G. (1970). The Mathematical Theory ofNonuniform Gases, 3rd ed.
Cambridge: Cambridge University Press.
Crowe, C. T. (1991). The State-of-the-Art in the Development of Numerical Models for Dispersed
Two-Phase Flows. Proceedings of the First International Conference on Multiphase
Flows, Tsukuba. 3, 49.
Crowe, C. T., Sharma, M. P. and Stock, D. E. (1977). Particle-Source-In Cell (PSI CELL) Model
for Gas-Droplet Flows. Trans. ASME, J. Fluids Eng., 99, 325.
Culick, F. E. C. (1964). Boltzmann Equation Applied to a Problem of Two-Phase Flow. Phys.
Fluids,!, 1898.
Darcy, H. (1856). Les Fontaines Publiques de la Ville de Dijon. Paris: Victor Dalmon.
Delhaye, J. M. and Archard, J. L. (1976). Proceedings of CSNI Specialists9 Meeting, Toronto. Ed.
Bannerjee and Weaver. AECL, 1, 5.
Elghobashi, S. E. (1994). Numerical Models for Gas-Particle Flows. Seventh Workshop on
Two-Phase Flow Predictions, Erlangen.
Ergun, S. (1952). Fluid Flow Through Packed Columns. Chem. Eng. Prog., 48, 89.
Ergun, S. and Orning, A. A. (1949). Fluid Flow Through Randomly Packed Columns and Fluidized
Beds. I&EC, 41, 1179.
Fitzgerald, T. J. and Crane S. D. (1980). Cold Fluidized Bed Modeling. Proceedings of the Sixth
International Conference on Fluidized Bed Combustion, Atlanta, Georgia. 3, 815. U.S.
DOE.
Gidaspow, D. (1993). Hydrodynamic Modeling of Circulating and Bubbling Fluidized Beds. In
Paniculate Two-Phase Flow. Ed. M. C. Roco. Boston: Butterworth-Heinemann.
Glicksman, L. R. (1984). Scaling Relationships for Fluidized Beds. Chem. Eng. Sci., 39, 1373.
Glicksman, L. R. (1988). Scaling Relationships for Fluidized Beds. Chem. Eng. Sci., 43, 1419.
Glicksman, L. R., Hyre, M. R. and Farrell, P. A. (1994). Dynamic Similarity in Fluidization. Int. J.
Multiphase Flow, 20/S, 331.
Hinze, J. O. (1959). Turbulence. New York: McGraw-Hill.
Horio, M., Nonaka, A., Sawa, Y. and Muchi, I. (1986). A New Similarity Rule for Fluidized Bed
Scale-Up. AIChEJ., 32, 1466.
Huang, X. Q. and Zhou, L. X. (1991). Simulation of 3-D Turbulent Recirculating Gas-Particle
Flows by an Energy-Equation Model of Particle Turbulence. FED, 121, Gas-Solid Flows,
ASME, 261.
References 241
Hunt, M. L. (1989). Comparison of Convective Heat Transfer in Packed Beds and Granular Flows.
In Annual Review of Heat Transfer. Ed. C. L. Tien. Washington: Hemisphere.
Ishii, M. (1975). Thermo-Fluid Dynamic Theory of Two-Phase Flow. Paris: Eyrolles.
Jenkins, J. T. and Savage, S. B. (1983). A Theory for the Rapid Flow of Identical, Smooth, Nearly
Elastic, Spherical Particles. /. Fluid Mech., 130, 187.
Kozeny, J. (1927). Uber kapillare Leitung des Wassers im Boden (Aufstieg, Versickerung und
Anwendung auf die Bewasserung). Ber. Wien. Akad. 136a, 271.
Launder, B. E. and Spalding, D. B. (1972). Mathematical Models of Turbulence. London: Academic
Press.
Launder, B. E. and Spalding, D. B. (1974). The Numerical Computation of Turbulent Flows.
Computer Methods in Applied Mechanics and Engineering, 3, 269.
Lun, C. K. K., Savage, S. B. and Jeffery, D. J. (1984). Kinetic Theories for Granular Flow: Inelastic
Particles in Couette Flow and Slightly Inelastic Particles in a General Flow Field. /. Fluid
Mech., 140, 223.
Oseen, C. W. (1927). Hydrodynamik. Leipzig: Akademische Verlagsgescellschafe.
Patankar, S. V. (1980). Numerical Heat Transfer and Fluid Flow. Washington: Hemisphere.
Prandtl, L. (1925). Bericht uber Untersuchung zur ausgebildeten Turbulenz. ZAMM, 5, 136.
Reif, F. (1965). Fundamentals of Statistical and Thermal Physics. New York: McGraw-Hill.
Reynolds, O. (1900). Papers on Mechanical and Physical Subjects. Cambridge: Cambridge
University Press.
Savage, S. B. (1979). Gravity Flow of Cohesionless Granular Materials in Chutes and Channels.
J. Fluid Mech., 92, 53.
Savage, S. B. (1983). Granular Flows at High Shear Rates. In Theory of Dispersed Multiphase Flow.
Ed. R. E. Meyer. New York: Academic Press.
Savage, S. B. and Jeffery, D. J. (1981). The Stress Tensor in a Granular Flow at High Shear Rates.
J. Fluid Mech., 110, 255.
Savage, S. B. and Sayed, M. (1984). Stresses Developed by Dry Cohesionless Granular Materials
Sheared in an Annular Shear Cell. /. Fluid Mech., 142, 391.
Scheidegger, A. E. (1960). The Physics of Flow Through Porous Media. Toronto: University of
Toronto Press.
Slattery, J. C. (1967a). General Balance Equation for a Phase Interface. I & EC Fund., 6, 108.
Slattery, J. C. (1967b). Flow of Viscoelastic Fluid Through Porous Media. AIChEJ., 13, 1066.
Soo, S. L. (1965). Dynamics of Multiphase Flow Systems. I & EC Fund., 4, 426.
Soo, S. L. (1989). Particulates and Continuum: Multiphase Fluid Dynamics. New York:
Hemisphere.
Sun, J. and Chen, M. M. (1988). A Theoretical Analysis of Heat Transfer due to Particle Impact. Int.
J. Heat & Mass Transfer, 31, 969.
Taylor, G. I. (1921). Diffusion by Continuous Movements. Proc. London Math. Soc, 20, 196.
Tchen, C. M. (1947). Mean Value and Correlation Problems Connected with the Motion of Small
Particles in a Turbulent Field. Ph.D. Thesis. Delft University, Netherlands.
Whitaker, S. (1969). Advances in Theory of Fluid Motion in Porous Media. / & EC, 61, 4.
Zeininger, G. and Brennen, C. E. (1985). Interstitial Fluid Effects in Hopper Flows of Granular
Materials. ASME Cavitation and Multiphase Flow Forum, Albuquerque, N. Mex.
Zhang, J., Nieh, S. and Zhou, L. (1992). A New Version of Algebraic Stress Model for Simulating
Strongly Swirling Turbulent Flows. Numerical Heat Transfer, PartB: Fundamentals, 22,49.
Zhou, L. (1993). Theory and Numerical Modeling of Turbulent Gas-Particle Flows and
Combustion. Boca Raton, Fla.: CRC Press.
Problems
5.1 Show that dJ/dt = (V U)J, where / is the Jacobian determinant.
5.2 Prove that the Maxwell-Boltzmann distribution can be expressed as Eq. (5.24). Several
steps are suggested:
(1) Use Stirling's approximation formula (lnjc! = x lnx x) to show that
In WMB = n + 2_.(i In gx - n{ In n{)

(2) On the basis of the Lagrangian method (Lagrangian multipliers) and the conservation
of total number of particles and total energy of the system, show that the Maxwell-
Boltzmann distribution can take the form
1
he**
where A and B are constants to be determined.
(3) From the conservation of total number of particles, show that
1
(4) Note that the averaged translational energy of a single particle is
Show that B = 1/kT.

5.3 Derive Eq. (5.38) on the basis of the Maxwell-Boltzmann velocity distribution.
5.4 Show that for a steady, incompressible, isothermal flow, the ^-equation can be expressed
by Eq. (5.75). The derivation procedure is suggested as follows:
(1) Subtract the time-averaged momentum equation from the instantaneous momentum
equations.
(2) Multiply the equation by u[.
(3) Take the time average over the resulting equation.
5.5 Show that for a steady, incompressible, isothermal flow, the 6-equation can be expressed
by Eq. (5.80). The derivation procedure is suggested as follows:
(1) Differentiate the instantaneous momentum equations with respect to x\.
(2) Multiply the equation by ivdUJdxx.
(3) Take the time average over the resulting equation.
(4) Differentiate the time-averaged momentum equations with respect to x x.
(5) Multiply the equation by ivWJdxx.
(6) Subtract (5) from (3).
5.6 Show that, for flow through a circular tube, the velocity distribution can be expressed by
Eq. (5.318).
5.7 Design a scale prototype for a dilute solid fuel combustor with the scale of 1:10. The
prototype (P) operates at the temperature Tp = 1, 150 K and the pressure pv = 10 atm. The
averaged air velocity in the prototype Vp = 6 m/s. The operating temperature of the model
(M) is TM = 300 K. The physical properties of the combustion products are assumed to be
the same as that of air. It is further assumed that the air obeys the equation of state of an
ideal gas. Thus, /xP = 4.56 x 10~5 kg/m-s and /zM = 1.85 x 10~5 kg/m-s. Determine the
particle size and operational pressure and velocity of the air in the model.
5.8 Applying the volume-averaging theorems to Eqs. (5.14) and (5.18), verify that the volume-
averaged momentum and energy equations for phase k in a multiphase flow can be given
by Eqs. (5.123) and (5.126), respectively.
5.9 Consider an isothermal, steady gas-solid laminar pipe flow under the microgravity condi-
tion. The flow is regarded as axisymmetric. Both the gas and particle phases can be treated
Problems 243
as a pseudocontinuum. Derive the governing equations, in cylindrical coordinates, which

account for the motion of the gas and particle phases.
5.10 A golf ball with diameter dv travels in the air at uniform velocity U = uoex. Consider the
initial velocity of the ball Uv0 = uvOex + i>pOey with initial angular velocity Q = 20ez. The
gravity acts in the y -direction. The angular velocity is assumed to be constant during the
motion of the ball. Derive the governing equations which describe the motion of the ball.
CHAPTER 6
Intrinsic Phenomena in a Gas-Solid Flow
6.1 Introduction
There are inherent phenomena associated with gas-solid flows. Such phenomena
are of considerable interest to a wide spectrum of process applications involving gas-
solid flows and can be exemplified by erosion, attrition, pressure-wave propagation, flow
instability, and gas-solid turbulence modulation.
Mechanical erosion may cause severe damage to a pipe wall or any moving part in a
gas-solid system. Particle attrition yields fine particles, which may alter the flow conditions
of the system and may become a source of particulate emission. Understanding the basic
modes of mechanical wear and the mechanism of attrition is essential to the control of their
behavior. Pressure waves through a gas-solid suspension are directly related to nozzle flows
(such as in jet combustion) and measurement techniques associated with acoustic waves and
shock waves. Understanding the propagation of the speed of sound and normal shock waves
in a dilute gas-solid suspension is, therefore, important. Row instability such as wavelike
motion represents an intrinsic feature of a gas-solid flow. An instability analysis can be
made using the pseudocontinuum approach. For most gas-solid flows, particles move in a
strong turbulent gas stream. The gas turbulence can be significantly modulated by the solid
particles. Thus, understanding the particle-turbulence interactive phenomenon is necessary.
This chapter addresses the various phenomena indicated. In addition, the thermodynamic
laws governing physical properties of the gas-solid mixture such as density, pressure,
internal energy, and specific heat are introduced. The thermodynamic analysis of gas-solid
systems requires revisions or modifications of the thermodynamic laws for a pure gas
system. In this chapter, the equation of state of the gas-solid mixture is derived and an
isentropic change of state is discussed.
6.2 Erosion and Attrition

The mechanical erosion of a solid surface such as a pipe wall in a gas-solid flow
is characterized by the loss of solid material from the solid surface due to particle impacts.
The collisions of the particles either with other particles or with a solid wall may lead to
particle breakup, known as particle attrition. Pipe erosion and particle attrition are major
concerns in the design of a gas-solid system and during the operation of such a system.
The wear of turbine blades or pipe elbows due to the directional impact of dust or granular
materials, the wear of mechanical sieves by the random impact of solids, and the wear of
immersed pipes in a fluidized bed by both directional and random impacts are examples
of the erosion phenomenon in industrial systems. The surface wear associated with the
erosion phenomenon of a gas-solid flow has been exploited to provide beneficial industrial
applications such as abrasive guns, as well.
The most commonly used pipe materials may be classified into four categories based
on the mechanical erosion modes: metals such as copper, aluminum, and steel; ceramics
244
6.2 I Erosion and Attrition 245
and glasses; plastics such as polyvinyl chloride (PVC) and acrylic (Plexiglas); and
rubbers.
6.2.1 Ductile Erosion and Brittle Erosion

When a particle strikes a solid surface, the extent and nature of the damage on
the surface depend on the normal compressive force F n , tangential cutting force F t , area
of contact Ac, duration of contact tc, angle of incidence au shape of the particle, and the
materials of the particle and the solid surface. The mechanisms of mechanical erosion may
be explained in terms of two basic modes, i.e., the ductile mode and the brittle mode.
Ductile erosion is the loss of material from a solid surface due to the tangential cutting
force of the particle impact. When a solid surface is struck by a particle, elastic and plastic
deformations occur. The yield stress for ductile removal YD must be greater than the onset
yield stress or one-half of the yield strength in simple tension [Timoshenko and Goodier,
1951]. YD is typically greater than or nearly equal to the pressure for plastic deformation,
0.35 percent of the modulus of elasticity [Goldsmith and Lyman, I960].
A ductile rupture may be characterized by a maximum erosion occurring at some
small angle of incidence, usually between 20 and 30. Typical ductile damages due to
single-particle impacts are illustrated in Fig. 6.1 (a) [Hutchings, 1987]. In this figure, it
can be seen that rounded particles striking at shallow impact angles lead to ploughed
craters, while impacts from sharp-edged particles cause indentations with a pronounced
chip of material still attached. In both cases, the chips can be detached by only a small
number of subsequent impacts. In a surface damaged by multiple impacts of hard parti-
cles, flake- or platelet-shaped debris may be formed as a result of collisions with round
particles, whereas bulky-shaped debris is likely to be caused by the impacts of angular
particles.
Brittle erosion is the loss of material from a solid surface due to fatigue cracking and
brittle cracking caused by the normal collisional force Fn. Materials with very limited
capacity for elastic and plastic deformation, such as ceramics and glass, respond to particle
impacts by fracturing. The yield stress for brittle failure F B for normal impacts is about
(a) (b)
Figure 6.1. Examples of damage sites due to single-particle impacts: (a) On mild steel, a
ductile metal; (b) On poly crystalline alumina, a brittle ceramic (from Hutchings, 1987;
reproduced with permission).
246 61 Intrinsic Phenomena in a Gas-Solid Flow
three times the yield stress [Tabor, 1951], or stress to cause fracture at static conditions,
or 0.43 percent of the shear modulus. 7B is also equal to [(1 2v)/1.86] times the yield
strength in simple tension, where v is the Poisson ratio. A typical example of brittle erosion
is shown in Fig. 6.1(b), where the surface is characterized by the formation of isolated pits
from which some grains appear to have been removed by intergranular fracture [Hutchings,
1987].
The fractures on a plane surface, created by the collisions of hard spherical particles
at low-impact velocities, may form a conical crack according to the Hertzian quasi-static
stress theory. In a multiple-impact situation, the conical cracks meet those extending from
neighboring impact sites, and then the brittle material becomes detached. Once appreciable
damage is done, the cracking mechanism may be altered because the particles no longer
strike on a plane surface; nevertheless the brittle removal continues by the successive
formation and intersection of cracks.
A probable mechanism of erosion for plastics is illustrated in Fig. 6.2 [Briscoe and Evans,
1987]. Initially, a series of plastically deformed grooves can be formed by the abrasion of
particle flows. The subsequent directional or random impacts of particles may push the
deformed grooves from side to side. The fatigue limits of the plastics would eventually
make the ridges between the grooves detach to form ribbonlike debris. Brittle cracks also
occur when the wear tracks interact.
The essential material property of rubbers is their low elastic modulus, which ensures
that the contact deformation remains elastic over a very wide range of contact conditions.
The abrasive wear of rubbers is due to either fatigue of the material or tearing by a cutting
force from impacts with sharp-edged particles.
In general, erosion via tangential cutting tends to be more prominent for ductile materials
such as metals and plastics than for brittle materials such as ceramics and glasses. In practice,
erosion is a result of the combined effect of both the ductile and brittle modes. The extent
(a)
(b) (d)
(c) (e)
Figure 6.2. Schematic diagram of probable mechanism of plastic fatigue wear (from
Briscoe and Evans, 1987): (a) Formation of plastically deformable grooves in series; (b)
Deformation of the grooves pushed in one direction; (c) Sway back to the opposite direction;
(d) Deterioration of ridges after repeated fluttering; (e) Detachment of ridges in the form of
band-shaped debris.
62 I Erosion and Attrition 247
of erosion for a small deviation from the elastic states can be accounted for by a linear
combination of that due to the ductile and brittle modes.
6.2.2 Locations of Erosive Wear

Abrasive erosion in a straight pipe caused by particle impacts is usually insig-
nificant compared to the wear at bends such as elbow connectors for pneumatic conveying
pipelines and the outer wall near the entrance of a cyclone separator. The angle at which
maximum erosion occurs primarily depends on the material properties. For ductile materi-
als, the impact angles are typically in the range of 20to 45, with the smaller angles being
related to more angular eroding particles. On the other hand, for brittle materials, maximum
erosion is caused by brittle fractures under conditions of normal impingement where the
impact angle approaches 90. An example of both cases can be seen in Fig. 6.3, in which
the eroding particles are 100 /xm sharp-edged silicon carbide particles with a velocity of
150 m/s, the ductile material is aluminum, and the brittle material is alumina [Finnie et ah,
1967]. In this example, the erosion is defined as the mass of material removed per unit mass
of colliding particles.
The location of the most severe erosion on the extrados of a pipe bend also depends
on the particle flow pattern inside the bend section. Particle trajectories in a right angle
connector are shown in Fig. 6.4. For large or heavy particles, the inertia of the particles plays
a dominant role in determining the trajectories of these solids and the fluid drag forces only
slightly affect their motion. Thus, the paths of large particles in a bend are nearly straight,
leading to a steep angle of impact on the pipe wall (see Fig. 6.4, path a). For smaller or
lighter particles, both the drag force and inertia are important in the particle motion, and
thus the paths of the particles may deflect, with the particle-wall impact occurring at a
Ductile mode
Brittle mode
0.0016 / \
\
1 \
/ \
J \
<g 0.0012 1 \
'
\
1 \
j \

\
\
0.0008 1
/
I
-f /
11
0.0004 -h /
1
/ i
1 1 1 1 i i
0 30 60 90
Impact angle, degree
Figure 6.3. Dependence of erosion on impact angle (from Finnie et ai, 1967).
Figure 6.4. Particle trajectories in a right-angled pipe bend: (a) Heavy and/or large particle;
(b) Medium-sized particle; (c) Light and/or small particle.
shallower angle, closer to the exit of the bend connector (see Fig. 6.4, path b). With further
reduction of particle size or weight, the particles may closely follow the fluid motion and
may not contact the wall at all (see Fig. 6.4, path c).
As mentioned, the erosion of a solid surface depends on the collisional force, angle of
incidence, and material properties of both surface and particles. Although abrasive erosion
rates cannot be precisely predicted at this stage, some quantitative account of erosion modes
which relates various impact parameters and properties is useful. In the following, a simple
model for the ductile and brittle modes of erosion by dust or granular materials suspended
in a gas medium moving at a moderate speed is discussed in light of the Hertzian contact
theory [Soo, 1977].
When the deviation from the elastic state of the material surface is small, the Hertzian
theory can estimate the force of impact, contact area, and contact duration for collisions
between spherical particles and a plane surface using Eqs. (2.132), (2.133), and (2.136),
respectively. To account for inelastic collisions, we may introduce r* as the ratio of the
reflection speed to the incoming speed, V. Therefore, we may write
Fntc = 2.94(5 /16)mV(l+r (6.1)

Ft = fFn (6.2)
where / is the coefficient of friction.

5i
Actc = Aoto(suioii) (6.3)
where <X[ is the impact angle and
Aoto = 25 (6.4)
where k = (1 v)/nE\ subscripts " 1 " and "2" denote each of the colliding pair; and N\m
is the impact number, defined by
( I , v 4
(6 5)
\T +
\T) -
y K2 y tC\ I
which correlates the dynamic and material properties of the impacting system.
In the following, the effects of gliding, scattering, and lifting of the particles by the gas
stream are neglected and the mechanical efficiency of impact is assumed to be 100 percent.
Thus, for ductile erosions, the wear in volume loss per impact equals the work of ductile
wear by the tangential force divided by the energy required to remove a unit volume of
material1 in Fductile erosion mode. Therefore,
F V we have y
(* v V/ A & (y DA O fo,. ,-A
wD = I YD)Vcosa[Actc = cosai 1 (sinai) 5 (6.6)
^\4 / 6 V ^'
where u;D represents the wear in volume loss per impact and 6D denotes the energy required
to remove a unit volume of material in ductile mode. In a dimensionless form, Eq. (6.6)
can be expressed as
ED = c o s t a l - ^ D(sinofi)"5] = D (a if KD) (6.7)
where ED is a dimensionless ductile erosion parameter, defined by
ED =
and KD is the ductile resistance parameter, defined by
YD
K (6.9)
Substituting Eqs. (6.1), (6.3), and (6.4) into Eqs. (6.8) and (6.9) yields
ED =
2.94(5/16)(l
and
i -i
Equation (6.11) indicates that, since (I H-r*)" 1/ 5 is nearly unity, J^D is only related to
the material properties and the collisional speed. Thus, K& can be treated as a parametric
constant and the function D characterizes the ductile wear at various angles of impact via
Eq. (6.7). Hence, the maximum value of ED occurs at an angle of impact am given by
5(sina m )T
KD = - 2 (6.12)
4 snr am + 1
1.0
0.8
-6
0.4
0.2
0.0
10 20 30 40 50 60 70 80 90
Maximum eroding angle ccm, degree
Figure 6.5. Relationship between the ductile resistance parameter and the angle of impact for
maximum erosion (from Soo, 1977).
0 10 20 30 40 50 60 70 80 90
Impact angle a{, degree
Figure 6.6. Normalized ductile erosion energy function varying with impact angle and ductile
resistance parameter (from Soo, 1977).
In terms of Eqs. (6.12) and (6.7), the relations KD and EDm versus am are given in
Fig. 6.5, where EDm = ED(am, KD). The ratio of ED to EDm at various angles of impact ax
and KD is plotted in Fig. 6.6, which indicates the maximum eroding angle for a given KD
or the locations of the most severe ductile erosions for a given particle flow pattern.
Similarly, for brittle erosions, the wear in volume loss per impact w;B is expressed
by
1 (Fn \ FntcV
= I B ]V sma[A ctc = 1 (6.13)
\ A )
where eB is the energy required to remove a unit volume of material in brittle erosion mode.
By putting Eq. (6.13) into a dimensionless form, we may define a dimensionless erosion
parameter EB for brittle erosion as
EB = (6.14)
F nt c V 2.94(5/16)(l + r*)mV2
and a brittle resistance parameter KB as
= YB (6.15)
so that
EB = sinofi[l - = EB(oti, KB) (6.16)

Equation (6.16) indicates that the maximum brittle wear occurs at ofi = 90, i.e., normal
collision. The variation of the ratio of EB to B m (=EB(oim, &B)) with ax as a function of
KB is plotted on the basis of Eq. (6.16) as given in Fig. 6.7. The figure reflects that for a
given KB, the degree of abrasive damage due to brittle erosion may be estimated from the
given particle flow pattern.
The preceding delineation of ductile and brittle modes by directional impacts repre-
sents the idealized components of surface wear by dust particles. In reality, most cases
for surface wear result from the effects due to a combination of these two modes. Effects
of gliding, scattering, and lifting of the particles by the gas stream may also significantly
influence the impacts of particles. For example, the Saffman force due to the boundary
layer motion of the gas may produce a gliding (barely touching) or lifting (nontouch-
ing) motion of particles, whereas the deflective motion of a gas flow may produce a cen-
trifugal force on the particles, thereby increasing the compressive stresses for the particle
impacts.
0 10 20 30 40 50 60 70 80 90
Impact angle a{, degree
Figure 6.7. Brittle erosion energy function varied with impact angles and brittle resistance
parameters (from Soo, 1977).
Example 6.1 In a survey of turbine bucket erosion, the friction coefficient can be expressed
by [Soo, 1977]
where
Assuming that the erosion is purely ductile, derive an expression for the impact angle
yielding maximum wear.
Solution From Eq. (6.6), we have
(smof) 5 (E6.3)
)
Now we redefine ED and KD by
W D
ED = ^ v (E6.4)
and
_ YD Aoto
~ /o Fntc
Substituting Eqs. (E6.1), (E6.4), and (E6.5) into Eq. (E6.3) yields
5
ED = cosofi J f ^ A _ ^ D (sinofi)-5 I (E6.6)
Note that, at the maximum eroding angle am for a given K&, dE^/dai = 0. Thus, the
maximum eroding angle is given by
(cos m ) 3 / 5 /5sin 2 a m +
6.2.3 Mechanisms of Attrition

Attrition denotes a breakup process of solid particles by collisions. Attrition can
be further classified into fragmentation and abrasion. If the sizes of broken pieces of a
particle in a breakup process are of the same order of magnitude as the initial particle, this
process is termed fragmentation. If the broken particles are at least an order of magnitude
smaller than the initial particle, the breakup process is called abrasion. Thus, in general,
fragmentation deals with bodily damage, whereas abrasion is associated with the blunting
of corners and edges on the surface of a particle.
Particle attrition can be caused by the relative motion of mechanical parts such as a
blade in the flow of bulk particles; by the impact of particles with a solid wall or with other
62 I Erosion and Attrition 253
particles; and by the volumetric change due to chemical reactions or phase change such as
devolatilization of coal particles in a coal combustion process. The degradation in particle
size by attrition would not only change the particle size distribution but also alter the flow
pattern of solids, or even produce fine or ultrafine particles which usually present difficulties
in transport, separation, and other handling processes.
For particle surface abrasions, a satisfactory empirical formulation was suggested by
Gwyn (1969) as
w=afi (6.17)
where w is the attrited mass fraction of the sample particles; e is the shear strain on the
particle surface; and a and /3 are empirical constants. A theoretical basis for Eq. (6.17) was
explored through a surface abrasion model [Paramanathan and Bridgwater, 1983], which
indicates that the parameter p is related to the material properties. In the following the
surface abrasion model is introduced.
Assume that the rate of abrasion is proportional to the reduction in radius of a particle
raised to an arbitrary power, i.e.,
da n
= -A(ai-a)-B (6.18)
d
where ax is the initial radius of a particle and a is the radius of the particle at strain e; A
and B are constants depending on the attrition mode and material properties; and B ^ 1.
Thus, by integrating Eq. (6.18), we have
Consider particles with an initial size distribution in the sieve range from a\ to ai and define
an as the initial radius of a particle, which shrinks to a value a\ at strain e. Hence, we have
(on - ai) B + 1
= (6 20)
fl + 1 ^ '
As a result, the mass fraction attrited from other sizes to a\, w, can be expressed by
/ (6.21)
where /MO^) is the mass density function of initial particle sizes. Furthermore, if the
material held on the top sieve produces negligible amounts of fines, the normalization of
tfi) yields
l (6.22)
J Q\
Assuming that the initial size distribution is uniform, Eq. (6.22) gives
fM(a{) = ^ (6.23)
a2 a\
Therefore, we have
w= ^^ (6.24)
a2-ai
l n A + + 1)
in w = B ^ - ln(a2 - ,) + - i - In, (6.25)
D -f- 1 > "t" 1
Now, defining a and /? as
Eq. (6.25) is reduced to Eq. (6.17).
6.3 Thermodynamic Properties of a Gas-Solid Mixture

The analysis of a multiphase flow system is complex, in part because of the diffi-
culties in assessing the dynamic responses of each phase and the interactions between the
phases. In some special cases, the gas-solid mixture can be treated as a single pseudo-
homogeneous phase in which general thermodynamic properties of a gas-solid mixture can
be defined. This treatment provides an estimate for the "bulk" behavior of the gas-solid
flow. The following treatment is based on the work of Rudinger (1980).
6.3.1 Density, Pressure, and Equation of State

Assume that the particles can be regarded as molecules of a second gaseous species.
Hence, multiple or interactive collisions among particles, and forces between particles and
gas molecules can be neglected. However, the finite volume of particles may be taken into
account. It is shown later that with this type of treatment the mixture behaves as a van der
Waals gas without correction for the interaction force between the molecules.
Denote ap as the volume fraction of particles in a gas-solid mixture. The volume-
averaged densities of both gas and particle phases are then given by
(p) = (1 - a p )p; (pp) = appp (6.27)
where p and p p are the material densities of gas and particles, respectively. The bulk density
of the mixture is defined by
Pm = (P> + (Pp) = (1 - Qfp)p + CXppp (6.28)
It is convenient to express the particle loading by the particle mass fraction, which
depends on the material densities of the phases as well as the volume fractions of the phases
and is defined by
0 = W = Ê (6.29)
pm 1 - av + a p
where is the ratio of material density of particle to gas. Hence, an alternative form of
Eq. (6.28) is
J_ = -0) + (6 30)
Pm P Pp
6.3 I Thermodynamic Properties of a Gas-Solid Mixture 255
Since the particles are treated as a second "gas" phase, it is important to consider their
contribution to the pressure of the mixture. The molecular weight of this "gas" is
(631)
where mH = 1.66 x 10~27 kg is the mass of a hydrogen atom. If the molecular weight of
the gas is denoted by wg, the molecular weight of the mixture is given by
-L = -* + <Iz (6.32)
Thus, the ratio of the pressure of the gas p to the pressure of the mixture pu becomes
p (l-(f))/wg ( 0 6wgmH\~l ,,,
= 77T = 1+ T> --TT- ( 6 - 33 )
PU
Note that the material density of a gas phase at a temperature of 300 K and pressure of
1 atm is given by
Po = miiWgL (6.34)
25 3
where L = 2.69 x 10 m" is the Loschmidt number. Substituting Eq. (6.34) into Eq. (6.33)
yields
(6.35)
If the contribution of the particles is less than 1 percent of the total pressure (i.e.,
0.99), the particle diameter (measured in micrometers) must satisfy the condition
-1/3
(Aim) (6.36)
\Po 0
where dm is the particle diameter below which the particle contribution to the pressure of
the mixture exceeds 1 percent at the atmospheric pressure. The relationship between d m
and 0 is shown in Fig. 6.8 for various values of pp/po- It is noted that the contribution of the
particles to the gas-solid mixture pressure can be neglected even at very high density and
mass fraction so long as the particle is larger than a few hundredths of a micrometer. Thus,
for practical purposes, the gas-solid mixture pressure in a gas-solid flow can be given by
the gas-pressure alone.
Assume that the gas obeys the ideal gas law so that
p = pRMT (6.37)
where RM is the gas constant. For a gas-solid mixture in thermal equilibrium (i.e., TP = T),
the equation of state for the mixture may be obtained from Eq. (6.37) and Eq. (6.30) as
l0(P/P)
where 0, which is a constant in a closed system, is a parameter rather than a variable in the
equation of state for the mixture. Hence, the equation of state for a gas-solid mixture does
not have the same form as that for an ideal gas.
0.020
0.015
0.010
0.005 -
0
0 0.2 0.4 0.6 0.8 1.0
Massfractionof particles <>|
Figure 6.8. Relationship between dm and <>
/ (from Rudinger, 1980).
It should, however, be noted that the preceding results are valid only when the collisional
interacting forces between the particles can be neglected; it is clear, however, that for high
particle concentration conditions such as those in fluidized beds these forces cannot be
neglected. A simple way to verify the significance of the interacting forces between the
particles is to examine the average spacing between them in a given system. Assume that,
on average, the particles are arranged in a regular cubic array of side /. The volume fraction
occupied by a particle is determined by
As an example, for mass fraction 0 = 0.1 and f = 1,000, the volume fraction av is about
10~4. The average spacing between particles becomes about 17 particle diameters. Thus,
for this example we can consider the direct interactions among particles to be insignificant.
Example 6.2 In a gas-solid mixture, the particle density is 2,400 kg/m 3 and the gas
density is 1.2 kg/m3. For a particle mass fraction of 99 percent, what is the corresponding
particle volume fraction? What is the corresponding particle volume fraction, if the particle
mass fraction is 50 percent?
Solution From Eq. (6.29), the particle volume fraction can be expressed in terms
of material density ratio of particle to gas and particle mass fraction as
The material density ratio of particle to gas from the given conditions is 2,000. Thus, for a
particle mass fraction of 99 percent, we have
0 0.99
ttp = =
= 0 0.01x2,000 + 0.99
6.3 I Thermodynamic Properties of a Gas-Solid Mixture 257
and for a particle mass fraction of 50 percent, the volume fraction of particles becomes
0 0.5
0*r> = = 5 x 10"4 (E6.10)
(1-0K+0 0.5x2,000 + 0.5
This example reflects that even for a very dilute suspension of gas-solid flow where
the particle volume fraction is negligibly small, the particle mass fraction may still be
significant, irrespective of the size of the particles.
6.3.2 Internal Energy and Specific Heats

The internal energy per unit volume of a mixture e^ is given by
e
M = (1 ~~ <P)e H~ 0^p = (1 0)^v^ + (/>cTp (6.40)
where c v is the specific heat of gas at constant volume and c is the specific heat of the
particles. Hence, the specific heat of the mixture at a constant volume can be defined as
c V M = (1 cj))c\ + (j)c (6.41)
Similarly, the specific heat of the mixture at constant pressure can be defined as
CpM = (1 0)c p + (pc (6.42)
From Eqs. (6.41) and (6.42), the ratio of the specific heats can be expressed as
(p -\- 8(p
(6.43)
=r
where 8 denotes the ratio of the specific heats of particles to gas at constant pressure
c/cp.
On the basis of Eq. (6.43), the relationship between }î and (p is plotted for various
values of <5, as given in Fig. 6.9. The figure shows that a large value of <p and/or 8 leads to
0.2 0.4 0.6 0.8 1.0

Mass fraction of particles (|>
Figure 6.9. Relationship between yu and 4> at various 8 with y = 1.4 (from Rudinger, 1980).
YM of unity, which implies an isothermal flow. Thus, the flow with a high particle loading
and/or a high particle thermal capacity behaves as an isothermal flow because the large heat
capacity of the particles significantly offsets the temperature variations in the gas induced
by expansion or compression.
6.3.3 Isentropic Change of State

The equation of state in an isentropic process of a gas-solid mixture can be ob-
tained in terms of an energy conservation relationship. When applying the first law of
thermodynamics to a gas-solid mixture, we have
dq = d^M T dPm (6.44)

Pm
where q is heat absorbed per unit mass of the mixture. Thus, for an isentropic process, we
have
deM = T dpm (6.45)
Pm
Assuming T = 7 p , Eq. (6.40) yields
Substitution of Eqs. (6.46) and (6.38) into Eq. (6.45) gives

( l - 0 ) c v + 0c dT = 1 dA (6 4 ? )
(1 - 0)/?M T 1 - 0p m /Pp Pm
The relationship for the ideal gas gives
/? M = Cp cy (6.48)
Therefore, we have
(1 (j))Cy + 0C CvM 1 //r ACk,
= = (6 49)
~T\^ T" *
For convenience, let
X=0 (6.50)
where, for a closed system, both p p and 0 are constant (note that, in this case, av is a variable
rather than a constant). Substitution of Eqs. (6.49) and (6.50) into Eq. (6.47) yields
1 dT 1 dX
(6.51)
Y M - 1 T 1 - X X
which gives
= A (6.52)
where A is a constant. Using the equation of state to eliminate T from Eq. (6.52) leads to
X
=B (6.53)
l-X
6.4 I Pressure Waves Through a Gas-Solid Suspension 259
where B is a constant. In another form, Eq. (6.53) becomes

n \-YbA
-YM
C (6.54)
-(ppm/ppj
where C is a constant.
Equation (6.54) accounts for the change of state in the isentropic processes of a gas-
solid mixture in which the effect of afiniteparticle volume is considered. An example using
this equation to obtain the speed of sound in a gas-solid mixture is introduced in the next
section.
6.4 Pressure Waves Through a Gas-Solid Suspension

The propagation of pressure waves such as acoustic wave, shock wave, and Prandtl-
Meyer expansion through a gas-solid suspension is a phenomenon associated primarily with
the transfer of momentum although certain processes of energy transfer such as kinetic en-
ergy dissipation and heat transfer between gas and solids almost always occur. Typical
applications of the pressure wave propagation include the measurements of the solids con-
centration and flow rate by use of acoustic devices as well as detonation combustion such
as in a rocket propellant combustor or in the barrel of a gun.
When a pressure wave travels through a gas-solid mixture, the interaction of the pressure
wave with the suspension can be strong. In general, the traveling speed depends not only
on the local concentration of solids and the properties of the gas and solid phases but also
on the wave frequency. The propagation of different types of pressure waves through the
same mixture behaves differently because of the difference in the dynamic response of
the mixture. The speed of sound represents the propagation of periodic pressure waves of
small amplitude. The speed of a shock wave, having a front of discontinued pressure rise
followed by an extended relaxation zone, represents the propagation of a sharp pressure
jump which produces a transition from one flow condition to another. For simplicity we
consider only the cases of plane waves through a gas-solid suspension, in the following
discussion.
6.4.1 Acoustic Wave

In order to obtain governing equations for the motion of particles and the speed
of sound when an acoustic wave passes through a gas-solid suspension, it is necessary
to determine the interacting force between the solid and gas phases which accounts for
the relative oscillating motion of the particles. Under the condition of creeping flow, an
expression for the interacting force between an oscillating solid sphere and an infinitely
large incompressible fluid at rest was obtained by Lamb (1932). Thus, by substituting
the resultant interacting force into the BBO equation (see 3.4), the oscillating motion of a
spherical particle may be described by the modified equation. When this equation is coupled
with the equation of a plane wave in a mixture, the speed of sound in a gas-solid suspension
can be obtained.
First, consider the case of a ball pendulum oscillating in an infinite mass of fluid. The
fluid is incompressible and originally at rest. It is convenient to set the origin at the mean
position of the center, and the x-axis in the direction of the oscillation. For small motion
of the fluid where the inertia can be neglected and in the absence of external forces, we
have
VU=0 (6.55)
and
= - - V p + vV 2U (6.56)
at p
where U denotes the fluid velocity. The boundary condition on the surface of the sphere is
given by
U = UV =
V = Vv = 0 (6.57)
W = Wv = 0
where l/ p is the velocity of the sphere.
With a lengthy derivation, the analytic solution for the preceding problem was provided
by Lamb (1932). The resultant force on the sphere due to oscillating motion is obtained
as
F = - V P 3 (~ + 4 = ) - - - H 3 P ^ (l + 4=)u (6.58)
where Nw is a dimensionless parameter expressed as
codl
Nw = (6.59)
8v
The first term on the right-hand side of Eq. (6.58) gives the correction for the inertia of
the sphere, commonly referred to as the virtual mass term. It should be noted that, as a
result of viscous and oscillating effects, the amount of replaced fluid mass increases. The
restriction here is that dUp/dt = icoUp, i.e., pure oscillating motion of the particles. The
second term is the drag force. When the oscillation period is made infinitely long (i.e.,
co -> 0), Eq. (6.58) leads to the Stokes drag force.
Consider a gas-solid suspension which is in a state of steady dilute flow with no interpar-
ticle collision or contact. In this situation, the linear particle velocity is practically identical
to the superficial particle velocity. The motion of a spherical particle in an oscillating flow
field can thus be given by
PpdUp_(l 9 \d(UUp) 9 co / 1 \
(6.60)
which is a modified BBO equation derived by replacing the Stokes drag and carried mass
terms in Eq. (3.101) by Eq. (6.58). If we further approximate the pressure term by
-Wp * p^- (6.61)

at
Equation (6.60) becomes

p p dt/ p /I 9 \d(U-Up) 9 co / 1 \
+ ( p)
p dr ~U 4VN^j d* 4^1\CJ
d
dT (6 62)
+ / / -
Now we consider the case of one-dimensional oscillation. If we neglect the effects due
to the volume occupied by the solid particles and scattering, for a gassolid suspension with
solid-to-gas mass ratio of mp, the equation of a plane wave in the mixture may be expressed
by [Soo, 1990]
d2U\ 6mp (3F\ (6 63)
J U J -
where F is the total particle-fluid interacting force (on the particle) including the carried
mass force, Basset force, and drag force. F may be expressed by Eq. (6.58) when Re p is in
the Stokes regime.
For a harmonic oscillation, we may express U and Uv by
fiz
where ft is the attenuation coefficient. Thus, Eq. (6.62) becomes
Uv = B(U - Uv) - iA(U - Uv) + U+ C(l - i)(U - Up) (6.65)
where
> V , >& %
(6.66)
A is the friction force coefficient; B is the inertia correction coefficient; and C is the
resistance coefficient due to the Basset force. Equation (6.65) gives
(6.67)
where ^ = B 4- C, and r\ = A + C. The speed of sound in the mixture can be obtained

from Eq. (6.63) by noting that
Substitution of Eq. (6.68) into Eq. (6.63) gives

/92f/\ ypfd2U\ (d*Uv p3f/\
Substituting the real part of Eq. (6.67) into Eq. (6.69) gives
p i yP
D mD \ = (6.70)
where D is the real part of Eq. (6.67) and is equal to
D = (6.71)
Hence, the speed of sound in the mixture is given by
al
(6.72)
where am is the speed of sound in the mixture and ag is the speed of sound in the pure gas,
which is given by
al = (6.73)
g
P
Moreover, substituting Eq. (6.67) and the expression of U in Eq. (6.64) into Eq. (6.69), for
f$ag/co < 1, an expression for Pag/co can be obtained as
pag P_ (6.74)
O) 2 a* Pv
An example of dispersion of sound in an air-magnesia mixture with solid-to-air mass
ratio, 7wp, of 0.3 for pp/p (=f) of 100 and 1,000 is illustrated in Fig. 6.10 [Soo, I960].
The figure gives the dispersion of sound for various values of Nw when the heat transfer
is neglected. Under the same experimental condition, the relationship between Pa g/co and
N^ is shown in Fig. 6.11 [Soo, I960]. It is noted that the preceding model is valid only
1.00
0.76
10"55 10"44
10 10'33
10 10'22 10'
10 10"' 10 101 102
N
Figure 6.10. Dispersion of acoustic waves in the air-magnesia mixture (from Soo, 1960).
10
10 10 10 10 10 10 10
Figure 6.11. Attenuation per wavelength in the dispersion of acoustic waves in the
air-magnesia mixture (from Soo, 1960).
for dilute suspensions where the influence of particle contact is negligible. For high solids
concentration situations such as those in fluidized beds or packed beds where the effect of
particle contact may play a significant role in the sound propagation, the speed of sound is
possibly higher than that in a pure gas phase situation.
Another approach for estimating am is based on the pseudothermodynamic properties of
the mixture, as suggested by Rudinger (1980). The equation for the isentropic changes of
state of a gas-solid mixture is given by Eq. (6.53). Note that for a closed system the material
density of particles and the mass fraction of particles can be treated as constant. Hence, in
terms of the case for a single-phase fluid, the speed of sound in a gas-solid mixture can be
expressed as
dP
' (6.75)
which, with the substitution of Eq. (6.53), gives
pp\dXJs PpA(l-A)
The subscript "s" denotes entropy. From Eqs. (6.28) and (6.50), the speed of sound in a
mixture is obtained as
It is noted from Eq. (6.72) and Eq. (6.77) that am < ag under dilute suspension situations.
Example 6.3 In a gas-solid suspension, the speed of sound depends on the frequency
of the acoustic wave. Derive the expressions for the speed of sound under two extreme
conditions: (1) very low frequency (co -> 0) and (2) very high frequency (co -> oo). Use
the results to estimate the speed of sound in a gas-solid suspension with a particle-to-gas
mass ratio of 1 and a particle-to-gas density ratio of 2,000.
Solution The general form of the speed of sound in a gas-solid suspension is

given by Eq. (6.72), in which the parameters and r] are expressed by
(E6.ll)
where C is given by Eqs. (6.66) and (6.59) as

~8v~
(E6.12)
4/>PVK
For the acoustic wave with very low frequency (co 0),C -> oo. Hence, Eq. (E6.ll)
becomes
^ 2C n& --C2 (E6.13)
Substituting Eq. (E6.13) into Eq. (6.72) yields

! r / ^ \ i -1
(E6.14)
For the acoustic wave with very high frequency {co oo), C -> 0. Hence, Eq. (E6.ll)
becomes
(E6.15)
Substituting Eq. (E6.15) into Eq. (6.72) yields
Jl V / _ \ 1 1
ag ipvlP (E6.16)
For the gas-solid suspension with mv = 1 and pp/p = 2,000, from Eqs. (E6.14) and
(E6.16), we have
1
= 1 +m -Z-\\ = 1 + 1 x 1 - = 0.71 (E6.17)
for very-low-frequency acoustic waves, and
Pp
1
1 - = 1.00 (E6.18)
2x2,000+1 2,000
for very-high-frequency acoustic waves.

This example shows that in a dilute gas-solid suspension, for a very-low-frequency

perturbation, the particles may follow the changes in motion of the surrounding gas. The
speed of sound in this case is known as the equilibrium speed of sound; it can be estimated
from Eq. (E6.14). On the other hand, for a very-high-frequency perturbation in a dilute
gas-solid suspension, the particles do not follow the changes in motion of the surrounding
gas. In this case, the speed of sound is known as the frozen speed of sound and has almost
the same value as that for the gas alone.
6.4.2 Normal Shock Wave

Consider the propagation of a one-dimensional normal shock wave in a gas medium
heavily laden with particles. Select Cartesian coordinates attached to the shock front so that
the shock front becomes stationary. The changes of velocities, temperatures, and pressures
of gas and particle phases across the normal shock wave are schematically illustrated in
Fig. 6.12, where the subscripts "1," "2," and "oo" represent the conditions in front of,
immediately behind, and far away behind the shock wave front, respectively. As shown
in Fig. 6.12, a nonequilibrium condition between particles and the gas exists immediately
behind the shock front. Apparently, because of the finite rate of momentum transfer and
heat transfer between the gas and the particles, a relaxation distance is required for the
particles to gain a new equilibrium with the gas.
The elapsed time for particles to pass through the shock front may be approximated by
dv/U. Since U is of the same order of magnitude as the speed of sound in the gas, the ratio
of the flying time to the Stokes relaxation time of a particle can be expressed by
(6.78)
where Ts is the Stokes relaxation time. Calculation using Eq. (6.78) based on some represen-
tative values reveals that, even for particles as small as 0.1 /im, the elapsed time for a particle
to pass through the shock front is about three orders of magnitude smaller than the Stokes
relaxation time [Rudinger, 1980]. Thus, as illustrated in Fig. 6.12, a particle passes through
the shock front without yielding a significant change in its velocity. This is also the case for
the change in the particle temperature when the particle passes through the shock front.
To analyze the shock wave behavior in a gas-solid flow, several assumptions are made:
(1) The gas follows the equation of state of an ideal gas.
(2) The particles are uniformly distributed in the gas upstream of the shock wave.
(3) The Brownian motion of the particles does not contribute to the pressure of the
system so that the pressure of the mixture can be given by the gas pressure alone.
(4) The thermal diffusivity of the particles is high so that the internal temperature of
a particle is uniform.
(5) The effect of thermal radiation is negligible. (It should be noted, however, that
a cloud of particles is usually a better emitter and absorber of radiation than a
pure gas. Thus, the hot particles downstream of the shock may preheat the cold
particles upstream of the shock by radiation. This effect becomes significant as
the surface area of the particles increases.)
(6) The particle volume fraction is much less than unity.
Shock front
u2
-
Z
(a)
, Shock front
(b)
Shock front
P<~
Pi
(c)
Figure 6.12. Schematic flow relations across a plane shock wave in gas-solid suspensions:
(a) Changes of velocities across a plane shock wave; (b) Changes of temperatures across a
plane shock wave; (c) Change of pressure across a plane shock wave.
The following analysis is based on the method suggested by Carrier (1958). The law of
conservation of mass requires
/ = (1 - a?)pU pU = const (6.79)
and
/ p = ap/op/7p = const (6.80)
where / and / p are the mass fluxes of the gas and particles. The conservation of momentum
of the gas-solid mixture yields
(pU)U + (ap/t>p/p)I/p + p = const (6.81)

and the conservation of energy gives

l
-(pU)U2 + l-{avPvVv)Ul + (pU)cpT + (apPpUp)cTp = const (6.82)
Note that the equilibrium between the gas and solid phases indicates the conditions
U = Uv and T = Tv. Substituting these conditions as well as Eqs. (6.29), (6.37), and (6.79)
into Eq. (6.81) and Eq. (6.82) gives
+ ^ - = const (6.83)
1-0 U
U
and
1
U2 + 2 f cpp + r-^c ) T = const (6.84)
1-0 V 1-0
By use of Eq. (6.43), Eq. (6.84) becomes
U2 + m
RMT = const (6.85)
1-0 KM- 1
Now, consider the velocities and temperatures in two equilibrium regions, one in front
of the shock and the other far behind the shock. Applying Eq. (6.83) and Eq. (6.85) to these
two regions yields
> /i 1-0 Uno

and
2+ jT^RuTl = 1 V2 + J^ M r o o (6.87)
1-0 1K
KMM - 1-0 KK M - 1
Combining Eq. (6.86) and Eq. (6.87) gives
2
UooUi = -m^RMTl + ^-Ul (6.88)
1
KM + KM + 1
Thus, UOQ can be evaluated from Eq. (6.88). The gas speed just after the shock can also be
obtained from Eq. (6.88), by setting 0 = 0, as
U2UX = -^y-RuTi + ^rUl (6.89)

K+ 1 K+1
A comparison of Eq. (6.88) and Eq. (6.89) indicates that the states of the region far behind
the shock and that just after the shock differ and vary with the particle mass fraction and/or
particle thermal capacity. In the transition region in which the velocity and temperature of
the particles are changing from an equilibrium before the shock to a new equilibrium after
the shock, the changing processes can be described on the basis of the equation of motion
and the equation of energy for the particle phase in the gas-solid mixture. Assuming that
U2 = UOQ, T2 = r ^ , and only the drag force and the convective heat transfer are important,
the equation of motion for a particle can be given by
f V ^ ff - Upf (6.90)
and the equation of energy for a particle can be given by
^ = TtdvK (6.91)
where Nup is the particle Nusselt number and K is the thermal conductivity of gas. The
boundary conditions are given by
f/plz=0 =
(6.92)
U,plz->oo (= = Uo ( = Tpoo) =
The rates of change in the particle velocity and temperature are therefore dependent
on the values of the drag coefficient C D and the heat transfer coefficient, respectively. In
a gas-solid flow, both the drag coefficient and the heat transfer coefficient are functions
of the particle Reynolds number and the volume fraction of solids. These relations are
primarily represented by empirical or semiempirical relations (see 1.2 and 4.3). In a
very dilute suspension, C D and Nup for a single particle can be applied to Eq. (6.90) and
Eq. (6.91). As an example, for a mixture of air and 10 /im glass beads with a particle to
gas mass flow ratio of 0.2 and a Mach number of 1.5, on the basis of the equations given,
different relations for drag and heat transfer coefficients are shown to significantly affect
the pressures, velocities, and temperatures, as shown in Fig. 6.13 [Rudinger, 1969]. The
1.4
>s -
=*
1.3 - A
B, .
1.5 - 3.0
/// D ^
^ =
Particles /
D
1.0 ^ ^ \ ^ Particle *
1.1 -
^ -
^^5. velocity
B
1.0 1 i 1 i 0.5 1 1 1 1 1 2.0

0.10 0.20 0.30 0 0.10 0.20 0.30
Distance behind shock wave, m Distance behind shock wave, m
(a) cD NUp
(b)
A Standard curve 2
_ 1/2
B Standard curve 2+0.6 Pr"3 Rep
C 24/Rep (Stokes) 2
D 24/Ren (Stokes) 2+0.6 Pr1/3 Re m
Figure 6.13. Effect of CD and Nup on the dynamic properties of phases over a plane shock in a
gas-solid suspension with a Mach number of 1.5, 10 ^m glass beads, and mp = 0.2 (from
Rudinger, 1969): (a) Temperature distributions; (b) Pressure and particle velocity distributions.
0.4
0.2
0.0
Figure 6.14. Effect of particle volume fraction on the equilibrium velocity behind a normal
shock in a gas-solid suspension (from Rudinger, 1965).
effects of the particle volume fraction and particle to gas mass flow ratio on the equilibrium
velocity, UQQ, behind a normal shock wave as a function of the velocity in front of the
shock, U\9 are shown in Fig. 6.14 [Rudinger, 1965]. It is seen that the equilibrium velocity
decreases with increasing mass ratio of particle to gas and increases with increasing particle
volume fraction. It is noted that with an increase in the particle volume fraction, particle-
particle collisions and other types of particle interactions become increasingly important
and thus the assumption of negligible particle interactions used in the calculation is no
longer feasible.
We introduce only the normal shock waves in an infinitely large, dilute, and uniformly
distributed gas-solid suspension, which represents the simplest case for the propagation
of a normal shock. For a normal shock passing through a one-dimensional duct with var-
ied cross-sectional area, a set of equations for both gas and particle phases that may be
solved by using the method characteristics can be derived [Soo, 1990]. Examples using
the method of characteristics were reported for the centered expansion wave and shock
wave generated by an impulsively driven piston in a suspension of 10 fim glass beads in air
[Rudinger and Chang, 1964]. The normal shock wave analysis can be extended to oblique
shocks. For oblique shocks, the shock front is not perpendicular to the direction of the
fluid motion. However, for large particles, the velocity vectors of the gas and particles are
no longer parallel after the oblique shock because of the inertia of the particles. There-
fore, the flow after the shock becomes two-dimensional, thus requiring more complicated
mathematical treatments [Morgenthaler, 1962; Peddieson, 1975; Chang, 1975].
6.5 Instability
An instability analysis of a multiphase flow is usually based on the continuum as-
sumption and the equations of motion of the phases. Two different approaches are common.
One approach is to study the wave propagation speed from the equations of motion using
an analogy to the surface wave situation. The other approach is to study the growth rate
of the perturbation wave when a small perturbation wave is introduced to the system. The
criterion for stability can then be derived. A perturbation wave can be expressed by
y = yoei(kr~ct) (6.93)
where k is the wave number and C is the wave speed. It is noted that when the imaginary
part of C is positive, the fluctuation will grow exponentially. Thus, an instability criterion
can be determined from the analysis of the imaginary part of the wave speed.
In gas-solid multiphase flows, the wave propagation method is commonly used to study
the stability of stratified pipe flows, where an analogy to gas-liquid wave motion with a
free surface is prominent. The perturbation method is commonly used to study the stability
of a fluidized bed. In the following, both methods are introduced.
6.5.1 Wave Motions in Stratified Pipe Flows

A fundamental account of unsteady motion in a dense suspension requires knowl-
edge of phase interactions and origins of wavy stratified motion [Zhu, 1991; Zhu et al.,
1996]. Flow stratification involves a dense phase of solids flowing along the bottom of a
horizontal pipe with wave motions, or a layer of the dense phase of solids flowing near the
wall of a vertical pipe with wave motions. The former was analyzed, in part, on the basis
of the theoretical work of de Crecy (1986) on stratified liquid-vapor flows.
6.5.1.1 General Considerations for Stratified Pipe Flows

To simplify the analysis of wave motions in stratified pipe flows, several assumptions can
be made:
(1) A stratified pipe flow is composed of a dilute flow and a dense flow with an
infinitesimal thickness of interface.
(2) Each phase (gas or solids) is continuous in any cross section.
(3) Regionally averaged flows are unsteady and one-dimensional.
(4) The area and shape of the cross section of the pipe are constant, and the pipe wall
is impermeable.
An instantaneous regional average of phase j (j = gas or solids) in the region k is defined

as
(6 94)
i '
Ak(z,t)
From Delhaye (1981), the instantaneous regional-averaged continuity equation projected
along the tube axis can be expressed as
/ ^ j k dC (6.95)
C(zJ) dt nk nkc
6.5 I Instability 111
Ak(z,t)
Figure 6.15. General configuration for regional averages.
and the momentum equation projected along the tube axis can be given by
a a - A k (p jk / Zj ) + A k (p jk ) - Ak(nz r j k nz
a7 Yz
dC dC
- nz-(nk- {p]kl - r j k (6.96)
Jc( nk nkc JCUCk(z,t)
where Ak, Ai, C k , C, #tz, n kc , and wk are defined in Fig. 6.15.

The integrals in the preceding equations can be simplified. It is noted that (1) C(z,t)
represents the common boundary of interface A{ with the cross section plane; (2) Ck(z, t) +
C(z,t) is the enclosed curve in the cross section and Ck is on the wall of the pipe; and
(3) wkc, located in the cross section plane and directed away from region k, is the unit vector
normal to C. The mass transfer rate through the interface is expressed by
I/
c(z,t) dt
dm]k
r
nk
dC
nkc
(6.97)
where m^ is the mass transfer rate from phase j to phase 1 in region k, and m}\k is the mass
transfer rate from phase 1 to phase j in region k. The momentum transfer rate due to mass
transfer through the interface is given by
N-\
dC
(6.98)
JCc(z,t) dt nk - nkc J2
From assumption (4), nk nz = 0 on Ck. Hence, the integral with pressure can be written
as
f dC
dC ff dC
/ (z k)Pjk ^ //
= ^jki n nk (6.99)
JCUCk(z,t) ^ k ' Wkc JC(z,t)
where P^ is the pressure of phase j in region k at the interface. Furthermore, the limiting
form of the Gauss theorem gives [Delhaye, 1981]
L C(z,t)
Hz.nk^L_
n
k * Hkc
= _l^
^Z
(6.100)
Equation (6.99) is then reduced to

f dC dAk
/ (wz k)Pjk = -^jki"^ (6.101)
Jcuc k (z,o k nkc oz
Moreover, we may express the interfacial stress by
LJC(z,t)
nz - (nk Tjk)
% ' Wkc
= -kzjkiC (6.102)
where Tjk is the stress tensor of phase j in region k; ijki is the interfacial stress of phase j in
region k projected on the pipe axis; and 6k = 1 for one region and 1 for the other; and the
wall shear stress or frictional force at wall is given by
I ck(z,o
nz (nk Tjk)
dC
nk nkc
= CkXjf%w
where Xjf is the fraction of the frictional perimeter of phase j , and tjkW is the wall shear
(6.103)
stress of phase j in region k projected on the pipe axis.

The volume fraction (or area fraction in this case) of phase j in the region k is
with
^ k = l (6.105)
Now, define the intrinsic average x(x]k) of phase j in the region k as

1
(6.106)
The relationship between regional average and intrinsic average is
\^jk) : C^jk \-*jk) (6.107)
To simplify the problem further, we adopt a concept similar to that for the intrinsic averaged
velocity and intrinsic velocity tensor as in Eqs. (5.129) and (5.130). Thus, Eqs. (6.95) and
(6.96), respectively, become
ad ,. 9
d
- Aktfjk1" (Pjk) + A k V (pjk)' (W]k) = y W - rfijki)C (6.108)
dt dz ^
and
- A k jk '(p jk }(/ Z
d
+ ' (pk) - - Aka-]k (n z r j k n z )
Yi 01
N-l
( ^-miklWik)C-Piki-^- (6.109)
where the velocity tensor may be approximated as l (WjkWjk) ^ l (W^)1 (Wjk).

The expansion rate projected on the pipe axis for the gas phase (j = gas) may be ex-
pressed by
yW(n z r k nz> = Fk((Wpk) - l {Wk)) (6.110)

01
where xk is the stress tensor of the gas phase in region k. It is noted that the combination
of the last two terms on the left-hand side of Eq. (6.109) represents the effective total stress
projected along the pipe axis, Tpk. Hence, for the solid phase in a dense suspension region,
we have
nz) = ~ - '(Wvk)) (6.111)
where Cpp is the momentum transfer coefficient due to phase expansion.

For subsequent discussions, the symbolx ( ) is removed with an understanding that the
velocities and densities involved are intrinsically averaged.
6.5.1.2 Stratified Horiiontal Pipe Flows

A stratified horizontal pipe flow is schematically illustrated in Fig. 6.16. In addition to
the assumptions given in 6.5.1.1, other assumptions are introduced in the analysis of the
stratified horizontal pipe flow:
(1) The pressure distribution over the cross section is induced by gravity.
(2) The mass transfer between phases can be neglected.
(3) The presence of particles in the dilute region is negligible, whereas particles in the
dense region can be regarded as in a moving bed state.
In this section, we only discuss the case of stratified horizontal pipe flow in a rectangular
duct. The expansion rate projected on the pipe axis for the gas phase (j = gas) is small (as
7//////////////7/7/7/////////
Figure 6.16. Phase interaction in stratified horizontal pipe flow.

is true for low-velocity channel flows), i.e.,
Akak^/iz r k nz) 0 (6.112)
In the dilute region (denoted by subscript "1") ot\ 1; the regional cross-sectional area
is given by
A /z
where A is the cross-sectional area of the duct; yx is the depth of interface; and h is the
height of the duct. From Assumption (1), the local pressure is given by
Pi=Pi-g I" pdy (6.114)
where Px is the pressure at the interface. The regional averaged pressure is thus obtained as
Thus, Eqs. (6.108) and (6.109) yield
0 (6.116)
dt dz
and
9 9 2 S ^
9/ * dz 2dz
= ^ M i + I ( T i C H-CiTgw) (6.117)
p 9z p
where rgw is the wall shear stress of the gas phase in the dilute region projected on the pipe
axis. Using Eqs. (6.113) and (6.116), we have
^ A TTT'J^ ^. T T T (7/1] tTTO (7/1] /I I .,. ^ Jl TTT .> " /I 1 S - ^ ^ Q^
dz dt dz h \ dt dz (o.llo)
Equation (6.117) then becomes
dt h \ dt dz J \ h) dz p dz
+ -(TiC+CiTgw)=0 (6.119)
In the dense region (denoted by subscript "2"), the solid particles are in a moving bed
state. We may further assume that there is no slip in motion between the gas and solids
so that we can treat the mixture as a pseudosingle-phase flow (i.e., (#2 = 1)- The regional
cross-sectional area is given by
Â = I:
h
The regional averaged pressure of the mixture is expressed as
(p2) = P{ + - r (g r pmdy)dy = Pi + ^ (6.121)
Thus, the continuity and momentum equations of the mixture can be obtained from
Eqs. (6.108) and (6.109) as
= 0 (6.122)
01 OZ
and
+ (-tiC+C mrmw)=0 (6.123)

An
where rmw is the wall shear stress of the mixture in the dense region projected on the pipe
axis.
For the internal unsteady wave motion, it may be reasonable to assume that the total
mass flow is constant, i.e.,
pAxW + pmA2Wp = const (6.124)
Combining Eq. (6.119) with Eq. (6.123) and using Eq. (6.124) yield a wave equation at the
interface as
Ct- + C z- = C c (6.125)
at dz
where
2(pW - PmWp)
~ h
(pW2 - pmW2) u
Cz = ^ ^ + pg + (An - P) J (6.126)
r _ _^_ _ g CmTmw)
c
" dz A
According to the preceding model, there is always an unsteady flow since C t # 0. The
wave speed of the interface is obtained as
(An -
In addition, a steady wave motion, C c = 0, leads to a balance between pressure drop and
wall friction as
- = (6.128)
dz A
Wall Core
region/ region
R
Figure 6.17. Phase interaction in stratified vertical pipe flow.
6.5.1.3 Stratified Vertical Pipe Flows

A stratified vertical pipe flow is schematically illustrated in Fig. 6.17. In addition to the
assumptions given in 6.5.1.1, other assumptions are introduced in the analysis of the
stratified vertical pipe flow:
(1) The flow is axially symmetric.

(2) The flow comprises a dilute suspension in the core region and a dense flow in the
wall region.
(3) The solids concentration in each region is uniformly distributed.
(4) There is no phase change between the gas and solids.
In terms of Eqs. (6.108) and (6.109), the continuity and momentum equations of each
phase are given in the following for each region.
A. GAS PHASE IN THE CORE REGION

a a c
A cac + AcacWc = (mwc - mcw) (6.129)
at dz p
and
d
V] + Acacg + - Acacpc
p dz
cc c + ode
_ ^L (6.130)
P P
(Wpc-Wc)"
where the first term
p on the right-hand side of Eq. (6.130) represents the drag force between
the gas phase and the particle phase, and F gpc is the coefficient of momentum transfer from
particles to gas due to drag in the core region. TCI is the interfacial stress of the gas phase in
the core region projected on the pipe axis.
6.51 Instability 277
B. PARTICLE PHASE IN THE CORE REGION

d d C
Acapc + -~ cWpC = (m pwc - rap ) (6.131)
at az pp
and
d d
AcavcWpc + --AcQ Acapcg
dt dz
_ Pgc nxf ur \ (6.132)
V" c " vc)
PP
Pv
where Tpci is the interfacial stress of the particle phase in the core region projected on the
pipe axis and F pgc is the coefficient of momentum transfer from gas to particles due to drag
in the core region.
C. GAS PHASE IN THE WALL REGION

d d C
- mwc) (6.133)
at dz p
and
3 d Id
- H
dz p dz
(mcwWc - -
p dz
(6.134)
where F gpw is the coefficient of momentum transfer from particles to gas in the wall region;
rwi is the interfacial stress of the gas phase in the wall region projected on the pipe axis;
and rw is the wall shear stress of the gas phase projected on the pipe axis.
D. PARTICLE PHASE IN THE WALL REGION
8
,pw + wpw pw = pcw Wpwc
c (6.135)
~ PP
and
3 9 2 19/ , 9W p w \
pw + W P W A- A n
"pw + ^ w " pp
a? 'VO! W
P
9^^ dz )
Fpgw
p ^pcw - pwc pw
c TpwiC C w Xw P r p w
(6.136)
p c
-P Pv
where F pgw is the coefficient of momentum transfer from gas to particles in the wall region;
Tpwi is the interfacial stress of the particle phase in the wall region projected on the pipe
axis; and rpw is the wall shear stress of the particle phase projected on the pipe axis. The
last term on the left-hand side of Eq. (6.136) represents the expansion rate of the particle
phase due to interparticle collisions.
Using the preceding continuity and momentum equations for each phase in each region,
the equations for the interface can be obtained. For the gas phase in the core region, from
Eqs. (6.129) and (6.130), we have
dt V Wcp J dz \WC dt Wc Wcp dz J ac acWc

(6.137)
where
(mwc
Q = ~ mcw)C (6.138)
P
and
Mc = ^ ( W p C - Wc) + (mwcWw - mcwWc)~ - (6.139)
P P P
For the particle phase in the core region, from Eqs. (6.131) and (6.132), we have
(6.140)
oi oz \wvc oi vvvc/ apc apcKKpc
where
gp = ^ ^ (6.141)
and
F C x ciC
c pc pwc pw pcw pc
pp pp pp
In the wall region, note that Aw = A Ac. Therefore, for the gas phase in the wall region,
from Eqs. (6.133) and (6.134), we have
dAc ( P{~ Pw\ dAc ( 1 dWw g 1 dpw\
h I W w H - I I h - (- --- - IAc
dt \ Wwp J dz \Ww dt Ww Wwp dz J
=2^+^-(4-d-jr+4-+w-M)A (6 143)
-
where
wg w
Mw = - ^ ( W p w - Ww) + (mcv,Wc - mwcWw) + (6.144)
p p p p
For the particle phase in the wall region, from Eqs. (6.135) and (6.136), we have
Q
PW
dt V PP^PW dz ) dz V^PW dt ^pw PpWpw dz2 J c
GP Mpw / i a^ p w g cPp a 2 w p w \
; ^ H - h r-T" M (6.145)
where
CwXwpTpw
Mpw = ^(WW- Wpw) + (mpcwWpc - mpwcWW- + ^ ^ - (6.146)
/Op Pp Pp Pp
It is noted that Eqs. (6.137), (6.140), (6.143), and (6.145) refer to the identical interfacial
area, which has the following common form of expression
d
-^ + F1d-^ + F2Ac = F3 (6.147)
Therefore, the equation of the interface is a wave equation, reflecting that the flow is in
an unsteady wave motion. Furthermore, because the corresponding coefficient of each
term in these four equations should be equal, several relations among phase quantities
result:
1. Relation for the slip velocity between phases in the core region
Wc-Wpc = iszA (6.148)
2. Relation for the slip velocity between phases in the wall region
Ww - Wpw = - % - ^ j W + i Z * * (6.149)
3. Relation for Wpw and Wpc
- ^ d ^ L = Wpw-Wpc (6.150)
ppWpW dz
4. Relation for Wc and W^
dt wc dt )
^ \wpQ
r ^ wj
^ 4 ^w dz
(6.15.)
P c
5. Relation for Ww and Wpw
_j_dw^\ /j i_ = i ddPw
Pw cCp p
wvw dt Wv dt ) V^PW w w / P ^ W dz ppwvv, dz2
TT T r\ . I ' 1 TTT TTTT I O
(6.152)
6. Relation for Wpw and
/J l_\ Cpp d2Wpw

dt Wpc 8f ; \W
\Wpw W)8 W 92
(6.153)
7. Relation between momentum transfer and mass transfer in the core region
2e_j*c_ = 2 e p _j^ (6154)

ac acWc Qfpc Qfpc^pc
8. Relation between momentum transfer and mass transfer in the wall region
2A + J ^ _ = 2ep_ + j ^ _ (6.155)
9. Relation between momentum transfer and mass transfer of particle phase between
two regions
*pc
' pw +
*PCi
'pw dt 77^M (6.156)
It is important to note that these 9 relationships are essential to the closure of the
problem. There are a total of 14 unknowns, namely, Ww, Wc, Wpc, Wpw, mcw, m wc , mpcw,
^pwc5 Pc> Pw, Pi, Ac, ofpc, and ofpw. Eight independent equations representing four pairs of
continuity and momentum equations (Eqs. (6.129) through (6.136)) are initially developed.
Converting the momentum equations into wave equations and equating the coefficients
of these wave equations yield a total of 14 independent equations, namely, 4 continuity
equations, 1 wave equation, and 9 relations, and hence the closure of the problem.
6.5.2 Continuity Wave and Dynamic Wave

Waves in gas-solid flow generated by a perturbation to the flow can propagate in
the gas-solid medium as indicated in 6.4. The propagation may yield continuous changes
in local volume fractions of phases. It may also yield a step (sharp) change in local gas
void fraction or solids holdup with afinitediscontinuity, known as a shock (or shock wave).
"Waves" or "shocks" exist in many different forms depending on the origin of the disturbance
[Wallis, 1969]. Basically, two classes of waves can be distinguished, i.e., continuity waves
and dynamic waves. Although wave phenomena in gas-solid flow are two-dimensional or
three-dimensional in nature, for simplicity only the one-dimensional wave is considered in
the following discussion.
6.5.2.1 Continuity Waves

A continuity wave is generated when a perturbation in a gas-solid flow yields a local change
in the solids holdup or gas voidage which propagates through the vessel at a certain velocity.
In other words, the steady-state gas voidage or solids holdup simply propagates into another
one, as illustrated in Fig. 6.18, and the propagation process does not introduce any dynamic
effect of inertia or momentum [Wallis, 1969].
Un U, + 8!/,
///////////////Y////////////
Figure 6.18. Continuity wave propagation due to the variation of volume fraction (from
Wallis, 1969).
Consider an incompressible gas-solid flow in a tube of constant cross-sectional area.

Denoting U, Up, and Um as the superficial velocities for gas, solids, and a mixture of gas
and solids, respectively, Um is related to U and Up by
Um = U + Uv (6.157)
The continuity wave velocity can be derived by considering the solids holdup variation
across a control volume enclosed by the aa and bb planes given in Fig. 6.18. The figure
shows the propagation of a wave with velocity Vc from the aa plane where the solids holdup
is ap to the bb plane where the solids holdup is ap + 8ap. A solids flow approaching the
moving wave front would be equal to that leaving the moving front, i.e.,
Up - apVc = Up + 8UP - Vc(ap + 8ap) (6.158)
which yields
Vc = ^ (6.159)
dap
Note that Up can also be expressed by
Up = upap (6.160)
where up is the particle velocity. Equations (6.159) and (6.160) yield
^ (6.161)
An unsteady-state mass balance for solids in a differential element with a small pertur-
bation in a system otherwise in equilibrium yields
(top (KOpWp)
dt dz
Equation (6.162) leads to
hwP = 0 (6.163)
dz) \dz
Comparing Eqs. (6.161) and (6.163) yields
(6.164)
Equation (6.164) expresses the continuity wave velocity as the ratio of the change of the
voidage with time to the change in voidage with distance, which provides an alternative
definition for Vc [Rietema, 1991].
Define the slip velocity Vs as up u where u = U/a. In equilibrium conditions, Um
is constant throughout the system, indicating that (an + apup) or (up aVs) is constant.
Thus, Eq. (6.161) can be expressed by
Vc = Up + a ^
p (6.165)
dav
For batch-solid operation, uv = 0. Thus, Eq. (6.165) can be reduced to
Vc = (l-a) (6.166)
da
Consider a particulate fluidization system (see 9.3.2) where the Carman correlation given
for a < 0.7 [Carman, 1937] is applicable
Thus, for this system, Vc can be expressed from Eqs. (6.166) and (6.167) as
Vc = U{3-2a) (6.168)
a
6.5.2.2 Dynamic Waves
Wallis (1969) defined the dynamic wave in one-phase flow as being that which occurs
whenever there is a net force on the flowing medium produced by a concentration gradient.
For a two-phase flow, i.e., gas-solid flow, theflowmedium refers to the gas phase and
the concentration refers to the solids holdup. Thus, to analyze dynamic waves, one can
examine the wave equation obtained from the perturbation of the momentum and mass
balance equations for the gas and solid phases. The analyses given later for both 6.5.2.2
and 6.5.2.3 follow those of Rietema (1991).
The mass balance equations for gas and solid phases yield
and
9 ( 1 a )
~ ^ = 0 (6.170)
In terms of the simplified two-fluid model, the momentum balance of the gas phase can be
given as
a^-+apg + Fi=0 (6.171)

02
where p represents the pressure of the mixture of gas and solids, F\ is the interacting
force between gas and solids, and the inertial terms in the gas phase are neglected. The
pseudosteady-state condition is applied to Eq. (6.171) considering the assumption that the
dynamic acceleration is negligible because of the low gas density and small variation of the
gas velocity. The momentum balance of the solid phase can be expressed by
(1 - c x ) p p ^ + (1 - a)ppup^- + (1 - a)^- + ^ + (1 - a)ppg -F{ = 0 (6.172)

01 02 02 02
where az is the particle normal stress due to particle-particle interactions and the variation
of particle shear stresses rxz and ryz is assumed to be negligible compared with the variation
of GZ in the z-direction. Furthermore, we may define the elasticity modulus of the particle
phase as
daz
E (6 173)
= ~it -
dz
Combining Eqs. (6.171), (6.172), and (6.173) yields
(1 - a)pp [ ^ + M p ^ ] - E^ + (1 - aO(pp - p)g - = 0 (6.174)

Therefore, we obtain three independent equations (Eqs. (6.169), (6.170), and (6.174)) with
three dependent variables, i.e., a, w, and wp.
In order to obtain an expression for the dynamic wave speed, consider a perturbation
imposed on a steady-state flow yielding a, w, and wp, as
a = a0 + a' (6.175)
/
w = w+ w (6.176)
f
uv = up + u v (6.177)
The superscript "0" indicates the steady-state value and the prime indicates the perturbed
value. The perturbation of F-Ja yields
. (6.178)
a \ a
Consider a particulate fluidization system where U = 0 and that the Carman correlation
for F{ given in the following under the condition of a < 0.7 [Carman, 1937] is applicable
Thus, Eqs. (6.169), (6.170), and (6.174) through (6.178) lead to
a z - 8 , . = 0 (6.180)
The first two terms establish the basic wave equation describing the dynamic wave propa-
gation with velocity Vd expressed by
(6.181)
The coefficient in the third term, (3 2a)u, is the continuity wave velocity as given by
Eq. (6.168). Thus, Eq. (6.180) can be given by
= 0 (6.182)
where
180^(1-g)
C = - (6.183)
rf2p(O)3
6.5.2.3 Instability Analysis of a Batch-Solid Fluidization System

The instability analysis of Eq. (6.182) for a' can yield a stability criterion accounting for
the conditions under which bubbles will form in a fluidized bed. In this analysis, when
bubbles form, the bed is regarded as unstable. Thus, the instability analysis gives rise to
the condition for minimum bubbling of the bed (also see 9.4).
First, let us express the voidage perturbation a' in the form of a Fourier series as
a = exp[at + icoh(t - z/Uh)] (6.184)
at
where e is the amplitude of perturbation; a > 0 indicates the perturbation will yield an
unstable bed, while for a < 0 the perturbation will be damped out yielding a stable bed.
C/h is the propagation velocity of the perturbation and co^ is its frequency. Substituting
Eq. (6.184) into Eq. (6.182) yields
f ^ \z0 (6>185)
which gives rise to

Q2 / yl _ TJ2S
/n 2
1 c h (6.186)
^ 4 \u-vZ,
)
for the real part and
for the imaginary part. For the perturbation to grow, a has to be positive and, hence,
vc>ub (6.188)
From Eq. (6.186), a>l > 0 andthus we have
V2 > U2 > V 2 (6.189a)
c h d
or
Vc < Ul < V2 (6.189b)
Equations (6.188) and (6.189) yield the criterion given for a bed to be unstable, i.e., a bed
with a bubble formation:
Vc > V d (6.190)
Likewise, for perturbation to be damped out, a < 0, and hence
Vc < Uh (6.191)
Equations (6.189) and (6.191) yield the criterion given for a bed to be stable, i.e., a bed
without a bubble formation:
Vc < Vd (6.192)
Thus, the condition for the minimum velocity for bubble formation or the minimum bubbling
point can be obtained when
Vc = V d (6.193)
6.6 Particle-Turbulence Interaction

When particles are introduced into a turbulent flow, they are dispersed through
turbulent fluctuations of the gas. The addition of particles can enhance or reduce the gas
turbulence, affecting the behavior of the mean flow field (e.g., drag reduction in pipe flows).
In very dilute gas-solid flows (e.g., ap <3C 10~5), the particle dispersion by turbulence can be
studied by assuming that the turbulent flow is not influenced by the presence of the particles
(i.e., one-way coupling). When the particle concentration is high (e.g., ap > 10~5) such
that the particle momentum loss or gain transforming to the turbulence can no longer be
neglected (i.e., two-way coupling), the modulation of the turbulence by the particles should
be considered.
One of the earliest models for turbulence modulation in homogeneous dilute particle-
laden flows is Hinze's model (1972), in which the assumption of "vortex trapping" of
particles is employed. On the basis of this model, the particle turbulent kinetic energy, k v,
is determined by the local gas turbulent kinetic energy k as
^ = 1+ - (6.194)
where r s is the Stokes relaxation time defined by Eq. (3.39) and r e is the eddy existence
time defined by Eq. (5.253). On the other hand, the effect of particles on gas turbulence
can be given by
S&V1 (6.195)
p )
where k0 is the local gas turbulent kinetic energy in the absence of the particles. Equation
(6.194) indicates that kv is always smaller than k, and in the same turbulent flow, the larger
the particle size the smaller the value of kv. Moreover, according to Eq. (6.195), particles
should always reduce gas turbulence. However, many experimental results, especially for
large particles, do not support this indication [Tsuji and Morikawa, 1982; Lee and Durst,
1982; Tsuji et al., 1984; Zhou and Huang, 1990]. In actual applications, particles are
not always trapped by eddies, and turbulence modulation is a result of multiple particle-
turbulence interaction mechanisms. Therefore, Hinze's model seems to be oversimplified
to describe the general particle-turbulence interactions.
On the bais of an extensive review of experimental results of turbulence modulation in
dilute suspension pipe flows and jet flows, Gore and Crowe (1989) proposed a critical ratio
of particle diameter to a characteristic integral length scale of turbulence by the following
relation
'0.1 (6.196)
where /e is the Eulerian integral length scale. The turbulence intensity is enhanced when
the ratio given in Eq. (6.196) is greater than the critical value and is suppressed when the
ratio is less than the critical value. This criterion indicates that the turbulence is attenuated
by small particles while it is amplified by large particles.
One predominant mechanism for the enhancement of turbulence in the presence of
large particles is due to wake shedding, i.e., vortex shedding from the particle wake. With
Rep < 110, there is no wake shedding; with Rep > 400, wake shedding occurs. In the range
400 < Rep < 1,000, the wake shedding frequency is given by [Achenbach, 1974]
(6 197)
-
where St is the Strouhal number and / is the wake shedding frequency. For a typical
range of large particles (e.g., 500 /xm < dv < 5,000 /xm), the corresponding wake shedding
frequencies are in the range where a considerable particle effect on the turbulence intensity
is observed. On the basis of this observation, Hetsroni (1989) suggested that large particles,
with Rep > 400, exhibit wake shedding, leading to an enhancement of the turbulence; i.e.,
the wake shedding by large particles causes energy to be transferred from the average
velocity to the fluctuating velocity.
In general, at least six mechanisms, which are not independent of each other, contribute
to turbulence modulation in gas-solid suspension flows:
(1) Dissipation of turbulent kinetic energy by the particles.

