Escolar Documentos
Profissional Documentos
Cultura Documentos
Copyright
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, electronic, mechanical, photocopying,
recording, scanning or otherwise, except under the terms of the Copyright, Designs and Patents
Act 1988 or under the terms of a licence issued by the Copyright Licensing Agency, 90
Tottenham Court Road, London, UK WIP 9HE, without the permission in writing of the
publisher.
Other Wiley Editorial Offices
A catalogue record for this book is available from the British Library
ISBN 0-471-98285-7
Typeset in 10/12.25pt Times from the author's disks by Pure Tech India Ltd, Pondicherry
Printed and bound in Great Britain by Bookcraft (Bath) Ltd, Midsomer Norton, Somerset
This book is printed on acid-free paper responsibly manufactured from sustainable forestry,
in which at least two trees are planted for each one used for paper production.
Contents
Preface
>
vi
CONTENTS
4 Constitutive Relations-Plasticity
4.1 Introduction
4.2 The general Framework of Plasticity
4.2.1 Phenomenological aspects
4.2.2 Generalized plasticity
4.2.3 Classical theory of plasticity
4.3 Critical State Models
4.3.1 Introduction
4.3.2 Critical state models for normally consolidated clay
4.3.3 Extension to sands
4.4 Advanced Models
4.4.1 Introduction
4.4.2 A generalized plasticity model for clays
4.4.3 A generalized plasticity model for sands
4.4.4 Anisotropy
4.5 Modified Densification Model
4.5.1 Densification model for cyclic mobility
References
CONTENTS
5.3
5.4
5.5
5.6
5.7
vii
viii
CONTENTS
7.3.1
CONTENTS
Author Index
Subject Index
Preface
Although the concept of effective stress in soils is accepted by all soil mechanicians,
practical predictions and engineering calculations are traditionally based on total
stress approaches. When the senior author began, in the early seventies, the application of numerical approaches to the field of soil mechanics in general and to soil
dynamics in particular, it became clear to him that a realistic prediction of the
behaviour of soil masses could only be achieved if the total stress approaches were
abandoned. The essential model should consider the coupled interaction of the
soil skeleton and of the pore fluid. Indeed, the phenomena of weakening and
of 'liquefaction' in soil when subjected to repeated loading such as that which occurs
in earthquakes, can only be explained by considering this 'two-phase' action and
the quantitative analysis and prediction of real behaviour can only be achieved by
sophisticated computation. The simple limit methods often applied in statics are no
longer useful.
It therefore seems necessary at the present time to present, in a single volume, the
basis of such computational approaches because a wider audience of practitioners
and engineering students will require the knowledge which hitherto has only been
available through scientific publications scattered throughout many journals and
conferences. The present book is an attempt to provide a rapid answer to this need.
The multiple authorship not only ensures a speedy result, it also introduces
members of the research team which, during the last decade, have focused attention
on the subject which has developed practical computer codes which are now available
to both practitioners and researchers.
Since 1975 large number of research workers, both students and colleagues, have
participated both at Swansea and elsewhere in laying the foundations of numerical
predictions which were based largely on concepts introduced in the early forties by
Biot. However, the total stress calculation continues to be used by some engineers for
earthquake response analysis, often introduced with the linear approximations. Such
simplifications are generally not useful and can lead to erroneous predictions.
In recent years, centrifuge experiments have permitted the study of some soil
problems involving both statics and dynamics. These provide a useful set of benchmark predictions. Here a validation of the two-phase approach was available and
close agreement between computation and experiment was found. A very important
landmark was a workshop held at the University of California, Davis, in 1993, which
PREFACE
xi
Figure 1.1 The Vajont reservoir, failure of Mant Toc in 1963 (Oct. 9th): (a) hypothetical slip
plane; (b) downhill end of slide (Miiller, 1965) Plate 1 shows a photo of the slides (front page)
tion, however, escapes immediate attention. This is due to the 'weakening' of the soilfluid composite during the periodic motion such as that which is involved in an
earthquake. However, it is this rather than the overall acceleration forces which
caused the collapse of the Lower San Fernando dam. What appears to have happened here is that during the motion the interstitial pore pressure increased, thus
reducing the interparticle forces in the solid phase of the soil and its strength.
Such strength reduction phenomena are mainly evident in essentially non-cohesive materials such as sand
and slit. Clays in which negative, capillary pressure provide an apparent cohesion are less liable to such
strength reduction.
This phenomenon is well documented and in some instances the strength can drop
to near zero values with the soil then behaving almost like a fluid. This behaviour is
known as soil liquefaction and Plate 2 shows a photograph of some buildings in
Niigata, Japan taken after the 1964 earthquake. It is clear here that the buildings
behaved as if they were floating during the active part of the motion.
In this book, we shall discuss the nature and detailed behaviour of the various
static, quasi-static and dynamic phenomena which occur in soils and will indicate
how a computer based, finite element, analysis can be effective in predicting all these
aspects quantitatively.
1.2
T H E NATURE OF SOILS
Solid
/ I
Figure 1.3 Various idealized structures of fluid saturated porous solids: (a) a granular
material; (b) a perforated solid with interconnecting voids
Such drained behaviour, however, seldom occurs even in problems which we may
be tempted to consider as static due to the slow movement of the pore fluid and.
theoretically, the infinite time required to reach this asymptotic behaviour. In very
finely grained materials such as silts or clays this situation may never be established
even as an approximation.
Thus, in a general situation, the complete solution of the problem of solid material
deformation coupled to a transient fluid flow needs generally to be solved. Here no
short-cuts are possible and full coupled a n ~ l ~ v s eofs equations which we shall introduce in Chapter 2 become necessary.
We have not mentioned so far the notion of so called undrained behaviour, which
is frequently assumed for rapidly loaded soil. Indeed, if all fluid motion is prevented,
by zero permeability being implied or by extreme speed of the loading phenomena, the pressures developed in the fluid will be linked in a unique manner to
deformation of the solid material and a single-phase behaviour can again be specified. While the artifice of simple undrained behaviour is occasionally useful in
static studies, it is not applicable to dynamic phenomena such as those which
occur in earthquakes as the pressures developed will, in general, be linked again to
the straining (or loading) history and this must always be taken into account.
Although in early attempts to deal with earthquake analyses and to predict the
damage and response, such undrained analyses were invariably used, adding generally a linearization of the total behaviour and an heuristic assumption linking
the pressure development with cycles of loading, the behaviour predictions were
poor. Indeed recent comparisons with centrifuge experiments confirmed the inability
of such methods to predict either the pressure development or deformations
(VELACS
Arulanandan & Scott, 1993). For this reason we believe that the
only realistic type of analysis is of the type indicated in this book. This was demonstrated in the same VELACS tests to which we shall frequently refer in later
chapters.
-
1.3
note
The essential concepts defining the stresses which control strength and constitutive
behaviour of a porous material with internal pore pressure of a fluid appear to have
been defined, at least qualitatively towards the end of the last century. The work of
Lye11 (1871), Boussinesq (1876) and Reynolds (1886) was here of considerable note
for problems of soils. Later, similar concepts were used to define the behaviour of
concrete in dams (Levy, 1895 and Fillunger, 1913a, 1913b and 1915) and indeed for
other soil or rock structures. In all of these approaches the concept of division of the
total stress between the part carried by the solid skeleton and the fluid pressure is
introduced and the assumption made that the strength and deformation of the
skeleton is its intrinsic property and not dependent on the fluid pressure.
If we thus define the total stress a by its components aq using indicia1 notation
these are determined by summing the appropriate forces in the i-direction on the
projection, or cuts, dx, (or dx, dy, dz in conventional notation). The surfaces of cuts
are shown, for two kinds of porous material structure, in Figure 1.3 and include the
total area of the porous skeleton.
In the context of the finite element computation we shall frequently use a vectorial
notation for stresses, writing
This notation reduces the components to six rather than nine and has some computational merit.
Now if the stress in the solid skeleton is defined as the effective stress d again over
the whole cross-sectional area then the hydrostatic stress due to the pore pressure,
p acting, only on the pore area should be
where n is the porosity and Sji is the Kronecker delta. The negative sign is introduced
as it is a general convention to take tensile components of stress as positive.
The above, plausible, argument leads to the following relation between total and
effective stress with total stress
(1.4)
The above arguments do not stand the test of experiment as it would appear that,
with values of porosity n with a magnitude of 0.1-0.2 it would be possible to damage a
specimen of a porous material (such as concrete for instance) by subjecting it to
external and internal pressures simultaneously. Further, it would appear from equation (1.3) that the strength of the material would be always influenced by the pressure p.
Fillunger introduced the concepts implicit in (1.3) in 1913 but despite conducting
experiments in 1915 on the tensile strength of concrete subject to water pressure in
the pores, which gave the correct answers, he was not willing to depart from the
simple statements made above.
It was the work of Terzaghi and Rendulic in 1934 and by Terzaghi in 1936 which
finally modified the definition of effective stress to
where n,,.is now called the effective areu coef$cient and is such that
1.3.2
Let us now consider the effect of the simultaneous application of a total external
hydrostatic stress and a pore pressure change; both equal to A p , to any porous
material. The above requirement can be written in tensorial notation as requiring
that the total stress increment is defined as
In the above, the negative sign is introduced since 'pressures' are generally defined as
being positive in compression, while it is convenient to define stress components as
positive in tension.
It is evident that for the loading mentioned, only a very uniform and small
volumetric strain will occur in the skeleton and the material will not suffer any
damage provided that the grains of the solid are all made of an identical material.
This is simply because all parts of the porous medium solid component will be
subjected to an identical compressive stress.
However, if the microstructure of the porous medium is composed of different
materials, it appears possible that non-uniform, localized stresses, can occur and that
local grain damage may be suffered. Experiments performed on many soils and rocks
and rock-like materials show, however, that such effects are insignificant. Thus in
general the grains and hence the total material will be in a state of pure volumetric
strain
CONCEPTS OF EFFECTIVE S T R E S S
where K, is the average material bulk modulus of the solid components of the skeleton.
Alternatively, adopting a vectorial notation for strain in a manner involved in (1.1)
where E is the vector defining the strains in the manner corresponding to that of stress
increment definition. Again, assuming that the material is isotropic, we shall have
Those not familiar with soil mechanics may find the following hypothetical experiment illustrative. A block of porous, sponge-like rubber, is immersed in a fluid to
which an increase in pressure of A p is applied as shown in Figure 1.4. If the pores are
connected to the fluid, the volumetric strain will be negligible as the solid components
of the sponge rubber are virtually incompressible.
Frame after load
Negligible deformation
(a)
AP
(b)
Surface
membrane
Figure 1.4 A Porous Material subject to external hydrostatic pressure increases A,. and (a)
Internal pressure increment A,; (b) Internal pressure increment of zero
10
If, on the other hand, the block is first encased in a membrane and the interior is
allowed to drain freely, then again a purely volumetric strain will be realised but now
of a much larger magnitude.
The facts mentioned above were established by the very early experiments of
Fillunger (1915) and it is surprising that so much discussion of "area coefficients"
has since been necessary.
From the preceding discussion it is clear that if the material is subject to a
simultaneous change of total stress Ao and of the total pore pressure Ap, the
resulting strain can always be written incrementally in tensorial notation as
or in vectoral notation
with
The last term in (1.1 la) and (1.1 1b), AEO, is simply the increment of an initial strain
such as may be caused by temperature changes etc., while the penultimate term is the
strain due to the grain compression already mentioned viz Eq. 1.10 D is a tangent
matrix of the solid skeleton implied by the constitutive relation with corresponding
compliance coefficient matrix D-' = C. These of course could be matrices of constants, if linear elastic behaviour is assumed, but generally will be defined by an
appropriate non-linear relationship of the type which we shall discuss in Chapter 4
and this behaviour can be established by fully drained @ = 0) tests.
Although the effects of skeleton deformation due to the effective stress defined by
(1.6) with n,,= 1 have been simply added to the uniform volumetric compression, the
principle of superposition requiring linear behaviour is not invoked and in this book
we shall almost exclusively be concerned with non-linear, irreversible, elasto-plastic
and elasto-viscoplastic responses of the skeleton which, however, we assume incremental properties.
For assesment of strength of the saturated material the effective stress previously
defined with n,,.= 1 is sufficient. However, we note that the deforination relation of
(1.1 1) can always be rewritten incorporating the small compressive deformation of
the particles as (1.12).
It is more logical at this step to replace the finite increment by an infinitesimal one
and to invert the relations in (1.1 1) writing these as
11
and the new form eliminates the need for separate determination of the volumetric
strain component. Noting that
6..6..
- 3
!I !I -
or
we can write
or simply
which is the tangential bulk modulus of an isotropic elastic material with X and
being the Lame's constants. Thus we can write
11
The reader should note that in (1.12) we have written the definition of the effective
stress increment which can, of course, be used in a non-incremental state as
12
assuming that all the stresses and pore pressure started from a zero initial state, (for
example, material exposed to air is taken as under zero pressure). The above definition corresponds to that of the effective stress used by Biot (1941) but is somewhat
more simply derived. In the above, cr is a factor which becomes close to unity when
the bulk modulus Ks of the grains is much larger than that of the whole material. In
such a case we can write. of course
recovering the common definition used by many in soil mechanics and introduced by
von Terzaghi (1936). Now, however, the meaning of a is no longer associated with an
effective area.
It should have been noted that in some materials such as rocks or concrete it is
possible for the ratio K,/K, to be as large as 113 with a = 213 being a fairly common
value for determination of deformation.
We note that in the preceding discussion the only assumption made which can be
questioned, is that of neglecting the local damage due to differing materials in the soil
matrix. We have also implicitly assumed that the fluid flow is such that it does not
separate the contacts of the soil grains. This assumption is not totally correct in soil
liquefaction or flow in soil shearing layer during localization.
1.3.3
The interstitial space, or the pores, may in a practical situation be filled with two or
more fluids. We shall, in this section, consider only two fluids with the degree of
saturation by each fluid being defined by the proportion of the total pore volume n
(porosity) occupied by each fluid. In the context of soil behaviour discussed in this
book the fluids will invariably be water and air respectively. Thus we shall refer to
only two saturation degrees S, that for water and S, that for air, but the discussion
will be valid for any two fluids.
It is clear that, if both fluids fill the pores completely, we shall always have
Clearly this relation will be valid for any other pair of fluids e.g. oil and water and
indeed the treatment described here is valid for any fluid conditions.
13
The two fluids may well present different areas of contact with the solid grains
of the material in the manner illustrated in Figure 1.5 (a) and (b). The average
pressure reducing the interstitial contact and relevant to the definition of
effective stress found in the previous section (Equations (1.16) and (1.17)) can thus
be taken as
I-' = XwPw
+ XaPa
(1.19)
The individual pressures pw and pa are again referring to water and air and their
difference i.e.
and
Occasionally the contact of one of the phases and the solid may disappear entirely as
shown in Figure 1.5a giving isolated air bubbles and making in this limit
Figure 1.5 Two fluids in pores of a granular solid (water and air). (a) air bubble not wetting
solid surface (effective pressure p = p w ; (b) both fluids in contact with solid surfaces (effective
X, pa
pressure p = X, p,
14
Whatever the nature of contact, we shall find that a unique relationship between p,
and the saturation Sw can be written i.e.
Indeed, the degree of saturation will similarly affect flow parameters such as the
permeability k to which we shall make reference in the next chapter, giving
Many studies of such relationship are reported in the literature (Liakopoulos, 1965;
Neuman, 1975; Van Genuchten et al., 1977; Narasimhan & Witherspoon 1978; Safai
& Pinder, 1979; Lloret & Alonso, 1980; Bear et al. 1984; Alonso et al., 1987). Figure
1.6 shows a typical relationship.
The concepts of dealing with the effects of multiple pore pressure by introducing
an average pressure and using the standard definition of effective stress (1.19, 1.16
and 1.17) were first introduced by Bishop (1959). Certainly the arguments for
thus extending the original concepts are less clear than is the case when only a
single fluid is present. However, the results obtained by this extension are quite
accurate. We shall therefore use such a definition in the study of partially saturated
media.
In many cases occurring in practice, the air pressure is close to zero (atmospheric
datum) as the pores are interconnected. Alternatively, negative air pressure occurs as
cavitation starts and here the datum is the vapour pressure of water. In either case the
effect of p, can be easily neglected as the water pressure simply becomes negative
from Equation (1.24). Such negative pressures are responsible for the development of
certain cohesion by the soil and are essential in the study of free surface conditions
occurring in embankments, as we shall see later
Figure 1.6 Typical relations between pore pressure head. h, = p, / X , , Saturation, S, , and
relative permeability, k, = k , (S,)/k,(l) (Safai & Pinder 1979). Note that relative permeability decreases very rapidly as saturation decreases
REFERENCES
15
REFERENCES
Alonso E. E., Gens A. and Hight D. W. (1987) Special problems of soil: general report. Proc.
9th Euro. Con/.' Int. Soc. Soil Meel?. Found. Eng., Dublin.
Arulanandan K. and Scott R . F. (Eds) (1993) Proceedings of the VELACS symposium. 1. A.
A. Balkema, Rotterdam.
Bear J., Corapcioglu M. Y. and Balakrishna J., (1984) Modeling of centrifugal filtration in
unsaturated deformable porous media, A h . Water Resour., 7 , 150-167.
Biot M. A. (1941) General theory of three-dimensional consolidation, J. Appl. P h j ~ . 12,
, 155164.
Biot M. A. (1955) Theory of elasticity and consolidation for a porous anisotropic solid, J. Appl.
Phys, 26, 182-1 85.
Biot M. A. (1956a) Theory of propagation of elastic waves in a fluid-saturated porous solid.
part I-low-frequency range, J. Acoust. Soc. Am., 28, No. 2, 168-178.
Biot M. A. (1956b) Theory of propagation of elastic waves in a fluid-saturated porous solid.
part-11-higher frequency range, J. Acoust. Soe. Am., 28, No. 2, 179-191.
Biot M. A. (1962) Mechanics of deformation and acoustic propagation in porous media.
J. Appl. Phys., 33, No. 4, 1482-1498.
Biot M. A. and Willis P. G. (1957) The elastic coefficients of the theory consolidation, J. Appl.
Mech., 24. 59&601.
Bishop A. W. (1959) The Principle of Effective Stress, Teknisk Ukeblad, 39, 859-863.
De Boer R. (1996) Highlights in the historical development of the porous media theory.
Applicd Mec1z~1nic.sReview, 49, 20 1-262.
De Boer R., Schiffman R. L. and Gibson R. E. (1996) The origins of the theory of consolidation: the Terzaghi-Fillunger dispute, Gkotechnique, 46, No. 2, 175-1 86.
Boussinesq J. (1876) Essai theorique sur I'equilibre d'elasticite des massif pulverulents. Mem.
savants h a n g e r s , Acad. Belgique, 40, 1-180.
Fillunger P. (1913a) Der Ayftrieh in Tcrlsperrcv~,0sterr. Wochenschrift offentlichen Baudienst.
532-556.
Fillunger P. (1 9 13b) Der Aujirieh in Tul~sperren,Osterr. Wochenschrift offentlichen Baudienst.
567-570.
Fillunger P. (1915) Versuch iiher die Zugfestigkeit he; crllseitigem Wasserdruck, 0sterr.
Wochenschrift offentl. Baudienst, H29, 443448.
Leliavsky S. (1947) Experiments on effective area in gravity dams, Trans. Am. Soc. Civil
Engrs., 112, 444.
Levy M. M. (1895) Quelques Considerations sur la construction des grandes barrages, C o w
ptes Rendus De L'Academie Des Sciences Serie I-Mathematique. 288.
Liakopoulos A. C. (1965) Trczn.sient,flow through unsrrturated porous media, D. Eng. Dissertation, University of California, Berkeley, USA.
Lloret A. and Alonso E. E. (1980) Consolidation of unsaturated soils including swelling and
collapse behaviour, GPotechnique, 30, 449447.
Lyell C. (1871) Student's elements of geology, London.
McHenry D. (1948) The effect of uplift pressure on the shearing strength of concrete-R.48.
International Congress Large Dams, 3rd, Stockholm, Vol. I.
Miiller L. (1965) The Rock slide in the Vajont Valley, Fels Mecharzik, 2, 148-212.
Narasimhan T. N. and Witherspoon P. A. (1978) Numerical model for saturated-unsaturated
flow in deformable porous media 3. Applications, Water Resources Res., 14. 1017-1034.
Neuman S. P. (1975) Galerkin approach to saturated-unsaturated flow in porous media in
Finite Elernents in jluids, Wiley, London.
Reynolds 0. (1886) Experiments showing dilatancy, a property of granular material. Proc. R.
Inst., 11, 354363.
16
Safai N. M. and Pinder G. F. (1979) Vertical and horizontal land deformation in a desaturating porous medium, Adv. Water Resources 2, 19-25.
Seed H. B. (1979) Consideration in the earthquake resistant design of earth and rockfill dams,
GCotechnique, 29, No. 3, 215-263.
