1

Overview

1–1
Chapter 1: OVERVIEW 1–2

§1.1 WHERE THE BOOK FITS

This book is an introduction to the analysis of nonlinear elastic structures by the Finite Element
Method (FEM). It embodies five Parts:
I Overview of Nonlinear Problems.
II Formulation of Geometrically Nonlinear Finite Elements.
III Solution Methods.
IV Application to Stability Analysis.
V Nonconservative Problems.
This Chapter presents an overview of where the book fits within the larger scope of Mechanics. It
is assumed that the reader has a good idea of what finite elements are, so this aspect is glossed over.

§1.2 WHERE THIS MATERIAL FITS

The field of Mechanics can be subdivided into three major areas:

 Theoretical
Mechanics Applied (1.1)
Computational

Theoretical mechanics deals with fundamental laws and principles of mechanics studied for their
intrinsic scientific value. Applied mechanics transfers this theoretical knowledge to scientific and
engineering applications, especially as regards the construction of mathematical models of physical
phenomena. Computational mechanics solves specific problems by simulation through numerical
methods implemented on digital computers.

REMARK 1.1
Paraphrasing an old joke about mathematicians, one may define a computational mechanician as a person who
searches for solutions to given problems, an applied mechanician as a person who searches for problems that
fit given solutions, and a theoretical mechanician as a person who can prove the existence of problems and
solutions.

§1.2.1 Computational Mechanics

Several branches of computational mechanics can be distinguished according to the physical scale
of the focus of attention:

 Nanomechanics and micromechanics

  Solids and Structures

Computational Mechanics Continuum mechanics Fluids (1.2)



 Multiphysics
Systems

Nanomechanics deals with phenomena at the molecular and atomic levels of matter. As such it
is closely interrelated with particle physics and chemistry. Micromechanics looks primarily at the

1–2
1–3 §1.2 WHERE THIS MATERIAL FITS

crystallographic and granular levels of matter. Its main technological application is the design and
fabrication of materials and microdevices.
Continuum mechanics studies bodies at the macroscopic level, using continuum models in which
the microstructure is homogenized by phenomenological averages. The two traditional areas of
application are solid and fluid mechanics. The former includes structures which, for obvious
reasons, are fabricated with solids. Computational solid mechanics takes a applied-sciences ap-
proach, whereas computational structural mechanics emphasizes technological applications to the
analysis and design of structures.
Computational fluid mechanics deals with problems that involve the equilibrium and motion of
liquid and gases. Well developed related areas are hydrodynamics, aerodynamics, atmospheric
physics, and combustion.
Multiphysics is a more recent newcomer. This area is meant to include mechanical systems that
transcend the classical boundaries of solid and fluid mechanics, as in interacting fluids and structures.
Phase change problems such as ice melting and metal solidification fit into this category, as do the
interaction of control, mechanical and electromagnetic systems.
Finally, system identifies mechanical objects, whether natural or artificial, that perform a distin-
guishable function. Examples of man-made systems are airplanes, building, bridges, engines,
cars, microchips, radio telescopes, robots, roller skates and garden sprinklers. Biological systems,
such as a whale, amoeba or pine tree are included if studied from the viewpoint of biomechanics.
Ecological, astronomical and cosmological entities also form systems.1
In this progression of (1.2) the system is the most general concept. Systems are studied by de-
composition: its behavior is that of its components plus the interaction between the components.
Components are broken down into subcomponents and so on. As this hierarchical process continues
the individual components become simple enough to be treated by individual disciplines, but their
interactions may get more complex. Thus there is an art in deciding where to stop.2

§1.2.2 Statics vs. Dynamics

Continuum mechanics problems may be subdivided according to whether inertial effects are taken
into account or not:

Statics
Continuum mechanics (1.3)
Dynamics

In dynamics the time dependence is explicitly considered because the calculation of inertial (and/or
damping) forces requires derivatives respect to actual time to be taken.
Problems in statics may also be time dependent but the inertial forces are ignored or neglected.
Static problems may be classified into strictly static and quasi-static. For the former time need not

1 Except that their function may not be clear to us. “The usual approach of science of constructing a mathematical model
cannot answer the questions of why there should be a universe for the model to describe. Why does the universe go to
all the bother of existing?” (Stephen Hawking).
2 Thus in breaking down a car engine, say, the decomposition does not usually proceed beyond the components you can
buy at a machine shop.

