INTERNATIONAL J O U R N A L F O R NUMERICAL METHODS I N ENGINEERING, VOL.

1 75-100 (1969)

ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS

'INITIAL STRESS', FINITE ELEMENT APPROACH
0. C. ZlENKlEWICZ,* S. VALLlAPPANt AND I. P. KINGS
* Profissor q/ Civil Eir,~~iriceritig
t Senior Researcli Assistant
$ Former Reswrcli Fellow
University of Wales, S,vun.sea

SUMMARY
The paper presents first a general formulation of the elasto-plasticmatrix for evaluatingstress increments
from those of stresses for any yield surface with an associated flow rule. A new 'initial stress' computa-
tional process is proposed which is shown (I) to yield more rapid convergence than alternative approaches
(2) to permit large load increments without violating the yield criteria and thus simply to establish lower
bound solutions. Several solutions showing stress distribution, strain developnient and growth of plastic
enclaves are given both for the von Mises and for Coulomb (Drucker) type yield surfaces. Load reversal
and thermoplastic behaviour are dealt with.

INTRODUCTION
It is probably true to say that the recent development of numerical methods in general, and of
the finite element method in particular, permits solutions to be obtained for any rationally
conceived constitutive laws of the material behaviour. Of practical interest in this context are
the numerous problems in which plasticity plays a dominating part. Such problems range in
application from machine technology, through structural applications, to geophysics. The
materials for which solutions are required may exhibit a variety of yield surfaces. One of the
objectives of this present paper is to provide a unified treatment of general viability so that with
a minimum of programming effort, diverse situations can be accommodated.
In this context it is essential that the method should be able to deal with problems of ideal or
work hardening plasticity.
As the interest in a particular solution may be in the prediction of displacements and strains
at various stages of the loading, in the development of plastic zones or in the prediction of residual
strain distribution on load removal it is important that all these should be adequately represented.
In particular it is important that the method should be able to follow both loading and unloading
cycles without difficulty, reproducing fully the elasto-plastic behaviour.
Quite frequently the only information required by the designer is that of determining the
collapse situation. Here quite simple computational processes suffice and there is no apparent
need to go through all the stages of a complete elasto-plastic solution. However, for various
important situations only bounding answers can be achieved and indeed kinematic (upper)
bound is usually the only one readily achieved. Here it is hoped that the methodology presented
will be of interest as it will invariably provide an assessment of the equilibrium (lower) bounds.
Various computational procedures have been used with success for a limited range of elasto-
plastic problems utilizing the finite element approach. Two main formulations appear. In the
first, during an increment of loading, the increase of plastic strain is computed and treated as an
initial strain for which the elastic stress distribution is adjusted.',' This approach manifestly
Received 23 May 1968
75
76 0. C. ZIENKIEWICZ, S. VALLIAPPAN AND I. P. KING

fails if ideal plasticity is postulated or if the degree of hardening is small. The second approach
is that in which the stress-strain relationship in every load increment is adjusted to take into
account plastic deformations. The work of Pope,3 Swedlow,’ Marcal and King,’ Reyes and
Deere6 and Popov and others’ falls into this category. With a properly specified elasto-plastic
matrix this incremental elasticity approach can successfully treat ideal as well as hardening
plasticity.
From the computational point of view the ‘incremental elasticity’ process has one serious
disadvantage. At each step of computation the stiffness of the structure is changed and iterative
processes of solution are necessary to avoid excessive computer times. In this paper an alternative
approach which we shall refer to as the ‘initial stress’ process is developed. By using the fact
that even in ideal plasticity increments of strain prescribe uniquely the stress system (while the
reverse is not true for ideal plasticity) an adjustment process is derived in which ‘initial stresses’
are distributed elastically through the structure.
This approach permits the advantage of initial processes (in which the basic elasticity matrix
remains unchanged) to be retained. The process appears to be the most rapidly convergent. It
will be found that no special treatment of unloading cycles is now required. The method will be
fully described later but is in principle similar to the treatment of cracking materials described
elsewhere.*
A comprehensive bibliography on the finite element method in general and on non-linearity
in particular will be found in a text.’ No description of the finite element method or of the
nomenclature is thus called for here.

SOME BASIC CONCEPTS OF PLASTICITY

Yield surface
It is quite generally postulated, as an experimental fact, that yielding can occur only if the
stresses {CT] satisfy the general yield criterion

In this a vectorial notation is used for stress components and K is a hardening parameter.
This yield condition can be visualised as a surface in n-dimensional hyper space of stress with
the position of the surface dependent on the instantaneous value of the parameter K (Figure 1).