Seed H. B., Idriss I. M., Lee K. L. and Makdisi F. I. (1975) Dynamic analysis of the slide in the
Lower San Fernando dam during the earthquake of February 9, 1971, J. Geotech. Eng.
Div., ASCE, 101, No. 9, 889-911.
Serafim J. L. (1954) A subpresseo nos barreyens-Publ. 55, Laboratorio Nacional de Engenheria Civil, Lisbon.
Serafim J. L. (1964) The 'uplift area' in plain concrete in the elastic range-C. 17, Int. Congr.
Large Dams, 8th, Edinburgh, Vol. V.
Skempton A. W. (1960) Effective Stress in Soils, Concretes, and Rocks, Proc. Con$ Pore
Pressures and Suction in Soils, 4-16.
Terzaghi K . von (1936) The shearing resistance of saturated soils, Proc. 1st ICSMFE, 1 , 5 4 5 6 .
Terzaghi K . von and Rendulic L. (1934) Die wirksame Flachenporositat des Betons, Z. Ost.
Ing.-u. Archit Ver., 86, No. 112, 1-9.
Van Genuchten M. T., Pinder G. F. and Saukin W. P. (1977) Modeling of leachate and soil
interactions in an aquifer - EPA-60019-77-026, Proc. 3rd Annual Municipal Solid Waste
Res. Symp., 95-103.
Zienkiewicz 0.C. (1947) The stress distribution in gravity dams, J. Inst. Civ. Eng., 27,244271.
Zienkiewicz 0. C. (1963) Stress analysis of hydraulic structures including pore pressures
effects, Water Power, 15, 104108.
19
We recall first the effective stress and constitutive relationships as defined in equations (1.16), of the previous chapter which we repeat below.
This effective stress is conveniently used as it can be directly established from the
total strains developed.
However, it should be remembered here this stress definition was derived in the
first chapter as a corollary of using the effective stress defined as below
which is responsible for the major part of the deformation and certainly for failure.
In soils, the difference between the two effective stresses is negligible as cu z 1.
However, for such materials as concrete or rock the value of o in the first expression
can be as low as 0.5 but experiments on tensile strength show that the second
definition of effective stress is there very much more closely applicable as shown by
Leliavsky (1947), Serafim (1954) etc.
For soil mechanics problems, to which we will devote most of the examples, o = 1
will be assumed. Constitutive relationships will still, however, be written in the
general form using an incremental definition
(2.3b)
The vectorial notation used here follows that corresponding to stress components
given in (I. I). We thus define the strains as
In the above, D is the 'tangent matrix' and deOis the increment of the thermal or
similar autogeneous strain and of the grain compression mp/3Ks. The latter is
generally neglected in soil problems.
If large strains are encountered, this definition needs to be modified and we must
write
where the last two terms account for simple rotation (via the definition in (2.6b) ) of
existing stress components and are known as the Zaremba (1903a, 1903b)-Jaumann
(1905) stress changes. We omit here the corresponding vectorial notation as this is
not easy to implement.
The large strain rotation components are small for small displacement computation and can frequently be neglected. Thus in the derivations that follow we shall do
so-though their inclusion presents no additional computational difficulties and they
are included in the computer program.
The strain and rotation increments of the soil matrix can be determined in terms of
displacement increments du; as
and
The comma in the suffix denotes differentiation with respect to the appropriate
coordinate specified. Thus
If vectorial notation is used, as is often the case in the finite element analysis, the so
called engineering strains are used in which (with the repeated index of du;,; not
summed.)
+ du,,;
dy,, =
ddu,
ax
21
+ ddu,
ay
We shall usually write the process of strain computation using matrix notation as
d~ = Sdu
where u
[u,u?,u,]
(2.8)
T
(2.9)
with corresponding changes for three dimensions (as shown in Zienkiewicz and
Taylor 1989).
We can now write the overall equilibrium or momentum balance relation for the
soil-fluid 'mixture' as
where w;
dw;
= --etc.
dt
where ps is the density of the solid particles and n is the porosity (i.e. the volume of
pores in a unit volume of the soil).
The second equilibrium equation ensures momentum balance of the fluid. If
again we consider the same unit control volume as that assumed in deriving
(2.11) (and we further assume that this moves with the solid phase) we can
write
In the above we consider only the balance of the fluid momentum and R
represents the viscous drag forces which, assuming the Darcy seepage law, can be
written as
Note that the underlined terms in (2.12) represent again the convective fluid
acceleration and are generally small. Also note that, throughout this book, the
permeability k is used with dimensions of [length]3[time]/[mass]which is different
from the usual soil mechanics convention k' which has the dimension of velocity, i.e.,
[length]/[time]. Their values are related by k = kl/p;g' where p[ and g' are the fluid
density and gravitational acceleration at which the permeability is measured.
The final equation is one accounting for mass balance of the flow. Here we balance
the flow divergence w;,~by the augmented storage in the pores of a unit volume of soil
occurring in time dt. This storage is composed of several components given below in
order of importance:
(i) the increased volume due to a change in strain i.e.: Siideii= d~~~= mTdc
(ii) the additional volume stored by compression of void fluid due to fluid pressure
increase: ndp/Kf
(iii) the additional volume stored by the compression of grains by the fluid pressure
increase: (1 - n)dp/Ks
and
(iv) the change in volume of the solid phase due to a change in the intergranular
effective contact stress (4,= og + Sup): - Bqd~,/K, = - 5
(drii +
Ks
g)
Here KT is the average bulk modulus of the solid skeleton and E;; the total
volumetric strain.
Adding all the above contributions together with a source term and a second-order
term due to the change in fluid density in the process we can finally write the flow
conservation equation
23
or in vectorial form
where
In (2.16) the last two (underlined) terms are those corresponding to a change of
density and rate of volume expansion of the solid in the case of thermal changes and
are negligible in general. We shall omit them from further consideration here.
Equations (2.1 I), (2.13) and (2.16) together with appropriate constitutive relations
specified in the manner of (2.3) define the behaviour of the solid together with its
pore pressure in both static and dynamic conditions. The unknown variables in this
system are:
The pressure of fluid (water), p = p,
The velocities of fluid flow wi or w
The displacements of the solid matrix u; or u.
The boundary condition imposed on these variables will complete the problem.
These boundary conditions are:
(1) For the total momentum balance on the part of the boundary I?, we specify the
total traction t ; ( t )(or in terms of the total stress avn, (a.
G) with n; being the ith
component of the normal at the boundary and G is the appropriate vectorial
equivalence) while for I?,, the displacement ui(u),is given.
(2) For the fluid phase, again the boundary is divided into two parts r, on which the
values of p are specified and r,,,where the normal outflow w, is prescribed (for
instance, a zero value for the normal outward velocity on an impermeable
boundary).
Summarising, for the overall assembly, we can thus write
and
Further
and
and
25
where KT is (as defined in Eq. (1.10) the bulk modulus of the solid phase and p is once
again Lame's constant. B has, of course, a value approaching unity for soil but can be
considerably lower for concrete or rock. Further, for unsaturated soils, as will be seen
from the next section, the value will be much lower (Terzaghi, 1925, Lambe and
Whitman, 1969 and Craig, 1992).
sTu- pii + pb = 0
(2.20b)
The second equation is obtained by coupling (2.13) and (2.16) using the definition
(2.14) and thus eliminating the variable w; (w). We now have, omitting density
changes
This reduced equation system is precisely the same as that used conventionally in the
study of consolidation if the dynamic terms are omitted or even of static problems if
the steady state is reached and all the time derivatives are reduced to zero. Thus the
formulation conveniently merges with procedures used for such analyses. However,
some loss of accuracy will be evident for problems in which high-frequency oscillations are important. As we shall show in the next section, these are of little importance for earthquake analyses.
In eliminating the variable w;(w) we have neglected several terms but have
achieved an elimination of two or three variable sets depending on whether the two
or three-dimensional problem is considered. However, another possibility exists for
obtaining a reduced equation set without neglecting any terms provided that the fluid
(i.e. water in this case) is compressible.
With such compressibility assumed, Equation (2.16) can be integrated in time
provided that we introduce the water displacement u ? ( u R ) in place of the velocity
w;(w). We define
where the division by the porosity n is introduced to approximate the true rather than
the averaged fluid displacement. We now can rewrite (2.16) after integration with
respect to time as
27
Table 2.1 Comparative sets of coupled equations governing deformation and flow
u - w - p equations (exact) ((2.11) (2.13) and (2.16))
sTu- pii + pb = 0
u - U, only convective terms neglected (3.72)
sTu+ a Q ( a - n ) v ( v T u )+ a e n V ( v T u ) - (1 - n)pii - p i J + pb = 0
=0
In place of (2.13):
with
Taking K,
--+
cc
a periodic solution arises after the dissipation of the initial transient in the form
= jjell"'
P = ~ e ~ etc.
l''~
and a system of ordinary linear differential equations is obtained in the frequency
domain which can be readily solved by standard procedure.
The boundary conditions imposed are as shown in Figure 2.1. Thus at
s=Lu=Ow=Oandatx=O~,=qp=O.
In Figure 2.1 (taken from Zienkiewicz et al. 1980) we show a comparison of various
numerical results obtained by the various approximations:
(i) exact equation (Biot's, labelled B)
(ii) the u - p equation approximation (labelled Z)
(iii) the undrained assumption (w = 0) and
(iv) the consolidation equation obtained by omitting all acceleration terms,
labelled C.
The reader will note that the results are plotted against two non-dimensional
coefficients:
29
where k' and k are the two definitions of permeability discussed earlier.
In the above
where L is a typical length such as the length of the one-dimensional soil column
under consideration, g is the gravitational acceleration,
is the speed of sound, T is the natural vibration period and T is the period of
excitation.
The second non-dimensional parameter is defined as
Figure 2.2 Summarizes the conclusions by indicating three zones in which various
approximations are sufficiently accurate.
We note that, for instance, fully undrained behaviour is applicable when
II, <
and when 111> lo2 the drainage is so free that fully drained condition
can be safely assumed.
To apply this table in practical cases, some numerical values are necessary. Consider, for instance, the problem of the earthquake response of a dam in which the
typical length is characterized by the height L = 50 metres, subject to an earthquake
in which the important frequencies lie in the range
we have T = = 0.1s
the parameter 112is therefore in the range
3.9
< II~< 39
Figure 2.1 The soil column -variation of pore pressure with depth for various values of rl and
B (Biot theory) - - - - z (u-p approximation theory)
c (Consolidation
"2
theory) (Solution (C) is independent of xz).Reproduced from Zienkiewicz (1980) by permission
of the Institution of Civil Engineers
31
Drained (influence
of z, negligible)
Figure 2.2 Zones of sufficient accuracy for various approximations: Zone 1, B = Z = C, slow
phenomena (w and u can be neglected) Zone 2, B = Z # C, moderate speed (w can be neglected)
Zone 111 B # Z # C, fast phenomena (w cannot be neglected only full Biot equation valid).
Definition as in Figure 2.1. Reproduced from Zienkiewicz(1980) by permission of the Institution
of Civil Engineers
AI = k p ~ , Z /=~2~k 2
pT/~f2
A2 = J L ~ / V , Z=
k=&lpg, k - kinematic permeability, f = 2L/ V, , V: = ( D + k f / n ) / p -- Pkf /pin -- kf / p l
(speed of sound in water), P = pf / p , n
0.33, P
0.33
In the previous, fully saturated, analysis, we have considered both the water
pore pressure and the solid displacement as problem variables. In the general case of
non-linear nature which is characteristic for the problems of soil mechanics both the
effective stresses and pressures will have to be determined incrementally as the solution
process (or computation) progresses step by step. In many soils we shall encounter a
process of 'densification' implied in the constitutive soil behaviour. This means that
the history of straining (associated generally with shear strain) induces the solid matrix
to contract (or the material to densify). Such densification usually will cause the pore
pressure to increase, leading finally to a decrease of contact stresses in the soil particles
to near-zero values when complete liquefaction occurs. Indeed generally failure will
occur prior to the liquefaction limit. However, the reverse may occur where the soil
'dilation' during the deformation history is imposed. This will imply development of
negative pressures which may reach substantial magnitudes. Such negative pressures
cannot exist in reality without the presence of separation surfaces in the fluid which is
contained in the pores, and consequent capillary effects. Voids will therefore open up
during the process in the fluid which is essentially incapable of sustaining tension. This
opening of voids will probably occur when zero pressure (or corresponding vapour
pressure of water) is reached. Alternatively air will come out of solution-or indeed
ingress from the free water surface if this is open to the atmosphere. The pressure will
not then be vapour pressure but simply atmospheric.
We have shown in Chapter 1 that, once voids open, a unique relationship exists
between the degree of saturation Swand the pore pressures pw (see Figure 1.6). Using
this relation, which can be expressed by formula or simply a graph we can modify the
equation used in Section 2.2 to deal with the problem of partial saturation without
introducing any additional variables assuming that the air throughout is at constant
(atmospheric) pressure. Note that both phenomena of densification and dilation will
be familar to anybody taking a walk on a sandy beach after the tide has receded leaving
the sand semi-saturated. First note how when the foot is placed on the damp sand, the
material appears to dry in the vicinity of the applied pressure. This obviously is the
dilation effect. However if the pressure is not removed but reapplied several times the
sand 'densgies' and becomes quickly almost fluid. Clearly liquefaction has occurred. It
is surprising how much one can learn by keeping one's eyes open!
The presence of negative water pressures will, of course, increase the strength of
the soil and thus have a beneficial effect. This is particularly true above the free water
surface or the so called phreatic line. Usually one is tempted to assume simply a zero
pressure throughout that zone but for non-cohesive materials this means almost
instantaneous failure under any dynamic load. The presence of negative pressure in
the pores assures some cohesion (of the same kind which allows castles to be built on
the beach provided that the sand is damp). This cohesion is essential to assure the
structural integrity of many embankments and dams.
33
Before proceeding, we must note that the effective stress definition is modified and
the effective pressure now becomes (viz ch. 1 sect. 1.3.3)
neglecting the weight of air. The correction is obviously small and its effect insignificant.
However (2.21) will now appear in a modified form which we shall derive here.
First, the water momentum equilibrium, Equation (2.8), will be considered. We
note that its form remains unchanged but with the variable p being replaced by p,.
We thus have
As before, we have neglected the relative acceleration of the fluid to the solid
Equation (2.14), defining the permeabilities, remains unchanged as
However, in general, only scalar, i.e. isotropic, permeability will be used here
where I is the identity matrix. The value of k is, however, dependent strongly on S,
and we note that:
(v) and a new term taking into account the change of saturation: ndSw
Adding to the above, as in Section 2.2, the terms involving density changes, on
thermal expansion, the conservation equation now becomes:
Now, however,
place
which of course must be identical with (2.17) when Sw= 1 and xW= 1 i.e. when we
have full saturation. The above modification is mainly due to an additional term to
those defining the increased storage in (2.17). This term is due to the changes of the
degree of saturation and is simply:
The final elimination of w in a manner identical to that used when deriving (2.21)
gives, (neglecting density variation):
( k j - p , , , - swpfiij
P
+ S,pfbj)),;+aEii + Q*
+ SO = 0
-
35
(2.33~)
The small changes required here in the solution process are such that we shall find it
useful to construct the computer program for the partially saturated form, with the
fully saturated form being a special case.
In the time-stepping computation, we still always assume that the parameters
S,, k , and C, change slowly and hence we will compute these at the start of the
time interval keeping them, subsequently, constant.
Previously, we mentioned several typical cases where pressure can become negative
and hence saturation drops below unity. One frequently encountered example is
that of the flow occuring in the capillary zone during steady state seepage. The
solution of the problem can of course be obtained from the general equations simply
by neglecting all acceleration and fixing the solid displacements at zero (or constant)
values.
If we consider a typical dam or a water retaining embankment shown in Figure 2.3
we note that, on all the surfaces exposed to air, we have apparently incompatible
boundary conditions. These are:
p, = 0 and w, = 0 (i.e. net zero inflow)
Clearly both conditions cannot be simultaneously satisfied and it is readily concluded that only the second is true above the area where the flow emerges. Of course
when the flow leaves the free surface, the reverse is true.
Figure 2.3 A partially saturated dam. Initial steady-state solution. Only saturation (a) and
pressure contours (b) are shown. Contour interval in (b) is 75 kPa. The Phreatic line is the
boundary of the fully saturated zone in (a)
Computation will easily show that negative pressures develop near the surface
and that, therefore, a partially saturated zone with very low permeability must
exist. The result of such a computation is shown in Figure 2.3 and indeed it will
be found that very little flow occurs above the zero-pressure contour. This contour
is in fact the well known phreatic line and the partially saturated material
procedure has indeed been used frequently purely as a numerical device for its
determination (see Desai, 1977a and b, Desai and Li 1983 etc.). Another example
is given in Figure 2.4. Here a numerical solution of Zienkiewicz et a[., (1990b) is
given for a problem for which experimental data are available from Liakopolous
(1965).
In the practical code used for earthquake analysis we shall use this partially
saturated flow to calculate a wide range of soil mechanics phenomena. However,
for completeness in Section 2.4 we shall show how the effects of air movement can be
incorporated into the analysis.
2.4
2.4.1
This part of the chapter is introduced for completeness-though the effects of the air
pressure are insignificant in most problems. However, in some cases of consolidation
and confined materials, the air pressures play an important role and it is useful to
have means for their prediction. Further, the procedures introduced are readily
37
2.4.2
The dynamics of the total mixture can, just as in Section 2.3, be written in precisely the
same form as that for a single fluid phase (see (2.1 1)). For completeness we repeat
that equation here (now, however, a priori omitting the small convective terms)
noting that
For definition of effective stress we use again (2.24) now, however, without equating
the air pressure to zero i.e. writing
P = xwpw
+ (1 - xw)pa
(2.36)
For the flow of water and air we can write the Darcy equations separately, noting that
Finally, the mass balance equations for both water and air have to be written. These
are derived in a manner identical to that used for equation (2.30). Thus for water we
have
n .
w;.; aSW&+ Sw-pw
Kw
a-n
Pw
+xwpw+ nSw + nSw- + So = 0
n
VT w+crSwmi.+Sw--t),+Kw
a-n
Pw
PW So
KS xwpW+ nSw nSw-
Ks
Pw
=0
(2.41~)
(2.41b)
n
v;.; aSaii,+ S, -pa
Ka
VTv
a-n
Ks
Pa
+ -~ a p a+ nSa + nSaPa + i0= 0
n
a-n
+ crSami.+ S, -pa
+xada + nSa + nSaPa + so = 0
Ka
Ks
-
Pa
(2.42~)
(2.42b)
Now, in addition to the solid phase displacement u;(u),we have to consider the water
presure pw and the air presure pa as independent variables.
However, we note that now (see (1.16))
and that the relation betweenp, and Sw is unique and of the type shown in Figure 1.6
of Chapter 1. p, now defines Sw and from the fact that
Sw
+ S, = 1
and
xW+ xa = I
(2.44)
39
THEORY
We have now the complete equation system necessary for dealing with the flow of air
and water (or any other two fluids) coupled with the solid phase deformation.
2.5
It has already been indicated in Section 2.1 that the governing equations can alternatively be derived using mixture theories. The classical mixture theories (see Green
(1969) and Bowen (1980,1982), Morland (1972)) start from the macro-mechanical level,
i.e. the level of interest for our computations, while the so called hybrid mixture theories
(viz. Whitaker (1977) Hassanizadeh and Gray (1979a, 197913, 1980, 1990) start from
micro-mechanical level. The equations at macro-mechanical level are then obtained by
spatial averaging procedures. Further there exists a macroscopic thermodynamical
approach to Biot's theory proposed by Coussy (1995). All these theories lead to a similar
form of the balance equations. This was in particular shown by de Boer et al. (199 1) for
the mixture theory, the hybrid mixture theories and the classical Biot's theory.
Where the theories differ, is in the constitutive equations, usually obtained from
the entropy inequality. This is shown here in particular for the effective stress
principle, because this was extensively discussed in Chapter 1. Let's consider first
the fully saturated case. For instance, Runesson (1978) shows that the principle of
effective stress follows from the mixture theory under the assumption of incompressible grains. This means that cu in equation (1.16a) or (2.1a)
g!! = g..
r,
11
VP
(2.45)
Only in this case the two formulations given by Biot's theory and the mixture theory
coincide.
In the hybrid mixture theories, the concept of the solid pressure ps i.e. the pressure
acting on the solid grains is introduced (Hassanizadeh and Gray, 1990). From the
application of the second principle of thermodynamics, it appears that
and the effective stress principle in the form of (2.46) can be derived following
Hassanizadeh and Gray (1990). Equation (2.48) holds also under non-equilibrium
conditions, it has, however, to be assumed that the solid grains remain incompressible, the same assumption as with the mixture theory.
