2-Plasticiy Theory and Application - A. MENDELSON - 3

MACMILLAN SERlE~ IN
APPLIED MECHANICS
F R E D LANDIS, EDITOR
The Analysis of Stress and Deformation

by George W. Housner and Thad Vreeland, Jr.
Analytical Methods in Vibrations by Leonard Meirovitch
Continuum Mechanics by Walter Jaunzemis
Plasticity: Theory and Application by Alexander Mendelson
Statics of Deformable Bodies by Nils 0. Myklestad
ALEXANDER MENDELSON
National Aeronautics and Space Administration
PLASTICITY:
Theory and
Application
MACMILLAN COMPANY, NEW YORK
COLLIER-MACMILLAN LIMITED, LONDON

PREFACE
With the advent of the jet age followed closely by the space age, the theory
of plasticity has been brought forcibly into the forefront of engineering appli-
cation and design. Modern aircraft, missiles, and space vehicles must be
designed on the basis of minimum weight, which invariably means designing
into the plastic range to obtain maximum load to weight ratios. Moreover,
the facts of economic life have made the saving of material and more efficient
design a necessity for even the more earthbound industrial applications.
This book is the result of the author's teaching for several years of a
graduate course in plasticity for engineers at Case Institute of Technology.
It was soon realized that although a number of excellent books on plasticity
were available, none of them adequately met the requirements of the course.
The available books were either too theoretical and mathematical for the
average engineer and designer, or their main emphasis was placed on problems
of large plastic deformations such as are encountered in metal-forming
processes. Very little has been published in textbook form on the most
,important class of elastoplastic problems, in which the plastic strains are
of the same order of magnitude as the elastic strains, which are of such prime
'concc~rn to today's engineer. Furthermore, where such problems are treated,
usual assumptions of perfect plasticity are used, no attempt being made
take into account the strain-hardening properties of real materials.
A set of mimeographed notes was prepared which included the basic theory
placed primary emphasis on the solution of elastoplastic problems for
vii
viii Preface
materials with strain hardening. In particular, it was emphasized that with

the present availability of high-speed computing facilities, many of the
simplifying assumptions hitherto commonplace in plasticity calculations were
no longer necessary. The present book is based on these notes.
Following a brief introduction, Chapter 2 discusses some of the basic
experiments concerning the elastoplastic behavior of metals. Chapters 3 and 4
describe the basic properties of the stress and strain tensors. Tensor notation
is introduced and is frequently used together with the longhand notation,
but a knowledge of tensor properties is not needed. Chapter 5 describes
briefly the elastic stress-strain relations. Chapter 6 discusses the various
yield criteria and their experimental verification. In Chapter 7 the plasticity
flow rules, or stress-strain relations, are derived and discussed, including a
new set of equations which relate plastic strain increments to total strains
rather than stresses.
A series of practical problems for both ideally plastic and strain-hardening
materials is presented in Chapters 8 through 11. Chapter 8 deals with prob-
lems of spheres and cylinders. Chapter 9 is devoted entirely to the powerful .
method of successive elastic solutions, by means of which a large class of
otherwise intractable problems can be solved. First introduced by Ilyushin
some twenty years ago, this method has not yet gained wide acceptance in
this country and, to the author's knowledge, is not even mentioned in any
other current book in the English language.
Chapter 10 discusses plate problems, both for the plane stress and plane
strain cases. Chapter 11 gives the general solution to the elastoplastic torsion
problem. The theory of the slip-line field as appiied to the plane strain problem
of plastic-rigid materials is then presented in Chapter 12, and limit analysis
of framed structures in Chapter 13. Chapter 14 discusses problems of creep
at elevated temperatures and shows how the previously discussed plasticity
methods can be applied to creep problems. It is realized that to treat these
last three subjects adequately would require a book for each of them. It is
hoped, however, that sufficient information is furnished herein to provide
the reader with a worthwhile introduction to, and basic understanding of,
these subjects.
In the author's experience the material included can be covered adequately
in a one-semester graduate course. Chapters 3 through 5 may be omitted by
those familiar with basic elasticity theory. Sections 6.3, 6.5, 7.6, 7.8, 12.6,
12.8, and 13.6 may also be omitted on a first reading or if time is short. It is
hoped that this book will be found useful as a graduate text and as an aid to
engineers and designers faced with the problem of designing into the plastic
range.
The author would like to acknowledge his appreciation to his colleagues
Preface ix
and co-workers at the Lewis Research Center of the National Aeronautics

and Space Administration whose helpful discussions and combined efforts
over the years are in no small measure responsible for much that is in this
book. In particular, thanks are due to S. S. Manson, M. H. Hirschberg, and
E. Roberts. A special debt of gratitude is owed to Professor R. H. Scanlan
for his continuous encouragement and interest. But above all I offer humble
thanks to Him who " ... favours man with knowledge and enables him to
achieve understanding." 31"::l'?Wm
A.M.
Cleveland
CONTENTS
Chapter 1. Introduction 1
Chapter 2. Basic Experiments 4

2-1 TENSILE TEST 4
2-2 TRUE STRESS-STRAIN CURVE 7
2-3 COMPRESSION TEST AND THE BAUSCHINGER EFFECT.
ANISOTROPY 13
2-4 EFFECTS OF STRAIN RATE AND TEMPERATURE 15
2-5 INFLUENCE OF HYDROSTATIC PRESSURE. INCOMPRESSIBILITY 16
2-6 IDEALIZATION OF THE STRESS-STRAIN CURVE.
DYNAMIC AND KINEMATIC MODELS 16
2-7 EMPIRICAL EQUATIONS FOR STRESS-STRAIN CURVES 20
3. The Stress Tensor 24

3-1 TENSOR NoTATION 25
3-2 STRESS AT A POINT 27
3-3 PRINCIPAL STRESSES. STRESS INVARIANTS 30
3.:..4 MAXIMUM AND OCTAHEDRAL SHEAR STRESSES 34
3-5 MOHR'S DIAGRAM 37
3-6 STRESS DEVIATOR TENSOR 39
3-7 PURE SHEAR 41
xi
xii Contents
Chapter 4. The Strain Tensor 44

4-1 STRAIN AT A POINT 44
4-2 PHYSICAL INTERPRETATION OF STRAIN COMPONENTS 48
4-3 FINITE DEFORMATIONS 51
4-4 PRINCIPAL STRAINS. STRAIN INVARIANTS 53
4-5 MAXIMUM AND OCTAHEDRAL SHEAR STRAINS 55
4-6 STRAIN DEVIATOR TENSOR 58
4-7 COMPATIBILITY OF STRAINS 59
Chapter 5. Elastic Stress-Strain Relations 64

5-1 EQUATIONS OF ELASTICITY 64
5-2 ELASTIC STRAIN ENERGY FUNCTIONS 67
5-3 SOLUTION OF ELASTIC PROBLEMS 68
Chapter 6. Criteria for Yielding 70

6-1 EXAMPLES OF MuLTIAXIAL STRESS 70
6-2 EXAMPLES OF YIELD CRITERIA 71
6-3 YIELD SuRFACE. HAIGH-WESTERGAARD STRESS SPACE 79
6-4 LODE'S STRESS PARAMETER. EXPERIMENTAL VERIFICATION OF
YIELD CRITERIA 88
6-5 SUBSEQUENT YIELD SURFACES. LOADING AND UNLOADING 92
Chapter 7. Plastic Stress-Strain Relations 98

7-1 DISTINCTION BETWEEN ELASTIC AND PLASTIC
STRESS-STRAIN RELATIONS 98
7-2 PRANDTL-REUSS EQUATIONS 100
7-3 PLASTIC WORK. TWO MEASURES OF WoRK HARDENING 104
7-4 STRESS-STRAIN RELATIONS BASED ON TRESCA CRITERION 108
7-5 EXPERIMENTAL VERIFICATION OF PRANDTL-REUSS EQUATIONS 109
7-6 GENERAL DERIVATION OF PLASTIC STRESS-STRAIN RELATIONS 110
7-7 INCREMENTAL AND DEFORMATION THEORIES 119
7-8 CONVEXITY OF YIELD SURFACE. SINGULAR POINTS
7-9 PLASTIC STRAIN-TOTAL STRAIN PLASTICITY RELATIONS
7-10 COMPLETE STRESS-STRAIN RELATIONS. SUMMARY
Chapter 8. Elastoplastic Problems of Spheres and Cylinders
8-1 GENERAL RELATIONS

8-2 THICK HOLLOW SPHERE WITH INTERNAL PRESSURE AND
THERMAL LOADING
Contents xiii
8-3 HOLLOW SPHERE. SPREAD OF PLASTIC ZONE.

PRESSURE LOADING ONLY 141
8-4 HOLLOW SPHERE. RESIDUAL STRESSES. PRESSURE LOADING 145
8-5 HOLLOW SPHERE. THERMAL LOADING ONLY 148
8-6 HOLLOW SPHERE OF STRAIN-HARDENING MATERIAL 150
8-7 PLASTIC FLOW IN THICK-WALLED TUBES 156
Chapter 9. The Method of Successive Elastic Solutions 164

9-1 GENERAL DESCRIPTION OF THE METHOD 164
9-2 THIN FLAT PLATE 172
9-3 THIN CIRCULAR SHELL 183
9-4 LONG SOLID CYLINDER 193
9-5 ROTATING DISK WITH TEMPERATURE GRADIENT 197
9-6 CIRCULAR HOLE IN UNIFORMLY STRESSED INFINITE PLATE 208
Chapter 10. The Plane Elastoplastic Problem 213

10-1 GENERAL RELATIONS 213
10-2 ELASTOPLASTIC THERMAL PROBLEM FOR A FINITE PLATE 218
10-3 ELASTOPLASTIC PROBLEM OF THE INFINITE PLATE WITH A
CRACK 223
10-4 STRAIN-INVARIANCE PRINCIPLE 230
Chapter 11. The Torsion Problem 234

11-1 TORSION OF PRISMATIC BAR. GENERAL RELATIONS 234
11-2 ELASTICITY SOLUTIONS 240
11-3 MEMBRANE ANALOGY 245
11-4 ELASTOPLASTIC TORSION. PERFECT PLASTICITY 246
11-5 ELASTOPLASTIC TORSION WITH STRAIN HARDENING 248
11-6 BAR WITH RECTANGULAR CROSS SECTION 250
11-7 BAR WITH CIRCULAR CROSS SECTION 253
12. The Slip-Line Field 260

12-1 PLANE STRAIN PROBLEM OF A RIGID PERFECTLY PLASTIC
MATERIAL 260
12-2 VELOCITY EQUATIONS 266
12-3 GEOMETRY OF THE SLIP-LINE FIELD 268
12-4 SOME SIMPLE EXAMPLES 272
12-5 NUMERICAL SOLUTIONS OF BOUNDARY-VALUE PROBLEMS 276
12-6 GEOMETRIC CoNSTRUCTION oF SLIP-LINE FIELDS 279
12-7 COMPLETE SOLUTIONS. UPPER AND LOWER BOUNDS 284
12-8 SLIP LINES AS CHARACTERISTICS 285
Contents
xiv
300
Chapter 13. Limit Analysis
300
13-1 DESIGN OF STRUCTURES
301
13-2 SIMPLE TRUSS
305
13-3 PURE BENDING OF BEAMS
13-4 BEAMS AND FRAMES WITH CoNCENTRATED LOADS
307
13-5 THEOREMS OF LIMIT ANALYSIS
312
13-6 METHOD OF SUPERPOSITION OF MECHANISMS
318
323
13-7 LIMIT DESIGN
327
Chapter 14. Creep
327
14-1 BASIC CONCEPTS
331
14-2 MuLTIDIMENSIONAL PROBLEMS
333
14-3 UNIAXIAL CREEP IN INFINITE STRIP
335
14-4 CREEP IN ROTATING DISKS
347
Index
CHAPTER 1
INTRODUCTION
The history of plasticity as a science began in 1864 when Tresca [1] pub-
lished his results on punching and extrusion experiments and formulated his
famous yield criterion. A few years later, using Tresca's results, Saint-Venant
[2] and Levy [3] laid some of the foundations of the modern theory of plasti-
city. For the next 75 years progress was slow and spotty, although important
contributions were made by von Mises [4], Hencky [5], Prandtl [6], and others.
It is only since approximately 1945 that a unified theory began to emerge.
Since that time, concentrated efforts by many researchers have produced a
voluminous literature which is growing at a rapid rate. Brief but excellent
historical sketches are furnished by Hill [7] and Westergaard [8].
The theories of plasticity fall into two categories: physical theories and
mathematical theories. The physical theories seek to explain why metals flow
plastically. Looking at materials from a microscopic viewpoint, an attempt
made to determine what happens to the atoms, crystals, and grains of a
when plastic flow occurs. The mathematical theories, on the other
are phenomenological in nature and attempt to formalize and put into
form the results of macroscopic experiments, without probing very
into their physical basis. The eventual hope, of course, is for a merger
two approaches into one unified theory of plasticity which will both
the material behavior and provide the engineer and scientist with the
tools for practical application. The present treatise is concerned
the second of these categories, i.e., the mathematical theories of plasticity
1
2 Introduction [Ch. 1
and their application, as distinct from the physical theories. The latter belong
to the realm of the metal physicist or solid-state physicist.
We start by defining roughly and intuitively what is meant by a metal
flowing plastically. If one takes a thin strip of a metal such as aluminum and
clamps one end and applies a bending force to the other end, the end of the
strip will deflect. Upon removal of this force, if this force is not too large, the
end of the strip will spring back to its original position, and there will be no
apparent permanent deformation. If a sufficiently large load is applied to the
end, the end will not spring back all the way upon the removal of the load
but will remain permanently deformed, and we say that plastic flow has
occurred. Our objective in this case will not be to determine why the perma-
nent deformation took place but to describe what has happened in terms of
stresses, strains, and loads. Solutions of this particular problem can be found,
for example, in references [9] and [10].
In short, plasticity is the behavior of solid bodies in which they deform
permanently under the action of external loads, whereas elasticity is the
behavior of solid bodies in which they return to their original shape when the
external forces are removed. Actually, however, the elastic body is an ideali-
zation, because all bodies exhibit more or less plastic behavior even at the
smallest loads. For the so-called elastic body, however, this permanent
deformation is so small as to be practically not measurable, if the loads are
sufficiently small. Plasticity theory thus concerns itself with situations in
which the loads are sufficiently large so that measurable amounts of perma-
nent deformation occur. It should.further be noted that plastic deformation
is independent of the time under load. Time-dependent deformations are
discussed briefly in Section 2.4 and in Chapter 14.
The theory of plasticity can conveniently be divided into two ranges. At
one end are metal-forming processes such as forging, extrusion, drawing,
rolling, etc., which involve very large plastic strains and deformations. For
these types of problems the elastic strains can usually be neglected and the
material can be assumed to be perfectly plastic. At the other end of the scale
are a host of problems involving small plastic strains on the order of the
elastic strains. These types of problems are of prime importance to the
structural and machine designer. With the great premium currently placed
on the saving of weight in aircraft, missile, and space applications, the de-
signer can no longer use large factors of safety and "beef up" his design. He
must design for maximum load to weight ratio, and this inevitably means
designing into the plastic range. Even in more prosaic industrial applications
the competitive market is forcing the application of more efficient design.
In this book emphasis will be placed primarily on the second type of prob-
lem, i.e., the elastoplastic problems, where the plastic strains are of the same
Ch. 1] Introduction 3
order of magnitude as the elastic strains. Problems of large plastic deforma-

tions will be treated only briefly, as will problems of creep and limit design.
In Chapter 2 some simple experiments to determine several basic facts
about the elastoplastic behavior of metals will be discussed.
References
1. H. Tresca, Sur l'ecoulement des corps solids soumis a de fortes pression,
Compt. Rend., 59, 1864, p. 754.
2. B. de Saint-Venant, Memoire sur l'etablissement des equations differen-
tielles des mouvements interieurs operes dans les corps solides ductiles au
dela des limites oil l'elasticite pourrait les ramener a leur premier etat,
Compt. Rend., 70, 1870, pp. 473-480.
3. M. Levy, Memoire sur les equations generales des mouvements interieurs
des corps solides ductiles au dela des limites oil l'elasticite pourrait les
ramener a leur premier etat, Compt. Rend., 70, 1870, pp. 1323-1325.
4. R. von Mises, Mechanik der festen Koerper im plastisch deformablen
Zustant, Goettinger Nachr., Math.-Phys. Kl., 1913, pp. 582-592.
5. H. Hencky, Zur Theorie plastischer Deformationen und der hierdurch im
Material hervorgerufenen Nebenspannungen, Proceedings of the 1st Inter-
national Congress on Applied Mechanics, Delft, Technische Boekhandel en
Druckerij, J. Waltman, Jr., 1925, pp. 312-317.
6. L. Prandtl, Spannungsverteilung in plastischen Koerpern, Proceedings of the
1st International Congress on Applied Mechanics, Delft, 1924, pp. 43-54.
7. R. Hill, The Mathematical Theory of Plasticity, Oxford University Press,
London, 1950.
8. H. M. Westergaard, Theory of Elasticity and Plasticity, Harvard University
Press, Cambridge, 1952.
B. W. Shaffer and R. N. House, The Elastic-Plastic Stress Distribution
Within a Wide Curved Bar Subjected to Pure Bending. J. Appl. Mech., 22,
No. 3, 1955, pp. 305-310.
B. W. Shaffer and R. N. House, Displacements in a Wide Curved Bar,
J. Appl. Mech., 24, No. 3, 1957, pp. 447-452.
CHAPTER 2
BASIC
EXPERIMENTS
In this chapter the results of some basic experiments on the behavior of

metals is presented. The stress-strain curve in tension, one of the basic
ingredients necessary in applying plasticity theory, is described in some detail.
The effects of reverse loading, strain rate, temperature, and hydrostatic pres-
sure are briefly discussed. Idealizations of the stress-strain curve and various
models of material behavior are described.
2-1 TENSILE TEST
The simplest and most common experiment, as well as the most important,
is the standard tensile test. A cylindrical test specimen such as shown in
Figure 2.1.1 is inserted into the tensile machine, the load is increased, and the
readings of the load, the extension of the gage length inscribed on the speci-
men, and/or the decrease in diameter are recorded. A typical load extension
diagram is shown in Figure 2.1.2.
The nominal stress, defined as the load divided by the original cross-
sectional area, is plotted against the conventional or engineering strain, defined
as the increase in length per unit original length. Nominal stress is represented
by
p
a=-
n Ao
4
Sec. 2-1] Tensile Test 5
p
FIGURE 2.1.1 Tensile specimen.
and conventional strain by

l - 10
e=-- (2.1.2)
lo
Initially the relation between stress and strain is essentially linear. This linear
part of the curve extends up to the point A, which is called the proportional
limit. It is in this range that the linear theory of elasticity, using Hook's law,
is valid. Upon further increase of the load, the strain no longer increases
linearly with stress, but the material still remains elastic; i.e., upon removal
of the load the specimen returns to its original length. This condition will
FIGURE 2.1.2 Conventional stress-strain curve.

6 Basic Experiments [Ch. Z
prevail until some point B, called the elastic limit, or yield point, is reached.
In most materials there is very little difference between the proportional limit
A and the elastic limit B. For our purpose, we shall consider them to be the
same. Furthermore, the values of these points depend on the sensitivity of the
measuring instruments. For some materials the yield point is so poorly defined
that it is arbitrarily taken to be at some fixed value of permanent strain, such
as 0.2 per cent. The stress at this point is usually called the offset yield strength,
or the proof strength. Beyond the elastic limit, permanent deformation, called
plastic deformation, takes place. The strain at the elastic limit is of the order
of magnitude of 0.001, or 0.1 per cent. As the load is increased beyond the
elastic limit, the strain increases at a greater rate. However, the specimen will
not deform further unless the load is increased. This condition is called work
hardening, or strain hardening. The stress required for further plastic flow is
called flow stress. Finally a point is reached, C, where the load is a maximum.
Beyond this point, called the point of maximum load, or point of instability, the
specimen" necks down" rapidly and fractures at D. Beyond C a complicated
triaxial state of stress exists. The point C therefore represents the limit of the
useful part of the tensile test as far as plasticity theory is concerned. The stress
at the maximum load point Cis called the tensile strength, or ultimate stress.
If at any point between the elastic limit B and the maximum load point C
the load is removed, unloading will take place along a line parallel to the
elastic line, as shown in the figure by B' C'. Part of the strain is thus recovered
and part remains permanently. The total strain can therefore be considered
as being made up of two parts, ee, the elastic component, and eP, the plastic
component:
(2.1.3)
Upon reloading, the unloading line, B'C', is retraced with very minor
deviations. Actually a very thin hysteresis loop is formed, which is usually
neglected. Plastic flow does not start again until the point B' is reached. With
further loading, the stress-strain curve is continued along B'C as if no
unloading had occurred. Point B' can thus be considered as a new yield point
for the strain-hardened material.
A few materials, such as annealed mild steel, exhibit a sharp drop in yield
after the upper yield point B is reached, as shown by the dashed line. The
specimen will then extend at approximately constant load to a strain of about
10 times the initial yield before the load will start increasing again as the
material begins to work harden. The flat portion of the curve, called the
lower yield, actually represents an average of a series of unstable jumps
between the upper and lower yields caused by the propagation of Luder bands
across the specimen. The upper yield point is very sensitive to small bendin~
Sec. 2-2] True Stress-Strain Curve 7
stresses or to irregularities in the specimen as well as to the rate of loading.

Furthermore, very little plastic flow takes place at the upper yield point. The
lower yield point should therefore always be used for design purposes and for
plastic flow calculations.
2-2 TRUE STRESS-STRAIN CURVE
We have discussed the plot of the nominal stress versus the conventional
strain. It is evident, however, that this nominal stress is not the true stress
acting in the specimen, since the cross-sectional area of the specimen is
decreasing with load. At stresses up to and near the yield, this distinction is
of no importance. At higher stresses and strains this difference becomes
important. The true stress can readily be obtained from the nominal stress
as follows. If small changes in volume are neglected, i.e., the material is
assumed to be incompressible, then
Aolo =AI
where A 0 and /0 are the original cross-sectional area and gage length and
A and l are the current values. If P is the load, then the true stress a is
P PI
a=-=-
A A 0 l0
The nominal stress an is an = P/A 0 and the conventional strain is

e = (l/10 ) - I. Therefore,
(2.2.1)
In a somewhat similar fashion, one recognizes that the conventional or

engineering strain cannot be completely correct, since it is based on initial
length, whereas the length is continuously changing. A different definition
was therefore introduced by Ludwik [1] based on the changing length. Thus
increment of strain for a given length is defined as
dl
de= T (2.2.2)
the total strain in going from some initial length /0 to the length l is
(2.2.3)
8 Basic Experiments [Ch. 2
e is called the natural, logarithmic, or true strain and it represents a sort of

average strain in going from the length 10 to the length /. Its relation to the
conventional strain is readily found, since l/10 = 1 + e:
e = ln (1 + e) (2.2.4)
For small strains the two are practically identical, and for most problems
considered the conventional strain will be used. The natural strain, however,
has several advantages. Natural strains are additive, but conventional strains
are not. Second, if a ductile material is tested in compression and in tension,
the true-stress versus true-strain curves are almost identical, whereas they are
quite different if conventional strain is used. Finally, the incompressibility
condition to be used later becomes simply
(2.2.5)
whereas in terms of conventional strains it is
(1 + el)(1 + e2 )(1 + e3 ) - 1 =0 (2.2.6)

which reduces to
only in the case of small strains.

If a plot is now made of true stress versus true strain for the tensile test
previously described, the curve will be essentially the same up to and slightly
above the yield point. Beyond this point the two types of plots will diverge.
The true stress will always increase until the rupture point and does not have
a maximum at the point where the load starts dropping. The true stress at
the point of maximum load can be found as follows. Since
P = aA
at the point of maximum load
dP = a dA + A da = 0
da
or
a
Also A 0 l0 =AI
Ad!+ IdA= 0
dA dl
or A -~
da dl
Hence -=-=de
a l
da
or -=a
de
da a
(2.2.7)
de = 1 +e
On a plot of a versus e, the value of a at which the load is a maximum
occurs where the slope is equal to the stress; i.e., one must draw a tangent to
that point of the curve for which the subtangent is equal to 1, as shown in
Figure 2.2.1. Discussions of the stress-strain curve and the strain distributions
FIGURE 2.2.1 True stress-strain curve.
in the neck of a tensile specimen after necking has started can be found in
"'""·l'-'11~·"~ [2] and [3].
Alternatively, the true stress-strain curve can be obtained by measuring the
ctttamt>tr.al strain rather than the longitudinal strain, provided the tensile sped-
has a circular section. Thus, if en is the strain in the diametral direction,
D- D 0
en= Do (2.2.8)
D 0 is the initial diameter and Dis the diameter at the true stress, a. The
diametral strain is
(2.2.9)
and from equation (2.2.5) the longitudinal true strain is
e- Do
= - 2-Bv = 21nD (2.2.10)
The true strain at any load can therefore be determined by measuring the
change in diameter of the specimen.
From equation (2.2.10) it is seen that the true strain can also be written
A
e = ln_Q (2.2.11)
A
The quantity on the right of equation (2.2.11) is called the true reduction in
area. Equation (2.2.11) states that the true strain is equal to the true reduction
in area.
Figure 2.2.2 (from reference [4]) shows the true stress-strain curves for a
320
"'r0
)(
0.80 1.00
Natural Strain- E
FIGURE 2.2.2 True stress-strain curves for several materials.

1,000r------------------------,
,.., SAE 3140 240,000 psi

'~
><
'iii
0.
b" Hard-drown copper
(/)
100
(/)
~
+-
(/)
Q)
::l
~
o Necking
x Fracture
1.0
Natural strain 8
FIGURE 2.2.3 True stress-strain curves on log-log coordinates.
variety of materials. The ends of the curves represent the points of fracture
and the circle on each curve represents the maximum load point or instability
point for that curve. To show the complete curves to fracture, the abscissa is
such that the elastic parts of the curves are too small to be seen. These curves
are also shown replotted on log-log coordinates in Figure 2.2.3. Note that
most of the curves appear as straight lines on this log-log plot. This indicates
that they can be represented by an equation of the form
a= Aen (2.2.12)
where A and n are material constants with n the slope of the curve when
plotted on log-log coordinates. A is called the strength coefficient and n is
called the strain-hardening exponent.
It follows from equation (2.2. 7) that for a material which behaves according
equation (2.2.12), the true strain at the point of maximum load is given by
e=n (2.2.13)
simple relation (2.2.13) has been found useful in fracture studies. It also
a simple method for determining the instability point on the true
curve .
.12q1uatton (2.2.12) will, of course, not fit all materials, nor will it be valid
all strains or very large strains. However, Marin [5] has studied 31
different materials and found that the average deviation between the theoreti-
cal values of e as given by equation (2.2.13) and the actual values was 2 per
cent.
A single quantity which represents the ability of a material to deform
plastically is the ductility of the material. The most common measure of
ductility is the per cent elongation in the tensile test, i.e., the per cent strain at
fracture. Thus, if 11 is the gage length at fracture and /0 is the initial gage
length, then the per cent elongation is
I,- lo
e1 = - - X 100 (2.2.14)
lo
Together with the per cent elongation as given by equation (2.2.14), one
must also specify the initial gage length /0 , since the per cent elongation will
depend on the gage length used. This is due to the fact that once the instability
point is reached and necking starts, most of the deformation occurs in the
smallest cross section, with only a relatively small amount of additional de-
formation occurring throughout the rest of the gage length. The longer the
gage length used, the smaller the per cent elongation will be. The ductility is
therefore reported as the per cent elongation for a given gage length.
A better measure for ductility, however, is the true strain at fracture:
(2.2.15)
Alternatively, equation (2.2.15) can be written in terms of the reduction in area

at fracture. From equation (2.2.11) it follows that
where A 0 is the initial area and A 1 is the area at fracture. As mentioned pre-
viously, ductility is an important property which measures the ability of a
material to deform. A material with low ductility cannot deform very much
under load and will behave in a brittle fashion. Unexpectedly high loads may
cause such a material to fracture, whereas a material with high ductility will
deform under similar loads without fracturing. A cyclic load above the
will cause a low-ductility material to fail in relatively few cycles, whereas
high-ductility material will fail after a much larger number of cycles (at
for very low cycle fatigue). In metal-forming processes such as rolling,
ing, forging, etc., a sufficient amount of ductility is needed to prevent
during the forming process.
Sec. 2-3] Compression Test and the Bauschinger Effect 13
2-3 COMPRESSION TEST AND THE

BAUSCHINGER EFFECT. ANISOTROPY
If instead of a tensile test one runs a compression test and plots nominal
stress against conventional strain, a different curve will be obtained than for
the tensile test. However, if the true stress is plotted against the true strain,
practically identical curves are usually obtained. The yield points in tension
and compression will, for example, generally be the same. If, however, a
metal is first deformed by uniform tension and the load is removed and the
specimen is reloaded in compression, the yield point obtained in compression
will be considerably less than the initial yield in tension. This has been ex-
plained as being the result of the residual stresses left in the material due to
the tensile deformations [6]. A perhaps better explanation is based on the
anisotropy of the dislocation field produced by loading [7]. This effect is
called the Bauschinger effect, and is present whenever there is a reversal of the
stress field. The Bauschinger effect is very important in cyclic plasticity
studies. Unfortunately, however, it enormously complicates the problem and
is therefore usually ignored.
There are several simplified models used to describe the Bauschinger effect.
These are illustrated in Figure 2.3.1 (from reference [8]). At one extreme it is
assumed that the elastic unloading range will be double the initial yield stress.
If the initial yield stress in tension is a 0 , then the specimen will yield in com-
pression after being stressed in tension to a = a 1 when
This is shown as path ABCDE in Figure 2.3.1. According to this theory, then,
the total elastic range of the material remains constant, the initial compressive
yield being reduced by the same amount as the tensile yield is raised.
At the other extreme there is isotropic hardening. This theory assumes
the mechanism that produces hardening acts equally in tension and
cornp1res1dOJ11. Thus compressive yielding will occur when
shown by the path ABCFG. This is the simplest of the theories to apply and
the one most frequently used.
'£e:tw1~P.n these theories there is a theory which assumes that the tensile and
mJ>re1>Sl,,e yields are independent of each other. The compressive yield
is independent of the amount of tensile hardening and remains at
a= - a0
G
FIGURE 2.3.1 Theories for Bauschinger effect.
as shown by ABCHI. Actually experiments indicate that the compressive

yield stress usually lies between points Hand D of Figure 2.3.1, such as at J.
It should be noted that in this figure the curves after yielding are shown as a
set of parallel lines for simplicity. Actually a real stress-strain curve will show
continuous curvature and varying slope after yielding when the load is
reversed.
As an allied effect to the Bauschinger effect, any initial isotropy which is
present is usually destroyed upon loading into the plastic range; i.e., if
originally the tensile yield point was the same in all directions, it will no
longer be so. Both the compressive and tensile yield values are changed in all
directions by plastic yielding in one direction. Thus plastic deformation is
anisotropic. For example, cold-rolled sheet has markedly different properties
in the thickness direction than in the plane of the sheet, and usually somewhat
different yield points in the rolling than in the transverse direction.
We see that the material may have initial anisotropy due to the manufac-
turing process, and it may also develop anisotropy due to plastic yielding.
For small plastic strains the second effect is probably not too important. As
for the first effect, the material being used can be tested for anisotropy. If a
large amount of anisotropy is found, a much more complicated
theory of plasticity may have to be used.
Sec. 2-4] Effects of Strain Rate and Temperature 15
2-4 EFFECTS OF STRAIN RATE AND

TEMPERATURE
Tests on the effect of the rate of straining and of temperature on the prop-
erties of mild steel were carried out by Manjoine [9], among others. The effect
of increasing the strain rate is generally to increase the tensile yield, as shown
in Figure 2.4.1. For materials with a lower yield, such as mild steel, the stress-
~---10 3 sec-1
~----10- 2 sec- 1
e
FIGURE 2.4.1 Effect of strain rate.
strain curve may approach that of a perfectly plastic material. For other
materials the reverse will be true, and the strain hardening will increase with
strain rate [16]. These effects are important in some metal-forming processes
which are performed at very high strain rates. These types of processes will
not be discussed in this text.
Temperature has a very important effect on metal properties. At very low
temperatures metals which are very ductile can become very brittle. This is
illustrated in Figure 2.4.2 (from reference [10]). The temperature at which the
ductility changes so rapidly is called the transition temperature. Such strong
temperature and strain-rate effects occur more generally in metals with
body-centered-cubic structures.
100
0~
0Q) 80
0 100min-1
.s 60
c
0
:;::: 40
u
::J
"0
Q)
20
0:
-200 -150 -100 -50 0 oc

FIGURE 2.4.2 Effect of temperature.
At the other end of the time and temperature scales is the phenomenon of
creep. Creep is a continuous deformation with time under constant load and
occurs primarily at high temperatures, although some metals, e.g., lead, will
creep at room temperature. Although it is questionable whether plasticity
theories can be applied to the creep phenomenon, it is the usual practice to
do so and, in Chapter 14, we shall describe how this is done. A typical set of
creep curves is shown in Figure 2.4.3.
Time
FIGURE 2.4.3 Creep curves.
2-5 INFLUENCE OF HYDROSTATIC PRESSURE.

INCOMPRESSIBILITY
Bridgman, in a series of classical experiments [11, 12] in which he carried

out tensile tests under conditions of hydrostatic pressures up to 25,000 atm,
showed that hydrostatic pressure has negligible effect on the yield point until
extremely high pressures are reached. Furthermore, the shape of the stress-
strain curve remains unaltered in the small-strain range. The major effect of
hydrostatic pressure is to increase the ductility of the material and to permit
much larger deformations prior to fracture.
It has also been shown that the density, and consequently the volume, does
not change even for very large plastic deformations. Thus, in the plastic range,
a material can be considered as incompressible. These two experimental facts,
i.e., the lack of influence of hydrostatic pressure and incompressibility, are
very important in the development of plastic flow theories.
2-6 IDEALIZATION OF THE STRESS-STRAIN

CURVE. DYNAMIC AND KINEMATIC MODEL
Because of the complex nature of the stress-strain curve, it has

customary to idealize this curve in various ways [13]. Figure 2.6.1
Sec. 2-6] Idealization of the Stress-Strain Curve 17
r--p
(a)
t
cr
0
7tan- 1 mE p
_i__.___ _ _ __
(c)
(d)
(e)
2.6.1 Idealized stress-strain curves: (a) perfectly elastic, brittle;

perfectly plastic; (c) rigid, linear strain hardening; (d) elastic, perfectly
(e) elastic, linear strain hardening.
idealized curves as well as corresponding dynamic models which can be used

to describe the material behavior. It will be shown subsequently that with the
use of modern computing machinery, these idealizations are in many cases
not necessary.
The models shown in Figure 2.6.1 are designated as dynamic models. They
replace stresses by forces and strains by displacements. To devise models of
u =OP
1-
0 e =Iii OR
(a)
T----'---------,--------'------T
(b) IR Pe
T-----'-------~-------L------T
(c) IR
T------~----------.--------~-----T
(d) IR PI!
T-------''--------,---------'----T
(e) IR p.
T----------~---,------------L-T
(f)
(g)
(h) IR
T----------~----------~-----------------'-------T
(i) IR
T-------~----------~-----------------'-------T
0
FIGURE 2.6.2 Kinematic model.
Sec:. 2-6} Idealization of the Stress-Strain Curve 19
this type to represent combined stresses acting in several directions would be

extremely difficult. For this reason Prager [14] introduced ingenious kinematic
models in which both the stresses and strains are represented by displacements.
Figure 2.6.2 illustrates this type of model for the case of the rigid linear~
hardening plastic material (8]. The model is taken to be a slotted bar, as
shown. The bar is free to move along its length on the frictionless table T.
But for the bar to move, the pin P must engage the end of the bar. Initially
pin P is at the center R of the slot and this point is marked as point 0 on the
table. The distance from P to either end of the slot is taken equal to the yield
stress u 0 of the rigid linear~hardening material.
The distance OP of the pin P from the fixed point 0 is taken equal to the
stress. Then the distance OR from the center of the slot R to the fixed point 0
is proportional to the strain, i.e., e = OR/m, where tan - l m is the slope of
the plastic stress-strain curve. Thus plastic flow will take place when the pin
is engaged at one end or the other of the slot. Figures 2.6.2 and 2.6.3 illustrate
the different positions of the kinematic model and the corresponding stress-
strain diagram. Note that for this particular model it has been assumed that
the elastic unloading range EG is equal to twice the initial yield, so that the
yield point in compression G is less than the initial yield C.
A~----+------------+F~-----
FIGURE 2.6.3 Stress-strain curve for model of Figure 2.6.2.

In a two-dimensional state of stress, two stress variables, a 1 and a 2 , are

specified instead of the single variable a. Therefore, if the position of the pin
is to indicate the state of stress, it must be free to move in two directions. In
this case the rigid region must be a two-dimensional frame rather than a slot.
This will be discussed in greater detail in Chapter 6, after the discussion of
yield surfaces.
2-7 EMPIRICAL EQUATIONS FOR STRESS-STRAIN

CURVES
It is sometimes useful to represent the stress-strain curve of a given material

by an equation obtained empirically by fitting the experimental data. Equa-
tion (2.2.12) is such an equation which will frequently fit most of a given
stress-strain curve, but, as was previously mentioned, will not usually fit at
the low-strain and high-strain ends of the stress-strain curve. One of the
first such empirical equations was proposed by Ludwik [1]. It has the form
(2.7.1)
A frequently used form, due to Ramberg and Osgood [15], is
(2.7.2)
Some other forms that have been proposed are
a = a + (b - a)(l - e-n•) (2.7.3)
a= c(a + e)n 0 s n:::; I (2.7.4)
(2.7.5)
where e is the base of natural logarithms, e0 the yield strain, a 0 the yield stress,
E the elastic modulus, and m, n, k, a, b, and c constants.
It is also possible to fit the plastic part of the stress-strain curve as
ately as desired by a polynomial of arbitrary degree, i.e.,
where e0 is the yield strain. For linear strain hardening all the CO(~ffilcieJilf
beginning with a2 are zero.
Ch. 2] Problems 21
Problems
1. Show that natural strains are additive whereas conventional strains are not.
2. Assume that a material behaves elastically up to the point of instability.
Show that the natural strain at this point is unity.
3. Derive equation (2.2.13).
4. Let the stress-strain curve of a material be given by a = Aen, where e is the
conventional strain. Show that at the point of instability
n
e=--
1- n
5. In a standard tensile test using a t-in.-diameter specimen with a l-in. gage
length, the following data were obtained. At a load of 10,000 lb, the conven-
tional strain was 0.10, and at a load of 12,000 lb, the conventional strain was
0.60. Find the true stresses and strains for these two conditions. Determine
the strength coefficient A, the strain-hardening exponent n, the change in
gage length at the maximum load, and the maximum load assuming equation
(2.2.12) to hold.
6. A tensile load is applied to a thin-walled hollow circular cylinder. Determine
the change in wall thickness and in mean radius at the point of maximum
load, if the stress-strain curve is given by a = Aen, where e is the conven-
tional strain and a is the true stress.
7. Derive the incompressibility conditions (2.2.5) and (2.2.6).
8. The following data were obtained in a tensile test on a 0.505-in.-diameter
specimen:
Diameter, in. Load, lb Diameter, in. Load, lb
0.487 6,750 0.419 11,000

0.481 9,250 0.402 10,800
0.472 10,400 0.375 10,200
0.463 10,900 0.361 9,700
0.450 11,100 0.354 9,500
0.438 11,200 0.326 8,950
Fracture
(a) Plot the true stress-strain curve.

(b) Determine the strength coefficient A and the strain-hardening exponent n.
(c) Determine the maximum load from the stress-strain curve and compare
it with that obtained using equation (2.2.13).
Consider a material whose stress-strain curve is given by a = 30,000 +
1.5 X 106 e, a > 30,000 psi. If a tensile specimen of this material is stretched
to a strain of 0.004 in./in., at what stress will it yield in compression when
the load is reversed, for each of the assumptions in Figure 2.3.1?
For the dynamic models of Figure 2.6.1, show the relations between the
constants and the parameters of the stress-strain curve. Denote the
constants by k (k1 and k2 for the last model), the weight of the block
by W, the friction coefficient by p,, and the force by P. For example, for the
first model, the equation of the stress-strain curve is a = Ee and the corres-
ponding model equation is P = kx. Thus
11. For the kinematic model of Figure 2.6.2, show that e = ORfm.
12. Describe a kinematic model similar to that shown in Figure 2.6.2 for iso-
tropic hardening.
13. Sketch typical stress-strain curves that would be obtained using Ludwik's
expression for the following cases:
(a) n = 1.
(b) 0 s n < 1.
(c) ao = 0, n = 0, t, 1.
References
1. P. Ludwik, Elemente der technologischen Mechanik, Springer, Berlin, 1909.
2. J. D. Lubahn and R. P. Felgar, Plasticity and Creep of Metals, Wiley, New
York, 1961.
3. G. E. Dieter, Jr., Mechanical Metallurgy, McGraw-Hill, New York, 1961.
4. H. Schwartzbart and W. F. Brown, Jr., Notch-Bar Tensile Properties of
Various Materials and their Relation to the Unnotch Flow Curve and
Notch Sharpness, Trans. ASM, 46, 998, 1954.
5. J. Marin, Mechanical Behavior of Engineering Materials, Prentice-Hall,
Englewood Cliffs, N.J., 1962.
6. R. Hill, The Mathematical Theory of Plasticity, Oxford Univ. Press, London,
1950.
7. D. Mclean, Mechanical Properties of Metals, Wiley, New York, 1962.
8. J. N. Goodier and P. G. Hodge, Jr., Elasticity and Plasticity, Wiley,
York, 1958.
9. M. J. Manjoine, Influence of Rate of Strain and Temperature on
Stresses of Mild Steel, J. Appl. Mech., 11, A-211, 1944.
10. A. W. Magnusson and W. M. Baldwin, Low Temperature Brittleness,
Mech. Phys. Solids, 5, 172, 1957.
11. P. W. Bridgman, The Effect of Hydrostatic Pressure on the Fracture
Brittle Substances, J. Appl. Phys., 18, 246, 1947.
12. P. W. Bridgman, Studies in Large Plastic Flow and Fracture with
Emphasis on the Effects of Hydrostatic Pressure, McGraw-Hill, New
1952.
13. W. Johnson and P. B. Mellor, Plasticity for Mechanical Engineers,
Nostrand, Princeton, N.J., 1962.
14. W. Prager, The Theory of Plasticity-A Survey of Recent Ac:hu~ve:me~n
Proc. Inst. Mech. Engrs., London, 169, 41, 1955.
Ch. 2] General References 23
15. W. Ramberg and W. R. Osgood, Description of Stress-Strain Curves by

Three Parameters, NACA Technical Note No. 902, July 1943.
16. T. A. Trozera, 0. D. Sherby, and J. E. Dorn, Effect of Strain Rate and
Temperature on the Plastic Deformation of High Purity Aluminum, Univ.
Calif. (Berkeley) Tech. Rept., Ser. 22, Issue 44, Contract N7-0!{R-295,
Dec. 1955.
General References
Drucker, D. C., Stress-Strain Relations in the Plastic Range-A Survey of
Theory and Experiment, ONR Rept. NR-D41-D32, 1950.
Goodier, J. N., and P. J. Hodge, Jr., Elasticity and Plasticity, Wiley, New York,
1958.
Hill, R., The Mathematical Theory of Plasticity, Oxford Univ. Press, London,
1950.
Johnson, W., and P. B. Mellor, Plasticity for Mechanical Engineers, Van Nostrand,
London, 1962.
CHAPTER 3
THE STRESS
TENSOR
It is assumed that the reader is familiar with the basic concepts of the theory
of elasticity, including the definitions of stress and strain. However, to avoid
having the student refer to other texts and to refresh the memory of those
readers who have not recently done any work in elasticity, we shall briefly
review in the next three chapters some of these basic concepts, with particular
emphasis on those properties of the stress and strain tensors which are par-
ticularly important in the development of plasticity theory. The reader
thoroughly familiar with elasticity theory may skip directly to Chapter 6.
Stress and strain are second-order tensors, and although we shall not con-
cern ourselves with tensors and their properties as such, it is important that
the student be familiar with the subscript notation known as tensor notation.
This notation is not only a time saver in writing out long formulas or expres-
sions, but it is also extremely useful in derivations and in the proof of
theorems. Furthermore, a major part of the past and present literature on the
subject utilizes tensor notation, and a knowledge of this notation is "'~~''"T''"
in following the literature. In solving specific problems the usual <vu")';u<u,...
notation must, however, always be used. We shall therefore start with a
description of tensor notation.
24
Sec. 3-1] Tensor Notation 25
3-1 TENSOR NOTATION
A tensor is a system of numbers or functions which transform according

to a certain law, when the independent variables undergo a linear trans-
formation. We shall not concern ourselves here with the transformation laws
for tensors but will merely record a few elementary properties, including the
tensor notation. We shall accept the fact that stress and strain are tensors.
A subscript notation is used which is really very simple. The coordinate
axes are designated by the letter x, with a latin subscript. Thus x 1 does not
mean just one quantity, but three quantities, xl> x 2 , and Xs, where x 1, x 2 , and
Xs are used instead of x, y, and z (orr, 8, z, etc.). Any other subscript, such
asj, k, l, m, etc., can be used equally well. For two-dimensional problems the
subscript is understood to have a range of only two rather than three.
A double subscript indicates a system of nine components if the range of
each of the subscripts is three, or a system of four components if the range is
two. For example, the stress tensor is designated by a" and stands for nine
components:
all a12 als] [ax Txy Txz]

ali = [ a21 a22 a2s = Tyx ay Tyz (3.1.1)
asl as2 ass Tzx Tzy az
Similarly, the nine components of the strain tensor are designated by e!i.
Two subscripted quantities are said to be equal if their corresponding com-
ponents are equal. Thus if A!i = B 1h then All = Bll, A 12 = B 12, etc.
If two subscripted quantities are added, their corresponding components
are added. Thus
All + Bll = Cll

A12 + B12 = C12
A 21 + B21 = C 21 etc.
one-subscript system is called a system of first order, a double subscript

one of second order, etc. It is evident from the definitions of equality
addition above, that these can apply only to systems of equal order.
now come to the only "tricky" part of tensor notation-the summation
. Whenever a subscript is repeated, this indicates summation over
of the subscript. Thus
26 The Stress Tensor [Ch. 3
Such a subscript is called a dummy subscript and it must be a letter not a

number; i.e., a 11 does not mean summation.
A more complicated example is given by the increment of work per unit
volume:
(3.1.2)
The advantage here becomes apparent, since we have written down one term
instead of nine. Also the work increment stands out as the scalar product of
the strain increment and the stress. Furthermore, rli de!J represents the work
increment without being tied to any particular system of axes. It would thus
include r 11 de 11 + r 22 de 22 + r 33 de 3s, where these are the principal stresses
and strains.
A system having any number of subscripts is said to be symmetric in two of
these subscripts if the components of the system are unaltered when the two
subscripts are interchanged. Thus a second-order system is called symmetric if
The stress and strain tensors are usually symmetric.

A system is said to be skew-symmetric or antisymmetric if the interchange
of the indices changes the signs of the components. Thus for a second-order
skew-symmetric system
This shows immediately that
A 11 = -A 11 = 0
Azz = -A22 = 0
A33 = -Ass= 0
Therefore a skew-symmetric tensor of second order is characterized by

three quantities:
P1 = Asz = -A2s
P2 = A13 = -Asl
Ps = A1z = -A21
It can readily be shown that every second-order tensor A 11 may be

posed into the sum of a symmetric tensor eli and a skew-symmetric
Sec. 3-2] Stress at a Point 27
The simplest second-order symmetric tensor is the Kronecker delta or

substitution tensor, defined by
(3.1.3)
= [~ ~]
It is called the substitution tensor because
8t,AJ =At (3.1.4)

i)lfAilc = Aile
Finally, the convention is used to designate partial differentiation by a

comma. Thus
(3.1.5)
3-2 STRESS AT A POINT
Consider a body as shown in Figure 3.2.1 subjected to a system of external

forces P 1 P 2 ••• P 8 • Now consider a plane AB passing through the body
dividing it into parts I and II, as shown. If we consider part I, it is seen that
FIGURE 3.2.1 Loaded body.

it is in equilibrium under the action of forces P1 , P2 , P3, P 4 , and the force P 12

that part II exerts on part I, P 12 being the resultant of the continuous distri-
bution of forces on the plane AB that part II exerts on part I. If a small area
b..A is taken in this plane with a force b..P acting on it, then the unit stress
acting at this point is defined as
. b..P
p = 1lm AA (3.2.1)
t.A->0 i l
The important thing to note here is that the unit stress, p, must be referred to
a particular plane. For any other plane passing through the same point, it is
obvious from consideration of Figure 3.2.1 that the force distribution on this
plane, and hence the unit stress, will be different.
The unit stress, p, of course, need not be perpendicular to the plane AB.
In practice, therefore, the stress, p, is decomposed into two components, one
normal to the plane of reference, called the normal stress, and one parallel
to this plane, called the shearing stress. The normal stress is taken as positive
when it is tensile in nature and negative when it is compressive.
To completely specify the stress at a point it is necessary to specify the
stresses at that point on three mutually perpendicular planes passing through
the point. The stress on any arbitrary plane through the point can then be
determined in terms of the stresses on the three perpendicular planes, as will
shortly be shown. Let the three mutually perpendicular planes be the planes
perpendicular to the x, y, and z coordinate axes. Then the stresses acting on
these planes at their point of intersection are as designated in Figure 3.2.2.
The stresses as shown are all positive. The subscripts denote the direction of
the stress. The first subscript designates the normal to the plane under con-
sideration, and the second subscript designates the direction of the stress.
Thus r xy denotes a shearing stress acting on the face of the element that is
perpendicular to the x axis, the stress acting in the direction of the y axis.
As mentioned previously, the normal stress is taken positive when it nrr•rlllt'Po
tension and negative when it produces compression. The positive mrectwtts
of the components of shearing stress on any side of the cubic element
taken as the positive directions of the coordinate axes, if a tensile stress on
same side would have the positive direction of the corresponding axis.
It is seen from the figure that the complete specification of the stress at
point is given by the nine quantities
Sec. 3-ZJ Stress at a Point Z9
FIGURE 3.2.2 Convention for stresses.
It is customary in engineering practice to delete the second subscript on the

normal stress and write ax instead of a xx, etc. In tensor notation the nine
'!'"""'·~"""" are designated simply by a 11 or, by some authors, TIJ·
If one considers an infinitesimal rectangular parallelepiped surrounding a
point in a body, then it readily follows, as is shown in standard texts,
static equilibrium of forces and moments requires that the stresses at
point satisfy the following equations:
OCT 07'
X OT zx YX
-
ax + oy
- +-=
oz -Fx
(3.2.2)
0Txz + O'Tyz + OCTz _ -F.

ox oy oz - ~
, are the components of the body forces per unit volume. Also
Txz = 7zx (3.2.3)

In tensor notation these become simply
(3.2.4)
The second line of(3.2.4) expresses the fact that the stress tensor is symmetric.
There are therefore in general only six independent components of stress at a
point rather than nine. (Note: There are some peculiar conditions for which
the stress tensor will not be symmetric, as in the case when body moments
act [1 ].)
3-3 PRINCIPAL STRESSES. STRESS INVARIANTS
If we are given the six components of stress at a point with respect to some
coordinate system (x, y, z), we can determine the stresses acting on any plane
through this point. This can be done by consideration of the static equilibrium
of an infinitesimal tetrahedron formed by this plane and the coordinate
planes, as shown in Figure 3.3.1. In this figure we have shown the stresses
acting on the three coordinate planes. These stresses are assumed to be known.
We wish to find the stresses acting on the plane ABC whose normal ON has
direction cosines l, m, and n. Let the area of the infinitesimal triangle ABC be
designated by ~A. Then the areas of the faces AOB, COB, and AOC are equal
to m ~A, l ~A, and n ~A, respectively. Now let the stress vector acting on
the face ABC be designated by Sand its x, y, and z components by Sx, Sy,
and S2 as shown in Figure 3.3.l(b). Then for equilibrium of forces in the
x direction,
or
Similarly,
Sy = lrxy + may + llTzy
S 2 = lrxz + mryz + llUz
To obtain the normal stress Sn acting on this plane we project the

Sx, Sy, and S2 onto the normal ON, to get
Sec. 3-3] Principal Stresses. Stress Invariants 31
z OD ..L ABC
N .I= 00
OA
00
m=-
oc
00
n = 08
c
y
(a)
z
c y
X
(b)
FIGURE 3.3.1 Forces on infinitesimal tetrahedron.
from (3.3.1) results in
(3.3.3)
, to obtain the resultant shear stress s. acting on this plane,

s; = S2 - s;
= s; + s~ + s; - s~ (3.3.4)
Equations (3.3.1) give the x, y, and z components of the stress acting on this
plane, and equations (3.3.3) and (3.3.4) give the normal and shear stresses.
Equations (3.3.1) can also be considered as the boundary conditions that
have to be satisfied by the stress components u 11 at any point on the boundary
of the body. Thus if the element of area ABC is considered to be an element
of the boundary whose normal has the direction cosines /, m, and n, and
Sx, Sy, and S2 are the components of the applied boundary forces at the
point 0, then equations (3.3.1) are precisely the boundary conditions that
must be satisfied by the stress tensor. In tensor notation, if we replace l, m,
and n by /1 , 12 , and /3 , we can write (3.3.1) as
(3.3.5)
Suppose the plane element ABC of Figure 3.3.1 is so oriented that the
resultant stress Son this plane element is normal to the plane; i.e., S = Sn
and Ss = 0. The plane is then called a principal plane at the point, its normal
direction is called a principal direction, and the stress S = Sn is called a
principal stress. At every point in a body there are at least three principal
directions. These principal stresses and principal directions can readily be
found as follows. Assume the element ABC to lie in a principal plane at
point 0 so that S = Sn. Then S has the same direction cosines l, m, and n
as the normal. The components of S in the x, y, and z directions are then
Sx =IS
Sy = mS
Sz = nS
and equations (3.3.1) give immediately
l(ux- S)+ mryx + nTzx = 0

hxv + m(uy - S) + nrzy = 0
hxz + mryz + n(uz- S) = 0
or in tensor notation (replacing l, m, and n by /1 , 12 , and /3 )
For equation (3.3.6) to have a nontrivial solution for l, m, and n, the

minant of the coefficients must vanish, resulting in
ax- S Tyx Tzx

iai} - 3;1SI = Txy O'y - s Tzy =0
7" xz 7" YZ O'z- S
Sec. 3-3] Principal Stresses. Stress Invariants 33
Expanding the determinant gives a cubic equation for S:
(3.3.9)
where
12 = 'T;y + 'T~z + 'T;x- (axay + UyUz + UzUx) (3.3.10)

1s = UxUyUz + 2Txy'Tyz'Tzx- (ax'T~z + Uy'T~x + Uz'T;y)
It can be proved [2] that the cubic equation (3.3.9) has three real roots and
consequently there are (at least) three principal stresses, which will be desig-
nated by a 1 , a 2, and as. Substituting any of these solutions back into equations
(3.3.6) enables one to solve for the corresponding direction cosines l, m, and
n, if in addition the identity !2 + m 2 + n 2 = 1 is used. If the three roots
a 1 , a 2, and as are distinct, the three corresponding principal directions will
be unique and orthogonal. If two of these roots are equal, one direction will
be unique but the other two directions can be any two directions, orthogonal
to the first. If all three roots are equal, there are no unique principal directions
and any three directions can be chosen. This corresponds to a state of hydro-
static stress.
Suppose instead of the axes x, y, and z, a different set of axes, x', y', and z',
were chosen at the point 0. Then the equation for determining the principal
stresses, (3.3.9), would be the same, except that 11 , 12, and 1s would be defined
in terms of the stresses a~, a~, a~, etc., with respect to the new coordinate
the principal stresses are physical quantities and obviously do not

on the coordinate axes chosen. Hence the numbers 11 , 12, and 1s
""'''"'""'in (3.3.9) must be the same, no matter what coordinate axes are
in order that they give the same values for a 1 , a 2, and as. Thus
ly for 12 and 1s. 11 , 12, and 1s are therefore called the first, second,
third invariants of the stress tensor, respectively. We will show later
the first and second invariants are particularly important in plasticity
If we choose the principal directions as the directions of the coordinate

axes, then the stress invariants take on the simple form
/1 = a1 + az + as
l2 = -(ala2 + a2as + aaal) (3.3.11)
It is to be noted that the invariants 11 , 12 , and Is appearing in (3.3.11) are

three independent quantities which specify the state of stress just as well as
al> a 2 , and as. That is, given a 1 , a 2 , and a 3 , we can calculate 11> 12 , and 13 , and
given 11 , 12 , and 13 , we can calculate al> a 2 , and a 3 , one set of these numbers
uniquely determining the other set. Furthermore, any three independent
combinations of these invariants will obviously also be invariants and can
specify the state of stress just as a 1 , a 2 , and a3 do.
3-4 MAXIMUM AND OCTAHEDRAL SHEAR

STRESSES
Let us take the coordinate axes in the principal directions. Then the shear
stresses referred to these axes are zero and the normal and shear stresses on
some oblique plane, with direction cosines with respect to these axes of!, m,
and n, are, from equations (3.3.3) and (3.3.4),
Sn = + m 2 a 2 + n2 as
l2a1
(3.4.1)
s; = sr + s~ + sg - s~
But, from (3.3.1), with the shear stresses zero,
Therefore,
We already know that on the principal planes the shear stress is a minirnu1n---'
zero. Let us now find the planes for which it is a maximum; i.e., we seek
values of l, m, and n, such that s. as given by equation (3.4.2) is a u"'""w'"'"
In addition to equation (3.4.2), there is a restriction on the direction
/
2
+ m2 + n2 = 1; i.e., only two of them can be independent.
into (3.4.2) n2 = 1 - m 2 - f2, differentiating the resultant equation
Sec. 3-4] Maximum and Octahedral Shear Stresses 35
respect to I and m, and equating these derivatives to zero, the following

equations are obtained for I and m:
l[(al - aa)f2 + (a2 - aa)m 2 - ·Hal - aa)] =0

(3.4.3)
m[(al - a 3)f2 + (a 2 - a 3)m 2 - !(a2 - a 3)] =0
One solution is obviously I = m = 0 and n = ± 1. Another solution is ob-

tained by taking I = 0 but not m. Then from the second equation,
m = ± Vl/2. Also, taking m = 0 gives, from the first equation, I= ± Vl/2.
There are in general no solutions for I and m both nonzero except in the
special case a 1 = a 2 •
If the above calculations are carried out by eliminating from equation
(3.4.2) first m and then I, the following table is obtained:
0 0 ±1 0 ±V1f2 ±V1/2
m 0 ±1 0 ± v'1/2 0
n ±1 0 0 0
The first three columns obviously give the direction cosines of the coordinate
planes which are principal planes, and therefore the shearing stresses on these
planes are zero; i.e., they are a minimum. The three last columns give direc-
tion cosines of 45° angles. These planes, therefore, bisect the angles between
the coordinate axes. On these planes the shearing stresses are maximum.
Designating these stresses by r 1 and substituting these direction cosines into
equation (3.4.2), the values of the shearing stresses are obtained as
rl = ±-t(a2 - aa)
T2 = ±-t(al - aa) (3.4.4)
Ta = ±!(al - a2)
the maximum shearing stress acts on the plane bisecting the ar1gle
the largest and smallest principal stresses and is equal to half the
...~·~"11"" between these principal stresses.
we compute the normal stresses on these planes and designate them by
we get, from (3.4.1),
N1 = !(a2 + aa)
N2 = !(al + aa) (3.4.5)
Na = !h + a2)
so that the normal stress on each of these planes is equal to the average of the
principal stresses on the two planes whose angle it bisects.
If the plane ABC of Figure 3.3.1 is so oriented that OA = OB = OC, then
the normal ON will make equal angles with all the axes, and
1
l=m=n= ±v'"3 (3.4.6)
The normal stress acting on this plane is
(3.4.7)
That is, the normal stress on this plane is equal to the mean stress, and the
shear stress is
s; = t(ar + a~ + a5) - }(al + a2 + as) 2

= -b[(al - a2)2 + (a2 - as) 2 + (as - a1)2] (3.4.8)
or, making use of (3.4.7),
It is apparent that a tetrahedron similar to this one can be constructed in

each of the four quadrants above the xy plane and in each of the four quad-
rants below the xy plane. On the oblique face of each of these eight tetrahedra
the condition I2 = m 2 = n 2 = t will apply. The difference between the
tetrahedra will be in the signs attached to !, m, and n. The eight tetrahedra
together will form an octahedron, and on each of the eight planes forming
the faces of this octahedron the normal stress and the shear stress will be as
given by equations (3.4. 7) and (3.4.8). These planes are called the octahedral
planes, and the shear stresses acting on them are called the octahedral shear
stresses. We shall designate this stress by Toot· Thus
Tact= t[(a1 - a2) 2 + (a 2 - as) 2 +(as - a 1 ) 2]1 12
= ~3 ((al - am? + (a2 - am)2 + (as - am)2]li2 (3.
In terms of the stress invariants the octahedral shear stress can be written
'Tact=
v'2 (J21 + 312)1/2
3
Sec. 3-5] Mohr's Diagram 37
and in terms of generalnonprincipal stresses it becomes
T~ot = M(ax + ay + az) 2 + 3(-r~y + T~z + r;x) - (axay + ayaz + azax)]

= i[(axay) 2 + (ayaz) 2 + (azax? + 6(-r;,y + T~z + r;x)]
(3.4.11)
which gives the octahedral shear stress in terms of the stress components
referred to an arbitrary set of axes.
3-5 MOHR'S DIAGRAM
The result of the previous section can be obtained graphically by the use
of a diagram proposed by Mohr [3]. This approach is particularly useful in
discussing a pair of parameters called Lode's variables which are frequently
encountered in plasticity theory. A full discussion of Lode's variables will be
postponed until later. Mohr's diagram can be obtained in the following way.
Let us take the principal directions for the coordinate axes and consider the
normal and shear stresses acting on an arbitrary plane oblique to these axes.
Instead of Sn and Ss we will for convenience use a and, r for the normal and
shear stresses on this plane. Then in terms of the principal stresses equations
(3.4.1) and (3.4.2) become
= a f2al + m 2a2 + n2as

2 (3.5.1)
-r 2 +a = f2ar + m 2 a~ + n 2 a~
With the identity f2 + m 2 + n2 = 1, we have three equations to solve for

2 2 2
/ , m , and n • Solving these results in
f2 = -r
2
+ (a - a2)(a - as)
(a 1 - a2 )(a1 - as)
m 2 = -r 2 + (a - as)(a - a 1) (3.5.2)
(a 2 - as)(a 2 - a 1 )
n2 = -r
2
+ (a - a 1)(a - a 2)
(as - al)(as - a2)
loss of generality we can order the axes so that

Then since !2 2 0 and (a 1 - a 2)(a 1 - as) 2 0, from the first of equations

(3.5.2),
(3.5.3)
Equation (3.5.3) can also be written
(3.5.4)
If the equality sign is used in equation (3.5.4), this equation represents a

circle in the -r - a plane, as shown by C1 in Figure 3.5.1. The center is
FIGURE 3.5.1 Mohr's diagram.
obviously at 0, (a 2 + as)/2, and it crosses the a axis at a = a 2 and a = as.