(2) Increase in the apparent viscosity due to the presence of particles.
(3) Vortex shedding or the presence of the wake behind the particles.
(4) Gas moving with the particles as carried mass of the particles.
(5) Enhancement of the velocity gradients between particles.
(6) Preferential concentration of particles by turbulence, i.e., the particles selectively
concentrated in particular structures by turbulence may cause rapid attenuation of
that structure or trigger a new instability [Squires and Eaton, 1990].
At this stage, complete modeling including all of the preceding mechanisms is not pos-
sible because of the complex coupling relationships of the turbulent interactions and the
lack of knowledge of the sources of turbulence generation. Nevertheless, simple mecha-
nistic modeling accounting for a few of the predominant mechanisms for the modulation
of turbulence is possible. In the following, a simple mechanistic model developed by
Yuan and Michaelides (1992) is introduced. This model includes only two predominant
mechanisms of turbulence modulation in the dilute gas-solid flows, namely, (1) turbu-
lence reduction due to the kinetic energy dissipation from an eddy, for the acceleration of
a particle, and (2) turbulence enhancement due to the wake of the particle or the vortex
shedding.
Consider a particle with velocity wp entering an eddy of velocity u and interacting with
it for a period of TJ which can be estimated from Eq. (5.251). It is assumed that the velocity
is scalar in the modeling and the only interacting force between the particle and gas is
the drag force F&. The rate of work done by the gas is equal to F\yu and the change of
kinetic energy of the particle is Fûv. The rate of energy dissipation e is thus expressed
by
e = FD(u - up) = ^Cvdlp{u - up)2\u - up\ (6.198)

To simplify the following calculation, the drag coefficient is expressed by
CD = ^ (6.199)
Rep
where C, a function of Rep in principle, is regarded as constant during the interaction.
During the particle-eddy interaction, the equation of motion for the particles can be
expressed by
p ^ E = C^piu - uv)\u - p| (6.200)

Combined with Eq. (6.199), Eq. (6.200) gives the approximation of the particle velocity
during the time interval X{ as
uv = wpo + (u - Wpo) 1 - expf j , t < Ti (6.201)
The total energy dissipation from the eddy, which equals the total work performed by the
eddy on the particle during the particle-eddy interaction, is thus obtained by integrating
Eq. (6.198) with respect to t
(6.202)
In this model, the velocity disturbance by the particle is from both the wake behind
the particle (Rep > 20) and the vortex shedding (Rep > 400). Hence, the changes in the
kinetic energy associated with the turbulence production are proportional to the difference
between the squares of the two velocities and to the volume where the velocity disturbance
originates. It is further assumed that the wake is half of a complete ellipsoid, with base
diameter of dp (same as the particle diameter) and wake length of/w. Thus, the total energy
production of the gas by the particle wake or vortex shedding is
= ^dpwp(u2 - u2v0) ^dp wpVs(2u - Vs) (6.203)
Combined with Eq. (6.202), Eq. (6.203) yields the total turbulence modulation as
AEt = Ap - AEd = ^d^pVs(2u - Vs) - ^d3pPpV? [l - e x p ( - ^ ) J (6.204)
For the case of very small particles where ti ^> ts, the particle velocity approaches the
gas velocity and the wake disappears. The asymptotic expansion of Eq. (6.204) for fine
particles is thus given by
(6.205)
In the case of large particles where T{ <. is, the particle velocity does not change appreciably
during the particle-eddy interaction and the production term predominates. Consequently,
the asymptotic value of AEt for large particles equals A p . The asymptotic expansion of
Eq. (6.202) shows that fine particles will cause turbulence reduction, which is proportional
to the cube of the particle diameter; whereas large particles will predominantly cause an
increase in turbulence, which is proportional to the square of the particle diameter. This
indication is in agreement with the data compiled by Gore and Crowe (1989).
Despite the importance of particle-turbulence interaction in the dispersed multiphase
flows, understanding of the phenomena is still lacking, in part because of the difficul-
ties in obtaining comprehensive experimental data to substantiate the sources of turbu-
lence generation. New avenues of gaining understanding may come from computational
fluid mechanics using numerical simulations such as direct numerical simulation (DNS)
[McLaughlin, 1994], large-eddy simulation (LES) and discrete vortex simulation (DVS).
These numerical simulations may provide some insights into the physical mechanisms of
the particle-turbulence interaction in dispersed multiphase flows.
Nomenclature
A Constant, defined by Eq. (6.52) Specific heat of gas at constant
A Friction force coefficient pressure
Ac Contact area Specific heat of mixture at
Ac Cross-sectional area in the core constant pressure
region Specific heat of gas at constant
Ai Interfacial area volume
Ak Cross-sectional area of phase k Specific heat of mixture at
Cross-sectional area in the wall constant volume
Aw
region D Parameter, defined by Eq. (6.71)
a Particle radius Particle diameter below which the
Speed of sound in a pure gas in particle contribution to the
a\
front of the shock wave pressure of mixture exceeds 1
percent at atmospheric pressure
a% Speed of sound in a pure gas
Particle diameter
am Speed of sound in a gas-solid
Modulus of elasticity of the
mixture
particle phase
B Constant, defined by Eq. (6.53)
Brittle erosion parameter, defined
B Inertia correction coefficient by Eq. (6.14)
C Common boundary of interface Brittle erosion parameter with
with cross section plane ofi =am
C Coefficient of wave equation ED Ductile erosion parameter, defined
C Constant, defined by Eq. (6.54) by Eq. (6.8)
and Eq. (6.183) Ductile erosion parameter with
C Resistance coefficient due to a{ = am
Basset force Internal energy per unit volume of
CD Drag coefficient mixture
ck Common boundary of pipe wall in Force
region k with cross section plane Drag force
Momentum transfer coefficient Coefficient of momentum transfer
due to phase expansion from particles to gas in the core
c Specific heat of particles region
Nomenclature 289
p Me Equilibrium shock Mach number

Coefficient of momentum transfer
from particles to gas in the wall Mpc Parameter, defined by Eq. (6.142)
region Mpw Parameter, defined by Eq. (6.146)
Fi Interacting force between gas and Mw Parameter, defined by Eq. (6.144)
particle racw Gas mass transfer rate from the
Fk Coefficient of momentum transfer core region to the wall region
from particle to gas mH Mass of a hydrogen atom
Fn Normal collisional force
p Coefficient of momentum transfer
7?2jk Mass of phase j in region k
from gas to particle in the core ^jkl Mass transfer rate from phase j to
region phase 1 in region k
Ajlk Mass transfer rate from phase 1 to
Coefficient of momentum transfer
from gas to particle in the wall phase j in region k
region mp Mass ratio of particle to gas
Coefficient of momentum transfer mpCw Particle mass transfer rate from
fpk
from gas to particle the core region to the wall region
Ft Tangential cutting force pwc Particle mass transfer rate from
f Coefficient of friction the wall region to the core region
mwc Gas mass transfer rate from the
f Force per unit mass wall region to the core region
f Wake shedding frequency Ni m Impact number, defined by
fu Mass density function Eq. (6.5)
h Height of the duct Nw Parameter, defined by Eq. (6.59)
I Unit tensor Nup Particle Nusselt number
J Mass flux of gas Pressure at interface
JP Mass flux of particles Pressure of phase j in region k at
K Thermal conductivity the interface
KB Brittle resistance parameter, P Pressure of gas
defined by Eq. (6.15) Pressure in front of the shock
Pi
KD Ductile resistance parameter, wave
defined by Eq. (6.9) Regional averaged pressure in
Pi
k Parameter, defined as (1 v)/nE dilute region
k Wave number P2 Regional averaged pressure in
k Turbulent kinetic energy of gas dense region
k0 Turbulent kinetic energy of gas in Pc Pressure of gas in the core region
absence of the particles Pressure of phase j in region k
kv Turbulent kinetic energy of PM Pressure of the mixture
particles Pw Pressure of gas in the wall
L Loschmidt number region
I Interparticle distance Q Parameter, defined by Eq. (6.138)
Eulerian integral length scale
u
/w Wake length
q Heat absorbed by unit mass of
gas-particle mixture
Mc Parameter, defined by Eq. (6.139) Gas constant
Rep Particle Reynolds number, based Vd Dynamic wave velocity

on particle diameter and relative Vk Volume occupied by phase k
velocity Vp Particle velocity component
r Radial coordinate Vs Slip velocity
r* Ratio of reflection speed to W Gas velocity component
incoming speed Wc Axial gas velocity in the core
St Strouhal number region
T Absolute temperature of gas Wjk Axial velocity of phase j in region
T\ Fluid temperature in front of the k
shock wave Wk Axial velocity of gas phase in
Tp Absolute temperature of particle region k
Tv \ Particle temperature in front of the Wik Axial velocity of phase 1 in
shock wave region k
t
Time Wp Axial particle velocity
tc
Contact time duration Wpc Axial particle velocity in the core
U
Superficial velocity of gas region
U
Gas velocity component Wpk Axial particle velocity in region k
U\
Gas velocity component in front Wpw Axial particle velocity in the wall
of the shock wave region
Ui Gas velocity component Ww Axial gas velocity in the wall
immediately behind the shock region
wave w Attrited mass fraction of the
/oo Gas velocity component far sample particles
behind the shock wave von Brittle wear in volume loss per
Upoo Particle velocity component far impact
behind the shock wave ifD Ductile wear in volume loss per
U Velocity vector of fluid impact
4 Velocity of perturbation Wg Molecular weight of gas
Um Superficial velocity of the wM Molecular weight of a
gas-solid mixture gasparticle mixture
Up Superficial velocity of solids Wp Molecular weight of "particulate
Up Particle velocity component gas"
Up Velocity vector of particles X Parameter, defined by Eq. (6.50)
Upt Particle terminal velocity TB Yield stress for brittle removal
u Linear fluid velocity YD Yield stress for ductile removal
ui Fluid velocity in front of the vi Depth of interface
shock wave z Axial coordinate
wp Particle velocity Greek Symbols
V Incoming speed of particle before
a Volume fraction of gas phase
impact
a Empirical constant, defined by
V Gas velocity component
Eq.(6.17)
Vc Continuity wave velocity
Nomenclature 291
a' Voidage perturbation V Kinematic viscosity of gas

ac Volume fraction of gas phase in V Poisson's ratio
the core region p Density of fluid
a\ Impact angle Pjk Density of phase j in region k
Ojk Volume fraction of phase j in Pm Density of the mixture
region k Pp Density of particles
am Maximum eroding impact angle a Normal stress
Qfp Volume fraction of the particle Particle normal stress
phase
rci Interfacial stress of the gas phase
QfpC Volume fraction of particle phase in the core region projected on the
in the core region pipe axis
Qfpw Volume fraction of particle phase Te Eddy existence time
in the wall region TgW Wall shear stress of the gas phase
ofw Volume fraction of gas phase in in the dilute region projected on
the wall region the pipe axis
p Attenuation coefficient Particle-eddy interaction time
y Ratio of specific heats of gas tjki Interfacial stress of phase j in
YM Ratio of specific heats of region k projected on the pipe axis
gas-particle mixture tjkw Wall shear stress of phase j in
A d Total energy dissipation from the region k projected on the pipe axis
eddy during the particle-eddy ^mw Wall shear stress of the mixture in
interaction the dense region projected on the
A p Total energy production of gas by pipe axis
the particle wake or vortex
Tp Flying time
shedding
rPd Interfacial stress of the particle
AEt Total turbulence modulation
phase in the core region projected
during the particle-eddy
on the pipe axis
interaction
Wall shear stress of the particle
8 Ratio of specific heats of particle
phase projected on the pipe axis
to gas at constant pressure
^pwi Interfacial stress of the particle
e Shear strain on particle surface
phase in the wall region projected
6 Rate of energy dissipation on the pipe axis
6B Energy required to remove a unit ^s Stokes relaxation time
volume of material in brittle mode
rw Wall shear stress of the gas phase
6D Energy required to remove a unit projected on the pipe axis
volume of material in ductile mode
Interfacial stress of the gas phase
6k Sign parameter in a two-region in the wall region projected on the
model for region k pipe axis
f Ratio of density of particle to gas "7""jk Stress tensor of phase j in region k
r) Parameter, defined by Eq. (6.67) Tk Stress tensor of the gas phase in
Parameter, defined by Eq. (6.67) region k
/A Dynamic viscosity of gas 0 Mass fraction of particles
Fraction of frictional perimeter of co Angular frequency

phase j &>h Frequency of perturbation
References
Achenbach, E. (1974). Vortex Shedding from Spheres. /. Fluid Mech., 62, 209.
Briscoe, B. J. and Evans, P. D. (1987). The Wear of Polymers by Particle Flows. In Tribology in
Paniculate Technology. Ed. Briscoe and Adams. Philadelphia: Adam Hilger.
Carman, P. C. (1937). Fluid Flow through Granular Beds. Trans. Instn. Chem. Engrs., 15, 150.
Carrier, G. F. (1958). Shock Waves in a Dusty Gas. /. Fluid Mech., 4, 376.
Chang, S. S.-H. (1975). Nonequilibrium Phenomena in Dusty Supersonic Flow Past Blunt Bodies of
Revolution. Phys. Fluids, 18, 446.
de Crecy, F. (1986). Modeling of Stratified Two-Phase Flow in Pipes, Pumps, and Other Devices.
Int. J. Multiphase Flow, 12, 307.
Delhaye, J. M. (1981). Basic Equations for Two-Phase Flow Modeling. In Two-Phase Flow and
Heat Transfer in the Power and Process Industries. Ed. Bergles, Collier, Delhaye, Hewitt
and Mayinger. Washington: Hemisphere.
Finnie, I., Wolak, J. and Kabil, Y. (1967). Erosion of Metals by Solid Particles. /. Mater. Sci., 2, 682.
Goldsmith, W. and Lyman, P. T. (1960). The Penetration of Hard Steel Sphere into Plane Metal
Surface. Trans. ASME, J. Appl. Meek, 27, 717.
Gore, J. P. and Crowe, C. T. (1989). Effect of Particle Size on Modulating Turbulent Intensity. Int. J.
Multiphase Flow, 15, 279.
Gwyn, J. E. (1969). On the Particle Size Distribution Function and the Attrition of Cracking
Catalysts.A/C7*/.,15,35.
Hetsroni, G. (1989). Particle-Turbulence Interaction. Int. J. Multiphase Flow, 15, 735.
Hinze, J. O. (1972). Turbulent Fluid and Particle Interaction. Prog. Heat Mass Transfer, 6,433.
Hutchings, I. M. (1987). Surface Impact Damage. In Tribology in Paniculate Technology.
Ed. Briscoe and Adams. Philadelphia: Adam Hilger.
Lamb, H. (1932). Hydrodynamics, 6th ed. Cambridge: Cambridge University Press.
Lee, S. L. and Durst, F. (1982). On the Motion of Particles in Turbulent Duct Flows. Int. /.
McLaughlin, J. B. (1994). Numerical Computation of Particle-Turbulence Interaction. Int. J.
Multiphase Flow, 20/S, 211.
Morgenthaler, J. H. (1962). Analysis of Two-Phase Flow in Supersonic Exhausts. In Detonation and
Two-Phase Flow. Ed. Penner and Williams. New York: Academic Press.
Paramanathan, B. K. and Bridgwater, J. (1983). Attrition of Solids-II. Chem. Eng. Sci., 38, 207.
Peddieson, J. (1975). Gas-Particle Flow Past Bodies with Attached Shock Waves. AIAA J., 13, 939.
Rietema, K. (1991). The Dynamics of Fine Powders. London: Elsevier Applied Science.
Rudinger, G. (1965). Some Effects of Finite Particle Volume on Dynamics of Gas-Particle Mixture.
AIAA J., 3, 1217.
Rudinger, G. (1969). Relaxation in Gas-Particle Flow. In Nonequilibrium Flows. Ed. P. P. Wegener.
New York: Marcel Dekker.
Rudinger, G. (1980). Fundamentals of Gas-Particle Flow. New York: Elsevier Scientific.
Rudinger, G. and Chang, A. (1964). Analysis of Nonsteady Two-Phase Flow. Phys. Fluids, 7, 1747.
Soo, S. L. (1960). Effect of Transport Process on Attenuation and Dispersion in Aerosols. /.
Acoustical Society of America, 32, 943.
Problems 293
Soo, S. L. (1977). A Note on Erosion by Moving Dust Particles. Powder Tech., 17, 259.
Soo, S. L. (1990). Multiphase Fluid Dynamics. Beijing: Science Press; Brookfield, USA: Gower
Technical.
Squires, K. D. and Eaton, J. K. (1990). Particle Response and Turbulence Modification in Isotropic
Turbulence. Phys. Fluids, A2, 1191.
Tabor, D. (1951). The Hardness of Metals. Oxford: Oxford University Press.
Timoshenko, S. and Goodier, J. E. (1951). The Theory of Elasticity, 2nd ed. New York:
McGraw-Hill.
Tsuji, Y. and Morikawa, Y. (1982). LDV Measurements of an Air-Solid Two-Phase Flow in a
Horizontal Pipe. /. Fluid Mech., 120, 385.
Tsuji, Y, Morikawa, Y and Shiomi, H. (1984). LDV Measurements of an Air-Solid Two-Phase
Flow in a Vertical Pipe. /. Fluid Mech., 139, 385.
Walk's, G. B. (1969). One-Dimensional Two-Phase Flow. New York: McGraw-Hill.
Yuan, Z. and Michaelides, E. E. (1992). Turbulence Modulation in Particulate Flows a Theoretical
Approach. Int. J. Multiphase Flow, 18, 779.
Zhou, L. X. and Huang, X. Q. (1990). Predictions of Confined Turbulent Gas-Particle Jets by an
Energy Equation Model of Particle Turbulence. Science in China, 33, 52.
Zhu, C. (1991). Dynamic Behavior of Unsteady Turbulent Motion in Pipe Flows of Dense
Gas-Solid Suspensions. Ph.D. Dissertation. University of Illinois at Urbana-Champaign.
Zhu, C , Soo, S. L. and Fan, L.-S. (1996). Wave Motions in Stratified Gas-Solid Pipe Flows.
Proceedings ofASME Fluids Engineering Division Summer Meeting, San Diego,
California. 1, 565. ASME.
Problems
6.1 Consider the surface wear by random motion of particles in a fluidized bed. The random
motion is assumed to have the intensity of (w'2). Determine the relationship between ED
and KD for overall ductile wear and the relationship between EB and KB for overall brittle
wear.
6.2 For a general surface erosion, the wear is composed of both ductile and brittle modes.
Thus, the combined directional wear can be expressed as the weighted summation of both
modes. Consider a case where the weighted factor for ductile wear is (ai) = A sinc^ and
the weighted factor for brittle wear is rj{a\) = B cosc^. Both A and B are constant. What
is the extent of the total erosion? Derive an expression for the maximum eroding angle in
terms of KD, KB, A, and B.
6.3 Consider a closed system of a dust-laden air with an initial pressure of 1 atm. The dust
particles are 10 /xm in diameter and the particle density is 2,000 kg/m 3. The relative ratio
of the specific heat of particle to gas is 50. The mass loading ratio of particle to gas is 10.
Assuming the system undergoes an isentropic compression to a final state with a pressure of
20 atm, calculate the volume fraction of solids and estimate the averaged distance between
particles in the final state.
6.4 Show that, from Eq. (6.74), the asymptotic value of the attenuation parameter for the acoustic
wave propagation through a dilute gas-solid suspension can be expressed as
for small Nw, and
for large N^. Assume that p p /p ^> 1 and m p (p/p p ) <$C 1.

6.5 Consider an acoustic wave with a frequency of 1,000 Hz passing through dust-laden air
at ambient conditions. Assume that mv = 0.5, dp 5 /xm, p p/p = 1,500, 5 = 10,
y = 1.5 x 10~5 m2/s, and y = 1.4. (a) Estimate the speed of sound using Eq. (6.72).
Is the result close to the "frozen speed of sound" or the "equilibrium speed of sound?"
(b) Estimate the speed of sound using Eq. (6.77). Compare the results between (a) and (b).
6.6 Considering a batch-solid fluidized bed, show that Eq. (6.180) can be derived from
Eqs. (6.169), (6.170), and (6.174) through (6.179).
6.7 Assume that the bed expansion characteristics in gas-solid particulate fluidization can be
described by an equation of the Richardson-Zaki form as given by
%- = otn (P6.3)
t/pt
where n is the modified Richardson-Zaki index for gas-solid systems. Express the elastic
modulus E in terms of the gas and particle properties and the bed voidage for the minimum
bubbling point.
6.8 For plastic beads with p p = 500 kg/m3 falling in air at ambient conditions, estimate the
range of variation of the shedding frequency of the particle wake when the particle Reynolds
number, based on the particle terminal velocity, varies from 500 to 1,000. For this particle
Reynolds number range, what is the corresponding range of variation for the particle sizes?
PART II
System Characteristics
CHAPTER 7
Gas-Solid Separation
7.1 Introduction
Separation processes are central to gas-solid flow systems concerning dust re-
moval, particulate collection, sampling, particle recirculation, and other operations. Gas-
solid separation can be achieved by application of the principles involving centrifugation,
electrostatic effects, filtration, gravitational settling, and wet scrubbing. Gas-solid separa-
tors applying these principles employ rotary flow dust separators, electrostatic precipitators,
filters, settling chambers, and scrubbers. In order to yield highly efficient solids collection
or removal processes, multistage gas-solid separators based on a combination of several of
these components are also commonly employed.
In this chapter, the collection mechanisms, types, and collection efficiencies of gas-solid
separators are discussed. Specifically, separation by rotatingflow,exemplified by a cyclone
separator, is presented. Tangential flow cyclones, which are the most commonly used cy-
clones, are described. The principles of electrostatic precipitation are illustrated. Comments
are presented with respect to the difficulties involved in obtaining an accurate estimation
of the collection efficiency of an electrostatic precipitator. Factors contributing to these
difficulties include the geometric complexity of the system, significant flow disturbance
due to electric wind, and poor predictability of particle charges. Concepts on filtration
are also introduced. A filtration process can collect particles of almost all sizes; however,
the pressure drop across the filter may vary significantly with the internal structure of the
filter, particle deposition mode (cake or in-depth deposition), and quantity of accumulated
particles. Wet scrubbing, which produces wet sludge from the impaction of solids with
liquid droplets, is also delineated in this chapter.
7.2 Separation by Rotating Flow

Separation based on rotatingflowprinciples is one of the most common operations
involved in gas-solid flows. This section describes the fundamental rotating flow principles
and their applications to cyclone operation. The efficiency of dust collection in cyclones is
also described.
7.2.1 Mechanism and Type of Rotary Flow Dust Separators

In a rotary flow dust separator, the rotating flow of the gas-solid suspension can
be initiated as a result of the design of the separator, such as the tangential inlet of the
separator, the guide vanes at the inlet, and relative rotating cylinders. As a result of the
rotating flow, the particles are subjected to a centrifugal force which is usually at least two
orders of magnitude greater than the gravitational force. Thus, even light particles can be
easily directed toward the wall by the centrifugal force and collected in the dust bunker
while the clean gas escapes through the gas outlet.
297
298 7 / Gas-Solid Separation
Exit
pipe
Inlet of
gas-solid
flow
Dust
bunker
llected
particles
Figure 7.1. Schematic diagram of the tangential inlet cyclone.
Separating devices using this mechanism consist of reverse-flow cyclones and uniflow
cyclones. The reverse-flow cyclones can be represented by the tangential inlet cyclone
and the tornado dust collector, as shown in Figs. 7.1 and 7.2, respectively. The suspension
flow is fed into the cyclone tangentially so that the flow begins rotating by following the
inner contour of the cyclone. The particles accumulated on the wall slide down along the
cone into the dust bunker. Typical uniflow cyclones include the axial flow cyclones and
the rotating cylinder separators. In the axial flow cyclone, as shown in Fig. 7.3, a rotating
flow is created by the guide vanes. The solid particles are forced toward the outer wall by
the centrifugal force and are separated from the gas in the coaxial outlets. In the rotating
cylinder separators represented by the rotex centrifugal separators, as shown in Fig. 7.4,
the dust-laden gas flows through the rotating cylinders and starts rotating by the effect of
wall friction at the cylinders. The separation of particles is achieved by the relative inertia
between gas and particles and by the effect of gravity.
The cyclone dust collector is one of the simplest dust collectors: no moving parts and
ease in maintenance. In spite of the simple structure, the centrifugal force on the solids in
the cyclone dust collector can easily reach 300 to 2,000 times the gravitational force and a
7.2 I Separation by Rotating Flow 299
Clean
Recirculating system for

the secondary gas flow
Tornado
Primary
cyclone
Gas-solid
flow
Figure 7.2. Schematic diagram of the tornado dust collector (after Ogawa, 1984).
Locus of
Guide particle
Dust Clean
\ / vanes / gas
Figure 7.3. Schematic diagram of the axialflowcyclone (from Ogawa, 1984).

Dust-laden Clean
Figure 7.4. Schematic diagram of the Rotex centrifugal separator (from Ogawa, 1984).
high separation efficiency can be achieved. Although the separation efficiency of a cyclone
with guide vanes may be low in comparison with that of the tangential inlet cyclone as a
result of the slow rotating gas flow caused by the guide vanes, the coaxial structure of the
outlets for solids and gas flow in the cyclone with guide vanes prevent the reentrainment
of the solids which may occur in a tangential inlet cyclone. The cyclone can be operated
for high-temperature gas. The advantages of cyclones also include low operating costs,
reliability, and suitability for high-temperature operations.
7.2.2 Flow Field in a Cyclone

Visualization and measurements of velocity distributions reveal that the tangential
velocity field in a cyclone basically consists of a core of solid body rotation in the central
region and a free vortex flow in the outer region [Ogawa, 1984; Rietema, 1962]. Various
vortex models based on this experimental finding have been developed. One of the simplest
vortex models is Rankine's combined vortex flow model, which considers the flow field
comprising the forced vortex region and the free vortex region, as shown in Fig. 7.5. To
describe this model, we denote Uy V, and W as the gas velocity components with respect
to the cylindrical coordinates r, 0, and z, respectively. The velocity in the forced vortex
region (0 < r < rf) is expressed by
V = cor (7.1)
and that in the free vortex region (r > rf) is given by
Vr = r = corf (7.2)
where rf is the radius of the free vortex, co is the angular velocity, and T represents the
72 I Separation by Rotating Flow 301
P-Po
Static pressure
Forced vortex
Free vortex
Figure 7.5. Tangential velocity and pressure distributions based on Rankine's combined
vortex model.
circulation. The corresponding static pressure distribution can also be obtained as
P= (7.3)
for the forced vortex region and
P = P0+-Z-[-2--t (7.4)
for the free vortex region, where po is the static pressure at the center of the cyclone and p
is the gas density.
Another commonly used model is based on the general differential balance of mass
and momentum [Burgers, 1948]. Consider a steady, incompressible, and axially symmetric
flow in which the body forces are negligible. In cylindrical coordinates, the equation of
continuity of the fluid can be given as
du dw u
1 H = o (7.5)
dr dz r
and the Navier-Stokes equations can be expressed by
dU dU V2 1 d2u d 2u 1 du u

p dr2 dz 2
' r dr
2 2
dv dv uv fd v dv i ay y
(7.6)
dr dz r ~o z^ ~r~dr~~V2
T Jr
d2 w ldW\
l7J
where v is the kinematic viscosity of the fluid. A typical solution is obtained if U, V, W,

and p are in the forms
U = -Ar, V = V(r), W = 2Az
/2
p pA\r + 4z)
/ dr
where A is a constant. It can be easily verified that these equations satisfy the equation of
continuity and the equations of motion in the axial and radial directions. Substitution of
the preceding expressions into the equation of motion in the tangential direction yields
dV fd2V ldV V\
-Ar - AV = v( - 2+ - - (7.8)
dr \ dr r dr rl )
which has a solution
M)]
where C is a constant representing the maximum circulation in the system. The circulation
is thus given by
+ exp(f) (7.10,
dr r 2nv \ 2v )
Now, define the dimensionless velocities and coordinate as
v (7J1)
"- '-h '"='-R
where Uw and Vw are the radial and tangential gas velocities near the cyclone wall, re-
spectively (just outside the wall surface boundary layer), and R is the radius of the cross
section of the cyclone. Note that both Vw and R are functions of the axial coordinate z only.
Neglecting the thickness of the wall boundary layer, the dimensionless velocity profiles are
obtained from Eqs. (7.7) and (7.9) as
{7 U)
r*{ l-exp(-Rew/2) ) '
where Rew is the Reynolds number, defined as
Rew = -Z (7.13)
V
In terms of Eq. (7.12), the variations of the radial profiles of the dimensionless tangential
velocity V* with Rew are plotted as shown in Fig. 7.6. It is seen that the dimensionless
tangential velocities reach a maximum at r* < 0.5 for values of Rew varying from 10 to 30.
In an actual operation, the gas-solid flow in a cyclone separator is turbulent. The tur-
bulence effects on the distributions of the gas velocity and static pressure in the preceding
two models are, however, excluded. The strongly swirling turbulent gas-solid flows in tan-
gential inlet cyclones were accounted for numerically by Zhou and Soo (1990) using the
k-e turbulence model. Their typical results along with laser Doppler velocimetry (LDV)
measurements for the radial variations of the axial velocity, tangential velocity, and pres-
sure in the cyclone are shown in Fig. 7.7. It is seen that there are an upward flow region
Figure 7.6. Vortex flow distributions (from Burgers, 1948).
in the near-axial core region and a downward flow region in the wall region. Also, the
pressure increases radially in the cyclone. The k-e model is seen to predict the velocity and
pressure in the wall region well, but not so precisely in the near-axial core region because
of the inability of the model to simulate the nonisotropic turbulence of strongly swirling
flow.
For the engineering design of tangential inlet cyclones, the variation of the tangential
velocity in the radial direction can be correlated in a form
= const (7.14)
where the vortex exponentftis related to the absolute gas temperature T and the cyclone
radius R by [Alexander, 1949]
0.3
? = l-(l-0J4R^-) (7.15)
For most tangential inlet cyclones,ftvaries from 0.5 to 0.7.
7.2.3 Collection Efficiency of Cyclones

The collection efficiency usually depends on the shape of the cyclone, size and
density of the particles, and incoming tangential velocity. The collection efficiency of a
10
0L )
^ ^
10
0
r
(c)
Figure 7.7. Calculated and experimental results for radial variations of the axial velocity,
tangential velocity, and pressure in a cyclone (from Zhou and Soo, 1990): (a) Axial velocity,
m/s; (b) Tangential velocity, m/s; (c) Pressure distributions, Pa.
tangential inlet cyclone with the geometric configuration given in Fig. 7.8 can be estimated
by using a simple model suggested by Leith and Licht (1972). In this model, the following
characteristics are assumed:
(1) Particles are spherical.
(2) The radial velocity of the gas is zero.
(3) The radial velocity of the particles is nearly constant.
(4) The drag force follows Stokes's law.
(5) There is no slip in the tangential direction between the particles and the gas.
(6) The tangential velocity can be expressed by Eq. (7.14).
(7) The effect of the electrostatic charge is negligible.
(8) The concentration of uncollected particles is uniformly distributed as a result of
turbulent mixing.
Figure 7.8. Geometric configuration of the tangential inlet cyclone.
Figure 7.9. Horizontal cross section of a cyclone.
Consider a horizontal cross section of a cyclone, as shown in Fig. 7.9. During a time
interval dt, the number of particles collected, dN, within a differential volume dr (r dO) dz
is expressed by
2
-dN = [R - (R - dr)2]n dz (7.16)
where n is the particle number density. The total number of particles in the sector from
which particles are removed is
N = R 2ndz (7.17)
Therefore, the fraction of particles collected within dt is obtained as
The relationship between dr and dt is determined from the equation for the radial motion
of the particle, which is obtained by the force balance between the drag force and the
centrifugal force, i.e.,
^=V- (7.19)
TS dt r
where ts is the Stokes relaxation time defined by Eq. (3.39). Substituting Eq. (7.14) into
Eq. (7.19) yields
which gives the traveling time t for particles moving from r 0 at t = 0 to r as
(7.2D
where V\ is the inlet velocity.

Assume that the initial radial position of the uncollected particles is at the vortex center
where r0 = 0. Substitution of Eqs. (7.20) and (7.21) into Eq. (7.18) yields
Thus, the collection efficiency, r]c, can be obtained by integrating the preceding equation
from t = 0 to the mean residence time, tm, as
(7.23)
where No is the number of particles at t = 0, and Nm is the remaining uncollected particles

at time tm.
The mean residence time tm of gas can be related to the geometric configuration of the
cyclone and the inlet flow rate. It is reasonable to take tm as the average of the minimum
residence time tm\ and the maximum residence time tm2, where tm\ is the average time
required for the gas to descend from the midlevel of the entrance to the exit at the bottom
of the cyclone. Thus, we have
VA
where A\ is the cross-sectional area of inlet and s, a, R, and r e are dimensions of the cyclone,
defined in Fig. 7.8.
In order to estimate tm2, we need to identify the lowest point to which the gas descends.
This point does not necessarily correspond to the bottom of the cyclone as characterized by
the depth H. The deepest descending length / is given by an empirical equation [Alexander,
1949] as
() 3 (7.25)
which is independent of the inlet flow rate. Hence, we have
where
H)[]
and h, H, and rb are dimensions of the cyclone defined in Fig. 7.8.
A cyclone should be designed in such a way that its physical length (H s) is near /.
If this physical length is longer than /, the space below the vortex turning point is wasted.
If the physical length is shorter than /, the full separation potential of the cyclone is not
realized. For a properly sized cyclone, tm2 is expressed by
nR2(h -s) + ^R2(H - h) (l + ^ + r-\ - nrl{H - s)] (7.28)
Therefore, tm is given as
tm = ~(tml + tm2) (7.29)
where tm\ is given by Eq. (7.24) and tm2 is given by either Eq. (7.26) or Eq. (7.28). From
Eqs. (7.29) and (7.23), the collection efficiency of a cyclone can be calculated.
It is noted that the particle-wall interaction in the boundary layer and the electrostatic
effect due to the electrostatic charge carried by the particles may strongly affect particle
collection and reentrainment in the cyclone and, consequently, affect the collection effi-
ciency. In the presence of electrostatic charges of particles and an external electric field, the
collection efficiency of a tangential inlet cyclone with a steep cone is given by [Soo, 1989]
Q
where pp is the density of particles; a is the sticking probability of the particles to the wall
or the layer of collected particles; F m is the maximum circulation in the system; Q is the
volumetric flow rate of gas; 60 is the permittivity of the vacuum; q/m is the charge-to-mass
ratio of particles; and E is the electric field strength. The first term in the bracket on the
right-hand side of Eq. (7.30) represents the collection by centrifugal force exerted by the
vortex; the second term is the collection by space charge of the particles; and the last term
accounts for the collection by the applied electric field. It can be shown that the sticking
probability a depends not only on the contact surfaces and deposit forces but also on the
particle turbulent diffusivity and turbulent viscosity [Zhou and Soo, 1991].
The cut-off size of a cyclone, which is an important performance index of the cyclone, is
defined as the size of particles below which the collection efficiency is less than 50 percent.
In terms of Eq. (7.23), the cut-off size of a cyclone, dpc, in the absence of electrostatic
charges in a gas-solid flow can be estimated by
In the presence of electrostatic charges of particles and an external electric field, the cut-off
size of a tangential inlet cyclone with a steep cone is given from Eq. (7.30) as
(7.32>
where \i is the viscosity of the gas.
Example 7.1 In a test of the performance of a cyclone, a dust-laden gas with the particle
mass flow rate of 600 kg/h is used. Given that the dust escape rate is 30 kg/h, what is the
overall collection efficiency? Plot the fractional efficiency curve to identify the cut-off size
of this cyclone. The results of particle size analysis (by weight percentage [wt%]) at the
inlet and outlet of the cyclone are given in Table E7.1.
Table E7.1. Results of the Particle Size Analysis
at the Inlet and Outlet of the Cyclone
Particle size analysis (wt%)

Size range
Qim) Inlet Outlet
0-5 1.5 28.5
5-10 1.7 20.0
10-15 2.2 13.0
15-20 3.4 10.0
20-30 9.5 14.7
30-40 15.0 8.5
40-50 19.0 3.3
50-60 22.0 1.3
>60 25.7 0.7
Solution The overall collection efficiency is
mass rate of collected particles 600 30

rjc = - = = 95 percent (E7.1)
mass rate of inlet particles 600
The fractional collection efficiency at each size interval, ryCi, is calculated by using a similar
formula, and the results are given in Table E7.2.
In terms of the averaged particle diameters and the corresponding fractional collection
efficiency, the fractional efficiency curve is obtained as shown in Fig. E7.1, from which the
cut-off size of this cyclone is estimated as 9 /xm.
7.3 I Electrostatic Precipitation 309
Table E7.2. Fractional Collection Efficiency for Dust Collection in Each Particle Size Range
Outlet
dp range ave. of ^ Inlet Inlet mass Outlet mass ??ci
(Aim) (/xm) (wt%) (kg/h) (wt%) (kg/h) (%)
0-5 2.5 1.5 9.0 28.5 8.55 5.0

5-10 7.5 1.7 10.2 20.0 6.0 41.2
10-15 12.5 2.2 13.2 13.0 3.9 70.5
15-20 17.5 3.4 20.4 10.0 3.0 85.3
20-30 25 9.5 57.0 14.7 4.4 92.3
30-40 35 15.0 90.0 8.5 2.6 97.1
40-50 45 19.0 114.0 3.3 1.0 99.1
50-60 55 22.0 132.0 1.3 0.4 99.7
>60 60 25.7 154.2 0.7 0.2 99.9
100
80
60
50
40
20
20 40 60
dp9\im
Figure E7.1. Fractional collection efficiency for a cyclone.
7.3 Electrostatic Precipitation

An electrostatic precipitator is a gas-solid separator in which particles suspended
in a gas stream are charged and removed by electrostatic force. Since the separation force
is directly applied to the particles without the need of accelerating the gas phase, an elec-
trostatic precipitator usually requires much less hydraulic power than other separation sys-
tems. Hence, electrostatic precipitation is widely recognized as an important technique of
gas-solid separation. This technique is characterized by low pressure drop, relative high
collection efficiency for fine particles, and high adaptability to a wide variety of effluents,
either wet, dry, or corrosive.
7.3.1 Mechanism of an Electrostatic Precipitator

The complete process of particle removal by electrostatic precipitation commonly
includes three steps: (1) particle charging; (2) particle migration in an external electric
field; and (3) particle discharging and collection. This removal process is illustrated in
Fig. 7.10.
The electrostatic charges of particles in a gas stream can be acquired naturally through
triboelectric charging via mutual impacts or by collisions with other solid materials such
as pipe walls and valves (see 3.5.2). However, to reach a high collection efficiency in an
electrostatic precipitator, the particles are typically artificially charged to a saturated degree
by passing through a corona discharge area where the gas is highly ionized. The corona
can be either positive or negative, but the negative corona is much more stable and is easier
to control than the positive corona. Consequently, the negative corona is more frequently
used in industrial electrostatic precipitators. When particles are exposed to unipolar ions
in an electric field, impacts of ions and particles take place as a result of two basic mech-
anisms: field charging and diffusion charging. In a field-charging process, the motions
of ions are induced by the applied electric field. The particles are charged as a result of
the ion bombardments. This charging mechanism predominates when particles are larger
than 1 /zm in diameter. In a diffusion-charging process, the particles are charged by the
Brownian motion of ions. This mechanism occurs even in the absence of external elec-
tric fields and predominates when particles are smaller than 0.2 /xm in diameter. In most
cases for particles inside electrostatic precipitators, the saturation charge is attained within
a second.
The charged particles migrate toward the collecting electrodes in the applied electric
field by the Coulomb force. Upon contact with the collecting electrodes, the particles lose
their charges and thus can be removed either by adhesion to the collecting surface or by
Dust-laden
Collecting Discharge
electrode electrode
Particle
Gas ion
or ^
electron
Negatively
charged
particle
(a) (c)
Figure 7.10. Mechanism of fine solids collection in electrostatic precipitation (after Ogawa,
1984): (a) Particle charging; (b) Particle migration; (c) Particle discharging and collection.
DC high-voltage
cable
Discharge
electrode
Collecting
/êlectrode
Collecting
plate
Weight
Heavy
plumment
(weight)
(b)
Figure 7.11. Basic types of electrodes (after Ogawa, 1984): (a) Cylinder-type precipitator;
(b) Plate-type precipitator.
sliding into the discharge hoppers. Thus, the separation of particles from their carrying gas
is achieved. Typical corona dischargers in electrostatic precipitators are either thin metal
wires or metal needles, and typical collecting electrodes are plates or cylinders. Electrostatic
precipitators may thus be divided into two types, cylinder-type and plate-type, as shown in
Fig. 7.11. The cylinder-type electrostatic precipitators, sometimes used as wet precipitators,
may achieve a collection efficiency as high as 99.9 percent. The plate-type precipitators,
commonly used for dry particle-laden gas with large flow rates, can reach a collection
efficiency of 95 percent [Ogawa, 1984].
Similarly to other separation methods, the electrostatic precipitation method has its
own shortcomings. For certain types of particles, electrostatic charging is difficult without
additional chemical conditioning. To handle a high flow rate of particulate-laden gas, the
size of the precipitator, i.e., total area of collecting electrodes, has to be large enough to
ensure a high collection efficiency. The method requires a regulated high-voltage power
supply, which is usually above 10 kV, and cleaning of the collecting plates is generally
difficult. Consequently, the equipment and operating costs of this separator are relatively
high.
7.3.2 Migration Velocity and Electric Wind

To estimate the particle migration velocity, it is assumed that (1) particles are
spherical and have the same size; (2) all particles are charged to the same extent; (3) the
particle motion is governed by the Coulomb force and the Stokes drag only; and (4) the
direction of the applied electric field is perpendicular to the direction of the suspension flow.
Voltage = 15 kV
# = 2 0 mm
-30 -20 -10 0 10 20 30

r, mm
(a) (b)
Figure 7.12. Experimental system and results describing electric wind (from Kercher, 1969):
(a) Test system; (b) Velocity distributions of electric wind.
The equation for the particle motion in the direction of the electric field is given by
m- = qE 3jtjjidpUv (7.33)
at
where m is the particle mass and q is the electric charge carried by the particle. The particle
velocity, Up, is obtained by integrating Eq. (7.33) as
V
1 exp( (7.34)
37TfJidp
Define the migration velocity Um as the terminal velocity of particles in the electric field.
qE
(7.35)
The value of Um measured in a practical situation is higher than that calculated from the pre-
ceding equation because of the existence of the electric wind from the discharge electrode
to the collecting electrode. The electric wind is caused by the motion of ions induced by the
strong corona discharge field. The intensity of the electric wind is related to both the proper-
ties of the gas and the intensity of the electric field. As illustrated in Fig. 7.12, thefluidveloc-
ity of electric wind can be on the order of 10 m/s [Kercher, 1969]. Hence, the motion of par-
ticle migration in the electrostatic precipitator can be strongly affected by the electric wind.
It can be shown that for fine particles {dv < 10/xm) in an electrostatic precipitator the
migration velocity of the particle is usually much higher than the corresponding particle
terminal velocity in the gravitational field. For this reason, fine particles are commonly
separated by electrostatic precipitation.
7.3.3 Collection Efficiency of Electrostatic Precipitators

The collection efficiency of the wire-plate-type of electrostatic precipitator can
be estimated by the widely used Deutsch theory [Deutsch, 1922]. To obtain the Deutsch
equation, the following assumptions are made: (1) because of the turbulent mixing ef-
fect, particle concentration at any cross section of the precipitator is uniformly distributed;
(2) gas velocity is constant everywhere except in the boundary layer near the walls; (3) U m
is constant and is relatively small compared to the gas velocity; (4) no particle reentrainment
occurs inside the precipitator; and (5) no disturbing electrical effect on the gas flow occurs.
Consider a one-dimensional flow of charged particles passing through the gap between
two parallel rectangular plates acting as collecting electrodes. The plates are L in length,
H in height, with a distance between the plates of / w. The mass of particles collected, dM,
over a distance dx is expressed by
dM = 2H dx avpvUm dt = 2Havpv ^ dx2 (7.36)
where av is the particle volume fraction. On the other hand, from the mass balance of
particles, we have
dM = - d a p Pp/w// dx (7.37)
Combining Eq. (7.36) with Eq. (7.37) yields
**~2 (7.38)
which leads to
fî) ,7.39)
where aPi and apo are the particle volume fractions at the inlet and the outlet, respectively.
Therefore, the collection efficiency of an electrostatic precipitator, r)e, is obtained as
<*pi-a po / UmA\
ric = = 1 - expl (7.40)
where A is the area of the electrode. Equation (7.40) is called the Deutsch equation. It indi-
cates that the collection efficiency of an electrostatic precipitator increases with increasing
area of the electrodes and the migration velocity, and decreasing flow rate of gas. The actual
collection efficiency of an electrostatic precipitator is also greatly affected by the adhesion
of particles on the collecting surface. This adhesion may dramatically change the original
electric field by sharply increasing the resistivity of the electrode; severe "coating" and
high resistivity of particles may even reverse the ionization or reverse the corona discharge
[White, 1963; Svarovsky, 1981; Ogawa, 1984].
Example 7.2 Consider a dust-laden gas passing through an electrostatic precipitator.

Assume that all particles have the same charge-to-mass ratio, q/m, and the electric field
is uniform. Use the Deutsch equation to estimate the overall collection efficiency if the
particle size distribution based on weight percentage (wt%) can be expressed by the mass
density function, /M, as
0 < dp < dpm - 8
/M = (E7.2)
dpm + 8 < dp < oo

where dvm and 8 are the density function parameters.
314 71 Gas-Solid Separation
Solution From Eq. (7.35), the migration velocity can be expressed as
(E7-3>
Therefore, the fractional collection efficiency is given by Eq. (7.40) as
Given the mass density function of particles, the overall collection efficiency of an electro-
static precipitator can be expressed as
>7e= r furjeiddp (E7.5)

Jo
To simplify the derivation, let
The overall collection efficiency of this electrostatic precipitator is

l
" 3 2
3-1 4
3 r?+
[(i _ /j2) + 2fix - x 2 ]exp(- 2 x 2 ) i c
- I) 2 ) - * | ^ e x p ( - $ 2 ( / + I) 2 ) (E7.7)
7.4 Filtration
In filtration, solid particles are separated from the gas stream when the gas-solid
suspension passes through a single-layer screen, a multiple-layer screen, or a porous, per-
meable medium which retains the solid particles. This separation method is simple and has
a high efficiency in collecting particles of all sizes. However, in this type of separation
method, a high pressure drop may develop as a result of the accumulation of collected
particles in the filter, which may yield high energy consumption and require high material
strength of filters for the operation.
7.4.1 Mechanisms of Filtration and Types of Filters

On the basis of particle collection characteristics, filtration can be classified into
cake filtration and depth filtration. In cake filtration, particles are deposited on the front sur-
face of the collectingfilter,as shown in Fig. 7.13 (a). Filtration of this type is achieved mainly
as an effect of screening. This is by far the most common type employed in the chemical
and process industries. In depth filtration, particles flow through the filter and are collected
7.4 I Filtration 315
Filter Fiber of
cake
v / filter
Gas-solid W^ /r\ Gas-solid },

flow SSL V flow
-O o
'
O'oa " O
(a) (b)
Figure 7.13. Basic types of filtration: (a) Cakefiltration;(b) Depth filtration.
during their passage, as illustrated in Fig. 7.13(b). Examples of filters in cake filtration in-
clude fabric filters, sieves, nuclepore filters, and membrane filters. Filters for depth filtration
include packed beds, meshes, fluidized beds, and glass-fiber filters. In both cake filtration
and depth filtration, continuous particle collection increases the flow resistance through the
filter, and, thus, the collection efficiency decreases. Efficient operation of the filter requires
periodic removal of the particles deposited on the filter or replacement of the filter.
Filtration is a physical separation whereby particles are removed from the fluid and
retained by the filters. Three basic collection mechanisms involving fibers are inertial im-
paction, interception, and diffusion. In collection by inertial impaction, the particles with
large inertia deviate from the gas streamlines around the fiber collector and collide with the
fiber collector. In collection by interception, the particles with small inertia nearly follow
the streamline around the fiber collector and are partially or completely immersed in the
boundary layer region. Subsequently, the particle velocity decreases and the particles graze
the barrier and stop on the surface of the collector. Collection by diffusion is very important
for fine particles. In this collection mechanism, particles with a zig-zag Brownian motion in
the immediate vicinity of the collector are collected on the surface of the collector. The effi-
ciency of collection by diffusion increases with decreasing size of particles and suspension
flow rate. There are also several other collection mechanisms such as gravitational sedimen-
tation, induced electrostatic precipitation, and van der Waals deposition; their contributions
in filtration may also be important in some processes.
Filters can be grouped on the basis of different characteristics. For example, they can be
grouped on the basis of the design capacity of the filters according to the particle concen-
tration or loading [Svarovsky, 1981]; or they can be grouped on the basis of the materials
of the filters, e.g., fabric or nonfabric [Cooper and Freeman, 1982]. Most filters in use are
bag filters, which are fabric. The common fabric materials include cotton, polyester, wool,
asbestos, glass, acrylic, polytetrafluoroethylene (Teflon), poly (m-phenylene isophthalate)
(Nomex), polycaprolactam (Nylon), and polypropylene.
7.4.2 Pressure Drop in a Filter

The flow pattern in a filter is complex as a result of the intricate structure of the
fiber, which is further complicated by the particle deposition. Consequently, the pressure
Figure 7.14. Mapping of flow area around a single fiber.
drop across a dust filter is difficult to predict accurately. One approach to quantify the
pressure drop is to treat the flow as comprising several simple flows around the fibers.
Assume that the fibers in a filter are cylindrical; they are parallel to each other and
are uniformly assembled. Consider cake filtration in which particles are collected with
the deposited particles forming a layer of porous structure as shown in Fig. 7.13(a). Thus,
to account for the total pressure drop, three basic flow modes are pertinent: (1) flow is
parallel to the axis of fibers; (2) flow is perpendicular to the axis of the cylinder; and
(3) flow passes through a layer of a homogeneous porous medium. In the analysis of the
first two modes given later, Happel's model [Happel, 1959] is used, while for the third
mode, Ergun's approach [Ergun, 1952] is used.
First, consider the case where the flow is parallel to the cylinders. It is assumed that the
fluid is moving through the annular space between the cylinder of radius a and the fluid
envelope of equivalent radius b, as shown in Fig. 7.14. Assume that the fluid motion is
in the creeping flow regime so that inertia terms can be omitted from the Navier-Stokes
equations. Thus, in cylindrical coordinates, we have
1 d / dW\ _ 1 dp
(7.41)
r dr \ dr ) /x dz
with the boundary conditions
W\r=a=0; (7.42)
V^ Jr=b
The axial fluid velocity inside the annular space is obtained as
(7.43)
4/x dz \ a,
The fluid flow rate through the entire annulus is given by
Q = _JL^R (7.44)
8/x dz
Denote off as the volume fraction of fibers and A as the total cross-sectional area. In this
case, off = a2/b2 and A = nb2. From Eq. (7.44), the pressure drop over the filter thickness
L is obtained as
Ap QIJL 4off
(7.45)
L ~ Aa2 (2af - a2/2 - 3/2 - lna f )
Now, consider the case where the flow is perpendicular to the cylinders. The cylinder
may be treated as being stationary as a result of the force balance between the pressure
gradient and the drag force. Thus, the pressure drop may be obtained from the drag force
on a single cylinder. Alternatively, we can consider the case where a cylinder of radius a
moves perpendicular to its axis in a fluid cell of radius b. It is assumed that no shear stress
exists on the outside cylindrical fluid surface. The drag force in this case is noted to be the
same as that for a fluid moving perpendicular to a stationary cylinder of an identical radius.
In cylindrical coordinates, the continuity equation can be given as
dU U 13V
+ - + -=0 (7.46)
dr r r 80
where U and V are the radial and tangential components of the fluid velocity, respectively.
The momentum equation is given by
(7.47)
It is convenient to introduce the stream function, i/r, defined by

v
u = --; = ~ir ( 7 - 48 )
r 80 dr
Equation (7.47) reduces to
vV = 0 (7.49)
which has a general solution of the form
+ l-C2r (\nr - ^j +C3r + y \ (7.50)
where the C's are constants to be determined by the boundary conditions. The boundary
conditions in this case are
U\r=a = UocosO; V\r==a = -UosinO
dV idU V\ (7.51)
+- _ - _ =0
8r r 80 rjr=b
where Uo is the velocity of the moving cylinder. Therefore, the C's are determined as
(7.52)
(7.53)
-a22hbAJ]
-n 4
Un 0 \[ bh anA4 1I"11""1'
(7.55)
2(a4 + b4)[ a
318 71 Gas-Solid Separation
Therefore, the drag force on a single cylinder, FD, is found to be
Furthermore, the averaged pressure gradient over a single cylinder may be estimated from
the force balance on the cylinder as
dp FD 4nU0\. ln b +, a4 I]"1 (7 57)

[ -
2 2
Note that the volume fraction of the filter is f = a /b and the superficial velocity is
U = Q/A. Thus, for a flow which is transverse to the cylindrical fibers, the pressure drop
over a filter of thickness L is given by
A/7 = l 8 G /W
n o l fn
,
+
L a2 A \ \+afJ
For flow through a layer of a porous medium, the pressure drop can be estimated from
Eq. (5.358), as discussed in 5.6. In the case of a random assemblage of fibers with a layer
of dust, the total pressure drop for a fabric filter in cake filtration may be expressed in the
form
Apt = Ap3 + yi Api + y2Ap2 (7.59)
where Api, Ap 2 , and A/?3 are the pressure drops contributed by each of the three different
basic flow modes given. y\ and yi are the weighting factors. In random assemblage flows,
it is necessary to give twice the weight to the correlation for the flow perpendicular to cylin-
ders as to that for theflowparallel to cylinders since the model for the flow perpendicular to
cylinders does not distinguish between crossed and parallel arrangements of the cylinders
[Happel, 1959].
Example 7.3 Consider a clean house bag filter filled with randomly assembled fibers.
The porosity of the filter is 0.97, and the fibers have an effective diameter of 10 /xm. The
length of the filter is 0.1 m. If the flow velocity of the dust-laden air is 0.2 m/s, estimate the
pressure drop across the filter. The viscosity of air is 1.8 x 10~5 kg/m s.
Solution The overall pressure drop is a combined result of flow parallel to the
fibers and flow perpendicular to the fibers. The pressure drop of flow parallel to the fibers
is estimated from Eq. (7.45) as
\6LUfiaf
Ap|, = df (2af - a\jl - 3/2 - In a f )
16 x 0.1 x 0.2 x 1.8 x 10~5 x 0.03

= 836 Pa (E7.8)
10- 10 (2 x 0.03 - (0.03)2/2 - 1.5 - In0.03)
and the pressure drop of flow perpendicular to the fibers is estimated from Eq. (7.58) as
(
Ap = 2
-l
32 x 0.1 x 0.2 x 1.8 x 10"5 x 0.03 / 1 - 0.03 2 \
= IF* (ln0-03 + TToo^J
= 1,378 Pa (E7.9)
For aflowthrough a random assemblage of fibers, it is necessary to give twice the weight
to the pressure drop from the flow perpendicular to fibers as that from the flow parallel to
fibers; therefore, the overall pressure drop is obtained as
Apt = Apu + 2Ap = 836 + 2 x 1,378 = 3,592 Pa (E7.10)
7.4.3 Collection Efficiency of Fabric Filters

The collection efficiency of a fiber is defined as the ratio of the number of particles
striking the fiber to the number which would strike it if the streamlines were not diverted
[Dorman, 1966]. For particles with dv > 1 /xm, the particles are collected primarily by in-
ertial impaction and interception. Collection by diffusion is only important for submicron
particles. In the following, we only discuss the methodology of collection efficiency by
interception. Details about the filtration of fine and ultrafine powders or granular filtration
in general can be found in Davies (1973), Fuchs (1964), Dickey (1961), Matteson and Orr
(1987), and Tien (1989).
Consider the case of flow transverse to a cylindrical fiber, as shown in Fig. 7.15. The
Gas
streamline
/ Fiber
collector
Figure 7.15. Interception of particles by a fiber.

collection efficiency of a fiber, rj[, can be directly given by its definition
m= - (7.60)
a
where y is the limiting width of the particle stream to be collected, y can be calculated from
the particle trajectories in theflowaround the fiber. In general, the equation of motion for the
particle is coupled with the momentum equation for the gas phase. The complexity of the
coupled equations requires, in most cases, that the particle trajectory be solved numerically.
To determine the collection efficiency for a bag filter of thickness L, height H, and width
VK, it is assumed that the fibers are loosely packed and, thus, the interaction between the
fibers can be neglected. Consider theflowmoving transverse to the cylinders. The interstitial
velocity u is given by
u = -^ (7.61)
1 -af
where U is the superficial velocity of the dust-laden gas. In an element of thickness dL,
the number of fibers Nf is
(7.62)
nal
Within a time interval dt, the number of particles collected by a single fiber is
-dNi =2yudtWn (7.63)
Thus, within a time interval dt the total number of particles collected by all fibers in the
element becomes
-dN = 2yu dtWnNf (7.64)
The total number of particles entering within time dt is
N = U dtHWn (7.65)
Thus, we have
N na2(l - off)
Integrating Eq. (7.66) over L yields
lyoifL
M ~n o,, , , (7-67)
No
where NL is the number of particles at the filter outlet. Therefore, by substituting Eq. (7.60)
into Eq. (7.67), the collection efficiency of the bag filter, rjf, is obtained as
(7.68)
7.5 Gravity Settling and Wet Scrubbing

The most classical method for solids removal from a gaseous stream is gravity
settling. In gravity settling, the natural gravitational force separates particles from their
carrying medium on the basis of the difference in material density. Although gravity settling
devices or gravity settlers have the advantages of simple structure and economical operation,
7.5 I Gravity Settling and Wet Scrubbing 321
their development has been nearly halted by the low collection efficiency (typically 20-60
percent for dv > 40 /zm) and the large collection chamber requirement compared to that of
cyclones, fabric filters, wet scrubbers, and electrostatic precipitators. Nevertheless, gravity
settlers are still used in industry as precleaners to remove large particles.
Wet scrubbing is another technique to remove solids from a gas stream. In wet scrubbing,
solid particulates collide with liquid droplets generated from the nozzles in the scrubbers.
The collected particles and liquid droplets then form a slurry which flows to the slurry
collection chamber. Water is the most commonly used working liquid in wet scrubbing. The
scrubbers can be installed inside existing separators such as cyclones, gravity settlers, and
bed filters, or as an independent unit. The advantages of wet scrubbing include low pressure
drop, relatively high efficiency for fine particle collection, no particle reentrainment, and
possible removal of some gaseous pollutants simultaneously by using chemical solutions
as the collecting liquid. The disadvantage of this separation method is the necessity for the
treatment of the wet sludge generated from the wet scrubbing process.
7.5.1 Gravity Settling Chambers

In gravity settling separation, the particle-laden gas is fed horizontally into a large
expansion chamber. The enlargement of the cross section of the flow stream significantly
reduces the gas velocity. The particles settle downward on the collection surface or hopper
collectors at the bottom of the chamber. The decrease in the gas stream velocity reduces
the reentrainment of collected particles and increases the residence time of the particles so
that they may have sufficient time to settle by gravity.
There are two basic types of gravity settlers: the simple expansion chamber or horizontal
flow settling chamber (Fig. 7.16(a)) and the multiple-tray settling chamber or Howard
settling chamber (Fig. 7.16(b)). In the multiple-tray settling chamber, the presence of thin
collection plates yields a slightly larger linear gas velocity and significantly shorter vertical
settling distance compared to those in the simple expansion chamber. Thus, the overall
collection efficiency of the multiple-tray settling chamber is higher than that of the simple
expansion chamber. For a given collection efficiency, the size of the multiple-tray settler is
smaller than that of the simple expansion settler. A slight variation of the gravity settling
chamber, known as an elutriator, is also frequently used in industry. A system of three
elutriators in series is schematically illustrated in Fig. 7.17.
For a given type of gravity settler, the pick-up velocity and the minimum diameter of
particles which can be collected must be defined. The pick-up velocity is the onset velocity
for particle reentrainment and depends on the particle diameter. The gas flow rate is to
be controlled in such a way that the average gas velocity in the settling chamber is much
smaller than the pick-up velocity. The minimum diameter of collected particles can be
evaluated by the following analysis.
Consider a small spherical particle settling in a two-dimensional (2-D) rectangular-
shaped simple expansion chamber. The settling chamber is of length L and height H.
Assume that (1) the gas velocity U is uniformly distributed throughout the chamber;
(2) there is no slip between the particle and gas in the horizontal direction; (3) the par-
ticle settles down at its terminal velocity; and (4) the particle is subjected to the Stokes drag.
Thus, the residence time of the small particle, tT, is given by
tr = ^ (7.69)
Dust-laden
gas
I I I I
Dust collection hoppers
(a)
Multiple tray
Dust-laden Clean
gas
(b)
Figure 7.16. Typical gravity settling chambers: (a) Horizontal flow settling chamber;
(b) Howard settling chamber.
Dust-laden
air
Large Small Fine

particles particles particles
Figure 7.17. Elutriators in series.

and the settling time, fs, can be estimated by
*s = 77- (7.70)
where y is the initial vertical distance between the particle and the collection plate and U pt
is the particle terminal velocity given by Eq. (1.7). To ensure that the particles are collected,
the maximum settling time of the particle, tsm, should be less than the particle residence
time in the chamber, i.e.,
tsm = ^-< tT (7.71)

uPt
Substituting Eq. (1.7) and Eq. (7.69) into Eq. (7.71) leads to the following expression for
the diameter of the particle that can be collected as
When the gas velocity in the settler is sufficiently high, the particles deposited on the
collection plates can be picked up by the gas stream. This reentrainment can substantially
reduce the collection efficiency. By neglecting interparticle friction, the pick-up velocity,
/pP, may be estimated as [Zenz and Othmer, 1960]
(7.73)
In actual applications, the gas flow in a gravity settler is often nonuniform and turbulent;
the particles are polydispersed; and the flow is beyond the Stokes regime. In this case, the
particle settling behavior and hence the collection efficiency can be described by using the
basic equations introduced in Chapter 5, which need to be solved numerically. One common
approach is to use the Eulerian method to represent the gas flow and the Lagrangian method
to characterize the particle trajectories. The random variations in the gas velocity due to
turbulent fluctuations and the initial entering locations and sizes of the particles can be
accounted for by using the Monte Carlo simulation. Examples of this approach were
provided by Theodore and Buonicore (1976).
7.5.2 Mechanisms of Scrubbing and Types of Scrubbers

Wet scrubbing uses liquid droplets to remove fine dust in a gas stream. In all types
of wet scrubbing, the basic cleaning mechanism involves the attachment of particles to the
droplets. The function of the droplets in scrubbers is similar to that of spherical fibers in
filtration. Likewise, the primary collection mechanisms in scrubbing are similar to those in
filtration, i.e., inertial impaction, interception, and diffusion [e.g., Fan, 1989]. Secondary
collection mechanisms include thermophoresis due to temperature gradients, coagulation
of particles due to particle electrification, and particle growth due to liquid condensation.
Almost all particle-separating devices can be converted into wet scrubbers by adding
liquid spraying systems. Three types of commonly used scrubbers are the spray chamber,
cyclonic scrubber, and venturi scrubber. Figure 7.18 shows a simple spray chamber in
which water is sprayed through a series of nozzles into a settling chamber. The dust-laden
gas is fed from the bottom of the chamber and exits from the upper portion of the chamber.
Clean gas
t
Spray liquid
Spray
Dust-laden
gas
If
I Slurry
Figure 7.18. Spray chamber.
The collection of particles is achieved in a countercurrent flow between the water droplets
and the particulates. In a cyclonic scrubber, water is injected into the cyclone chamber from
sprayers located along the central axis, as shown in Fig. 7.19. The water droplets capture
particles mainly in the cross-flow motion and are thrown to the wall by centrifugal force,
forming a layer of slurry flow moving downward to the outlet at the bottom of the cyclone.
Another type of scrubber employs a venturi, as shown in Fig. 7.20. The velocity of the
gas-solid suspension flow is accelerated to a maximum value at the venturi throat. The inlet
of the water spray is located just before the venturi throat so that the maximum difference in
velocity between droplets and particles is obtained to achieve higher collection efficiency
by inertial impaction. A venturi scrubber is usually operated with a particle collector such
as a settling chamber or cyclone for slurry collection.
7.5.3 Modeling for Scrubbing and Collection Efficiency

In wet scrubbing, water droplets collide with particles in threeflowmodes of opera-
tion: (1) cross-flow; (2) cocurrent flow; and (3) countercurrent flow. Here the mathematical
Clean
Water
spray
Swirling
Dust-laden dust
gas
Slurry
outlet
Water
inlet
Figure 7.19. Cyclone scrubber.
model proposed by Fonda and Herne [Ogawa, 1984] is introduced to describe each of these
flow modes.
A model for wet scrubbing in a cross-flow is illustrated in Fig. 7.21. Consider a rectan-
gular scrubbing domain of length L, height H, and width of unity in Cartesian coordinates.
Assume that the gas-solid suspensionflowis moving horizontally, and that the solid particles
are spherical and of uniform size. The particle concentration across any plane perpendicular
to the flow is assumed to be uniform. The water droplets fall vertically and are uniformly
distributed in the flow system.
The mass balance of particles in the volume element in Fig. 7.21 can be given as
Hppn - UvHpv(n + ^ {1.1 A)
where dw is the averaged diameter of the water droplets; a w is the volume fraction of droplets;
3/2dw is the ratio of droplet cross-sectional area to droplet volume so that (3/2J w )// dx aw
represents the total cross-sectional area of the water droplets in that volume element; t] w is
the collection efficiency of a single water droplet; Uv and Uw are velocities of particles and
jet
Dust-laden
gas
Settling chamber
Slurry
Figure 7.20. Venturi scrubber.
droplets, respectively; and p p is the particle density. Thus, we have

dn 3 Uw 3 Qw H
(7.75)
ft 2dW Ury 2dyr Q\\ L
where Qv and Q^ are the flow rates of the particles and the droplets, respectively; and av{
is the particle volume fraction at inlet. The total collection efficiency, t] s, is thus obtained
from the integration of Eq. (7.75) as
(7.76)
where nx and n0 denote the particle number densities at inlet and outlet, respectively.
7.5 I Gravity Settling and Wet Scrubbing 327
Water droplets
Gas-solid
flow
Figure 7.21. Cross-flow mode.
Clean gas Water droplets

U fi
Gas-solid
flow
Figure 7.22. Countercurrent-flow mode.
Equation (7.76) indicates that the collection efficiency increases with the height of the
scrubbing chamber and with increasing ratio offlowrate of droplets toflowrate of particles.
For the vertical countercurrent scrubbing system shown in Fig. 7.22, the gas-solid sus-
pensionflowsupward from the bottom of the chamber and the water droplets fall downward.
It is assumed that both the gas-solid suspension velocity and droplet velocity are unchanged
in the process of scrubbing. It is also assumed that the no-slip condition exists between the
particles and gas. The particle concentration across any horizontal plane is assumed to be
uniform.
Similarly to the case of cross-flow, the mass balance of particles in the volume element
shown in Fig. 7.22 is expressed by
UpLppn - UvLpv ( n + - ^ dy J - UwLrjwaw ( j pvn dy = 0 (7.77)

\ ay J \Zaw J
where a w is
(7.78)
- UV)L
y A
H
Water droplets
fiw- -*-

Gas-solid Clean gas
flfVW
p
o
Figure 7.23. Cocurrent-flow mode.
Equation (7.77) thus becomes

d/i _ 3 f/w wQ?pi dj
(7.79)
which leads to
r]wapiH
(7.80)
It is noted from Eq. (7.78) that the maximum droplet holdup can be reached when the up-
ward gas-solid mixture velocity Uv approaches the sedimentation velocity of the droplets,
assuming that no coalescence of droplets occurs. Under this condition, the collection effi-
ciency is maximum.
In the case of cocurrent flow scrubbing, as shown in Fig. 7.23, it is assumed that the
velocity of the gas-solid suspension is always higher than the droplet velocity. The volume
fraction of droplets is then given by
Ofw = (7.81)
(Uv - UW)L
which results in the collection efficiency in cocurrent flow as
= 1-exp -- (7.82)
2 (/p - Uw) QP )
It is interesting to note that the volume fraction and collection efficiency in cocurrent
scrubbing are in the same form as those in countercurrent scrubbing even though the flow
directions of the dust-laden gas in these two cases are opposite to each other.
Nomenclature
A Constant, defined by Eq. (7.7) a Radius of fiber
A Area of electrode a Dimension, defined by Fig. 7.
A{ Area, dimension, defined by b Equivalent radius of flow cell
Fig. 7.8 around a single fiber
Nomenclature 329
Static pressure at the center of the

c Constant, defined by Eq. (7.9)
cyclone
df Diameter of fiber
dp Particle diameter Q Gas flow rate
dpC Cut-off size of a cyclone Qv Particle flow rate
dpm Parameter, defined by Eq. (E7.2) Gw Droplet flow rate
dw Diameter of water droplet q Charge carried by a particle
E Electric field strength R Radius of cross section of the
FD Drag force cyclone
fu Mass density function R Maximum radius of the cyclone
g Gravitational acceleration Rew Reynolds number based on Uw
H Height of a rectangular collecting r Radial coordinate
plate r* Dimensionless radial coordinate
H Cyclone dimension, defined by fo Radial location of particles at t = 0
Fig. 7.8 >b Cyclone dimension, defined by
H Height of a scrubber Fig. 7.8
H Height of the settling chamber rQ Cyclone dimension, defined by
H Height of the bag filter Fig. 7.8
h Cyclone dimension, defined by Radius of free vortex
Fig. 7.8 s Cyclone dimension, defined by
L Length of a rectangular collecting Fig. 7.8
plate T Absolute gas temperature
L Length of a scrubber t Time
L Thickness of the filter tm Mean residence time
L Length of the settling chamber tm\ Minimum residence time
I Deepest descending length twl Maximum residence time
/w Distance between two collecting Residence time of a particle
plates k Settling time of a particle
m Mass of a particle Maximum settling time of a particle
N Number of particles u Superficial gas velocity
No Number of particles at t = 0 u Gas velocity component
Ni Number of particles collected by a u* Dimensionless radial gas velocity
single fiber u0 Velocity of the moving cylinder
N{ Number of fibers Velocity of electric wind
NL Number of particles at outlet um Particle migration velocity in the
Nm Number of particles at t = tm electric field
n Particle number density up Particle velocity
Particle number density at the f/pp Pick-up velocity
separator inlet t/pt Particle terminal velocity
n0 Particle number density at the
Radial gas velocity near the
separator outlet
cyclone wall
V Pressure 11 Velocity of droplets
u Interstitial gas velocity Weighting factors, defined by

V Gas velocity component Eq. (7.59)
Dimensionless gas velocity A/7 Pressure drop
component 8 Parameter, defined by Eq. (E7.2)
Tangential gas velocity at the Permittivity of a vacuum
cyclone inlet Collection efficiency of a cyclone
Tangential gas velocity near the Fractional collection efficiency at
cyclone wall size intervals
w Gas velocity component Collection efficiency of an
w Width of the bag filter electrostatic precipitator
y Limiting width of the particle Collection efficiency of a filter
stream Collection efficiency of a fiber
Initial vertical distance between Collection efficiency of a scrubber
the particle and collection plate
Collection efficiency of a droplet
Axial coordinate
Tangential coordinate
Dynamic viscosity of gas
Greek Symbols V Kinematic viscosity of gas
Fiber volume fraction Parameter, defined by Eq. (7.27)
ttp Particle volume fraction Parameter, defined by Eq. (E7.6)
<*pi Particle volume fraction at inlet P Density of gas
Otpo Particle volume fraction at outlet Pv Density of particles
<*w Volume fraction of droplets Stokes relaxation time
P Vortex exponent a Sticking probability
r Circulation Stream function
rm Maximum circulation CO Angular velocity
References
Alexander, R. M. (1949). Fundamentals of Cyclone Design and Operation. Proc. Australas. Inst.
Min. Met., 152, 203.
Burgers, J. M. (1948). A Mathematical Model Illustrating the Theory of Turbulence. In Advances in
Applied Mechanics, 1. Ed. von Mises and von Karman. New York: Academic Press.
Cooper, D. W. and Freeman, M. P. (1982). Separation. In Handbook of Multiphase Systems. Ed. G.
Hetsroni. New York: McGraw-Hill.
Davies, C. N. (1973). Air Filtration. London: Academic Press.
Deutsch, W. (1922). Bewegung und Ladung der Electrizitatstrager im Zylinderkondensator. Ann.
Phys., 68, 335.
Dickey, G. D. (1961). Filtration. New York: Reinhold.
Dorman, R. G. (1966). Filtration. In Aerosol Science. Ed. C. N. Davies. New York: Academic Press.
Ergun, S. (1952). Fluid Flow Through Packed Columns. Chem. Eng. Prog., 48, 89.
Fan, L.-S. (1989). Gas-Liquid-SolidFluidization Engineering. Stoneham, Mass.: Butterworths.
Fuchs, N. A. (1964). The Mechanics of Aerosol. Oxford: Pergamon Press.
Happel, J. (1959). Viscous Flow Related to Arrays of Cylinders. AIChEJ., 5, 174.
Kercher, H. (1969). Elektischer Wind: Ruckspriiken und Staubwiderstand als Einflu^groên im
Elektrofilter. Staub, 29, 314.
Problems 331
Leith, D. and Licht, W. (1972). The Collection Efficiency of Cyclone Type Particle Collectors: A
New Theoretical Approach. AIChE. Symp. Sen, 68(126), 196.
Matteson, M. J. and Orr, C. (1987). Filtration: Principles and Practices, 2nd ed. New York: Marcel
Dekker.
Ogawa, A. (1984). Separation of Particles from Air and Gases, II. Boca Raton, Fla.: CRC Press.
Rietema, K. (1962). Proceedings of Symposium on Interaction Between Fluid and Particles.
London: Institute of Chemical Engineers.
Hemisphere.
Svarovsky, L. (1981). Solid-Gas Separation. New York: Elsevier Scientific.
Theodore, L. and Buonicore, A. J. (1976). Industrial Air Pollution Control Equipment for
Particulates. Cleveland, Ohio: CRC Press.
Tien, C. (1989). Granular Filtration of Aerosols and Hydrosols. Stoneham, Mass.: Butterworths.
White, H. J. (1963). Industrial Electrostatic Precipitation. Reading, Mass.: Addison-Wesley.
Zenz, F. A. and Othmer, D. F. (1960). Fluidization and Fluid-Particle Systems. New York: Reinhold.
Zhou, L. X. and Soo, S. L. (1990). Gas-Solid Flow and Collection of Solids in a Cyclone Separator.
Zhou, L. X. and Soo, S. L. (1991). On Boundary Conditions of Particle Phase and Collection
Efficiency in Cyclones. Powder Tech., 64, 213.
Problems
7.1 Show that Eq. (7.7) represents a set of nontrivial solutions of Eq. (7.6).
7.2 A tangential inlet cyclone with steep cone is operated under the following conditions:
R = 0.12 m, Q = 0.20 m3/s, Tm = 4.2 m2/s, H = 0.87 m, dv = 10/xm, p p = 2,000
kg/m3, q/m = 10~4 C/kg, avpv/p = 4, and a = 0.05. The carrier gas is air at ambient
conditions. Calculate the collection efficiency. Assuming all conditions except particle
diameter remain unchanged, estimate the cut-off size of this cyclone.
7.3 A cyclone has an overall collection efficiency of 90 percent for dust-laden air at 1 atm and
25C.The mass loading of dust to air is 5. The results of the particle size analysis are given
in Table P7.1.
Table VIA. Data on Size Distribution of

Particles at Inlet and Outlet of the Cyclone
Particle size analysis (wt%)

Size range
(/Ltm) Inlet Outlet
0-5 8.0 76.0

5-10 1.4 12.9
10-15 1.9 4.5
15-20 2.1 2.1
20-25 2.1 1.5
25-30 2.0 0.7
30-35 2.0 0.5
35-40 2.0 0.4
40-45 2.0 0.3
>45 76.5 1.1
Plot the fractional efficiency curve and estimate the cut-off diameter of this cyclone.
7.4 Consider a parallel-plate electrostatic precipitator across which a 10 kV potential is applied.