1–3
Chapter 1: OVERVIEW 1–4

be considered explicitly; any historical time-like response-ordering parameter (if one is needed)
will do. In quasi-static problems such as foundation settlement, creep deformation, rate-dependent
plasticity or fatigue cycling, a more realistic estimation of time is required but inertial forces are
still neglected.

§1.2.3 Linear vs. Nonlinear

A classification of static problems that is particularly relevant to this book is


Linear
Statics
Nonlinear

Linear static analysis deals with static problems in which the response is linear in the cause-and-
effect sense. For example: if the applied forces are doubled, the displacements and internal stresses
also double. Problems outside this domain are classified as nonlinear.

§1.2.4 Discretization methods

A final classification of CSM static analysis is based on the discretization method by which the
continuum mathematical model is discretized in space, i.e., converted to a discrete model of finite
number of degrees of freedom:



Finite Element Method (FEM)




Boundary Element Method (BEM)
Finite Difference Method (FDM)
Spatial discretization method (1.4)

 Finite Volume Method (FVM)



 Spectral Method
Mesh-Free Method

For linear problems finite element methods currently dominate the scene, with boundary element
methods posting a strong second choice in specific application areas. For nonlinear problems the
dominance of finite element methods is overwhelming.
Classical finite difference methods in solid and structural mechanics have virtually disappeared
from practical use. This statement is not true, however, for fluid mechanics, where finite difference
discretization methods are still dominant. Finite-volume methods, which address finite volume
method conservation laws, are important in highly nonlinear problems of fluid mechanics. Spectral
methods are based on transforms that map space and/or time dimensions to spaces where the problem
is easier to solve.
A recent newcomer to the scene are the mesh-free methods. These are finite different methods on
arbitrary grids constructed by a subset of finite element techniques

§1.2.5 FEM Variants

The term Finite Element Method actually identifies a broad spectrum of techniques that share
common features outlined in introductory FEM textbooks. Two subclassifications that fit well

1–4
1–5 §1.3 THE FEM ANALYSIS PROCESS

VERIFICATION
Mathematical Discretization + solution error
model
IDEALIZATION
REALIZATION
FEM
SOLUTION

Ideal Discrete Discrete

physical model solution
system

IDEALIZATION & VERIFICATION

DISCRETIZATION
solution error

generally irrelevant

Figure 1.1. The Mathematical FEM. The mathematical model (at top) is the source of the process.
Discrete model and solution follow from it. The ideal physical system is inessential.

applications to structural mechanics are


 Displacement 
 Stiffness
Equilibrium
FEM Formulation FEM Solution Flexibility (1.5)

 Mixed Mixed (a.k.a. Combined)
Hybrid

Using the foregoing classification, we can state the topic of this book more precisely: the compu-
tational simulation of nonlinear static structural problems by the Finite Element Method. Of the
variants listed in (1.5), emphasis is placed on the displacement formulation and stiffness solution.
This combination is called the Direct Stiffness Method or DSM.

§1.3 THE FEM ANALYSIS PROCESS

A model-based simulation process using FEM involves doing a sequence of steps. This sequence
takes two canonical configurations depending on the environment in which FEM is used. These
are reviewed next to introduce terminology.

§1.3.1 The Mathematical FEM

The process steps are illustrated in Figure 1.1. The process centerpiece, from which everything
emanates, is the mathematical model. This is often an ordinary or partial differential equation in
space and time. A discrete finite element model is generated from a variational or weak form of
the mathematical model.3 This is the discretization step. The FEM equations are processed by an
equation solver, which delivers a discrete solution (or solutions).

3 The distinction between strong, weak and variational forms is discussed in advanced FEM courses. In the present book
such forms will be stated as recipes.

1–5
Chapter 1: OVERVIEW 1–6

generally
Ideal irrelevant
Mathematical
model

CONTINUIFICATION

SOLUTION

Physical FEM Discrete Discrete

system model solution

IDEALIZATION & VERIFICATION

DISCRETIZATION
solution error

simulation error= modeling + solution error

VALIDATION

Figure 1.2. The Physical FEM. The physical system is the source of
the process. The ideal mathematical model is inessential.