Therefore the region defined by equation (3.5.4) lies outside this circle and
includes it as a boundary, since it represents circles with radii equal to or
greater than (a 2 - as)/2.
Now considering the second equation of (3.5.2), since a 2 - a 3 > 0 and
a 2 - a 1 < 0, the numerator must be less than or equal to zero:
or
which represents a family of circles with centers at (a1 + as)/2 and

Sec. 3-6] Stress Deviator Tensor 39
equal to or less than (al - a3)/2, as shown by the interior of circle c2 in

Figure 3.5.1. Finally, the third of equations (3.5.3) gives
or (3.5.6)
which represents the region exterior to the circle C3 with center at (a1 + a 2 )/2
and crossing the a axis at a 1 and a 2 •
It follows therefore from (3.5.4), (3.5.5), and (3.5.6) that the admissible
values of rand a lie in the shaded region shown in Figure 3.5.1 bounded by
the circles C1 , C2 , and C3 • The maximum shearing stress, as is clear from the
figure, is represented by the largest ordinate, AB, which is the radius of the
circle C2 and is therefore equal to (a 1 - a 3 )/2. To determine the orientation
of the plane that has this shearing stress, we make use of equation (3.5.2).
The value of a corresponding to the maximum shearing stress is equal to
(a1 + a 3)/2 (the center of the circle C2 ). Substituting these values of rand a
into equations (3.5.2), we get
We have thus obtained the same values for the maximum shearing stress as
in the previous section.
3-6 STRESS DEVIATOR TENSOR
It is convenient in plasticity theory to split the stress tensor into two parts,
one called the spherical stress tensor and the other the stress deviator tensor .
.The spherical stress tensor is the tensor whose elements are am oli. where am
the mean stress, i.e.,
(3.6.1)
(3.6.2)
(3.6.2) it is apparent that am is the same for all possible orientations

axes; hence the name spherical stress. Also, since am is the same in all
directions, it can be considered to act as a hydrostatic stress. Now, it was

shown in Section 2.5 that even very large hydrostatic pressure has a negligible
effect on yielding and plastic flow. Therefore, in plastic flow considerations
one can consider, if desired, the stress system obtained by subtracting the
spherical state of stress from the actual state, rather than working with the
actual state of stress. We therefore define a stress tensor called the stress
deviator tensor sjj as follows:
'T'xy (3.6.3)
It is apparent that subtracting a constant normal stress in all directions will

not change the principal directions. The principal directions are therefore
the same for the deviator tensor as for the original stress tensor. In terms of
the principal stresses the deviator tensor is
slf = ["' ~ "•

0
0'2 -
0
am
u,~J
2a 1 - a2 - a3
0 0
2
2a 2 - a3 - a1
0 0
3
2a 3 - a1 - a 2
0 0
3
To obtain the invariants of the stress deviator tensor,

S' + tf1 in equation (3.3.9). This results in
Sec. 3-7] Pure Shear 41
where
Jl = 0
'2 = ten + 3/2) (3.6.6)
J3 = -i-7(2Jr + 9/1/2 + 27/3)
One advantage of using the stress deviator tensor is now apparent. The
first invariant of this tensor is always zero. This can also be seen by taking
the sum of the diagonal elements in equation (3.6.3) or (3.6.4).
The invariants 1 2 and J 3 can, of course, be written in terms of the stress
components a 11 • For example,
(3.6.7)
Or, in terms of principal stresses,
l2 = i[(al - a2) 2 + (a2 - a3) 2 + (a3 - a1) 2]

(3.6.8)
J 3 = (a1 - am)(a 2 - am)(a 3 - am)
In terms of the principal stress deviators Sb S 2, and S 3 defined by

S1 = a1 - am,
J2 = -(S1S2 + S2S3 + S3Sl)
(3.6.9)
13 = sls2s3
Finally, comparing (3.6.7) to (3.4.11) we see that
(3.6.10)
This relationship between J 2 and the octahedral shear stress is sometimes

used to lend credence on physical grounds to some plasticity theories, as
will be discussed later.
PURE SHEAR
important state of stress is the one designated as pure shear, or simple

If at a point in a body a set of coordinate axes can be found such that
= ay = a 2 = 0, then the point is said to be in a state of pure shear. The
state in a cylindrical bar in torsion is an example of pure shear. Another
is the stress state
a3 =0 (3. 7.1)
It can readily be shown [4] that a necessary and sufficient condition for a
state of pure shear to exist is
(3.7.2)
In conclusion it should be emphasized that all the equations developed

and presented in this chapter are independent of the material properties and
are therefore equally valid for bodies which behave elastically or plastically.
Problems
1. Write out the expression for the work increm.-nt
dW = T!J defJ
2. Given the tensor TucTicJ·
(a) What is the order of this tensor?
(b) If i, j, and k range from 1 to 3, how many components does this tensor
have?
(c) Write out the components.
3. The general Hooke's law is given by
(a) Write out the equations this represents.

(b) What is the order of this tensor?
4. Prove that every second-order tensor afi may in a unique way be decomposed
into the sum of a· symmetric tensor e11 and a skew-symmetric tensor Pu·
5. Verify relations (3.4.4).
6. Verify equations (3.4.6), (3.4.7), (3.4.8), (3.4.10), and (3.4.11).
7. Discuss the Mohr diagram for the case a 2 = a 3 and determine the orienta-
tion of surface elements experiencing extreme shearing stresses. Consider
also the case where a1 = a2 = a 3.
8. Derive equations (3.6. 7), (3.6.8), and (3.6.9).
9. Let ax = ay = a. = 0, Txy = Tyz = Tzx = a. Calculate the octahedral shear
stress, J 1 , J2, and J 3 , the principal stresses, the greatest shearing stress, and
the direction cosines for the greatest principal stress.
10. Show that subtracting a hydrostatic stress from a given stress state will not
change the principal directions.
11. Show that the second and third invariants of the stress deviator tensor can
also be written
J2 = t(S~ + S~ + S~)
13 = -tcsr + s~ + s~)
12. Determine the invariants of the stress tensor and the octahedral stresses
the case of uniaxial state of stress.
13. The case of plane stress is partially characterized by a 2 = Txz = Tyz =
Determine the invariants of the stress tensor, the principal stresses,
maximum shear stresses, and the octahedral stresses for this case.
Ch. 3] General References 43
14. Given the following stress tensor at a point:

10,000 1,000 -8,000)
1,000 -6,000 6,000
(
-8,000 6,000 20,000
Determine the stress invariants, the stress deviator tensor, the principal
stresses, the maximum shear stress, and the plane upon which it acts and
the normal and shear stresses acting on a plane with direction cosines
1 1 1
1= v3 m = v2 n = v6
15. A circular cylinder 2 in. in diameter is subjected to a tensile force of 30,000 lb.,
a bending moment of 35,000 in.-lb, and a twisting moment of 50,000 in.-lb.
Determine the stress invariants, the principal stresses, the maximum shear
stress, and the stress deviator tensor.
16. Show that if a uniform tension a 1 is applied to the two opposite sides of a
square and a uniform compression -a1 is applied to the other two sides, a
state of pure shear is obtained.
References
L I. S. Sokolnikoff, Mathematical Theory of Elasticity, 2nd ed., McGraw-Hill,
New York, 1956, p. 42.
2. Ibid., p. 17.
3. 0. Mohr, Abhandlungen aus dem Gebiet der technischen Mechanik, 2nd ed.,
Wilhelm Ernst und Sohn, Berlin, 1914, p. 192.
4. C. E. Pearson, Theoretical Elasticity, Harvard University Press, Cambridge,
1959, p. 57.
General References
Sokolnikoff, I. S., Mathematical Theory of Elasticity, 2nd ed., McGraw-Hill,
New York, 1956.
and J. N. Goodier, Theory of Elasticity, McGraw-Hill, New
CHAPTER 4
THE STRAIN
TENSOR
When the relative position of any two points in a continuous body is

changed the body is said to be deformed or strained. If the distance between
every pair of points in a body remains constant during motions of the body,
the body is called rigid. The displacements of a rigid body consist of trans-
lations and rotations; translations and rotations are therefore called rigid body
displacements. The analysis of strain consists of the study of the deformations
of bodies which is essentially a geometric problem and is unrelated to the
material properties. The specification of strain at a point is therefore the
same for elastic and for plastically deforming bodies.
4-1 STRAIN AT A POINT
We start the analysis of the strain at a point in a body by considering

body referred to an orthogonal set of axes as shown in Figure 4.1.1. '"'v'"''""'
two arbitrary neighboring points P 0 and P in the unstrained body.
straining, the points move toP~ and P', respectively. The coordinates of
are x 0 , y 0 , and z 0 and of P, x, y, and z. The coordinates of P~ and P'
x~, y~, and z~ and x', y', and z', respectively. The vector A deforms into
vector A'. The components of A are Ax, Ay, and A 2 and the components
are
44
Sec. 4-1] Strain at a Point 45
Pt(x,y,z)---.
--- -- --
A
Po(xo, Yo· zol
FIGURE 4.1.1 Deformed body.
The components of displacement of the point P 0 are
Uo =X~- Xo
Vo = Y~- Yo (4.1.1)
w0 = z~ - z0
The components of displacement of the point P are
u = x'- x
v = y'- y (4.1.2)
w = z'- z
We assume that the displacements are single-valued continuous functions

the coordinates x, y, and z. (Actually, as will be apparent later, we have to
that all their derivatives through the third are also continuous.) We
therefore expand the displacements at Pin a Taylor series about P 0 as
8u ou ou
U = Uo + ox Ax + oy Ay + oz Az
(4.1.3)
The Strain Tensor [Ch. 4
46
where higher-order terms are neglected, since P is taken in the neighborhood
of P • Making use of (4.1.1) and (4.1.2),
0
, ) ( , ) 8u A A 8u A
( X - X - Xo - Xo = -ox x + -8u
By Y + -8z
2
(4.1.4)
, ) ( , ) 8v A 8v A Bv A
( Y - Y - Yo-Yo=-
ox x +-
By +-
YBz
2
A ow A ow A
(
I )
z - z -
( I
Zo - Zo
)
= -ow
ox x + -By Y + -8z z
Now the change in the components of the vectors A and A' can be written
oAx = A~ - Ax = (x' - x~) - (x - Xo) = (x' - x) - (x~ - Xo)

oAy = A~ - Ay = (y' - y~) - (y - Yo) = (y' - y) - (y~ -Yo) (4.1.5)
oAz = A~ - A. = (z' - z~) - (z - Zo) = (z' - z) - (z~ - Zo)
Substituting into (4.1.4) gives

8u 8u 8u
oAx = ox Ax + By Ay + Fz Az
8v 8v 8v (4.1.6)
oAy = ox Ax + By Ay + oz Az
ow ow ow
oAz = ox Ax + By Ay + oz Az
or, in tensor notation, simply
The transformation (4.1.6) is called an affine transformation, and if it

further assumed that the u1 and their derivatives are small, then it be1~011lles
an infinitesimal affine transformation (1). The tensor
8u 8u 8u
8u1 8u1 8u1 By 8z
ox
8x1 8x 2 8x 3
8v 8v 8v
8u 2 8u 2 8u 2
ox By 8z
Ut,J = Bx 1 8x 2 8x3
ow ow ow
8u 3 8u3 8u 3
ox By 8z
8x 1 8x 2 8x 3
Sec. 4-1] Strain at a Point 47
is called the relative displacement tensor. In general, as can be seen, it is not

symmetric.
Nothing has been said so far as to whether the displacements of P 0 and P
represent rigid body motions or not, or what part of them does. Since rigid
body motions play no part in the analysis of strain, it is best to eliminate
them from consideration at the start. This can be done as follows. A rigid
body motion, as mentioned earlier, is characterized by the fact that the length
of a line joining any two points remains unchanged. Consider the vector A
shown in Figure 4.1.1 and assume the vector A' to be obtained from A by
rigid body displacements. Then
A2 = A' 2 = (Ax + 8Ax) 2 + (Ay + 8Ay) 2 + (Az + 8A 2) 2

= A; + A~ + A; + 2(Ax 8Ax + Ay 8Ay + A 8Az) 2
= A2 + 2(Ax 8Ax + Ay 8Ay + Az 8Az)
where we have neglected higher-order terms in 8Ai since we are considering

only infinitesimal transformations. Therefore,
or (4.1.8)
But, from (4.1.6),

(4.1.9)
or
equation (4.1.9) must be true for all values of Ax, Ay, and A 2 , a necessary
sufficient condition that the transformation (4.1.6) represent a rigid body
is
ou = ov =ow=
ox oy oz 0
(4.1.10)
ou ov ou ow
oy = -ox oz -ox
48 The Strain Tensor [Ch. 4
That is, for rigid body motion the tensor u1,1 of equation (4.1.7) is skew-
symmetric.
Now every second-order tensor can be decomposed into a symmetric
tensor and a skew-symmetric tensor in one and only one way (see Problem 4,
Chapter 3). It follows therefore that if we decompose the tensor u1,1 into
symmetric and skew-symmetric parts, the skew-symmetric part will represent
rigid body motions, whereas the symmetric part will represent pure deforma-
tion. Therefore, let
Ut,J = -!(ut,J + UJ,t) + !(Ut,J - UJ,t) (4.1.11)
or
Ut,f = 8 tJ + Wtj
where
ou .1(8u + ov) .1(8u + ow)
ox 2
oy ox oz ox
2
Btj = .1(ou + ov) ov l.(ov + ow) (4.1.12)

2
oy ox By 2
oz By
.1(Bu + 8w) .1(BvBz + 8w) 8w
2
Bz Bx 2
8y 8z
0 .1(Bu _ Bv) .l(Bu- ow)
2
8y 8x 2 Bz Bx
wlf = tCv
Bx _ By
Bu) 0 .1(8vBz _ ByBw)2
t(Bw _ Bu) .lew_ Bv) 0

Bx Bz 2 By Bz
e11 is called the strain tensor and w 11 is called the rotation tensor. For pure
deformation therefore, equations (4.1.6) become
oA 1 = euA 1
4-2 PHYSICAL INTERPRETATION OF STRAIN

COMPONENTS
The physical meaning of the components of the strain tensor (4.1.12)

readily be determined as follows. Assume the vector A to be parallel to
x axis so that Ay = Az = 0 and Ax = A. Then equation (4.1.14)
immediately
Sec. 4-2] Physical Interpretation of Strain Components 49
Thus ex represents the extension or change in length per unit length of

a vector originally parallel to the x axis. ey and ez obviously have similar
interpretations.
To interpret the meaning of the strain component elf = ezy, use is made of
Figure 4.2.1. Consider two vectors A and 13 initially parallel to the y and z
FIGURE 4.2.1 Shear deformation.
axes, respectively. Hence if i, j, and k represent the unit vectors in the x, y,

and z directions, then A =jAy and 13 = kB2 • After deformation, these
vectors are deformed into A' and B' with components
A' = i oAx + j(Ay + oAy) + k oAz

(4.2.2)
B' = i oBx + j oBy + k(Bz + oBz)
Denote the angle between A' and B' by e. Then from the definition of the
product
I cos e= A'. Jj' = oAx oBx + (Ay + oAy) oBy + (Bz + oBz) oAz
;:;; Ay oBy + Bz oAz (4.2.3)
we have neglected the products of the vector increments, assuming

to be small. Therefore,
.ii'. B'
[(0Ax)2 + (Ay + oAy)2 + (oAz)2]112[(oBx)2 + (oBy)2 + (Bz + oBz)2]112

Ay oBy + B oAz = -oBy + -oAz
2
--"-~"-;--'-:::----'::.___.....::: (4.2.4)
AyBz Bz Ay
50 The Strain Tensor [Ch. 4 r
We have thus neglected all increments except 8A 2 and 8By and we can now
draw the figure as shown in Figure 4.2.2.
say ___ ,o'

-::...-:::.:::.. - - i1 I
R'
,,
0 1
1
p'''
f..(-8Az
0 p y
Now, from equation (4.1.14),
8A 2 = e 2 yAy
(4.2.5)
8By = By 2 B 2
Also cos 8 = cos (7r/2 - a), where a is the decrease in the right angle between
A and B. Or
cos 8 = sin a ~ a
for small angles. From (4.2.4) and (4.2.5),
a = Byz + By =
2 2ey 2
Hence eyz represents half the decrease in the right angle between two vectors
that were initially directed along the positive y and z axes. Also, from Figure
4.2.2 and equation (4.2.5),
SA
L POP' ~ tan POP' = Ayz = Bzy
L ROR' ~ tan ROR' = TSB z

= Byz
Thus angles POP' and ROR' are equal. If we now rotate the
R 'OP' Q' through an angle ey 2 about the origin, we obtain a figure such
that of Figure 4.2.3. This represents a sliding or shear ofthe elements
to the xy plane, the amount of the sliding being proportional to the
of the element from the xy plane. Hence eyz is called a shear strain.
derivations can be made for exy and exz·
Sec. 4-3] Finite Deformations 51
R'
y
4.-3 FINITE DEFORMATIONS
In the previous discussion we assumed that the displacements and their

derivatives were small; i.e., we considered only infinitesimal strains. If the
strains become too large, it is obvious that the previous formulation can no
longer be correct. In this case the strains are no longer linearly related to the
derivatives of the displacement. Furthermore, the equilibrium equations must
be satisfied in the deformed body and must therefore be written in terms of
the deformed coordinates if these are considerably different from the uncle-
formed ones.
There are two methods of describing the deformation of a continuous body,
when the deformations are large, the Lagrangian and the Eulerian. The Lag-
rangian method uses the initial coordinates of each particle to describe the
deformation. The Eulerian method uses the coordinates of the particles in the
deformed state to describe the deformation. We shall derive the elements of
the Eulerian strain tensor using tensor notation for brevity. The power of this
notation for derivation purposes, as was mentioned earlier, will be evident.
Let the coordinates of two neighboring particles before deformation be
a1 and a1 + da 1• After deformation has taken place, let the coordinates of
these particles be x 1 and x 1 + dx1• The initial distance between the particles is
(4.3.1)
the final distance between them is given by
(4.3.2)
the Eulerian description we wish to describe everything in terms of the

coordinates x 1• We therefore write
a1 = a1(x1, x2, x3 )
(4.3.3)
da 1 = a1,1 dx1 = a1," dx"
Therefore,
We can also write
Therefore,
(4.3.4)
Now the displacements are given by
U1 = X1 - al
a1 = X1 - U1
or
Hence
ds 2 - ds~ = [oil, - (81J - u1,JX811, - ul,~c)] dx1 dx1c

= [o11c - (o11c - u~c, 1 - u1,1c + u1,tu1,1c)] dx1 dx1c
= (ulc,J + Ut,lc - U1,JU1,1c) dx1 dx1c
= 2E1 ~c dx1 dx1c (4.3.5)
The tensor
(4.3.6)
is called the Eulerian strain tensor. Some typical terms in engineering notation
are
E = EJu _ 1.[(8u) 2+ (8v)2 + (8w) 2]

2
X OX OX OX OX
(4.3.7)
Exy
1
= 2
(8uEJy + OXEJv) - J.(EJu EJu
2 OX EJy +
EJv EJv
OX EJy +
ow ow)
OX EJy
We see that if the derivatives are small, so that their products can be neglected,
these reduce to the same expressions as previously obtained for the infini-
tesimal strains. The simple physical interpretations given in Section 4.2 for
small strains are, however, no longer applicable in this case. Thus if E 1 is the
change in length per unit length of an element finally parallel to the x axis,
it can be shown [2) that
Sec. 4-4] Principal Strains. Strain Invariants 53
which reduces to E 1 ~ Ex only if Ex « 1. Similarly, the decrease in the right

angle between two elements finally directed along the y and z axes is no
longer 2Eyz but is given by
which again reduces to a = 2Eyz only if Ey, E2 , and a are small.

In spite of the foregoing, however, it is possible sometimes to treat prob-
lems involving large strains using the equations for infinitesimal strains.
This is possible if we do the problem incrementally a small step at a time and
after each step change the coordinates to correspond to the deformed body
[4]. We are then solving, in essence, a successive series of small strain prob-
lems. We will elaborate on this concept more fully in the treatment of creep
problems.
4-4 PRINCIPAL STRAINS. STRAIN INVARIANTS
In the study of the stress at a point we found that there exist at least three
mutually orthogonal planes which have no shear stress acting on them, i.e.,
the principal planes. We now ask ourselves, do there exist in a similar fashion
planes with no shoor strain? By this we mean planes whose normals will not
change orientation when the body is strained. Thus a vector A, which was
originally normal to such a plane, will either shorten or lengthen but will not
change direction. The answer to this question is yes. Such planes, as in the
case of stress, are called principal planes, the normal directions to these planes
are the principal directions, and the corresponding strains are called principal
strains. To find these directions and the corresponding strains we proceed as
follows.
Consider a vector A normal to the plane ABC as shown in Figure 4.4.1.
Upon straining it is assumed that A changes length by an amount BA but its
direction remains the same; i.e., BA is in the same direction as A. Since A
and BA are parallel, the components of A and of BA are proportional; i.e.,
(4.4.1)
Now by definition the strain e in the direction of A is
BA (4.4.2)
e=A
54
l
z
X
FIGURE 4.4.1 Principal strain vector.
Hence from (4.4.1),

(4.4.3)
Writing out equation (4.1.14) we have

oAx = BxAx + BxyAy + BxzAz
(4.4.4)
oAy = BxyAx + eyAy + ByzAz
oAz = BxzAx + ByzAy + ezAz
Substituting (4.4.3) into (4.4.4) gives
(ex - e)Ax + + BxzAz =

BxyAy 0
(4.4.5)
BxyAx + (ey - e)Au + ByzAz = 0
BxzAx + ByzAy + (ez - e)Az = 0
To illustrate again the power of the tensor notation, equations (4.4.3) can
be written
and substituting into (4.1.14) gives immediately
or
which is the same as (4.4.5).

Sec. 4-5] Maximum. and Octahedral Shear Strains 55
The three homogeneous equations (4.4. 5) will have a nonvanishing solution

if, and only if, the determinant of the coefficients vanish. Thus
Byz =0 (4.4.6)
Equation (4.4.6) is of exactly the same form as (3.3.8) with stresses replaced
by strains. All the remarks and derivations made there, therefore, apply here
just as well. Thus (4.4.6) will have three real roots el> e2 , and e3 , corresponding
to the three principal strains. The invariants appearing in the cubic equation
(4.4.7)
are
I~ = Bx + By + B 2
I~ = e;y + e~ 2 + e;x - (exBy + ByB + e ex)

2 2 (4.4.8)
I~ = BxByBz + 2exyBy B x - (exB~z + ByB;x + e 2 e~y)
2 2
or in terms of the principal strains
+ e 2 + e3
I{ = e 1
If = -(e 1 e 2 + e 2 e 3 + e 3 e 1 ) (4.4.9)
I~ = e 1e 2 e 3
We note that if equations (4.4.5) are divided by A, then, since Ax/A = l,

AyjA = m, and AjA = n, Ax, Ay, and A 2 can be replaced in these equations
by l, m, and n, respectively. By substituting into these equations the principal
strains as obtained by solving the cubic (4.4.7), the principal directions can
be obtained, just as was done for the stress tensor.
MAXIMUM AND OCTAHEDRAL SHEAR

STRAINS
Just as in the case of stress, there exists a direction at every point in a

strained body for which the shear strain is a maximum. To find this direction
we can proceed as follows. Let the coordinate axes be taken in the directions
of the principal shear strains and consider a vector A, as shown in Figure
4.5.1, having direction cosines l, m, and n with respect to these axes, which
56
3
FIGURE 4.5.1 Maximum shear vectors.
are designated by 1, 2, and 3. Let the vector A suffer a small strain so that the
point P moves to Q. The strain is thus composed of two parts: a linear strain
of amount e = RQJA, and a rotation or shear of amount 8 = RPJA, for
small 8. To calculate the linear strain, e, we have
2
(A + RQ)2 = (A1 + 8A1) 2 + (A2 + 8A2)2 + (A3 + 8A3) (4.5.1)
A 2 + 2A(RQ) ;;:::: Ai + A~ + A~ + 2(A 1 8A 1 + A2 8A 2 + A3 8A3)
where squares of small quantities have been neglected. Therefore,
(4.5.2)
From the basic equation (4.1.14), which is given in expanded form in (4.4.4),
we have
8A1 = e1A1
8A 2 = e2A2
8A3 = e3A3
Making use of the relations A 1/A = 1, A2/A = m, and A 3 /A = n and sub-

stituting (4.5.3) into (4.5.2) results in
Sec. 4-5] Maximum and Octahedral Shear Strains 57
Equation (4.5.4) gives the linear strain for a vector having direction cosines
!, m, and n with respect to the principal strain directions, in terms of the
principal strains. To determine the shear 8 = RP/A, we have
2 2 2 2
2 +
e (PQ)
82 _ - --
A2
_ - 1-
A
) +-
A
(SA
2) + -3)
A
(SA (SA
or (4.5.5)
or (4.5.6)
Comparing equations (4.5.6) with (3.4.2), it is seen that they are of identical
forms, with s. replaced by 8 and the principal stresses by the principal
strains. The maximum shear strains and the corresponding directions can
therefore be obtained in the exact same way as for the stresses. Thus desig-
nating the maximum shear strains by Y1> y 2 , and y 3 , we can write directly, by
analogy to (3.4.4),
Y1 = ±!(e2 - ea)
Y2 = ±!(el - ea) (4.5.7)
Ya = ±!(el - e2)
and the same table of direction cosines is applicable as given on page 35.
Thus the maximum shearing strain acts on a plane bisecting the angle between
the maximum and minimum principal strain directions and is equal to half
the difference of these strains.
If we consider the octahedral planes for which I = m = n = ± ljv3, we
see from (4.5.6) that the shear strain on these planes, which we designate
Yoct> is given by
Y;ot = j-(e~ + e~ + e~) - !(el + 82 + ea) 2

= t[(el - e2) 2 + (e2 - ea) 2 + (ea - e1) 2] (4.5.8)
in exact analogy to the octahedral shear stress. In terms of the invariants of

the strain tensor in analogy to (3.4.10) we have
Y;ct = ~-[1{2 + 31~] (4.5.9)
and in terms of nonprincipal strains this becomes

It is also apparent by the same analogy [comparing equations (4.5.4) and

(4.5.5) with (3.5.1)] that a Mohr's diagram can be constructed for the strains
in an identical fashion to the stresses.
4-6 STRAIN DEVIATOR TENSOR
As in the case of the stress tensor, the strain tensor can be separated into
two parts, a spherical part q11 and deviator part e11 • The spherical part is
given by
(4.6.1)
where Bm = t(e1 + Bz + B3) is the mean strain. The deviator strain then
becomes
2Bx- B2 - By
Bxy Bxz
3
2By- Bz- Bx
Bxy Byz (4.6.2)
3
2B 2 - B x - By
Bxz Byz
3
or, in terms of the principal strains,
2B 1 - Bz - B3
0 0
3
2B 2 - B3 - B1
el1 = 0
3
0
2B3 - B1 - Bz
0 0
3
If we consider a rectangular parallelepiped with sides equal to a, b, and c,

so that its initial volume is abc, then after straining its volume will be
or abc + abc(e 1 + e2 + B3 )
Sec. 4-7] Compatibility of Strain 59
neglecting products of the strains. The change in volume per unit volume is
then
The spherical strain tensor is thus proportional to the volume change. The
deviator strain tensor then represents a pure distortional strain.
By analogy with the stress deviator tensor discussed in Section 3.6, the
invariants of the strain deviator tensor are
J{ = 0
J~ = -!(I? + 3!~) (4.6.4)
J~ = -f?- (2!? + 9I{I~ + 27!~)
or
J~ = ifh - e2) 2 + (e2 - ea) 2 + (ea - e1) 2]
= i[(ex - ey) 2 + (ey - e2 )2 + (ez - ex) 2 + 6(e~y + e;z + e~x)]
= -(e1e2 + e2es + e3 e1 ) (4.6.5)
J~ = e1e2ea
where
Also it follows that
(4.6.6)
4-7 COMPATIBILITY OF STRAIN
The strain tensor eu defines six strain components in terms of three dis-
placements; i.e.,
8u 8v
e =- e =-
x ox Y By
(4.7.1)
_(ou ov)
1
exy- 2 8y + OX eyz =
1
2
(8vf}z + ow)
By
_
exz-
1
2
(8u ow)
f}z + ox
It is obvious that if the displacements are specified as continuous functions

of x, y, and z, then one can use equations (4.7.1) to compute the strain com-
ponents uniquely. Let us now consider the inverse problem, i.e., to calculate
the displacements u, v, and w in a body given the strain components e11 • Here
several problems are encountered. In the first place, it is apparent that the
solution will not be unique, for the strains represent pure deformation,
whereas the displacements include rigid body motion which have no effect
on the strains. The problem can, however, be made unique by specifying the
rigid body motion, i.e., specifying the displacement and rotation at some
point in the body.
However, a more difficult problem is encountered in calculating the dis-
placements from the strains. Equations (4.7.1) are six equations for the three
unknowns u, v, and w. It is evident, therefore, that these equations will not
have a solution for any arbitrarily chosen strains, but that some restrictions
must be placed on the strains in order that equations (4.7.1) have a solution.
This can also be seen from the following physical considerations. Assume
the body is divided into infinitesimal cubes. Let all these cubes be separated
from each other and let each of the cubes be subjected to some arbitrary
strains. It is obvious that if we now try to put all the cubes together, we will,
in general, no longer be able to fit all the cubes together the way they were
before, to produce a continuous body. Between some of the cube boundaries
there will be gaps; others will overlap. This shows that there must be some
relationships between the strains at the different points of a body in order
that the body remain continuous after straining, i.e., that the displacements
be continuous functions of the coordinates. These relationships are called
the compatibility relations, or, sometimes, the continuity relations.
The compatibility equations are found in all standard texts on elasticity.
They are derived, for example, in a particularly elegant fashion in reference
[3]. In engineering notation these equations are
(4.7.2)
8 ( 8eyz
8ezx
- --
ox - +8exy)
8x + oy -
8z
These equations were first derived by Saint-Venant.

Problems 61
Equations (4. 7.2) are necessary and sufficient conditions that the strain
components give single-valued displacements for a simply connected region.
For a multiply connected region, however, these conditions are necessary but
generally not sufficient.
Just as for the stress equations of Chapter 3, it should also be emphasized
here that all the relations presented in this chapter, including the com-
patibility relations (4.7.2), are independent of the material properties and
therefore hold for both elastically and plastically behaving materials.
Note on Shear Notation
The shear strains defined herein differ from the shear strains defined by
many authors by a factor of 1/2. Thus Timoshenko, for example, defines the
shear strains by
ou ov
YxY = oy + ox
ou ow
Yxz = OZ + OX
ov ow
YYz = oz + oy
The factor 1/2 used herein is necessary for the tensor definition of strain as
exemplified by equation (4.1.12). Care must be exercised in reading the
literature to determine which definition of shear strain is used.
Problems
1. Obtain the equations representing rigid body displacements; i.e., show that
U = a + WyZ - WzY
V = b +W X
2 - WxZ
W = C + WxY - WyX
where a, b, and c are constants and

2. Let the rectangle of Figure 4.2.3 be a square. Show that the shear strain is
equal to the extension of the diagonal of this square.
3. The displacement vector at a point in a body is given by
il = C[i(lOx + 3y) + j(3x + 2y) + k(6z)]
where C is a constant. Show that there is no rotation and compute the

principal strains.
4. Given the strain tensor
-0.005 -0.004
-0.~04
e11 =
[
0.001
0 o.Ll
Determine:
(a) The principal strains.
(b) The direction cosines of the principal directions.
(c) The largest shearing strain.
(d) The octahedral shear strain.
5. If a rectangular parallelepiped with initial dimensions of 1 in. in the x direc-
tion, 2 in. in the y direction, and 3 in. in the z direction is strained to the
condition of Problem 4, determine the final dimensions.
6. Derive the Lagrangian strain tensor.
7. Show that in plane polar coordinates the infinitesimal strains are given by
ou 1 ov u
e, = or 89
= ro(J + r
ov 1 ou v
8 9
' = or + o(J - r r
where u and v are the displacements in the radial and tangential directions,
respectively.
8. Assuming axial symmetry, derive the compatibility equation in plane polar
coordinates.
9 Verify equations (4.5.9) and (4.5.10).
10. Verify equations (4.6.4) and (4.6.5).
References
1. I. S. Sokolnikoff, Mathematical Theory of Elasticity, McGraw-Hill, New
York, 1956, p. 21.
2. Ibid., p. 31.
3. Ibid., p. 25.
4. W. Prager and P. G. Hodge, Theory of Perfectly Plastic Solids, Wiley, New
York, 1951, p. 118.
References 63
General References
Novozhilov, V. V., Theory of Elasticity, Pergamon Press, London, 1961.
Sokolnikoff, I. S., Mathematical Theory of Elasticity, McGraw-Hill, New York,
1956.
Timoshenko, S., and J. N. Goodier, Theory of Elasticity, McGraw-Hill, New
York, 1951.
CHAPTER 5
ELASTIC
STRESS-STRAIN
RELATIONS
In the previous chapters the states of stress and of strain at a point in a

body have been defined and some of their properties discussed. Equations
(3.2.2) give the conditions of equilibrium that must be satisfied by the stresses
at every point of a body, whereas equations (4. 7.2) represent the compatibility
conditions that must be satisfied by the strains. The boundary conditions in
terms of surface forces are given by equations (3.3.1). There remains to dis-
cuss the relations between the stresses and the strains at every point of the
body. We shall discuss briefly these relations for an elastic body, as a prel-
ude to the general elastoplastic stress-strain relations to be subsequently
discussed.
5-l EQUATIONS OF ELASTICITY
Hooke first proposed a linear relation between stress and strain for a load
applied in one direction. The generalization of Hooke's law to three dimen·
sions is given in Problem 3, Chapter 3. For an isotropic material this becomes,
using the tensor subscript notation and including thermal strains,
where E is the elastic modulus, J-t Poisson's ratio, a the coefficient of

64
Sec. 5-l] Equations of Elasticity 65
thermal expansion, T the temperature above some arbitrary reference tem-

perature, G the shear modulus related to E and 1-" by the well-known relation
E
G = 2(1 +f.") (5.1.2)
and
(5.1.3)
Equation (5.1.1) can be solved for the stresses to give
(5.1.4)
where
8= Bx + By + B = 2 If
(5.1.5)
.\ = 1-"E
(1 + f.")(l - 2fl,)
In engineering notation equations (5.1.1) become
1
By = E [ay - f."( ax +a 2 )] + aT
(5.1.6)
1 1 + 1-"
Byz = 2G 7'yz = ~ 7'yz
1 1 + 1-"
Bzx = 2G,7'zx = ~ 1'zx
It readily follows from (5.1.6) that
1- 2fl,
8= - E - 0 + 3aT (5.1.7)
1 - 2fl,
Bm = --E-am +aT (5.1.8)
66 Elastic Stress-Strain Relations (Ch. 5
where em and am are the mean strain and mean stress, respectively. Finally,
r
combining (5.1.8) and (5.1.1) results in
(5.1.9)
where e!J and Sli are the strain deviator and stress deviator tensors, respec-
tively. Thus the deviators of the stress and strain tensors are related to each
other, in the elastic case, by the simple equation (5.1.9), whereas the spherical
stress components are related to the spherical strain components by equation
(5.1.8).
It also follows that
JJ. = Jl = 0
(5.1.10)
J~ = 4~2J2
2 1 2
(5.1.11)
Yoct = 4G2 '~"oot
It should be noted that nothing in the foregoing discussion requires that

E, G, fJ-, or a be constant throughout the body. They may, for example, be
functions of temperature, so that if the body is not at a uniform temperature,
these constants may have different values at different points in the body.
It is probably worthwhile to briefly discuss the origin of the aT terms
appearing in equations (5.1.1). We start by defining the free thermal expansion
of a material as that part of the expansion which is uninfluenced by stress
but is due to temperature rise alone. We also make the following four
assumptions:
1. The material is isotropic and, therefore, the free thermal expansion is

the same in all directions.
2. The free thermal expansion is directly proportional to the tetnpt~ratur,e.,
3. The principle of superposition of strains holds.
4. The thermal expansion does not influence the shear strains.
Let us now denote the strains due to temperature rise by primes and
strains due to stresses by double primes. Then assumptions 1 and 2
us to write
e~ = e~ = e~ = aT
Sec. 5-2] Elastic Strain Energy Functions 67
where a is the coefficient of linear thermal expansion. From Hooke's law

we of course have
and now making use of assumption 3, the total strains are as given by the
first three of equations (5.1.6), whereas Hooke's law and assumption 4 give
the last three of equations (5.1.6). Note that assumption 2 may be modified
to the extent that a may be a function of temperature.
5-2 ELASTIC STRAIN ENERGY FUNCTIONS
If a body is deformed by the action of external forces, then the latter do

work on the body. If the body is in equilibrium so that none of this work
goes into kinetic energy, then this work is stored as strain energy of deforma-
tion in the body. The elastic strain energy can be written as follows [1]:
(5.2.1)
If we substitute into (5.2.1) equation (5.1.4), there results
U = Gelie11
A
+Z 8 2
-
2G + 3A aT8 (5.2.2)
2
Making use of the definitions of the invariants of the strain and strain
deviator tensors, equation (5.2.2) can be written after some algebraic manip-
ulations as
U=
20
+ 3A (I~
6
2
- 3aTI~) + 2GJ~ (5.2.3)
Now I~ represents the spherical state of strain which is proportional to the

change, whereas J~ is an invariant of the deviator strain tensor which
repres(mts a pure distortional strain. Equation (5.2.3) can therefore be con-
as the sum of two energy terms. The first term is the energy involved
68 Elastic Stress-Strain Relations [Ch. 5
in changing the volume; the second term is the energy of distortion, which is
designated by Ua and can be written
Ua = 2GJ~ = 3Gy~ot
or, from (5.1.10),
Ud = 1G J.2 = 4G
3 2
Toot (5.2.4)
2
5-3 SOLUTION OF ELASTIC PROBLEMS
To solve an elastic problem we have to find the stresses and the strains
which will satisfy the previously derived equations. Thus the stresses must
satisfy the equilibrium equations, (3.2.2) and (3.2.3), as well as the boundary
conditions (3. 3. 5). The strains must satisfy the compatibility equations (4. 7.2).
Finally, the stresses must be related to the strains through the stress-strain
relations (5.1.1) or their equivalent. The problem of finding a set of stresses
and strains satisfying the above relations is known as the .first boundary-value
problem of elasticity. Alternatively, it is possible to reduce the above set of
equations to three equations, called the Navier equations, involving only the
displacement u1 [2].
where V 2 is the Laplacian operator,;'\ and G are Lame's constants previously

defined, F 1 are the body forces per unit volume, and 8 is the invariant, defined
in (5.1.5) and which can also be written
This last formulation is most convenient if the displacements are specified

on the boundary rather than the forces. The problem is then known as the
second boundary-value problem of elasticity.
From the previous derivations it is apparent that both the equilibrium
equations and compatibility relations are independent of the properties of .
the material under consideration. The equilibrium equations express the
static equilibrium of an element of the body and the compatibility relations
express the continuity of the body. These equations will therefore hold
whether the body behaves elastically or whether plastic flow occurs. The same
holds for the boundary conditions. The difference between the elastic problem
References 69
and the plastic problem occurs in the stress-strain relations. If plastic flow
occurs, the linear generalized Hooke's law no longer holds and the relations
between stress and strain become nonlinear in a very complicated manner.
What happens to the relations between stress and strain when plastic flow
occurs will be the subject of Chapters 6 and 7.
Problems
1. Show how equation (5.1.4) can be obtained from equation (5.1.1).
3. Given the following stress tensor at a point:
10,000 1,000 -8,000
1,000 -6,000 6,000
-8,000 6,000 20,000
Assume the material behaves elastically with an elastic modulus of 30 x 106
and Poisson's ratio of 0.3. Determine the strain deviator tensor. Calculate
the mean strain assuming the temperature rise at the point is 300°F and that
a = 7.5 x 10- 6 in./in.;aF.
5. Prove that the axes of principal stress coincide with the axes of principal
strain for an isotropic material obeying Hooke's law.
6. Prove that for the material of Problem 5, Mohr's stress diagram will be
similar to Mohr's strain diagram.
References
York, 1956, p. 84.
2. Ibid., p. 73.
General References
Hoffman, 0., and G. Sachs, Introduction to the Theory of Plasticity for Engineers,
McGraw-Hill, New York, 1953.
Sokolnikoff, I. S., Mathematical Theory of Elasticity, McGraw-Hill, New York,
1956.
CHAPTER 6
CRITERIA FOR
YIELDING
6-1 EXAMPLES OF MULTIAXIAL STRESS
In Chapter 2 the stress-strain curve for a material in simple tension was

discussed and it was shown that there exists a yield point at which the material
will begin to deform plastically. In this case the stress is uniaxial and this
point can readily be determined. But what if there are several stresses acting
at a point in different directions? What combination of these stresses will
cause yielding? We know, for example, that for a hydrostatic stress, i.e.,
equal stresses in all directions, yielding does not occur even for very large
values of stress. As another example involving a test which can be performed
without too much difficulty, consider a thin-walled cylinder which is being
pulled axially by a load P, is being twisted by a twisting moment T, and is
pressurized internally (see Figure 6.1.1) by a pressure p.
By varying the pressure p, the axial pull P, and the torque T, it is possible
to get various combinations of stresses, which will also of course result in
FIGURE 6.1.1 Combined stresses in thin-walled cylinder.
70
Sec. 6-2] Examples of Yield Criteria 71
different principal directions. The question here is: For what combinations
of loads will the cylinder begin to yield plastically?
Another simple example is the plane stress problem of a thin rotating disk
with or without temperature gradients. At every point of the disk (except
possibly at the rim) there exists a state of biaxial stress. The question again
arises: For which states of biaxial stress will the disk deform plastically? The
criteria for deciding which combination of multiaxial stresses will cause
yielding are called yield criteria. The first step of any plastic flow analysis is
to decide on a yield criterion. The next step is to decide how to describe the
behavior of the material after yielding has started. In this chapter we shall
discuss the choice of a yield criterion.
6-2 EXAMPLES OF YIELD CRITERIA
Numerous criteria have been proposed for the yielding of solids, going as
far back as Coulomb in 1773. Many of these were originally suggested as
criteria for failure of brittle materials and were later adopted as yield criteria
for ductile materials. Some of the more common ones will be briefly discussed.
Although some of these theories are no longer in use, they are included here
both for their historic interest and to give the reader a feeling for the type of
approach used in promulgating yield criteria.
Maximum Stress Theory, or Rankine Theory
This theory assumes that yielding occurs when one of the principal stresses
becomes equal to the yield stress in simple tension a 0 , or the yield stress in
compression ao,c· Thus if a 1 is the maximum principal stress and a 2 is the mini-
mum principal stress, yielding will occur in tension when a 1 = a 0 and it will
occur in compression when a 2 = a 0 , 0 • For a material with the same yield in
tension and compression, this criterion becomes
or (6.2.1)
A simple plot illustrating this criterion for the case of biaxial stress with
aa = 0 is shown in Figure 6.2.1. The coordinates are the remaining principal
stresses a 1 and a 2 • Yielding occurs when the state of stress is on the boundary
of the rectangle, for then one of the stresses is at the yield point in tension or
compression. For example, consider a thin-walled cylinder subjected to an
72 Criteria fo:r Yielding [Ch. 6
0"2
uo
o,c /
....1/-' cro
CTj
cro,c
FIGURE 6.2.1 Maximum stress theory.
increasing internal pressure p. Let a 1 be the circumferential stress and a 2 the

axial stress. Then a 1 = 2a 2 • As the pressure is increased from zero, the
stresses follow the dashed line of Figure 6.2.1, as shown, a 1 always being
equal to twice a 2 • At point 1 the cylinder is still elastic, neither stress having
reached the value of a 0 • At point 2, a 1 is equal to a 0 and yielding begins even
though a2 is only !a 0 • This maximum stress criterion, however, shows very
poor agreement with experiment and is rarely used.
Maximum Strain Theory, or Saint-Venant Theory
This theory assumes yielding will occur when the maximum value of the
principal strain equals the value of the yield strain in simple tension (or
compression), e0 = a 0 /E. Thus if e 1 is assumed to be the largest strain in
absolute value, yielding will occur when
(6.2.2)
or, for the biaxial case with a 3 = 0,
Ee 1 = a 1 - fl-Uz = ±a0 for la1! :2:: iazl

Ee 2 = a2 - wr 1 = ±a0 for ja2 j :2:: !a1!
A plot in the a 1 a 2 plane showing the boundary at which yielding begins is

shown in Figure 6.2.2. This theory also does not agree well with most
ments. It has, however, been used in the design of guns, since some
mental results on thick-walled cylinders are in agreement with this theory
crz
FIGURE 6.2.2 Maximum strain theory.
Maximum Shear Theory, or Tresca Criterion
This theory (sometimes called the Coulomb theory) assumes that yielding
will occur when the maximum shear stress reaches the value of the maximum
shear stress occurring under simple tension. The maximum shear stress is
given by equation (3.4.4) and is equal to half the difference between the
maximum and minimum principal stresses. For simple tension, therefore,
since a 2 = a 3 = 0, the maximum shear stress at yield is !a0 • The Tresca
criterion then asserts that yielding will occur when any one of the following
six conditions is reached:
(6.2.4)
For the biaxial case with a3 = 0, we have
if a1 > 0, a2 < 0
if a1 < 0, a2 > 0
if a 2 > a1 > 0
(6.2.5)
if a 1 > a 2 > 0
if a 1 < a 2 < 0
if a 2 < a1 < 0
74 Criteria for Yielding [Ch. 6
A plot in the a 1a 2 plane for this yield criterion is shown in Figure 6.2.3. It is
to be noted that one limitation of this theory is the requirement that the yield
FIGURE 6.2.3 Maximum shear stress theory.
stresses in tension and compression be equal. The Tresca criterion is in fair

agreement with experiment and is used to a considerable extent by designers.
It suffers, however, from one major difficulty-it is necessary to know in
advance which are the maximum and minimum principal stresses. For the
case of pure shear (see Section 3.7),
the Tresca criterion predicts yielding to occur when
or
That is, the yield stress in pure shear is t the yield stress in simple tension.
Maximum Strain Energy Theory, or Beltrami's Energy Theory
This theory assumes that yielding will occur when the total strain
per unit volume equals the total strain energy per unit volume at yielding
uniaxial tension or compression. The total strain energy at yield in
tensile test is
and the total strain energy U is given by
U= -! (a1e1 + a 2 s 2 + asss)
= 2~ [a~ + a~ + a~ - 2p,(a 1 a 2 + a 2 as + asa1)]
The yield criterion becomes
(6.2.6)
For the biaxial case this becomes
(6.2.7)
This is the equation of an ellipse in the a 1a 2 plane. It is apparent from

(6.2.6) that yielding can occur under sufficiently high hydrostatic pressure
a 1 = a 2 = a 3 , which, as has been shown, is contrary to experiment. This is
also apparent from equation (5.2.4), where it is shown that only the distor-
tional strain energy can contribute to yielding. This theory has therefore
been superseded by the theory described next.
Distortion Energy Theory, or the von Mises Yield Criterion
The distortion energy theory (also associated with Hencky) assumes that
yielding begins when the distortion energy equals the distortion energy at
yield in simple tension. Thus from equation (5.2.4),
_lr_3 2
Ud - 2G"2- Toot
40
At th~ yield point in simple tension, from (3.6.8),
Therefore the yield condition becomes
(6.2.8)
for the biaxial case,

(6.2.9)
FIGURE 6.2.4 Distortion energy theory.
This plots as an ellipse, called the von Mises ellipse, in the a 1 a 2 plane, as
shown in Figure 6.2.4. For the case of pure shear
J2 = i-[(al- a2) 2 + (a2 - a3) 2 + (a3- a1)2]

=a~= k2
and the von Mises criterion would predict yielding to occur when
or k= ~3
That is, the yield stress in pure shear is l/V3 times the yield stress in simple
tension. Thus the von Mises criterion predicts a pure shear yield stress which
is about 15 per cent higher than predicted by the Tresca criterion. It will
subsequently be shown that this is the maximum difference between the two
criteria.
The von Mises yield criterion usually fits (but not always) the experimental
data better than the other theories, and it is usually easier to apply than
Tresca criterion because no knowledge is needed regarding the
magnitudes of the principal stresses. For these reasons, this criterion
widely used at the present time. If, however, the relative magnitudes of
principal stresses are known, as, for example, in the case of the
tube discussed in Chapter 8, the Tresca criterion is easier to apply.
Von Mises originally proposed his criterion because of mathematical
convenience. Hencky later showed that it was equivalent to assuming that
yielding will take place when the distortion or shear strain energy reaches a
critical value, as shown above. Also, since the octahedral shear stress is
equal to
which for simple tension at yield becomes
'roct,O =
v2
3 Go
then equation (6.2.8) can be written
Toot = Toot,O
That is, yielding will occur when the octahedral shear stress reaches the
octahedral shear stress at yield in simple tension.
Alternatively, the criterion (6.2.8) can be looked upon as stating that
yielding will occur when the second invariant J 2 of the stress deviator tensor
reaches a critical value, i.e., the value of J 2 at yield in simple tension. The
assumption that the yield criterion should depend on the invariants of the
stress deviator tensor is generally accepted, as will be discussed in the next
section.
Mohr's Theory of Yielding
Mohr extended the maximum shear stress theory by assuming that the
critical shear stress is not necessarily equal to the maximum shear stress but
depends also on the normal stress acting on the shearing plane. In general,
the greater the normal stress, the lower is the critical shear stress. Mohr's
theory therefore takes into account the effect of mean stress, which has
experimentally been shown to be important in some cases, particularly with
regard to fracture. Mohr's theory can also take into account differences in the
points in tension and compression, as exemplified by the internal friction
This is a special case of Mohr's theory, in which the critical shear stress is
ouu"'""'u to be a linear function of the normal stress acting on the plane of
Criteria for Yielding [Ch. 6
78
maximum shear. This results in the following conditions for the biaxial case
[la]: If a 1 > a 2 > a 3 , a 2 = 0,
ao (6.2.10)
a l - - a3 = ao
ao,c
2
a3
2
where a ,c is the yield stress in compression. If a 1 > > a , a = 0, then
0
(6.2.11)
If a 1 > a 3 , both negative, then
a3 = -ao,c
A plot in the a 1 a 3 plane is shown in Figure 6.2.5.
cr1 =- ao,c
FIGURE 6.2.5 Internal friction theory.
The various yield criteria listed by no means exhaust the available list.
Besides, there are other, more complicated theories which attempt to take
into account anisotropy. For example, a yield criterion proposed by
[2] for anisotropic materials with different yields in tension and compression
has the form (for biaxial stress)
Sec. 6-3] Yield Surface. Haigh-Westergaard Stress Space 79
where a 11 and a 12 are the tensile yields in the a 1 and a 2 directions, a 01 is the
compressive yield in the a 1 direction, and To is the shear yield strength.
6-3 YIELD SURFACE. HAIGH-WESTERGAARD

STRESS SPACE
In Section 6.2 we discussed several yield criteria and also plotted several
two-dimensional plots for biaxial stress cases, showing the curves at which
yielding takes place. In the most general case, the yield criterion will depend
on the complete state of stress at the point under consideration and will
therefore be a function of the nine components of stress at the point. Since
the stress tensor is symmetric, we can reduce this function to a function of the
six independent components of the stress tensor. This yield criterion for a
virgin material is then essentially the extension of the single yield point of the
uniaxial tensile test to the six-component stress tensor. For a material loaded
to the initial yield, it can be expressed by the relationship
F(alf) = K (6.3.1)
where K is a known function, or if desired,
(6.3.2)
Equation (6.3.1) represents a hypersurface in the six-dimensional stress

space and any point on this surface represents a point at which yielding can
begin. For example, for the simple tensile test at the yield point a 0 , the point
ax= a 0 , ay = az = 'Txy = 'Txz = 'Tyz = 0, must lie on this surface, and the
point ax = ay = az = 'Txz = 'Tyz = 0, 'Txy = To, which represents a thin-
walled tube loaded in torsion to the torsional yield To, must also lie on this
surface.
The function appearing in (6.3.1) is called the yield function and the surface
described by (6.3.1) or (6.3.2) in the stress space is called the yield surface.
Without specifying yet any particular form for this surface, the equation
describing it can be simplified somewhat by making use of some of the
previously discussed assumptions regarding the yielding of metals. If, as
isotropy is assumed so that rotating the axes does not affect the yielding,
can choose the principal axes for the coordinates, and then (6.3.1) can be
(6.3.3)
Furthermore, since it is always assumed that hydrostatic tension or compres-

sion does not influence yielding, we can assume that only the stress deviators
enter into the yield function and write
(6.3.4)
Alternatively, since S1o S 2 , and S3 can be written in terms of the invariants

J10 J2, and J 3 , where
J1 = s1 + s2 + s3 = o
J2 = -!-(S~ + S~ + S5) = -(S1S2 + S2S3 + S3Sl) (6.3.5)
J3 = tcs~ + s~ + s~) = s1s2sa
we can write
(6.3.6)
Subject therefore to the above two assumptions, the yield criterion has been
reduced to a function of the two nonzero invariants of the stress deviator
tensor. We note in passing thatj(J2 , J 3 ) is symmetric in the principal stresses,
which is to be expected since in an isotropic material all of the principal
stresses must play the same role in yielding. Thus whatever yield function is
chosen, it must be symmetric in the principal stresses.
The two most widely used yield criteria, the Tresca maximum shear
criterion and the von Mises yield criterion, discussed in Section 6.2, are
specific cases of (6.3.6). The von Mises criterion is by far the simplest one
that can be associated with equation (6.3.6),
(6.3.7)
where k is the yield in pure shear.

The Tresca criterion can be written in the form
and we see the great complexity of the Tresca criterion in its general form
compared to the von Mises criterion. Only in the case, as previously discussed,
where the maximum and minimum principal stresses are known a priori can
the Tresca criterion be reduced to the simple form
(
We can learn a great deal about the yield surface defined by equation
(6.3.3) from simple geometric considerations. We introduce the (a 1 , a 2, as)
coordinate system, which represents a stress space called the Haigh- Wester-
gaard stress space [3]. Every point in this space having coordinates al> a 2 , and
as is a possible stress state. Consider a line ON as shown in Figure 6.3.1
FIGURE 6.3.1 Haigh-Westergaard stress space.
passing through the origin, and having equal angles with the coordinate axes.
Then for every point on this line the stress state is one for which
(6.3.10)
Thus every point on this line corresponds to a hydrostatic or spherical stress

state, the deviatoric stresses sl = (2a1 - a2 - as)/3, etc., being equal to zero.
Furthermore, if we consider any plane perpendicular to ON, the equation
of this plane will be
(6.3.11)
where p is the distance along the normal from the origin to the plane. Hence
the spherical component of the stress tensor increases linearly with the
distance of the plane from the origin. On the plane passing through the
origin, the spherical stress is zero, the equation being a 1 + a 2 + as = 0. This
plane is called the 1r plane.
Now consider any arbitrary stress state such as at point P with stress
:;comr>onrents a 1 , a 2 , and as. The stress vector OP can be decomposed into two
cornp<ments, the component A parallel to ON and the component 1i per-
pertdli~Ul;ar to ON, the latter parallel to the 1r plane. The component A can
readily be determined by projecting the components of OP-at. a 2, and a 3-

on to 0 N. Thus
1 1 1
A = VJ a 1 + yj a2 + yj a 3
or A= v3 am (6.3.12)
To determine Jj we have
= ar + a~ + a~ - 3a~
= (al - am)2 + (a2 - am)2 + (aa - am)2
= sr + s~ + s~
= 212 (6.3.13)
The components of Jj are therefore the stress deviators St. S 2 , and S 3 • We

have thus decomposed the arbitrary stress state at P into a spherical part A
and a deviatoric part JJ, the latter parallel to the 1r plane. Furthermore, if we
take a different stress state P' lying on a line through P parallel to ON, then
obviously the projection of the vector OP' onto the 7T plane will be the same
as the projection of the vector OP onto this plane. The two states of stress,
at P and at P', will therefore have the same deviatoric components and will
differ only in their spherical parts. In fact, all points on the line through P
parallel to 0 N will correspond to the same deviatoric states of stress.
Since it is assumed that yielding is determined by the deviatoric state of
stress only, it follows that if one of the points on the line through P parallel
to ON lies on the yield surface, they must all lie on the yield surface, since
they all have the same deviatoric stress components. Hence the yield surface
must be composed of lines parallel to ON; i.e., it must be a cylinder
generators parallel to ON. The only assumption needed to arrive at
conclusion is the dependence of the yield surface on the deviatoric
components only.
The intersection of this yield cylinder with any plane perpendicular to
will produce a curve called the yield locus. Since this curve will be the
for all planes perpendicular to the cylinder, we need consider only the
locus on one such plane. For this purpose we choose the 7T plane, on which,
was pointed out previously, the spherical stress state is zero. Figure 6.
shows the 7T plane as the plane of the paper and the projections upon
plane of the coordinate axes a 1 , a 2 , and a 3 • Since this plane makes
angles with the coordinate axes, their projections upon this plane must
FIGURE 6.3.2 1T plane.
equal angles with each other, 120°, as shown. It can be shown from symmetry
considerations that the yield locus must have the same shape in each of the
twelve 30° sectors dividing the 1r plane as shown in the figure. This follows
from the following considerations. The yield surface must be symmetric in
the principal stresses since it certainly does not matter, for example, if we
interchange the values of a 2 and as. It follows, therefore, that the lines
bisecting the angle between any two principal axes in the plane must be lines
of symmetry. The lines a 1 , - a 1 ; a 2 , - a 2 ; and as, -as are therefore lines of
symmetry and we now have six symmetric sectors. Furthermore, if we assume
equal yielding in tension and compression, then if we go from a point Q on
the yield locus to the point Q', where all the stresses have reversed signs, we
should again be on the yield locus. Therefore, if a 1 goes to -a1 , we must
have symmetry about a line perpendicular to the a 1 axis; if a 2 goes to - a 2 ,
we must have symmetry about a line perpendicular to the a 2 axis; and if as
goes to -as, we must have symmetry about the line perpendicular to the as
axis. We have thus divided the yield locus into 12 symmetric sectors, each of
, and we need only consider the stress states lying in one of these sectors.
To determine the location on the 1r plane of the projection of the point P
on the yield cylinder, we proceed as follows. Let r and 8 be the polar
~·~··"""""' of the point Pin the 1r plane, 8 being measured from the horizon·
axis, and let a and b be the horizontal and vertical components of r, as
in Figure 6.3.3. P has components a 1 , a 2 , and as, and in the 1r plane
will project as V2/3 a 1 , V2/3 a 2 , and V2/3 as, respectively. This can
seen in Figure 6.3.4, which shows the plane containing the vector a 1 the
Criteria £or Yielding [Ch. 6
84
FIGURE 6.3.3 Projection of point P on 7T plane.
normal to the plane, ON, and the projection of a 1 on the 7T plane. It is obvious
1
from the figure that the projection of a 1 on the 7T plane is equal to V2{3 a •
From Figure 6.3.3:
a = V2{3 a2 cos 30° - V2/3 a 1 cos 30° = (a 2 - lll)fVZ
b = V2/3 as - V2f3 a 2 sin 30° - V2f3
a 1 sin 30°
= l/v'6 (2as - a 2 - a1) (6.3.14)
rz = az + bz =
(a 2 -a
2 1
)2 + l(2as - az - a1)2
2
= i-[(al - az) 2 + (az - as) + (as - a1)2]
2
= [(al - am) 2 + (az - am) 2 + (as - am) ]
= 21 (6.3.15)
2
1T plane
FIGURE 6.3.4 Projection of a1 on 7T plane.

Let us now consider the yield locus for the von Mises yield criterion
[equation (6.3.7)]:
From (6.3.15) we have

(6.3.16)
The yield locus is therefore a circle of radius r = V2/3 a 0 , as shown in

Figure 6.3.5. The yield surface in the Haigh-Westergaard stress space will
FIGURE 6.3.5 von Mises circle and Tresca hexagon.
then be a circular cylinder whose axis is the line ON equally inclined to the
stress axes.
The yield locus for the Tresca maximum shear criterion is a regular hexagon
inscribed in the von Mises circle, as shown in Figure 6.3.5. This can be proved
as follows.
Consider the sector between the a 2 and -a1 axes in Figure 6.3.5. For any
stress state in this sector,
The Tresca yield condition for this sector is then a 2 - a 1 = a 0 • This will be
represented by a line parallel to the a 3 axis, since it is independent of a 3 , the
distance of this line from the a 3 axis being equal to l/v2 (a2 - a 1 ) =
1/Vl a0 [by equation (6.3.14)]. In a similar manner, a straight line is obtained
for each sector, thus forming the hexagon shown in the figure. The corners of
hexagon will touch the von Mises circle, as can be seen again from the
sector, since the radius to the corner is equal to
r = (1/V2)a0 = yi3 a 0
cos 30°
which is the radius of the von Mises circle. In the stress space the Tresca yield
surface is a regular hexagonal cylinder inscribed in the von Mises cylinder, as
shown in Figure 6.3.6.
FIGURE 6.3.6 Tresca and von Mises cylinders.
It is now evident from Figure 6.3.5 that the maximum difference between
the von Mises and Tresca criteria occurs at 8 = 0° and is equal to
as previously mentioned.
The line at 8 = oo corresponds to pure shear stress states. This follows
from (6.3.15), for e = 0:
and the mean stress is

am = t(al + Uz + as) = t(al + az)
Therefore, if we subtract the mean stress from the stress components a 1 , a 2 ,
and as, we get a stress state -!-(a 1 - a 2), f(a 2 - a 1), 0, which corresponds to a
state of pure shear (seep. 41). In then plane itself, since a 1 + a 2 + as = 0,
it follows that as = 0 and a 1 = - a 2 • For any other plane parallel to the
n plane, we have a case of pure shear, as indicated, with the addition of a
spherical state of stress of amount !(a1 + a 2 ).
The line at 8 = 30° corresponds to a uniaxial stress, since
2_a--"3'---a-=2'---a-=1 = 1
Uz- a1
or
If we subtract a hydrostatic stress of a 2 , we get a stress state a 1 - a 2 , 0, 0.

It can now readily be shown that if the yield locus is assumed to be convex
and one circumscribes the von Mises circle by a regular hexagon, then all
possible yield loci must lie between the two regular hexagons inscribed in,
and circumscribing, the von Mises circle. By the locus being convex, we mean
simply that any straight line in the 7T plane may pierce the locus in at most
two points. It will be proved later that the yield surface must indeed be
convex; for the present we shall merely assume it. Let the point A on the a 3
axis in Figure 6.3. 7 lie on the yield locus. By the symmetry conditions pre-
viously discussed, the points B and C must also lie on the yield locus. The
FIGURE 6.3.7 Bounds on yield loci.
curve CAB is a convex piecewise smooth curve passing through CAB and
having the proper symmetries. Any other curve through these points passing
inside CAB will obviously not be convex. CAB is therefore a lower bound
for the yield loci.
Now draw a horizontal line symmetric about the a 3 axis intersecting the
adjacent axes of symmetry at C' and B'. Then it follows that no piecewise
smooth curve passing through C'AB' can lie outside C'AB', since then it
would not be convex. Thus C'AB' represents an upper bound to the yield
locus. The rest of the upper and lower bounds can be constructed from
symmetry and are shown as the two regular hexagons in Figure 6.3.7. Thus
It has been shown that all conceivable yield loci satisfying the conditions of
, equal yield in tension and compression, independence of hydro-
stress, and convexity must lie between two regular hexagons as shown;
convex curves are, of course, admissible. The usual Tresca yield locus is
represen1ted by the inner hexagon, and the von Mises circle circumscribes the
inner hexagon and is circumscribed by the outer one. If the von Mises yield
surface is taken as a reference, then the maximum deviation of any admissible
yield surface is about 15.5 per cent.
6-4 LODE'S STRESS PARAMETER.