Let us consider now the partially saturated case. There are again several conflicting expressions in literature. The first expression for partially saturated soils
was developed by Bishop (1959) and may be written as follows, see Equations
(2.36, 2.24 and 1.19)
which considering thermodynamic equilibrium conditions or non-equilibrium conditions but incompressible solid grains reduces to
the solid pressure (2.51) coincides with that of equation (2.49) if xW=Sw but this is of
course not often the case. Finally, Coussy (1995), using the Clausius-Duhem inequality, obtains
where
is the capillary pressure increment. This equation has hence an incremental form and
differs substantially from the previous ones. The practical implication of these
different formulations for slow phenomena have been investigated in detail by
Schrefler and Gawin (1996). It was concluded that in practical soil mechanics situations the resulting differences are small and appear usually after long lasting variations of the moisture content. Only several cycles of drying and wetting would
produce significant differences.
The formulations of the effective stress principle in finite form coincide if the
Bishop parameter x = Sw and the solid grains are incompressible, i.e. a = 1 this
ofcourse limits the applicability. This assumption is now made and the governing
41
THEORY
equations are derived again, using the hybrid mixture theory, as has been done by
Schrefler (1995) and Lewis and Schrefler (1998). Isothermal conditions are assumed
to hold, as throughout this book. For the full nonisothermal case the interested
reader is referred to Lewis and Schrefler (1998).
We first recall briefly the kinematics of the system.
2.5.1
Kinematic equations
As indicated in Chapter 1, a multi-phase medium can be described as the superposition of all 7r phases, 7r = 1,2, . . . K , i.e. in the current configuration each spatial
point x is simultaneously occupied by material points X" of all phases. The state of
motion of each phase is, however, described independently.
In a Lagrangian or material description of motion, the position of each material
point x" at time f is a function of its placement in a chosen reference configuration,
X" and of the current time t
x; = x;
(x;,
x;,x;,t )
To keep this mapping continuous and bijective at all times, the Jacobian of this
transformation must not equal zero and must be strictly positive, since it is equal to
the determinant of the deformation gradient tensor FT
where U" is the right stretch tensor, V" the left stretch tensor, and the skew-symmetric tensor R" gives the rigid body rotation. The differentiation with respect to the
appropriate co-ordinates of the reference or actual configuration are respectively
denoted by comma or slash, i.e.
Because of the non-singularity of the Lagrangian relationship (2.55), its inverse can
be written and the Eulerian or spatial description of motion follows
The material time derivative of any differentiable function f"(x, t ) given in its
spatial description and referred to a moving particle of the 7r phase is
where i is the local value of the velocity field of the .ir phase in a fixed point in space, i is
the flux vector associated with a,g the external supply of @ and G is the net production
of Q. The relevant thermodynamic properties @ are mass, momentum, energy and
entropy. The values assumed by i, g and G are given in Table 2.2 (Hassanizadeh and
Gray (1980, 1990) and Schrefler (1995)). The constituents are assumed to be microscopically non-polar, hence the angular momentum balance equation has been
omitted. This equation shows, however, that the stress tensor is symmetric.
1
i
E
0
t "I
tmr - q
cP
+ 0.5r.r
g
0
g
g.i+h
0
0
0
43
THEORY
The averaged macroscopic mass balance equations are given next. For the solid
phase this equation reads
where u is the mass averaged solid phase velocity and p" is the intrinsic phase
averaged density. The intrinsic phase averaged density p" is the density of the .ir
phase averaged over the part of the control volume (Representative Elementary
Volume, REV) occupied by the .rr phase. The phase averaged density p,, on the
contrary, is the density of the .ir phase averaged over the total control volume. The
relationship between the two densities is given by
where nSwpwew(p)= -m is the quantity of water per unit time and volume, lost
through evaporation and vw the mass averaged water velocity.
For air, this equation reads
where va is the mass averaged air velocity. The linear momentum balance equation
for the fluid phases is
t;,,
+ p"(b;
a;)
=0
where t" is the partial stress tensor, p"b" the external momentum supply due to
gravity, p"a" the volume density of the inertia force, p"eT(pr) the sum of the momentum supply due to averaged mass supply and the intrinsic momentum supply due to a
change of density and referred to the deviation i" of the velocity of constituent .ir
from its mass averaged velocity, and accounts for the exchange of momentum due
to mechanical interaction with other phases. pke"(pr) is assumed to be different from
zero only for fluid phases. For the solid phase the linear momentum balance equation
is hence
The average angular momentum balance equation shows that for non-polar media
the partial stress tensor is symmetric ti: = t; at macroscopic level also and the sum of
the coupling vectors of angular momentum between the phases vanishes.
2.5.4
Constitutive equations
Constitutive models are here selected which are based on quantities currently measurable in laboratory or field experiments and which have been extensively validated.
Most of them have been obtained from the entropy inequality, see Hassanizadeh and
Gray (1980, 1990). It can be shown that the stress tensor in the fluid is
r..S!I = a - - (1 - n)pshii
1
11
(2.68)
with p, = pwSw + p a s a in the case of thermodynamic equilibrium or for incompressible solid grains, (2.51). The sum of (2.67), written for air and water and of (2.68)
gives the total stress a, acting on a unit area of the volume fraction mixture
This is the form of the effective stress principle employed in the following, as already
explained. Moist air in the pore system is assumed to be a perfect mixture of two ideal
gases, dry air and water vapour, with T = ga and r = g w respectively. The equation
of a perfect gas is hence valid
where M , is the molar mass of constituent T , R the universal gas constant, and Q the
common absolute temperature. Further, Dalton's law applies and yields the molar
mass of moisture
Water is usually present in the pores as a condensed liquid, separated from its vapour
by a concave meniscus because of surface tension. The capillary pressure is defined as
pc = pg - pw, see Equation (2.54).
The momentum exchange term of the linear momentum balance equation for
fluids has the form
where v"5s the velocity of the r phase relative to the solid. It is assumed that R"
is invertible, its inverse being ( R " ) - ' =
and KT is defined by the following relation
45
THEOR Y
k,
K$ = -(p", 7 " ) T )
PT
where p, is the dynamic viscosity, k the intrinsic permeability and T the temperature
above some datum. In the case of more than one fluid flowing, the intrinsic permeability is modified as
where k'" is the relative permeability, a function of the degree of saturation. For the
water density, the following holds,
The linear momentum balance equation for the solid phase is obtained in a similar
way, taking into account equations (2.68) instead of (2.67). By summing this momentum balance equation with equation (2.76) written for water and air, and by
taking into account the definition of total stress (2.69) assuming continuity of stress
at the fluid-solid interfaces and by introducing the averaged density of the multiphase medium,
we obtain the linear momentum balance equation for the whole multi-phase
medium
The mass balance equations are derived next. The macroscopic mass balance equation for the solid phase (2.61), after differentiation and dividing by pS, is obtained as
This equation is used in the subsequent mass balance equations to eliminate the
material time derivative of the porosity. For incompressible grains, as assumed here,
DSPS~r
- 0. For compressible grains, see equation (2.89) and related remarks.
The mass balance equation for water (2.63) is transformed as follows. First
Equation (2.75), the material time derivative of the water density with respect to the
moving solid phase and the relative velocity vWS,are introduced. Then the derivatives
are carried out, the quantity of water lost through evaporation is neglected and the
material time derivative of the porosity is expressed through Equation (2.79), yielding
To obtain the equations of Section 2.4.1, further simplifications are needed, which
are introduced next.
An updated Lagrangian framework is used, where the reference configuration is
the last converged configuration of the solid phase. Further, the strain increments
within each time step are small. Because of this we can neglect the convective terms in
all the balance equations. Neglecting, in the linear momentum balance equation
(2.78) the relative accelerations of the fluid phases with respect to the solid phase,
yields the equilibrium equation (2.28a)
The linear momentum balance equation for fluids (2.76) by omitting all acceleration
terms, as in Chapter 2, can be written for water
where
qWv? = W ; and
kwii
k,krw
=!Jw
NOMENCLA TURE
47
where
The phase densities appearing in Chapter 2 are intrinsic phase averaged densities as
indicated above.
The mass balance equation for water is obtained from Equation (2.80), taking into
account the reference system chosen, dividing by pw, developing the divergence term
of the relative velocity and neglecting the gradient of water density. This yields
where the first of Equations (2.84) has been taken into account. This coincides with
Equation (2.35a) for incompressible grains except for the source term and the secondorder term due to the change in fluid density. This last one could be introduced in the
constitutive relationship (2.75)
Similarly, the mass balance equation for air becomes
where again the first of equations (2.86) has been taken into account and the gradient
of the water density has been neglected. Similar remarks as for the water mass
balance equation apply. In particular, the constitutive relationships for moist air,
Equations (2.70) and (2.71), have been used.
Finally, if for the solid phase the following constitutive relationship is used (see
Lewis and Schrefler (1998))
where Ks is the bulk modulus of the grain material, then the mass balance equations
are obtained in the same form as in Chapter 2 (with xW= S,). However, this is not in
agreement with what was assumed for the effective stress.
2.5.6
As this section does not follow the notations used in the use of the book, we summarise below
for purposes of nomenclature
a" mass averaged acceleration of .rr phase
aTS acceleration relative to the solid.
fji
fi
aflaxi
aflaxi
CONCLUDING REMARKS
49
2.6
CONCLUDING REMARKS
The equations derived in this chapter together with appropriately defined constitutive laws allow (almost) all geomechanical phenomena to be studied. In the next
chapter, we shall discuss in some detail approximation by the finite element method
leading to their solution.
REFERENCES
Biot M. A. (1941) General theory of three-dimensionalconsolidation, J. Appl. Phys., 12,155-164.
Biot M . A. (1955) Theory of elasticity and consolidation for a porous anisotropic solid, J. Appl.
Phys., 26, 182-185.
Biot M. A. (1956a) Theory of propagation of elastic waves in a fluid-saturated porous solid,
part I-low-frequency range, J. Acoust. Soc. Am., 28, No. 2, 168-178.
Biot M. A. (1956b) Theory of propagation of elastic waves in a fluid-saturated porous solid,
part 11-low-frequency range, J. Acoust. Soc. Am., 28, No. 2, 179-191.
REFERENCES
51
Lambe T. W. and Whitman R. V. (1969) Soil Mechanics, (SI Version), John Wiley & Sons,
New York.
Leliavsky S. (1947) Experiments on effective area in gravity dams, Trans. Am. Soc. Civil
Engrs., 112, 444.
Lewis R.W. and Schrefler B.A. (1987) The Finite Element Method in the Deformation rind
Consolidation ofPorous Media, J . Wiley & Sons, Chichester.
Lewis R.W. and Schrefler B.A. (1998) The Finite Element Method in the Static and Dynamic
Deformation and Consolidation of Porous Media, J . Wiley & Sons, Chichester.
Li X. K. and Zienkiewicz 0. C. (1992) Multiphase flow in deforming porous-media and finiteelement solutions, Comp. Struct., 45, No. 2, 21 1-227.
Li X. K., Zienkiewicz 0. C. and Xie Y. M. (1990) A numerical model for immiscible 2-phase
fluid-flow in a porous medium and its time domain solution, Int. J. Num. Meth. Eng.. 30,
NO. 6, 1195-1212.
Liakopoulos A. C. (1965) Transient flow through unsaturated porous media, D.Eng. Dissertation, University of California, Berkeley, USA.
Morland L. W. (1972) A simple constitutive theory for fluid saturated porous solids,
J. Geophys. Res., 77, 890-900.
Runesson K. (1978) On non-linear consolidation of soft clay, Ph.D. Thesis, Chalmers University of Technology, Goeteborg.
Sandhu R. S. and Wilson E. L. (1969) Finite element analysis of flow in saturated porous
elastic media, ASCE EM, 95, 641-652.
Schrefler B. A. (1995) Finite Elements in Environmental Engineering: Coupled thermo-hydromechanical process in porous media involving pollutant transport, Archives of Computer
Methods in Engineering, 2, 1-54.
Schrefler B.A. and Gawin D. (1996) The effective stress principle: incremental or finite form?.
Int. J. Nurn. Anal. Meth. Geom., 20, 785-814.
Schrefler B. A. and Zhan X. (1993) A fully coupled model for waterflow and airflow in
deformable porous media, Water Resources Res., 29, No. 1, 155-167.
Serafim J. L. (1954) A subpresseo nos barreyens - Publ. 55, Laboratorio Nacional de Engenheria Civil, Lisbon.
Simoni L. and Schrefler B. A. (1991) A staggered finite element solution for water and gas flow
in deforming porous media, Commun. Appl. Num. Meth.. 7, 213-223.
Skempton A. W. (1954) The pore pressure coefficients A and B, Gkotechnique. 4.
143-147.
Terzaghi K. von (1925) Erdbaumechanik auf bodenphysikalischer Grundlage, Deuticke.
Vienna.
Whitaker S., (1977). Simultaneous heat mass and momentum transfer in porous media:
a theory of drying, Advances in Heat Transfer, 13, Academic Press. New York.
Zaremba S. (1903a) Le priniple des mouvements relatifs et les equations de la mecanique
physique. Reponse a M. Natanson, Bull. Int. Acad. Sci. Cracovie, 614-621.
Zaremba S. (1903b) Sur une generalisation de la theorie classique de la viscosite, Bull. Int.
Acad. Sci. Crucovie., 380-403.
Zienkiewicz 0. C. (1982) Field equations for porous media under dynamic loads in Num.
Meth. in Geomech., D. Reidel, Boston U.S.A.
Zienkiewicz 0 . C. and Shiomi T. (1984) Dynamic Behaviour of saturated porous media: The
generalized Biot formulation and its numerical solution, Int. J. Nurn. Anal. Geotnech.. 8,
71-96.
Zienkiewicz 0. C. and Taylor R. L. (1989) The Finite Element Method - Volume I : Baslc
Formulation and Linear Problems (4th edn), McGraw-Hill Book Company, London.
Zienkiewicz 0. C., Chang C. T. and Bettess P. (1980) Drained, undrained, consolidating and
dynamic behaviour assumptions in soils, GPotechnique, 30, No. 4. 385-395.
3.1
The general procedures of the Finite Element discretization of equations are described
in detail in various texts. Here we shall use throughout the notation and methodology
introduced by Zienkiewicz & Taylor (1989 and 1991) which is the most recent (fourth)
edition of the first text for the finite element method published in 1967.
In the application to the problems governed by the equations of the previous
chapter we shall typically be solving partial differential equations which can be
written as
where A, B are matrices of constants and L is an operator involving spatial differentials such as d l d x , d l d y , etc., which can be, and frequently are, non-linear.
The dot notation implies time differentiation so that
a@
-=&
at
d2@
dl, z & etc.
--
54
where the shape functions are specified in terms of the spatial coordinates i.e.
and &;
= 6;(t)
(3.4b)
are usually the values of the unknown function at some discrete spatial points called
nodes which remain as variables in time.
(ii) Inserting the value of the approximating function 6 into the differential
equations we obtain a residual which is not identically equal to zero but for which
we can write a set of weighted residual equations in the form
Where M , C and P are matrices or vectors corresponding in size to the full set of
numerical parameters 6k.A very suitable choice for the weighting function Wj is to
take them being the same as the shape function Ni:
Indeed this choice is optimal for accuracy in so called self-adjoint equations as shown
in the basic texts and is known as the Galerkin process.
If time variation occurs, i.e. if the parameters Giare time dependent, equation (3.6)
which is now an ordinary differential equation requires solution in the time domain.
This can be, once again, achieved by discretization in time and use of finite elements
there although many alternative approximations (such as the use of finite differences
or other integration schemes) are possible.
Usually, the parameters & represent simply the values of
at specified pointscalled nodes and the shape functions are derived on a polynomial basis of interpolating between the nodal values for elements into which the space is assumed
divided.
Typical finite elements involving linear and quadratic interpolation are shown in
Figure 3.1.
The present chapter will be divided into two sections.
In Section 3.2 we shall consider solution of the approximation based on the u-p
form in which the dependent variables are the displacement of the soil matrix and the
+"
55
Figure 3.1 Some Typical two dimensional elements for linear and quadratic interpolations
pore pressure characterized by a single fluid, i.e. water. However, we shall allow
incomplete saturation to exist assuming that the air pressure is zero.
The formulation thus embraces all the features of the u-p approximation of
sections 2.2.2, 2.3.1 and 2.3.2 and is the basis of a code capable of solving all lowfrequency dynamic problems, consolidation problems and static drained or
undrained problems of soil mechanics. Only two dimensions will be considered
here and in the examples which follow, but extension to three dimensions is obvious.
The code based on the formulation contained in this part of the chapter is named
SWANDYNE (indicating its Swansea origin) and its outline was presented in literature by Zienkiewicz et al. (1990a).
Section 3.3 of this chapter is intended for the solution of dynamic problems where
high-frequency effects predominate. The variable here will be u and U i.e. displacement
vectors of the solid and of the pore fluid (water). The code based on this form is fully
explicit and it is named GLADYS-2E as it was developed in Glasgow (see Chan et al.,
1991) following the work described by Zienkiewicz and Shiomi (1984).
The time-step limitations of such explicit codes are severe and the code is therefore
limited to relatively short time durations. On the other hand, the implicit form in
Section 3.2. allows much longer periods to be studied. Indeed, with suitable accuracy
control such codes can be used both for earthquake phenomena limited to hundreds
of seconds or consolidation problem with a duration of hundreds of days.
3.2
3.2.1
We will report here the basic governing equation derived in the previous chapter.
However, we shall limit ourselves to the use of the condensed, vectorial form of these
which is convenient for finite element discretization. The tensorial form of the
equations can be found in Section 3.4.
56
In the above and in all following equations, the relative fluid acceleration terms are
omitted as only the u-p form is being considered.
The strain matrix S is defined in two dimensions as (see (2.10))
{ : z } = Sdu
Here u is the displacement vector, p the total density of the mixture (see (2.19))
= XwPw
= 0.
(3.14)
The effective stress a" is computed from an appropriate constitutive law generally
defined as 'increments' by (2.2)
where D is the tangent matrix dependent on the state variables and history and
corresponds to thermal and creep strains.
EO
57
The main variables of the problem are thus u and p,. The effective stresses are
determined at any stage by a sum of all previous increments and the value of p,
determines the parameters S, (saturation) and X , (effective area). On occasion the
approximation
can be used.
An additional equation is supplied by the mass conservation coupled with
fluid momentum balance. This is conveniently given by (2.27) which can be
written as
with k = k ( S , ) .
The contribution of the solid acceleration is neglected in this equation. Its inclusion
in the equation will render the final equation system non-symmetric (see Leung,
1984) and the effect of this omission has been investigated in Chan (1988) who
found it to be insignificant. However, it has been included in the force term of the
computer code SWANDYNE-I1 (Chan, 1995) although it is neglected in the left
hand side of the final algebraic equation when the symmetric solution procedure
is used.
The above set defines the complete equation system for solution of the problem
defined providing necessary boundary conditions have been specified as in (2.18) and
(2.19) i.e.
and
where q, is the influx i.e. having an opposite sign to the outflow w,.
The total boundary I? is the union of its components i.e.
58
We assume here that the expansion is such that the strong boundary conditions
(3.18) are satisfied on ru and r, automatically by a suitable prescription of the
(nodal) parameters. As in most other finite element formulation, the natural boundary condition will be obtained by integrating the weighted equation by parts.
To obtain the first equation discretized in space we premultiply (3.8) by (NU)=and
integrate the first term by parts (see for details Zienkiewicz and Taylor, 1989 or other
texts) giving:
and the 'load vector'f ( I ) , equal in number of components to that of vector u contains
all the effects of body forces, and prescribed boundary tractions i.e.
At this stage it is convenient to introduce the effective stress (see (3.12)) now defined
to allow for effects of incomplete saturation as
where
59
60
+,+,,
From the initial conditions, we have the known values of a,, 6,. We assume that the
above equation has to be satisfied at each discrete time i.e. t, and t,+l, we can thus
write:
and
From the first equation, the value of the acceleration at time t, can be found and this
solution is required if the initial conditions are different from zero.
61
The link between the successive values is provided by a truncated series expansion
taken in the simplest case as GN22 as Eq(3.34) is a second-order differential equation
j and the minimum order of the scheme required is then two: as (p > j)
Alternatively, a higher order scheme can be chosen such as GN32 and we shall
have:
In this case, an extra set of equations is required to obtain the value of the highest
time derivatives. This is provided by differentiating (3.35) and (3.36).
and
In the above equations, the only unknown is the incremental value of the highest
derivative and this can be readily solved for.
Returning to the set of ordinary differential equations, we are considering here i.e.
in (3.23), (3.27) and (3.28) and writing (3.23) and (3.28) at the time station t,+l, we
have:
62
u,+1
= u,
u,+1 = u,,
+ Au,
+&at
+ p,au,at
and
where Au, and Ap, are as yet undetermined quantities. The parameters PI,P2 and PI
are usually chosen in the range of 0 to 1. For P2 = 0 and PI = 0, we shall have an
explicit form if both the mass and damping matrices are diagonal. If the damping
matrix is non-diagonal, an explicit scheme can still be achieved with PI = 0 thus
eliminating the contribution of the damping matrix. The well known central difference scheme is recovered from (3.41) if PI = 1/2,P2 = 0 and this form with an
explicit u and implicit p scheme has been considered in detail by Zienkiewicz et al.