On the left Figure 1.1 shows an ideal physical system. This may be presented as a realization of
the mathematical model; conversely, the mathematical model is said to be an idealization of this
system. For example, if the mathematical model is the Poisson’s equation, realizations may be a
heat conduction or a electrostatic charge distribution problem. This step is inessential and may be
left out. Indeed FEM discretizations may be constructed without any reference to physics.
The concept of error arises when the discrete solution is substituted in the “model” boxes. This
replacement is generically called verification. The solution error is the amount by which the
discrete solution fails to satisfy the discrete equations. This error is relatively unimportant when
using computers, and in particular direct linear equation solvers, for the solution step. More
relevant is the discretization error, which is the amount by which the discrete solution fails to
satisfy the mathematical model.4 Replacing into the ideal physical system would in principle
quantify modeling errors. In the mathematical FEM this is largely irrelevant, however, since the
ideal physical system is merely that: a figment of the imagination.

§1.3.2 The Physical FEM

The second way of using FEM is the process illustrated in Figure 1.3. The centerpiece is now
the physical system to be modeled. Accordingly, this sequence is called the Physical FEM. The
processes of idealization and discretization are carried out concurrently to produce the discrete
model. The solution is computed as before.
Just like Figure 1.1 shows an ideal physical system, Figure 1.1 depicts an ideal mathematical
model. This may be presented as a continuum limit or “continuification” of the discrete model. For
some physical systems, notably those well modeled by continuum fields, this step is useful. For

4 This error can be computed in several ways, the details of which are of little practical importance.

1–6
1–7 §1.3 THE FEM ANALYSIS PROCESS

others, notably complex engineering systems, it makes no sense. Indeed FEM discretizations may
be constructed and adjusted without reference to mathematical models, simply from experimental
measurements.
The concept of error arises in the physical FEM in two ways, known as verification and validation,
respectively. Verification is the same as in the Mathematical FEM: the discrete solution is replaced
into the discrete model to get the solution error. As noted above this error is not generally important.
Substitution in the ideal mathematical model in principle provides the discretization error. This is
rarely useful in complex engineering systems, however, since there is no reason to expect that the
mathematical model exists, and if it does, that it is more relevant than the discrete model. Validation
tries to compare the discrete solution against observation by computing the simulation error, which
combines modeling and solution errors. Since the latter is typically insignificant, the simulation
error in practice can be identified with the modeling error.

§1.3.3 Synergy of Physical and Mathematical FEM

The foregoing physical and mathematical sequences are not exclusive but complementary. This
synergy5 is one of the reasons behind the power and acceptance of the method. Historically the
physical FEM was the first one to be developed to model very complex systems such as aircraft.
The mathematical FEM came later and, among other things, provided the necessary theoretical
underpinnings to extend FEM beyond structural analysis.
A glance at the schematics of a commercial jet aircraft makes obvious the reasons behind the
physical FEM. There is no differential equation that captures, at a continuum mechanics level,6 the
structure, avionics, fuel, propulsion, cargo, and passengers eating dinner.
There is no reason for despair, however. The time honored divide and conquer strategy, coupled
with abstraction, comes to the rescue. First, separate the structure and view the rest as masses
and forces, most of which are time-varying and nondeterministic. Second, consider the aircraft
structure as built of substructures7 wings, fuselage, stabilizers, engines, landing gears, and so on.
Take each substructure, and continue to decompose it into components: rings, ribs, spars, cover,
plates, etc, continuing through as many levels as necessary. Eventually those components become
sufficiently simple in geometry and connectivity that they can be reasonably well described by
continuum mathematical models provided, for instance, by Mechanics of Materials or the Theory
of Elasticity. At that point, stop. The component level discrete equations are obtained from a
FEM library based on the mathematical model. The system model is obtained by going through
the reverse process: from component equations to substructure equations, and from those to the
equations of the complete aircraft. This system assembly process is governed by the classical
principles of Newtonian mechanics expressed in conservation form.
This multilevel decomposition process is diagrammed in Figure 1.3, in which the intermediate

5 This interplay is not exactly a new idea: “The men of experiment are like the ant, they only collect and use; the reasoners
resemble spiders, who make cobwebs out of their own substance. But the bee takes the middle course: it gathers its
material from the flowers of the garden and field, but transforms and digests it by a power of its own.” (Francis Bacon).
6 Of course at the atomic and subatomic level quantum mechanics works for everything, from landing gears to passengers.
But it would be slightly impractical to model the aircraft by 1036 interacting particles.
7 A substructure is a part of a structure devoted to a specific function.