Figure I . Yield surface and normality criterion in two dimensional stress space
 ELASTO-PLASTIC s o L u r i o N s OF ENGINEERINGPKOULEMS 77

Flow Rule
von Mises" first suggested the basic constitutive relation dcfinine the plastic strain increments
in relation to the yield surface. Heuristic argniiients for the validity of the relationship proposed
have been given by various workers in the field".'2 and at the present time the following hypothe-
sis appears to be generally accepted:
If 6 { ~ denotes
] ~ the increment of plastic strain then

or for any component n

dF
SEnp = A-
070,

In this 1is a proportionality constant, as yet undetermined. The rule is known as the norrnality
principle because relation (2) can be interpreted as requiring the normality of the plastic strain
increm,ent vector to the yield surface in the hyper space of n stress dimensions.

Total Stress-strain relations

During an infinitesimal increment of stress, changes of strain are assumed to be divisible into
elastic and plastic parts. Thus

a{&}= a{&}, + 6 { E } p (3)

The elastic strain increments are related to stress increments by a symmetric matrix of constants
[D] known as the 'elasticity matrix'. Thus

S{E), = [D]-'S{o} (4)

We can thus write (3) as

When plastic yield is occurring the stresses are on the yield surface given by (1). Differentiating
this we can write

or

A=?dK.- I
in which
OK i
78 0. C. ZIENKIEWICZ, S. VALLIAPPAN A N D I. P. KING

Equations ( 5 ) and (6) can be written in a single symmetric matrix form as

‘
I^.
aF
on1
I

This form is convenient for use directly provided that ‘A’ is not zero as shown in a particular
form by Marcal and King.5 Alternatively 3, can be eliminated (taking care not to multiply or
divide by A which may be zero in general). This results in an explicit expansion which determines
the stress changes in terms of imposed strain changes.

with b{c} = [D]:p S ( E }

The elasto-plastic matrix [D]jp takes the place of the elasticity matrix [D] in incremental
analysis. It is symmetric, positive definite, and the expression (9) is valid whether or not ‘A’
takes on a zero value.

Significance of parameter ‘A’

Clearly for ideal plasticity with no hardening ‘A’ is simply zero.
If hardening is considered, attention must be given to the nature of the parameter (or para-
meters) K on which the shifts of the yield surface depend.
With a ‘work hardening’ material K is taken to be represented by the amount of plastic work
done during plastic deformation. Thus

substituting the flow rule (2) we have simply

By equation (7) we now see that L disappears and we can write

a strictly determinate form if explicit relationship between F and K is known.

 ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS 79

An illi 1m-nt ive e . w i iple

To illustrate some of the concepts consider the well known von Mises yield surface. This is
given by
F = [HUl - + -:(a2 - a$ + $(a, - o,)2 + 30: + 30: + 3.35 - (5 (14)

in which suffixes 1, 2, 3 refer to the normal stress components and 4, 5, 6 to shear stress com-
ponents.
On differentiation it will be found that

in which the dashes stand for deviatoric stresses i.e.

The quantity = 0 (K) is the uniaxial stress at yield. If a plot of the uniaxial test giving a
versus the plastic uniaxial strain eUpis available then

and
d-F - dZ
- -- .
di? -1- -H'
du - d u - denp 0 - 5

in which H'is the slope of the plot at the particular value of 2.

On substituting into (13) we obtain after some transformation simply

A = H' (17)
This re-establishes the well known Prandtl-Reuss stress strain relations.

'Corners' of a yield surface

It happens, not infrequently, that the yield surface is defined not by a single continuous (and
convex) function but by a series of functions:

F,, F , . . . Fn
the state of strain below the yield limit being defined by negative values of all the functions F.
For most of the bounding surface only a single condition such as F,,, = 0 will define the yield
surface and the previously written flow rules (and elasto-plastic matrices) apply.
At a 'corner' of the yield surface we may have however the condition that
Fh= ... =F,,,=O
80 0. C. ZIENKIEWICZ, S. VALLIAPPAN A N D I. P. KING

Here the Koiter generali~ation‘~

replaces equation (2) giving

i18)

where are positive constants (Figure 2).

