Documente Academic
Documente Profesional
Documente Cultură
0. C. ZIENKIEWICZ
Department of Civil Engineering, University College of Swansea, Swansea SA2 8PP, UK
In this paper we present the survey of research carried out over the past ten years at University
College of Swansea under the guidance of the author to determine a rational approach to the study
of foundation and other soil mechanics problems. The paper starts with a description of the need
for numerical approaches utilizing finite element or similar methodology and discusses various
constitutive models for static soil behaviour. Plasticity is adopted to describe the non-linear
characteristics of soil. A series of tests on ideally elasto/plastic, associative and non-associative,
models and on an extended critical state model show that with the latter it is possible to obtain
good predictions of the behaviour for drained and undrained behaviour of normally consolidated
materials and indeed to extend the results to over-consolidated situations. The remainder of the
paper concerns itself with the cyclic and transient load behaviour. Here the well known increase of
pore pressure under repeated loading has to be accounted for as this can either lead to liquefaction
or a very considerable weakening of the material. Two alternative approaches are proposed. In the
first a concept of an autogenous densification of the material is introduced to supplement the
original elasto/plastic models and this is shown to be effective in predicting liquefaction of sands.
An alternative model modifies the critical state using methods proposed by Mroz to describe
behaviour of clays more accurately. Finally, the paper deals with shakedown or ratchetting type
problems in which it is possible to obtain collapse without material deterioration merely by a
sufficient number of cyclic load repetitions. The numerical methods of dealing with such problems
are discussed.
0141-1187/80/010023~9/%02.00
0 1980 CML Publications Appl. Ocean Res. 1980, Vol. 2, No. 1 23
Constitutive laws and numerical analysis Jbr soil foundations: O. C. Zienkiewicz
can be reasonably assured by limit computations while such 'effective stress' conditions. In other publications it is
very simplified one-dimensional consolidation type so- shown how the knowledge of such basic constitutive laws
lutions are adequate for prediction ofdeformatfons. In the allows undrained, consolidated, or dynamic behaviour to
view of the author this is a considerable oversimplification be readily deduced 2- 5.
and potentially dangerous when large structures are When choosing the plasticity model the most impor-
considered. While limit load computations are well tant feature must be its ability to reproduce the collapse or
developed for two-dimensional 'constant cohesion' type peak stress situation which is the best known quantity for
materials obeying purely cohesive type behaviour of soils and which corresponds to a Mohr Coulomb surface
associated plasticity, the real situation is not so readily in the principal stress space. We shall propose to make a
treated and some of the difficulties are cited below: (1) selection from a range of three basic general possibilities
under undrained conditions normally consolidated clays to define reasonably well plastic soil behaviour.
show an apparent cohesion increasing with depth for A. Ideal associative elasto-plasticity. The Mohr
which few solutions of limit type are available; (2) in Coulomb surface is assumed to act as plastic yield and
drained conditions with very small cohesion existing the potential surface. This together with its triaxial stress
limit load calculations are generally not applicable due to section is shown in Fig. l(a).
the non-associativity of the flow rule and hence only in B. Ideal non-associative, elasto-plasticity. Here the mo-
certain cases can reasonable predictions be accepted; (3) del A is extended by defining a set of plastic potential
overconsolidated behaviour of clays can not readily be surfaces of the same type but not parallel to the Mohr
treated in terms of total stress analysis using simple limit Coulomb yield surface Q(a). With the flow rule given as:
theorems; (4) for complex load and non-homogeneous
materials as well as for three-dimensional conditions very
few limit solutions are applicable and here full defor- dep= ~Q),
O,a" (3)
mation solutions are desirable.
Where deformations have to be assessed a need for this allows a more realistic dilatancy (or in fact a zero
some solutions of the boundary value problems exist and dilatancy) to be imposed on the material during yield thus
here frequently the resort is made to linear finite element approximating in a better way the true behaviour. This
analysis with suitable adjusted moduli. It is a contention model is shown in Fig. l(b).
of the author that in the present state of development it is C. An associative strain hardening plastic, critical state
economically feasible to treat the deformation and col- model. This is based on the well known classic model
lapse behaviour in a unified way without excessive cost derived by Roscoe and his collaborators 5 - 7. We shall use
and thus avoiding the difficulties mentioned above. This,
however, requires the development of reasonable con-
stitutive models and we shall now turn to their
determination. [ ol
F(o, e p) = a (1)
e =e e +e p (2) <~O2=O3
As it is possible to describe fully the behaviour of
Triaxiol section OAB
saturated soil defining the law governing the drained or Figure l(a) Model A - - ideal associated plasticity with
(skeleton) component we shall be only concerned with Mohr Coulomb yield surface.
Load q (Ib/in.2}
0 100 200
Critical state B
"~'
- ~-C
. _1'2 q R=5~ bC'~ k
"-" -- = 5ft.-
Triaxial section
AO. ~ ~~.P~p I / OAB o 2.4-- Mesh
~.