The interplate spacing is 1 cm. The particle density is 2,000 kg/m3. The total surface area is
10 m2, and the flow rate is 0.5 m 3/s. Use the test results provided in Table P7.2 to estimate
Table P7.2. Data on Properties of
Particles Collected in the Electrostatic
Precipitator
Um (cm/s) wt%(< dp)
5 2.2
10 5.9 10
20 8.3 50
40 11.2 90
80 13.8 99.6
>80 14.0
(a) Charge-to-mass distribution versus particle size

(b) Overall collection efficiency
(c) Size distribution (wt%) at the outlet
7.5 Consider a clean house bag filter filled with randomly assembled fibers. The porosity of the
filter is 0.97, and thefibershave an effective diameter of 10 /xm. The length of thefilteris 0.1
m. The viscosity of air is 1.8 x 10~5 kg/m s. For a velocity of the dust-laden air of 0.2 m/s,
and a collection efficiency of a single fiber of 1.50 percent, estimate the overall efficiency
of the filter. To reach a collection efficiency of 99.99 percent, what is the minimum length
of this type of filter? At this minimum length, what is the overall pressure drop?
7.6 On the basis of the Navier-Stokes equation given in Eq. (7.6) for the gas flow in a cyclone
separator, derive the static pressure distribution equations, i.e., Eqs. (7.3) and (7.4), for this
separator. It is assumed that the radial component of the gas velocity is negligible.
CHAPTER 8
Hopper and Standpipe Flows
8.1 Introduction
Granular materials in the food, pharmaceutical, chemical, and fuel industries are
routinely processed through hoppers and standpipes. When the solids flow in a vertical
or inclined pipe is dominated by gravity and the solids are in a nearly packed condition,
the flow is known as a standpipe flow or downcomer flow. In a standpipe flow, the cross-
sectional area of the standpipe is unchanged and the gas pressure at the bottom is higher
than that at the top. The downward transport may also be achieved with higher pressure at
the top. In this case, the solids transport is referred to as a discharged flow. The discussion
of discharged flow is, however, beyond the scope of this chapter; readers interested in
discharged flow may refer to the work of Berg (1954), and Knowlton et al. (1986, 1989).
Hoppers can be classified into two categories, namely, mass flow hoppers and core (or
funnel) flow hoppers. A mass flow hopper is characterized by the simultaneous movement
of every particle of the material during the discharge process, as shown in Fig. 8.1 (a). On
the other hand, in a core flow hopper, there exist stagnation zones in the material discharge
process, as shown in Fig. 8.1(b). Advantages of mass flow hoppers are overwhelming
compared to those of core flow hoppers. Problems like flooding, segregation, inconsistent
flow rate, arching, and piping are less likely to occur in mass flow hoppers. There are two
mainflowpatterns of hopper flows. Thefirsttype is illustrated by conical or axisymmetrical
hoppers where stresses, stress distribution, andflowpattern of solids are symmetric about the
central axis of the hopper. The second type is the two-dimensional or plane-strain hoppers,
in which stresses and solids flow are independent of the third transverse dimension.
In this chapter, the mechanics of hopper flow and standpipe flow along with their oper-
ational characteristics are described. Problems such as segregation, inconsistent flow rate,
arching, and piping that disrupt and obstruct theflowof bulk solids in hoppers are discussed.
Remedies with respect to the use of flow-promoting devices such as vibrators and aerating
jets to reinitiate the flow are presented. The importance of the flowability of the solids to
be handled in relation to the hopper design is emphasized.
8.2 Powder Mechanics in Hopper Flows

The onset of powder motion in a hopper is due to stress failure in powders. Hence,
the study of a hopper flow is closely related to the understanding of stress distribution in a
hopper. The cross-sectional averaged stress distribution of solids in a cylindrical column
was first studied by Janssen (1895). Walker (1966) and Walters (1973) extended Janssen's
analysis to conical hoppers. The local distributions of static stresses of powders can only
be obtained by solving the equations of equilibrium. From stress analyses and suitable
failure criteria, the rupture locations in granular materials can be predicted. As a result, the
flowability of granular materials in a hopper depends on the internal stress distributions
determined by the geometry of the hopper and the material properties of the solids.
333
334 81 Hopper and Standpipe Flows
stagnant zone
(a) (b)
Figure 8.1. Hopper types based on flow patterns: (a) Mass flow hopper; (b) Core (funnel) flow
hopper.
Figure 8.2. Stress components in a plane-strain problem.
8.2.1 Mohr Circle for Plane Stresses

Consider a plane-strain problem in the x-z plane, as shown in Fig. 8.2. The stress
tensor, expressed in Cartesian coordinates, takes the form
0 rx
T = 0 ay 0 (8.1)
0 oz
where the o 's are normal stresses (compressive stresses are considered as positive) and the
r's are shear stresses. From Hooke's law (Eq. (1.50)) and assuming no displacement in the
y -direction, we have
ay = (8.2)
8.2 I Powder Mechanics in Hopper Flows 335
Figure 8.3. Equilibrium of stress components on a differential element.
where v is Poisson's ratio. In addition, from the conservation of angular momentum, T is

found to be a symmetric tensor so that TXZ = rzx. Thus, the plane stress tensor in Eq. (8.1)
depends solely on crx, crz, and rxz.
Since a tensor is invariant with respect to the selection of the reference coordinates, the
stresses on any plane can be expressed in terms of ox, a z , and rxz. As shown in Fig. 8.3, the
force balance of the differential element results in the stress relationships on the BC plane,
as given by
a = ax cos2 fi + a z sin2 /? + 2rxz sin ft cos ft

(8.3)
x= TXZ(COS2 P sin2 p) + (az ax) sin /3 cos
where P is the angle between the normal of the BC plane and the x-axis. From Eq. (8.3),
two perpendicular planes can be found on which shear stresses vanish (i.e., r = 0). The
directions of these planes are known as the principal directions and the corresponding normal
stresses as the principal stresses. The angle for principal directions, /3pr, is determined from
Eq. (8.3) as
2rx;
fipr = - tan - l (8.4)
which yields the principal stresses as
+ tfz
(8.5)
If the principal directions are taken as the x- and z-axes, Eq. (8.3) is reduced to
a = crx cos 2 ft + oz sin 2 ft

(8.6)
x = (crz <JX) sin fi cos ft
which is an equation of a circle in the a-x coordinates, as shown in Fig. 8.4. This circle is
known as the Mohr circle. The direction and magnitude of stresses on any plane can also
be determined graphically from the Mohr circle. As shown in Fig. 8.4, the normal stresses
on the principal planes are of maximum and minimum values.
Figure 8.4. Mohr circle and Mohr-Coulomb failure envelope.
8.2.2 Mohr-Coulomb Failure Criterion and Coulomb Powders

The most common failure criterion for granular materials is the Mohr-Coulomb
failure criterion. Mohr introduced his theory for rupture in materials in 1910. According to
his theory, the material fails along a plane only when a critical combination of normal and
shear stresses exists on the failure plane. This critical combination, known as the Mohr-
Coulomb failure criterion, is given by
r = c + a tan rj (8.7)
where c is the cohesion defined as the resistance of the material to shear under zero normal
load and is a result of the intermolecular cohesive forces, frictional forces, and other forces
acting on the material. t] is the angle of internal friction of the material, which corresponds
to the maximum static friction condition as the bulk solids start to slide on themselves at
the state of incipient failure.
The Mohr-Coulomb failure criterion can be recognized as an upper bound for the stress
combination on any plane in the material. Consider points A, B, and C in Fig. 8.4. Point A
represents a state of stresses on a plane along which failure will not occur. On the other
hand, failure will occur along a plane if the state of stresses on that plane plots a point on the
failure envelope, like point B. The state of stresses represented by point C cannot exist since
it lies above the failure envelope. Since the Mohr-Coulomb failure envelope characterizes
the state of stresses under which the material starts to slide, it is usually referred to as the
yield locus, YL.
A rigid-plastic powder which has a linear yield locus is called a Coulomb powder.
Most powders have linear yield loci, although, in some cases, nonlinearity appears at low
compressive stresses. A relation between the principal stresses in a Coulomb powder at
failure can be found from the Mohr circle in Fig. 8.4 as
(<73 + at) _ 1 - sin r\

(8.8)
(<j\ + <rt) 1 + sin rj
where at is the tensile strength of the powder. For cohesionless powders (at = c = 0),
Eq. (8.8) reduces to
(T3 1 sin rj
(8.9)
o\ 1 + sin rj
An important application of this equation is to distinguish between two extreme failure
conditions, known as the active and passive failures. First, the active and passive states of
stress may be explained as follows: Consider a cohesionless Coulomb powder. If the powder
is assembled in a large container in successive horizontal layers without disturbance, there
will be no shear stresses along the horizontal and vertical planes inside the powder because
of the symmetry of the problem. Thus, at any point, the horizontal and vertical normal
stresses are the principal stresses at that point. In this case, if the major principal stress is
the horizontal stress, crh, the powder is said to be in a passive state of stress. On the other
hand, if the major principal stress is the vertical stress, a v , the powder is in an active state
of stress. Thus, Eq. (8.9) can be written for each state as
<7h _ 1 + sin Y]
= Kp for passive states where a h > ov
ov 1 sin T)
(8.10)
crh 1 sin r)
= Ka for active states where <JV
<JV 1 + sin rj
where Kv and Ka are the ratios between the principal stresses in the passive and active
states, respectively. When the Mohr circle for an active or passive state of stress touches
the yield locus, a state of active or passive failure takes place along the plane represented in
the Mohr circle by the radial line which passes the point of intersection between the circle
and the yield locus, as shown in Fig. 8.5.
During the filling of a hopper an active state of stress exists since the powder tends to
compress vertically. This case is also referred to as the "static state" of stress since the
powder is at rest. On the other hand, during discharging from the hopper, the powder tends
to expand in the vertical direction. In that case, the horizontal stress becomes the major
principal stress, leading to a state of passive stress. The passive state of stress in this case
may also be referred to as the "dynamic state" of stress because of the flow of powder.
8.2.3 Static Stress Distributions in Standpipes and Hoppers

So far, we understand that the flowability of powders depends on their "failure"
stresses from the Mohr-Coulomb failure criterion. Therefore, analyses of powder flows
ah a
(a) (b)
Figure 8.5. Mohr circles for two cases: (a) Active failure; (b) Passive failure.
hydrostatic
increasing bin
diameter
(a) (b)
Figure 8.6. Janssen's model for stress distribution in cylindrical bins: (a) Schematic of the
forces applied to a differential slice; (b) Wall pressure distribution; Eq. (8.16).
in bins and hoppers should be closely related to the stress analyses of the systems. By
treating a powder as a pseudocontinuum solid, the static stress distribution inside the hopper
or standpipe, in both active and passive states, may be obtained from the equations of
equilibrium and the appropriate boundary conditions. In the following sections, simple
analytical methods to evaluate cross-sectionally averaged stresses in a cylindrical vertical
column and a conical hopper are presented. For more details on these methods and other
more sophisticated ones, the reader may refer to specialized references on the subject [e.g.,
Home and Nedderman, 1978; Nedderman and Tiiztin, 1979; Drescher, 1991; Nedderman,
1992].
8.2.3.1 Averaged Stresses in a Cylindrical Column

A simple model for the estimation of stress distribution in cohesionless bulk solids stored
in a tall cylindrical bin was presented by Janssen in 1895. The system of consideration is
shown in Fig. 8.6(a). In Janssen's model, it is assumed that
(1) Stresses in the bulk solids are functions of z only.
(2) <rr and oz are principal stresses, which are related by
ox Koz (8.11)
where K equals Kv or Ka according to the passive or active state of stresses.
(3) The bulk density of solids, a p p p , is constant.
(4) Wall friction is fully mobilized and the solids are on the point of slipping in the
bin.
From Assumption (4), the wall shear stress, r w, can be related to the horizontal stress,
<jr,by
t w = f wa T (8.12)
where / w is the coefficient of friction at the wall. Therefore, a cross-sectionally averaged

force balance in the z-direction leads to
dcrz 4r w
=apPp8- = (8.13)
where D is the diameter of the cylinder bin. The boundary condition is

az = 0 at z = 0 (8.14)
Thus, oz is obtained as
(8.15)
Substituting Eq. (8.15) into Assumption (2) yields
(8.16)
Equations (8.15) and (8.16) show that, as z increases, az and ox approach the maximum
values of
0ivPvgD
(8.17)
4/ w
Moreover, the horizontal wall pressure, a r, may be estimated from Eq. (8.16). The results
are illustrated in Fig. 8.6(b).
It should be noted that the stresses in Janssen's model are cross-sectionally averaged.
Therefore, Janssen's results in principle are not applicable for local stress analysis. This
argument can be clearly shown by the inconsistency of Eqs. (8.15) and (8.16) with the
equations of equilibrium.
8.2.3.2 Averaged Stresses in a Storage Hopper

Janssen's analysis can be extended to the stress analysis of conical hoppers. Walker (1966)
proposed a differential slice with vertical sides, as shown in Fig. 8.7(a), where the cross-
sectionally averaged force balance in the vertical direction results in
N--^-=a vpvg (8.18)
where oz is the average vertical stress on the horizontal plane and N is a parameter related
to the geometry of the hopper system. In Walker's model, N is given by
N= (8.19)
(a)
Figure 8.7. Types of differential slices used for hopper analysis: (a) Walker (1966);
(b) Walters (1973).
Figure 8.8. Mohr circle for the state of stresses near the hopper wall.
where <pw is the hopper half-apex angle; A is a distribution factor, defined as the ratio of the
vertical normal stress at wall, <xzw, to cross-sectional averaged vertical normal stress, az\
and B is the ratio of the vertical shear stress at wall, r zw , to vertical normal stress at wall,
a zw , which can be evaluated using the Mohr circle for stresses at the wall (see Fig. 8.8) as
r zw sin rj sin 2(ww + y)
B = = (8.20)
<TZW 1 sin K) cos 2(<pw + y)
where y is the angle between the major principal plane and the hopper wall. Consider the
boundary condition at z = zo as oz = cr0, where a 0 is the surcharge on the top of the hopper.
Then, integrating Eq. (8.18) gives
N-i
azw = Aaz = (8.21)
The normal stress on the hopper wall, <xw, can be determined by

1 + sing cos 2y
tfw o z w :
(8.22)
1 sin r\ cos 2(cpw + y)
Walters (1973) considered the more accurate differential slice with sides aligned with
the hopper walls as shown in Fig. 8.7(b). His results are similar to those of Walker with a
slight difference in the parameter N. In Walters's approach, N is given by
(8.23)
It should be noted that the main objective in all of the foregoing analyses is to estimate
the upper bounds of wall pressure distributions in bins and hoppers for both active and
passive stresses. Simplicity of calculations in these methods makes them appropriate for
structural design purposes of bins and hoppers. However, they are not meant to introduce
a complete solution for the stress field within the bulk solids.
8.2.4 Stress Distribution in a Steady Hopper Flow

For a slow and steady motion of powders, the local stresses inside the powder
are approximated to be pseudostatic and, consequently, are governed by the equation of
equilibrium as
:0 (8.24)
where F represents the body force. Let us consider a gravity flow of a powder in a hopper.
We may express the equation of equilibrium in cylindrical coordinates with the z-axis in
the same direction as the gravity in the form
dar drrz ar
(8.25)
and
daz rrz (8.26)
The axisymmetric nature of conical hoppers results ineo =0 and, according to Eq. (2.20),
erg = (orr + crz) v. To close the problem, one more relation is required. In a continuum solid,
this relation is known as the compatibility requirement, i.e., the relationship of strains. This
relation, with the aid of constitutive relations between stress and strain (e.g., Hooke's law),
provides an additional equation for stress so that the problem can be closed. However, the
compatibility relation for a continuum solid may not be extendable to the cases of powders.
Thus, additional assumptions or models are needed to yield another equation for stresses
in powders.
It was recognized by Jenike (1961) that the powder undergoes a continuous shear de-
formation without change of local stresses in a steady hopper flow. This indicates that the
bulk solid in a steady motion can be regarded as a plastic, as illustrated in Fig. 8.9. Mohr
circles which represent the steady flow of any packed particulate materials under different
consolidating conditions are tangential to a straight line (known as the effective yield locus)
that passes through the origin as shown in Fig. 8.9. Thus,
sint]k=
where r)k is the kinematic angle of internal friction. Jenike (1961) also introduced the
theory of radial stress field in conical hoppers. In this theory, the stress is considered to
be a function only of the radial distance from the virtual apex of the hopper. Hence, with
appropriate boundary conditions, the local stress distributions in a steady hopper flow can be
obtained by solving the system of equations which consists of the equation of equilibrium,
ideal plastic
deformation
a3 70 ax a
Figure 8.9. Effective yield locus and evaluation of YQ and a\ for steady hopper flows.
Eq. (8.24), and the modified Mohr-Coulomb yield (or failure) criterion, Eq. (8.27). It should
be noted that other yield criteria, such as the von Mises criterion, are used to model the
flow of bulk solids in hoppers, and more conditions may need to be imposed, such as
the Levy flow rule, in order to close the system of equations [Cleaver and Nedderman,
1993].
8.2.5 Flowability of Powders in Hopper Design

Powderflowabilityis a major concern in hopper design to prevent problems such as
arching, insufficient flow rate and capacity, piping, segregation, andfloodingfrom occurring
inside hoppers. The initiation of powder flow is due to the incipient failure where the forces
acting on the solid produce stresses in excess of the strength of the solid. The rupture
of powders depends on the mechanical properties of bulk solids, including [Johanson,
1981]
(1) Angle of internal friction, rjs
(2) Kinematic angle of internal friction, r]k
(3) Kinematic angle of wall friction, ??w
(4) Bulk density, a p p p
(5) Unconfined yield strength, Yo
(6) Major principal stress, o\
(7) Cohesion, c
The angle of friction is related to the corresponding coefficient of friction by
r]i = tan"1 /i, i = s,k, andw (8.28)
In addition to the foregoing properties, there are other important flow properties such as the
angle of repose, angle of fall, compressibility, and dispersibility which provide additional
measures for the flowability of granular materials when discharged out of the hopper [Carr,
1965]. In general, flow properties are influenced by the conditions under which they are
handled, such as moisture content, storage time at rest, and consolidating pressure. All of
these properties can be determined with simple bench-top equipment such as the Jenike
translatory shear cell [Jenike, 1961], the rotational split-level shear cell [Peschl, 1989], and
other shear testing devices [Kamath et al., 1993].
A simple approach for the design of massflowhoppers was suggested by Jenike (1964a);
it is based on two conditions: (a) occurrence of flow at the hopper walls and (b) no arching
and/or piping. The first condition can be satisfied by applying the limiting relation between
the angle of the hopper wall, <pw, and the angle of wall friction, rjw, whereas the second
one can be met if the stress in a stable arch or pipe exceeds the unconfined yield strength
of the solids. In order to define the "flow" or "no-flow" criterion of solids in hoppers, let
us first examine the case where the bulk material could form a stable arch near the hopper
outlet. As shown in Fig. 8.10, the lower surface of the arch is free of stresses while the
major stress, ora, is acting on the plane perpendicular to that free surface. The Mohr circle
representing this state of stress must pass through the origin. When such a circle touches
the yield locus of the material, the major stress is then called the unconfined yield stress,
Yo, which represents in this case the strength of the material required to support the stable
arch. Thus, the critical state of stress <ra = 7 0 represents that needed to initiate the flow
and make the arch collapse. Dependence of the minimum size of hopper outlet required to
ca
Figure 8.10. Arching (or doming) at the hopper outlet.
tan (Vff)
Figure 8.11. Flow factor jfand material flow function FF.
prevent arching on the unconfined yield stress of the material can be obtained by studying
the equilibrium of forces in a stable arch (see Problem 8.3).
It is convenient to introduce the concepts of material flow function, FF, and flow factor,
ff. The materialflowfunction, FF, relates the unconfined yield stress, YQ, to the correspond-
ing major consolidating stress, o\, and is determined experimentally from the yield locus
of the material, as shown in Fig. 8.9. The material flow function is presented as a plot of
Fo versus o\, as shown in Fig. 8.11. The flow factor, ff, is defined by
ff=^- (8-29)
and it can be regarded as a measure of the stresses that might develop whenflowis interrupted
by the formation of a stable arch. Jenike (1964a) observed that the ratio o\ /aa is constant for
a given <pw, rj^, and ^ w. Thus, when plotted on the same coordinate with the flow function,
the flow factor is a straight line passing through the origin with a slope of l/ff, as shown
in Fig. 8.11. The intersection between ff and FF represents the critical condition marking
flow or no-flow of solids. The flow criterion is satisfied for the conditions where FF lies
below ff. The critical value of a a corresponding to the point of intersection can be used to
determine the minimum width of the hopper outlet required to ensure the collapse of the
2
ft?
1.5
10 20 30 40 50
Half angle of hopper (pw
Figure 8.12. Function H((p^) for circular and square outlets (from Jenike, 1964b).
arch by the relation [Jenike, 1964b]
"min = (8.30)
where dm[n is the minimum outlet dimension of the hopper to prevent arching, and H(cpw)
is a function of the hopper angle and is provided graphically for the cases of mass flow
hoppers with circular and square outlets, as shown in Fig. 8.12.
The introduction of the flow factor makes it unnecessary for designers actually to solve
the stress field equations in order to determine the dimensions of the hopper outlet. Jenike
(1964b) and several codes of practice for silo design [e.g., BMHB, 1988] provide designers
with graphs, obtained numerically, ofjâs a function of <pw, %, and t]w. The standard design
procedure of hoppers using these design graphs is illustrated in the following example.
Example 8.1 Three sets of yield loci under different consolidation conditions are obtained
for a sample of powder of bulk density 1,500 kg/m3, as shown in Fig. E8.1. If a conical
hopper is to be designed, determine the wall slope of the hopper and the opening size
necessary to ensure a steady mass flow. The angle of wall friction is 15. The design
diagram for mass flow conical hoppers is given in Fig. E8.2 [BMHB, 1988].
Solution The kinematic angle of internal friction can be determined from the
Mohr circle, which is tangential to the yield locus at the end point. This Mohr circle yields
the major consolidating stress o\ and minor consolidating stress <J3. Thus, % is found to
be 30, either from Eq. (8.27) or from a tangent of the Mohr circle which passes through
the origin, as shown in Fig. E8.1.
In order to determine FF, two Mohr circles are drawn for each yield locus, as shown
in Fig. E8.1. The circle passing through the origin and being tangential to the yield locus
gives the unconfined yield strength, YQ, whereas the circle which is tangential to the yield
locus at the end point yields the corresponding o\. Thus, from the three sets of yield loci,
three sets of 7 0 and o\ are obtained. Consequently, FF is determined as shown in Fig. E8.3.
1.5 o consolidating stress, o x
unconfined yield stress, Y o

1.0
0.5
0 0.5 1.0 1.5 2.0 2.5
Normal stress c , kN/m 2
Figure E8.1. Shear test results and flow property measurements.
10 20 30 40 50
Half angle of hopper <pw
(a)
10 20 30 40 50
Half angle of hopper <pw
(b)
Figure E8.2. Design diagram for mass flow conical hopper (from BMHB, 1988): (a) Flow
factor; (b) Angle of wall friction.
1.5
= 1.23 kN/m 2
. 1.0
0 0.5 1.0 1.5 2.0 2.5 3.0

2
Consolidating stress cx, kN/m
Figure E8.3. Flow factor ff and material flow function FF.
From Fig. E8.2, for y/w = 15 andrj^ = 30, the half angle of hopper walls, ^w , is found
to be 35 and the corresponding flow factor, jff, is 1.88. The intersection between FF andj^
in Fig. E8.3 gives the critical major stress in a stable arch aac as 1.23 kN/m2. The value of
H((pw) = 2.52 can be obtained from Fig. 8.12 for the conditions of a circular opening and
cpw = 35. The minimum diameter of the conical hopper opening can thus be obtained by
using Eq. (8.30) as
aac//(<pw) (1.23 x 1,000) x (2.52)
= 0.21 m (E8.1)
1,500x9.81
Therefore, to ensure mass flow of the given powder in a conical hopper, the half angle of the
hopper should be no more than 35 and the circular opening should be larger than 0.21 m
to prevent the formation of a stable arch that can span the opening.
8.3 Hopper and Standpipe Flow Theories

In this section, we introduce the theories of moving bed flows in hoppers and
general standpipes. The hopper flow theory may be extended to the cases of moving bed
flows through a discharge valve. A unique hopper-standpipe-<iischarger flow system is then
discussed as an example of the application of these theories. Multiplicity of steady flows
in a standpipe system is also illustrated.
8.3.1 Moving Bed Flows in a Feed Hopper

Let us consider moving bed flows through a conical feed hopper where the gas
moves upward relative to the moving bed of solids, although the gas motion relative to the
hopper wall may still be downward. It is convenient to write equations of motion in spherical
coordinates with the origin at the vertex and the polar axis along the axis of symmetry of the
cone. The flow is assumed to be invariant with respect to the azimuthal angle, 0. Moreover,
the volume fraction of gas in the moving bed, a mb, is constant.
8.3 I Hopper and Standpipe Flow Theories 347
The continuity equations for the gas and the particles, respectively, are
Y
dR R R d<p R
0 (832)
=
For the momentum equation of gas in a moving bed flow, the inertial and gravitational
terms can be neglected. The pressure drop is therefore balanced by the drag forces, which
give
MR) (8.33)
| ^ K p - ^ ) (8.34)
R dip
where ft is a coefficient which depends on the voidage of the mixture. The momentum
equations of solids are expressed by [Nguyen et ai, 1979; Chen et aL, 1984]
L 2 a G( T cot
<\ D R^~^T
dR dip + R^ t R - e ~ P + R^ V i + (! ~ mb)Pp<^ cos <^
U J\ I\ OiD t\
1
OA K ocp K
+ (1 - a m b ) p p I W R P - ^ j f ^ ^ 1
To close the problem, constitutive relations of powders must be introduced for the internal
connections of components of the stress tensor of solids and the linkage between the stresses
and velocities of solids. It is assumed that the bulk solid material behaves as a Coulomb
powder so that the isotropy condition and the Mohr-Coulomb yield condition may be used.
In addition, ae has to be formulated with respect to the other stress components.
The isotropy condition is that in the (R,<p) plane, the principal directions of stresses and
the rate of deformation coincide [Jenike, 1961]. Thus, the isotropy condition is written as
duRp 1 dUyy WRp
ofr-gy _ 3R R d(p R
~j 77; u \ 7,
TR<p / <py _ ^ p , J_ d^Rp
2V~9^ R~ ~R~dy~
The Mohr-Coulomb yield condition is given by
. 2 /o oo\
sin
^k (8.38)
Here, the Haar-von Karman hypothesis, which requires ae to be equal to one of the principal
stresses in the (R, <p) plane, is adopted. Thus, it is reasonable to assume a0 to be equal to the
major principal stress since the deformation of solids in a converging cone is compressive
in the azimuthal direction when the solids move down the hopper. This gives rise to
a, = ^^(l+sin % ) (8.39)
So far, we have nine equations (Eqs. (8.31)-(8.39)) for nine variables (uR, u<p, wRp, u^,
p,o^,OQ,Oy, and TR^). Therefore, the problem is closed and, in principle, can be solved
numerically. However, analytical solutions are available with further simplifications of the
model equations.
To obtain an analytical solution of the preceding problem, the stresses in terms of the
mean stress am and the stress angle \jr may be expressed by [Sokolovskii, 1965]
<fy = <rm(l + sin r/k cos 2^)
(o.4U)
?R<P = -<7m sin r]k sin 2f\ oe = a m (l + sin rjk)
where am = {aR-\-a(p)/2 and TT/2 x/r is the acute angle between the radial direction and the
major principal stress axis in the (R, <p) plane. Furthermore, variables may be approximated
by using a perturbation technique with respect to the polar angle <p [Nguyen et ai, 1979].
The expansion of variables in powers of cp terminates at terms of o((p). Thus, we have
p p(R); p p(R)
Hence, Eq. (8.31) and Eq. (8.32) reduce to
(g42)
Integrating the preceding equations yields
= ^ <8.44,
where the subscripts "i" and "o" denote the inlet and outlet locations, respectively, and the
f/'s represent the superficial velocities.
Similarly, for an ideal gas in an isothermal process, Eq. (8.33) becomes
f URpi URiPi
dR R2 V 1 - <*mb amb P .
which gives the pressure distribution in the mass hopper flow as
MK)
where A and B are defined by
1 - Of mb E/Rpi Qf
It is noted that the trigonometric functions in Eq. (8.35) can be expanded in powers of
(p as
(0 1
cos 2 ^ 1; sin 2 ^ 2 ^ w ; cot<p ^ - (8.49)
where <pw is the half angle of the feed hopper and ^w is the value of \jr at the wall of the
feed hopper. Equation (8.35) is thus simplified to
. dam Aam ( ^r w \
(1 - sin^ k )- 1 + sin 77k + (1 - amb)pvg
aK K \ <pw J
+ (1 - a m b )p p M R p -| + ^(MRp - MR) = 0 (8.50)

Cl/v
Substituting Eqs. (8.44) and (8.45) into the preceding equation gives
.- . .dam 4o-m / \/rw\
(1 - sin rjk)- 1 + sin 7/k + (1 - amb)pvg
dR R \ (pwj
l ? A( B\
(1 - amb)R 5 + "^
R2\ 1 - -p 1 = 0 (8.51)
On the basis of Eq. (8.51), distributions of the mean stress am can be determined and, thus,
stress distributions inside the hopper can be evaluated from Eq. (8.40).
Example 8.2 Consider a conical feed hopper with half angle 10. The inlet and outlet
diameters of the hopper are 1.0 and 0.5 m, respectively. The particles, 200 /im beads with a
density of 2,300 kg/m3, are fed into the hopper at the mass flow rate of 100 kg/m2 s under
an inlet pressure of 105 Pa. The particles are in a moving bed motion with a particle volume
fraction of 0.5. The pressure at the hopper outlet is 1.2 x 105 Pa. Determine the air flow
rate in this hopper flow. The coefficient /3 is given as 102,700 kg/m3 s.
Solution In spherical coordinates with origin at the vertex, the distances of the
hopper inlet and outlet from the vertex are given by
Ri = -2 = - ^ - = 2.84 m, Ro = - ^ - = - ^ - = 1.42 m (E8.2)

tan<pw tan 10 tan<pw tan 10
The superficial velocity of particles at the inlet is calculated by
(E83)
where the minus sign indicates that the particles flow in the opposite direction of the polar
axis.
The relationship between the inlet and outlet pressures is given by Eq. (8.47). Substi-
tuting all the given conditions into Eq. (8.47) leads to
105 - 1.2 x 105 + B In -

V l . 2 x l O 5 -B/
= 1.027 x 105 x 2.842 ( ^ ^ - ) ( - ) (E8.4)
from which B is obtained as 0.23 x 105 Pa. Thus, from Eq. (8.48), the superficial velocity
of gas at the inlet is solved as
f -0-043
The overall gas flow rate in the hopper is thus obtained by
Q = \URi\7tr? = 0.01 X J T X 0.52 = 0.0079 m 3 /s (E8.6)
8.3.2 Standpipe Flows

As defined in 8.1, standpipe flow refers to the downward flow of solids with
the aid of gravitational force against a gas pressure gradient. Although the gas flow with
respect to the downward-flowing solids is in the upward direction, the actual direction of
flow of gas relative to the wall can be either upward or downward [Knowlton, 1986]. Solids
fed into the standpipe are often from hoppers, cyclones, or fluidized beds. A standpipe
can be either vertical or inclined and its outlet can be simply an orifice or be connected
to a valve or a fluidized bed. There can be aeration along the side of the standpipe.
Figure 8.13 illustrates a typical standpipe application in catalytic cracking and regeneration
systems. In this application, solids are fed into the vertical standpipes from fluidized beds
and through slide valves. It should be noted that side aeration is employed for this standpipe
flow.
The model given here for flows in a vertical standpipe follows closely the approach of
Ginestra et al. (1980), Chen et al. (1984), and Jackson (1993). In this model, it is assumed
that
(1) Both solids and gas are regarded as a pseudocontinuum throughout the standpipe
system.
(2) Motions of solids and gas are steady and one-dimensional (in the axial direction).
(3) Solids can flow in either moving bed mode or dilute suspension mode. Solids
stresses among particles and between particle and pipe wall are considered in the
moving bed flows but neglected in the dilute suspension flows.
(4) The gas can be regarded as an ideal gas, and the transport process is isothermal.
The cylindrical coordinate system is selected for the standpipe, as shown in Fig. 8.14.
For a one-dimensional steady motion of solids, the momentum equation of the particle
phase can be written as
yOp(l a)uzp = ppg(l a) h r D (8.52)
where apz is the normal stress of solids; rpw is the shear stress of solids at the pipe wall; F&
is the drag force per unit volume; and Rs is the radius of the standpipe. Under low particle
Reynolds number conditions where the drag is dominated by the viscous effects, FD can be
expressed in terms of the relative velocity of gas and solids as
FD = P(uz-uzp) (8.53)
83 I Hopper and Standpipe Flow Theories 351
Flue
gas
Hydrocarbon
product
Fluidized bed
regenerator
Fluidized bed
reactor
Standpipe
Aeration
Slide valve
Hydrocarbon Air
feed
Figure 8.13. Typical solidsflowloop in a catalytic cracker with underflow standpipes.
1 t
Figure 8.14. One-dimensional standpipe flow.

For uniform gas-solid flows, the well-known Richardson and Zaki correlation [Richardson
and Zaki, 1954; Growther and Whitehead, 1978] given next can be applied
(8.54)
' can be expressed as

-a)
In Eq. (8.54), Upt is the particle terminal velocity, and n is the modified Richardson-
Zaki index for gas-solid systems. Thus, on the basis of Eqs. (8.52) and (8.53), a general
momentum equation for the solid phase becomes
/i \ dWzP /i \ dapz 2r p w
p p ( l Qf)wzp = p p g(l a) - h p(u z wzp) (8.56)
It should be mentioned that for flows at high Reynolds numbers, inertia effects must
be included in Eq. (8.53), hence, Eq. (8.56) should be modified accordingly. The intersti-
tial velocities of solids and gas are related to the superficial velocities of solids and gas,
respectively, by
v
(8.57a)
(1 - a)
uz = (8.57b)
a
It is noted that from the continuity equation, the superficial velocity of solids remains
constant since the cross-sectional area of the standpipe is unchanged and the solid material
is incompressible. However, the superficial gas velocity varies along the standpipe as a
result of the pressure variation and compressibility of gas. From Assumption (4), Uz, p,
and p at any location can be related to those at the inlet of the standpipe by
= = (8.58)
Uz Po Po
where subscript "0" denotes the location at the inlet of the standpipe (z = 0). Therefore,
Eqs. (8.57b) and (8.58) yield
uz = ^ (8.59)
ap
Since the particles are sliding down at the wall, we have
Tpw = /w^prw (8.60)
where / w is the coefficient of kinematic friction between the solids and the pipe wall, while
Oprw represents the normal stress in the solids on the wall surface (in the radial direction).
From Assumption (2), the normal stress in the radial direction is constant. Thus, aprw can
be related to apz by Eq. (8.11) as
apz (8.61)
where K is given by Eq. (8.10). Thus, substitution of Eqs. (8.57a), (8.59), (8.60), and (8.61)
into Eq. (8.56) gives
^ + ^ l + 4 y (8.62,
dz R J [ ap \ o t \
Equation (8.62) is the general momentum balance equation. For moving bed solids trans-
port in the standpipe, a is constant with a value of amb- Thus, Eq. (8.62) can be simpli-
fied to
2
* + J^ = ppg{1 _ amb) + P\U*PO _ "sJ (8 .63)
dz Rs [ambp 1-ambJ
It is noted that, in the preceding equation, the drag force represented by the second term
on the right-hand side may be positive or negative depending on the compressibility of gas
reflected in the value of p. Thus, <7pz may increase or decrease along the pipe depending on
the effect of the drag force. For a suspension flow mode of solids transport, crpz is negligibly
small. Consequently, Eq. (8.62) reduces to
g ( l - a)3 dz pvg(l -a)[ap 1-<

By equating the pressure gradient and the gas-particle drag force, the general momentum
balance for the gas phase can be expressed as
- T = P(UzP - uz) (8.65)

dz
In an alternative form, Eq. (8.65) becomes
dp
(8.66)
dz 11 a ap J
Equation (8.66) is directly applicable to the suspension flow mode of solids transport. For
a moving bed flow, Eq. (8.66) can be simplified to
(8.67)
C^mb ^mbP J
For suspension flows, solutions for a and p can be obtained by solving numerically
the governing equations, i.e., Eqs. (8.64) and (8.66). For moving bed flows, analytical
expressions for the solutions of the governing equations, i.e., Eqs. (8.63) and (8.67), can be
obtained. Integrating Eq. (8.67), p can be expressed in a dimensionless form
P
p* - 1 + a In = bz* (8.68)
Ll-a J
where
p* = ; z* = ; a = ~^~ ; b= (8.69)
Po H Uzvamb A ) ( l <*mb)
where H is the height of the standpipe.

p p Integrating
gg Eq.
q (8.63)
() yyields
U
z
*f fa M 1
where 1 + dl 1 exp(cx) dx } exp(-cz*) (8.70)

a* _ w L VP*2fwKH
c_
) \ j _^J po
Hpvg(\-amhY Rs ' pvgH(l-amh)
The bulk flux of solids in a standpipe depends not only on the geometry of the standpipe
but also on its feed devices and flow control valve, located at the top and bottom of the
standpipe, respectively. The flow pattern of solids in the pipe may vary with the pipe
distance, depending on the variation of the slip velocity in the pipe, wzp uz. The effect of
gas compression due to pressure increase may also be very significant. As a result, p varies
along the standpipe, which in turn changes a and, consequently, the slip velocity. Thus, for
a standpipe flow, it is possible that several flow patterns exist in various axial locations as
solids travel through the standpipe.
8.3.3 Hopper-Standpipe-Discharger Flow

As mentioned, the flow rate in a standpipe depends on the solid feed device as
well as the flow control valve. In this section, we discuss the gas-solid flows in a simple
standpipe system where the feed device is a mass flow hopper and the solid flow regulator
is a discharge orifice [Chen et al., 1984]. As shown in Fig. 8.15, the entrance of the vertical
standpipe is connected to a conical hopper feeder of half angle <pw, while a discharge orifice
is located at the bottom of the standpipe. Here, two representative solids flow patterns are
considered. One is a dilute suspension flow, and the other is a solid moving bed. In this
case, the following additional assumptions are needed:
(1) The angle of the feed hopper is small so that a continued flow of solids to the
standpipes can be established without creating a stagnant region in the hopper.
(2) The solid flow through the discharge orifice is characterized by the flow through a
conical converging channel due to the accumulation of solids on the orifice plate.
The half angle of the channel is the angle of friction for the particles, as shown in
Fig. 8.15.
(3) For a given type of solids, the control variables of the standpipe system are the
orifice opening and the pressure drop from the top of the hopper to the bottom of
the orifice.
The general system in Fig. 8.15 consists of six sections or regions: the hopper section,
the transition region I (between hopper and standpipe), the dilute transport region in the
standpipe, the moving bed region in the standpipe, the transition region II (between the
standpipe and discharge orifice), and the discharge section.
In the hopper section (1-2), the pressure drop is given by
where R2 = Rs/smcpw, and A and B are given by Eq. (8.48) with the subscript "i" repre-
senting "2." The velocities of gas and particles are given by Eqs. (8.44) and (8.45) with the
subscripts "i" and "o" representing "2" and "1," respectively.
l\
Hoppei flow \
M
A ^ -*^/ X r a n s i t i o n recrinn T
3
Suspension flow
Standpipe flow
Moving bed flow
5
^ j Transition region II
1
Discharge flow L M . Stagnant region of solids

i
7\ V
2rjk x Orifice
Figure 8.15. Configuration of hopper-standpipe-discharge system.
In the transition region I (2-3), the mass balance of solids requires
(1 - a 3 )w zp3 A s = - ( 1 - a2)uRv2Ach (8.73)
Note that wRp2 is the radial velocity in spherical coordinates while wzp3 is the axial velocity
in cylindrical coordinates; a 2 = mb; and A ch is the surface area of the spherical cap near
the exit of the hopper, whereas As represents the cross-sectional area of the standpipe. Ach
is related to As by
ACh 2(1 cos<pw)
(8.74)
As sin2 cpw
2(1 -
(8.75)
-a3 J sin2cpw
Similarly, the mass balance of gas (assumed incompressible in this region) requires
(8.76)
sin2
By neglecting the wall friction in the transition region, the momentum balance of solids in
the axial direction is given by
(1 - a3)pPzp3 + 0pz3 = (1 - <Xmb)Pp4p2 + apR2 (8.77)

where a pR2 can be obtained by integrating Eq. (8.51). The pressure drop over this region
can also be obtained from the momentum balance of gas as
p3 - Pi = amhpul2 - a3pu2z3 (8.78)
In the dilute transport region in the standpipe (3-4 in Fig. 8.15), the mass balance of
solids and gas requires
zn =
a
J
uz = ^ (8.80)
ap
The momentum equation of solids for suspension flow is given by
"ZP , ^ = 1 + rf r(wz - uzv) (8.81)

g(l-a)dz pp^(l-of)
and the momentum equation of gas is given by Eq. (8.65). Thus, phase velocities, pressure
distribution, and particle concentration are obtained from solving the coupling equations:
Eqs. (8.65), (8.79), (8.80), and (8.81).
In the moving bed region in the standpipe (4-5), the solids volume fraction remains
constant. The velocities of solids and gas are expressed by
=
1 -amb
uz = _i_ii_i (8.83)
OimbP
The pressure distribution in this region can be obtained from Eq. (8.68), and the normal
stress distribution is obtained from Eq. (8.70).
In transition region II (5-6), the mass balance of solids requires
w
zp5As = WRP6AC0 (8.84)
Note that wzp5 is the axial velocity in the cylindrical coordinate while WRP6 is the radial
velocity in the spherical coordinate; A& represents the cross-sectional area of the standpipe
whereas A co is the surface area of the spherical cup near the inlet of the discharge orifice
region. Aco is related to As by
Aeo = 2(1 - C O . ft)
As sin r]k
Thus,
=- = r-2 (8.86)
w R6 sin z rjk
By neglecting the wall friction in the transition region, the momentum balance of solids in
the axial direction is given by
(1 - amb)pvu2zv5 + apz5 = (1 - ofmb)PpWRp6 + <TPR6 (8.87)
The pressure drop over this region is obtained from the momentum balance of gas as
p6 - p5 = amhpu2z5 - Qf mb pw| 6 (8.88)
In the discharge orifice region (6-7 in Fig. 8.15), the pressure drop is given by
where Re = Rs/sinr]^ and /?7 = r d /sin%. rd is the radius of the discharge orifice. A and
B are given by Eq. (8.48) with the subscript "i" representing "6." The velocities of gas and
particles are given by Eqs. (8.44) and (8.45) with the subscripts "i" and "o" representing
"6" and "7," respectively.
Thus, the total pressure drop over the system is
6
Ap = Pl-p7 = ^ ( / ? i - pi+1) (8.90)
and the mass flow rates depend on the size of the discharge orifice.
8.3.4 Multiplicity of Steady Standpipe Flows

For a standpipe system, different flow patterns of steady-state flows may exist,
depending upon the ranges of operational parameters of the system. This phenomenon is
commonly known as the steady-state multiplicity. We introduce the general concepts of the
steady-state multiplicity based on the standpipe system in 8.3.3, in which the solids in the
feed hopper are kept in a steady moving bed motion.
There are four possible flow regimes for standpipe flows, as shown in Fig. 8.16. Regime 1
represents a moving bed flow mode throughout the standpipe. If a suspension flow occurs,
the gas volume fraction a increases with z as the particle accelerates away from the hopper
outlet until the gravity and drag forces are in balance. Then, a decreases as a result of gas
compression until amb is reached and the moving bed flow begins. If the location where a
reaches amb is in the region above the discharge cone, the flow is in Regime 2, while if it
is inside the discharge cone region, then the flow is in Regime 3. If it never reaches a mb
within the length of the standpipe, the flow is in Regime 4. The flow behavior in each of
the four regimes can thus be simulated using the model described in 8.3.3. Consequently,
the multiplicity of steady standpipe flows for this standpipe system can be characterized by
two parameters, namely, the ratio of discharge opening diameter to the standpipe diameter
y and the pressure drop over the system Ap.
For a system with a fixed y, the steady-state multiplicity may be revealed by the curve
of superficial velocity of solids Uv versus Ap, as illustrated in Fig. 8.17, where a hysteresis
loop is formed between two branch (upper and lower) curves [Chen et al., 1984]. The upper
branch corresponds to Regimes 2, 3, or 4, whereas the lower one corresponds to Regime 1.
At Ap less than (A/?)i, the system is operated on the upper branch of the curves, while it
is operated on the lower branch of the curves at Ap larger than (Ap)2. Between these two
U
Regime 1
u u
Regime 2 Regime 3
Figure 8.16. Flow regimes for standpipe without aeration (from Chen et al., 1984).
Regime 4
Suspension
transport
Moving bed
transport
Reversed gas flow

starts
Ap
(Ap), (Ap)2
Figure 8.17. Schematic illustration of multiplicity of steady standpipe flows (after Chen et al.,
1984).
values of Ap, the system can be operated at one of the multiple superficial velocities of
solids. Furthermore, the transition from Regime 1 to other regimes (or vice versa) may not
be continuous. For a system initially operated in Regime 1, as Ap decreases, U v follows
the lower curve until the point of (Ap)i where a sudden rise of Uv to the upper curve takes
place, indicating a sudden transition from Regime 1 to Regime 2 where a suspension section
in the standpipe occurs. In contrast, if the system initially has a section of suspension flow,
as Ap increases, Uv follows the upper curve until the point of (Ap)2, where a sudden drop
of Up to the lower curve takes place. This indicates a transition from Regime 2 to Regime 1
where the solids suddenly fill the suspension section into a moving bed. In addition, it is
noted that the standpipe system may not be operated at a very high Ap. The upper limit
(A/?)3 corresponds to the start of upward gas flow relative to the pipe wall.
In standpipe systems, aeration from the side is often implemented to provide a lubricating
effect for the solids flow. Aeration is also used in process applications to provide a stripping
gas to remove undesirable gas entrained by the solid particles in the standpipe flow. With
side aeration, the multiplicity of flows in the standpipe becomes much more complicated.
The flow patterns of solids and gas depend not only on Ap and y but also on the aeration
locations and aeration rates. For a single aeration point, the possible number offlowregimes
increases to 12 [Mountziaris and Jackson, 1991].
8.3.5 Leakage Flow of Gas in a Standpipe

Standpipes are used in the gas-solid circulating flow systems in which a cyclone-
standpipe-valve configuration for the separation and reintroduction of particles, as shown
in Fig. 8.18, is constantly encountered. The satisfactory operation of the cyclone-standpipe
system depends on a small leakageflowof gas up the standpipe. A simple method to estimate
Valve
Air + particles
Air
Pressure
Figure 8.18. Schematic configuration of a cyclone-standpipe-valve system.

this leakage flow of gas in the standpipe is delineated by Soo and Zhu (1993) as given in
the following.
Consider the flow of cohesionless material in the system. The particle mass flow through
a valve can be modeled as the flow of solids through an orifice where the Bernoulli relation
holds [Li et ai, 1982]. Thus the particle mass flow in the standpipe can be expressed as
(8.91)
where C o is an orifice coefficient, Ao is the valve opening area, As is the cross-sectional

area of the standpipe, and apmf is the volume fraction of particles at minimum fluidization.
When the volume fraction of particles in the standpipe is higher than 0.2, the pressure head
due to gravity in the standpipe can be estimated by [Chong et aL, 1987]
(8.92)
where i\> is the height of bulk solids in the standpipe.

The momentum equation of gas in the standpipe can be expressed as [Soo, 1989]
(1 - av)p (M + MP) - ^ = 0 (8.93)
where uv is downward and u is assumed upward, and is the relaxation time constant of
the momentum transfer from particles to gas, which is given by
L75ap(u + up)
(l-Op)H ^ '
where \i is the gas viscosity. For simplicity, we assume that the particle volume fraction in
the standpipe is constant. Substituting Eq. (8.92) into Eq. (8.93) yields
(8.95)
z
where A and B are given by
1.75(1- p )pJ p ' 1.75 V p ; ^ '

The validity of Eq. (8.95) is subjected to the condition that \u + wp| is less than the terminal
velocity of the particles. Equation (8.95) further indicates the condition at which the leakage
flow of gas downward along the standpipe occurs when
(8.97)
The typical pressure profile of the standpipe and the valve is shown in Fig. 8.18. As can
be seen in the figure, the pressure head over the standpipe A/?s, in the "absolute" sense, is
equal to the sum of the overall pressure head, A/?t, and the pressure drop over the valve,
Ap 0 . From Eq. (8.92), the particle volume fraction in the standpipe can be estimated
from the measured height of the bulk particles. Thus, given the particle mass flow rate, the
leakage flow of gas in the standpipe can be estimated from Eq. (8.95).
8.4 I Types of Standpipe Systems 361
Example 8.3 In an operation of gas-solid circulating flow in a cyclone-standpipe-valve

system, the particles are 100 [im glass beads with a density of 2,500 kg/m3. The particle
volume fraction in the standpipe is 0.55. The gas is air with a viscosity of 1.8 x 10~5 kg/m-s
and a density of 1.2 kg/m3. The particle mass flow is 70 kg/m2-s. The height of solids in
the standpipe is 1.4 m. The total pressure head over the standpipe and valve is 4,500 Pa.
Estimate the leakageflowof air in the standpipe. If the particle volume fraction at minimum
fluidization is 0.5 and the area ratio of valve opening to pipe cross section is 0.6, what is
the orifice coefficient of this valve?
Solution The interstitial velocity of particles in the standpipe is obtained by
p
pvotv 2,500x0.55 '
150ofp/x _ 150 x 0.55 x 1.8 x 10~5 _
~ 1.75(1 -av)pdv " 1.75 x 0.45 x 1.2 x 10~4 ~ ' m/S
(E8.8)
B = 11_z^( m , = f _ ^ ( 1 0 - 4 x 2 500x9.8 , = a 5 2 5 m 2 / s 2
1.75 V P J 1-75 V 1.2
Thus, the interstitial velocity of leakage flow of air is estimated from Eq. (8.95) as
+ -525 - 15 71+
- * X a 0 S 1 = -0.018 m/s
(E8.9)
The minus sign means that the leakage flow of air is directed downward.
The pressure drop over the valve is given by
Ap0 = avpvgzb - Apt = 0.55 x 2,500 x 9.8 x 1.4 - 4,500 = 14,365 Pa (E8.10)
Thus, on the basis of Eq. (8.91), the orifice coefficient of the valve can be calculated as
Co = otvpvuv I *
= 70 x (0.6V2 x 2,500 x 0.5 x 14, 365 ) ~ l = 0.019
8.4 Types of Standpipe Systems

Standpipes are applied to fluidized bed systems in several ways. The following
describes industrial applications of standpipes. An application of nonmechanical valves
which control the solids flow rates is also given.
8.4.1 Overflow and Underflow Standpipes

Standpipes are commonly grouped into overflow and underflow types, as illustrated
in Fig. 8.19. An overflow standpipe is structured with solids overflow from the top of a
I Pressure profile
1 in standpipe
Pressure profile
in column
0
AP,
o
P%
Pressure
(a)
Pressure profile
0 Pressure profile
in column
T
8
CO
a
Pressure
xxxxxxxxxxxxx>
(b)
Figure 8.19. Schematic diagrams and pressure distributions: (a) Overflow standpipe;
(b) Underflow standpipe (from Knowlton, 1986).
fluidized bed to a standpipe. On the other hand, a standpipe is termed an underflow type if
solids are introduced from the bottom of a bed.
In general, the outlet of an overflow standpipe is immersed into a bed in order to provide
an adequate hydrostatic head or seal pressure in the standpipe, as shown in Fig. 8.19(a).
The overflow pipe can be operated in either a moving bed or fluidized bed mode. Usually,
solids occupy only a fraction of the pipe near the standpipe outlet. The solidsflowrate in the
pipe depends on the overflow rate from the fluidized bed. Consequently, no valve is needed
at the pipe outlet. In the overflow standpipe system where solids transfer from one fluidized
bed to the other, the pressure drops across the lower fluidized bed, the grid (distributor),
the upper fluidized bed, and the height of solids in the standpipe. For a given standpipe
system and given particles, the pressure distribution profile depends on the gas velocity.
The difference in the pressure distribution reflects the difference in solids concentration,
bed height for the lower bed, solids concentration for the upper bed, and solids holdup
height in the standpipe.
Solids in an underflow standpipe are often discharged to a freeboard region of a lower
fluidized bed via a valve which regulates the solids flow rate, as illustrated in Fig. 8.19(b).
Solids in the standpipe are usually in a moving bed mode with an adequate height to provide
a seal pressure for the downward solids flow. If the solids size is not uniformly distributed in
the fluidized bed, the underflow pipe will basically discharge small particles. The pressure
distribution in Fig. 8.19(b) shows the pressure drops across the gas distributor in the upper
fluidized bed, underflow standpipe, discharge valve, and freeboard region at the bottom of
the standpipe. The gradients of the pressures across the distributor and standpipe are higher
than those of other parts of pressure distributions shown in the figure.
For an underflow standpipe system, the pressure distribution varies with the gas velocity.
By defining the slip velocity Us as Uv U, Kojabashian (1958) indicated that, because of
gas compression, when the decrease of voidage is greater than the increase of gas density, U v
decreases and U increases, and hence Us decreases. In a bubbling fluidized standpipe flow,
when Us decreases to a value less than the minimum fluidization velocity, defluidization in
the standpipe occurs. In this case, if the gas velocity relative to the solid velocity is upward
or Us is positive and directed upward, the pressure in the defluidized region increases
with decreasing pipe height, as illustrated in Fig. 8.20(a). On the other hand, if the gas
u
\ iu
1 1
Fluidized \
bed
flow Fluidized
\ v flow
Bubbles
u
Packed 11 I \ flow
bed \
flow !
\
\
\
/ \
(a) (b)
Figure 8.20. Pressure profiles in an underflow standpipe (from Knowlton, 1986): (a) Without
pressure reversal; (b) With pressure reversal.
velocity relative to the solid velocity is downward or Us is negative and directed downward,
the pressure in the defluidized region decreases with decreasing pipe height (i.e., pressure
reversal occurs), as shown in Fig. 8.20(b).
In Fig. 8.20(b), there are two distinctive regions for pressure distributions, namely, a
standpipe flow condition (i.e., solids flow against the gas pressure gradient) and a discharge
flow condition (i.e., solids flow coinciding with the gas pressure gradient). However, con-
sidering the overall gas pressure gradient between the inlet and outlet, the overall pipe flow
is a standpipe flow. It is noted that the solids flow needs to overcome the pressure drop
further through the flow control valve. Hence, it is necessary to have the overall pressure
buildup sufficiently high to yield a sealing effect which prohibits the gas from flowing back
into the standpipe.
The pressure drop in the overflow and underflow standpipes can be predicted using Eqs.
(8.64) through (8.67) considering either the moving bed flow or the suspension flow. The
pressure drops across the fluidized bed and grid or distributor can be evaluated, respectively,
by Eqs. (9.7) and (P9.7) given in Chapter 9.
8.4.2 Inclined Standpipe and Nonmechanical Valves

In practice, it is common to have standpipes with inclined configurations.
Figure 8.21 illustrates two examples of applications using an inclined standpipe for solids
transport to a fluidized bed. Figure 8.21(a) shows a fluidized bed with an inclined pipe as
a solid feeder connected to a vertical standpipe. The system operates with some gas, in
bubble form, bypassing the fluidized bed to the standpipe. Gas bubbles move upward from
the upper part of the inclined pipe while a packed bed in segregated flow moves downward
along the lower part of the pipe. Figure 8.21(b), on the other hand, shows a fluidized bed
with a vertical pipe as a solid feeder connected to an inclined standpipe. The vertical pipe
provides a higher seal pressure for the downward flow than that in Fig. 8.21(a).
CD
0
(b)
Figure 8.21. Two applications of angled standpipe for solids flow to a fluidized bed (from
Knowlton, 1986): (a) With segregated gas bubble flow; (b) Without segregated gas bubble flow.
L-valve
opening Aeration
A
L-valve Stagnant
cross section solids
(a)
Aeration
(c)
Figure 8.22. Nonmechanical valves and loopseal: (a) L-valve; (b) J-valve; (c) Loopseal.
In standpipe systems, the solids flow rate can be controlled by using nonmechanical
valves {e.g., L-valve and J-valve), as shown in Fig. 8.22(a) and (b). In such arrangements,
only aeration gas is needed to control the solids flow. Arrangements of this nature are known
as nonmechanical valves; as the name implies, they use nonmechanical means to control
the solids flow rate. By adjusting the gas aeration rate, the solids flow into a dense or dilute
environment can be controlled [Knowlton, 1986]. Using an L-valve for low solids flow
rates, a stagnant solids region will form and the size of the region will reduce as the solids
flow rate increases, as shown in Fig. 8.23. The loopseal shown in Fig. 8.22(c) provides a
viable means for gas seal when solids are transferred to a vessel. This is achieved by moving
up the solids in a short standpipe at the end of the L-valve and introducing the solids to the
vessel by an inclined pipe. The solids flow rate can also be controlled by aeration along the
loopseal.
The L-valve is the most commonly used nonmechanical valve. The pressure drop across
the L-valve can be evaluated by Eqs. (10.9) and (10.10), given in Chapter 10.
Low
solids
flow
Stagnant solids
Aeration
Medium
solids
flow
Stagnant solids
Aeration
High
solids
flow
Stagnant solids
Figure 8.23. Stagnant regions in an L-valve for three solids flow rates (from Knowlton, 1986).
Nomenclature
A Parameter, defined by Eq. (8.48) B Ratio of vertical shear stress at
A Parameter, defined by Eq. wall to vertical normal stress at
(8.96) wall
Ao Valve opening area B Parameter, defined by Eq. (8.48)
Ach Surface area of spherical cap at B Parameter, defined by Eq. (8.96)
outlet from feed hopper Co Orifice coefficient
Aco Surface area of spherical cap at c Cohesion
inlet from discharge cone D Column diameter
As Cross-sectional area of "min Minimum outlet dimension of
standpipe the hopper
Nomenclature 367
dP Particle diameter uz Gas velocity in the axial

F Body force vector direction
FD Drag force ^zp Particle velocity in the axial
FF Material flow function direction
Gas circumferential velocity
ff Flow factor LI m
component
A Kinematic coefficient of friction
Particle circumferential velocity
A Static coefficient of friction
/w Coefficient of friction at wall
Yo Unconfined yield strength
8 Gravitational acceleration Zb Height of solids in standpipe
H Height of standpipe
Hi?*,) Function, defined graphically in
Fig. 8.12 Greek Symbols
K3 Principal stress ratio in an active
a Volume fraction of the gas phase
state
a Volume fraction of gas in a
Kp Principal stress ratio in a moving bed
passive state Volume fraction of the particle
P
N Parameter, defined by Eq. (8.19) phase
or Eq. (8.23)
n Modified Richardson-Zaki tfpmf Volume fraction of solids at
index for gas-solid systems minimum fluidization
Pressure P Coefficient for the drag between
P
the particulate phase and
P* Dimensionless pressure interstitial gas
Q Gas flow rate
P Angle between x-axis and the
Rs Radius of standpipe normal to an inclined plane
r* Radius of discharge orifice A* Angle for principal directions
T Stress tensor Y Angle between the major
U Superficial gas velocity principal stress and hopper wall
Up Superficial particle velocity Y Ratio of discharge opening
f/pt Particle terminal velocity diameter to standpipe diameter
f/zp Superficial particle velocity in A Ratio of vertical normal stress at
one-dimensional axial flow wall to averaged vertical normal
uR Superficial gas velocity in stress
one-dimensional radial Apo Pressure drop over orifice
flow Aps Pressure drop over standpipe
Us Slip velocity Apt Overall pressure drop
uz Superficial gas velocity in rj Angle of internal friction of
one-dimensional axial flow solids
U Interstitial gas velocity % Kinematic angle of internal
uv Particle velocity friction of solids
UR Gas radial velocity component Static angle of internal friction
URV Particle radial velocity of solids
component Kinematic angle of wall friction
Relaxation constant, defined by &h Horizontal normal stress

Eq. (8.94) am M e a n stress
p Density of gas G0 Surcharge on top of the hopper
pp Density of particles crt Tensile strength
a Normal stress <J V Vertical normal stress
cra Major principal stress in a r Shear stress
stable arc which spans the li Dynamic viscosity of gas
hopper outlet v Poisson's ratio
aac Critical major principal stress in cpw Hopper half-apex angle
a stable arc which spans the i/r Angle between R-direction and
hopper outlet the normal to the major
<7i Principal stress (i = 1, 2, or 3) principal stress axis
References
Berg, C. H. O. (assigned to Union Oil Company of California) (1954). U.S. Patent 2,684,868.
BMHB (British Material Handling Board) (1988). Draft Design Code of Silos, Bins, Bunkers and
Hoppers. Berkshire: Milton Keynes.
Carr, R. L. (1965). Evaluating Flow Properties of Solids. Chemical Engineering, 72 (2), 163.
Chen, Y. M., Rangachari, S. and Jackson, R. (1984). Theoretical and Experimental Investigation of
Fluid and Particle Flow in a Vertical Standpipe. I&EC Fund., 23, 354.
Chong, Y. O., Teo, C. S. and Leung, L. S. (1987). Recent Advances in Standpipe Flow. / . Pipelines,
6,121.
Cleaver, J. A. S. and Nedderman, R. M. (1993). Theoretical Prediction of Stress and Velocity
Profiles in Conical Hoppers. Chem. Eng. Sci., 48, 3693.
Drescher, A. (1991). Analytical Methods in Bin-Load Analysis. Amsterdam: Elsevier.
Ginestra, J.C., Rangachari, S. and Jackson, R. (1980). A One-Dimensional Theory of Flow in a
Vertical Standpipe. Powder Tech., 27, 69.
Growther, M. E. and Whitehead, J. C. (1978). Fluidization of Fine Particles at Elevated Pressure. In
Fluidization. Ed. Davidson and Kearins. Cambridge: Cambridge University Press.
Home, R. M. and Nedderman, R. M. (1978). Stress Distribution in Hoppers. Powder Tech., 19, 243.
Jackson, R. (1993). Gas-Solid Flow in Pipes. In Paniculate Two Phase Flow. Ed. M. C. Roco.
Boston: Butterworth-Heinemann.
Janssen, H. A. (1895). Versuche iider Getreidedruck in Silozellen. Vercin Deutcher Ingenieure, 39,
1045.
Jenike, A. W. (1961). Gravity Flow of Bulk Solids. Bulletin No. 108, Utah Engineering Experiment
Station.
Jenike, A. W. (1964a). Steady Gravity Flow of Frictional-Cohesive Solids in Converging Channels.
Trans. ASME, J. Appl. Mech., 31, 5.
Jenike, A. W. (1964b). Storage and Flow of Solids. Bulletin No. 123, Utah Engineering Experiment
Station.
Johanson, J. R. (1981). How to Predict and Use the Properties of Bulk Solids. In Solids Handling.
Ed. McNaughton and the Staff of Chemical Engineering. New York: McGraw-Hill.
Kamath, S., Puri, V. M., Manbeck, H. B. and Hogg, R. (1993). Flow Properties of Powders Using
Four Testers: Measurement, Comparison and Assessment. Powder Tech., 76, 277.
Problems 369
Knowlton, T. M. (1986). Solids Transfer in Fluidized Systems. In Gas Fluidization Technology.

Ed. D. Geldart. New York: Wiley.
Knowlton, T. M., Findlay, J. G., Sishtla, C. and Chan, I. (1986). Solids Pressure Reduction Without
Lockhoppers: The Restricted Pipe Discharge System. 1986 AIChE Annual Meeting,
Miami, Fla.: Nov. 2-7.
Knowlton, T. M., Findlay, J. G. and Chan, I. (1989). Continuous Depressurization of Solids Using a
Restricted Pipe Discharge System. 1989 AIChE Annual Meeting, San Francisco, Calif.:
Nov. 5-10.
Kojabashian, C. (1958). Properties of Dense-Phase Fluidized Solids in Vertical Downflow. Ph.D.
Dissertation. Massachusetts Institute of Technology.
Li, X. G., Liu, D. L. and Kwauk, M. (1982). Pneumatically Controlled Multistage Fluidized Beds.
II. Proceedings of Joint Meeting of Chemical Engineers, SIESC and AIChE. Beijing:
Chemical Industry Press.
Mountziaris, T. J. and Jackson, R. (1991). The Effects of Aeration on the Gravity Flow of Particles
and Gas in Vertical Standpipes. Chem. Eng. Sci., 46, 381.
Nedderman, R. M. (1992). Statics and Kinematics of Granular Materials. Cambridge: Cambridge
University Press.
Nedderman, R. M. and Tiiziin, U. (1979). A Kinematic Model for the Flow of Granular Materials.
Nguyen, T. V., Brennen, C. and Sabersky, R. H. (1979). Gravity Flow of Granular Materials in
Conical Hoppers. Trans. ASME, J. Appl. Mech., 46, 529.
Peschl, I. A. S. Z. (1989). Measurement and Evaluation of Mechanical Properties of Powders.
Powder Handling & Processing, 1, 135.
Richardson, J. F. and Zaki, W. N. (1954). Sedimentation and Fluidization. Trans. Instn. Chem.
Engrs., 32, 35.
Sokolovskii, V. V. (1965). Statics of Granular Media. New York: Pergamon Press.
Hemisphere.
Soo, S. L. and Zhu, C. (1993). Unsteady Motion of Dense Suspensions and Rheological Behavior.
In Paniculate Two-Phase Flow. Ed. M. C. Roco. Boston: Butterworth-Heinemann.
Walker, D. M. (1966). An Approximate Theory for Pressures and Arching in Hoppers. Chem. Eng.
Sci., 21, 975.
Walters, J. K. (1973). A Theoretical Analysis of Stresses in Axially-Symmetric Hoppers and
Bunkers. Chem. Eng. Sci., 28, 779.
Problems
8.1 A cohesionless material of weight density 12 kN/m3 and an angle of internal friction of 30
is stored in a tall cylindrical bin of diameter 1.8 m and an angle of wall friction 22. Use
Janssen's formula to determine the depths at which the radial stress in the material reaches
70 percent of its maximum value for both active and passive cases.
8.2 A conical hopper of a half angle 20 and an angle of wall friction 25 is used to store a
cohesionless material of bulk density 1,900 kg/m3 and an angle of internal friction 45. The
top surface of the material lies at a level 3.0 m above the apex and is free of loads. Apply
Walker's method to determine the normal and shear stresses on the wall at a height of 1.2 m
above the apex if the angle between the major principal plane at that height and the hopper
wall is 30. Assume a distribution factor of 1.1.
8.3 Considering the equilibrium of a stable arch, show that the minimum diameter of the orifice
in a conical hopper, dmin, which is defined to account for the arching, can be related to the
unconfined yield stress of the bulk material by the relation
4 = 2Y 0/apppg (P8.1)
Neglect the force exerted on the arch by the weight of the material above the arch in the
equilibrium equation.
8.4 It is required to design a mass flow conical hopper with the volume capacity of 100 m3 to store
a cohesionless material of bulk density 1,700 kg/m3 and an angle of internal friction 40.
Four sets of shear tests have been conducted on the material and results for the unconfined
yield strength and the corresponding consolidating stress are as follows:
(1) (2) (3) (4)
7 0 (kN/m 2 ) 0.90 1.35 1.75 1.95

ax (kN/m2) 1.10 1.75 2.75 3.25
Determine the half angle of the hopper and the minimum outlet diameter for the two cases
where the hopper wall material has an angle of wall friction of (i) 10 and (ii) 30.
8.5 Consider a conical hopper flow where there is no-slip between the gas and the particle.
Determine the distribution of mean stress of particles in the hopper flow. The particles can
be assumed to be in a moving bed condition.
8.6 For a fluidized standpipe, the drag force of particles balances the pressure head as a result of
the weight of solids. If the Richardson and Zaki form of equation (Eq. (8.55)) is proposed
for the drag force, derive an expression for the leakageflowof gas in this standpipe. Discuss
the effect of particle size on the leakage flow assuming all other conditions are maintained
constant.
CHAPTER 9
Dense-Phase Fluidized Beds
9.1 Introduction
Contact schemes of gas-solid systems can be classified on the basis of the state of
solids motion. For a batch-solids system, the gas at a low velocity merely percolates through
the voids between packed particles while the particles remain motionless. The solids in this
case are in the fixed bed state. With an increase in the gas velocity, particles move apart
and become suspended; the bed then enters the gas fluidization state. Under relatively low
gas velocities, a dense gas-solid suspension characterizes the bed; on the bed surface, some
solids entrainment occurs. Gas bubbles or voids of bubble shapes can usually be clearly
distinguished in the dense suspension. Under relatively high gas velocities, significant
solids entrainment occurs and bubbles or voids are much less distinguishable in the dense
suspension. These states of the beds are referred to as dense-phase fluidization. Further
increase in the gas velocity will intensify the solids entrainment and the dense bed surface
gradually becomes undistinguishable. At high gas velocities, particles are fully entrained,
and the whole bed transforms to a "lean" gas-solid suspension. This state of fluidization
is maintained by means of external recirculation of the entrained particles. Such operation
is termed lean-phase fluidization. This chapter focuses on dense-phase fluidization. Lean-
phase fluidization or transport is discussed in Chapters 10 and 11. As indicated in the
Preface, the correlation equations presented in the text, unless otherwise noted, are given
in SI units.
Dense-phase fluidized beds with bubbles represent the majority of the operating inter-
ests although the beds may also be operated without bubbles. The bubbling dense-phase
fluidized bed behavior is fluidlike. The analogy between the bubble behavior in gas-solid
fluidized beds and that in gas-liquid bubble columns is often applied. Dense-phase flu-
idized beds generally possess the following characteristics, which promote their use in
reactor applications:
(1) Capability of continuous operation and transport of solids in and out of the bed
(2) High heat transfer rates from bed to surface and from gas to particle leading to
temperature uniformity in the bed
(3) High mass transfer rates from gas to particle
(4) Applicability over a wide range of particle properties and high solids mixing rates
(5) Simplicity in geometric configuration and suitability for large-scale operation
Examples of applications of gas-solid dense-phase fluidization are given in Table 9.1.
9.2 Particle and Regime Classifications and Fluidized

Bed Components
The transport behavior of fluidized bed systems can be described in light of the
properties of the fluidized particles and the flow regimes. In the following, particle and
regime classifications along with general components in a fluidized bed are given.
371
372 9 / Dense-Phase Fluidized Beds
Table 9.1. Examples of Applications of Gas-Solid Dense-Phase Fluidization
Types of applications
Physical Metallurgical and

operations Chemical syntheses mineral processes Other applications
Heat exchange Phthalic anhydride synthesis Uranium processing Coal combustion

Solids blending Acrylonitrile synthesis Reduction of iron Coal gasification
oxide
Particle coating Maleic anhydride synthesis Pyrolysis of oil shale Fluidized catalytic
cracking
Drying Ethylene dichloride synthesis Roasting of sulfide Incineration of
ores solid waste
Adsorption Methanol to gasoline process Crystalline silicon Cement clinker
production production
Solidification Vinyl acetate synthesis Titanium dioxide Microorganism
& granulation production cultivation
Particle growth Olefin polymerization Calcination
6,000
D
4,000 Spoutable
^ 2,000
Sf A
c 1,000 Aeratable
i
" 500
C
Cohesive
100 , ,,, I I
lO" 5 10" 4
,m
Figure 9.1. Geldart's classification offluidizedparticles (from Geldart, 1973).
9.2.1 Classification of Fluidized Particles

Particles can be classified into four groups (i.e., Groups A, B, C, and D) based on
their fluidization behavior [Geldart, 1973]. This classification, known as Geldart's classifi-
cation, is shown in Fig. 9.1, where particles are classified in terms of the density difference
between the particles and the gas (p p p), and the average particle diameter dv.
Figure 9.1 was obtained empirically and has been widely adopted in the fundamental
research and design of gas-solid fluidized beds. However, it should be noted that some
fluidization variables other than (p p p) and dv may alter the demarcation of particle
9.2 I Particle and Regime Classifications and Fluidized Bed Components 373
groups. For example, for fluidization using different gases, the cohesive nature of fluidized
particles due to the surface or interparticle contact forces can vary significantly because
of the difference in the type of gases adsorbed on the particles. Likewise, gas velocity,
temperature, and pressure may also dramatically affect the fluidized particle characteristics,
which in turn alter the particle group demarcation.
Group C in Fig. 9.1 comprises small particles (dp < 20 /im) which are cohesive. In
fluidization with Group C particles, the interparticle contact forces such as the van der
Waals force, capillary forces, and electrostatic forces dominate the hydrodynamic forces.
The interparticle contact forces are affected by particle properties such as hardness, electrical
conductivity, magnetic susceptibility, surface asperities, and moisture content, as well as
by gas properties. Group C particles are difficult to fluidize and gas channeling is the
most common characteristic when fluidizing these particles. When Group C particles are
fluidized, the bed expansion can be very high.
Group,A particles, with a typical size range of 30-100 //,m, are readily fluidized. In
fluidization with Group A particles, the hydrodynamic forces are important; the interparticle
contact forces may also play a significant role. Beds with Group A particles can be operated
in both the particulate fluidization regime where bubbles are not present, and the bubbling
fluidization regime where bubbles are present (see 9.2.2 for regime classification). Thus,
for Group A particles, the gas velocity for minimum fluidization, (/ mf, is smaller than that
for the minimum bubbling, f/mb. For bubbling fluidization with Group A particles, there
exists a maximum stable bubble size.
For Group B particles, there is no particulate fluidization regime. In this case, U mf equals
Umh. The bubble size increases with the bed height and bed expansion is moderate. For
Group B particle fluidization, there exists no maximum stable bubble size.
Group D comprises coarse particles (dp > 1 mm) which are commonly processed by
spouting. When Group D particles are fluidized, the bed expansion is low and the particle
mixing is not as good as that for Group A and B particles.
For demarcation between Group C and A particles, Molerus (1982) proposed a semiem-
pirical criterion based on a balance between the van der Waals force and the hydrodynamic
forces as given by
(9.1)
where F H is the cohesive force between particles with a value ranging from 8.8 x 10~8 N
(for hard material) to 3.7 x 10~7 N (for soft material) and K is a constant (K = 0.01). The
subscript CA refers to the C-A boundary.
For demarcation between Group A and B particles, an empirical equation with air as the
fluidizing gas under ambient conditions was suggested by Geldart (1973):
[(pp-p)Ll7dp]AB = 0.906 (9.2)
Equation (9.2) can be modified to apply to elevated pressures and temperatures as well as
to different gases as given by [Grace, 1986]
-1.275
(9.3)
where the subscript AB represents the A-B boundary; and Ar is the Archimedes number
(seeEq. (9.12)).
Increasing gas velocity
Fine
Fixed particles Particulate Bubbling Turbulent Fast Dilute
bed fluidization fluidization fluidization fluidization transport
Coarse particles (U^ = I

Small D and/or
large Htl D ratio
Slugging
Cohesive powder Channeling

flow
Dense-phase fluidization _ Lean-phase fluidization _ J
Figure 9.2. Interrelationship of various regimes including fixed bed, dense-phase fluidization,
and lean-phase fluidization.
For demarcation between Group B and D particles, Grace (1986) considered the fast
bubble characteristics for Group B particles in which the bubble rise velocity is higher
than the interstitial gas velocity, and the slow bubble characteristics for Group D par-
ticles in which the bubble rise velocity is lower than the interstitial gas velocity (see
9.4.2). Thus, by setting the bubble rise velocity and the interstitial gas velocity equal,
a criterion for demarcation between Group B and D particles can be obtained as [Grace,
1986]
P/(P P ~ P)
ArBD = 1.581 x 107 for P P - P >219 (9.4)
[1 - 219p/(p p - p)]2
9.2.2 Fluidization Regimes

Regime classification for dense-phase fluidization in general can be based on bub-
ble behavior. Figure 9.2 shows the interrelationship of the regimes in dense-phase fluidiza-
tion with those in lean-phase fluidization, which covers fast fluidization (Chapter 10) and
dilute transport (Chapter 11). Dense-fluidization regimes include particulate fluidization,
bubbling fluidization, and turbulent fluidization. In a broad sense, "dense-phase fluidiza-
tion" also encompasses the slugging, spouting, and channeling regimes or conditions of
operation. The fundamental distinction between these regimes or conditions of operation
is briefly described later.
The particulate fluidization regime, as shown in Fig. 9.3(a), is bounded by the minimum
fluidization velocity and the minimum bubbling velocity. In particulate fluidization, all the
gas passes through the interstitial space between the fluidizing particles without forming
bubbles. The bed appears homogeneous. This regime exists only in a bed with Group A
particles under a narrow operating range of gas velocities. At high pressures or with gases
of high density, the operating range of this regime expands. In a bed of coarse particles
9.2 I Particle and Regime Classifications and Fluidized Bed Components 375
: :O '
t
Gas
(b)
T t r
Gas
Gas
(f)
(d)
Figure 9.3. Various flow regimes or patterns in dense-phase fluidization: (a) Particulate
fluidization; (b) Bubbling fluidization; (c) Turbulent fluidization; (d) Slugging; (e) Spouting;
(f) Channeling.
(Group B and D particles), bubbles appear as soon as the gas velocity reaches Umf and,
thus, the particulate regime does not exist for these particles.
The bubbling fluidization regime, as shown in Fig. 9.3(b), is reached with an increase in
the gas velocity beyond Um\y. Bubbles form and induce vigorous motion of the particles. In
the bubbling fluidization regime, bubble coalescence and breakup take place. With increas-
ing gas velocity, the tendency of bubble coalescence is enhanced. Two distinct phases, i.e.,
the bubble phase and the emulsion phase, are present in this regime.
The turbulent fluidization regime, as shown in Fig. 9.3(c), is realized when the gas
velocity increases beyond the bubbling fluidization regime. In the turbulent fluidization
regime, the bubble and emulsion phases may become indistinguishable with increased
uniformity of the suspension. In contrast to that in the bubbling regime, the tendency
for bubble breakup is enhanced as the gas velocity increases in this regime, resulting in
reduced presence of large bubbles. In addition, the surface of the bed becomes very diffused.
Increased freeboard activities are evidenced in this regime.
Slugging, as shown in Fig. 9.3(d), refers to thefluidizationregime which is characterized
by the presence of bubbles or slugs with sizes comparable to the bed diameter. Slugging
occurs only in a fluidized bed of a small bed diameter and/or large height/diameter ratio.
Slugs are formed by the coalescence of bubbles/voids. Their rise velocity varies with the
bed diameter. As slugs break up under sufficiently high gas velocities, the slugging regime
transforms to the turbulent fluidization regime or fast fluidization regime. This regime may
also be present in other fluidized vessels such as standpipes.
Spouting occurs, as shown in Fig. 9.3(e), when the gas is injected vertically at a high
velocity through a small opening into the bed, typically one containing Group D particles. At
the center of the spouted bed, the gas jet penetrates the whole bed and carries some particles
upward, forming a dilute flow core region. Particles, reaching the top of the core region,
fall back to the top of the annular dense region located between the core region and the wall.
Particles in the annular region move downward in a moving-bed mode and recirculate to the
core region, forming a solids circulatory pattern. Thus, the particle mixing in a spouted bed,
as induced by the axial spout, is more regular and cyclic than that in bubbling or turbulent
fluidized beds.
Channeling, as shown in Fig. 9.3(f), is caused typically by aggregation effects of cohesive
(Group C) particles due to interparticle contact forces. In addition to the particle size and
density, the particle shape may have effects on channeling behavior. Channeling can also
occur in a bed of noncohesive particles when gas distribution is nonuniform across the
distributor.
The presence of internals in a bed, such as baffles and heat exchangers, or variation
in geometric configurations, such as tapered geometry, can affect the flow behavior. The
classification scheme in Fig. 9.3 is, however, generally applicable to these situations.
9.2.3 Components in a Dense-Phase Fluidized Bed