EXPERIMENTAL VERIFICATION OF
YIELD CRITERIA
The first investigation of a yield criterion was performed by Tresca in 1864,

in which he measured the loads required to extrude metals through dies of
various shapes. On the basis of these experiments he arrived at the maximum
shear stress criterion, previously discussed. It is evident that according to the
Tresca criterion the intermediate stress has no effect on yielding. The von
Mises criterion, on the other hand, gives equal weight to all three principal
stresses. The simplest and most common type test specimen used to check
these criteria is the thin-walled tube shown in Figure 6.1.1. The first such
experiments were run by Lode [5], who tested tubes of steel, copper, and
nickel under various combinations oflongitudinal tension and internal hydro-
static pressure. Lode devised a very sensitive method of differentiating
between the Tresca and von Mises criteria by determining the effect of the
intermediate principal stress on yielding. According to the Tresca criterion,
if a 1 :?:: a 2 :?:: a 3 the yield criterion is given by
To account for the influence of the intermediate stress in the von Mises
criterion, Lode introduced the parameter p.,, called Lode's stress parameter
(not to be confused with Poisson's ratio, p.,):
fk=
a2 - !(al + a 3)
!h- aa)
p, is thus the ratio of the difference between the intermediate stress and
average of the largest and smallest stresses to half the difference
the largest and smallest stresses, and is therefore a measure of the effect of
intermediate stress. Comparing (6.4.2) to (6.3.15) for the case a 1 :?:: a 2 :?::
we see that
p, = -v'3 tan()
Sec. 6-4] Lode's Stress Parameter. Verification of Yield Criteria 89
Thus if 8 varies from oo to 30°, f.L will vary from 0 to -I. All combinations
of loading from pure shear to simple tension are therefore included in the
range 0 to - I.
By means of (6.4.2), the von Mises criterion can be written
a 1 -as 2
(6.4.4)
ao = (3 + f.L2)1/2
The difference between the Tresca and von Mises criteria is then determined
by how much the right side of (6.4.4) differs from I. For f.L = -1, equation
(6.4.4) agrees with. (6.4.1). This is the case of simple tension, and the two
criteria agree as expected. The maximum difference in the range 0 to -1
obviously occurs when f.L = 0, which is the case of pure shear. The difference
is then 2/v'3, which agrees again with the previous results.
Lode ran a series of tests covering the range of f.L from 1 to -1 (- 30° :::; 0
:::; 30°). A f.L of 1 corresponds to uniaxial compression. The results are shown
in Figure 6.4.1. It is seen that the data favor the von Mises criterion even
1.30
I
• Steel
1.25 - o Copper
+ Nickel 0
1.20
0" 1-0"3 2
..J.----a:o- ="'3 +fJ \
n~1.15
/../ 0~
:f:\ "'~
le Oo
v J
1.10 u 0 0 :
·, ._,
1.00
1.00 v
-1.0 -0.8 -0.6 -0.4 -0.2
t o-;;;3 =\0\
0
fL
0.2 0.4 0.6
""
0.8
~ 1.0
+
FIGURE 6.4.1 Lode's test results.
appreciable deviation occurs. These deviations were partially attri-

to lack of isotropy of the material. Some of the results are also plotted
Figure 6.4.2 in the a 1 a 2 plane, since as is zero for these tests. The inter-
of the von Mises circular cylinder with the a 1 a 2 plane is obviously
ellipse, and the intersection of the Tresca cylinder is a hexagon (no longer
, as shown in Figures 6.2.3 and 6.2.4. Figure 6.4.2 shows that the
plot between the hexagon and the ellipse, although generally closer to
ellipse.
cr1 lcro
FIGURE 6.4.2 Lode's test results.

In 1931 Taylor and Quinney [6) published their classical experiments,
which were intended to settle this question. They used copper and steel
tubings which were very nearly isotropic and tested them very carefully in
combined tension and torsion. They concluded that the deviations from the
von Mises criterion were real and could not be explained on the basis of lack
of experimental accuracy or isotropy. Their results are shown in Figure 6.4.3.
0.4
Txy 03
cro .
0.2
• Copper
0.1 >< Aluminum
o Mild steel
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
crxlcr0
FIGURE 6.4.3 Results of Taylor and Quinney.
The theoretical curves are obtained as follows. Using the coordinate
of Figure 6.1.1, for tension and torsion loads only, the stresses are ax
Txy, all the others being zero. The Tresca criterion then becomes
2 J4a~ + 2
Txy = ao
or
Sec. 6-4] Lode's Stress Parameter. Verification of Yield Criteria 91
The von Mises criterion, on the other hand, becomes
(6.4.6)
Both (6.4.5) and (6.4.6) plot as ellipses in the axrxy plane, as shown in Figure
6.4.3. It is seen in the figure that the data fit the von Mises criterion consider-
ably better than the Tresca criterion, although appreciable deviations from
the von Mises criterion do sometimes occur. Figure 6.4.4. shows a replot of
0 1.0
r---~--~---.----,----/~
/
/ 0
/
Tresca-.."'>/
/
/0
/
/
/
/0
/
/
/
/
/
/
/-
FIGURE 6.4.4 Results of Taylor and Quinney.
some of these results in the a 1 a 2 plane, and Figure 6.4.5 shows similar results
obtained by Ros and Eichinger [7]. Other tests of similar nature can be found
in the literature.
0 1.0
r---.----r--~------~
/
/
/
/o
/
/
/
/
/
/
h
FIGURE 6.4.5 Results of Ros and Eichinger.

Attempts have been made to improve the correlation of the data by includ~
ing the effect of the third invariant J 3 into the yield criterion [8]. It seems,
however, that from an engineering viewpoint the accuracy of the von Mises
criterion for yielding is amply sufficient, considering the general scatter and
lack of uniformity in the properties of nominally the same material obtained
from different batches from the manufacturer. Therefore, the search for more
accurate theories, particularly since they are bound to be more complex,
seems to be a rather thankless task. In what follows, therefore, we shall
generally use the von Mises yield criterion and occasionally the Tresca
criterion.
6-5 SUBSEQUENT YIELD SURFACES. LOADING

AND UNLOADING
So far we have discussed the initial yield surface at which a material will
first start yielding. For a perfectly plastic material, this yield surface remains
fixed, as is seen in the uniaxial tensile test, where the stress after yielding re-
mains constant at the yield stress [Figure 2.6.1(d)). However, for a material
that strain hardens the yield surface must change for continued straining
beyond the initial yield. We know that for the uniaxial case, if a material is
strained beyond the yield point to some point such as B' in Figure 2.1.2, the
load removed so that the stress state moves to C', and then the load is in~
creased again, yielding will not take place until the point B' is reached again.
Thus the yield point has been raised in the work-hardened material. In
same way the yield surface in the case of multiaxial stress must "move or·
in some way, at least at the point where yielding initially took place.
In equation (6.3.1) we defined a yield function by the relation
such that whenever the function F became equal to the constant K, yielding
would begin. K then represented an initial yield surface in the stress space.
We can now generalize this type of relation to subsequent yield surfaces.
After yielding has occurred, K takes on a new value (or values), Cle}Jendmg
on the strain-hardening properties of the material. If the material is uu•.v"'""'"
and then loaded again, additional yielding will not occur until the new
of K is reached. The function F can then be looked upon as a loading ru11rn~nn;
or loading surface, which represents the load being applied, and the
K is a yield function, or strain-hardening function, and will depend on
complete previous stress and strain history of the material and its
Sec. 6-5] Subse'quent Yield Surfaces. Loading and Unloading 93
hardening properties. We can now distinguish three cases for a strain-

hardening material:
(1) F = K
oF
dF= ~da11 > 0
ualf
This constitutes loading.
oF
(2) F = K dF= -dalf
oajj
=0 (6.5.2)
This is called neutral loading.
=
oF
(3) F K dF= ~dalf < 0
ualf
This constitutes unloading,

Geometrically the conditions (6.5.2) are readily visualized. F = K means
the stress state is on the yield surface. dF > 0 means the stress state is "moving
out" from the yield surface and plastic flow is occurring. dF < 0 means the
stress state is "moving in" from the yield surface and unloading is therefore
taking place. dF = 0 corresponds to the case of the stress state moving on
the yield surface and is called neutral loading. For a strain-hardening material
no plastic flow occurs. If F < K, the stress state is an elastic one. For per-
fectly plastic materials plastic flow occurs for
F=K dF= 0 (6.5.3)
The case dF > 0 does not exist.

The above discussion is illustrated in Figure 6.5.1. Point P represents the
FIGURE 6.5.1 Stress increment vector for loading,

existing state of stress lying on the yield surface. If the stress state changes to
Q, so that the vector da points" outward" from the cylinder, then loading is
taking place. If Q lies on the surface, we have neutral loading, and if Q is
inside the surface, we have unloading. As loading continues, the point
representing the new existing state of stress will again lie on the surface,
which will move and/or change its shape correspondingly. A similar picture
can be drawn in the 7T plane.
Assume the material obeys the von Mises criterion, so that the initial yield
locus is a circle of radius vi a 0 in the 7T plane. Suppose that straining takes
place to some point a~ > a 0 and the material is then unloaded. If we now
assume that the material remains isotropic, just as it was originally, then the
new yield locus is a circle of radius vi a~, which is larger than, but concentric
with, the original yield circle. We have thus assumed that the material strain
hardens isotropically, as shown for the uniaxial case by the curve ABCFG of
Figure 2.3.1. For isotropic hardening, therefore, the yield cylinder will expand
with stress and strain history but will retain the same shape as initially. For
a Tresca material, the subsequent yield loci will be a series of concentric
regular hexagons. This is shown in Figure 6.5.2.
FIGURE 6.5.2 Subsequent yield loci.
The assumption of isotropic hardening is the simplest one, mathematically,

to use. However, it does not take into account a Bauschinger effect. The
Bauschinger effect would tend to reduce the size of the locus on one side as
that on the other side is increased. The yield surface would thus change shape
as the yielding progresses. Experiments to verify this effect are described in
reference [9]. Several tests were carried out with aluminum alloy tubes
various ratios or torsion and tension to obtain an initial yield surface.
unloading and loading again in a specified manner, subsequent yield
were obtained. The results plotted in the ra plane are shown in Figure 6.5
Without going into a detailed discussion of this data, it is evident that
initial von Mises yield ellipse does not just grow symmetrically but that
definite Bauschinger effect exists.
To account for the Bauschinger effect, Prager (reference [14] of Chapter
Sec. 6-5] Subsequent Yield Surfaces. Loading and Unloading 95
second -yield locus
·;;;
Q,
-10
-20
---
-25~~~--~~--~~--~~~
FIGURE 6.5.3 Initial and subsequent yield loci.
introduced the kinematic model, discussed for the one-dimensional case in

Section 2.6 and illustrated for that case in Figures 2.6.2 and 2.6.3. In this
model the total elastic range is maintained constant by translating the initial
yield surface without deforming it. The model is represented by a rigid frame
having the shape of the yield surface, as shown for the Tresca criterion in
Figure 6.5.4. The state of stress before yield occurs is represented by the
position of a pin free to move within the frame. As the pin contacts the side
of the frame, yielding occurs. The frame is assumed to be constrained against
rotation and to be perfectly smooth, so that only forces normal to the frame
can be transmitted to it. As the pin pushes against the frame it causes it to
translate in a direction normal to the surface at the point of contact. At
corners, if the motion of the pin engages both sides, the frame translates in
the direction of the motion of the pin. Depending on the material being
described by the model, i.e., rigid perfectly plastic, rigid strain hardening, or
· elastic perfectly plastic, the state of stress and the state of strain are repre-
in the model in different ways. For example, for a rigid strain-hardening
........,...u, the displacement of the center of the frame relative to the origin is
·Pr<>pc1rtionlll to the total strain, and the state of stress is represented by the
\'PO!nncm of the pin relative to the origin (compare to Figure 2.6.2).
Figure 6.5.4 (from reference [4]) shows the translating of the Tresca hexagon
means of this mode 1 for a stress path OP. This type of hardening is called
itn~!mtltic hardening because ofthe type of model used to represent it. It takes
FIGURE 6.5.4 Kinematic model for Tresca criterion.
into account a Bauschinger effect but, because it maintains the total elastic
range constant, it probably overcorrects somewhat for the Bauschinger effect,
as discussed in Section 2.3. Attempts have therefore been made to improve
on the kinematic hardening model. This type of model, however, is much
more difficult to handle mathematically, and therefore the isotropic hardening
assumption is still generally used. For small plastic strains it probably gives
answers that are sufficiently accurate.
Problems
1. Compare the values of the pure shear yield strengths based on the six -·!!ories
of yielding listed (exclude Mohr's theory). Assume p, = 0.3 and a 0 ,c = 1;JSa0.
For this case, r = a1 = - a2.
2. A circular shaft 10 in. in diameter has a tensile yield strength of 90,000 psi
and a compressive yield strength of 120,000 psi. Determine the twisting
moment M 1 required to produce yielding based on:
(a) The maximum stress theory.
(b) The maximum shear theory.
(c) The distortion energy theory.
(d) The internal friction theory.
4. It was shown in Section 6.3 that if the von Mises and Tresca criteria
assumed to agree for the case of uniaxial tension, then they will disagree
the case of pure shear and the von Mises yield circle will circumscribe
Tresca hexagon. Show that if it is assumed that the von Mises and
criteria agree for the case of pure shear, then they will disagree for the
of uniaxial tension and the Tresca hexagon will circumscribe the von
ellipse.
5. Derive the general form of the Tresca yield criteria as given by "'Y.''""'v'l
(6.3.8).
References 97
6. Derive the equation for the Tresca yield criterion in the sector between the
- a 1 and a 3 axes as was done in the text for the sector between the a 2 and
-a1 axes.
8. Show that Lode's stress parameter /h = 1 corresponds to uniaxial compres-
sion, /h = -1 to uniaxial tension, and /h = 0 to pure shear.
9. Derive equations (6.4.5) and (6.4.6).
References
1. J. Marin, Mechanical Behavior of Engineering Materials, Prentice-Hall,

Englewood Cliffs, N.J., 1962, p. 117.
1a. Ibid., p. 122.
2. J. Marin, Theories of Strength for Combined Stresses and Nonisotropic
Materials, J. Aeronautical Sci., 24, No. 4, 1956, pp. 265-269.
3. H. M. Westergaard, On the Resistance of Ductile Materials to Combined
Stresses, J. Franklin Inst., 189, 1920, pp. 627-640.
4. P. M. Naghdi, Stress-Strain Relations in Plasticity and Thermoplasticity,
Inst. Eng. Res., Univ. Calif. (Berkeley), Ser. 131, Issue 9, March 1960, p. 18.
5. W. Lode, Versuche ueber den Einfluss der mittleren Hauptspannung auf das
Fliessen der Metalle Eisen Kupfer und Nickel, Z. Physik., 36, 1926, pp.
913-939.
6. G. I. Taylor and H. Quinney, The Plastic Distortion of Metals, Phil. Trans.
Roy. Soc., London, A230, 1931, pp. 323-362.
7. M. Ros and A. Eichinger, Versuche Zur Klaerung der Frage der Bruchgefahr
III, Metalle, Eidgenoss. Material pruf. und Versuchsanstalt Industriell Bauwerk
und Gewerbe, Diskussionsbericht No. 34, Zurich, 1929, pp. 3-59.
8. F. D. Stockton and D. C. Drucker, Fitting Mathematical Theories of Plasti-
city to Experimental Results, J. Col!. Sci. (Rheology Issue), 5, 1950, pp.
239-250.
9. P. M. Naghdi, F. Essenberg and W. Koff, An Experimental Study of Initial
and Subsequent Yield Surfaces in Plastic:,;., J. Appl. Mech., 25, 1958, pp.
201-209.
General References
1950.
Marin, J., The Mechanical Behavior of Engineering Materials, Prentice-Hall,
Englewood Cliffs, N.J., 1962.
CHAPTER 7
PLASTIC
STRESS-STRAIN
RELATIONS
7-1 DISTINCTION BETWEEN ELASTIC AND

PLASTIC STRESS-STRAIN RELATIONS
In the previous sections the relations between stress and strain in the
elastic range were discussed and also the stress states at which plastic flow
or yielding will begin. There is one more ingredient necessary in constructing
a plasticity theory-the relations between stress and strain when plastic flow
is occurring. These relations are the subject of this chapter.
Whereas the strains are linearly related to the stresses by Hooke's law in
the elastic range, the relation will generally be nonlinear in the plastic range,
as is evident from the uniaxial stress-strain curve. A more complicated
distinction between elastic and plastic stress-strain relations arises from the
fact that whereas in the elastic range the strains are uniquely determined
by the stresses, i.e., for a given set of stresses we can compute the strains
directly using Hooke's law without any regard as to how this stress state
was attained, in the plastic range the strains are in general not uniquely
determined by the stresses but depend on the whole history of loading or
how the stress state was reached. This can readily be illustrated by considering
a thin-walled tube in tension and torsion as in the experiments of Taylor and
Quinney.
Consider the initial yield curve to be as shown in Figure 7.1.1. Let the
specimen be strained in uniaxial tension beyond the initial yield to some
98
Sec. 7-1] Distinction Between Elastic and Plastic Stress-Strain Relations 99
Txy
subsequent yield
E
8 A C a-x
FIGURE 7.1.1 Effect of loading path on plastic strains.
point C, where CDE defines the subsequent yield curve. The plastic strains
will then be
e~ = ep
e~ = ef = -1-eP
e~y = e~z = efx =0
Let the specimen now be unloaded to the point B and let us apply a shear
stress increasing from B to D on the new yield locus. The plastic strains will
still be as given above. Any other path could have been used in arriving at
D from C such as OCFD, as long as we do not move outside the yield locus.
Now suppose that the specimen were first stressed in shear to the pointE on
the new yield locus and then, by any other path inside EDC, such as EGD,
were stressed to the point D. The plastic strains would be
e~y = ?'P
e~ = e~ = ef = e~z = e~z =0
which is obviously completely unrelated to the previous strain state. Thus
even though the same stress state at D exists for both loading paths, and
.!herefore the elastic strain states are the same, the plastic strain states are
different.
Because of the above illustrated dependence of the plastic strains on the
loading path, it becomes necessary, in general, to compute the differentials
or increments of plastic strain throughout the loading history and then
obtain the total strains by integration or summation. However, there is at
least one important class of loading paths for which the plastic strains are
independent of the loading path and depend only on the final state of stress.
These are the so-called radial or proportional loading paths, in which all the
--r
100 Plastic Stress-Strain Relations [Ch. 7
stresses increase in the same ratio. These will be more fully discussed subse-
quently.
I
7-2 PRA.NDTL-REUSS EQUATIONS
The first approach to plastic stress-strain relations was suggested by

Saint-Venant in 1870 [1], who proposed that the principal axes of strain
increment coincided with the principal stress axes. The general three-dimen-
sional equations relating the increments of total strain to the stress deviations
were given by Levy in 1871 [2] and independently by von Mises in 1913 [3].
These are known as the Levy-Mises equations. These equations are
(7.2.1)
or
where S11 is the stress deviator tensor and d'A is a nonnegative constant which
may vary throughout the loading history. In these equations the total strain
increments are assumed to be equal to the plastic strain increments, the
elastic strains being ignored. Thus these equations can only be applied to
problems of large plastic flow and cannot be used in the elastoplastic range.
The generalization of equations (7.2.1) to include both elastic and plastic
components of strain is due to Prandtl [4] and Reuss [5] and are known as
the Prandtl-Reuss equations.
Reuss assumed that the plastic strain increment is at any instant of loading
proportional to the instantaneous stress deviation; i.e.,
(7.2.2)
or defJ = Su d'A
Equations (7.2.1) can then be considered as a special case of (7.2.2) where

the elastic strain components are neglected.
Equations (7.2.2) state that the increments of plastic strain d!:)pend on the
current values of the deviatoric stress state, not on the stress increment
required to reach this state. They also imply that the principal axes of stre~.L
and of plastic strain increme'lt tensors coincide. The equations themselves
merely give a relationship between the ratios of plastic strain increments in
the different directions. To determine the actual magnitudes of the increments
a yield criterion is required, as will shortly be shown.
Sec. 7-2] Prandtl-Reuss Equations 101
If the principal directions are considered, equations (7.2.2) can be written
def _ de~ _ de~ _ d'A (7.2.3)

S1 - S2 - Sa -
or de~= Sa d'A
or (7.2.4)
The numerators of the first three terms of (7.2.4) are the diameters of the
three Mohr's circles for the plastic strain increments and the denominators
are the diameters of Mohr's stress circles, as shown in Figure 7.2.1. ~9'~~t:vns
(7.2.4) therefore imply that the Mohr's circles of stress and plastic strain
jncre111en!_ are similar. Also from the relations for the principal shears,
equations (7.2.4) can be considered as stating that the ratios of the three
principal plastic shear strain increments to the principal shear stresses are
constant at any instant.
Equations (7.2.2) can be written in terms of the actual stresses as
de~ = !d'A [ax - !(ay + a 2 ))
de~ = !d'A [ay - !(a2 + ax)]

de~ = !d'A [a 2 - !(ax + ay)]
(7.2.5)
de~y = d'A Txy
de~z = d'A Tyz
de~x = d'A Tzx
Therefore, if d'A were known, we would have the desired stress-strain rela-
tions. To determine d'A use is made of the yield criterion as follows. By means
of equations (7.2.2),
(de~ - de~) 2 + (de~ - den 2 + (de~ - de~) + 6(de~y)2 + 6(de~2) + 6(dekx) 2

2 2
= (d'A) 2 [(ax - ay) 2 + (ay - a 2) 2 + (az - ax)2 + 6-r;y + 6-r~z + 6-r~x]

(7.2.6)
The bracketed quantity on the right side of equation (7.2.6) is seen to be

proportional to the square of the octahedral shear stress [equation (3.4.11)],
and the left side of equation (7.2.6) is proportional to the square of the incre-
ment of octahedral plastic shear strain defined by [see equation (4.5.10)]
(dyS) 2 =i-[(de~ - de~)2 + (de~ - den 2 + (de~ - de~) 2

+ 6(de~y) 2 + 6(de~z) + 6(de~x) ]
2 2
(7.2.7)
Plastic Stress-Strain Relations [Ch. 7
102
FIGURE 7.2.1 Mohr's circles for stress and plastic strain increments.
The constant dil now becomes
dil =dyE
'~'oct
(7.2.8)
where J 2 is the second invariant of the stress deviator tensor.

It is convenient to define an equivalent or effective stress and an equivalent
or effective plastic strain increment as
3
= v2 '~'oct
(7.2.9)
and
dep =3v2 [(de~ - ds~) 2 + (de~ - de~)

2
+ (de~ - de~)2
2
+ 6(de~y) 2 + 6(de~ 2) 2 + (de~x) ]11 2
= v2 dyE
Sec. 7-2] Prandtl-Reuss Equations 103
For a uniaxial tensile test in the x direction the equivalent stress and equiva-
lent plastic strain increment reduce to
(7.2.11)
The convenience of the above definitions now becomes apparent. The equiva-
lent or effective stress, a., and the equivalent or effective plastic strain incre-
ment, dev, will henceforth be used in this text rather than the octahedral
shear stress and octahedral plastic shear strain increment.
The constant dl.. can therefore be written
(7.2.12)
and the stress-strain relations (7.2.5) become
(7.2.13)
p 3 dep
dByz = -2- Tyz
Ue
p 3 dev
dBzx = 2-- Tzx
Ue
or
p 3 dep S
dBjj = - - IJ
2 Ue
If one compares equation (7.2.9) for the equivalent stress u8 with equation
(6.2.8), which gives the von Mises yield criterion, it is seen that just as yielding
be~ins
(7.2.14)
where u 0 is the yield stress in simple tension. The equivalent stress is thus the
same as the von Mises yield function, and since equations (7.2.13) make use
of this function, the original Prandtl-Reuss assumptions imply the von Mises
yield criterion. This will also be shown subsequently from other considera-
tions.
It also follows from (7.2.14) that for a perfectlyplastic material the Prandtl-:-
R._~u~s eg_uaticmsmay bewritten
:d P
• eli= 23 dep
~ lj
S (7.2.15)
For a material that work hardens, however, a. may be greater than a 0 , and
it is now necessary to find the relation between the equivalent stress ae and
the equivalent plastic strain increment, dev· Before this is done, we introduce
the concepts of plastic work and the measures of work hardening.
7-3 PLASTIC WORl{. TWO MEASURES OF

WORK HARDENING
An important concept that is frequently used in plasticity theory is the

concept of plastic work. The work done per unit volume on an element during
straining is
dW = a11 dell
= (def, + defJ)
ali
= dw• + dWP (7.3.1)
But dW• = a11 def1 is recoverable elastic energy, whereas the plastic deforma-
tion is an irreversible process from which the energy cannot be recovered.
The remainder of the work done is called the plastic work per unit volume,
(7.3.2)
which in turn can be written
(7.3.3)
or, in terms of the principal stresses,
Now let us consider again a plot in the 7T plane where the axes are taken to
be the principal stress deviators S 1 , S 2 , and Sa (for a point lying in the
7T plane, since a 1 + a 2 + a a = 0, S 1 = a 1 , S 2 = a 2 , Sa = a a). Since for plastic
strain increments def + de~ + de~ = 0, the same plot in the 7T plane can
Sec. 7-3] Plastic Work. Two Measures of Work Hardening 105
be used for both the stress deviators and the plastic strain incrementsl pro-
vided the latter are multiplied by a constant to give them the dimensions of
stress. For this constant we choose 2G. The stress deviator vector and the
plastic strain increment vector can then be plotted on the same plot, as
Rdo
FIGURE 7.3.1 Stress and plastic strain increment vectors in rr plane.
shown in Figure 7.3.1. From (7.3.3) the plastic work increment is the scaler
product of the two vectors, or
dw p = OP·RQ = (OP)(RQ) (8- ·'·)

2G 2G cos '~'
But
OP =.VSf + s~ + s~ =Viae
and
Hence the plastic work increment can be written
(7.3.4)
If RQ is parallel to OP, this reduces to
(7.3.5)
In particular, equation (7.3.5) will be valid for the Prandtl-Reuss relations

of Section 7.2. Making use of equation (7.3.5), the Prandtl-Reuss relations
(7.2.13) can now also be written
(7.3.6)
We next consider the question of how the amount of work hardening or

strain hardening that has taken place in a given material due to plastic flow
is measured. For this purpose two work-hardening hypotheses, known as the
two measures of work hardening, have been proposed.
The first hypothesis assumes that the amount of hardening depends only
on-the total plastic work, and is jndependent of the strain path [6]. This is
called the equivalence of plastic work. The implication is that the resistance
to further yielding depends only on the amount of work which has been done
on the material. This amount of work is measured by the yield criterion.
Thus, as in (6.5.1), the yield criterion is written
(7.3.7)
where K keeps changing as the material work hardens, and, assuming

isotropic hardening, F(atJ) remains the same. By the above hypothesis, K is
a function of the plastic work done per unit volume and we can write
\
F( ali) = f( WP) (7.3.8)

!
where
If the equivalent stress ae is taken as the yield function, we can write
(7.3.9)
The functional relationship between the yield function and the plastic work
can be obtained experimentally and then the plastic strain increments can
be calculated using equation (7.3.6).
The second hypothesis uses the equivalent plastic strain previously defined
as a measure of work hardening; i.e., --
(7.3.10)
where dep is given by equations (7.2.10). The yield function is then assumed
to be a function of the equivalent plastic strain. Thus
Sec. 7-3] Plastic Work. Two Measures of Work Hardening 107
Again the functional relationship can be determined by experiment. If, as

before, the equivalent stress is used for the yield function, then
(7.3.12)
and the plastic strain increments can be computed from (7.2.13).

For the case of the von Mises criterion assuming isotropic hardening, these
two hypotheses are obviously equivalent. Since
equation (7.3.9) can be written
(7.3.13)
which is the same as (7.3.12). In general, however, equations (7.3.11) and

(7.3.8) need not be equivalent, because of anisotropy and the Bauschinger
effect. A detailed discussion of the conditions under which the two are
equivalent is given in reference [7]. Equations (7.3.8) and (7.3.11) are some-
times referred to as the work-hardening and strain-hardening hypotheses,
respectively [8].
The formulation (7.3.11) is simpler to use. For the case of the Prandtl-
Reuss equations and the von Mises yield criterion, equation (7.3.12) is used
almost exclusively in conjunction with equations (7.2.13). In actual applica-
tion, the experimental relationship given by equation (7.3.12) is taken from
the uniaxial tensile stress-strain curve, as shown in Figure 7.3.2. The abscissa
<re
e =E +ep
FIGURE 7.3.2 Relation between equivalent stress and equivalent plastic strain.
and ordinate of the uniaxial stress-strain curve are replaced by ev = Jdev

and a., respectively. In terms of the slope of this curve, equations (7.2.13)
can be written
3 da 8 S
dBtJP = 2 (7.3.14)
H'ae iJ
where H' = da8

dep
7-4 STRESS-STRAIN RELATIONS BASED ON

TRESCA CRITERION
In Section 7.2 it was shown that the Prandtl-Reuss relations are associated
with the von Mises yield criterion. The stress-strain relations generally used
with the Tresca criterion are simpler in nature. However, as discussed in
Section 6.2, to use the Tresca yield criterion it is necessary to know which is
the maximum and which is the minimum principal stress. Assuming that a 1
is the maximum principal stress and a 3 is the minimum principal stress, an
equivalent stress can be defined for the Tresca criterion by
(7.4.1)
For the uniaxial tensile test in the 1 direction, aT becomes equal to a 1 • The
Prandtl-Reuss equations can now be written
(7.4.2)
with dev defined as previously. However, it can easily be shown (see Problem
5) that equations (7.4.2) are inconsistent with the definition (7.4.1). In spite
of this inconsistency, equations (7.4.2) are sometimes used to good advantage
[9]. Since in using the Tresca criterion it is assumed that the middle stress has
no effect on yielding, it is reasonable to assume also that there is no plastic
flow in that direction. A consistent set of relations can therefore be obtained
by assuming
d81p = v3dBp
T
ds~ =0
de~= - ~?, dep = -def
with dev = Vf V(def)2 + (de~) 2 + (dsn 2

Sec. 7-5] Experimental Verification of Prandtl-Reuss Equations 109
Equations (7.2.13) and (7.4.3) are known as the flow rules associated with
the von Mises and Tresca criteria, respectively. In Section 7.6 we will show
that these are special cases of a more general flow rule. In that section the
results of the previous sections will be essentially rederived based on a more
unified approach.
7-5 EXPERIMENTAL VERIFICATION OF

PRANDTL-REUSS EQUATIONS
The first experimental investigation to determine the validity of the plastic

stress-strain relations was made by Lode (reference [5] of Chapter 6). As
explained in Chapter 6, Lode tested tubes of steel, copper, and nickel under
combined tension and internal pressure. In addition to the stress parameter
p, defined by (6.4.2), Lode introduced the plastic strain parameter v defined by
2de~ - de~ - def de~ - -!(de~ + def) (7.5.1)

V = dsap - de 1p = 2 (dsap - de1P)
1
From equations (7.2.3),
2de~ - de~ - def

de~- def
or v=p, (7.5.2)
If the Prandtl-Reuss relations are valid, then equation (7.5.2) should be

satisfied over the whole range of experiments. The results of Lode's tests are
shown schematically in Figure 7.5.1. Although the relation appears to be
approximately satisfied, there is a definite deviation which cannot be ac-
counted for by experimental scatter. This deviation was confirmed by Taylor
and Quinney (reference [6] of Chapter 6), whose results on tension-torsion
tests are shown roughly in Figure 7.5.2. Although some of these deviations
can be attributed to lack of isotropy, it appears that the Prandtl-Reuss
relations are not quite correct. Prager [10] has shown that the data can be
made to fit quite well by including la in the yield function. However, the
relations are too cumbersome for general use.
From a practical viewpoint the Prandtl-Reuss relations appear to be
sufficiently accurate, just as the von Mises yield criterion is sufficiently
accurate, and the small deviations of the experimental data are not large
enough to make additional complexity worthwhile. These relations will
therefore henceforth be used most frequently, the Tresca criterion and its
-1 0
JL
FIGURE 7.5.1 Results of Lode's tests.
-1.0
FIGURE 7.5.2 Results of Taylor and Quinney's tests.
associated flow rule being used occasionally. It is to be noted that the assumed
coincidence of the principal stress axes and principal plastic strain increment
axes are borne out very well by the experiments of Taylor and Quinney.
7-6 GENERAL DERIVATION OF PLASTIC

STRESS-STRAIN RELATIONS
In Section 7.5 the Saint-Venant-Levy-Mises and the Prandtl-Reuss rela-

tions were described as originating basically from an assumption that "the
Sec. 7-6] General Derivation of Plastic Stress-Strain Relations 111
maximum shear and maximum slide velocity are co-directional," as Saint-

Venant expressed it (see reference [11]). It was also shown that these relations
imply the von Mises yield function.E_l_!hi~~ection the general equations for
d~!t;:rmining the plastic stress-strain relations for any yield criterion will be
deriv~d based on fl_unified approach due to Drucker [12, 13].
We start with a more precise definition of work hardening which is due to
Drucker, and it will subsequently be shown that together with two additional
assumptions it is sufficient to obtain the most general form of the stress-
strain relations.
Suppose we have a given state of stress and then some external agency
applies an additional set of stresses and then slowly removes them. Work
hardening implies that for all such added sets of stresses the material will
remain in equilibrium, and
1. Positive work is done by the external agency during the application of
the set of stresses.
, . 2. The net work performed by it over the cycle of application and removal
k, is zero or positive.
It should be emphasized that the work referred to is not the total work
done by all the forces acting; it is only the work done by the added set of
forces on the displacements which result. In other words, work hardening
means that useful net energy over and above the elastic energy cannot be
extracted from the material and the system of forces acting on it.
This definition can be put into mathematical language as follows: Suppose
that to a state of stress ali and strain elf some external agency applies small
surface forces so that the stress at each point is changed by an amount da1i
and the strain by an amount de 1i. Part of de 11 is elastic and part may be plastic;
i.e., de 1i = defi + defJ. Now suppose these added forces are removed, thus
releasing the elastic strain increments, defi· It then follows from implication 1
that for work hardening
and, from implication 2,
or
dati( defi + defJ) > 0} (7.6.1)
dalf defJ:?: 0
Equations (7.6.1) represent the mathematical definition of work hardening.

The second of equations (7.6.1) is sometimes referred to as the uniqueness
condition.
To obtain the general stress-strain relations, we use the above definition

plus two basic assumptions. These are:
1. A loading function exists. At each stage of the plastic deformation there
exists a functionf(a!i) so that further plastic deformation takes place only for
/(aiJ) > K. Both/ and K may depend on the existing state of stress and on the
strain history.
2. The relation between infinitesimals of stress and plastic strain is linear;
i.e.,
1d p--
i eli clfkl dakz (7.6.2)
Although equation (7.6.2) seems very reasonable, it should be noted that

there appears to be no theoretical justification for it. It is purely an assump-
tion. Although the C1 i~cz may be functions of stress, strain, and history of
loading, (7.6.2) implies that they are independent of the da~cz·
From assumption 1 it follows that for plastic deformation to take place
or (7.6.3)
and from the linearity assumption 2 it follows that the superposition principle
may be applied to the stress and strain increments. Thus if dai1 and da71 are
two increments producing plastic strain increments, defJ' and de[j', then an
increment dalf = dai1 + da71 will produce an increment, de[j + de[j'.
Now assume that for a given state of stress a~c 1 , an increment of stress da~cz
producing plastic flow is imposed. This increment da~c 1 can be decomposed
into two parts da~ 1 and daZ 1 such that da~ 1 produces no plastic flow and daZ 1
is proportional to the gradient of f(ali). Geometrically this means that the
vector da~c 1 is decomposed into a component tangent to f and a component
perpendicular to j, as shown in Figure 7.6.1.
FIGURE 7.6.1 Decomposition of stress increment vector.

Since the increment dak 1 produces plastic flow, from (7.6.3) we have
-
8f
8 akl
J
uakl = -88f (-'~uaklI + uakl
J II)
> 0 (7.6.4)
akl
But the increment da~ 1 produces no plastic flow. Therefore,
(7.6.5)
Also, da~ 1 has been taken proportional to the gradient off; therefore,
(7.6.6)
where a is a scaler > 0 and from (7.6.4), (7.6.5), and (7.6.6) it follows that
Hence
(7.6.7)
Equation (7.6.7) proves that the proposed decomposition is possible.

Comparing (7.6.2) and (7.6.7) [realizing that dak 1 in (7.6.2) can be replaced
by da~ 1 since da~ 1 produces no plastic flow] we see that every component of
defJ must be proportional to a, or
.;)
(7.6.8)
or, combining with (7.6.7) gives
(7.6.9)
where g11 depends in general on stress, strain, and history of loading.

Now the second of conditions (7.6.1) can be written
(7.6.10)
But da;1 produces no plastic flow, so that the increment dau = Cdaj1 + da;'1
for any value of C, positive or negative, will produce the same plastic incre-
ment defJ. We can therefore write the strain-hardening condition as
(7.6.11)
But daj1 defJ must vanish;, otherwise C could be chosen (a large negative
number) so as to violate (7.6.11). Therefore,
But
Hence
Comparing with (7.6.5) it is seen that
(7.6.12)
where G is a scalar which may depend on stress, strain, and history. Substi-
tuting (7.6.12) into (7.6.9) gives
(7.6.13)
or (7.6.14)
which is the general stress-strain relation consistent with the original assump-
tions.
Let us take some specific examples. For the von Mises yield function, let
f = J2 = ![(al - a2) 2 + (a2 - as) 2 + (as - a1) 2]
!f = ~{al -
ua 1
!(a2 + as)]
Therefore,
where ld'A = G df,\and we recognize the Prandtl-Reuss equations. For the

Tresca yield condition, assuming it is known which is the maximum principal
stress a 1 and minimum princip[tl stress a 3 , we have
f= ·Ha1- aa)
of= 0 of
oa2 oa3
Then
def =-! d'A
de~= 0
de~= --!-d'A
which are the same as equations (7.4.3). We note that the form of the flow
rule or plastic stress-strain relations associated with the Tresca criterion is
entirely different than that for the von Mises; thus each yield condition has
an associated flow rule, as was pointed out in Section 7.4. This is sometimes
ignored and, for example, the Tresca criterion has been used with the
von Mises flow rule. There is, however, no theoretical justification for that
type of assumption .
.\._\· . It is worth noting one ot.her importan.t fact from the previous derivation .
. Since da;, defJ = 0 and da;1 is tangent to the yield surface, it follows that
I "
l! defJ is normal to the yield surface, for the above equation merely represents a
h ..
dot product of two vectors. This can also be seen from (7.6.14), since defJ is
equal to the gradient of jtimes a scalar. It also follows from the above that !i
the Prandtl-Reuss equations imply the use of the von Mises criterion. .i
To summarize: Starting with the definition of work hardening and postu- '
lating the existence of a loading function and linearity between increments
of stress and increments of strain, we can arrive at the general flow rule
(7.6.14) for a strain-hardening material. It can also be shown that the plastic
strain increment vector must always be normal to the yield surface. The
scalar G, which depends in general on the stress, strain, and history, must be
determined from experiment, and its derivation will be discussed shortly.
Perfectly Plastic Material
For this case the work done by an external agency which slowly applies
and removes a set of stresses is zero over the cycle, or
(7.6.15)
It should be remarked that this equation is not the same as the second of
(7.6.1) with the equality sign. In (7.6.1) the equality sign is used only when
defJ = 0.
For ideal plasticity it is also assumed thatf(a11) exists and is a function of
stress only, and that plastic flow takes place without limit when f(a 11) = K
and the material behaves elastically when f(a 11) < K. For plastic flow,
therefore,
8/
df= -dali =0 (7.6.16)
8a 11
Comparing (7.6.15) and (7.6.16) it is seen that
8
deft = dA / (7.6.17)
8a11
where dA. is a scalar.
Determination of the Function G. Effective Stress and Effective Strain
For (7.6.14) to be of any practical use, it must be related somehow to the

experimental uniaxial stress-strain curve. What we are looking for is some
function of the stresses, which might be called the effective stress, and some
function of the strains or strain history, which might be called effective strain,
so that results obtained by different loading programs can all be correlated
by means of a single curve of effective stress versus effective strain. This curve
should preferably be the uniaxial tensile curve.
The definition of effective stress can be arrived at rather simply; since it
should reduce to the stress in the uniaxial tension test, it is a quantity
which will determine whether plastic flow takes place or not, and it must be a
positively increasing function of the stresses during plastic flow. Now the
loading function f( a11) also, by definition, determines whether additional
plastic flow takes place. It is also a positively increasing function as long as
plastic flow takes place and, if unloading takes place, plastic flow is not
resumed until the highest previous value of/is exceeded. The loading function
f(a 1) must therefore be some constant times the effective stress to some
power; i.e.,
(7.6.18)
For example, if we assume again
then
and for the uniaxial tensile test a. = a 1 • Therefore,
n=2 c= 1/3
which agrees with the previous definition in equation (7.2.9).

The definition of effective plastic strain, ep, is not quite as simple. There are
two methods generally used. One defines the effective strain increment in
terms of the plastic work per unit volume; i.e.,
(7.6.19)
and since
(7.6.20)
For example, iff= J 2 , it can readily be shown that
(7.6.21)
and, iff= a 1 - a3 with a 1 > a 2 > a3 as for the Tresca criterion, then
(7.6.22)
Equation (7.6.21) expanded becomes
dep = 'V1 [(de~) 2

+ (de~) 2 + (def) 2 + 2(de~y) 2 + 2(de~2) 2 + 2(de~x) 2]11 2
(7.6.23)
and, in terms of principal strain increments,
deP = 'V1 [(def) 2 + (de~)2 + (de~)2]112

2 .
= v'3 [(def)2 + (de~) + def de~]1
2 12
where the incompressibility condition def + de~ + de~ = 0 has been used.
A second method for arriving at (7.6.21) is sort of intuitive. One seeks to
find a definition of effective plastic strain increment which when integrated
is a function of a. only. The simplest combination of plastic strain increments

which is positive increasing and has the correct "dimension" is
To make this definition agree for simple tension we must have
Therefore,
C= vt
dep = VidefJ defJ
and, for f = 12,
(7.6.24)
so that the integrated effective strain is a function of effective stress only; i.e.,
(7.6.25)
It should be noted here that the definition (7.6.21) for dev has been derived
for f = 12 only. Drucker has shown that it is reasonably correct for almost
any /(12 , 13). The second intuitive approach for defining dev is, of course, not
based on any specific loading function.
We are now in a position to determine the function G. It should first be
realized that for the previous formulation to agree with the uniaxial tensile
curve, da./dev must be the slope of that curve (in the plastic range). Substi-
tuting the basic equation
into (7.6.21) gives
(7.6.26)
or (7.6.27)
and the general plastic stress-strain relation becomes

Sec. 7-7] Incremental and Deformation Theories 119
or (7.6.29)
where a~ = da./dsv is the slope of the uniaxial stress-plastic strain curve at

the current value of a •. As an example, for f = J 2 , equation (7.6.29) gives
(7.6.30)
\Equations (7.6.30) constitute the flow rule (or plastic stress-strain relations)
1associated with the von Mises yield criterion. They are the well-known
Prandtl-Reuss relations we obtained previously. If we replace the plastic
strain increments in the above equations by total strain increments, the
Levy-Mises relations are obtained which are valid only if the plastic strains
are so large that the elastic strains can be neglected.
As a final note, a general flow law such as (7.6.14) can also be obtained on
the basis of a hypothesis that there exists a plastic potential (similar to the
strain energy density function) which is a scalar function of stress, g(a 11), from
which the plastic strain increments can be obtained by partial differentiation
with respect to the stresses. Thus
dsft = ~g d(3 (7.6.31)

uaij
where df3 is a nonnegative constant. The plastic potential g(ali) was first
introduced by Melan [14]. By comparison with (7.6.14), it would appear that
the plastic potential should play the same role as the yield function, and
indeed Bland [15] has proved that they must be the same function, so that gin
(7.6.31) can be replaced by f; (7.6.31) and (7.6.14) are then the same.
7-7 INCREMENTAL AND DEFORMATION

THEORIES
Equations such as (7.6.30) are called incremental stress-strain relations
because they relate the increments of plastic strain to the stress. To obtain
the total plastic strain components, one must integrate these equations over
the whole history ofloading. Hencky [16] proposed total stress-strain relations
whereby the total strain components are related to the current stress. Thus,
instead of (7.6.30), one would have
(7.7.1)
The plastic strains then are functions of the current state of stress and are
independent of the history of loading. Such theories are called total or
clf![ormation theories in contrast to the incremental or flow theories previously
described. This type of assumption greatly simplifies the problem; however,
as was previously shown, the plastic strains cannot in general be independent
of the loading path and deformation theories cannot generally be correct.
There has often been a tendency therefore to ignore all deformation theory
as of little value.
It can easily be shown, however, that for the case of proportional or radial
loading, i.e., if all the stresses are increasing in ratio, the incremental theory
reduces to the deformation theory. For if a11 = Ka~, where aZ· is an arbitrary
reference state of stress (nonzero) and K is a monotonically increasing func-
tion of time, then S11 = KS?1 and a. = Ka~ and (7.6.30) becomes
3dep
dBt·P1 = - so
- i}
2a~
which can be immediately integrated to give
(7.7.2)
so the plastic strain is a function only of the current state of stress and is
independent of the loading path.
Furthermore, it has been proposed by Budiansky [17] that there are ranges
of loading paths other than proportional loading for which the basic postu-
lates of plasticity theory are satisfied by deformation theories. Budiansky's
theory postulates the occurrence of corners or singular points on the succes-
sive yield surfaces and, although the existence of such singular points has as
yet not been established experimentally, one cannot rule out the possibility
of loading paths other than proportional loading for which total plasticity
theories may give satisfactory answers.
From a practical viewpoint, there are a great many engineering problems
where the loading path is not far from proportional loading, provided one is
careful when unloading occurs to separate the problem into separate parts, the
loading parts, and the unloading parts.
Sec. 7-8] Convexity of Yield Surface. Singular Points 121
Problems of plastic flow in thermally stressed disks and cylinders have been
handled in this way and good results obtained using deformation theory.
On the other hand, it will subsequently be shown that with the present
widespread availability of high-speed computors, many simplifying assump-
tions heretofore made, including the use of deformation theories under
doubtful conditions, are often unnecessary.
7-8 CONVEXITY OF YIELD SURFACE.

SINGULAR POINTS
In Section 6.3 the statement was made that the yield surface was convex.
A proof, as given in reference [13], will now be presented. Consider some state
of stress a~ inside the loading surface, as shown in Figure 7.8.1. Let some
FIGURE 7.8.1 Stress path produced by external agency.
external agency add stresses along some arbitrary path inside the surface
until a state of stress au is reached which is on the yield surface. Only elastic
changes have taken place so far. Now suppose the external agency to add a
very small outward pointing stress increment dati which produces small
plastic strain increments defJ, as well as elastic increments. The external
agency then releases the dati and the state of stress is returned to a~ along an
elastic path. The work done by the external agency over the cycle is
(7.8.1)
If the plastic strain coordinates are superimposed on the stress coordinates,

as in Figure 7.8.2, oW may be interpreted as the scalar product of the vector
122 Plastic Stress-Strain Relations (Ch. 7
FIGURE 7.8.2 Stress and plastic strain increment vectors.
a11 - at and the vector defJ plus the scalar product of dali and de{j. Now,
from the strain-hardening definition equation (7.6.1),
or ldatJlldefJI COS 8~ 0 (7.8.2)

or
That is, the vectors dali and deft make an acute angle with each other. In a
similar fashion, since the magnitude of a11 - at can always be made larger
than the magnitude of da11 , it follows that
(ali - a~)defJ ~ 0
or laii - a~lldefJI COS rp ~ 0
Hence
1T 1T
-- < .1. <-
2-'f'-2 (7.8.3)
Thus the vector ali - at makes an acute angle with the vector deft for all
choices of at. Therefore, all points a~ must lie on one side of a plane per-
pendicular to deft, and, since deft is normal to the yield surface, this plane will
be tangent to the yield surface. This must be true for all points ali on the yield
surface, so that no vector a11 - at can pass outside the surface intersecting
the surface twice, as shown in Figure 7.8.3. The surface must therefore be.
FIGURE 7.8.3 Convex surface, only acute angles possible.

Sec. 7-9] Plastic Strain-Total Strain Plasticity Relations 123
convex. On the other hand, if the surface is not convex, there exist some
a
po1n:ts 11 and at a at
such that the vector 11 - forms an obtuse angle with
the vector da1J> as shown in Figure 7.8.4. This completes the convexity proof.
FIGURE 7.8.4 Surface not convex, obtuse angles possible.
Equation (7.6.14) implies that the yield surface has a unique gradient. It
may happen, however, that the yield surface has vertices or corners where the
gradient is not defined. For example, the Tresca hexagon has no unique
normal at the corners, where two of the stresses are equal. Such points are
called singular points or singular yield conditions. Such points can be treated
by introducing an auxiliary parameter as described in reference [18].
7-9 PLASTIC STRAIN-TOTAL STRAIN

PLASTICITY RELATIONS
The Prandtl-Reuss equations relate the plastic strain increments to the

stresses. We shall now derive a similar set of equations involving only strains.
These equations enable one to compute the plastic strain increments from
the total strains without recourse to the stresses. In effect, they provide a
simple method for separating the total strains into their elastic and plastic
components. The advantage of this formulation will become evident later
when certain iterative methods for solving plasticity problems are discussed.
Assume some loading path to a given state of stress and total plastic strains
er,. Let the load be increased by a small amount, producing additional plastic
strains Llsf,. The total strains can now be written
(7.9.1)
where ef1 is the elastic component of the total strain, ef, is the accumulated
plastic strain up to (but not including) the current increment of load, and
..........------------------
124
Plastic Stress-Strain Relations {Ch. 7
1
~sfJ is the increment of plastic strain due to the increment of load. sfJ is
presumed to be known, ~sfJ is to be computed. Define modified total strains
as follows:
(7.9.2)
Then
(7.9.3)
Subtracting the mean strain from the diagonal components of both sides of
equation (7.9.3) results in
(7.9.4)
where er, is the elastic strain deviator tensor and e;, is the modified strain
deviator tensor. From Hooke's law and the Prandtl-Reuss relations,
1 S 1
e,,e = 2G 11 = 2G ~A us,,
A P
Hence
(7.9.5)
(7.9.6)
We now define an equivalent modified total strain by
so that, from (7.9.6),
and, from (7.9.5),

Sec. 7-9] Plastic Strain-Total Strain Plasticity Relations 125
or, in expanded form,
A p D.Bp (
LJ.Bx = 3- 2BxI -
I
By -
I)
Bz
Bet
A p D.Bp (
LJ.By = - 2ByI -
I
Bz -
1)
Bx
3Bet
A p D.Bp (
U.Bz = 3- 2BzI -
I
Bx -
I)
By
Bet
(7.9.10)
with Bet given by (7.9.7) or alternatively by
Bet
V2 [(BxI
= T -
1)2
By + ( I
By -
1)2
Bz + ( I
Bz -
I
Bx
)2 + 6( I
Bxy
)2 + 6(I
Byz
)2 + 6( I
Bzx
)2]1/2
(7.9.11)
and the primed quantities are the modified total strains as given by equation
(7.9.2).
Equations (7 .9 .1 0) are equivalent to the Prandtl-Reuss equations. The
stresses do not appear in these equations and the increments of plastic strain
can be computed from the total strains. Note that since they have been
derived by use of the Prandtl-Reuss equations, they implicitly make use
of the von Mises yield function. It should also be emphasized that the equiva-
lent total strain defined by (7.9.7) is a purely mathematically defined quantity
without any direct physical meaning, even in the uniaxial case. However, it
can be related to the uniaxial stress-strain curve as follows: From equation
(7.2.12),
(7.9.12)
Substituting this value for L\,\ into equation (7.9.8) gives
1 O'e ) Set
( + 3G L\sp = L\sp
or
(7.9.13)
Referring to the uniaxial stress-strain curve as shown in Figure 7.9.1, let
FIGURE 7.9.1 Relation between Set. ae, and D.sp.
L\aebe the increment in stress to which corresponds a plastic strain increment

L\sPand let ae be the stress at the end of the increment. Then Bet is the sum of
the plastic strain increment plus the total elastic strain multiplied by
2/3(1 + fl-). Solving equation (7.9.13) for L\sP results in
(7.9.14)
a. can now readily be eliminated from equations (7.9.13) or (7.9.14) as

follows. Let the stress preceding the increment of load be ae,i-1• i.e.,
ae = ae,i- 1 + L\ae. Then expanding a. in a Taylor series about ae,i- 1 gives
approximately
ae = Ue,i-l + (ddae).
Sp 1-1
L\ep + ...
Sec. 7-10] Complete Stress-Strain Relations. Summary 127
where higher-order terms in LlBp have been neglected. Substituting into

(7.9.14) and solving for LlBp gives
LlB _ Bet - %[(1 + p,)/E] O'e,l-1

(7.9.16)
P- 1 + j[(1 + p,)/E] (dae/dBp)t-1
For linear strain hardening, equations (7.9.15) and (7.9.16) are obviously1
exact. Equation (7.9.16) shows how Bet is related to LlBp through the geometry!
of the uniaxial stress-strain curve. We shall use this relationship subsequently
together with equations (7.9.10) and (7.9.11) to solve specific problems. For
want of a better name we will refer to equations (7.9.10) as the plastic strain-
total strain equations.
If one desires to use the total or deformation theory of plasticity, it can be
shown [19] that it is only necessary to remove the primes and increment
symbols from equations (7.9.9) through (7.9.11); i.e.,
where
Bet = V je11eu
(7.9.17)
7-10 COMPLETE STRESS-STRAIN RELATIONS.

SUMMARY
In the previous sections the relations between the increments of plastic

strain and the stresses at any instant were discussed in some detail. The
fundamental problem in applying plasticity theory is to determine the total
plastic strain as a function of the history of loading or history of stress.
Suppose a body is loaded along some specified load path to some final load
condition. To calculate the plastic strains at this final load condition it is
theoretically necessary, in general, to integrate the infinitesimal plastic strain
increments over the actual loading path. Although this can be done in rela-
tively simple cases, it is usually more expeditious to assume the load applied
in small finite increments and calculate the finite increments of plastic strain
................... ------------~,
for each of the load increments. All these increments of plastic strain are then
I
added to give the total plastic strain. The integration is thus replaced by a
summation.
Let the total loading path be divided into N increments of load. Assume
that the plastic strains have been computed for the first i - 1 increments of
load and we now wish to compute them for the ith increment of load. The
total strains at the end of the ith increment can be written with thermal
strains included, as
(7.10.1)
where 0 = a 11 = ax + ay + a 2 • The first two terms on the right side of

equation (7.10.1) represent the elastic part of the total strain, the third term
is the thermal strain, the fourth term is the plastic strain accumulated in the
first i - 1 increments of load, and the fifth term is the plastic strain due to the
ith increment of load. In expanded form these equations are
(7.10.2)
In the above equations the sums are known and the problem is to calculate
the plastic strain increments for the current or ith increment of load, and the
corresponding stresses. To do this it is necessary to use one or another of the
plastic stress-strain relations discussed in previous sections. A yield criterion
must be chosen and the associated flow rule as given by equation (7.6.14).
In particular, since we shall concern ourselves only with the von Mises and
Tresca yield criteria, the following relations previously derived will be used.
For the von Mises criterion:
ae = J
v'3:J; = 2 Tact
= V!St}Stj
= J2 [(axay) 2 + (ay - a 2) 2 + (azax)2 + 6(T;y + T~z + T;x)]1 12
(7.10.3)
= vi[(Lle~) 2 + (Lle~)2 + (LleD 2 + 2(Lle~y) 2 + 2(Lle~z? + 2(Llefx) 2] 112
= ~) [(Lle~) 2 + (Lle~) 2 + (Lle~y) 2 + (Lle~2)2 + (,:lefx) 2 + Lle~ Lle~]ll 2
= ~2, [(Lle~ - Llef) 2 + (Lle~ - LleD 2 + (,:lef - LleD 2
(7.10.4)
or
(7.10.5)
_11-nd Llep is related to a0 through the uniaxial tensile stress-strain curve as

shown in Figures 7.3.2 or 7.9.1.
130
Alternatively we define
Plastic Stress-Strain Relations [Ch. 7 l
1-1
e;, = el, - 2:
k=1
ilB&,k = B;, - s,,Bm
1-1
BtJ = B11 - L
k=l
ilBfJ,k
Bm = j-B11 = j(Bx + By + B2)
= v2 [(
3 I
Bx - ByI )2 + ( ByI - Bz1)2 + ( BzI - I )2
Bx
(7.10.6)
Then
or
A p
UBx = -!1Bp
3
(
2 BxI -
I
By -
I)
Bz
Bet
A p
UBy = -
!1Bp (2 ByI -
I
Bx -
I)
Bz .
3Bet
A P
UBz = -!1Bp
3
(
2BzI -
I
Bx -
I)
By
Bet
(7.10.7)
and ilBv, Bet• and a8 are related to each other by
Furthermore, for small increments

where a 8 , 1 _ 1 is the value of the equivalent stress at the end of the (i - 1)st
increment of load and (da./dep) 1 _ 1 is the slope of the uniaxial tensile curve
replotted as true stress versus true plastic strain. _Equation (7.10.9) is exact
for)inear_strain hardening. The above relations are shown graphically in
Figure 7.9.1.
If the deformation or total theory of plasticity is used, all the above
relations are valid if the Ll's are removed from all the previous equations
and the primes are removed from equations (7.10.6) through (7.10.7).
Equation (7.10.8) becomes
(7.10.10)
and by the use of (7.10.10), the uniaxial stress-strain curve can be replotted
as a curve of eP versus e 81 as shown in Figure 7.10.1. This curve can then be
used instead of the original stress-strain curve.
24 ~10-3 /
....
~ 20
c
·et; 16
-+-
c
/
~ 12
>
·:; /
cr /
Q) 8
'E /
~
4
v
0 4 8 12 16 20 24X10- 3
Equivalent plastic strain, ep
FIGURE 7.10.1 Equivalent total strain-equivalent plastic strain curve.
For the Tresca criterion,assume a1 > a2 > a 3• Then
(7.10.11)
or, alternatively,
and the relation between aT and Llep is taken from the uniaxial tensile curve.
The stress-strain relations discussed in this chapter are just one of four
sets of relations that must generally be satisfied in solving an elastoplastic
problem. The other three sets of relations are the same as for any elasticity
problem. These are
1. The equations of equilibrium of stresses.
2. The strain-displacement or compatibility relations.
3. The boundary conditions.
To obtain a complete solution we must find a set of stresses and strains
which satisfy these four sets of relations. In the next several chapters it will
be shown how these relations are adapted to specific problems and how
solutions to these problems can be obtained. In all that follows, as in the
preceding, it is assumed that the material is homogeneous, isotropic, and
strain hardens isotropically.
Problems
1. Show that the equivalent stress a 6 defined by equation (7.2.9) can also be
written
a.= VJStJSij = VJ(Si + s~ + S5
2. Show that the equivalent plastic strain increment dev defined by equation
(7.2.10) can also be written
dev = V j.delJ de~ = V j.[(def)2 + (de~) 2 + (den 2]

3. Show that equation (7.3.5) is valid for the Prandtl-Reuss relations.
4. Show that equation (7.3.3) follows directly from equation (7.3.2).
5. Show that the stress-strain relations (7.4.2) are inconsistent with the
definition (7.4.1).
6. Derive equations (7.6.21) and (7.6.22) making use of equations (7.6.20).
7. Show that the following expressions for the effective plastic strain increment
are equivalent:
dep = vt [(de~) 2
+ (de~) 2 + (de~) 2 + 2(de~y) 2 + 2(de~.) 2 + 2(de~x)2] 1 1 2
= ~'3 [(de~) 2 + (de~) 2 + de~ de~ + (de~y) 2 + (de;2 ) 2 + (de~x) 2] 1 1 2
v2
= - - [(de~ - de~) 2 + (de~ - den 2 + (de~ - deD 2
3
+ 6(de~y) + 6(de;2 ) 2 + 6(de~x)2]112
2
8. From the fact that the plastic strain increment vector is normal to the yield
surface, prove that the Prandtl-Reuss equations imply the use of the
von Mises yield criterion.
9. Derive equation (7.6.30) from equation (7.6.28).
References 133
10. Show that the Prandtl-Reuss relations imply that the principal axes of
stress and of plastic strain increment coincide.
11. Derive equations (7.2.12) and (7.2.13) using tensor notation only.
12. Determine the equivalent stress aT for the Tresca criterion by means of
equation (7.6.18). Assume a1 > a2 > as. Also determine the effective plastic
strain increment by the two methods described by equation (7.6.20) and
what follows.
13. Prove that
where S11 is the stress deviator tensor.