Figure 2. Corners in a yield surface. Graphical interpretation of Koiter’s criterion

Matrices of type equation (8) can once again be written now with several undetermined para-
meters A. Procedures similar to those above will yield new forms for the elasto-plastic matrix
applicable at such corners.
The computation of singular points on yield surfaces is best avoided by a suitable choice of
continuous surfaces which usually can with a good degree of accuracy represent the true
conditions.
App Iica t ion
The ‘elasto-plastic’ matrix has been given in a general form in equation (10). Particular forms
will obviously depend on the problem at hand. A generalized stress vector was used throughout
as the process is equally applicable to full three dimensional stress fields, to special two dimen-
sional situations or to moment-curvature relations in plate bending. An explicit form of a plane
strain matrix for a Prandtl-Reuss material has recently been published by Yamada and other^.'^
In this paper various forms of this matrix will be used in different examples.

THE ‘INITIAL STRESS’ COMPUTATIONAL PROCESS

The expressions derived in the previous section describe fully the stress-strain relation in the
elasto-plastic state. The essential non-linearity is evident from the equation (10) with the ‘elasto-
plastic’ matrix being dependent on the state of total stress. Some ‘piecewise’ linearization
process is then required. In the ‘incremental elasticity’ method small load increments are
prescribed and in each the material treated as quasi-elastic, with a constant ‘elasto-plastic’
matrix. The numerical values prescribing this matrix may either correspond t o the initial stress
values of the increment or by adjustment may be made to coincide to average stresses in the
increment (5). Whichever approximation is used different elastic problems are posed at each
load increment necessitating changes of the stiffness matrices at each step. This computational
 ELMTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS 81

inconvenience is overcome by an essentially different approach which we shall call the ‘initial
stress’ method.
The ‘initial stress’ process which is used in this paper approaches the solution of a non-linear
problem as a series of approximations. In the first place during a load increment a purely elastic
problem is solved determining an increment of strain A{&}‘ and of stress A{a}’ at every point
of the continuum (or structure).
The non-linearity implies however that for the increment of struin found, the stress increment
will in general not be correct. If the true increment of stress possible for the given strain is A{.}
then the situation can only be maintained by a set of body forces equilibrating the ‘initial’ stress
system A{a}’-A{c}.
At the second stage of the computation this body force system can be removed by allowing
the structure (with unchanged elastic properties) to deform further. An additional set of strain
and, corresponding, stress increments is caused. Once again these are likely to exceed those
permissible by the non-linear relationship and the redistribution of equilibrating body forces has
to be repeated.
If the process converges then finally within an increment the full non-linear compatibility and
equilibrium conditions will be satisfied just as they are in an ‘incremental elasticity’ solution.
It appears ‘a priori’ in an elastic-plastic situation that the process is a natural one making use
of the fact that for any prescribed strain the increment of stress is a determinate one and one that
changes slowly. Indeed application shows that convergence is rapid, three or four cycles of re-
distribution (iteration) being necessary in any increment.
As for each cycle the same elastic problem is being solved then, clearly, if use is made of a partial
invertion of the elastic equation very rapid computer times will result.
Obviously the process is of a general applicability not limited to the elasto-plastic situation.
Indeed if the situation is such that a limit on stress is imposed without a corresponding strain
relationship a single increment of load can be used to achieve a final solution, as for instance has
been done in ‘tension cut-off‘ situations.’
In full elasto-plastic situations it is generally necessary to proceed in a series of load increments
to follow the appropriate flow rules. If however a single load increment is used it will be found that
an approximate lower bound is achieved, the final solution satisfying equilibrium and yield criteria
but not necessarily following the current strain development. The use of such bounding solutions
in practice is important.
For the elasto-plastic situation the steps during a typical load increment can be summarized
as follows:
1. Apply load increment and determine elastic increments of stress and strain
{ A E ’ } ~which correspond.
2. Add to stresses existing at start of increment {ao}to obtain {a’}. Check whether
F{a’} <O (with K referring to the initial value at start of increment). If above satisfied
only elastic strain changes occur and process is stopped, if not proceed to 3.
3. If F{o‘}2 0 and also F{ao}= 0 (i.e. element was in yield at start of increment) find

with [D]:.,. computed from equation (9) with stresses {a’}.