Bed components in a dense-phasefluidizedbed include the gas distributor, cyclone,
dipleg, heat exchanger, expanded section, and baffles, as shown in Fig. 9.4. The basic
function of each component in a fluidized bed is described in the following.
The gas distributor yields a desired distribution of fluidizing gas and supports particles in
the bed. Generally, a minimum pressure drop across the distributor is required to ensure the
uniformity of the gas distribution in the bed [Problem 9.11]. Gas distributors can be designed
in many ways; examples are sandwiched, stagger perforated, dished, grated, porous, tuyere,
and cap types [Kunii and Levenspiel, 1991]. The initial bubble size from the distributor
strongly varies with the distributor design. The initial bubble size may significantly affect
the bubble size in the bed.
The cyclone separates solid particles from the outlet gas. Several cyclones may be com-
bined to form a multistage cyclone system. More than one multistage cyclone system can
be placed inside or outside the bed. Other types of gas-solid separators {e.g., filters) are
also used for gas-solid separation although the cyclone is most widely adopted in fluidized
beds. The dipleg/standpipe returns the particles separated by the cyclone into the dense bed.
The outlet of a dipleg may be located in the freeboard or immersed in the dense bed.
The heat exchanger removes generated heat from or adds required heat to the fluidized
bed by a cooling fluid or heating fluid. Heat exchangers can either be immersed in the dense
bed or be placed both in the dense bed and in the freeboard, which is the region above the
dense bed. Possible erosion and corrosion of heat exchangers require special attention.
The expanded section reduces the linear gas velocity in the freeboard so that settling of
the particles carried by the fluidizing gas can be efficiently achieved. The expanded section
92 I Particle and Regime Classifications and Fluidized Bed Components 317
outlet gas
cyclone
expanded section
dipleg
freeboard
surface of
dense bed baffles
emulsion phase
cooling (heating)
fluid dense bed
bubble
heat exchanger
^ gas distributor
plenum
t
inlet gas
Figure 9.4. Schematic diagram of a dense-phasefluidizedbed.
may not always be necessary, depending on the operating conditions and design of the
gas-solid separator.
The baffles restrictflow,enhance the breakup of bubbles, promote gas-solid contact, and
reduce particle entrainment. Commonly, baffles refer to any internals other than diplegs and
immersed heat exchangers, although the latter may also perform the functions of baffles.
Like heat exchangers, baffles can be installed in the dense bed and/or in the freeboard.
Baffles can be designed with a number of variations, e.g., horizontal and vertical grates,
fins of different sizes and directions, solid or mesh, or even constructed with a shape like a
pagoda [Jin et al, 1980]. The benefits of baffles are more distinct for beds of coarse particles
(Groups B and D) than for fine particle beds (Group A), because the bubbles in the latter
case are smaller.
Dense-phase fluidization can also be conducted in the presence of force fields other than a
gravitational field. Such force fields include vibrational, acoustic, centrifugal, and magnetic
fields. Operations with applications of thesefieldsare known, respectively, as vibrofluidized
beds [Mori et al, 1992], acoustic fluidized beds [Montz et al, 1988; Chirone et al, 1992],
378 91 Dense-Phase Fluidized Beds
centrifugal fluidized beds [Kroger et al., 1979] (see Problem 9.1), and magnetofluidized
beds [Rosensweig, 1979; Liu etaU 1991] (see Problem 9.2).
9.3 Minimum Fluidization and Particulate Fluidization

The state of fluidization begins at the point of minimum fluidization. This section
provides a detailed account of the characteristics of minimum fluidization and particulate
fluidization.
9.3.1 Minimum Fluidization

Incipient or minimum fluidization can be characterized by the relationship of the
dynamic pressure drop, Ap^, to the gas velocity. The dynamic pressure gradient, dp^/dH,
can be related to the total pressure gradient, dp/dH, by
dpd
(9.5)
'dH
For a gas-solid system, pg is negligibly small compared to (dp/dH). Consequently,
dp d /d// can be approximated by dp/dH. The relationship of pressure drop through the
bed, A/?b, and superficial gas velocity U for fluidization with uniform particles is illustrated
in Fig. 9.5. In the figure, as U increases in the packed bed, A/?b increases, reaches a peak,
and then drops to a constant. As U decreases from the constant A/?b, A/?b follows a different
path without passing through the peak. The peak under which the bed is operated is denoted
the minimum fluidization condition, and its corresponding superficial gas velocity is defined
as the minimum fluidization velocity, Umf.
For particles with nonuniform properties, the hysteresis effect noted previously may also
occur; however, the transition of Apb from the fixed bed to the fluidized bed is smoother
and Umf can be obtained from the interception of the A/?b li n e fr the fixed bed with that for
the fluidized bed. Thus, the expression for Umf can be analytically obtained on the basis of
Fluidized
bed
U
Figure 9.5. Determination of Umf from the pressure drop variation with the gas velocity.
9.3 I Minimum Fluidization and Paniculate Fluidization 379
the equivalence of the pressure drop for a fixed bed and that for a fluidized bed under the
minimum fluidization condition.
The pressure drop in the fixed bed can be described by the Ergun equation, as given by
Eq. (5.377). Under the minimum fluidization condition, we have
1 n
where (p is the sphericity of the particles and Hmf is the bed height at minimum fluidization.
It is noted that most fluidized particles are not spherical, and for most cases, the equivalent
particle diameter refers to the volume diameter (Eq. (1.3)). For a bed with a wide range of
particle sizes, DeBroucker's mean diameter (Eq. (1.40)) may be used for Eq. (9.6).
In a fully fluidized bed, the pressure drop (cross-sectionally averaged) counterbalances
the weight of the pseudocontinuum of the gas-solid mixture, which yields
(9.7)
Usually, the bed voidage a is considered independent of the bed height and is given by
J (9 8)
-
where Mp is the total mass of particles in the bed; A is the cross-sectional area of the bed;
and Hf is the expanded bed height.
From Eqs. (9.5) and (9.7), we have
^ v (9.9)
Under the minimum fluidization condition, Eq. (9.9) gives
amf)g (9.10)
Equating Eqs. (9.6) and (9.10) results in
<L^!> H l ; f (9.1.)
where the Archimedes number (Ar) and the particle Reynolds number at minimum fluidiza-
tion (Repmf) are defined as
Ar =
p(pv ~ tf
and
nils /7_
(9.13)
Equation (9.11) can be written as Ar = &iRepmf + A^RCpmf- In terms of this form, a fre-
quently used semiempirical correlation relating Repmf to Ar is given by [Wen and Yu,
1966]
Repmf = \/(33.7) 2 + 0.0408Ar - 33.7 (9.14)
Ht9 m
Particle: 60 urn ballotini
0.05 U, m/s
Figure 9.6. Bed expansion curve for typical Group A particles (from Geldart, 1986).
Following a similar approach, other correlations have been proposed [e.g., Babu et ah,
1978; Thonglimp, 1981; Grace, 1982].
9.3.2 Particulate Fluidization

For a bed with Group A particles, bubbles do not form when the gas velocity reaches
Umf. The bed enters the particulate fluidization regime under this condition. The operation
under the particulate fluidization regime is characterized by a smooth bed expansion with
an apparent uniform bed structure for Umf < U < Umh, where /mb is the superficial gas
velocity at the minimum bubbling condition. The height of the bed expansion in terms of
a can be estimated by [Abrahamsen and Geldart, 1980a]
a3 (pP - p)gdl - p)gdl
= 2W(U - Umf) (9.15)
- OL /x 1 - amf ix
The gas velocity for the upper bound of the particulate fluidization regime is /mb. The
operational range of the particulate fluidization regime is quite narrow, as shown in Fig. 9.6.
It is seen that for U > Umf, //f monotonically increases with U until the minimum bubbling
point is reached. Further increase in the gas velocity leads to a decrease in the bed height
as bubbling fluidization begins.
The particulate fluidized regime will occur when the following criterion is met [Wilhelm
and Kwauk, 1948]
Fr m f <0.13 (9.16)
where the Froude number at minimum fluidization Frmf is defined by
Frmf = (9.17)
gdP
9.4 I Bubbling Fluidization 381
Considering more parameters, an improved criterion was suggested by Romero and

Johanson (1962) as
(^)^100 (9.18)
9.4 Bubbling Fluidization

In a bubbling fluidized bed, the behavior of bubbles significantly affects the flow
or transport phenomena in the bed, including solids mixing, entrainment, and heat and
mass transfer. Theories accounting for the multibubble behavior are primarily based on
those for single-bubble behavior. Hence, in this section, fundamentals pertaining to single-
bubble behavior in a fluidized bed such as the bubble wake, cloud, shape, and rise velocity
are introduced. A theory describing the flow field surrounding a single rising bubble and
characteristics of bubble formation and jetting is presented. In addition, the behavior of
multibubbles in free bubbling conditions including bubble interaction, bubble size, rise
velocity, and volume fraction in the bed is discussed in conjunction with the two-phase
theory of fluidization. It should be noted that free bubbling refers to bubbling fluidization
without the internals constraint.
9.4.1 Onset of Bubbling

Bubbles are formed as a result of the inherent instability of gas-solid systems. The
instability of a gas-solid fluidized bed is characterized by fast growth in local voidage in
response to a system perturbation. Because of the instability in the bed, the local voidage
usually grows rapidly into a shape resembling a bubble. Although it is not always true,
the initiation of the instability is usually perceived to be the onset of bubbling, which
marks the transition from particulate fluidization to bubbling fluidization. The theoretical
explanation of the physical origin behind and prediction of the onset of the instability of
gas-solid fluidized beds has been attempted [e.g., Anderson and Jackson, 1968; Verloop and
Heertjes, 1970; Rietema and Piepers, 1990]. The efforts have been focused on the primary
forces behind the stability among the interparticle contact forces, particle-fluid interaction
forces, and particle-particle interaction via particle velocity fluctuation.
In theoretical analyses, Jackson (1963) predicted that a fluidized bed would always be
unstable if the interaction forces among particles were not considered. Garg and Pritchett
(1975) indicated that a fluidized bed may be stabilized with the inclusion of a particle
concentration gradient driven force in the momentum equation of the particle phase. A
simple criterion for bed instability, based on a stabilization mechanism of the particle-
fluid interaction forces, was established by Foscolo and Gibilaro (1984). Batchelor (1988)
established another simple criterion for bed stabilization due to hydrodynamic forces by
considering the particle diffusion effect due to random fluctuation of the particle velocity. It
is noted, however, that, for some Group A particles {e.g.,fluidcatalytic cracking [FCC] and
fine sand particles), the inherent hysteresis phenomenon in the variation of pressure drop
with gas velocity in the particulate fluidization regime given in Fig. 9.5 has been strongly
linked to the interparticle contact forces as the factor responsible for bed stability for fine
particles [Tsinontides and Jackson, 1993].
While the issue concerning origin of the bed instability is yet to be clarified, a mechanistic
interpretation of the minimum bubbling point based on the wave concept appears viable. In
terms of Wallis's concepts of continuity wave and dynamic wave, the minimum bubbling
point in a fluidized bed corresponds to a condition in which the continuity wave velocity
and dynamic wave velocity are equal (see 6.5.2.3 and Problem 9.3). Accordingly, the
minimum bubbling velocity Umb can be expressed in terms of the elasticity modulus of the
bed as [Rietema, 1991]
(9.19)
It should be noted that the elasticity modulus E is not merely a property of the solid material
in the bed. In general, E is a complex function of the structure of packing, material properties
of packing particles, particle size, and particle contact and cohesion forces between particles.
In practice, the minimum bubbling velocity can be estimated by [Abrahamsen and
Geldart, 1980a]
= 2.O7exp(O.7160f) .0.347
(9.20)
where 0f is the mass fraction of the fine powder smaller than 45 i. The preceding expres-
sion becomes invalid when Um\y calculated is less than Umf.
9.4.2 Single Bubble in a Fluidized Bed

In the experimental study of single-bubble behavior, the gas-solid fluidized bed
to which a single bubble is introduced is usually maintained at the minimum fluidization
condition. Characteristics of a single gas bubble in a gas-solid fluidized bed are similar to
those in a liquid medium [Clift and Grace, 1985; Fan and Tsuchiya, 1990; Krishna, 1993].
9.4.2.1 Bubble Configurations

Most bubbles in gas-solid fluidized beds are of spherical cap or ellipsoidal cap shape.
Configurations of two basic types of bubbles, fast bubble (clouded bubble) and slow bubble
(cloudless bubble), are schematically depicted in Fig. 9.7. The cloud is the region established
Cloud Gas through flow

Circulatingcloyjd. . / ^ - r - - ' - - ? ^ .
Wake
Wake
(a) (b)
Figure 9.7. Bubble configurations and gas flow patterns around a bubble in gas-solid fluidized
beds: (a) Fast bubble (clouded bubble) Uh > Umf/amf; (b) Slow bubble (cloudless bubble)
(a) (b)
Figure 9.8. (a) An NO2 bubble rising in a two-dimensional ballotini bed showing the cloud
region of the bubble; (b) A two-dimensional bubble rising through a layer of black particles
showing the bubble wake in a gas-solid fluidized bed (from Rowe, 1971; reproduced with
permission).
by the gas which circulates in a closed loop between the bubble and its surroundings. The
cloud phase can be visualized with the aid of a color tracer gas bubble. For example, when
a dark brown NO2 bubble is injected into the bed (as shown in Fig. 9.8(a)), the light brown
color surrounding the bubble represents the cloud region [Rowe, 1971]. When the bubble
rise velocity is higher than the interstitial gas velocity, a "clouded" bubble, in which the
circulatory flow of gas takes place between the bubble and the cloud, as shown in Fig. 9.7(a),
forms. The cloud size decreases as the bubble rise velocity increases (Problem 9.4). As
the bubble rise velocity becomes significantly higher than the interstitial gas velocity (i.e.,
Ub ^> Umf/amf), the cloud size becomes so thin that most of the gas circulates inside the
bubble. Bubbles influidizedbeds with Group A or B particles are typically characterized by
clouded bubbles. When the interstitial gas velocity is greater than the bubble rise velocity,
it yields a "cloudless" bubble in which the emulsion phase gas flows through the bubble
phase, as shown in Fig. 9.7(b). This gas throughflowin the bubble is also known as invisible
bubble flow, which is distinguished from visible bubble flow. Bubbles in Group D particle
fluidized beds are typically characterized by cloudless bubbles.
The wake angle 0w defined in Fig. 9.7 is closely related to the shape of a bubble. A
smaller #w represents a flatter bubble, whereas a large 0w represents a more rounded one. A
large wake angle #w yields a smaller wake fraction / w which is defined as the volume ratio
of wake to bubble. As shown in Fig. 9.9, #w is related to particle diameter dv. An increase in
0.7
0.6
0.5
0.4
0.3
Synclyst catalyst
(rounded)
0.2
0.1
0.0
10
dp9 m
Figure 9.9. Relationship of the wake fraction and the wake angle for bubbles in a
three-dimensional fluidized bed, based on the data of Rowe and Partridge (1965) (from Clift
and Grace, 1985).
dp results in an increase in #w [Grace, 1970; Clift et a/., 1978]. It is noted that the apparent
viscosity of the bed measured experimentally increases with particle diameter for Group A
and Group B particles [Geldart, 1986].
The bubble wake plays an important role in solids movement or solids mixing in the
bed and the freeboard. A bubble wake in a single-phase fluid is defined as the streamline
enclosed region beneath the bubble base. In a gas-solid fluidized bed, the emulsion phase
can be treated as a pseudosingle-phase fluid. Hence, a bubble wake is defined as the region
enclosed by streamline of the pseudofluid behind the bubble base. In the bed, the wake rises
with the bubble and thereby provides an essential means for global solids circulation and
induces axial solids mixing. In the freeboard, particles carried along in the bubble wake are
the primary origin of the particles there (see 9.6). The bubble wake fractions in gas-solid
fluidized beds are generally much smaller than those in low-viscosity liquid systems.
Figure 9.8(b) shows a photograph of a large two-dimensional bubble followed by a wake
of black particles. It is seen that the bubble wake has a cusp-shaped end. Compared to
the geometry of the bubble wake in gas-liquid or slurry bubble column systems [Fan and
Tsuchiya, 1990], that of the bubble wake in a gas-solid fluidized bed is more regular and
distinct. It is noted that the precise flow structure inside the bubble wake in a gas-solid
fluidized bed is not fully clear. Studies of bubble wake lead to two different views. The
first view is conceived on the basis of an analogy of a gas bubble in liquids with that in
gas-solid fluidized beds [Clift, 1986]. Reb in a gas-solid fluidized bed, defined in terms
of the apparent density and apparent viscosity of the gas-solid suspension, is typically
on the order of 100, on the basis of the apparent viscosity of gas-solid suspensions on
the order of 1 kg/m-s [Grace, 1970; Clift and Grace, 1985]. With this order of magnitude
for Reb under the analogy, the rising bubble in a gas-solid fluidized bed would yield a
vortical motion of gas and particles in its wake. The second view indicates that the wake
flow is nonvortical, with gas and solids entering the wake from its outer boundary, flowing
inward radially toward the central region, and then directed vertically downward out of the
wake [Kozanoglu and Levy, 1991]. This view could be interpreted to indicate that for a
bubble with large 0W, which is common in gas-solid fluidized beds, the curvature around
the bubble edge is not sufficiently large to generate a significant amount of vorticity [Leal,
1989; Fan and Tsuchiya, 1990] and, thus, the accumulation of vorticity in the wake does
not occur. Since the flow pattern in the wake strongly depends on the curvature of the
bubble edge, which varies with the shape of the bubble, a complete analogy of bubbles in
liquids with those in a gas-solid fluidized bed requires a similarity in the Weber number in
addition to similarity in Ret,. The Weber number varies with the surface tension around the
bubble. However, surface tension does not exist for bubbles in a gas-solid fluidized bed,
and, thus, the analogy discussed in the first view may not strictly hold. For the second view,
computation of the vorticity based on a realistic bubble shape with a precise edge curvature
is needed to quantify the vorticity generation. Thorough experiments are necessary to clarify
this important, complex wake flow problem.
In gas-solid fluidized beds wake shedding has also been observed. The shed wake
fragments are banana-shaped and the shedding may occur at fairly regular intervals [Rowe
and Partridge, 1965; Rowe, 1971].
9.4.2.2 Flow Around a Bubble in a Fluidized Bed

To describe the particle and gas flows around the bubble in fluidized beds, the pioneering
model of Davidson and Harrison (1963) is particularly noteworthy because of its fundamen-
tal importance and relative simplicity. On the basis of some salient features of this model,
a number of other models were developed [e.g., Collins, 1965; Stewart, 1968; Jackson,
1971]. The material introduced later follows Davidson and Harrison's approach.
Consider a bubble rising in afluidizedbed. It is assumed that the bubble is solids-free, is
spherical, and has a constant internal pressure. Moreover, the emulsion phase is assumed to
be a pseudocontinuum, incompressible, and inviscid single fluid with an apparent density
of p p (l amf) + pamf. It should be noted that the assumption of incompressibility of the
mixture is not strictly valid as voidage in the vicinity of the bubble is higher than that in the
emulsion phase [Jackson, 1963; Yates et al., 1994]. With these assumptions, the velocity
and pressure distributions of the "fluid" in a uniform potential flow field around a bubble,
as portrayed by Fig. 9.10, can be given as [Davidson and Harrison, 1963]
(9.21)
and
PPf = PPf |z=o + ( ^ ) ( r - ^ f ) cos 0 (9.22)

V 9Z / o o \ r
/
where the subscripts "pf" and "oo" represent the pseudofluid and the undisturbed condition,
respectively; /boo is the rise velocity of an isolated bubble; and R\> is the bubble radius.
The pressure far away from the rising bubble in a fluidized bed at minimum fluidization can
Figure 9.10. Potential flow around a spherical bubble in a two-dimensional projection.
be estimated by
(9.23)
V 8z ) Hf
Ppf = PpfUo ~ (Pp ~ PX 1 ~ amf)g K - -^ \ COS ^ (9.24)
From the discussion in 6.3.1, it is reasonable to neglect the solids contributions to the
total pressure. Hence, the pressure of gas p ^ ppf. Furthermore, it is assumed that the
particle velocity Uv ^ UVf. In order to determine the local gas velocity, the pseudofluid
medium, i.e., the emulsion phase, is treated as a porous medium so that the Darcy law (i.e.,
Eq. (5.316)) can be applied. Thus, the local gas velocity is obtained as
(9.25)
where k is the specific permeability of the porous medium. In the undisturbed region which
1
1
6
2
I
2
Dimensionless pressure difference
Figure 9.11. Pressure distribution in the vicinity of a rising three-dimensional bubble with a
comparison of the experimental data by Reuter (1963a, b) and the Davidson-Harrison model
prediction (from Stewart, 1968).
is far away from the bubble, it gives
(9.26)
Substituting Eqs. (9.23), (9.24), and (9.26) into Eq. (9.25) yields
U
U=\-2- ) - ( t/boo * cos (9
amf/J (9.27)
v.\^ ubo y - ^ i l sir
boo '
Figure 9.11 shows a comparison between the measured dynamic pressure and the cal-
culated results based on Eq. (9.24). The profile indicates that there is a local high-pressure
region near the bubble nose and a local low-pressure region around the bubble base, i.e.,
the wake region. The low pressure in the wake region promotes pressure-induced bubble
coalescence in the bed.
(a) (b)
Figure 9.12. Formation of bubbles and a bubble jet in afluidizedbed: (a) Bubble; (b) Jet.
9.4.3 Bubble!Jet Formation and Bubble Coalescence and Breakup

When gas enters the orifice of the distributor, it can initially form bubbles or jets,
as illustrated in Fig. 9.12. The formation of bubbles or jets depends on various parameters
including types of particles, fluidization conditions surrounding the orifice, orifice size, and
presence of internals or walls in the bed [Massimilla, 1985]. The initial bubble or jet is
then transformed into a chain of bubbles. The jet is defined as an elongated void which is
appreciably larger than a bubble; it extends permanently to some distance from the orifice
into the inner bed. In general, bubbles tend to form in the presence of small particles, such
as Group A particles [Rowe et al., 1979]; jets tend to form in the presence of large particles,
such as Group D particles, when the emulsion phase is not sufficiently fluidized or when
internals are present and disrupt the solids flow to the orifice region [Massimilla, 1985].
The initial bubble can be formed either directly with little necking or with distinguishable
necking; the neck is a small elongated void which provides a path for the gas flow for initial
bubble formation. In the presence of neck, when the initial bubble is detached from the
neck, the neck size may reduce and then grow to form the next bubble. A jet undergoes
dynamic variations in location due to pressure fluctuations of the bed. It is noted that jet
and spout (see 9.8) are different in that the spout penetrates the bed and is the main source
for bed aeration, whereas the jet usually ends in the distributor region and is part of the
bubble formation process in the bed and the bed (or emulsion phase) aeration occurs mainly
through bubble action.
Bubbles may coalesce to form large bubbles or break up into smaller ones, as shown
in Fig. 9.13. This bubble interaction leads to bubbling behavior in a multibubble system
significantly different from that in a single-bubble system. The coalescence of bubbles in a
gas-solid medium is similar to that in a liquid or liquid-solid medium [Fan and Tsuchiya,
1990]. The coalescence usually takes place with the trailing bubble overtaking the leading
bubble through its wake region as a result of the regional minimum pressure (see 9.4.2.2).
When two bubbles coalesce, the volume of the resulting bubble may be greater than the
sum of the two coalescing bubbles [Clift and Grace, 1985]. It is noted that a high-voidage
region which is different from the cloud region may exist around a bubble [Yates et ah,
1994]. Upon coalescence, the gas in this high-voidage region may be added to the volume
9.41 Bubbling Fluidization 389
(a) (b)
Figure 9.13. Bubble coalescence and breakup processes: (a) Bubble coalescence; (b) Bubble
breakup.
of the resulting bubble. On the other hand, the breakup of a large bubble may start from
an indentation on the upper boundary of the bubble resulting from the disturbance induced
by relative motion with the particles, which may eventually break the bubble. Similarly to
bubble coalescence, when a bubble splits, the total volume of the new bubbles can be less
than the volume of the original parent bubble [Grace and Venta, 1973].
9.4.4 Bubble/Jet Size and Bubble Rise Velocity

The volume-equivalent diameter, or volume diameter (see Eq. (1.3)), of an initial
bubble formed at a single upward orifice, d\>[, may be related to the volumetric gas flow rate
through the orifice Qor by [Davidson and Harrison, 1963; Miwa et al., 1972]
,-0.2
(9.28)
For a jet discharging from a single jet nozzle, the jet penetration length Lj, which is the
averaged distance from the nozzle to the tip of the elongated void, may be estimated, for
50 < dp < 3,800 /xm, 208 < pp < 11,750 kg/m3, and ambient conditions, by [Yang and
Keairns, 1978]
(9.29)
p p - pj gd
390 9 / Dense-Phase Fiuidized Beds
For jets discharging from multiple orifices such as jets above a distributor plate in a fiuidized
bed, Lj may be estimated by [Yang and Keairns, 1979]
- Pj
Equation (9.30) is for 50 < dv < 830 /xm, 1,000 < p p < 2,635 kg/m3, and ambient condi-
tions. The ultimate bubble size in a multibubble fiuidized bed results from the balance of
the rates of bubble coalescence and breakup.
Bubble stability would lead to a maximum (stable) bubble size closely associated with
the growth rate of the disturbance on the bubble [Clift et al., 1974]. The growth rate of
the disturbance increases with decreasing kinematic viscosity of the gas-solid medium
surrounding the bubble. Thus, a medium with small kinematic viscosity yields a small sta-
ble bubble size. Accordingly, Group A particles with relatively small medium viscosities
yield relatively small maximum stable bubble sizes; Group D particles with relatively large
medium kinematic viscosities yield maximum stable bubble sizes which are effectively
too large to be realized in a gas-solid fiuidized bed [Clift, 1986]. The maximum stable
bubble sizes for Group B particles are between those for Group A and D particles. It is
generally regarded that the maximum stable bubble size does not exist for Group B and D
particles.
In bubbling fiuidized beds under ambient pressure and low gas velocity conditions where
the bubble size increases with the gas velocity, the bubble size may be estimated, for
U -Umf < 0.48 m/s, 0.3 m < D < 1.3 m, and 60 /xm < dp < 450 /xm, by [Mori and
Wen, 1975]
4 = 4m - (4m - 4 i ) e x p f - 0 . 3 ^ (9.31)
where H is the vertical distance from the gas distributor; 4 is the volume diameter of
bubbles averaged over a cross-sectional area of the bed; 4 i is the initial bubble size at the
distributor; and 4m is the maximum bubble size. 4 i is given by
f 138g-2 [A d (U - Umf)]0A for perforated plates

4i = < (9.32)
[ 0.376(t/ Umf)2 for porous plates
where A& is the distributor plate area per orifice evaluated by dividing the total distributor
plate area by the total number of orifices on the distributor. 4m can be expressed by
dbm = lA9[D2(U-Umf)]0A (9.33)
The bubble size may also be estimated on the basis of the correlation of Darton et al.
(1977) under conditions of free bubbling beds without slugging and maximum stable bubble
size as
4 = 0.54(/ - Umf)\0.4/
0A rr , A //"^~\0.8 0-0.2
(H + 4 V A d ) V ' 2 (9.34)
For dense-phase fluidization at high gas velocities and/or under elevated pressures, the
bubble size may not always increase with the gas velocity. With 0.1 MPa < p < 7.1 MPa
(in either the bubbling or the turbulent regime), the bubble size can be empirically correlated
., 1994]
( Pj
\ 0.06
O.Ob
" \ fjj TJ \0.42
x exp|-1.4 x l(T 4 f^ - 0.25(/7 - Umf)2 - 0.1 (U - U mf)] (9.35)

L \PJ P* J
where pa is the ambient pressure. Equation (9.35) covers the fluidization range for Group B
particles and the lower part of Group D particles. This equation does not distinguish the
distributor effects. The volume diameter of bubbles averaged over the entire bed, 4 b > c a n
be given by
ii rr
4b-/ dbdH (9.36)
Example 9.1 In a fluidized bed of Group B particles with // f 1 m, calculate the average
bubble sizes in the bed under the conditions of 0.1 MPa < p < 2.1 MPa and 0.05 m/s
< U -Umf < 1.8 m/s.
Solution Since the operating conditions cover both the low and high gas velocity
ranges, Eq. (9.35) is used. Thus, 4b can be obtained in terms of Eq. (9.36) as
006
4b = 0.21//f08 m (U-Um{)0
\pj
x exp[-1.4 x 10"4f -^] - 0.25(U - Umf)2 - 0.1-?-(fJ - Umf)
L \PJ P*
The average bubble sizes calculated by Eq. (E9.1) under the given conditions are shown in
a three-dimensional plot given in Fig. E9.1. It is seen that the bubble size variations are quite
different in the bubbling regime and the turbulent regime, which are clearly demarcated by
the onset velocity of the transition from the bubbling regime to the turbulent regime, i.e., Uc
(see 9.5). The applicable ranges of Eqs. (9.31) and (9.34) are also indicated in the figure.
The rise velocity of a single spherical cap bubble in an infinite liquid medium can be
described by the Davies and Taylor equation [Davies and Taylor, 1950] (Problem 9.6).
Experimental results indicate that the Davies and Taylor equation is valid for large bubbles
(4oo > 0.02 m, in general) with bubble Reynolds numbers greater than 40, while for bubbles
in fluidized beds, the bubble Reynolds numbers are typically on the order of 10 or less [Clift,
1986]. By analogy, the rise velocity of an isolated single spherical cap bubble in an infinite
gas-solid medium can be expressed in terms of the volume bubble diameter by [Davidson
and Harrison, 1963]
/boc=0.71V / 4^ (9.37)
In applying Eq (9.37), the infinite medium condition can be assumed when d^/D < 0.125;
Represented by
Eqs. (9.31) and (9.34)
rV
Bubbling Turbulent
fluidization u=u fluidization
regime > regime
\16i
O.l\|
iff
0.12-
0.10-
a
lii
0.08 J
0.06-
0.04:
0.02- f-..
Y
n 1.2 1.4
1.6 J
pip* 1.0
17 0.6 0.8
19 0.4 U-UM9 m/s
Figure E9.1. Three-dimensional diagram of the bubble size variation as a function of the gas
velocity and the operating pressure (Hf = 1 m; calculated from Eq. (E9.1)).
at db/D > 0.6, the wall effect becomes so significant that the bubble should be regarded as
a slug [Clift, 1986, see 9.7].
In a free bubbling bed, the average bubble rise velocity U^ can be described by [Davidson
and Harrison, 1963]
?bb (9.38)
where (U Umf) is added to the rise velocity of a single bubble to account for multibubble
effects.
9.4.5 Gas Flow Division and Bed Expansion

The distribution of gas flow in the fluidized bed is important for the analysis of
the fundamental characteristics of transport properties in the bed. One common method
to estimate the gas flow division is based on the two-phase theory of fluidization, which
divides the superficial gas flow in the bed into two subflows, i.e., bubble phase flow and
emulsion phase flow, as shown in Fig. 9.14. According to the theory, theflowvelocity can
be generally expressed as
U = ahUbb (9.39)
Bubble
Phase
\ I Emulsion
Phase
U
Figure 9.14. Gas flow distribution in the bed based on the two-phase model.
where C/em is the superficial gas velocity for the emulsion phase flow and c^bb is that for
the bubble phase flow. The variation of the relative magnitude of each flow depends on the
gas velocity and gas and particle properties. The simple two-phase theory is to express U em
as Umf and aÛ^ as (U Umf) [Toomey and Johnstone, 1952]. Thus, the volumetric gas
flow rate to the bubble phase, Q^, becomes
= (U- Umf)A (9.40)
Equation (9.40) may overestimate the actual bubble phase flow in a fluidized bed as a result
of two prevailing effects of the gas flow pattern: (1) significant emulsion phase flow or
invisible gas flow through the bubble and (2) larger interstitial gas flow in the emulsion
phase than Umf [Clift and Grace, 1985]. The theoretical analyses of invisible gas flow
through the bubble indicate that the extent of the invisible gas flow through the 2-D and 3-D
bubbles may vary from 0.5Umf to 4Umf, depending on the specific stream function used in
characterizing the flow field [Davidson and Harrison, 1963; Partridge and Rowe, 1966].
To account for the reduced bubble flow, Eq. (9.40) can be modified as [Peters et a/.,
1982]
= Y(U-Umf)A (9.41)
Bubbling
fluidization
regime
Turbulent
fluidization
regime
Figure 9.15. Typical variation of Y with pressure and gas velocity (dp = 0.5 mm,
pp = 2,500 kg/m 3 , Hmf = 0.71 m, T = 20C, and<p = 1) (from Cai et al.9 1993).
With Y = 1, Uem = Umf. Since Qh = abUbhA, we have
Y = (9.42)
U-Umf
Figure 9.15 illustrates the variation of Y with the gas velocity and pressure, which shows
that the deviation of the gas flow distribution from the simple two-phase theory (Eq. (9.40))
becomes significant with high gas velocities at high pressures. In other words, more gas
flows to the emulsion phase under these conditions. Under very low gas velocities in the bub-
bling regime, the prediction of Qb by Eq. (9.40) becomes more accurate for high pressures.
Empirically, Y can also be estimated by 2.27Ar~0-21 [Geldart, 1986].
In dense-phase fluidization, the expansion of the bed is caused by two factors: emul-
sion phase expansion and presence of bubbles. Thus, the bed expansion height Hf can be
expressed as a linear addition of the contributions of these factors
Hf = //em + Hfah (9.43)
where // em is the expansion height of the emulsion phase, which is larger than // mf for fine
particles (Group A). For Group B and D particles, // em = Hmf. Under very high operating
pressures, some Group B particles may yield // em > //mf. From Eqs. (9.39) and (9.43), it
gives
êm^bb
Ubh-(U-Uem)
For the emulsion phase expansion of Group A particles, // em can be correlated by
[Abrahamsen and Geldart, 1980b]
Hm 2.54poolVOO66exp(Q.O90f)
Equation (9.45) assumes that there is no velocity effect on // em .

When Uem or U^ in Eq. (9.44) is unknown, the determination of bed expansion height,
Hf, can be estimated by empirical correlations. At ambient temperature, the correlation for
//f of Babu et al. (1978) covers a wide range of particle properties (0.05 < dv < 2.87 mm;
257 < p p < 3,928 kg/m3) and operating pressures (0.1 < p < 7.0 MPa)
Hmf ~ +
t/f V 1 2 6 ( }
When operated under high-temperature (up to 1,258 K) conditions, the bed expansion
can be estimated by [Cai et al., 1993]
w~ = l + >""^ < 9 - 47 )
^mf
where wg is the molecular weight of gas and t/mf is the minimum fluidization velocity at am-
bient temperature. Equation (9.47) covers the same ranges of dp, p p , and p as Eq. (9.46) does.
Example 9.2 Determine the gas flow rate in the bubble phase and emulsion phase for a
fluidized bed combustor under the following two operating conditions: (a) T = 1,173 K,
p = 1.1 MPa, D = 1.2 m, Mp = 1,600 kg, dp = 0.51 mm, <p = 0.84, p p = 2,422 kg/m3,
and U = 1.0 m/s; (b) T = 293 K, p = 0.1 MPa, and other conditions are the same as (a).
Solution (a) Knowing dp, p p , p, and fi, the minimum fluidization velocity under
both the given and ambient conditions can be obtained from Eq. (9.14) as Umf = 0.077 m/s
and U^f = 0.122 m/s, respectively. Knowing dp, p p , p, /x, and <p9 the voidage at the
minimum fluidization condition can be calculated from Eq. (9.11) as amf = 0.422. Thus,
the bed height at minimum fluidization is given by
f = _ ^2 = 1.01 m (E9.2)
A t p p (l - anrf)
_ x i 22 x 2,422 x (1 - 0.422)
4
The expansion height of the bed can be calculated by Eq. (9.47) as
2l 4 x
- (l.O - 0.122) 0738 x (510 x 10- 6 ) 1 0 0 6 x 2,422 0376
(28.9x11)0-126
(E9.3)
The bubble fraction in the bed can be obtained from Eq. (9.43) using // em = //mf for Group
B particles as
ab l ^ l _ 1 2 i = 0.380 (E9.4)
/If l.OJ
The average bubble size in the bed can be calculated by Eq. (E9.1) as
4 b = 0.21 x 1.6308 x (1.0 - 0.077) 006 x exp[-1.4 x 10~4 x I I 2
- 0 . 2 5 x ( 1 . 0 - 0 . 0 7 7 ) 2 - 0 . 1 x 11 x (1.0 - 0.077)] = 0.100 m (E9.5)
The average bubble rise velocity in the bed can be obtained by Eq. (9.38) as
Ubb = 1.0 - 0.077 + 0.71 x V9.807x 0.100 = 1.63 m/s (E9.6)
Thus, Y can be evaluated from Eq. (9.42) as given by
(b) Repeating the procedure for the conditions of 293 K and 0.1 MPa yields Y = 0.754.
The gas flow distributions between the two phases for these two cases can be summarized
as follows:
Conditions Bubble phase, ahUhh Emulsion phase, U Y(U Umi)
1,173 K, 1.1 MPa 0.619 m/s 0.381 m/s

293 K, 0.1 MPa 0.662 m/s 0.338 m/s
9.5 Turbulent Fluidization

The turbulent regime is often regarded as the transition regime from bubbling
fluidization to lean-phase fluidization. At relatively low gas velocities, bubbles are present
in the turbulent regime, while at relatively high gas velocities in the turbulent regime,
the clear boundary of bubbles disappears and the nonuniformity of solids concentration
distribution yields distinct gas voids which become less distinguishable as the gas velocity
further increases toward lean-phase fluidization.
9.5.1 Regime Transition and Identification

In bubbling fluidization, bubble motion becomes increasingly vigorous as the gas
velocity increases. This behavior can be reflected in the increase of the amplitude of the
pressure fluctuations in the bed. With further increase in the gas velocity, the fluctuation
will reach a maximum, decrease, and then gradually level off, as shown in Fig. 9.16. This
fluctuation variation marks the transition from the bubbling to the turbulent regime.
The onset velocity of the transition to the turbulent regime is commonly defined to be
the gas velocity corresponding to the peak, Uc, whereas the leveling point, U^, may be
recognized as the onset of the turbulent regime proper [Yerushalmi and Cankurt, 1979].
9.51 Turbulent Fluidization 397
Uc Uk Superficial gas
velocity
Figure 9.16. Variations of pressure fluctuation with the gas velocity for dense-phase fiuidized
beds with FCC particles (from Yerushalmi and Cankurt, 1979).
However, on the basis of direct observation of bed phenomena, appreciable variations in

bubble behavior occur at gas velocities around Uc. Specifically, the bubble interaction is
dominated by bubble coalescence at gas velocities less than UCi while it is dominated by
bubble breakup at gas velocities greater than Uc. Furthermore, there is no significant flow
structure change observed at gas velocities from Uc to U^ and even higher [Brereton and
Grace, 1992]. At /k, the bed is usually already highly turbulent and may even be close to
the transition to lean-phase fluidization [Bi and Fan, 1992]. U^ was reported to vary with
axial location [Chehbouni et aL, 1994]. The plot in Fig. 9.16 for the amplitude of pressure
fluctuation appears to be typical only of Group A particles. For Group B and D particles,
the amplitude of pressure fluctuation at velocities beyond Uc may decrease without having
a distinct leveling-off point for U^ [Satija and Fan, 1985]. Thus, for all practical purposes,
Uc is defined as the onset velocity for the transition to the turbulent regime. The upper
bound of the turbulent regime is discussed in Chapter 10.
Examining the variation of bubble behavior across Uc, a criterion for the bubbling-
turbulent regime transition is proposed by Cai et al. (1992) as
(9.48)
where N# is the bubble number per unit bed volume. The conditions for the bubbling and
the turbulent regimes can be expressed by
32NB
<0 for the bubbling regime
3U2
(9.49)
d2NB
> 0 for the turbulent regime
dU2
Thus, to predict the onset velocity for the regime transition using Eq. (9.48), a mechanistic
model accounting for NB for a given fluidized bed operation is needed (Problem 9.9).
The specific flow and transport characteristics of turbulent fluidization, which are dis-
tinctly different from those of bubbling fluidization operated at relatively low gas velocities,
Table 9.2. Examples ofKDf for Beds with Internals
Internals
2.32 x 10 3 Vertical tubes with tube diameter 30 mm and
tube distance (center-to-center) 120 mm
1.64 x 10~3 Vertical tubes with tube diameter 20 mm and
tube distance (center-to-center) 60 mm
3.42 x 10~4 Pagoda baffles [Jin et ai, 1980]
can be described as follows:

(1) The upper surface of the turbulent bed exists but becomes more diffused with a
large particle concentration in the freeboard.
(2) The bubble size is small. Bubbles split and redisperse frequently and often appear
in more irregular shapes.
(3) Bubbles/voids move violently, rendering it difficult to distinguish the emulsion
(continuous) phase from the bubble/void (discrete) phase in the bed.
(4) Bubble motion appears to be more random with enhanced interphase exchange
and hence intimate gas-solid contact and high heat and mass transfer.
9.5.2 Determination of Transition Velocity

While the efforts to develop a mechanistic model for predicting Uc are limited,
empirical correlations are commonly employed. For Group A and B particles within the
range of 293 < T < 773 K and 0.1 < p < 0.8 MPa, the correlation of Cai et al. (1989)
can be used:
where KDf is the parameter accounting for the effects of the bed geometry and internals.
For beds without internals, KDf can be expressed by
1
0.211 2.42 x _ ,
.j i)
Equation (9.51) is applicable for 57 mm < D < 475 mm. For beds with internals, values for
KDf are given in Table 9.2, which shows that KDf is highly dependent on internals geometry.
Equation (9.50) shows the functionality dependency of Uc on the operating variables.
The equation reflects that the turbulent regime transition occurs at a lower gas velocity when
the particle size is smaller or the operating pressure is higher. The influence of internals or
baffles on Uc is also revealed in the equation as well as in Fig. 9.17 for pagoda baffles and
vertical tubes. The significant reduction of Uc in a bed with internals or baffles is evident.
This is due to the fact that the internals or baffles induce bubble breakup, leading to the
dominance of bubble interactions via breakup over coalescence at a relatively lower gas
velocity. The magnitude of the pressure fluctuation reduces in the presence of the baffles
and hence smoother operation results in baffled beds than in beds without baffles.
9.5 I Turbulent Fluidization 399
O Without internals
With pagoda baffles
With vertical tubes
400
300
J 200
100
I I
0.0 0.2 0.4 0.6 0.8 1.0
U,m/s
Figure 9.17. Effect of immersed internals on the flow regime transition (from Cai, 1989).
9.5.3 Hydrodynamic Characteristics

As a result of the existence of bubbles/voids in turbulent fluidized beds, the hy-
drodynamic behavior of turbulent fluidization under relatively low gas velocity conditions
is similar, to a certain extent, to that of bubbling fluidization. However, distinct differ-
ences exist under relatively high gas velocity conditions, thus rendering many correlations
previously developed for the bubbling regime invalid for the turbulent regime.
The bubble/void size in the turbulent regime tends to decrease with an increase in the
gas velocity due to the predominance of bubble breakup over bubble coalescence, as shown
in Fig. E9.1. This trend is opposite to that exhibited in the bubbling regime. However,
similarly to the bubbling regime, an increase in the operating pressure at a constant gas
velocity or constant excess gas velocity {i.e., U Umf) decreases the bubble size. Thus,
the bubble/void size in the turbulent bed can eventually be reduced to such a magnitude,
with sufficiently high gas velocities and high pressures, that it marks a gradual transition
to lean-phase bubble-free fluidization. For bubble/void size prediction, Eq. (9.35) can be
applied to both the bubbling and turbulent fluidization regimes in beds without internals.
For bubble rise velocity prediction, Eq. (9.38) can be employed provided that bubble/void
sizes are known.
As more small bubbles/voids with low rise velocities and long residence times are present
in the turbulent bed, they give rise to more significant dense bed expansion in the turbulent
bed than in the bubbling bed. In general, Eqs. (9.46) and (9.47) can be used to account for
the height of bed expansion for the turbulent bed under relatively low gas velocity conditions
where the surface of the dense bed is distinguishable.
The gasflowdistribution in the turbulent regime can be estimated from bubble properties
and bed expansion, following Eq. (9.42). The general variations of the gas flow distribution
in the turbulent regime can also be accounted for by Fig. 9.15. It is recognized that increasing
the operating pressure promotes gas entry into the emulsion phase, which eventually results
in undistinguishable bubble and emulsion phases in the bed, leading to bubble-free lean-
phase fluidization.
9.6 Entrainment and Elutriation

The freeboard region in a fluidized bed, as shown in Fig. 9.18, accommodates
particles that are being entrained from the dense bed. Entrainment refers to the ejection of
particles from the dense bed into the freeboard by the fluidizing gas. Elutriation refers to
the separation of fine particles from a mixture of particles, which occurs at all heights of
the freeboard, and their ultimate removal from the freeboard. The terms entrainment and
elutriation are sometimes used interchangeably. The carry-over rate relates to the quantities
of the particles leaving the freeboard. Coarse particles with a particle terminal velocity
higher than the gas velocity eventually return to the dense bed, whilefineparticles eventually
exit from the freeboard. The freeboard height required in design consideration is usually
Fine exit
Axial 4
distance
TDH
Solids
holdup
gas
Figure 9.18. Solids holdup distribution in the freeboard region of a fluidized bed.
9.61 Entrainment and Elutriation 401
Bed surface
Bed surface
: . ;. :
(a) (b)
Figure 9.19. Two main mechanisms of solids injection into the freeboard (from Kunii and
Levenspiel, 1991): (a) From the bubble roof; (b) From the bubble wake.
higher than the transport disengagement height (TDH), defined as a height beyond which
the solids holdup, solids entrainment, or carry-over rate remains nearly constant. The
nonuniformity of the solids holdup distribution in the freeboard region, as indicated in
Fig. 9.18, is in part due to the nonuniformity of particle properties in the dense bed. As
shown in the figure, coarse particles are present primarily in the region of the freeboard
close to the dense bed surface, whereas the upper part of the freeboard consists mainly of
fine particles.
9.6.1 Mechanisms of Solids Ejection into the Freeboard

Particles are ejected into the freeboard via two basic modes: (1) ejection of particles
from the bubble roof and (2) ejection of particles from the bubble wake, as illustrated in
Fig. 9.19. The roof ejection occurs when the bubble approaches the surface of the bed, and
a dome forms on the surface. As the bubble further approaches the bed surface, particles
between the bubble roof and surface of the dome thin out [Peters et al., 1983]. At a certain
dome thickness, eruption of bubbles with pressure higher than the surface pressure takes
place, ejecting the particles present on top of the bubble roof to the freeboard. In wake
ejection, as the bubble erupts on the surface, the inertia effect of the wake particles traveling
at the same velocity as the bubble promptly ejects these particles to the freeboard. The gas
leaving the bed surface then entrains these ejected particles to the freeboard.
In free bubbling fluidization, especially at a high velocity, bubble coalescence frequently
takes place near the bed surface. The coalescence due to an accelerated traveling bubble
yields a significant ejection of wake particles from the leading bubble. The wake ejection in
this case becomes the dominant source of particles present in the freeboard in free bubbling
fluidized beds [Pemberton, 1982]. The wake ejection in the turbulent regime is much more
pronounced than that in the bubbling regime because of higher bubble rise velocities.
The fluid-particle and particle-particle interaction and gravitational effects are key factors
dictating the solids holdup profile in Fig. 9.18. Particles may also form agglomerates or
clusters in the freeboard.
9.6.2 Correlations and Modeling

The total solids entrainment flux, /h> at a given axial distance above the bed surface,
h, can be correlated to the solid entrainment flux on the bed surface, Jo, and that above the
TDH, /oo, by [Wen and Chen, 1982]
Jh = J^ + (/ 0 - /oo) exp(-* A A) for A < TDH (9.52)

where kA is the solids concentration decay coefficient, which is a function of gas velocity
and particle properties, and h is the axial distance above the bed surface. Similarly, the flux
of solid component / can be written as
Ah = /ioo + C/io - Jioo) exp(-kiAh) (9.53)

where Ji^ is the removal rate of solids component /. By assuming that /ioo is proportional
to the weight fraction of this component in the bed, 0 i? we have
^rîB0i (9.54)
A at
where M-x is the mass of the solid component / in the bed, k& is the elutriation rate constant of
component /, and A is the cross-sectional area of the bed. Likewise, for total solid particles,
the equation becomes
]T (9.55)
where k& can be estimated by [Geldart et al., 1979]
= 23.7 exp - 5 . 4 - ^ - (9.56)

pU \ U J
for small particles (60 < dv < 350 /xm) in the bubbling or turbulent fluidization regime
(0.6 < U < 3 m/s); or by [Colakyan and Levenspiel, 1984]
(9.57)
for elutriation of relatively coarse particles (0.3 < dp < 1 mm) at high gas velocity (0.9 <
U < 3.7 m/s).
The TDH for Group A particles can be determined by [Horio et al., 1980]
TDH = 4.47 yfa\s (9.58)
where d^ is the bubble size at the bed surface. Once given TDH at a gas velocity U\, the
TDH at a different gas velocity /2 can be calculated by [Fournol et al., 1973]
- ^ (959)
The variation of the TDH with respect to the bed diameter as well as the gas velocity
is illustrated by Fig. 9.20. It shows that a higher TDH results from an increase in the gas
velocity at a given D. In addition, the TDH may also increase under elevated pressures
[Chan and Knowlton, 1984].
9.71 Slugging 403
40.0
10.0
1.0
0.40
0.03 0.10 1.0 10.0
Bed diameter D, m
Figure 9.20. Variation of TDH with bed diameter for FCC particles (from Zenz and Weil,
1958).
9.7 Slugging
When bubbles grow to sizes comparable to the bed diameter, slugging occurs,
as shown in Fig. 9.21. Slugging is most frequently encountered in small-diameter beds
with large bed heights, especially when large/heavy particles are employed. There are
requirements of a minimum slugging velocity and a minimum bed height for a slug flow
to take place. The minimum gas velocity for slugging can be estimated by [Stewart and
Davidson, 1967]
Umsi = Umf + 0.07 \fgb (9.60)
whereas the minimum bed height required for slugging is given by [Baeyens and Geldart,
1974]
Hmsl = 1.34D,0.175 (9.61)
3
for 0.05 < D < 0.3 m, 850 < pp < 2,800 kg/m , and 0.055 < dp < 3.38 mm. Unlike a con-
stant pressure drop pattern for the bubblingfluidizedbed, the total pressure drop in a slugging
bed increases with an increase in the gas velocity as a result of strong particle-wall friction
and momentum dissipation from the gas to the solids.
9.7.1 Shapes and Rise Velocities of Single Slugs

The slugs may appear in different forms. As shown in Fig. 9.21, Type (a) is the
round-nosed slug, which occurs in systems of fine particles; Type (b) is the wall slug (also
known as the half slug), which takes place in beds with rough walls, large dp/D ratios,
T Gas
Gas
T Gas
(a) (b) (c)
Figure 9.21. Types of slugs (after Kunii and Levenspiel, 1991): (a) Round-nosed slug;
(b) Wall slug (half slug); (c) Square-nosed slug (or plug).
angular particle shapes, and relatively high gas velocities; and Type (c) is the square-
nosed slug, which appears in coarse particle systems where the particle bridging effect is
significant.
The rise velocity of a single isolated Type (a) slug can be given by [Hovmand and
Davidson, 1971]
(9.62)
The Type (b) slug can be treated as a full round-nosed slug in a column of diameter ID.
Thus, the typical rising velocity is approximately \[l times the velocity given by Eq. (9.62).
However, for Type (c) slug (plug), the situation becomes quite different. The particles will
"rain" through the plug continuously, and the particles are in a state of bridging instead of
truly fluidizing. The plug rising velocity is then a function of interparticle forces, internal
friction of particles, and gas velocity. In general, the rise velocity of a plug is lower than
that calculated from Eq. (9.62).
9.7.2 Continuous Slugging

In a continuous slugging flow, particularly of Types (a) and (b), two adjacent
slugs can coalesce to form a larger slug in a way similar to but simpler than that of bubble
coalescence. If the distance between the tail of the leading slug and the nose of the following
slug is larger than 2D, the leading slug has almost no effect on the trailing one [Hovmand and
Davidson, 1971]. Coalescence between slugs is less frequent than that in the comparable
bubbling bed. Furthermore, a smaller wake volume and more ordered wake motion or less
wake dispersion for the slugging bed are observed than in the bubbling bed.
971 Slugging 405
Applying the simple two-phase theory to the slugging bed, the average slug velocity in
continuous slugging flows can be expressed by [Ormiston et a/., 1965]
Usi = tfdoo + U - Una (9.63)
The maximum height of expansion of the slugging bed, Hmax, can also be obtained by
//max = //mf + #maxC*sl (9.64)
where asi is obtained by considering the overall gas balance as
U = Uslasl + UM (9.65)
Substituting Eqs. (9.63) and (9.65) into Eq. (9.64) yields
"max (9.66)
Usloo
:
Example 9.3 Verify Stewart's criterion (Eq. (9.60)), which states that slug flow occurs if
U-Um{> 0.07^/JD (E9.8)
with the following assumptions (see Fig. E9.2):
(1) The slug is approximated as a paraboloid revolution so that the slug volume is given by
V* = ^ (E9.9)
ID
Figure E9.2. Schematic representation of the slugflowconditions given in Example 9.3.

(2) The distance between consecutive slugs is ID.
Solution Substituting Eqs. (9.62) and (9.63) into Eq. (9.65) gives
U-Umf = 0.351 (E9.10)
The slug voidage in a continuous stream with separation distance 2D, as shown in Fig. E9.2,
is given by
_ volume of slug _ (TT/8)D 3 _ j.
asl
~~ total volume ~ (n/4)D2(2D + D) ~ 6 '
Substituting Eq. (E9.ll) into Eq. (E9.10) yields the condition for slug flow given by
Eq. (E9.8).
9.8 Spouted Beds

In a spouted bed, gas enters the bed through a jet nozzle of diameter Dx, forming
a spout of diameter Ds in the center of the bed, as shown in Fig. 9.22. The surrounding
annular region forms a downward-moving bed. Particles are entrained into the spout from
the bottom and side wall of the spout. Part of the gas seeps into the annular region through
the spout wall, whereas the other part leaves the bed from the top of the spout. The particles
carried into the spout disengage from the gas in a solid disengagement fountain just above
the bed and then return to the top of the annular region. Group D particles are commonly
used for the spouted bed operation, although spouting of Group B particles is also possible
Fountain
Bed
surface
Annulus
Spout
Gas
Figure 9.22. Schematic diagram of a spouted bed (after Mathur and Epstein, 1974).
9.81 Spouted Beds 407
Figure 9.23. Variation of the pressure drop with the gas flow rate in a spouted bed (from
Mathur and Epstein, 1974).
if the nozzle diameter does not exceed 25.4 times the particle diameter [Chandnani and
Epstein, 1986]. In some cases, a fluid-spout or aerated-spout bed operation is conducted
with secondary gas, which is introduced into the bed through the side wall of the conical
base to increase the gas-particle contact in the annular region.
9.8.1 Onset of Spouting

The process of spouting onset can be schematically depicted as in Fig. 9.23, which
shows the variation of the pressure drop with the gas flow rate in the process. With an increase
in the gas velocity, a central dilute flow core region due to gas penetration is formed under
the bed surface, and the pressure drop increases, following the line AB. As the gas velocity
further increases, the pressure drop decreases until point C, when the core region penetrates
the entire bed, forming a spout. A slight further increase in the gas velocity leads to a
steep decrease in the pressure drop to point D, beyond which the pressure drop levels off.
However, as the gas velocity decreases from point D, the pressure drop traces a different
path (DEFA), marking a hysteresis loop. The minimum spouting velocity, f/msp, is defined
as the gas velocity at point E.
Unlike the minimum fluidization velocity, the minimum spouting velocity Umsp depends
not only on the particle and gas properties but also on the bed geometry and static bed height.
By reducing the gas velocity or by increasing the bed height, spouting can be diminished.
Umsp can be determined by [Mathur and Gishler, 1955]
for D < 0.4 m (9.67)

V
where // sp is the spouted bed height. For D > 0.4 m, Umsv can be estimated by multiplying
the value of Umsp, calculated in terms of Eq. (9.67), by a factor of 2D [Fane and Mitchell,
1984].
9.8.2 Maximum Spoutable Bed Depth and Spout Diameter

Spouting cannot be maintained once the bed height exceeds a certain level, com-
monly known as maximum spoutable bed depth, Hm. At Hm, the spout starts to collapse and
the bed starts to change from a spouted bed to a fluidized bed. As indicated in Eq. (9.67),
the minimum spouting velocity increases with an increase in the bed height. Consequently,
Umsp achieves its maximum at Hm. The maximum value of the minimum spouting velocity
can be related to the minimum fluidization velocity by
OAnspXnax = b (9M)
where b is a function of the particle properties and spouting vessel geometry, varying from
0.9 to 1.5 [Epstein and Grace, 1997]. Substituting Eq. (9.67) into Eq. (9.68) yields
( 9 m
The spout diameter varies along the bed height. The average spout diameter D s can be
estimated by [McNab, 1972]
Ds = 2.0/-49 D a68 /p b 0 - 41 (9.70)

where / is the mass flux of gas and Pb is the bulk density of particles. This equation is valid
only for ambient temperature. For a spouted bed of D < 0.16 m at elevated temperatures,
Ds is given by [Wu et aL, 1987]
r r0.433r)0.583 0.133-1
Ds = 5.61 - ^ (9.71)
L (PbP) 0283 J
The voidage of the annular region is near the minimum spouting condition, which is very
close to amf. Thus, the annular region can be treated as a loosely packed bed. The voidage
in the spout decreases roughly linearly with the height [Epstein and Grace, 1997].
9.8.3 Fountain Height

The fountain height is the distance from the surface of the annular region to the
fountain top. The shape of the fountain is approximately parabolic for most cases. To
estimate the fountain height, // F , Grace and Mathur (1978) proposed an expression based
on a momentum integral over the fountain, i.e.,
(9.72)
[2g(pp-p)
where VSHO is the particle velocity along the spout axis at the bed surface and a?o is the
spout voidage at the bed surface level.
9.8.4 Gas Flow Distribution

The axial gas flow distribution in the annulus in a spouted bed under the condition
Hsp = Hm can be derived by considering the boundary condition of Ua = Umf at zs H m,
Nomenclature 409
as [Mamuro and Hattori, 1970; Problem 9.10]

3
(9.73)
where /a is the superficial gas velocity in the annulus, based on the annular cross-sectional
area, in the spouted bed at z s. Epstein et al. (1978) noted that though Eq. (9.73) was derived
on the assumption that Hsp = Hm, it applies unchanged even when // sp < Hm. That is, the
flow in the annulus at any bed level is independent of // sp . When // sp < // m , the superficial
gas velocity in the annulus, based on the annulus cross-sectional area, in the spouted bed at
zs = // sp , i.e., f/aH, can be expressed from Eq. (9.73) by [Epstein et ah, 1978]
3
(9.74)
Nomenclature
A Cross-sectional area of the bed Frmf Froude number at minimum
Ao Total cross-sectional area of the fluidization
orifice in the distributor /w Wake fraction, defined as the
Ar Archimedes number volume ratio of wake to bubble
Aa Cross-sectional area of annulus 8 Gravitational acceleration
in a spouted bed h Axial distance above the bed
Ad Distributor plate area per orifice surface
Vertical distance from gas
cd Drag coefficient H
4 Volume diameter of bubble distributor
4b Averaged volume diameter of Emulsion phase expansion height
bubble in the bed H Fountain height of a spouted
4i Initial volume diameter of bubble bed
formed at the distributor Hf Expanded bed height
4m Maximum volume diameter of Hm Maximum spoutable bed depth
bubble #ma: Maximum expanding height of a
4s Volume diameter of bubble at the slugging bed
bed surface Hmb Bed height at minimum bubbling
4oo Volume diameter of an isolated Hmf
Bed height at minimum
bubble
fluidization
4r Diameter of the orifice Hms\
Minimum bed height for slugging
Particle diameter // sp
Spouted bed height
Bed diameter /
Mass flux of gas
A Diameter of the entering orifice h Total solids entrainment flux
of a spouted bed Jo
Solid entrainment flux on the bed
Ds Spout diameter
surface
E Young's modulus Joo Solid entrainment flux above the
Cohesive force between adjacent TDH
particles KD{ Parameter, defined by Eq. (9.50)
k Specific permeability of a porous Superficial gas velocity in the

medium annulus, based on the
kA Solids concentration decay cross-sectional area of the
coefficient annulus, in the spouted bed at zs
kiA Solids concentration decay #aH Superficial gas velocity in the
constant of component / annulus, based on the
Elutriation rate constant of cross-sectional area of the
kiB
component / annulus, in the spouted bed at // sp
Penetration length of a jet Local bubble rise velocity
M{ Total mass of the particle Uboc Rise velocity of an isolated
component / in the bed bubble
Mp Total mass of the particles in the Average bubble rise velocity in
bed the whole bed
NB Bubble number per unit bed Transition velocity between
volume bubbling and turbulent
fluidization
P Total pressure TJ
Superficial gas velocity in the
Pa Ambient pressure
emulsion phase
Pi Dynamic pressure
Gas velocity corresponding to the
Pvt Pressure of pseudofluid k
pressure fluctuation leveling
Gb Volumetric gas flow rate to the point in the pressure fluctuation
bubble phase versus velocity plot
Gor Volumetric gas flow rate through Superficial gas velocity at
the orifice minimum bubbling conditions
Rb Bubble radius Superficial gas velocity for
Re Radius of bubble cloud minimum fluidization
Reb Bubble Reynolds number U
mf Superficial gas velocity for
Re p f Particle Reynolds number minimum fluidization at ambient
conditions
Repmf Particle Reynolds number at
minimum fluidization velocity Minimum gas velocity for
slugging
r Radial coordinate
Minimum spouting velocity
r Radial distance
Gas velocity at the orifice
rc Radius of frontal curvature of a TJ
u
v Particle radial velocity
spherical cap bubble
component in the r-direction
n Inner radius of the centrifugal
f/nf Velocity component of
fluidized bed ^ pr
pseudofluid in the r -direction
r0 Outer radius of the centrifugal Terminal velocity of particle
t/nti
P11
fluidized bed component /
t Time f/sl Average slug rise velocity in the
TDH Transport disengagement height bed
U Gas velocity component in the ^sloo Velocity of an isolated slug
r -direction V Gas velocity component in the
U Superficial gas velocity 0-direction
References 411
Particle velocity component in Pressure drop across the bed

the #-direction 6 Polar coordinate
y
pf Velocity component of os Conical angle of the spouted bed
pseudofluid in the #-direction n
C7W Wake angle
VSHO Particle velocity along the spout Dynamic viscosity
axis at the bed surface
Viscosity at ambient conditions
wg Molecular weight of gas Apparent viscosity of a gas-solid
Me
Y Parameter, defined by Eq. (9.41) suspension
Zs Axial distance from gas inlet of Density of fluid
P
the spouted bed Gas density at ambient conditions
Pa
Pb Bulk density of particles
Greek Symbols Pe Apparent density of a gas-solid
a Bed voidage suspension
Volume fraction of the bed Pp Density of particles
occupied by bubbles 0 Vertical stress
Oimb Bed voidage at minimum 4H Mass fraction of particles less
bubbling conditions than 45 /xm
amf Bed voidage at minimum Weight fraction of particle
fluidization component / in the bed
ao Spout voidage at the bed surface <P Sphericity of particles
Volume fraction of the bed CO Angular velocity of the
occupied by slugs centrifugal fluidized bed
References
Abrahamsen, A. R. and Geldart, D. (1980a). Behavior of Gas Fluidized Beds of Fine Powders. Part
I. Homogeneous Expansion. Powder Tech., 26, 35.
Abrahamsen, A. R. and Geldart, D. (1980b). Behavior of Gas Fluidized Beds of Fine Powders. Part
II. Voidage of the Dense Phase in Bubbling Beds. Powder Tech., 26, 47.
Anderson, T. B. and Jackson, R. (1968). A Fluid Mechanical Description of Fluidized Beds:
Stability of the State of Uniform Fluidization. I&EC Fund., 1, 12.
Babu, S. P., Shah, B. and Talwalker, A. (1978). Fluidization Correlations for Coal Gasification
Materials: Minimum Fluidization Velocity and Fluidized Bed Expansion Ratio. AIChE
Symp. Ser., 74(176), 176.
Baeyens, J. and Geldart, D. (1974). An Investigation into Slugging Fluidized Beds. Chem. Eng. Sci.,
29, 255.
Batchelor, G. K. (1988). A New Theory of the Instability of a Uniform Fluidized Bed. /. Fluid
Mech., 193, 75.
Bi, H.-T. and Fan, L.-S. (1992). Existence of Turbulent Regime in Gas-Solid Fluidization. AIChE /.,
38, 297.
Brereton, C. M. H. and Grace, J. R. (1992). The Transition to Turbulent Fluidization. Trans. Instn.
Chem. Engrs., 70, Part A, 246.
Cai, P. (1989). Flow Regime Transition in Dense-Phase Fluidized Beds. Ph.D. Dissertation.
Tsinghua University (a cooperative program with Ohio State University), People's
Republic of China.
Cai, P., Chen, S. P., Jin, Y., Yu, Z. Q. and Wang, Z. W. (1989). Effect of Operating Temperature and
Pressure on the Transition from Bubbling to Turbulent Fluidization. AIChE Symp. Ser., 85
(270), 37.
Cai, P., DeMichele, G., Traniello Gradassi, A. and Miccio, M. (1993). A Generalized Method for
Predicting Gas Flow Distribution Between the Phases in FBC. In FBCs Role in the World
Energy Mix. Ed. L. Rubow. New York: ASME.
Cai, P., Jin, Y, Yu, Z. Q. and Fan, L.-S. (1992). Mechanistic Model for Onset Velocity Prediction for
Regime Transition from Bubbling to Turbulent Fluidization. I&EC Res., 31, 632.
Cai, P., Jin, Y, Yu, Z. Q. and Wang, Z. W. (1990). Mechanism of Flow Regime Transition from
Bubbling to Turbulent Fluidization. AIChE J., 36, 955.
Cai, P., Schiavetti, M., DeMichele, G., Grazzini, G. C. and Miccio, M. (1994). Quantitative
Estimation of Bubble Size in PFBC. Powder Tech., 80, 99.
Chan, I. H. and Knowlton, T. M. (1984). The Effect of System Pressure on the Transport
Disengaging Height (TDH) Above Bubbling Gas-Fluidized Beds. AIChE Symp. Ser, 80
(241), 24.
Chandnani, P. P. and Epstein, N. (1986). Spoutability and Spouting Destabilization of Fine Particles
with a Gas. In Fluidization V. Ed. Ostergaard and Sorensen. New York: Engineering
Foundation.
Chehbouni, A., Chaouki, J., Guy, C. and Klvana, D. (1994). Characterization of the Flow Transition
Between Bubbling and Turbulent Fluidization. I&EC Res., 33, 1889.
Chirone, R., Massimilla, L. and Russo, S. (1992). Bubbling Fluidization of a Cohesive Powder in an
Acoustic Field. In Fluidization VII. Ed. Potter and Nicklin. New York: Engineering
Foundation.
Clift, R. (1986). Hydrodynamics of Bubbling Fluidized Beds. In Gas Fluidization Technology.
Ed. D. Geldart. New York: John Wiley & Sons.
Clift, R. and Grace, J. R. (1985). Continuous Bubbling and Slugging. In Fluidization, 2nd ed.
Ed. Davidson, Clift and Harrison. London: Academic Press.
Clift, R., Grace, J. R. and Weber, M. E. (1974). Stability of Bubbles in Fluidized Beds. I&EC Fund.,
13, 45.
Clift, R., Grace, J. R. and Weber, M. E. (1978). Bubbles, Drops and Particles. New York: Academic
Press.
Colakyan, M. and Levenspiel, O. (1984). Elutriation from Fluidized Beds. Powder Tech., 38, 223.
Collins, R. (1965). The Rise Velocity of Davidson's Fluidization Bubble. Chem. Eng. Sci., 20,
788.
Darton, R. C , La Nauze, R. D., Davidson, J. F. and Harrison, D. (1977). Bubble Growth Due to
Coalescence in Fluidized Beds. Trans. Instn. Chem. Engrs., 55, 274.
Davidson, J. F. and Harrison, D. (1963). Fluidized Particles. Cambridge: Cambridge University
Press.
Davies, L. and Taylor, G. I. (1950). The Mechanics of Large Bubbles Rising Through Extended
Liquids and Through Liquids in Tubes. Proc. R. Soc. London, A200, 375.
Epstein, N. and Grace, J. R. (1997). Spouting of Particulate Solids. In Handbook of Powder Science
and Technology, 2nd ed. Ed. Fayed and Otten. New York: Chapman & Hall.
Epstein, N., Lim, C. J. and Mathur, K. B. (1978). Data and Models for Flow Distribution and
Pressure Drop in Spouted Beds. Can. J. Chem. Eng., 56, 436.
References 413
Fan, L.-S. and Tsuchiya, K. (1990). Bubble Wake Dynamics in Liquids and Liquid-Solid
Suspensions. Boston: Butterworths.
Fane, A. G. and Mitchell, R. A. (1984). Minimum Spouting Velocity of Scaled-up Beds. Can. J.
Chem. Eng., 62, 437.
Foscolo, P. U. and Gibilaro, L. G. (1984). A Fully Predictive Criterion for the Transition Between
Particulate and Aggregative Fluidization. Chem. Eng. Sci., 39, 1667.
Fournol, A. B., Bergougnou, M. A. and Baker, C. G. I. (1973). Solids Entrainment in a Large Gas
Fluidized Bed. Can. J. Chem. Eng., 51, 401.
Gabavcic, Z. B., Vukovic, D. V., Zdanski, F. K. and Littman, H. (1976). Fluid Flow Pattern,
Minimum Spouting Velocity and Pressure Drop in Spouted Beds. Can. J. Chem. Eng.,
54, 33.
Garg, S. K. and Pritchett, J. W. (1975). Dynamics of Gas-Fluidized Beds. /. Appl. Phys., 46,
4493.
Geldart, D. (1973). Types of Gas Fluidization. Powder Tech., 7, 285.
Geldart, D. (1986). Gas Fluidization Technology. New York: John Wiley & Sons.
Geldart, D., Cullinan, J., Georghiades, S., Gilvray, D. and Pope, D. J. (1979). The Effect of Fines on
Entrainment from Gas Fluidized Beds. Trans. Instn. Chem. Engrs., 57, 269.
Grace, J. R. (1970). The Viscosity of Fluidized Beds. Can. J. Chem. Eng., 48, 30.
Grace, J. R. (1982). Fluidized-Bed Hydrodynamics. In Handbook of Multiphase Systems. Ed. G.
Hetsroni. Washington, D.C.: Hemisphere.
Grace, J. R. (1986). Contacting Modes and Behavior Classification of Gas-Solid and Other
Two-Phase Suspensions. Can. J. Chem. Eng., 64, 353.
Grace, J. R. and Mathur, K. B. (1978). Height and Structure of the Fountain Region Above Spouted
Beds. Can. J. Chem. Eng., 56, 533.
Grace, J. R. and Venta, J. (1973). Volume Changes Accompanying Bubble Splitting in Fluidized
Beds. Can. J. Chem. Eng., 51, 110.
Horio, M., Taki, A., Hsieh, Y. S. and Muchi, I. (1980). Elutriation and Particle Transport Through
the Freeboard of a Gas-Solid Fluidized Bed. In Fluidization. Ed. Grace and Matsen.
New York: Plenum.
Hovmand, S. and Davidson, J. F. (1971). Pilot Plant and Laboratory Scale Fluidized Reactors at
High Gas Velocities: The Relevance of Slug Flow. In Fluidization. Ed. Davidson and
Harrison. London: Academic Press.
Jackson, R. (1963). The Mechanics of Fluidized Beds: The Motion of Fully Developed Bubbles.
Trans. Instn. Chem. Engrs., 41, 22.
Jackson, R. (1971). Fluid Mechanical Theory. In Fluidization. Ed. Davidson and Harrison.
New York: Academic Press.
Jin, Y, Yu, Z. Q., Shen, J. Z. and Zhang, L. (1980). Pagoda-Type Vertical Internal Baffles in
Gas-Fluidized Beds. Int. Chem. Eng., 20, 191.
Kozanoglu, B. and Levy, E. K. (1991). Transient Mixing of Homogeneous Solids in a Bubbling
Fluidized Bed. AIChE Symp. Ser, 87 (281), 58.
Krishna, R. (1993). Analogies in Multiphase Reactor Hydrodynamics. In Encyclopedia of Fluid
Mechanics. Supplement 2. Advances in Multiphase Flow. Ed. N. P. Cheremisinoff.
Houston: Gulf Publishing.
Kroger, D. G., Levy, E. K. and Chen, J. C. (1979). Flow Characteristics in Packed and Fluidized
Rotating Beds. Power Tech., 24, 9.
Kunii, D. and Levenspiel, O. (1991). Fluidization Engineering, 2nd ed. Boston:
Butterworth-Heinemann.
Leal, L. G. (1989). Vorticity Transport and Wake Structure for Bluff Bodies at Finite Reynolds
Number. Phys. Fluids A, Fluid Dynamics, 1, 124.
Liu, Y. A., Hamby, R. K. and Colberg, R. D. (1991). Fundamental and Practical Developments of
Magnetofluidized Beds: A Review. Powder Tech., 64, 3.
Mamuro, T. and Hattori, H. (1970). Flow Pattern of Fluid in Spouted Beds: Correction. /. Chem.
Eng. Japan, 3, 119.
Massimilla, L. (1985). Gas Jets in Fluidized Beds. In Fluidization, 2nd ed. Ed. Davidson, Clift and
Mathur, K. B. and Epstein, N. (1974). Spouted Beds. New York: Academic Press.
Mathur, K. B. and Gishler, P. E. (1955). A Technique for Contacting Gases with Coarse Solid
Particles. A/C/*/., 1,157.
Matsen, J. M., Rossetti, S. J. and Halow, J. S. (1986). Fluidized Beds and Gas Particle Systems. In
Encyclopedia of Chemical Processing and Design. Ed. J. J. McKetta. New York: Marcel
Dekker.
M'cNab, G. S. (1972). Prediction of Spouted Diameter. Brit. Chem. Eng. Proc. TechnoL, 17, 532.
Miwa, K., Mori, S., Kato, T. and Muchi, I. (1972). Behavior of Bubbles in Gaseous Fluidized Beds.
Int. Chem. Eng., 12, 187.
Molerus, O. (1982). Interpretation of Geldart's Type A, B, C and D Powders by Taking into Account
Interparticle Cohesion Forces. Powder Tech., 33, 81.
Montz, K. W., Beddow, J. K. and Butler, P. B. (1988). Adhesion and Removal of Paniculate
Contaminants in a High-Decibel Acoustic Field. Powder Tech., 55, 133.
Mori, S., Iwasaki, N., Mizutani, E. and Okada, T. (1992). Vibro-Fluidization of Two-Component
Mixtures of Group C Particles. In Fluidization VII. Ed. Potter and Nicklin. New York:
Engineering Foundation.
Mori, S. and Wen, C. Y (1975). Estimation of Bubble Diameter in Gaseous Fluidized Beds. AIChE
J., 21, 109.
Ormiston, R. M., Mitchell, F. R. G. and Davidson, J. F. (1965). The Velocities of Slugs in Fluidized
Beds. Trans. Instn. Chem. Engrs., 43, T209.
Partridge, B. A. and Rowe, P. N. (1966). Analysis of Gas Flow in a Bubbling Fluidized Bed When
Cloud Formation Occurs. Trans. Instn. Chem. Engrs., 44, 349.
Pemberton, S. T. (1982). Entrainment from Fluidized Beds. Ph.D. Dissertation. Cambridge
University.
Peters, M. H., Fan, L.-S. and Sweeney, T. L. (1982). Reactant Dynamics in Catalytic Fluidized Bed
Reactors with Flow Reversal of Gas in the Emulsion Phase. Chem. Eng. Sci., 37, 553.
Peters, M. H., Fan, L.-S. and Sweeney, T. L. (1983). Study of Particle Ejection in the Freeboard
Region of a Fluidized Bed with an Image Carrying Probe. Chem. Eng. Sci., 38, 481.
Reuter, H. (1963a). Druckverteilung um Blasen im Gas-Feststoff-Flie^bett. Chem.-Ing.-Tech.,
35, 98.
Reuter, H. (1963b). Mechanismus der Blasen im Gas-Feststoff-Flie^bett. Chem.-Ing.-Tech., 35,
219.
Rietema, K. (1991). The Dynamics of Fine Powders. London: Elsevier Applied Science.
Rietema, K. and Pipers, H. W (1990). The Effect of Interparticle Forces on the Stability of
Gas-Fluidized Beds. I. Experimental Evidence. Chem. Eng. Sci., 45, 1627.
Romero, J. B. and Johanson, L. N. (1962). Factors Affecting Fluidized Bed Quality. Chem. Eng.
Prog. Symp. Ser., 58(38), 28.
Rosensweig, R. E. (1979). Fluidization: Hydrodynamic Stabilization with a Magnetic Field.
Science, 204, 57.
Problems 415
Rosensweig, R. E. (1985). Ferro Hydrodynamics. Cambridge: Cambridge University Press.