14. Show that equations (7.9.7) and (7.9.11) are equivalent.
References
1. B. Saint-Venant, Memoire sur 1' etablissement des equations differentielles

des mouvements interieurs operes dans les corps solides ductiles au dela des
limites ou I'elasticite pourrait les ramener a leur premier etat, Compt. Rend.,
70, 1870,pp. 473-480.
2. M. Levy, Memoire sur les equations generales des mouvements interieurs
des corps solides ductile au dela limites ou l'elasticite pourrait les ramener
a leur premier etat, Compt. Rend., 70, 1870, pp. 1323-1325.
3. R. von Mises, Mechanik der festen Koerper in Plastisch deformablem
Zustand, Goettinger Nachr. Math. Phys., Kl., 1913, pp. 582-592.
4. L. Prandtl, Spannungsverteilung in plastischen Koerpern, Proceedings of the
1st International Congress on Applied Mechanics, Delft, Technische Boek-
handel en Druckerij, J. Waltman, Jr., 1925, pp. 43-54.
5. E. Reuss, Beruecksichtigung der elastischen Formaenderungen in der
Plastizitaetstheorie, Z. Angew. Math. Mech., 10, 1930, pp. 266-274.
6. R. Hill, The Mathematical Theory ofPlasticity, Oxford Univ. Press, London,
1950, p. 25.
7. D. R. Bland, The Two Measures of Work-Hardening, 9th International
Congress of Applied Mechanics, Univ. de Bruxelles, 1957, pp. 45-50.
8. H. Ford, Advanced Mechanics of Materials, Wiley, New York, 1963, p. 416.
9. A. M. Wahl, Effect of Transient Period in Evaluating Rotating Disk Tests
Under Creep Conditions, J. Basic Eng., 85, 1963, pp. 66-70.
10. W. Prager, Strain Hardening Under Combined Stress, J. Appl. Phys., 16,
1945, pp. 837-840.
11. I. Todhunter and K. Pearson, A History of the Theory of Elasticity and
Strength of Materials, Vol. II, Part 1, Cambridge Univ. Press, 1893, p. 166.
12. D. C. Drucker, Some Implications of Work Hardening and Ideal Plasticity,
Quart. Appl. Math., 7, 1950, pp. 411-418.
13. D. C. Drucker, A More Fundamental Approach to Plastic Stress-Strain
Relations, 1st U.S. Congress of Applied Mechanics, ASME, New York,
1952, pp. 487-491.
14. E. Melan, Zur Plastizitaet des raeumlichen Kontinuums, Ingr.-Arch., 9,
1938, pp. 116-126.
......................................;
15. D. R. Bland, The Associated Flow Rule of Plasticity, J. Mech. Phys. Solids,
6, 1957, pp. 71-78.
16. H. Z. Hencky, Zur Theorie Plastischer Deformationen und der hierdurch
im Material hervorgerufenen Nachspannungen, Z. Angew. Math. Mech.,
4, 1924,pp. 323-334.
17. B. Budiansky, A Reassessment of Deformation Theories of Plasticity,
J. Appl. Mech., 26, 1959, pp. 259-264.
18. W. T. Koiter, Stress-Strain Relations, Uniqueness, and Variational Theo-
rems for Elastic-Plastic Materials with a Singular Yield Surface, Quart.
Appl. Math., 11, 1953, pp. 350-354.
19. A. Mendelson and S. S. Manson, Practical Solution of Plastic Deformation
Problems in Elastic-Plastic Range, NASA Tech. Rept. R-28, 1959.
General References
Drucker, D. C., Stress-Strain Relations in the Plastic Range, a Survey of Theory
and Experiment, Office of Naval Research, Contract N7 onr-358, NR-()41-()32,
Dec. 1950.
1950.
Johnson, W., and P. B. Mellor, Plasticity for Mechanical Engineers, Van Nostrand,
Princeton, N.J., 1962.
Naghdi, P. M., Stress-Strain Relations in Plasticity and Thermoplasticity, Office
of Naval Research, Contract Nonr-222 (69), Tech. Rept. No. 9, 1960.
CHAPTER 8
ELASTOPLASTIC
PROBLEMS OF
SPHERES AND
CYLINDERS
Spheres and cylinders are widely used as pressure vessels, in the chemical
industry, for example, as well as many other places. The loads involve high
pressures and sometimes high temperatures and high temperature gradients.
The elastic stress and strain distributions are relatively simple to obtain,
particularly since the loading is usually reasonably symmetric. The solutions
in the elastoplastic range, however, become complicated, and so simplifying
assumptions of various types are made. These usually involve assuming the
material to be incompressible in both the elastic and plastic ranges, and
assuming it to be perfectly plastic in the plastic range. With these assumptions
closed-form solutions can be obtained. We shall first present some of these
classical solutions. Subsequently it will be shown how these problems can be
solved without the usual simplifying assumptions.
For later use we record here the equilibrium, compatibility, strain-
displacement, and stress-strain relations in spherical coordinates and polar
coordinates assuming spherical and axial symmetry, respectively.
Spherical Coordinates
The stresses are designated by a, and a 8 = a,p and the strains by e, and
e8 = e,p. The equilibrium equations reduce to
(8.1.1)
135
136 Elastoplastic Problems of Spheres and Cylinders [Ch. 8
where Fr is the body force per unit volume. The strains are related to the
displacements by
du u
e =-
r dr Be= f = Bq, (8.1.2)
where u is the radial displacement. Combining both of equations (8.1.2) gives

the compatibility equation
dee Be - er 0
-
dr + -r- - - (8.1.3)
Because of symmetry the shear stresses and shear strains are zero as well as
the tangential displacements.
The stress-strain relations are
1
er = E (ar - 2p..ae) + aT+ e{.'
(8.1.4)
1
Be = E [(1 - p..)ae - p..ar] + aT + e~
where e;: and e~ are the total plastic strains. From the incompressibility
condition it follows that
e{.' = -2e~ (8.1.5)
For the von Mises yield criterion, the equivalent stress becomes
(8.1.6)
so that the yield criterion is
(8.1.7)
and the equivalent plastic strain increment is
de P-- ldePIr (8.1.8)
The Prandtl-Reuss relations thereupon reduce to
derp = 2d d ar - ae
- eo
P
= eP Iar- ae I
= dep sgn (ar - ae)
Sec. 8-1] General Relations 137
where sgn stands for "the sign of." We note that if the plastic strains vary
monotonically with the applied load, equation (8.1.9) can be integrated to
give
(8.1.10)
Note also that the Tresca yield criterion in this case coincides with the
von Mises criterion.
Polar Coordinates
We assume axial symmetry and either plane strain or plane stress. The
equilibrium equations then become
(8.1.11)
where Fr is the body force per unit volume, The strain-displacement relations
and corresponding compatibility equation are
(8.1.12)
which are the same as equations (8.1.2) and (8.1.3). The stress-strain relations
are given by
(8.1.13)
For the case of plane stress, uz = 0, and for the case of plane strain ez = 0
or ez = constant for generalized plane strain. In both cases the shear stresses
and strains are zero.
The von Mises and Tresca criteria do not coincide in this case as they do
for the case of spherical symmetry. The yield criteria and corresponding
plasticity relations will be described subsequently as they are used. Several
examples will now be discussed beginning with the case of a thick hollow
sphere.
8-2 THICK HOLLOW SPHERE WITH INTERNAL

PRESSURE AND THERMAL LOADING
Consider a sphere with inner radius a and outer radius b, subjected to an
internal pressure p and a radial temperature distribution T(r). It is obvious
that complete symmetry about the center will exist so that the radial and any
two tangential directions will be principal directions. Equations (8.1.1)
through (8.1.10) apply. We start by finding the elastic solution. Substituting
the stress-strain relations (8.1.4) (with the plastic strains set to zero) into the
compatibility equation (8.1.3) and making use of the equilibrium equation
(8.1.1 ), the following solution for the stresses can readily be obtained:
ar = --2E 1
1 --1-L-
r3
J' aT r dr + -23 ( 1 -
a
2
3
-ar3 ) C1 + -Cr32
(8.2.1)
where C1 and C2 are integration constants. Note that E and a have been
assumed constant in obtaining the above solution. The constants C1 and C2
can be obtained using the boundary conditions
ar(a) = -p
(8.2.2)
ar(b) = 0
resulting in
c2 = -pa 3
_ 3E
Cl - -1-- b3
1
3
fb aT r2 dr + -23 b3pa3 3 (8.2.3)
-1-L -a a -a
For convenience the following dimensionless quantities are now introduced:
b r P=p
{3=-a p=-a
ao
(8.2.4)
EaT = ar
S r- ae
T = (1 Se =-
- /-L)ao ao ao
where a 0 is the yield stress in uniaxial tension. Equation (8.2.1) can now be
written in the dimensionless form as
(8.2.5)
Sec. 8-2] Hollow Sphere with Internal Pressure and Thermal Loading 139
The strains can be computed from (8.1.4) and the displacement from
(8.1.2).
We note that in the case of pressure loads only, the stress distribution is
independent of Poisson's ratio [see equations (8.2.5) and (8.2.4)]. The as-
sumption that is often made that the material is incompressible in the elastic
range (p, = -!), as well as the plastic range, therefore leads to no error in the
elastic stress distribution. In the case of temperature loads, however, assuming
p, = t instead of 0.3, for example, results in approximately a 30 per cent
error in the elastic thermal stresses. The strains are not independent of
Poisson's ratio even for the case of pressure loading. In what follows, the
effect of Poisson's ratio is always taken into account.
The conditions for the onset of yielding in the sphere can now be investi-
gated. In terms of the dimensionless stresses defined in (8.2.4), the yield
criterion (8.1.7) is written
From equations (8.2.5) the yield condition becomes
As a specific example, assume a temperature distribution resulting from an

outward flow of heat due to an inner surface temperature of T 0 and outer
surface temperature of zero. This steady-state temperature distribution will
be given by
T = T 0a (~ _ 1)
b- a r
or 'T =~ (~-
f3- 1 p
1) (8.2.7)
EaT0
where To = (1 - p,)ao
Evaluating the integrals and substituting into the yield condition (8.2.6)
results in
Consider first the case of pressure only. Then the yield condition becomes
3[3 3P
2p3(f33 - 1) = 1 (8.2.9)
Yielding will first occur at the smallest value of p, i.e., p = 1, and the dimen-
sionless pressure necessary to first cause yielding, the critical pressure, will be
2(fi3- 1)
Porlt = 3[33 (8.2.10)
A plot of the ratio of this critical pressure as a function of the ratio of the
outer to the inner radii f3 is shown in Figure 8.2.1. For a given value of {3,
yielding will start at the inner surface at a pressure as given by equation
(8.2.10) or Figure 8.2.1. AsP is increased, the plastic zone will spread from
Peri!
0.8
2/3
0.6 t:-------===:::::=====
0.4
0.2
1.5 2.0 2.5 3.0 3.5

f3
FIGURE 8.2.1 Variation of critical pressure with {3, pressure loading only.
the inner surface toward the outer surface. Note that as f3 approaches infinity
Porit approaches t, so that if the pressure is equal tot the yield stress, yielding
is sure to take place, no matter what the dimensions of the sphere.
Considering the case of temperature only, equation (8.2.8) gives
(8.2.11)
For this case yielding will also first start at the inner surface. However, if
both pressure and temperature are present, yielding may start at any radius,
depending on the relative values of P, T 0 , and {3.
As an example let f3 = 2. Then for the case of internal pressure alone,
equation (8.2.10) gives
7
Porlt = 12 = 0.583
For temperature alone, equation (8.2.11) gives
To,orlt = 1.4
Sec. 8-3) Hollow Sphere. Spread of Plastic Zone. Pressure Loading Only 141
so that assuming E = 31 x 106 , a 0 = 31,000 psi, f1- = 0.3, and a = 7.5 x

IQ- 6 per °F, the temperature difference between the inner and outer surfaces
for yielding to start is 130°F. For both pressure and temperature acting,
equation (8.2.8) gives, for f3 = 2,
Jl2P + (7p 2 - 12)r0 J = 7p 3
For a value of r 0 = 0.4, which corresponds to only a 37°F temperature

difference T0 , yielding will first occur at p = 1 for a value of
Porlt = 0.75
compared to 0.583 for pressure alone. The effect of the temperature gradient
in this case has been to retard the onset of yielding.
A complete discussion of the effects of temperature and pressure on
yielding is given in reference [1].
So far, only the start of yielding has been considered. The spread of the
plastic zone through the sphere is investigated next. The pressure problem
and the temperature problem will be discussed separately in Sections 8.3, 8.4,
and 8.5 under the assumption that the material is perfectly plastic. The general
solution for strain-hardening materials under combined pressure and thermal
gradient is presented in Section 8.6.
8-3 HOLLOW SPHERE. SPREAD OF PLASTIC

ZONE. PRESSURE LOADING ONLY
When only internal pressure is acting, yielding will begin at the inner
surface at a pressure given by equation (8.2.10); i.e.,
(8.3.1)
As the pressure increases, the plastic zone will spread outward toward
the outer surface. Let the radius to the end of the plastic zone be r 0 • Since the
material is assumed perfectly plastic, at every point in the plastic region the
equivalent stress is equal to the yield stress and since for this case a8 >a.,
(8.3.2)
in the plastic region. Substituting into the equilibrium equation (8.1.1) gives
dSr 2
dp =p
or Sr = 2ln p +C
But at
p= 1 Sr = -P
Therefore,
C = -P
and S, = 2ln p- P }
P < Pc (8.3.3)
8 8 = (21n p + 1) - P -
Equations (8.3.3) give the stresses in the plastic region. Note that no stress-
strain relation was needed to obtain these stresses. The problem is therefore
called statically determinate.
At the plastic zone boundary, i.e., at p = Pc, the radial stress is
Sr,c = 2ln Pc - P (8.3.4)
We can now consider the elastic part of the sphere as a new sphere with
inner radius r c and outer radius b, with an internal pressure given by equation
(8.3.4). Since at this new inner radius the sphere is just at the yield point,
equation (8.3.1) must apply with {3 replaced by f3c = b/rc, and -Pcrit replaced
by Sr,c· Hence
2 f3c 3 - 1
2lnpc- P = -3~
or P = 2 ln Pc + i( 1- ;~)
= 2ln ~
a
+ ~3 (1 - br~)
3 (8.3.5)
Equation (8.3.5) gives the pressure required to cause the plastic zone to reach
a radius rc or, alternatively, for a given internal pressure P, equation (8.3.5)
could be solved for the plastic zone radius rc. A plot of the pressure versus
the plastic zone radius is given in Figure 8.3.1 for f3 = 2.
When rc becomes equal to b, the sphere is completely plastic. This will
occur at a pressure [from (8.3.5)]
P = 2ln {3 (8.3.6)
Sec. 8-3] Hollow Sphere. Spread of Plastic Zone. Pressure Loading Only 14 3
Pc
FIGURE 8.3.1 Plastic zone radius versus applied pressure, f3 = 2.
From (8.3.3) and (8.3.6) the stresses will be
S, = 2 ln p - 2 ln f3
= 2ln~
(8.3.7)
r
= 2ln b
r
S 0 = 1 + 2ln b
As a check, note that S 0 - S, = 1 for all r.

So far the stresses in the plastic part of the sphere have been computed.
The stress distribution in the elastic part of the sphere can be readily obtained
by considering the elastic portion of the sphere to be a new sphere with inner
radius r 0 , outer radius b, and with the pressure at the inner radius equal to
the critical pressure for a sphere with these dimensions. The elastic solution,
equations (8.2.5) (without temperature terms), can be used for this fictitious
sphere replacing a by rc and f3 by {3 0 • Thus
or
(8.3.8)
Equations (8.3.3) and (8.3.8) give the complete stress distribution in the
sphere for given ratio of applied internal pressure to yield strength, with the
plastic zone boundary Pc given by equation (8.3.5).
To calculate the strains and displacements in the sphere, the stress-strain
relations and strain-displacement relations are used. For convenience we
define, as was done for stresses in equations (8.2.4), "dimensionless" strains
and displacements as follows:
(8.3.9)
To compute the strains before yielding begins, equations (8.2.5) (with tem-
perature terms deleted) are substituted into the stress-strain relations (8.1.4),
resulting in
(8.3.10)
and from the strain-displacement relation (8.1.2),
(8.3.11)
When yielding starts P is equal Pcrit given by equation (8.2.10), and the
displacement at the inner surface, p = 1, is
Note that if incompressibility had been assumed in the elastic region, p, = f,

the first term on the right of equation (8.3.12) disappears. For a Poisson's
ratio of 0.3 and f3 of 2, the error in the displacement of the inner surface
the beginning of yield would be about 7 per cent. At the outer surface the
error is 38 per cent.
As yielding progresses to some radius r 0 , we can consider as before a
sphere with inner radius r 0 , outer radius b, and critical pressure
2 {3~ - 1
Pcr!t =3 -r
Sec. 8-4] Hollow Sphere. Residual Stresses. Pressure Loading 145
acting at the inner radius rc. From (8.3.11) the displacement at r:::: rc is
obtained by replacing f3 by f3c and p by rfrc, resulting in
3
2 [ (1 - 2/k)P
U = 3{3~ + -l+ftf3
2 - p2 ] (8.3.13)
When the plastic zone reaches the outer radius b, p = [3, f3c = I, and
(8.3.14)
The error in assuming fL = -!is about 30 per cent, for fL = 0.3.

In all the previous computations it was assumed that the dimensions of the
sphere do not change as the pressure increases. This is, of course, not true,
since the sphere grows with increase in internal pressure, the inner radius
becoming a + Ua and the outer radius b + ub. A rigorous analysis would
therefore have to take into account the change of dimensions of the shell.
This can become particularly important for large strains.
8-4 HOLLOW SPHERE. RESIDUAL STRESSES.

PRESSURE LOADING
If the pressure is removed from the sphere discussed in Section 8.3 after
plastic flow has occurred over part of the sphere, residual stresses will result.
To find the residual stresses it is necessary to superpose on the stress system
due to the internal pressure p and temperature T a completely elastic stress
system due to a pressure -p and temperature -T. This will be correct as
long as yielding in reverse does not occur; i.e., the residual stresses are not
large enough to produce yielding. To see this, consider two stress systems
satisfying the following two sets of equations:
I I I
dar 2 ar - ae- 0
-+
dr r
-
de 8 e~ - e~
dr = - r -
(8.4.1)
a;(a) = -p
a;(b) = 0
and
II II II
dar 2 ar - ao- 0
-+
dr r -
ll II II
dBr Br - eo
dr = --r-
(8.4.2)
eoII = E
1 [(1 - P,) aoII - p,arII] - a T
a;(a) = p
a;{b) = 0
The primed system corresponds to the system of stresses in the sphere with
temperature T and internal pressure P. The double-primed system corre-
sponds to the stresses in a sphere with temperature - T and internal pressure
-P. If the two systems are added together, there is obtained a system of
stresses ar = a~ + a;, etc., satisfying the following equations:
(8.4.3)
eo = ~ [(1 - p,)ao - p,ar] + e~

ar(a) =0
ar(b) = 0
Thus the resultant system corresponds to the unloaded sphere with permanent
plastic strains due to the first system. If plastic flow occurs during the un-
loading, the elastic double-primed system can no longer be added to the
original system, but it is necessary to solve another plastic flow problem for
the new plastic strains.
For the case of pressure loading only, the elastic stresses due to a pressure
equal to - P are, from (8.2.5),
2p3 + [33
Sll- P
8 - - 2p3(!33 - 1)
Sec. 8-4] Hollow Sphere. Residual Stresses. Pressure Loading 147
Adding to the stresses given by (8.3.3) and by (8.3.8) gives for the residual
stresses
p3 - (33 )
s~ = 2ln p - p ( 1 + p3((33 - 1)
= 2 In p - ~3 Pcrlt
_!_ ( 1 - \)
P
or s; = ~3 [3ln p - _!_
Porit (1 - ~)]
P
}
p < Pc (8.4.5)
s~ = ~3 [~2 + 1
3ln p - _!_ (1 + 2 p3)]
Pcrit
-
s; = j [(P: 1
t - P~) {: 3 - ; 3 )] }
2 P 1 1 p :::0: Pc
(8.4.6)
S~ = -3 [(Pcrit - P~ )(2p + (3 )]
3 3
The superscript r is used in the above equations to indicate residual stresses.

When p = 1 (at the inner surface), S~ = 0, as expected, and
(8.4. 7)
an~ since P :::0: Parit• a residual compressive stress results. Upon reapplication
of a pressure less than or equal to the original maximum, only elastic strains
will occur. The shell has thus been strengthened by the initial pressurization.
If the material work hardens, an even greater strengthening can be achieved.
In the above derivation it has been assumed that no plastic flow takes place
during the unloading; i.e., there is no yielding in compression due to the
residual stresses. If such yielding occurs, then not only is our assumption
that the unloading is elastic violated, but the situation may be dangerous
with regard to the safety of the sphere. The maximum value of applied
pressure P such that if the sphere is unloaded there will be no reversed plastic
flow is called the shakedown pressure, P•. This pressure can be found as
follows. For reversed yielding the yield criterion can be written
s;- s~ = 1 (8.4.8)
The maximum residual stress will occur at p = 1, where s; = 0. From

(8.4.7) it therefore follows that
Ps = 2Porit (8.4.9)
Ps
1.6
4/3t------=:::::::::=====
1.2
1.5 2.0 2.5 3.0 3.5

{3
FIGURE 8.4.1 Variation of shakedown pressure with thickness ratio for hollow
sphere with internal pressure.
As long as the applied pressure is less than twice the critical pressure, the
residual stresses will be elastic. Making use of equation (8.3.1) the shakedown
pressure can be written directly as a function of the thickness ratio f3:
(8.4.10)
Figure 8.4.1 shows the shakedown pressure as a function of {3.
8-5 HOLLOW SPHERE. THERMAL LOADING ONLY
For the case of a temperature gradient as given by equation (8.2.7),

yielding will occur as given by equation (8.2.11). Then
(8.5.1)
2.0
1.6
+-
'Ei. 1.2
~ ~----------------
0.8
0.4
234567 8
{3
FIGURE 8.5.1 To,orlt as a function of {3.

Sec. 8-5] Hollow Sphere. Thermal Loading Only 149
Yielding will first occur at r = a, (p = 1), and the critical temperature

difference at which yielding will first start is given by
2({33 - 1) 2((3 2 + (3 + 1)
= (3(2(32 _ (3 _ 1) (8.5.2)
To,orit
(3(2(3 + 1)
A plot of To,crit versus (3 is shown in Figure 8.5.1.

If To exceeds To,cr1t, the plastic zone will spread outward to some radius r 0 •
Within this zone, i.e., for r:::; r 0 , the yield criterion JS0 - S,J = 1 will apply.
But since in this region the tangential stress will be a large compressive stress
and the radial stress will be a small compressive stress, the yield criterion
can be written
(8.5.3)
The equilibrium equation now becomes
dS, 2
dp-- p
or S, = -2ln p +C
and since S,(1) = 0, C = 0. Hence
S, = -2lnp }
(8.5.4)
S0 = -1-2ln p P <
-
Pc
Note that the stresses in the plastic region are independent of the temperature.
The radius of the plastic zone, r 0 , of course, depends on the temperature.
FIGURE 8.5.2 Two plastic zones due to temperature gradient.

As the temperature is further increased, a second plastic zone unconnected

to the first may start at a new radius, depending on the value of f3 as shown in
Figure 8.5.2. This is due to the fact that since there are no external forces
acting on the sphere, the resultant force acting on any cross s'ection must
vanish. Thus the tangential stress varies from compression at the inner
surface to tension at the outer surface. The inner surface will begin flowing
plastically in compression, but if the temperature gradient is high enough, the
outer surface will start flowing plastically in tension, thus producing two
plastic zones, one in tension and one in compression. A detailed discussion
is given in reference [1].
8-6 HOLLOW SPHERE OF STRAIN-HARDENING

MATERIAL
We now consider the general case of a hollow sphere of strain-hardening

material with both pressure and thermal loads. Equations (8.1.1) through
(8.1.9) apply and in addition the assumption is made that the plastic strains
are varying monotonically, so that equation (8.1.10) may be used. If this is
not the case, equation (8.1.9) is used instead, and the calculation performed
in steps or increments as described subsequently for more general types of
problems. In addition, we use the dimensionless quantities defined by
equations (8.2.4) and (8.3.9). The equilibrium compatibility and stress-strain
relations are now written
dSr + 2(Sr - Se) =O

dp p
dEe + Ee - Er = 0
dp p (8.6.1)
Er = Sr - 2/hSe + (1 - fh)T + Ef
Ee = (1 - !h)S8 - fhSr + (1 - f.L)r + E~
Substituting the last two of equations (8.6.1) into the second, combining
with the first, and integrating results in the following equations:
(8.6.2)
Sec. 8-6] Hollow Sphere of Strain-Hardening Material 151
(8.6.3)
(8.6.4)
where
(8.6.5)
The boundary conditions used in deriving the above equations were
Sr(I) = -P
(8.6.6)
S//3) = 0
For the elastic case equations (8.6.2) reduce to (8.2.5). For the case of a
perfectly plastic material, the solution was given in the previous sections. We
shall consider here only the case of a strain-hardening material.
To obtain a complete solution to the problem, it is necessary to determine
the plastic strain distribution E~ through the sphere. This will, of course,
depend on the stress-strain curve of the material. The plastic strain distri-
bution can be obtained in the following manner. The equivalent stress is
related to the equivalent plastic strain through the stress-strain curve of the
material. Thus
or, for this case,

lSI = /()E~j)
where f is a known function representing the stress-strain curve. It therefore
follows that
lSI ;::: I
(8.6.7)
)SI ~I
Also, from equation (8.6.3),
(8.6.8)
A complete solution can now be obtained by an iterative or successive

approximation method. One chooses a distribution of E~ (say zero). Sis com-
puted using equation (8.6.3), merely to determine its sign at the different
radial positions. A first approximation to Sis then obtained from equation

(8.6.7), and a first approximation to the plastic strain distribution can be
calculated from (8.6.8). A better value of S can then be computed from
(8.6.7) and the next approximation forE~ obtained from (8.6.8). If the process
converges, we will thus obtain the proper values of E~ and S such that (8.6.7)
and (8.6.8) are satisfied simultaneously. The individual stresses S, and S8 can
then readily be computed. Thus a complete solution is obtained in both the
elastic and plastic regions. There is no need to treat the two regions separately
as was done for the perfectly plastic material in previous sections. It should
be noted, however, that equations (8.6.7) and (8.6.8) apply only for lSI > 1;
for lSI ::; 1, Ef is set equal to zero. The integrands Ef/p appearing in the
previous equations are therefore generally zero over part of the integration
range. A similar technique is described in reference [9].
As a specific example, consider a sphere made of a material whose stress-
strain curve is given by the following equation:
ae = 30,000 + 136,000e~12 ae;::: 30,000
Equation (8.6.7) now becomes
S = ~~~ (1 + 0.1434\efil 12)
Results of calculations performed by the iterative procedure described are

shown in Figures 8.6.1 and 8.6.2 for f3 = 2 and temperature distribution
given by equation (8.2.7). In performing the calculations, the thickness of
the sphere was divided into 40 equally spaced intervals and Simpson's rule
2.0
p
c 1.5
p
FIGURE 8.6.1 Variation of plastic zone radius with applied pressure for different
temperature gradients:
f3 = 2, T = -r 0 (f3/p - 1)/(/3 - 1), ae = 30,000 + 136,000 ev 1' 2 •
FIGURE 8.6.2 Variation of plastic strain with radius for various pressures:
{3 = 2, To = 0.
was used to perform the integrations. The cases shown are for illustrative
purposes only. Any combination cf geometry, loading, and material proper-
ties can be used and a rapid solution obtained. The time required to obtain
a complete solution for a given loading condition, using a high-speed digital
computer, is on the order of a few seconds.
This type of successive approximation method will be discussed at greater
length in Chapter 9, where several numerical examples will be given. Right
now it will be shown that if the material strain hardens linearly the solution
can for some cases be obtained in closed form.
For linear strain hardening it follows from Figure 8.6.3 that
Ev = 1- m (I Sl - 1) (8.6.9)
m
where the strain-hardening parameter m is defined as the ratio of the slope of

the strain-hardening part of the stress-strain curve to the elastic modulus.
Then, from (8.1.10),
1-m S
= --(1-ISj)- 1-m(S - s )
Ep
r m lSI = -m- -
lSI (8.6.10)
Consider the case of pressure loading only. a8 will always be positive and
ar will always be negative, so that a 8 - ar > 0. Therefore, S/1 S I +1 and
1-m
Ef = ---m (1 - S) (8.6.11)
"e
~ e
FIGURE 8.6.3 Stress-strain curve for linear strain hardening.
If the plastic zone extends to p = Pc• then, from (8.6.1I), making use of the
first of equations (8.6.1), it follows that
I- m I- m
Pr;P
--lnp- --(S
m 2m r
+ P) P ~ Pc
...!.. dp = (8.6.I2)
fl p { I - m In
m
_ I - m (S + P) P;::: Pc
Pc 2m r,c
where Sr,c is the value of S, at p = Pc· Substituting into (8.6.3) and (8.6.4)
results in
(8.6.I3)
I- m 3
S = 2(1 - tL)m (1 - S) + 2ps Cl (8.6.14)
and, since S = 1 when p = Pc• we have
(8.6.15)
(8.6.16)
At the onset of yield when Pc = 1, we have Sr,c = -P and therefore

which is the same result previously obtained for the perfectly plastic material.
Obviously the onset of yield depends only on the yield stress. AsP is increased,
the plastic zone spreads to Pc and -Sr,c can be considered to be the critical
pressure acting on a sphere with inner radius Pc and outer radius fie = fJ/ Pc·
Thus
(8.6.18)
Hence
2 fJ3 · [( 1 I - m ) 1- m 1 - m fJ~ - 1]
- p,)m P - (1 - p,)m In Pc - 3(1 - p,)m ~
3
3 Pc = fJ 3 - 1 + 2(1
(8.6.19)
or
p _ 4(1 - p,)m[(fJ3 - 1)/{13 ] p~ + 2(1 - m) ln Pc + i(l - m)(fJ~ - 1)/fJ~
- 2m(l - p,)+ (1 - m)
(8.6.20)
which relates the pressure P to the plastic zone radius Pc· Note that if m = 0,
2{1~-1
P = 2lnp 0 + 3~
which is the value previously obtained for the perfectly plastic material
[equation (8.3.5)].
As an example, for fJ = 2, m = 0.1, and p, = 0.3, the pressure required
for yielding of the complete sphere, Pc = fJ, is 1.83, compared to 1.39 for a
perfectly plastic material. It thus takes a 32 per cent higher pressure for the
strain-hardening sphere to yield completely as compared to the perfectly
plastic sphere.
To obtain the stresses we substitute into equations (8.6.2). Thus
Sr = -P+(l-l)c +- 1 1-mlnp _l-m(S +P)J}

p3 1 _-(
/1- m c 2m 1
r,c
P;:: Pc
-P+ ( 1 +-1) +1 --1 -/1- [1- --mlmn pc - -1 2m
- (mS +P)J
S0 =
2p 3 C1 - r,c
(8.6.21)
S, = -P + 2(1 - m) ln P +!(I - p,)m(p~jp3)(p3 - 1) }

1-m + 2(1 - p,)m 1 - m + 2(1 - p,)m
P ~ Pc
So= -P + 2(1 - m) In + .2. (1 - p,)m(p~jp3)(2p3 + 1)
1 - m + 2(1 - p,)m P 3 1 - m + 2(1 - p,)m
(8.6.22)
Note that if m = 0 these reduce to the previously obtained values for the
perfectly plastic material. Thus to obtain the complete stress distribution,
Pc (or P) is obtained from (8.6.20), Sr,c from (8.6.18), and then the stresses
from (8.6.21) and (8.6.22).
8-7 PLASTIC FLOW IN THICK-WALLED TUBES
A considerable amount of work has been done on the problem of plastic

flow in a thick-walled tube under internal pressure with and without tempera-
ture gradients. Solutions have been obtained, for example, in references [1]
through [8]. These solutions differ in the yield criteria used and in the plastic
stress-strain relations. Some solutions use the von Mises yield criterion and
the associated flow rule [3]. Others use the Tresca criterion and its flow rule
[7]. Reference [4] uses the Hencky total strain relations. In other papers
complete incompressibility is assumed in both the elastic and plastic regions.
Of the references cited, only [8] takes into account strain hardening of the
material.
There are three cases that can be treated: (1) plane strain, e z = 0; (2)
generalized plane strain, ez = constant =1 0; and (3) tube with open ends,
P = 0. We shall present a general solution for a strain-hardening material,
including radial temperature gradients, which can take into account any of
these cases.
The Tresca criterion and its associated flow rule will be used, since in this
case it offers some simplifications. For this purpose it will be assumed that
a 8 > az > ar· It is shown by Koiter [7] that this is true for a large range of
conditions.
We introduce the same dimensionless quantities as in the problem of the
sphere; i.e.,
EaT EaT0 r b
p=- (3=-
T = (1 - tL)ao y = (1 + tL)ao a a
where a and b are the internal and external radii, a 0 is the yield stress, and e0
is the yield strain.
Sec. 8-7] Plastic Flow in Thick-Walled Tubes 157
Equations (8.1.11), (8.1.12), and (8.1.13) are now written
dSr Se- Sr
dp = - p -
(8.7.2)
Er = Sr - P-(Se + Sz) + (1 - P-) T + Er
Ee = Se - P-(Sr + Sz) + (1 - f-t) T + Ef (8.7.3)

Ez = Sz - f-1-(Sr + Se) + (1 - f-t) T + €~
If the Tresca criterion and its associated flow rule are used, then, assuming
Se > Sz > S,
(8.7.4)
and S = Se- Sr = 1 at yielding
For boundary conditions it is assumed that
Sr(a) = -P
(8.7.5)
Sr(b) =0
and the conditions at the end of the tube are determined by case I, 2, or 3
above. For plane strain Ez = 0. For generalized plane strain, Ez is a constant
which can be determined from the end loads on the tube. Thus let the axial
force acting on the tube be F. Define
Then it readily follows from the condition
F* = J: Szp dp
and the third of equations (8.7.3) that
(8.7.6)
If the axial force is due to internal pressure only, then F* = P/2 and
€z = f32 ~ 1 [o - 2p,)P + 2(1 - 11-) J: TP dp] (8.7.7)
For a tube with open ends, P = 0 and
€z
2(1 - !1-)
= [32 - 1 1 TP dp
Jp (8.7.8)
In any case €z is a known constant. From the last of equations (8. 7.3),
(8.7.9)
Substituting this relation into the first two of equations (8.7.3), making use
of the equilibrium and compatibility equations (8.7.2), and integrating results
after some algebraic manipulations in the following solution:
sr = p - -1
-- Jp TP dp + -1- Jp€P dp + ( 1 - -
1) C1
p2 p2
1
1 _ f/-2 1 ..L.
p p2
(8.7.10)
so = P - 'T
1
+ _!_ Jp TP dp + -1 - - (€r + Jp €r dp) + (1 + P\)c1
p2 p2 1 f/-2 1 p
2p - + 2p2JP TP dp + -1--2
s = 2P 'T
1
-11- 1
p
€r + 2p2 c1 (8.7.11)
(8. 7.12)
Yielding will begin at p = 1 when S = 1, so that the critical pressure will be
and, from (8.7.12),
Therefore,
Pcrlt
2
-1(
= {32P 1 + 'T - {3 2 2 1 J/3 TP dp )
_
1
(8. 7.13)
For P .::; Pcrlt> we have the elastic solution, which agrees with the classical
elastic solution. For P ;:::: Pcrlt a plastic zone will spread out to some radius
Pc· The solution for general strain hardening can be obtained by an iterative
or successive approximation method, as indicated for the sphere.
Let the stress-strain curve be given by a relation of the form
(8.7.14)
To relate Er to Ep, the two methods indicated in Section 7.6 may be used. If
the definition
is used, then from (8.7.4),
(8. 7.15)
On the other hand, if the definition (7.6.20) is used, then it follows that
(8.7.16)
The two definitions differ by the familiar constant 2/V3 and either one can
be used. Since the definition based on the plastic work increment appears
to be more consistent with the Tresca criterion, we shall use it, and assume
for the case under consideration that (8.7.14) may be written
lSI ;::: 1
(8.7.17)
lSI ::;; 1
To use the successive approximation method, it is preferable, as was done the

for case of the sphere, to rewrite equations (8.7.11) and (8.7.17) as follows:
(8. 7.18)
S = /(IErl} sgn S
Er = o
lSI;:::
lSI::;;
1}
1
(8.7.19)
An initial distribution of Er (such as zero) is assumed. The signs of S through-

out the cross section of the tube are then determined from (8. 7.11 ), and the
actual values of S are calculated from (8. 7.19). A better approximation can
now be obtained for the Er using equation (8.7.18). The process is repeated
until convergence is obtained.
For the cases of the sphere and the tube heretofore discussed, the successive
approximation method has been found to converge fairly rapidly using the
techniques described. However, this may not always be true. A general dis-
cussion of the convergence of the successive approximation method is given
in Chapter 9.
For the case of linear strain hardening, a solution can be obtained in

closed form for the above problem. As for the sphere problem, equation
(8.6.11) is used:
1-m
ef =- - ( 1 - S)
m
(8.7.20)
PEf 1-m
J P dp = ---,:n- (In Pc -
1
Sr,c - P)
Therefore,
2
c1
1
= ,82 - 1
[JP TP dp
1
,8 (1 - m)
+ p - m(l - /h2) (In Pc - Sr,c - P)
] (8.7.21)
Also, from (8.7.11),
s = - T + 2p2JP1 TP dp + 2p2 (P + C1) + m (11-m

-/h
2) (1 - S) P :0; Pc
(8.7.22)
When p = p0 , S = 1. Therefore, from (8.7.22),
c1 =
p2 [ 1 + T(Pc) - p~
2 2 fPc
1 TP dp
] - p (8.7.23)
Substituting into the expression for C1 , (8.7.21), we get
2 fPc ]
~2 [ 1 + T(p0) - p~ 1
TP dp
=
,82 [
,82 - 1 p + ,82lJP1 TP dp 1-m
- m(l - /h2) (ln Pc - Sr,c - P)
] (8.7.24)
As a check, at the onset of yield, Pc = 1, Sr,c = -P, and (8.7.24) reduces to

(8.7.13) for the critical pressure.
At p = Pc we can consider a new tube with inner radius Pc and outer
radius ,8 0 with Sr,c equal to -Porlt· Thus
,8~ - 1 [ 2 fPc ]
Sr,c = - 2,8~ 1 + T(Pc) - ,8~ _ 1 Pc TP dp (8.7.25)
Solving (8.7.24) for P gives

Equation (8.7.26) gives the relationship between the plastic zone radius Pc
and the applied pressure P, for a given temperature distribution r. For a
perfectly plastic material this reduces to
P = In Pc - Sr,c
{3~ - 1 [ 2 (f3c ]
=In Pc + 2[3~ 1 + r(pc) - [3~ _ l JPc rp dp (8.7.27)
To obtain the stresses we now substitute into equations (8. 7.1 0). Thus,
since
P€~ 1-m
J 1
- dp
p
= - - (In p - Sr -
m
P) P :o; Pc
P €~ 1- m
= ---,:;:z- (In Pc - Sr,c - P)
J 1
Pdp P :2: Pc
then, from (8.7.10),
2 2
p - 1 JP pr dp p)
Sr = m(l
l
- 11-
2
) (
C1 - -
2.- -
1
2 - 2
-11-m p p 1 P
1-m
+ 1 2 (In P - P)
-11-m
P :o; Pc
So = m(l -
1 _ 11-2m
!1-2) (c1 P2 + 1 + _!_ JP pr dp + p - r)
P2 P2 1 P2
+ 11-m
-11-m
(1 + ln p 2 - P)
and (8.7.28)
Sr = -2
IJP rp dp + (l 1-m (In Pc 2)
·
- Sr,c - P)
p 1 -11-m
P :2: Pc
So = - 'T + 2p1 JP1 rp dp + (I I-11-m
- m
2) (In Pc - Sr,c - P)
+ ~ + ( 1 + :2)c1
To obtain the complete stress distribution, we compute P or Pc from (8.7.26),
C1 from (8.7.23), Sr,c from (8.7.25), and the stresses from (8.7.28).
Problems
1. Show that for the sphere with radial symmetry, the von Mises yield criterion
becomes
and the Prandtl-Reuss equations reduce to
de;' = - 2de% = dev sgn (a, - a0 )
2. Explain why one would expect the Tresca and von Mises yield criteria to
coincide for the case of a sphere with radial symmetry.
3. Derive equations (8.2.1) and (8.2.3).
4. Obtain the equations for all the strains and displacements in the ·sphere
before yielding begins for pressure loading only, for thermal loading only,
and for the case when both thermal and pressure loading exist.
5. Show that the steady-state temperature distribution in a sphere of inner
radius a and outer radius b is equal to
T = Toa (!?. _ 1)
b- a r
if the inner and outer surfaces are kept at temperatures of T 0 and zero,
respectively.
6. Using equations (8.3.3) and (8.3.8), show that the stresses are continuous
across the elastoplastic boundary.
7. Compute the displacements and strains in a sphere with pressure loading
only, for r :o; rc. Assume a perfectly plastic material and that the dimensions
remain fixed. Determine the error in the displacements of the inner radius
for the fully plastic case if J1, is assumed to equal 0.5 instead of 0.3.
8. Show that for a hollow sphere with a temperature distribution given by
equation (8.2. 7), the tangential stress is compressive and the radial stress
is near zero in the region adjacent to the inner circumference, so that the
yield criterion in this region can be written
Se- S, = -1
9. Starting with equations (8.6.1), derive equations (8.6.2) and (8.6.4) using
boundary conditions (8.6.6).
11. Derive equations (8.6.12).
13. Derive equations (8.7.10) through (8.7.12).
14. Show that the definition (7.6.20) for the equivalent plastic strain increment
leads to equation (8.7.16) for the case of a tube with the Tresca criterion
and associated flow rule, if a 0 > a 2 > a,.
General References 163
15. Plot the plastic zone as a function of the applied pressure for a sphere with
linear strain hardening. Assume f3 = 2, m = 0.1, and 1'- = 0.3. Compare
the results with those for a perfectly plastic material.
16. Repeat Problem 15 for a tube.
17. Perform complete numerical analysis of the problem of the sphere,
a :$ r :$ b, T(a) = T0 , and T(b) = 0. Assume E = 30 x 10 6 , 1'- = 0.3,
a= 10- 5 , a.= 30,000 + 136,000 (ep + 10- 4 ) for ae > 30,000, and E, a,
and 1'- are independent of temperature.
References
1. W. Johnson and P. B. Mellor, Plasticity for Mechanical Engineers, Van
2. R. Hill, The Mathematical Theory of Plasticity, Oxford Univ. Press, London,
1950.
3. P. G. Hodge and G. N. White, A Quantitative Comparison of Flow and
Deformation Theories of Plasticity, J. Appl. Mech., 17, 1950, pp. 180-184.
4. D. N. de G. Allen and D. G. Sopwith, The Stresses and Strains in a Partially
Plastic Thick Tube Under Internal Pressure and End-Load, Proc. Roy. Soc.
(London), A205, 1951, pp. 69-83.
5. M. C. Steele, Partially Plastic Thick-Walled Cylinder Theory, J. Appl. Mech.,
19, 1952, pp. 133-140.
6. R. Hill, E. H. Lee, and S. J. Tupper, The Theory of Combined Plastic and
Elastic Deformation with Particular Reference to a Thick Tube Under
Internal Pressure, Proc. Roy. Soc. (London), A191, 1947, pp. 278-303.
7. W. T. Koiter, On Partially Plastic Thick-Walled Tubes, Biezeno Anniversary
Volume in Applied Mechanics, N. V. De Technische Uitgeverij H. Starn,
Haarlem, 1953, pp. 232-251.
8. D. R. Bland, Elastoplastic Thick-Walled Tubes of Work-Hardening Material
Subject to Internal and External Pressures and to Temperature Gradients,
J. Mech. Phys. Solids, 4, pp. 209-229.
9. I. S. Tuba, Elastic-Plastic Analysis for Hollow Spherical Media Under
Uniform Radial Loading, J. Franklin Inst., 280, 1965, pp. 343-355.
General References
1950.
Johnson, W., and P. M. Mellor, Plasticity for Mechanical Engineers, Van Nost-
rand, Princeton, N.J., 1962.
CHAPTER 9
THE METHOD
OF SUCCESSIVE
ELASTIC
SOLUTIONS
9-1 GENERAL DESCRIPTION OF THE METHOD
In Chapter 8 it was indicated how the sphere and tube problems can be
solved for arbitrary strain hardening by a successive-approximation method.
This method is nothing more than the extension of Picard's method (see
reference [1]) of successive approximations to nonlinear equations. The
method was apparently first used in plastic flow problems by Ilyushin [2]
in his treatment of a thin shell. Ilyushin refers to it as the method of successive
elastic solutions, since each iteration involves essentially the solution of an
elastic problem.
Before the advent of modern high-speed computing machinery, this
method could be used only for relatively simple problems. However, with
current widespread availability and use of digital computers it now becomes
possible to solve simply and quickly many problems whose elastic solution
can be obtained by numerical methods.
Before proceeding to describe this method for the general elastoplastic
problem, we shall first give an illustration of the method of successive
approximations for a simple differential equation [3]. Consider the equation
dy
- -y = 0 y(O) = 1 (9.1.1)
dx
164
Sec. 9-1] General Description of the Method 165
The solution is known to bey = ex. To find the solution by Picard's method,
we proceed as follows. Integrate (9.1.1) to give
(9.1.2)
Assume as a first approximation for y,
y<l) = 1
Substitute this value for y on the right side of (9.1.2) and calculate a second
approximation for y:
y< 2l = 1 + J: y< 1l dx = 1 +x
Substitute the second approximation for y and calculate the third approxi-
mation:
y<3l = 1 + rx y<2l dx =
Jo
1 + X + X2
2
Continue in this way to get
y<4l =
rx
+ Jo y<3l dx = 1 + x + 2! + IT
x2 x3
x2 xn
y<n + l) = 1 +
ix
o
y<n) dx = 1 + X + - + · · · + -
2! n!
As n gets larger and larger, it is seen that the infinite series for ex is approached.
The exact solution can thus. be approached as closely as desired by taking
more and more approximations.
This technique can be directly extended to the general elastoplastic problem
in the following manner. For convenience the pertinent equations given in
previous chapters will be repeated here. The equilibrium and compatibility
equations are independent of the plasticity relations and are given by equa-
tions (3.2.2) and (4.7.2); i.e.,
Oax + OTxy + OTxz = -Fx

ox oy Bz
(9.1.3)
166 The Method of Successive Elastic Solutions [Ch. 9
and
(9.1.4)
~ ( _ OBzx + OBxy + OByz) = o2ey

oy oy oz ox oz ox
2
~ ( _ OBxy + OByz + OBzx) = o ez
oz 8z ox 8y ox oy
The stress-strain relations depend on the plasticity theory used and we can
write
(9.1.5)
1 p A 1-
Byz = 2G Tyz + Byz + UByz
_ 1 p A P
Bzx - 2G Tzx + Bzx + UBzx
where s~, e~, etc., are the total accumulated plastic strains up to, but not
including, the current increment of loading, b..s~, b..s~, b..sf, etc., are the
plastic strain increments due to the current increment of loading. The plastic
strain increments are related to the stresses through the yield criterion and
the associated flow rule. For definiteness we shall consider the Prandtl-Reuss
relations, but any other set of relations can be used equally well. Thus
(9.1.6)
where
l1Bp =
v2v(l1B~-l1B~) +(l1B~ -1:1BD +(1:1B~ -l1B~) + 6[(l1B~y) +(l1B~z) +( l1B~
2 2 2 2 2
x?J
3
(9.1.7)
Alternatively, the plastic strain increments can be related to the modified

total strains as described in Section 7.9; i.e.,
A P -
LJ.Bx - -l1Bv (2 Bx' - By' - Bz')
3Bet
A P -
uBy - -l1Bv (2 By' - Bz' - Bx' )
3Bet
(9.1.8)
where
ByI = By - Byp etc.

etc. (9.1.9)
Finally, the equivalent plastic strain increment llBp is related to the equivalent
stress a. or the equivalent total strain Bet through the stress-strain curve. In
addition, of course, the boundary conditions must always be satisfied. For
complete generality, the full three-dimensional equations have been written
out above. In practice only one- and two-dimensional problems can usually
be solved, which is equally true for elasticity problems.
The method of successive approximations now proceeds as follows. The
loading path is divided into a number of increments. For the first increment
of load, a distribution is assumed for the plastic strain increments llB~, llB~,
etc. The total plastic strains B~, etc., are zero. The set of equations (9.1.3),
(9.1.4), and (9.1.5) are now solved as for any elasticity problem and a first
approximation obtained for the stresses and total strains. At the same time,
using the assumed values of the plastic strain increments, an equivalent
plastic strain increment llBp is computed by the first of equations (9.1.7).
From the stress-strain curve the corresponding value of a. can be determined.
A new approximation can now be obtained for the individual plastic strain
increments using the Prandtl-Reuss relations (9.1.6).
Using these new plastic strain increments, equations (9.1.3) through (9.1.5)
are solved again as a new elastic problem. A second and presumably better
approximation is obtained for the stresses and total strains. At the same time,
using these last values of the plastic strain increments a new approximation
is computed for the equivalent plastic strain increment llBp by the first of
equations (9.1.7). Using this value of llBp, a new value is obtained for a8
from the stress-str.ain curve. New approximations are now obtained for the
plastic strain increments !lBL etc., using the Prandtl-Reuss relations (9.1.6).
The process is continued until convergence is obtained; i.e., the differences
between two successive sets of strain increments are less than some prescribed
values. The calculation scheme is illustrated by the flow diagram of Figure
9.1.1.
In this manner the solution is obtained for the first increment of loading.
For the next increment of load, an exactly similar calculation is made except
that B~, B~, etc., are no longer zero but are equal to the known values of
!lB~, llBP, etc., obtained for the first increment ofloading. The complete stress
I
-- p
6eij,n f- 6ep f-- ""e ~
p
6elj,n+1 1----
Differential
Equations ""ij 1-
FIGURE 9.1.1 Block diagram for computing plastic strain increments by

successive approximations.
and strain history can thus be obtained for any loading and unloading path.
If a deformation type of theory of plasticity is used, only one loading step
is required for the calculation.
The question of course which immediately arises is: What about conver-
gence? Will this method always converge and, if not, under what conditions
will it converge? A rigorous discussion of the convergence problem for the
complicated set of nonlinear equations representing the elastoplastic problem
is beyond the scope of this book. Our discussion will therefore be based on
experience, together with some qualitative observations.
From the previous description of the method and Figure 9.1.1, it is seen
that the equivalent plastic strain increment !:iep is computed from the individ-
ual plastic strains by means of the first of equations (9.1.7), and the equiva-
lent stress a. is then determined from !:ieP and the stress-strain curve. This is
illustrated by path OABC of Figure 9.1.2. On the other hand, it would seem
more direct to compute a. from the second of equations (9.1.7) and then to
determine l:ieP from a. and the stress-strain curve as shown by the flow
diagram of Figure 9.1.3 and path OCBA of Figure 9.1.2. However, experience
FIGURE 9.1.2 Determination of a. from ~e 11 •

-4.__6_e_/J_.n___,H.__o_.E_.___.H.__"'_ii_-'H'-__"'e_-'H'--6-eP_ _,H &,}, n+1 ~

FIGURE 9.1.3 Block diagram for nonconvergent scheme.
has shown that this latter scheme will generally not converge, whereas the
former method will converge provided the loading increment is made sufficiently
small. The reason for this can be seen qualitatively from Figure 9.1.2. Since
the stress-strain curve is very flat in the plastic region, it follows that a small
error in D..eP will produce a smaller error in a., but a small error in a. will
produce a much larger error in D..eP, and so the second method described will
not converge. The reason for the suggested order of the computations for
the sphere and tube of Chapter 8 [equations (8.6.8) and (8.7.18)] now becomes
apparent.
It was mentioned above that the first method described will generally
converge if the increment is made small enough. The question of what
increment size is sufficiently small for a given problem can usually be deter-
mmed only by trial and error; i.e., one picks an increment size and if the
process diverges the increment size is reduced. As a rough rule of thumb it
has been found that if D..eP is less than about 0.3 per cent, convergence will
usually take place, whereas if it is greater than 0.3 per cent, the process may
diverge. The reason for this divergence can again be seen in a qualitative
way from the Prandtl-Reuss relations. Thus consider the equations (9.1.6),
In using this equation the value of D..eP is computed from the previous values
of D..e~, D..e~, etc., using the first of equations (9.1.7). The value of a. is then
obtained from the stress-· train curve, using this computed value of D..eP, as
shown by path OABC of Figure 9.1.2. The stresses appearing in this equation
are obtained from the solution of equations (9.1.3) through (9. 1.5), using the
previous values of the plastic strain increments, as indicated by Figure 9.1.1.
If we compute a. by the second of equations (9.1.7) using these stresses, the
value of a. will not in general agree with the value of a. obtained from the
stress-strain curve, until convergence is obtained. There is thus an incon-
sistency in the Prandtl-Reuss equations (9.1.6), this inconsistency diminishing
as convergence is approached. If the increments of strain are too large, this
initial inconsistency will produce divergence.
In the second method described above and shown in Figure 9.1.3, this
type of inconsistency is avoided, since a. is computed from the calculated
stresses and D..eP is then obtained from the stress-strain curve. However, as
pointed out above, the second method diverges because of the flatness of the
stress-strain curve.
Both these difficulties can, however, be avoided by using the plastic
strain-total strain equations described in Section 7.9 and given by equations
(9.1.8) and (9.1.9). In using these equations the total strains are obtained
from the solution of equations (9.1.3) through (9:1.5) with assumed or
previously calculated approximate values of tle~, tle~, lle%, etc. The modified
or shifted total strains e~, e~, e~, etc., are then computed from the first of
equations (9.1.9) and the equivalent total strain from the last of equations
(9.1.9). The equivalent plastic strain increment is then determined from eet
and the stress-strain curve using the relation (7.9.16); i.e.,
(9.1.10)
where ae,t- 1 is the value of ae at the end of the previous increment ofloading,
and, if we are dealing with the first increment, ae,t _ 1 is equal to the yield
stress. The new values of the plastic strain increments are then computed
from equations (9.1.8), and the process is continued until convergence is
obtained. A flow diagram of the calculation scheme is shown in Figure 9.1.4.
~ toe f.
1;,n
H H H H H
D. E. G" ••
I}
e,j- let toe
p
H{. r
FIGURE 9.1.4 Block diagram for rapidly convergent scheme.
In this case, the values of e~, e~, e~, etc., appearing in equations (9.1.8) are
consistent with the value of eet appearing in these equations. Also, examination
of (9.1.1 0) shows that since the denominator of the right side is approximately
unity, a plot of !lev versus eet will have a slope of approximately unity, so that
a small error in eet will produce the same order-of-magnitude error in !lev,
without any magnification.
As a result, it has been found from experience that this last method is
convergent for even large size increments and converges more rapidly than
the first method described when that method converges. The plastic strain-
total strain method is therefore generally recommended for use wherever
possible.
We shall now present a series of examples of the use of the successive
approximation method. These examples are taken from references [4] and [5]
and also unpublished results. In all the examples of this chapter the von Mises
criterion and the Prandtl-Reuss equations will be used.
9-2 THIN FLAT PLATE
As a first example, consider the uniaxial case of a thin infinite plate of

width 2c with a temperature distribution T(y) across the width, as shown in
Figure 9.2.1. For a thin infinite strip, as shown, the only nonzero stress is
FIGURE 9.2.1 Thin infinite strip with temperature distribution across the width.
assumed to be ax(Y) and the stress-strain relation can be written
-
ex - E1 Gx + a T + exP (9.2.1)
Introduce the dimensionless quantities
aT p e~ y
T =- Ex=- 7]=- (9.2.2)
eo eo c
where a 0 is the yield stress at some reference temperatures T 0 , e0 = a 0 fE0

the yield strain at the reference temperature, and E 0 the modulus of elasticity
at the reference temperature. Note that the modulus and yield stress may be
functions of temperature. Equation (9.2.1) may now be written
Ex= H
s +T+ Ep (9.2.3)
where E~ = Ev for the uniaxial case under consideration, with Ev taken

positive for tension and negative for compression.
The assumed stress distribution, i.e., ax = ax(y), all other stresses being
zero, satisfies the equilibrium equations identically. Also, all the compatibility
equations are identically satisfied, except one, i.e.,
(9.2.4)
which gives
(9.2.5)
Sec. 9-2] Thin Flat Plate 173
Equation (9.2.5) expresses the usual assumption of simple beam theory-that

plane sections remain plane.
To determine the constants a and b, the boundary conditions are used.
If it is assumed that at the ends of the strip, an axial force P and a bending
moment M are applied, then it follows that for equilibrium
l p
J-1
SdrJ =-
aoch
=P*
(9.2.6)
where h is the thickness of the plate. Combining equations (9.2.5), (9.2.3),

and (9.2.6) enables one to obtain the constants a and b:
(9.2.7)
where
(9.2.8)
Substitution of (9.2.7) into (9.2.5) now gives
Ex= (A 1 - A 2'r)) [fl H(r + Ep)d'r) + P*]
- (A 2 - A 3 'r)) [f 1
H(r + Ep)'r) d'rj + M*] (9.2.9)
or, for the mechanical strain, defined byE = Ex - r,
E = (A 1 - A 27]) (f 1
Hr d7] + P*)
- (A 2 - A 3 7])(f 1
Hr7] d7] + M*)
+ (A 1 - A 27]) f 1
HEp d7] - (A 2 - A 3 7]) f 1
HEp'l) d7]
-T (9.2.10)
If E if a constant, independent of temperature, then H = I, A 1 = !, A 2 = 0,

Aa = t, and
+ 2I J1-1 Ep d7] + 23 7) J1
-1 Ep'l) d'l) (9.2.11)
It is also to be noted that if the temperature distribution is symmetric (an

even function of y), then J~ 1 T'l) d7J = 0. Furthermore, if there are no external
forces and moments and the temperature distribution is symmetric, equation
(9.2.11) reduces to the simple form
(9.2.12)
In equations (9.2.10) through (9.2.12) the plastic strain ratio Ep will, in

general, be a nonlinear function of the total mechanical strain ratio E. They
will be related to each other through the stress-strain curve of the material,
as shown in Figure 9.2.2. The mechanical strain e is the abscissa, the
equivalent stress ae is the ordinate, and the plastic part of the strain is eP. If the
stress-strain curve of the material is temperature dependent in the range of
temperatures being considered, a different stress-strain curve, similar to
Figure 9.2.2, must be used for each point of the plate. This does not appreci-
ably complicate the analysis.
Equations (9.2.10) through (9.2.12) therefore represent integral equations
for the solution of the elastoplastic strains during the loading of a thermally
stressed plate made of a work-hardening material with temperature-dependent
properties. For the general case of nonlinear strain hardening, these equations
are nonlinear and can be solved by the successive-approximation method
described in Section 9.1, as will be shown. If the strain hardening is linear,
these equations are linear Fredholm equations of the second kind and the
solution can often be obtained in closed form.
60 X10 3
i -- /_
c-ep
~
I
1/
0 0.004 0.008 0.012 0.016 0.020
Strain e
FIGURE 9.2.2 Typical stress-strain curve.
Nonlinear Strain Hardening

The solution for the case of nonlinear strain hardening can readily be
obtained by successive approximations as follows:
1. Assume "P to be zero everywhere.
2. Calculate E from equations (9.2.10), (9.2.11), or (9.2.12). This gives the
elastic solution.
3. For each value of " read "P from the stress-strain curve. A plot of "P
versus "' which can be obtained from the stress-strain curve, would be
convenient for this purpose. If the material properties are temperature-
dependent, a different curve is used at every station.
4. Using equation (9.2.10), (9.2.11), or (9.2.12) calculate better approxi-
mations to E. The integrals appearing in these equations are evaluated
numerically.
5. Repeat steps 3 and 4 until convergence is obtained. It will be found
that the foregoing process converges fairly rapidly, a half a dozen iterations
generally being sufficient for practical purposes.
To illustrate the procedure, consider the following example: Let the
temperature be given by
T = 600 (y 2 - j-)
E = 28 X 106
c = 1
a= 9.5 X I0- 6
a 0 = 28,000
.......
CJ')
TABLE 9.2.1 Plastic Strain Calculation for Flat Plate

e ep e ep e ep e ep e ep
n TJ ep Eq. (9.2.12) Fig. 9.2.2 Eq. (9.2.12) Fig. 9.2.2 Eq. (9.2.12) Fig. 9.2.2 Eq. (9.2.12) Fig. 9.2.2 Eq. (9.2.12) Fig. 9.2.2
X 10- 3 X 10- 3 X 1Q-3 X 10- 3 x 10- 3 X 10- 3 X 10- 3 x 1o- 3 X 10- 3 X 10- 3
1 0 0 1.9 0.7 1.7 0.5 1.6 0.4 1.6 0.4 1.6 0.4
2 0.1 0 1.8 0.6 1.6 0.4 1.5 0.4 1.5 0.4 1.5 0.4
3 0.2 0 1.7 0.5 1.5 0.4 1.4 0.3 1.4 0.3 1.4 0.3
4 0.3 0 1.4 0.3 1.2 0.2 1.1 0.1 1.1 0.1 1.1 0.1
5 0.4 0 1.0 0 0.8 0 0.7 0 0.7 0 0.7 0
6 0.5 0 0.5 0 0.3 0 0.2 0 0.2 0 0.2 0
7 0.6 0 -0.2 0 -0.4 0 -0.4 0 -0.5 0 -0.5 0
8 0.7 0 -0.9 0 -1.1 -0.1 -1.2 -0.2 -1.2 -0.2 -1.2 -0.2
9 0.8 0 -1.7 -0.5 -1.9 -0.7 -2.0 -0.8 -2.0 -0.8 -2.0 -0.8
10 0.9 0 -2.7 -1.5 -2.9 -1.6 -3.0 -1.7 -3.0 -1.7 -3.0 -1.7
11 1.0 0 -3.8 -2.5 -4.0 -2.7 -4.0 -2.7 -4.1 -2.8 -4.1 -2.8
T = 5.7(-rp - t). Stress-strain curve of Figure 9.2.2.

and let P and M be zero. Then
7" = 5.7(TJ 2 - !) (9.2.13)
Assume a stress-strain curve independent of temperature, as shown in

Figure 9.2.2. We can then use equation (9.2.12), which becomes
(9.2.14)
By using Figure 9.2.2 and the procedure outlined above, Table 9.2.1 was
constructed. Eleven equally spaced stations were used and the integral was
evaluated by the trapezoidal rule. If greater accuracy is desired, a greater
number of stations and a more accurate integration formula can be used.
However, the accuracy of Table 9.2.1 is sufficient for engineering purposes,
as can be seen in Figures 9.2.3 and 9.2.4, where the results of Table 9.2.1 as
well as the results obtained by using 51 stations and Simpson's rule are
4.0 X10- 3
--elastic
- - - f i r s t approximation
3.0 - second and third
approximations
2.0
;;;t:::-- .........
1.0
"-":::.;
~
.,~
~
0
~
~~
"'
c
0 '~)
t; -1.0
~\_
~.
-2.0
~~
\\~[\
-3.0
~~
-4.0
\
-5.0
0 0.2 0.4 0.6 0.8 1.0 ,,
Distance from plate center line
FIGURE 9.2.3 Strain distribution in infinite strip.
100 X10
--elastic
- ---first approximation
- - second and third
60
==1'---r--.,.......
approximations
20
r-- - ~
·v; '~ ~
': 0
~
b" -20 ~,,
lil
~
~ r- ,_
'\
ti 1\.
-60
\
-100
1\
-140
0 0.2 0.4 0.6 0.8 1.0
Distance from plate center line
FIGURE 9.2.4 Stress distribution in infinite strip.
plotted. Table 9 .2.1 was constructed in approximately 1- hour using a desk

calculator. Ev at each station was computed from the equation of the stress-
strain curve with Ev, of course, being set equal to zero at those stations
where \E\ ~ 1.
Linear Strain Hardening
For linear strain hardening the solution can often be obtained in closed
form. We shall consider only the simple type of temperature distribution of
the previous example. A more general formulation is given in reference [5].
Thus let T = T 0 (ry 2 - t). To will be called the loading parameter. For linear
strain hardening it can readily be shown that (assuming constant material
properties)
Ep = (1 - m)(E + 1) (9.2.15)
where m, the strain-hardening parameter, is the ratio of the tangent modulus

to the elastic modulus. The minus sign is used when E is positive (tension),
and the plus sign when E is negative (compression). Furthermore, since \"\
must be greater than 1 for plastic flow to occur, Ev = 0 if \E\ ~ 1. Substituting
into (9.2.12) gives
(9.2.16)
where
K(''), g)= 0
K(TJ, g) = 1
If 1€1 < 1 everywhere, no plastic flow will occur and the first term on the
right of (9.2.16) gives the elastic solution. The maximum value of 1€1 will
obviously occur at TJ = 1 and yielding will therefore begin at the edge,
TJ = 1, when E = -1, i.e., in compression. The value of To for yielding to
begin is then
'To,crlt = -!
If To < t, no yielding will occur. As the loading parameter To is increased
beyond t, a compressive plastic zone will spread inward from the edge. The
center of the plate, TJ = 0, will meanwhile be in a state of tension, the tensile
stress increasing as To increases. Eventually at some value of To designated
by To,c a tensile plastic zone will begin spreading from the center toward the
compressive zone spreading from the edge.
Let YJ 1 be the value of TJ at the edge of the tensile plastic zone extending
from the center, and let YJ 2 be the value of TJ at the edge of the compressive
zone spreading from the edge; i.e., for 0 < TJ < YJ 1 there is a tensile plastic
zone with € ;:::: 1, and for YJ 2 ::; TJ ::; 1 there is a compressive plastic zone with
E ::; -1. As long as To ::; '~"o,c• YJ 1 = 0, and equation (9.2.16) becomes
1
€ = -To(TJ 2 - t) + (1 - m) r (€ + 1)dg
Jn2
1
-To(TJ 2 - t) + (1- m)(1- ")2) +(1-m) r Edg
Jn2
(9.2.17)
Let us now apply the method of successive approximations to equation

(9.2.17). Thus let
€< 1 > = - To(YJ 2 - !) + (1 - m)(1 - "12)
€(
2) = - 'To(TJ 2 - t) + (1 - m)(l - ")2) + (1 - m) r
1
Jn2
€(
1) dg
- '~"o(TJ 2 - t) + (1 - m)(1 - TJ 2 ) + (1 - m)
x [;o (TJ~- ") 2) +(I - m)(1 - YJ 2) 2]
€< 3> = -To(TJ 2 - t) + (1 - m)(1 - "12) + (1 - m) 2(1 - YJ 2) 2
+ (1 - m) 3 (1 - ") 2 )3 + (1 - m) ;o (TJ~ - YJ 2 )
x [1 + (1 - m)(1 - "12)]
e< 4l = -T0 (77 2 - !) + (1 - m)(1 - 77 2) + (1 - m) 2(1 - 712) 2

+ (1 - m) 3 (1 - 712)3 + (1 - m) 4 (1 - 712) 4
+ (1 - m) ~o (71~ - 712)[1 + (1 - m)(1 - 712)
+ (1 - m)2(1 - 77 2) 2]
or
n
e<n+ 1l = -To(71 2 - j-) + (1 - m)(1 - 77 2) L (1 -
!=0
m)f(1 - 712)1
n-1
- m) ~ (71~ - 712) ~ (1 - m)f(l - 712) 1
0
+ (1
1
and
lim e<n+1) = e = -To(7l2 _ j-) + (1 _ m) 1 - 712 + (To/3)(71~ - 77 2)
n->oo 1 - (1 - m)(1 - 772)
(9.2.18)
For To > To,c• a tensile plastic zone will extend from 71 = 0 to 71 = 77 1.