(1982) and Leung (1984). However, such schemes are only conditionally stable and
for unconditional stability of the recurrence scheme we require
1
P2
> PI > 2
and
81 2 5
63
w!lj
=0
(3.44b)
J=
where
which are the well known expressions for tangent stiffness matrix. The underlined
term corresponds to the 'initial stress' matrix evaluated in the current configuration
as a result of stress rotation defined in (2.5).
Two points should be made here:
(a) that in the linear case a single 'iteration' solves the problem exactly
(b) that the matrix can be made symmetric by a simple scalar multiplication of the
second row (providing KT is itself symmetric).
In practice it is found that the use of various approximations of the matrix J is
advantageous such as, for instance, the use of 'secant' updates (see for instance
Crisfield (1979), Matthies and Strang (1979) and Zienkiewicz and Taylor (1991)
64
65
Quadratic for u
Linear for p
a
(ii)
Biquadratic for u
Bilinear forp
D)0
(ii)
Linear for u
Linear forp
(ii)
b
Linear (with cubic bubble) for u
Linear f o r p
Figure 3.2 Elements used for coupled analysis, displacement (u) and pressure @) formulation
(a) (i), quadratic for u; (ii), linear for p; (b) (i), biquadratic for u; (ii), bilinear for p: (c) (i) linear
for u; (ii), linear for p: (d)(i), linear (with cubic bubble) for u; (ii), linear for Element (c) is not
fully acceptable at incompressible-undrained limits.
Thus for known Au increment the 0'' are evaluated by using an appropriate
tangent matrix D and an appropriate stress integration scheme.
Further we note that if p,,. 0 (full saturation)
then we have
>
However, when negative pressures are reached i.e. when p < 0, the values of
S,, , x,,,,
k , have to be determined from appropriate formulae or graphs.
The reader will observe that when p,,. 2 0 we regain the fully saturated behaviour
described in Section 2.2 of the previous chapter.
66
we have
(3.51)
67
This limit was re-established later by Zienkiewicz and Xie (1991) who replace the
leading coefficient of (3.52) as a result of a more detailed analysis by
Whatever the form of error estimator adopted, the essence of the procedure is
identical and this is given by establishing a priori some limits or tolerance which
must not be exceeded, and modifying the time steps accordingly.
In the above we have considered only the error in one of the variables i.e. u but in
general this suffices for quite a reasonable error control.
The tolerance is conveniently chosen as some percentage 7 of the maximum value
of norm llul12 recorded. Thus we write
In the standard treatment of consolidation equation (see for instance Lewis and
Schrefler, 1998) the acceleration terms are generally omitted a priori. However,
there is no disadvantage in writing the full dynamic formulation for solving such a
problem. The procedure simply reduces the multiplier of the mass matrix M to a
negligible value without influencing in any way the numerical stability, provided of
course that an implicit integration scheme is used.
68
and
with the effective stresses given by (3.27) and are defined incrementally as
dd'
de")
(3.57)
First we observe that the equations are uncoupled and that the second of these i.e.
(3.56) can be solved independently of the first for the water pressures. Indeed in this
solution the negative pressure zones and hence the partially saturated regions can be
readily determined following the procedures outlined in the previous chapter.
With p,,, determined as
the first equation (3.55) coupled with the appropriate constitutive law (3.57) can be
solved once the history of the load applied has been specified.
The solution so obtained is of course the well known, drained, behaviour.
The case of undrained behaviour is somewhat more complex. We note that with
k = 0 i.e. with totally impermeable behaviour
H = 0 and f ( 2 )
=0
(3.59)
which on integration establishes a unique relationship between u and p,, which is not
time dependent
69
However, this process may be time consuming even if large time steps At are used. A
simpler procedure is to use the GNOO scheme with
Equations (3.41) and (3.61) now become, for the undrained problem,
If the material behaviour is linearly elastic, then the equation can be solved directly
yielding the two unknowns u,,, and p;+, and if the material is non-linear, an
iteration scheme such as the Newton Raphson, Quasi-Newton, Tangential Matrix
or the Initial Matrix method can be adopted. With a systematic change of the
external loading, problems such as the load-displacement curve of a non-linear soil
and pore-fluid system can be traced.
(3.62)
70
We can now reduce the governing u-p equations (3.23) and (3.28) to the form given
below
and
where p = p,,.
and K
1,
B~DB~O
is the well known elastic stiffness matrix which is always symmetric in form. S and H
are again symmetric matrices defined in (3.31) and (3.30) and Q is as defined in (3.29)
The overall system can be written in the terms of the variable set [ii,plT as
Once again the uncoupled nature of the problem under drained condition is evident
(by dropping the time derivatives) giving
It is interesting to observe that in the steady state we have a matrix which, in the
absence of fluid compressibility, results in
which only can have a unique solution when the number of ii variables nu is greater
than the number of p variables n,. This is one of the requirements of the patch test of
Zienkiewicz et al. (1985), Zienkiewicz et al. (1986) and Zienkiewicz and Taylor (1989)
and of the Babuska-Brezzi (Babuska, 1973 and Brezzi, 1974) condition.
71
Indeed such damping matrices have a physical significance and are always introduced
in earthquake analyses or similar problems of structural dynamics. With the lack of
any special information about the nature of damping it is usual to assume the so
called 'Rayleigh damping' in which
where cr and p are coefficients determined by experience (see for instance Clough
and Penzien, 1975 or 1993). In the above, M is the same mass matrix as
given in (3.24) and K is some representative stiffness matrix of the form given
in (3.47).
3.3
3.3.1
We shall now return to the original equations of Chapter 2 Section 2.2 without the
introduction of the approximation used in Section 3.2.2. Thus (2.1 l), (2.13) and
(2.16) are repeated below as:
omitting now only the convective acceleration, density variation and source terms,
i.e., the terms underlined in the above equations. Here for brevity the equations are
now given only in tensorial notation.
72
we can integrate, (3.72~)in time and find a direct expression for the pressure provided
that the compressibility, l / Q , is different from zero (see (2.17)). Thus we have
Inserting this into (3.72a and b), we obtain the following system
rJ,'.
'J
a;,
= 01. .+ ,$..
+ a Q ( ' ~ ~ k+knu&),;
(3.75)
VP
pui
=0
(3.76a)
and
The vector form of the above equation is presented in Table 2.1 of Chapter 2. The
discretization of the above leads to the equation originally used by Ghaboussi and
Wilson (1972). However, this system is inconvenient as second derivatives in time of
both variables occur in each of the equations and thus completely diagonal matrices
cannot be obtained by mass lumping. A very simple modification can be made here
as suggested by Zienkiewicz and Shiomi (1984), which leads to complete matrix
decoupling of higher derivatives and which is therefore ideally suited for explicit
schemes.
Now in place of relative displacements of water, we shall use its total displacement
defined as
thus
where the relative displacements are divided by the porosity n, equation (3.73), to
approximate the average true displacement.
Starting from (3.76), after some algebraic manipulations we have
and
73
and
Multiplying (3.79b) by n and subtracting from (3.79a) we find that the first equation
is now free from the acceleration of the fluid displacement and becomes
In the final equation system (3.80) only ii, occurs in the first equation and only
the second, thus leading to a convenient diagonal form in discretization.
a in
and
74
3.3.3
where ni are the directions of the normal at the boundary. Linearizing the constitutive equation we have
D ~ i j k l= D ~ i j l k
we have
Note that
75
76
77
oi)dfl = ,
-MflLULi
~;np~N;dfl6~,
-M~U
Collecting all the terms and displaying in matrix form which was first presented by
Shiomi (1983) and Zienkiewicz and Shiomi (1984):
It is of interest to observe that the mass matrix of the system does not couple
variables ti and U and here can be easilty diagonalized.
With full symmetry of the system, we can use the generalized Newmark procedure,
now applied in terms of combined displacement variable:
for
if
P2
= 0 it disappears
78
Although the mass matrices can be diagonalized rather simply, if the Newmark
parameter p or the Generalized Newmark parameter PI is non-zero then the inclusion of the damping matrix may destroy this diagonal structure. The following two
cases can be identified.
If anisotropic permeability is used with cross-coupling terms i.e. kv # 0 for i # j,
then the damping matrix will assume a block diagonal structure of 4 x 4 for twodimensions and 6 x 6 for three-dimensions if the variables u and U are numbered
next to each other.
On the other hand, if the cross-coupling terms are all zero, then the block diagonal
structure will consist of 2 x 2 matrices whether it is two-dimension or threedimension.
In any case, if Rayleigh damping with the full stiffness matrix is used then the
diagonal structure will be destroyed, (see Chan et al., 1991), hence only the diagonal
contribution of the stiffness matrix is included in the solution matrix while the full
matrix is retained in the right hand side during the calculation of the residuals.
Having dealt with the problem raised by the Rayleigh damping, the off-diagonal
term on the solution matrix can also be removed by further lumping (see, for
instance, Simon et al., 1986).
One of the major constraints on an explicit scheme is the limitation of the explicit
time step. However, for many non-linear and transient problems, the use of a small
time step is a positive advantage as the non-linearity may impose a large number of
iterations in typical implicit schemes. In the case when the fluid bulk modulus Kfis
much bigger than the bulk modulus of the soil matrix KT the critical time step is
found to be:
where e is the minimum length between nodes and C1 is a constant depending on the
type of element used (and such other factors as porosity). Therefore by a suitable
reduction in the value of Kf,
an incrase in the critical time step length is possible and this
was shown in Chan et al. (1991). Generally, the results are not adversely affected until
the value of Kfbecomes comparable to the bulk modulus of soil matrix K T .
3.4
The equation numbers given here correspond to the ones given earlier in the text.
79
(5 - (5..
(5"
IJ
(3.11b)
= 0..
+ ~16..
11
(3.12b)
!IP
and
n;w;
= n;kii(-p
,,.,
,,+ S,,.p/-6,)= w,
on I? = I?,,.
The summation range for the upper case indices will depend on the number of
nodes with solid displacement and pore water pressure degrees-of-freedom (dof)
respectively.
Applying Green's identity to the internal force term (first term on the left-hand
side)
80
Rearranging
NZjordR
+[
I,
N;pN;dR]GLi
NipbidR +
rr
N;iidT.
The definition of the B matrix in equation (3.21) is not needed in tensorial form.
0..lJ-
MKLsL;+
1,
o'llJ.
N;, ,$dR
1
(~
do; = D i j ~ i(N;,kd",
1,
N; (kl(-pw,
ax6..
d'
~ ~ , ~fi)
f I0 r
-
+ NtidiiKk)
(3.23b)
d&li)
(3.27b)
,
+ Sivpfbj),; + a&,;+ P<
+ i0)dR = 0
Q
Neglecting source term and integrating by part the first part of the first term
-
Ira
N$nikvp.,,jdT.
1,
b , , k u p M . , , N$(kgS,vpfb,),i Niazi,; + N g f i dR = 0
Q*
.
CONCLUSIONS
81
3.5
CONCLUSIONS
REFERENCES
Babuska I. (1971) Error Bounds for finite element methods, Num. Math., 16, 322-333.
Babuska I. (1973) The finite element method with Lagrange Multipliers, Num. Math., 20,
179-192.
Bellman R. (1960) Introduction to Matrix Analysis (1st edn), Mcgraw-Hill Book Company,
London.
Bergan P. G . and Mollener (1985) An automatic time-stepping algorithm for dynamic problems, Comp. Meth. Appl. Mech. Eng. 49, 299-318.
Brezzi F. (1974) On the existence, uniqueness and approximation of saddle point problems
arising from Lagrange multipliers, R. A. I. R. 0 . Anal. Numtr., 8, No. R-2, 129-151.
Chan A. H. C. (1988) A unqied Finite Element Solution to Static and Dynamic Geomechanics
problems, Ph.D. Dissertation, University College of Swansea, Wales.
Chan A. H. C. (1995) User Manual for DIANA SWANDYNE-11, School of Civil Engineering, University of Birmingham, December, Birmingham.
Chan A. H. C., Famiyesin 0 . 0. and Muir Wood D. (1991) A Fully Explicit u-w Schemes
for Dynamic Soil and Pore Fluid Interaction, APCOCM Hong Kong, 11-13 Dec., 1,
881-887.
Clough R. W. and Penzien J. (1975) Dynamics of Structures, McGraw-Hill, New York.
Clough R. W. and Penzien J. (1993) Dynamics of Structures (2nd edn), McGraw-Hill, Inc.,
New York.
82
Crisfield M. A. (1979) A faster modified Newton-Raphson Iteration, Comp. Meth. Appl. Mech.
Eng., 20, 267-278.
Crisfield M. A. (1991) Non-linear Finite Elenzent Analysis of Solid.7 and Structures, 1, John
Wiley & Sons, Chichester.
Crisfield M. A. (1997) Non-linear Finite Element Analysis of Solids and Structures, 2. John
Wiley & Sons, Chichester.
Dahlquist G. (1956) Convergence and stability in the numerical integration of ordinary
differential equations, Math. Scnnd., 4, 33-53.
Dahlquist G. (1959) Stability and error bounds in the numerical integration of ordinary
differential equations, Kungl. Teknisku Hogskolans Handlingar, 130.
Dahlquist G. (1978) On accuracy and unconditional stability of linear multistep methods for
second order differential equations, BIT, 18, 133-1 36.
Dewoolkar M. M. (1996) A study of seismic effects on centiliver-retaining ~valls with
saturated backfill. PhD Thesis, Dept of Civil Engineering, University of Colorado, Boulder,
USA.
Gantmacher F. R. (1954a) Applications of the theory of matrices: Second part of ' A Theory o f
Matrices', translated and revised by Brenner J. L. et al., Interscience Publishers, Inc.
Gantmacher F. R. (1954b) The Theory of Matrices-English translation by Hirsch K. A, 1,
Chelsea Publishing Company.
Gantmacher F. R. (1954~)The Theory of'Matrices-English translation by Hirsch K. A. 2,
Chelsea Publishing Company.
Ghaboussi J. and Wilson E. L. (1972) Variational formulation of dynamics of fluid saturated
porous elastic solids, ASCE E M , 98, No. EM4, 947-963.
Hurwitz A. (1895) ~ b e die
r Bedingungen, unter welchen eine Gleichung nur Wurzeln mit
negativen realen Teilen besitzt, Math. Ann., 46, 273-384.
Hurwitz A. (1933) ~ b e die
r Bedingungen, unter welchen eine Gleichung nur Wurzeln mit
negativen realen Teilen besitzt, Mathematische Werke, Basel, 2, 533-545.
Katona M. G. (1985) A general family of single-step methods for numerical time integration of
structural dynamic equations, NUMETA 85, 1, 213-225.
Katona M. G. and Zienkiewicz 0.C. (1985) A unified set of single step algorithms Part 3: The
Beta-m method, a generalisation of the Newmark scheme, Int. J. Num. Meth. Eng., 21,
1345-1 359.
Lambert J. D. (1973) Computational Methods in Ordinary Differential Equutions, John Wiley
and Sons Ltd., Chichester.
Leung K. H. (1984) Earthquake response of saturated soils and liquefaction, Ph.D. Dissertation, University College of Swansea, Wales.
Lewis R. W. and Schrefler B. A. (1998) The Finite Element Method in the Static and Dynamic
Deformation and Consolidation of Porous Media, John Wiley & Sons, Chichester.
Li X. K. and Zienkiewicz 0. C. (1992) Multiphase flow in deforming porous media and finiteelement solutions, Comp. Struct., 45, No. 2, 21 1-227.
Li X., Zienkiewicz 0. C. and Xie Y. M. (1990) A numerical model for immiscible two-phase
fluid flow in a porous medium and its time domain solution, Int. J. Num. Meth. Eng., 30,
1195-1212.
Matthies H. and Strang G. (1979) The solution of nonlinear Finite Element equations, Int.
J. Num. Meth. Eng., 14, 1613-1626.
Newmark N. M. (1959) A method of computation for structural dynamics, Proc. ASCE, 8,
67-94.
Park K. C. (1983) Stabilization of partitioned solution procedure for pore fluid-soil interaction
analysis, Int. J. Num. Meth. Eng., 19, 1669-1673.
Park K. C. and Felippa C. A. (1983) 'Partitioned analysis of coupled systems', Chapter 3 in
Computational Methods,for Transient Analysis, Elsevier Science Publishers B. V.
83
REFERENCES
Paul D. K. (1982) Efficient dynamic solutions for single and coupled multipled field problems.
Ph.D. Dissertation, University College of Swansea, Wales.
Routh E. J. (1877) Stability of a given state of motion-the advanced part of a treatise on the
dynamics of a system of rigid bodies (6th edn), Dover, London.
Routh E. J. (1930) Stability of a given state of motion-the advanced part of a treatise on the
dynamics of a system of rigid bodies, Dover, London.
Routh E. J. (1955) Stability of a given state of motion-the advanced part of a treatise on the
dynamics of a system of rigid bodies, Dover, New York.
Schrefler B. A. (1995) Finite Element in Environmental Engineering: Coupled thermo-hydromechanical process in porous media involving pollutant transport, Archives of Computational Methods in Engineering, 2, 1-54.
Schrefler B. A. and Zhan X. (1993) A fully coupled model for waterflow and airflow in
deformable porous media, Water Resources Res., 29, No. 1, 155-167.
Shiomi T. (1983) Nonlinear behaviour of soils in earthquake-C/Ph/73/83, Ph.D. Dissertation,
Univ. Coll. of Swansea, Wales.
Simon B. R., Wu J. S. S., Zienkiewicz 0 . C. and Paul D. K. (1986) Evaluation of u-w and u-p
FEM for the response of saturated porous media using I-dimensional models, Int. J. Nunz.
Anal. Geomech., 10, 461482.
Whitman R. V. (1953) After Marcusion (1995): an example ofprofessional modesty, The Earth.
Engineers and Education, MIT, 200-202.
Wood W. L. (1984a) A further look at Newmark, Houbolt, etc.. time-stepping formulae, Int.
J. Num. Meth. Eng., 20, 1009-1017.
Wood W. L. (198413) A unified set of single step algorithms Part 2: Theory, Int. J. Num. Meth.
Eng., 20, 2303-2309.
Wood W. L. (1985a) Addendum to 'a unified set of single step algorithms, Part 2: Theory', Int.
J. Num. Meth. Eng., 21, 1165.
Wood W. L. (1985b) A unified set of single-step algorithms Part 4: Backward error analysis
applied to the solution of the dynamic vibration equation-Numerical Analysis Report
6185, Department of Mathematics, University of Reading, Reading.
Wood W. L. (1990) Practical Time-stepping Schemes, Clarendon Press, Oxford.
Zienkiewicz 0. C. (1977) The Finite Element Method (3rd edn), McGraw-Hill Book (UK) Ltd,
London.
Zienkiewicz 0 . C., Hinton E., Leung K. H. and Taylor R. L. (1980a) Staggered, Time
Marching Schemes in Dynamic Soil Analysis and Selective Explicit Extrapolation Algorithms in Proceedings of the Second Symposium on Innovative Numerical Analysis for the
Engineering Sciences, University of Virginia Press.
Zienkiewicz 0 . C., Wood W. L. and Taylor R. L. (1980b) An alternative single-step algorithm
for dynamic problems, Earthquake Engineering & Structural Dynamics, 8, 31-40.
Zienkiewicz 0 . C., Leung K. H., Hinton E. and Chang C. T. (1982) Liquefaction and
permanent deformation under dynamic conditions Numerical solution and Constitutive
relations, Chapter 5 in Soil Mechanics- Transient and Cyclic loads, John Wiley, Chichester.
Zienkiewicz 0 . C. (1984) 'Coupled Problems and their Numerical Solution,' Chapter 1 in
Numerical Methods in Coupled Systems, John Wiley and Sons Ltd., Chichester.
Zienkiewicz 0 . C. and Shiomi T. (1984) Dynamic Behaviour of saturated porous media: The
generalized Biot formulation and its numerical solution, Int. J. Num. Anal. Geomech.. 8.
71-96.
Zienkiewicz 0 . C., Wood W. L., Hine N. W. and Taylor R. L. (1984) A unified set of single
step algorithms Part I: General formulation and applications, Int. J. Num. Meth. Eng., 20.
1529-1552.
Zienkiewicz 0 . C. (1985) The coupled problems of soil-pore fluid-external fluid interaction:
Basis for a general geomechanics code, ICONMIG 5, 1731-1 740.
-
84
4.1 INTRODUCTION
The purpose of constitutive models is to capture some of the main features of the
mechanical behaviour of solids under given conditions of temperature, velocity of
load application, level of strain, nature of stress conditions, etc.