Yield and J' ~ °1>oz=o3
mtential / / " I
a
i Data • c = 10 Ib/in 2
2.s -- ~,= 20 o
E = 30,000 Ib/in. 2
F l a , e v ) ~ v=O.3 o
Critical state" J~ o1<02 0'3
,,,~l",.uz-
Figure 2. Axi-symmetric Jboting (uniform load) drained
Figure l(c) Model C - - strain hardening critical state load-deformation behaviour for three material models. A,
associated plasticity with Mohr Coulomb critical state Ideal associated plasticity (Mohr Coulomb); B, ideal non-
surface. associated plasticity; C, extended critical state.
I
Winnicki, 1978); Critical state with Mohr Coulomb cut-off:
E=12000 kN/m2, cp=30”, v=O.3, H=25. Non-
associated flow rule f
6
Computations
Same remarks should be made with regard to the
numerical finite element processes used in the solution of
realistic problem utilizing above models. The choice of
model may to same extent be influenced by campu-
tatianal techniques used’. Thus if tangential methods are Figure 7(a). A three-dimensional, non-linear Fourier Scotts
used the critical state model is preferable to a non- solution for a footing problem. Element mesh in r-z plane.
associative one as tangent matrix is then symmetric. On Footing: E =2.0 x lo6 kN/m’; v=O.3.
the other hand, with initial stress techniques (or equiva- Soil: E = 1.O x lo4 kN/m2; v = 0.3.
lent viscaplastic processes) the ideally plastic model is fly =cahesion = 50 kN/m’; K, = 1.0; p = weight = 20
preferable as mare rapid convergence occurs and nan- kN/m2
assaciativity is not important.
In the context of two-dimensional plane strain or axi-
symmetric, static or dynamic, analysis, the solution
techniques are widely available at moderate cast with all
the approaches. In full three-dimensional situations,
Limit load ratio
however, we find that the cast is still large and refinements
of numerical methodology are proceeding. Here iterative 3D/2D =l-72
procedures will .ultimately became the basis and distin-
ctions between tangent and initial stress methods will
disappear.
In the intermediate case of axi-symmetric bodies sub-
ject to non-axi-symmetric loads such as occur frequently
in offshore platform analysis, an alternative to full three-
dimensional analysis exists using a Fourier type expan-
sion in the circumferential direction. This reduces the cast
of full three-dimensional analysis considerably and details
of the process are described elsewhere. However, it should
be noted that for such methods an initial stress technique
is essential and hence preference far simple, non-strain ttlement of the
hardening models exists I3 . In Fig. 7 same results of a nan-
linear analysis carried out far a three-dimensional loading
of a footing are shown.
special problems. Here tests indicate that in bath clays Figure 7(b). Behaviour of problem of Fig. 7(a) first
and sands cyclic stress reversals progressively increase the increasing vertical load, OA’ (or OA”), then the horizontal
pare pressure and thus, for undrained conditions, will lead load up to collapse.
.::f
The work of Seed, Finn and others ~8-21 has indicated
that under cyclic loading dry soils (sands) densify while
saturated ones develop considerable pore pressure. If the
i0'- A . . . . . . . .
amount of densification can be related in some manner to
the stress and strain path the material undergoes, both
effects can be predicted by the same phenomenon. Clearly
when the materials are saturated and the skeleton shows a
°°' I/b xAAAAA A A
o., v ,,v vvVW" tendency to contract it will transfer a considerable
proportion of its mean total stress onto the fluid or water
in the pores. The quantification of these effects is simple
and straightforward in the analysis once such 'initial'
Time, s strain can be predicted. Writing thus that the total strain
(2) can be given as:
e =e e +e p +e a (4)
7
analysis. Zienkiewicz et al. 14 suggest such a law for a
particular sand and show that its effects can be taken quite
simply into account in the dynamic non-linear com-
putation. In Fig. 8, we show an analysis of a quay wall in
2 .®©
which such densification is now introduced in addition to
elastic-plastic strain of model B discussed previously. The
formula used here relates autogenous strain to (a) the ratio
.,t ....