Evaluate stress which has to be supported by body forces
{ A o ” ) ~= {Ao’)~ - { A o } ~
82 0. C. ZIENKIEWICZ, S. VALLIAPPAN AND I. P. KING

Store current stress {g} = {a’} -

and current strain {E} = {E‘} + A{E’}~
4. If F{g}>O but F{a,}<O
find the intermediate stress value at which yield begins and compute increment {Ag}l
by equation (9) starting from that point. Then proceed as in 3.
5. Compute nodal forces corresponding to the equilibrating body forces. These are given
for any element by*
{ P } l ‘ = J[BIT{ A d } 1 d(vo1)
6. Resolve using original elastic properties and the load system {P} to find { A d ) 2 and
W 1 2 .
7. Find current value of K
8. Repeat steps 2 to 6, etc.

The cycling is terminated when the nodal forces of 5 reach sufficiently small values. If this is
not achieved in a predetermined number of cycles (20 in our case) collapse condition is deemed to
have been achieved and the process is stopped.
This brief description shows the necessary modifications to any standard finite element elastic
program to enable it to deal with the elasto-plastic situation.
The process is illustrated graphically in a two dimensional stress space in Figure 3. Note that
after a few cycles the resulting stress is always brought back to the yield surface.

Ul (€J

Figure 3. Graphical interpretation of the ‘initial stress’ process

It is evident (vide step 2) that if an unloading occurs the program will automatically follow
the elastic unloading process until a new intersection with the yield surface again occurs. Thus
complete cycles of load can be dealt with.
By ‘load‘ increments, implicitly any set of external loads or internal initial strains is meant.
Thus thermal problems fall readily into the method described.

* The standard notation of the finite element process is used here, reference 9 in which the matrix [B] defines
strains in terms of nodal displacements-see Appendix.
 ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS 83

Any specified increments can be used from the start of loading. It is convenient however to
start the incremental process only when first yield has occurred and in the program this allows
the subsequent load increments to be related to the load at which first yield is noted.

SPECIAL FORMS OF THE ELASTO-PLASTIC RELATIONSHIP

Most generally the yield criterion is established in the ‘six dimensional‘ stress space as a function
of all the six stress components. When dealing with more restrictive problems such as prescribed
by cases of plane stress, plane strain or axial symmetry an appropriate specialization of the yield
surface to the more limited freedom has to be made.22 On occasion the special form may be
described directly as, for example, when yield surfaces for plates and shells are considered.”
Alternatively a purely formal operation will yield the appropriate result.
Consider the general relationship (8) written in terms of the six three dimensional stress com-
ponents listed as
01 =a, 02 =ay 0 3 =a, 6 4 = r x y 6 5 =Tyz 66 =tzx

and let us now turn our attention to the special cases.

Plane Stress
If z is chosen as the direction normal to the plane we note immediately that
da, = az, = dr,, = 0
and that appropriate columns may be deleted from the relationship. The rows corresponding to
these stress components cease to be of interest in the two dimensional analysis and we are left
with the special form of equation (8)

CDI-

where [D] stands for the simple elasticity plane stress matrix and F is the ‘cross-section’ of the
yield surface with a, =,.z, = r,, = 0. The same transformation equation (10) will obviously
be still used to eliminate A.
Although a plastic strain e z P now occurs it is not necesary to record this as with a work
hardening situation assumed, its contribution to equation (1 1) is zero.

Plane strain
Once again two shear stress components (r,, and rZx)become zero and rows and columns
corresponding to these can be omitted. The normal stress a, is however no longer zero and the
condition that
E, = 0
has to be imposed.
84 0. C. ZIENKIEWICZ, S. VALLIAPPAN AND I. P. KING

The equation can now be written as

D o 1- '

----
symmetric

If the matrix of elastic constant [Do]-' is written as

[Do] - = [uij l

then it is easy to verify that on elimination, equation (20) can be reduced to

P I-

iymmetric (A - '(")z)
a44 8%

In this [DI-' is the usual reduced plane strain elastic matrix. For an isotropic material

1 V
E
a14 = a24 = - -E and aS4= 0

Elimination of 1 can still be carried out in the manner of equation (10) noting however that
appropriate substitutions have to be made. In particular A is now replaced by

and does not become zero even for ideal plasticity.

 ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS 85

As oZis now no longer zero it is necessary to keep a record of it in the computation as plastic
strains in the normal direction will now occur.

Axial symmetry
Here the situation is once again more simple as four stress and strain components have non
zero values and only two shear stress and strain components vanish. The form of the relationship
will be identical with that of equation (20) but with the fourth non zero strain component.