1–7
Chapter 1: OVERVIEW 1–8

ENT
M PON
at ical CO EVEL
hem L
Matmodel
ent
ponns
ary Com atio
Libr en t equ
FEM Com
pon e
TEM
ret
discdel
mo
SYS EL
LEV e
plet
Comution
s o l
em
Systrete
discdel
m o
l
sica
Phy tem
sys

Figure 1.3. Combining physical and mathematical modeling through multilevel FEM.
Only two levels (system and component) are shown for simplicity.

substructure level is omitted for simplicity.

REMARK 1.2
More intermediate decomposition levels are used in some systems, such as offshore and ship structures, which
are characterized by a modular fabrication process. In that case the decomposition mimics the way the system is
actually constructed. The general technique, called superelements, is outlined in the IFEM book. In nonlinear
analysis the analysis procedure is more complex since the static condensation technique of linear analysis no
longer works.

REMARK 1.3
There is no point in practice in going beyond a certain component level while considering the complete model,
since the level of detail can become overwhelming without adding significant information. Further refinement
or particular components is done by the so-called global-local analysis techniques. This technique is an
instance of multiscale analysis.

For sufficiently simple structures, passing to a discrete model is carried out in a single idealization
and discretization step, as illustrated for the truss roof structure shown in Figure 1.4. Multiple
levels are unnecessary here. Of course the truss may be viewed as a substructure of the roof, and
the roof as a a substructure of a building.

§1.4 INTERPRETATIONS OF THE FINITE ELEMENT METHOD

Just like there are two complementary ways of using the FEM, there are two complementary
interpretations for teaching it. One interpretation stresses the physical significance and is aligned
with the Physical FEM. The other focuses on the mathematical context, and is aligned with the
Mathematical FEM.

1–8
1–9 §1.4 INTERPRETATIONS OF THE FINITE ELEMENT METHOD

member

support

joint
Physical System

IDEALIZATION

;;
;;
;;

;;
;;
;;
Figure 1.6. The idealization process for a simple structure. The physical system,
here a roof truss, is directly idealized by the mathematical model: a
pin-jointed bar assembly. This coalesces with the discrete model.

§1.4.1 Physical Interpretation

The physical interpretation is aligned with the view of Figure 1.2. This interpretation has been
shaped by the discovery and extensive use of the method in the field of structural mechanics. This
relationship is reflected in the use of structural terms such as “stiffness matrix”, “force vector” and
“degrees of freedom.” This terminology carries over to non-structural applications.
The basic concept in the physical interpretation is the breakdown (≡ disassembly, tearing, partition,
separation, decomposition) of a complex mechanical system into simpler, disjoint components
called finite elements, or simply elements. The mechanical response of an element is characterized
in terms of a finite number of degrees of freedom. These degrees of freedoms are represented as
the values of the unknown functions as a set of node points. The element response is defined by
algebraic equations constructed from mathematical or experimental arguments. The response of
the original system is considered to be approximated by that of the discrete model constructed by
connecting or assembling the collection of all elements.
The breakdown-assembly concept occurs naturally when an engineer considers many artificial and
natural systems. For example, it is easy and natural to visualize an engine, bridge, aircraft or
skeleton as being fabricated from simpler parts.
The underlying theme is divide and conquer. If the behavior of a system is too complex, the
recipe is to divide it into more manageable subsystems. If these subsystems are still too complex
the subdivision process is continued until the behavior of each subsystem is simple enough to fit a
mathematical model that represents well the knowledge level the analyst is interested in. In the finite
element method such “primitive pieces” are called elements. elements The behavior of the total
system is that of the individual elements plus their interaction. A key factor in the initial acceptance
of the FEM was that the element interaction can be physically interpreted and understood in terms
that were eminently familiar to structural engineers.