Roughly speaking, models can be classified into two main groups:
(i) Micromechanical or physical models, based on the behaviour of grains or
particles, and
(ii) Macromechanical or phenomenological models.
Most of the models used in computer codes are of the second class.
All materials present a response which depends on time to a greater or lesser
degree. For instance, a specimen of soft clay subjected to a constant vertical stress
shows a vertical deformation which increases monotonically with time.
However, most of the geomaterials under normal engineering conditions present a
mechanical behaviour which depends more on the level of stress, pore pressure, past
history, direction of load increment and material structure than on time. In fact, the
major part of the time dependence observed is generally connected with the pore
water flow.
For these, plasticity based theories provide a consistent framework in which the
behaviour can be accurately understood and predicted.
It can be said that history of plasticity began in 1773 with the work of Coulomb in
soils, applied later by Poncelet and Rankine to practical soil mechanics problems.
It was not until almost a century later, in 1864, when Tresca, based on experimental results on punching and extrusion tests, proposed a yield criterion dependent
on the maximum shear stress. Later, St. Venant in 1870 introduced the concept of
86
CONSTITUTIVE RELATIONS
PLASTICITY
4.2
4.2.1
The uniaxial behaviour of materials shows that irreversible strain develops in a way
which depends on the type of material. In the case of metals such as mild steel, the
observed behaviour in tension is schematized in Figure 4.1, where it can be seen that
the response is elastic and linear until a point A is reached, from which plastic or
irreversible strain upon unloading appears. If the specimen is subjected to an increasing
strain, the stress does not change until point E. Along the plateau ABDE, the material
behaviour is known as perfectly plastic. If the specimen is unloaded, both loading and
unloading follow the same path, without irreversible deformation. The level of stress
at which plastic strains appears does not change, and the material does not harden.
Once a certain level of strain has been reached (E), the stress again increases. If we
unload at some point F and then reload again, the material is able to resist a higher
load until new plastic strains develop (hardening). Finally, a maximum load is
reached from which the stress decreases until the material fails.
In the case of soft soils such as saturated clays, the stress-strain curve is different,
as plastic strain are present from very early stages of the test (Fig. 4.2).
87
Stress
Strain
Figure 4.1
Stress
Strain
4.2.2
Generalized plasticity
In the following, boldface characters will be used for tensors, uppercase (such as D)
denoting fourth-order tensors Dijkl,and lower case (such as a) for second-order
tensors a,-.
88
CONSTITUTIVE RELATIONS
PLASTICITY
Strain
Basic theory
If the response of the material does not depend on the velocity at which the
stress varies the relationship between the increments of stress and strain can be
written as
89
where Q, is a function of the increment of the stress tensor d u and variables describing
the 'state' (or history) of the material. This is a general relation embracing most nonlinear, rate-independent constitutive laws.
An inverse form is
= @(Ada)
where
90
CONSTITUTIVE RELATIONS
Figure 4.4
PLASTICITY
Strain
This kind of constitutive law has been defined by Darve (1990) as incrementally
non-linear.
There are several alternatives to introduce the dependence on the direction of the
stress increment, among which it is worth mentioning the multilinear laws proposed
by Darve and co-workers in Grenoble (Darve and Labanieh, 1982), or the hypoplastic laws of Dafalias (1986) or Kolymbas (1991). However, the simplest consists of
defining in the stress space a normalized direction n for any given state of stress a
such that all possible increments of stress are separated into two classes, loading and
unloading
deL = CL : d a for n : d a > 0 (loading)
dev = Cv : d a for n : d a < 0 (unloading)
(4.10)
91
and
where n , ~and ngU are arbitrary tensors of unit norm and H L j u two scalar functions
defined as loading and unloading plastic moduli. It can be very easily verified that
both laws predict the same strain increment under neutral loading where both
expressions are valid and hence non-uniqueness is avoided. As for such loading,
the increments of strain using the expressions for loading and unloading are
and
It follows that material behaviour under neutral loading is reversible, and it can
therefore be regarded as elastic. Indeed, the tensor Ce characterizes elastic material
behaviour, and it can be very easily verified that for any infinitesimal cycle of stress
(do, -du) where d u corresponds to neutral loading conditions, the accumulated
strain is zero.
This suggests that the strain increment can be decomposed into two parts
where
and
We note that irreversible plastic deformations have been introduced without the
need for specifying any yield or plastic potential surfaces, nor hardening rules. All
92
CONSTITUTIVE RELATIONS
PLASTICITY
that is necessary to specify are two scalar functions HLluand three directions, ngLlu
and n.
To account for softening behaviour of the material, i.e., when HL is negative,
definitions of loading and unloading have to be modified as follows:
dcL = CL : d u for n : d u e > 0 (loading)
dew = C u : d u for n : d u e < 0 (unloading)
(4.17)
We note here (and in what follows) that in matrix notation the product forms are
written simply as deL = CL.du for; nT.du' > 0 etc.
(n : D') : (C' : d u )
+ (n : D')
: dXngLlu
(4.21)
where we have taken into account that the product De : C' is the fourth-order
identity tensor.
Substituting now
93
we obtain
n : D'
: d~ =
(HLIU n : D'
: ngLlu)dX
(4.24)
and
+ dXngLlu
we have,
du
= D' : d~ - dXD' : ~ , L , u
= D' : d~ -
(n : Dr : d ~ )
(D' : ngLIU)
HL/U n : D' : ~ , L / u
= Del' : d~
4.2.3
94
CONSTITUTIVE RELATIONS
PLASTICITY
where we have assumed that there is a set of scalar internal variables K accounting for
the material state and characterizing the size (and shape) of the yield surface. Sometimes, as will be discussed later, f depends also on the tensor variable a , as in the case
of kinematic and anisotropic hardening models, for instance. Here we will restrict the
discussion to the isotropic case stated above.
In the interior of the yield surface, there is no plastic deformation, and, consequently, the plastic modulus is H = m.
The loading-unloading direction is given by the normal to the surface
where
The direction of plastic flow is similarly derived from a plastic potential surface
g(u) = 0 passing through the stress point considered,
Both surfaces can coincide, and the flow rule is then said to be associative, or can
be different in which case there is a non-associative flow rule.
Therefore, the material behaviour predicted by Classical Plasticity models presents
a sharp transition from the elastic to the elastoplastic regime, with a discontinuity in
the derivative of stress-strain curves.
The plastic modulus is obtained through application of the so called 'consistency
condition', i.e., the requirement that during yield the stress point should always
remain on the yield surface. A certain 'hardening law' has to be introduced, relating
d~ to either incremental plastic work or to the increment of plastic strain.
95
stress states in the interior of this surface, material behaviour is elastic and
K is constant, the material cannot sustain a higher stress and failure
takes place. This is the reason why the yield surface is also known as the failure
surface. Care should be taken, however, as in the case of materials with hardening,
these surfaces can be different.
The scalar K usually characterizes the size of the surface. This is, of course, a
simplification, and more complex descriptions are available, such as
f (au,K ) < 0. If
+ +
02
03
= u,,
JZ and J3 the second and third invariants of the deviatoric stress tensor s,
(4.38)
96
CONSTITUTIVE RELATIONS
PLASTICITY
At this stage it is convenient to define also the Lode's angle 19often used instead of J 3 .
with
-d~,,
(4.43)
and in Figure 4.5 we show the yield surface in the space of two stress invariants, the
second or deviatoric invariant and the first or mean, hydrostatic stress invariant.
With each of these is associated an appropriate strain component -n,, being the
component in the direction of decreasing volumetric strain if plasticity is assumed
associated. Thus the three stages of loading P I ,P2 and P3 correspond to increasing
values of ti as shown in Figure 4.5 (b).
It has to be noticed that plastic strain appears from the beginning of the test, as the
initial stress is on the yield surface. If, for instance, we unload at P2, there will exist a
permanent deformation even when the stress has come back to the original state.
Deviatoric stress
Deviatoric stress
Axial strain
(b)
Figure 4.5 Typical hardening behaviour of clay. (a) Yield surfaces (b) Stress-strain curve
showing permanent strain upon unloading
97
t Deviatoric stress
Deviatoric stress
-Hydrostatic stress
Axial strain
(a)
('J)
Deviatoric stress
Deviatoric stress
Axial strain
-Hydrostatic stress
(a)
(b)
The process of increasing the size of the yield surface in this case is known as
hardening. Comparing the conditions at PI and Pz, the elastic domain is bigger in the
latter, and the material is harder in this sense. Notice that slopes of the stress-strain
curves contradict this definition, as the incremental response of the material is harder
in the first case.
Hardening is not a common feature of all materials. Indeed, in the case shown in
Figure 4.6, the size did not change and failure takes place as soon as the yield surface
is reached.
In another loading case, the size of the yield surface may decrease, as shown in
Figure 4.7, and softening behaviour occurs.
Some frequently used failure and yield criteria. Pressure independent criteria :
von Mises-Huber yield criterion
von Mises yield criterion assumes that plastic strain appears whenever the second
invariant of the stress tensor reaches a critical value Y * ,
98
Taking into account that the condition J2 = constant, corresponds to stress states
a1 ,az,
a3 such that the distance to the hydrostatic axis a1 = a2 = a3 is constant, von
Mises criterion is represented in principal stress axes as a cylinder of radius
= &!Y which is schematized in Figure 4.8(a). In the same figure we show a
plane perpendicular to the hydrostatic axis, which is referred to as the II plane. Its
intersection with the von Mises cylinder is a circle, which is shown in Fig.4.8(b).
A simple method of determining the constant Y is to perform a tension test
a2 = a3 = 0 and to determine the instant at which plastic strain develops. If the
value of limiting tensile stress is a y then we obtain
from which
Figure 4.8
ll plane
von Mises - Huber yield criterion. (a) In the principal stress space (b) Section by
99
Figure 4.9
= 0,
which corresponds in the al, a2 axes to an ellipse with principal axes at 45" (Figure
4.9).
Tresca criterion The Tresca criterion, proposed in 1864, is based on the assumption
that plastic straining of a material appears when the maximum shear strain reaches a
critical value Y. This condition, expressed in terms of the principal stresses reads
(urnax - ffrnin)
(4.50)
Substituting now the maximum and minimum principal stresses by their values in
terms of the invariants I , , J2 and Lode's angle 8
Noting that
100
CONSTITUTIVE RELATIONS
Figure 4.10
PLASTICITY
Tresca Yield criterion. (a) In principal stress axes (b) In the Il plane
When plotted in the space of principal stresses, the Tresca yield criterion is a
hexagonal prism, with its axis coincident with the hydrostatic axis a , = 02 = a3
(Figure 4.10a). The section by the II-plane is a regular hexagon as can be seen in
Figure 4.10(b).
Finally, the plane stress condition a2 = 0 is represented by
= C.
a,, tan q5
(4.55)
to describe the conditions under which failure takes place in soils. He assumed that
failure occurs on a plane on which the shear stress T , and the normal stress a,7
(compression negative) fulfill the above condition. Although it is not advisable to
think of it as a yield surface, it has been used frequently in engineering practice, and
most finite element codes include it.
and
101
102
CONSTITUTIVE RELATIONS
Figure 4.13
PLASTICITY
and
From above, using the relationships between principal stresses and invariants, it is
easy to obtain
and. when plotted in the space of principal stresses, consists of a cone in which the
axis is coincident with the hydrostatic axis.
The section of this cone through the II plane is a circle and when plotted in the
mean hydrostatic pressure-deviatoric stress plane, the intersection with it consists of
two lines with identical slope (compression and extension). Therefore, the friction
angles corresponding to compression and extension are different, and, in fact, given
the parameter a and a value of Lode's angle 19,if the intersections of the DruckerPrager cone and the Mohr-Coulomb surfaces are to coincide for a certain value of
Lode's angle 8,the relationship between the friction angle and a is
103
These relationships have to be taken into account when trying to use the DruckerPrager criterion for plane strain conditions and what is known from experiments is
cohesion and angle of friction.
Under cylindrical triaxial conditions, i.e., uT = ( u l ,uz = u3)the angles of friction
in compression and extension are different, and can be obtained using the above
relationship with 0 = 7r/6 and 0 = -7r/6
ac =
2sin4
cos 2 - 1sin 2 sin4
JS
from where
and
In a similar way,
and
104
CONSTITUTIVE R E L A T I O N S PLASTICITY
Finally, it can be seen that, for a given value of a, the relationship between angles
of friction in extension and compression is
It is left to the reader as an exercise to demonstrate that there is a value of sin& for
which the friction angle in extension reaches 7r/2.
It is interesting to mention that 'rounded' Mohr-Coulomb surfaces have been
proposed in the past (Zienkiewicz and Pande, 1977), in which the slope M is assumed
to vary as:
In this way, the yield surface is smooth. and coincides with the Mohr-Coulomb
original surface in both triaxial compression and extension conditions.
and hence
105
dn dg = 0 with
(b)--.
~ E Pd u
dn # 0
ddl
--
Computational aspects
Most of the expressions given in the above sections simplify in the case of isotropic
materials, as all necessary items can be defined in terms of stress invariants. The yield
surface, for instance, can be expressed as
106
CONSTITUTIVE RELATIONS
PLASTICITY
af
=-
du
and n,
ag
where
3d3
J3
sin 38 = - -2 5,312
Differentiating this expression, we arrive at
107
we have, finally,
n = C~nl C m
+C m
where
and
It can be seen that the set of constants { C ; )depends on the yield criterion chosen,
being independent of the vectors ni.
Next, we will obtain the explicit form of nl, n2 and n3.
108
Vector
CONSTITUTIVE RELATIONS
nl
is given by
nl
= = - -
or.
d
--{a\-
aa,
+a,. + a,)
we arrive at
and
=1
PLASTICITY
109
Therefore,
Vector n3 is given by
where
1
J3 = j~ i j ~ j k ~ k i
Constants CI , C2, and C3 depend on the yield criterion chosen. In the case of the
von Mises yield criterion, which can be written as
f=J2-y2=0
we find
(4.106)
110
CONSTITUTIVE RELATIONS
PLASTICITY
and then C2 is
4.3
4.3.1
0; =
a.
4. Three-dimensional effects
5. Anisotropy:
Material
Induced by loading.
111
0
0
The last section was devoted to the introduction of elastoplastic constitutive models
in the framework of Generalized Plasticity, and it was shown there how Classical
Plasticity models can be considered as particular cases of the theory. The simple models
of Tresca and von Mises present severe limitations when applied to geomaterials in
general and soils in particular. Drucker and Prager proposed, in 1952, an elasticperfectly plastic constitutive model with an associated flow rule which could be applied
to limit analysis problems. However, this model is not able to describe plastic deformations inside the yield surface cone, as occur in common engineering situations. Moreover, the associated behaviour is not valid, as it would predict large dilatancy at failure.
Later, in 1957, Drucker, Gibson and Henkel introduced an elastoplastic model
including two fundamental ingredients, a closed yield surface which consisted of a
cone and a circular cap, and a hardening law dependent on density, paving the way to
modern plasticity.
At the same time, extensive research on the basic properties of soils in triaxial
conditions was carried out at Cambridge University (Henkel 1956, Henkel, 1960 and
Parry, 1960), and these ideas were further elaborated, arriving not only at practical
expressions describing volumetric hardening, but also at the concept of a line in the
( e , p', q) space where the residual states lie. This line was referred to as the Critical
State Line and is one of the basic ingredients of Critical State Theory introduced by
the Cambridge group (Roscoe, Schofield and Thurairajah, 1963; Roscoe, Schofield
and Wroth, 1958; Schofield and Wroth, 1958 and Roscoe and Burland, 1968).
The purpose of this Part is to describe classical elastoplastic models for soils,
together with their limitations, which made it necessary to introduce new concepts.
4.3.2
112
CONSTITUTIVE RELATIONS
Figure 4.14
PLASTICITY
The simplest case in which the first mechanism occurs is hydrostatic loading of a
soil specimen, in which the confining pressure is varied. The process is carried out
slowly enough to prevent the development of intersticial pore pressure (drained
conditions).
If the initial state of stress is
the stress path will consist on a segment of straight line along the hydrostatic axis
d,= d!= d', or along the axis q = 0 if we are using the plane (p',q ) (Figure 4.14). In
above, we have considered compressions as negative.
The change of volume can be described either by the volumetric strain,
where we have used the minus sign for consistency with the definition of p', or the
change of voids ratio e
113
Figure 4.15 Hydrostatic compression test on a normally consolidated clay. (a) Experimental
results (b) Idealized behaviour
It can be seen that unloading and reloading branches differ, although volumetric
strain developing in the branch 2-3-4 can be considered to be reversible. However,
irreversible plastic deformation occurs from 1 to 2 and from 4 to 5. This behaviour is
typical of soft clays, and can be sketched as shown in Figure 4.15(b).
If time effects can be neglected, the response of a soft cohesive soil can be idealized
in the In($) e plot as a line of slope X (Points 1-2-5-8)
-
or, alternatively
X dp'
d ~ ?=, -1+ep1
This line is often referred to as the 'Normal Consolidation Line', and is one of the
basic ingredients of modern plasticity models for soils. The parameter X depends on
the type of soil, and it can be related to the Plasticity Index PI by the empirical
relation (Atkinson and Bransby, 1978)
114
CONSTITUTIVE RELATIONS
Figure 4.16
PLASTICITY
surfacefi, and therefore, no plastic strain is produced in the process. The solution is
to use yield surfaces intersecting the hydrostatic axis q = 0. Loading from 1 to 2
expands the yield surface, which can be assumed to harden as the soil densifies.
The observed volumetric strain can be decomposed into elastic and plastic parts
according to
d ~ ,=, det
A dp'
+ d ~ := ----1 + e p'
dp'
ltey'
K
= --
K,. =
1+e
K
Once the stress point reaches the yield surface again, a plastic strain develops, the
yield surface expands and the soil continues hardening.
A simple law relating the size of the yield surface, which will be denoted by p, to
the plastic volumetric strain is obtained as
from where
115
The subscript 'c' refers to consolidation, and the process 1-2-5 is referred to as
'isotropic consolidation'.
If a soil specimen which has been subjected in the past to a consolidation pressure
of p: = p i is tested at a lower pressure pl, it will be possible to observe in the curve
e - lnp' a change of slope such as depicted in Figure 4.15(a) in the branch 3-4-5. This
soil is referred to as 'overconsolidated', while soils at the normal consolidation line
are called 'normally consolidated'. Both concepts can be easily understood in the
framework of plasticity, as overconsolidated soils are characterized by the stress state
being inside the yield surface at the initial state.
The 'Overconsolidation Ratio', or OCR, is a parameter measuring the degree of
overconsolidation,
Triaxial test
So far, we have only considered stress paths where no shear strain is induced, unless
the soil is anisotropic. It was seen that isotropic compression results in densification
and hardening of soil, and it was mentioned that another mechanism causing
densification was shear. Here we will concentrate in shear behaviour of normally
consolidated clays subjected to symmetric or cylindrical triaxial stress conditions
116
Two main types of tests are currently used: (i) Consolidated drained; and (ii)
Consolidated undrained. In the first case, a saturated soil specimen is brought to
an initial state of hydrostatic stress
I T
u1= (ah,a; , a,,)
= -pl(l, 1,l)
117
In this figure, it can be seen that both the hydrostatic pressurep' and the deviatoric
stress are increasing. The stress path can be studied either in the space of principal
stresses, or, alternatively, in the spaces ( I ; ,J 2 , 19) or (p',q, 8). The last is very convenient, as these invariants are given by
and
with
(u', - .;)I
(u', - u i )
(4.127)
which is precisely the stress induced by the vertical load applied through the ram. In
above, both stresses are negative, and we have supposed that the absolute value of d,
is higher than that of a',.
The measures of strain are those work-associated top' and q
Concerning Lode's angle, it is kept constant during the test, provided that
(d,-a',) does not change sign during the test. This fact occurs when applying
compression-extension cycles.
The stress path in the (p', q) plane is shown in Figure 4.19, where it can be seen
that, due to its positive slope, and for a given absolute value of the deviator q, the
angle AOB is smaller than AOB'. Therefore, if the failure surface is of MohrCoulomb type, M,. 2 Me, the soil will fail earlier in extension. The slope of the stress
path can easily be obtained as
118
CONSTITUTIVE RELATIONS
PLASTICITY
and
d' is constant,
ir
Figure 4.19 Consolidated drained stress path in the p'
Figure 4.20
-q
plane
Figure 4.21
119
There is a tendency of the soil to compact as the test proceeds, caused by the
increase of p' and a rearrangement of soil particles.
Failure takes place at a certain value of stress ratio q = M for tests performed at
different confining pressures.
Soil strength and compaction depend on confining pressure, and increase with it.
A second type of triaxial test is the Consolidated Undrained (CU) test, where, after
consolidation, the drainage valve is closed to prevent dissipation of pore pressure.
The test has to be carried out slowly enough for the pore pressure to be homogeneous
through the specimen. During the test, measurements of pore pressure, axial strain,
vertical stress and cell pressure are taken to monitorize the stress path. Figure 4.21
shows typical results obtained in drained consolidated clays.