PLANE STRESS AND STRAIN WITH von MISES YIELD CRLTERON-EXAMPLES

Perfarated plate with and without strain hardening
This plane stress test example was studied by both the finite element m e t h 0 d ~ 9 'and
~ experi-
mentally16 and is therefore of value in assessing the accuracy and efficiency of the different
approaches. In Figure 4 (see page 86) the mesh of simple triangular elements is shown together
with plastic enclaves for both strain hardening and ideal plasticity cases. The results are in
substantial agreement with those of the previous investigations.
In solutions (b) and (c) increments of load equal to 0.2 of the load at first yield were used.
It was of interest to investigate a one step process in which the final load was reached in a
single increment. Figure 4 (d) shows the final zone of plasticity thus obtained which,
surprisingly, differs but little from the more correct small increment solution.
In Figure 5 (see page 87) the development of the maximum strain is compared with results
of the finite element 'partial stiffness' method as used by Marcal and King' and experimental
results.16 The results of the initial stress method are slightly closer to experiment for the same
load increments. The single step solution gives again an exceedingly good estimate of this
strain. Indeed the surprising insensitivity of the results to the magnitude of the increment is
illustrated in Figure 6 (see page 87).
If large increments are used directly with the partial stiffness approach5 then meaningless
results are in general obtained.
Table I shows some comparison of the computational times involved in the two approaches.
Table I. Comparison of Computer Time

No.of Computer Time in mts.

Problem
No. of
Nodcs
No. of
Elements
-
H'
E
Load
Increment
No. of
Increments
Interations
within an
I.C.T. 1905
Initial Partial
Increment Stress Method Stiffness Method

Q1L 7 3 18 31
5 23
0*2L 7 4 21 43
6 29
0.032 0.4L 4 6 21 -
1L 30 31 failed to
~~

converge
149 4 22 30
O*IL 7 6 30
0 Q2L 7 S 21 50
I 31

z
- 0.IL 14 5 12 -
94 I49 0
26
z: 26 0.2L I1 6 61 69

68 g Note: L = Load at fint yield.

z- 5
5
86 0. C. ZIENKIEWICZ, S. VALLIAPPAN, AND I. P. KING

2 IT rneon
1

(a) Finiteelement division (149

elements, 94 nodes)

(b) Spread of plastic zones for

18 mn various ratios of ume.,,/ZV
Ideal plasticity
E = 7000 kg/mms Y = 0-2
b, = 24.3 kg/mmP

0.92

1-00

1.08

0.9

(d) Load of 0.98 applied ir (c) As (b) but with strain

single increment straifi hardening.
hardening material Constant slope H/E =
0.032

lOmm
I

Figure 4. Perforated tension strip (plane strip)

 ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PRORLEMS 87

1.25
/
/
/

0.75.

0.50
- EXPERIMENTAL RESUCT FROM
THEOCARIS P MARKETOS 16
"0" INITIAL STRESS METHOD
--X--.PARflAL STIFFNESS METHO0
SOLUTION FOR YNGLE
A STEP m PLASTIC REGION

1.5 2.0 2.5 3.0 3-5 4.0 4

'E
Q
Y

Figure 5. Perforated plate-strain hardening material

Development of maximum strain at point of first yield
H / E = 0.032
Load increment = 0.2 x first yield load

1.00

0.75

-
-
-0
FROM THEOCARIS 6 MARKETOS
x YIELD LOAD
INCREMENT ~ 0 . 1
-X - INCREMENT = O ~ YIELD
X LOAD
0.2 5 --A - INCREMENT = O ~ YIELD
X LOAD-

0 2

Figure 6. Perforated plate-strain hardening material

As Figure 5 showing the effect of load increment on 'initial stress' solution
88 0. C. ZIENKIEWICZ, S. VALLIAPPAN AND I. P. KING

Figure 7 shows how closely the stresses at a particular element follow the yield criterion specified.
This indeed appears to be a particularly satisfactory feature of the method used.

Figure 7. Perforated plate-strain hardening material

Effective stress versus effective strain in element which yielded first

Notched plane strain specimen-perfect plasticity

Figures 8 and 9 show some results for this case and compare the results with those previously
~ b t a i n e d .Once
~ again the insensitivity to load increment size should be noticed.

Deep cantilever beam. Plane stress and ideal plasticity

The third example concerns the behaviour of a simple cantilever beam under a uniform load on
its top surface. Figure 10 shows the rather coarse mesh used and the spread of the plastic zones.
The ‘collapse’ load was estimated by conventional beam theory with a ‘hinge’ developing
at the support.
Plastic zones (Figure 10) and deflections (Figure 11) are shown for different ratios of the actual
to the ‘collapse’ load. When the load was incremented up to the full value no convergence could be
obtained with full 20 cycles and it is presumed that the state of steadily increasing deformation
was reached. The last point represents a good estimate of the lower bound of the collapse load.