Equation (9.2.16) now becomes
Or E= -To(77 2 - j-) + (1 - m)(1 - 7l2- 711) + (1 - m) (fl + {: E ag)

(9.2.19)
and a similar successive approximation as above leads to
e= -To(7l2 _ 1_) + (1 _ m) 1 - 711 - 712 + !To(71~ - 7l2 - 71~ + 771)

3 1 - (1 - m)(1 + 711 - 712)
(9.2.20)
The above derivations of equations (9.2.18) and (9.2.20) are actually not
necessary, since the integral equations (9 .2.19) and (9 .2.17) are Fredholm
equations of the second kind with degenerate kernels and the solutions can
therefore be obtained directly by standard methods (see, for example,
reference [6]). However, the successive approximation method was used in
obtaining the solution to show how this method leads to the exact solution
for this more complicated case.
Equations (9.2.18) and (9.2.20) give the complete solution to the problem
if 77 1, 77 2, and To,c were known. These can be determined as follows. If
To :::; To,c, there is one region of plastic flow extending from 77 2 to 1. At
must equal -1. Therefore, substituting 'rJ = 'r) 2, € = -1, into

'rJ = 'r) 2, €
equation (9.2.18) a relationship is obtained between T 0 , 'r) 2 , and m:
i(I - m)'TJ~ + m('TJ~ - -D = __!_

To
(9.2.21)
Thus for given values of m and To, 'TJ 2 can be determined or, more simply,
for given values of m and 'r) 2 , To can be calculated directly. For a perfectly
plastic material, m '= 0, and
'TJ2--
- ( 3 )l/3 (9.2.22)
2T 0
Once 'r) 2 is known, the complete strain distribution is obtained from (9.2.18),
the plastic strains and the stresses from (9.2.15) and (9.2.3).
If To is greater than To,c, a tensile plastic zone will spread from the center
to 'rJl· To determine the values of 'r) 1 and 'r) 2 in this case, one first sets 'rJ = 'TJ 1
and € = 1 in (9.2.20) and then 'rJ = 'r) 2 and E = -1 in (9.2.20). This results
in the following two simultaneous equations for determining 'TJ 1 and 'TJ 2 :
=
1 -To('r)t + j-) + C
(9.2.23)
-1 = -To('r)~ + j-) + C
where
C = (1 _ m) 1 - 'TJ1 - 'TJ2 + iTo('TJ~ - 'TJ2 - 'f)~ + 'r) 1)
1 - (1 - m)(1 + 'TJ1 - 'TJ2)
If the second equation is subtracted from the first, there results
2 = To('TJ~- 'r)i) (9.2.24)
The value of To,c at which the center begins to flow plastically can now be
found. For this condition, 'TJ 1 = 0 and (9.2.24) gives
'TJ2=[f (9.2.25)
To,c
Substituting this value for 'TJ 2 together with 'rJl = 0 into either of equations
(9.2.23) results in
2 2
Tg c +61 - 2m To c +9 (1 - 2m) To c - 32 (1-- -
m) = 0 (9.2.26)
• m · m • m
Only one of the roots of(9.2.26) will be physically meaningful. For a perfectly
plastic material, m = 0, and
32
To,c = 9 = 3.555 (9.2.27)
Figure 9.2.5 shows the variation of To,c with m. The value of To,c is rather
insensitive to the strain-hardening parameter m.
1'6~
0 3.4
0.
tJ>
:gc 3. 2 L,__ __j__ ___!_ __ j __ _ J ._ ___J
.3 0 0.04 0.08 0.12 0.16 0. 20

Strain-hardening parameter, m
FIGURE 9.2.5 Variation of To,c with m.
If To is greater than To,c, two regions of plastic flow will exist. The corre-
sponding values of 71 1 and 71 2 are found by solving simultaneously (9.2.24) and
one of (9.2.23). Thus eliminating 71 1 from the first of (9.2.23) by means of
(9.2.24) results in the following quintic for 71 2 :
5
712 +
( 3m
4(1 - m) +2
1-
mT0
m) 712 + 3m1 ( m + To3 ) 712
4 3
m + 3/To + 4(1 - m)] + (m + 3/T0) + 32(1 - m) /T&

2 2 =0
2
[2(1 - m) mT02 712 12m(1 - m)
(9.2.28)
A plot of 71 1 and 71 2 showing the growth of the plastic regions is shown in

Figure 9.2.6.
1.0
0.8
0.6
"7
0.4
0.2
2 4 6 8 10 20 40 60 80100
Loading parameter To
FIGURE 9.2.6 Growth of plastic region with loading parameter T 0•

Sec. 9-3] Thin Circular Shell 183
9-3 THIN CIRCULAR SHELL
The next problem we shall consider is that of a thin circular shell with an
axial temperature gradient. We shall use the deformation theory of plasticity
for this problem and assume a one-step thermal load, although this is not
essential.
The equilibrium equations for the shell are given in reference [7]. These are
Nx = 0
Rd M 2
(9.3.1)
[2 dx2 x + Ne = 0
h/2
where Nx =
J -h/2
h/2
ax dz
N8 =
J -h/2
h/2
a8 dz (9.3.2)
Mx
J
=
-h/2
axzdz
where R is the mean radius of the shell and l is the characteristic length,
equal to
(9.3.3)
h is the thickness, x is the dimensionless axial coordinate, the actual coordi-

nate divided by /, and z is the radial coordinate measured from the middle
surface, positive inward.
The stress-strain relations are
ax = - E 2 [ex - aT-
1 -tL
e~ + fL(ee - aT- e~)]
(9.3.4)
ae = - E 2 [ee - aT -
1 -tL
e~ + fL(Bx - aT - e~)]
and a2 is assumed to be zero for a thin shell.

The strain-displacement relations are
(9.3.5)
w
ee = - R
where u is the axial displacement of a point on the middle surface of the shell
and w is the radial displacement, positive inward. Substituting (9.3.5) into
(9.3.4) gives
ax= 1 -E fl-2 [ ldu

2
d w z- aT -ex+
dx- !2dx2 P
f1-
( w
-R- eeP -a T)]
(9.3.6)
2
ae = __!!_2
1 - t-t
[-~
R
- aT- e8P + rn (.E!!._-
l dx
d w2 z - aT-
!2 dx
eP)]
X
From the first of (9.3.1), using (9.3.2),
ldu w + (1
dx = t-t R + tt)aT + 7i1 J"'
-n/
2
2
(e~ + tte~)dz (9.3.7)
and from (9.3.2), making use of (9.3.7),
(9.3.8)
Also eliminating dujdx from (9.3.5) by means of (9.3.7) gives
w
ex = t-t R + (I + tt)aT + 7i1 J"' 2(e~ + ttendz -
2
-nt J21 ddx2w z

2
(9.3.9)
and, since a 2 = 0, the third stress-strain relation gives
and using (9.3.4) results in
e2 = -1-t-t -+-f1-aT - -
-f1-- ( ex+ ee) +11-t-t 1 --2t-t
- (ex+
p
e8P) (9.3.10)
1-t-t
Finally from the second of (9.3.1), making use of (9.3.8), there results
(9.3.11)
12[2 fh/2
where P = -h
3 (e~ + f.Le:)z dz
-h/2
(9.3.12)
4R fh/2
Q = -h e~ dz
-h/2
Equation (9.3.11) is recognizable as the equation for the deflections of a

beam on an elastic foundation, the right side representing the load. If P and
Q are zero, the problem is elastic. If P and Q are not zero, the problem is an
elastoplastic problem which can be solved by successive approximations.
We note first that the elastic solution can be obtained in closed form. The
homogeneous solution can be written
WH = C1 COS X COSh X + C2 COS X sinh X

+ C3 sin x cosh x + C4 sin x sinh x (9.3.13)
A particular solution can be obtained by the use of the Green's function.

Without going into the details, it can be verified by differentiation that a
particular solution is given by
Wp = -4R J: aT(g)G(x- g)dg- f 2

Peg) d Gc:x2- e) dg
-f QWG(x- e)de (9.3.14)
where G(x - e), the Green's function, is
G(x - g) = t[sin(x - g) cosh(x - g) - cos(x - g) sinh(x - e)] (9.3.15)
The complete solution is then

W = WH + Wp (9.3.16)
where the constants cl to c4 are determined from the boundary conditions.

The method of solution will be illustrated for a single step load using the
plastic strain-total strain relations; i.e.,
(9.3.17)
To facilitate the solution, the stress-strain curve can be cross-plotted to

obtain a strain-strain curve as shown in Figure 9.3.1.
28 X10- 3
24 /
....
o;"'
.~ 20 /
....~ /
"'
016 /
....0 /
+-
-::: 12 /
!!!
0
/
>
·::; 8 /
rr
w /
4
/
0 4 8 12 16 20 24X10- 3
Equivalent plastic strain Bp
FIGURE 9.3.1 Equivalent total strain versus equivalent plastic strain.
The solution now proceeds as follows. Assume as a zeroth approximation

that B~ and B~ are zero everywhere. w can then be computed from (9.3.13)
through (9.3.16), and the total strains Bx, B8 , and Bz from (9.3.5), (9.3.9), and
(9.3.10). Bet is computed from the last of (9.3.17) and Bp determined from
Figure 9.3.1. First approximations for the plastic strains are now computed
from the first two equations of (9.3.17). P and Q are then computed from
(9.3.12) and w from (9.3.13) through (9.3.16). The process is repeated until
convergence is obtained. For every iteration, new values of C1 , C2 , C3 , and
C4 must be determined from the boundary conditions. Once convergence
has been obtained, the stresses can be computed from (9.3.4).
As a specific example consider a thin circular shell with the following
geometric and physical properties:
L = 48 in. R = 12 in. h = 2 in. ""= 0.3 E = 28 x 106

a = 9.5 X 10- 6
Then from (9.3.3), l = 3.81 in. and 0 ::; x ::; 12.6. Assume a temperature
distribution given by T = 2.21x 2 (corresponding to a 350°F rise from one
end of the shell to the other end), and boundary conditions (fixed ends)
w(O) = w'(O) = w(12.6) = w'(12.6) = 0 (9.3.18)

The functions P and Q become
(e~ + 0.3ef)z dz
1
P = 21.8 J -1
(9.3.19)
Q = 24 f 1
-1
e[ dz
From the first two boundary conditions,
Also
4R J: aT(g)G(x - g)dg = 0.000252x2 + 0.000252 sin x sinh x
and the equation for w becomes
W = C2 (cos X sinh X - sin X cosh X)

+ C4 sin X sinh X (9.3.20)
- 0.000252x 2 - l 1(x) - l2(x)
where
I1(x) = LX Q(e)G(x - e)dg
l2(x) = J: Peg) d G~x;

2
g) dg
(9.3.21)
Twenty-one stations were taken along the length of the shell and six stations
through the thickness. For the zeroth approximation, it is assumed that
e~ and e: and hence 11 and / 2 are zero. The function w(x) is calculated from
(9.3.20) with the constants C2 and C4 determined from the last two boundary
conditions of equation (9.3.18). The strains ex, e0 , and e 2 are then computed
from equations (9.3.5), (9.3.9), and (9.3.10). First approximations to e~ and
e: are obtained by computing Bet from the third of equations (9. 3.17), reading
ev from Figure 9. 3.1 and calculating e~ and e[ from the first two of equations
(9.3.17). With these values of e~ and e:, P and Q are computed from equations
(9.3.19), 11 and 12 from (9.3.21), and w from (9.3.20) with new values com-
puted for c2 and c4. New values are then computed for Bx, eo, and Bz and the
process continued until convergence is obtained.
Wherever derivatives of w are needed, the following relations are useful:

dl1 = {x Q(g) dG(x - g) dg
dx Jo dx
(9.3.22)
Approximation
--
---
0
1
-- 4 through 7
40 X10 3
20 II r\
1{1
~4
0
·u;
~
a.
b" -20 "~

~
... \\
~
::l
"'
~ -40
·v;
....
::l
lf
0
b -60
U)
U)
~
+-
~ -80
.Q
><
<(
-100
-120
-140
-160
0 2 4 6 8 10 12 14
Dimensionless axial distance x
(a)
FIGURE 9.3.2 Variations of stresses and strains in shell.

r T
20 Xf0 3
0
...,_ ~ ~
·v;
a.
f.- Approximation
\
be;,
~ -20 I- --o l\
.g ---1
f.- - - 4 through 7 \ ~-
:J
V>
<ll -40 \ '\'
"0
·u;
+-
:J
\
0
b -60
~
~
t;
0 -80
~
<ll
0\
c
~-100
-120
I
-140
0 2 4 6 8 10 12 14
(b)
I I
Approximation
2.0 x10- 3 - I--
--o
---1
- - 4 through 7
.4~
"tl
.,><
I
1.0
!Jl
t 0 ~l'
.g
:J
V>
<ll
] -1.0
+-
:J
0
c
~ -2.0
·et;
]~ -30 .
-4.0
\
-5.0
0 2 4 6 8 10 12 14
Dimensionless axial· distance x
(c)
2.0 x1Q- 3
Approximation
h.
t:l
--o
I
Q)
---1
1.0
"' - - 4 through 7
"'
-,
(.)
0
't:
:::>
If) 0
Q)
'"0
·u;
+-
:::>
0 -1.0 ~\
c
0
c
Th
·e ~
t; -2.0
0
~ ~
Q)
Ol
c
-3.0
\
~
-4.0
'
0 2 4 6 8 10 12 14
(d)
FIGURE 9.3.2 (continued)
and
~~ = ! sin(x - g) sinh(x - e)
~:~ = ! [sin(x - e) cosh(x - g) - cos(x - e) sinh(x - g) (9.3.23)
d 3G
dx3 = cos(x - e) cosh(x - e)
All the integrals in the previous equations were evaluated using the trape-
zoidal rule. The results for this problem are shown in Figure 9.3.2. An
abbreviated calculation setup for one of the successive approximations at the
last two stations is given in Table 9.3.1. As can be seen from the figures, the
differences between the seventh and the fourth approximations are very small.
From an engineering viewpoint the first approximation is actually sufficient.
This is probably due to the fact that the total strains do not change much,
because of the plastic flow. This will generally be true when the only loads
are thermal loads.
TABLE 9.3.1 Calculation of Plastic Strains in Thin Shells

ePX e: /1 /2
Sta- prev. prev. prev. prev. w
tion X z aT approx. approx. approx. approx. (9.3.20)
X IQ-a XIQ-a X 10-a x to-a x to-a
20 11.97 -1.0 -3.008 0.550 -0.800 -24.t4 -581.1 -O.Ot372
-0.6667 0.480 -0.770
-0.3333 0.420 -0.740
0 0.365 -0.725
0.3333 0.3t0 -0.710
0.6667 0.260 -0.700
1.0 0.2t5 -0.695
2t t2.60 -1.0 -3.333 -3.20 -2.500 -1113 -t932 0
-0.6667 -1.57 -2.270
-0.3333 -0.175 -2.075
0 0.940 -2.030
0.3333 2.170 -2.300
0.6667 3.800 -2.750
1.0 5.500 -2.850
Q p /1 I•
Sta- Bx- aT e6 - ixT ePX e: w
tion (9.3.9) (9.3.5) (9.3.t7) (9.3.17) (9.3.t9) (9.3.2t) (9.3.20)
x to-a x to-a x to-a x to-a x to-a x to-a

20 1.314 -1.865 0.600 -0.780 -0.03295 -0.003013 -22.56 -545.6 -0.014t4
l.ltO 0.505 -0.745
0.9t00 0.420 -0.705
0.7089 0.340 -0.670
0.5100 0.270 -0.650
0.3tOO 0.200 -0.640
O.tOOO 0.135 -0.650
2t -4.414 -3.333 -3.350 -2.500 -0.1130 0.06t30 -t043 -2t95 0
-2.490 -1.625 -2.280
-0.5700 -0.175 -2.075
1.349 0.985 -2.030
3.270 2.285 -2.325
5.t90 4.000 -2.750
7.110 5.800 -2.850
Plastic Strain Charts
Instead of using equations (9 .3 .17) and the strain-strain curve of Figure

9.3.1, it is sometimes more convenient to use a plastic strain chart, as shown
in Figure 9.3.3. The ordinate and abscissa are the total mechanical strains,
and the curves shown are two intersecting families representing curves of
constant 8~ and constant 8~, respectively. Thus for a given pair of values of
8x and 8 8 , the values of 8~ and 8~ can be read directly from this chart. Inter-
polation between curves may be necessary. This procedure avoids the

l J 1 6 Jo-
l r8-
b- -l- -<.2-
~- t-8- -g- -2l- t-o-ood- 8 r-- 8- _()_
N 0 ~
0 0
0
eeP
0
0
0
0
l{),--10
0 ~ l{) l{)
4X 110-3 N
f--r- I -I 0 0 0
0 o_ 0 c:5 0
c:5 c:5
0
c:5
I I I l -,o.o~25
...- IJ- f...-- I I - -- 0.0020
2 L-- I
lL
'(.--"
l-
--
/
l
-
........-;
/
7
v I-
-r--
..__
\ r--4..
r- I I
0.0?15
J.
r--
....... Vf I -r ..-
v r--
\ ~ \
o.r101_
v/ I / ~r-....~ ~ \ \ I 0.0005
-- f.- ~
/r
J 1; ' .\ ~ - .1
I
0~00b1
0 exP
0.0001_
-I
~.0005-
-
0 1/
\ :---.. sf\. lfJ- ___..... t-
I I
I r-\ \ t" r- -~ v I /'1 I I
:..- ~.00110-
\ "\- ......... i / / f1' I l/
vI
-2 I 1-- -\._ / ....1-
.-1v
0.~01~-
I
--
I - I- l£- I--
\ I
-; 1--
t--
- f--.
1
I_ _., t--
I 1 1-'
I
f-""' 0.0:020 -
1
I
l j
l 1/ /
-4
- [...--
l-lo.oo25
I I
t--
-4 -2 2
FIGURE 9.3.3 Plastic strain chart.
necessity of computing ez and ee~o reading eP from the strain-strain curve

(Figure 9.3.1), and finally computing e~ and However, constructing such e:.
a chart involves a considerable amount of labor, and a different chart must
be constructed for each stress-strain curve used. However, if many similar
computations are to be made using the same stress-strain curve, such a chart
can be very time saving, particularly if a digital computer is not being used.
To obtain this chart, the stress-strain relations with az = 0 are written
(9.3.24)
Sec. 9-4] Long Solid Cylinder 193
Also, the Prandtl-Reuss relations solved for the stresses are
(9.3.25)
Substituting (9.3.25) into (9.3.24) gives
_ T _ (1 + 2(2 - p,) ~) P + 2(1 - 2p,) ae P

8x a - 3E 8p 8x 3E 8p 89
(9.3.26)
89
_
a
T _
-
(I + 2(23E- p,) ~) 8p
P
89
+ 2(1 - 2p,) ~
3E 8p
P
8x
With the above equations and a given stress-strain curve, a two-parameter

family of curves can be plotted giving the total strains for any pair of plastic
strains 8~ and8:. Thus
1. An arbitrary convenient value is chosen for 8:.
2. A series of values are chosen for 8~. For each of these values, (a) 8p is
computed from the first of (9.1.7) (without the ~'s), (b) a. is then read from
the stress-strain curve, and (c) 8x - aT and 8 9 - aT are computed from
(9.3.26). This gives one curve ofthefamily.
3. To obtain the other curves, the process is repeated for new values of
8:. The limiting curve of zero plastic strain is an ellipse about the origin as
shown. Any point inside this ellipse corresponds to zero plastic strain.
9-4 LONG SOLID CYLINDER
Several problems involving rotational symmetry will next be considered.

We start with the problem of a long solid circular cylinder with a radial
temperature distribution. The cylinder, having a radius R, is assumed in a
state of generalized plane strain; i.e., the axial strain 8z is assumed to be a
constant. The ends of the cylinder are assumed to be unloaded; hence the
solution will not be valid close to the ends.
The equilibrium, compatibility, and stress-strain relations for this problem
can be written
(9.4.1)
d8 9 89O - 8r
-
dr + - -
r =
(9.4.2)
ar = 11(8 - 3o:T) + 2G(er - o:T- ef)

a0 = 11(8 - 3o:T) + 2G(e0 - o:T- ef) (9.4.3)
Uz = 11(8 - 3o:T) + 2G(ez - o:T + ef + en
where
II= p,E E
(1 + p,)(1 - 2p,) G = 2(1 + p,)
To save writing, er and e: are here defined as the total accumulated plastic
strains, including those for the current increment of load; i.e.,
1-1
er = 2 b.ef.k + b.er,l
k=1
(9.4.3a)
1-1
e: = 2 b.e:,k + b.e:,l
k=1
where the summations are assumed to have already been computed and
b.e:.i and b.ef. 1 are to be determined. Assuming E, p,, and ez constant, and
substituting (9.4.3) into (9.4.1) and eliminating er by the use of (9.4.2), a
differential equation is obtained for e0 , which upon integration results in
e0 = -11-+ p, 1 ir o:Tr dr + -1-

-2
- -
2p, 21 ir err dr
p
-p,r o 1 -p,r o
(9.4.4)
+ 1 --2p,
-1 - p, r2 o o
ir
- -1 r -'--drdr
r
J' ep - e: + c1 +-
r2
C2
For a solid cylinder C 2 must vanish. Now substituting equation (9.4.4) into
the compatibility relation (9.4.2) and solving for e, gives
e, = -e0 + -11-+ p, 1 - 2p, P 1 - 2p,

- o:T + - - - e, + - - -
-p, 1 -p, 1 -p,
i' o
ef - e: dr
r
+ 2C1 (9.4.5)
To determine ez use is made of the fact that ez is a constant and that for
unloaded ends
(9.4.6)
Substituting into the third of the stress-strain relations (9.4.3) and solving
for e2 results in
ez = ; 2 [f o:Tr dr- f (er + enr dr] (9.4.7)

Sec. 9-4] Long Solid Cylinder 195
The constant C1 is determined from the boundary condition a,(R) = 0.

Substituting the first of equations (9.4.3) into this condition gives for C1 :
1 - 3p, 1 fR 1 - 2p, fR ef - eC
C1 = 1 _ f.t R 2 Jo aTr dr + 2( 1 _p,) Jo r dr
+ 2(1
1 1
- p,) R2 JorR (e, + eoP)r dr
p
(9.4.8)
To summarize, the total strains are given by
e0 = -1-+-f.t21 aTrdr
1 -p,r o
J,' + 21( 1- -p,
2p,)2
r
1 J,' (e, + e )rdr
o
p p
0
1 - 2p, f' ef - eC
+ 2(1 - p,) Jo -,.- dr + cl
(9.4.9)
-eo
1 + p, T
+ -1--
-p,
1 - 2p, p 1 - 2p,
a + -1-- e, + -1--
-p, -p,
f' o
ef - eC d
r
r + 2c 1
e2 =; 2 [LR aTr dr- J: (ef + eC)r dr]

where the double integral in (9.4.4) has been integrated by parts.
Equations (9.4.9) can now be solved by successive approximations as for
the previous examples. The zeroth approximation to the total struins is
obtained from equations (9.4.9) and (9.4.8) by assuming ilef and ileC zero.
The total equivalent strain is computed from equations (9.1.9) and the
equivalent plastic strain increment determined from the stress-strain curve
and equation (9.1.10). First approximations are then calculated for Llef and
ileC from the first two equations of (9.1.8) (with x replaced by randy by B).
TABLE 9.4.1 Calculation of Plastic Strains in Long Solid CyJinder
prev.
.p
r
·~
prev.
.,
Eq. Eq.
•e
Eq.
••
Eq. Fig.
Bet Bp eP
r
Eq.
•f
Eq.
n r aT approx. approx. (9.4.9) (9.4.9) (9.4.9) (9.1.9) 9.3.1 (9.1.8) (9.1.8)
x to-a X 10-a x to-a x to-a x to-a x to-a x to-a X 10-a

1 0 9.50 0 0 9.457 9.457 8.9t2 0.363 0 0 0
2 0.75 9.50 0 0 9.462 9.452 0.363 0 0 0
3 0.80 9.29 0 0 9.085 9.439 0.310 0 0 0
4 0.85 8.81 0 0 8.240 9.393 0.668 0 0 0
5 0.90 7.68 0 0 6.236 9.299 1.926 0.91 -0.907 0.545
6 0.95 7.06 0 0 5.274 9.t09 2.494 t.40 -t.397 0.754
7 1.000 6.83 0 0 5.045 8.911 2.578 1.48 -1.477 0.738
196 The Method o£ Successive Elastic Solutions [Ch. 9
These values of t.er and t.eC are substituted into equations (9.4.8) and (9.4.9)
and new approximations obtained for e" e8 , and e2 • The process is repeated
as many times as necessary to obtain the desired degree of convergence. After
1100
l;'- 1000
~ \.
::J
+-
2 900
\
Q)
\
a.
E
~ 800 1\
\
700
0 0.2 0.4 0.6 0.8 1.0
Radius, in.
FIGURE 9.4.1 Temperature distribution in long solid cylinder.
Approximation
0.010 0
----- 1
- - 4ond5
~ \
0.009
~
"'\; ee
1\
0.008
\
\
"'0.007
c
~
0.006
\\
0.005
\'\
\ er
\\
0.004
\
0.003
0 0.2 0.4 0.6 0.8 1.0
Radius, in.
{a)
FIGURE 9.4.2 Strain and stress distributions in long solid cylinder.
Sec. 9-5] Rotating Disk with Temperature Gradient 197
100 J1o 3 --o

---- 1
Approximation
- - 4 and 5
/
80
(
1/
60
·' (]"8-
I
I
f_,.-
20
0
I <rr
1:-=-- =--· =-== -=-= =-....::-: F---= =---= "--=1 ~~ j;'

-20
0 0.2 0.4 0.6 0.8 1.0
Radius, in.
(b)
convergence, the stresses can be computed from (9.4.3). The load is then
incremented and the process repeated.
The above calculations have been carried out for a 1-in.-radius cylinder
with a temperature distribution as shown in Figure 9.4.1, using the strain-
strain curve of Figure 9.3.1. The computations are shown in Table 9.4.1 for
one iteration and the results are plotted in Figure 9.4.2. Very little difference
is found for this problem between the deformation theory and the incre-
mental theory assuming the temperature distribution applied in several steps.
9-5 ROTATING DISK WITH TEMPERATURE

GRADIENT
The previous problem was one of plane strain. We now consider a problem
of plane stress-a rotating disk with a temperature gradient. Two methods
will be presented for solving this problem. Both use successive approxima-
tions. The first converts the problem to solving a set of integral equations in
the strains similar to the cylinder problem of the previous section. The
second method converts the differential equations to finite-difference form
and solves the resulting finite-difference equations. The integral-equation
formulation has the advantages of being conceptually simpler and of including
the boundary conditions automatically. On the other hand, it is difficult by
this approach to take into account variations of material properties with
temperature as well as variations in disk thickness. These can readily be
taken into account by the finite-difference method.
Integral-Equation Formulation
We start as usual with the equilibrium and compatibility equations. These

can be written
(9.5.1)
dee+ ee- er O (9.5.2)

dr --r- =
where h is the disk thickness, p the density, and w the rotational speed. For
the plane stress problem a 2 = 0, and the stress-strain relations can be written
[see equations (9.3.4)]
E
ar = - - -2 [e,
1 -p,
+ p,ee - (er + p,ef) - (1 + p,)aT]
(9.5.3)
ae = - E [ee
1 -p,2
+ p,e, - (e~ + p,ef) - (1 + p,)aT]
The terms er and e~ are here defined, as for the cylinder, by equations (9.4.3a).
Proceeding as for the case of the cylinder in Section 9.4, equations (9.5.3)
are substituted into (9.5.1) and e, eliminated by the use of (9.5.2), to give a
differential equation in ee. Assuming that h, E, and p, are constant, this
differential equation is integrated to give
1 - p, 2 pw 2r 2 1 + p, J' 1- p, Jr eP - e~
ee =- - E - --
8
+ --r2
-
o
aTrdr + - -
2 o
r
r
dr
1 + t-t
+-y,:z J' (e, +eer d. +y+f2
0
p P) 1 cl c2 (9 ,5,4)
TABLE 9.5.1 Calculation of Plastic Strains in Rotating Disk.
eP
r e: er ee ez eet ep e: e:
prev. prev. Eq. Eq. Eq. Eq. Fig. Eq. Eq.
n r aT approx. approx. (9.5.5) (9.5.4) (9.5.7) (9.1.9) 9.3.1 (9.1.8) (9.1.8)
x 10-s x 10- 3 x 10- 3 X 10- 3 x 10- 3 x 10- 3 x 10- 3 X 10-s x 10- 3 x w- 3
1 0 0.95 1.567 1.567 3.757 3.757 -3.247 4.669 3.441 1.720 1.720
2 0.5 1.045 1.515 1.517 3.780 3.784 -3.033 4.543 3.317 1.657 1.660
3 1.0 1.33 1.459 1.351 3.983 3.845 -2.490 4.270 3.048 1.573 1.474
4 1.5 1.71 1.376 1.144 4.250 3.942 -1.776 3.919 2.721 1.466 1.252
5 2.0 2.47 1.256 0.7269 4.862 4.104 -0.3885 3.277 2.124 1.298 0.8067
6 2.5 3.80 1.089 0.1111 6.022 4.381 1.913 2.389 1.304 1.047 0.1508
7 3.0 5.415 1.070 -0.4288 7.614 4.797 4.371 2.035 0.9972 0.9899 -0.3908
8 3.5 7.315 1.410 -1.175 9.773 5.362 6.964 2.578 1.469 1.371 -1.142
9 4.0 9.5 1.904 -2.298 12.28 6.076 10.00 3.623 2.446 1.908 -2.279
10 4.5 11.88 2.290 -3.637 14.83 6.911 13.51 4.896 3.664 2.303 -3.620
11 5.0 14.25 2.498 -4.996 17.18 7.824 17.18 6.235 4.982 2.491 -4.982
-
~
(Q
(Q
The disk is assumed to be solid; hence C2 must vanish. To obtain en equation

(9.5.4) is substituted into the compatibility equation (9.5.2), resulting in
I - fl-2 pw2r2
-eo- - E - - - +(I + p,)aT + ef + p,e~
2
i
r eP eP
+ (I - p,) r -
0
dr + C1 (9.5.5)
o r
ez is now obtained by substituting equations (9.5.3) and az = 0 into the

stress-strain relation
(9.5.6)
to give
(9.5.7)
Finally, the constant C1 is evaluated from the known rim loading. If the rim
stress due to the rim loading is ar(R), the first of equations (9.5.3) gives
(R being the radius of the disk)
ar(R) = --p,
EI2 [er + p,e 0 - (ef + p,en - (I + p,)aT]r=R (9.5.8)
and substituting equations (9.5.4) and (9.5.5) into (9.5.8) gives for C1 :
- 2 J0
I rR ep -
r r
ep
0
dr
I
+ 2R 2 J0
rR ( ef + ef)r dr
] (9.5.9)
1600
v
lL.
0
i
::J
1200
/
v
b..... 800
(l)
0. /
E
~ 400 /
0
!'-"-" - ... v 2
Radius, in.
3 4 5
FIGURE 9.5.1 Temperature distribution in rotating disk.

The solution is now obtained by successive approximations exactly as in

the previous example for a solid cylinder. By starting with assumed values
of Lle; and Lle: equal to zero, e8 , e" and e, are computed from (9.5.4), (9.5.5),
and (9.5.7), eet from the last of equations (9.1.9), Llev from the stress-strain
curve and equation (9.1.10), and Lle; and Lle: from the first two of equations
(9.1.8), with x replaced by randy by 8. New values of e" e8 , and e, are com-
puted from (9.5.4), (9.5.6), and (9.5.7), and the process is repeated.
Approximation
0
---- 1
-- 4
--- 12 with intermediate
extrapolation or 40
without extrapolation
0.018
er
0.016 I
II
0.014
//
II
0.012 VI
/
IJ
'<)
.s
~ 0.010
+-
(/) ,~
,J/ eg
0.008
) /
~f
v
/, /~
0.006
Lf1 v_ v
-~- li':P" ----- ~
~ k-;:::' ~
.- --
~
0.004
·-
0.002
0 1.0 2.0 3.0 4.0 5.0
Radius, in.
(a)
FIGURE 9.5.2 Strain and stress distributions in rotating disk.

Approximation
0
-----1
-.---4
- - - - - 1 2 with intermediate
extrapolation or 40th
without extrapolation
120 X10 3
~ ~ r-..
...............
80
40
- -= f::.::::-
\
t---
F.};::. ;-\ ,,
·~
"""'
~
k\' ~- ~ ......_......_
\~ f-:::::,_~
0
~
.~
"r
·v; -40
0.
vl
V1
\ '~
\ ""'"" "e
~-
-
~
+-
(J)
-80
\
1\
-100 \
1\
-140
\
\
-180
0 1.0 2.0 3.0 4.0 5.0
Radius, in.
(b)
As an example, the solution obtained in this manner for a 10-in.-diameter

parallel-sided disk is shown in Figure 9.5.2. The value of pw 2 was taken as
1,500 and a temperature was assumed as shown in Figure 9.5.1. The strain-
strain curve of Figure 9.3.1 was used. Again little difference was found
between deformation theory and incremental theory. The computations for
one iteration are shown in Table 9.5.1. As seen from Figure 9.5.2, a straight-
Sec. 9-5) Rotating Disk with Temperature Gradient 203
forward application of this method requires about 40 iterations for accurate

results. The reason a greater number of iterations is required here compared
to the cylinder problem of Section 9.4 is that there are external loads acting
besides the thermal loads and the strains are larger. However, the convergence
can be greatly increased by performing three or four iterations, taking the
differences between successive iterations for the various strains, and extra-
polating to a zero difference, as shown in Figure 9.5.3. A straight line is
5.2 X10- 3
0 _Br
~\
o 8e
5.0 \ \
\
4.8
\
', \
ll
\
·~ 4.6
+-
({)
' \. \
'\.
\
4.4
\ \
I~
4.2
'\
~
4.0
jq
0 0.04 0.08
Se
FIGURE 9.5.3 Variation of strain with change of strain.
drawn as shown and the intercept at zero oe is obtained. This furnishes a

new starting estimate. Three or four more successive approximations are
carried out, and another, similar, extrapolation is made. This technique
reduced the number of successive approximations for this problem from
40 to 12. However, it should be noted that if a high-speed digital computor
is used, the time per iteration is on the order of 0.1 sec and the number of
iterations required is of secondary importance.
Finite-Difference Formulation
If it is desired to take into account variations of E, p,, or h along the disk,

then it is necessary to use the finite-difference method [8]. This formulation
is very general and can take into account not only initial variations in the
disk thickness, but also the changes in thickness and radius during the
loading process, if the disk should grow and change shape. In addition, we
shall also consider the case of a disk with a central hole.
We start again with the equilibrium and compatibility equations (9.5.1)
and (9.5.2). It is convenient this time to solve the problem in terms of stresses
rather than strains, since the boundary conditions are generally given in
terms of stresses. The stress-strain relations are written as
Br = ~ (ar - p,a8) + exT+ ef + ~ef
Be = i (ae - p,ar) + exT+ e~ + ~e~ (9.5.10)
where ef and e~ represent the total plastic strains up to the current increment
of loading, and ~ef and ~e~ are the increments of plastic strain due to the
current increment of loading.
Substituting (9.5.10) into the compatibility equation (9.5.2) gives the
compatibility equation in terms of stresses:
Equations (9.5.1) and (9.5.11) are two equations for the two stresses ar and a8.
We proceed by putting them into finite-difference form as follows. Let the
disk radius be divided into N intervals (not necessarily equal). There are thus
N + 1 stations, the first station being at the center for a solid disk, or at the
inner radius for a hollow disk. The last station is at the outer radius. Equa-
tions (9.5.1) and (9.5.11) are written in finite-difference form at the midpoints
of these intervals. Thus at the midpoint of the (i - 1)st interval,
(9.5.12)
In this manner equations (9.5.1) and (9.5.11) can be written
Ciar,i - Diae,i = Fiar,i-1 + GiaB,i-1 -Hi

(9.5.13)
C[ar,i- D;ae,i = F/ ar,i-1 - G;ae,i-1 + H[ - P{
where
Fil - f-1-i-1 -
---
bi
Ei-1
2
Hi = ~ (ri - ri-1)(pihirt + Pi-1hi-1rt-1)
(9.5.14)
It is to be noted that if the disk dimensions do not change appreciably

with loading, all these coefficients except P[ depend only on the initial geom-
etry, material properties, and operating conditions of the disk and are
evaluated once and for all. Only P[ is a function of the plastic flow. Also
for a solid disk, ri_ 1 can equal zero and so some of the primed coefficients
become infinite. This can be avoided by assuming for the first radial station
some small nonzero number, rather than zero.
Equations (9.5.13) give the stresses at the ith station in terms of the stresses
at the (i - l)st station. Solving for the stresses at the ith station gives
ar,i = lll,iar,i-1 + l12,iae,i-1 + m1,i (9.5.15)

ae,i = /21,iar,i-1 + l22,iae,i-1 + m2,i
where
l _ D;Gi + DiG;
12
'i - cin; - c;ni
l _ c;Gi + CiG; (9.5.16)

22
'i - cin;- c;ni
ciP; - Hie; - H; ci
m2 ,i = cin; - c; Di
or, in matrix notation,
[::::J = [;:::: ;::::J [::::~:] + [::::]

or ai = L1a1-1 + Mi (9.5.17)
where a1, a1_1> L~o and M1 are the indicated matrices. Equation (9.5.17)
represents a linear recurrence relation.between the stresses at the ith station
and the stresses at the (i - l)st station. Obviously by successive application
of (9.5.17), the stresses at the ith station can be linearly related to the stresses
at the first station. Let this linear relation be written
(9.5.18)
where A1 and B1 are as yet unknown and a 1 are the radial and tangential
stresses at the first station. Substituting (9.5.18) into (9.5.17) gives
A1a1 + Bt = Lt(Ai-lal + B1-1) + Mt

or (AI - A1-1 L1)a1 = L1B1-1 - B1 + M1 (9.5.19)
Now a 1 will depend on the boundary conditions and is completely arbitrary,

whereas (9.5.19) must be true for all values of a 1 . It therefore follows that
both sides of the equation must vanish identically. Hence
A1 = A1-1L1
(9.5.20)
Also, for the second station, equation (9.5.18) gives
and equation (9.5.17) gives
Hence (9.5.21)
Beginning therefore with A 2 and B2 as given by (9.5.21), all the other A's
and B's can be computed successively by the recurrence relations (9.5.20).
For the last station [the (N + l)st], rN+l equals R, the disk radius, and
equation (9.5.18) becomes
or
ar,N+l] = [all,N+l a12,N+l] [ar,l] + [bl,N+l] (9.5.22)
[ae,N+l a21,N+l a22,N+l ae,l b2,N+l
Therefore,
For a solid disk ar, 1 = a 8, 1 ; therefore,
ar(R) - b1 N+l
Ur,l = ae,l = ' (9.5.23)
au,N+l + a12,N+l
For a disk with a central hole of radius R 0 , with inner prescribed pressure
we get
Ur,l = ar(Ro),
(9.5.24)
Thus a 1 is now known. The stresses at every station can then be directly
computed by means of (9.5.18).
To summarize then, the L1 and M1 matrices are computed from (9.5.16)
and (9.5.14), the A1 and B1 matrices from (9.5.20) and (9.5.21), and then the
stresses are computed from (9.5.18), using either (9.5.23) or (9.5.24).
It is to be noted that this straightforward procedure takes into account
with equal ease variations of E, h, p, or even f1- along the radius of the disk.
Furthermore, if the dimensions of the disk are changing during the plastic
flow process, this can readily be taken into account. For if r is the current
radius to a given point P, and r' was the radius to the point P before plastic
flow took place, then approximately
r = r'(1 + e&) (9.5.25)
and, similarly, if H was the underformed thickness at the point P and h is

actual thickness after deformation, then approximately
H
(9.5.26)
h = (1 + ef)(l + e&}
Thus at any stage of the plastic flow process the values of r1 and h1 appearing
in equations (9.5.14) can be corrected by means of (9.5.25) and (9.5.26).
The finite-difference formulation presented will of course give directly and
quickly the elastic solution for a disk of arbitrary profile with variable
properties, if P{ is set equal to zero in equations (9.5.16). For the plastic
problem P[ is not zero, and its values can be determined by successive
approximations, as thoroughly described in the previous examples. We shall
return to this formulation at a later time in discussing the creep of a rotating

disk. Solution by this method of the disk problem of the previous example
gave almost identical answers.
9-6 CIRCULAR HOLE IN UNIFORMLY STRESSED

INFINITE PLATE
As a final example of a problem involving rotational symmetry, consider
the case of a thin infinite plate uniformly stressed containing a circular hole of
radius a. Solutions to this problem by iterative methods similar to that dis-
cussed herein are given in references [9] and [10]. The present solution is
taken from reference [10].
The equilibrium and compatibility equations are given by (9.4.1) and
(9.4.2); i.e.,
(9.6.1)
deo eo - er O
-
dr + - -r =
The stress-strain relations for the case of plane stress are
(9.6.2)
Substituting (9.6.2) into the second of equations (9.6.1), the compatibility

equation is obtained in terms of stresses. A stress function c? is now introduced
defined by
(9.6.3)
The equilibrium equation is identically satisfied by the stress function, and

the compatibility equation becomes
£.[l d (r de?)] = - g(r) (9.6.4)

dr r dr dr
(9.6.5)
Sec. 9-6] Circular Hole in Uniformly Stressed Infinite Plate 209
Integrating results in
~ = Ar2 + Blnr + C- far-:1 fra r fra g(r)drdrdr

1
(9.6.6)
The triple integral can be somewhat simplified by making use of (9.6.5) and
integrating by parts. The constants A and B are determined from the bound-
ary conditions
(9.6.7)
where G is the applied uniform stress at infinity. Note that

00 Goo < G0 , where
Go is the yield stress. Then
A=~ (9.6.8)
2
and the constant C can arbitrarily be set equal to zero without affecting the
solution.
FIGURE 9.6.1 Tangential strain concentration factors at hole.

Detailed calculations were carried out by Tuba [10] using the successive-
approximation technique with linear strain hardening for various values of
the strain-hardening parameter m (ratio of slope of the strain-hardening part
of curve to slope of the elastic part), and for various ratios of a oo to the yield
stress a 0 • His results, taken from reference [10], are shown in Figures 9.6.1
through 9.6.3. In these figures the stress and strain concentration factors due
m=O
m=0.1
3.0
m=0.2
m =0.3
Kee m=0.4
m=0.6
m =0.8
2.0 m =1.0 (elastic)
m=0.8
m=0.6
m=0.4
Ko-e m =0.3
m=0.2
m=0.1
1.0 m=O
FIGURE 9.6.2 Equivalent stress and strain concentration factors at hole.
to the hole are plotted as functions of the dimensionless parameters m and A.

The various parameters are defined as follows:
slope of plastic part of stress-strain curve

m= --~--~----~~E=---------------
Sec. 9-6] Circular Hole in Uniformly Stressed Infinite Plate 211
0 1.0
m
FIGURE 9.6.3 Equivalent stress and strain concentration factors at hole.
K = e8(a) = e8(a)
Be e8 (oo) [(I - p..)/E]a"'
K _ ee(a) _ e.(a)
Be - e.(oo) - i[(I + p..)/E]a"'
where Be = v; V(er - e8 ) 2 + (er - e2) 2 + (e 8 - e2) 2
It is also noted in reference [10] that if the load is increased monotonically,

the deformation and incremental theories give the same results for this
problem. However, since in this case we have proportional loading at the
hole where the stress concentration factors are determined, this is not too
surprising.
In Chapter 10 the much more difficult two-dimensional plane elastoplastic
problem is formulated and several examples of plate problems ate solved.
Problems
1. Use the successive-approximation method to solve the differential equation
-dy + y
dx
= 0 y(O) = 1
Show that in the limit the exact solution is obtained.

4. Show that the method of successive approximations leads to equation
(9.2.20).
7. Obtain the equations corresponding to (9.5.4) through (9.5.9) for a hollow
disk with inner radius a and outer radius R.
8. Derive formulas (9.5.25) and (9.5.26).
9. Derive equations (9.6.4), (9.6.6), and (9.6.8).
References
1. E. L. Ince, Ordinary Differential Equations, Dover, New York, 1944.

2. A. A. Ilyushin, Some Problems in the Theory of Plastic Deformation,
RMB-12, translation from Prikl. Math. Mech., 7, 1943, pp. 245-272, by
Grad. Div. Appl. Math., Brown Univ., 1946.
3. S. S. Manson, Thermal Stress and Low Cycle Fatigue, McGraw-Hill, New
York, 1966, p. 100.
4. A. Mendelson and S. S. Manson, Practical Solution of Plastic Deformation
Problems in the Elastic-Plastic Range, NASA Tech. Rept. R-28, 1959.
5. A. Mendelson and S. W. Spero, A General Solution for the Elastoplastic
Thermal Stresses in a Strain-Hardening Plate with Arbitrary Material
Properties, J. Appl. Mech., 29, 1962, pp. 151-158.
6. S. G. Mikhlin, Integral Equations, Pergamon Press, London, 1957, p. 19.
7. C. T. Wang, Applied Elasticity, McGraw-Hill, New York, 1953.
8. A. Mendelson, M. H. Hirschberg, and S. S. Manson, A General Approach
to the Practical Solution of Creep Problems, Trans. ASME, 81D, 1959,
pp. 585-598.
9. E. A. Davis, Extension of Iteration Method for Determining Strain
Distributions to the Uniformly Stressed Plate with a Hole, J. Appl. Mech.,
30, 1963, pp. 21{}--214; discussions, ibid., 31, 1964, pp. 362-364.
10. I. S. Tuba, Elastic-Plastic Stress and Strain Concentration Factors at a
Circular Hole in a Uniformly Stressed Infinite Plate, J. Appl. Mech., 32,
1965, pp. 71{}--711.
CHAPTER 10
THE PLANE
ELASTOPLASTIC
PROBLEM
By plane elastoplastic problems we mean the usual generalized plane

strain or plane stress problems. Generalized plane strain is characterized by
the stresses and strains being functions of x and y only and the strain in the
z direction, ez, being equal to a constant. This is the type of problem en-
countered in long cylinders under certain loading conditions such as discussed
in Sections 8. 7 and 9.4. Generalized plane stress problems are encountered in
thin plates and, in this case, the stresses and strains are taken as the average
values through the thickness of the plate and uz is assumed to be zero. In
both types of problems the shear stresses and strains in the z direction are
assumed to be zero.
In all that follows, it is assumed that the material is homogeneous, isotropic,
and strain hardens isotropically. The material properties such as modulus
of elasticity, Poisson's ratio, and coefficient of linear thermal expansion are
assumed independent of the temperature, and body forces are not considered.
Rotational symmetry, as for the cylinder and disk problems of Chapter 9, is
not assumed to exist. As before, the von Mises yield criterion and the asso-
ciated flow rule will be used.
213
214 The Plane E1astop1astic Problem [Ch. 10
The equilibrium, compatibility, and stress-strain relations for the plane

stress problem are
(10. 1.1)
(10.1.2)
(10.1.3)
To these we append the Prandtl-Reuss relations,
(10.1.4)
or the corresponding plastic strain-total strain relations,
A
uBy
p
= -~Bp ( 2B y - I I
B2 -
I )
Bx
3Bet
where, as previously defined,

1-1
B~ = L ~e~,k
k=l
is the plastic strain accumulated during the first i - 1 increments of load

and ~e~ is the unknown plastic strain increment occurring during the ith or
current increment of load. The subscript i has been deleted [see equations
(7.10.2)] since no confusion can arise. Similar definitions hold for ~e~ and
e~y· The definitions of some of the other quantities entering into the above
equations will be also repeated here, for convenience. Thus
e~ = ex - B~ etc.
(10.1.6)
B~t = 3v2 [( I
Bx -
1)2
By +( I
By -
1)2
Bz
+ (BzI -
1)2
Bx
+ 6(Bxy
I )2]1/2
(a 2
ae-- X + a Y2 - a XY
a + 3r 2XY ) 1 ' 2
and ~Bp is related to a. through the uniaxial stress-strain curve as shown in

Figure 7.9.1 or to Bet by the uniaxial stress-strain curve and the relation
(10.1.7)
All these relations have been previously described in detail.

For the case of plane strain, all the above relations remain unchanged
except for equations (10.1.3) and (10.1.4) and the definition of a •. Since a, is
no longer zero,
(10.1.8)
where Bz is a constant. Also

216 The Plane Elastoplastic Problem [Ch. 10
and
(10.1.10)
We now proceed to introduce a stress function, as is common in plane

problems of elasticity. First by substituting equations (10.1.3) or (10.1.8)
into (10.1.2), the compatibility equation is obtained in terms of stresses.
Making use of the equilibrium equation (10.1.1) to eliminate the shear stress
and performing some algebraic manipulations results in
V 2 (ax + ay) = -E'i2(cxT) - [g(x, y) + Llg(x, y)] (10.1.11)
where V2 =(8 /8x'

2 2
) + (82J8y 2 ) is the Laplacian operator in two dimensions
and
E=E for plane stress
- E (10.1.12)
E=-1- for plane strain
-~.t
(10.1.13)
plane strain
(10.1.14)
) = _E_ (8 ( Lle~) + 8 ( Lle~) _ 2 8 ( Lle~y))
2 2 2
Ll (x
g ' Y - 1 - ft 2
8y 2 28x ox 8y
- ~.tE V1 2 ( Lle~ + Llet)
1 - { t2
Define a stress function cp by the relations

The equilibrium equations (10.1.1) are identically satisfied and the com-
patibility equation (10.1.11) becomes
(10.1.16)
where V' 4 is the biharmonic operator:
The boundary conditions to be satisfied by the stress function rp are (see, for
example, reference [1])
(10.1.17)
where Tx and Ty are the x and y components, respectively, of the external

force acting at a point of the boundary. The integrations are performed along
the boundary from some arbitrary point. The constants C1 and C2 are
arbitrary and can be set equal to zero without affecting the stresses. For a
multiply connected region, however, the constants will be different for each
of the contours and they can be arbitrarily chosen only on one contour. On
the other contours they must be determined so that the displacements are
single-valued. This greatly complicates even the elastic problem, and multiply
connected regions will not be considered here.
Alternatively, by means of equations (10.1.17), the boundary conditions
can be written in terms of rp and its normal derivative on the boundary [1],
rp =
J (orp
s
-
dx
-
ox ds
+ -orp -dy) ds = f(s) + Ca
oy ds
(10.1.18)
drp = orp dy _ orp dx = h(s)
dn ox ds oy ds
so that if orp(s)joy and orp(s)joy are known from (10.1.17), orpjdn and rp(s)
can be computed from (10.1.18). For an unloaded boundary, Tx = Ty = 0
and consequently j 1(s) = j 2 (s) = 0, so that orp(s)joy = C1 and orp(s)jox = C2 ,
and, since the stresses depend only on the second derivatives of rp, we can
arbitrarily subtract C1 y + C2x + Cs from f (if the region is simply con-

nected) so that
8f(s) = 8f(s) =
0
8y 8x
(10.1.19)
or df(s) = f(s) = 0
dn
To solve the plane elastoplastic problem it is therefore necessary to solve

the inhomogeneous biharmonic equation (10.1.16) subject to the boundary
conditions (10.1.17) or (1 0.1.18) and the appropriate plasticity relations.
Note that the plane strain and plane stress problems differ primarily in
the definition of the g function appearing in (1 0.1.11) and the inclusion of
az in the definition of a 8 and in the Prandtl-Reuss equations. The calculation
procedures for the two problems therefore differ in only minor details, as
will be shown subsequently.
Solutions to several specific plate problems will now be presented using
the successive-approximation technique described in Chapter 9.
10-2 ELASTOPLASTIC THERMAL PROBLEM FOR

A FINITE PLATE
The first problem considered is that of a thin rectangular plate with a

temperature distribution T(x, y) and no external loads. The geometry and
the coordinate system used is shown in Figure 10.2.1. Since there are no
r
2b '-------~x
1!-----~
~---------2o--------~~
FIGURE 10.2.1 Flat plate and coordinate system.
external loads, the boundary conditions (10.1.19) are used; i.e., f and its
normal derivative are zero on the boundaries.
Equation (10.1.16) with the right side zero is the classical biharmonic
equation. The elastic thermal stress problem with g = D..g = 0 has been
Sec. 10-2] Elastoplastic Thermal Problem for a Finite Plate 219
solved by a variety of methods, including energy, collocation, eigenfunction,

and finite-difference methods. In all cases the solution is actually only
approximate, although in theory the exact solution can be approached as
closely as desired. In solving the plastic flow problem, one or more of these
methods can also be used, together with the successive-approximation tech-
nique. The simplest and most straightforward approach is to use finite
differences, as will now be described.
The plate is divided into a grid of n x m stations. If symmetry exists about
the x and y axes with the origin taken at the center of the plate, only one
quadrant of the plate need be considered. At each point of the grid, equation
(10.1.16) is written in finite-difference form. For example, at the station
designated by i, j in Figure 10.2.2, equation (10.1.16) becomes
•
t; /+2
•
/-1,)+1
•
i,/+1
• it1,/+1
8
• • • • •
i-2,/ ;-1, j t;) i+1,j 1+2, j
•
i-1,/-1
•
i, j-1
•
it1, /-1
•
i, j-2
FIGURE 10.2.2 Finite difference net for station (i,j).
1
84 [r/>t-2,, + r/>1.1-2 + r/>1.1+2 + r/>1+2,,
+ 2(</>t-1,1-1 + <Pt-1,1+1 + 4>t+1,j-1 + 4>t+1,1+1)
+ 20r/>t,1- 8(</>t-1,1 + 4>t,1-1 + 4>t.1+1 + 4>t+1,1)]
= - E'v 2 ( aT) 1,1 - g1,1 - ll.g1,1 (10.2.1)
where o is the grid spacing.
Equations similar to (10.2.1) are written for each of then x m stations.

There then results n x m linear equations for the n x m unknown rp's,
assuming the right sides are known. These equations can now be solved by
of the numerous methods of solving large sets of simultaneous linear
algebraic equations. Once these are solved, the stresses can be computed
from the relations
O'y,!j = cfl+1,j + cfl-1,1-

<')2
2,P!,j
(10.2.2)
Txy,!j =
cfl-1,}+1 - cf!-1,1-1 -
4S2
cfi+1,J+1 + cfl+1,1-1
The strains are computed from (10.1.3) and the plastic strain increments
from either (10.1.4) or (10.1.5), together with the stress-strain curve. The
function 11g is now changed and the solution obtained again. The process is
continued until convergence is obtained. It is to be noted that only the right
sides of the set of n x m simultaneous equations change from iteration to
iteration. It should also be noted that although equation (10.2.1) has been
written for equal spacing between stations, it is possible to write the finite-
difference formulas for unequal spacing.
Once the calculation has converged for a given increment of load (in this
case, thermal load) and /1g determined, /1g is added to g, the load is incre-
mented, and a new calculation started to determine the value of the plastic
strain increments and stresses due to the new increment of load. It is to be
remembered that at any station for which ae is less than the yield stress, the
plastic strain increments are set equal to zero.
As an example, such a solution was obtained for a square plate with a
parabolic temperature distribution given by T = T0 (y 2 - f), the constant T0
being raised in increments until large zones of plastic flow occurred. Linear
strain hardening was assumed with the strain-hardening parameter m taken
to be 0.1. Some of the results are shown in Figures 10.2.3 and 10.2.4. Only
one quadrant of the plate is shown, the other three quadrants being identical
because of symmetry. In these computations 20 stations were taken in the
x direction and 20 in the y direction, resulting in a set of 400 simultaneous
equations to solve.
Figure 10.2.3 shows the rate of growth of the regions of incipient flow
The curves are the loci of all the points of incipient plastic flow for a ·
value of loading parameter To = T 0 Erxja 0 • Plastic flow starts first at
centers of the four sides of the plate and moves rapidly inward. Plastic
does not start at the center of the plate until it is well developed at the
Once plastic flow has started at the center, however, the rate of growth
this zone is greater than at the sides.
Figure 10.2.4 shows the plastic strain trajectories for the maximum
Sec. 10-2] Elastoplastic Thermal Problem for a Finite Plate 221
1.0
/ /
v //
0.9 V/
//
v ./
0.8
To ... v v/
~-
I
v / v
0.7
~;.... v 1/ /
0.6
i-
-- ~ !--"" .......-i><-15
f- 20 To v
/
vv
/ V_ v
y 0.5 2b
-
20
0.4
== =-1F-
I r-
r-..... 157¥ v 11
I-
........
1'- I I /
.'\ f/
-
~
0.3
= ~1
1\ II I
0.2 r-... \ \ I
\ I I
0.1
\ \ I ij
1\ \ I
1_ l 1 I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
X
FIGURE 10.2.3 Curves of incipient plastic flow.
1.0
r-g k-
..-
......... v _,...v v /_ ~ '/-
r- 7_ f.-
..-
v . . . v v ~ r.%'V I I I I I
0.9
J- f- "5
0.8 F= F= -4 /
~ /
6~
k-
1---::
- ~ t% ~
/
/
/': b( V' r'---0.5
"/
v v h Vj~f/
v
./
- -incipient
plastic
flow
~ -:l ~ % 1'-1
0.7 r-
f.- 1- ...... --
~ ......... )(::: v/ "'--1.5
L ~
0.6
J- f-
J- 1- - ~ ~/ "-2
Dimensionless
plastic strain,
Ep / ~
I%- !1-j
1.% ~ v
vv
y 0.5
I ''. '// '/ II"1/
t--t--, lfJ IJ J J J
0.4
r-- r-._0.5
"" \ 1.5:-' fL '/ v vv v
r- 1-- 1"'- Ill rn l_; L L /_
0.3
- I-r-.... '\ I\ 0.5_::' fL 1/ 1/ 1/ I!
0.2
!"'--
: , 2. 2\
1.5 "
~
1 \
\ \
I
II I I
1--c h( I ,12/3 ~ I I
I I ,,
0.1 I'-- I lit II I 415 /6
'i
......... A<
~~2.4
2.5
\ I I I 1/
~~ \ I \ 1 J J _j ~
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
X
FIGURE 10.2.4 Equivalent plastic strain trajectories.