It can be seen how the effective stress path bends towards the origin as a consequence of pore pressure increase caused by the tendency of soil to compact. Again the
failure takes place at a line of slope M,., which coincides with that obtained in drained
tests. This test has been classically used to characterize soil behaviour under 'fast'
loading, where pore pressures do not have time to dissipate, in short-term stability
analysis. Of course, this is a simplification, and a complete coupled analysis should be
performed instead.
It has to be noticed that soil strength is lower in undrained than in drained
conditions because of the generated pore pressures.
Undrained behaviour of normally consolidated clays provides an interesting illustration of the shortcomings of the Mohr-Coulomb criterion when used as a yield surface.
Figure 4.22 compares the predicted behaviour of such a model with that observed
in the laboratory. It can be seen how the model overestimates soil strength because it
120
CONSTITUTIVE RELATIONS
PLASTICITY
.......... Predicted
Experiment
p'o
cannot predict the pore pressures caused by plastic volumetric strain which develop
during the test. In the Mohr-Coulomb model, no plastic strain is produced until the
yield surface is reached. In addition to that, if the flow rule is associative, dilation and
negative pore pressures will develop at failure, and the stress path will turn to the
right following the yield surface (Bl-C). This process will be endless, and the
deviatoric stress will keep increasing continuously. In reality, the process is stopped
by cavitation of the pore fluid.
121
Figure 4.23
,4
Figure 4.24
122
CONSTITUTIVE RELATIONS
PLASTICITY
was given by
6 WP = Mp'd&r
from where
= d&,P/d&,P
is
obtained as
where it is interesting to note that dilatancy is zero at the Critical State Line.
The normal to the plastic potential surface is proportional to
and
7
we obtain
123
where p:. is the abscissa at which the surface intersects the hydrostatic axis q = 0.
This surface has been depicted in Figure 4.25, where it can be seen that the normal
to the surface at p' = p', is not directed along the axis. Therefore, the normal will not
be uniquely defined in three-dimensional stress conditions, although it can be
assumed that the surface is rounded off in the proximity of the axis so that the
normal is directed along it.
If the flow rule is assumed to be associated, the yield surface coincides with g.
This model was further elaborated by Burland (1965), who suggested an ellipse as a
yield surface. The work dissipation was given by
PC
Figure 4.25 Yield and Plastic potential surfaces of Cam clay Model
124
CONSTITUTIVE RELATIONS
PLASTICITY
with
The yield surface can easily be obtained by integration of the above and is given by
4.3.3
Extension to sands
The models described so far are able to predict with reasonable accuracy the behaviour of normally consolidated clays. They depart from reality (i) when applied to
overconsolidated soils, as it is not possible to reproduce inelastic strain which develops inside the yield surface, and (ii) when applied to sands.
The behaviour of granular soils depends mainly on density, and two extreme
classes of behaviour can be identified.
Sands at very dense states can be prepared in the laboratory by vibration and not
by compaction, as occurs in the case of cohesive soils. When sheared in triaxial
conditions, the behaviour is similar to that depicted in Figure 4.27.
In the first part of the test the sand contracts, reaching a minimum void ratio e,
and, from there, it dilates. Concerning the deviatoric stress, it increases until a peak is
reached, and then it softens. Finally, it stabilizes at residual conditions, where the
plastic flow takes place at constant volume. Therefore, a critical state exists for sands.
This fact had been first established by Rowe (1962). The results sketched in the figure
follow the ideas of Taylor (1948), who suggested that the moment at which the stress
ratio (deviatoric stress in the plot) reaches the value at which residual conditions will
take place later, the volumetric strain presents a peak. From the point of view of
elastoplasticity, it can be assumed that dilatancy is always zero at the line 7 = Mg,
Axial strain
Figure 4.27
*
Dense
Axial strain
125
2
3
'
Axial strain
Axial strain
Figure 4.28
either before reaching the critical state or at that point. In fact, several investigators
proposed a separate denomination of this line referring to it as the 'Characteristic
State Line' (Habib and Luong, 1978) or the Line of Phase Transformation' (Ishihara,
Tatsuoka and Yasuda, 1975).
An important difficulty encountered is that the specimen is no longer homogeneous long before residual conditions, as the strain concentrates in the shear bands.
Therefore, the observed softening is rather of a structural than a material nature.
If such behaviour is modelled by a basic critical state model such as described in
the preceding section, the best choice is to assume the sand as overconsolidated, as
shown in Figure 4.28.
The behaviour is elastic from 1 to 2, where the yield surface is reached. As the soil
dilates, it softens, and the stress path follows from 2 to 3, where it stabilizes at critical
state conditions.
Under undrained conditions, the results present more important discrepancies.
Dense sands separate at the beginning of the test from the undrained stress path of an
elastic material, which is a vertical segment on the @', q) plane (Figure 4.29(a)).
If such a process is reproduced with a critical state model, the results will be similar
to those shown in Figure 4.29 (b).
It can be seen there that the difference from the observed behaviour is important.
as the strength is underestimated.
On the other side of the density spectrum, very loose sands present liquefaction
under undrained conditions. The phenomenon consists of a sudden drop of resistance of the soil, which behaves as a viscous fluid (Figure 4.30).
It is important to note that the behaviour in the descending branch corresponds to
increasing values of the stress ratio, and, therefore, it is not sound to assume this
behaviour as softening. The separation from the vertical of the stress path shows that
plasticity is present from the beginning of the test.
126
CONSTITUTIVE RELATIONS
PLASTICITY
Axial strain
(4
Axial strain
(b)
Figure 4.29 Undrained behaviour of dense sand in CU triaxial test. (a) Experimental results
(b) Predicted
Such behaviour cannot be described by the models presented above, especially the
dramatic loss of resistance. If the material is assumed normally consolidated, the
results will be similar to those shown in Figure 4.21.
All these limitations motivated further research to extend the range of application
of CS models. The three fundamental ingredients were: (i) hardening laws depending
on deviatoric and volumetric plastic strain; (ii) non-associative plastic flow rules; and
(iii) plastic deformations existing throughout the process.
127
Axial strain
I
Axial strain
Figure 4.30
The first is required to cross the line 77 = M,, as the plastic modulus is zero there
otherwise. Deviatoric hardening was introduced by Nova (1977) and Wilde (1977),
who assumed a hardening parameter of the type
Y=E~+D<
where
(4.149)
The size of the yield surface was made to depend on Y, and the plastic modulus in
triaxial conditions was found to be proportional to
ag
-+
ap'
D-ag
ay.
can be crossed, without the plastic modulus is zero. Failure will occur when
128
CONSTITUTIVE RELATIONS
PLASTICITY
which happens at a stress ratio higher than M,. Once there, if D is kept constant, the
path will not return to the CSL, and failure will take place with dilation. Another
possibility, proposed by Nova (1982), consists of making them drop D to zero, which
results in a discontinuity of slope but with the desired result of coming back to CS
conditions. Finally, a hardening law with saturation can be assumed to hold for D
This law was suggested by Wilde (1977) and applied to a bounding surface model
by Pastor, Zienkiewicz and Leung (1985).
If a negative value of D is assumed, then the plastic modulus becomes zero at a
stress ratio lower than critical, and liquefaction-like behaviour can be modelled in the
softening regime. As discussed above, it is more sound to assume this process to be of
the hardening type, as the stress ratio is continuously increasing.
The second ingredient is a non-associative flow rule, as suggested by Poorooshasb,
Holubec and Sherbourne (1966 and 1967), Nova and Wood (1979), Nova (1982),
Zienkiewicz, Humpheson and Lewis (1975), and Pastor, Zienkiewicz and Leung (1985).
The plastic potential and flow rules can be determined from experiment, as shown
in Nova and Wood (1978), where surfaces were defined by different analytical
expressions valid for different ranges of the stress ratio.
Pastor, Zienkiewicz and Leung (1985) used as plastic potential a simplification of
that proposed by Nova and Wood (1979), assuming now that a single expression was
valid for the full range of stress ratios
dg = (1
CSL
+a). (Mg
rl)
CSL
..'
Figure 4.31
Plastic potential and yield surfaces for (a) loose sands (b) dense sands
ADVANCED MODELS
129
4.4
4.4.1
ADVANCED MODELS
Zntvoduction
So far, we have discussed in the previous sections some classical plasticity models for
soils which have proven to reproduce accurately enough the behaviour of soil under
monotonic loading. They incorporated as basic ingredients a plastic potential and a
yield surface, the latter being allowed to expand or contract depending on whether
the material was hardening or softening. However, material remained elastic within
the yield surface, where no plastic deformation can develop. The immediate consequence is that these models are unable to reproduce either the behaviour of overconsolidated soils, or phenomena occuring during cyclic loading, such as pore
pressure generation in fast processes or densification. Indeed, both phenomena are
related, as the latter is a direct consequence of the tendency of soil to compact. In
fact, it can be shown that variation of pore pressure and densification in undrained
conditions are related by the expression
where n is the porosity, Kf the bulk modulus of pore fluid and KT that of the soil
skeleton.
As these are key aspects which reproduce failure of the soil caused by liquefaction
or cyclic mobility phenomena (Martin et ul., 1975), it motivated an important effort
of research, which proceeded along two main lines.
The first consisted in developing new cathegories of models embracing classical
plasticity as a particular case, and the second was based on introducing the volumetric deformation which is produced by cyclic shearing of a soil as an 'autogenous
volumetric strain', from which suitable densification laws were produced. Here, we
could mention the work of Bazant and Krizek (1976) and Cuellar et 01. (1977) who
developed densification laws in the context of the endochronic theory, and that of
Zienkiewicz and co-workers at Swansea University, who developed simple densification models and implemented them into coupled numerical models (Zienkiewicz er
al., 1978 and Zienkiewicz et a/., 1982). These will be described next.
130
CONSTITUTIVE RELATIONS
PLASTICITY
Densification models
Phenomena such as liquefaction can be thought of being caused by: (i) an accumulation of pore pressure with the number of cycles; and (ii) liquefaction in the last cycle
as it occurs in monotonic loading. This interpretation has motivated the development
of the so called densification models, where simple elastoplastic behaviour of soil and
the accumulation of pore pressure were taken into account by two different mechanisms. The constitutive equation was written as
da'
D,,.(d
dq)
(4.159)
<
a parameter defined as
where I@/ is the amplitude of the stress cycles (deviatoric) and a,,,o the initial value of
the mean effective stress at the beginning of the cyclic loading process.
The 'densification' model differs from other elasto-plastic models described in this
text by a rather arbitrary separation of effects. For this reason we defer the discussion
of its most recent forms to the end of this chapter. We must, however, remark that it
is capable of modelling with good accuracy the liquefaction phenomena encountered
in earthquakes and because of its simplicity it deserves wider use.
ADVANCED MODELS
131
132
CONSTITUTIVE RELATIONS
PLASTICITY
Bounding surface
occur in cyclic loading, the difference is that bounding surface models are able to
introduce plastic deformations by using some interpolation rules relating the stress
point P(C in Figure 4.32) to an image of it on the BS, PBs(Bin Figure 4.32). Simple
interpolation rules were proposed by Dafalias and Herrmann (1982), and by Zienkiewicz, Leung and Pastor (1985). There, to obtain the image point PBS,a line was
drawn passing through the origin and point P, its intersection with the bounding
surface being taken as the image point. Directions n and n, in P were assumed to be
those at Pss and the plastic modulus was interpolated according to a simple law
where 6 is the distance from the origin to the stress point P, and So the distance
between the origin and the image point PBs, y being a parameter of the model
(Figure 4.32).
The main shortcoming of early BS models was their inability to reproduce plastic
deformations which develop when unloading, and it was overcome within the more
general framework of generalized plasticity (Pastor, Zienkiewicz and Leung, 1985).
Here, the model was of bounding surface type for loading, but plastic deformations
during unloading were introduced within the more general framework of generalized
plasticity. A further step was given by Pastor, Zienkiewicz and Chan, introducing a
full generalized plasticity model in Pastor and Zienkiewicz (1986) and Pastor, Zienkiewicz and Chan (1990), which was applied by the authors to reproduce the
behaviour of both cohesive and frictional soils under monotonic and cyclic loading.
ADVANCED MODELS
133
were u is the unit tensor along direction of loading increment. If constitutive tensors
along positive and negative directions of the principal directions are N+ and N-, a
simple incrementally non-linear law could be given by
where
The model proved to reproduce well the behaviour of soils under both monotonic
and cyclic loading (Darve, Flavigny and Rojas 1985).
Dafalias and co-workers presented extensions of the bounding surface model
within the framework of hypoplasticity (Dafalias, 1986; Wang, Dafalias and Shen,
1990).
Hypoplastic models have been introduced also in Karlsruhe by Kolymbas and coworkers (Wu and Kolymbas, 1990; Kolymbas and Wu, 1993), producing general
expressions for the constitutive tensor.
CONSTITUTIVE RELATIONS
134
4.4.2
PLASTICITY
where
E,.
d& dt,
d : de = (p'q).
(2:)
defined as
If the ratio between the plastic increments of volumetric and deviatoric strain is
assumed to be the same then the ratio of total (elastic plus plastic) increments
observed in laboratory tests, direction n, can be immediately obtained.
To this end, experiments carried out by Balasubramanian and Chaudhry (1978)
using constant p'lq stress paths suggest that dilatancy can be approximated by a
straight line in the p'-q plane (Figure 4.33).
Dilatancy is therefore expressed as
where M, is the slope of the critical state line in the p'-q plane, cr a material constant
and 7) is the so called stress ratio defined as
135
ADVANCED MODELS
0.5
I .O
Figure 4.33 Dilatancy of soft Bangkok clay (after Balasubramanian and Chaudry)
with
This law can also be used to describe dilatancy of granular materials, as it was
suggested in Pastor, Zienkiewicz and Leung (1985) according to test results reported
by Frossard (1983).
Concerning direction n, we will assume that the flow rule is associative, following
Atkinson and Richardson, who performed experiments in three cohesive soils and
found little discrepancies from plastic potential and yield surfaces (Atkinson and
Richardson, 1985).
Therefore, we will have
with
136
CONSTITUTIVE R E L A T I O N S PLASTICITY
and
In what follows, we will drop the subindex 'g' referring to the plastic potential
surface as it coincides with the yield surface.
To obtain the plastic modulus for virgin loading, we will consider an isotropic
consolidation test of a normally consolidated specimen, for which the increments of
volumetric elastic and total strain are given by
K
dp'
d ~ ;=
, -l+epl
and
x dp'
de,. = -l+epl
from which the plastic volumetric strain increment is
~EP=
. (AI + eK ) dp'
p'
--
Comparing now the above equation to the general expression for the plastic strain
increment
1
d~~ = -n.(n
HL
: da')
(4.183)
which particularizes to
for the stress path considered here, it can be concluded that the plastic modulus H L is
given by
The parameters X and K are the slopes of the normal consolidation and elastic
unloading lines in the (e, In p') plane, and Ho is a material constant.
To generalize this expression of the plastic modulus to other conditions than
isotropic compression paths, we will make the assumption that plastic modulus
depends on the mobilized stress ratio, decreasing as the later increases until reaching
a value of zero at the critical line (7 = q/pl = M).
137
ADVANCED MODELS
Therefore,
and
where M,. is the slope of the critical state line obtained in standard compression
triaxial tests and 0 Lode's angle
and
K
1
dEt = -dp'
1 ep'
138
CONSTITUTIVE RELATIONS
PLASTICITY
where
The model presented so far concerns normally consolidated clays under virgin
loading. In order to assess model performance, their predictions will be compared
against a full set of tests carried out by Balasubramanian and Chaudhry in 1978 for
soft Bangkok clay.
The proposed model has five parameters, i.e., two elastic constants, the slope of the
critical state line M, the constant a characterizing dilatancy and Ho, which appears in
the plastic modulus. T o determine them the following procedure may be followed:
The elastic constants can be easily determined from unloading-reloading tests.
Here, they were found from the constant p' tests reported.
The slope M of the critical state line on the (p' - q) plane is found from drained,
undrained or constant p' tests.
The parameter a controlling dilatancy can be found from dilatancy plots, and it is
given by
Experiments
- predicted
1r: ,\y--
Figure 4.34 Constant p' test on Bangkok clay (After Balasubramanian and Chaudhry)
::(':
100.
. :sl
Experimants
I:
Predicted
p'Wa)
100.
200.
300.
400.
0.10
0.20
Figure 4.35 Consolidated undrained tests on Bangkok clay (After Balasubramanian and
Chaudhry)
ADVANCED MODELS
139
Figure 4.36 Consolidated drained tests on Bangkok clay (After Balasubramanian and
Chaudhry 1978)
The results obtained with the proposed model are shown in Figure 4.34, 4.35 and
4.36. First of all, Figure 4.34 shows the constant p' tests together with the experimental results. Next, the model is applied to simulate the consolidated drained
behaviour of Bangkok clay (Figure 4.35), and, finally, we present the results obtained
for Consolidated undrained tests (Figure 4.36)
Overconsolidated clays
The model described in the previous section can be extended to describe the behaviour of overconsolidated clays. To this end, we will introduce a function accounting
for memory of past history, which will consist of storing the past event of maximum
intensity. The mobilized stress function proposed in Pastor, Zienkiewicz and Chan
(1990) is
where f (7) has been given in the previous section and CMAX is the maximum value
previously reached by the mobilized stress function.
In the above, we have introduced a deviatoric strain hardening function g(<)
(Wilde, 1977),
where
140
CONSTITUTIVE RELATIONS
PLASTICITY
and
Therefore, two additional parameters y and 00 are needed to extend the range of
application of the model to overconsolidated clays.
It should be noted that for first or virgin loading of clays above expressions reduce
to those previously proposed for normally consolidated clays.
Figures 4.37 and 4.38 show the behaviour of normally and heavily
consolidated Weald clay reported by Henkel (1956), together with the model predictions.
It is important to note that in overconsolidated clays the peak value of the stress
ratio may be higher than M, then decreasing to reach it as a residual state.
For cyclic loading it is possible to obtain quite satisfactory results for clays
using simple elastic unloading and thus avoid the introduction of additional
parameters. We show in Figure 4.39 the performance with the above assumption
for cyclic tests with constant stress amplitude carried out by Taylor and Bacchus
(1969).
Finally, Table 4.1 gives the parameters used in the simulations described above.
Bangkok clay
Weald clay
Figure 4.37 Behaviour of normally consolidated Weald clay (after Henkel 1956)
141
ADVANCED MODELS
r z
Axial strain
10
Number of cycles
20
Figure 4.39 Behaviour of clay under two-way strain-controlled triaxial loading (data from
Taylor and Bacchus 1969)
142
CONSTITUTIVE RELATIONS
PLASTICITY
which coincides with the projection of the critical state line on the plane (pl-q).
This line has also been referred to as the 'characteristic state line' (Habib and
Luong, 1978) or the 'line of phase transformation' (Ishihara, Tatsuoka and Yasuda,
1975) and plays an important role in modelling sand behaviour as will be shown later.
It has to be noted that this line is not the critical state line, which will be reached at
residual conditions. Whether the critical state line existed or not has been a matter of
discussion during past years, due to the difficulty of obtaining homogeneous specimens at failure after shear bands have developed. However, recent experiments
carried out at Grenoble by Desrues have shown that inside the shear band a critical
void ratio is reached.
During a test, this line can be crossed a first time, with the specimen still far from
the residual state. If shearing continues, the stress path will finally approach the
critical state line.
Therefore, the condition 77 = M, represents two different states at which dilatancy
is zero, the 'characteristic state' and the critical state.
The direction of plastic flow n , ~can be determined in the triaxial space by similar
procedures used in cohesive soils, giving
with
with
ADVANCED MODELS
143
where
where the size of both surfaces is characterized by the integration constants pi, andpk.
Both surfaces were depicted in Figures 4.31(a) and (b) for a medium-loose sand,
together with experimental data obtained from accoustic emission (Tanimoto and
Tanaka, 1986). Similar yield surfaces were proposed by Nova (1982).
To derive a suitable expression for the plastic modulus HL. it is necessary to take
into account several well established experimental facts:
(i) residual conditions take place at the critical state line
(ii) failure does not necessarily occur when this line is first crossed.
(iii) The frictional nature of material response requires the establishment of a
boundary separating impossible states from those which are permissible.
A convenient law was introduced in Pastor and Zienkiewicz (1986) in the form
where
together with
144
CONSTITUTIVE RELATIONS
PLASTICITY
having thus
Figures 4.40(a), (b) and (c) show, respectively, the stress paths, deviatoric stress vs.
axial strain and pore pressures obtained by Castro, together with the model predictions, which agree well with them.
(ii) At the other end of the density range, peaks exist in deviatoric stress during drained
shear of very dense sands, this effect developing progressively as density is increased.
The factor H, is introduced in the expression giving a plastic modulus to account for:
crossing of the Characteristic State line (7 = M,) without immediately producing
failure,
reproduction of softening
residual conditions taking place at the critical state line.