Deep cantilever beam-load reversal

Up to now only a monotonic load increase was considered. To illustrate that the process is
available for load cycling without any modification the load on the previous example was first
removed from a particular increment and then entirely reversed.
Figure 12 shows the deflection-load path and Figure 13 the stress patterns. It is of interest to
note the residual stress distribution-not dissimilar to that predicted by conventionalbeam theory.
 2 u mean

t f +
(a) Mesh 149 elements 94 nodes

(b) Zone of yielding for various ratiosof

Q,/CTY

E = 7000 kg/mm*
v = 0.2
18mm
0” = 24.3 kg/mm*

I
I
I
I
1-
I

2.0

1.5 / o<

/
A
,
#
.
*AQ
B/ ’*O’ AARAL X - PARTIAL STIFFNESS METHOD
b Ib &.+fivr -- 0 INITIAL STRESS METHOD (0.2 INC )

0.5 -A- HlTlAL STRESS METHOD (0.1 INC)

INITIAL STRESS METHOD- ONE
-0-
INCREMENT IN PLASTIC RANGE

Figure 9. Notched specimen. Development of maximum strain for different load increments
90 0. C. ZIENKIEWICZ, S. VALLIAPPAN AND I. P. KING

4Y W

0.690.9451.0
Figure 10. Cantilever Beam-Plane stress, ideal plasticity. The spread
of plastic zones for different ratios of w/w,when w, is calculated as
from plastic beam theory. w, = collapse load

I-c - Not converged

0.E

0.E

- % 0.4
4

0-2

0 0.2 0.4 0.6 08 1.0

Deflection at C

Figure 11. Cantilever beam

Deflection versus w/w,
 ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS 91

-0.8 -0.6 -0.4 -0.2 0 02 0.4 0.6 0.8

Deflection at C
Figure 12. Cantilever beam
Deflections for load reversal

0-4 At start of yield

X---X At m s t load

+-A Reversal
0-0 Residual

Section AA Section BB
Figure 13. Cantilever beam
a, stress distribution at various stages of loading-unloading
 92 0. C. ZIENKIEWICZ, S. VALLIAPPAN AND I. P. KING

20ft
4 *
lo

/
Elastic
0.71T

I.OT

\
Elastic plastic

Figure 14. Thermal stress (plane strain)

(a) Mesh used 142 elements 88 nodes
E = 30 x 106 lb/in'
v = 0-3
3"= 1,665,000 lb/in*
(b) Final temperature state
(c) Spread of plastic zones
(d) Stress distribution
Elastic and Elasto-plastic (both for same temperature state)

Thermo-plastic behaviour
To illustrate that no difficulties are encountered with thermo-plastic behaviour an illustrated
problem is shown in Figure 14. A steady temperature shown is for simplicity assumed to be
applied in several increments.
 ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS 93

The yield stress was taken as temperature independent although inclusion of temperature
effects on the yield surface would have caused no major difficulties.

PLANE STRAIN WITH COULOMB TYPE YIELD SURFACE-EXAMPLES

For concrete, rock and soil the yield surface depends not only on the deviatoric but also on the
‘hydrostatic’ stress components. The best known ‘law’ approximating to the true observed
behaviour is that of Coulomb which specifies yield when the following relation between shear and
normal stresses on any plane is reached.
rsn= C + ontan 4 (22)
A more ‘manageable’ form approximating to this relationship was suggested by Drucker‘’ and
can be written in the following forms

F = ctJ, + J2f - K = 0
where J , is the first invariant
J, = 0, + av + bz
and J2 is the second invariant given as

in which CL and K are constants depending on cohesion and friction of the material. The
constants of equation (23) are related to the standard ones of equation (22) as

tan 4 3c
01= K=
J(9 + 12 tan’ 4) J(9 + 12 tan’ 4)
Other possible forms are discussed in some detail by Bishop” but for the present purposes
the formulation given by Drucker will suffice.
It is a simple matter of algebra to evaluate the appropriate differentials of the yield surface
F = 0 and use in the general formulation already discussed.
A word of warning is perhaps due at this stage. In the first place the yield surface of ideal
non-hardening type has been assumed. This is not true in general for such materials as a t first
yield work hardening followed by work softening occurs in practice. The latter presents several
difficultie~’~and will be dealt with in another publication. For practical purposes the approxi-
mation of ideal plasticity is valid providing the total strains developed are limited.
The second difficulty is that of the ‘normality’ principle built into the general program as
described. This for a material with an appreciable value of 4implies a continuing volume increase
in plastic deformation-a fact at variance with experimental evidence.” Alternative ‘non-asso-
ciated’ flow rules could be adopted2’ but these are by no means verified.
Three problems of the category described are solved.