-r 0 = 20. Ev is the summation of the equivalent plastic strain increment

divided by the yield strain. The curves shown are the loci of all points of
constant equivalent plastic strain. The dashed curves, labeled curves of
incipient plastic flow, which represent the elastoplastic boundary, correspond
to the curves labeled -r 0 = 20 in Figure 10.2.3.
Since both Figures 10.2.3 and 10.2.4 show only one quadrant, it is apparent
that for the entire plate there will be five regions of plastic flow. There is a
region about the center almost circular in shape and four regions identical in
shape along the four sides of the plate. There is no plastic flow at the corners
of the plate. Maximum plastic flow occurs at the centers of the four sides,
where the plastic straining is roughly three times as great as at the center of
the plate. Furthermore, the strain gradient is considerably steeper than that
near the center.
The accuracy of the method was checked by comparing the results of the
elastic solution (g = 0) with those obtained by other methods, and by com-
paring the results of the plastic solution obtained at the center section of a
3 x 1 plate with the closed-form solution for the infinite plate discussed in
Section 9.2. The agreement was excellent, as can be seen in Figure 10.2.5.
SandE are defined in Section 9.2; i.e.,
€
e -aT
= ....::X::.._--.-=,-
Uo/E
1.2 2
infinite strip
0 3x1plate
- ..............
0.8 1\ <=
·~ 1
\ t;
"'vi'
0.4 i
\
VI
VI
~ 0
.2
VI
""' '\ \
VI c
~
+-
VI 0 1\ ,g -1 1\.
I~
'()
VI
VI ~
0 1\
~
c
·~ -0.4
\
(
+-
~ -2 \
c
<))
E
0
Q.
E
1\
0 -0.8 \ .o
u -3 \
\ 0
,',!
c
\
-1.2
"-...," ~
0
-4 \
['., <))
:0::
-1.6 -5
0 0.2 0.4 0.6 0.8 i.O 0 0.2 0.4 0.6 0.8 1.0
Dimensionless coordinate, y
(a) (b)
FIGURE 10.2.5 Comparison of finite-difference elastoplastic stress-strain solutions

for 3 x 1 plate with closed-form solution for infinite strip: (a) dimensionless
stress; (b) dimensionless strain.
Sec. 10-3] Elastoplastic Problem of the Infinite Plate with a Crack 223
Since the solution obtained was an incremental one using finite increments
of load, it was thought worthwhile to run an experiment to determine the
effect of increment size f1r 0 on the final solution. It was found that f1r 0 = 2.5
was approximately the largest increment size that produced no appreciable
difference in the final stress and strain distribution at r 0 = 20. Calculations
were also performed using the two iterative techniques described in Section
9.1, that is, using equations (10.1.4) to calculate the plastic strain increments
and using equations (10.1.5). As expected for very large load increments, the
first method (using the Prandtl-Reuss relations) diverged, whereas the second
method, using the plastic strain-total strain equations, always converged.
For those cases where both methods converged they gave identical answers.
The solution to this problem can be found discussed in greater detail in
reference [2].
The above calculations were performed for a very thin plate. For a very
thick plate, a condition of plane strain would exist at planes far from the
surfaces. To obtain a solution to this problem only minor modifications of
previous calculation method are necessary. Thus g and f1g must be computed
using equations (I 0.1.14) instead of (10.1.13). Also it is now necessary to
compute a 2 • This can be done as for the case of the cylinder in Section 9.4.
From the third of equations (10.1.8),
(10.2.3)
Substituting into the relation
L a2 dA = 0 (10.2.4)
where A is the cross-sectional area of the plate, enables one to determine e2 , •

which is a constant, and hence a 2 • The rest of the computations then proceed
as for the case of plane stress. The above holds for generalized plane strain.
For plane strain, e2 = 0 and a 2 is computed directly from (10.2.3).
10-3 ELASTOPLASTIC PROBLEM OF THE

INFINITE PLATE WITH A CRACK
The elastic solution of the infinite plate with a crack was first obtained by
Inglis [3] and was used by Griffith [4] in his theory of brittle crack propagation.
It was pointed out by Orowan [5] that, for ductile materials, the plastic strain
energy was a major factor in the energy balance of the system and could not
be ignored in any analysis of crack stability. Since no solution was available
for the plastic strain field, various assumptions were introduced to take into
account the plastic zone at the tip of a crack, such as Irwin's "equivalent
crack length" [6] and Neuber's "plastic particle" [7].
In the present section it will be shown how the previous method can be
used to obtain at least approximately the elastoplastic strain field in the
vicinity of a crack, for a strain-hardening material. Since the solution can be
obtained only numerically, the presence of the mathematical singularity at
the tip of the crack makes accurate answers very difficult to obtain. The
solution to be presented is therefore not intended to provide accurate quanti-
tative results, nor is it necessarily, or even probably, the best method of
solving this problem. It is intended primarily to provide qualitative informa-
tion on the effect of strain hardening and on the differences that might be
expected between plane stress and plane strain solutions.
Consider the case of an infinite plate with a central crack 2 units long, with
a uniaxial tensile load at infinity perpendicular to the plane of the crack.
As before, all the stresses are made dimensionless by dividing by the yield
stress and all the strains are divided by the yield strain. The tensile stress at
infinity is also divided by the yield stress. Since half the crack length is taken
as unity, the x andy coordinates are dimensionless in terms of the half-crack
length. We shall attempt a solution using finite differences, as in the case of
the rectangular plate of Section 10.2.
We are faced, however, with two problems in applying the finite-difference
methods previously described. In the first place, it is obviously impossible to
cover an infinite region with a finite-difference grid. Second, the crack tip is a
singular point of the stress field and, as pointed out in reference [8], the error
in the finite-difference formulation in the vicinity of the singularity spreads
to other points. The best procedure in this type of problem is to subtract out
the singularity, if possible, giving a new problem with different boundary
conditions, but which will be well behaved.
In the present problem we attempt to minimize both the above difficulties
by subtracting out the elastic solution from the problem. The elastic solution
contains a singularity at the crack tip and also satisfies the boundary condi-
tions. We are thus left with a well-behaved problem with homogeneous
boundary conditions. Furthermore, it has been shown experimentally by
Dixon [9] that for most materials (mild steel, which has a lower yield point,
is one exception), the strain field outside the plastic zone is the same as the
elastic strain field; i.e., the elastic solution prevails outside the plastic
The assumption is therefore made that the infinite plate can be replaced by
finite rectangle with an edge crack upon whose boundaries the elastic
field acts. Because of symmetry, only one quadrant of the plate need
considered. We proceed, therefore, in the following manner.
A rectangular section is chosen as shown in Figure 10.3.1. On the upper,

left, and right boundaries of this rectangle, the elastic stress field for the
~I _l __
-1
X
FIGURE 10.3.1 Rectangular boundary for finite-difference grid.
infinite plate acts. The lower boundary is a line of symmetry. The differential
equation to be satisfied is equation (10.1.16) without the temperature term;
i.e.,
V44> = -g(x, y) - ~g(x, y) (10.3.1)
Let
(10.3.2)
where 4>e is the elastic solution to the problem. Then 4>e satisfies the differential
equation
(10.3.3)
and the boundary conditions. 4>P then satisfies
(10.3.4)
homogeneous boundary conditions. If it is assumed that the elastic

-~ ..•uv·u prevails on the upper, left, and right boundaries of the rectangle
in Figure 10.3.1, then we areleft with the problem of finding a function
</>v, satisfying equation (10.3.4), symmetric about the lower boundary and
equal to zero along the crack and on the other three boundaries of the quad-
rant. This function can be found using finite-difference and successive approxi-
mations just as was done for the thermal stress problem of Section 10.2.
The rectangle shown is covered by a grid and the differential equation
(10.3.4) is written in finite-difference form for every point of the grid. The
solution then proceeds by successive approximations as previously described.
The second iterative method employing the plastic strain-total strain equa-
tions was used. One additional step is, however, needed. Mter the function
<Pv is obtained by solving the set of simultaneous equations, the elastic stress
function, cp 8 , must be added to <l>v before computing the total stresses. The
rest of the iterative scheme proceeds exactly as before. Both the plane stress
and plane strain solutions can be readily obtained, as described in Section 10.2.
The elastic stress function </>e has been obtained in many ways. Using
Muskhelishvili's solution [10], it can be written
<Pe = ~ Re [ -zz + z2 + 2zvz 2 - 1 - 2ln (z + VZ'2=1)] (10.3.5)
where Re stands for the "real part of," z is the complex variable x + iy, z is
its conjugate, and P is the stress at infinity divided by the yield stress. For
every iteration the value of cp 8 , as computed from equation (10.3.5), is added
to the <Pv values at every station before the stresses are computed.
The first computation made was to determine the validity of the assumption
that the elastic stress field prevails outside the plastic zone. For this purpose
three different-sized rectangles were chosen to enclose the plastic zone, as
shown in Figure 10.3.2. For each of these rectangles a plane stress calculation
was made by the method described to determine the extent of the plastic zone.
Linear strain hardening was assumed with the strain-hardening parameter m
taken equal to 0.1, and the load was raised in steps in four increments from
a value of 0.2 times the yield stress to 0.5 times the yield stress. The elasto·
plastic boundaries at a load of 0.5 times the yield stress are shown in the
figure. It is seen that there is not too great a difference in the plastic zone size
for all three rectangular boundaries used, particularly the last two, indicating
that the effect of the plastic zone on the stress field outside the zone dies out
rapidly, as indicated by Dixon's experiments [9].
Also shown in Figure 10.3.2 is the elastic yield locus, i.e., the locus of
points which are just at the yield stress as computed elastically. This
commonly assumed to be the boundary of the plastic zone. It can be seen
from the figure that this assumption can be appreciably in error. This
further illustrated in Figure 10.3.3, where the elastoplastic boundary is
o small grid
t::. medium grid
o large grid
r-------------, -- - elastic yield locus
I r------------------1 !
0 I I I
II :
' II ot::. II
[p o oo ~:::.
I
0
II ;srJJJ- . . . . , I '
t \,
0 D
y
Ii
0
°;"& i I
0
I II :1~: :. I
~ tt I
I
1
00
I
'i \mokt;p / oH i '
0.8 0.9 1.0 1.1 1.2 1.3 1.4
X
FIGURE 10.3.2 Effect of grid size on plastic zone.
1.0
0 1-----'----::0.
0.8 0.9 1.1 1.2
X
Variation of plastic zone shape with strain-hardening parameter.

FIGURE 10.3.4 Growth of plastic zone size with load; plane stress.
as a function of the strain-hardening parameter m. Figure 10.3.4 shows the

growth of the plastic zone with load for a strain-hardening parameter,
m = 0.1. It is seen that the shape of the elastoplastic boundaries remain
similar to each other as the load increases, again in agreement with the
experimental results of reference [9].
All the previous results were for the case of plane stress. The plane strain
case provides no additional difficulties. One uses the definition of the functions
g and t:.g given in equations (10.1.14) and computes uz from (10.2.3). Figure
10.3.5 shows the growth of the plastic zone with load for the case of plane
strain for linear strain-hardening with m equal to 0.1. Comparison of Figures
10.3.4 and 10.3.5 shows that the areas covered by the plastic zones for plane
strain are considerably smaller than those for plane stress for the same load
ratios.
As mentioned at the beginning of this section, the above results cannot be
looked upon as quantitatively accurate without further verification. This is
due to several factors. In the first place, to ensure that the assumption that
the elastic solution prevails on the boundary of the rectangle is reasonably
correct, the rectangle size was kept at roughly twice the plastic zone size.
This meant that a majority of the grid points did not fall within the plastic
zone. To keep the number of simultaneous equations to be solved at a
reasonable value, the grid spacing was necessarily very coarse with respect
to the plastic zone size. Thus, although as many as 800 simultaneous equations
P=0.7
::;;..-_(,..
- _u_ _ _ _ ,___ _L _ _ _ j_
0.9 1.1 1.2 1.3 1.4

X
FIGURE 10.3.5 Growth of plastic zone size with load; plane strain.
were solved for one particular case, less than 300 of the grid points were
within the plastic zone. It is, of course, possible to improve this situation by
using a variable grid spacing, which was not done in the above calculations.
Second, the effect of the singularity on the results of the computation is
not at all clear. The elastic solution cf>., which is subtracted as previously
described, has a square-root stress singularity. If the plasticity solution has
the same order singularity, then the technique used in subtracting the singu-
larity would be effective. However, it is likely that the order of singularity
changes as plastic flow progresses. This has been indicated in results obtained
by Swedlow [11]. If this is true, then subtracting the elastic singularity may
not be so effective and errors due to the singularity may be propagated
through the solution. This requires further investigation.
A solution using the successive-approximation technique and finite differ-
ences was also obtained in reference [12] for a finite plate with symmetric
edge cracks. Again, because of the coarseness of the grid and the stress
singularities, the results cannot be expected to be accurate, but do give an
indication of the shape and growth of the plastic zone.
A somewhat different technique was employed in reference [11] for solving

the problem of a center-cracked symmetrically loaded plate. An equation
similar to (10.3.1) was derived in terms of the rate of change (or increment)
ofthe stress function. For the case ofp1ane stress, this equation is quasilinear;
i.e., it is linear in~ although nonlinear in cfo. Assuming the stresses and strains
are known at a given time, the increment in </>, and consequently in the stresses
and strains, can be computed using the stress-strain curve. Both finite-
difference and finite-element methods [15] were used in reference [11]. The
latter method is said to give better results. The plane strain problem, however,
cannot readily be solved by the method of reference [11].
10-4 STRAIN-INVARIANCE PRINCIPLE
A study of the previous examples, as well as other examples involving

primarily thermal loading, has led to an interesting observation of practical
importance. For thermal stress problems without additional loads, the total
strains do not change very much because of the plastic deformation compared
to the elastically computed strains. Hence an elastic computation gives
approximately the correct values of the total strains for these types of prob-
lems. This has been called the strain-invariance principle [13, 14].
To compute the plastic strains using strain invariance, the plastic strain-
total strain relations are used. If eb e 2 , and e 3 are known from the elastic
solution (principal strains are used for brevity), then eet is computed from
(10.4.1)
ev can then be obtained from the plot of ev versus eet derived from the
stress-strain curve. The plastic strains are computed from
(10.4.2)
The stresses are computed from the stress-strain relations
a1 = ,\(el + e2 + ea - 3aT) + 2Gh - aT - ei)

a2 = ,\(el + e2 + ea - 3aT) + 2G(e2 - aT- e~) (10.4.
a 3 = ,\(e 1 + e2 + e3 - 3aT) + 2G(e3 - aT- en
This method only applies for a one-step loading problem.
Sec. 10-4] Strain-Invariance Principle 231
It should be noted that the total strains, being a solution of the elastic
problem, satisfy the compatibility relations. The stresses, however, do not
satisfy equilibrium equations. If the plastic zone is relatively small compared
to the elastic zone, then, since the elastic stresses do satisfy the equilibrium
equations, equilibrium will be satisfied approximately on the whole, and the
strain-invariance answers would be expected to be reasonably correct.
However, if the plastic zone is very large, then equilibrium on the whole
will not be satisfied and the strain-invariance answers may be appreciably in
error.
The results of strain-in variance answers for the plate problems of Sections
9.2 and 10.2 are shown in Figures 10.4.1 and 10.4.2. Figure 10.4.1 shows a
---strain-hardening error
- - - - strain-invariance error
- · - error due ta neglecting
strain-hardening and
assuming strain- invariance
'
\
\
\ 0.1_.----
---
,,,........
\\
.........
.,.,.. .....
'· __ _.5::--------
....
Loading parameter r0
Per cent error in plastic strain using strain invariance; infinite

10
___ p at e
------ --
neglecting strain hardening x
3 1 1
or-----~--------~~~-=----------------------
\
'-----
""'---- --- _a'~ 3 x 1 plate
---·
neglecting strain hardening
- · - 1x1 plate
____ .........- ..---·

......... .
10 20
Loading parameter r 0
FIGURE 10.4.2 Same as Figure 10.4.1; 1 x 1 and 3 x 1 plates.
plot of the error in the maximum plastic strain assuming strain invariance,
as a function of the loading parameter r 0 , for the thin infinite strip of Section
9.2. Also shown is the error in neglecting strain hardening for a strain-
hardening parameter m = 0.1 (i.e., assuming m = 0 instead of m = 0.1).
It is seen that the error in neglecting strain hardening is on the conservative
side (gives larger plastic strains than actually exist); the error in strain
invariance is on the nonconservative side. It is interesting to note also, as is
shown in the figure, that if one makes both assumptions, i.e., neglects strain
hardening and assumes strain invariance, the errors tend to cancel each other.
In any case, from an engineering viewpoint a 10 per cent or even a 20 per cent
error is not very large, considering our inexact knowledge of material
behavior. Figure 10.4.2 shows similar results for the square plate and the
3 x 1 plate of Section 10.2.
Problems
2. Obtain the finite-difference equation corresponding to equation (10.2.1)
assuming the grid spacing in the y direction is twice that in the x direction.
3. Determine the expression for e. for the case of generalized plane strain by
means of equations (10.2.3) and (10.2.4).
4. Show that the stress function given by equation (10.3.5) satisfies the boundary
conditions for an infinite plate with a crack 2 units long and loaded at infinity
with a tensile stress P normal to the plane of the crack.
References 233
References
1. I. S. Sokolinkoff, Mathematical Theory of Elasticity, McGraw-Hill, New
York, 1956, p. 260.
2. E. Roberts, Jr., and A. Mendelson, Analysis of Plastic Thermal Stresses
and Strains in Finite Thin Plate, NASA TN D-2206, 1964.
3. C. E. Inglis, Stresses in a Plate Due to the Presence of Cracks and Sharp
Corners, Trans. Inst. Naval Arch., 55, No. 1, 1913, pp. 219-239.
4. A. A. Griffith, The Phenomena of Rupture and Flow in Solids, Phil. Trans.
Roy. Soc. London, A221, 1921, pp. 163-198.
5. E. Orowan, Fundamentals of Brittle Behavior in Metals, Fatigue and Fracture
of Metals, M.I.T. Press, Boston, Mass., and John Wiley, New York, 1952,
pp. 139-169.
6. Fracture Testing of High Strength Sheet Materials, A Report of a Special
ASTM Committee, ASTM Bull. 243, 1960.
7. H. Neuber, Theory of Notch Stresses, ABC TR-4547, 1958.
8. L. Fox, Numerical Solution of Ordinary and Partial Differential Equations,
Pergamon Press, London, 1962, p. 301.
9. J. R. Dixon, The Effect of Local Plastic Deformation on the Stress Distri-
bution Around a Crack, NEL Rpt. No. 71, Dec. 1962.
N. I. Muskhelishvili, Some Basic Problems in the Mathematical Theory
of Elasticity, P. Noordhoff, Groningen, 1953.
J. L. Swedlow, M. L. Williams, and W. H. Yang, Elasto-Plastic Stresses
and Strains in Cracked Plates, Proceedings of the International Conference on
Fracture, Sendai, Japan, Sept. 1965; also Galcit SM 65-14, July 1965.
I. S. Tuba, A Method of Elastic-Plastic Plane Stress and Strain Analysis,
J. Strain Anal., 1, No. 2, 1966, pp. 115-120.
A. Mendelson and S. S. Manson, Practical Solution of Plastic Deformation
Problems in the Elastic-Plastic Range, NASA TR R-28, 1959.
S. S. Manson, Thermal Stress and Low Cycle Fatigue, McGraw-Hill, New
York, 1966, p. 196.
M. J. Turner, R. W. Clough, H. C. Martin, and L. J. Tupp, Stiffness and
Deflection Analysis of Complex Structures, J. Aerospace, Sci., 23, No. 9,
1965, pp. 805-823.
CHAPTER 11
THE TORSION
PROBLEM
In this chapter several problems of torsion will be considered. Although

these problems are in general two-dimensional, they are usually simpler
than plane stress or plane strain problems. This is due to the fact that the
governing differential equation is of the second order rather than the fourth
order, as is the case for the plane problems. Furthermore, for bars of circular
cross section the loading is radial, so that incremental plasticity theories
need not be used at least for this case.
11-1 TORSION OF PRISMATIC BAR. GENERAL

RELATIONS
Consider a prismatic bar subject to a twisting couple as shown in Figure

11.1.1. One end of the bar is assumed fixed against rotation but not ·
warping, and at the other end a couple M with a moment along the z axis is
applied. The elastic solution for a beam of arbitrary cross section was first
obtained by Saint-Venant [1], using the semiinverse method. In this metno,a.
some simplifying assumptions concerning the stresses and/or
are first made and a solution to the given problem obtained. If this solllltlOill
satisfies the necessary equations of equilibrium and compatibility and
stress-strain relations, as well as the boundary conditions, then the sotlltlC>ll
is a correct one.
234
Sec. 11-1] Torsion of Prismatic Bar. General Relations 235
y
z
''f--'r-------
1 ,'
I ,
FIGURE 11.1.1 Prismatic bar subject to twisting couple.
For the case of the beam twisted as shown, it is assumed that the angle of
twist of a given cross section is directly proportional to its distance from the
origin. Thus if the angle of twist per unit length is a, then a section a distance
z from the origin will have rotated an amount
1{1 = az (11.1.1)
Now consider a point P originally at location (x, y, z). This point will
rotate to a position P 1 with coordinates (xi> Y1> z1 ), where
x1 = x +u JI=y+v z1 = z +w
u, v, and w being the displacements, assumed small. The projection of these
points on the xy plane are shown in Figure 11.1.2. If the polar coordinates
of the point P are r and 8, then we have
x = r cos 8 y = r sin 8
(11.1.2)
x1 = rcos(8 + 1{1) Y1 = r sin (8 + 1{1)
FIGURE 11.1.2 Projection of twisted cross section.

236 The Torsion Problem [Ch. 11
and if it is assumed that tjJ is small, so that
cos"'~ 1 sino/~ tjJ
then
u = x1 - x = - r!f sin e = -yo/

(11.1.3)
v = Y1 - y = r!f cos e = x!f
or u = -yza
v = xza (11.1.4)
and w = w(x, y, a)
It is to be noted that equations (11.1.4) are essentially assumptions, since

they stem directly from the assumption (11.1.1).
The displacement in the z direction w(x, y, a) is called the warping function.
For the elastic problem it is directly proportional to the angle of twist per
unit length a. For the plasticity problem this is in general no longer true.
Equations (11.1. 4) can be substituted into the strain-displacement relations
(4.1.12) to obtain the strains. Thus
Bxz
. 1(
= 2 - ya + 0:
8 )
(11.1.5)
Byz =
1(Xa + 8w)
2 oy
Substituting equations (11.1.5) into the compatibility equations (4.7.2), it is

seen that the first three and the last of these equations are identically satisfied
and the other two become
~ (- Oeyz + Oexz) = O
ox ox oy
_ ~ ( _ Oeyz + Oexz) = O
oy ox ay
8s os
or _J!3. - ___3.!, = constant
ox ay
From (11.1.5) it follows that the constant in (I 1.1. 7) must equal a. Therefore,
the compatibility equation for this problem becomes
0Byz 0Bxz
---=a (11.1.8)
ox By
It should be emphasized again that the shear strains eyz and exz are the
components of the strain tensor as defined by equation (4.1.12) and are there-
fore equal to one half the engineering shearing strains used by other authors.
Looking now at the equilibrium equations (3.2.2), it is seen by virtue of
(11.1.5) and the stress-strain relations that the first two are identically satisfied
and the third becomes
(11.1.9)
which is the equilibrium equation for the problem.

We now turn to the stress-strain relations. These can be written
(11.1.10)
where e~2 and e~2 are the accumulated plastic components of the total shear
strains. Substituting (11.1.10) into (11.1.8), the compatibility equation is
obtained in terms of the stresses:
OTyz
- - - - = 2Ga
OTxz + g (X, y) (11.1.11)
ox oy
X
g ( ,y
) = 2G (8e~
ay
2 _ 8efz)
ox (11.1.12)
We now introduce a stress function r/> such that
orf> - orf> (11.1.13)

Txz = oy ox
Then the equilibrium equation (11.1.9) is identically satisfied and the com-
patibility equation (11.1.11) becomes
(11.1.14)
For the elastic problem g(x, y) is equal to zero.

The boundary conditions for the problem can be obtained directly from
equations (3.3.1). The first two of these are identically satisfied and the third
reduces to
{11.1.15)
From Figure 11.1.3 it can be seen that
FIGURE 11.1.3 Geometry of boundary.
dy
I = cos (n, x) = ds
(11.1.16)
dx
m = cos (n, y) = - ds
and {11.1.15) becomes, upon combining with {11.1.13),
or 4> equals a constant along the boundary. In the case of simply connected
boundaries, e.g., solid bars, this constant can be chosen arbitrarily, since we
are interested only in the derivatives of 4>. It is, therefore, for convenience
chosen to be zero, so that the boundary condition becomes
4>=0 on the boundary

It should be noted that equation (11.1.15) is equivalent to the statement

that the resultant shear stress at the boundary is tangent to the boundary.
This must always be true if the lateral surface of the bar is force-free. It also
follows, since equation (11.1.17) and consequently equation (11.1.15) hold
for any line </> = constant in the cross section, that the resultant shear stress
at any point is tangential to the </> = constant line passing through that point.
The lines </> = constant are called the stress trajectories. Furthermore, from
(11.1.13) it follows that the resultant shear stress rat any point is given by
(11.1.19)
or r = jgrad </>I
The resultant shear stress is thus equal to the gradient of the stress function <fo.
Let us now calculate the resultant forces and moments acting on any cross
section. The force in the x direction is given by
Q., = JJr xz dx dy = JJ~t dx dy

= J[</>(x, A) - <fo(x, B)]dx =0 (11.1.20)
where the double integral is taken over the area of the cross section and
~(x, A) and <fo(x, B) are the values of</> at two opposite points of the boundary
at a given value of x, and are consequently equal to zero because of the
boundary condition (11.1.18). In a similar fashion, it follows that the y
component of the resultant stress is zero. Finally, the torque acting on the
section is computed from
M = JJ(ryzX- TxzY)dx dy
= - JJ(:~x + ~ty)dxdy (11.1.21)
Integrating by parts and making use again of the boundary condition

(11.1.18) results in
(11.1.22)
If a solution is obtained for equation (11.1.14) subject to the boundary

condition (11.1.18), then the stresses can be computed from (11.1.13) and the
strains from (11.1.1 0). (The calculation of the plastic strains will be discussed
subsequently.) The equilibrium and compatibility equations, the stress-strain
relations, and the boundary conditions will all be satisfied. The assumption
(11.1.1), or equivalently (11.1.4), thus leads to the correct solution of the
torsion problem. It should be noted, however, that the solution requires the
same stress distributions to act on every cross section, including the end
sections. However, by Saint-Venant's principle, if the beam is sufficiently
long, the solution will be valid for all cross sections far enough away from
the ends for any stress distribution acting on the ends, provided the resultant
force is zero and the resultant moment is given by (11.1.22).
Equation (11.1.14) is the well-known Poisson's equation, which is encoun-
tered so frequently in mathematical physics. Its solution can be obtained by
many different techniques. For the elasticity problem, the right side is a
constant, and solutions, at least for simple shapes, can readily be obtained.
For the plasticity problem the right side is a function of the plastic strains
and is therefore unknown until the solution is obtained. (The problem is
nonlinear.) For materials with or without strain hardening we can use the
method of successive elastic solutions described in Chapter 9. For perfectly
plastic materials, a simpler, more specialized approach, to be described sub-
sequently, can sometimes be used. In either case the elasticity solution is a
prerequisite to the plasticity solution, and we shall therefore first present some
solutions to the elasticity problem.
11-2 ELASTICITY SOLUTIONS
Equation (11.1.14), with g(x, y) equal to zero, can sometimes be solved

by guessing a solution, if the boundary of the cross section is of a simple
shape. For example, if the bar has an elliptic cross section, the equation of
the boundary curve being
then choosing for a stress function
(11.
Sec. 11-2] Elasticity Solutions 241
the boundary condition (11.1.18), <P = 0, will obviously be satisfied, and

substituting (11.2.2) into (11.1.14) results in
(11.2.3)
Hence
(11.2.4)
is the solution for the elastic case.

The torque required to produce the angle of twist per unit length a is
obtained by substituting the solution (11.2.4) into (11.1.22), giving
(11.2.5)
The constant GJ is called the torsional rigidity, or torsional stiffness, of the

bar, for obvious reasons.
To obtain the shear stresses, the solution (11.2.4) is differentiated to give
(11.2.6)
and the resultant shear stress is
(11.2.7)
The maximum stress will occur on the boundary at the point closest to the
axis of the bar at x = 0, y = b. Thus
(11.2.8)
To obtain the warping function w(x, y, a), we first compute the strains from
(11.1.10):
(11.2.9)
and then, from equation (11.1.5),
(11.2.10)
Therefore,
b2- a2
w = b2 + a2 axy + constant
and since at the origin w must vanish, the constant must equal zero, and
b2- a2
w(x, y, a) = b 2 + a2 axy (11.2.11)
The lines of constant w are therefore equilateral hyperbolas.

If the cross section is a circle, then b is equal to a in all the previous formu-
las, resulting in
Txz = -Gay (11.2.12)

Tyz = Gax
T =Gar
Tmax = Gaa
w=O
Next consider a bar whose cross section is an equilateral triangle as shown

in Figure 11.2.1. The origin of x and y axes is taken at the centroid. Then the
equation of the straight lines representing the three sides are
X=-
a
y'j
(11.2.13)
Sec. 11-2] Elasticity Solutions 243
FIGURE 11.2.1 Equilateral triangle.
If we therefore choose as a stress function
f will be zero on the boundary. Substituting into (11.1.14) [with g(x, y) = 0]

gives for C:
C=-Ga
V3
2a
Hence
f = -Ga (v. r3 x
2a
- a) ( y - - x
V3 - -
3
2a)(y + -=
x 2a)
V3 + -3 (11.2.14)
is the solution to the problem. The stresses, the torque, and the warping
function can be computed as before.
As a final example, consider a bar of rectangular cross section, as shown in
Figure 11.2.2. For this case the solution is not as simple as for the previous
y
a
X
FIGURE 11.2.2 Bar of rectangular cross section.

cases, since it is no longer possible to guess a solution. Equation (11.1.14)

must be solved by one of the available methods. For example, the method
of separation of variables is used in reference [2], resulting in an infinite-series
solution, whereas in reference [3] the Green's function is used, resulting in a
double infinite series. Here the finite-difference method will be used, since
it is relatively simple and can be directly extended to elastoplastic problems.
We note first that because of symmetry, only one quadrant of the section
need be considered. For a square cross section the diagonals are also lines
of symmetry and only one octant is used. The quadrant is divided into a
grid of n x m points as shown in Figure 11.2.3. At each of the grid points,
m
0
• • • 0
• •
• • • • • • •
• • • • • • •
(t;J)
• • • • • • • • • j
• • • • • • • • • /-1
• • • •
0
• •
• • • • • • • • •
• • • • • • • • • 3
• • • • • • • • 2
2 3 i -1 n
FIGURE 11.2.3 Finite-difference grid for rectangle.
such as the point designated by (i,j), equation (11.1.14) is written in finite-

difference form:
tPt+l,J + tPt-1,1 + tPt,J-1 + tPt,J+l - 4</>I,J = -2Gah 2 (11.2.15)
where h is the grid spacing assumed constant and is the same in both the
x andy directions.
An equation such as (11.2.15) can be written for each of the n x m grid
points, resulting in a set of n x m simultaneous linear equations for the
unknown values of 4> at each of the points. Once the fs are determined as
the solution of this set of equations, the shear stress can be obtained from
equations (11.1.13) by numerical differentiation and the torque for a given
angle of twist per unit length a from equation (11.1.22) by numerical inte-
gration.
Actually the number of equations to be solved is (m - 1) x (n - 1) rather
than m x n, since the boundary conditions require 4> to be zero at the
Sec. 11-3] Membrane Analogy 245
and right boundaries of the quadrant. For a square section, only the points
on the diagonal and to the right of the diagonal (or left) need be considered,
resulting in -!-n(n - 1) equations. Along the lower boundary, because of
symmetry, equation (11.2.15) becomes
cfot+1,1 + cfot-1,1 + 2cfot,2 - 4cfot,1 = -2Grxh 2 (11.2.16a)
and along the left boundary,
(11.2.16b)
and at the center,
(11.2.16c)
Along the diagonal of a square,
2cfot+1,1 + 2cfot,t-1 - 4cfot.t = -2Grxh2 (11.2.16d)
Such a solution was obtained for a square cross section using an 11 x 11

grid as shown in Figure 11.2.3. The 55 resulting equations were solved by a
straightforward Gauss-Seidel process (see reference [9]). The maximum shear
stress, which occurs at the centers of the sides, and the torque are compared
below with the values from reference [2]:
Finite
Ref. [2] Difference
Tmax/Grxa 1.351 1.343

M/Grxa 4 2.250 2.244
cp1,1/Grxa 2 0.589 0.589
It is seen that the solution with this many grid points is sufficiently accurate.
11-3 MEMBRANE ANALOGY
Solutions of the elastic torsion problem can also be obtained experi-

mentally by means of the membrane analogy suggested by Prandtl [4]. Con-
a membrane such as a soap film having the same shape as the cross
section of the bar being twisted. If the edges of this membrane are fixed and a
pressure is applied to one side, the membrane will deflect by an amount given
by the solution of the following equations (reference [2], p. 269):
(11.3.1)
where z is the deflection of a point of the membrane, p the applied pressure,

and S the constant tension per unit length in the membrane. Comparison of
equations (11.3.1) and (11.1.14) [with g(x, y) equal to zero for the elastic
problem] shows the analogy immediately. The deflection z of the membrane
corresponds to the stress function, and if 2Ga is equal to pjS, z is equal to<{>.
The maximum slope of the membrane at any point is proportional to the
resultant shear stress at the point and the volume under the membrane
(between the membrane and the z = 0 plane) is proportional to the torque
producing a twist of a per unit length. The membrane analogy is useful for
determining the stress function for complicated shapes and has been used
very successfully. Its extension to plastic torsion for perfectly plastic materials
will be discussed in Section 11.4.
11-4 ELASTOPLASTIC TORSION. PERFECT

PLASTICITY
If the applied torque is sufficiently large, plastic flow will occur. Since the
maximum stress will always occur at the boundary (see Problem 2 and
reference [5]), a plastic zone will start at some point on the boundary and
spread toward the interior as the torque is increased. Additional plastic
zones may subsequently start at other points in the cross section. For the
case of the torsion problem, the yield criteria of von Mises and of Tresca
both reduce to
where k is the yield stress in simple shear. According to the von Mises
criterion, k is equal to a 0 /V3, and according to the Tresca criterion it is
equal to a 0 /2, where a 0 is the yield stress in simple tension. If the material is
perfectly plastic, then equation (11.4.1) must hold everywhere in the plastic
region. In terms of the stress function <{>,
Sec. 11-4] Elastoplastic Torsion. Perfect Plasticity 247
in the plastic region, whereas in the elastic part of the bar
V2 cfo = -2Ga (11.4.3)
and cfo = 0 on the boundary of the bar. The elastoplastic boundary is unknown
and is determined from the conditions that cfo and its first derivatives (the
shear stresses) are continuous across this boundary and that the resultant
shear stress is less than or equal to k inside the elastic region. Equation
(11.4.2) can also be written
Igrad cfol = k (11.4.4)
inside the plastic region. In other words, the slope of the cfo surface is a
constant, equal to k, in the plastic region, it is not greater than kin the elastic
region, and the height and slope of the cfo surface are continuous across the
elastoplastic boundary.
The above conditions on cfo for a perfectly plastic material suggest the
extension of the membrane analogy to a partially plastic bar [6, 7]. A roof
of constant slope, proportional to k, is erected with the membrane as its base.
As the membrane is pressurized and deflects, it will approach the roof. The
region of the membrane corresponding to the region of the bar flowing
plastically will be pressed against the roof and will have the same slope as the
roof. The rest of the membrane, corresponding to the elastic region, will not
be touching the roof and will have a smaller slope.
The membrane-roof analogy furnishes a simple physical and intuitive
picture of the growth of the plastic zones as the torque is increased. To obtain
quantitative results will usually entail a considerable amount of labor. For
the case of complete plastic yielding, the solution becomes much simpler.
In this case the membrane will be in contact with the whole roof, and it is
no longer necessary to use a membrane. Instead, one constructs a roof of the
proper slope. This can be done by simply heaping dry sand onto a plate whose
shape is similar to the cross section of the bar. Since the torque is equal to
twice the volume of sand (see Problem 5), the torque required to produce
complete yielding can readily be determined. Thus for a circle of radius a,
the volume of the sand hill (in this case, a cone) is
where his the height of the heap. Since the slope of the sand hill corresponds
to the shear yield stress k,
h
k=-a
and the torque is given by

(11.4.5)
Similarly, for a rectangle with dimensions a x b [8],
(11.4.6)
Figure 11.4.1 shows the sand-hill analogies for the above two cases.
FIGURE 11.4.1 Sand-hill analogies for circular and rectangular cross sections.
11-5 ELASTOPLASTIC TORSION WITH

STRAIN HARDENING
The torsion problem for strain-hardening materials has received relatively

little attention. In this section it will be shown that the method of successive
elastic solutions or successive approximations can readily be adapted to the
torsion problem. The perfectly plastic case then becomes a simple limiting
case of the more general problem of linear strain hardening. The plastic
strain-total strain equations will be used, and for convenience the following
dimensionless quantities are introduced.
U=-r/>-
- 2Ge0 a
P-
€y =-
e~z
eo
_ '~'xz 1 + P, _ Tyz 1 + p,
Tx = 2Geo = -ao- T xz TY = 2G
eo
= --
ao
Tyz
~
~p-
= ~ M *- M
= 2Ge0 a3
eo
g = xja 'TJ = yja
Sec. 11-5] Elastoplastic Torsion with Strain Hardening 249
where e0 and a 0 are the yield strain and yield stress, respectively, related to
each other by a 0 = Ee0 and a is a characteristic linear dimension of the
cross section.
The system of equations to be solved for a simply connected cross section
can now be written
(11.5.2)
(11.5.3)
U=O on boundary
=
au (11.5.4)
Ty
- ag
(11.5.5)
2 .;-2--2
Et = VJ V Ey + Ex (11.5.6)
(11.5.7)
(11.5.8)
The relationship expressed in equation (11.5.7) is obtained from the

uniaxial stress-strain curve and, as is evident, relations in terms of plastic
strains-total strains are being used.
The successive approximation method proceeds in the usual manner. The
plastic strains are assumed to be zero everywhere. Equations (11.5.3) are
solved by any available method. The stresses, the total strains, and equivalent
total strain are computed by means of equations (11.5.4) through (11.5.7)
with the help of the stress-strain curve. If at any point in the cross section
the equivalent plastic strain as computed from (11.5.7) is negative, this point
in the elastic region, and the plastic strains at this point are set equal to
zero. Otherwise, new approximations to the plastic strains are calculated
by means of equations (11.5.8). One then returns to equation (11.5.2) and
process repeated until convergence is obtained. The method will be illus-
trated for bars of rectangular and circular cross sections. For a circular cross
with linear strain hardening the solution can be obtained in closed
11-6 BAR WITH RECTANGULAR CROSS SECTION
The elastic solution for a bar with a rectangular cross section by means of
finite differences was presented in Section 11.2. To obtain the elastoplastic
solution the function g(x, y) is subtracted from the right side of equation
(11.2.15), which in terms of the dimensionless quantities defined in (11.5.1)
becomes
where
(11.6.2)
and H is the grid spacing divided by a.

The only difference between equations (11.6.1) and (11.2.15) is in the
numbers appearing on the right side, which now depend on the plastic
strains and change from iteration to iteration. Equations (11.6.1) can there-
fore be solved in the same way as equations (11.2.15) for the elastic case.
Once the values of U are determined at all the grid points, corresponding
values of plastic strains are computed by means of equations (11.5.4) through
(11.5.8), as fully described in Section 11.5. The g 1,j are then recomputed and
equations (11.6.1) solved again, the process being repeated until convergence.
is obtained.
m
1.0
2.0
1.6
)(
0
E
...
0 2 3 4
{3
FIGURE 11.6.1 Variation of Tmax with {3.
Sec. 11-6] Bar with Rectangular Cross Section 251
The results of such a calculation for a square cross section with 11 x 11

grid points as shown in Figure 11.2.3 are shown in Figures 11.6.1 through
11.6.4. In these calculations linear strain hardening was assumed. Equation
(11.5.7), relating the equivalent plastic strain to the equivalent total strain,
can then be written (see Section 7.9)
Et - i(l + p,) (11.6.3)

Ep = 1 + i(l + p,)(m/1 - m)
2.4
X
0
n_E
"'1.6
0 2 3 4
{3
FIGURE 11.6.2 Variation of E~ax with {3.
FIGURE 11.6.3 Variation of M* with {3.

where the strain-hardening parameter m is the ratio of the slope of the linear
hardening curve to the slope of the elastic curve, as previously defined. For
the perfectly plastic case m is equal to 0, and for the elastic case m is equal to 1.
FIGURE 11.6.4 Variation of elastic-plastic boundary with (3.
TABLE 11.6.1 Summary of Results for Torsion of a Square

m (3 M* Tmax €~ax
0 2 1.786 0.751 0.820

3 1.918 0.751 1.824
4 1.955 0.751 2.851
5 1.977 0.751 3.959
0.05 2 1.813 0.785 0.758

3 1.997 0.825 1.623
4 2.094 0.862 2.434
5 2.166 0.899 3.240
6 2.228 0.934 4.003
0.10 2 1.838 0.818 0.701

3 2.073 0.893 1.478
4 2.223 0.963 2.209
5 2.347 1.032 2.919
6 2.465 1.099 3.618
0.20 2 1.890 0.881 0.600

3 2.220 1.022 1.250
4 2.471 1.156 1.870
5 2.717 1.290 2.488
6 2.966 1.426 3.116
Sec. 11-7] Bar with Circular Cross Section 253
The figures show the effects of the strain-hardening parameter and the angle
of twist on the maximum stress, the maximum plastic strain, the size of the
plastic zone, and the torque. The results are also summarized in Table 11.6.1.
Although the calculations are described using deformation theory, a
similar calculation, increasing a in steps, gave almost identical results. This
is in agreement with similar calculations in reference [11]. As shown in
reference [12], incremental and deformation theories give identical results for
a perfectly plastic material of any cross section or a strain-hardening material
of circular cross seption. For strain-hardening materials of noncircular cross
sections they will yield different results. It appears, however, that the differ-
ences will in general be slight.
11-7 BAR WITH CIRCULAR CROSS SECTION
For a bar with a circular cross section the solution is greatly simplified.
In particular, for the case of linear strain hardening, a closed-form solution
can be obtained. In polar coordinates the displacements are
Ur = 0 u8 = arz (11.7.1)
and the only nonzero strain is
(11. 7.2)
The stress-strain relation can therefore be written
(11.7.3)
The von Mises equivalent stress reduces to
(11. 7.4)
and the equivalent plastic strain is
(11.7.5)
(11.7.6)
Let
S6 r
f1e
2Geo
-
= a=p
(11.7.7)
where a is the radius of the bar. Then equation (11.7.6) can be written in
dimensionless form as
(11.7.8)
The stress-strain curve can be written in dimensionless form as
(11.7.9)
and combining with (11.7.8) results in
(11.7.10)
which can be solved iteratively for Ep.

Equation (11.7.10) is valid only in the plastic region. Let this region extend
between p = Pc and p = 1. To determine the position of the elastoplastic
boundary, i.e., p0, let eP = 0 when a 6 = a 0 or when
Hence from (11.7.8)
2(1 + p,)
Pc = V3 f3
which depends only on Poisson's ratio and the yield strain but not on
stress-strain curve. The value of f3 at which plastic flow just starts is
from equation (11.7.11) by setting Pc equal to 1. Thus the critical value
will be
f3 - 2(1 + p,)
"- v3
and the critical angle of twist per unit length will be
(11.7.13)
1 a0
or aG=--
o V3 a
To summarize, the strain-hardening solution is found as follows. The

elastoplastic boundary Po is first determined from equation (11.7.11). The
stress and strain in the elastic region for p :::; Po are then computed from
equations (11.7.2) and (11.7.3) with s:z
set equal to zero. In the plastic region
p > p0 , equation (11. 7.1 0) is solved, usually by an iterative method. s:z
can
then be computed from equation (11. 7.5) and the shear stress from (11. 7.3).
Once the shear stress is known throughout the section, the torque can be
computed by integration.
Let us now consider the case of linear strain hardening. Equation (11. 7.9)
for the stress-strain curve can be written
(11. 7.14)
Hence equatio~ (11. 7.1 0) becomes
V3 {3p - 2(1 + p,) (11.7.15)

€p = 3 + 2(1 + p,)m/(1 - m) P ~Po
Note that the critical value of p is obtained when the numerator of (11. 7.15)
vanishes, which results again in equation (11.7.11).
Once the equivalent plastic strain is known from equation (11.7.15), the
shear strain and the stress are computed from (11.7.5) and (11.7.3).
have thus obtained a complete solution in closed form. To compute the
define To = Tozf2Ge 0 • Then
M *- M s -2
= 2-G
s0a
- 7r
o
ToP 2dp il (11. 7.16)
tf3p P:::; Po
To = { 21[3P- P
€o P ~Po
results in
M* = 7rt - v'3 1r[tA(l - p~) + tB(l - p2)]
where A= v'3{3 (11.7.17)

3 + 2(1 + t-t)mf(l - m)
B = _ 2(1 + t-t)
3 + 2(1 + t-t)m/(1 - m)
Note that for Pc = 1 (no plastic flow) the torque reduces to the elastic torque
as given in equation (11.2.12). For a perfectly plastic material,
m =0 A= ~3 B = -i(l + t-t)
and M* = 27T(l + t-t) (1 - .1 3)

y' 4Pc
3 3
or
3v'3
3
M = 21ra a 0 [l _12v'3a
1 (Ga)3] 3
a0 (11.7.18)
and we recover the classical solution as given in reference [10].
20
18
16
14
12
.810
0 0.2 0.4 0.6 0.8 1.0

Pc
FIGURE 11.7.1 Variation of Pc with {3.
Results of computations using the above formulas are shown in Figures

11.7.1 through 11.7.3. Figure 11.7.1 shows the elastop1astic boundary as a
function of (3. Figure 11.7.2 shows the effect of the strain-hardening para-
meter on the shear stress for (3 = 5.0, and Figure 11. 7.3 shows the effect of
these parameters on the torque.
0 0.2 0.4 0.6 0.8

p
FIGURE 11.7.2 Variation of -r8 with p for various m: (3 = 5.0.
m
0.20
0 4 8 12 16
{3 I
.I
FIGURE 11.7.3 Variation of M* with (3 for various m. I
I I
Problems
1. Show that the loading is radial for the torsion problem of a bar with a circular
cross section, so that total plasticity theories may be used, as long as there is
no unloading.
2. Show that the maximum shear stress for a solid bar of elliptic cross section
under torsion occurs on the boundary at the point closest to the axis of the
bar.
3. Determine the stresses, the torque, and the warping function for the triangular
cross section bar of Figure 11.2.1.
4. Show that if the lateral surface of the bar is stress-free, the resultant shear
stress must be tangent to the boundary.
5. Show that if pjS is equal to 2Ga, the applied torque acting on a bar is equal
to twice the volume between a membrane of the same shape as the cross
section of the bar and the z = 0 plane.
6. Determine the torque acting on a bar of circular cross section by calculating
the volume under a membrane of the same shape.
7. Show that for the torsion problem the yield criteria of von Mises and Tresca
both reduce to equation (11.4.1).
8. Using the Saint-Venant assumptions (11.1.4), show that for a bar of circular
cross section the radial and tangential displacements become
llr =0 Uo = arz
and consequently the only nonzero strain is e92 = tra.
9. Calculate the torques required to produce a twist of 0.004 rad/in. in a 2-in.-
diameter shaft if the material is perfectly plastic, and if the material strain
hardens with m = 0.1. Assume E = 30 x 106 , p, = 0.3, and a0 = 30 x 103 •
References
1. B. Saint-Venant, Memoire sur la torsion des prismes, Mem. A cad. Sci. Math.
Phys., 14, 1856, pp. 233-560.
2. S. Timoshenko and T. N. Goodier, Theory of Elasticity, McGraw-Hill,
New York, 1951, p. 275.
3. V. Kantorovich and V. I. Krylov, Approximate Methods of Higher Analysis,
P. Noordhoff, Groningen, 1958, p. 70.
4. L. Prandtl, Zur Torsion von prismatischen Staeben, Physik. Z., 4, 1
pp. 758-759.
York, 1956, p. 117.
6. A. Nadai, Der Beginn des Fliessvorganges in einem tortierten
Angew. Math. Mech., 3, 1923, p. 442-454.
7. A. Nadai, Theory of Flow and Fracture of Solids, Vol. 1, lVHAJJ'""'".LJ""'
New York, 1950.
8. W. Johnson and P. B. Mellor, Plasticity for Mechanical Engineers,
Nostrand, Princeton, N.J., 1962, p. 132.
General References 259
9. R. S. Varga, Matrix Iterative Analysis, Prentice-Hall, Englewood Cliffs,

N.J., 1962.
10. W. Prager and P. G. Hodge, Jr., Theory of Perfectly Plastic Solids, Wiley,
New York, 1951, p. 72.
11. J. H. Huth, A Note on Plastic Torsion, J. App/. Mech., 22, 1955, pp. 432-434.
12. W. Prager, An Introduction to the Mathematical Theory of Plasticity,
J. Appl. Phys., 18, 1947, pp. 375-383.
General Reference
Johnson, W. and P. B. Mellor, Plasticity for Mechanical Engineers, Van Nostrand,
Prager, W., and P. G. Hodge, Jr., Theory of Perfectly Plastic Solids, Wiley New
York, 1951.
CHAPTER 12
THE SLIP-LINE
FIELD
12-1 PLANE STRAIN PROBLEM OF A RIGID-

PERFECTLY PLASTIC MATERIAL
In the previous chapters it was shown how the successive-approximation

method can be applied to a variety of problems, including plane strain and
plane stress elastoplastic problems. In these problems the constrain_ts imposed
by the elastic parts of the material prevented unrestrained plastic flow. In
many metal-forming processes, such as rolling, drawing, forging, etc., large
unrestricted plastic flows occur except for very small elastic zones. For such
problems it may not be unreasonable to neglect the elastic strains and assume
the material to be rigid-perfectly plastic, as shown in Figure 2.6.1(b). Any
elastic part of the body is then assumed to act as a rigid inclusion and the
plastic parts can flow freely at constant equivalent stress.
A great deal of work has been done on solutions of this type of problem
under conditions of plain strain, using the theory of slip lines. This
will be devoted to a brief discussion of this theory. We begin by writing
equations of plane strain for a rigid-plastic body. By plane strain is
the condition wherein the displacements all occur in parallel planes in
body, say, planes parallel to the xy plane, and all stresses and strains
independent of z; i.e.,
260
Sec. 12-1] Plane Strain Problem of a Rigid-Perfectly Plastic Material 261
ex= ex(X, y) ey = ey{x, y) ax= ax(X, y) (12.1.1)

ay = ay(x, y) a.,= a.,(x, y) Txy = Txy(X, y)
Since Txz = Tyz = 0, it follows that the z direction is a principal direction
and a., is a principal stress.
For a rigid-plastic material, the elastic strains are neglected, so that the
total strains and strain increments are equal to the corresponding plastic
strains and strain increments, and the Levy-Mises relations result in [see
equation (7.2.5)]
dex = i- dt\[ax - t(ay + a.,)])

dey = i- dt\[ay - !(ax + a.,)]
(12.1.2)
de., = i- dt\[a., - -!(ax + ay)]
dexy = d;\ Txy
and since de., = 0, the last equation gives

(12.1.3)
and also the mean stress is
(12.1.4)
The von Mises yield criterion for this case becomes
(12.1.5)
where k is yield stress in simple shear and T has been written for brevity
instead of T xy· The equilibrium equations to be satisfied are
(12.1.6)
Equations (12.1.6) and (12.1.5) represent three equations in the three

unknowns ax' ay, and T, If the boundary conditions are given only in terms
of stresses, these equations are sufficient to give the stress distribution without
any reference to the stress-strain relations. Such problems are called statically
We pointed out a similar situation in discussing the sphere
in Section 8.3. However, if displacements or velocities are specified
262 The Slip-Line Field [Ch. 12
over part of the boundary, then the stress-strain relations must be used to
relate the stresses to the strains and the problem becomes much more com-
plicated.
The principal stresses in the plastic field can be written in terms of am and k
as follows:
(12.1.7)
or, from (12.1.4) and (12;1.5),
(12.1.8)
The next step is to :find the prindpal directions. We define the first principal
direction as the direction of the maximum principal stress. Let cp be the angle
between the first principal direction and the x axis as shown in Figure 12.1.1.
FIGURE 12.1.1 Principal directions and ex and ,8 lines.
Then from equation (3.3.6) for the prinCipal directions it follows that
2T
tan2</> = - - -
axay
which gives two values of cp differing by 90°. The second principal direction
is taken 90° counterclockwise from the first.
Sec. 12-1] Plane Strain Problem of a Rigid-Perfectly Plastic Material 263
Having determined the principal stresses and directions, the maximum and
minimum shearing stresses and directions can readily be determined. These
shearing stresses act on the planes bisecting the principal directions as
described in Section 3.4. Their values are given by
(12.1.10)
The maximum shear directions will be designated by the a and (3 directions.

a, called the first shear direction, is taken 45° clockwise from the first principal
direction, as shown in Figure 12.1.1, and (3, the second shear direction, is 90°
counterclockwise from the first shear direction or 45° counterclockwise from
the first .principal direction, as shown.
Let (} be the angle which the first shear direction makes with the x axis
(measured counterclockwise). Then
(} = rP - 45°
1
tan 2B = -tan 2rP
and, from (12.1.9),

a -a
tan 2B = Y x (12.1.11)
27'
It follows therefore that
cos 2(} = i
(12.1.12)
. 2B- ay- ax
Sill - 2k
At every point in the plastic field, the angle which the maximum shear
direction makes with the x axis is determined by equations (12.1.11) or
(12.1.12). If curves are now drawn in the xy plane such that at every point
of each curve the tangent coincides with one of maximum shear directions,
then two families of curves called shear lines, or slip lines, will be obtained.
Obviously, since the maximum and minimum shear directions at a point are
orthogonal to each other, the two families of slip lines will form an orthogonal
set. These two families of curves will be called the a lines and (3 lines, respec-
tively.
It should be carefully noted that along an a line a is varying and (3 is
constant, and along a (3 line (3 is varying and a is constant. a and (3 are merely
or curvilinear coordinates used to designate the point under
consideration, just as x and y designate the point. Thus the point P shown in
Figure 12.1.2 can be designated P(xl> y 1 ) or P(a 3 , {3 2).
{3 lines
FIGURE 12.1.2 Families of a and (3 lines.
The normal stresses acting on the maximum shear planes equals the
average of the principal stresses, as shown by equation (3.4.5). Thus the
stresses acting normal and tangential to the a and (3 lines are given by
a" = ap = !(a1 + aa) = am

(12.1.13)
Tap = !(a1 - a 3) = k
as illustrated in Figure 12.1.3.
L---------------------~x
FIGURE 12.1.3 Stresses normal and tangential to a and (3lines.
Finally, we can express ax, ay, and -r in terms of am and 0, as follows:
ax=am-ksin2(}
ay = am + k sin 20
-r = kcos 20
All these results can also be obtained by use of Mohr's diagram.

Sec. 12-1] Plane Strain Problem. of a Rigid-Perfectly Plastic Material 265
From (12.1.14) it is seen that the state of stress can be determined in terms
of two independent quantities, am and B. The equilibrium equations can be
written in terms of these quantities by substituting (12.1.14) into (12.1.6).
Thus
Bam - 2k (cos 28 BB + sin 28 88 ) = 0

Bx Bx By
(12.1.15)
8
Bam + 2k (cos 28 BB - sin 28 B ) =0
By By Bx
or defining
we can write
8 8
Bx - cos 28 B - sin 28 B =0
ox Bx By
(12.1.16)
8
Bx - sin 28 B + cos 28 B8 = 0
By Bx By
Now the choice of the x andy axes is arbitrary. If we choose the x andy axes
at a given point to coincide with the a and f3 directions at this point, then
B = 0 and
B B
By = Bf3
and equations (12.1.16) become
(12.1.17)
Equations (12.1.17) are called compatibility equations (not to be confused

with the strain compatibifity equations). Each equation contains derivatives
in only one direction. Integrating,
along the a curve

(12.1.18)
along the f3 curve
where C1 and C2 are constants. These equations were first derived by Hencky
in 1923, [14].
From equations (12.1.18) it is apparent that if X and (J are prescribed on

the boundary, it should be possible to proceed along constant IX and fllines
to determine X = amf2k and (J, If the displacements or velocities are prescribed
over part of the boundary, as is frequently the case, these equations are not
sufficient to obtain a solution and the velocity equations following must also
be used.
12-2 VELOCITY EQUATIONS
The Levy-Mises relation (12.1.2) can also be written
dex - dey _ ax - ay
(12.2.1)
dexy - 1'xy
In addition, the incompressibility condition with de 2 = 0 is
(12.. 2.2)
It is convenient to divide the strains by dt, the increment of time, and write
these equations in terms of velocities. Of course, these equations remain
homogeneous in t, which acts merely as a scaling parameter. Then
where Vx is the velocity in the x direction; i.e., Vx = dufdt and similarly

Vy = dvfdt. Equations (12.2.1) and (12.2.2) become
(8vxf8x) - (8vyj8y) _ ax - ay
(8vxf8y) + (8vyj8x) - 27'
(12.2.3)
Now since the principal axes of stress and of plastic strain increment
coincide (see Section 7.2), it follows that the maximum shear stress lines and
maximum shear velocity lines coincide, or that the stress slip lines are the
same as the velocity slip lines. Also the strain rates normal to the IX and {3
directions are equal to the mean strain rates [see equation (4.5.4)]. Therefore,
Sec. 12-2] Velocity Equations 267
There are no extensions, only shearing flows in the slip directions-hence

their name.
Now consider the velocities in the slip directions. From Figure 12.2.1,
FIGURE 12.2.1 Velocities in a and f3 directions.
Vx = Va COS 8- Vp sin 8
(12.2.5)
Vy = Va sin 8 + Vp COS 8
If 8 = 0, the x axis will coincide with the a direction and the condition that
the normal strain rates be zero can be written
Ba = (8Vx)
ax 0=0
= 0
ep = (8Vy)
8y 0=0
= 0
or, from (12.2.5),

8v" 88
8x - Vp 8x =0
(12.2.6)
v ae + 8vp
« 8y 8y
= 0
or, since the x direction is the same as the a direction and the f3 direction is
the same as the y direction,
8v" 88
- - Vp- = 0
8a 8a
(12.2.7)
If f3 is kept constant in the first equation and a in the second equation, we

can write
dv"- v13 d8 = 0 along an a line
(12.2.8)
dv 13 + v" d8 = 0 along a f3 line
These are the compatibility equations for the velocities first derived by
Geiringer in 1930, [15].
If the problem is statically determined, the slip line field and the stresses
can be found from equations (12.1.18) (or their equivalent) and the stress
boundary conditions. The velocities can then be computed from (12.2.8) (or
their equivalent) using the velocity boundary conditions, since d8 will now
be known from the stress solution. If, however, the problem is not statically
determined, which means that the stress boundary conditions are insufficient
to obtain a unique slip-line field, then equations (12.1.18) must be solved
simultaneously with (12.2.8) using both the stress boundary conditions and
the velocity boundary conditions. This is an extremely difficult problem and
must usually be done by trial and error. A slip-line field satisfying all stress
conditions is assumed. The velocities are then computed and a check made
to see if the velocity boundary conditions are satisfied. If not, the slip-line
field is modified and the procedure repeated as often as necessary. This is
obviously a very laborious process, since the construction of just one slip-line
field is a lengthy task.
It is worthwhile to note some of the differences between Hencky's stress
equations (12.1.18) and Geiringer's velocity equations (12.2.8).
1. Hencky's equations relate two unknowns, x and 8, by two equations.
Geiringer's equations relate three unknowns, Vx, Vy, and d8, by two equations.
2. Hencky's equations give the stress state all along a known slip line if the
stress state is known at one point on the slip line. Geiringer's equations will
not give the velocities along a known slip line, if they are known at one point.
3. Hencky's equations force certain restrictions on the geometry of the
slip-line field, as we will shortly see. Geiringer's equations place no restriction
on the geometry of the slip-line field, except through the boundary conditions.
12-3 GEOMETRY OF THE SLIP-LINE FIELD
Hencky's equations, as mentioned above, impose some rather severe

restrictions on the geometry of the slip-line field, which are a great aid
computation. We will list a number of these but prove only a few to
the method of attack.
Sec. 12-3] Geometry of the Slip-Line Field 269
1. Hencky's first theorem states that the angle between two slip lines of
one family at the points where they are cut by a slip line of the other family
is constant along their lengths. This is shown in Figure 12.3.1, the angles
81 and 82 being equal.
FIGURE 12.3.1 Demonstration of Hencky's first theorem.
2. All a lines (f31ines) turn through the same angle in going from one~ line
(a line) to another.
3. If one a line (f3 line) is straight between two ~ lines (a lines), then all
a lines (f3lines) are straight between these two~ lines (a lines). Furthermore,
these straight segments have the same length.
4. If both the a and ~ lines are straight in a certain region, all the stresses
in the region are constant. This is called a field of uniform stress state.
5. Along a straight shear line the state of stress is constant.
6. If the state of stress is constant along a curve, then either the curve is
embedded in a field of constant stress or else the curve is a straight shear line.
7. The radii of curvature of the a lines (f3lines) where they intersect a given
~line (a line) decrease in direct proportion to the distance traveled in the
positive direction of the f3line (a line). Therefore, if the plastic zone extends
far enough, the radii of curvature eventually become zero, so that neighboring
slip lines run together and the solution ends at the envelope of the slip lines.
This is Hencky's second theorem.
8. As we proceed along a given slip line of one family, the centers of curva-
ture of the slip lines of the other family form an involute of this slip line.
9. The envelope of the slip lines of one family is the locus of the cusps
of the slip lines of the other family.
270 The SUp-Line Field [Ch. 12
10. The envelope of the slip lines of one family is a limiting line across
which the shear lines of the other family cannot be continued.
11. If the radius of curvature of an a line (ftline) jumps discontinuously as
it crosses a {3line (a line), all a lines ({3lines) crossing the {3line (a line) will
suffer the same jump in radius of curvature. This also means that the deriva-
tives ofthe stresses are discontinuous across the slip line.
There are many similar theorems, but they are not of practical interest.
Hencky's first theorem can easily be proved as follows. Referring to Figure
12.3.1, along the a line AD, the first of Hencky's equations (12.1.18) gives
and along the {3line CD, the second equation gives
Therefore,
(12.3.1)
Also along AB,
xA + eA = xB + eB
and along BC
Xa - Ba = XB - eB
Therefore,
(12.3.2)
or, comparing with (12.3.1),
(12.3.3)
which proves Theorem 1.

Theorem 2 is a direct corollary, since from equation (12.3.3),
Theorem 3 is also a direct corollary, since if one of the lines is straight, say
AD, then Bn - ()A = 0 and therefore 80 - (JB = 0, so the other line is also
straight.
Similarly Theorems 4 through 6 follow directly from Hencky's equations
(12.1.18). Theorems 7 through 10 are based on the theory of plane curves.
Thus let the radii of curvature of the a and {3 lines at the point A in Figure
12.3.2 be designated by Ra and Rp, respectively. At the point B the
Sec. 12-3] Geometry of the Slip-Line Field
Rrt.-6f3
Rrt. (3
--r~
~----~r-~-----
8 c
0 ----- 1
FlGURE 12.3.2 Proof of Hencky's second theorem. 1
1
of curvature R" has decreased by an amount Af3, and at the point D, Rp has
decreased by an amount Ao:, to first order of small quantities. In the limit 1
1
(12.3.5)
1
8Rp = -l
00:
1
or, more conveniently for computational purposes, since Ao: = Ro:Ae and
Af3 = -RpAO, equations (12.3.5) can be written 1
AR"- Rp MJ = 0 along a f31ine 1

(12.3.6)
ARp + R" AO = 0 along an o: line
1
Equations (12.3.5) are the mathematical statement of Theorem 7, and

1
hence Theorem 7 is proved.
The proofs of the other theorems are given in Prager and Hodge [1] and 1
will not be given here. Figure 12.3.3 illustrates some of these properties.
ABCD and A'B'C' D' are neighboring o: shear lines. AA', BB', CC', and DD' 1
are infinitesimal arcs of the f3lines. The center of curvature of these arcs form
an involute PQRS of the slip line ABCD. (An involute is the curve obtained 1
by unwinding a flexible string originally lying on the curve, so that the string
is always tangent to the curve. The original curve is the evolute of the involute 1
and is the locus of the radii of curvature of the involute.) At the point Twhere
the involute PQRST meets the slip line ABCDT, the distance between the ; ~
I
Z7Z The Slip-Line Field [Ch. lZ
FIGURE 12.3.3 Involute and limiting line. (Reference [1].)
neighboring slip lines becomes zero, and so does the radius of curvature of
the (3 line through T, as stated in Theorem 7. It is a point on the envelope of
the a lines and is a cusp of the (3 line.
12-4 SOME SIMPLE EXAMPLES
State of Uniform Stress
If the stress is constant throughout the field, the slip lines form two sets
of orthogonal straight lines. This follows directly from Hencky's equations
(12.1.18), for if X is constant, then 8 is constant. This is the converse of
Theorem 4 of Section 12.3.
Centered Fan
Consider a slip-line field composed of a set of radial lines originating from

a point and a set of concentric circular arcs as shown in Figure 12.4.1. Let
((lines
FIGURE 12.4.1 Centered fan.

Sec. 12-4] Some Simple Examples 273
the straight lines be the a lines and the circular arcs the f3 lines. Then from
Hencky's first equation, since 8 is constant along an a line, X must also be
constant along an a line, and, from the second equation, since 8 varies linearly
with distance along a f3 line, X must vary linearly with distance along a f3line.
Thus the mean stress is constant in the radial direction and varies linearly
with the angle measured from the x axis. To find the stress components we
then make use of equations (12.1.14). This type of slip-line field is called a
centered fan. Note that the center of the fan is a singular point of the stress
field, since it can have any one of an infinity of values.
Indentation by a Punch
We now consider a problem which combines, or "patches" together, the

state of uniform stress and the centered fan-the indentation of a semi-
infinite body by a flat rigid punch in the form of an infinite strip. Figure
12.4.2 shows a typical plane [2]. It is assumed that the surface of the punch
Y ---C( lines
----13 lines
' '\
'\
F '
FIGURE 12.4.2 Slip lines under lubricated flat punch.
and body are perfectly lubricated, so that there is no friction between them.
It is also assumed that there is a constant pressure over the face of the punch.
For the boundary conditions we have that over the segment AB there is a
uniform pressure kp, and the rest of the boundary is stress-free. We consider
only the case of incipient plastic flow, since once plastic flow progresses, the
shape of the boundary GABC changes considerably and it is necessary to
satisfy the boundary conditions on the deformed boundary.
Assume now that plastic flow occurs over a segment AG of the free bound-
ary as shown. The length of this segment is as yet not known. From the
boundary condition on this segment
onAG
From the yield condition,
it follows that
ax= ±2k
Intuitively we would expect ax to be compressive and we tentatively assume
onAG (12.4.1)
Since the shear stress is zero, AG is a principal direction and the slip lines
must be at ± 45° with AG. This also follows from the last of equations
(12.1.14). The a lines make 45° angles with AG and the {31ines 135° (or -45°),
as shown in Figure 12.4.2.
Consider the triangular region AGFformed by AG and the slip lines AFand
GF. By Theorem 6 of Section 12.3, this is a constant stress regipn. The slip
lines are straight lines with 8 = 71"/4. The mean stress X is a constant and must
satisfy equations (12.1.18) throughout this region. Since on the boundary AG
it follows that
X- _1.} 2
8 =i inAGF (12.4.2)
Now consider the boundary AB. Since it has been assumed that there is
no friction, r = 0 along this boundary, so that
a =
: =
-kp} alongAB
0
(12.4.3)
Therefore,
(ax~ ayr = k2
or ax= ay ± 2k = k(2- p)
where the plus sign has been chosen (see Problem 7). It follows then, just as
for the segment AG, that AB is a principal direction and the slip lines make
± 45° angles with AB. This time the {3lines are at 45° with AB and the a lines
Sec. 12-4] Some Simple Examples 275
are at 135° (or -45°). As before, region AEB is a constant stress region, the
slip lines being straight lines with 8 = ;br. From the boundary condition,
X=
Ux + Uy
4k = -2-
1- p} inAEB (12.4.5)
8= tn-
Now AF and AE are straight a slip lines and it follows from Theorem 3 of
Section 12.3 that all the shear lines in between these two are straight, or
region F AE is a centered fan. The stresses are then constant along any radial
line from A to the arc FE and vary linearly along any arc such as IJ from the
value
1-p
X= -! along AF to X= - -
2
along AE
Similar results hold in the regions GBD and BDC.