Rowe, P. N. (1971). Experimental Properties of Bubbles. In Fluidization. Ed. Davidson and
Harrison. New York: Academic Press.
Rowe, P. N., Macgillivray, H. J. and Cheesman, D. J. (1979). Gas Discharge from an Orifice into a
Gas Fluidised Bed. Trans. Instn. Chem. Engrs., 57, 194.
Rowe, P. N. and Partridge, B. A. (1965). An X-Ray Study of Bubbles in Fluidized Beds. Trans.
Instn. Chem. Engrs., 43, 157.
Satija, S. and Fan, L.-S. (1985). Characteristics of the Slugging Regime and Transition to the
Turbulent Regime for Fluidized Beds of Large Coarse Particles. AIChE /., 31, 1554.
Stewart, P. S. B. (1968). Isolated Bubbles in Fluidized Beds: Theory and Experiment. Trans. Instn.
Chem. Engrs., 46, T60.
Stewart, P. S. B. and Davidson, J. F. (1967). Slug Flow in Fluidized Beds. Powder Tech., 1, 61.
Thonglimp, V. (1981). Contribution a L etude Hydrodynamique des Couches Fluidisees pour un
Gaz Vitesse Minimale de Fluidisation et Expansion. Docteur-Ingenieur thesis. Institut
National Polytechnique, Toulouse.
Toomey, R. D. and Johnstone, H. F. (1952). Gaseous Fluidization of Solid Particles. Chem. Eng.
Prog., 48, 220.
Tsinontides, S. C. and Jackson R. (1993). The Mechanics of Gas Fluidized Beds with an Interval of
Stable Fluidization. /. Fluid Mech., 255, 237.
Verloop, J. and Heertjes, P. M. (1970). Shock Waves as a Criterion for the Transition from
Homogeneous to Heterogeneous Fluidization. Chem. Eng. Sci., 25, 825.
Wen, C. Y. and Chen, L. H. (1982). Fluidized Bed Freeboard Phenomena: Entrainment and
Elutriation. AIChE /., 28, 117.
Wen, C. Y. and Yu, Y. H. (1966). Mechanics of Fluidization. Chem. Eng. Prog. Symp. Ser., 62(62),
100.
Wilhelm, R. H. and Kwauk, M. (1948). Fluidization of Solid Particles. Chem. Eng. Prog., 44, 201.
Wu, S. W. M., Lim, C. J. and Epstein, N. (1987). Hydrodynamics of Spouted Beds at Elevated
Temperatures. Chem. Eng. Commun., 62, 251.
Yang, W. C. and Keairns, D. L. (1978). Momentum Dissipation of and Gas Entrainment into a
Gas-Solid Two-Phase Jet in a Fluidized Bed. In Fluidization. Ed. Davidson and Keairns.
Cambridge: Cambridge University Press.
Yang, W. C. and Keairns, D. L. (1979). Estimating the Jet Penetration Depth of Multiple Vertical
Grid Jets. I&EC Fund., 18, 317.
Yates, J. G., Cheesman, D. J. and Sergeev, Y. A. (1994). Experimental Observations of Voidage
Distribution Around Bubbles in a Fluidized Bed. Chem. Eng. Sci., 49, 1885.
Yerushalmi, J. and Cankurt, N. T. (1979). Further Studies of the Regimes of Fluidization. Powder
Tech.,24,lSl.
Zenz, F. A. and Weil, N. A. (1958). A Theoretical-Empirical Approach to Mechanism of Particle
Entrainment from Fluidized Beds. AIChE J., 4, 472.
Problems
9.1 The dense-phase fluidized bed can be operated in the presence of an external force field
other than the gravitational field, such as the centrifugal or magnetic field. Consider a
centrifugal fluidized bed in which the dense bed is situated in a rotary cylinder (see Fig. P9.1)
[e.g., Kroger et ai, 1979]. Gas enters the bed through the cylindrical wall to fluidize the
particles inside the cylinder. Generally, the centrifugal acceleration is significantly larger
than the gravitational acceleration so that the dense bed expands toward the central axis of
the cylinder. Particulate and bubbling fluidization in the radial direction can be achieved
in such a cylinder. Derive the relationship between the pressure drop and the minimum
fluidization velocity for a centrifugal fluidized bed shown in Fig. P9.1.
Dense bed
Rotating fluidized
bed
Fixed plenum
vessel
Gas distributor
Inlet
'//////////// /AV/ /V // /J //////////
Direction of bed
expansion
Figure P9.1. Cross-sectional view of the centrifugal fluidized bed.
9.2 Fluidization behavior of a dense bed with ferromagnetic particles can be significantly altered
by exerting a magnetic field on the bed (see Fig. P9.2) [e.g., Rosensweig, 1979, 1985].
When applying an alternating magnetic field to the fluidized bed, the gas bubble formation
may be impeded and, thus, the particulate fluidization regime can be largely extended.
Solids can flow through this magnetically "stabilized" bed with almost no backmixing.
The magnetofluidized bed has been applied to aerosol filtration, petroleum refining, and
fermentation. Under the conditions of no bubbles and of a uniform magnetic field, derive
the continuity and momentum equations for the gas and the particles in a magnetofluidized
bed and interpret the physical meaning of each term in the equations.
9.3 The minimum bubbling velocity, Um\,, given in Eq. (9.19) can be derived from the instability
analysis based on the wave concept as described in 6.5.2. A wave in a gas-solid medium
can be generated by perturbation. A continuity wave yields a local change in the solids
holdup or gas voidage which propagates through the vessel at a certain velocity. The
propagation process does not introduce any dynamic effect of inertia or momentum. The
continuity wave velocity Vc for batch-solid operation can be expressed by
dU
da (P9.1)
where U and a are related by
U 1 a3
(P9.2)
10 1 - a
Problems All
Stable
gas-solid
emulsion
Coil
Magnetic line
of force
Gas distributor
Gas
Figure P9.2. Schematic diagram of a magnetofluidized bed.
The propagation of a dynamic wave produces dramatic changes in local solids holdup and
momentum in a gas-solid flow. Thus, to analyze the dynamic wave, one can examine the
wave equation obtained for the perturbation of the momentum and mass balance equations
for each of the gas and solid phases. Accordingly, the dynamic wave velocity, Vd, can be
expressed by
(P9.3)
The condition for the minimum velocity for bubble formation corresponds to Vc = Vd.
Derive Eq. (9.19) on the basis of the preceding information.
9.4 Derive the "cloud" thickness and volume ratio of cloud to bubble for a single bubble in a
fluidized bed in terms of the Davidson-Harrison (1963) model. Since the cloud is the region
established by the gas circulating in a closed loop between the bubble and its surrounding,
the following assumptions can be employed: (1) zero radial gas velocity outside the surface
of the cloud sphere (r = Rc); (2) uniform gas velocity at far distance from the bubble, i.e.,
Uhoo = Umf/amf.
9.5 For a fluidized bed with a bed diameter of 0.15 m, a porous plate distributor, and Group
B particles, calculate the bubble sizes at (1) H = 0.25 m and U - Umf = 0.028 m/s and
(2) H = 0.4 m, and U - Umf = 0.039 m/s under the operating pressure of 0.1 to 6.1 MPa
using Eqs. (9.31), (9.34), and (9.35). Compare the calculated results with the following
experimental data [after Cai et a/., 1994]:
p (MPa) 0.1 0.6 1.1 1.6 2.1 3.1 4.1 5.1 6.1
4 (mm) at (1) 27 30 30 31 29 29 27 26 22
4 (mm) at (2) 45 47 52 49 45 45 39
9.6 The rise velocity of a single isolated spherical cap bubble in an infinite liquid medium can
be expressed as
Uboo = - (P9.4)
3
where rc is the radius of the frontal curvature of the spherical cap bubble. Derive this
expression in terms of the following assumptions: (1) Pressure in the vicinity of the bubble
nose is constant. (2) A potential flow field around a sphere is valid in the vicinity of the
bubble nose.
9.7 The bed-collapsing technique is an important experimental method to determine such hy-
drodynamic properties as bubble velocity, emulsion phase gas velocity, bubble fraction,
and emulsion phase voidage [e.g., Abrahamsen and Geldart, 1980b]. This method portrays
these properties in terms of a bed-collapsing profile in the time domain. In the experiment,
the bed is fluidized at a given gas velocity and then the gas supply is suddenly shut off.
The variation curve of the collapsed bed height with respect to time is recorded. A typical
variation curve for Group A particles is shown in Fig. P9.3. As can be seen in the figure, the
curve reflects two stages of gas escaping behavior, i.e., the bubble escape stage (t 0 < t < t\)
and the hindered sedimentation (or emulsion gas escape) stage. During the first stage, all
the bubbles and part of the gas in the emulsion phase escape the bed. The first stage is char-
acterized by a rapid bed collapse or a decrease in bed height with time. During the second
stage, the gas in the emulsion phase continues to escape. The relationship between the bed
height and the time is linear in this stage until the bed approaches the packed bed state. On
the basis of the information provided, determine /bt> aem, a^, and Uem from Fig. P9.3.
H4
Bubble escape
stage
Hindered sedimentation
stage (emulsion gas
escape)
h h t
Figure P9.3. Typical results from bed-collapsing experiments.
9.8 The wall effect on the performance of fluidized beds can be reflected by the differences of
the overall hydrodynamic behavior in beds with different diameters. Practically, in order to
scale up a pilot fluidized bed to a full-size commercial unit, the size of the pilot unit should
be large enough to exclude the wall effect. A critical scale-up diameter (D cs) for the pilot
fluidized bed is thus defined as the minimum bed diameter beyond which the bed behavior
is almost independent of the bed size. Calculate the critical scale-up diameter for fluidized
beds without internals, using the following criterion:
D=DJ
< 2 percent (P9.5)
Problems 419
9.9 It was observed by Cai et al. (1990) that as the predominant bubble interaction varies from
coalescence to breakup, the prevailing flow regime undergoes a transition from bubbling to
turbulent fluidization. On the basis of this observation, a criterion for the regime transition
can be established according to the variation of the bubble number per unit bed volume,
NB, with the gas velocity, as given by Eq. (9.48). Express Eq. (9.48) in terms of the
operating variables of the bed, using the simple two-phase theory (Eq. (9.40)). Assume that
the relationship between the bed voidage, a, and the operating variables is known to be
a= f(U,p,ix,dp,pp) (P9.6)
9.10 Consider the superficial gas velocity Ua in the annular region of a spouted bed (see Fig. P9.4).
Ua is defined in terms of the cross-sectional area of the annulus. Assume that (1) the annular
region is a uniformly packed bed; (2) the flow is only a function of height z s; (3) Darcy's
law holds for the pressure drop; (4) the Janssen model given in Eq. (8.11) can be applied
in such a way that the vertical solids stress in the annulus at zs is proportional to the square
of the radial gas velocity on the inner wall of the annular region; and (5) the friction at
the vertical boundaries of the annulus can be neglected, (a) Derive the governing equation
for Ua. (b) Solve the equation from (a) for Ua (see Eq. (9.73)). The following boundary
conditions are applicable for the condition of Hsp = Hm: (l)atzs = Hm,Ua = f/mf; (2) at
zs = Hm, da/dzs = 0; (3) at zs = 0, Ua = 0.
Spout Annulus Vessel wall
Radial component p + dp c + da
of stress due to U+dUa
solid particles \
Dynamic pressure due

to gas cross-flow
Figure P9.4. Schematic diagram illustrating the force balance in the annular region of a
spouted bed (from Mamuro and Hattori, 1970).
9.11 In industrial fluidized bed reactors, the bed height is commonly fixed by an overflow weir.
Thus, as the gas velocity, U, increases from /mf, the apparent bed density decreases. An
important design principle for the gas distributor is to ensure its sufficient pressure drop for a
uniform gas distribution, i.e., without gas channeling, and stable bed operation. Specifically,
the total pressure drop across both the distributor and the bed should be in an increasing
trend with an increase in the gas velocity. Suppose that the pressure drop across a perforated
distributor, Apdistributor, with a total orifice area of Ao can be expressed by
p(UA/A0)2
Apdistributor (P9.7)
where C& is the drag coefficient. Verify that for a uniform gas distribution and stable bed
operation, the ratio of the pressure drop across the distributor, A/?distributor, to that across the
bed, A/?b, satisfies the inequality [Matsen et al., 1986]
^distributor > ^
APh ~ 2Uh
Assume that the simple two-phase theory can be applied and the following relationship
holds
^ = 0 (P9.9)
U
where f/boo is the rise velocity of an isolated bubble.
CHAPTER 10
Circulating Fluidized Beds
10.1 Introduction
As discussed in Chapter 9, dense-phase fluidization other than particulate fluidiza-
tion is characterized by the presence of an emulsion phase and a discrete gas bubble/void
phase. At relatively low gas velocities in dense-phase fluidization, the upper surface of
the bed is distinguishable. As the gas velocity increases, the bubble/void phase gradually
becomes indistinguishable from the emulsion phase. The bubble/void phase eventually
disappears and the gas evolves into the continuous phase with further increasing gas ve-
locities. In a dense-phase fluidized bed, the particle entrainment rate is low and increases
with increasing gas velocity. As the gas flow rate increases beyond the point corresponding
to the disappearance of the bubble/void phase, a drastic increase in the entrainment rate of
the particles occurs such that a continuous feeding of particles into the fluidized bed is re-
quired to maintain a steady solids flow. Fluidization at this state, in contrast to dense-phase
fluidization, is denoted lean-phase fluidization.
Lean-phase fluidization encompasses two flow regimes, i.e., the fast fluidization regime
and the dilute transport regime. The fast fluidization regime is characterized by a heteroge-
neousflowstructure, whereas the dilute transport regime is characterized by a homogeneous
flow structure. The circulating fluidized bed (CFB) is a fluidized bed system in which solid
particles circulate between the riser and the downcomer. The fast fluidization regime is
the principal regime under which the circulating fluidized bed is operated. The operating
variables for a circulating fluidized bed system include both the gas flow rate and the solids
circulation rate, in contrast to the gas flow rate only in a dense-phase fluidized bed system.
The solids circulation is established by a high velocity of gas flow. Note that the dilute
transport regime is discussed in Chapter 11, in the context of pneumatic conveying.
Theflowproperties, such as solids holdup, can be established with various combinations
of gas flow rate and solids circulation rate [Lim et al, 1995]. High operating gas velocities
for lean-phase fluidization yield a short contact time between the gas and solid phases,
which is desirable for process applications involving chemical reactions. The system has
been widely used in the petrochemical industry (e.g., in fluid catalytic cracking (FCC)
[King, 1992]) and the utility industry (e.g., in coal combustion [Basu and Fraser, 1991]).
Other commercial applications of the CFB systems include maleic anhydride production
from n-butane [Contractor et al, 1993] and Fischer-Tropsch synthesis of liquid fuels from
carbon monoxide and hydrogen [Shingles and McDonald, 1988]. The operating conditions
of the CFB systems can be significantly different, depending on process applications. For
example, for coal combustion, the gas velocity and solids flow rate are typically 5-8 m/s
and less than 40 kg/m2 s, respectively; for FCC, on the other hand, the gas velocity and
solids flow rate are substantially higher, e.g., 15-20 m/s (at the riser exit) and greater than
300 kg/m2 s, respectively [Werther, 1993]. The flow behavior of the CFB systems can
also be significantly different, depending on design or operating variations such as insertion
of internals [Wu et al, 1990; Jiang et al, 1991; Zheng et al, 1991], superimposition of
421
422 101 Circulating Fluidized Beds
different fluidized particles [Johnson. 1982; Toda et ah, 1983; Fan et ah, 1985; Satija and
Fan, 1985], and tangential gas injection [Ilias et al, 1988].
This chapter discusses basic flow characteristics of a CFB operation. Material is in-
troduced to illustrate the flow phenomena and operating constraints of the riser and the
effects caused by the intricate interplay with the accompanying components in an inte-
grated gas-solid flow loop. The phenomenological analyses of the system are a major part
of this chapter. More rigorous analyses involving computational fluid dynamics are being
developed. The flow regimes and their transitions, hydrodynamic behavior, solids flow
structure, and flow models are specifically described. Again, as indicated in the Preface,
the correlation equations presented in the text, unless otherwise noted, are given in SI units.
10.2 System Configuration

The riser, gas-solid separator, downcomer, and solids flow control device are the
four integral parts of a CFB loop. The riser is the main component of the system. In the riser,
gas and solids commonly flow cocurrently upward, although they can also flow cocurrently
downward. This chapter covers only the cocurrent gas and solid upward operation. In this
operation, as shown in Fig. 10.1, the fluidizing gas is introduced at the bottom of the riser,
where solid particles from the downcomer are fed via a control device and carried upward
in the riser. Particles exit at the top of the riser into gas-solid separators. Separated particles
then flow to the downcomer and return to the riser.
The entrance and exit geometries have significant effects on the gas and solid flow
behavior in the riser. The efficiency of the gas-solid separator can affect the particle size
distribution and solids circulation rate in the system. In a CFB system, particle separation
is typically achieved by cyclones (see Chapter 7). The downcomer provides hold volume
Outlet gas Outlet gas Outlet gas Outlet gas

ifl
Particles^
and particle
clusters
V \7 \7
Aeration II Aeration Aeration

I I U- gas
t
Inlet gas Inlet gas
(a) (b) (c) (d)
Figure 10.1. Various configurations of circulating fluidized bed systems: (a) Mechanical valve
with reservoir; (b) Nonmechanical valve with reservoir; (c) Nonmechanical valve without
reservoir; (d) Mechanical feeder without reservoir.
103 I Flow Regimes and Transitions Between Regimes 423
and a static pressure head for particle recycling to the riser. The downcomer can be a
large reservoir which aids in regulating the solids circulation rate (see Figs. 10.1 (a) and
(b)). In applications involving chemical reactions, the downcomer may be used as a heat
exchanger or as a spent solid regenerator. The downcomer can also simply be a standpipe
(see Chapter 8) which provides direct solids passage to the riser (see Figs. 10.1(c) and (d)).
The key to smooth operation of a CFB system is the effective control of the solids
recirculation rate to the riser. The solids flow control device serves two major functions,
namely, sealing riser gas flow to the downcomer and controlling solids circulation rate.
Both mechanical valves or feeders (see Figs. 10.1 (a) and (d)) and nonmechanical valves
(see Figs. 10.1(b) and (c)) are used to perform these functions. Typical mechanical valves
are rotary, screw, butterfly, and sliding valves. Nonmechanical valves include L-valves,
J-valves (see Chapter 8), V-valves, seal pots, and their variations. Blowers and compressors
are commonly used as the gas suppliers. Operating characteristics of these gas suppliers
which are directly associated with the dynamics and instability of the riser operation must
be considered (see 10.3.3.2).
10.3 Flow Regimes and Transitions Between Regimes

The flow behavior in the riser varies with gas velocity, solids circulation rate, and
system geometry. On the basis of the flow behavior, the fast fluidization regime can be
distinguished from neighboring regimes.
10.3.1 Flow Regimes and Regime Diagrams

The fast fluidization regime is characterized by a dense region at the bottom of the
riser and a dilute region above it [Li and Kwauk, 1980]. The interrelationship of the fast
fluidization regime with other fluidization regimes in dense-phase fluidization and with the
dilute transport regime is reflected in the variations of the pressure drop per unit length of
the riser, Ap/Az; gas velocity; and solids circulation rate [Bai et al., 1993], as given in
Fig. 10.2.
The upper diagram of Fig. 10.2 shows the typical variations of the pressure drop per
unit length with the gas velocity for a constant solids circulation rate of J p\ at a. lower part
of the riser, (Ap/ AZ)L (curve a-b-c-d), and an upper part of the riser, (Ap/Az)u (curve
a-b-c-d')- It is seen in the figure that at high gas velocities the wall friction is the dominant
component for (Ap/Az); (Ap/Az) increases with an increase in gas velocity. At relatively
low gas velocities, the solids holdup is the dominant component for (Ap/Az); (Ap/Az)
decreases with an increase in gas velocity. Point b represents the minimum Ap/Az marking
the onset of the dilute transport regime where the pressure drop is dominated by the wall
friction. In a dilute flow (curve a-b), there is no axial variation in (Ap/Az); i.e., (Ap/Az) L
and (Ap/Az)u are the same for a given gas velocity, noting that the location for (Ap/Az) L
is beyond the acceleration region. Curve b-c represents the transition from the dilute
transport regime to the fast fluidization regime; in the transition, the annular-core flow
pattern prevails with little axial variation in (Ap/Az). In the fast fluidization regime (curve
c-d or c-d')> there is a variation in (Ap/Az) between the upper and the lower parts of the
riser; i.e., (Ap/Az\ and (Ap/Az)u are different. The variation of (Ap/Az)L with the
gas velocity is much sharper than that of (Ap/Az)\j. At point d, the flow transition to
dense-phase fluidization occurs. The gas velocities corresponding to points c and d or d! are
Ap/Az
* Lean-phase fluidization
Dilute transport
regime
Figure 10.2. Variations of pressure drop per unit riser length with solids circulation rate and
gas velocity for variousfluidizationregimes (after Bai et ai, 1993).
denoted Ufd and t/tf, respectively. For a given gas velocity, Ap/Az increases as the solids
circulation rate (/ p ) increases. The curve of Ap/Az for Jv = 0 in the diagram represents
the wall friction due to gas flow alone.
The lower diagram of Fig. 10.2 reveals the effect of the gas velocity U, at the solids
circulation rate of Jv\, on theflowregime map plotted in terms of Jv versus U. It is seen in the
diagram that as / p decreases, the applicable range of the gas velocity for the fast fluidization
regime decreases; eventually, this regime diminishes when the solids circulation rate and
the gas velocity reach / p?tr and Utr, respectively. Therefore, for a system to be operated
in the fast fluidization regime, the solids circulation rate and the gas velocity need to be
maintained greater than JvM and UiT, respectively. With decreasing gas velocity at / p i ,
the transition from fast fluidization to dense-phase fluidization can be a fuzzy or a sharp
process, depending on variables that include column size and particle properties. For a
small column with large particles, the transition to dense-phase fluidization goes through
choking (Problem 10.7) and is a relatively sharp process which is reflected by a steep slope
of the (A/?/ Az) L curve at Utf, as given in the figure. For a large column with small particles,
the transition to dense-phase fluidization does not go through choking (also noted as the
nonchoking transition; see 10.3.2.2) and is a fuzzy process; in this case, the slope of the
( A / ? / A Z ) L curve at /# would be less steep. Choking occurs in a vertical gas-solid flow
system when the solids circulation rate reaches the saturation solids carrying capacity at
a given gas velocity, leading to a collapse of a relatively dilute suspension to a relatively
dense suspension [Zenz and Othmer, 1960; Bi et ai, 1993; also see 11.2.4]. The dense
suspension in the choking condition is usually characterized by the presence of solids slugs
[Yang, 1975] yielding considerable pressure fluctuations.
10.3.2 Regime Transition Determination

The boundaries of the fast fluidization regime can be determined from three inde-
pendent approaches: phenomenological, statistical, and structural. The phenomenological
approach utilizes the relationship among the pressure drop or bed voidage, gas velocity, and
solids circulation rate, based on the overall fluidization behavior. This approach expresses
the regime boundaries in terms of the transport velocity, gas velocity, and solids circulation
rates, as given in Fig. 10.2. The statistical approach considers that the property fluctua-
tions are dictated by bubbles under low-velocity conditions and solids assemblage under
high-velocity conditions. This approach utilizes the amplitude and standard deviation of
the voidage, pressure, or differential pressure fluctuations as well as their correlation and
power spectra as criteria to define the regime boundaries. The structural approach recog-
nizes the differences in flow structure of the flow system in the mesoscale such as bubbles
and clusters and defines the flow regimes on the basis of these differences. In the following,
the quantitative accounts of the transport velocity and the regime boundaries are discussed.
10.3.2.1 Transport Velocity

As noted in 10.3.1, the transport velocity, UtT, marks the lower limit of the gas velocity
for fast fluidization operation. The characteristics and the prediction of U tr are given in the
following discussion.
A. CHARACTERISTICS OF THE TRANSPORT VELOCITY

The transport velocity can be characterized from the voidage and gas velocity relationship
for a riser flow with Group A particles. In general, the voidage-gas velocity curve in a
logarithmic plot spanning several flow regimes exhibits a straight line within each flow
regime [Avidan and Yerushalmi, 1982]. Figure 10.3 shows such a relationship for FCC
particles. It is seen that a sharp change in the slope occurs at a point noted as U iT which
marks the onset of the dependency of the solids circulation rate on the voidage-gas velocity
relationship.
The transport velocity can also be evaluated from the variations of the local pressure
drop per unit length (Ap/Az) with respect to the gas velocity and the solids circulation
rate, / p . An example of such a relationship is shown in Fig. 10.4. It is seen in the figure
that, along the curve AB, the solids circulation rates are lower than the saturation carrying
capacity of the flow. Particles with low particle terminal velocities are carried over from the
riser, while others remain at the bottom of the riser. With increasing solids circulation rate,
more particles accumulate at the bottom. At point B in the curve, the solids fed into the riser
are balanced by the saturated carrying capacity. A slight increase in the solids circulation
rate yields a sharp increase in the pressure drop (see curve BC in Fig. 10.4). This behavior
reflects the collapse of the solid particles into a dense-phase fluidized bed. When the gas
Jp = 50 75 100 125
1.0 150 kg/m-8
> = 152mm
0.9
Particle: FCC
p p = 1,070 kg/m 3
0.8 Circulating system
dp = 49 jum
1 0.7
(based on Eq. (10.1))
0.6
0.5 i , . I
0.1 1.0 10.0

Superficial gas velocity, m/s
Figure 10.3. Characteristics of the overall voidage in various fluidization regimes for FCC
particles (after Avidan and Yerushalmi, 1982).
6,000
Catalyst particles
pp = 1,450 kg/m3
5,000 7=1.0m/s dv = 49 jim
? C7=1.34m/s D= 15.20 cm
Z= 8.50 m
4,000 z2 = 1.62 cm
3.14 cm
<f 3,000
2,000
1,000
0.0 20.0 40.0 60.0 80.0
Solids circulation rate, kg/m s
Figure 10.4. Local pressure drop as a function of solids circulation rate at various gas
velocities (after Yerushalmi and Cankurt, 1979).
10.3 I Flow Regimes and Transitions Between Regimes All
velocity is equal to or higher than the transport velocity, e.g., curve EF in Fig. 10.4, there is
no longer a sharp increase in the slope of the / p versus (Ap/Az) relationship. Thus, UtT is
characterized as the lowest gas velocity at which the discontinuity in the curve of J v versus
(Ap/Az) disappears [Yerushalmi and Cankurt, 1979].
B. PREDICTION OF THE TRANSPORT VELOCITY

There has been no definitive theory proposed to predict the transport velocity. A simple
empirical equation proposed by Bi and Fan (1992), given next, can be used for estimation
of the transport velocity:
Retr = 2.28Ar0419 (10.1)
where Retr = pUixdv/iJL and Ar is defined by Eq. (9.12). This correlation, covering Re tr
from 2.42 to 2,890, is applicable for particles of sizes from 23.6 to 5,000 /xm and of densities
from 660 to 4,510 kg/m3, i.e., Groups A, B, and D particles, in risers of diameters from
0.05 to 0.3 m. Equation (10.1) is graphically shown in Fig. 10.5. In the figure, Geldart's
demarcations of Groups A, B, and D particles in terms of Ar which are applicable for
particles with (pv p)/p ranging from about 1,000 to 2,000 [Grace, 1986] are also shown
along with the variation of the Reynolds number based on single particle terminal velocity,
Ret, with Ar. This figure indicates that for Ar less than 125 (Group A particles), for a
given Ar, the Ret is lower than Retr, and for 125 < Ar < 1.45 x 105 (Group B particles),
Ret is virtually of the same order as Re tr. However, at Ar larger than 1.45 x 105 (Group D
particles), Ret is greater than Retr. Equation (10.1) clearly indicates that UiT varies solely
with the gas and solid particle properties.
10J
Group D
:
.p Data obtained
10"' under various
operating conditions
10"1 1 10 102 103 104 105 106 107 108

Ar
Figure 10.5. Variation of Re tr and Re t with the Archimedes number (from Bi and Fan, 1992).
428 10 I Circulating Fluidized Beds
10.3.2.2 Minimum and Maximum Gas Velocities for Fast Fluidization

As noted, the fast fluidization operation is confined by a certain range of gas velocities and
solids circulation rates. For a given solids circulation rate in the fast fluidization regime,
the gas velocity needs to be greater than UtT. The analysis of the regime boundaries for
fast fluidization relies primarily on empirical approaches. An empirical correlation which
is applicable to both choking and nonchoking transitions from dense-phase fluidization to
fast fluidization, i.e., lower bound, was proposed by Bi and Fan (1991):
0.311
(10.2)
pUtf
This correlation can be applied to both Group A and Group B particles in small risers
(D < 0.3 m). It is noted that operation with choking is unstable. Another type of unstable
operation is caused by system design and operation (see 10.3.3.2). The lower bound of
the gas velocity due to instability caused by system design and operation is greater than or
equal to that due to choking.
The following empirical correlation can be used to estimate the upper bound of fast
fluidization operation for Group A particles in risers of D < 0.2 m [Bi and Fan, 1991]:
r \ 0.542
-4H Ar0105 (10.3)

The fast fluidization regime is thus bounded by U > UtT and /p,min < Jp < /P,max for
a given gas velocity (U > Uir), or Utf <U < Ufa for a given solids circulation rate. Note
that replacing Utf and U^ with U for a given value of U, the values of / p obtained from
Eqs. (10.3) and (10.2) represent / p , m i n and / p , max , respectively. The values for / p , m i n and
/P,max would merge to a value JvM when U approaches Utr. Thus, using Eqs. (10.1), (10.2),
and (10.3) for UtT, Utf, and Uf&, respectively, a quantitative flow regime diagram can be
constructed. More studies, however, are needed on flow regime boundaries for high gas
velocities (> 10 m/s) and high solids circulation rates (> 100 kg/m2 s).
Example 10.1 Determine the operating boundary of the fast fluidization regime for FCC
particles in a CFB system. The mean diameter of the FCC particles is 65 /xm and the
density is 1,500 kg/m3. Air is used as the fluidizing medium (p = 1.18 kg/m3, /x = 1.82 x
10"5 kg/m s).
Solution Calculating Retr from Eq. (10.1) yields Retr = 6.96, and Utr = 1.65
m/s. Calculating Ret from Eq. (1.7) yields Upt = 0.189 m/s and Ret = 0.798.
For a given system, Utf and Ufa are functions of the solids circulation rate. Rearranging
Eq. (10.2) gives
Substituting all values given in this example into Eq. (E10.1) yields Utf = 0.978/ p 0237 .
Likewise, rearranging Eq. (10.3) gives
^fd [2,1.op ' Ar " ygdp) Jp' (blU.Zj
and Ufd can be calculated as Ufd = 0.763/ p 0351

The calculated results of /tf and Ufd as functions of / p are plotted in Fig. E10.1
with t/tf replaced by Utr, the value for / p (=/ p , tr ) can be calculated from Eq. (E10.1) as
9.09 kg/m2 s. The values for UtY and JvM are also indicated in the figure. At gas velo-
cities higher than /tr, the lower bound of the fast fluidization regime marks the transition to
dense-phase fluidization. The bed undergoes transition from the fast fluidization regime to
the dilute phase transport as the gas velocity reaches Uf&. The solids flow pattern and axial
voidage profiles pertaining to these regimes are also given in Fig. E10.1.
60
Dense-phase
fluidization
^40
o
a 20
o
C/3
a a
Transition Dilute transport
regime
'p,tr
Transition and dilute
transport regimes
1 , 1 .
1.2 1.6 2.0 2.4 2.8 3.2
Figure E10.1. Fast fluidization regime boundaries for Example 10.1.
10.3.3 Operable Fluidization Regimes

The riser cannot be considered as an isolated entity in the CFB loop. When no
particles are introduced into or discharged from the loop, the solids mass flow rate in the
riser is equal to that in the downcomer. For a given quantity of solids in the loop, the
presence of fewer solids in the riser implies the presence of more solids in the downcomer.
Likewise, the pressure drop across the riser must be balanced by that imposed by the flow
through its accompanying components such as the downcomer and the recirculation device.
Furthermore, theflowcharacteristics of the riser can be significantly affected by the behavior
of the accompanying components in the loop.
10.3.3.1 Overall Pressure Balance Around a CFB Loop

Typical pressure profiles in a CFB loop with a nonmechanical valve (see Chapter 8) are
shown in Fig. 10.6. In this figure, line a-b-c-d represents the pressure drop across the riser,
AP, 4 -
3 -
2 -
1 -
Aeration
"gas 0
Pa
0 2 4 6 8 10
/
Gauge pressure, kPa
Inlet gas
Figure 10.6. Pressure profile around a circulating fluidized bed loop.
indicating a smooth drop of the gradient from the bottom to the top. Line e-f illustrates a
small pressure drop across the cyclone and connecting pipes, and line f-g-h-i-/? a depicts
a pressure buildup by gravity of the particles in the downcomer and the hopper above it.
Line /? a-j-a describes the pressure drop across the solids recirculation and control devices.
The pressure drops across the downcomer, the solids circulation and control device, and the
riser are major elements in the pressure balance around the CFB loop. The measurements
of Rhodes and Laussmann (1992) showed that Apsp/ApT = 2-5 and (Ap\y + A/?ib)/ Apr =
1-5 under the operating range of the gas velocity from 2 to 5 m/s, and the solids circulation
rate from 5 to 50 kg/m2 s for nonporous alumina particles with a density of 2,456 kg/m3
and a mean diameter of 75 /xm (see notations defined in the figure). The balance of the
pressure around the CFB loop requires quantitative information for the pressure drops in
each component. In the following, the pressure drop on each loop component is described.
A. PRESSURE DROP ACROSS THE RISER

For the gas flow in a riser, energy in the gas phase is partially transferred into the solids
phase through gas-particle interactions and is partially dissipated as a result of friction.
Under most operating conditions, gravitational effects dominate overall gas phase energy
consumption. Thus, neglecting the particle acceleration effects, the pressure drop in the
riser can be approximated by the equation
Apr = / p p (l - a)g dz = pJl - ao)gZ (10.4)

Jo
where Z is the riser height.
B. PRESSURE DROP THROUGH THE CYCLONE

The pressure drop through a cyclone has been analyzed in Chapter 7. Empirically, the
pressure drop can be expressed by
ApCy = kpUcy (10.5)
where Ucy is the inlet gas velocity. The value for the coefficient, k, may vary from 1 to 20,
depending on the design [Perry et ai, 1984].
C. PRESSURE DROP ACROSS THE DOWNCOMER

Maintaining the downcomer or a standpipe under a moving bed condition allows a large
pressure buildup along the downcomer. For a given quantity of particles in the downcomer,
the pressure at the bottom of the downcomer is closely associated with the relative velocity
between the gas and particles. The pressure drop rises with the relative velocity as the
particles are in a moving packed state. The maximum pressure drop in the downcomer
is established when particles are fluidized, a state which can be expressed in terms of the
pressure drop under the minimum fluidization condition as
A/?splmax = L sp (l - amf)pvg (10.6)
where L sp is the height of the solids in the downcomer. Details of the pressure drop analysis
in a downcomer or standpipe are given in Chapter 8.
D. PRESSURE DROP THROUGH THE SOLIDS FLOW CONTROL DEVICES

The solids flow in the downcomer can be either in a dense fluidized state or in a moving
packed state. If the particles in the downcomer are fluidized, the pressure drop through the
mechanical solids flow control devices can be expressed as [Jones and Davidson, 1965]
Jpf) (10.7)
2p p (l - a w ) \CAJ
where Wwp is the solids feeding rate and Ao is the opening area for the mechanical valve.
The coefficient, C o , can be taken as O.7M3.8 over a variety of systems and control device
opening configurations [Rudolph etaL, 1991]. For moving packed flow through an orifice,
the pressure drop across the orifice can be calculated from [Leung and Jones, 1978]
f-^ ) + C o A o [2p p (l - amf)A/7CV]0-5 (10.8)
tan o j
where 8 is the angle of internal friction of particles.
The solids flow rate can be controlled by nonmechanical valves such as the L-valve,
as noted in Chapter 8. The L-valve has a long horizontal leg. Thus, it is convenient to
characterize the pressure drop across an L-valve by two terms. One term is the pressure
drop through the elbow (Ap\v). This term can be described by the equations developed
for the mechanical valve because the solids flow patterns between the two are similar
except that the opening area for the L-valve is actuated by the amount of external aera-
tion and the amount of gas entrainment into the valve [Yang and Knowlton, 1993]. The
other term is the pressure drop caused by the gas-solid flow in the horizontal leg (A/?ib).
Solids are transferred into the horizontal leg with very low initial velocity and are then
accelerated. There is no reliable correlation for the pressure drop which occurs during the
acceleration of the solids. Yang and Knowlton (1993) suggested the following equation to
describe the pressure drop for an L-valve, including both the elbow and the horizontal leg
contributions:
(Aplv + Aplh) = - i - ^ (10.9)

2 p ( l - a m f ) \COAOJ
Co = 0.5 can be used here. The opening area for the L-valve (see Fig. 8.22(a)), AO9 is
determined by the external aeration rate and can be evaluated by
gt0.177(^/4)Lh
0710tf
where D h is the diameter of the horizontal section of the L-valve and L h is the horizontal
length of the L-valve. Equation (10.10) is applicable to the following conditions: D^ <
152.4 mm; Wwp < 2.5 kg/s; 175 /xm < dv < 509 /xm; 1,230 kg/m3 < p p < 4, 150 kg/m3.
The total gas flow rate, Qu consists of external aeration, Qext and the gas entrained by
the moving particles in the L-valve, <2ent- The external aeration rate, Qext, is an operating
variable. QQnt can be approximated in most cases by the following equation, assuming
no-slip between gas and moving particles [Yang and Knowlton, 1993]:
W
WP
6ent= n amf (10.11)
Pp(l -amf)
Any changes in the gas velocity and/or solids circulation rate in the riser will change
the pressure drop in each component of the system. Figure 10.7 shows the effects of the
solids circulation rate on the pressure drop in each component for a given solids inventory
in a CFB system with an L-valve. Comparisons of the pressure drops for each component
clearly indicate the important roles of (Ap\ y + A/?ib) and A/?sp in the loop pressure balance.
Generally, the pressure drop across the downcomer, A/?sp, is larger than that across the
L-valve. Ap sp shows a slight increase with the solids circulation rate. For a given solids
circulation rate, A/?sp remains essentially unchanged with an increase in the gas velocity.
A decrease in the pressure drop across the L-valve with an increase in the solids circulation
rate is also indicated in the figure. As expected, ApY increases with an increase in the solids
circulation rate.
10.3.3.2 Operating Constraints of Fast Fluidization

In practice, unstable operation may occur at a higher gas velocity than that at choking or at
nonchoking transition to dense-phase fluidization. Thus, the minimum operable gas velocity
for a given solids circulation rate can be higher than Utf for fast fluidization operation in
some CFBs. The factors contributing to this unstable situation are
(1) Insufficient pressure head established at the bottom of the downcomer
(2) Limited solids circulation rate which can be delivered by the downcomer
(3) Insufficient pressure head provided by gas suppliers
These factors can be illustrated in two aspects as given in the following.
12
(3 Q
*+ ^~--
1 --J,
J
= 40.2 kg/m2-s
-20.1kg/m 2 .s
A
' ^
> Ap
2.0 2.5 3.0 3.5 4.0 4.5

Figure 10.7. Effects of gas velocity on the pressure drop of various system components in a
CFB loop (after Rhodes and Laussmann, 1992).
A. UNSTABLE OPERATIONS RELATED TO DOWNCOMER OR SOLIDS FLOW

CONTROL DEVICES
Operating constraints related to the downcomer or solids flow control devices in CFB sys-
tems include limitations concerning the maximum available pressure drop across the down-
comer and the maximum solids circulation rate, which can be delivered by the downcomer
and the solids circulation and control device.
In CFB systems, as noted, the pressure drop in the downcomer is affected by the flow in
the other components. Under normal operating conditions, the particles in the downcomer
are in a moving bed state; the pressure drops across the downcomer and solids control
device adjust accordingly, on the basis of the resulting relative velocity between gas and
particles. However, there is a maximum pressure drop that the downcomer can develop,
which is equal to the pressure drop of a fluidized bed at its incipient fluidization state,
(A^/Lsp)mf. Under normal operating conditions, the pressure drop across the downcomer
increases with the relative velocity. When the relative velocity is higher than the minimum
fluidization velocity of the bed materials, bubbles (or slugs) will be generated. The bubble
motion, in turn, may hinder the particle flow and reduce the solids flow rate, leading to
unstable feeding in the system (see Chapter 8). In this case, the operation of the CFB loop
is limited by the solids feeding rate from the downcomer if the feeding rate at the incipient
fluidization condition is less than the solids circulation rate desired for a given gas velocity.
The constraint of the pressure drop across the downcomer is graphically illustrated in
Fig. 10.8. For a given solids inventory in the downcomer and given gas and solidsflowrates,
the pressures at the bottom of the riser and the downcomer can be determined at steady state.
Under normal operating conditions (point A in the figure), the pressure drop across the riser
is balanced by the pressure drop across the recirculation loop. If a small reduction in gas
velocity takes place, the flow in the riser responds by moving upward along the pressure
drop curve of the riser to point B. On point B of line AB, the decrease in the gas velocity
causes the pressure drop across the riser to rise by 8Pr, which has to be balanced by the
Maximum pressure drop

available from the downcomer
and recirculation device
Pressure drop across

the downcomer and
recirculation device
Pressure drop
across the riser
Superficial gas
velocity
Figure 10.8. Operational instability due to imbalance of pressure around the CFB loop (after
Bi and Zhu, 1993).
pressure drop through the circulating loop. Within a certain range of operating conditions,
the pressure drop across the downcomer can be adjusted accordingly, and the operating state
can be shifted to a new balance point, i.e., point B. However, when the pressure drop through
the riser is higher than the maximum pressure drop that is available from the downcomer and
solids recirculation devices, e.g., the solids flow control valve, the system starts to oscillate,
causing unstable operation (point C) [Bi and Zhu, 1993]. It is noted that unstable operation
related to downcomer limitations will not be present when a screw feeder is used. Using
the screw feeder, the downcomer is decoupled hydrodynamically from the riser [Hirama
et al., 1992], and the control of the solids feeding rate is through mechanical means which
do not depend on the downcomer pressure drop.
It appears that the fast fluidization regime may exist only within a narrow range in a
system when the operating instability indicated is encountered. Improvements in operating
strategy and unit design can expand the operating range of the gas velocity for the fast
fluidization regime. These improvements include increased solids inventory level in the
downcomer and reduced resistance through the solids recirculation device. For a given
system, the operable limits can be predicted by considering the hydrodynamics of each
component in the loop. This calculation procedure is illustrated in Example 10.2.
Example 10.2 A circulating fluidized bed is operated using FCC particles with a mean
diameter of 65 /xm and a density of 1,500 kg/m3. The following geometric and operating
conditions are given:
am{ = 0.48, /x = 1.82 x 10" 5 kg/m-s, p = 1.18kg/m3, Upt = 0.189m/s
Riser: D = 100mm, Z = 8 m, AY = 7.85 x 10~3 m2
Downcomer: D sp = 200mm, Asp = 3.14 x 10" 2 m 2
L-valve: D h = 60mm, L h = 300mm, L z = 500mm, Ax = 2.83 x 10~3m2
Solids inventory: M = 20, 30, and 40 kg
Also, Qext can be estimated by [Rhodes and Cheng, 1994]

n
D2
Jx
* (E10.3)
where A\ is the cross-sectional area of the L-valve; AT is the cross-sectional area of the riser;
Asp is the cross-sectional area of the downcomer; L z is the length of the vertical section of
the L-valve; D is the riser diameter; and Dsp is the diameter of the downcomer. Express the
pressure drops around the riser, cyclone, downcomer, and L-valve in the circulating fluidized
bed loop. Also calculate the maximum operable solids circulation rate as a function of gas
velocity.
Solution The maximum solids circulation rate through an L-valve is limited by

the pressure drop across the downcomer under minimum fluidizing conditions. Thus, the
pressure balance can be written in terms of voidage and solids inventory in the downcomer
as (see Fig. 10.6)
p p g(l - a mf )L sp = Apiv + Apib + P P g(l - OLQ)Z + Aph + Apcy (E10.4)

Rearranging this equation yields
L sp Apiv + A/?ib + A/?h + Apcy
1 ao = (1 amf) (B1U.5)
Z PvgZ
Calculate Apcy from Eq. (10.5) by taking the coefficient k as 10 and inlet velocity Ucy as
U, i.e.,
Apcy = kpU2y = U.SU2 (E10.6)
(A/?iv + A/?ib) can be evaluated from Eq. (10.9) on the basis of the following information:
(a) Solids circulation rate through the L-valve is
Wwp = ^D2JV = 7.85 x 10~ 3 / p (E10.7)
(b) With Wwv from Eq. (El0.7) and values provided by this example, Eq. (10.11) can be
expressed by
W
WP
Gent = /t amf = 4.838 x 10" 6 / p (E10.8)
(c) From Qt = QQnt + QQXt and Eqs. (E10.3) and (E10.8), Eq. (10.10) becomes
Gt - 0.177(7rDl/4)Lh 4.838 x 10" 6 / p + QQXt - 1.501 x 10~4
An, = =
0.710/pt 0.134
Substituting the expressions obtained into Eq. (10.9), (Ap\v + Apib) can be given as
(E10.10)
The solids inventory level in the downcomer can be evaluated from a mass balance. Ne-
glecting the solids inventory in the connecting tubes and the cyclone, L sp can be obtained
as
_ M/pp - L h Ai(l - amf) - LzAi(l - amf) - ZA r (l - a0)
SP
A s p ( l - Qfmf)
= 0.0408M - 3.846(1 - a0) - 0.072 (E10.ll)
On the basis of the calculated values from the preceding expression, and assuming A/?h ^ 0,
Eq. (El0.5) gives rise to
1 - ao = 2.65 x 10~3M - 0.25(1 - a0) - 4.68 x 10"3
-0.024f. n . n y , ^ _150A) ~ L007 x 10 4f/2

" (E10-12)
Rearranging Eq. (El0.12) yields
1 - ao = 2.12 x 10~3M - 8.05 x 10" 5 (/ 2
The sum on the right-hand side of Eq. (E10.13) is a function of operating variables. To
determine the operating limit for the solids circulation rate, another equation linking the
bed voidage, solids circulation rate, and gas velocity is necessary.
Consider a riser where the voidage in the bottom dense region is uniform and the top
dilute region behaves as the freeboard of a dense-phase fluidized bed. The axial profile of
the voidage in the top dilute region can be expressed by [Kunii and Levenspiel, 1990] (see
10.4.1)
01
= e-a(z-Zi) (E10.14)
where a is a decay constant and can be taken as 0.5 for this example. Integrating Eq. (E10.14)
over the top dilute region and combining with the voidage in the bottom dense region give
the overall voidage as
1 a=
" ^ ? " ( ! " z) (c **" aa) + (1" *a) (Eiai5)
a e in Eq. (E10.15) is the voidage at the riser exit, which can be evaluated by 1 JP/(U
Uvi)pv. The height of the bottom dense region, zi9 can be determined from Eq. (E10.14)
as
zi = Z - - In ( ~aa ) (E10.16)
a \ot* aQJ
Substituting Eq. (E10.16) into Eq. (E10.15) yields
1 _ ao = (i _ aa) + _L L _ a) _ (a* _ a) in f^ZÂ] (E10.17)
a0 in Eq. (El0.17) is also a function of the gas velocity and solids circulation rate.
10.3 I Flow Regimes and Transitions Between Regimes 437
Equations (E10.13) and (E10.17) can be solved simultaneously to obtain the relationship be-
tween the minimum operable gas velocity and solids circulation rate, as given in Fig. El0.2.
To illustrate the dependence of operable velocities on system designs, the results for three
solids inventory heights in the downcomer are given as shown in Fig. El0.2. It is seen in
the figure that for a given solids circulation rate, a higher inventory level yields a lower
minimum operable velocity.
150
"a 125
100
75
50
Solids inventory
M=20kg
25 M=30kg
M=40kg
1.0 2.0 3.0 4.0

Figure E10.2. Maximum solids circulation rates as a function of gas velocity for Example 10.2.
B. UNSTABLE OPERATION RELATED TO THE BLOWER OR COMPRESSOR

Another operational limit in the CFB system involves gas suppliers. Three types of gas
suppliers, i.e., a reciprocating compressor, a blower with throttle valve, and a compressor,
are commonly used in the CFB system. For blower operation, as the gasflowrate decreases,
the pressure head of the blower increases. For compressor operation, the pressure head of
the compressor can be maintained constant with variable gas flow rates. The interactive
behavior between a CFB system and a blower can be illustrated in Fig. 10.9, where dashed
curves refer to the blower characteristics and solid curves refer to the riser pressure drop.
At point A, the pressure drop across the riser matches the pressure head provided by a
blower; thus, a stable operation can be established. Since the pressure drop across the riser
in fast fluidization increases with a decrease in the gas flow rate at a given solids circulation
rate, a reduction in the gas flow rate causes the pressure drop to move upward on the curve
in the figure to point B with an increase in the pressure drop of 8pr. In the case shown
in Fig. 10.9(a), with the same reduction in the gas flow rate, i.e., 8Q, the increase in the
pressure drop, 8pT, from point A to point B is greater than that which can be provided by
Pressure head of the

Pressure drop blower available to
across a riser maintain suspension
Pressure head of the

blower available to
maintain suspension
Gas flow rate Gas flow rate
(a) (b)
Figure 10.9. Operational instability or stability due to insufficient pressure head provided by
gas blowers (after Wen and Galli, 1971): (a) 8ph < &pT; (b) 8ph > 8pr.
the blower, 8p\>. As a result, the gas flow rate is further reduced and the reduction may
continue to the condition where the gas flow rate is no longer sufficient to suspend the solid
particles [Doig and Roper, 1963; Wen and Galli, 1971]. Another situation is illustrated in
Fig. 10.9(b). In this case, the increase in 8pT is less than that which can be provided by the
blower, 8p\,, with the same reduction in the gas flow rate. An increase in the gas flow rate
results from the blower's response; a steady operation is then reestablished.
10.4 Hydrodynamic Behavior in a Macroscale

The hydrodynamics of a circulating fluidized bed can be analyzed from both the
macroscopic and mesoscopic points of view. The nonuniformity of the solids concentration
in the radial and axial directions represents macroscopic behavior. The existence of solid
clusters characterizes mesoscopic behavior (see 10.5). The hydrodynamic behavior in a
macroscale is discussed in the following.
10.4.1 Axial Profiles of Cross-Sectional Averaged Voidage

The axial profile of the cross-sectional averaged voidage in the riser is typically
S-shaped [Li and Kwauk, 1980], as shown in Fig. 10.10 for Group A particles. This profile
reflects an axial solids concentration distribution with a dense region at the bottom and a
dilute region at the top of the riser. The boundary between the two regions is marked by the
inflection point in the profile. An increase in the gas flow rate at a given solids circulation
rate reduces the dense region [from (a) to (c) in Fig. 10.10], whereas an increase in the solids
circulation rate at a given gas flow rate results in an expansion of the dense region [from (c)
to (a) in Fig. 10.10]. When the solids circulation rate is very low and/or the gas velocity is
very high, the dilute region covers the entire riser (see curve (d) in Fig. 10.10). For given
gas and solids flow rates, particles of high density or large size yield a low voidage in the
bottom of the riser.
10.4 I Hydrodynamic Behavior in a Macroscale 439
(c) (d)
Dense
region
I Acceleration
/ region -X
Voidage
Figure 10.10. Typical axial voidage profiles for Group A particles (after Li and Kwauk, 1980;
Yang, 1992).
The S-shaped profiles of the voidage can be represented by an equation of the following
form [Li and Kwauk, 1980]:
a - aa z- z (10.12)
a* -a
Equation (10.12) contains four parameters: a a , *, Zo, and z\. The value of Z o represents
the length of the transition region between the dilute and dense regions. Z o approaching
0 indicates a clear interface between the dense and dilute region, while Z o approaching
infinity implies a uniform axial profile of the voidage. Zo can be empirically correlated by
[Kwauk, 1992]
Z o = 500exp[-69(a* - a a )] (10.13)
where a a is the asymptotic voidage in the bottom dense region and a* is the asymptotic
voidage in the top dilute region. Their values can be obtained from the following empirical
correlations [Kwauk, 1992]:
l - a a = 0.2513
Ar 1 -
(10.14)
for a a , and
/18Re*+2.7Re* 1 - 6 8 7 V a 6 2 2 2
1 - a* = 0.05547
V Ar Ppl-a*,
(10.15)
for a*. Equations (10.14) and (10.15) were obtained on the basis of experimental data for
FCC catalyst, fine alumina, coarse alumina, pyrite cinder, and iron ore concentrate, ranging
from Group A to weakly Group B particles. These correlations fit the experimental data
well in the range of a a = 0.85-0.93 and a* = 0.97-0.993.
Of primary significance is the location of the inflection point zi? which is closely as-
sociated with the solids quantity in the downcomer and solids circulation rate in the CFB
system. Consider the pressure balance across the circulating fluidized bed loop as given by
(1 - ao)ppgZ = A/?sp - Apiv - Aplb - Aph - Apcy (10.16)
The pressure drops in Eq. (10.16) can be obtained by assuming that the particles in the
downcomer are in the incipient fluidization state. Neglecting the solids holdup in the
connecting tube between the riser and the cyclone and in the cyclone, the mass balance of
solid particles in the CFB loop can be expressed as
M
ATZ(1 - a0) + A sp L sp (l - a sp ) + AxLh{\ - am) + AXLZ(\ - alz) = (10.17)
Pp
Assuming the voidage in an L-valve as amf, Eqs. (10.6), (10.16), and (10.17) yield the
following expression for (1 ao):
Mg - (A/?iv + Apib + A/7h + Apcy)Asp - Ai(l - amf)(Lh + Lz)ppg
i ao = ( 1 U . 1 5 (1U15J J
ppg(AT + Asp)Z
Thus, the inflection point can be obtained by substituting ao from Eq. (10.18) into Eq. (10.21)
given in 10.4.3.
Recent experimental observations have indicated that the voidage in the bottom dense re-
gion is very uniform along the bed height, and the voidage increases exponentially along the
bed height in the top dilute region. Thus, alternatively, the top dilute region can be regarded
to be the freeboard of the dense-phase fluidized bed. Consequently, the well-established
entrainment model (see 9.6), which was originally developed for particle entrainment in a
freeboard region of a bubbling bed, can be adopted to describe the top dilute region [Rhodes
and Geldart, 1987; Kunii and Levenspiel, 1990].
The axial profile of the voidage or solids concentration given in Fig. 10.10 is influenced
not only by the gas velocity, solids circulation rate, and particle properties, but also by the
riser entrance and exit geometries. For smooth entrance and exit geometries, the end effects
are minimized and axial voidage exhibits a typical S-shaped profile, shown in Fig. 10.10.
The shape of the profile may vary, however, for nonsmooth entrance and exit geometries.
Figure 10.11 shows the gas and solid flow rate effects on the axial voidage profiles for an
abrupt exit. Considerable impaction of particles on the top end of the riser yields a sharp
decrease in the voidage in the exit region as evident in the figure. It is seen in the figure
that the axial voidage profile varies more significantly when the gas and solid flow rates are
higher.
10.4.2 Radial Profiles of Voidage and Solids Flux

A comprehensive account of radial voidage distribution requires a recognition
of the lateral movement of solids in addition to their axial movement. One of the most
important and yet least understood aspects of riser hydrodynamics is the lateral solids dis-
tribution mechanism [Kwauk, 1992]. Typical experimental findings for the radial voidage
10.4 I Hydrodynamic Behavior in a Macroscale 441
10
a
2
I
H3 5
8
7
6 IF"
"iwJ
HI
5-33-7
A
kg/m -s
9
31
4 -Wf 6A 53
I Aji
&=>^>
o 7.5
V 92
66
89
ffi 2
Riser 1
i i i
0
100 200 300 400 500
Suspended solids density, kg/m
(a) (b)
Figure 10.11. Effects of gas and solidsflowrates on the axial profile of the cross-sectional
averaged voidage in a riser with an abrupt exit (from Brereton and Grace, 1993b): (a) Abrupt
exit geometry; (b) Voidage profiles.
1.00
3 8 9
^ r>
0.90 0.90
Oz = 0.6m
o z = 1.4m
0.80 \ 0.80
oz = 2.45 m
AZ = 5.97 m
D = 50mm
Quartz sand D = 400 mm
^p = 56îm,Pp = 2,500 kg/m3 _ Quartz sand
0.70 0.70 ^ = 56 urn, pp = 2,500 kg/nS
J p = 100 kg/m2s
C/=3.9m/s / p = 75kg/m2.s
C/=5.0m/s
0 60 I,I, I , I , 0.60
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
r/R r/R
Figure 10.12. Typical radial profiles of voidage (after Hartge et ai, 1986).
profile are shown in Fig. 10.12. Figure 10.12(a) shows the results for a small CFB unit,
while Fig. 10.12(b) shows results for a large unit under similar operating conditions. Note
that both results are time-averaged. When the radial profile is normalized with respect
to the cross-sectional averaged voidage at the corresponding axial location, the results of
60
U= 2.4 m/s, Jp = 20 kg/m -s
D= 150 mm, z= 1.5 m
40 FCC, d =
Upflow
20 0 Net flow
O Downflow
3 0
-20
-40
0.0 0.2 0.4 0.6 0.8 1.0
rlR
Figure 10.13. Radial profiles of local solids flux (after Herb et al, 1992).
Figs. 10.12(a) and (b) along with other data can be empirically correlated by [Zhang et al.,
1991]
ot-a , (p - - (10.19)
At the center of the riser, Eq. (10.19) gives rise to

ac = a 0191 (10.20)
This correlation has been verified for a wide range of operating conditions and Group A
and Group B particles. Measurements of radial solids concentration profiles in a large-scale
CFB combustor also confirm the validity of this correlation [Werther, 1993].
The net solids flux at a given location in a riser can be obtained on the basis of the
net effect of the measured upward and downward local solids flux [Rhodes et al., 1992;
Herb et al., 1992]. Figure 10.13 shows the typical results of the solids flux measured in a
riser using FCC particles with positive values indicating the upflow. Both the upflow and
downflow are observed through the entire cross section. However, the downflow dominates
in the wall region. The point where the net solids flux is zero defines the interface between
the wall and core regions. The thickness of the wall region mainly depends on the solids
concentration in the suspension. The axial profile of the solids concentration reflects that
the thickness of the wall region decreases with the bed height. The thickness of the wall
region also depends on the riser diameter [Werther, 1993]. It is noted that the wall region
occupies, in the case of Fig. 10.13, about 31 percent of the total cross-sectional area.
10.4.3 Overall Solids Holdup

The overall solids holdup is defined as the particle volume fraction over the entire
riser. Figure 10.14 shows the typical effect of the solids circulation rate on the overall
10.5 I Local Solids Flow Structure AA3
1
1 ' 1
0.08 FCC
o [/=2.0m/s
i
0.06 - / Region m
n
Region 11,
d
0.04 o _
Region I
ja^D o
0.02 -
0.00 | 1 i
10 20 30
Solids circulation rate, kg/m -s
Figure 10.14. Overall solids holdup as a function of solids circulation rate for Group A
particles (after Jiang et al., 1993).
solids holdup for FCC particles. In the figure, three distinct regions can be identified on the
basis of riser flow behavior. In Region I, there is a nearly linear relationship between the
solids circulation rate and the overall solids holdup. The implication of this relationship is
that the average particle velocity essentially remains constant with the change in the solids
circulation rate for a given gas velocity. Region I represents the dilute transport regime. In
Region II, the high solids circulation rate leads to the formation of a particle wall layer and
particle refluxing, resulting in a steeper increase in the solids holdup as / p increases. This
region is narrow at low gas velocities and becomes relatively wider at higher gas velocities.
The flow in this region is characterized by core-annular flow. Region III represents the
fast fluidization regime. The flow at the bottom of the riser is in a turbulencelike state.
This flow region is characterized by a decrease in the slope of the solids holdup versus / p
relationship.
The prediction of overall solids holdups can be obtained by integrating the axial profile
of the solids concentration over the axial length. In terms of Eq. (10.12), the overall solids
holdup can be expressed as
aQ -aa l+exp[(Z-Zi)/Zo]\
(10.21)
a* -a. 1 + exp(-Zi/Z 0 ) )
The equation is applicable for a a = 0.85-0.93. For comparison, the typical solids con-
centration ranges for the bubbling and turbulent regimes are, respectively, 0.4-0.55 and
0.22-0.4 [Kunii and Levenspiel, 1990].
10.5 Local Solids Flow Structure

The preceding sections are concerned with the time-averaged flow behavior in a
macroscale. The time-variant flow behavior is complex. Analyses of the instantaneous flow
structure require recognition of the following factors:

(1) Particles migrate to the wall region by means of particle-particle collisions and
diffusion, and through particle-wall collision effects which tend to widen the
particle velocity distribution in the radial direction.
(2) No-slip for the gas phase on the wall results in a low velocity and low turbulent
intensity region in the vicinity of the wall.
These factors lead to localized particle accumulation in the wall region. The particle
accumulation alters the large-scale motion in the gas-solid flow, which in turn affects the
cluster size and motion.
10.5.1 Transient Nature of Solids Flow

In fast fluidization, particle collisions and particle-turbulence interactions yield
high concentrations of solids in the low-velocity wall region. As the solids concentration
reaches a certain level, the solids accumulate at various locations on the wall in the form
of a thin, dense layer. The solids layer, identified with arrows in Fig. 10.15, is of a wavy
shape. As the wavy solids layer builds up, more particles are moving into the wall region.
The wavy solids layer moves downward as a result of the net effect of gravitational and drag
forces imposed by the flow stream in the core region. When the layer becomes highly wavy
and the frontal face (leading edge of the layer) becomes bluff, particles in the wavy layer
are swept away from the wall region in much the same way as turbulent bursts occurring
in the boundary layer [Praturi and Brodkey, 1978]. The bursting process is quite sudden
compared to the growing process, and considerable amounts of solids are entrained from
the wall region to the core region during this process. A complete cycle of the evolution
%':'/] ' . * 1 :
' *. ** *' *" *. ** ] J .*. '- '*

!> f . . ' * ' ." *
', * " ' ' *
' .* 7ji*:i'
*"< ' /"^
. *..*& ".! f.- . * * ; ! ". . * *'. * " *" y'?J

** *i >"ic ' *: . "V *
*.' * *"*$ ': * '
*. ** * * * ' !**".* ':
sm? .** * !
s i *" * 1 "" * .' *v . ' '..'
. *'**i'* ; f .** J A '
s **^.*
"
. * *
##
'-"' *T *
i/'."' 5
* "* *y * *'
: ; ?& v
:*.*."''" I. *, * - ;;';.'; J
1
\'-}$ i
' . '
|:;:V; 1 ./.!
' _t .
". . t
(a) (b) (c) (d) (e)
Figure 10.15. Evolution process of wavelike solids flow structure inferred from the solids
concentration measurements and visual observations (from Jiang et al, 1993): (a) Particle
interaction yielding a thin, dense, wavy solids layer in the wall region; (b) Solids accumulation
and descent of enlarged solids layers; (c) Shaping as a bluff solids layer; (d) Solids layer
bursting; (e) Solids sweeping to the core region.
10.5 I Local Solids Flow Structure 445
of such a wavy solids layer is illustrated in Fig. 10.15. The measurement of local voidage
suggests that the solids wavy layer is fragmented into relatively small clusters during the
ejecting process as a result of large eddies. The bursting causes a surge in the local solids
concentration, which in turn enhances the local turbulence. Thus, the bursting phenomenon
evolves into a large-scale structure characteristic of the core region, while the flow in the core
region affects the frequency of bursting. The existence of the transient wavy solids layer
and the event of bursting can be detected from the measurements of the instantaneous local
solids concentration variation [Jiang et al., 1993], visualization, or computation [Gidaspow
et al., 1989]. The preceding represents one cluster formation mechanism in the top dilute
region of the riser. As research continues in this area, more will be unraveled on the general
mechanisms of wavy flow and clustering phenomena.
It is noted that a cluster as referred to here is a lump of solid particles over which
flow properties such as voidage do not vary substantially. It is formed mainly as a result
of hydrodynamic effects. The mechanism of particle clustering is different from that of
agglomeration, in which particles adhere to one another mainly by surface attraction (e.g.,
van der Waals force and electrostatic forces), and mechanical or chemical interaction [Horio
and Clift, 1992].
10.5.2 Characterization of Intermittent Solids Flow

As noted, a high degree of voidage fluctuation is associated with the bursting of
wavy solids layer into clusters. The intermittency of the bursting and the cluster velocity
are elaborated in the following.
10.5.2.1 Intermittency Index

The intermittent nature of the local solids flow behavior in a riser can be quantified on the
basis of the intermittency index defined as [Brereton and Grace, 1993a]
y= - (10.22)
where a is the standard deviation of the voidage fluctuations at a given point in the riser
and <TS is the standard deviation of the voidage fluctuations for an ideal cluster flow (see
Fig. 10.16(a)) with an identical time-averaged voidage at the same location. The ideal
cluster flow refers to a fluidization state in which the flow field comprises only clusters and
voids. In the ideal cluster flow, the local voidages are either 1 or amf, yielding [Brereton
and Grace, 1993a] the standard deviation of the voidage variation for this flow as
crs = 7 ( 1 - < * ) ( < * - a m f ) (10.23)
where a is the time-averaged local voidage. For fast fluidization in a riser, the values for
y vary between 0 and 1. For an ideal cluster flow, the value for y is 1 (see Fig. 10.16(a)).
Figure 10.16 shows two other ideal gas-solid flow structures, i.e., core-annular flow
(Fig. 10.16(b)) and uniform gas-solid flow (Fig. 10.16(c)). It is noted that at a given
location there is no time variation in the voidage in either one of the flows and, thus, the
value of the intermittency index for these ideal flows is zero.
The general trend of the intermittency index obtained experimentally for, as an
example, U 6.5 m/s and Jv = 42 kg/m2 s (case (a)) and 62 kg/m2 s (case (b)), is given
in Fig. 10.17. The results indicate that the intermittency index varies between 0.1 to 0.7.
These values imply that the gas-solid flow in the riser is never solely in the ideal
Y=1.0
(a) (b) (c)
Figure 10.16. Illustration of the definition of intermittency index for three limiting cases (after
Brereton and Grace, 1993a): (a) Ideal clusterflow;(b) Core-annular flow; (c) Uniform
dispersed flow.
0.7
oz = 0.533 m / <^<o
0.6 oz = 1.448 m 0.6
A z = 2.362 m
OZ = 0.533 m
0.5 oz = 1.448 m
A z- 2.362 m /^V
0.4 0.4 / ^
0.3 0.3
0.2
10.2
0.1 0.1
no | I . I . I .
0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
r/R r/R
(a) (b)
Figure 10.17. Intermittency index versus radial position at three axial positions (from Brereton
and Grace, 1993): (a) U = 6.5 m/s, / p = 42 kg/m2 s; (b) U = 6.5 m/s, / p = 62kg/m2 s.
core-annular nor in ideal clustering structure [Brereton and Grace, 1993a]. In the core-
annular region of the real riser flow, the increase in the intermittency index along the radial
direction reflects a variation in the solids flow pattern from a relatively uniform flow in the
core region to a transient wavy flow in the annular region. When the dense region is well
established at the bottom of the riser, the clusterlike flow structure dominates across the
entire cross section, yielding a large value of the intermittency index with a small variation
along the radial direction (see case (b) in the figure with z = 0.533 m). These results
indicate that for high solids concentrations, the clusterlike flow dominates the core-annular
flow. The core-annular flow pattern prevails for low solids concentrations, especially in the
top dilute region of the riser.
10.5.2.2 Descending Velocity and Characteristic Length of Clusters

As a result of the turbulent flow nature, there are significant variations in size and shape
of the wavy solids layers and of the ejecting locations of the wavy solids layer. Thus, the
velocities measured for descending clusters vary appreciably for a given gas velocity and
solids circulation rate. The experimental results obtained by Schnitzlein and Weinstein
(1988) and Jiang et al. (1993) reveal a slight decrease of the descending velocity with the
gas velocity. The descending velocity is about 1.8 m/s over a range of gas velocities from
1.5 to 5.0 m/s and is not sensitive to the solids circulation rate. Bader et al. (1988) and
Horio et al. (1988) have also estimated the falling velocity to be in the range of 0.5 to 1.8
m/s for FCC particles. For sand particles, Wu et al. (1991), who recorded the solids flow
by using high-speed cinematography at the wall, reported that there were widely distributed
velocities with an average of 1.26 m/s and that the velocity was independent of the solids
concentration.
The coverage time for each wavy solid layer multiplied by the descending velocity gives
the characteristic length of the wavy solids layer. Jiang et al. (1993) reported that the most
probable length observed is around 20 cm.
10.6 Mathematical Models of Fast Fluidization

On the basis of the observations in the macroscale, the flow of a fast fluidized bed
can be represented by the core-annulus flow structure in the radial direction, and coexistence
of a bottom dense region and a top dilute region in the axial direction. Particle clusters
are an indication of the heterogeneity in the mesoscale. A complete characterization of the
hydrodynamics of a CFB requires the determination of the voidage and velocity profiles.
There are a number of mathematical models accounting for the macro- or mesoaspects of
the flow pattern in a CFB that are available. In the following, basic features of several types
of models are discussed.
10.6.1 Models Based on the Concept of Cluster

The cluster concept evolved as a result of the recognition of a large slip velocity
between the gas and solid particles in the circulating fluidized bed [Yerushalmi et al., 1978;
Yerushalmi and Cankurt, 1979]. Li and Kwauk (1980) and Li et al. (1988) proposed
that the flow field in the fast fluidized beds consists of a dilute suspension phase and a
cluster phase. Considering the size of cluster in fast fluidization to be larger than the
Kolmogorov dissipative scale, Li et al. (1988) assumed that the cluster diameter is inversely
proportional to the total energy input into the system and developed a model to estimate
the cluster diameter based on the assumption that the steady-state condition follows the
energy minimization principle. Under typical operating conditions of a fast fluidized bed,
the results calculated from these models showed that the cluster diameter is in the range of
20 to 80 mm. A wide range of cluster sizes (as large as 80 mm) was also experimentally
observed by Horio and Kuroki (1994) using particle imaging techniques and Soong et al.
(1995) using optical probe techniques.
Consideration of particle clustering effects is of relevance to the modeling of the hydro-

dynamic behavior in the riser. Thus, when the cluster properties are known for a given riser
operation, models based on the concept of clusters can be applied.
10.6.2 Models Based on the Core-Annular Flow Structure

Various models based on the core-annular flow structure have been proposed.
Bolton and Davidson (1988) extended Rhodes's entrainment model [Rhodes and Geldart,
1987] by taking into account the falling particle film adjacent to the wall. Horio et al.
(1988) demonstrated that the annular flow structure cannot be rigorously explained without
taking the cluster effects into consideration. Thus, they incorporated the cluster concept
into the annular model originally developed by Nakamura and Capes (1973) for pneumatic
transport systems. In the model of Horio et al. (1988), the equation of Richardson and Zaki
(1954) was applied to estimate the effective cluster size in the wall or annular region and
core region. To close the model, they used the minimum pressure drop principle. Senior and
Brereton (1992) proposed a core-annular model based on the observed physical conditions
for the annular region. The mechanisms and relationship for interchange of solid particles
between the annular and the core regions were given.
The consideration of the core-annular structure is of significance to the modeling of the
hydrodynamic behavior in the top dilute region of the riser. Thus, when the dilute region is
predominantly present in the fast fluidized bed, models based on the core-annular structure
can be applied. The core-annular model is a one- (axial)-dimensional model. The mass and
momentum equations for the annular and the core regions can be obtained by using a radial-
property-averaging procedure such as that given by Delhaye et al. (1981). The resulting
equations are, however, complex. Simplified models have been proposed. The model by
Bolton and Davidson (1988) is simple to apply and is feasible in its mechanistic account of
the large gas-solid slip velocity in the system; this model is, therefore, introduced here.
Figure 10.18 shows the conceptual configuration of Bolton and Davidson's model. In
the model, the flow structure is idealized as a dilute core region surrounded by a dense
particle film which falls adjacent to the wall. The mathematical model for a circular riser
can be based on the following assumptions.
(1) The flow is a one-dimensional, steady flow.