of the deviatoric to mean stresses and (b) the total length
of the strain path to be measured in absolute terms. Thus
dd=mde~ mr=0, 1, 1, 0, 0, O)
A
de~ =f(~c)d~c f(t¢) = - -
I+B~c
(5)
A i~¢ 4 =ec
d~c = g(#/a,.)d( g(6/a,.) = exp (Ta/am)
the analysis of an undrained kind using the drained soil were applied to soils 23. This final model is described
properties and thus progressively weakening the material. qualitatively in the mean stress section of the critical state
This insertion of pore pressure obviously will lead to the plasticity model in Fig. 10. In this model we retain the
same answers providing it can be related to the strain path normal critical state concepts but use the previous
and stress levels and perhaps as such presents a more plasticity surface not as a yield condition but as an
direct approach. If drainage does however occur, it envelope of all yield conditions defined by ellipses of
appears necessary to use the densification concept di- smaller size (usually given as a fraction of the exterior
rectly. In Fig. 9 we show how the proposed densification ellipse size) which are capable of moving in the interior as
formula takes into account experimental data observed in the stress point moves within. By imposing a suitable flow
a sand. rule it is easy to achieve compaction or densification of the
material both for increasing and decreasing stress mo-
Modified material model II - - isotropic hardening concepts tions within the overall critical state ellipse. This model
The modified material model discussed in the preceding has now been considerably refined and is well capable of
section is perhaps too pragmatic and lacks the aesthetic responding to load cycles. In Fig. 11 we show some results
appeal of a single model in which the densification effects tfiken from a recent paper 17 in which the densification
could be incorporated. The origin for the development of with increasing cycle number is indicated. The model is
such a model now appears in two sources. In the first place not only capable of showing this densification but im-
we have the original work of Mroz ~5 where a series of proves the general performance of the critical state models
kinematically hardening surfaces nesting within each if refinement is needed and adapts this model better to
other was suggested to describe metal plasticity. In the
second overlay concepts introduced by Zienkiewicz et al. f.~c
(¢)
-B (o)
02 / A m
om
(¢v)
.0.1
to (b)
o~, (EJ f //
s j ~ v • experimental data /
0.01 0.1
' 1.'o '
100
Figure 9. Autogenous volumetric strain versus ~: (N.G.I. Figure 10. Critical state plasticity, (a) single surface, (b)
sand) two surfaces.
2-
-L
I-
lillE/ 'c) 4.
I -Io X = 100 41
Eq103 w°"
- 21
$
Ks= 103
(d)
-q
Figure 11. Cyclic loading under drained triaxial conditions. (a) Shear stress vs. volumetric strain; (b) volumetric strain vs.
number of cycles; (c) shear stress vs. shear strain; (d) shear strain (at end of cycle) vs. number of cycles.
o0;
~ / R,g,(~ elements (ep~ese~1 p,c~tcr~ ,r t • o~o )s,s
o~
~1 ¢ ~= ib i i / ~ ~ 5C,I p'ope,!,es
N° boll-CyCles
0 2 t ~ ; ,,o ,~ 't ?
Mo~,-Coulomb e:30" '~:O" ~ ~-C ¢':2a"
( / , X ,oo . . . . f ....... . , 30°
'
/H
.
I
of ox,symmelr,c
Moment kN ~
mesh
~
lol
J BuI* rnoO~lus : 5000
~
E
020
o,o
0 50
I\.\ \
\.
-~
\.
Slroln- hofaenlng
\
[Initial
Moment = 200
- X : 2~ 0 ~ : ~ 0
vertlcot
Stro,e hordemng -- X
kNm
pressure
S -h
v,~th M-C [~=O')r.ul-ofI C -c
k
: 200 k%/w, 2
per metres1rp
: 25 0.
o~=Oy =
X : 25C
s!eess : ".
. /,.,./ ¢ = 20"
[ , ~ " p. . . . . (...... )
. \
~,0 i';';-- ,~- ..............
\.
X :500
(el
{b)
?. , w2
C/$p,ocement clue I0 Verhccl O,$DJocemenl clue Io vefhCal
I~e o~ly Se(} bea -~ Ioocl ,°~lY
2" -+-010
C 20
v, o ~,ae,,o~ - MOb,-Coulomb /
¢111n¢01 slale
t 0 ~0 ./L// ~7_.
it)
Cychc momenl ]. Cychc moment ~7,
(el e~2:[
N" of holf - c y c l e s
o,o i 2. ., . ~ . , . ,,0 t~ , t t
,
I X = 50 0 melre
~- ' ~, : 30"
No o, t ~ c ,y,d.,
o~o
-~
'~ 0 11 ~2~--_....... -.~-'-....... x : 20001
I 200 w ~ / ~ ? ...... , ........ p ........ (gl
Figure 12. Cyclic loading of a Jootin9 under undrained conditions. (a) Finite element mesh and soil properties; (b)
displacements due to vertical loading followed by a monotonic moment; (c) displacement due to vertical loadingjbllowed by a
cyclic moment; (d) vertical central displacement due to cyclic moment for Mohr Coulomb and strain-hardenin 9 yield
conditions; (f) excess pore pressure due to moment after 3/4 cycle (kN/m2); (g) variation of excess pore pressures due to cyclic
moment.
overconsolidated conditions. In ref. 17, a full description collapse. This is a well known phenomenon of ratchetting
of the performance of the model is given and in the near and in Fig. 12 we show an example using a total stress
future we expect that this model will be able to be adjusted analysis which shows how at loads well below those of
so as to give all observable pressure rise characteristics. collapse continuing deformation may occur for a typical
foundation due to reversals of wave loading ultimately
Shakedown and ratchettin9 problems causing unserviceability or indeed incremental collapse.
With a structural system subject to loads during which In the case illustrated a large number of re-analyses were
plastic deformation occurs and if further, some loads are carried out. This is feasible in a simple case but is very
subject to reversal while others remain constant, pro- costly. For practical purposes more sophisticated ap-
gressive plastic deformation may develop in each cycle proaches will have to be developed to tell the engineer
even if the total load combination is well below that of whether such progressive collapse is likely to occur. This