1–9
Chapter 1: OVERVIEW 1–10

§1.4.2 Mathematical Interpretation

This interpretation is closely aligned with the configuration of Figure 1.1. The FEM is viewed as
a procedure for obtaining numerical approximations to the solution of boundary value problems
(BVPs) posed over a domain . This domain is replaced by the union ∪ of disjoint subdomains
(e) called finite elements. In general the geometry of  is only approximated by that of ∪(e) .
The unknown function (or functions) is locally approximated over each element by an interpolation
formula expressed in terms of values taken by the function(s), and possibly their derivatives, at a
set of node points generally located on the element boundaries. The states of the assumed unknown
function(s) determined by unit node values are called shape functions. The union of shape functions
“patched” over adjacent elements form a trial function basis for which the node values represent the
generalized coordinates. The trial function space may be inserted into the governing equations and
the unknown node values determined by the Ritz method (if the solution extremizes a variational
principle) or by the Galerkin, least-squares or other weighted-residual minimization methods if the
problem cannot be expressed in a standard variational form.

§1.5 THE SOLUTION MORASS

In nonlinear analysis the two FEM interpretations are not equal in importance. Nonlinear analysis
demands a persistent attention to the underlying physics to avoid getting astray as the “real world”
is covered by layer upon layer of mathematics and numerics.
Why is concern for physics of paramount importance? A key component of finite element nonlinear
analysis is the solution of the nonlinear algebraic systems of equations that arise upon discretization.

FACT

The numerical solution of nonlinear systems in “black box” mode is much

more difficult than in the linear case.

The key difficulty is tied to the essentially obscure nature of general nonlinear systems, about which
very little can be said in advance. And you can be sure that Murphy’s law8 works silently in the
background.
One particularly vexing aspect of dealing with nonlinear systems is the solution morass. A deter-
minate system of 1, 1000, or 1000000 linear equations has, under mild conditions, one and only
one solution. The computer effort to obtain this solution can be estimated fairly accurately if the
sparseness (or denseness) of the coefficient matrix is known. Thus setting up linear equation solvers
as “black-box” stand-alone functions or modules is a perfectly sensible thing to do.
By way of contrast, a system of 1000 cubic equations has 31000 ≈ 10300 solutions in the complex
plane. This is much, much larger than the number of atoms in the Universe, which is merely 1050

8 If something can go wrong, it will go wrong.

1–10
 1–11 §1.6 HISTORICAL BACKGROUND

give or take a few. Suppose just several billions or millions of these are real solutions. Which
solution(s) have physical meaning? And how do you compute those solutions without wasting time
on the others?
This combinatorial difficulty is overcome by the concept of continuation, which engineers also call
incremental analysis. Briefly speaking, we start the analysis from an easily computable solution
— for example, the linear solution — and then try to follow the behavior of the system as actions
applied to it are changed by small steps called increments. The previous solution is used as a starting
point for the iterative solution-search procedure. The underlying prescription: follow the physics.
This technique is interwined with the concept of response explained in Chapter 2.
REMARK 1.4
Not surprisingly, incremental analysis was used by the aerospace engineers that first used the finite element
method for geometrically nonlinear analysis in the late 1950s. Techniques have been considerably refined
since then, but the underlying idea remains the same.

We conclude this overview with a historical perspective on nonlinear finite element methods in
solid and structural mechanics, along with a succint bibliography.