(2) The particle film region is of annular cross section with an outer diameter of D
and an inner diameter of Dc, and the core region is of circular cross section with
a diameter Dc.
(3) The core region consists of a dilute upward-flowing suspension of solids with a
velocity of upc and a volume fraction of apc.
(4) The annular region consists of a dense downward-flowing suspension of solids
with a velocity of wpw and a volume fraction of a pw .
(5) All gas passes through the core at a velocity of wfc.
(6) There are no radial variations in terms of velocity and volume fraction across either
the annular region or the core region.
(7) Particles are transported by turbulent diffusion from the core region to the surface of
the falling film, where they are trapped and are carried downward. The implication
of this assumption is that the net upward particle flow will decrease along the riser
height in the core.
{'
Ik
;:)*
A-'i-i
t.''.: ;.- ....'.' : ;'f. ;
^Tt
k - ; . : ; . ; : &
Figure 10.18. Conceptual configuration of the core-annular model.
Considering an element of dz in the riser, the mass balance of the solids in the core
region gives
= -kd(avc - a
where &d is the deposition coefficient accounting for the turbulent diffusion of particles in
the core region to the wall region; avcoo is an equilibrium solids concentration in the core
region where deposition and entrainment occur at the same rate. Note that Eq. (10.24) is
developed by assuming that Dc can be approximated by D. The changes in the particle
velocity in the core region are likely to be small. Therefore, it is reasonable to assume that
wpc is independent of the bed height. Also, assuming that particles do not interfere with
each other significantly, the upward solids velocity in the core region can be taken as
Wpc = Mfc ~ Upt (10.25)

Equation (10.24) can be solved with respect to avc to yield
e~KdZ
)e~ (10.26)
+ - oi
pcoo
where Kd = 4k^/Duvc. a?pco is the solids concentration in the core region at z = 0, and
aPcoo represents the asymptotic concentration at large z. The upward flow of solids, We,
defined as the total flow rate through the cross section of the core region, can be expres-
sed as
We = ^D2cupcapc (10.27)
From Eqs. (10.26) and (10.27), it yields

We = WQOO + (W e o - Weoo)e~KdZ (10.28)
On the other hand, in terms of the mass balance across a cross section of the riser, the
upward solids flow rate We can also be evaluated from
n-DlJ = W - W = W - W = W - W (10.29)
p e w Q0 w0 eoo woo
Substituting Ww in Eq. (10.29) into Eq. (10.28) yields

Ww = Wwoo + (Ww0 ~ Wwoo)e-*dZ (10.30)
Equation (10.30) implies that the particle downward flow rate along the wall decays expo-
nentially with the bed height. The deposition coefficient, &d, can be related to the amplitude
of the gas velocity fluctuations as [Pemberton and Davidson, 1986; Bolton and Davidson,
1988] (Problem 10.8)
/
(10.31)
l+St/12
where u' is the amplitude of turbulent velocity fluctuations, correlated by U (1-2.8
and Re = UDpvj\x. St is the Stokes number, defined as pvdÛ/IS/xD. For small particles,
the Stokes number is small, and u' = O.lf/, which expresses a typical velocity fluctuation
for a pipe flow. These simplifications give
kd = Om^/nU (10.32)
Thus, A^ can be given as
- (10.33)
u upc
Once We and Ww are known, the cross-sectional averaged solids holdup, ap, can be
calculated by using the following equation
ap = _ ^ _ + _ ^ _ (10 .34)
where pw is the falling velocity of the film and can be estimated on the basis of the
information given in 10.5.2.2.
Example 10.3 A riser is of 0.15 m in diameter and 8 m in height. Particles with a mean
diameter of 200 fim and a density of 384 kg/m3 are used in the riser, which operates at U =
2.21m/sand/ p = 3.45kg/m 2 s. The gas used is air. For this operating condition, Davidson
(1991) reported a particle downward velocity, wpw, of 0.5 m/s and a particle downward flow
rate, Ww, of 0.2 kg/s in the annular region. Assume that the solids volume fraction in the
central core region, o?pc, is 0.015. Calculate the cross-sectionally averaged solids holdup
and the decay constant, Kd, defined in Eq. (10.33) in terms of the core-annular model.
Solution Assuming that the slip velocity between particles and air in the core
region is equal to the particle terminal velocity and Dc can be approximated by D, the
particle velocity in the core region can be determined from Eq. (10.25), i.e.,
M = Mfc - Upt = Upt = . 2 ^ m g - 0.331 = 1.91 m/s (E10.18)

1 Qfpc 1 0.015
where Upt is calculated from Eq. (1.7). WQ can be calculated from Eq. (10.29), which is
rearranged to yield
We = -D2JP + WW = - 0 . 1 5 2 x 3.45 + 0.2 = 0.261 kg/s (E10.19)

4 ^ 4
With the expression of wpc in Eq. (E10.18) and the value of We in Eq. (El0.19) for the
operating condition given, Eq. (10.34) becomes
1 (We Ww\ 1 /0.261 0.2
^J 384 x 0.0176 V ^ T + 05
3.86 x 10 + 592 io_2
(E10.20)
pc
The decay constant, Kd, can be calculated from Eq. (10.33) as
10.6.3 Models Based on the Axial Profiles of Solids Holdup

These types of models extend the approach of the conventional fluidized bed with
a freeboard of varying solids concentrations [Kunii and Levenspiel, 1990] to the circulating
fluidized bed. The two-phase theory used for the bubbling bed is adopted to account for the
bottom dense region, while the entrainment model proposed by Wen and Chen (1982) is
extended to the top dilute region of the fast fluidized bed. Some descriptions of the model,
considering the coexistence of the bottom dense region and top dilute region, are given in
10.4.1. In the absence of the radial variation information, this type of model can provide
a reasonable approximation of the gross hydrodynamic behavior.
10.6.4 Two-Phase Flow Models and Computational Fluid Dynamics

One-dimensional flow models are adopted in the early stages of model develop-
ment for predicting the solids holdup and pressure drop in the riser. These models consider
the steady flow of a uniform suspension. Four differential equations, including the gas
continuity equation, solids phase continuity equation, gas-solid mixture momentum equa-
tion, and solids phase momentum equation, are used to describe the flow dynamics. The
formulation of the solids phase momentum equation varies with the models employed [e.g.,
Arastoopour and Gidaspow, 1979; Gidaspow, 1994]. The one-dimensional model does not
simulate the prevailing characteristics of radial nonhomogeneity in the riser. Thus, two- or
three-dimensional models are required.
For the motion of a gas-solid suspension in the riser, both the gas and particle velocities
have local averaged and random components. Thus, it is desirable to develop a mechanistic
model which incorporates a variety of interactive effects due to both the gas and particle
velocity components (see Chapter 5) as given in the following [Sinclair and Jackson, 1989]:
(1) Interactions between particles and gas resulting from the mean slip velocity that
yield the drag force driving the nonrandom part of the particle motion
(2) Interactions of the particles with the fluctuating component of the gas velocity,
which lead to particle turbulent diffusion and induce an exchange of the kinetic
energy between the fluctuating components of the velocity of the two phases,
which results in either damping the fluctuations of the gas velocity and enhancing
fluctuations in particle velocity, or vice versa
(3) Interactions of the fluctuating part of the particle motion with the mean particle
motion through interparticle collisions, which generate pressure and stresses in the
particle assembly, consequently yielding apparent viscosity of the particle phase
(4) Interactions between the turbulent fluctuations of the gas velocity and the mean
motion of the gas, which yield the gas Reynolds stresses
The two-fluid models considering these effects for the gas-solid flow in a riser can be
described by Eqs. (5.168) through (5.170). In these models, the gas and solid are treated as
two interpenetrating fluids (see 5.3).
For equation closure, constitutive equations for the solid phase stresses are needed. From
the physical point of view, the solid stresses are a by-product of the microscale mechanisms
of the momentum transfer between the particles. The random motion of particles generates
an effective pressure in the solid phase, together with an effective viscosity which resists
shearing of the particle assembly. The kinetic energy of this random motion can be anal-
ogous to that of the thermal motion of molecules in a gas, and correspondingly it can be
characterized by a granular temperature defined as proportional to the mean square of the
random component of the particle velocity. Both the effective pressure and the viscosity
are functions of the granular temperature. Thus, it is required to have separate differential
equations representing a balance for the pseudothermal energy; that is, the energy of random
particle motion (see 5.5).
To simulate the granular flow while adopting the gas kinetic analogy, Sinclair and
Jackson (1989) pointed out that lateral segregation is induced by stresses which are caused
by interactions between the random motion and mean motion of the particle in the parti-
cle assembly. This consideration led to the development of a model for fully developed
vertical flows by taking into account the particle phase interactions. Louge et al. (1991)
considered the effect of gas phase turbulence in the modeling; however, it appears that
interactions among particles are the key to lateral solids segregation. Pita and Sundare-
san (1993) have proposed a two-fluid model for developing flow, and their computation
revealed profound effects of the entry configuration on the internal recirculation and lateral
segregation. Tsuo and Gidaspow (1990) have presented a two-dimensional unsteady-state
model. Their simulation characterizes the transient cluster motion for the dilute flow and
the annular flow structure for the dense flow and indicates that the transient integration
does not converge toward a steady state over a long period, reflecting the dynamic be-
havior of the flow. Furthermore, their model indicates that the two-fluid model with the
kinetic energy balance of the particle velocity fluctuations can be used to account for the
general behavior of the riser flow. An improved model, however, is needed to describe
Nomenclature 453
the detailed flow structure in the core and annular regions and the dynamics of the cluster
formation.
Nomenclature
Ao Opening area for the turbulent regime or choking to
mechanical valve or L-valve the fast fluidization regime
Ai Cross-sectional area of the Solids flux at radial position r
L-valve p, tr Solids circulation rate
Ar Cross-sectional area of the riser corresponding to the condition
A sp Cross-sectional area of the
downcomer k Coefficient, defined by
Eq. (10.5)
Ar Archimedes number
h Deposition coefficient, defined
a Decay constant, defined by by Eq. (10.24)
Eq. (E10.14) Parameter, defined as 4kd/Duvc
Kd
b Coefficient, defined by Lh Horizontal length of the
Eq. (P10.8) L-valve
Co Valve aperture discharge L sp Solids inventory height in the
coefficient downcomer or standpipe
D Riser diameter Lz Length of the vertical section
Dc Diameter of the central core of the L-valve
region in the core-annular flow M Solids inventory in the CFB
model loop
Dg Gas turbulent diffusion m Coefficient, defined by
coefficient Eq. (P10.8)
>h Diameter of the horizontal Pa Pressure at the L-valve inlet
section of the L-valve Po Pressure at the riser inlet
Dp Particle turbulent diffusion Gent Gas entrainment rate in the
coefficient solids flow control valve
Dsp Diameter of the downcomer Gext External gas aeration rate to the
solids flow control valve
Dy Diameter of the solids flow
Gt Total gas aeration rate to the
control valve
solids flow control valve
dp Particle diameter
R Radius of the riser
Ff Friction force
Ret Particle Reynolds number,
g Gravitational acceleration based on particle diameter and
/p Solids circulation rate or solids particle terminal velocity
flux Retr Particle Reynolds number,
J$, min Minimum solids circulating based on particle diameter and
rate for transition from the fast transport velocity
fluidization regime to the dilute r Radial position
transport regime St Stokes number
Jp, max Maximum solids circulating
rate for transition from the
U Superficial gas velocity Characteristic length, defined

Ucy Gas velocity at the inlet of a by Eq. (10.12)
cyclone
Ufd Upper bound of the gas
velocity for the fast fluidization Greek Symbols
regime Local voidage
Umf Minimum fluidization Cross-sectional averaged
velocity voidage
/pt Particle terminal velocity Asymptotic voidage in the
Utf Lower bound of the gas upper dilute region
velocity for the fast fluidization Asymptotic voidage in the
regime lower dense region
UiT Transport velocity Local voidage at the center of a
u Linear gas velocity riser
u' Gas velocity fluctuation Voidage at the riser exit
Mfc Gas velocity in the core region amf Bed voidage at minimum
of the core-annular model fluidization
Voidage in the horizontal
uv Particle velocity
section of the L-valve
wpc Particle velocity in the core
region of the core-annular a\z Voidage in the vertical section
model of the L-valve
w pw Particle velocity in the wall ao Overall voidage
region of the core-annular Cross-sectional averaged solids
model holdup
a pc Solids volume fraction in the
We Solids entrainment rate in the
central core region
core region
Asymptotic solids volume
WQO Solids entrainment rate in the fraction in the central core
core region at z = 0 region at z = 0
WQ0Q Asymptotic limit of the Asymptotic solids volume
*pcoo
solids entrainment rate in the fraction in the central core
core region region, defined by Eq. (10.24)
Ww Solids flow rate in the wall
region Volume fraction of the solid
WWQ Solids flow rate in the wall phase in the wall region
region at z = 0 y Intermittency index, defined by
Wwp Solids feeding rate Eq. (10.22)
Wwoo Asymptotic limit of the solids Ap/Az Local pressure drop
flow rate in the wall Apcy Pressure drop across a cyclone
region Apcw Pressure drop across a solids
z Distance above the distributor flow control valve
zi Location of inflection point for Pressure drop between the riser
fast fluidization exit and the inlet of a cyclone
Z Riser height Pressure drop across the
References 455
horizontal section of the V* Dynamic viscosity of gas

L-valve P Density of fluid
Apiv Pressure drop across the elbow A Density of particles
of the L-valve * J
G Standard deviation of solids
ApY Pressure drop across a riser concentration fluctuations at a
A/?sp Pressure drop across a given point
downcomer or standpipe Standard deviation of solids
8 Angle of internal friction concentration fluctuations for
8pb Pressure perturbation in the fully segregated two-phase
blower flow with identical
8pT Pressure perturbation in the time-averaged solids
riser concentration at the same point
V Kinematic viscosity of gas
SQ Perturbation in the gas flow rate
References
Arastoopour, H. and Gidaspow, D. (1979). Vertical Pneumatic Conveying Using Four
Hydrodynamic Models. I & EC Fund., 18, 123.
Avidan, A. A. and Yerushalmi, J. (1982). Bed Expansion in High Velocity Fluidization. Powder
Tech., 32, 223.
Bader, R., Findlay, J. and Knowlton, T. M. (1988). Gas-Solids Flow Pattern in a 30.5cm Diameter
Circulating Fluidized Bed. In Circulating Fluidized Bed Technology II. Ed. Basu and
Large. Toronto: Pergamon Press.
Bai, D., Jin, Y. and Yu, Z. (1993). Flow Regimes in Circulating Fluidized Beds. Chem. Eng.
TechnoL, 16, 307.
Basu, P. and Fraser, S. A. (1991). Circulating Fluidized Bed Boilers: Design and Operations.
Boston: Butterworths.
Bi, H. T. and Fan, L.-S. (1991). Regime Transition in Gas-Solid Circulating Fluidized Beds. 1991
AlChE Annual Meeting, Los Angeles, Calif., Nov. 17-22.
Bi, H. T. and Fan, L.-S. (1992). Existence of Turbulent Regime in Gas-Solid Fluidization. AIChEJ.,
38, 297.
Bi, H. T., Grace, J. R. and Zhu, J.-X. (1993). Types of Choking in Vertical Pneumatic Systems. Int.
J. Multiphase Flow, 19, 1077.
Bi, H. T. and Zhu, J. (1993). Static Instability Analysis of Circulating Fluidized Beds and Concept
of High-Density Risers. AIChEJ., 39, 1272.
Bolton, L. W. and Davidson, J. F. (1988). Recirculation of Particles in Fast Fluidized Risers. In
Circulating Fluidized Bed Technology II. Ed. Basu and Large. Toronto: Pergamon Press.
Brereton, C. and Grace, J. R. (1993a). Microstructural Aspects of the Behavior of Circulating
Fluidized Bed. Chem. Eng. Sci., 48, 2565.
Brereton, C. and Grace, J. R. (1993b). End Effects in Circulating Fluidized Bed Hydrodynamics. In
Circulating Fluidized Bed Technology IV. Ed. A. A. Avidan. New York: AIChE
Publications.
Contractor, R. M., Patience, G. S., Garnett, D. I., Horowitz, H. S., Sisler, G. M. and Bergna, H. E.
(1993). A New Process for n-Butane Oxidation to Maleic Anhydride Using a Circulating
Fluidized Bed Reactor. In Circulating Fluidized Bed Technology IV. Ed. A. A. Avidan.
New York: AIChE Publications.
Davidson, J. F. (1991). The Two-Phase Theory of Fluidization: Successes and Opportunities. AIChE
Symp.Ser,H7(2Sl),l.
Press.
Delhaye, J. M., Giot, M. and Riethmuller, M. L. (1981). Thermohydraulics of Two-Phase System for
Industrial Design and Nuclear Engineering. Washington, D.C.: Hemisphere; New York:
McGraw-Hill.
Doig, I. D. and Roper, G. H. (1963). The Minimum Gas Rate for Dilute-Phase Solids Transportation
in Gas Stream. Australian Chem. Eng., 1, 9.
Fan, L.-S., Toda, M. and Satija, S. (1985). Apparent Drag Reduction Phenomenon in the Defluidized
Packed Dense Bed of the Multisolid Pneumatic Transport Bed. Chem. Eng. Sci., 40,
809.
Gidaspow, D. (1994). Multiphase Flow and Fluidization: Continuum and Kinetic Theory
Descriptions. San Diego, Calif.: Academic Press.
Gidaspow, D., Tsuo, Y. P. and Luo, K. M. (1989). Computed and Experimental Cluster Formation
and Velocity Profiles in Circulating Fluidized Beds. In Fluidization VI. Ed. Grace, Shemilt
and Bergougnou. New York: Engineering Foundation.
Grace, J. R. (1986). Contacting Modes and Behavior Classification of Gas-Solid and Other
Two-Phase Suspensions. Can. J. Chem. Eng., 64, 353.
Hartge, E.-U., Li, Y. and Werther, J. (1986). Analysis of the Local Structure of the Two Phase Flow
in a Fast Fluidized Bed. In Circulating Fluidized Bed Technology. Ed. P. Basu. Toronto:
Pergamon Press.
Herb, B., Dou, S., Tuzla, K. and Chen, J. C. (1992). Solid Mass Fluxes in Circulating Fluidized
Beds. Powder Tech., 70, 197.
Hirama, T., Takeuchi, H. and Chiba, T. (1992). On the Definition of Fast Fluidization in a
Circulating Fluidized Bed Riser. In Fluidization VII. Ed. Potter and Nicklin. New York:
Engineering Foundation.
Horio, M. and Clift, R. (1992). A Note on Terminology: Clusters and Agglomerates. Powder Tech.,
70, 196.
Horio, M. and Kuroki, H. (1994). Three-Dimensional Flow Visualization of Dilutely Dispersed
Solids in Bubbling and Circulating Fluidized Beds. Chem. Eng. Sci., 49, 2413.
Horio, M., Morishita, K., Tachibana, O. and Murata, N. (1988). Solid Distribution and Movement in
Circulating Fluidized Beds. In Circulating Fluidized Bed Technology II. Ed. Basu and
Large. Toronto: Pergamon Press.
Ilias, S., Ying, S., Mathur, G. D. and Govind, R. (1988). Studies on a Swirling Circulating Fluidized
Bed. In Circulating Fluidized Bed Technology II. Ed. Basu and Large. Toronto: Pergamon
Press.
Jiang, P. J., Bi, H., Jean, R.-H. and Fan, L.-S. (1991). Baffle Effects on Performance of Catalytic
Circulating Fluidized Bed Reactor. AIChE J., 37, 1392.
Jiang, P. J., Cai, P. and Fan, L.-S. (1993). Transient Flow Behavior in Fast Fluidization. In
Circulating Fluidized Bed Technology IV. Ed. A. A. Avidan. New York: AIChE
Publications.
Johnson, E. P. (1982). Better Fluid-Bed Units Ready to Make Debuts. Chemical Engineering, 89
(7), 39.
Jones, D. R. M. and Davidson, J. F. (1965). The Flow of Particles from a Fluidized Bed Through an
Orifice. Rheologica Acta, 4, 180.
References 457
King, D. (1992). Fluidized Catalytic Crackers: An Engineering Review. In Fluidization VII. Ed.
Potter and Nicklin. New York: Engineering Foundation.
Kunii, D. and Levenspiel, O. (1990). Entrainment of Solids from Fluidized Beds. I. Holdup of
Solids in the Freeboard. II. Operation of Fast Fluidized Beds. Powder Tech., 61,
193.
Kwauk, M. (1992). Fluidization: Idealized and Bubbleless, with Applications. Beijing: Science
Press.
Leung, L. S. and Jones, P. J. (1978). Flow of Gas-Solid Mixture in Standpipes: A Review. Powder
Tech., 20, 145.
Li, Y. and Kwauk, M. (1980). The Dynamics of Fast Fluidization. In Fluidization. Ed. Grace and
Matsen. New York: Plenum.
Li, J., Li, Y. and Kwauk, M. (1988). Energy Transport and Regime Transition in Particle-Fluid
Two-Phase Flow. In Circulating Fluidized Bed Technology II. Ed. Basu and Large.
Toronto: Pergamon Press.
Lim, K. S., Zhu, J. X. and Grace, J. R. (1995). Hydrodynamics of Gas-Solid Fluidization. Int. J.
Multiphase Flow, Ills, 141.
Louge, M., Mastorakos, E. and Jenkins, J. T. (1991). The Role of Particle Collisions in Pneumatic
Transport. /. Fluid Mech., 231, 345.
Nakamura, K. and Capes, C. E. (1973). Vertical Pneumatic Conveying: A Theoretical Study of
Uniform and Annular Particle Flow Models. Can. J. Chem. Eng., 51, 39.
Pemberton, S. T. and Davidson, J. F. (1986). Elutriation from Fluidized Beds. II. Disengagement of
Particles from Gas in the Freeboard. Chem. Eng. Sci., 41, 253.
Perry, R. H., Green, D. W. and Maloney, J. O. (1984). Perry's Chemical Engineer's Handbook, 6th
ed. New York: McGraw-Hill.
Pita, J. A. and Sundaresan, S. (1993). Developing Flow of a Gas-Particle Mixture in a Vertical
Riser. AIChEJ., 39, 541.
Praturi, A. and Brodkey, R. S. (1978). A Stereoscopic Visual Study of Coherent Structures in
Turbulent Shear Flow. /. Fluid Mech., 89, 251.
Rhodes, M. J. and Cheng, H. (1994). Operation of an L-Valve in a Circulating Fluidized Bed of Fine
Solids. In Circulating Fluidized Bed Technology IV. Ed. A. A. Avidan. New York: AIChE
Publications.
Rhodes, M. J. and Geldart, D. (1987). A Model for the Circulating Fluidized Bed. Powder Tech., 53,
155.
Rhodes, M. J. and Laussmann, P. (1992). A Study of the Pressure Balance Around the Loop of a
Circulating Fluidized Bed. Can. J. Chem. Eng., 70, 625.
Rhodes, M. J., Wang, X. S., Cheng, H., Hirama, T. and Gibbs, B. M. (1992). Similar Profiles of
Solids Flux in Circulating Fluidized Bed Risers. Chem. Eng. Sci., 47, 1635.
Richardson, J. F. and Zaki, W. N. (1954). Sedimentation and Fluidization, Part I. Trans. Instn. Chem.
Engrs., 43, 35.
Rudolph, V, Chong, Y. O. and Nicklin, D. J. (1991). Standpipe Modelling for Circulating Fluidized
Beds. In Circulating Fluidized Bed Technology HI. Ed. Basu, Horio and Hasatani. Oxford:
Pergamon Press.
Satija, S. and Fan, L.-S. (1985). Characteristics of Slugging Regime and Transition to Turbulent
Regime for Fluidized Beds of Large Coarse Particles. AIChEJ., 31, 1554.
Schnitzlein, M.G. and Weinstein, H. (1988). Flow Characterization in High-Velocity Fluidized Beds
Using Pressure Fluctuations. Chem. Eng. Sci., 43, 2605.
Senior, R. C. and Brereton, C. M. M. (1992). Modelling of Circulating Fluidized-Bed Solids Flow

and Distribution. Chem. Eng. Sci., 47, 281.
Shingles, T. and McDonald, A. F. (1988). Commercial Experience with Synthol CFB Reactors. In
Circulating Fluidized Bed Technology II. Ed. Basu and Large. Toronto: Pergamon Press.
Sinclair, J. L. and Jackson, R. (1989). The Effect of Particle-Particle Interaction on the Flow of Gas
and Particles in a Vertical Pipe. AIChE /., 35, 1473.
Soong, C. H., Tuzla, K. and Chen, J. C. (1995). Experimental Determination of Cluster Size and
Velocity in Circulating Fluidized Bed. In Fluidization VIII (preprints). Ed. Large and
Laguerie. Tours, France: Conference Publication.
Tsuo, Y. P. and Gidaspow, D. (1990). Computation of Flow Pattern in Circulating Fluidized Beds.
AIChE J., 36, 885.
Toda, M., Satija, S. and Fan, L.-S. (1983). Fundamental Characteristics of a Multisolid Pneumatic
Transport Bed: Minimum Fluidization Velocity of the Dense Bed. In Fluidization IV. Ed.
Kunii and Toei. New York: Engineering Foundation.
Weinstock, J. (1978). Analytical Approximations in the Theory of Turbulent Diffusion. Phys.
Fluids, 21, 887.
Wen, C. Y. and Chen, L. H. (1982). Fluidized Bed Freeboard Phenomena: Entrainment and
Elutriation. AIChE /., 28, 117.
Wen, C. Y. and Galli, A. F. (1971). Dilute Phase Systems. In Fluidization. Ed. Davidson and
Werther, J. (1993). Fluid Mechanics of Large-Scale CFB Units. In Circulating Fluidized Bed
Technology IV. Ed. A. A. Avidan. New York: AIChE Publications.
Wu, R. L., Lim, C. J., Grace, J. R. and Brereton, C. M. H. (1991). Instantaneous Local Heat Transfer
and Hydrodynamics in a Circulating Fluidized Bed. Int. J. Heat & Mass Transfer, 34,
2019.
Wu, R. L., Grace, J. R. and Lim, C. J. (1990). A Model for Heat Transfer in Circulating Fluidized
Beds. Chem. Eng. Sci., 45, 3389.
Yang, W. C. (1975). A Mathematical Definition of Choking Phenomenon and a Mathematical
Model for Predicting Choking Velocity and Choking Voidage. AIChE J., 21, 1013.
Yang, W. C. (1976). A Criterion for Fast Fluidization. Proceedings of the Third International
Conference on the Pneumatic Transport of Solids in Pipes (Pneumotransport 3), Bath,
England. E5 - 49.
Yang, W. C. (1992). The Hydrodynamics of Circulating Fluidized Beds. In The Supplements of the
Encyclopedia of Fluid Mechanics. Ed. N. P. Cheremininoff. Houston: Gulf Publishing.
Yang, W. C. and Knowlton, T. M. (1993). L-Valve Equations. Powder Tech., 77, 49.
Yerushalmi, J. and Cankurt, N. T. (1979). Further Studies of the Regimes of Fluidization. Powder
Tech.,24,\Sl.
Yerushalmi, J., Cankurt, N. T, Geldart, D. and Liss, B. (1978). Flow Regimes in Vertical
Gas-Solids Contact System. AIChE Symp. Ser, 74 (176), 1.
Yuu, S., Yasukouchi, N., Hirosawa, Y. and Jotaki, T. (1978). Particle Turbulent Diffusion in a Dust
Laden Round Jet. AIChE J., 24, 509.
Zenz, F. A. and Othmer, D. F. (1960). Fluidization and Fluid-Particle Systems. New York: Reinhold.
Zhang, W., Tung, Y. and Johnsson, F. (1991). Radial Voidage Profiles in Fast Fluidized Beds of
Different Diameters. Chem. Eng. Sci., 46, 3045.
Zheng, C. G., Tung, Y, Xia, Y. S., Bin, H. and Kwauk, M. (1991). Voidage Redistribution by Ring
Internals in Fast Fluidization. In Fluidization 91 Science and Technology. Ed. Kwauk and
Hasatani. Beijing: Science Press.
Problems 459
Problems
10.1 Calculate the transport velocity of particles with
(1) dp = 80 fim and p p = 1,500 kg/m3 (Group A)
(2) dp = 800 /xm and p p = 1,500 kg/m3 (Group B)
(3) dp = 4 mm and p p = 1,010 kg/m3 (Group D)
for a fluidizing medium with p = 1.18 kg/m3 and /z = 1.82 x 10~5 kg/m s
10.2 Assuming the riser flow to be characterized by one-dimensional steady flow and applying
the two-fluid model to the flow, one can obtain the momentum equation for the gas-solid
mixture as
dw
t +a
where Ff is the friction force. Use assumptions that the variation of gas momentum and
the friction can be neglected to prove the following equation and discuss the effects of the
axial variations of voidage on the pressure drop:
dp _ / / \ 1 d(l-g)
i fir i 7v\
f_
L (P10.2)
dz \ l - a j pp dz " ' v ^pJO
10.3 In the core-annular flow model, the mass balance of the solid phase can be expressed as
i = rf-apcUpc + (1 - r]2)UpWapw, r\ = (P10.3)

PP D
and
(1 - a) = apCrj2 + Qfpw(l - ry2) (P10.4)
Derive the following equation from Eqs. (P10.3) and (P10.4)
:
+ (1 - ^72) ( 1 - ) <VPP (P10.5)
10.4 Calculate the axial profiles of the cross-sectional averaged voidage in a CFB system with
FCC particles of dp = 200 fim and p p = 1,500 kg/m3 (amf = 0.48; Upt = 0.88 m/s) under
the following conditions:
Riser diameter D = 80 mm, and height Z = 8 m
Downcomer diameter Dsp = 250 mm
L-valve diameter Dh = 50 mm, length L h = 300 mm, and L z = 500 mm
Cross-sectional area of the cyclone Ac = 6,000 mm2
Solids inventory M = 60 kg
Gas velocity U = 3.5 and 4.5 m/s
Solids circulation rate Jp = 80 kg/m2 s
The properties of the fluidizing medium are p = 1.18 kg/m3 and /x = 1.82 x 10~5 kg/m s.
Also, discuss the solids inventory effects on the voidage profiles.
10.5 The mass balance equations for a gas-solid flow in a riser can be described as
^-(pa) + -^-(rpaur) + ^-(pauz) = 0 (P10.6)

dt r dr dz
for the gas phase and
^-[p p (l - a)] + -^-[rpp(l - a)uvA + ^-[/P (1 - ot)uvz] = 0 (P10.7)
t r r dz
for the solid phase. Here uz and ur are linear gas velocities in the z and r directions,
respectively. wpz and wpr are particle velocities in the z and r directions, respectively.
For the core-annular model, the one-dimensional continuity equations for the core and the
annular regions can be obtained from volume averaging over Eqs. (P10.6) and (P10.7).
With the assumptions given in 10.6.2, derive one-dimensional continuity equations for the
core and annular regions based on Eqs. (P10.6) and (P10.7).
10.6 The reduced solids flux profile, / p r / / p , can be empirically correlated by [Rhodes et al.,
1992]
R , . - (P10-8)
where / pr is the solids flux at radial position r. Assuming that all gas goes through the
core region and the slip velocity between particle and gas in the core region is equal to the
particle terminal velocity, determine the parameter b in the preceding equation. Also prove
that the thickness of the annular region can be expressed by
10.7 In the Davidson and Harrison's (1963) maximum stable bubble size model, the bubble
disintegration takes place when the relative velocity between the bubble and the particles
exceeds the particle terminal velocity. Considering that, for a vertical gas-solid flow system,
choking occurs when the maximum stable bubble size is equal to the column size, Yang
(1976) obtained the following choking criterion for fine particles fluidization:
< 0.35 without choking (P10.10)
> 0.35 with choking (P10.11)
Derive Eqs. (P10.10) and (P10.ll) on the basis of the conditions stated. In deriving these
equations, consider also the simple two-phase theory (see 9.4.5), which gives the particle
velocity above the gas slug as U Umf.
10.8 The particle turbulent diffusion coefficient Dp can be related to the gas turbulent diffusion
coefficient Dg by [Yuu et al, 1978]
Z>p/Dg = l + S t / 1 2 (P10.12)
D g can be expressed empirically by [Weinstock, 1978]
/k0 (P10.13)
In Eq. (P10.13), &o is the eddy number, which is assumed to be D/4, and u* = u'/2. On the
basis of the information given, verify that the deposition coefficient kd can be expressed by
Eq. (10.31) if the Sherwood number, defined askdD/Dp, for the turbulent flow is assumed
to be 4.
CHAPTER 11
Pneumatic Conveying of Solids
11.1 Introduction
In solids processing systems, it is commonly required to transport solids from one
location to another. Solids can be transported via various means, including (1) pneumatic
conveying, in which solids are transported in a pipe or channel by gasflowthrough blowing
or suction; (2) gravity chutes, where solids transport downward by gravitational force; (3)
air slides, where solids, partially suspended in a channel by the upward flow of air through
a porous wall, flow at a small angle to the horizontal; (4) belt conveyors, where solids
are conveyed horizontally, or at small angles to the horizontal, on a continuous moving
belt; (5) screw conveyors, in which solids are transported in a pipe or channel by a rotating
helical impeller; (6) bucket elevators, in which solids are carried upward in buckets attached
to a continuously moving vertical belt; and (7) vibratory conveyors, where solids flow is
activated by jigging action provided by angled spring supports.
This chapter focuses on pneumatic conveying as the means for solids transport. Examples
of solid materials that are commonly transported via pneumatic conveying areflour,granular
chemicals, lime, soda ash, plastic chips, coal, gunpowder pellets, ores, and grains [Stoess,
1983; Williams, 1983; Konrad, 1986; Soo, 1990; Marcus et al, 1990]. Materials to be
conveyed are usually dry and readily free-flowing. Some sluggish and damp materials can
also be free-flowing with proper aeration. The advantages of pneumatic conveying include
extreme flexibility in routing and spacing, safe working conditions, and low maintenance
cost. The disadvantages include relatively high power consumption compared to that of
other bulk solids transport systems and significant wearing and abrasive effects due to high
collision velocity of solids in the dilute-phase conveying systems. The transport phenomena
associated with the pneumatic conveying of solids are complex, and, therefore, special
requirements are needed in the design and operation of those systems. Variables governing
the phenomena encompass the gas velocity, characteristics of the solids (size, density,
distribution, and shape), solids loading, pipe size and configuration, solids feeding device,
and transport direction [Yang, 1987].
11.2 Classifications of Pneumatic Conveying Systems

Pneumatic conveying systems can be classified on the basis of the angle of incli-
nation of pipelines, operational modes {i.e., negative- or positive-pressure operation), and
flow characteristics (i.e., dilute or dense phase transport; steady or unsteady transport). A
practical pneumatic conveying system is often composed of several vertical, horizontal, and
inclined pipelines. Multiple flow regimes may coexist in a given operational system.
11.2.1 Horizontal and Vertical Transport

In a horizontal conveying pipeline, motion of the particles is not always straight
in the horizontal direction. Particles constantly fall down to the bottom of the pipe by
461
462 11 /Pneumatic Conveying of Solids
gravity, slide along or collide with the bottom wall of the pipe, and reenter the gas stream.
Consequently, the particle concentration is higher at the bottom than at the top of the pipe
except for dilute-phase flows of fine particles at very high transport velocities. Thus, when
the gravity effect is significant, the gas-solid horizontal pipe flows are essentially two-
dimensional regardless of the shape of the pipe [Roco and Shook, 1984; Soo and Mei,
1987]. In a horizontal pneumatic transport, the gas flow must overcome the drag forces,
the frictions, and the forces associated with particle reentrainment and acceleration. To
establish a fully suspended particle flow, a high gas velocity is required to provide upward
lift of the particles at the bottom of the pipe.
In vertical pneumatic transport, particles are always suspended in the gas stream mainly
because the direction of the gravity is in line with that of the gas flow. Hence, to trans-
port the same solids mass flow rate in a suspended flow, it takes lower gas velocity for
upward vertical conveying than for horizontal conveying. For most cases in vertical
pneumatic conveying, the radial particle concentration distribution is near-uniform and,
thus, gas and solids can be reasonably treated as being in one-dimensional flow. The drag
force, wall friction, and gravity combine to produce a pressure drop in upward vertical
conveying that is higher than that in horizontal conveying for given gas and solids flow
rates.
Inclined pipelines require the highest gas velocity to convey the same amount of solid
materials when compared to both vertical and horizontal conveying. In this case, gasflowhas
to overcome the forces associated with horizontal conveying and the tendency of materials
to slide back down the incline. In practice, a slope of less than 15 or more than 80 from
the horizontal need not be considered an incline [Williams, 1983].
11.2.2 Negative- and Positive-Pressure Pneumatic Conveyings

There are two major operational modes of pneumatic conveyor systems, i.e.,
negative-pressure or vacuum conveying and positive-pressure conveying, as shown in
Fig. 11.1. The operating pressure for the former is lower than ambient pressure while
it is higher for the latter. Vacuum conveyors are in general used for transport of materials
from several feeding points to a common collection point. Because of the employment
of exhaust systems in this type of operation, vacuum conveyors have a definite limita-
tion of operating pressure (about 0.4 atm) which restricts the capacity and the length
of a pipeline. Negative-pressure conveying is extremely useful in the transport of toxic
and hazardous materials. This type of conveyance not only provides dust-free feeding
but also prevents the escape of solids through leakage (if any) in the pipeline. Positive-
pressure conveying is commonly used to handle larger solids loadings and longer con-
veying distances without theoretical limitations. This type of operation is ideally suited
for applications with multiple discharges, where solids are picked up from one location
and delivered to several receiving locations. As expected, positive-pressure systems are
in general more complicated than negative-pressure systems. Both negative- and positive-
pressure operational modes are used for horizontal as well as vertical pneumatic
transport.
In a pneumatic conveying system, either positive or negative, materials are first fed into
a conveying pipeline and then carried along the pipeline by the gasflow.After reaching the
destination, particles are separated from the gas stream and collected in receiving bins. The
11.2 / Classifications of Pneumatic Conveying Systems 463
Transport r
pipe rI
I.I Exhauster
| Solids fan
(a)
Solids
Feeder (g) Transport Air

<
pipe
Blower
\ Solids
(b)
Figure 11.1. Pneumatic conveying systems: (a) Negative conveying system; (b) Positive
conveying system.
major components in a pneumatic conveying system therefore include a gas mover, particle
feeder, conveying conduit, gas-solid separator, and solids collector.
Blowers, fans, and vacuum pumps can be used as gas movers to provide necessary
energy to transport the solids. In a negative-pressure conveying system, the vacuum pump
is located downstream of the material collection device. In the positive-pressure conveying
systems, the blowers or fans are installed upstream of the solids feeding equipment. The
selection of a gas mover is primarily based on gas flow rate and pressure drop requirements
for the successful transport of solids.
Uniform feeding of the solids and mixing of the solids and carrier gas are essential to
the effective operation of pneumatic conveying. In a negative-pressure system, the material
feeding and gas and solids mixing are relatively independent of the feeding arrangement; in
a positive-pressure system, an air-lock type of solids feeder must be used to discharge the
materials into the conveying lines. Rotary feeders are most frequently used in the vacuum- or
low-positive-pressure systems. The particle size and abrasive properties have an important
bearing on the design of the feeding vanes. A high solids feeding rate can be obtained by
using a fluidized bed or a standpipe feeder.
For a positive-pressure system with dust-free solids, the gas-solid mixture can be directly
charged into terminal receiving bins, and the exiting gas passing through the receiving bins
can be discharged through the top vent into the surrounding atmosphere. The primary
gas-particle separator in pneumatic conveying is a cyclone. Chapter 7 provides details on
gas-solid separation.
11.2.3 Dilute Flow Versus Dense Flow

The term dense-phase flow in contrast to dilute flow has been loosely used in the
literature. However, there are several criteria commonly employed to define the operating
ranges for dense-phase pneumatic conveying, in either vertical or horizontal transport. They
include (i) a mass flow ratio of solids to gas that is greater than a certain value, e.g., 15
[Marcus et al, 1990], 20 [Leva, 1959], or 80 [Kunii and Levenspiel, 1969]; (ii) a high
volumetric concentration of solids in the pipeline, typically greater than 40 or 50 percent
[Chen et al., 1979]; (iii) the conveying of solids by a pipeline that is filled with solids at one
or more cross-sectional areas [Konrad et al., 1980]; (iv) particle-particle interaction that
is strong and cannot be neglected in comparison with the particle-fluid interaction [Soo,
1989]; and (v) transition demarcated by the minimum pressure on the curve of pressure
drop against gas velocity at a constant solids feed rate (see Fig. 11.3 in 11.2.4). Criteria
(i) and (ii) are intuitive but without significant analytical justification. Criterion (iii) is only
applicable to plug flow and is not specific enough for general use. Criterion (iv) provides
general concepts for a dense suspension but lacks quantitative specification. Criterion (v)
provides a quantitative means of demarcating the dense and dilute suspension flow. Since
the pressure variation due to transition of these suspension flows is closely associated with
the particle-particle interactions, Criteria (iv) and (v), in effect, are similar. They are the
preferred criteria used in this chapter.
Specifically, Crowe (1982) indicated that the dilute phase can be distinguished from the
dense phase on the basis of the effect of the particle-particle collisions on the particle motion.
The dilute and the dense gas-particle flows can be directly analogous to, respectively, the
free-molecule and continuum flow regimes of a single-phase fluid. In this analogy, the
parameter corresponding to the Knudsen number is the Stokes number, which is based
on the distance a particle travels between collisions. Consider a particle that has a speed
\U Up\ relative to the gas. The particle travels a distance of |U Up|r^ before it stops
in a coordinate system moving with the gas flow, where Trp is the particle relaxation time.
If this distance is small compared with the particle-particle collision distance in the same
coordinate system, one can regard the particle cloud as dilute. General comparisons of
dilute and dense suspensions are given in Table 11.1.
There are significant differences in dynamic behavior (e.g., effects of particle-particle
interactions on drag force and mean free path between collision, velocity, and concentration
distributions) between vertical and horizontal pneumatic conveying. These differences yield
distinctly different operating characteristics; e.g., the onset velocity for the dense-phase flow
in horizontal conveying differs from that in vertical conveying. The following discusses
flow regime and regime transition in both horizontal and vertical conveying systems.
11.2.4 Flow Regimes and Regime Transitions

The flow pattern in a horizontal gas-solid pipe flow varies with the solids concen-
tration. In a horizontal pipeflowat a given gasflowrate, the following flow patterns sequen-
tially occur as the particle loading, defined as the massflowratio of particles to gas, increases:
(1) Dilute suspension flow
(2) Sedimentation of particles and dune formation
(3) Stratified flow
(4) Plug flow
11.2 / Classifications of Pneumatic Conveying Systems 465
Table 11.1. Comparisons of Dilute and Dense Suspensions.
Dilute suspension Dense suspension
Relative motion between particles Large Small

Particle-particle interactions Weak Strong
Particle diffusivity Large Small
Apparent particle viscosity Due to gas-particle Due to particle-particle interaction and
interaction gas-particle interaction
Flow regime in applications Steady, turbulent Pseudolaminar to unsteady
and stratified
Motion above minimum Stable Unstable
transport velocity
Analogy in molecular system Rarefied gas flow Molecular theory of liquids
or kinetic theory of gas
Source: S. Soo's Particulates and Continuum, Hemisphere, 1989.
(5) Moving bed flow at relatively high gas velocity or solid ripples traveling along the
top of a stationary solid layer at relatively low gas velocity
As shown in Fig. 11.2(a), at a very low solids concentration, particles are fully suspended
and quite uniformly distributed throughout the pipe, as is characteristic of dilute suspension
flow. With increasing particle loading, sedimentation of particles occurs, as shown in
Fig. 11.2(b). Particles tend to settle at the bottom of the pipe and may slide over other
particles. In this flow, particle dunes are formed and particle flow takes place by way of
moving from one dune to another and undergoing alternate acceleration and deceleration.
In a low-velocityflow,dunes are formed more readily and the length and height of the dunes
are correspondingly larger. An increase in the particle loading yields a stratified flow with a
wavy interface, as shown in Fig. 11.2(c). A further increase in the particle loading leads to the
formation of a solid plug characterized by the intermittentflowof gas and solids in alternating
plugs, as shown in Fig. 11.2(d), and eventually results in a complete blockage of the pipe. As
can be seen from Fig. 11.2(e), at a relatively high gas velocity a moving bedflowoccurs; at a
relatively low gas velocity, the solids travel in the form of ripples at the top portion of the pipe
while a large portion of solids remain stationary at the bottom. Thus, dense-phase pneumatic
transport through horizontal pipes is usually unstable, with large pressure fluctuations.
The flow regime diagram for a pneumatic conveying system can generally be described
in terms of the variation of the pressure drop per unit length with the superficial gas veloc-
ity given in Fig. 11.3, in a manner similar to that for a circulating fluidized bed given in
Fig. 10.2. The effect of the solids mass flux, shown in Fig. 11.3, indicates that an increase
in the solids mass flux increases the pressure drop at a given gas velocity. It is seen in
the figure that the pressure drop per unit length decreases and then increases as the gas
velocity increases. The minimum pressure drop point marks the dense flow and dilute
flow transition for the pneumatic conveying system. For vertical conveying, the choking
phenomenon discussed in 10.3.1 may also occur, depending on the pipe size and particle
properties [Leung, 1980]. For horizontal conveying, the saltation phenomenon will take
place at the minimum pressure drop point for coarse particles, where particles start settling
466 11 / Pneumatic Conveying of Solids
' J - ' . . ' . '
''.!
(a)
(b)
. ' . ^ f a - - . . . .v ../:.. . .' . > . ; . . . .
(c)
(d)
:j?.';.-?:':>::-::-:-?.y.-'?:':y
:;}:;.;;:'.; V.';i::v|: s-* : .i
(e)
Figure 11.2. Flow regimes for horizontal conveying of solids: (a) Dilute suspension flow;
(b) Sedimentation and dune formation; (c) Stratified flow with wavy interface; (d) Plug flow;
(e) Moving bed flow.
at the dilute and dense flow transition. For fine particles, the saltation phenomenon will
take place at gas velocity higher than that for the minimum pressure drop point [Marcus
et al, 1990]. The flow regimes in pneumatic conveying of fine particles for both vertical
and horizontal systems can be further described by the diagrams in log-log scales, as given
in Fig. 11.4. The diagrams illustrate that the power-law relationship (or linear relationship
in log-log scales) between the pressure drop and the superficial gas velocity holds for dilute
suspension flow of fine particles. The diagrams further indicate that, on the basis of visual
observations of theflowregimes, dense-phase flow is typically represented by plug flow and
moving bed flow. For comparisons the pressure drop for thefixedbed is also indicated in the
figure. The unsteady-state plug-dune flow for horizontal conveying and the unsteady-state
nonuniform dispersed flow for vertical conveying denoted by broken lines in the diagrams
appear to be the regimes which are in need of further study.
11.3 Pressure Drop

For transport of a gas-solid suspension, it is important to be able to evaluate the
pressure drop in order to estimate the power consumption. In a pneumatic conveying system,
particles are usually fed into the carrying gas and accelerated by the gas flow. The power
consumption (or pressure drop per unit pipe length) for the acceleration of particles can be
significant compared with that for a fully developed flow condition. Moreover, the pressure
11.3 / Pressure Drop 467
Dense flow Dilute flow
Gas alone (J = 0)
Superficial gas
velocity
Figure 11.3. General flow regime diagram for a pneumatic conveying system (from Marcus
et aU 1990).
drop over a bend or branch is also high because of the deceleration (by the particle-wall
collision) and reacceleration of particles. In the following, a general expression of pressure
drop in one-dimensional pipe flows is first introduced. A unique phenomenon of drag
reduction associated with very dilute suspension flows is then discussed. In addition to
pressure drops for solids acceleration, the acceleration length in the developing flow region
is also illustrated. The calculation of the pressure drop over a bend is included in this section;
however, a detailed discussion of velocityfieldsof gas-solidflowsin a bend is given in 11.5.
11.3.1 General Pressure Drop in One-Dimensional Flow

The role of the pressure gradient may be shown in the momentum equation of a
gas-solid mixture. Consider a steady pipe flow without mass transfer and with negligible
interparticle collisions. From Eq. (5.170), the momentum equations for the gas and particle
phases can be given by
V Jm = -Vp + V . T e + F A + Otpg (11.1)
and
V- = -FA 1 - a)ppg (11.2)
where Te denotes the effective stress tensor per unit volume; FA accounts for the interfacial
interactions between the gas and solids; FE describes the electrostatic forces of particles;
and Jm and Jmp are the momentum fluxes of gas and particle phases, respectively. Jm and
J m p are defined by
Jm = paUU - pD[(Va)U + UVa] (11.3)
Log (superficial
gas velocity)
(a)
Log (superficial
gas velocity)
(b)
Figure 11.4. Flow regime diagrams on log-log scales for pneumatic conveying of fine
particles (from Marcus etaL, 1990): (a) Horizontal flow; (b) Vertical flow.
and
= Ppd - a)UpUp + PpDv[(Va)Uv + UpWa] (11.4)
where D and D p represent the diffusivities of gas and particle phases, respectively. The
momentum equation of the mixture is obtained by
V (Jm +/mp) - V T e - FE - Otpg - (1 - a)Ppg = -Vp (11.5)
For one-dimensional horizontal pipe flows without electrostatic effects, using Gauss's the-
orem, Eq. (11.5) yields
P\ Pi = (11.6)
where A is the cross-sectional area, Cw is the perimeter, and x is the axial coordinate.
Equation (11.6) shows that the axial pressure drop is balanced by the change of axial
viscous force, wall friction, and increase of momentum fluxes of the phases.
11.3.2 Drag Reduction

On the basis of Eq. (11.6), it can be deduced that, with the addition of particles into
a flow stream, the pressure loss would be higher than that of a single-phase flow. However,
experimental data have revealed that under certain operating conditions (typically, the mass
loading ranging from 0.5 to 2), the pressure drop of the two-phase mixture is much less
than that of a particle-free flow [Halstrom, 1953; Thomas, 1962; Peters and Klinzing, 1972;
Kane, 1989]. This unique phenomenon, termed drag reduction, is illustrated in Fig. 11.5
for the acceleration region [Shimizu et al., 1978] and in Fig. 11.6 for the fully developed
region [Kane and Pfeffer, 1973]. The phenomenon reflects that it is possible to consume less
energy to transport a two-phase flow than to transport a single-phaseflow.The effect of drag
reduction is significant for small particles (less than 200 /xm) in turbulent pipe flow. The
decrease in the pressure drop can be up to 30 percent [McCarthy and Olson, 1968; Boyce
and Blick; 1969]. It is generally recognized that the effects of particle addition on the drag
reduction are mainly through turbulence modulation. The phenomenon of drag reduction
is reported to be linked to various factors associated with particle motion near the wall
region, including particle spinning, deposition, agglomeration, and electrostatics [Kane,
1989]. These factors ultimately affect the velocity profiles of gas and particles, and particle
concentration distributions in the gas-solid flow system. The basic mechanisms of drag
reduction are not yet fully known, in part because of a shortage of reliable measurements
which quantify these factors. In the following a phenomenological model which accounts
for the drag reduction phenomenon from the viewpoint of turbulent intensity attenuation
due to the effect of particle addition is introduced.
2.0
1.5
1.0 Pipe diameter

76.2 mm
50.8 mm
0.5 I
0 1 2 3 4 5 6
m
Figure 11.5. Pressure drop ratio as a function of mass flux ratio for 10 /xm zinc particles in the
acceleration region of Re = 53,000 (after Shimizu et al, 1978).
A 19,000 < Re < 24,000

B 17,000 < Re < 19,000
C 15,000 < Re < 17,000
D 10,000 < Re < 15,000
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2

m
Figure 11.6. Pressure drop ratio as a function of mass flux ratio for 36 iim glass beads in a
fully developed region (after Kane and Pfeffer, 1973).
It is recognized that within the range where drag reduction occurs, the solids concentra-
tion is so dilute that the averaged distance between particles is usually 10 or more particle
diameters. Therefore, under this flow condition, interparticle effects can be neglected. Con-
sider the case of a fully developed horizontal pipe flow with negligible electrostatic effects.
From Eq. (11.6), the pressure drop depends only on the wall friction, as given by
A/? _ Cw(rgp + rwp)
(11.7)
AJC ~ A
where rgp denotes the friction induced by the gas phase in the presence of solid particles; rwp
represents the friction due to collisions of particles with the wall, which can be expressed
by [Soo, 1962]
' (11.8)
The equation indicates that r wp can be estimated from the particle volume fraction, the
particle velocity, and the intensity of the turbulent motion of particles at the wall. If we
approximate rgp by the friction due to the gas flow without particles, the ratio of r wp to rgp
can be roughly estimated by
(11.9)
where m is the mass flux ratio of particle to gas; / is the friction factor of a single-phase
pipe flow; the first pair of brackets is the turbulent intensity of the gas phase; and the second
one is the ratio of the velocity fluctuation of particles to that of gas.
For the Reynolds number range typical of drag reduction (Re~ 105), / is about 0.02
from the Moody chart (see Fig. 11.7). The typical turbulent intensity of gas in a pipe flow
is about 5 percent. Using the Hinze-Tchen model (see 5.3.4.1), the ratio of the velocity
fluctuation of the particles to that of the gas may be given by Eq. (5.196) as
_i
(11.10)
2
i
0.01
0.009
0.00001
0.008
10 10' 10
Reynolds number Re
Figure 11.7. Moody chart for pipe friction with smooth and rough walls (from Moody, 1944).
where Ts is the Stokes relaxation time defined by Eq. (3.39) and Xf is the characteristic time
of the turbulent fluctuation of gas and is given by
Tf=-7= (H.H)
where /m is the mixing length of turbulence. For a fully developed turbulent pipe flow, / m
near the wall is about 0.4, where is the distance from the wall [Schlichting, 1979]. Thus,
we may take lm = 0.4dp, where dv is the particle diameter. Substituting Eq. (3.39) and
Eq. (11.11) into Eq. (ll.lO) yields
= ii ^p4> (11.12)
i U' L IL 72 p D
7.2 Dd U
where Re is the Reynolds number, based on the pipe diameter Dd and averaged flow velocity
U. Hence, Eq. (11.9) becomes
'-wp
(11.13)
r
gp 7.2 p Dd U
Substitution of Eq. (11.13) into Eq. (11.7) gives
Ap _ Cwrgp
(11.14)
Ax ~ A
It is noted from measurements that within the range of particle loading where the drag
reduction takes place, the velocity profile of the gas phase is nearly unaltered by the presence
472 7 7 / Pneumatic Conveying of Solids
of solid particles [Soo et ah, 1964]. Thus, from Eq. (11.14), for drag reduction to occur,
the turbulent viscosity of the gas phase in the presence of particles, />igp, has to satisfy the
condition
where fiQ is the effective turbulent viscosity in a single-phase flow. Equation (11.15) sug-
gests that the drag reduction is due to the reduction of the turbulent viscosity of the gas
phase by the presence of particles in the stream. More discussion on the gas turbulence
modulation by particles is given in 6.6. Other explanations of the drag reduction phe-
nomenon are linked to the non-Newtonian nature of the gas-solid mixtures [Thomas, 1962;
Peters and Klinzing, 1972]. In addition, there is also an observed reduction in the con-
vective heat transfer coefficient in pipe flows when drag reduction occurs [Tien and Quan,
1962].
Example 11.1 Estimate the ratios of r wp to rgp for the following cases: (1) for 100 /xm
glass beads in air with a mass flux ratio of 1, pp/p of 2,000, Dd of 50 mm, and U of
15 m/s; and (2) for 10 /xm alumina in air with a mass flux ratio of 1, pp/p of 2,400, Dd of
50 mm, and U of 30 m/s. The turbulence intensity is assumed to be 5 percent. The pipes
are hydraulically smooth. The air kinematic viscosity is 1.5 x 10~ 5 m 2 /s.
Solution For case (1), Re can be calculated as
v 1.5 x 10~5
From the Moody chart (Fig. 11.7), the friction factor of smooth pipe for Re of 5 x 104 is
0.021. Thus, the ratio of rwp to rgp is obtained from Eq. (11.13) as
lwp ^ Am *M\ | Repp dp

U [ 7.2 p Dd U
4x1
" 0.021^ L 7.2 ' 5 x 10- 2
= 0.15 (El 1.2)
For case (2), Re can be calculated as 105, which gives rise to / of 0.018. Thus, on the
basis of a similar procedure to case (1), t w p /r g p is obtained as 0.34. Both cases show that
rgp plays a dominant role in the determination of r wp.
11.3.3 Pressure Drop and Acceleration Length in Developing Regions

The pressure drop over the developing regions of gas-solidflowsis often significant
or even dominant. Typical developing regions of pneumatic conveying systems include the
regions at the particle entrances (i.e., solids feeding ports) and regions immediately beyond
pipe bends. In the developing regions of pneumatic conveying systems, particles are usually
11.3 /Pressure Drop 473
accelerated. The acceleration length of solid particles can be much longer than that of gas
as a result of the higher inertia of solid particles.
Consider a one-dimensional gas-solid pipe flow where the electrostatic effect is negli-
gible. The momentum equation of the particle phase can be expressed as [Klinzing, 1981]
dl/ p 3 (U - Up)2 . 2fvU2v

sin
4 pvdp # ^ Dd (11.16)
where the drag coefficient C D is in general a function of the particle volume fraction; f) is
the inclined angle of the pipe ()8 = 0 for horizontal pipe flows and p = 90 for upward
pipe flows); and / p denotes the friction coefficient of particles at the wall. Noting that
dx = Up dt, the acceleration length L a is obtained from Eq. (11.16) as
u=
where UV{ and /poo represent the particle velocities at the inlet and in the fully developed
region, respectively. The pressure drop over the acceleration region can be expressed by
fLa a A ^ fLa2fJUdx [L* ifpJpUpdx
= / apPpgsmpdx+ / + / Dd
Jo Jo Dd Jo
(11.18)
where J and Jv are mass fluxes of gas and particles, respectively.

The acceleration length and the pressure loss in that length can also be estimated by
using empirical correlations. Rose and Duckworth (1969) reported that
and
(^)() (11.20)
The gas-solid flow through a bend represents another typical case of developing regions,
in which particles are first decelerated by gas impingement and collision with the wall and
then accelerated by the gas after the bend turn. A general empirical relation for the estima-
tion of the additional pressure drop due to particles over a bend can be used [Klinzing, 1981]
ApL \DdJ \Dd/

where RB is the bend radius; A/?L is the pressure drop over an equivalent length of a straight
pipe; and a, b, and c are empirical constants. For coarse particles (1-3 mm) and for ap < 5
percent, Schuchart( 1968) suggested that a = 210, b = 1.15, and c = 0. ApL in Eq. (11.21)
may be determined by [Schuchart, 1968]
:D (11.22)
where A is an empirical constant and L B is the equivalent length of a straight pipe.

It is noted that the total pressure drop over a bend includes contributions from both the
gas and particle phases. The pressure drop of gas over a bend, ApBg, can be estimated by
[Ito, 1959; 1960]
0.029 + 0.304[Re(D d /2/? B ) 2 r 0 - 25 / / Tj2N
L' B-\^ ~(pU2\
(1L23)
D&J \ I J
The preceding equation is valid for 0.034 < Re(Dd/27?B)2 < 300. When Re(D d /2/? B ) 2 <
0.034, the bend yields the same pressure drop as for the straight pipe.
Example 11.2 Use Eq. (11.17) to derive a general expression for the acceleration length
for dilute gas-solid pipe flows. Assume that the Stokes drag coefficient can be used. The
friction coefficient of particles at the wall can be estimated by [Konno and Saito, 1969]
/ p = 0.0285^^ (El 1.3)
Solution For diluteflows,the particle volume fraction av < 1 and, thus, the drag
coefficient Co can be approximated to be independent of ap. Substituting Eq. (El 1.3) into
Eq. (11.17) and noting that CD = 24/Re p , we have
= f (
P (U
~ U^ JjL\ UpdUp (EH.4)
where rs is the Stokes relaxation time defined by Eq. (3.39). Equation (El 1.4) can be
integrated directly to become
lUpi A-BUV B2 \A-BUVOOJ B

where
U
(El 1.6)
11.4 Critical Transport Velocities

For a horizontal flow, the direction of the gravitational force is perpendicular to
that of the main stream of the flow. Gravitational settling of particles occurs when the gas
velocity is not sufficiently high. The minimum transport velocity and pickup velocity are
defined to account for the transition of the settling phenomenon. The minimum transport
velocity is the minimum mean gas velocity required to prevent the accumulation of a layer
of stationary or sliding particles on the bottom of a horizontal conveying pipeline. The
minimum transport velocity is equivalent to the saltation velocity, below which saltation
occurs in the pneumatic conveying system. The gas velocity required to pick up the particles
at rest is higher than the saltation velocity because additional interparticle forces or particle-
wall forces (such as cohesive and van der Waals forces) must be overcome during the particle
pick-up. In this section, simple models for the minimum transport velocity and pick-up
velocity are presented.
77.4/ Critical Transport Velocities 475
11.4.1 Minimum Transport Velocity

The vertical motion of solids in a horizontal suspension flow is strongly influenced
by the ratio of the terminal settling velocity to the friction velocity [Blatch, 1906; Chien
and Asce, 1956]. In a circular pipe, the mean stream velocity can be related to the friction
velocity by [Taylor, 1954]
= 5 log (BM) _3.90 (11.24)
where Uf is the friction velocity based on the mixture density and is defined by
/ *w ^ / AiAp
\ J a p +( \ ) p Y 4L[ap + ( l ) p ]
The friction velocity at the minimum transport condition may be related to the system
configuration and the operating condition by a two-step correlation method, i.e., first, to
obtain the velocity at the minimum transport condition under infinite dilutions and, second,
to correct for the concentration dependence [Thomas, 1962]. The functional dependence
of Uf to the solids concentration is given by
where Upt is the particle terminal velocity and Ufo is the friction velocity at the minimum
transport condition and zero solids concentration. Furthermore, Ufo can be expressed by
IO>dj (11.27)
for dp > 5fi/pUfo, and
^=OmfiEÊ\ ' (11.28)

for dn <
Example 11.3 Consider a gas-solid horizontal pipe flow. The pipe diameter is 50 mm.
The particle used is 50 /xm glass bead with the density of 2,500 kg/m3. The average particle
volume fraction is 0.1 percent. The gas density and kinematic viscosity are 1.2 kg/m3
and 1.5 x 10~5m2/s, respectively. Estimate the minimum transport velocity and power
consumption per unit length.
Solution The particle terminal velocity Upt can be calculated from Eq. (1.7) as
= (*?)* = (50 x 10-*)'x (2,500- 1.2) x 9.8 =

p
ISpv 18 x 1.2 x 1.5 x 10"5
The corresponding particle Reynolds number at terminal velocity, i.e., Re t, is 0.67, which
falls within the correct Ret range for Eq. (1.7) to be applied to calculate Upt. The friction
velocity at the minimum transport condition and zero solids concentration, /fo, can be
obtained from Eq. (11.28) as
J J = .93m/s
The corresponding dpUfo/v is 3.1, which verifies that Eq. (11.28) is a correct equation to
use. The friction velocity at the minimum transport condition, Uf, can thus be given by
Eq. (11.26) as
^Y^\ = 0.93^1 + 2.8
The minimum transport velocity U can be determined from Eq. (11.24) as
U = Uf\5log(^] -3.90 =4.91og(3.33 x 10 3t/) - 3.82 (Ell.10)
which yields U = 19.8 m/s.

The power consumption per unit length (i.e., Ap/L) is estimated from Eq. (11.25) as
-Z- = [appp + (1 - av)p] l- = [0.001 x 2,500 + 1.2]- = 280Pa/m
(Ell.11)
11.4.2 Pick-up Velocity

Particle pick-up is affected by two major factors, i.e., hydrodynamics and particle-
wall collisions. When the hydrodynamic forces overcome the gravitational force, the par-
ticle can be lifted away from the wall. The bouncing effect of particle-wall collisions can
also hinder particles from settling. For fine particles, the particle motion is controlled by
gas motion and turbulent diffusion. In this case, the influence of particle-wall collisions
is less significant since the particles follow the gas flow after wall collisions. For large
particles, the particle motion is dictated by particle inertia and is not strongly influenced by
flow turbulence or by the changes in the mean flow. Thus, the motion of large particles is
dominated by the particle-wall collision.
The mechanisms of a single particle-wall collision are given in Chapter 2. A particle-
wall collision in pneumatic transport systems is a complex process. The bouncing charac-
teristics depend on many parameters, including impact angle, translational and rotational
velocities of the particle before collision, physical properties of the wall and particles, and
wall roughness and particle shape.
Saltation of solids occurs in the turbulent boundary layer where the wall effects on the
particle motion must be accounted for. Such effects include the lift due to the imposed
mean shear (Saffman lift, see 3.2.3) and particle rotation (Magnus effect, see 3.2.4), as
well as an increase in drag force (Faxen effect). In pneumatic conveying, the motion of a
particle in the boundary layer is primarily affected by the shear-induced lift. In addition,
the added mass effect and Basset force can be neglected for most cases where the particle
11.4/ Critical Transport Velocities All
Gravity (mg)
Figure 11.8. Forces acting on the particle on an inclined surface.
acceleration is moderate. In terms of the preceding considerations, the motion of a particle

in the wall region of an inclined pipe, as shown in Fig. 11.8, can be expressed as
d/D Tt
m- dt = -pd'CD\U - UV$U - Up) -Ff-mg sin/? (11.29)
for the x -direction and
=
d(u - uv) (U - Uv)-mgcos0
& z dy
(11.30)
for the j-direction. Ff is a friction force, which can be expressed as

Ff = f(mg cos ft - F$ (11.31)
where Fi is the lift force.
A single particle, initially at rest at the bottom wall of the conveying pipe, may undergo
various states of motion when the gas velocity increases. Single-particle motion can occur
in one or a combination of the following states [Zenz, 1964]:
(1) Rolling, sliding, or bouncing along the bottom of the pipe in the direction of the
flow.
(2) Swept up without saltation and obviously rolling and bouncing.
(3) Picked up from rest at the bottom of a pipe and transported in suspension.
The incipient gas velocities for each state have been analyzed by Halow (1973) and are
elaborated in the following.
The incipient gas velocity for a particle rolling along the wall can be determined by the
torques exerted on the particle. Assuming that the drag force, gravitational force, and lift
force are exerted through the central point of the sphere, the balance of the torques about
the uppermost point of the particle-wall contact plane (see Fig. 11.8) can be given by
FD cos 0 + F\ sin 6 = mg sin(0 + ft) (11.32)
where 0 is an inclined angle between a line through the particle center and perpendicular
to the contact surface and a line connecting the particle center and the uppermost contact
point on the wall. Any imbalance of the torques will lead to a rolling of the particle.
The particle will slide when the drag force exceeds the frictional force between the
particle and the pipe wall in the direction parallel to the flow. From Eq. (11.29), this
condition can be expressed as
-UP\(U - Up) > Ff + rag sin (11.33)
for uphill sliding and
^pd%CD\U -UP\(U -Up) < Ff-mg sin0 (11.34)

for downhill sliding.
At the point where the particle just begins to be picked up from rest, the upward forces
have to be equal to the downward forces so that dVv/dt = 0, i.e.,
d(U - /p)
- UP\(V - Vp) (U -Uv)-mgcosfi =0
dy
(11.35)
To develop a criterion to predict the incipient points for rolling, sliding, and pick-up,
the gas velocity profile in the boundary layer of the wall region is required. The preceding
discussion refers to a flowing gas stream with a single particle in the wall region. When
a layer of particles is encountered at the bottom of the pipe, the minimum gas velocity
required to pick up a particle from rest and transport it in suspension can be 2.5 times larger
than that for the single-particle case as a result of the extra energy needed for overcoming
the interparticle forces [Zenz, 1964]. Similar accounts can be extended to other situations
involving finite solids concentrations in the wall region. The incipient gas velocities for
picking up a particle from rest are of practical importance and have received much attention
in theoretical treatment.
11.5 Flows in Bends

The gas-solid flow in a pipe bend is of major interest because it contributes sig-
nificantly not only to pressure drop effects but also to erosion problems that result from the
directional impacts of the solids. Curved pipes or bends provide a large degree of flexibility
to gas-solid transport systems by allowing routing and distribution. A basic description of
the flow of a suspension through bends is thus considered critical from the viewpoint of
transport system design. In the following, the flow field of a single-phase flow in a curved
duct is first discussed. The motion of a single particle in such a system is then examined by
analyzing its momentum transfer in the bend. The pressure drop of gas-solid flow in the
bend may be linked to the particle deceleration induced by the bend.
11.5.1 Single-Phase Flow in a Curved Pipe

It is shown in Fig. 11.9 that when a pipe is curved, the distribution of velocity in
a bend is appreciably altered by the secondary flow and the highest axial velocity occurs
11.5 / Flows in Bends 479
(a)
(b)
Figure 11.9. Schematic representation of flow in curved pipes: (a) Axial velocity distribution;
(b) Velocity distribution in the shedding layer; (c) Notations of pipe cross section.
near the outer wall. The secondary flow takes place primarily in a layer called the shedding
layer near the wall. The thickness of the shedding layer 8 is defined as being equal to the
distance from the wall to the point where the peripheral velocity component W changes its
sign, as shown in Fig. 11.9(b).
Let RB be the radius of curvature of the pipe axis and R^ be the radius of the circular
cross section of the pipe. Define U as the axial velocity component and %(=Rd r) as the
distance normal to the wall. Denote 0 as the angle in the transverse plane with respect to
the outward direction of the symmetry line and 0 as the angle measured in the plane of the
curved pipe axis, as shown in Figs. 11.9(a) and (c). Assume that the changes of the flow
pattern along the axis of the bend can be neglected. Thus, the momentum integral equations
for the shedding layer are given by [Ito, 1959]
2 sin u2
P ^ I w di= = -zeRd + P8-^-UI # -PITsin0 I ^ ( 1L36 )
ou Jo ^B KB JO
and
r8
P-^ ( y " U^Wd^ = " T ^d - -rh (H.37)

60 Jo RB aq>
where subscript " 1 " denotes the value at the edge of the shedding layer and the pressure
gradient is given by
d p
dcp
-
7tRd
2R
* fn-AO
Jo
f (11.38)
The velocity profile of U can be estimated to be proportional to f1/7 throughout the

shedding layer, i.e.,
J (11.39)
On the other hand, W increases from zero at the wall to a positive value and then decreases
to zero at the edge of the shedding layer. Thus, W is assumed to have the form
where B is a factor to be determined later. The tangential velocity near the wall V t is thus
expressed by
Vt = y/u2 + W2 y / \ + B2U{ I | J (11.41)
The shear stress at the wall can be given in terms of Vt and f as [Schlichting, 1979]
/ \ 1/4
7/4
rw = 0.0225pV t f-J (11.42)
where v is the kinematic viscosity. Substitution of Eq. (11.41) into Eq. (11.42) yields
rw = 0.0225 p(Uxy/\ + B2)1/A(-\ (11.43)
With the relationships of T#/T0 = W/U and r | + T^ = r2, it gives rise to

tw BTW
(1L44)
Since the maximum velocity in the secondary flow is lower than the axial velocity at the
edge of the shedding layer, B is much smaller than unity. Thus, we have
U &Um, W ^ BUm, r 0 & rw, xe Br w
/4
(11.45)
11.5/Flows in Bends 481
where Um is the averaged axial velocity and Re is based on Um and pipe diameter 2Rd. As,
in arriving at Eq. (11.36) for the momentum balance, the value for each term in the equation
is taken to be of the same order of magnitude, B should have the asymptotic form
B2^ (11.46)
and the thickness of the shedding layer can be estimated by
B and 8 can also be determined numerically by solving the momentum integral equations,
Eqs. (11.36) and (11.37), with the assumed profiles, Eqs. (11.39) and (11.40), and with
the wall shear stress, Eq. (11.44). The axial component in the central core of the fluid is
assumed to be in the form
U = Um + Ar cos 0 (11.48)
where the constant A is determined by the condition of continuity of the secondary flow.
Thus, an expression for the friction factor of a curved pipe can be obtained from its
definition given below
Thus, / B can be expressed by

1/10
where C is a constant with a value of about 0.3 [Ito, 1959].
11.5.2 Particulate Flow in a Bend

When a dilute gas-solid suspension flow passes through a curved pipe, the particle
velocity is lowered by the wall friction, the effects of gravitation, and collision with the
wall. The analysis of flows in pipe bends may be represented by three typical arrangements
of bends:
(1) A pipe bend in a horizontal plane
(2) A pipe bend in the vertical plane with a horizontal approach flow
(3) A pipe bend in the vertical plane with a vertical approach flow
Without losing generality, in this section we only consider case (3), where the pipe bend is
located in the vertical plane with a vertical gas-solid suspension flow at the inlet, as shown
in Fig. 11.10. It is assumed that the carried mass and the Basset force are neglected. In
addition, the particles slide along the outer surface of the bend by centrifugal force and by
the inertia effect of particles. The rebounding effect due to particle collisions with the wall
is neglected.
The equation of motion of a single particle in the direction of the pipe axis is expressed
by
Up dUp 3 p C D ,TT TT X2
= (U Un) cos d>
dp
mUn
Figure 11.10. Pipe bend in the vertical plane with vertical approach flow.
where / p is the sliding frictional coefficient between the particles and the pipe wall. The
last term in the preceding equation requires the normal force against the wall to be positive
within the range of 0 considered, i.e.,
p
(11.52)
Hence, once the local gas velocity and drag coefficient are given, the variation of the particle
velocity along the bend can be obtained by numerical integration of Eq. (11.51).
In order to estimate the extent of decrease in the particle velocity due to the pipe curvature,
it is assumed that the drag force is negligible in comparison with the sliding force of the
particle on the bend surface. With this consideration, Eq. (11.51) is reduced to
-^ + 2/p/p2 = -2(Rd - / p sin0) (11.53)
With the boundary condition of Up = /po at 0 = 0, the preceding equation gives rise to a
solution for Up as
oS* + (1-2/?) sin*)
(11.54)
which yields the expression for the particle velocity at the exit of a right angle bend as
2
-2/P
(11.55)
+ 4/ p 2
11.6 Fully Developed Dilute Pipe Flows

To analyze the dynamic behavior of gas-solid pipe flows, the most common and
easiest system to consider is a dilute gas-solid pipe flow which is fully developed and is
subject to the effects of electrostatic force and gravitational force. The fully developed flow
here refers to the situation where the velocity profiles of both gas and particles are unchanged
along the axial direction. The system of this nature was analyzed by Soo and Tung (1971).
In this section, the analysis of Soo and Tung (1971) is presented. It is assumed that no
particles are deposited on the wall surface of the pipe (or the particle deposition rate is zero).
Moreover, the pipe flow is considered to be turbulent, as is true for most flow conditions.
11.6/Fully Developed Dilute Pipe Flows 483
Figure 11.11. General coordinates for the analysis of fully developed dilute pipe flows.
For dilute suspensions, particle-particle interactions can be neglected. The extent of

transfer of particles by the gradient in the particle phase density or volume fraction of
particles is proportional to the diffusivity of particles D p . Here Dp accounts for the random
motion of particles in the flow field induced by various factors, including the diffusivity of
the fluid whether laminar or turbulent, the wake of the particles in their relative motion to the
fluid, the Brownian motion of particles, the particle-wall interaction, and the perturbation
of the flow field by the particles.
The flow system consists of a circular pipe with an inclined angle of j$ to the direction
of gravitational acceleration g, as shown in Fig. 11.11. Thus, for ft of 90 the flow is
horizontal and for f$ of 0 or 180 the flow is vertical.In Fig. 11.11, z, r, and 0 are the axial,
radial, and azimuthal coordinates; U, V, W are the conjugate components of velocity of
the fluid; and Up9 Vp, and Wp are those of the particle phase. The particles are assumed to
be monodispersed and spherical.
11.6.1 Basic Equations and Boundary Conditions

To yield a constitutive relation for the mass flux of particles, it is convenient to
begin with the equation of motion of a single particle, which is expressed by
TC
(11.56)
6 - *> U + F G
where FG is the gravitational force including buoyancy effect and is given by
- - ) g (11.57)
and FE denotes the electrostatic force as

FE = qE = -qVVE (11.58)
where q is the electrostatic charge carried by each particle and VE is the electric potential.
For a steady and fully developed flow in an infinitely long pipe, VE satisfies the Poisson
equation in the form
(11.59)
rdr\ dr ) ' r2 dO2 m 0
where Q is the permittivity in a vacuum.
It is convenient to express Eq. (11.56) in an alternative form

dtfp 7-7Ep / p\ qVVE / i i m
l = + 1 - )g - m (11.60)
dt Trp V Pp/
where r^ is the particle relaxation time, given by
1 _ 3 p \U-U0\ 3 M C D R ^ ( n 6 1 )
where Rep is defined by Eq. (3.2). It should be noted that, in general, the drag coefficient
CD is a function of Rep and the local volume fraction a.
For a fully developed flow, dUp/dt is zero and, thus, Eq. (11.60) becomes
)g (11.62)
Pp
which leads to an expression of the particle mass flux due to relaxation under the gravitational
force and electrostatic force as
(1 - a)pv(Uv -U) = (l- cOpptrp | Y l - \ - ^ ^ 1 (11.63)
Since a fully developed motion excludes deposition of the solid particle phase under
field forces (i.e., zero deposition rate), the flux due to diffusion must be equal to the flux
due to relaxation under the field force, i.e.,
-Z>PV(1 - a)pv + (1 - a)pp(/p - 0 = 0 (11.64)
where Dp is the particle diffusivity. Thus, we obtain a relation for the mass flux of particles
in the radial direction
3 (1L65)
Trp(l -a) dr m dr \( pp )
and in the azimuthal direction
(11.66)
- a)r dO mr dO \ pp
It is seen that Eqs. (11.65) and (11.66) are not independent because both can be integrated
to the form
^ l n ( l - a ) = Ci - - V E - g r s i n c o s 6 > ( 1 - ) (11.67)
tip rn V PP/
where the constant of integration C\(0) = C\(r) = C\ is determined by the boundary
conditions of particle concentration (see Example 11.4 and Problem 11.5). To obtain the
velocity profiles of both gas and particle phases, we need to examine further the momentum
equations for each phase. A fully developed flow requires that
~=0; ~=0; V = V p = W = Wp = 0 (11.68)

ot oz
Therefore, we only need to investigate the axial component of the momentum equations.
Equation (11.59) indicates that the electrostatic force does not contribute to the momentum
transfer in the axial direction. For high Reynolds number and Froude number of the gas
phase, the gravity effect does not influence the velocity distribution of that phase. The
Froude number here is defined as
(1L69)
Sk
where R& is the pipe radius and Uo is the velocity at the centerline of the pipe. It is assumed
that the gas velocity is a function of the radial coordinate only.
The momentum equation for the gas in the axial direction is thus expressed by
{
-^*(U-Up) = 0 (11.70)
where /xe is the effective viscosity of turbulence.

The momentum equation for the particle phase in the axial direction is given by
(11.71)
where TP is the shear stress of particles. Since the transport of momentum of the particle
phase in the mixture for a dilute suspension is mainly caused by the diffusion of particles
through the gas phase, the viscosity of the particle phase is given by
p p ) p (11.72)
where A is a constant with the value of about 1. Therefore, we have
rpzr = (1 - a ) p P > p ^ ; r pz , = (1 - oc)PpDp^ (11.73)

or rou
so that Eq. (11.71) gives
^)^-^
Now we have four independent equations (Eqs. (11.67), (11.59), (11.70), and (11.74)) for
four independent variables: volume fraction a, electric potential VE gas velocity U, and
particle velocity Uv. We may further assume that U and r of the gas are not influenced by
the presence of the particles so that we may directly use the velocity profile for a single-
phase pipe flow. For a fully developed turbulent pipe flow for a single phase, we have, for
r < Rd-8,
H-i)"'
where the thickness of the laminar sublayer S is given by
S 60
(11.76)
with Re defined by
R e = ^ ^ (11.77)
Within the laminar sublayer, i.e., Rd 8 < r < Rd, we have
= 0.0225 ( 1 - ] R e 3 / 4 (11.78)
UQ \ RdJ
The boundary conditions are given by
(1) Electric potential at the wall
where VEW(#) becomes constant for a conducting pipe.

(2) Slip velocity of particles along the wall
dUx
(11.80)
where L p is the length scale of the slip velocity at the wall.

(3) Total flow rates of gas and particles
apUrdrdO = J (11.81)
Jo Jo
Jo
and
/ / (l-a)pvUprdrd0 = Jv (11.82)
Jo Jo
where / and Jp are the given total massflowrates for the gas phase and the particle
phase, respectively.
Example 11.4 Consider a dilute gas-solid flow in a horizontal pipe made of electrically
conducting materials. The pipe is well grounded. Theflowis fully developed. Show that
(1) the particle concentration is exponentially distributed and is a function of the vertical
distance only and (2) the ratio of particle volume fraction at the top to that at the bottom
depends on the density ratio of gas to particle, particle diffusivity, and pipe diameter.
Solution For a horizontal pipe, p = 90. Since the pipe is well grounded, the
electric potential VE is zero. Therefore, in terms of Eq. (11.67), the distribution of the
particle volume fraction can be expressed by
(l (E1L12)
where y(= r cos 6) is the vertical coordinate and apo represents the particle volume fraction
at the centerline of the pipe. Note that the integral constant C\ in Eq. (11.67) in this case
can be related to a^ by
^ (Ell.13)
Equation (El 1.12) shows that the particle concentration is exponentially distributed and
varies only with respect to the vertical distance. From Eq. (11.67), we also have
^ = e x p trp 1 (Ell.14)
which depends on the density ratio of gas to particle, particle diffusivity, and pipe diameter.
11.6.2 Characteristic Relations

To generalize the analysis, various equations and boundary conditions are ex-
pressed in dimensionless forms. Consider the following dimensionless quantities
r* r
ap
R& 1-,
(11.83)
U
u* =
where subscript "0" represents r = 0. In addition to the Reynolds number and Froude
number, defined, respectively, by Eq. (11.77) and Eq. (11.69), three dimensionless numbers
of relevance are given in the following:
(1) Electrodiffusion number NED
m
mDDv]j e0
which is the ratio of the displacement by electrostatic repulsion to that by diffusion.
(2) Diffusion response number N D F
n T
(11.85)
which represents the ratio of the relaxation time to the diffusion time (Rl/D v).
(3) Momentum transfer number N m
(11.86)
^d
which gives the ratio of the relaxation time to the transport time (R&/ Uo).
Using the preceding definitions, the dimensionless forms of Eqs. (11.59), (11.67), and
(11.74) can be given by
1 s /.aw\. i8>w .... . (1187)
r* 9r* V dr* J r*2 dO1
1
(11.88)
^~2W
r-U*)=0 (11.89)
where rj and y are parameters defined by
and
The boundary conditions now become
(1) Electric potential at the wall
(2) Slip velocity of particles along the wall
(11.93)
where Knp is the Knudsen number for particle-fluid interaction, defined by
Knp = ^ (11.94)
(3) Total flow rate of particles
For vertical pipe flows (77 = 0) with negligible electric charge effect (N ED = 0), the
particle velocity distributions for various values of N D F, Knp, and y are shown in Fig. 11.12.
Figure 11.12(a) illustrates that a small N D F leads to a particle velocity profile close to the
gas velocity profile. As expected, a larger Knp gives rise to a larger slip velocity of particles
at the wall and results in a flatter velocity profile as shown in Figs. 11.12(a) and (b). In
Fig. 11.12(c), it is interesting to note that as y increases, the direction of the particleflowmay
change from concurrent to countercurrent to the gas flow, and, thus, the net characteristic
massfluxratio of particle to gas may become zero (fluidized bed) or negative (countercurrent
flow). For a large, negative y9 the velocity for the particle phase is higher than that for the
gas phase as a result of the gravity effect (downward flow).
The effects of electrostatic charges of particles can be shown in Fig. 11.13, which illus-
trates the velocity and concentration profiles in vertical pipe flows (rj = 0) with negligible
gravity effect for various values of NED, NDF, and Knp. The influence of the electrostatic
charge on the magnitude and the profile of the particle velocity is far less significant than that
of the particle concentration. The smaller the diffusion response number N DF , the smaller
is the velocity lag at the axis. The Knudsen number is seen again as a key parameter
accounting for the velocity slip at the boundary.
The mass flow distribution for various values of rj in horizontal pipe flows (y = 0) with
N D F of 0.01 and Knp of 0.1 is illustrated in Fig. 11.14, which is compared to the experimental
data of Wen (1966). The figure shows that a higher rj gives rise to a sharper change in the
mass flux profile, namely, a much higher mass flux at the bottom of the horizontal pipe
N D F = 0.1/ /
0 01
0.5 - / - //
/
.//0.001
1.0
0.0 0.2 0.4 ,0.6 0.8 1.0
(a)
0.0 0.2 0.4 .0.6 0.8 1.0
(b)
0.0 0.2 0.4 0.6 0.8 1.0
(c)
Figure 11.12. Particle velocity distribution in vertical pipe flows with negligible electric
charge effect (from Soo, 1990): (a) Effect of N D F at Knp = 0.1 and y = 1.0; (b) Effect of N D F
at Kn p = 1.0 and y = 1.0; (c) Effect of y at Kn p = 0.1 and N D F = 0.01.
than that at the top. This trend can be understood from Eq. (11.88), which indicates that
the particle concentration profile is determined by rj in the absence of electrostatic effect.
Moreover, in this case, a higher particle mass loading always leads to more nonuniformity
in concentration distribution, resulting in a higher value of rj.
11.6.3 Temperature Distributions of Phases

For a fully developed gas-solid suspension pipe flow where the overall heat transfer
rate between the suspension and the pipe wall is dominated by both thermal convection and
1.0
=1.0
N R n=1.26
0.5
i \ i
1.2
0.0 0.5 1.0
(a)
NED = 0 Nm=1.26
1.0
0.5
1.2
0.0 0.5 1.0
(b)
Figure 11.13. Velocity and density distributions in vertical pipe flows at negligible gravity
effect (from Soo, 1990): (a) N D F = 0.25; (b) N D F = 0.025.
o Silica sand, 20 m/s, m = 1.5

(Wen, 1966)
n Silica sand, 10 m/s, m = 0.5
(Wen, 1966)
Figure 11.14. Mass flow distributions in horizontal pipe flows at NED = 0, NDF = 0.01, and
Knp = 0.1 (from Soo, 1990).
radiation, a thermal analysis of the flow can be made with the following assumptions:
(1) Velocity and volume fraction distributions of phases are independent of the tem-
perature distributions of phases.
(2) Transport properties are constant.
(3) Thermal conductivity of solid particles is very high (Bi <^ 1; see Eq. (4.5) for Bi),
so that a uniform temperature is maintained in the particle.
(4) Effect of collisions of particles with the pipe wall is negligible.
(5) Particles and wall are perfect gray and gas is transparent for thermal radiation.
(6) Gravitational force is neglected so that the flow is axisymmetric.
The energy equation for a particle can be expressed by
dTp
mc^- = hpSp(T - 7p) + ahR (T* - 714) (11.96)
dr
where c is the specific heat of particles; hp is the convective heat transfer coefficient of the
particle; Sp is the particle surface area; a b is the Stefan-Boltzmann constant; and 6R is the
emissivity of the particle. For a steady dilute suspension flow, Eq. (11.96) leads to
dT
L
mcup + npbp(Tp T)+ GbeR{Tv T w)=0 (11.97)
which describes the energy equation for the particle phase.