A circular underground opening

A circular excavation was studied under the same yield criterion by Reyes and D e e d by
a rather elaborate ‘incremental elasticity’ approach. Figure 15 shows the finite element mesh
and the spread of plastic zones. The opening is assumed so deep that uniform initial stresses
 Vertical displacement = 0
1.000 lb/in2
50ft

--- From Reference 6.

cu
I .E

soft1

Horizonial
displacement = 0

X I
Figure 15. A circular underground opening

(a) Mesh and boundary condition 140 elements 87 nodes

Initial stress o,, = 250 lb/in* u,,"= lo00 Ib/in*
E = 500,0001b/in2 Y =0 2
c = 2801b/ina Q = 30"
(b) Spread of plastic zones and cnmparison6.
 ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS 95

existed before excavation with horizontal component equal to 0.25 of the vertical. The spread of
plastic zones is due to progressive removal of boundary loads on excavation.
Figure 16 shows the stress distribution resulting as well as a stress distribution which could have
arisen under purely elastic conditions. Note the appreciable reduction of tensile stresses which now
have to be limited to a value C/tan 4 by the yield criterion.

A lined tunnel
This example only differs from the previous in two respects. Firstly a more realistic tunnel shape
is adopted and secondly the tunnel is lined with a material assumed to be elastic throughout.
The excavation loads are ‘externally supported’ during the process of lining and any stresses in
this are due to subsequent removal of the support.
Figures 17 and 18 show similar results to those of the previous two figures.

A strip foundation
Figure 19 illustrates the solution to this problem of a strip loading on a half space.
The spread of plastic zones can be observed and contrasted with the usual mechanism of
failure assumption.
Bounds for the collapse load on this example have been estimated according to Finnz3 and
are
Lower bound 18,040 Ib.
Upper bound 46,300 lb.
In the computation the problem did not converge at a load of 23,000 lb., hence the highest
estimate of the lower bound is given by the previous step, i.e. 20,OOO Ib.

(a) (b)

Figure 16. A circular underground opening

(a) Elastic stress distribution on full load
(b) Elasto-plastic stress distribution
96 0. C. ZIENKIEWICZ, S. VALLIAPPAN AND I. P. KING

Gmnd surface
t
A \\\ \\\ \\\ \\\ "

I Horizontal displacement = 0
Vertical displacement = 0

Figure 17. A lined tunnel

(a) Mesh 153 elements 94 nodes

Lining E = 3 x lo" lb/in* Rock E = 5 x 10lb/in*
Y = 0.15 v =020
C = 1401b/in2
4 = 30"
Initial stress a, = yh with origin of h at 400 ft.
= 0.2 yh y = 150 lbs/ftS

(b) Spread of plastic zones

 ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS 97

+ + -c

- -c +
+
+
+
--c

+
-- - +
-I-
+ +

+
L

-4-
+

--
-a.
t

+
+
x
-c
c +

+ +

+ +

- - - +
t
+
+
&
 24ft

Horizontol I
displacement = 0

-i- .$
\ Horizontal displacement = 0
Vertical displacement = 0

Figure 19. A strip foundation

(a) Mesh 134 elements 80 nodes

E = I x 1Oa1b/in*
Y = 0-35
C = 7 Ib/in*
$6 = 35"
(b) Spread of plastic zones
 ELASTO-PLASTIC SOLUTIONS OF ENGINEERING PROBLEMS 99

APPENDIX
Determination of nodal forces corresponding to ‘initial’ stresses.
The displacement vector {f) at any point of an element is defined in terms of its nodal
displacements { as

{fl = “I (61‘
where [N]are the chosen ‘shape functions’.
Thus the total strains {E} will now also be defined by substitution into strain-displacement
relations as
(4 = CBI (61‘
If now a system of initial stresses {AQ) exists within an element and is to be balanced by a set of
nodal forces {P},’ we must have for any variation S{6)’ the equality of internal and external
work.
Thus

{Pi)” 6{6)’ = /{An}‘ 6 { ~ }d(vo1.) = [/(A{Q}’ [B] d(vol.))] 6{a}‘

or

{Pi}‘ = SIB]‘ A{Q} d(vo1.)