§1.6 HISTORICAL BACKGROUND

In the history of finite element methods the year 1960 stands out. The name “finite element method” appears
for the first time in the open literature in an article by Clough [52]. And Turner, Dill, Martin and Melosh [187]
publish a pioneering paper in nonlinear structural analysis. The then-five-year-old “direct stiffness method”
(what we now call displacement-assumed finite element method) was applied to
“problems involving nonuniform heating and/or large deflections . . . in a series of linearized steps. Stiffness
matrices are revised at the beginning of each step to account for changes in internal loads, temperatures, and
geometric configuration.”
Thirty years and several thousand publications later, computerized nonlinear structural analysis has acquired
full adult rights, but has not developed equally in all areas.
The first fifteen years (1960-1975) were dominated by formulation concerns. For example, not until the
late 1960s were correct finite-deflection incremental forms for displacement models rigorously derived. And
interaction of flow-like constitutive behavior with the spatial discretization (the so called “incompressibility
locking” effects) led to important research into constitutive equations and element formulations.
While the investigators of this period devoted much energy to obtaining correct and implementable nonlinear
finite-element equations, the art of solving such equations in a reliable and efficient manner was understandably
neglected. This helps to explain the dominance of purely incremental methods. Corrective methods of Newton
type did not get much attention until the early 1970s, and then only for geometrically nonlinear problems. At
the time of this writing, progress in numerical solution techniques has been uneven: well developed for certain
problems, largely a black art in others. To understand the difference, it pays to distinguish between smooth
nonlinearities and rough nonlinearities.

§1.6.1 Smooth Nonlinearities

Problem with smooth nonlinearities are characterized by continuous, path-independent nonlinear relations at
the local level. Some examples:

1. Finite deflections (geometric nonlinearities). Nonlinear effects arise from strain-displacement equations,
which are well behaved for all strain measures in practical use.

1–11
Chapter 1: OVERVIEW 1–12

2. Nonlinear elasticity. Stresses are nonlinear but reversible functions of strains.

3. Follower forces (e.g., pressure loading). External forces are smooth nonlinear functions of displacements.

A unifying characteristic of this problem is that nonlinearities are of equality type, i.e., reversible, and these
relations are continuous at each point within the structure. Mathematicians call these smooth mappings.
It is important to point out, however, that the overall structural behavior is not necessarily smooth; as witnessed
by the phenomena of buckling, snapping and flutter. But at the local level everything is smooth: nonlinear
strain-displacement equations, nonlinear elasticity law, follower pressures.
Methods for solving this class of problems are highly developed, and have received a great deal of attention
from the mathematical and numerical analysis community. This research has directly benefitted many areas
of structural analysis.
Let us consider finite deflection problems as prototype. Within the finite element community, these were
originally treated by purely incremental (step-by-step) techniques; but anomalies detected in the mid-1960’s
prompted research into consistent linearizations. A good exposition of this early work is given in the book by
Oden [123]. Once formulation questions were settled, investigators had correct forms of the “residual” out-
of-balance forces and tangent stiffness matrix, and incremental steps began to be augmented with corrective
iterations in the late 1960s. Conventional and modified Newton methods were used in the corrective phase.
These were further extended through restricted step (safeguarded Newton) and, more recently, variants of the
powerful conjugate-gradient and quasi-Newton methods.
But difficulties in detecting and traversing limit and bifurcation points still remained. Pressing engineering
requirements for post-buckling and post-collapse analyses led to the development of displacement control,
alternating load/displacement control, and finally arclength control. The resultant increment-control methods
have no difficulty in passing limit points. The problem of reliably traversing simple bifurcation points without
guessing imperfections remains a research subject, while passing multiple or clustered bifurcation points
remains a frontier subject. A concerted effort is underway, however, to subsume these final challenges.
These reliable solution methods have been implemented into many special-purpose finite element programs,
and incorporation into general-purpose programs is proceeding steadily.

REMARK 1.5
As noted above, incremental methods were the first to be used in nonlinear structural analysis. Among the pre-1970
contributions along this line we may cite Argyris and coworkers [9, 10], Felippa [62], Goldberg and Richard [82], Marcal,
Hibbitt and coworkers [89,108,109], Oden [122], Turner, Martin and coworkers [110,187,188],

REMARK 1.6
The earliest applications of Newton methods to finite element nonlinear analysis are by Oden [122], Mallet and Marcal [107],
and Murray and Wilson [117,118]. During the early 1970s Stricklin, Haisler and coworkers at Texas A&M implemented
and evaluated self-corrective, pseudo-force, energy-search and Newton-type methods and presented extensive comparisons;
see Stricklin et. al. [175–178], Tillerson et. al. [185], and Haisler et. al. [85]. Almroth, Brogan, Bushnell and coworkers
at Lockheed began using true and modified Newton methods in the late 1960s for energy-based finite-difference collapse
analysis of shells; see Brogan and Almroth [34], Almroth and Felippa [5], Brush and Almroth [37], and Bushnell [38–39].
By the late 1970s Newton-like methods enjoyed widespread acceptance for geometrically nonlinear analysis.