145
ADVANCED MODELS
(a)
p'
392 KPa
200
400
600
Mean e r l t c b v c confinmg pressure p' (kPa )
Figure 4.40 Undrained behaviour of Banding sand (data from Castro 1969) Computed
results shown by solid line. (a) p'-q plot (b) Deviatoric stress vs. shear strain plot (c) Pore
pressure vs. shear strain plot
146
Figure 4.41
1948)
10
I5
Axial strain (Of;)
10
15
Axial strain ("h)
20
5
10
IS
20
A x d strain (",I)
10
10
15
Axial strain ('14)
20
Drained behaviour of dense and loose sand (experimental data from Taylor,
To illustrate the role of the plastic modulus in the transition from softening to
hardening regimes, let us consider a drained triaxial test (Figure 4.41). During a first
part of the path, both H I ,and H, are positive and decrease in a monotonous way. At
17 = M,, i.e., when crossing the characteristic state line, H,, becomes zero, while H, is
still positive. If the process continues, a moment arrives at rl, where
with
If the test is run under displacement control, the deviatoric stress does not change
for an infinitesimal variation of the strain
dp'
= dq = 0
ADVANCED MODELS
147
Meanwhile, H, has decreased, and consequently, the plastic modulus becomes negative. The soil has entered the softening regime, and from this moment the deviatoric
stress will present a descending branch.
The deviatoric strain hardening function H , will vanish as deformation progresses,
reaching a final asymptotic value of zero at 17 = M,, this time at the critical state line.
During the softening process,
and
The overall response is governed by the ratio between the width of the shear band
and the length of specimen. This effect is similar to what can be observed in
numerical computations and which has been referred to as mesh-dependence.
148
"
r)
2 -50
Hollow cylinder
Hostun sand
conf.press.207kPA
- Experiment
o
Prediction
-3
-2
Axial strain (%)
Test HH3
Hollow cylinder
Hostun sand
conf.press.207kPA
&
in
in
z2 -100
-1
- Experiment
'tiu
Prediction
.m
-150
"
(a)
(b)
Figure 4.42
and
dp' = 0
Figures 4.40(a), (b) and (c) show how relative density influences the undrained
behaviour of sand, together with predictions of the proposed model.
149
ADVANCED MODELS
Three-dimensional behaviour
So far we have considered the triaxial response of soil (compression and extension).
However, proposed relations have been made dependent not only on I; and J2 (or,
alternatively, p' and q), but also on the third invariant or equivalently on Lode's
angle 6.
Hence, the soil response can be generalized to any path out of the triaxial plane.
Denoting triaxial stress parameters p', q and 19by a*
= d a * : n*
n,
=n,
da
(4.226)
150
CONSTITUTIVE RELATIONS
PLASTICITY
Figure 4.43
Constant b tests on Reid sand (experiments from Saada and Bianchini 1989)
151
ADVANCED MODELS
500 -
PREDICTED
EXPERIMENTAL - - - - - - - - - -
Figure 4.44
Experiments with pure rotation paths show a plastic volumetric strain under
drained conditions and pore pressure generation under undrained loading.
Generalization of the proposed model for these phenomena can be made by considering several mechanisms, as it was proposed in Pastor. Zienkiewicz and Chan ( 1 990)
using models able to introduce strain and load-induced anisotropy (Pastor, 1991 ).
152
CONSTITUTIVE RELATIONS
EXPERIMENTAL
PLASTICITY
-
PREDICTED
0.60
I
Mean confining pressure
Figure 4.45 Undrained behaviour of loose sand under reversal of stress (experiments after
lshihara and Okada, 1982)
=H
(5) 1 1
'
for
>I
ADVANCED MODELS
153
where
and
where
Figure 4.45 shows a prediction of this model extension for the experimental results
of Ishihara and Okada.
It is possible now to model cyclic phenomena as liquefaction and cyclic mobility
which appear in loose and medium sands under cyclic loading, and which are
responsible of catastrophic failure of structures subjected to earthquakes.
Both phenomena are largely caused by the overall tendency of medium and loose
sands to densify when subjected to drained cyclic shearing. If the load is applied fast
enough or the permeability is relatively small, this mechanism causes progressive
pore pressure build-up leading to failure.
In the case of very loose sands, liquefaction takes place following a series of cycles
in which the stress path migrates towards lower confining pressures. Figure 4.46
shows the results obtained by Castro (1969) in his pioneering work.
Denser sands do not exhibit liquefaction but cyclic mobility. Failure here is
progressive since the stress path approaches the characteristic state line by its shift
caused by pore pressure build up. Deformations during unloading cause the stress
path to turn towards the origin, and strains produced during the next loading branch
are of higher amplitude.
Figure 4.47 shows both the experimental results obtained by Tatsuoka on Fuji
river sand.
154
CONSTITUTIVE RELATIONS
Experimental
PLASTICITY
Experimental
(a) 2
(bi
0
pt (kg I cm-)
5
Axial stram (%)
10
Predicted
Predicted
2
3
p' (kg 1 cm-)
10
Figure 4.46 Liquefaction of loose banding sand under cyclic loading (predictions and
experimental data from Castro, 1969)
(a)
(b)
Prcdicted
Predicted
OE$l
m o
Y,
0-
-I
(c)
p'1(kg I cm-)
2.5
I 10.0
i?
j
(4
Experimental
Exper~mental
-5
"
Y,
I
,
p' (kg 1 cm-)
-I
2.5
-10.0
0
Axial stram (%)
Figure 4.47 Cyclic mobility of loose Niigata sand (predictions and experimental data from
Tatsuoka, 1972)
155
ADVANCED MODELS
Fig.4.41(loose) Fig.4.42
Fig.4.45
Fig.4.46
Fig.4.47
Table 4.2 above gives the model parameters used in the preceding simulations.
The model can be further elaborated as shown in Pastor, Zienkiewicz and Chan
(1987) and Pastor, Zienkiewicz, Xu and Peraire (1993) by improving the way in which
the history ofpast events is taken into account. To thisend, two elements are introduced:
(i) a surface defining the maximum level of stress reached, and
(ii) the point at which last reversal took place.
Directions n and n,, and the plastic modulus HL depend on the relative position of
the stress C with respect to the point at which the load was reversed, B, and an image
point, D, defined on the same mobilized stress surface as B.
To obtain the values of H L , n and n,, suitable interpolation rules are used. In
particular, n is interpolated from -n to n using a linear law. The direction of plastic
flow is obtained again by defining a suitable dilatancy at C, d g which is interpolated
from an initial value dgoto
156
CONSTITUTIVE RELATIONS
Figure 4.48
PLASTICITY
Interpolation rule
where the constant C,(O < C, < 1) varies with the density, being close to zero for
medium-loose sands.
The plastic modulus is interpolated between an initial value HUoand its final value
at the image point on the mobilized stress surface H D . The initial value can be
assumed to be infinite to decrease a possible accumulation of plastic strain under
very low amplitude cycles.
Figure 4.49 shows the densification of a loose sand under cyclic loading, and it can be
observed how the volumetric and deviatoric plastic strain produced decreases with
the number of cycles.
It should be mentioned here that since the simple models we have described here
were proposed, several improvements and modifications have been introduced,
157
ADVANCED MODELS
Stress ratio
- 1 .0
1 .0
2.0
Figure 4.49
particularly at CERMES (Paris) (Saitta, 1994). where research recently finished has
succeeded to include state parameters describing, in a consistent way, the behaviour
of sand under different conditions of confining pressure and relative density (Bahda,
1997; Bahda, Pastor and Saitta, 1997).
4.4.4
Anisotropy
Introductory remarks
Anisotropy in geomaterials is caused either by the arrangement of particles such as
occur in natural deposits in which the grains may have their major axes on the
bedding planes, or by the spatial distribution of contacts and contact forces. In the
first case, it is found that the strength is higher when tested along the deposition
direction. This effect can introduce important errors if not taken into account. For
instance, if the number of cycles to liquefaction is determined using a standard
triaxial testing machine to evaluate the liquefaction potential of a natural sand
158
CONSTITUTIVE RELATIONS
PLASTICITY
deposit, the value obtained will be greater, and, therefore, the strength will be
underestimated.
Several theories have been proposed within the framework of plasticity to describe
both initial and stress-induced anisotropy. Basically, anisotropy has been
approached most of the times by changing the position, orientation and shape of
isotropic yield, loading or plastic potential surfaces, in such a way that those changes
were dependent on tensors such as stress or strain and not only on their invariants.
Initial or fabric anisotropy could be reproduced as well by introducing initial
movements and distortions on the surfaces.
Combination of kinematic and anisotropic hardening laws proposed by Mroz
(1967), have provided a suitable way to model anisotropic behaviour of soils
(Mroz, Norris and Zienkiewicz, 1979; Hashiguchi, 1980; Hirai, 1987; Ghaboussi
and Momen, 1982; Liang and Shaw, 1991; Cambou and Lanier, 1988 and di Prisco,
Lanier and Nova, 1993). Surfaces can be allowed to expand following isotropic
hardening rules and to translate and rotate under anisotropic hardening laws.
If only initial or fabric anisotropy is to be considered, a simple way to introduce
anisotropic surfaces is to define a modified second invariant of deviatoric stress tensor
159
ADVANCED MODELS
place at all possible contact planes within the material. Of all possible active planes
only a reduced number of sampling planes is considered. The overall response is
obtained by a process of numerical integration extended to sampling planes.
Alternatively, if attention is focused only on planes normal to XY, YZ and ZX,
and their responses are grouped together, one finally arrives at a three-mechanism
model (Aubry, Hujeux, Lassoudiere and Meimon, 1982 and Matsuoka, 1974).
Both multi-laminate and multi-mechanism models of the type described above can
produce plastic strain under pure rotation of principal stress axes (Pande and
Sharma, 1983; Pastor, Zienkiewicz and Chan, 1990).
Alternatively, the behaviour of the material can be assumed to be caused by
superposition of responses to variations in 4,o;and 4 , and can then be generalized
to more general stress conditions. This has been proposed by Darve and Labanieh
(1982) and applied to complex stress paths, including anisotropy effects.
Finally, it should be mentioned that material fabric plays a paramount role on
geomaterials anisotropic response and it is in turn modified by the deformation
process. The fabric may be approximated by a second-order tensor, which can be
incorporated into the constitutive equations. An interesting way has been recently
proposed by Pietruszczak and Krucinski (1989), and consists on adding two components to obtain the increment of plastic strain. First one corresponds to an isotropic
hardening mechanism, and the second accounts for deviations of isotropy, which are
made dependent on fabric tensor.
Proposed approach
It has been mentioned above that material structure or fabric has to be incorporated
in the constitutive equations to account for both initial and induced anisotrpy. Here,
it will be assumed that fabric can be described by a second-order structure tensor A,
which will determine its type of symmetry. If Q is a rotation or reflection tensor, the
class of symmetry will be defined by the set of operators Q which fulfill
(a
0
0 cos0 -sin0
s i n O:oc
160
CONSTITUTIVE RELATIONS
PLASTICITY
Now, the structure tensor can be used to define a fourth-order anisotropy tensor B
from which a modified second invariant Ji can be derived as suggested by Hill (1950);
Nova and Sacchi (1982) and Nova (1986). Following Cowin (1985), B can be
expressed as a combination of terms listed below.
(i) 6 @ 6
(ii) 6 @ A A t 3 6
(iii) 6 t3 A2 A'
(iv) A t3 A
(v) ABA'
(vi) A'
A'BA
A'
where rows and columns include components of the tensor with their components
ordered as
and, therefore, component 1133 is located at first row and third column.
ADVANCED MODELS
161
In the above,
It can be seen that the resulting anisotropy tensor depends only on five constants,
and that the form proposed by Nova and Sacchi (1982) is a particular case of the
above expression in which BI and B4 have been made one and zero respectively.
So far, only JS has been extended to account for anisotropy. However, geomaterial
behaviour is also dependent on first and third invariants, and anisotropy should also
be reflected on them. New invariants fl and J-; can be introduced in a similar manner,
by defining B' and B"' which are tensors of orders two and six.
The first anisotropy tensor B' would be dependent on 6 and A, and B'" on double
tensorial products of S, A and
such as S @ S @ S, A @ S 8 6 , etc.
Therefore, the extended set of invariants is given by
Finally, constitutive laws derived for isotropic materials in terms of I,'.J: and J ;
can be generalized to anisotropic situations by substituting them by modified forms
given above.
An interesting particular case is obtained when A is taken as
Then, the constitutive law can be seen to be dependent on joint stress-strain invariants, as proposed by Baker and Desai (1984).
As mentioned above, the result of substituting stress invariants by their modified
forms can be viewed as introducing a rotation and a distortion of yield and plastic
potential surfaces, and indeed some kinematic hardening models in which a back
stress is introduced can be considered as particular cases of the theory outlined
above. They present, however, the advantage of being simpler to develop.
If the three tensors introduced to produce the modified invariants are defined as:
162
CONSTITUTIVE RELATIONS
PLASTICITY
it can be checked that this choice corresponds to a pure rotation of the yield and
plastic potential surfaces. In the case of generalized plasticity, where surfaces are not
introduced, it will be a rotation of directions n and n , ~ , " (Pastor, 1991).
The model described in Pastor (1991) and Pastor, Zienkiewicz, Xu and Peraire
(1993) introduced a rotation characterized by a new direction of the plastic potential
and yield surface axes given by a unit tensor a given by:
q, g) can be
Alternatively, a modified form of the classical set of invariants (j',
obtained. Care should be taken when defining a new Lode's angle as the trace of the
new modified deviatoric stress is not zero. A possible solution consists of defining a
new deviatoric stress t as
from which
and
ADVANCED MODELS
163
and e can now be introduced into the generalized plasticity model described in the
preceding section, to describe the anisotropic behaviour of sand.
where M, depends only on Lode's angle 0 and q' is the stress ratio. The tests
performed by Yamada and Ishihara (1979) consisted of proportional, radial
paths performed at constant p' and 0. A detailed description of both the testing
procedures and the results obtained is given there. The samples exhibited a
strong anisotropy as grains were arranged such that their long axes were
horizontal. Therefore, different behaviour was observed along paths such as
ZC and YC which have the same value of Lode's angle. However, no such
dependence was found for M,, which will be assumed to be independent of
anisotropy.
0
A similar analysis may be carried out to study how soil dilatancy, is affected by
soil anisotropy. Miura and Toki (1984) analyzed the experiments of Yamada and
Ishihara (1979) and concluded that soil anisotropy did not greatly affect the
dilatancy behaviour of sand, confirming the linear relation between dilatancy
and stress ratio proposed above for isotropic soils.
It was found that the parameter cu depended on 0, and the following relationship
was proposed
164
CONSTITUTIVE R E L A T I O N S PLASTICITY
with axis X I coinciding with direction of deposition, while isotropic materials will be
characterized by
As the material is loaded, deformation will produce rotation of grains and rearrangement of microstructure. Therefore, A will change, and dA will be a function of
the increment of strain.
General expressions for this functions have been suggested by Pietruszczak and
Krucinski (1989). Here we will assume that dA may be expressed as
where CCloand CClIare two material parameters and E is the accumulated deviatoric
strain.
If CLloand C,,, are taken as zero and one respectively, and no initial anisotropy
exists, A will coincide with EP, and the modified invariants will be functions of both
the stress and mixed stress-strain invariants.
To show the performance of the proposed model, Figure 4.50 reproduces both the
results obtained by Yamada and Ishihara (1979) and the model predictions (Pastor,
1991). The tests were run on the Fuji River, with specimens constructed by pluviation
of sand through water to simulate the natural deposition process. This resulted on a
highly anisotropic structure which was modified by subsequent loading. Cubic samples of sand were tested on a true triaxial apparatus along different paths. Those
depicted in Figure 4.50 correspond to compression and extension along the vertical.
It is interesting to notice how the anisotropic structure induced by the deposition process resulted in a higher overall stiffness for the specimen tested in com-
165
MODIFIED DENSIFICATION M O D E L
4 (kglcm-)
Figure 4.50
...
Predicted
-Experiment
pression along the vertical, and how the situation was reversed when loaded in
extension.
4.5
4.5.1
Introduction
The densification model is firstly introduced with the classical Mohr-Coulomb
criteria for shear strength (Zienkiewicz rt ul, 1978). That model is suited to simulate
undrained tests in which the excess pore pressure is built up due to the constant cyclic
load, until the full liquefaction takes place or the shear stress reaches the phase
transformation line. It cannot, however, simulate the cyclic mobility behaviour
after the stress point reaches the phase transformation line. The most important
application is the soil layer problem. After the stress reaches the failure line of the
Mohr-Coulomb densification model in the middle of the soil layer, the wave does not
transfer upward and no response acceleration is observed. This can happen if the soil
layer consists of very loose sand but the dense soil layer resists and transfers the
166
CONSTITUTIVE RELATIONS
PLASTICITY
upward wave propagation. The modification of the densification model for this kind
of phenomena is introduced here. The definition and simulation of a triaxial test is
presented.
where 71 is the stress ratio against the current mean stress and H is the stress ratio
against the initial mean stress. Most experimental data refers to 0 and very few data
are available for the relation between the dilatancy and rl. But if the dilatancy
depends on the stress ratio, it should theoretically be 77.
Cyclic mobility can be included in the function h(r1). The empirical formulation
can be used for this purpose.
In order to simulate a triaxial undrained test, the above function can be modified
further to avoid numerical difficulty. The dependence of ti is included in the function
h; that is h ( 7 , K ) . Therefore, the function K ( K )is defined as
Definition of function
Before cyclic
mobility
r
After cyclic mobility
4q3 K )
=Y,P(~)IXK)
167
where A,, B,,, y,,are the model parameters of the densification model. 'n' is the index
for the strain region, n = 1 before cyclic mobility, n = 2, after cyclic mobility, and
n = 3 when K > k,. as shown in Table 4.5
a ( 7 ) is is the differential of the stress path shown in Figure 4.51 and is a function of
the stress ratio q(= ~ / a : , )It
. is defined as an empirical function and described in the
table as { ( a l711,
, ( a Zq2),
, . . . (a,,,G ) ) .The different function is used before and after
the liquefaction as show in Table 4.4.
0.0
0.1
Figure 4.51
Table 4.4
0.2
0.3
0.4
0.5
0.6
o:,,
(MPa)
Model of stress path for dilation behaviour
Definition of a ( q )
Function a ( q )
Before cyclic mobility After cyclic mobility
For loading
Defined by table
(e.g. Table 4.7)
For unloading
lrll
0.1 q'?
lrll
1 -%
11,
lrll
'11
where 1 7 ~= tan& and q, = tan$, , & is the phase transfer line and $f is the failure line.
P(K) depends on the damage parameter.
Table 4.5
Definition of
2 K,.
(a)
Loading
I - C~L(O)
C ~ L ( O ) +I + B l K
cru
+ 1 +1B-Z (Czc
K KO)
-
Unloading
1.0
I CZL
C2Lt1+BI(~-~U)
-
168
CONSTITUTIVE RELATIONS
Figure 4.52
Table 4.5
(b)
Contu' for k
PLASTICITY
> kc
After cyclic mobility (n=3)
Loading
Unloading
Here KO is the damage parameter when cyclic mobility takes place. The coefficients
C7L and C2L' are coefficients for the loading process and the unloading process
respectively. An example of P ( K ) is shown in Figure 4.52. The function P ( K ) reaches
a certain value when the damage parameter ti becomes large.
169
stiffness at low shear stress. This is well represented. Furthermore the enlargement of
the strain after cyclic mobility takes place also shows good agreement, as shown
Figure 4.55.
Table 4.6 Parameters of modified densification model
c
4
4
0.OMPa
41.8 deg.
25.0 deg.
yl
AI
B1
4.0 x lo7
1.0 x lo-'
2.0
Table 4.8
25.0 y3
0.016 A3
8.0
B3
y2
A?
B2
b
G
471)
Function C4L
Experiment (q = 0.147)
Calculation
0.147)
0.01 5
Vi
-g2 ez:2
2 rc
m
0.000
-0.015
0.00
0.02
0.04
0.06
0.08
Effective mean stress o',(MPa)
0.10
2.50
0.50
90Mpa
170
CONSTITUTIVE RELATIONS
PLASTICITY
Experiment (q=O.147)
Calculation (q=0.147)
0.015
-z2 r_
P.
2
V)
0.000
-0.0 15
-4
Figure 4.54
-2
0
Shear strain y(%)
10
Number of cycles N
100
Figure 4.55 Enlargement of shear strain due to accumulation of damage (experiment and
simulation)
REFERENCES
171
REFERENCES
Anandarajah, A. and Dafalias. Y. F. (1986) Bounding surface plasticity 111: Application to
anisotropic cohesive soils J. Eng. Mech. ASCE, 112, 12, 1292-1 3 18.
Atkinson, J. H. and Bransby, P. L. (1978) The Mechanics qfSoi1.s; An Introduction to Criticcrl
State Soil Mechanics. McGraw Hill, London.