REFERENCES
1. R. H. Gallagher, J. Padlog and P. B. Bijlard, ‘Stress analysis of heated complex shapes’, J. Am. Rocket SOC.,
32, 700-707 (1962).
2. J. H. Argyris, ‘Elasto-plastic matrix displacement analysis of three dimensional continua’, Jf. R. ueronuut.&c.,
69,633-5 (1965).
3. G. G. Pope, ‘A discrete element method for analysis of plane eIast+plastic strain problems’, R.A.F. Furdorough
T.R.65028 (1965).
4. T.L. Swedlow, M.L. Williams and W.M.Yang, ‘Elasto-plastic stresses in cracked plates’, Culcit, Report SM,
65-19, California Institute of Technology (1965).
5. P. V. Marcal and I. P. King, ‘Elastic-plastic analysis of two dimensional stress systems by the finite element
method‘, Int. J. mech. Sci.,9, 143-155 (1967).
6. S. F. Reyes and D. U. Deere, ‘Elasto-plastic analysis of underground openings by the finite element method’,
Proc. 1st. int. Congr. Rock Mechunics, 11,477-86, Lisbon (1966).
7. E. P. Popov, M. Khojasteh-Bakht and S. Yaghmai, ‘Bending of circular plates of hardening material’, Intern.
J. Sol. Struct., 3, 975-988 (1967).
8. 0.C . Zienkiewin, S. Valliappan and I. P. King, ‘Stress analysis of rock at a “no tension” material‘, Giorechnique
18, 1,5666 (1968).
9. 0. C . Zienkiewin, The finite element method in structural and continuum mechanics, Mffiraw-Hill, New York
and London, 1967.
10. R. von Mises, ‘Mechanik der plastichen fonnanderung der kristallen’ 2.ungew. Math. Mech., 8, 161-85 (1928).
11. D. C. Drucker, ‘A more fundamental approach to plastic stress-strain solutions’, Proc 1st U.S.nutn. Cong.
uppl. Mech., 487-91 (1951).
12. W. Prager, ‘A new method of analysing stress and strain in work hardening plastic solids’, Revue MPc. uppl.,
Buc.. 23, 493-6 (1956).
13. W. T. Koiter, ‘Stress-strain relations, uniqueness and variational theorems for elastic plastic materials with a
singular yield surface,’ Q . uppl. Math., 11,350-54 (1953).
14. Y. Yamada, N. Yoshimura and T. Sakurai, ‘Plastic Stress-Strain Matrix and its Application for the Solution
of Elastic-Plastic Problems by the Finite Element Method’, Int. J. Mech. Sci., 10, 343-354 (1968).
100 0. C. ZIENKIEWICZ, S. VALLIAPPAN AND 1. P. KING

15. J. N. Goodier and P. G. Hodge, Jr., The Mathematical Theory of Elasticity andplasticity, Wiley, New York
(1958).
16. P.S. Theokaris and E. Marketos, ‘Elastic-plastic analysis of perforated thin strips of strain-hardening material‘,
J . Mech. Phys. Sol., 12, 377-90 (1964).
17. D. C. Drucker and W. Prager. ‘Soil mechanics and plastic analysis or limit design’, Q . appl. Moth, 10, 157-165
’ 11952).
\----I-

18. A. W. Bishop, ‘The strength of soils as engineering materials’, Ghtechnique, 16,91428 (1966).
19. A. G. Maier, ‘Onelastic-plastic structures with associated stress-strain solutions allowing for work softening’,
J. Italian Assoc. Th. appl. Mech., 2, 1, 55-64 (1967).
20. A. C. Palmer, ‘On stress-strain relations for soils’, Brown Univ. Reporr No. 57. G.P. 1115/19, March (1965)
21. Z. Mroz, “on associated laws in plasticity’, J. MPc. Phys. Arms., 2, 7, 2 1 4 1 (1963).
22. S. Valliappan, Non-Linear Stress Analysis of two-dimensionalProblems with Special Reference to Rock and Soil
Mechatucs, Ph.D. Thesis. University of Wales (1968).
23. W. D. Liam Finn, ‘Applications of limit plasticity in soil mechanics’, J. Soil Mech. Fdns. Div. Am. Soe. civ.
Engrs, 93, SM5 Part I, 101-120 (1967).