REMARK 1.7
Displacement control strategies for finite element post-buckling and collapse analysis were presented by Argyris [11] and
Felippa [62] in 1966, and generalized in different directions by Sharifi and Popov [165,166] (fictitious springs), Bergan et.
al. [25,26], (current stiffness parameter), Powell and Simons [142] and Bergan and Simons [27] (multiple displacement
controls). A modification of Newton’s method to traverse bifurcation points was described by Thurston [184]. Arclength
control schemes for structural problems may be found in the following source papers: Wempner [194], Riks [154], Schmidt

1–12
1–13 §1.6 HISTORICAL BACKGROUND

[160], Crisfield [48,49], Ramm [146], Felippa [65–66], Fried [73], Park [133], Padovan [130,131], Simo et.al. [167], Yang
and McGuire [199], Bathe and Dvorkin [20]. Other articles of particular interest are Bathe and Cimento [18], Batoz and
Dhatt [22], Bushnell [39], Bergan [27], Geradin et al. [79,80]. Meek and Tan [112], Ramm [146,147], Riks [155–157], Sobel
and Thomas [170], Zienkiewicz [200,201,203]. Several conferences have been devoted exclusively to nonlinear problems
in structural mechanics, for example [12,28,17,125, 179,180,198]. Finite element textbooks and monographs dealing rather
extensively with nonlinear problems are by Oden [123], Bathe [19], Bushnell [40], White [196] and Zienkiewicz [202].

REMARK 1.8

In the mathematical literature the concept of continuation (also called imbedding) can be traced back to the 1930s. A survey
of the work up to 1950 is given by Ficken [70]. The use of continuation by parameter differentiation as a numerical method
is attributed to Davidenko [54]. Key papers of this early period are by Freudenstein and Roth [72], Deist and Sefor [58] and
Meyer [114], as well as the survey by Wasserstrom [191]. This early history is covered by Wacker [190].

REMARK 1.9

Arclength continuation methods in the mathematical literature are generally attributed to Haselgrove [87] and Klopfestein
[99] although these papers remained largely unnoticed until the late 1970s. Important contributions to the mathematical
treatment are by Abbott [1], Anselone and Moore [8], Avila [14], Brent [31], Boggs [29], Branin [30], Broyden [35,36],
Cassel [43], Chow et. al. [45], Crandall and Rabinowitz [47], Georg [77,78], Keller and coworkers [44,56,57,93–96],
Matthies and Strang [111], Moore [115,116], Pönish [139,140], Rheinboldt and coworkers [59,113, 149–150], Watson
[192] and Werner and Spence [195]. Of these, key contributions in terms of subsequent influence are [45,94,149]. For
surveys and edited proceedings see Allgower [2,3], Byrne and Hall [42], Küpper [104,105], Rall [145], Wacker [190], and
references therein. Textbooks and monographs dealing with nonlinear equation solving include Chow and Hale [46], Dennis
and Schnabel [61], Kubı́ček and Hlaváček [102], Kubı́ček and Marek [103], Ortega and Rheinboldt [128], Rabinowitz [143],
Rall [144], Rheinboldt [153], and Seydel [164]. Of these, the book by Ortega and Rheinboldt remains a classic and an
invaluable source to essentially all mathematically oriented work done prior to 1970. The book by Seydel contains material
on treatment of conventional and Hopf bifurcations not readily available elsewhere. Nonlinear equation solving is interwined
with the larger subject of optimization and mathematical programming; for the latter the textbooks by Gill, Murray and
Wright [81] and Fletcher [71] are highly recommended.