The energy equation for the gas phase is expressed by
apcpU = - (rK e) + nhpSJTp - T) (11.98)

dz r dr \ dr J
where Ke is the effective thermal conductivity of gas, which includes the effect of eddy
diffusion of heat, and n is the particle number density. For given gas and solids velocities
and volume fraction distributions, Eqs. (11.97) and (11.98) can be solved numerically with
the following boundary conditions:
dT\ (dTA _ _ _
= 0
~dr~)r=o ' \~dr~)r=0~0' Tr R
\= = T
^ T
p\r=R-Tw ( n 9 9 )
T\z=zo = To, Tp\z=o = To; Îzôo = ?w> ^plzôo = T^

The analytical solution of this problem is difficult because of the highly nonlinear ra-
diation term in Eq. (11.97). However, under some conditions, the term can be simplified.
For instance, when the temperature difference between particles and pipe wall is small (i.e.,
\Tp Tw\ ^C 7"w), the radiation term in Eq. (11.97) can be linearized to
w ^ p = Uw ~ ^p)(^w + ^p)(/ w ' P / ~ ^ w ^ w ^p) (ll.lUUj
Defining nondimensional temperatures
= ; ip = (11.101)
Eqs. (11.97) and (11.98) can be expressed in dimensionless form as

dT;
where
Z* = j , A=
and
(
U* = (r*K* J + ~a)fe(C-^*) (11.104)
y
3z* r* dr* \ 3r* / a
where
Hi
RePr K pc/
and K* is a function of r* whereas Pr is the Prandtl number of the gas and Re is defined by
Eq. (11.77). The boundary conditions can be expressed in dimensionless form as
*11\ =o, (d-^-) =0; n r . = i=o, r p v = i=o

3r
/r*=0 V^*/r*=0 (11.106)
It is noted that now the heat transfer of the system is governed by two coupled linear
equations, Eqs. (11.102) and (11.104). The last boundary condition in Eq. (11.106) suggests
the solution technique for Eqs. (11.102) and (11.104), i.e., separation of variables. Thus,
we have
oo
r = 5^C n * n exp(-^*) (11.107)
and
00
T; = ]Tc n /? p n exp(-^z*) (11.108)

n=0
where tfn(r*) and Rpn(r*) satisfy
(11.109)
P\ PI
and
n= 0 (11.110)
r* dr* \ dr* ) a
with the boundary conditions
= 0; *n|r*=i=0, Rpn\r*=l=0 (11.111)

r*=0
Substitution of Eq. (11.109) into Eq. (11.110) yields
(11.112)
with
= 0, *nlr'=l=0 (11.113)
' r*=0
where all /Ts are constants and a, K*, U*, and U* are given functions of r* (or relations
independently solved from the continuity and momentum equations).
The problem imposed by Eq. (11.112) with boundary conditions of Eq. (11.113) is noted
as the Sturm-Liouville boundary-value problem [e.g., Derrick and Grossman, 1987], if
<l (1L114)
TT^
Since U*, /3i, and ^2 ar e uniquely characteristic for the operating system, the validity of the
preceding inequality is largely dependent on the values of A2. As indicated in Eq. (11.107),
only the first few successive eigenvalues starting from the smallest one are needed. There-
fore, Eq. (11.114) would hold true for the first few eigenvalues. Thus, we have
1
^ ft L | *fo * \
P
(11.115)
which leads to Eq. (11.110) in the form of the Sturm-Liouville problem
where
A + f t
(11.117)
with the boundary conditions given in Eq. (11.113).

In principle, eigenfunction Rn and its eigenvalue An can be obtained numerically by
solving Eqs .(11.113) and (11.116). The coefficients are determined in terms of the boundary
conditions in Eq. (11.106) and the orthogonality property in the Sturm-Liouville theorems
as
r -h (11.118)
/o
In addition, the heat flux to the gas phase at the wall is given by
The preceding analysis is based on Tien's model of heat transfer by a gas-solid sus-
pension in turbulent pipe flow [Tien, 1961]. However, nonuniform distribution of solids,
slip between solids and gas, and effect of thermal radiation were excluded in Tien's
work.
Nomenclature
A Cross-sectional area /mp Momentum flux tensor of
A Constant, defined by Eq. particles
(11.22) K Thermal conductivity of gas
A Constant, defined by Eq. Effective thermal conductivity
(11.48) of gas
A Constant, defined by Eq. Knp Knudsen number for
(11.72) particle-fluid interaction,
B Factor, defined by Eq. (11.46) defined by Eq. (11.94)
Bi Biot number L Pipe length
CD Drag coefficient La Acceleration length
Cw Perimeter LB Equivalent length of straight
c Specific heat of particles pipe
Cp Specific heat of gas at constant Lp Length scale of slip velocity at
pressure wall
D Diffusivity of gas /m Mixing length of turbulence
>d Pipe diameter m Mass of a particle
Dp Diffusivity of particles m Mass flux ratio of particle to
dp Particle diameter gas
E Electric field strength vector N Normal impact force
Fr Froude number, defined by Eq. N DF Diffusion response number,
fll 69) defined by Eq. (11.85)
NED Electrodiffusion number,
FD Drag force
defined by Eq. (11.84)
Ff Friction force Nm Momentum transfer number,
F\ Lift force defined by Eq. (11.86)
FA Interfacial force vector n Particle number density
between gas and particles Pr Prandtl number
FE Electrostatic force vector
FG Gravitational force vector P
q Electrostatic charges carried
including buoyancy effect
by a particle
f Friction factor Bend radius
RB
/B Friction factor of a bend Pipe radius
/P Friction coefficient of particles
Re Reynolds number
at wall
Gravitational acceleration r Cylindrical coordinate
g
hp Convective heat transfer sv Particle surface area
coefficient T Gas temperature
J Mass flux of gas To Temperature at the inlet
Jv Mass flux of particles T
P Particle temperature
J
v Total flow rate of particles Tw Wall temperature
/m Momentum flux tensor of gas U Gas velocity component
Nomenclature 495
U Mean stream velocity tfpw Volume fraction of particles at

Uo Gas velocity at pipe centerline wall
U\ Gas velocity at the edge of P Inclined angle of pipe
shedding layer y Parameter, defined by Eq.
Uf Friction velocity based on the (11.90)
mixture density Ap Pressure drop
Ufo Friction velocity at minimum Ap0 Pressure drop in a single gas
transport and zero solids flow
concentration A/?a Pressure drop over the
Um Averaged axial velocity of gas acceleration region
Up Particle velocity component A/?Bg Pressure drop of gas over a
(/pi Particle velocity at inlet bend
Upt Particle terminal velocity Additional pressure drop due
to particles over a bend
/pw Particle velocity at wall
Pressure drop over an
UPoo Particle velocity in the fully equivalent length of a straight
developed flow pipe
/* Dimensionless gas velocity 8 Shedding layer thickness
component o Permittivity in a vacuum
/p Dimensionless particle 6R Particle emissivity
velocity component Y] Parameter, defined by Eq.
U Gas velocity vector (11.91)
Up Particle velocity vector 0 Cylindrical coordinate
u' Gas velocity fluctuation 6 Inclined angle, defined by
Wp Particle velocity fluctuation Fig. 11.8
w pw Particle velocity fluctuation at 0 Angle in the transverse plane
wall with respect to the outward
V Gas velocity component direction of the symmetry line,
VE Electric potential defined by Fig. 11.9
VEW Electric potential at wall /x Dynamic viscosity of gas
Vp Particle velocity component /xe Effective turbulent viscosity in
Vt Tangential velocity near the a single phase flow
wall /xgp Turbulent viscosity of the gas
phase in the presence of
W Gas velocity component
particles
Wp Particle velocity component
/xp Viscosity of particles
x Axial coordinate
v Kinematic viscosity of gas
x Cartesian coordinate
Distance normal to wall
y Cartesian coordinate
p Density of gas
Greek Symbols pp Density of particles
ah Stefan-Boltzmann's constant
a Volume fraction of gas
re Effective stress tensor per unit
ap Volume fraction of particles
volume
Characteristic time of Wall shear stress

turbulent fluctuations of Shear stress due to collisions
gas of particles with the wall
Shear stress of particles Shear stress along #-axis
rgp Wall shear stress of gas in the Shear stress along 0-axis
presence of particles Angle measured in the plane of
Particle relaxation time the curved pipe axis, defined
Stokes relaxation time by Fig. 11.9
References
Blatch, N. S. (1906). Discussion: Water Filtration at Washington, D.C. Trans. Am. Soc. Civil Eng.,
57, 400.
Boyce, M. P. and Blick, E. F. (1969). Fluid Flow Phenomena in Dusty Air. ASME Paper No.
69-WA/FE-24.
Chen, T. Y, Walawender, W P. and Fan, L. T. (1979). Solids Transfer Between Fluidized Beds: A
Review. J. Powder Bulk Solids Tech., 3, 3.
Chien, N. and Asce, A. M. (1956). The Present Status of Research on Sediment Transport. Trans.
Am. Soc. Civil Eng., 121, 833.
Crowe, C. T. (1982). Review: Numerical Models for Dilute Gas-Particle Flows. Trans. ASME, J.
Fluids Eng., 104, 297.
Derrick, W. R. and Grossman, S. I. (1987). Introduction to Differential Equations with
Boundary-Value Problems, 3rd ed. St. Paul, Minn.: West Publishing.
Haag, A. (1967). Velocity Losses in Pneumatic Conveyer Pipe Bends. British Chemical
Engineering, 12, 65.
Halow, J. S. (1973). Incipient Rolling, Sliding, and Suspension of Particles in Horizontal and
Inclined Turbulent Flow. Chem. Eng. ScL, 28, 1.
Haltsrom, E. A. N. (1953). Design of Experimental Apparatus for the Study of Two-Phase Flow in
Circular Straight Pipe. M.S. Thesis. Princeton University.
Ito, H. (1959). Friction Factors for Turbulent Flow in Curved Pipes. Trans. ASME, J. Basic Eng.,
81D, 123.
Ito, H. (1960). Pressure Loses in Smooth Pipe Bends. Trans. ASME, J. Basic Eng., 82D, 131.
Kane, R. S. (1989). Drag Reduction by Particle Addition. Viscous Drag Reduction in Boundary
Layers. Vol. 123. Progress in Astronautics and Aeronautics. Ed. Bushnell and Hefner.
Washington, D.C: AIAA.
Kane, R. S. and Pfeffer, R. (1973). Characteristics of Dilute Gas-Solids Suspensions in Drag
Reduction Flow. NASA CR-2267, 1973.
Klinzing, G. E. (1981). Gas-Solid Transport. New York: McGraw-Hill.
Konno, H. and Saito, S. (1969). Pneumatic Conveying of Solids Through Straight Pipes. J. Chem.
Eng. Japan, 2, 211.
Konrad, K. (1986). Dense-Phase Pneumatic Conveying: A Review. Powder Tech., 49, 1.
Konrad, K., Harrison, D., Nedderman, R. M. and Davidson, J. F. (1980). Prediction of the Pressure
Drop for Horizontal Dense-Phase Pneumatic Conveying of Particles. Pneumotransport,
5, 225.
Kunii, D. and Levenspiel, O. (1969). Fluidization Engineering. New York: Wiley.
References 497
Leung, L. S. (1980). Vertical Pneumatic Conveying: A Flow Regime Diagram and a Review of
Choking Versus Nonchoking Systems. Powder Tech., 25, 185.
Leva, M. (1959). Fluidization. New York: McGraw-Hill.
Marcus, R. D., Leung, L. S., Klinzing, G. E. and Rizk, F. (1990). Pneumatic Conveying of Solids.
New York: Chapman & Hall.
McCarthy, H. E. and Olson, J. H. (1968). Turbulent Flow of Gas-Solids Suspensions. I& EC Fund.,
7,471.
Moody, L. F. (1944). Friction Factors for Pipe Flow. ASME Trans., 66, 671.
Peters, L. K. and Klinzing, G. E. (1972). Friction in Turbulent Flow of Gas-Solid Systems. Can. J.
Chem.Eng., 50,441.
Rose, H. E. and Duckworth, R. A. (1969). Transport of Solid Particles in Liquids and Gases. The
Engineer, 227, 392.
Roco, M. and Shook, C. (1984). Computational Method for Coal Slurry Pipelines with
Heterogeneous Size Distribution. Powder Tech., 39, 159.
Schuchart, P. (1968). Widerstandsgesetze Beim Pneumatischen Transport in Rohrkrummern.
Chem.-Ing.-Tech., 40, 1060.
Shimizu, A., Echigo, R., Hasegawa, S. and Hishida, M. (1978). Experimental Study on the Pressure
Drop and the Entry Length of the Gas-Solid Suspension Flow in a Circular Tube. Int. J.
Soo, S. L. (1962). Boundary Layer Motion of a Gas-Solid Suspension. Proceedings of Symposium
on Interaction Between Fluids and Particles. Institute of Chemical Engineers,
London.
Hemisphere.
Soo, S. L. (1990). Multiphase Fluid Dynamics. Beijing: Science Press; Brookfield, Vt: Gower
Technical.
Soo, S. L. and Mei, R. (1987). Dense Suspension and Fluidization. Proceedings of Powder and Bulk
Solids, 1987 Conference, Rosemont, 111.
Soo, S. L., Trezek, G. L., Dimick, R. C. and Hohnstreiter, G. F. (1964). Concentration and Mass
Flow Distributions in a Gas-Solid Suspension. I&EC Fund., 3, 98.
Soo, S. L. and Tung, S. K. (1971). Pipe Flow of Suspensions in Turbulent Fluid: Electrostatic and
Gravity Effects. Appl Sci. Res., 24, 83.
Stoess, H. A. (1983). Pneumatic Conveying, 2nd ed. New York: John Wiley & Sons.
Taylor, G. I. (1954). The Dispersion of Matter in Turbulent Flow Through a Pipe. Proc. R. Soc.
London, A223, 446.
Thomas, D. G. (1962). Transport Characteristics of Suspensions. Part IV. Minimum Transport
Velocity for Large Particle Size Suspension in Round Horizontal Pipes. AIChE J., 8, 373.
Tien, C. L. (1961). Heat Transfer by a Turbulently Flowing Fluid-Solids Mixture in a Pipe. Trans.
ASME, J. Heat Transfer, 83C, 183.
Tien, C. L. and Quan, V. (1962). Local Heat Transfer Characteristics of Air-Glass and Air-Lead
Mixtures in Turbulent Pipe Flow. ASME Paper No. 62-HT-15.
Wen, C. Y. (1966). In Pneumatic Transportation of Solids. Ed. Spencer, Joyce, and Farber.
Washington, D.C.: Bureau of Mines Information Circular, U.S. Deptartment of Interior.
Williams, O. A. (1983). Pneumatic and Hydraulic Conveying of Solids. New York: Marcel
Dekker.
Yang, W. C. (1987). Pneumatic Transport in a 10cm Pipe Horizontal Loop. Powder Tech., 49, 207.
Zenz, R A. (1964). Conveyability of Materials of Mixed Particle Size. 1 & EC Fund., 3, 65.
Problems
11.1 In a gas-solid horizontal pipe flow, the particle used is 100 /xm glass bead with a density of
2,500 kg/m3. The pipe diameter is 50 mm. The average particle volume fraction in the pipe
is 1 percent. The gas density and kinematic viscosity are 1.2 kg/m3 and 1.5 x 10~5 m2/s,
respectively. Estimate the minimum transport velocity and power consumption per unit
length. Repeat this problem for a 10 /zm glass bead size.
11.2 For a single-phase flow in a bend, the momentum integral equations for the shedding layer
of the secondary flow can be expressed by Eqs. (11.36), (11.37), and (11.38) [Ito, 1959].
On the basis of these equations and the approximations in Eq. (11.45), (a) prove Eq. (11.46)
and (b) show that the thickness of the shedding layer can be estimated by Eq. (11.47).
11.3 The pressure drop of a gas flow over a 90 bend may be estimated by either Eq. (11.23) or
Eq. (11.50). For the ratios of the pipe radius to the bending radius, /?d/^B, of 0.1, 0.05,
and 0.01, compare the pressure drop predictions using these two equations for the Reynolds
numbers ranging from 4,000 to 105.
11.4 Consider a pneumatic transport of grains in a sheet metal duct. The sliding frictional coeffi-
cient / p is about 0.36 [Haag, 1967]. To ensure a smooth operation over a 90bend, the grain
particles should always be sliding against the wall during the turn. Estimate the minimum
particle velocity required for such a smooth operation. The pipe diameter is 0.1 m. The
bend radius is 0.5 m.
11.5 Consider a dilute gas-solid flow in a horizontal rectangular pipe made of electrically con-
ducting materials. The pipe is well grounded. The flow is fully developed. It is assumed that
the particle volume fraction distribution in the vertical direction is the same as that in a cir-
cular pipeflow.Find out (a) the cross-sectionally averaged particle volume fraction in terms
of the particle volume fraction at the centerline and (b) the vertical location at which the
particle volume fraction represents the cross-sectionally averaged particle volume fraction.
11.6 Consider a dilute gas-solid flow in a pipe in which the solid particles carry significant
electrostatic charges. It is assumed that (a) the flow is fully developed; (b) the gravitational
effect is negligible; and (c) the flow and the electrostatic field are axisymmetric. Derive an
expression to describe the radial volume fraction distribution of the particles and identify
the radial locations where the particle volume fractions are maximum and minimum in the
distribution. Also, if the electrostatic charge effects are negligible, derive an expression to
describe the radial volume fraction distribution of the particles.
11.7 Consider a fully developed dilute gas-solid flow in a vertical pipe in which solid particles
carry significant electrostatic charges. The particle charges vary radially. It is assumed that
(a) theflowand the electrostatic field are axisymmetric and (b) the radial charge distribution
C(r)y defined as q(r)ap(r), is known. Derive an expression for the radial volume fraction
distribution of the particles.
CHAPTER 12
Heat and Mass Transfer Phenomena in

Fluidization Systems
12.1 Introduction
Among many gas-solid flow systems involving heat and mass transfer operations,
thefluidizationsystem is one of the most frequently encountered. The heat and mass transfer
behavior in a gas-solid fluidized bed is important when physical or chemical operations
are conducted in the bed, such as drying, coal combustion, polymerization reaction, and
chemical synthesis. The fluidized bed is characterized by a high efficiency of heat exchange
between particles and the fluidizing gas. The inherent intensive gas-solid contact or rapid
mixing between gas and solid phases contributes to this efficient heat transfer. The fluidized
bed possesses high heat capacity, and uniform temperature in the bed can generally be
maintained. Thus, temperature control in the bed can be effectively carried out. Comparable
characteristics are also exhibited for thefluidizedbed mass transfer between gas and particles
and between the bed and gas or particles.
The heat and mass transfer properties can be represented by heat and mass transfer
coefficients, which are commonly given in empirical or semiempirical correlation form.
The transfer coefficient is defined in terms of flow models under specific flow conditions
and geometric arrangements of the flow system. Thus, when applying the correlations, it is
necessary to employ the sameflowmodel to describe the heat and mass transfer coefficients
for conditions comparable to those where the correlations were obtained. An accurate
characterization of the heat and mass transfer can be made only when the hydrodynamics
and underlying mechanism of the transport processes are well understood.
The governing heat transfer modes in gas-solid flow systems include gas-particle heat
transfer, particle-particle heat transfer, and suspension-surface heat transfer by conduction,
convection, and/or radiation. The basic heat and mass transfer modes of a single particle in a
gas medium are introduced in Chapter 4. This chapter deals with the modeling approaches in
describing the heat and mass transfer processes in gas-solid flows. In multiparticle systems,
as in the fluidization systems with spherical or nearly spherical particles, the conductive
heat transfer due to particle collisions is usually negligible. Hence, this chapter is mainly
concerned with the heat and mass transfer from suspension to the wall, from suspension
to an immersed surface, and from gas to solids for multiparticle systems. The heat and
mass transfer mechanisms due to particle convection and gas convection are illustrated. In
addition, heat transfer due to radiation is discussed.
12.2 Suspension-to-Surface Heat Transfer

A mechanistic account of suspension-to-surface heat transfer is necessary to quan-
tify the heat transfer behavior accurately and to assess the form of dependency of dimen-
sionless groups in the correlations. In the following, the modes and regimes of suspension-
to-surface heat transfer along with the three mechanistic models accounting for this heat
transfer behavior are described.
499
500 12 I Heat and Mass Transfer Phenomena in Fluidization Systems
12.2.1 Heat Transfer Modes and Regimes

Heat transfer between a fluidized bed and an immersed surface can occur by three
modes, namely, particle convection, gas convection, and radiation.
Particle convective heat transfer is due to the convective flow of solid particles from the
bulk of the bed to the region adjacent to the heat transfer surface. Solid particles gain heat by
thermal conduction from the heated surface (assuming the surface has a higher temperature
than the solids). As the particles return to the bulk of the bed, the heat is dissipated. The
capacity of moving particles to transfer heat is reflected by the extent of heat conduction by
the suspension. Gas convective heat transfer is caused by gas percolating through the bed
and also by gas voids in contact with the surface. Radiative heat transfer is due to radiant
heat transmitted to fluidized particles or solid surfaces from a heat transfer surface at high
temperature. The total heat transfer coefficient, h, can be estimated from the summation
of the individual heat transfer coefficients for particle convection, h pc; gas convection, hgc;
and radiation, hr, although their precise relationship may not be additive, i.e.,
*pc *gc (12.1)

The heat transfer coefficients are strongly influenced by the operating conditions of
the fluidized bed. Variations in the operating conditions such as bubbling and spouting
yield a varied bed structure and hence varied heat transfer coefficients. Understanding the
governing mechanisms in heat transfer is important to the development of simplified heat
transfer models or equations. The relative importance of the heat transfer modes for the
suspension-to-surface heat transfer in gas-solid fluidized beds is illustrated in Fig. 12.1
[Flamant et aL, 1992]. The figure indicates that the governing modes of heat transfer
mechanism in the fluidized bed depend on both particle size and bed/surface temperature.
particle convection /
2,000 radiation / particle convection
gas convection
/
radiation
1,000
convection -
n
<S0^ | 1 1 1 1
2 3
dn, nun
Figure 12.1. Heat transfer diagram for various governing modes (from Flamant et ah, 1992).
72.2 / Suspension-to-Surface Heat Transfer 501
Fixed bed
uoptr u
Figure 12.2. Typical dependence of the heat transfer coefficient on gas velocity in dense-phase
fluidization systems (from Gel'Perin and Einstein, 1971).
It is seen that particle convection is important for almost all the conditions, except at low
bed temperatures in a bed of large particles, where the gas convection becomes important.
In most dense gas-solid fluidization systems, particle circulation {e.g., that induced by
the bubbles) is the primary cause of particle convective heat transfer. When particles initially
reach the heat transfer surface, the temperature gradient is high and yields high heat transfer
rates. As time progresses, the temperature gradient reduces, yielding low heat transfer rates.
Thus, the heat transfer rate is high when there is an extensive solids exchange between the
bulk and the surface. Increasing particle circulation enhances the heat transfer rate.
Development of a mechanistic model is essential to quantification of the heat transfer
phenomena in afluidizedsystem. Most models that are originally developed for dense-phase
fluidized systems are also applicable to otherfluidizationsystems. Figure 12.2 provides basic
heat transfer characteristics in dense-phase fluidization systems that must be taken into
account by a mechanistic model. The figure shows the variation of heat transfer coefficient
with the gas velocity. It is seen that at a low gas velocity where the bed is in a fixed bed
state, the heat transfer coefficient is low; with increasing gas velocity, it increases sharply
to a maximum value and then decreases. This increasing and decreasing behavior is a result
of interplay between the particle convective and gas convective heat transfer which can be
explained by mechanistic models given in 12.2.2, 12.2.3, and 12.2.4.
Numerous models have been developed to describe heat transfer behavior in fluidiza-
tion systems. They can be broadly categorized into the following three models: the film
model, the single-particle model, and the emulsion phase/packet model. Each model has
its limitations in application. Accordingly, a given model may be more suitable to some
fluidization conditions than the others. For example, the film model and the single-particle
model are more suitable for a particulate fluidized bed [Gel'Perin and Einstein, 1971]
than for a system that contains gas bubbles. The components contributing to heat transfer
given by Eq. (12.1) can be evaluated separately by different models. The most challeng-
ing aspect of applying these models, however, lies in the determination of the flow and
thermal properties of the region directly adjacent to the surface where heat transfer takes
place.
12.2.2 Film Model

In a well-fluidized gas-solid system, the bulk of the bed can be approximated to be
isothermal and hence to have negligible thermal resistance. This approximation indicates
that the thermal resistance limiting the rate of heat transfer between the bed and the heating
surface lies within a narrow gas layer at the heating surface. Thefilmmodel for the fluidized
bed heat transfer assumes that the heat is transferred only by conduction through the thin
gas film or gas boundary layer adjacent to the heating surface. The effect of particles is
to erode the film and reduce its resistive effect, as shown by Fig. 12.3. The heat transfer
coefficient in the film model can be expressed as
h=j (12.2)
where K is the gas thermal conductivity and 8 is the boundary layer thickness, which
depends on the velocity and physical properties of the fluid and also on the intensity of
motion of the solid particles that erode the boundary layer. With an increase in the gas
velocity, the particles near the surface move more vigorously but the local concentration of
the particles decreases. This interplay results in a maximum in the h-U curve, as shown in
Fig. 12.2.
The model based on the concept of pure limiting film resistance involves the steady-state
concept of the heat transfer process and omits the essential unsteady nature of the heat trans-
fer phenomena observed in many gas-solid suspension systems. To take into account the
unsteady heat transfer behavior and particle convection in fluidized beds, a surface renewal
model can be used. The model accounts for the film resistance adjacent to the heat transfer
Growing gas film
Heat transfer by conduction

through the gas film
Surface
Descending particles
scour away the film
Figure 12.3. Conceptual representation of the thin-film model (from Levenspiel and Walton,
1954).
122 I Suspension-to-Surface Heat Transfer 503
surface, induced by the particle-gas emulsion phase instead of the pure gas phase. The model
is capable of explaining the initial increase in the observed heat transfer with the gas velocity
beyond the point of onset of fluidization. Thus, the models suggest different mechanisms
useful to describe high heat transfer coefficients in fluidized beds [Botterill, 1975].
The film model discounts the effects of thermophysical properties such as the specific heat
of solids and, thus, is not able to predict the particle convective component of heat transfer.
Moreover, the measured temperature gradient extends deeper into the bed from the heating
surface than that which can be predicted by the film model [Baeyens and Goossens, 1973].
Thus, models based on heat conduction through the gas boundary layer cannot be useful for
estimating the contribution of the particle convective component of heat transfer, and the
emulsion phase/packet model given in 12.2.4 should be used to describe the temperature
gradient from the heating surface to the bed.
12.2.3 Single-Particle Model

The single-particle model [Zabrodsky, 1963] postulates that the moving solid par-
ticles play an important role in heat transfer by thermal conduction. The model takes into
account the thermal conduction through the layer of gas at the heating surface. Thus, al-
lowance is made for the influence of the thermophysical properties of the solid material.
In this model, the contributions from gas convection and bed-to-surface thermal radiation
are excluded. The high heat transfer coefficient shown in Fig. 12.2 is due to the high tem-
perature gradient during heating of moving solids. The presence of the maximum on the
heat transfer curve with gas velocity is a result of the simultaneous effect of the rise in
temperature gradient and the fall in concentration of solid particles.
The simplest model of this kind can be represented by one in which an isolated particle
surrounded by gas is in contact with or in the vicinity of the heating surface for a certain
time, during which the heat transfer between the particle and the heating surface takes place
by transient conduction, as shown in Fig. 12.4. In terms of the model, the Fourier equation
of thermal conduction can be expressed as
p p c ^ = KVV% (12.3)
for the particle phase and
p c p ^ = KV2T (12.4)
ot
for the gas phase.
The boundary and initial conditions for Eqs. (12.3) and (12.4) are given by the following:
(1) At the symmetric plane
T = 0;
7T= (12 5)
-
dn dn
where dTv/dn and dT/dn are the normal derivatives with respect to the temperature
symmetric plane.
(2) At the heating surface
r p = Ts; T = Ts (12.6)
For the case without particle-surface contact, only T = TS should be imposed.
symmetric plane 1
heating
surface
Particle
symmetric plane 2
(a)
symmetric plane 1
heating
surface
symmetric plane 2
(b)
Figure 12.4. Transient heat conduction in the single-particle model (from Botterill, 1975):
(a) Heat conduction with particle-surface contact; (b) Heat conduction without particle-surface
contact.
(3) At the gas-particle interface

dTp _ dT
p (12.7)
~dn7 ~ ~dnf
where dTv/dnf and 3T/dnf are the normal derivatives with respect to the gas-
particle interface.
(4) Far inside the bed
T = Th (12.8)
(5) At t = 0 (initial condition)
(12.9)
Sample isotherm simulation results for a glass sphere of 200 /xm in contact with a heat
transfer surface surrounded by static air are shown in Fig. 12.5 for two contact times, i.e., 1.2
ms and 52.4 ms [Botterill and Williams, 1963]. The initial temperature difference between
the sphere and the surface is 10C. It is seen that at the instant of contact heat begins to
flow around the upper surface of the sphere and significant heat transfer takes place at
12.2 I Suspension-to-Surface Heat Transfer 505
40 80 120 160 200 240

Distance from the heat transfer surface, urn
(a)
0 40 80 120 160 200 240

Distance from the heat transfer surface, urn
(b)
Figure 12.5. Isotherms for a glass sphere of 200 ^m in contact with a heat transfer surface
surrounded by air at an initial temperature difference of 10C at contacttime: (a) 1.2 ms;
(b) 52.4 ms (from Botterill and Williams, 1963).
the point of contact between the heat transfer surface and the sphere. Compared to that in
air, the temperature rise in the sphere is relatively slow as a result of larger heat capacity
but increases significantly as the contact time increases. The model can be extended from
a single particle to a single layer of particles at the surface. The model requires precise
information on the position and the residence time of the particle near the surface, and this
requirement may limit its usage. It is applicable only when the heat from the heat transfer
surface does not penetrate beyond the single particle layer. The depth of penetration into the
bed (<$em) can be estimated from the temperature gradient at the heating surface. In terms
of Eq. (4.26), we have
OC ( Z c)1/2 (12.10)
where DtQm is the thermal diffusivity of the emulsion phase and tc is the averaged contact
(or residence) time of the particle near the heating surface. Equation (12.10) yields
1/2
ocFo (12.11)
where Fo is the Fourier number (Dtemtc/d2). From this analysis, it is evident that the single-
particle models are suitable only for low Fourier numbers, i.e., large particles with a short
contact time. To expand the range of applicability, the heat diffusion equation for multiple
particle layers has been solved [Gabor, 1970].
12.2.4 Emulsion Phase/Packet Model

In the emulsion phase/packet model, it is perceived that the resistance to heat
transfer lies in a relatively thick emulsion layer adjacent to the heating surface. This approach
employs an analogy between a fluidized bed and a liquid medium, which considers the
emulsion phase/packets to be the continuous phase. Differences in the various emulsion
phase models primarily depend on the way the packet is defined. The presence of the
maxima in the h-U curve is attributed to the simultaneous effect of an increase in the
frequency of packet replacement and an increase in the fraction of time for which the heat
transfer surface is covered by bubbles/voids. This unsteady-state model reaches its limit
when the particle thermal time constant is smaller than the particle contact time determined
by the replacement rate for small particles. In this case, the heat transfer process can be
approximated by a steady-state process. Mickley and Fairbanks (1955) treated the packet
as a continuum phase and first recognized the significant role of particle heat transfer since
the volumetric heat capacity of the particle is 1,000-fold that of the gas at atmospheric
conditions. The transient heat conduction equations are solved for a packet of emulsion
swept up to the wall by bubble-induced circulation. The model of Mickley and Fairbanks
(1955) is introduced in the following discussion.
Consider a packet of emulsion phase being swept into contact with the heating surface
for a certain period. During the contact, the heat is transferred by unsteady-state conduction
at the surface until the packet is replaced by a fresh packet as a result of bed circulation,
as shown in Fig. 12.6. The heat transfer rate depends on the rate of heating of the packets
(or emulsion phase) and on the frequency of their replacement at the surface. To simplify
the model, the packet of particles and interstitial gas can be regarded as having the uniform
thermal properties of the quiescent bed. The simplest case is represented by the problem of
one-dimensional unsteady thermal conduction in a semiinfinite medium. Thus, the govern-
ing equation with the boundary conditions and initial condition can be imposed as
0 < X < O Q
dt cpdx2 ~ ~
T = TS at x = 0
T = Too atx = oo (12.13)

T = r^ at t = 0
where Ktm and p em are the conductivity and density of the emulsion phase, respectively.
WMHI:
iilil
\
Surface illlii
mmmm
HP
Figure 12.6. Conceptual representation of the emulsion-contact model of Mickley and
Fairbanks (1955).
The solution can be obtained as
T = T8 + (T00 (12.14)
where erf(x) is the error function of argument x. Therefore, the instantaneous heat flux is
expressed by
T T
(12.15)
Consequently, the instantaneous local heat transfer coefficient is obtained as
(12.16)
TCt
Furthermore, the area-averaged local heat transfer coefficient can be expressed by

1
A / (12.17)
m JO
where Am is the area of the packet in contact with the heating surface, ^r(r) represents
the frequency of occurrence in time of the packet of age r, and S is the area mean stirring
factor, defined as
The preceding model successfully explains the role played by the particles in the heat
transfer processes occurring in the dense-phase fluidized bed at voidage a < 0.7. But
it predicts very large values when the contact time of particles with the heating surface
decreases because the nonuniformity of the solids concentration near the wall is not taken
into account in this model.
As a result of the time-dependent voidage variations near the heating surface, the ther-
mophysical properties of the packet differ from those in the bed, and this difference has
not been included in the packet model. The limitation of this model lies in not taking into
account the nonuniformity of the solids concentration near the heating surface. Thus, the
packet model under this condition is accurate only for large values of Fourier number, in
general agreement with the discussion in 4.3.3.
An important variation of the model of Mickley and Fairbanks is the film-penetration
model developed for more accurate prediction. This model, originally proposed for the
case of a single-phase liquid flow [Toor and Marchello, 1958], was extended to gas-solid
suspensions by Yoshida et al. (1969) by treating packets as a continuum with finite thickness
(<$em)- The film-penetration mechanism, analogous to the single-phase flow, includes two
extremes of emulsion behavior. On one extreme, the packet contacts the heating surface for
a short time so that all the heat entering the packet is used to heat it (penetration theory)
while none passes through it. On the other extreme, the packet stays at the surface long
enough to achieve steady-state and simply provides resistance for heat conduction.
In the heat transfer process a thin layer of emulsion of thickness <5em is in contact with a
heat transfer surface and after a time tc, is replaced by a fresh element of emulsion from the
bulk of the suspension, as shown in Fig. 12.7. Mathematically, the governing equation can
be expressed by Eq. (12.12), with x, which is the distance from the heat transfer surface,
confined to
0<x<8em (12.19)
The governing equation must be solved with two boundary conditions and an initial condi-
tion which are identical to Eq. (12.13) except
(12.20)
For convenience, we define the thermal diffusivity of the emulsion phase as
D tem = ÊL - ^ (12.21)

PemC PPC(1 - Qf mf )
The solution for the instantaneous local heat transfer coefficient can be obtained as
[Yoshida et al, 1969]
_^_ / *2 M 2\l *2
XP ^ ^ ^ ; x < - < OO (12.22)
^/7tDtemt V AemJ_
-
7r 2 D tem m 2 \
OO /
h em
; 0 if
(12.23)
<5em n=\ ^
5 2
em / ^tem^
Next packet of emulsion

to contact the surface
Or
Heating
surface
Direct contact of particle

with the surface
Thickness of
gas film, 5
Equivalent thickness
of emulsion layer, 8 ^
Figure 12.7. Conceptual representation of thefilm-penetrationmodel for

suspension-to-surface heat transfer (from Yoshida et ai, 1969).
where the dimensionless group Dtemt/8lm characterizes the intensity of renewal or the
magnitude of contact time. Generally, when tc ^> 8lm/Dtem, the film theory holds, and when
(
c ^ <$em/Aem> the penetration theory is valid. The time-averaged heat transfer coefficient
h is obtained as
(12.24)
-
where 1(0 is Jo
the age distribution function, representing the fraction of surface occupied
by packets of age between t and t + dt. Thus, the age distribution function of emulsion
elements on the surface needs to be defined. Two commonly used age distribution functions
for random surface renewal and for uniform surface renewal are discussed in the following.
The mode of random surface renewal exists for a surface that is in the main body of a
suspension and is continually contacted by rising bubbles. The age distribution of elements
on the surface is represented by that of a continuous-stirred tank reactor (CSTR), as given
by
(12.25)
Therefore, we obtain the following expressions for the average heat transfer coefficient for
two extreme cases: For rapid renewal when tc <^ <5gm/Dtem, we have
28emn
h = (12.26)
n=\
and for slow renewal when tc ^> 8lm/Dtcm, we have
h = (12.27)
In the mode of uniform surface renewal, all elements of emulsion contact the surface for
the same length of time; such a situation is representative of emulsion flowing smoothly
past a small heat transfer surface. Here, the age distribution function is represented by that
of a plug flow reactor, as given by
for 0 < t < tc
1(0 = 0 for t > tc
(12.28)
Thus, for rapid renewal, we have
(12.29)
and for slow renewal, we have
(12.30)
Different forms of thefilm-penetrationmodel and the controlling mechanisms under various

conditions are illustrated in Fig. 12.8.
It should be noted that the preceding treatment of the emulsion phase/packet model is
suitable for a system with homogeneous emulsion phase (i.e., particulate fluidization). The
model needs to be modified when applied to the fluidized bed with a discrete bubble phase.
Rapid renewal of Slow renewal of

emulsion layer emulsion layer
Conduction through
emulsion layer
Uniform surface
renewal
1.0 2.0 3.0 4.0
Figure 12.8. Heat transfer coefficient as a function of (AemO^Vêm for random and uniform
surface renewal (from Toor and Marchello, 1958; Yoshida et ai, 1969).
Modifications should account for the time fraction of the heating surface covered by the
bubbles/voids and changes in the pseudothermal properties.
Example 12.1 In fluidization of microspherical catalysts, the particle and gas proper-
ties are given as [Yoshida et al., 1969] dv = 152 /xm, p p = 1,540 kg/m3, Umf = 0.02 m/s,
amf = 0.505, c = 920 J/kg K, Kv = 0.139 W/m K, and K = 2.62 x 10"2 W/m K. The
following conditions are also given: fluidizing velocity U =0.1 m/s and Nup = 2.33.
(a) Assuming the film model applies, find the film thickness.
(b) Using the emulsion phase model and assuming random surface renewal, determine
the mean surface renewal time with rapid replacement and the effective thickness
of the emulsion layer with slow replacement.
(c) Calculate the mean surface renewal time, assuming uniform surface renewal.
Solution
(a) The value for h can be calculated from Nup as
and assuming the film model, we have
(b) Assuming random surface renewal, for rapid renewal, we have (from Eq. (12.26))
h = ^b= (E12.3)
Combining this equation with Eq. (12.21) yields
_ Kemppc(l - aM) _ [Kamf + Kp(l - am{)]ppc(l - amf)
tc
~ h* ~ Jfl
[2.62 x 10"2 x 0.505 + 0.139 x (1 - 0.505)] x 1,540 x 920 x (1 - 0.505)
= 0.36 s (E12.4)
For slow renewal, from Eq. (12.27), we have
_ Kem _ Kamf + Kp(l-am{) _ 2.62 x 10~2 x 0.505 + 0.139 x (1 - 0.505)
h h 402
= 204/xm (E12.5)
(c) Assuming uniform surface renewal, from Eq. (12.29), we have
_ 4Kemppc(l - ttmf) _ 4[Kamf + Kp(\ - Qfmf)]ppc(l - a m f )
7th2 ~ nh2
_ 4 x [2.62 x 10"2 x 0.505 + 0.139 x (1 - 0.505)] x 1,540 x 920 x (1 - 0.505)
" 3.14 x4022
= 0.45 s (E12.6)
12.3 Heat Transfer in Dense-Phase Fluidized Beds

As noted, most of the heat transfer models and correlations for gas-solid fluidiza-
tion systems were originally developed for dense-phase fluidized beds (see Chapter 9). In
the following, the behavior of heat transfer coefficients between the suspension (or bed)
and the particle, between the suspension (or bed) and the gas, and between the suspension
(or bed) and the wall or heat transfer surface are discussed.
12.3.1 Particle-to-Gas and Bed-to-Gas Heat Transfer

The particle-to-gas heat transfer can be obtained by unsteady-state experiments
which measure the time required for cold particles of temperature r po , mass M, and surface
area Sv to reach the bed temperature when they are introduced into the bed. Assuming that
the bed is well mixed and the bulk solids temperature in the bed is the same as the gas
temperature, the heat balance gives
cMdTv = hgvSv(Tb - Tp)dt (12.31)
where /zgp is the local particle-to-gas heat transfer coefficient. Integrating the preceding
equation yields
e o ( ^ ^ ) (12.32)
r b -r p o
The particle-to-gas heat transfer coefficient in dense-phase fluidization systems can be
determined from the correlation [Kunii and Levenspiel, 1991]
Nugp = ^ 2 + (0.6 ~ 1.8) Re^ f/2 Pr 1 ' 3 (12.33)
where Repf = dvU pj [i\ Pr = QVII/K\ and U is the superficial gas velocity. Equation (12.33)
indicates that the values of the particle-to-gas heat transfer coefficient in a fluidized bed lie
between those for afixedbed with large isometric particles (with a factor of 1.8 in the second
term; Ranz, 1952) and those with a factor of 0.6 in the second term of the equation. It is
noted that the equation with the factor of 0.6 coincides with that for the single-particle heat
transfer coefficient, i.e., Nup, obtained with Rep based on the relative velocity of particle to
gas as given in Eq. (4.40).
For suspension-to-gas (or bed-to-gas) heat transfer in a well-mixed bed of particles,
the heat balance over the bed under low Biot number (i.e., negligible internal thermal
resistance) and, if the gas flow is assumed to be a plug flow, steady temperature conditions
can be expressed as
cpUp dT = -hhgSB(T - r b ) dH (12.34)
where dT is the change in gas temperature when flow is through a layer of bed with height
dH; hbg is the suspension-to-gas heat transfer coefficient; and 5B is the surface area of
particles per unit bed volume as given by
SB (12.35)
12 3 I Heat Transfer in Dense-Phase Fluidized Beds 513
i
10
Eq. (12.37)
1.0 Range of Nubg from

reported literature data
/
0.1 , 1
1.0 10 100
Re pf orRe p
Figure 12.9. Particle-to-gas and bed-to-gas heat transfer coefficients under various flow
conditions (from Kunii and Levenspiel, 1991).
Integrating Eq. (12.34) gives
H (12.36)
Ji-ThJ \Upcv,
where T\ is the gas temperature at the inlet.
The range of data for the bed-to-gas heat transfer coefficient reported in the literature,
which were primarily determined on the basis of Eq. (12.36), is shown in the shadowed area
in Fig. 12.9. It can be seen from thefigurethat, under high Reynolds numbers (Repf > 100),
values of h^g are very close to those of hgv determined by Eq. (12.33) as the plug flow
assumption for the gas phase in the bed is realistic. However, this is not the case under low
Reynolds numbers (Repf < 100). Values of Nubg under low Reynolds numbers, as in fine
particle fluidization, are smaller than Nugp in terms of Eq. (12.33) and are much smaller
than the value of 2 for an isolated spherical particle in a stationary condition expressed by
Eq. (4.15). Nubg under this Reynolds number range follows the correlation given by
= 0.03 Reif3 (12.37)

pi v /
It should be mentioned that this deviation is more model-dependent than mechanistic be-
cause the real gas-solid contact is much poorer than that portrayed by plug flow, on which
Eq. (12.36) is based [Kunii and Levenspiel, 1991]. The deviation can also be related to the
effects of the particle boundary layer reduction due to particle collision and the generation
of turbulence by bubble motion and particle collision [Brodkey et al., 1991].
12.3.2 Bed-to-Surface Heat Transfer

For the bed-to-surface heat transfer in a dense-phase fluidized bed, the particle
circulation induced by bubble motion plays an important role. This can be seen in a study of
heat transfer properties around a single bubble rising in a gas-solid suspension conducted
D
115
Rising bubble axis close

to probe
Bubblerisingto side of
A the probe
100
1.0 2.0
Time, s
Figure 12.10. Probe-to-bed heat transfer coefficient variations in a fluidized bed (from Tuot
and Clift, 1973).
by Tuot and Clift (1973). Using a sensitive probe of low heat capacity with fast response
time, they observed that the heat transfer coefficient increased as the bubble rose toward
the probe (points A to B on the solid line in Fig. 12.10). The increase results from some
particle movement close to the probe surface as the bubble approaches from beneath. As
the bubble envelopes the probe, the heat transfer coefficient decreases (point C on the
solid line) because of the lower thermal conductivity and heat capacity of the gas. Further
rising of the bubble leads to a peak in the signal behind the bubble (point D) which is
the effect of the high concentration of particles carried by the wake passing the probe.
The relatively slow decay beyond point D to a new steady value is due to the turbulence
in the medium. The dashed line in Fig. 12.10 shows the heat transfer signal caused by a
bubble rising to the side of the probe and the maximum is again the effect of the high
concentration of particles carried in the wake of the bubble. Thus, it is evident that the
bubble wake plays a significant role in particle circulation and hence the heat transfer in
gas-solid fluidization.
The surface heat transfer is contributed to by three major components, as given in
Eq. (12.1), i.e., the particle convective component, the gas convective component, and
radiation. In gas-solid fluidization systems, radiation may be neglected when the bed tem-
perature is lower than 400 C. The significance of particle convection and gas convection
depends mostly on the types of particles used. As a rule of thumb, particle convection
is the dominant mechanism for small particles (dv < 400 /xm); it usually plays a key role
for Group A particles (see 9.2.1). Gas convection becomes dominant for large particles
(dp > 1,500 fim) and for high-pressure or high-velocity fluidizations; it usually plays a key
role for Group D particles [Maskaev and Baskakov, 1974]. For Group B particles, both
components are significant.
12.3.2.1 Particle Convective Component

Particle convection, caused by the particle motion within the bed, is concerned with the
heat transfer from a surface when it is in contact with the particulate emulsion phase instead
of the void/bubble phase. Thus, the heat transfer coefficient of particle convection can be
123 I Heat Transfer in Dense-Phase Fluidized Beds 515
defined as
v=
= : ^ \ -, ( 12 - 3g )
particle convective heat resistance
where the particle convective heat resistance can be further divided into the following two
series resistances: (1) average packet (particulate phase) resistance, 1/AP, and (2) film
resistance, l/hf. Thus, Eq. (12.38) can be expressed by
V =T 7 ^ r (12.39)
1/Ap + l/Af
where bubble fraction a\> can be calculated from Eq. (9.43) and /zp can be calculated from
hp = - [C h{dt (12.40)
where h\ is the instantaneous heat transfer coefficient averaged over the contact area.
Considering the thermal diffusion through an emulsion packet and assuming that the
properties of the emulsion phase are the same as those of minimum fluidization, h x can be
expressed by [Mickley et al., 1961]

2 yi
The surface-emulsion phase contact time tc can be estimated by assuming that the time
fraction for the surface to be covered by bubbles equals the bubble volume fraction in the bed
to = ^ (12.43)
/b
where / b is the bubble frequency at the surface. Equations (12.42) and (12.43) yield
2
22 /w P (i mf) c/y/
b y
Furthermore, Kem can be expressed by [Ranz, 1952]
Kem = Ke + 0.1 pcpdpUmf (12.45)
where Ke is the effective thermal conductivity of a fixed bed with stationary gas. In the
fixed bed, heat transfers through the gas and solid via parallel paths, as seen in Fig. 12.11 (a).
For this case, the effective conductivity is given by
Ke = amfK + (1 - amf)Kp (12.46)
To account for the actual geometry and small contact region between the adjacent particle
in the fixed bed with stationary gas, as shown in Fig. 12.1 l(b), Eq. (12.46) was modified
by Kunii and Smith (1960) to
where <pb is the ratio of an equivalent thickness of gas film to particle diameter.
Heat flow Heat flow
, Through the contact point
Gas
Solid
Through the gas layer
, Through the gas

(a) (b)
Figure 12.11. Conceptual representation of models for evaluating the effective thermal
conductivity of a fixed bed (from Kunii and Levenspiel, 1991): (a) Simple parallel path model;
(b) Modified model of Kunii and Smith (1960).
For film resistance, the film heat transfer coefficient can be expressed by
Af = ^ (12.48)
where is a factor ranging from 4 to 10 [Xavier and Davidson, 1985]. Thus, the particle
convective heat transfer component hvc can be calculated from Eqs. (12.39), (12.44), and
(12.48).
12.3.2.2 Gas Convective Component

The gas convective component is concerned with the surface heat transfer due to gas per-
colating through the particulate phase and the surface heat transfer due to gas bubbles.
For small particles, though the gas convective component is small in the bulk (i.e., in-bed
region), it can be important in the freeboard region. The heat transfer coefficient of gas
convection in general varies with the geometry of the heat transfer surface. However, it can
be approximated without treating specific surface geometries, as suggested by Baskakov
et al. (1974) for 0.16 mm < dv < 4 mm
^ = 0.009 A r ^ P r 1 / 3 (12.49)
K
or as proposed by Denloye and Botterill (1978) for 103 < Ar < 2 x 106 and for operating
pressure up to 1 MPa
(12.50)
K
where the coefficient 0.86 has the unit of meter~ 1/2
The heat transfer rate for the gas convective component can be regarded as comparable
to that at incipient fluidizing conditions. Thus, assuming hgc = /zmf, Xavier and Davidson
(1985) simulated the system by considering a pseudofluid with the apparent thermal con-
ductivity K& of the gas-solid medium flowing at the same superficial velocity and the same
inlet and outlet temperatures as the gas. Therefore, the heat conduction of the fluid flowing
12 3 I Heat Transfer in Dense-Phase Fluidized Beds 517
through a cylindrical column can be expressed by Eq. (4.43) as
dz pcpU \dr2 r dr J
Here the axial conduction is considered negligible in comparison with the convection due
to the flowing gas. This assumption may be invalid for fluidizations with very low gas
velocities.
For a vertical heating surface, the boundary conditions can be imposed as
0 <r < , z = 0, T = Ti
2
D (12.52)
0 < z < L, r = , T =T S
If T2 is the mean temperature of the gas leaving the bed at z = L, the temperature distribution
in the bed can be obtained as
where A.n are eigenvalues of the eigenequation /o(A) = O, with the first three being
A.i = 2.450, X2 = 5.520, and k3 = 8.645.
12.3.2.3 Radiative Component

Radiative heat transfer plays an important part in many fluidized bed processes operated at
high temperatures, such as coal combustion and gasification. When treating a fluidized bed
as a whole "solid" gray body, the radiative heat transfer coefficient hT between the fluidized
bed at temperature Tb and a heating surface at temperature Ts is defined as
hT = - ^ = <7bbs (7b2 + Ts2) (7i, + Ts) (12.54)

T T
where Jr is the radiant heat flux, <7b is the Stefan-Boltzmann constant, and 6bs is the
generalized emissivity, which depends on the shape, material property, and emissivity of
the radiating and receiving bodies [Baskakov, 1985]. For two large, parallel, and perfect
gray planes, we have
6bs = ( l / 6 b + 1/S - I)" 1 (12.55)
The effective bed-to-surface emissivity is larger than the particle emissivity, e p, because of
the multiple surface reflections. For the same reason, an irregular surface has a higher emis-
sivity than a smooth surface of the same material. It can be recognized that the importance
of radiative heat transfer significantly increases with the temperature. Increasing particle
size also increases the relative radiative heat transfer [Baskakov et al., 1973].
The general bed emissivity for bed-to-surface radiation may be estimated from the emis-
sivity of an individual particle as [Grace, 1982]
6 b s ^ 0.5(1+6 P ) (12.56)
Another expression for ebs is suggested by Baskakov (1985)
i T ^ 6b6s (12.57)
l/b + l/e s - 1
/e 1
Equation (12.57) is valid when eb and es are greater than 0.8.
The radiant heat transfer coefficient becomes important at temperatures above 600C,
but it is difficult to predict. In general, depending on particle size, hY increases from being
approximately 8-12 percent of the overall heat transfer coefficient at 600C to being 20-30
percent of h at 800C [Baskakov et a/., 1973].
Example 12.2 An exothermic chemical reaction occurs in a gas-solid fluidized bed re-
actor. To maintain the reaction temperature at 350 K, 50 MJ/h of heat of reaction must be
removed from this reactor. Determine the heat transfer area needed in this reactor with the
following data: U = 0.25 m/s; p = 0.716 kg/m3; cv = 2,185 J/kg K; K = 0.034 W/m K;
/z = 1.09 x 10~5 kg/m s; p p = 2,600 kg/m3; c = 840 J/kg K; Kv = 1.9 W/m K; dv =
0.5 mm; a mf = 0.47; a b = 0.6; fb = 3;Umf = 0.192m/s; ^ = 0.85, = 6 . The cooling wa-
ter temperature is 293 K.
Solution To determine the heat transfer area, the overall heat transfer coefficient
given in the following needs to be calculated first from Eq. (12.1). Radiative heat transfer
would be low for the temperature of 350 K and, thus, it is neglected.
(1) The particle convective component hvc may be calculated from Eq. (12.39). The
heat transfer coefficients of film can be calculated from Eq. (12.48)
Similarly, the heat transfer coefficients of the packet can be calculated from
Eqs. (12.42), (12.45), and (12.47) and can be obtained as 759 W/m 2 K. Thus,
hpc is found to be 106 W/m 2 K.
(2) The gas convective component can be calculated by Eq. (12.50) as
hgc = 0.86 Ar 039 K /^fdv = 0.86 x (19.2 x 10 3 ) 039 x 0.034/^0.5 x 10" 3
= 61.2 W/m 2 . K (E12.8)
After adding hv to hgCi the surface to bed heat transfer coefficient becomes 167
W/m2- K. Thus, the heat transfer area can be determined as
(E.2.9)
(350 293) x 167
12.3.3 Effect of Operating Conditions

A unique feature of the dense-phase fluidized bed is the existence of a maximum
convective heat transfer coefficient hmax when the radiative heat transfer is negligible. This
feature is distinct for fluidized beds with small particles. For beds with coarse particles,
the heat transfer coefficient is relatively insensitive to the gas flow rate once the maximum
value is reached.
For a given system, /zmax varies mainly with particle and gas properties. For coarse
particle fluidization at U > Umf, the heat transfer is dominated by gas convection. Thus,
^max can be evaluated from Eq. (12.50). On the other hand, hmax in afineparticle bed can be
reasonably evaluated from the equations for hpc. In general, /i max is a complicated function
123 I Heat Transfer in Dense-Phase Fluidized Beds 519
500 1
1
400 -
300 -
1 1 1 /
\ \ /
200 -
1 Y
/
/ 0.1 MPa
v 0.5 MPa -
J I /
/ A 1.0 MPa
o 2.0 MPa
- /s) 1
i
100 40 mm vertical flat _
surface and 0.475 mm
/c/^* glass beads-N2 system
i 1 i
0.1 0.2 0.3 0.4
U9 m/s
Figure 12.12. Effect of pressure on bed-to-surface convective heat transfer (from Xavier et al.,
1980).
of /*pc,max> hf, and other parameters. An approximation for this relationship was suggested
by Xavier and Davidson (1985) as
0.84
K (12.58)
*pc,max
The convective heat transfer hc (=/*pc + hgc) is influenced by both the pressure and the
temperature. An increase in pressure increases the gas density, yielding a lower U mf. Thus,
a pressurized operation enhances the convective heat transfer, as illustrated in Fig. 12.12.
Under vacuum operation, hc is lower than under ambient pressure operation as a result
of lower gas density and a decrease in Ke with decreasing pressure [Bock and Molerus,
1985]. For fluidized beds with small particles, increasing pressure enhances solids mixing
and hence particle convection [Borodulya et al., 1982]. For large particles, the heat transfer
is dominated by gas convection, which is enhanced with an increase in gas density through
the elevation of pressure. In general, the convective heat transfer significantly increases
with pressure for Group D particles; however, the pressure effect decreases with decreasing
particle size. For Group A and Group B particles, the increase in hc with pressure is
small.
1,000
Fluidizing gas: air, C0 2 , a 20C,O.lMPa
andAr b 20C,0.6MPa
c 600C,O.lMPa
b
100
10
0.1 1 10
Particle diameter, mm
Figure 12.13. Effect of temperature and pressure on hpc, hgc, and /zmax for bed-to-surface heat
transfer (from Botterill et al., 1981).
At high temperatures, the decreased gas density can decrease the gas convective compo-
nent hgc. On the other hand, the increased gas conductivity at high temperature can increase
hgc, KQ, and hvc. For a bed with small particles, the latter is dominant. Thus, a net increase
in hc with increasing temperature can be observed before radiation becomes significant.
For Group D particles, hc decreases with increasing temperature [Knowlton, 1992], The
effects of temperature and pressure on Apc, hgc, and hmax are illustrated by Fig. 12.13.
For radiative heat transfer, some differences between a fluidized bed and a fixed bed
exist [Kovenskii, 1980]. When a bed starts to fluidize from the fixed state, the radiative heat
transfer rapidly increases by 10-20 percent; it remains constant as the gas velocity increases
further in the bubbling bed.
Influidizedbed heat transfer, it is a common practice to use internals such as water-cooled
tubes in the bed. It is important to know the effect of immersed objects on the local fluidiza-
tion behavior and local heat transfer coefficient. Without loss of generality, we consider a
case in which a horizontal water-cooled tube is placed in a hot fluidized bed [Botterill et ai,
1984]. The interference of the immersed tube with the particle circulation pattern leads to an
increased average particle residence time at the heat transfer surface. The common forma-
tion of stagnant regions of particles on the top or nose of the horizonal tube is evidence of such
interference. Furthermore, at high gas flow rates, gas packets are frequently present at the
upper part of the horizontal tube. Thus, there is a significant variation in the heat transfer co-
efficient around the tube. As illustrated in Fig. 12.14, for fluidization in a bed of an immersed
horizontal tube, the heat transfer coefficient near the nose is considerably smaller than that
around the sides, indicating the reduction of convective heat transfer by the stagnation of gas
12.4 I Heat Transfer in Circulating Fluidized Beds 521
-30 30
600/^
) / 1
500 ^ 7
2SCX?\
-60 / XPX > 400 -J
x
60
( A/\ >r
200'^
iV
A\
-90 90
p.\p
-120 120
- A D t/=0.23m/s/
x
fif_ = 0.37 mm X / \
\ P 7/^T\
/ ^ "^ / /Gagflow\ \^ O C/=0.89m/s\/
-150 180 150
Figure 12.14. Variation of local heat transfer coefficient around a horizontal tube immersed in
a fluidized bed of alumina (from Botterill et aL, 1984).
near the nose. For large particles for which the interphase gas convective component of heat
transfer is significant, the circumferential variations would be somewhat similar with peaks
at the sides, but the variation is less than that with small particles, since the particle convec-
tion is no longer dominant. The overall heat transfer coefficient as a result of orientation usu-
ally does not vary much; that for a horizontal tube is slightly lower than that for a vertical tube.
It is noted that most of the models and correlations that are developed are based on
bubbling fluidization. However, most of them can be extended to the turbulent regime with
reasonable error margins. The overall heat transfer coefficient in the turbulent regime is a
result of two counteracting effects, the vigorous gas-solid movement, which enhances the
heat transfer and the low particle concentration, which reduces the heat transfer.
12.4 Heat Transfer in Circulating Fluidized Beds

The mechanism of heat transfer in circulating fluidized beds is described in this
section. Effects of the operating variables on the local and overall heat transfer coefficients
are discussed.
12.4.1 Mechanism and Modeling

The suspension-to-wall surface heat transfer mechanism in a circulating fluidized
bed (see Chapter 10) comprises various modes, including conduction due to particle clusters
on the surface or particles falling along the walls, thermal radiation, and convection due to
gas flow or particle clusters or particle flow. For modeling of heat transfer in a circulating
fluidized bed, the heat transfer surface is commonly portrayed to be covered alternatively
by cluster and dispersed particle phase [Subbarao and Basu, 1986; Wu et al., 1990]. Thus,
mechanistically, the packet model developed for dense-phase fluidized beds can be applied,
if the "packet" represents the "cluster."
In circulating fluidized beds, the clusters move randomly. Some clusters are swept from
the surface, while others stay on the surface. Thus, the heat transfer between the surface
and clusters occurs via unsteady heat conduction with a variable contact time. This part of
heat transfer due to cluster movement represents the main part of particle convective heat
transfer. Heat transfer is also due to gas flow which covers the surface (or a part of surface).
This part of heat transfer corresponds to the gas convective component.
12.4.1.1 Particle Convective Component

The particle convection is in general important in the overall bed-to-surface heat transfer.
When particles or particle clusters contact the surface, relatively large local temperature
gradients are developed. This rate of heat transfer can be enhanced with increased surface
renewal rate or decreased cluster residence time in the convective flowof particles in contact
with the surface. The particle-convective component hpc can be expressed by the following
equation, which is an alternative form of Eq. (12.39):
A <12 59)
7 ; '
Thus, hpc is determined from the wall (film) resistance, 1/Af, in series with the transient
conduction resistance of a homogeneous semiinfinite medium, \/hv. By analogy with
Eq. (12.48), hi can be expressed by [Gloski et al, 1984]
hf=r (12.60)
where <5* is the dimensionless effective gas layer thickness between wall and cluster (ratio of
gas layer thickness to particle size), which is mainly a function of cross-sectionally averaged
particle volume fraction [Lints and Glicksman, 1993]. The range of 5* is exemplified in
Fig. 12.15.
Similar to Eq. (12.42), hv can be expressed as [Lints and Glicksman, 1993]
For a very short cluster contact time or very large fluidizing particle, hf dominates the
particle convection and thus Eq. (12.59) yields
12.4.1.2 Gas Convective Component

In practice, hgc may be evaluated by one of the following approaches:
(1) Extended from Eq. (12.50) for hgc in dense-phase fluidized beds
(2) Approximated as that for dilute-phase pneumatic transport [Wen and Miller, 1961;
Basu and Nag, 1987]
(3) Estimated by the convective coefficient of single-phase gas flow [Sleicher and
Rouse, 1975]
12.4 I Heat Transfer in Circulating Fluidized Beds 523
1.5
A C/=2.3m/s
Particle: 182 urn sand
[/=3.0m/s
I 1.0 o [/=3.7m/s
[/=4.0m/s
(7=4.5m/s
v $
13 0.5 I
o
l
I
:
1 i i l 1 i
.005 .010 .015 .020
Averaged solids holdup (1 - oc)
Figure 12.15. Variation of gas layer thickness, 5*, between the cluster and wall with averaged
solids holdup (from Lints and Glicksman, 1993).
For high particle concentration on the surface with large dimension, any one of these
approaches is reasonable because of the small value of hgc. For low particle concentrations
and high temperatures, using these approaches may cause discrepancies in hgc.
12.4.1.3 Radiative Component

To understand the radiative heat transfer in a circulating fluidized bed, the bed can be
regarded as a pseudogray body. The radiative heat transfer coefficient is [Wu et al., 1989]
(12.63)
+ --l)(Tb-Ts)
where esus is the emissivity of the suspension.

An alternative treatment for radiative heat transfer in a circulating fluidized bed is to
consider the radiation from the clusters (hCT) and from the dispersed phase (i.e., the remaining
aspect of gas-solid suspension except clusters, hdT), separately [Basu, 1990]
hT = achC - ac)hdT (12.64)
where ac is the volume fraction of clusters in the bed. These two components can be defined
by
hcr = (12.65)
- + --i)(rb-rs)
(12.66)
Ts)
The emissivity of the cluster c can be determined by Eq. (12.56), and the dispersed phase
emissivity e^ is given by [Brewster, 1986]
+2 - (12.67)
where B is taken as 0.5 for isotropic scattering and 0.667 for diffusely reflecting parti-
cles.
12.4.2 Radial and Axial Distributions of Heat Transfer Coefficient

As opposed to the relatively uniform bed structure in dense-phase fluidization, the
radial and axial distributions of voidage, particle velocity, and gas velocity in the circulating
fluidized bed are very nonuniform (see Chapter 10); as a result the profile for the heat transfer
coefficient in the circulating fluidized bed is nonuniform.
In the axial direction, the particle concentration decreases with height, leading to a
decrease in the cross section averaged heat transfer coefficient. In addition, the influence
of the solids circulation rate is significant at lower bed sections but less significant at
upper bed sections, as illustrated in Fig. 12.16. In the radial direction, the situation is more
complicated as a result of the uneven radial distribution of the particle concentration as
well as the opposite solids flow directions in the wall and center regions. In general, the
coefficient is relatively low and approximately constant in the center region. The coefficient
increases sharply toward the wall region. Three representative radial profiles of the heat
transfer coefficient with various particle holdups reported by Bi et al. (1989) are shown in
a / a
t
300 300 - a I 300 -
(kg/m2.s) (kg/m2 s) (kg/m2.s)
V 84.4 0.820 ^-O v 133.8 0.800 V 133.8 0.930
//A V
A 68.2 0.850 // A 93.9 0.900 I A 93.9 0.950
250 250 / 250
o 56.3 0.905 41 h
I o
1
o 73.1 0.960 1 73.1 0.960
0.945 IT I a
1I
D 42.6 D 42.1 0.970 / 42.1 0.980
200 200 200

v
/4
150
wSr-
>-o
Jf 150u 7
;
1
r * ^ ^
V
A
/ A /
I7
150
i
r
100
0.0
-aD- -a
0.5 1.0
1001
r-
0.0
O
0.5 1.0
100
0.0 0.5 1.0

rlR rlR r/R
(a) (b) (c)
Figure 12.16. Radial distributions of overall heat transfer coefficient in a circulating fluidized
bed of dp = 280/urn and p p = 706 kg/m 3 (from Bi etal, 1989): (a) U = 3.7m/s, H = 1.25 m;
(b) U = 6.0 m/s, H = 1.25 m; (c) U = 6.0 m/s, H = 6.5 m.
72.4 / Heat Transfer in Circulating FluidizedBeds 525
Fig. 12.16 as described in the following:
(1) When the particle holdup is high, the contribution of hpc plays a dominant role
and /zgc is less important. The radial distribution of the heat transfer coefficient is
nearly parabolic, as shown in Fig. 12.16(a). Such a heat transfer profile is similar
to the solids concentration profile described in Chapter 10.
(2) As the gas velocity increases, the solids holdup decreases and, thus, hgc begins
to become as important as hpc. In the center region of the riser, hgc is dominant,
and its influence decreases with an increase in the solids holdup along the radial
direction toward the wall. In the near-wall region, hpc dominates the heat transfer.
The contribution of hpc decreases with a decrease in the particle concentration
toward the bed center. As a result, a minimum value of h appears air/R of about
0.5-0.8, as indicated in Fig. 12.16(b).
(3) With further decrease in the particle concentration at a > 0.93, hgc becomes dom-
inant except at a region very close to the wall. Thus, the heat transfer coefficient
decreases with increasing r/R in most parts of the riser, as shown in Fig. 12.16(c),
exemplifying the same trend as the radial profile of the gas velocity. In the region
near the wall hpc increases sharply, apparently as the effect of relatively high solids
concentration in that region.
12.4.3 Effect of Operating Parameters

The overall heat transfer coefficient can be influenced by the suspension density,
solids circulation rate, gas velocity, particle properties, bed temperature, pressure, and di-
mensions of the heating surface. The overall heat transfer coefficient increases with suspen-
sion density [Wu et al., 1989] and with particle circulation rate. The increase in gas velocity
appears to have two counteracting effects, namely, enhancing hgc through an increased gas
convection effect while reducing particle convective heat transfer through reduced particle
concentration, as illustrated in Fig. 12.17. When hvc dominates (in the near-wall region with
high particle concentration), h decreases with increasing U. On the other hand, h increases
with U if hgc is important (e.g., in the central region where the particle concentration is
small). Another reason for the decrease of h in the near-wall region is the reduced particle
downward velocity caused by increasing U, which results in prolonged particle-surface
contact.
In general, small/light particles can enhance heat transfer. The cluster formation in
small/light particle systems contributes to the enhancement of h pc. Also the gas film
resistance can be reduced by fluidizing with small particles [Wu et al., 1987]. When the
temperature is lower than 400C, the effect of bed temperature on the heat transfer coeffi-
cient is due to the change of gas properties, while hT is negligible. At higher temperatures, h
increases with temperature, mainly because of the sharp increase of radiative heat transfer.
Measurements of heat transfer in circulating fluidized beds require use of very small
heat transfer probes, in order to reduce the interference to the flow field. The dimensions of
the heat transfer surface may significantly affect the heat transfer coefficient at any radial
position in the riser. All the treatment of circulating fluidized bed heat transfer described
is based on a small dimension for the heat transfer surface. The heat transfer coefficient
decreases asymptotically with an increase in the vertical dimension of the heat transfer
surface [Bi et al., 1990]. It can be stated that the large dimensions of the heat transfer surface
175
O U=3.65m/s a = 0560
U= 6.00 m/s </p = 280 urn
A U= 7.33 m/s
100
I 1 I
75
0.0 0.2 0.4 0.6 0.8 1.0
rlR
Figure 12.17. Effect of gas velocity on the radial distribution of the heat transfer coefficient in
a circulating fluidized bed (from Bi et ai, 1989).
can prolong the residence time of particles or particle clusters on the surface, resulting in
lower renewal frequency and hence a low apparent heat transfer coefficient.
12.5 Heat Transfer in Spouted Beds

This section discusses the behavior of heat transfer between gas and particles, and
between the bed and the surface in a spouted bed.
12.5.1 Gas-to-Particle Heat Transfer

The heat transfer behavior in a spouted bed (see 9.8) is different from that in
dense-phase and circulating fluidized bed systems as a result of the inherent differences
in their flow structures. The spouted bed is represented by a flow structure that can be
characterized by two regions: the annulus and the central spouting region (see Chapter 9).
The heat transfers in these two regions are usually modeled separately. For the central
spouting region, the correlation of Rowe and Claxton (1965) can be used for RePf > 1,000
2 2
where acs is the gas volume fraction in the central spouting region. In the annulus, the heat
transfer can be described by using the correlations for fixed beds, for example, Littman and
Sliva's (1971) equation for Repf < 100
Nugp = 0.42 + 0.35 Re|,0.8
pf
(12.69)
Substituting the corresponding values for the spouted bed into Eqs. (12.32), (12.68), and
(12.69) reveals that the distance required for the gas to travel to achieve a thermal equilibrium
with the solids in the annulus region is on the order of magnitude of centimeters, while this
distance in the spout region is one or two orders of magnitude larger.
The importance of the intraparticle heat transfer resistance is evident for particles with
relatively short contacting time in the bed and those with a large Biot number. Thus, for a
shallow spouted bed, the overall heat transfer rate and thermal efficiency are controlled by
the intraparticle temperature gradient. This gradient effect is most likely to be important
when particles enter the lowest part of the spout and come in contact with the gas at high
temperature, while it is negligible when the particles are slowly flowing through the annulus.
Thus, in the annulus, unlike the spout, thermal equilibrium between gas and particles can
usually be achieved even in a shallow bed.
12.5.2 Bed-to-Surface Heat Transfer

Compared to the fluidized bed, a spouted bed with immersed heat exchangers is
less frequently encountered. Thus, the bed-to-surface heat transfer in a spouted bed mainly
is related to bed-to-wall heat transfer. The bed-to-immersed-object heat transfer coefficient
reaches a maximum at the spout-annulus interface and increases with the particle diameter
[Epstein and Grace, 1997].
Since the solid particles in the spouted bed are well mixed, their average temperature
in different parts of the annulus can be considered to be the same, just as in the case of a
fluidized bed. The maximum value of the heat transfer coefficient in the h-U plot is also
similar to that in a dense-phase fluidized bed [Mathur and Epstein, 1974].
12.6 Mass Transfer in Multiparticle Gas-Solid Systems

Values of the mass transfer coefficient for multiphase systems depend on the model
which defines this coefficient. Hence, mass transfer coefficients should be used in conjunc-
tion with the corresponding model equations.
12.6.1 Mass Transfer in Dense-Phase Fluidized Beds

In a dense-phase fluidized bed (see Chapter 9), mass transfer can take place
between the particle and the gas, between the bubble and the emulsion, and between the
bed and the surface. These processes are discussed in the following.
12.6.1.1 Gas-to-Particle Mass Transfer

Under fairly low gas velocity conditions where U is close to Umf or the bed is in particulate
fluidization, the plug flow assumption for the gas phase can be reasonably made [Wen and
Fane, 1982]. Considering the sublimation of species A from the solid phase to the gas
phase, the mass balance on the concentration for species A in the gas phase, CA, over the
incremental height d// can be expressed as
U-r=kfSB(CsA-CA) (12.70)
where CSA is the saturation concentration of species A on the solid surface and kf is the
particle-to-gas mass transfer coefficient.
80
Naphthalene particles in presence
of inert polystyrene beads
60
Q
1 40
Key dp, mm Key d , mm

20
O 0.95 2.05
1.06 3.70 7=0.33 m/s
A 1.19 X 4.38
A 1.33 V 5.18
1.55 6.17
T ,,,!
-2 J
10 J
10 io iou 10
SB ,m
Figure 12.18. Dependence of mass transfer coefficient on naphthalene-specific interfacial area

and particle size (after de Kok et al., 1986).
Assuming kf does not change with height, de Kok et al. (1986) studied the dependence
of the overall mass transfer coefficient on the specific interfacial area. Figure 12.18 depicts
the dependence, which shows that kf is constant up to 5B = 2 m - 1 and then decreases.
Integrating Eq. (12.70) yields
U - CA,in
In (12.71)
Equation (12.71) provides a means to quantify kf on the basis of experimentally measured

inlet and outlet concentrations of species A in the bed under low gas velocity conditions.
Wen and Fane (1982) proposed the following empirical correlations for overall mass transfer
coefficients in gas-solid fluidized beds, using the experimental data of Kato et al. (1970)
based on Eq. (12.71):
0.6 0.97 0.6
Sh = 0.43 Sc0 3 3 for 0.5 < Re pf -^ < 80 (12.72)
and
0.2
( d \0'6
0.6'
Sh = 12.5 Re pf [ ^ -^-J < 1,000 (12.73)
where Sh = kfdv/Dm, Sc = /x/(pD m ), He = x s // f , and xs is the volume fraction of ac-

tive solids taking part in the mass transfer. The definition for Sh given assumes that the
logarithmic mean fraction of the inert or nondiffusing component is unity. When the bed
is made up of all active particles, xs = 1.
Under relatively high gas velocity, e.g., bubbling conditions, the axial mixing effect
needs to be added to Eq. (12.70). Thus, we have
^ U ^ = k'MCX - CA) (12.74)

where D a is the axial dispersion coefficient, which can be evaluated empirically. Clearly, in
the formulation given by Eq. (12.74), the mass transfer resistance surrounding the bubble
phase is neglected.
12.6.1.2 Bubble-to-Emulsion Mass Transfer

A realistic treatment of mass transfer between the gas and solid phases requires consider-
ation of the bed structure comprising the bubble phase and the emulsion phase (see 9.4).
Considering bubbles containing species A passing through a fluidized bed where species
A is in depletion, the mass transfer, or the mass interchange coefficient from the bubble
phase to the emulsion phase, Aê, can be related to the difference in the concentration of
species A in the bubble phase, CA,b> and that in the emulsion phase, CA?e, by [Kunii and
Levenspiel, 1968]
=
'v^-dT " d T = ~Ub~^T = * be(CA ' b " CA'e) (12 75)
*
The mass transfer resistance between the bubble and emulsion phases can be described in
terms of bubble-to-cloud and cloud-to-emulsion resistances in series. Thus, we have
=
^be(CA,b - CA,e) ^bc(CA,b ~ Â,c) = ^ c e (C A ) c ~ C\,e) (12.76)
where A'bc and Kce are bubble-to-cloud and cloud-to-emulsion interchange coefficients,
respectively. Kbe relates to A^c and KCQ by the equation
+ (12 77)
IT = IT IT -
A
Abe ^bc ce
Define &be according to
^ ^ Abebe(CA,b - CA,e) (12.78)
where Abe is the bubble-emulsion interfacial area. Combining Eqs. (12.75) and (12.78)
yields
KbQ = kbQ^ = ^kbQ (12.79)
By experimentally measuring the concentration of species A in the gas phase and emulsion
phase and the inlet gas concentration in a gas-solid system without chemical reaction or
adsorption, Aê can be determined on the basis of Eq. (12.75).
A'be can also be accounted for mechanically from its relationship with A'bc and Kce given
in Eq. (12.77). Following the analysis of Davidson and Harrison (1963) for Kbc and that
of Kunii and Levenspiel (1968) for Kce for bubbling fluidized beds based on the bubble
configuration described in 9.4.2.1, a material balance on a single rising bubble gives
= (Q + ^ 5 b c ) ( C A , b - CA,C) (12.80)
where Q is the volumetric flow rate of circulating cloud gas illustrated in Fig. 9.7(a) and
Sbc is the interfacial area between the bubble and its cloud (^nd^). The gas flow velocity
through a bubble is assumed to be 3Umf (see 9.4.5). Thus, Q can be expressed as
Q = Î/mfdb (12-81)
For the mass transfer coefficient between the bubble and its cloud due to diffusion, kbc,
Davidson and Harrison (1963) also derived the following expression by assuming a spherical
cap bubble with 0w = 50(Problem 12.5)
tbc = 0 . 9 7 5 ^ / V 7 Y / 4 (12.82)
where Dm is the molecular diffusion coefficient of gas. Combining Eqs. (12.75) and (12.80)
yields
Q + ^bcSbc mar*
^bc = (12.83)
Vb
From Eqs. (12.81) and (12.82), Kbc in Eq. (12.83) can be expressed as
Kbc = 4.5 ( ^ ) + 5.85 f ^ J M (12-84)

Davidson and Harrison (1963) expressed the total interchange coefficient for mass trans-
fer from the bubble to the emulsion, Kbe, by using Eq. (12.84), which is reasonable for
very fast bubbles with negligible cloud. For bubbles with a large cloud, the cloud-emulsion
mass transfer coefficient, Kce, should also be considered, as indicated in Eq. (12.77).
For the cloud-emulsion mass transfer, the mass balance over the cloud-emulsion
interface gives
dNAc
- ^ = S ce ce (C A , c - CA,e) (12.85)
at
where kce is the cloud-emulsion mass transfer coefficient and SCQ is the interfacial area
between the cloud and the emulsion. Following the analogy of bubble contact by liquid, the
penetration model can be applied, leading to the following expression for kce (Problem 12.6)
kCe = (12.86)
V ^'e /
where Dem is the effective diffusion coefficient in the emulsion phase and te is the surface
renewal time in the penetration model for bubble in contact with the emulsion phase. For
a bubble with cloud, tt can be approximated by
^ ^ (12.87)
Since the mass transfer from bubble-to-cloud and to-emulsion is a continuous process,
Eq. (12.75) can also be written as
_ 1 d^c = Kce(c _ C ) (12 88)
Vb at
Combining Eqs. (12.85) and (12.88) yields
tfce = ^ (12.89)
For a thin-cloud bubble, 5ce/VJ, can be approximated by 6/d\>. From Eqs. (12.86), (12.87),
and (9.37) with Jboo replaced with a\, Eq. (12.89) becomes
2
(12.90)
Example 12.3 Determine the bubble-to-emulsion phase interchange coefficient for mass
transfer in the following gas-solid fluidization system operated at 400 C and the ambient
pressure: dv = 510 /im, p p = 2,242 kg/m3, shape factor = 0.84, U = 0.5 m/s, Mp = 50kg,
D = 0.3 m, and Dm = 3 x 10"5 m2/s. The fluidizing gas is air.
Solution The minimum fluidization velocity can be obtained by Eq. (9.14) as

Umf = 0.113 m. The bed voidage at minimum fluidization can be determined by the Ergun
equation (Eq. (9.11)) as amf = 0.422. The expanded bed height can be obtained from
Eq. (9.47) as Hf = 0.638 m. The average bubble size in the bed can be calculated by
Eq. (E9.1) as 0.091 m. D em can be approximated by [Kunii and Levenspiel, 1991]
>em = An (E12.10)
The bubble-to-cloud interchange coefficient is obtained by Eq. (12.84)
The cloud-to-emulsion interchange coefficient is calculated by Eq. (12.90)
/2
= 0.72,-' (E12.12)
Therefore, Eq. (12.77) gives the gas-to-emulsion phase interchange coefficient for mass
transfer
Khe = | - j - j | = 0.65 s"1 (E12.13)