The pressure p exerted by the punch to produce this state can readily be
determined. The line AF is an a line and line HIJK is a f3 line. The com-
patibility relation to be satisfied along the f31ine (Hencky's second equation) is
X +8= constant
Along HI,
X=-! and
AlongJK,
X= !(1 - p) and
Hence
-! + i = !(1 - p) + tn-
or p=2+7T (12.4.6)
The velocity distribution is readily determined from the Geiringer equations

(12.2.8). If the punch is moving with a velocity U0 in the negative y direction,
then region ABE moves as rigid body attached to the punch with the same
velocity. In region AEFG, Va equals zero and Vp equals U0/V2. Region AEF
thus moves out with velocity U0 /V2 and region AGF moves in the direction
FG with the same velocity.
The above solution was obtained by Prandtl [3]. An alternative solution,
given by Hill [4], assumes the rigid-plastic boundary to be HIJKLMNinstead
of GFEDC. An analysis similar to the previous one shows that regions

AHI and AJK are constant state regions and AIJ is a centered fan. The
stresses in these regions are the same as previously obtained but the velocities
are different, the outward flow velocity being twice that of the previous
solution.
Actually an infinity of solutions can be obtained between the two limiting
solutions discussed above. This illustrates one of the difficulties of the plane
strain solution for a rigid-plastic material. More than one solution (or no
solution) may be obtained for a given problem, and the "correct" solution
may be impossible to ascertain. The only truly satisfactory method is to
solve the complete elastoplastic problem using the Prandtl-Reuss relations.
This, of course, will in general be extremely difficult. It is often possible,
however, to determine the most probable solution, and sometimes a minimum
force criterion may be used.
In addition, we note the nonuniqueness of the boundary values due to the
quadratic yield conditions. Thus in equation (I2.4.1) the negative sign was
chosen for ax on the basis of intuition. If the plus sign had been chosen, the
pressure exerted by the punch, equation (I2.4.6), would have come out
negative, which is impossible. So we know that the negative choice was
correct. However, the correct choice of sign was not really known a priori,
and this will often be the case.
12-5 NUMERICAL SOLUTIONS OF

BOUNDARY-VALUE PROBLEMS
In the above example of the punch indentation, the slip lines and the
solution were obtained completely in closed form from the boundary condi-
tions. In general, however, numerical or graphical methods will be necessary.
In this section a brief discussion of the simplest numerical methods will be
presented. For this purpose we must distinguish among three types of
boundaries, as shown in Figures I2.5.I through I2.5.3. Figure 12.5.1 shows
a case where the boundary curve C0 is not a slip line. The values of x and 8
are given on this boundary and it is desired to construct the slip-line field~
We choose a number of stations on the arc C0 and try to construct the
lines passing through these points. The various shear lines are uvo•~J<.~'"
1, 2, 3, etc., and a grid point at the intersection of the ith a line with the jth
{3 line is designated i,j.
Now consider the points (I, I) and (2, 2) on the curve C0 • At the
(1, 1) draw a straight line with angle 811 , representing the a line through
point. At (2, 2) draw a line with angle 822 + n/2. The two lines intersect
Sec. 12-5] Numerical Solutions of Boundary-Value Problems 277
y
Co
L-----------------------------------~x
FIGURE 12.5.1 Numerical solution of first boundary-value problem.
(1, 2), which is a rough approximation to the true intersection point of the
a and f3 lines. (If the a and f3 lines happened to be straight lines, the inter-
section point would be exact.) The points (2, 3), (3, 4), (4, 5), etc., can be
determined approximately the same way. From Hencky's equations we have
Xu - Ou = X12 - 812
(12.5.1)
X22 + 822 = X12 + 812
We can therefore solve for x12 and 012, giving us
(12.5.2)
In the same way we find
We can now proceed to find X and 0 at (1, 3), (2, 4), (3, 5), and (4, 6). We thus
obtain the slip-line field and the stresses in the entire region bounded by C0
and the terminal slip lines AP and QP. A little reflection indicates that the
solution cannot be carried beyond region APQ without some additional
information. This leads us to the following theorem. Given an arc C0 which
is not a slip line and all the stresses acting at every point along the arc, then
the complete slip-line field and the corresponding stresses can be determined
within the region bounded by C0 and the intersecting terminal slip lines as
shown in Figure 12.5.1. Region APQ is called the region of influence of the
arc Co.
If the arc C0 is itself a slip line, then the previous method is obviously
inapplicable. If a second slip line, intersecting the first one, as shown in
Figure 12.5.2, is also given, a solution can be obtained. For if 8 is known
c
L---------------------------~~x
FIGURE 12.5.2 Second boundary-value problem.
along both slip lines, then 8 can be determined at the adjoining net points by
use of Hencky's first theorem. The complete slip-line field can then be con-
structed within the quadrilateral shown in Figure 12.5.2. To determine the
stresses, it is necessary to know the value of X at just one point on the bound-
ary slip line, for by use of Hencky's equations (12.1.18) X can then be
computed throughout the region.
Alternatively, if the curve C0 is a slip line, a solution can be obtained if on a
second intersecting curve, not a slip line, either X or 8 is specified. The solution
can then be obtained in the region indicated in Figure 12.5.3, using techniques
FIGURE 12.5.3 Third boundary-value problem.

Sec. 12-6] Geometric Construction of Slip-Line Fields 279
similar to those previously outlined. Details of the solution techniques for all
three types of boundary-value problems, as well as methods for improving
the accuracy, can be found in references [1], [4], and [5].
12-6 GEOMETRIC CONSTRUCTION OF

SLIP-LINE FIELDS
A geometric construction for the stress and velocity fields, which is fre-
quently very useful and leads to a better insight into the principles underlying
slip-line theory, has been suggested by Prager [6]. For this purpose we make
use of two planes, called the stress plane and the physical plane, as shown in
Figure 12.6.1. Consider a point P undergoing plastic flow. The stress vector
y
II
~---rrx
(a)
TI (b)
FIGURE 12.6.1 Stress plane (a) and physical plane (b).
acting at the point P will depend on the orientation of the area element
through the point P upon which it acts. This is shown in the physical plane
of Figure 12.6.1(b). The figure shows the traces of several area elements whose
~normals lie in the xy plane. These area elements actually contain the point P
but are here shown separated for clarity. The shaded side of a given trace
represents material, and the stresses shown are those transmitted from the
tlnshaded side to the shaded side. Instead of identifying an area element by
.he direction of its normal, it is convenient to identify it by the direction of
the trace of the element on the xy plane. Thus angles will be measured
counterclockwise from the negative y axis instead of the x axis.
On the stress plane, Figure 12.6.1(a), a Mohr circle is plotted for the
stress state at the point P of the physical plane. The Mohr's circle is con-
structed using the following convention. A shear stress which will cause the
element to rotate in a clockwise sense is considered as positive, counter-
clockwise as negative. Thus the stress state (ax, - T xy) on plane P A is shown
as point A of the stress plane. On plane PB, whose normal stress is ay, the
shear stress is positive (clockwise) and (ay, Txy) gives the point Bin the stress
plane. We note that the angle</> between the resultant stress Sand the normal
stress Sn in the stress plane is the same as the angle between the resultant
stress and the normal stress in the physical plane, but it is measured in the
opposite direction; i.e., if <{> is measured clockwise in the stress plane it is
measured counterclockwise in the physical plane as shown.
Once the points A and B are determined, the Mohr's circle can be con-
structed in the usual way by drawing a circle through A and B whose center
lies at the intersection of AB with the Sn axis. The stress vector acting on any
plane through P can now be found from the following consideration. If a
straight line is drawn in the stress plane through any stress point (e.g., A or
B) parallel to the trace in the physical plane upon which the stress acts, it
will intersect the Mohr's circle at a point P, and all such lines will intersect
the circle at this same point. The point P is called the pole of the Mohr's
circle. In Figure 12.6.1(a) the pole is obtained by drawing a vertical line
through A to the point P or a horizontal line through B. For example, to
obtain the stress vector on plane OC rotated a degrees counterclockwise from
OA, one draws a line through P on the Mohr's circle parallel to OC or a
degrees counterclockwise from P A. The point of intersection of this line with
the Mohr's circle gives the state of stress on the plane PC.
The top and bottom points of the Mohr's circle, labeled I and II, corre-
spond to the planes upon which the maximum and minimum shearing stresses
± k act. The directions of the traces of these planes are given by the
PI and PII. These directions are the first and secop.d shear directions or
a and (3 directions, respectively, as defined in previous sections.
The geometric construction of the slip-line field can now be o
making use of the following fact, proved by Prager in reference [6]. As
move along a slip line in the physical plane, the pole of the Mohr's
traces out a cycloid in the stress plane. It does this by rolling without
along the top tangent T = k if we move along an a line and along the
tangent T = - k for a (3 line. Thus assume the stress state is known at
point Pin the physical plane of Figure 12.6.2(b). The Mohr's circle and
pole P can be constructed as shown in Figure 12.6.2(a). The directions
I I I
1st
'\ P'
\
\
(a) (b)
FIGURE 12.6.2 Cycloid trace of pole on stress plane (a) and corresponding slip
line in physical plane (b). (Reference [5].)
the first and second slip lines at Pare also known from PI and PII. If we now
move along the second shear line (13 line) to the point P', the pole P of the
Mohr's circle will move toP' as the circle rolls on the tangent lines. = -k.
The tangents to the two slip lines atP' are given by the directions P'IandP'II,
as shown by the dashed lines in Figure 12.6.2(a). Alternatively, since
P'II is normal to the cycloid at P' (the point II is the instantaneous center of
rotation), the element of the slip line at P' in the physical plane is normal to
the element of the cycloid at P' in the stress plane. The slip lines at the point P"
can be established the same way. At the same time the stresses a~ and a~ are
determined from the positions of the center of the circle. It is apparent from
the above that cycloids generated in this fashion are the images in the stress
plane of the slip lines in the physical plane.
Prager has also shown [6] how to construct the velocity field at the same
01
(a) Slip lines ( b) Hodograph
FIGURE 12.6.3 Hodograph construction of velocity field.

time as the slip lines. Let P 1 , P 2 , and P 3 be neighboring points on the first
slip line, as shown in Figure 12.6.3. The line P 1 P 2 is in the first shear direction.
Let Va and v13 be the velocity components in the two slip directions at P 1 and
let Va + dva and v13 + dv 13 be the corresponding velocities atP2 • The condition
that there be no extensions, only shearing flows in the slip directions [see
equation (12.2.4)], requires that the projections of the velocity vectors at the
points P 1 and P 2 onto the line P 1 P 2 must be equal. This yields immediately
the Geiringer equation (12.2.8):
dva- v13 dB= 0

The second equation,
dv 13 + Va dB= 0
is obtained from similar considerations along a f3line.

Consider the velocity plane, called the hodograph, of Figure 12.6.3(b). The
vectors OP 1 and OP2 represent the velocities of the points P 1 and P 2 , respec-
tively. Since, as stated above, the projections of these two vectors onto the
line P 1P 2 of Figure 12.6.3(a) must be equal, the vector difference P 1P 2 must
be orthogonal to P 1P 2 in the physical plane. In other words, line elements
in the velocity plane must be normal to the corresponding line elements of
the physical plane. In general, therefore, corresponding elements of the slip-line_
field and hodograph are orthogonal, and hence corresponding elements of the
stress plane and hodograph are parallel.
In connection with the above, it should be pointed out that the velocity
fields under consideration need not be continuous, and velocity fields with
lines of discontinuity are often encountered. Such lines of discontinuity must
be shear lines (or the envelope of the shear lines). The velocity components
normal to any slip line is continuous across the slip line, but the tangential
velocity may have different values on both sides. It follows from the Geiringer
equations that the velocity jump must be constant along any slip line. This
also follows from the fact that the slip lines are the characteristics of the
governing differential equations, as will be shown in Section 12.7. In the
hodograph, the two sides of a line of velocity discontinuity are mapped
parallel curves, as shown in reference [6].
To illustrate how one begins to construct a slip-line field for given
conditions, consider a boundary arc along which the forces are
as shown in Figure 12.6.4(a). The material adjacent to this arc is known to
in a state of plastic flow, and from the boundary stresses the maximum
directions and hence the slip-line directions can be determined. A
circle can be constructed in the stress plane for each of the points A and
of the physical plane. Here a difficulty is immediately encountered, for it
( 0) (b)
FIGURE 12.6.4 Geometric construction of slip lines.
possible to draw two circles for each of the points. For example, in Figure
12.6.5 the stress vectorS acting at a point P has been drawn from the origin
to the point A of the stress plane. Through the point A, however, two circles
can be drawn, each with radius k and with their centers on the Sn axis, as
shown. One of these corresponds to a lower mean normal stress than the
other and these solutions are therefore called the weak and strong solutions,
respectively. As indicated on the figure, the poles and the shear directions will
be different for the two circles. Which of these solutions to choose is usually
determined from additional boundary conditions or from looking at the
boundary conditions as a whole, rather than at one point. This uncertainty
is due to the quadratic nature of the yield criterion, as was indicated for the
punch indentation problem of Section 12.5.
Returning now to Figure 12.6.4, the weak solution is adopted. The Mohr's
circles for points A and B are shown by the solid and broken-line circles,
respectively. As the solid-line circle rolls along the line r = k, the pole will
trace a cycloid which is the image of the first slip line through A. As the
broken-line circle rolls along s. = - k, the point B will describe a cycloid,
shown by the broken line, which will be the image of the second slip line
FIGURE 12.6.5 Strong and weak solutions.

through B. The intersection of the two cycloids at C will be the image of the
intersection of the two slip lines in the physical plane at C. The directions of
the normals to the cycloids at C fix the directions of the tangents to the slip
lines at C in the physical plane. The actual location of C in the physical plane
is found from the conditions that AD = CD and BE = CE. This amounts
to assuming that the slip-line arcs AC and BC are circular.
Once the slip lines have been established, the hodograph can be constructed
from the known velocities on the boundary using the orthogonality relation
between the slip line and hodograph elements.
The above is a brief outline of Prager's geometric construction of slip-line
fields. The interested reader is referred to references [6] through [9] for further
discussion and examples.
12-7 COMPLETE SOLUTIONS. UPPER AND

LOWER BOUNDS
The slip-line solutions discussed heretofore are incomplete in that no

attempt has been made to extend these solutions into the rigid regions
adjacent to the plastic regions to determine whether an equilibrium stress
distribution satisfying the boundary conditions and nowhere exceeding the
yield point exists in those regions. Such partial solutions are upper-bound
solutions; i.e., they overestimate the load necessary to produce the plastic
flow. This follows from the upper bound theorem, derived in reference [10].
We shall discuss the upper and lower bound theorems in greater detail in
Chapter 13. For our purpose here, the upper bound theorem can be stated
briefly as follows: A load which produces a kinematically admissible velocity
field will be equal or greater than the true load.
A kinematically admissible velocity field is one which satisfies the velocity
boundary conditions, is incompressible, and is continuous with continuous
first derivatives, except at certain discontinuity surfaces where the normal
velocity must be continuous but the tangential velocity may suffer a jump on
crossing the surface. Furthermore, on the boundary we must have
where X and Y are the boundary forces and dS is an element of boundary

area.
Since the slip-line solutions give kinematically admissible velocity
they are upper bound solutions. To obtain a complete solution, it is ne1~es:sar:y
Sec. 12-SJ Slip Lines as Characteristics 285
to determine if the stress fields in the rigid regions also satisfy equilibrium
and the boundary conditions and nowhere exceed the yield stress. If such a
statically admissible stress field exists in both the plastic and rigid parts, then
this solution constitutes a lower bound, and since the solution is both an
upper bound and a lower bound, it is a complete solution, giving the true
load for the problem.
The fact that a statically admissible stress field as defined above gives a
lower bound is expressed by the lower bound theorem [10], which can be
stated as follows: A load which produces a statically admissible stress field will
be equal to or less than the true load that will produce plastic flow.
Lower bound solutions are usually difficult to obtain since it would be
necessary to solve the equilibrium equations in the elastic or rigid region.
Upper bound solutions can be obtained from slip-line solutions and are of
greater value, in as much as they ensure that a certain operation will be
performed, since the calculated load will be greater than the required load.
Discussions of methods of obtaining upper bounds with examples can be
found in references [II] and [5].
12-8 SLIP LINES AS CHARACTERISTICS
The slip lines previously discussed are actually the characteristics of the
differential equations defining the problem. All the properties of charac-
teristics of hyperbolic equations and all the mathematical methods for
solving such equations can therefore be applied directly to the plane strain
problem of a rigid-perfectly plastic material. In this section we shall briefly
describe the origin and properties of characteristics to give the reader an
insight into the mathematical origin of slip-line fields. The reader interested
only in the physical description of the previous sections may skip to Chapter
13 without loss of continuity.
Consider the first-order differential equation
8u 8u
a-+b-=c (12.8.1)
ex ey
where a, b, and c are functions of u, x, andy but not of the partial derivatives
of u. Such an equation is called quasilinear. Introduce the standard notation
8u
8y =q (12.8.2)
Then (12.8.1) can be written

ap + bq = c (12.8.3)
We now pose the following question. Given a curve C in the xy plane

along which the function u is specified, does there exist a solution u of equa-
tion (12. 8.1) satisfying these "initial conditions" ? This is the two-dimensional
analogue of the ordinary first-order equation, where the initial value is
specified at one point. The above problem is called the Cauchy problem, and
the answer is that it depends on the curve C.
It is obvious that if given the values of u on C, p and q could be determined
on C such that (12.8.1) [or (12.8.3)] was satisfied, then u could be computed
a small distance away from C by use of a Taylor series expansion in two
variables. That is, if the values of x andy on Care designated by X 0 and y 0 ,
then
(12.8.4)
The value of u could thus be obtained at a neighboring curve. By repeating

this process, u could be determined over some region, the limits of which,
if any, are not yet known. The Cauchy problem, in this case, therefore
reduces to the problem of determining p and q on the curve C.
To find p and q we proceed as follows. Let s be the arc length along the
curve. Then if the derivatives p and q exist, we must have
du 8u dx 8u dy dx dy
ds = 8x ds + 8y ds = p ds + q ds (12.8.5)
Equations (12.8.5) and (12.8.3) give us two equations in the two unknowns
p and q. Solving these two equations gives
cdy- bdu
p =a dy- b dx
adu- cdx
q =a dy- b dx
Thus, given a, b, and c, the value of u on the curve C and the shape of the
curve, p and q can be computed from (12.8.6) and u in the neighborhood
of C obtained from (12.8.4). However, this will not be true for any curve C,
for if Cis such that the denominator in (12.8.6) vanishes; i.e., if
dy b
dx =a
Sec. 12-8] Slip Lines as Characteristics 287
then there will obviously be no solution unless the numerators also vanish.
A curve C whose equation satisfies (12.8. 7) is called a characteristic curve,
and if u is specified along such a curve, there will be no solution to the
problem unless u is specified so that
du c
or
dx =a (12.8.8)
causing the numerators in (12.8.6) to vanish. In the latter case, i.e., if u is

specified on C according to (12.8.8), there will be an infinity of solutions,
since (12.8.6) are then indeterminate.
To summarize: In answer to the question posed at the beginning, if u is
specified along a curve C, then a solution to (12.8.1) exists provided Cis not
a characteristic of the differential equation. If C is a characteristic, then
either there is no solution, if (12.8.8) is not satisfied, or there are an infinity
of solutions, if (12.8.8) is satisfied. Equation (12.8.8) is called the compatibility
equation.
Let us illustrate these results by a simple example [12]. Consider the
differential equation
8u + 8u = 1
ox 8y (12.8.9)
or p+q=1
Then a = b = c = 1 and, from (12.8. 7),
(12.8.10)
or
The characteristics are therefore straight lines with slopes of unity, as

shown in Figure 12.8.1. For a solution to exist, u must satisfy (12.8.8) or
u= x +B = y + (B - A) (12.8.11)
along these lines. Suppose now that u is given on the line segment 0 < x < 1,
y = 0. This line segment is the curve C of the previous discussion. Then
since Cis not a characteristic, we should be able to obtain a unique solution
for u. The simplest way to do this is to integrate along the characteristics.
Th11s considering the characteristic intersecting the x axis at x = x1 as shown,
FIGURE 12.8.1 Characteristics for equation (12.8.9).
the value of u specified at this point is U;, and since along this characteristic u
must satisfy (12.8.11 ), it follows that
B-A=Ut
so that
U = Y + Ut = X + Ut - Xt (12.8.11a)
for all
0 <X< 1
A solution has thus been obtained for the problem using the initial data
and the characteristic and compatibility equations (12.8.10) and (12.8.11).
These equations are the equivalent of the original differential equation,
which does not have to be used at all. This method of solving a partial differ-
ential equation by reducing it to a set of ordinary differential equations for
finding the characteristics and for integrating along the characteristics is
called the method of characteristics. Thus the introduction of characteristics
not only serves the purpose of determining whether a unique solution exists
but actually provides a method for obtaining the solution.
From the above example it is apparent that the solution is defined and is
unique only in the region bounded by the terminal characteristics at x = 0 and
x = 1. This region is called the region of influence. To determine the solution
outside this region requires additional data outside the range 0 < x < 1.
We can state this result as a general theorem.
A partial differential equation having real characteristics has a unique solution
within the region bounded by the curve C upon which the initial data is specified,
and the two characteristics intersecting the ends of C. If C coincides with
characteristic there is no unique solution.
Let us consider now the case when u is specified along one of the
teristics, say, the line x = y. For
specified on this line such that
U = X + Uo = Y + Uo
where u0 is the value of u at x = y = 0. The solution elsewhere, however, is

not unique. We can add, for example, a function K(x - y) to u, with K an
arbitrary constant and still satisfy the initial conditions along the line x = y.
Thus
U = X + Uo + K(x - y) (12.8.13)
will satisfy the differential equation (12.8.9) and also the initial values on
y = x. This illustrates the fact that if the initial conditions are specified on
characteristic, the solution is not unique. In this case we might say that the
solution is not defined uniquely at points not on this line because the terminal
characteristics coincide, and the region of influence is just the line itself.
Another very important property of characteristics can now be seen-the
possibilities of discontinuities in the solution which are propagated along
characteristics. For example, suppose on the initial line y = 0 of the previous
example we are given
u = /1(x) 0 < x < x1

(12.8.14)
u =!2(x) x1 < x < 1
where f 1 (x 1 ) i= fix 1 ); i.e., u is double-valued at this point. Then the solution

will be double-valued and hence discontinuous all along the characteristic
y = x - x 1 , passing through this point. The values of u to the left of this
line will be determined by u = f 1 (x) and to the right by u =fix).
In a similar fashion, discontinuities in the derivatives of u may be propa-
gated along characteristics. For example, if we are given that at y = 0,
u = 1
u=x+1
Then the solution using (12.8.11) is
u=y+l to the left of y = x - 1

(12.8.15)
u=x+1 to the right of y = x - 1
On the line y = x - 1, u is continuous, but the derivatives of u are not, for
8u =0 8u = 1 to the left of y = x - t
ox 8y
(12.8.16)
8u = 1 Bu =0 to the right of y = x - t
Bx oy
These derivatives are therefore discontinuous all along the characteristic

y =X-!.
In the above example the differential equation (12.8.9) was linear. It was
therefore possible to determine the characteristics once and for all independ-
ently of the solution u. Thus equation (12.8.7) could be integrated first and
then equation (12.8.8) solved for u along the characteristics. However, if the
equation were quasilinear, i.e., if a and b were functions of u, then equation
(12.8.7) could obviously not be solved for the characteristics without first
knowing the solution u. In this case it is necessary to solve (12.8. 7) and
(12.8.8) simultaneously; i.e., u and the characteristics must be determined
simultaneously.
Second-Order Equation
Let us now consider the second-order quasilinear partial differential

equation
(12.8.17)
By quasilinear we mean that the coefficients may all be functions of u,

8uj8x, 8uj8y, as well as of x andy. Using the standard notation,
(12.8.18)
equation (12.8.17) is written
Ar + Bs + Ct = D (12.8.19)
Now assume that on some initial curve C0 in the xy plane, the values of u,
and its normal derivative are given. Alternatively the partial derivatives p and
q are given, for then u and the normal derivative can be computed. The
values of u, p, and q along C0 are called a strip of first order and u, p, and
must obviously satisfy the condition
du = p dx + qdy
called the strip condition.
We now ask ourselves the same question as before. Does there exist
solution u of equation (12.8.17) satisfying these initial conditions? If
second derivatives r, s, and t on the curve can be determined so as to
(12.8.17), then as before u can be determined as well as its first derivatives

a small distance away from the curve by Taylor series expansions, and in
this way the solution can be continued into the region. Let us see then if the
second derivatives on the curve can be computed. On the curve
du = -dx
au + -dy
au
ox ay
ax8 udx + ox8 uoy dy

2
(~xu)
2
and d
u = 2
au) 8 u 8 u 2
d (oy = ox oy dx + oy2 dy
2
or rdx + sdy = dp
sdx + t dy = dq
(12.8.20)
and the differential equation is
rA + sB + tC = D
A, B, C, D, dq, and dp, being known on the curve, we have three equations
to solve for the unknown second derivatives r, s, and t. For example,
dx dp 0
0 dq dy
A D c
S= (12.8.21)
dx dy 0
0 dx dy
A B c
If the determinant in the denominator vanishes there will be no solution
unless the numerator also vanishes, in which case there are an infinite number
of solutions. If the denominator vanishes,
A(dy)2 - B dx dy + C(dx) 2 = 0
(12.8.21a)
(dy)
2
dy
A- -B-+C=O
dx dx
dy B ± VB 2 - 4AC
(12.8.22)
dx = 2A
Equation (12.8.22) defines two sets of curves (one using the plus sign, one
using the minus sign). These are called characteristic curves, and if the data
are given along one of these curves, no solution exists unless, as already
mentioned, the numerator determinants also vanish. For this case,
A dp dy + C dq dx - D dx dy = 0 (12.8.23)
Along the characteristic curves equation (12.8.23) must be satisfied. This is

the compatibility equation.
Equations (12.8.22) and (12.8.23) replace the partial differential equation
(12.8.17) and if the initial curve is not a characteristic, a solution can be
obtained by finding the characteristics from (12.8.22) and then integrating
along the characteristics using (12.8.23). If the equation is nonlinear, i.e.,
if A, B, and Care functions of u, p, and q, then (12.8.22) and (12.8.23) must
be solved simultaneously, as discussed for the first-order equation. If the
data are prescribed along one of the characteristic curves, there is no unique
solution.
The above discussion implies that (12.8.22) can be solved obtaining two
real families of curves. But this can only be true if B 2 - 4AC ;:o: 0. If
B 2 - 4AC < 0 there are no real characteristics. There are therefore three
types of equations.
1. B 2 - 4AC > 0. The equation is called hyperbolic. There are two sets
of characteristics.
2. B 2 - 4AC = 0. The equation is called parabolic. There is only one set
of characteristics.
3. B 2 - 4AC < 0. The equation is called elliptic. There are no real
characteristics.
The parabolic equation is similar to the hyperbolic with regard to the
properties and use of the characteristics. The first-order equations previously
discussed are actually parabolic, since they always have one family of charac-
teristics (if the coefficients are real). In fact, since there are no second deriva-
tive terms, B 2 - 4AC = 0.
The properties of the hyperbolic equation are then essentially the same as
previously discussed for the first-order equation. We shall briefly note some
of these properties and make a comparison with the elliptic case.
In the elliptic case, if u and its normal derivative is prescribed along any
curve C, then a solution can always be found in the neighborhood of C.
If the values are prescribed on a closed boundary, which is the usual case, the
solution will be determined within the bounded region. It is necessary, how-
ever, that u be regular on C, i.e., it possess derivatives of all orders.
In the hyperbolic (and parabolic) case, if the curve Cis not a characteristic,
a solution exists. The prescribed values in this case need not be regular and
the higher derivatives of the prescribed values may have finite jumps at points
on C. These discontinuities are then propagated along the characteristics
where they originate. If Cis a characteristic curve, there is no solution unless
u and its derivatives satisfy a compatibility relation, in which case there are
an infinity of solutions. It is also evident that closed boundaries are excluded,
for a given characteristic would then intersect the boundary twice, and values
of u and its derivatives could not be assigned arbitrarily to both points of
intersection.
Boundary-Value Problems
There are generally three types of boundary-value problems for hyperbolic

equations, [13].
I. Given u, p, and q on a curve C0 which is not a characteristic-and which
intersects each characteristic at most once, u can be determined in a triangular
region D 0 bounded by C0 and a characteristic of each family as shown in
Figure 12.8.2(a). The curves y constant and x constant are the characteristic
curves of each family. More specifically, the value of u at each point P of
Do is determined by the values of u and its derivatives on the portion CP of
C0 which is bounded by the characteristics through P. The segment CP is
called the domain of dependence of the point P.
2. A linear relationship a(8uj8n) + bu = c(x, y) is prescribed on an arc
C0 , and in addition u is prescribed on a characteristic arc Cc passing through
one end point of C0 • The solution can again be obtained in the triangular
region bounded by C0 , C0 , and the characteristic of the other family inter-
secting Cc and C0 , as shown in Figure 12.8.2(b).
3. u is prescribed on two intersecting characteristics Cc and C~, as shown
in Figure 12.8.2. Of course, u must satisfy the compatibility relation on both
characteristics. The solution is determined in the quadrilateral bounded by
the four characteristics as shown.
The reader will recognize the similarity between these three boundary-value
problems and the three boundary-value problems for slip-line fields discussed
in Section 12.5.
Finally, it follows that by combining these three basic types of boundary
conditions, more complicated conditions can be used. For example, in
Figure 12.8.2(d), u, p, and q may be given on C0 • But then only one relation
of the type a(8uj8n) + bu = c(x, y) can be prescribed on each of Cb and C~.
x=constant x=constant
jl=constont
(a) (b)
jl=constont
(c) (d)
FIGURE 12.8.2 Boundary-value problems (Reference [13].)
The solutions in the various regions shown can then be patched together along
characteristics so that u is continuous, while 8uj8x and 8uj8y may be dis-
continuous. Note again that closed boundaries are not admissible.
Numerical Solution
An analytical solution of the hyperbolic equation can be effected in theory,

by Reimann's method, which essentially reduces the boundary-value uHJu"~"'
of the first kind along an arbitrary curve C0 [Figure 12.8.2(a)] to a simpler
boundary-value problem of the third kind along two intersecting vlJ<Um'c~
teristics [Figure 10.8.2(c)]. This is essentially the counterpart of
function in solving elliptic equations. In practice, however, only a few
simple problems have been solved in this way. The widest use of the method
of characteristic is found in conjunction with numerical methods.
Suppose then that u, p, and q are given along a noncharacteristic arc C and
we wish to obtain the solution in the region bounded by C and the two
terminal characteristics as shown in Figure 12.8.2(a). It is desired to solve
equations (12.8.22) and (12.8.23) numerically. If A, B, and C are functions
only of x and y, (12.8.22) can be solved separately and the characteristics
constructed independently of the initial conditions. Such characteristics are
called fixed. If A, B, and C are also functions of u, p, and q, then (12.8.22)
and (12.8.23) must be solved simultaneously. (Note: If A, B, and C are
functions of u, p, and q but not of x andy, then it is possible to interchange
the roles of the independent and dependent variables and transform the
partial differential equation into a linear differential equation with x and y
as the dependent variables. This is called a hodograph transformation. The
solution is then obtained in the plane of u and 8u/8n, or p and q, called the
hodograph plane, instead of the xy plane.)
Equations (12.8.22) can be written
dy+ =f+ dx
(12.8.24)
dy- =f- dx
where
B + VB 2 - 4AC
2A
(12.8.25)
and (12.8.23) can then be written

D dy- - Af+ dp - C dq =0 along anf+
(12.8.26)
D dy + - Af- dp - C dq =0 along auf-
FIGURE 12.8.3 Numerical solution.

Let us then draw the curve C and two families of characteristics as shown
in Figure 12.8.3. Consider any two adjacent points P and Q on C and let the
j+ characteristic from P intersect the f- characteristic from Q at R. If the
characteristics are fixed, the position of the point R is known. If the charac-
teristics depend on the solution, the coordinates of R must be determined at
the same time that p and q are determined at this point. Equations (12.8.24)
and (12.8.26) are now written approximately as follows:
YR - YF = t(JR+ + f/)(xR - Xp)

YR - YQ = !(fR- + fQ-)(xR - xQ)
and (12.8.27)
t(DR + Dp)(YR - yp) - t(ARfR+ + Apf/)(PR - h)
- t(CR - Cp)(qR - qp) = 0
t(DR + DQ)(YR - yQ) - t(ARfR- + AQJQ-)(PR - PQ)
- t(CR - CQ)(qR - qQ) = 0
Also from the condition
du = p dx + qdy
UR - Up = t(PR + h)(XR - Xp) + t(qR + qp)(YR - YP)
(12.8.28)
or uR - uQ = t(pR + PQ)(xR - xQ) + t(qR + qQ)(YR - YQ)
Equations (12.8.27) and one of (12.8.28) furnish five equations for the five
unknowns xR, yR, PR, qR, and uR. These equations are nonlinear and they
usually have to be solved by some iterative method. Convergence is usually
rapid if the interval PQ is not too large.
In the same way, the solution at other grid points adjacent to the initial
curve, such as Sin the figure, can be obtained. From R and S we can proceed
to T, and so on. The solution is of course defined only in the region bounded
by the terminal characteristics, as previously explained.
Two First-Order Equations. Slip Lines
If we have two first-order equations of the form
au au & &
a 8x + b By + e 8x + f By =d
A 8u + B 8u + E dv. + F 8v = D
8x By 8x By
with
au au
du = ox dx + dy dy
(12.8.30)
ov ov
dv = ox dx + oy dy
then by a similar procedure the equation for the characteristics is
2
dy) dy .
(aE - eA) ( dx + (eB +fA - bE - aF) dx + bF - fB = 0 (12.8.31)
and the compatibility relation along the characteristics is
du dx dy 0
dv 0 0 dx
=0 (12.8.32)
d a b e
D A B E
It will now be shown that the slip lines discussed in Sections 12.1 through
12.7 are characteristics of the governing differential equations and all the
properties of characteristics discussed in Section 12.8 apply to them. We
start with equations (12.1.16), which are essentially the equilibrium equations
with the yield condition included:
ox 88 . 88
-
ox
- COS 28- -
dx
Sill 28-
oy
=0
(12.8.33)
ox . 88 88
-
ay
- Sill 28 -
ax
+ COS 28 -
ay
=0
These are two simultaneous quasilinear first-order equations, and if they are
hyperbolic we should be able to use the method of characteristics. Comparing
the coefficients with those in (12.8.29), it is seen that
a= 1 b=O e = -cos 28 f= -sin28 d=O

A=O B=1 E = -sin28 F = cos28 D = 0
and equation (12.8.31) for the characteristics becomes
sin 28 (;7x)
2
+ 2 cos 28 1x - sin 28 = 0
dy -cos 28 ± 1
or (12.8.34)
dx = sin28
or
2 sin 2
. a
ea= tan
l)
u
2 Sin u COS u
(12.8.35)
But tan 8 and tan(& + 'TI'/2) are the slopes of the slip lines, so that (12.8.35)
shows that the characteristics of the differential equations (12.8.33) are just the
slip lines. We can identify the ex slip lines with the plus characteristics and the
(3 slip lines with the minus characteristics, as defined by equations (12.8.35).
To determine the compatibility relation that must be satisfied along the
slip lines, equation (12.8.32) gives
dx dx dy 0
de 0 0 dx
=0 (12.8.36)
0 0 -cos 28
0 0 -sin 28
or dX - (sin 28 1x + cos 28)d8 = 0

and, from (12.8.34),
dx + d& = o
or X- 8 = constant along ex characteristic
(12.8.37)
X + 8 = constant along (3 characteristic
which are Hencky's equations, previously obtained.
Problems
2. Show that the principal stresses for plane strain problems are given by
equations (12.1.7) and (12.1.8).
3. Show that the Tresca criterion differs from equation (12.1.5) only by a
multiplicative constant.
5. Show that equations (12.1.14) can be obtained by means of equations
(12.1.4) and (12.1.12).
6. Prove Theorems 6 and 8 of Section 12.3.
7. Determine the velocity distribution for the condition of Figure 12.4.2 using
the Geiringer equations.
8. Obtain the complete solution for the rigid punch indentation of Figure
12.4.2 assuming the rigid plastic boundary to be HIJKLMN.
9. Show that the choice of a plus sign in equation (12.4.1) or a minus sign in
equation (12.4.4) would lead to solutions which violate the initial assump-
tions.
10. Prove that the point of intersection of a line drawn through the pole of the
Mohr's circle and the circle will give the stress state acting on a plane whose
trace is parallel to this line.
References
1. W. Prager and P. G. Hodge, Jr., Theory of Perfectly Plastic Solids, Wiley
New York, 1951.
2. P. G. Hodge, Jr., An Introduction to the Mathematical Theory of Perfectly
Plastic Solids, ONR-358, Feb. 1950.
3. L. Prandtl, Dber die Harte Plastischer Koerper, Goettinger Nachr., Math.
Phys. Kl., 1920, pp. 74~85.
4. R. Hill, The Plastic Yielding of Notched Bars Under Tension, Quart. J.
Mech. Appl. Math., 2, 1949, pp. 40-52.
5. E. G. Thomsen, C. T. Yang, and S. Kobayashi, Mechanics of Plastic Defor-
mation in Metal Processing, Macmillan, New York, 1965.
6. W. Prager, A Geometrical Discussion of the Slip-Line Field in Plane Plastic
Flow, Trans. Roy. Inst. Tech. (Stockholm), 65, 1953, pp. 1~26.
7. W. Prager, The Theory of Plasticity: A Survey of Recent Achievements,
Proc. Inst. Mech. Eng., 169, 1955, pp. 41~57.
8. H. Ford, The Theory of Plasticity in Relation to Engineering Application,
J. Appl. Math. Phys., 5, 1954, pp. 1~35.
9. J. M. Alexander, Deformation Modes in Metal Forming Processes, Proceed-
ings of the Conference on Technical Engineering Manufacture, Institute of
Mechanical Engineers, New York, 1958, Paper No. 42.
10. D. C. Drucker, H. J. Greenberg, and W. Prager, The Safety Factor of an
Elastic~Plastic Body in Plane Strain, J. Appl. Mech., 18, 1951, pp. 371~378.
11. W. Johnson and P. B. Mellor, Plasticity for Mechanical Engineers, Van
12. L. Fox, Numerical Solution of Ordinary and Partial Differential Equations,
Pergamon Press, London, 1962.
13. G. A. Korn and T. M. Korn, Mathematical Handbook for Scientists and
Engineers, McGraw-Hill, New York, 1961.
14. H. Hencky, Ueber einige statisch bestimmte Faelle des Gleichgewichts in
plastischen Koerpern Z. angew. Math Mech., 3, 1923, pp. 245~251.
15. H. Geiringer, Beit zum Vollstandigen ebenen Plastizitats-problem, Proc 3rd
Intern. Congr. Appl. Mech, 2, 1930, pp. 185~190.
General References
Prager, W., Introduction to Plasticity, Addison-Wesley, Reading, Mass., 1959.
Ford, H., Advanced Mechanics of Materials, Wiley, New York, 1963.
CHAPTER.13
LIMIT
ANALYSIS
13-1 DESIGN OF STRUCTURES
The theory of limit analysis is used primarily in the design of steel structures
composed of various elements such as beams, frames, girders, arches, etc.
For many years the basis for structural design has been the allowable stress
concept. The allowable stress was usually taken to be the yield stress of the
material, and the design stress was then taken to be some fraction of the
allowable stress, depending on the factor of safety used. In some applications
the allowable stress was governed by the possibility of buckling or fatigue.
The design methods used were always elastic.
It is apparent, however, that the important consideration in an engineering
structure is not whether the yield stress is exceeded at some point, but whether
the structure will carry the intended loads or perform its intended function,
and there is really no reason for assuming that the stress in the structure
should never exceed the elastic limit. As a matter of fact, it is fairly evident
that in almost all structures, local plastic flow will occur at stress raisers and
at points of discontinuity in the geometry, and, furthermore, residual stresses
as high as half the yield strength may already be in some elements as they
come from the steel mills, before the load is even applied.
The practice is therefore becoming more widespread to design structures
into the plastic range, the materials being assumed to behave in an elastic-
perfectly plastic manner. In these design procedures no attempt i~ made to
300
Sec. 13-2] Simple Truss 301
determine the stresses and strains in the structure, but rather what is sought
is the load-carrying capacity or limiting load at which the structure will
collapse. This type of analysis is called limit design or plastic design, and the
load at collapse is called the plastic collapse load.
For the great majority of problems this type of approach makes more
sense than a design based on an elastic analysis. Furthermore, since the stress
distribution is not SOJ.lght, it is much simpler. There are other important
problems, however, which cannot be resolved on the basis of simple limit
analysis. Among these are buckling, fatigue, and fracture.
The first use of plastic design in structures was apparently made by
Kazinczy in the design of apartment buildings in 1914 (see reference [1]).
Since then many contributions have been made to this theory both in this
country and abroad. Plastic design is already part of certain specifications in
some countries, and it is being used more and more by engineers in this
country.
13-2 SIMPLE TRUSS
To illustrate the concepts behind limit analysis a few simple examples

taken from reference [2] will be considered. Let us start with the problem
of the simple symmetric truss shown in Figure 13.2.1 and analyze this truss
2 2
L
______ lI
T2
F
FIGURE 13.2.1 Simple truss.
by the conventional elastic method and by the method of limit analysis.

Assume the truss to be made of mild steel with an elastic modulus E = 30 x
106 and a yield stress a 0 = 30,000. The stress-strain curve is assumed elastic-
perfectly plastic. Let L = 12 in. and cross-sectional area of all three members,
A = 1 in. 2 • The conventional analysis proceeds as follows. Let T1 be the
resultant force in bar 1 and T2 the resultant forces in each of the bars 2.
Then equilibrium of forces requires
l13.2.1)
302 Limit Analysis [Ch. 13
Since this is a redundant structure (having more members than necessary to

maintain equilibrium), the equilibrium equation (13.2.1) is obviously not
sufficient to obtain the solution and the continuity of displacement must be
used. The elongation of rod 1 is
(13.2.2)
and the elongation of each of rods 2 is (see Figure 13.2.2)
FIGURE 13.2.2 Elongation.
(13.2.3)
For continuity of displacements,
(13.2.4)
which results in
(13.2.5)
F
T2 = 2 + V2
(13.2.6)
2F
Tl = 2 + V2
Thus ifF = 20,000 lb,
T1 = 11,720 lb
(13.2.7)
T2 = 5,860 lb
Sec. 13-2] Simple Truss 303
and the elongations are
"' = 5,860 X 12 X Vl = O 00331 .

o2 30 X 1Q6 ' Ill.
(13.2.8)
o1 = v'2 (0.00331) = 0.00468 in.
This completes the elastic analysis, the stresses and deformation having been
determined. If the question is now asked, How strong is the truss?, the usual
answer is to quote a factor of safety giving the ratio of the maximum safe
load to the applied load. The conventional practice has been to take the
maximum safe load as that load which would just cause the maximum stress
to reach the yield stress. In this case, therefore, the factor of safety would be
a0 30,000
f = T 1 /A = 11,720 = 2 · 56 (13.2.9)
and the maximum safe load presumably is
Fmax = 2.56 X 20,000 = 51,200 lb (13.2.10)
However, if the material is ductile, as in the case of mild or structural steel,

this safety factor and the corresponding maximum safe load are not too
meaningful. The actual load that can be carried by the truss is considerably
greater. Assuming the material to be perfectly plastic, if the load is raised to
51,200 lb, the middle bar will just be at the yield. However, bars 2 will only
be at t their yield, since the stress in these bars is one half the stress in the
middle bar, as seen from equation (13.2.5). As the load is increased above
51,200 lb, the stress in the middle bar will remain the same, since it is already
at the yield, and the. additional load will be carried by the other two bars.
Furthermore, since the latter two bars are still elastic, the deformations of
the system will still be small. The load can be increased until the outside bars
just reach the yield point. From (13.2.1) the load at which this will occur is
F = a0 + v'.2 a 0 = (1 + V2)a0 = 72,400 lb (13.2.11)
This is the plastic collapse load. The deflections just as this load is reached
will be
o2 = 0.017 in.
(13.2.12)
o1 = v'.2 o2 = 0.024 in.
which are still small. If the load is increased beyond 72,400 lb, large deforma-
tions will take place and it can be assumed that the structure is no longer
usable.
We note that the factor of safety based on the load-carrying capacity is,
for the case F = 20,000,
72,400
fp = 20 000 = 3·62
'
rather than 2.56. Moreover, the load-carrying capacity or the plastic collapse
load is determined simply from the equilibrium equation (13.2.1) and a
knowledge of the yield stress. The detailed elastic analysis previously made
was not necessary. This is another great advantage of limit analysis.
Now consider the inverse problem, i.e., the problem of design rather than
analysis. Supposing we were given a truss such as shown in Figure 13.2.1,
and told to design such a truss to have minimum cross-sectional areas for a
working load of 20,000 lb, and a factor of safety of 3, taking all three bars
to have equal areas.
According to elastic design, the forces in the bars at a load of 3 x 20,000 =
60,000 lb would be, from (13.2.6),
60 000
T.2 = •
3.414
= 17 600 lb
'
T1 = 35,200 lb
To keep the stress in the middle bar below the yield stress of 30,000 psi, the
area would have to be
A 35,200 1 17 . 2
= 30,000 = . lll.
If plastic analysis were used, however, the stresses in all three bars could
be allowed to reach 30,000 psi at a load of 60,000 lb. Therefore, from
(13.2.1),
AC1 + v2) x 3o,ooo = 6o,ooo
or A .i
= 2 14 = 0.83 in. 2
Sec. 13-3] Pure Bending of Beams 30_5
Thus, using plastic design, 30 per cent of the material is saved. Furthermore,
under the working load of 20,000 lb, the actual stresses in the bars are
al = Tl = 11,720 = 14 100
A 0.83 '
a2 = 2= 7,050
which are well within the elastic range.
13-3 PURE BENDING OF BEAMS
The next problem considered is the pure bending of a beam. Consider the
beam shown in Figure 13.3.1. The conventions for positive moments and
FIGURE 13.3.1 Beam under pure bending.
deflections as shown in this figure will be used throughout. The bending

moment is given by
M = 2b Jh Zax dz
-h
(13.3.1)
With the usual assumption of beam theory that plane sections remain plane
the strain ex is given by
Bx = kz (13.3.2)
where k is the curvature of the middle surface.
If M is sufficiently small, the stresses will be elastic and
(13.3.3)
Substituting into (13.3.1) and integrating gives
(13.3.4)
The maximum stress will occur at the outer fibers and. will be tensile at
z = h and compressive at z = -h. When this maximum stress equals the
yield stress a 0 , we have, from (13.3.3),
ao = kEh (13.3.5)
and substituting into (13.3.4) gives for the moment when the outer fiber just
reaches the yield stress
(13.3.6)
On the basis of elastic analysis this is the maximum allowable moment.

However, on the basis of plastic analysis, it is evident that this is not the
maximum moment the beam can withstand, since if the moment is increased
beyond M., the additional load will be transferred to the inner fibers of the
beam. The stress distributions for various states of loading are shown in
Figure 13.3.2. The first figure (a) shows the distribution when M is less than
(a) (b) (c) (d)
FIGURE 13.3.2 Stress distributions in rectangular beam.
Me, (b) shows the distribution when M just equals Me, and as M is increased
more of the beam reaches the yield stress but the part between rh and - rh
is still elastic, as shown in (c). The moment for the condition shown in (c) is,
from (13.3.1),
rh kEz 2 dz + 4b
fh
M = 2b
J
-rh r/1
a0 z dz
but, from (13.3.5),
kE = ao
rh
Hence
Sec. 13-4] Beams and Frames with Concentrated Loads 307
As M increases, eventually r approaches zero, as shown in (d). At this time

the complete beam is flowing plastically and the moment can no longer be
increased. The value of the moment at this condition is called the yield
moment, limiting moment, or fully plastic moment. Denoting this moment by
M 0 , from (13.3.7)
(13.3.8)
Comparing equations (13.3.8) and (13.3.6), it is seen that the ratio of the
fully plastic moment to the maximum elastic moment is
~: = 1.5 (13.3.9)
This ratio is called the shape factor of the beam. It will, of course, be differ-
ent for different-shaped cross sections. For a circular cross section, for
Me 34 b h 2 D"o 40" 0 bhA(1+f) fO"oA3 T?_O"obhz 7;-bzhD"o
Mo 2bh 2D" 0 4D" 0 bhA(1+t) 4-0"oA3 2-{2

3 D"o
bh2 i bzhD"o
Mo 1+~j
1.50 1.70 2.34 2.00
Me 1+i-i
FIGURE 13.3.3 Shape factors for common beam sections.
example, it will be 1. 70. Figure 13.3.3 shows the values for several common
beam sections.
13-4 BEAMS AND FRAMES WITH

CONCENTRATED LOADS
In Section 13.3 the case of pure bending where all sections of the beam
behaved the same way was discussed. We now consider concentrated loads
where the moment distribution varies along the beam. Our purpose, as before,
will be to determine the collapse loads of the structure.
Cantilever Beam with Tip Load
Consider a cantilever beam with a concentrated load at the tip as shown in

Figure 13.4.1. The bending moment will obviously vary linearly from zero
~A fF
L~----)o')o-1
/ <
(a)
FIGURE 13.4.1 Collapse of cantilever beam.
at the tip to a maximum at the built-in end as shown. If M 0 is the yield

moment, then for some load F 0 for which
M = LFo = Mo (13.4.1)
the section at A will become fully plastic. Since perfect plasticity has been
assumed, the beam will now be able to rotate freely about the point A and
will thus have collapsed. We say that a yield hinge or plastic hinge has formed
at A. Such a hinge can be thought of as one which is locked in place as long
as the moment M is less than M 0 , and which becomes free to turn freely when
M becomes equal to M 0 • The collapse load, from (13.4.1), is obviously
2
Fo = M 0 = 2bh a0 (13.4.2)
L L
for a rectangular cross-sectional beam.
Indeterminate Beam
The indeterminate beam shown in Figure 13.4.2 is considered next. The

end A is built in and the end Cis simply ~upported so it can carry no moment,
F
L ---------;l>-1
A;~-<----L---)o-K---
'/!--------'--::--------;,
(a) B
(e)
FIGURE 13.4.2 Collapse of simple indeterminate beam.
the load F being applied at the middle. The elastic solution first requires
determining the reaction at C, which can be done by first calculating the
displacement at C with the support removed and calculating the reaction at
C necessary to bring the deflection back to zero with the force F removed.
The moment distribution can then be computed. To determine the collapse
load, however, none of this is necessary. The moment distribution when
everything is elastic is shown in (b), and as the force Fis increased, eventually
the section at A will become fully plastic, the moment at A being - M 0 • A
yield hinge is thus formed at A, just as in the case of the cantilever previously
discussed with the resultant moment distribution as shown in (c). Because
of the support at C, however, the beam will not yet collapse. As the load is
further increased, the point A cannot carry any more moment, but the rest
of the beam can and the moment at B will continue increasing until it reaches
the plastic moment M 0 , at which time a second yield hinge is formed. The
beam will then collapse, as shown in (e).
The calculation of the collapse load for this case is very simple. Let the
reaction at C be denoted by R. Then since the moment at A is - Ma and at

Bit is Ma,
MB = Ma = RL
MA = - Ma = 2LR - FaL
Hence
Fa= 3Ma R= Ma (13.4.3)
L L
For a beam of rectangular cross section it follows from (13.3.8) that
Fa= 6bft2aa
L
At this point it might be well to introduce a term frequently used in limit

analysis, the term mechanism, or kinematic mechanism. By a mechanism is
meant an articulated system which can deform without a finite increase in
load. Thus in the previous example when the second yield hinge was formed
at B, the system became a mechanism as shown in (e). There exist essentially
two hinged bars which can deform freely for an infinitesimally applied load.
We might say then that one of the objectives of limit analysis is determining
when a mechanism will be formed, and in the case when more than one
mechanism is possible for a system, to determine which is the collapse
mechanism. This last point will be illustrated in the next example.
Simple Frame
The example of a simple frame as shown in Figure 13.4.3 will now be

considered. Point 1 is built in and point 5 is pin supported and can therefore
Ia) I b) I c) (d)
FIGURE 13.4.3 Collapse of simple frame.

carry no moments. There are two concentrated forces, a vertical force of
amount 2F and a horizontal force 3F as shown. The critical points where the
moments will attain their largest absolute values are obviously the points
marked 1, 2, 3, and 4. It is seen that unlike the beam of the previous example,
where both critical points had to develop yield hinges, it is not necessary in
this case for all four critical points to develop yield hinges. As a matter of
fact, there are three distinct possibilities, as shown in (b), (c), and (d) of
Figure 13.4.3. Each of these three configurations is a possible mechanism.
The question then arises as to which is the correct one. This can be determined
as follows.
Let us consider each of these modes of collapse .in turn. Let the reactions
at point 5 be denoted by R 1 and R 2 , and consider the moments at the four
critical points.
M1 = 2R1 L - 8FL
M 2 = 2R 1 L - 2FL - 2R2 L
(13.4.4)
M 3 = R 1 L - 2R 2 L
M4 = -2R2 L
Eliminating the two reactions R 1 and R 2 gives
-M2 +2M3 - M4 = 2FL

(13.4.5)
- M1 + M2 - M4 = 6FL
If it is now assumed that the frame will collapse by the mechanism shown in
(c), then M 1 = -M0 , M 2 = M 0 , and M 4 = -Mo. From the second of
equations (13.4.5),
2L
F= Mo)
and from the first of (13.4.5), (13.4.6)
Ms = -!Mo
Now consider the mechanism shown in (b). Here
Ms = Mo
F= 2M0 (13.4.7)
L
which is impossible. The second mechanism shown in (b) is therefore not

possible.
Finally, consider the third mechanism shown in (d):
F= 5M0 (13.4.8)
8£
But this violates the original assumption for this case-that there is no yield
hinge at station 2; i.e., the moment at 2 is less than M 0 • This mechanism is
therefore also not admissible, leaving the mechanism of (c) as the only
possible one. The true collapse load is therefore given by equation (13.4.6).
Several results can now be noted. The collapse loads corresponding to the
cases of (d) and (b) are higher than the true collapse load given in (13.4.6).
It appears, then, that as the load is slowly increased from zero, the frame will
collapse the first chance it gets.
Second, if the load is decreased in these last two cases, so that the maximum
moment does not exceed M 0 , it would be necessary to divide by 12 and i,
respectively, as seen from (13.4.7) and (13.4.8), giving
and
FL 5
Mo 14
But these are less than the true collapse load of
It follows then that as the load is slowly increased, the frame will not
collapse for any load for which some equilibrium configuration can be found,
unless it produces a mechanism.
These two concepts can be formalized into basic theorems of limit analysis
to be discussed in Section 13.5.
13-5 THEOREMS OF LIMIT ANALYSIS
The previous problems were simple enough so that the correct collapse
load could be obtained in each case without too much difficulty. In more
complicated problems, however, it may be difficult or impossible to obtain
Sec. 13-5] Theorems of Limit Analysis 313
the collapse load exactly, and recourse must then be had to the upper and
lower bound theorems, which provide upper and lower bounds to the true
collapse load. These theorems are essentially the same as those presented in
Section 12.7 for the load to produce incipient plastic flow in a rigid-perfectly
plastic material. They were first presented by Gvozdev [3] and independently
proved by Hill [4, 5] for the rigid-perfectly plastic material, by Drucker et al.
[6, 7] for the elastic-perfectly plastic material, and in references (8] and [9] for
the special cases of beams and frames. We shall present the theorems in their
general form. The proofs, which are based primarily on the principle of
virtual work, can be found in the cited references and in a particularly elegant
form in reference [10].
Lower and Upper Bound Theorems
If an equilibrium distribution of stress can be found which balances the

applied load and is everywhere below yield or at yield, the structure will not
collapse or will just be at the point of collapse. This gives a lower bound on
the limit load, and is called the lower bound theorem.
The structure must collapse if there is any compatible pattern of plastic
deformation for which the rate at which the external forces do work is equal
to or exceeds the rate of internal dissipation. This gives the upper bound on
the limit load and is called the upper bound theorem.
The lower bound theorem merely says that the structure will withstand the
applied load by rearranging the internal stresses to best advantage, if at all
possible. It gives lower bounds on, or safe values of, the limit or collapse
loading. The maximum lower bound is the limit load.
The upper bound theorem says that if a path of failure exists, the structure
will take the path. It gives upper bounds on the limit load. The minimum
upper bound is the limit load. Since exact limit loads cannot usually be ob-
tained for practical problems, these two theorems enable one to bracket the
answer sufficiently close for engineering purposes.
In addition to the lower and upper bound theorems, the following theorems
are sometimes useful [11].
Addition of weightless material cannot result in a lower collapse load.
Increasing the yield strength of the material in any region cannot weaken
the body.
Residual or thermal stresses or deflections have no influence on the limit
load.
The first two of these are stated in the negative because the converses are
not necessarily true; i.e., addition of material or increase in yield strength
may, or may not, increase the strength of a body, but it will certainly not
decrease it.
The upper and lower bound theorems can be stated in a more precise and
elegant fashion by first introducing the following definitions and concepts [2],
which are useful in their own right.
Generalized stresses Q1 are used to designate the state of stress in a body.
These may be actual stresses, so that Q 1 = ax, Q 2 = ay, etc.; or they may
be moments, as in the case of beams, so that Q1 = M; or resultant forces
and moments as in the case of shells, where Q1 = N 8 , Q2 = Nq,, Q3 = M 8 ,
etc. The choice of generalized stresses to be used depends on the particular
problem and is a matter of convenience and is not necessarily unique. Once
such a choice has been made, however, the corresponding generalized strains
q1 are defined so that the increment of internal work done is given by
(13.5. 1)
For example, for the beam problems considered, the generalized stress at a
hinge is M, and the corresponding generalized strain is the rotation 8, the
internal work at the hinge being equal to M8.
Instead of using the increment of internal work as given by equation
(13.5.1), it is more convenient to introduce the dissipation function or specific
power of dissipation, given by
(13.5.2)
where the dots designate time derivatives. The dissipation function is uniquely
determined by the strain rates (even for singular yield loci [10]), and we can
consequently write
(13.5.3)
Consider an arbitrary structure of volume V and surface S subject to

certain geometric constraints and upon which a load distribution P 1 acts.
Let ,\ be a multiplier and consider the structure under the loads ,\P1 as ,\ is
slowly increased from zero. The safety factor fis defined as the smallest value
of,\ for which the structure can undergo an increase in deformation unrnmu
increase in load; i.e., the safety factor is the ratio of the collapse load to the
actual load.
We now define a statically admissible stress field as one which is in
equilibrium, is in equilibrium with the external loads ,\P~o and
exceeds the yield limit. The multiplier ,\- corresponding to such a
admissible stress field is called a statically admissible multiplier. The
bound theorem can now be stated as follows: The safety factor is the largest
statically admissible multiplier; i.e.,f?:. .\-.
A velocity field is called kinematically admissible if it satisfies the velocity
(or displacement) constraints and if the total external rate of work De done
by the actual loads on this velocity field is positive.
Let the generalized strain-rate vector associated with a given kinematically
admissible velocity field be designated by qf', where the asterisk is used to
indicate that this is not necessarily the true strain-rate vector but one that is
kinematically admissible. The internal dissipation function corresponding to
q'f can be determined from equation (13.5.3) and integrated over the complete
structure to obtain the total internal dissipation D 1• We now define a kine-
matically admissible multiplier .\ + as the ratio of the internal to the external
energy dissipations:
(13.5.4)
The upper bound theorem is then stated as follows: The safety factor is the
smallest kinematically admissible multiplier; i.e., f::;; .\ +. Both theorems can
be summarized by the relation
(13.5.5)
The above theorems furnish upper and lower bounds for the load. The
question of the uniqueness of the solution must still be considered. The
uniqueness of the safety factor follows from equation (13.5.5). It can also
be shown [4, 12, 13], using the principle of virtual work, that the stress field
at the start of plastic flow is unique except in the rigid regions and at stress
points which fall on the straight portions of yield loci (such as the Tresca
locus).
As a simple example of the use of these theorems, consider the indeter-
minate beam shown in Figure 13.5.1. The end A is built in and the end B is
A P 8
~! f f f f f !f f t f f !~~
(a)
(b)
FIGURE 13.5.1 Coilapse of beam with distributed load,

simply supported and the distributed load is of intensity P. The beam will
collapse when yield hinges are formed at A and at some as yet unknown point
designated by x = g, The correct solution can readily be obtained. Let R be
the reaction at the point B. Then the moment at any x is given by
M(x) = R(L - x) - tP(L - x) 2 (13.5.6)
Since yield hinges form at x = 0 and x = g,
M(O) = RL- tPL2 = -M0

(13.5.7)
M(t) = R(L - t) - tP(L - t? = Mo
Also, since at x = g, M must be a maximum,
M'W = -R + P(L- t) = 0 (13.5.8)
Solving equations (13.5.7) and (13.5.8) for R, g, and P,
RL
r = Mo = 2 + 2v.r2 = 4.82843
PL2 -
p = Mo = 6 + 4V2 = 11.65685 (13.5.9)
- g . ;-
g= £ = 2 - v 2 = 0.58579
Equations (15.5.9) give the exact answer to the problem. However, for more
complicated problems the exact solution may become very difficult. An
approximate solution can be obtained by finding upper and lower bounds.
We note first that if some arbitrary value g were chosen for the hinge, an
upper bound would result, for the correct value of g will cause the yield
limit to be reached at a lower load. Let us then arbitrarily choose the mid-
point g = t as the hinge point. Then if the hinge A rotates through an
8, g will rotate through an angle 28 and the total internal work will be
The external load P moves an average distance of !L8 and the external work
Equating the rates of work done, by the upper bound theorem
PL2- 3M
4 - 0
(13.5.11)
PV = 12
or p+ =-
Mo
Comparing with equation (11.5.9), it is seen that an upper bound has been
found which is reasonably close.
A lower bound to P is obtained as follows. Calculating the moment
distribution due to the load P [assuming M(O) = - M 0 ] results in
(13.5.12)
Now this moment distribution is not an admissible one (not statically

admissible), since it gives values greater than M 0 • Thus the maximum value
of m o.ccurs at x/L = i'-2 and is equal to 25/24. If we therefore multiply all
the moments and load by 24/25, an admissible moment distribution, which
by Theorem 2 is a lower bound, will be obtained. Hence
p = 24/25 X 12 = 11.52 (13.5.13)
is a lower bound for the collapse load.