§1.6.2 Rough Nonlinearities

Rough nonlinearities are characterized by discontinuous field relations, usually involving inequality constraints.
Examples: flow-rule plasticity, contact, friction. The local response is nonsmooth.
Solution techniques for these problems are in a less satisfactory state, and case-by-case consideration is called
for. The local and overall responses are generally path-dependent, an attribute that forces the past response
history to be taken into account.
The key difficulty is that conventional solution procedures based on Taylor expansions or similar differential
forms may fail, because such Taylor expansions need not exist! An encompassing mathematical treatment
is lacking, and consequently problem-dependent handling is presently the rule. For this class of problems
incremental methods, as opposed to incremental-iterative methods, still dominate.

REMARK 1.10

Earliest publications on computational plasticity using finite element methods are by Gallagher et. al. [74], Argyris [9,10],
Marcal [108], Pope [138] and Felippa [62]. By now there is an enormous literature on the numerical treatment of inelastic
processes, especially plasticity and creep. Fortunately the survey by Bushnell [40], although focusing on plastic buckling,
contains over 300 references that collectively embody most of the English-speaking work prior to 1980. Other important
surveys are by Armen [13] and Willam [197]. For contact problems, see Oden [126], Bathe and Chaudhary [19], Kikuchi
and Oden [97,98], Simo et. al. [167] Stein et. al. [173], Nour-Omid and Wriggers [121], and references therein.

1–13
Chapter 1: OVERVIEW 1–14

§1.6.3 Hybrid Approach

What does an analyst do when faced with an unfamiliar nonlinear problem? If the problem falls into the
smooth-nonlinear type, there is no need to panic. Robust and efficient methods are available. Even if the
whizziest methods are not implemented into one’s favorite computer program, there is a wealth of theory and
practice available for trouble-shooting.
But what if the problem include rough nonlinearities? A time-honored general strategy is divide and conquer.
More specifically, two powerful techniques are frequently available: splitting and nesting.
Splitting can be used if the nonlinearities can be separated in an additive form:

Smooth + Rough

This separation is usually done at the force level. Then the smooth-nonlinear term is treated by conventional
techniques whereas the rough-nonlinear term is treated by special techniques. This scheme can be particularly
effective when the rough nonlinearity is localized, for example in contact and impact problems.
Nesting may be used when a simple additive separation is not available. This is best illustrated by an actual
example. In the early 1970s, some authors argued that Newton’s method would be useless for finite-deflection
elastoplasticity, as no unique Jacobian exists in plastic regions on account of loading/unloading switches. The
argument was compelling but turned out to be a false alarm. The problem was eventually solved by “nesting”
geometric nonlinearities within the material nonlinearity, as illustrated in Fig. 1.1.
In the inner equilibrium loop the material law is “frozen”, which makes the highly effective Newton-type meth-
ods applicable. The non-conservative material behavior is treated in an outer loop where material properties
and constitutive variables are updated in an incremental or sub-incremental manner.
Another application of nesting comes in the global function approach (also called Rayleigh-Ritz or reduced-
basis approach), which is presently pursued by several investigators. The key idea is to try to describe the
overall response behavior by a few parameters, which are amplitudes of globally defined functions. The small
nonlinear system for the global parameters is solved in an inner loop, while an external loop involving residual
calculations over the detailed finite element model is executed occasionally.
Despite its inherent implementation complexity, the global function approach appears cost-effective for smooth,
path-independent nonlinear systems. This is especially so when expensive parametric studies are involved, as
in structural optimization under nonlinear stability constraints.

REMARK 1.11
For geometric-material nesting and subincremental techniques see Bushnell [38–41], and references therein. The global-
function approach in its modern form was presented by Almroth, Stern and Brogan [7] and pursued by Noor and coworkers
under the name of reduced-basis technique; see Noor and Peters [119] and Noor [120] as well as the chapter by Noor in this
volume. For perturbation techniques see the survey by Gallagher [75].

§1.6.4 Summary of Present Status

Solution techniques for smooth nonlinearities are in a fairly satisfactory state. Although further refinements
in the area of traversing bifurcation points can be expected, incremental-iterative methods implemented with
general increment control appear to be as reliable as an engineer user may reasonably expect.
For rough nonlinearities, case-by-case handling is still necessary in view of the lack of general theories and
implementation procedures. Separation or nesting of nonlinearities, when applicable, can lead to signifi-
cant gains in efficiency and reliability, but at the cost of programming complexity and problem-dependent
implementations.

1–14