+
> 6\72 072 s
12.6.1.3 Bed-to-Surface Mass Transfer

Mass transfer from a surface to the gas-solid suspension is significantly higher than that for
particle-free conditions. This increase is due to a possible reduction of the mass transfer
boundary layer and an increase in the interstitial gas velocity when particles are present.
The effect on heat transfer is even more significant as solid particles also act as heat carriers.
For gas-solid fluidized beds, Wen and Fane (1982) suggested that the determination of
the bed-to-surface mass transfer coefficient can be conducted by using the corresponding
heat transfer correlations, replacing the Nusselt number with the Sherwood number, and
replacing the Prandtl number by Sc(cpp)/(cpp)/(l a). Few experimental results on bed-
to-surface mass transfer are available, especially for gas-solid fluidized beds operated at
relatively high gas velocities.
12.6.2 Mass Transfer in Circulating Fluidized Beds

The effectiveness of the gas-solid mass transfer in a circulating fluidized bed (see
Chapter 10) can be reflected by the contact efficiency, which is a measure of the extent to
which the particles are exposed to the gas stream. As noted in Chapter 10, fine particles
tend to form clusters, which yield contact resistance of the main gas stream with inner
particles in the cluster. The contact efficiency was evaluated by using hot gas as a tracer
[Dry et al., 1987] and using the ozone decomposition reaction with iron oxide catalyst as
particles [Jiang etal, 1991 ]. It was found that the contact efficiency decreases as the particle
concentration in the bed increases. At lower gas velocities, the contact efficiency is lower
as a result of lower turbulence levels, allowing a greater extent of aggregate formation. The
contact efficiency increases with the gas velocity, but the rate of increase falls with the gas
velocity.
Kwauk et al. (1986) measured the gas-solid mass transfer coefficient using CCU as
a gas tracer in a circulating fluidized bed of Group B particles. The mass transfer
coefficient in their study is defined in terms of plug flow for the gas phase and com-
plete mixing for the solid phase. They reported an increase in the mass transfer coeffi-
cient with increasing gas velocity within the fast fluidization regime. Haider and Basu
(1988) used the sublimation of naphthalene particles to study gas-solid mass transfer
in a circulating fluidized bed. They indicated that the mass transfer coefficient from
particle to bed varies significantly with the slip velocity between the gas and the solid
phase, Us (=[(U/a) Uv/(l a)]). A smaller overall slip velocity yields a smaller gas-
solid mass transfer coefficient. They proposed a modified form of the Froessling equation
(Eq. (4.96)):
(12.91)
Dm
where a is the bed voidage.
Nomenclature
Abe Bubble-emulsion interfacial CA,UI M e t concentration of species A
area Â,out Outlet concentration of species
Am Area of packet in contact with A
the heating surface c Specific heat of particles
Ar Archimedes number cp Specific heat at constant
B Constant, defined by Eq. (12.67) pressure
CA Concentration of species A c pc Specific heat of clusters
C\ Saturation concentration of Da Axial dispersion coefficient of
species A gas
CA,b Concentration of species A in Dm Molecular diffusion coefficient
the bubble of gas
CA, C Concentration of species A in Dem Effective diffusion coefficient in
the cloud the emulsion phase
Dtem Thermal diffusivity of the
CA,Q Concentration of species A in
emulsion phase
the emulsion phase
Nomenclature 533
db Volume diameter of bubble Ka Apparent thermal conductivity

dc Diameter of bubble cloud of a gas-solid suspension
Fo Fourier number flowing through an empty bed
bc Bubble-to-cloud interchange
/b Bubble frequency at surface
coefficient for mass transfer
g Gravitational acceleration
be
Bubble-to-emulsion interchange
H Height
coefficient for mass transfer
He Effective bed height g Thermal conductivity of clusters
Hf Expansion bed height % Cloud-to-emulsion interchange
A Heat transfer coefficient coefficient for mass transfer
A Bed-to-surface heat transfer Ke Effective thermal conductivity
coefficient of a fixed bed with stagnant
Abg Bed-to-gas heat transfer gas
coefficient Ken Thermal conductivity of the
Ac Convective bed-to-surface heat emulsion phase
transfer coefficient Kf Thermal conductivity of gas film
Acr Radiative heat transfer Kp Thermal conductivity of
coefficient of clusters particles
Adr Radiative heat transfer &t>c Bubble-to-cloud mass transfer
coefficient of the dispersed coefficient
phase &be Bubble-to-emulsion mass
Af Gas film heat transfer coefficient transfer coefficient
AgC Gas convective component kce Cloud-to-emulsion mass
ofA c transfer coefficient
Agp Gas-to-particle heat transfer kf Gas-to-solid mass transfer
coefficient of the bed coefficient
Ai Instantaneous heat transfer K Gas-to-solid mass transfer
coefficient, defined by
coefficient
Eq. (12.74)
^max Maximum value of A
M Mass of particles
Ap Average heat transfer coefficient
Total mass of particles in the bed
between the particulate phase
and surface in the absence of NA?b Number of moles of species A
gas film resistance in the bubble
Apc Particle convective component AfA c Number of moles of species A
of Ac in the cloud
Apc,max Maximum value of Apc Nu b g Bed-to-gas Nusselt number
Ar Radiative heat transfer Nu, Particle-to-gas Nusselt number
coefficient N u gp Particle-to-gas Nusselt number
I(t) Age distribution function in the for a single particle
film-penetration model n Coordinate normal to the
/q Heat flux temperature symmetric plane
JT Radiant heat flux n' Coordinate normal to the
K Thermal conductivity of gas gas-particle interface
Pr Prandtl number Up Superficial particle velocity

Q Volumetric flow rate in and out Us Slip velocity between the gas
of a single bubble and solid phases
r Radial coordinate Umf Minimum fluidization velocity
R Radius of the bed Uopt Superficial gas velocity at
Re p Particle Reynolds number based n = nmSLX
on particle diameter and relative Bubble volume

velocity x Distance from surface
Repf Particle Reynolds number based xs Volume fraction of solids taking
on particle diameter and part in mass transfer
superficial gas velocity z Axial coordinate
S Area mean stirring factor
SB Surface area of solids per unit Greek Symbols
volume of the bed
S\jC Interfacial area between the a Bed voidage
bubble and cloud o?b Volume fraction of bubbles in
Sce Interfacial area between the the bed
cloud and emulsion ac Volume fraction of clusters in
Sp Surface area of particles the bed
Sc Schmidt number acs Volume fraction in the central
Sh Gas-to-particle Sherwood spouting region
number amf Bed voidage at minimum
tc Contacting time of clusters or fluidization
the particulate phase and 8 Boundary layer thickness
surface
<5* Dimensionless gas layer
te Surface renewal time in the
thickness
penetration model for bubbles
in contact with the emulsion 8C Time-averaged fraction of wall
phase area covered by clusters
T\ Gas temperature at the inlet of <$em Layer thickness of emulsion on
the bed the surface
T2 Gas temperature at the outlet of 6bs General bed emissivity for
the bed bed-to-surface radiation
r^ Temperature of the emulsion 6b Emissivity of bed suspension
phase when x -> oo c Emissivity of clusters
Tb Bed temperature d Emissivity of dispersed
Tp Temperature of particles phase
Ts Temperature of the heating p Emissivity of particle surface
surface es Emissivity of heat transfer
U Superficial gas velocity surface
U\y Bubble rise velocity 6SUS Emissivity of suspension
Uboo Rise velocity of an isolated phase
bubble Parameter, defined by
Eq. (12.48)
References 535
Pem Density of emulsion phase \lr(tc) Age distribution function in the

pp Density of particles packet model
crb Stefan-Boltzmann's <Pb Ratio of the equivalent thickness
constant of gas film to particle diameter
References
Baeyens, J. and Goossens, W. R. A. (1973). Some Aspects of Heat Transfer Between a Vertical Wall
and a Gas Fluidized Bed. Powder Tech., 8, 91.
Baskakov, A. P. (1985). Heat Transfer in Fluidized Beds: Radiative Heat Transfer in Fluidized Beds.
In Fluidization, 2nd ed. Ed. Davidson, Clift and Harrison. London: Academic Press.
Baskakov, A. P., Berg, B. V., Vitt, O. K., Filippovsky, N. R, Kirakosyan, V. A., Goldobin, J. M. and
Maskaev, V. K. (1973). Heat Transfer to Objects Immersed in Fluidized Beds. Powder
Tech.,H,213.
Baskakov, A. P., Vitt, O. K., Kirakosyan, V A., Maskaev, V. K. and Filippovsky, N. F. (1974).
Investigation of Heat Transfer Coefficient Pulsations and of the Mechanism of Heat
Transfer from a Surface Immersed into a Fluidized Bed. Proceedings of International
Symposium on Fluidization Applications. Toulouse: Cepadues-Editions.
Basu, P. (1990). Heat Transfer in High Temperature Fast Fluidized Beds. Chem. Eng. Sci.,
45,3123.
Basu, P. and Nag, P. K. (1987). An Investigation into Heat Transfer in Circulating Fluidized Beds.
Int. J. Heat & Mass Transfer, 30, 2399.
Bi, H.-T., Jin, Y., Yu, Z. Q. and Bai, D.-R. (1989). The Radial Distribution of Heat Transfer
Coefficients in Fast Fluidized Bed. In Fluidization VI. Ed. Grace, Shemilt and
Bergougnou. New York: Engineering Foundation.
Bi, H.-T., Jin, Y, Yu, Z.-Q. and Bai, D.-R. (1990). An Investigation on Heat Transfer Circulating
Fluidized Bed. In Circulating Fluidized Bed Technology III. Ed. Basu, Horio and
Hasatani. Oxford: Pergamon Press.
Bock, H.-J. and Molerus, O. (1985). Influence of Hydrodynamics on Heat Transfer in Fluidized
Beds. In Fluidization. Ed. Grace and Matsen. New York: Plenum.
Borodulya, V. A., Ganzha, V. L. and Kovensky, V I. (1982). Nauka I Technika. Minsk, USSR.
Botterill, J. M. S. (1975). Fluid-Bed Heat Transfer. London: Academic Press.
Botterill, J. M. S., Teoman, Y. and Yiiregir, K. R. (1981). Temperature Effects on the Heat Transfer
Behaviour of Gas Fluidized Beds. AIChE Symp. Ser., 77(208), 330.
Botterill, J. M. S., Teoman, Y and Yuregir, K. R. (1984). Factors Affecting Heat Transfer Between
Gas-Fluidized Beds and Immersed Surfaces. Powder Tech., 39, 177.
Botterill, J. M. S. and Williams, J. R. (1963). The Mechanism of Heat Transfer to Gas-Fluidized
Beds. Trans. Instn. Chem. Engrs., 41, 217.
Brodkey, R. S., Kim, D. S. and Sidner, W. (1991). Fluid to Particle Heat Transfer in a Fluidized Bed
and to Single Particles. Int. J. Heat & Mass Transfer, 34, 2327.
Brewster, M. Q. (1986). Effective Absorptivity and Emissivity of Particulate Medium with
Applications to a Fluidized Bed. Trans. ASME, J. Heat Transfer, 108, 710.
Press,
de Kok, J. J., Stark, N. L. and van Swaaij, W. P. M. (1986). The Influence of Solids Specific
Interfacial Area on Gas-Solid Mass Transfer in Gas Fluidized Beds. In Fluidization V. Ed.
Ostergaard and Sorensen. New York: Engineering Foundation.
Denloye, A. O. O. and Botterill, J. M. S. (1978). Bed to Surface Heat Transfer in a Fluidized Bed of
Large Particles. Powder Tech., 19, 197.
Dry, R. J., Christensen, I. N. and White, C. C. (1987). Gas-Solids Contact Efficiency in a High
Velocity Fluidized Bed. Powder Tech., 52, 243.
Epstein, N. and Grace, J. R. (1997). Spouting of Particulate Solids. In Handbook of Powder Science
and Technology, 2nd ed. Ed. Fayed and Otten. New York: Chapman & Hall.
Flamant, G., Fatah, N. and Flitris, Y. (1992). Wall-to-Bed Heat Transfer in Gas-Solid Fluidized
Beds: Prediction of Heat Transfer Regimes. Powder Tech., 69, 223.
Gabor, J. D. (1970). Wall-to-Bed Heat Transfer in Fluidized and Packed Beds. Chem. Eng. Prog.
Symp.Ser, 66(105), 76.
Gel'Perin, N. I. and Einstein, V. G. (1971). Heat Transfer in Fluidized Beds. In Fluidization. Ed.
Davidson and Harrison. New York: Academic Press.
Gloski, D., Glicksman, L. and Decker, N. (1984). Thermal Resistance at a Surface in Contact with
Fluidized Bed Particles. Int. J. Heat & Mass Transfer, 27, 599.
Grace, J. (1982). Fluidized-Bed Heat Transfer. In Handbook of Multiphase Systems. Ed. G.
Hetsroni. New York: McGraw-Hill.
Haider, P. K. and Basu, P. (1988). Mass Transfer from a Coarse Particle to a Fast Bed of Fine Solids.
AIChE Symp. Ser, 84(262), 58.
Higbie, R. (1935). The Rate of Absorption of a Pure Gas into a Still Liquid During Short Periods of
Exposure. Trans. AIChE, 31, 365.
Jiang, P., Inokuchi, K., Jean, R.-H., Bi, H.-T. and Fan, L.-S. (1991). Ozone Decomposition in a
Catalytic Circulating Fluidized Bed Reactor. In Circulating Fluidized Bed Technology III.
Ed. Basu, Horio, and Hasatani. Oxford: Pergamon Press.
Kato, K., Kubota, H. and Wen, C. Y. (1970). Mass Transfer in Fixed and Fluidized Beds. Inst. Chem.
Eng. Symp. Ser, 66(105), 86.
Knowlton, T. M. (1992). Pressure and Temperature Effects in Fluid-Particle System. In Fluidization
VII. Ed. Potter and Nicklin. New York: Engineering Foundation.
Kovenskii, V. I. (1980). Calculation of Emittance of a Disperse System. /. Eng. Physics, 38, 602.
Kunii, D. and Levenspiel, O. (1968). Bubbling Bed Model for Kinetic Processes in Fluidized Beds.
I&EC Proc. Des. Dev., 7, 481.
Kunii, D. and Levenspiel, O. (1991). Fluidization Engineering, 2nd ed. Boston:
Butterworth-Heinemann.
Kunii, D. and Smith, J. M. (1960). Heat Transfer Characteristics of Porous Rocks. AIChE J.,
6,71.
Kwauk, M., Wang, N., Li, Y, Chen, B. and Shen, Z. (1986). Fast Fluidization at ICM. In Circulating
Fluidized Bed Technology. Ed. P. Basu. Toronto: Pergamon Press.
Levenspiel, O. and Walton, J. S. (1954). Bed-Wall Heat Transfer in Fluidized Systems. Chem. Eng.
Prog. Symp. Ser., 50(9), 1.
Lints, M. C. and Glicksman, L. R. (1993). Parameters Governing Particle-to-Wall Heat Transfer in a
Circulating Fluidized Bed. In Circulating Fluidized Bed Technology IV. Ed. A. A. Avidan.
New York: AIChE Publications.
Littman, H. and Sliva, D. E. (1971). Gas-Particle Heat Transfer Coefficient in Packed Beds at Low
Reynolds Number. In Heat Transfer 1970, Paris-Versailles, CT 1.4. Amsterdam:
Elsevier.
Maskaev, V. K. and Baskakov, A. P. (1974). Features of External Heat Transfer in a Fluidized Bed of
Coarse Particles. Int. Chem. Eng., 14, 80.
Mathur, K. B. and Epstein, N. (1974). Spouted Beds. New York: Academic Press.
Problems 537
Mickley, H. S. and Fairbanks, D. F. (1955). Mechanism of Heat Transfer to Fluidized Beds. AIChE
J., 1, 374.
Mickley, H. S., Fairbanks, D. F. and Hawthorn, R. D. (1961). The Relation Between the Transfer
Coefficient and Thermal Fluctuations in Fluidized Bed Heat Transfer. Chem. Eng. Prog.
Symp. Ser., 57(32), 51.
Ranz, W. E. (1952). Friction and Transfer Coefficients for Single Particles and Packed Beds. Chem.
Eng. Prog., 48, 247.
Rowe, P. N. and Claxton, K. T. (1965). Heat and Mass Transfer from a Single Sphere to a Fluid
Flowing Through an Array. Trans. Instn. Chem. Engrs., 43, T321.
Sleicher, C. A. and Rouse, M. W. (1975). A Convective Correlation for Heat Transfer to Constant
and Variable Property Fluids in Turbulent Pipe Flow. Int. J. Heat & Mass Transfer,
18, 677.
Subbarao, D. and Basu, P. (1986). A Model for Heat Transfer in Circulating Fluidized Beds. Int. J.
Heat & Mass Transfer, 29, 487.
Toor, R. L. and Marchello, J. M. (1958). Film-Penetration Model for Mass and Heat Transfer.
AIChE J., 4, 97.
Tuot, J. and Clift, R. (1973). Heat Transfer Around Single Bubbles in a Two-Dimensional Fluidized
Bed. Chem. Eng. Prog. Symp. Ser., 69(128), 78.
Wen, C. Y. and Fane, A. G. (1982). Fluidized-Bed Mass Transfer. In Handbook of Multiphase
System. Ed. G. Hetsroni. Washington, D.C.: Hemisphere.
Wen, C. Y. and Miller, E. N. (1961). Heat Transfer in Solid-Gas Transport Lines. I&EC, 53, 51.
Wender, L. and Cooper, G. T. (1958). Heat Transfer between Fluidized-Solids, Beds, and Boundary
Surfaces: Correlation of Data. AIChE J., 4, 15.
Wu, R. L., Grace, J. R., Lim, C. J. and Brereton, C. M. H. (1989). Suspension-to-Surface Heat
Transfer in a Circulating Fluidized Bed Combustor. AIChE J., 35, 1685.
Wu, R. L., Grace, J. R. and Lim, C. J. (1990). A Model for Heat Transfer in Circulating Fluidized
Beds. Chem. Eng. ScL, 45, 3389.
Wu, R. L., Lim, C. J., Chaouki, J. and Grace, J. R. (1987). Heat Transfer from a Circulating
Fluidized Bed to Membrane Waterwall Surfaces. AIChE J., 33, 1888.
Xavier, A. M., King, D. F., Davidson, J. F. and Harrison, D. (1980). Surface-Bed Heat Transfer in a
Fluidized Bed at High Pressure. In Fluidization. Ed. Grace and Matsen. New York:
Plenum.
Xavier, A. M. and Davidson, J. F. (1985). Heat Transfer in Fluidized Beds: Convective Heat
Transfer in Fluidized Beds. In Fluidization, 2nd ed. Ed. Davidson, Clift, and Harrison.
London: Academic Press.
Yoshida, K., Kunii, D. and Levenspiel, O. (1969). Heat Transfer Mechanisms Between Wall Surface
and Fluidized Bed. Int. J. Heat & Mass Transfer, 12, 529.
Zabrodsky, S.S. (1963). Heat Transfer Between Solid Particles and a Gas in a Nonuniformly
Aggregated Fluidized Bed. Int. J. Heat & Mass Transfer, 6, 23 and 991.
Problems
12.1 Hot air at 473 K passes through a dense-phase fluidized bed from the bottom. The bed is
maintained at a constant temperature of 273 K. The expanded bed height is 3 m. Deter-
mine the bed-to-gas heat transfer coefficient and gas temperature at different heights of
the bed if the superficial gas velocity is 0.5 m/s. Also consider the following conditions:
a = 0.5, dv = 0.5 mm, kg = 0.026 W/m K, cp = 1,005 J/kg . K.
12.2 The bed-to-surface heat transfer coefficient may vary with the radial position in the bed.
Wender and Cooper (1958) correlated the data of coefficients for heat transfer to immersed
vertical tubes and proposed the following empirical correlation for 0.01 < Re p < 100
/ \ 0.43 / , TrX 0.23 / \ 0.8 / x 0.66
f = 3.51 xl0-CR(l- JCf) (**L\ (L) (*\ (Pm)

where CR is a parameter determined by Fig. PI2.1. Calculate the cross-sectional averaged
heat transfer coefficient for the conditions given in Example 9.2(b). Additional conditions
provided are c = 840 J/kg K, cv = 1,005 J/kg K, Kp = 1.9 W/m. K, K=0.0262 W/m K,
and tc =0.8s. Also, compare the results with those obtained by the approach given in
12.3.2.
Center of vessel Wall of vessel
0.2 0.4 0.6 0.8 1.0

Dimensionless position
Figure P12.1. Value of CR for nonaxial location of immersed vertical tubes (from Wender and
Cooper, 1958).
12.3 Derive the axial profile of the particle convective heat transfer coefficient in a circulating
fluidized bed of fine particles using the information given in 10.4.1. It can be assumed in
the derivation that particles in the bed are all in a cluster form.
12.4 Derive the radial profile of the particle convective heat transfer coefficient in a circulat-
ing fluidized bed of fine particles using the information given in 10.4.2, and simplify the
expression for the center region and the near-wall region, respectively.
12.5 The mass transfer coefficient between a bubble and its cloud due to diffusion, k bc, given
in Eq. (12.82) for a gas-solid fluidized bed can be derived on the basis of the mass
Problems 539
transfer analogy of gas bubble in liquids [Davidson and Harrison, 1963]. Consider gas
circulation inside the bubble and the diffusion of species A from the bubble to the cloud
through the bubble surface to follow the radial gradient of the gas concentration, C A.
The stream function, T/>\ for gas circulation near the bubble surface can be approximated
by
f = RhVsysin0 (P12.2)
where Rb is the bubble radius; y is the distance from the bubble surface in the direction
normal to the surface; 6 is the angle between a radius and the vertical axis of the bubble;
and Vs is the emulsion phase velocity at the bubble surface relative to the rise velocity of
the bubble. Vs is given by
Vs = V2gz (P12.3)
where z is the vertical distance from the nose of the bubble.

(a) Assuming that the diffusion along the circulation streamlines and the convection in
the radial direction are negligible, derive the following equation based on the mass
balance of species A in the bubble
(b) Convert Eq. (PI2.4) into the following equation
(P12.5)
by expressing y in terms of ^ and 0 in terms of <, where <J> is a parameter defined

as
= RlVs sin2 0 (PI 2.6)

b
d6
Specify the boundary conditions for Eq. (P12.5).
(c) Solve Eq. (PI2.5) to obtain the concentration distribution of species A near the bub-
ble surface and express the rate of mass transfer across the frontal surface of the
bubble.
(d) Obtain kt>c for the case of #w = 50, based on the following definition for fcbc
NA =fcbc7rdb2(CA,b- CA,C) (P12.7)
where N& is the mass transfer rate of species A.
12.6 Derive the interchange coefficient for mass transfer between a cloud and an emulsion
(Eq. (12.86)) considering Higbie's penetration model [Higbie, 1935] for mass transfer. The
Higbie penetration model assumes that the exposure time is the same for all the interface
elements moving along the periphery of the bubble.
APPENDIX
Summary of Scalar, Vector, and

Tensor Notations
The major notations of scalars, vectors, and tensors and their operations presented
in the text are summarized in Tables Al through A5. Table Al gives the basic definitions of
vector and second-order tensor. Table A2 describes the basic algebraic operations with vector
and second-order tensor. Tables A3 through A5 present the differential operations with
scalar, vector, and tensor in Cartesian, cylindrical, and spherical coordinates, respectively.
It is noted that in these tables, the product of quantities with the same subscripts, e.g.,
a\bu represents the Einstein summation and <5ij refers to the Kronecker delta. The boldface
symbols represent vectors and tensors.
Table Al. Definitions of Vector

and Second-Order Tensor
Vector
a = a\e\ + a2e2 + a3e3 =
Second-order tensor
Table A2. Algebraic Operations with Vector and Second-Order Tensor
Scalar product of two vectors

a b a\b\ + a2&2 + a^bi = a\b\
Vector product of two vectors
axb= a2 = (a2b3 - (a3bi - a{b3)e2 + {axb2 - a2bi)e3

bx b2 I
Dyadic product of two vectors
ab = a\b\e\e\ -\-a\ + a2bxe2ei + a2b2e2e2 + a2b3e2e3 + a3bxe3ei
+ a3b2e3e2 + a3b3e3e3
540
Appendix / Summary of Scalar, Vector, and Tensor Notations 541
Table A2. (cont.)
Scalar product of two tensors

T'.tJ X\\G\\
Vector product of a vector with a tensor

a - r = (aixn + a2x2i + a3x31)ei + (aixn + ^2^22 + ^3^32)^2 4-
Vector product of a tensor with a vector

r a = (nifli + ri 2 a 2 + xua3)e1 + (r 2 iai + T222 + ^3^3>?2 + (r3ifli + r32a 2 + x33a3)e3
Table A3. Differential Operations in Cartesian Coordinates (x, y, z)
Gradient of a scalar
85 ds ds
Vs = ? x + -e y +
dx dy dz
Divergence of a vector
3Ox 3y 3O,
3JC dy dz
Laplacian of a scalar
Laplacian of a vector
Divergence of a tensor
V - r = ( H H '
(dxxz 9r yz
Gradient of a vector
dax day a z 9fl x 9a y 9a z
~dxxx +
"a*"**** +
ay^ y< 2X H ^v^v H #v^z
dax 8av 9a z a^ av
_| _ _| Ig 0
H ez^z
zx zy
Substantial dz
derivatives8following
Z the gas flow
Dt dt Dt dt
Substantial derivatives following the particle flow
542 Appendix / Summary of Scalar, Vector, and Tensor Notations
Table A4. Differential Operations in Cylindrical Coordinates

(r, 9, z)
ds 1 ds ds
13 1 dae daz
V. = - _ ( , ) + - - + -
l d d s l d s d s
2 _ ( \
t) n 9 r)
2 2 2
r 3^ 3z r 3^ J Cr
2
3 / 1 3 , A 1 9 2 ^ , d2ae , 2 3 r
3r " r 3^ 3z r
13 13r, z 3rz
r 3r r 3^ 3z
daT da$ daz

Va = e veT 4- e^g + - - v
dr dr dr
\ daz
3a r 3a, 3a z
T-^z^r + - r - z ^ + ~r-ezez
dz dz dz
Table A5. Differential Operations in Spherical Coordinates (R,0,(p)
ds I ds 1 ds
1 9 / ,8s\ 1 3 ( ds\ 1
3 / 1 3 -, A + 1 3 f_:_JaR\ , 1 32aR
3/? V/f2 3 ' 7 ^sinfl 30 V 3^ 7 2 sin 2 e
2 3 ... 2 3ay
(asine)
Appendix / Summary of Scalar, Vector, and Tensor Notations 543
Table A5. (cont.)
1 3 1 3 1 d2a9
de
2 3 R 2cot6>
i?2sin6>
_^_3_/ 2daA J__9_/_l 9_

-^7 I ^ r I + jp 9J9 I sin^ 3^
2 daR 2cot0
+
RsiR0d9 dip R
3R 3(26 9a^
V= eReR+ e Ree + eRev
+
U ^ 1 - i C^R+ U ^ + ^' ^8 + - - *'
daK dae
aR ae
+
Index
abrasion Babinet's principle, 147

definition and effects of, 252-254 baffles, in dense-phase fluidized beds, 376, 377, 398
in particle flows, 246 bag filters, 315-319, 320
absorption, cross section for, 143 Bagnold number, 210
acceleration, of particles, 87, 88, 92, 93 barium titanate, permittivity of, 37
acceleration number, 108 Basset, Boussinesq, and Oseen (BBO) equation,
acicular particles, 4 107-108, 198-199, 259, 260
acoustic fluidized beds, 377 Basset coefficient, 108, 123
acoustic waves, 244, 259-265 Basset force, 87, 88-95, 107, 125, 199, 231, 261,
activated carbon, 25, 29 476,481
adsorption carried mass and, 92-95
chemical, 25 bed-to-gas heat transfer, in dense-phase fluidized
in dense-phase fluidized beds, 372 beds, 512-513
physical, 25 bed-to-surface heat transfer, 503, 513-518, 519, 520
aeration, in standpipes, 365, 366 coefficient for, 538
age distribution function, 510 in spouted beds, 527
agglomeration, of particles, 190, 445 bed-to-surface mass transfer, 531
air slides, solids transport by, 461 Beer-Lambert law, 14
Allen's equation, 7 belt conveyors, solids transport by, 461
alloys beryllia, thermal properties of, 35
Biot number, 131, 132, 159, 512, 527
as ferromagnetic materials, 38
blackbody
as poor electric conductors, 33
definition of, 34
yield strength of, 31
monochromatic-emissive power of, 144
alumina, properties of, 30, 31, 35, 37
blowers, in circulating fluidized beds, 423, 437^-38
aluminum, thermal properties of, 35
body forces. See field forces
aluminum alloys, properties of, 30, 31 Boltzmann constant, 114, 116, 171
American Society for Testing and Materials. See boron, elastic properties of, 30
ASTM boron nitride, thermal properties of, 35
Amontons's law of sliding friction, 64, 70, 76 Bose-Einstein statistics, 170
angle of repose. See flow properties of powders Bourger-Beer relation, 152
angular momentum, conservation of, 49 Boussinesq formulation, 176, 194, 198
arc discharge, from particle charges, 87 brass, yield strength of, 31
Archimedes number, 373-374, 379, 409, 427, 532 breakage, of particles, 46, 80
arching, at hopper outlets, 333, 342, 343, 346 brittle erosion, 245-247, 250-251
argon, use in gas adsorption, 13 brittle fracture, of particles, 3, 30-32
arithmetic mean diameter, of paniculate systems, 23 bronze, yield strength of, 31
ASTM Standard sieves, 5, 10, 11 Brunauer-Emmett-Teller (BET) isotherm, 26-28
attrition. See also abrasion; fragmentation Brunauer-Emmett-Teller (BET) method, for surface
mechanisms of, 252-254 area analysis, 13, 26
in particle collisions, 46, 190, 191 bubble column, 371,384
averages, 182-189 bubble Reynolds number, 391
ensemble average, 211 bubbles
intrinsic average, 182-183 breakup of, 375, 377, 389, 390
minimum control volume of phase average, in circulating fluidized beds, 433
183-187 circulation induced by, 506
phase average, 182-183 cloudless, 382, 383
averaging theorems, 182-189 cloud-type, 381, 382, 383,417
axial dispersion coefficient, 529 coalescence of, 375, 376, 387, 388-389, 390, 401
axial distribution, of heat transfer coefficient, 524, configurations of, 382-385
538 in dense-phase fluidized bed, 375, 376, 377,
axial flow cyclone, 299 385-389, 390, 401, 417, 513-514
545
546 Index
bubbles (cont.) charge transfer, 111-123

effect on heat transfer, 500, 501, 510, 513-514, basic mechanism of, 114
515 by collision, 119-123
effect on mass transfer, 538-539 momentum transfer and, 87-129
flow around, 385-388 by single impact, 119-120
formation of, 381 charge transfer coefficient, 119
fraction of, 396, 515 choking
frequency of, 515 in circulating fluidized beds, 425, 428, 432, 460
invisible gas flow through, 393 in pneumatic conveying, 465
jet formation from, 388, 389-392 circulating fluidized beds (CFB), 421^60
onset of bubbling, 381-382 applications of, 421
phase flow of, 392 axial profile of voidage in, 438^-40, 459
pressure distribution around, 387 blowers in, 423, 437-438
rise velocity of, 374, 381, 383, 389-392, 396, choking in, 425, 428, 432, 460
399,418,539 clusters and clustering in, 425, 445-446, 447,
size of, 373, 375-376, 381, 390-391, 392, 396 453,522, 532
stability of, 390 compressors in, 423, 437-438
wake angle of, 383-384 computational fluid dynamics for, 451-453
wake of, 381, 383, 385, 387, 514 core-annular flow model for, 448, 459
bubble-to-cloud interchange coefficient, 529, 531 core-annular flow structure in, 446, 447
bubble-to-cloud mass transfer, in dense-phase cyclones in, 422, 431
fluidized beds, 529-531, 538 dense-phase fluidization regime in, 423, 424, 425,
bubbling fluidization, in dense-phase fluidized beds, 427
373,374,375,381-396 deposition coefficient for, 449, 450
bucket elevators, solids transport by, 461 dilute transport regime of, 421, 443
Buckingham pi theorem, 230 downcomer in, 421, 422-423, 431, 432
buoyancy force, 93, 107 entrainment in, 440, 448, 449
Burke and Plummer's theory, 166, 226, 228-230
entrance geometry of, 440, 441
exit geometry of, 440, 441
cake filters, 314, 315,316
fast fluidization regime in, 423, 424, 425, 429,
calcium hydroxide, specific surface area of, 27-28
447^53
capillaric model, 224-225, 226
flow regimes and transitions in, 423-438
carbide (sintered), properties of, 30, 35
gas kinetic theory applied to, 452
Carman correlation, 282, 283
granular flow in, 452
carried mass, Basset force and, 92-95, 108
carried mass coefficient, 108 granular temperature in, 452
cascade impactor, particle sizing using, 17 heat transfer coefficient in, 524-525
catalytic cracking heat transfer in, 521-526
circulating fluidized bed use in, 421 hydrodynamics of, 422, 434, 440^42, 448, 451
dense-phase fluidized bed use in, 381 intermittency index for, 445447
standpipe use in, 350, 351 intermittent solids flow in, 445-447
cement clinker, mass distribution of, 21 internals for, 421
center of compression, 53-56 J-valve in, 423
centrifugal fluidized beds, 378, 415-416 L-valvein,423,431,435
centrifugal sedimentation, particle sizing by, 10, 12 mass transfer in, 532
centrifugal separators, 298, 300 maximum gas velocities in, 428-429
centrifugation, gas-solid separation by, 297 maximum stable bubble size in, 460
ceramics minimum gas velocities in, 428429, 437
as elastic-brittle particles, 29 nonmechanical valves in, 422,423, 429
pipe walls of, 244-245 operating conditions for, 421, 525526
yield strength of, 31 particle-particle collision in, 444, 452, 525
channeling, in dense-phase fluidized beds, 374, 375, particle turbulent diffusion coefficient for, 449
376 particle turbulent diffusion in, 448
charge generation, 111-123 particle-wall collisions in, 444, 525
corona charging, 115, 117-118 pressure balance in, 429-430
diffusion charging, 115, 116-117 pressure drop in, 423, 425,430-431, 433
field charging, 115-116 radial profile of voidage in, 440-442, 524
by ion collection, 115-119 reactor applications of, 421
in ionized gases, 117-118 risers in, 421, 422,423, 430-431, 434-436, 440,
saturation charge, 116 442,448
by surface contact, 111-115 seal pots in, 423
by thermionic emission, 119 solids circulation in, 524, 525
Index 547
solids flow control devices in, 422-423, 431, compressors, in circulating fluidized beds, 423,
433^37 437-438
solids flow structure in, 443-^47 conical hoppers
solids holdup in, 442^-43 design diagram for, 345
solids mixing in, 443-453 stress analysis of, 339
S-shaped profiles of voidage in, 439, 440 conservation of angular momentum, 49, 266
suspension-to-wall heat transfer in, 521-522 contact
system configuration in, 422-423 axisymmetric, 51, 52
tangential gas injection in, 422 elastic, 49-50
transport velocity in, 425,459 frictional, 63-71,76
two-phase flow models applied to, 451-^53 frictionless, 59-63
voidage in, 438-442, 445 inelastic, 49, 78-83
V-valve in, 423 nonsliding, 66-69
wall region of, 423, 424, 442, 444 normal, 59-63
wavy solids layer in, 444-445 oblique, 46, 63-64
cloud-to-emulsion interchange coefficient, 529, 530, plastic, 78-80
531,539 quasi-static, 51
clusters and clustering torsionless, 51
in circulating fluidized beds, 425,445-446,447, contact efficiency, effect on gas-solid mass transfer,
453, 522, 532 532
in dense-phase fluidized beds, 401, 521522, contact potential, 114
523-524 contact theories, collisional mechanics and, 46
role in heat transfer, 521-522, 523-524 continuity equation, 169, 175, 179, 195, 214
coal continuity waves
as elastic-brittle solids, 29 in dense-phase fluidized beds, 382
Rosin-Rammler distribution, 20, 21 generation of, 280-282
coal combustion continuum modeling. See also Eulerian continuum
circulating fluidized bed use in, 421 approach
dense-phase fluidized bed use in, 372 of multiphase flows, 164, 182-205
heat transfer in, 130,499, 517 convection
particle changes in, 253 heat transfer by, 130, 138-142
particle tracking in, 166 mass transfer by, 156-157
Sauter's diameter applied to, 6 copper
cobalt, as ferromagnetic material, 38 as isotropic material, 29
coefficient of diffusion, 198 thermal properties of, 33, 35
coefficient of kinetic friction, 64, 75, 76, 248, 252 copper alloys, elastic properties of, 30
coefficient of restitution, 47, 48, 78, 80-83 copper ball, restitution coefficient of, 82-83
coefficient of self-diffusion, 173-174 core-annular flow model, 448, 459
coefficient of thermal expansion, of particles, 32-33, core (funnel) flow hoppers, 333, 334
35 corona charging, 115, 117-118
coefficient of viscosity, 173 Coulomb force, 119,310
collinear impact, of spheres, 47-48 Coulomb powders, 336-337, 347
collision Coulomb's law, 104
charge transfer by, 119-123 Coulter principle, particle sizing using, 17
due to wake attraction, 93-95, 128 cracking mechanism, 246
of elastic spheres, 49-50, 72-78, 133-138 crack initiation and growth, 32
frequency of, 46, 172-173 creeping flows. See Stokes flows
heat conduction in, 133-138 crystalline particles, 4
of inelastic spheres, 78-83, 248 cyclones, 297. See also rotary flow dust separators;
interparticle kinetic theory modeling of, 164, 166, tangential flow cyclones
197 in circulating fluidized beds, 422, 431
interparticle particle viscosity due to, 202-204 collection efficiency of, 303-309
kinetic theory applied to, 210 cut-off size of, 307-308
speed of, 249 in dense-phase fluidized beds, 376
collisional deformation, of spheres, 49 flow field in, 300-303
collisional force, 49, 73, 87, 108, 203 geometric configuration of, 304, 305
erosion from, 248 principles of, 297-309
collisional heat transfer coefficient, 136, 137 reverse-flow cyclone, 298
collisional pair distribution function, 215-217 uniform cyclone, 298
collision mechanics, of solids, 46-86 cyclone scrubbers, 323, 325
compatibility requirement, 341 cylindrical column, stresses in, 338-339
548 Index
Darcy's law, 164, 166, 222-224, 226, 386, 419 two-phase theory of fluidization applied to, 375,
Davidson-Harrison model, for flow around a bubble 393,405, 420
in fluidized beds, 385, 387,417,460, 529, 539 vibrofluidized beds, 377
Davies and Taylor equation, 391 wall effect in, 403,418
DeBroucker's mean diameter, 24,40, 379 dense-phase transport modeling, 211-213
Debye shielding distance, 117 density, of gas-solid mixtures, 244, 254-257
deformation density functions, 18
elastic, 72-78 deposition coefficient, 449,450
ofparticles,3,24,28-32,46 depth filters, 297, 314-325
plastic. See plastic deformation deterministic trajectory method, for multiphase flow,
dendritic particles, 4 165, 205, 206-208
dense-phase fluidized beds, 371-420,421 detonation combustion, acoustic waves from, 259
acoustic fluidized beds, 377 Deutsch equation, 312, 313
applications of, 371, 372 diamagnetic solids, 37, 38, 39
baffles in, 376, 377, 398, 399 dielectric materials, electrical properties of, 37, 105
bed-collapsing technique applied to, 418 diffraction
bed expansion in, 378, 380, 392-396 effect on thermal radiation, 143
bed voidage in, 371, 388, 398, 399^00,408, 419 effects on scattering, 150
bubble behavior in. See bubbles from large spheres, 147148
bubbling fluidization in, 374, 375, 381-396 diffusion
centrifugal fluidized beds, 378 heat transfer by. See thermal diffusivity
channeling in, 374, 375, 376 mass transfer by, 156-157
components of, 376-378 diffusion charging, 115, 116-117
convective heat transfer coefficient of, 518 diffusivity, thermal. See thermal diffusivity
cyclones in, 376 dilute transport, 374, 421, 522
diplegs in, 376, 377 dimensional analysis, similarity and, 230-235
diplegs, in dense-phase fluidized beds, 376, 377
elutriation in, 400-402
direct numerical simulation (DNS), 174, 288
emulsion phase in, 392-393, 394-395
discharge flow, 364
entrainment in, 371,400-402
discrete-vortex simulation (DVS), 174, 288
flow around bubbles in, 385-388
dispersion forces, 25
fluidization regimes in, 374-376
displacement
force fields for, 377
from boundary point forces, 58
freeboard of, 376, 377, 384, 400-401, 440
radial displacement, 58
gas flow division in, 392-396
vertical displacement, 58, 60
heat exchangers in, 376
displacement-strain relationships, 52, 58
heat transfer in, 501, 508, 512-518 Doppler effect, 16
hydrodynamics of, 373, 399^00, 418 downcomer, in circulating fluidized beds, 421,
instability of, 381 422^23,431, 432,433^37
internals in, 376, 398, 399 drag coefficient, 7,40, 88, 94, 108, 123, 268, 287,
interparticle forces in, 372 484,494
jetting in, 381,388-389 drag force, 87-88, 96, 190, 247, 261, 267, 276-280,
lean-phase fluidization in, 371, 374, 396, 397 318,353,464,477,478
magnetofluidized beds, 378 drag reduction, 469-472
mass transfer in, 527-532 drying, in dense-phase fluidized beds, 372
minimum bubbling velocity in, 382, 416 ductile erosion, 24^-247, 249
minimum fluidization in, 378-380, 385 ductile fracture, of particles, 3, 30-32
minimum fluidization velocity in, 378, 395,407 dust explosion, from particle charges, 87
particle classifications in, 371-376 dust removal, 297
particle interaction in, 381 dynamic diameter, of particles, 6-9, 13
paniculate fluidization in, 373, 374-375, 380-381 dynamic waves, 282-284
reactor applications, 371 in dense-phase fluidized beds, 382, 417
regime classifications in, 371-376
regime transition in, 396-398 efficiency factors
slugging in, 374, 375-376,403-406 for absorption, 143
solids ejection in, 401 for extinction, 143, 144
solids mixing in, 371, 381, 384 for scattering, 143, 144, 151
spouting in, 372, 374, 376, 388,406-409 Einstein summation, 175, 219, 540
transition velocity in, 398 elastic-brittle particles, 29
transport disengagement height (TDH), 401, 402, elastic collision, 46
403 elastic constants, 29
turbulent fluidization in, 374, 375, 396-^00 elastic contact theory, of solids, 4950
Index 549
elastic deformation, 72-78, 245 Eulerian integral length scale, 285

ofparticles,3,28,29-30,31
elasticity, 50 fabric filters, 318
restoration of, 80 collection efficiency of, 319-320
elasticity modulus, 382 diffusion collection in, 319
elastic-plastic impact model, 80, 81, 83 inertial impaction collection in, 319
elastomeric deformation, 28 interception collection in, 319-320
elastoplastic particles, 29-30, 31 materials used in, 315
electrical conductivity, of particles, 35-36 far-field diffraction. See Fraunhofer diffraction
electric force, 101,105 fast fluidization regime, 421
electric porcelain, yield strength of, 31 in circulating fluidized beds, 423,424,425, 429,
electric wind, in electrostatic precipitation, 297, 432-438, 447^53
311-312 gas velocities for, 428-429
electrification. See charge generation mass transfer in, 532
electron microscopy, of particles, 6, 10 operating constraints of, 432438
electrophoresis, 108 fatigue, of particles, 30-32
electrostatic ball probe theory, 111, 119, 121-123 fatigue cracking, 245
electrostatic forces, 25,47, 87, 104, 107,108, 111, Faxen effect, 476
307, 308 Feret's diameter, 5-6
electrostatic precipitation and precipitators, 105 Fermi-Dirac distribution function, 114
collection efficiency in, 312-314 Fermi energy level, 111, 114, 115
cylinder-type precipitators, 311 ferrimagnetic and ferroelectric materials, 37, 39
gas-solid separation by, 297, 309-314, 315 fibrous particles, 4
migration velocity and electric wind in, 311-312 Fick's equation, 156, 158
particle charging in, 310 field charging, 115-116
particle discharging in, 310 field forces, 105-107. See also electric force;
plate-type precipitators, 311 gravitational force; magnetic force
principles of, 310-311 film model, for heat transfer, 501, 502-503
elutriation, in dense-phase fluidized beds, 400402 film-penetration model, for heat transfer, 508-510
elutriators, in settling chambers, 321, 322 filtration and filters
emulsion phase flow, 392, 394, 395 cake filters, 314, 315
emulsion phase/packet model, for heat transfer, 501, collection efficiency of, 319-320
503,506-511 fabric filters, 315, 318, 319-320
energy equation, 169-170, 196, 207-208, 213, mechanisms and types of, 314-315
214-215, 268 pressure drops in, 314, 315-319
entrainment Fischer-Tropsch synthesis, circulating fluidized bed
in dense-phase fluidized beds, 371,400-402,440 use in, 421
of fluid, 97 fixed beds, 371, 374, 379
model for, 448, 451 flaky particles, 4
-equation, 178, 201 flowability of powders, 337, 342-346
equation of equilibrium, 50, 51, 52, 341 arching effects on, 333, 342, 343, 346
equation of motion, 106, 108-110, 232, 481 Jenike's model for, 342
equation of state, for gas-solid mixtures, 244, particle shape effects on, 3
254-257 piping effects on, 333, 342
equilibrium speed of sound, in gas-solid mixtures, segregation effects on, 342
265 flow properties of powders
equivalent diameters, of nonspherical particles, 410 angle of fall, 342
Ergun equation, 164, 225-230, 234, 379 angle of repose, 342
application to suspensions, 167 compressibility, 342
flow-rate effects, 166, 225 dispersibility, 342
fluid-property effects on, 225-226 fluid catalytic cracking (FCC) particles, in fluidized
particle-property effects on, 229-230 beds, 381, 397,421, 426, 428,434,440, 443,
voidage of packing effects on, 226 459
erosion. See also flow properties of powders fluid element, of continuum, 164
abrasive, 246, 252-254 fluidization
brittle, 245-247 regimes of, in dense-phase fluidized beds,
ductile, 245-247 374-376
locations of, 247-252 two-phase theory of, 381
particle properties affecting, 3 fluidized beds. See also circulating fluidized beds;
of wall surfaces, 46, 244-254 dense-phase fluidized beds; lean-phase
Eulerian continuum approach, 164-165, 182, 183, fluidization
206, 323 bubbles in, 510-511. See also bubbles
550 Index
fluidized beds (cont.) glass beads, elastic displacement in, 62-63

dense-phase type. See dense-phase fluidized beds glass-ceramics, elastic properties of, 30
Ergun equation applied to, 164 glass powder, mass distribution of, 21
flow around bubbles in, 385 glasses
gas-particle heat transfer in, 499 index of refraction, 40
heat transfer coefficients in, 503 thermal properties of, 35
heat transfer in, 136,499-532 gliding, of particles, 249, 251
mass transfer in, 499-532 gradient-related forces, 95-97. See also pressure
particle interaction in, 256 gradient force; radiometric force; Saffman
particle properties affecting, 3 force
pressure drops in, 364 grain inertia flows, kinetic theory modeling of, 210
scaling relationships for, 232-235 grains, pneumatic transport of, 461, 498
speed of sound in, 263 granular materials
stability of, 270 flow of, 220
suspension-surface heat transfer in, 499, 500 hopper flows of, 333
temperature control in, 499 granular particles, 4
flux equation, 157 granular temperature, 222
fountain height, in spouted beds, 408 graphite
Fourier equation, 156, 503 properties, 30,31,35
Fourier number, 131, 132, 134, 135, 137, 142, 159, in suspension coolants, heat transfer by, 130
236, 506, 508 Grashof number, 158, 159
Fourier's law, 170 gravitational force, 92, 101,105,107, 108
fracture, of particles, 3, 25, 30-32 gravitational sedimentation, particle sizing by, 10,
fragmentation, definition and effects of, 252-254 12
Fraunhofer diffraction, particle sizing by, 10, 13-15, gravity chutes, solids transport by, 461
16,18 gravity settling and settlers
freeboard, in dense-phase fluidized beds, 376, 384 collection efficiency of, 321
Freundlich isotherm, 25 gas-solid separation by, 297, 315
friction, in powder flows, 341, 342 particle residence time in, 321, 323
frictional force, 64 particle settling time in, 323
Froessling equation, 158, 532 pick-up velocity in, 321
Froude number, 231, 235, 236, 380,409,484,485,
487 Haarvon Karman hypothesis, 347
frozen speed of sound, in gas-solid mixtures, 265 Hagen-Poiseuille equation, 223, 224-225, 226
Hamaker's constant, 102-103, 123, 128
gas adsorption, use in particle sizing, 10, 13 heat conduction, 3. See also thermal diffusivity
gas convection, heat transfer by, 500, 501 heat diffusion equation, 506
gas convective component, in heat transfer, 514, heat exchanger
516,522-523 in dense-phase fluidized beds, 371, 372, 376, 377
gas distributors, for dense-phase fluidized beds, 376 downcomer use as, 423
gas kinetic theory, granular flow and, 452 heat loss, by colliding spheres, 46, 80, 121
gas-solid mixture momentum equation, 451 heat transfer, 130-163,499-539
gas-solid separation, 297-332 in annulus region, 526-527
in circulating fluidized beds, 422 bed-to-gas type, 512-513
by electrostatic precipitation, 297, 309-314 bed-to-surface type, 503, 513-518, 519, 520
by filtration, 297, 314-320 in circulating fluidized beds, 521-526
by gravity settling, 297, 320-323 in collisions, 133-138
by rotating flow, 297-309 by convection, 130, 138-142, 500
by wet scrubbing, 297, 321, 323-324 in dense-phase fluidized beds, 501, 508, 512-518
gas-to-emulsion phase interchange coefficient, 531 emulsion phase/packet model for, 501, 503,
gas-to-particle heat transfer 506-511
in fluidized beds, 499, 526-527 film model for, 501, 502-503
in spouted beds, 526-527 film-penetration model for, 508-510
gas-to-particle mass transfer, in dense-phase gas convective component in, 514, 516, 522-523
fluidized beds, 527-529 gas-to-particle type, 526-527
gas turbulent diffusion coefficient, 453 mass transfer compared to, 130, 157-159
Gaussian distribution, 19, 21, 209, 210 modes and regimes of, 500-501
Gaussian quadrature formula, scattering integral particle convective component in, 514-516, 518,
approximation by, 153-154 522
Gauss's law, 112,115 by radiation, 499, 500, 523-524
Gauss theorem, 168, 189, 272,468 single-particle model for, 501, 503
Geldart's powder classification, 372, 427 of single spheres, 131-133
Index 551
in spouted beds, 526-527 inelastic contact, of solids, 49

by thermal conduction, 503 instability analysis, 244, 270-288
unsteady behavior of, 502 of batch-solid fluidization system, 284-285
heat transfer coefficient(s), 136, 137, 148, 268,499 intermittency index, for circulating fluidized beds,
axial distribution of, 524-525 445^47
for dense-phase fluidized beds, 512 internal energy, of gassolid mixtures, 244, 257258
in emulsion phase/packet model, 507 internals, in fluidized beds, 376, 398, 399, 412, 520
in film model, 502 intraparticle heat transfer resistance, 527
gas velocity effects on, 501 Invar, thermal properties of, 35
of particle convection, 514-515, 522 ion collection, charging by, 115
radial distribution of, 524-525 irregular particles, 4
in turbulent regime, 521 isentropic change of state, in gas-solid mixtures,
heat transfer probes, 514, 525 258-259
Hertzian contact theory, 46, 49, 50, 59-63, 64, 72,
74, 76, 78-79, 81, 133, 246, 248 Jacobian determinant, 168, 236
Higbie's penetration model, 509, 530, 539 Janssen's model, for stress distribution, 333,
Hinze's model, 285 338-339,419
Hinze-Tchen model, 165, 197-201,470 Jenike's model, for powder flowability, 342
Hooke's law, 29, 51,58, 334, 341. See also elasticity Jenike translatory shear cell, for powder flowability,
hoppers and hopper flows, 333-350 341,342
aerating jets for, 333 jetting
core (funnel) flow hoppers, 333, 334 in dense-phase fluidized beds, 376, 381, 388-389
flow patterns in, 333 jet formation in, 388-389
flow-promoting devices for, 333 jet size in, 389-392
hopper-standpipe-discharge flow, 354-357 J-valve
mass flow hoppers, 333, 334, 345 in circulating fluidized beds, 423
moving bed flows in, 346-350 in standpipe systems, 365
powder flowability in. See flowability of powders
powder mechanics in, 333-346 Kelvin effect, 44
stress distribution in. See stress distribution k-e-kp model, 165, 197, 201-202
stress failure in. See stress failure k-e model, for turbulent flow, 156, 165, 175, 176,
vibrators for, 333 177-178,179,180,201
horizontal flow settling chamber, 321, 322 ^-equation, 178, 181,201
Howard settling chamber, 321, 322 kinetic energy, loss due to collisions, 46, 48, 76, 80,
hydrodynamics 259
of circulating fluidized beds, 422 kinetic theory, transport coefficients and, 170-174,
of dense-phase fluidized beds, 372 197,200-201
fundamental equations for, 213, 214, 215 kinetic theory modeling
hydrogenation, heat transfer in, 130 collisional flux of fluctuation energy, 219, 222
collisional pair distribution function, 215-217
impact. See also collision collisional stress tensor, 217, 219, 220, 221
collinear, 47-48 for collision-dominated dense suspensions,
damage due to, 245 210-222
directional, 244, 251 constitutive relations, 217-222
duration of, 46, 72 dense-phase transport modeling, 211-213
dynamic, 46 energy dissipation rate, 219, 221
of elastic bodies, 46, 72 hydrodynamic equations, 213-215
force of, 46 for interparticle collisions, 164, 166
ofparticles,244,245,247 Kirchoff 's law, 34, 152
planar, 48 Knudsen diffusion, 156
random, 244 Knudsen number, 464, 488,494
single, charge transfer by, 119-120 Kolmogorov dissipative scale, 447
stereomechanical, 46,47-49 Kozeny theory, 166, 226-228
velocity of, 48, 72, 81, 82 Kronecker delta, 169, 239, 540
impact angle, 252 krypton, use in gas adsorption, 13
impact number, 249 Kutta-Joukowsky lift formula, 100
impulse, definition of, 46
impulse-momentum law, 46 Lagrangian correlation coefficient, 198, 200, 237
incident radiation, on dielectric spheres, 146, 147 Lagrangian trajectory model, 107, 164, 165166,
inclined standpipe, 364 167, 205-210, 323
index of refraction, 14, 40 deterministic trajectory method, 165
inelastic collision, 46 stochastic trajectory method, 165
552 Index
laminar flow, description of, 165 Mie scattering theory, 13, 16, 143, 144-148
laminar transport coefficients, 173, 179 migration velocity, in electrostatic precipitation,
Langevin model, 210 311-312
Langmuir isotherm, 25-26 Mindlin's theory for frictional contact, 46, 49, 50,
large-eddy simulation (LES), 174, 288 64,74
laser Doppler phase shift, particle sizing by, 1617 minimum fluidization, 378380, 385
laser Doppler velocimetry (LDV) measurements, on mixing length model, for turbulent flows, 165, 167,
cyclones, 302 175,176-177
lean-phase fluidization, 371, 374, 396, 397, 421 modular particles, 4
Lifshitz-van der Waals constant, 102-103, 124 Mohr circle, for plane stresses, 334-335, 337, 340,
lift force, 95,100,101 341,342,344
lifting, of particles, 251 Mohr-Coulomb failure criterion, for granular
light wavelength, 40 materials, 336-337, 342, 347
log-normal distribution, 19, 20, 21 molecular diffusion coefficient, of gas, 530
London's constant, 101, 126 molecular diffusivity, 157
loopseal, in standpipe systems, 365 momentum equation, 169, 175, 179, 195, 207, 214,
Lorentz constant, 33,41 271,317
Lorentz force, 105, 124 monochromatic emissivity, of particles, 34, 41
Lorenz-Mie theory, 14 Monte Carlo method, 153, 165, 206, 209, 323
Loschmidt number, 255 Moody chart for pipe friction, 470,471, 472
Love's stress function, 52, 85 moon dust, Rosin-Rammler distribution of, 20-21
L-valve multiple-tray settling chamber, 321
in circulating fluidized beds, 423,431-432,435
in standpipe systems, 365, 366 Navier-Stokes equations, 89, 97, 107, 109, 165,
198,301,316
macropores, 13 time-averaged, 174, 175-176
magnetic force, 101, 105, 108 Newton's equation for drag coefficient, 7, 88
magnetic properties, of solids, 3, 37-38 Newton's law of acceleration, 35, 203
magnetofluidized beds, 105, 378,416,417 nickel, as ferromagnetic material, 38
Magnus effect, 97,476 nonmechanical valves, 365, 366
Magnus force, 87, 97-101, 108, 124, 190 nonspherical particles, equivalent diameters, 4-10
maleic anhydride production, circulating fluidized Nusselt number, 132, 133, 138, 139, 141, 157, 159,
bed use in, 421 268,512,513,526,531,533
Martin's diameter, 5-6,40 nylon, properties of, 30, 31, 37
mass flow hoppers, 333, 334, 345
mass interchange coefficient, 529, 539 oblique, sliding contact, 6466
mass transfer, 156-159 optical microscopy, in particle sizing, 6, 10, 11
bubble-to-cloud type, 529-531 optical properties, of solids, 3, 39-^-0
in circulating fluidized beds, 532 oscillation, of spherical particles, 259-261
in dense-phase fluidized beds, 527-531 Oseen approximation, 98, 99
by diffusion and convection, 156-157 Oseen drag, 97, 100
in gas-solid systems, 527-532 overflow weir, in dense-phase fluidized beds, 419
gas-to-particle type, 527-529
heat transfer compared to, 130, 157-159 packed beds. See also fixed beds
in multiphase systems, 527-532 equations for flows through, 222-230
mass transfer boundary layer, 531 packet resistance, for heat conduction, 508
mass transfer coefficient, 159,499, 527, 528, 530 packets, clusters as, 522
material densities, 38-39 paramagnetic solids, 37, 38, 39
Maxwell-Boltzmann velocity distribution, 165, particle(s)
170-172,210,215,216 acceleration of, 87, 88, 92, 93
transport coefficients based on, 174 agglomeration of, 46
mesopores, 13 attrition of, 46
metals circulation of, effects on heat transfer, 501, 513,
electrical properties of, 35-36 514
yield strength of, 31 deformation of. See deformation
mica, properties of, 31, 37 in dense-phase fluidized beds, 371-376
MickleyFairbanks model for heat transfer, 506508 density, 38
microorganism cultivation, dense-phase fluidized diameter of, 255
bed use in, 372 eddy interaction with, 287
micropores, 13 electrification of, 46
microscopy, in particle sizing, 10, 11-12 fluid interactions of, 87101
microslip contact, tangential displacement from, 68 forces controlling motions of, 87
Index 553
Geldart's classification of, 372 acceleration length in, 472-473

interactions of, 87, 269, 381,452,464,465, 513 advantages and disadvantages of, 461
interparticle forces affecting, 101-107, 373 choking in, 465
intraparticle heat transfer, 499 classifications of, 461
motion of single, 87, 107-110 critical transport velocities in, 474-475
number density, 17, 18, 20, 23, 173, 185, 203, dilute suspension flow in, 464, 466,467
206,211 dilute transport regime in, 421
rotation of, 87, 97 drag reduction in, 469^472
shape effect, 3 flow characteristics in, 464-466
size and properties of, 3-45 flow regimes and transitions in, 464-466
sizing methods for, 10-17 fully developed dilute pipe flows in, 482^-93
trajectories of, 109-111 horizontal transport in, 461, 466, 468, 474
turbulent kinetic energy of, 285 inclination angle effects on, 477
viscosity from interparticle collisions, 202-204 minimum transport velocity in, 474, 475^76,
wall collisions of, 76-78, 82, 87, 205 498
particle clouds, thermal radiation with, 130, momentum integral equations for, 479-480
150-156 moving bed flow in, 465^-66
particle convective heat transfer, 501, 503, 514-516, negative- and positive-pressure systems in, 461
518,522 operational modes of, 461,462-4-63
particle convective heat transfer coefficient, axial particle sedimentation in, 464, 465,466
and radial profiles of, 538 particle-wall forces in, 474, 476,477
particle Reynolds number, 87, 125, 149, 160, 237, pick-up velocity in, 474, 476-478
268, 350, 379, 410,453, 534 pipe-bend flows in, 467, 473, 474, 478-482
particle terminal velocity, 7-8 plug flow in, 464, 466
particle-to-gas heat transfer, in dense-phase pressure drop in, 466-472
fluidized beds, 512-513 saltation in, 465,474
particle-to-gas mass transfer coefficient, 527 shedding layer in, 479-^80, 498
particle-turbulence interaction, 244, 285-288 single-phase flow in curved pipes in, 478-481
paniculate fluidization, in dense-phase fluidized slip velocity of particles in, 488
beds, 373, 374-375 stratified flow in, 464, 466
paniculate fluidized beds, heat transfer in, 501, 510 temperature distributions of phases in, 489-493
Peclet number, 139, 181 variables affecting, 461
permittivity, of particles, 36-37, 105 vertical transport in, 461, 466,468
perturbation method, for fluidized-bed instability wall region of, 469, 470
studies, 270 Poisson contraction, 29
phase average, 182-183 Poisson distribution, 172
Phase Doppler Particle Analyzer (PDPA), 17 Poisson equation, 115,483
phonons, heat conduction by, 33 Poisson's ratio, 29, 30, 42, 51, 63, 79, 84, 246, 291,
photophoresis, 96, 108 335,368
physical adsorption, of particles, 3, 25 poly dispersed particle systems, density functions
pick-up velocity affecting, 3
in gravity settling chambers, 321 polyethylene
in horizontal pneumatic conveying, 474, 476-478 elastic properties of, 30, 76, 79
piezoelectric materials, properties and uses of, 37 permittivity of, 37
pipes, stratified, wave motion in, 270-280 thermal properties of, 35
pipe walls polymerization, heat transfer in, 130, 499
erosion of, 244-254 polymers
materials used in, 244-245 in triboelectric series, 113-114
planar impact, of spheres, 48-49 use in fabric filters, 315
Planck intensity function, 152 yield strength of, 31
Planck's formula, 34, 142 polystyrene
plastic deformation, 121, 245 plastic deformation of, 29
onset of, 78-80 yield strength of, 31
of particles, 3, 28, 29-30, 47, 49, 80, 81 polyvinyl chloride, permittivity of, 37
plastics porcelain, permittivity of, 37
fatigue wear of, 246 pores, size classification of, 13
pipe walls of, 244-245 powder(s). See also Coulomb powders
Plexiglas, index of refraction, 40 flowability of. See flowability of powders
plugs flow properties of, 342
in pneumatic conveying, 464,466 motion onset in. See stress failure
square-nosed, 404 shear testing of, 341, 342
pneumatic conveying methods, 448, 461^498 sieve classification of, 10
554 Index
powder mechanics Richardson-Zaki correlation, 352, 367, 370, 448

angle of friction, 342 risers
bulk density, 342 circular, model for, 448
cohesion, 342 in circulating fluidized beds, 421, 422, 423, 429,
in hopper flows, 333-346 434-436,440,442
major principal stress, 342 unstable operation, 428,432, 433, 437
unconfined yield strength, 342 Rosin-Rammler distribution, 19, 20-21, 22-23,41
Prandtl number, 138, 139, 157, 158, 181,492,494, rotary flow dust separators, 297
531,534 axial flow cyclones, 299
pressure, of gas-solid mixtures, 244, 254-257 mechanism and types of, 297-300
pressure diffusion, 156 tangential flow cyclones, 297, 298, 302, 303-305
pressure distribution, 58 tornado dust collectors, 299
in frictionless contact, 61, 69 rubber
pressure drops abrasive wear of, 246
in circulating fluidized beds, 423, 425, 433 thermal properties of, 35
in fluidized beds, 364, 378-379
pressure gradient, force due to, 95, 96, 107 Saffman force, 87, 95-97, 108, 124, 125, 128, 190,
pressure waves. See also acoustic waves; shock 251,476
waves saltation, in pneumatic conveying, 465, 474
through gas-solid mixtures, 244, 259-269 saturation charge, 116
projected area diameter, 5-6 Sauter's diameter, 6, 13, 23, 24
pseudocontinuum model, 141-142, 150, 186, 338, scaling relationships
350 for fluidized beds, 232-235
instability analysis using, 244 for pneumatic transport, 230-231
PVC-acrylic alloy, yield strength of, 31 scanning electron microscopy (SEM), for particle
sizing, 10, 11, 12
radiant heating, of particles, 148-150 scattering
radiant heat transfer coefficient, 518 cross section for, 143
radiation, gas absorption of, 130 diffraction effects on, 150
radiation transport equation, 151-153 efficiency factor for, 143, 144
radiative heat transfer, 499, 500, 514, 517-518, geometric scattering, 143
523-524. See also thermal radiation independent vs. dependent scattering, 150-151
radiative heat transfer coefficient, 523 from large spheres, 146-148
radiometric force, 96 Mie scattering theory, 143, 144-148
random surface renewal, heat transfer and, 509 by multiple particles, 150
Rankine's combined vortex model, 300-301 by particle clouds, 155
rare earth elements, as ferromagnetic materials, 38 ofparticles,249,251
Rayleigh number, 158, 160 radiative transport by, 143
Rayleigh scattering, 14, 143, 144 Rayleigh scattering, 14, 143
reactors by single particles, 143-148, 150
circulating fluidized bed use in, 421 scattering integral, approximation of, 153-154
dense-phase fluidized bed use in, 371, 419 Schmidt number, 157, 158, 160, 534
reflection Schrodinger's wave equation, 165, 171
from diffuse spheres, 147 screw conveyors, solids transport by, 461
effect on thermal radiation, 143, 150 seal pots, in circulating fluidized beds, 423
from specularly reflecting spheres, 146-147 Sedigraph particle analyzer, 8
refraction, effect on thermal radiation, 143, 150 sedimentation, particle sizing by, 10, 11, 12-13
regime classifications, in dense-phase fluidized settling chambers, gas-solid separation by, 297
beds, 371-376 shape coefficients, 3
relaxation time shape factors, 3
of particle, 207, 231 shear strain, 51
Stokes relaxation time, 93, 108, 126, 199, 239, shear stress, 51, 56
265,285,306,330,471,497 shear testing
retardation effects, 101, 102 Jenike translatory cell for, 341, 342
Reynolds decomposition concept, application to of powders, 341,345
turbulent flow analysis, 174, 175, 193 rotational split-level shear cell for, 342
Reynolds number, 6, 88, 97, 100, 101, 107, 160, Sherwood number, 157, 158, 160, 531, 534
181, 231, 232, 234, 302, 329,470,471,484, shock waves
487,513 force generated by, 96, 108, 244
heat transfer and, 130, 139 propagation of, 108, 265-269
Reynolds stress, 175, 176, 177, 194, 195, 452 sieve standards, 4-5
Richardson-Dushman equation, 119 sieving, for particle sizing, 10, 18
Index 555
silica glass, elastic properties of, 30, 62 stress distributions in, 337-340
silo design, 344 system types, 361-366
silver, as electric conductor, 33 underflow standpipes, 361-364
similarity, dimensional analysis and, 230-235 static friction coefficient, 64, 66, 68
single-particle model, for heat transfer, 501, 503 steel
single-phase flows, modeling of, 167-181 elastic properties of, 30
skeletal density, 38, 39 thermal properties of, 35
slip velocity yield strength of, 31
cluster concept and, 447 steel ball, restitution coefficient of, 82
between particles, 451, 532 Stefan-Boltzmann constant, 161, 491,495, 517, 535
between phases, 279, 281 Stefan-Boltzmann's law, 34, 42, 45, 142
in standpipes, 363 Stefan flow, 158, 159, 160
slugging Stefan-Maxwell equation, 157
in circulating fluidized beds, 425 stereomechanical impact, 46, 4 7 ^ 9
continuous, 404-406 stochastic trajectory method, for multiphase flow,
in dense-phase fluidized beds, 374, 375-376, 392, 165, 205, 208-210
403-406 Stokes drag force, 89, 93, 95, 97, 100, 108, 109,
soda-lime glass, thermal properties of, 35 125,260,304,321,474
solids circulation Stokes expansion, 98, 99
in circulating fluidized beds, 524, 525 Stokes flows, 7, 88-90, 94, 108, 139, 140, 261, 323
solids mixing Stokes number, 231, 450, 453, 464
in circulating fluidized beds, 443-453 Stokes relaxation time, 93, 108, 126, 199, 239, 265,
in dense-phase fluidized beds, 371, 381, 384, 401 285,306,330,471,497
specific heat Stokes stream function, 97
of gas-solid mixtures, 244, 257-258 straight capillaric model, 224-225, 226
of particles, 32-33 strain(s)
ratio of, 257 deformation caused by, 28-29, 46,49
speed of sound, 244, 261-265 normal, 51
equilibrium speed of sound, 265 shear, 51
frozen, 265 stratified flows
in gas-solid mixtures, 244, 261-265 in horizontal pipes, 273-276
sphere(s), 4 wave motions in, 270-280
elastic, 69-71, 104, 133-138 stream function, 90, 91-92, 97, 317
frictional, 46, 49, 50, 64, 74-78 for linear motion of spheres, 89-92
frictionless, 59-63 stress(es)
motion of, 87 from boundary point forces, 56-58
oscillation of, 259-261 collisional, 104-105
perfect elastic spheres, 48, 49 deformation caused by, 28-29, 46, 49
perfect inelastic spheres, 48 general relations of, 50-52
rotation of, 97-101 normal, 55
spouted beds shear, 51, 56
in dense-phase fluidization, 372, 374, 376, 388, yield, 78, 79
406-409,419 stress distribution
heat transfer in, 526-527 equation of equilibrium, 341
spouting in hoppers, 337-340
effect on heat transfer coefficients, 500 Janssen's model (cylindrical column), 333,
fountain height, 408 338-339
maximum spoutable bed depth in, 408 in steady hopper flow, 340-342
minimum spouting velocity, 407 Walker's analysis, 339-340
onset of, 407 Walters' analysis, 339-340
spout diameter in, 406,408 stress failure
spray chamber scrubbers, 323-324 active failure, 337, 340
square-nosed slugs, formation in dense-phase incipient failure, 336
fluidized beds, 404 Mohr-Coulomb failure criterion, 336-337
standpipes and standpipe flows, 333, 350-354 passive failure, 337, 340
aeration of, 365, 366 of powders, 333
downcomer use as, 423 Strouhal number, wake shedding frequency and, 286
inclined standpipes, 364-366 Sturm-Liouville boundary-value problem, 493
leakage of flow of gas in, 359-361 surface abrasion model, 253
multiplicity of steady flows in, 357-359 surface area, determination in particles, 13
nonmechanical valves in, 364-366 surface diameter, of particles, 6
overflow standpipes, 361-364 surface mean diameter, of paniculate systems, 23
556 Index
surface renewal model, for heat transfer, 502 turbulence models and, 196-204
suspension density, effects on heat transfer for turbulent viscosity, 177
coefficient, 525 for viscosity, 173
suspension-surface heat transfer. See also transport disengagement height (TDH), in
bed-to-surface heat transfer dense-phase fluidized beds, 401,402, 403, 410
in fluidized beds, 499-511, 521-522 transport theorem, 167-168
suspension-to-gas heat transfer. See also bed-to-gas continuity equation in, 169, 175, 179, 195
heat transfer energy equation in, 169-170, 196, 207-208, 213,
in dense-phase fluidized beds, 512 214-215
momentum equation in, 169, 175, 179, 195, 207,
tangential displacements, 63-64 214,271
tangential flow cyclones, 297, 298, 300, 302, 303 transport velocity, in circulating fluidized beds, 425
collection efficiency of, 303-307 Tresca criterion, 78, 79
tangential tractions triboelectric charging, 112
in nonsliding contact, 66-69 triboelectric series, 112, 113-114
in sliding contact, 64-66 tungsten, elastic properties of, 30
Taylor expansion, 121, 173 turbulence
tensile strength, of particles, 29-30, 31 in cyclone separators, 302
thermal conductivity, 33, 174 modulation of, 286, 287
thermal convection particle interaction with, 285-288
heat transfer by, 130 turbulent dissipation rate, transport equation of, 177,
in a pseudocontinuum flow, 141-142 178
of spheres, 139 turbulent flow modeling, 174-179
thermal diffusivity, 133, 157 Hinze-Tchen model, 165, 197
thermal properties, of particles, 3, 13, 25, 32-35 k--kv model, 165, 197, 201-202
thermal radiation k- model, 165, 175, 176, 177-179
heat transfer by, 130, 142-152 mixing length model, 165, 175, 176-177
of particle clouds, 130,150-156 transport coefficients and, 196-204
of particles, 3,13,33-35 turbulent fluidization, in dense-phase fluidized beds,
in pseudocontinuum flows, 130 374,375, 396-400
through isothermal and diffuse scattering media, turbulent kinetic energy
154-156 equation for dissipation rate of, 179
thermionic emission, charging by, 119 equation of, 179
thermionic work function, 114 of particles, 285
thermodynamic properties transport equation of, 177-178
equation of state, 244, 254-257 turbulent mixing, 304, 312
of gas-solid mixtures, 244, 254-259 turbulent viscosity, 181
specific heat, 257-258 two-phase theory of fluidization, 375, 393,405, 420,
specific heat ratio, 257 451
speed of sound, 244, 261-265 Tyler Standard sieves, 4, 10, 11
thermophoresis effect, 96, 97,108, 323
time-averaging unsteady heat transfer behavior, 502
application to turbulent flow analysis, 174
of Navier-Stokes equations, 174, 175-176 Van der Waals adsorption, 24, 25
tornado dust collector, 299 Van der Waals forces, 87, 101-103, 108, 124, 128,
torsion, of elastic spheres in contact, 69-71, 75 191,315,373,445,474
traction, distribution of, 68, 69 macroscopic approach to, 102-104
trajectory modeling of multiphase flows, 205-210. microscopic approach to, 101-102
See also Lagrangian trajectory model venturi scrubbers, 323, 324, 326
deterministic trajectory model, 165, 205, 206-208 vibratory conveyors, solids transport by, 461
stochastic trajectory model, 165, 205, 208-210 vibrofluidized beds, 377
transmission, effect on thermal radiation, 143, 150 viscoelastic deformation, 28
transmission electron microscopy (TEM), for viscosity
particle sizing, 10, 11, 12 coefficient of, 173
transport coefficients effective, definition of, 179
based on Maxwell-Boltzmann distribution, 174 of gas-solid suspensions, 384
for diffusivity, 173-174 Volta potential, 114
for eddy diffusivity, 194 volume averaging, 165, 182-183
kinetic theory and, 170-174 volume-averaging theorems, 187-189, 190
for particle viscosity, 173 volume diameter, of particles, 6
for single-phase flows, 167 volume fraction, of particles, 268, 269, 272
for thermal conductivity, 174 volume mean diameter, of particulate systems, 23
Index 557
volume-time-averaged equations, 193-196, 206 in dense-phase fluidized beds, 382

Von Karman's constant, 180, 239 dynamic waves, 282-284
Von Mises criterion, 78 motion of, 270-280
vortex shedding, 286, 287 speed of, 281,283
V-valve, in circulating fluidized beds, 423 wave equation, 279
Weber number, 385
wake Wen-Yu equation, 379
from bubbles, 381, 387 wet scrubbing and scrubbers
shedding of, 286, 385 cocurrent-flow mode in, 324, 328
in slugging bed, 404 countercurrent-flow mode in, 324, 327
wake angle, of bubbles, 383-384 cross-flow mode in, 324
wake attraction, particle collision due to, 93-95, 128 cyclone scrubbers, 323
Walters's analysis, of stress distribution, 339-340 gas-solid separation by, 297, 321
walls mechanisms and types of, 323324
effect on dense-phase fluidized beds, 403, 418 modeling of, 324-328
particle collisions with, 76-78, 82, 87, 307 spray chamber scrubbers, 323-324
role in heat transfer, 521-522, 527 venturi scrubbers, 323, 324, 326
wall slugs, formation in dense-phase fluidized beds, Wiedemann-Franz law, 33
404 Wien's displacement law, 34,143
Walters's analysis, of stress distribution, 339-340 wood, elastic properties of, 30
wave equation, 279
waves. See also pressure waves yield strength, of particles, 29-30, 31
in circulating fluidized beds, 444 yield stress, 245
continuity waves, 280-282 Young's modulus, 29, 30,41, 51, 63, 83

Principles of Gas-Solid Flows-Fan, Zhu PDF

Uploaded by

Copyright:

Available Formats

Principles of Gas-Solid Flows-Fan, Zhu PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Principles of Gas-Solid Flows-Fan, Zhu PDF

Uploaded by

Copyright:

Available Formats

Principles of Gas-Solid Flows

Cambridge University Press

The Edinburgh Building, Cambridge CB2 2RU, UK

This publication is in copyright. Subject to statutory exception

First published 1998

This digitally printed first paperback version 2005

A catalogue recordfor this publication is available from the British Library

Library of Congress Cataloguing in Publication data

ISBN-13 978-0-521 -02116-6 paperback

Part I Basic Relationships

2 Collision Mechanics of Solids 46

3 Momentum Transfer and Charge Transfer 87

4 Basic Heat and Mass Transfer 130

5 B asic Equations 164

6 Intrinsic Phenomena in a Gas-Solid Flow 244

6.2.2 Locations of Erosive Wear 247

Part II System Characteristics

8 Hopper and Standpipe Flows 333

8.2.4 Stress Distribution in a Steady Hopper Flow 340

9 Dense-Phase Fluidized Beds 371

10 Circulating Fluidized Beds 421

11 Pneumatic Conveying of Solids 461

12 Heat and Mass Transfer Phenomena in Fluidization Systems 499

Appendix: Summary of Scalar, Vector, and Tensor Notations 540

Size and Properties of Particles

1.2 Particle Size and Sizing Methods

Table 1.1. Definitions of Particle Shape

Source: T. Allen's Particle Size Measurements, Chapman & Hall, 1990.

L Typical particle size range covered in this book J

- Aerosol reactor Fluidizedbed J

. Air pollution- Applications

1.2.1 Equivalent Diameters of a Nonspherical Particle

An equivalent diameter of a particle is usually defined in relation to a specific

1.2.1.1 Sieve Diameter

1.2.1.2 Martin's Diameter, Feret's Diameter, and Projected Area Diameter

1.2.1.3 Surface Diameter, Volume Diameter, and Sauter's Diameter

where V is the particle volume. The Sauter's diameter or surface-volume diameter is

1.2.1 A Dynamic Diameter

the terminal velocity of a sphere is related to its diameter by

C/2 = 3.o 3 d t i p v ~ P>* 500 < Re, < 2 x 105

Table El.l. Cumulative Weight Percentage Versus Terminal Velocity

Upt (m/s) Cumulative wt% Upt (m/s) Cumulative wt%

4.4 x 10"3 99.9 1.1 X 10-5 65.6

Table El.2. Mass Fraction (wt%) Versus

f/pt (m/s) dt (fim) /M (wt%)

Figure El.l. Mass fraction distribution based on data in Table El.2.

Table 1.2. Some Methods of Particle Size

Method Size range {iim)

1.2.2 Particle Sizing Methods

Table 1.3. Tyler Standard and U.S. ASTM Sieve Series

Tyler standard U.S. series ASTM standard

Source: A. G. Guy's Essentials of Materials Science, McGraw-Hill, 1976.

1.2.2.4 Gas Adsorption

1.2.2.5 Fraunhofer Diffraction

Incident light ^ , ^^^ _ ,

K < 0.3 Rayleigh scattering

The parameter K is defined as

where nx is the relative index of refraction of the particle.

Plane with a circular Observation