The collapse load is thus bounded by
11.52 ::;; p ::;; 12

or we can write (13.5.14)
p = 11.76 ± 0.24
which is sufficiently accurate for engineering purpose~.
For greater accuracy one could now take the value x/L =h.-, where the
previous moment distribution was a maximum, and using this location for
the hinge position obtain a new and better upper bound and then obtain as
before a new lower bound. The collapse load is the!l found to be
11.65674::;; p::;; 11.65714 (13.5.15)
For the above simple example there is little difference in the amount of
labor between the exact solution and the method of upper and lower bounds.
For more complex problems, however, the method of upper and lower bounds
is far superior.
13-6 METHOD OF SUPERPOSITION OF

MECHANISMS
In the simple examples previously discussed, it was possible to list all

collapse mechanisms and to determine which of these was the correct one.
For complicated structures this is no longer possible and more systematic
methods are needed to determine the limit loads. One such general method
will be briefly discussed in this section. The interested reader is referred to
references [2, 14, 15, 16] for more comprehensive treatments.
The method of superposition of mechanisms due to Symonds and Neal
[17, 18] basically involves combining or superposing various elementary
mechanisms to obtain the collapse mechanism, the true collapse mechanism
being distinguished by the lowest collapse load. The method will be described
by means of an illustrative problem taken from reference [17].
Consider the frame shown in Figure 13.6.1, loaded as indicated. Let the
fully plastic moment in the legs be M 0 and in the horizontal beams 2M0 •
There are 10 possible sections where plastic hinges can occur, as indicated
by the numbers in the figure. The hinge in the beam with the distributed load
may occur at some unknown distance x as shown.
The first step is to determine the number of linearly independent mechanisms
for the structure. It can be shown that this will equal the linearly independent
static equations of equilibrium, which in turn is equal to the number of
bending moments necessary to specify the bending moment diagram com-
pletely minus the degree of redundancy of the structure; i.e.,
n = m- r (13.6.1)
l
4P
p
----------1
15 17 81
X~ I
I I
I I
I I
III lII 12L

I I
I I
110 91
>I,
/
~I<~----2L------~~----2L
FIGURE 13.6.1 Two-bay frame.
Sec. 13-6] Method of Superposition of Mechanisms 319
where n is the number of independent equations of equilibrium or number

of independent mechanisms, m the number of bending moments necessary
to specify the bending moment diagram, and r the number of redundant
constraints. For example, the frame of Figure 13.4.3 has four bending
moments at the points I through 4, it has two redundant constraints, and it
therefore has two independent equations of equilibrium, as given by equations
(13.4.5). It also has two independent mechanisms, as shown in Figure
13.4.3(b) and (c). The mechanism of Figure 13.4.3(d) can be obtained by
combining the other two mechanisms.
For the problem under consideration, Figure 13.6.1, m is equal to 10, as
indicated, and the degree of redundancy r is equal to 6. Hence there are four
independent mechanisms. For these independent mechanisms we choose the
2P
(a) (b)
(c)
FIGURE
1 (d)
13.6.2 Elementary mechanisms.
basic or elementary mechanisms shown in Figure 13.6.2. There are three such
elementary mechanisms defined as follows:
I. Beam mechanisms as shown in (a) and (b). These are characterized by
one hinge at each end and one hinge under each concentrated load and one
hinge for a distributed load.
2. Frame or panel mechanisms as shown in (c). These are characterized
by hinges at the joints and supports but not under loads.
3. Joint mechanisms as shown in (d). These are fictitious mechanisms
representing the small rotation of a joint where three or more members are
connected.
There is also a fourth type of basic mechanism used in the analysis of

gabled frames, called a portal mechanism, which will not be used here.
The next step is to compute the collapse load for each of the elementary
mechanisms of Figure 13.6.2. For·this purpose the work equation obtained
from the principle of virtual work is used. The principle of virtual work
states that if a structure is in equilibrium under a system of external forces,
then the external work done by these forces during any virtual displacement
is equal to the internal work done by the stresses on the strains due to this
virtual displacement, the virtual displacement being any displacement com-
patible with the constraints. For the case of a frame structure composed of
rigid members, this principle takes the form of the following work equation:
(13.6.2)
where P1 are the k external loads, 81 the small displacements of a mechanism

motion, M 1 the plastic moments, and 81 the corresponding small rotations.
The use of this equation is further simplified by the fact that the product M 181
is always positive, so that we need not consider the signs of the M 1 or the 81•
Applying equation (13.6.2) to the mechanisms of Figure 13.6.2 gives for
Figure 13.6.2(a),
2P(L8) = M 0 8 + 2M0 (28) + 2M0 8
or PL = 3.5Mo (13.6.3)
In the above equation the hinge at section 2 was assumed to take place in
the vertical leg, which is weaker, having a plastic moment of M 0 rather than
2M0 as in the beam.
For the mechanism of (b),
4P (Tex) = 2Mo8 + 2Mo (2L2£8_ X

) + Mo 2Lxe_ X
PL = (8£ - x)L M
. 2x(2L- x) 0
For the mechanism of (c),

P(2L8) = 6M0 8
PL = 3Mo
There is no work equation for the mechanism of Figure 13.6.2(d).

Sec. 13-6] Method of Superposition of Mechanisms 321
The frame can fail in one of the three mechanisms analyzed or in some
combination of them. The correct failure mechanism is the one requiring the
smallest load P. Hence in combining the mechanisms we choose only those
combinations which are likely to give smaller loads than the mechanisms
already available. Examination of the work equation (13.6.2) shows that to
reduce the load the work done by the load should be made as large as possible
and the internal work of the plastic moments should be made as small as
possible. This leads to the following rules for combining mechanisms:
1. Two mechanisms should be combined so as to eliminate at least one

common hinge. Only then can the failure load be less than for each of the
individual mechanisms.
2. If necessary a joint connecting three or more members may be rotated
in order to reduce the internal work at· this joint. This corresponds to the
elementary mechanism of Figure 13.6.2(d).
3. For a beam with a distributed load, the hinge is first arbitrarily assumed
to be at the midpoint of the beam. If the actual collapse mode contains this
hinge, its correct position can then be determined. In p'ractice the variation
of this hinge position is small and can be neglected.
Let us apply these rules to the present problem. The mechanism of Figure
13.6.2(c) is most likely to be correct, since it corresponds to the smallest value
of P. If it is combined with mode (a), the hinge at section 2 is eliminated, so
that by rule 1 the resultant value of P should be lower. This combined mode
is shown in Figure 13.6.3. The work equation can be obtained for this new
mechanism of Figure 13.6.3, but it is usually simpler to modify one of the
4P
~-----2L----~~~----2L----~~
FIGURE 13.6.3 Combined mechanism.

previous work equations, here the one for the mechanism of Figure 13.6.2(c),
equation (13.6.5).
P(2UJ) + 2P(L8) = 6M0 8 - MoB + 2M0 (28) + 2M0 8
or PL = 2.75M0 (13.6.6)
This combined mechanism thus gives a lower collapse load and is therefore
more likely than any of the others. To determine whether it is correct, it is
necessary to check the other combinations. Neither mechanism (b) or (d)
of Figure 13.6.2 combines favorably with Figure 13.6.3, since no hinge is
thereby eliminated. However, both of these together will combine with
Figure 13.6.3 to eliminate the hinge of section 5 without introducing one at
10
//
~~~~----2L------~~-----2L------~
FIGURE 13.6.4 Combined mechanism.
section 6. This new combination is shown in Figure 13.6.4. Whether it is

advantageous or not must now be determined.
The work equation for this mechanism is
+ 4P( ~ ) =
8
4P(L8) Mo(58) + 2M0 (68)
Hence
PL = 2.83M0 (13.6.7)
In the above equation the hinge at section 7 was assumed at the midpoint
of the beam according to rule 3. The correct value of x can be determined
from the condition that P should be a minimum. This gives a value of x equal
to 0.982L instead of L. The corresponding value of PL remains the same as
given by equation (13.6.7) to three significant figures.
Sec. 13-7] Limit Design 323
The mechanism yielding the lowest collapse load is the one of Figure
13.6.3, and this is therefore the correct collapse mode, the load being given
by equation (13.6.6). However, to verify that this is the correct solution, it is
necessary to calculate the remaining moments to make sure that none of
them exceeds the yield moment. The four independent equations of equilib-
rium are
2PL = 2M3 - M2 - M4
2PL = 2M7 - Ma - Ms
(13.6.8)
2PL = M 2 - M1 + M 5 - M 1o + M 9 - Ms
0 = M4 + Ms- Ma
where for the second equation x was taken to be equal to L. Substituting
Ma = 2Mo
Ms = -Mo
into equations (13.6.8), and solving for the remaining moments gives
Ma = -Mo (13.6.9)
Thus none of the remaining moments exceeds the fully plastic moment for
that member, so that the failure mode of figure 13.6.3 is the correct one.
The above example illustrates the use of the method of superposition of
mechanisms. Obviously the method requires some skill in combining favor-
able mechanisms. Such skill can readily be acquired by experience and
practice. The method then offers a relatively simple procedure for finding the
collapse load of framed structures under proportional loading.
13-7 LIMIT DESIGN
Heretofore the problems discussed have been primarily those of analysis.

A given structure was analyzed to determine the maximum safe load. Of
equal importance is the problem of design where the loads are specified and
the best structure, in some sense, to carry these loads must be determined.
It should be first noted that for some design problems the methods pre-
viously discussed can be used directly: For example, if M 0 is the plastic
moment in one of the beams of the structure which is considered as a reference
beam and a;Mo are the plastic moments in the other elements of the structure,
where a1 corresponds to the ith element and the a1 are all specified, then the
design problem becomes one of finding the value of M 0 for collapse to occur
under the specified loads. This is called a restricted design problem [14]. To
find M 0 , the upper and lower bound theorems can be used as before, noting
that the lower bound theorem now gives an upper bound on the required value
of M 0 and the upper bound theorem gives a lower bound on the value of M 0 •
Thus in using the method of superposition of mechanisms of Section 13.6,
instead of combining mechanisms to obtain the lowest collapse load, we
combine the mechanisms to obtain the largest plastic moment. The procedures
for both types of problems are of course exactly the same.
For the illustrative problem of Section 13.6, as shown in Figure 13.6.1, the
highest value of the plastic moment M 0 is obtained for the combined mecha-
nism of Figure 13.6.3 and from equation (13.6.6) is given by
M0 = 0.364PL (13.7.1)
The choice of sections to be used in constructing the frame of Figure 13.6.1

is thus determined by equation (13.7.1).
In general, however, the designer seeks not only a safe design but one which
will also minimize the cost or the weight. It is to be realized that a minimum
weight design does not always result in a minimum cost design. However,
because of the many factors entering into the cost of a design, the methods
that have been developed for optimum design are primarily based on the
minimum-weight criterion. In any case this furnishes a good starting point
for the designer.
A design method for minimum weight was first discussed by Heyman [19],
followed by Foulkes [20], Livesley [21], and Heyman and Prager [22]. The
basic problem is one in linear programming, where one seeks to minimize a
weight function subject to inequality constraints [23]. However, the number
of variables and constraints is so large, except for relatively simple frames,
that a straightforward use of linear programming techniques would quickly
exhaust the capacity of even a large computer. The method of Heyman and
Prager [22], however, does not require as much computer storage and is
completely automatic. Furthermore, it can be used for analysis as well as
design. A complete description of the method is given in reference [22] and in
reference [14].
Problems
1. Consider a truss similar to the one shown in Figure 13.2.1, where the
between the vertical bar and each of the other two bars is 60°. Find
elastic and plastic safety factors if F = 20,000 lb.
References 325
2. Calculate the shape factor for a beam of circular cross section.

3. Determine the collapse load for the indeterminate beam of Figure 13.4.2
if the concentrated load F is located at a distance tL from the built-in end.
4. Obtain the collapse load for a beam built in at both ends with a concen-
trated load at the center.
5. Find the collapse load for the frame of Figure 13.4.3 if the vertical and
horizontal loads are equal.
6. Repeat Problem 5 with the vertical load equal to twice the horizontal load.
7. Determine the collapse load for a uniformly loaded beam built in at both
ends.
8. For the problem of Figure 13.6.1, show that if the distributed load is 5P
instead of 4P, the other loads remaining unchanged, the frame will fail by
the mechanism of Figure 13.6.4 at a load of 2.61M0 /L.
9. Repeat Problem 8 assuming the distributed load is 8P. Show that the failure
mechanism is the one of Figure 13.6.2(b) at a load of 1.75M0 /L.
10. Derive the equilibrium equations (13.6.8).
References
1. G. Kazinczy, Experiments with Clamped Girders, Betonszemle, 2, Nos. 4,
5, and 6, 1914.
2. P. G. Hodge, Jr., Plastic Analysis of Structures, McGraw-Hill, New York,
1959.
3. A. A. Gvozdev. The Determination of the Value of the Collapse Load for
Statically Indeterminate Systems Undergoing Plastic Deformation, Proceed-
ings of the Conference on Plastic Deformations, Akademiia Nauk SSSR,
Moscow, 1938, pp. 19-33. Translated into English by R. l'v!. Haythorn-
thwaite, Intern. J. Mech. Sci., 1, 1960, pp. 322-355.
4. R. Hill, On the State of Stress in a Plastic-Rigid Body at the Yield Point,
Phil. Mag., 42, 1951, pp. 868-875.
5. R. Hill, A Note on Estimating the Yield-Point Loads in a Plastic-Rigid
Body, Phil. Mag., 43, 1952, pp. 353-355.
6. D. C. Drucker, H. J. Greenberg, and W. Prager, The Safety Factor for an
Elastic-Plastic Body in Plane Strain, J. Appl. Mech., 18, 1951, pp. 371-378.
7. D. C. Drucker, W. Prager, and H. J. Greenberg, Extended Limit Design
Theorems for Continuous Media, Quart. Appl. Math., 9, 1952, pp. 381-389.
8. H. J. Greenberg and W. Prager, Limit Design of Beams and Frames, Trans.
ASCE, 117, 1952, p. 447. First published as Tech. Rept. Al8-J, Brown Univ.
Press, Providence, R.I., 1949.
9. M. R. Horne, Fundamental Propositions in the Plastic Theory of Structures,
J. Inst. Civil Engrs. (London), 34, 1949-1950, pp. 174-177.
10. W. Prager, An Introduction to Plasticity, Addison-Wesley, Reading, Mass.,
1959.
11. D. C. Drucker, Limit Analysis and Design, Appl. Mech. Rev., 7, No. 10,
1954, pp. 421-423.
12. R. Hill, On the Problem of Uniqueness in the Theory of a Rigid-Plastic
Solid, J. Mech. Phys. Solids, 4, 1956, pp. 247-255; 5, 1957, pp, 1-8, 153-161,
302-307.
13. R. M. Haythornthwaite and R. T. Shield, A Note on the Deformable

Region in a Rigid-Plastic Structure, J. Mech. Phys. Solids, 6, 1958, pp.
127-131.
14. C. E. Massonnet and M. A. Save, Plastic Analysis and Design, Vol. I, Ginn
(Blaisdell), Boston, 1965.
15. B. G. Neal, The Plastic Methods of Structural Analysis, Wiley, New York,
1956.
16. J. F. Baker, M. R. Horne, and J. Heyman, The Steel Skeleton, Vol. II,
Cambridge Univ. Press, New York, 1956.
17. P. S. Symonds and B. G. Neal, Recent Progress in the Plastic Method of
Structural Analysis, J. Franklin Inst., 252, 1951, pp. 383-407, 469-492.
18. B. G. Neal and P. S. Symonds, The Rapid Calculation of the Plastic Collapse
Load of a Framed Structure, Proc. Inst. Civil Engrs. (London), 1, 1952,
pp. 58-71.
19. J. Heyman, Plastic Design of Beams and Plane Frames for Minimum
Material Consumption, Quart. Appl. Math., 8, 1951, pp. 373-381.
20. J. Foulkes, The Minimum Weight Design of Structural Frames, Proc. Roy.
Soc. (London), Series A223, 1954, pp. 482-494.
21. R. K. Livesley, The Automatic Design of Structural Frames, Quart. J.
Mech. Appl. Math., 9, 1956, p. 257-278.
22. J. Heyman and W. Prager, Automatic Minimum Weight Design of Steel
Frames, J. Franklin Inst., 266, 1958, pp. 339-364.
23. G. B. Dantzig, Linear Programming and Extensions, Princeton Univ. Press,
General References
Baker, J. F., M. R. Horne, and J. Heyman, The Steel Skeleton, Vol. II, Cambridge
Univ. Press, Cambridge, 1956.
Commentary on Plastic Design in Steel, Progress Reports No. 1 and No. 2 of the
Joint WRC-ASCE Committee on Plasticity Related to Design, Proc. ASCE,
85, No. EM3, 1959.
Drucker, D. C., Plastic Design Methods, Advantages and Limitations, Brown
Univ; Div. of Appl. Math. Tech. Rept. No. 24, 1957.
Hodge, P. G., Jr., Plastic Analysis of Structures, McGraw-Hill, New York, 1959.
Massonnet, C. E., and M. A. Save, Plastic Analysis and Design, Vol. 1, Ginn
(Blaisdell), Boston, 1965.
Neal, B. G., The Plastic Methods of Structural Analysis, Wiley, New York,
2nd ed., 1963.
Van den Broek, J. A., Theory of Limit Design, Trans. ASCE, 105, 1940, pp.
638-661.
CHAPTER 14
CREEP
14-1 BASIC CONCEPTS
When a material stretches under constant load the phenomenon is called

creep. This was briefly mentioned in Section 2.4, where some typical creep
curves were shown. A test which is carried out at constant load is therefore
called a creep test and the measured strains are called creep strains. The slope
of the creep curve (strain versus time) at any point is called the creep rate.
The shape of the uniaxial creep curve as shown in Figure 14.1.1 has led to
a subdivision of the creep curve into three parts.
1. Primary creep, where the creep rate is decreasing rapidly.
2. Secondary (or steady state) creep, where the creep rate is essentially
constant.
3. Tertiary creep, where the creep rate increases very rapidly.
Since the primary part of the creep curve is usually of short duration and
the tertiary part leads quickly to rupture, the greatest interest usually lies
in the secondary part, although the primary part cannot always be neglected.
The classical creep experiments were carried out by Andrade [1]. in 1910.
Andrade concluded that at constant temperature and constant stress the
creep strain can be represented by an equation of the form
(14.1.1)
327
Creep [Ch. 14
328
Time
FIGURE 14.1.1 Typical uniaxial creep curve.
Equation (14.1.1) can also be approximated by writing
8 = 81 + 82
where 81 = (:3t1/3 (14.1.2)

82 = kt
8 and 8 are now recognizable as the primary and secondary components of

1 2
the total creep strain, for if we differentiate them with respect to time, we get
81 = tf:3t-2's (14.1.3)
82 = k
We see then that 81 is large for small times but becomes vanishingly small for
large times, whereas 82 is constant with time.
Many experiments on different materials have shown, however, that the
exponent i in equation (14.1.3) is not adequate and can vary between 0.4 and
0.85. Therefore, the relationship generally used is of the form
(14.1.4)
where q and (:3 depend on the material.

The constants (:3, k, and q will in general depend on the stress and the
temperature. The assumption is usually made, however, that q is independent
of the stress and that at constant temperature we can write
81 = (:3(a)tq (14.1.5)
e2 = k(a)t
Sec. 14-1] Basic Concepts 329
Equations (14.1. 5) imply that the curves for different stress levels are
geometrically similar, and this is approximately true.
In particular, the functions f3(a) and k(a) are frequently taken to be power
functions of the stress [2, 3]; i.e.,
{3( a) :;: Bam

(14.1.6)
k(a) = Kan
Such expressions often fit the data fairly well. For example, for a gas turbine-
alloy steel (Allegheny 418, 12% Cr, 3% W) the data could be fitted fairly
well by assuming
e1 = 3.36 X I0- 30 a 6t 112

(14.1.7)
e2 = 4.41 X I0- 32 a 6 •2 t
In performing calculations, it is the strain rate at any time which is of

importance. To determine the strain rate, we differentiate equations (14.1.5)
to get
81 = qf3tq-1
(14.1.8)
82 = k(a)
Alternatively, since from (14.1.5),
- (61)1/q
t- II (14.1.9)
we can substitute (14.1.9) into (14.1.8) to eliminate t, giving
(14.1.10)
Thus the creep rate can be written as a function of stress, temperature, and
time, or of stress, temperature, and strain.
Basically, however, expressions of this type pose a fundamental difficulty.
For a relation of the form
8 =f(a, T, t, e) (14.1.11)
implies that there exists a "mechanical equation of state." That is, the creep
rate at any time depends on the state of the system at that time and is inde-
pendent of how or by what path the system got to this state. There is con-
clusive evidence that this can generally not be true but may be true under
330 Creep [Ch. 14
certain restricted conditions. This is similar to the problem of using deforma-

tion theory in plasticity calculations. However, because there is at present
no other theoretical method that is practical, equations such as (14.1.11) are
generally used. However, the limitations should be kept in mind.
Creep data are almost always taken at constant stress. The question then
arises: How are these data to be used for the case where the stress is varying
with time? For example, suppose the stress in a body is constant at a value
a 1 up to a time t 1 • The stress then suddenly is changed to a 2 • At what rate will
the body begin creeping at the new stress a 2 ? To answer this question several
different cumulative creep laws have been proposed [11]. Three such laws are
illustrated in Figure 14.1.2. These are the strain hardening, time hardening, and
life-fraction laws. The strain-hardening law assumes that in going from one
stress level to the next the creep rate depends on the existing strain in the
material as indicated in Figure 14.1.2(b). The time-hardening rule assumes
that the creep rate depends upon the time from the beginning of the creep
5 35,000 35,000
30,000
~--
25,000 25,000
I 20,000 20,000
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
Time, hr Time, hr
(a) (b)
0 20 40 60 80 100 120 140

Time, hr
(c)
FIGURE 14.1.2 Proposed creep laws: (a) time-hardening rule; (b)

hardening rule; (c) life-fraction rule.
Sec. 14-2] Multidimensional Problems 331
process as shown in Figure 14.1.2(a). The life-fraction law assumes that the
creep rate depends upon the fraction of life used up, as shown in Figure
14.1.2(c).
Experimental data seem to lean toward the strain-hardening rule. However,
it has also been shown that if the stress is suddenly changed, a transient effect
takes place such that the strain rate is higher than that obtained even from
the strain-hardening rule. It has therefore been suggested by Rabotnov [4]
that a modification be made in the strain-hardening rule, as shown in Figure
14.1.3. Instead of moving from 0 to A, as in the strain-hardening rule, we
0"2
Time
FIGURE 14.1.3 Transient effect on creep rate.
move to a point D which has a higher strain rate. The point Dis taken as the
point at which the strain is a1 /a 2 times the strain at A. The validity of this
type of approach requires much more experimental verification.
If one uses the time-hardening rule, then the creep rates at any stress are
given by equations (14.1.8). On the other hand, if the strain-hardening rule
is used, the creep rates are given by equations (14.1.10). We note that for
steady-state or secondary creep, all the rules become the same, since all the
stress curves are then parallel straight lines. (This is not evident from Figure
14.1.2, since the primary parts of the curves have been greatly exaggerated).
14-2 MULTIDIMENSIONAL PROBLEMS
As mentioned before, all creep data are of the uniaxial type. How then do
we use this data for two-dimensional and three-dimensional problems? The
answer is, the same way as for any plasticity problem. Although the physical
processes involved in creep are undoubtedly different than in ordinary plastic
flow, we assume that the same relations hold for creep strains as for plastic
strains. For example, the Prandtl-Reuss relations can be used for computing
332 Creep [Ch. 14
the creep increments. Thus we assume an equivalent stress defined the same
way as in plasticity theory and an equivalent creep strain increment and write
etc. (14.2.1)
The computations are then performed as for any plasticity problem. Instead
of using the stress-strain curve, however, the creep curves, or equations such
as (14.1.8) or (14.1.10) representing these curves, are used. e0 and a. in these
equations are the equivalent strain rate and equivalent stress, respectively.
The successive-approximation method is very useful in making the calcula-
tions.
In general, solutions are obtained incrementally. One starts with a given
increment of time and solves the problem by successive approximations. The
next increment of time is then taken and the process repeated. The solution
can thus be extended to any time. For this purpose it is convenient to write
equations (14.1.8) as follows [using equation (14.1.6)]:
l::!ec = qBa':tq-l M
(14.2.2)
or l::!e 0 = Ka~ M
or, for the case of the strain-hardening rule (14.1.1 0),
(14.2.3)
We note that in the case of steady-state creep, the incremental approach is

not necessary, since the stress is constant and relations (14.1.8) and (14.1.10)
can be integrated directly (if we neglect elastic strains).
With this brief exposition of the creep equations we shall proceed to
several examples. The reader interested in more detailed discussions of creep
laws and their physical and phenomenological background is referred to the
general references at the end of the chapter.
Two examples will be considered. The first is the uniaxial problem of a
thin infinite strip with a parabolic temperature distribution across the width.
This is exactly the same as the plastic flow problem treated in Section 9.2.
Here we will show how the stresses relax with time due to creep. The second
problem discussed is that of a rotating disk similar to the problem of Section
9.5. The same method of solution, by successive approximations, will be used
here for the creep problems as was used in Chapter 9 for the plastic flow
problems. Both these problems are taken from reference [5].
Sec. 14-3] Uniaxial Creep in Infinite Strip 333
14-3 UNIAXIAL CREEP IN INFINITE STRIP
Consider the simple uniaxial case of a thin infinite plate of width 2c as

shown in Figure 14.3.1 with a temperature distribution T(y) across the
y=-c
FIGURE 14.3.1 Flat plate with temperature distribution.
width. Under these conditions, the only nonzero stress is
(14.3.1)
As in the usual theory of bending, it is assumed that plane sections remain

plane. This requires that
(14.3.2)
where c1 and c2 are constants to be determined. Let ex and ax be the total

strain and stress at the middle of the time interval M, e~ the total creep strain
up to time t, and ile~ the additional increment of creep strain during the time
interval M. Then
(14.3.3)
from which
aX =E (
eX - aT - eX0 -
ile")
~
2 (14.3.4)
The boundary conditions require that
fc axdY =0
(14.3.5)
fc axydy = 0
334 Creep [Ch. 14
The constants c1 and c2 of equation (14.3.2) can be determined by substituting

equations (14.3.2) and (14.3.4) into (14.3.5). Substituting these values of c1
and c2 back into equation (14.3.2) and assuming E to be constant results in
e =Bx - aT= 21 JcC -c

(aT+ B~ + !~eDdy
+ -23y3 Jc (aT+ e~ + !~e~)y dy - aT (14.3.6)

C -c
As a specific example, let
T = T 0 + 600(y 2 - t)
E = 28 x 10 6
(14.3. 7)
c= 1
a = 9.5 X 10-s
Substituting these values into equation (14.3.6) and noting the symmetry of
the problem results in
e = -0.0057(y 2 - -!-) + f (e~ + !~e~)dy (14.3.8)
This equation, along with the stress-strain relation (14.3.4) and some relation
between stress and creep rate, are all that are needed to solve the plate
problem. Let it be assumed that the relation between stress and creep rate is
of the form
(14.3.9)
Note that in equation (14.3.9) the sign of ~e~ must be taken the same as the
sign of ax, as indicated. It will generally be necessary to write the creep law
in this form unless the stress exponent is an odd integer. The procedure for
obtaining the solution to this problem is now as follows:
1. At the start of the first time interval M, e~ is known to be zero and
~e~ is assumed to be zero. Substituting these values into (14.3.8) gives the
elastic solution as a first approximation to the total strains.
2. Substitute this first approximation for the strain distribution into the
stress-strain relation (14.3.4) and solve for the first approximation to the
stress distribution.
3. Substitute this first approximation to the stress distribution into the
creep relation (14.3.9) and solve for the second approximation to the incre-
mental creep strains during the first time interval.
Sec. 14-4] Creep in Rotating Disks 335
4. These incremental strains are substituted into equation (14.3.8) and the
iteration proceeds from equation (14.3.8) to (14.3.4) to (14.3.9) back to
(14.3.8), etc., until the procedure converges to the correct set of incremental
strains and stresses.
5. At the start of the second time increment, the total strains e~ are now
known and are equal to the incremental creep strains developed during the
previous time increment. In fact, the total creep strains at the beginning of
any time interval will always be known and will be equal to the accumulated
incremental strains up to that time interval. The procedure for calculating
the average stresses and strains for the second or any other time interval is
then the same as in steps 1, 2, 3, and 4.
The results of this calculation are presented in Figure 14.3.2. This figure
y =0 center
"'
Q.
10 100 1,000 10,000

Time, hr
FiGURE 14.3.2 Stress relaxation in plate.
shows a plot of the variation of the stress with time at the center and at the
edge of the plate. For this problem as well as the subsequent ones, the
unknown integrals were evaluated numerically using Simpson's rule.
14-4 CREEP IN ROTATING DISKS
The simple example of the fiat plate of Section 14.3 involves uniaxial stress,
making it possible to determine directly the creep strains once the total
stresses and strains are known. In most cases of practical interest the stresses
are biaxial or triaxial in nature. The general procedure is then the same as
discussed in Chapter 9. As an example we consider the case of creep in a
rotating disk.
336 Creep [Ch. 14
We assume that the von Mises yield criterion and the Prandtl-Reuss
stress-strain relations are valid for both plastic flow and creep, as indicated
by equation (14.2.1). For problems of thin disks with radial symmetry these
relations are
(14.4.1)
where as usual
(14.4.2)
The increment in equivalent strain !le 0 will in general be a function of the

equivalent stress a., the total equivalent strain e0 , the temperature T, the time
t, and the strain history of the material. Such a general relationship can be
written
(14.4.3)
For problems of plastic flow without creep, the increment of equivalent

strain !le 0 is a function of the equivalent stress a. and the strain history or
loading path. For creep problems the strain path becomes time-dependent.
For this time dependency of the strain path a number of laws have been
proposed such as discussed in Section 14.1. For the examples to be presented
we shall use both the time-hardening and strain-hardening rules, although
any other law can be used as well.
We proceed essentially as in Section 9.5. The equilibrium equation for the
disk is given by
(14.4.4)
where h is the thickness, p the density, and w the rotational speed. The com-
patibility equation in terms of stresses is derived in Section 9.5 and in the
notation used here is
-d (aa
dr E
- -E + a T + sao +
- JLar A
usac)
1 + p, a, - aa e~ - e~ !le~ - !leg
= ---y- --r-- + --r- + r
Equations (14.4.4) and (14.4.5) must now be solved simultaneously for the
stresses ar and a 8 subject to the proper boundary conditions for the disk. At
any time t the total accumulated creep strains e~ and eg up to that time t are
known. The increments in creep strain lle~ and !leg for the next time interval
Llt are not yet known, but a set of values is assumed (such as the incremental
strains computed for the previous time interval). Since all the creep strains
now have known values, equations (14.4.4) and (14.4.5) form a linear pair of
equations which can readily be solved for the stresses ar and a 8 • From the
assumed values of lle~ and !leg, lle0 is calculated by the first of equations
(14.4.2). This value of flee corresponds to a particular value of a 8 as given by
equation (14.4.3). From these values of flee and a 8 as well as the stresses
computed from equations (14.4.4) and (14.4.5), new values for the creep
increments Lle~ and !leg are now computed from equations (14.4.1). These
better approximations to the incremental creep strains are put into equations
(14.4.4) and (14.4.5) and the process is repeated until convergence is obtained.
Equations (14.4.4) and (14.4.5) may be solved in a number of ways once
the incremental strains are assumed to be known. Two methods have been
described in Section 9.5, the finite-difference method and the integral-equation
method. The finite-difference method is generally preferable here, since the
disk thickness and other dimensions are changing with time. In connection
with this, it should be noted that since a disk under creep conditions will grow
and change dimensions with time, the values of r and h appearing in equations
(14.4.4) and (14.4.5) should be the true values at the time t and not the
original values at zero time. Thus if H is the original thickness at the
original radial position R, the current values h and r are given approximately
by [see equations (9.5.25) and (9.5.26)]
h= H
(1+ eg)(l + e~) (14.4.6)
r= R(l + eg)
Furthermore, the strains appearing in the above equations should be natural

strains. However, since the strain increments will generally be small and
the current dimensions of the disk will always be used, the error in using con-
ventional strains will be negligible.
Several cases of creep in rotating disks will now be considered. The first
case treats the same problem as that presented by Wahl et al. [6]. Since
reference [6] neglects the transient condition, the same will be done here, in
order to compare the results of the present method with those of that refer-
ence. This is done by neglecting the elastic strains, since the compatibility
relation (14.4.5) then becomes independent of time if the creep rate is given
338 Creep [Ch. 14
by a function of stress times a function of time. The same problem will then
be treated including the elastic strains, to determine the effect of neglecting
the transient stress distribution upon the creep strains. The second case will
treat a similar disk using more complicated creep laws. The solution to the
problem will thus be presented using both the time-hardening and strain-
hardening rules.
Case l(a). Creep in Rotating Disk Neglecting Transient Condition
Consider a constant-thickness constant-temperature disk of 12-in. outside

diameter and 2.5-in. inside diameter rotating at 15,000 rpm. Assume steady-
state creep with the creep rate as used in reference [6):
Neglecting the elastic strains is equivalent to letting E approach infinity in

equation (14.4.5), and since the stress distribution now becomes independent
of time, we can arbitrarily choose any time interval b.t and compute the
stresses. The strains will, of course, be directly proportional to the assumed
time interval b.t. One method of solution is therefore as follows. Solve
equation (14.4.1) for the stresses. This gives
(14.4.7)
Letting E approach infinity, integration of equation (14.4.5) and substitution

into (14.4.7) results in
(14.4.8)
where a is the inner radius of the disk.

Integrating (14.4.4) and substituting (14.4.8) into the resultant equation
gives
(14.4.9)
where Cis determined from the rim loading a,(b) and is given by
(14.4.10)
and b is the outer radius of the disk.

The solution now proceeds as follows:
I. Assume a value of l:lt, say, 1 hour.

2. Assume values for l:le~ and !leg, say, 0.001 everywhere.
3. Calculate
4. Calculate
l:le )l/6.2
ae = ( 4.41 X 10- 32 f:lt
5. Calculate C from (14.4.10).

6. Calculate a, and a8 from (14.4.8) and (14.4.9).
7. Calculate new values for l:le~ and !leg from (14.4.1).
8. Go back to step 3.
The results of carrying out this calculation are shown in Figure 14.4.1 and
14.4.2 and compared with the results of reference [6]. It is seen from the
2 3 4 5 6
Radius, in.
FIGURE 14.4.1 Creep stress in rotating disk neglecting elastic strains.

340 Creep [Ch. 14
0.04 o successive approximations

--Ref.(6)
c
~0.02
2 3 4 5 6
Radius, in.
FIGURE 14.4.2 Creep strains in rotating disk neglecting elastic strains.
figures that the agreement is excellent. The strains shown in Figure 14.4.2
were obtained by multiplying the strains computed using a M of 1 hour
by 180.
Case l(b). Creep in Rotating Disk Including Transient Conditions
The same disk as in Case 1(a) was considered, with the elastic strains
included this time. The modulus of elasticity E was taken to be 18 x 10 6 ,
corresponding to a disk temperature of 1000°F. The creep rate was the same
as previously given for Case 1(a). To start with, a time interval M of0.01 hour
was chosen, and as a first approximation the creep increments l:leie and l:lefe
were all taken to be constant at 0.00001, the total creep strains e~ and eg of
course being zero at zero time. The calculation then proceeded as follows:
1. !:lee was computed from the relation
2. Equations (14.4.4) and (14.4.5) were solved for the stresses ar and a8
by the method described in Section 9.5.
3. a. was computed from the relation
Clie )1/6.2
ae = ( 4.41 X 10 32 l:lt
4. New approximations were obtained for the strain increments l:le~ and
!:leg from (14.4.1).
5. Steps I through 4 were repeated until there was no change in two
successive computations of the strain increments.
The creep strains were thus computed at the end of O.OI hour. To obtain
the creep increments for the next O.OI hour the same procedure was followed,
except that the total creep strains e~ and eg were no longer zero but were
equal to the accumulated creep strains up to that time. A first approximation
to the creep increments <1e~ and <1eg was assumed (usually the values obtained
for the previous time interval), and steps I through 5 repeated. In this way
the incremental and total creep strains were computed up to I80 hours. The
time interval was arbitrarily chosen so that the stress during any interval
would not drop by more than approximately I,OOO psi. Thus this time
interval M was rapidly increased as the stress approached steady-state
conditions.
The results of this computation are shown in Figures I4.4.3 through I4.4.5.
Figure I4.4.3 shows the relaxation of the tangential stress with time at the
----neglecting transient conditions
8. 40
vi
V>
"'
~ 30 _____________ __ _
:~_:::_:::_'='-=---~-='-:::::::
20 tangential stress
at rim
10
0.001
Time, hr
FrouRE 14.4.3 Stress relaxations in rotating disk, case l(b).
inner and outer radii of the disk. It is seen that steady-state conditions are
obtained at about 10 hours and that the tangential stress at the inner surface
persists for a long time at a slightly higher value than that computed neglecting
the transient stress distribution.
The tangential and axial creep strains at the end of 180 hours are shown
plotted in Figure I4.4.4 together with the scatter band of test results given in
reference [6]. From Figure I4.4.3 it can be seen that after a long period of
stability the stresses start rising. As the disk grows due to creep, the centrifugal
loading increases and the thickness at the center decreases, as can be seen
342 Creep [Ch. 14
- - including transient
conditions
~ scatter- bond test
results, Ref. (6)
2 3 4 5 6
Radius, in
FIGURE 14.4.4 Strain distribution at 180 hours, case l(b).
i6.40[
e 6.20
<lJ
"0
".5
8 6.00 ----------------~
:~ 1.65~
2 1.45
<lJ
"0
·v;
E 1.25 ---------------~
.~ 1.00l
if
+-
0.90
£:~
f---Vl
·"o.so L---~-L-LJ-LILllui~~----L-~1~1-Lt~l~t~l~t~I~~--J-J-~I~I~I~tl
1 10 100 1,000
Time, hr
FTouRE 14.4.5 Variation of disk dimensions with time, case l(b).
from Figure 14.4.5. This increase in centrifugal loading and decrease in cross
section will eventually balance and finally surpass the stress relaxation, due
to the creep. The stresses will start increasing all over the disk and the disk
may start creeping at an accelerating rate. This is illustrated in Figure 14.4.3,
which shows that after a long period of stability the stresses gradually start
rising. The disk has now reached an unstable condition which may result in
failure.
Case 2. Nonlinear Creep Laws
In the previous examples the creep curves were assumed linear with the
creep rate independent of time. It will now be shown that more complicated
creep curves and creep laws can be used without any appreciable increase in
the complexity of the computations. Any nonlinear creep law can be used;
for illustration it will be assumed that the creep data can be represented by
(14.4.11)
The same disk as in case 1 will now be considered, with the creep data given
by equation (4.4.11 ), using both the time-hardening and the strain-hardening
rules previously mentioned.
For the time-hardening rule, since the creep rate depends on the actual
time elapsed, the creep rate at constant stress at the middle of the time
interval between t and t + l1t is obtained directly from (14.4.11):
!J.t)q -1
Ae0 = qBa~ (t +2 At (14.4.12)
where ae is the stress at the middle of the time interval and is assumed to be
constant during the interval. Solving (14.4.12) for a. gives
= (~)1/m ( !J.t)<l-q)/m
ae qB !J.t t + 2
(14.4.13)
For the strain-hardening rule, the creep rate depends upon the total
accumulated creep rather than the accumulated time. Eliminating the time
between equations ( 14.4.11) and (14.4.12) results in
_ -lim ( Llec )qfm ( Aec)(l-q)/m

ae - B q Llt ea +2 (14.4.14)
where e0 is the total accumulated creep strain up to the beginning of the time
interval under consideration and Lle0 is the strain increment during the time
interval. The solution to this problem is now obtained in exactly the same
way as for case l(b), except that in step 3, a. is computed by equation (14.4.13)
344 Creep [Ch. 14
for the time-hardening rule and by equation (14.4.14) for the strain-hardening
rule. For the time-hardening rule, track must be kept of the total elapsed
time, whereas for the strain-hardening rule, track must be kept of the total
accumulated strain.
Part of the results of this calculation for m equal to 6, q equal to ~' and
B equal to 1.5 x 10-ao are presented in Figures 14.4.6 and 14.4.7.
The creep curves represented by these constants are shown in Figure
14.4.8. As would be expected, Figure 14.4.6 shows that the stress does not
relax quite as rapidly using the strain-hardening rule as with the time-
hardening rule, since the creep rates will generally be lower for the strain-
hardening rule. This can be used to explain the results of Figure 14.4.7, which
- - time hardening
----strain hardening
40 .... ,.....
------ ---
'-.-. tangential stress at bore
·u;
-----------
0.
tangential stress ot rim
Time, hr
FIGURE 14.4.6 Stress relaxation for strain-hardening and time-hardening rules.
Time, hr
FIGURE 14.47 Creep at bore for strain-hardening and time-hardening rules.

References 345
120 140
Time, hr
FIGURE 14.4.8 Creep curved used with strain-hardening and time-hardening

rules.
shows that the tangential creep strain is essentially the same for both rules.
(Although the strain-hardening rule actually gave slightly lower creep strains,
the difference is too small to show up on the figure.) Although as seen from
Figure 14.1.2, in going to a given stress the creep rate will be smaller for the
strain-hardening rule than for the time-hardening rule, this is compensated
for in this problem by the fact that the stress does not relax as fast, so that
the total creep turns out to be about the same using either rule.
A comprehensive discussion of creep in rotating disks, including the effects
of the transient period, can be found in a series of papers by Wahl [6, 7, 8, 9,
10].
References
1. E. N. da C. Andrade, On the Viscous.Flow of Metal and Allied Phenomena,
Proc. Roy. Soc. (London), A84, 1910, p. 1.
2. F. H. Norton, Creep of Steel at High Temperatures, McGraw-Hill, New
York, 1929, p. 67.
3. R. W. Bailey, The Utilization of Creep Test Data in Engineering Design,
Proc. Jnst. Mech. Engrs., 131, London, 1935, pp. 186-205, 260-265.
4. Yu. N. Rabotnov, On the Equation of State of Creep, Proceedings of the
International Conference on Creep, Vol. 2, New York-London, 1963, p. 117.
346 Creep [Ch. 14
5. A. Mendelson, M. H. Hirschberg, and S. S. Manson, A General Approach

to the Practical Solution of Creep Problems, Trans. ASME, 81D, 1959,
pp. 585-598.
6. A. M. Wahl, G. A. Sankey, M. J. Manjoine, and E. Shoemaker, Creep
Tests of Rotating Disks at Elevated Temperature and Comparison with
Theory, J. Appl. Mech., 21, 1954, pp. 222-235.
7. A. M. Wahl, Analysis of Creep in Rotating Disks Based on the Tresca
Criterion and Associated Flow Rule, J. Appl. Mech., 23, 1956, pp. 231-238
8. A. M. Wahl, Stress Distribution in Rotating Disks Subjected to Creep
Including Effects of Variable Thickness and Temperature, J. Appl. Mech.,
24, 1957,pp, 299-305.
9. A. M. Wahl, Further Studies of Stress Distribution in Rotating Disks and
Cylinders Under Elevated Temperature Creep Conditions, J. Appl. Mech.,
25, 1958, pp, 243-250.
10. A. M. Wahl, Effects of the Transient Period in Evaluating Rotating Disk
Tests Under Creep Conditions, J. Basic Eng., 85, 1963, pp. 66-70.
11. H. R. Voorhees and J. W. Freeman, Notch Sensitivity of Heat-Resistant
Alloys at Elevated Temperature, Wright Air Development Center Tech. Rept.
No. 54-175, Part I, 1954.
General References
Arutyunyan, N. Kh., Some Problems in the Theory of Creep, translated by H. E.
Nowottny, Pergamon Press, London, 1966.
Finnie, J., and W. R. Heller, Creep of Engineering Materials, McGraw-Hill,
New York, 1959.
Hult, J., Creep in Engineering Structures, Ginn (Blaisdell), Boston, 1966.
Odquist, F. K. G., Mathematical Theory of Creep and Creep Rupture, Oxford
Univ. Press, London, 1966.
INDEX
Affine transformation, 46 Characteristics, 285-98

Alexander, J. M., 299 fixed, 295
Allen, D. N. de G., 163 method of, 288
Andrade, E. N. daC., 327, 345 region of influence, 288
Anisotropy, 13, 14 slip lines, 296-98
Arutyunyan, N. Kh., 346 Clough, R. W., 233
Associated flow rule, 115 Compatibility equations, 60, 165, 265,
287,292
Bailey, R. W. 345 Compression test, 13
Baker, J. F., 326 Coulomb, C. A., 71
Baldwin, W. M., 22 Coulomb's theory, 73
Bauschinger effect, 13 Creep, 16, 327-45
theories of, 13, 14 of infinite strip, 333-35
Beam(s) laws, 330, 331
indeterminate, 308-10 life-fraction, 330-331
pure bending, 305-307 strain-hardening, 3 30-3 31
with tip load, 308 time-hardening, 330-331
Beltrami's energy theory, 74 multidimensional problems, 331-32
Biharmonic operator, 217 primary, 327-29
Bland, D. R., 119, 133, 134, 163 rotating disks, 335-45
Boundary-value problems, 276, 293 secondary, 327-29
numerical solution, 276-78, 294-96 tertiary, 327
Bridgman, P. W., 16, 22 Creep curves, 16, 328, 330, 345
Brown, W. F., Jr. 22 Cylinder, hollow
Budiansky, B., 120, 134 critical pressure, 158
linear strain hardening, 160-61
Cauchy problem, 296 plastic flow, 156-61
Centered _fan, 272 strain-hardening material, 158-59
347
348 Index
Cylinder, long solid, 193-97 Ford, H., 133, 299

Foulkes, J., 324, 326
Dantzig, G. B., 326 Fox, L., 233, 299
Davis, E. A., 212 Fracture, true strain at, 12
Deformation Free thermal expansion, 66
finite, 51
pure, 48 Geiringer, H., 268, 299
Deformation theories, 119-21 Geiringer's equations, 268
Dieter, G. E., 22 Generalized strains, 314
Disk, rotating, 197-208 Generalized stresses, 314
creep of, 335-45 Goodier, J. N., 22, 23, 43, 63, 258
finite-difference formulation, 204- Green's function, 185
208 Greenberg, H. J., 299, 325
integral formulation, 198-203 Griffith, A. A., 223, 233
Displacements, rigid-body, 44 Gvozdev, A. A., 313, 325
Dissipation function, 314
Distortion energy, 68 Haigh-Westergaard stress space, 79-~8
Distortion energy theory, 75 Hardening
Dixon, J. R., 224, 226, 233 isotropic, 13, 94
Dorn, J. E., 23 kinematic, 95
Drucker, D. C., 23, 97, 111, 118, 133, Haythornthwaite, R. M., 326
134,299,313,325,326 Heller, W. R., 346
Ductility, 12 Hencky, H., 1, 3, 77, 134, 265,299
Hencky's equations, 265
Effective plastic strain increment, 102 Hencky's theorem, 269
Effective strain, 116 Heyman, J., 324, 326
Effective stress, 102, 116 Hill, R., 1, 3, 22, 23, 97, 133, 134, 163,
Eichinger, A., 91, 97 275, 299, 313, 325
Elastic limit, 6 Hirschberg, M. H., 212, 346
Elastoplastic problems, 2 Hodge, P. G., Jr., 22, 23, 62, 163, 259,
Elliptic equation, 292 271,299,325,326
Elongation, percent, 12 Hodograph, 281-84
Energy transformation, 295
distortion, 68 Hoffman, 0., 69, 97, 163
elastic, 67, 68 Hooke, R., 64
Equation of state, mechanical, 329 Hooke's law, 64, 65
Equilibrium equations, 29, 165 Horne, M. R., 325, 326
Equivalence of plastic work, 106 House, R.N., 3
Equivalent modified total strain, 124 Hult, J., 346
Equivalent plastic strain increment, Ruth, J. H., 259
102, 106 Hydrostatic pressure, effect of, 16
Equivalent stress, 102, 106, 107, 108 Hyperbolic equation, 292
Essenberg, F., 97 Hysteresis loop, 6
Felgar, R. P., 22 Ilyushin, A. A., 164, 212

Finite deformation, 51 Jnce, E. L., 212
Finnie, J., 346 Incompressibility condition, 16
Flow rule, 109 Incremental theories, 119-21
associated, 115 Inglis, C. E., 223, 233
for Tresca criterion, 108, 109 Instability point, 6, 11
for von Mises criterion, 100-104 Internal friction theory, 77
Flow stress, 6 Irwin, G. R., 224
Flow theories, 119-21 Isotropic hardening, 13, 94
Index 349
Johnson, W., 22, 23, 134, 163, 258, Marin, J., 11, 22, 97
259 Martin, H. C., 233
Massonnet, C. E., 326
Kantorovich, V., 258 Maximum load point, 6, 8, 9, 11
Kazinczy, G., 301, 325 Maximum shear theory, 73
Kinematically admissible multiplier, Maximum strain energy theory, 74
315 Maximum strain theory, 72
Kinematically admissible velocity field, Maximum stress theory, 71
284, 315 Mechanism(s), 30
Kinematic hardening, 95 elementary, 318
Kinematic mechanism, 310 beam, 319
Kobayashi, S., 299 frame, 319
Koff, W., 97 joint, 319
Koiter, W. T., 134, 156, 163 panel, 319
Korn, G. A., 299 portal, 320
Korn, T. M., 299 fictitious, 319
Kronecker delta, 27 linearly independent, 318
Krylov, V. 1., 258 rules for combining, 321
Melan, E., 119, 133
Mellor, P. B., 22, 23, 134, 163, 258,
Lee, E. H., 163 259
Levy, M., 1, 3, 100, 133 Membrane analogy, 245
Levy-Mises equations, 100, 261 Membrane-roof analogy, 247
Life-fraction law, 330-31 Mendelson, A., 134, 212, 233, 346
Limit analysis, 300-26 Metal forming processes, 2, 12, 15, 260
of beams, 305-310 Mikhlin, S. G., 212
with concentrated loads, 307-310 Minimum weight design, 324
in pure bending, 305-307 Mises, R. von, 1, 3, 77, 100, 133
design of structures, 300-301 Mises, von, criterion, 75, 115, 129
of frames, 310-12 flow rule, 100-1 04
of simple truss, 301-305 stress-strain relation, 100-104
theorems, 312-17 Mises, von, ellipse, 76
lower bound, 313-15 Mohr, 0., 43
upper bound, 313-15 Mohr's circle(s), 37-39, 280
Limit design, 301, 323, 324 for plastic strain increments, 101
restricted design problem, 324 pole of, 280
Livesley, R. K., 324, 326 for stress, 101
Loading, definition of, 93 Mohr's diagram, see Mohr's circles
Loading function, 92, 112 Mohr's theory, 77
Load point, maximum, 6, 8, 9, 11 Multiplier
Lode, W., 88, 89, 97, 109 kinematically admissible, 315
Lode's statically admissible 314, 315
strain parameter, 109 Muskhelishvili, N. I., 226, 233
stress parameter, 88
Lower bound theorem, 285, 313-15 Nadai, A., 258
Lower yield point, 6 Naghdi, P.M., 97, 134
Lubahn, J.D., 22 Navier equations, 68
Ludwik, P., 7, 20, 22 Neal, B. G., 318, 326
Necking, 6, 9, 12
Mclean, D., 22 Neuber, H., 224, 233
Magnusson, A. W., 22 Neutral loading, 93
Manjoine, M. J., 15, 22, 346 Norton, F. H., 345
Manson,S.S. 134,212,233,346 Novozhilov, V. V., 63
350 Index
Octahedral planes, 35 Pure deformation, 48

Odquist, F. K. G., 346 Pure shear, 41-42
Offset yield strength, 6
Optimum design, 324 Quinney, H., 90, 91, 97, 98, 109, 110
Orowan, E., 223, 233
Osgood, W. R., 20, 23 Rabatnov, Yu. N., 331, 345
Radial loading, 99, 120
Parabolic equation, 292 Ramberg, W., 20, 23
Pearson, C. E., 43 Rankine theory, 71
Pearson, K., 133 Reduction in area, 10, 12
Perfectly plastic material, 17, 115, 116 Reuss, E., 100, 133
Pi plane, 81-87 Rigid body, 44
Picard's method, 164 Rigid body displacements, 44
Plane elastoplastic problem, 213-30 Rigid-perfectly plastic material, 17, 260
Plane strain, 137, 213-18, 260 Roberts, E. Jr., 233
Plane stress, 137, 213-18 Ros, M., 91, 97
of finite plate, 218-23 Rotating disk, see Disk, rotating
of plate with crack, 223-30 Rotation Tensor, 48
Plastic collapse load, 301
Plastic deformation, 6 Sachs, G., 69, 97, 163
Plastic design, 301, 323, 324 Safety factor, 303, 304, 314, 315
Plastic hinge, 308 Saint-Venant, B. de, 1, 3, 100, 111,
Plastic moment, 307 133,234,258
Plastic potential, 119 Saint-Venant's principle, 240
Plastic strain charts, 191-93 Saint-Venant theory, 72
Plastic stress-strain relations, 98-134 Sand-hill analogy, 247-48
general derivation of, 110-119 Sankey, G. A., 346
Levy-Mises, 100, 261 Save, M. A., 326
Prandtl-Reuss, 100-104 Schwartzbart, H., 22
Plastic work, 104 Shaffer, B. W., 3
equivalence of, 106 Shape factor, 307
Plate, infinite, see Strip, infinite Shear lines, see Slip lines
Plate, infinite, withhole, 208-11 Shell, thin circular, 183-93
Polar coordinates, 137 Sherby, 0. D., 23
Power of dissipation, specific, 314 Shoemaker, E., 346
Prager, W., 19, 22, 62, 109, 133, 259, Simple shear, 41-42
271, 279, 280, 281, 284, 299, Singular yield conditions, 123
324, 325, 326 Slip lines, 263
Prandtl, L., 1, 3, 100, 133, 245, 258, compatibility equations, 265
275,299 geometry of, 268-72
Prandtl-Reuss equations, 100-104, 114, geometric construction, 279-84
115, 125, 167, 214 hodograph, 281
experimental verification of, 109, physical plane, 279
110 stress plane for, 279
Principal directions strong solution, 283
of slip line field, 262 weak solution in, 283
of strain tensor, 53-55 velocity equations, 268
of stress tensor, 30-34 Slip line theory, 260-99
Principal planes, 32, 53 complete solutions, 284
Proof strength, 6 lower bound theorem, 285
Proportional limit, 5 principal directions, 262
Proportional loading, 99, 120 principal stresses, 262
Punch indentatio!l, 273-76 shear directions, 263
Index 35 I
upper bound theorem, 284 hypothesis, 107

Sokolnikoff, I. S., 43, 62, 63, 69, 233, linear, 17, 20
258 Strain-hardening exponent, 11
Sopwith, D. G., 163 Strain-hardening hypothesis, 107
Spero, S. W., 212 Strain-hardening law, 330, 331
Sphere Strain-hardening parameter, 153
compatibility equation for, 136 Strain-invariance principle, 230-32
critical pressure, 140 Strain rate, effect of, 15
equilibrium equation, 135 Strain Tensor, 44-62
internal pressure, 138-48 Eulerian, 52
with linear strainhardening, 153-56 Lagrangian, 62
plastic flow in, 138-56 see also, Strain
Prandtl-Reuss relations, 136 Strength coefficient, 11
residual stresses, 145-47 Stress
shakedown pressure, 147 boundary conditions, 32
strain-hardening material, 150-56 convention, 28, 29
stress-strain relation, 136 deviator, 39-41
thermal loading, 148-50 invariants of, 40, 41
Spherical coordinates, 135-37 effective, 102, 116
Statically admissible multiplier, 314, equivalent, 102, 106, 107, 108, 116
315 generalized, 314
Statically admissible stress field, 285, invariants, 30, 33, 34
314 maximum shear, 34, 35
Statically determinate problem, 142, nominal, 4
261 normal, 28
Steele, M. C., 163 octahedral shear, 35, 36, 101
Stockton, F. D., 97 principal, 30-33
Strain, 44-63 principal directions, 32
compatibility of, 59-61 shear, 28
conventional, 4 maximum, 34, 35
deviator tensor, 58 octahedral, 35, 36, 101
modified, 124 spherical, 39
diametral, 9 true, 7, 8, 13
effective, 116 unit, 28
engineering, 4 Stress function, 216, 237
equivalent modified total, 124 Stress-strain curve, 4-11
generalized, 314 conventional, 5-7
invariants, 53, 55 dynamic models, 16, 17
logarithmic, 8 empirical equations, 20, 21
maximum shear, 55-57 idealizations, 16, 17
modified total, 124, 167 kinematic models, 18-20
natural, 8 strain-rate, effect of, 15
octahedral shear, 55, 57 true, 7-11
plastic increment, 101 Stress-strain relations
physical interpretation, 48-50 complete, 127-132
principal, 53-55 elastic, 64-69
shear, 50 general, 166
maximum, 55-57 incremental, 119, 120
octahedral, 55, 57 plastic, 98-134
octahedral plastic increment, 101 general, 114, 118
true, 8, 10, 13 general derivation of, 110-119
Strain Energy, see Energy Levy-Mises, 100, 261
Strain hardening, 6 Prandtl-Reuss, 100-104, 114, 115
352 Index
Stress-strain relations (continued) Todhunter, 1., 133

for Tresca criterion, 108, 109, Torsion of prismatic bar, 234-57
115 elasticity solution, 240-45
plastic strain-total strain, 123-127 circular cross section, 242
total, 119, 121 elliptic cross section, 240-42
Stress tensor, 24-43, see also, Stress equilateral triangle, 242, 243
Stress trajectories, 239 membrane analogy, 245
Strip, first order, 290 rectangular cross section 243-45
Strip, infinite thin, 172-82 general relations, 234-40
linear strainhardening, 178-82 plasticity solution, 246-58
nonlinear strainhardening, 175-78 circular section, 253-57
Strip condition, 290 membrane-roof analogy, 247
Successive approximations, method of perfect plasticity, 246-48
circular shell, 183-93 rectangular section, 250-53
convergence of, 169-71 sand-hill analogy, 247-48
creep problems, 332-45 strain hardening, 248-58
finite plate, 218-23 Torsional rigidity, 241
general description, 164-71 Total theories, 119-21
infinite plate with hole, 208-11 Transition temperature, 15
infinite thin strip, 172-82 Tresca, H., 1, 3
plate with crack, 223-30 Tresca criterion, 73
rotating disk, 197-208 flow rule, 108, 109
solid cylinder, 193-97 kinematic model, 95, 96
sphere, 150-53 stress-strain relations, 108, 109, 131
torsion of bar, 249-55 Trozera, T. A., 23
circular cross section, 253-55 Truss, simple, 301-305
rectangular cross section, 250-53 Tuba,I.S., 163,210,212,233
tube, 156-59 Tube (see Cylinder, hollow)
Successive elastic solutions, method of Tupp, L. J., 233
(see, Successive approxima- Tupper, S. J., 163
tions) Turner, M. J., 233
Superposition of mechanisms, method
of, 318-23 Ultimate strength, 6
Swedlow, J. L., 229, 233 Uniform stress state, 272
Symonds, P. S., 318, 326 Uniqueness condition, 111
Unloading, 6, 93
Taylor, G. 1., 90, 91, 97, 98, 109, 110 definition of, 93
Temperature Upper bound theorem, 284, 313-15
effect on metal properties, 15
transition, 15 Van den Broek, J. A., 326
Tensile strength, 6 Varga, R. S., 259
Tensile test, 4-7
Tensor, 24-27 Wahl, A.M., 133, 337, 346
notation, 24-27 Wang, C. T., 212
relative displacement, 47 Warping function, 236
rotation, 48 Westergaard, H. M., 1, 3, 97
skew-symmetric, 26 White, G. N., 163
substitution, 27 Williams, M. L., 233
summation convention, 25, 26 Work equation, 320
symmetric, 26 Work hardening, 6
Thomsen, E. G., 299 definition of, 111
Time-hardening law, 330, 331 measures of, 104-107
Timoshenko, S., 43, 63, 258 hypothesis, 106, 107
Index 353
Work-hardening hypothesis, 106, 107 convexity, 120-23

Work increment, 26 corners, 120, 123
singular points, 120, 123
subsequent, 92-96
Yang, C. T., 299 Yield criterion, 70-79
Yang, W. H., 233 Beltrami's energy, 74
Yield Coulomb's, 73
cylinder, 82, 86 distortion energy, 75
effect of hydrostatic pressure, 16 internal friction, 77
function, 79, 92 maximum shear, 73
hinge, 308 maximum strain, 72
locus, 82, 83, 87 maximum strain energy, 74
moment, 307 maximum stress, 71
point, 5, 6 von Mises, 75
lower, 6 Mohr's, 77
offset,. 6 Rankine's, 71
upper, 6 Saint-Venant's, 72
surface, 79, 92 Tresca's, 73

2-Plasticiy Theory and Application - A. MENDELSON - 3

Uploaded by

Copyright:

Available Formats

2-Plasticiy Theory and Application - A. MENDELSON - 3

Uploaded by

Document Information

Original Description:

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

2-Plasticiy Theory and Application - A. MENDELSON - 3

Uploaded by

Copyright:

Available Formats

MACMILLAN SERlE~ IN

The Analysis of Stress and Deformation

MACMILLAN COMPANY, NEW YORK

COLLIER-MACMILLAN LIMITED, LONDON

materials with strain hardening. In particular, it was emphasized that with

and co-workers at the Lewis Research Center of the National Aeronautics

Chapter 2. Basic Experiments 4

3. The Stress Tensor 24

Chapter 4. The Strain Tensor 44

Chapter 5. Elastic Stress-Strain Relations 64

Chapter 6. Criteria for Yielding 70

Chapter 7. Plastic Stress-Strain Relations 98

Chapter 8. Elastoplastic Problems of Spheres and Cylinders

8-1 GENERAL RELATIONS

8-3 HOLLOW SPHERE. SPREAD OF PLASTIC ZONE.

Chapter 9. The Method of Successive Elastic Solutions 164

Chapter 10. The Plane Elastoplastic Problem 213

Chapter 11. The Torsion Problem 234

12. The Slip-Line Field 260

order of magnitude as the elastic strains. Problems of large plastic deforma-

In this chapter the results of some basic experiments on the behavior of

2-1 TENSILE TEST

and conventional strain by

FIGURE 2.1.2 Conventional stress-strain curve.

stresses or to irregularities in the specimen as well as to the rate of loading.

2-2 TRUE STRESS-STRAIN CURVE

The nominal stress an is an = P/A 0 and the conventional strain is

In a somewhat similar fashion, one recognizes that the conventional or

e is called the natural, logarithmic, or true strain and it represents a sort of

whereas in terms of conventional strains it is

(1 + el)(1 + e2 )(1 + e3 ) - 1 =0 (2.2.6)

only in the case of small strains.

at the point of maximum load

FIGURE 2.2.1 True stress-strain curve.

and from equation (2.2.5) the longitudinal true strain is

FIGURE 2.2.2 True stress-strain curves for several materials.

,.., SAE 3140 240,000 psi

FIGURE 2.2.3 True stress-strain curves on log-log coordinates.

Alternatively, equation (2.2.15) can be written in terms of the reduction in area

2-3 COMPRESSION TEST AND THE

as shown by ABCHI. Actually experiments indicate that the compressive

2-4 EFFECTS OF STRAIN RATE AND

-200 -150 -100 -50 0 oc

2-5 INFLUENCE OF HYDROSTATIC PRESSURE.

Bridgman, in a series of classical experiments [11, 12] in which he carried

2-6 IDEALIZATION OF THE STRESS-STRAIN

Because of the complex nature of the stress-strain curve, it has

2.6.1 Idealized stress-strain curves: (a) perfectly elastic, brittle;

idealized curves as well as corresponding dynamic models which can be used

this type to represent combined stresses acting in several directions would be

FIGURE 2.6.3 Stress-strain curve for model of Figure 2.6.2.

In a two-dimensional state of stress, two stress variables, a 1 and a 2 , are

2-7 EMPIRICAL EQUATIONS FOR STRESS-STRAIN

It is sometimes useful to represent the stress-strain curve of a given material

A frequently used form, due to Ramberg and Osgood [15], is

Some other forms that have been proposed are

a = a + (b - a)(l - e-n•) (2.7.3)

a= c(a + e)n 0 s n:::; I (2.7.4)