Atkinson, J. H. and Richardson, D . (1985) Elasticity and normality in soil: experimental
examinations, GPotecliniqlre,35. 443449.
Aubry, D. Hujeux, J. C. Lassoudiere. F. and Meimon, Y. (1982) A double memory model with
multiple mechanisms for cyclic soil behaviour, Int. Synzp. Num. Models in Geomechanics.
R. Dungar, G . N. Pande and J. A. Studer, (Eds) Balkema, Rotterdam.
Bahda. F. (1997) Etude du comportement du sable au triaxial, PhD Thesis, ENPC-CERMES
Paris.
Bahda, F. Pastor, M. and Saitta, A. (1997) A double hardening model based on generalized
plasticity and state parameters for cyclic loading of sands, G . N. Pande and S. Pietruszczak.
Baker, R. and Desai, C. S. (1984) Induced anisotropy during plastic straining, Int. J. Num.
Anal. Methods Geon~ech.,8, 167-185.
Balasubramanian, A. S. and Chaudhry, A. R. (1978) Deformation and strength characteristics
of soft Bangkok clay, J. Geotech. En& Div. ASCE, 104, NO.GT9, 1153-1 167.
Banerjee, P. K. and Yousif. N. B. (1986) A plasticity model for the mechanical behaviour of
anisotropically consolidated clay, Int. J. Num. Anal. Methods Geomech, 10, 521-541.
Bardet, J. P. (1989) Prediction of deformations of Hostun and Reid Bedford sands with a
simple bounding surface plasticity model, in Constitutive Equations ,for Grunulcw Noncohesive Soils, A. Saada and G . Bianchini (Eds), 131-148, Balkema.
Baiant. Z. P. and Krizek, R. J. (1976) Endochronic constitutive law and liquefaction of sand.
Proc. Am. Soc. Civ. Eng. EM 102, 225-38.
Burland, J. B. (1965) Correspondence on 'The yielding and dilatation of clay'. Gkotechnique 15.
21 1-214.
Cambou, B. and Lanier J. (1988) Induced anisotropy in cohesionless soil: experiments and
modelling, Computers and Geotechnics 6, 291-3 1 1.
Castro, G . (1969) Liquefirction of'scmds, Ph.D. Thesis, Harvard Univ., Harvard Soil Mech.
Series no. 8 1.
Chan, A. H. C. Zienkiewicz, 0. C. and Pastor, M . (1988) Transformation of incremental
plasticity relation from defining space to general cartesian stress space, Comrnrm. Appl.
Num. Meth., 4, 577-580.
Cowin, S. C. (1985) The relationship between the elasticity tensor and the fabric tensor, Mech.
Materials, 4, 137-1 47.
Coulomb, C. A. (1773) MPm. M~rth.et Phys. 7 , 343.
Cuellar, V.. Baiant, Z. P., Krizek, R. J. and Silver, M. L. (1977) Densification and hysteresis of
sand under cyclic shear, J. Geotech. Eng. Div.. ASCE, 103, 9 18.
Dafalias, Y. F. (1986) Bounding surface plasticity. I: Mathematical foundation and
hypoplasticity, J. Eng. Mech. ASCE, 112, 966-987.
Dafalias, Y. F. and Herrmann, L. R. (1982) Bounding Surface Formulation of Soil Plasticity.
in Soil Mechanics-Transient and Cyclic L O N ~GP., N . Pande and 0. C. Zienkiewicz (Eds).
Ch. 10, 253-282, Wiley.
Dafalias, Y. F . and Popov, E. P. (1975) A model of non-linearly hardening materials for
complex loadings, Acta Mech. 21, 173-192.
Darve, F. (Ed.) (1990) Geonirrter-iuls. Constitutive Equations and Modelling. Elsevier Applied
Science.
Darve. F. and Labanieh, S. (1982) Incremental constitutive law for sands and clays: simulation
of monotonic and cyclic tests, Int. J. Numer. Anal. Meth. Geomechs. 6, 243-275.
172
CONSTITUTIVE RELATIONS
PLASTICITY
REFERENCES
173
174
CONSTITUTIVE RELATIONS
PLASTICITY
175
REFERENCES
5.1 INTRODUCTION
In this chapter we deal with the solution of static and quasi- static problems in which
dynamic (inertial) effects are negligible. In the latter category the phenomena associated with consolidation are typical and both fully and partially saturated cases will
be presented. Since finite deformation and partial saturated examples are only given
in this chapter, examples involving hypothetical dynamic behaviour using finite
deformation and under partially saturated condition respectively are also given.
In both classes of problems we are concerned with the deformation and movement
of the soil or of its associated foundation. An excessive amount of the latter is a
measure of failure generally or loss of servicability where the stress distribution or its
magnitude of pore pressures are indicative of the state of the material stressed. It is
the dejormution and displucement which are observable and must be determined. For
these the knowledge of the constitutive relation discussed in the previous chapter is of
paramount importance, but the simplest constitutive law, which answers the question
posed by the engineers and which provides the determination of failure, is to be used
at all times. Failure is sometimes associated with continuing displacement without a
load increment. This is a definition which is often accepted. However, on occasion a
finite displacement can be specified as failure by the engineer and knowledge of
displacements is important even if these are not excessive. In the first three chapters
of this book we have formulated the dynamic problem and its solution with the time
dependence being retained in the final discretized equations.
It is clear (as indicated earlier) (see Equation (3.66)) that the problems of consolidation can be directly solved by the code based on our formulation, as with slow
motion the dynamic effects become automatically negligible. Less computational
effort is involved if the acceleration terms are neglected and the G N l l scheme is
used for both the skeleton displacement and pore pressure. However, it is not
178
obvious that static problems can be directly dealt with by the general program
though it has been customary to create special programs for such analyses. In Section
3.2.4 we have shown how the static problem can be dealt with by the dynamic code
without any loss of efficiency when an appropriate time stepping scheme, i.e. the
GNOO scheme, is employed.
In this chapter, we shall introduce first in Section 5.2 some typical static problems
using a non-associative Mohr-Coulomb material model. Particular attention is paid
to the effect of the plastic dilatancy effect.
Since fully saturated behaviour is but a special case in our formulation, Sections 5.35.6 are devoted to various problems concerning consolidation under partially saturated conditions. First, the analysis of an isothermal drainage of water from a vertical
column of sand is compared with experimental results in Section 5.3. It is followed by
the modelling of subsidence due to pumping from a phreatic aquifer in Section 5.4. The
effect of air flow is considered in Section 5.5 which is the numerical modelling of air
storage in an aquifer. Then we return to the theme of bearing capacity in Section 5.6 by
the modelling of a flexible footing resting on a partially saturated soil.
Lastly, in Section 5.7, a comparison of consolidation and dynamic results for
various fully saturated and partially saturated problems using both small strain
and finite deformation formulation, is given. The comparison begins with the consolidation of a fully saturated soil column in Section 5.7.1. It is followed by the
comparison between the consolidation of a fully saturated and a partially saturated
soil column in Section 5.7.2. A two-dimensional consolidation example is introduced
in Section 5.7.3 for a soil layer consolidating under fully saturated and partially
saturated conditions, respectively. The comparison of the small strain and finite
deformation formulation under earthquake loading is given for a one-dimensional
soil column example in Section 5.7.4. Lastly in Section 5.7.5, the elasto-plastic largestrain behaviour of an initially saturated vertical slope subjected to gravitational load
and horizontal earthquake acceleration, followed by a consolidation phase, is given.
In the next section, we shall begin with some typical static problems.
STATIC PROBLEMS
179
where a', refers to the normal tensile effective stress. The superscript' implies that
these quantities are effective stress parameters. For the dry problem, the effective
stress will be the same as the total stress n. In this section, as effective parameters are
always used, there is no confusion if the superscript' is dropped.
Embankment
The first example is that of an undrained soil in an embankment in which the angle is
steeper than that of the internal friction and in which cohesion has been added. The
computation was carried out by Zienkiewicz et a1 (1975) from which Figure 5.1 is
taken. In this problem the collapse is achieved by progressively reducing the
cohesion. The value of the cohesion at the collapse point is identified as that obtained
by slip circle analysis (despite the coarse mesh employed). Further, the difference in
the collapse load between the assumptions of associative and non-associative
behaviour is very small.
Footing
This example, also taken from the same reference as the previous one, shows the
collapse of a footing exerting a uniform load on the soil stratum. Here no exact closed
180
-8.0
10
Cohesion : k ~ / m '
20 25 30
15
I
Cohesion : k ~ / m '
.2-13.0
K
b . 0
- 0
m
*
.8 - 5 . 0 -
II
-8
-6.0 - s ~ oconverged
t
Material propenles
E=Zx 1 0 J k ~ l r n L
v=0.25
@=2O
y-20 k ~ l r n '
STATIC PROBLEMS
181
Figure 5.1 Embankment deformation flow patterns and maximum effective shear strain
contours (gravity constant; progressive decrease of cohesion)
(a) Associative 0 = 4 = 20" (b) Non-associative 0 = 0 4 = 20". Reproduced from Zienkiewicz
(1975) by permission of the Institution of Civil Engineers
form solution exists, but for collapse, a mechanism suggested by Prandtl (1920) and
Terzaghi (1943) gives loads which are compared with the numerical ones in Figure
5.2. Once again very close results were obtained for the limit loads by both associated
and non-associated plasticity models.
The two previous examples suggest that numerical analysis adds little to the solution
of the problem from the point of view of practical geomechanics. Indeed, the main
addition appears to be only the increased cost of analysis and little additional information has been gained. This, however, is just not true as we shall illustrate in the
following sections with examples in which some degree of constraint exists.
182
Material properties
Plane strain
32 quadratic elements
20
Terzaghi
\wNot
Prandtl
Nan-associated flow'
rule ( q j
0")
!on\
I
Mohr-Coulomb material
:C
I0 lblih c$
.'\
JP"(~P'~
= 20"
Critical State 4
model
I
I
Figure 5.2 (a) Strip load on a foundation of a weightless c-4 material; ideal plasticity with
associated and non-associated (non-dilatant) flow rules and strain-hardening plasticity. Mesh
and load-deformation behaviour
STATIC PROBLEMS
183
(ii)
Figure 5.2(b) Relative plastic velocities at collapse (drained behaviour) (i) Associated
Mohr-Coulomb (dilatant) (ii) Non-associated Mohr-Coulomb (zero dilatancy) (iii) Strain
dependent critical state model. Reproduced from Zienkiewicz (1974) by permission of John
Wiley & Sons Limited
184
;,,
O r i g i n : & r Granular
l
fill material ) = 35
Granular material
0
Organic silt
10
20
) = 33
;25 lbln)
) = 20
= 125 1wrt2
y=llolb/f?
20
40
60
80
100
120
140
Method
Safety factor
Reference
Whitman-Bailey
Bishop
Fellenius
Sarma
Morgenstern-Price
Bell
Associated Mohr-Coulomb
Non-associated Mohr-Coulomb
1.24-1.26
1.33
1.09
1.542
1.557
1.49
1.165
1.165
Whitman-Bailey (1967)
Whitman-Bailey (1967)
Whitman-Bailey (1967)
Sarma (1973)
Sarma (1973)
Be11 (1968)
Figure 5.3 Layered embankment problem (a) geometry and material properties (b) finite
element mesh (53 quadratic isoparameric elements) (c) relative shear strain rate contours at
collapse (d) table of safety factors using various methods. Reproduced from Zienkiewicz
(1975) by permission of the Institution of Civil Engineers
STATIC PROBLEMS
1 85
;Prescribed load 4
Material properties
C = 10 m / m 2
Displacement v:m x 10
@=45
Figure 5.4 Axisymmetric sample between rough platens. Effect of degree of dilatancy 9 = 0
(zero dilatancy) and 9 = $J = 45" (fully associated flow rule). Reproduced from Zienkiewicz
(1978) by permission of John Wiley & Sons Limited
Figure 5.5 The Mohr-Coulomb trace in the mean stress - deviatoric stress plane. Changes of
total stress from point A. Reproduced from Zienkiewicz (1974) by permission of John Wiley &
Sons Limited
flow rule, the material will develop negative pore pressures during plastic straining
and will gain strength continuously.
Figure 5.6 shows the same footing problem as that in Figure 5.2 but now solved
(on the same mesh) introducing the undrained assumption. It will be observed that
for a non-associative behaviour with no dilatancy in failure, or for a critical state
model where failure occurs with no volume changes, the failure loads are almost
186
Pressure (lblinL)
IV up to = 3500 Iblin
Figure 5.6 Load deformation characteristics (undrained conditions) for plane footing.
Reproduced from Zienkiewicz (1974) by permission of John Wiley & Sons Limited
identical with those of total stress analysis with Tresca assumption. However the
non-associative, expanding material allows no overall failure.
It must be remarked that in the present case the failure loads are governed entirely by
the cohesion existing in the material as otherwise the strength would be simply zero
when gravity were absent giving an entirely different starting point (A) in Figure 5.5
for the material at different response.
5.2.4
In all of the previous examples, we have assumed that the full Mohr-Coulomb
relation is employed which requires the yield (andlor flow) surface giving, in the
absence of friction; the Tresca yield surface rather than the better known and simpler
von Mises one. However, with this type of definition, difficulties arise at corners
whlch need special treatment (see Zienkiewicz and Taylor, 1991) and many simplifications have been suggested in the literature. In Figure 5.7, we show the 7r plane
section of the Mohr-Coulomb surface for 4 = 20" and various approximations of it,
187
STATIC PROBLEMS
1" ,-
s-Zienkiewicz-Pande
ompromise envelope
Extension cone
Mohr-Coulomb
Figure 5.7 The -/r plane section of the Mohr-Coulomb surface with 4 = 20" and various
smooth approximations. Reproduced from Zienkiewicz (1978) by permission of John Wiley &
Sons Limited
.LoEompression cone
Compramise cone
E = 30000 Iblin2
n = 0.3
c = 10 lbIin.2
f = 20"
32 Parabolic elements used
Soil condition - drained
Figure 5.8 Load-deformation curves for ideal associated plasticity for various forms of the
Mohr-Coulomb approximation (solution on the same mesh as in Figure 5.2). Reproduced
from Zienkiewicz (1978) by permission of John Wiley & Sons Limited
188
I S O T H E R M A L DRAINAGE OF W A T E R
189
where p, is given in the unit of Pa and k,, is the relative permeability for the water
phase.
For the purpose of numerical modelling, the column was divided, in turn, into 10 and
20 four, eight and nine-noded isoparametric finite elements of equal size and different
time steps in the time domain with (At = IOs, Is or 0.5s) gave practically the same
results. At the beginning, besides uniform flow conditions i.e. unit vertical gradient of
the potential andp, = 0 at the top surface, static equilibrium was also assumed.
The boundary conditions were as follows:
For the lateral surface, the horizontal displacement and the fluid outflow is zero.
For the top surface, p , =pa,, where p,,, means atmospheric pressure. For the
bottom surface, pa = pa,, , p , = 0 for r > 150s, while waterp, was assumed to change
linearly from the initial value to zero for t < 150s and the base is fixed in both
displacement directions.
Liakopoulos did not measure the mechanical properties of the soil, so the Young
Modulus of the soil was assumed to be E = 1.3MPa, Poisson's ratio v = 0.4 and
Biot's constant cr = I, similar to Gawin er al. (1995), Schrefler and Zhan (1993) and
Schrefler and Simoni (1988).
The calculations were performed for one-phase flow with gas pressure fixed at the
atmospheric pressure in the partly saturated zone, as well as for two-phase flow. For
the latter case, the switching between saturated and unsaturated solution was performed at p, = 2kPa i.e. S = 0.998, which corresponds to bubbling pressure of the
analysed sand (Liakopoulos, 1965), as well as at p, = OPa, in order to analyse the
effect of the 'switching' value on the solution obtained.
The relative permeability of the gas phase was assumed to follow the relationship
proposed by Brooks and Corey (1966):
where S, is the equivalent value of saturation with the additional lower limit of
k,, 2 0.0001.
The resulting profiles of water pressure for the two-phase flow with switching at
pc2kPa shown in Figure 5.9 as solid curves with different symbols at different time
stations and for one-phase flow in dash-dot curves are compared with the experimental results of Liakopoulos (1965) in fine lines. The solution from one-phase flow
showed better agreement and this is because in Schrefler and Simoni (1988), the
Young's modulus value has been manually 'fitted' for the case.
The profiles of vertical displacements, water saturation and capillary pressure for
two-phase flow with switching pressure atp, = 2kPa (solid curves) andp, = OPa (dashdot curves) are compared with the results for one-phase flow (fine curves) in Figure 5.10.
190
Figure 5.9 Comparison of the numerical solutions for water pressure (two-phase flow with
switching at p, = 2kPa indicated by solid curves and one-phase flow solution indicated by
dash-dot curves) with experimental results of Liakopoulos (1965) in fine lines. Reproduced
from Gawin and Schrefler (1996) by permission of MCB University Press Ltd
There are some noticable differences between one and two-phase flow solutions
for vertical displacements as shown in Figure 5.10, although their final values are
similar. The differences between the one and two-phase flow solutions are more
appreciable for saturation of water and capillary pressure as shown in Figures 5.1 1
and 5.12. In the lower zone, where no gas flow occurs, the differences are small, but
higher up there is a qualitative change in the soil behaviour caused by the presence of
gas under pressure (see also Figure 5.13). The solution for two-phase flow with switching atp, = 2kPa is similar to the one-phase flow solution at the bottom of the column.
The one-phase flow solution tends to the two-phase flow solution with switching at
p, = OPa in the upper part of the sand column, while in the middle region it forms some
characteristic constant-value zones corresponding to switching values of capillary
pressure or saturation. In general, the gradients of saturation and capillary pressure
are higher for the two-phase solution in the upper zone where gas flow occurs. This is
qualitatively in accordance with the solution obtained by Schrefler and Zhan (1993).
The profiles of gas pressure for two-phase flow with switching at p, = 2kPa are
given as solid curves in Figure 5.13 while at pc = OPa, as dash-dot curves. The
discernable differences are caused by the different assumptions about gas flow i.e.
no gas flow for zones where degree of saturation is greater than 0.998 in the first case.
There is a small difference in the pressure amplitudes, nevertheless the qualitative
similarity of the gas pressure profiles in the zone, where gas flow occurs, is obvious.
This example shows that the modelling of the transition from fully saturated to
partially saturated condition and vice versa is very sensitive to the procedure adopted.
191
0.0045
5
0
0.2
0.4
0.6
0.8
1
Height (m)
Figure 5.10 Comparison of the two-phase flow solutions with switching at p, = 2kPa (solid
curves) and p, = OPa (dash-dot curves) with one-phase flow solution (fine curves) made for
vertical displacements. Reproduced from Gawin and Schrefler (1996) by permission of MCB
University Press Ltd
Saturation (-)
Figure 5.11 Comparison of two-phase flow solutions with switching at pc = 2kPa (solid
curves) and at p, = OPa (dash-dot curves) with one-phase flow solution (fine curves) made for
degree of saturation of water. Reproduced from Gawin and Schrefler (1996) by permission of
MCB University Press Ltd
192
0
0.2
Height (m)
0.4
0.6
0.8
Figure 5.12 Comparison of the two-phase flow solution switching at p, = 2kPa (solid curves)
and p, = 0 Pa (dash-dot curves) with one-phase flow solution (fine curves) made for capillary
pressure). Reproduced from Gawin and Schrefler (1996) by permission of MCB University
Press Ltd
Figure 5.13 Comparison of the gas pressure profiles (the two-phase flow solution) for
switching at p, = 2kPa (solid curves) and at p, = 0 Pa (dash-dot curves). Reproduced from
Gawin and Schrefler (1996) by permission of MCB University Press Ltd
193
MODELLING OF SUBSIDENCE
5.4
10
20
30
40
50
60
Distance (m)
70
80
90
100
Figure 5.14 Resulting profiles of saturation (full curves) compared with solution of Safai and
Pinder (1979) in broken curves. Reproduced from Gawin and Schrefler (1995) by permission of
John Wiley & Sons Limited
194
Distance (m)
Distance (m)
Figure 5.15 Resulting saturation and vertical displacement profiles (full curves) compared
with solution of Meroi (1993) in broken curves. Reproduced from Gawin and Schrefler (1995)
by permission of John Wiley & Sons Ltd
tion and relative permeability of water followed those proposed by Safai and
Pinder (1979). The relative permeability of gas proposed by Brooks and
Corey (1966) as given in (5.3) and (5.4) were again applied in the partially saturated
zone.
For numerical purposes, the aquifer was simulated by 50 eight-noded isoparametric
elements (five in height and ten in radial direction) using the same mesh as in
Meroi (1993). A three-by-three integration scheme was applied. Temporal discretization was performed with an initial step size of 1 min for the first 10 hours, 10 min for the
next 20 hours and then 1 hour for the rest of the 28 days-the total required time of
analysis.
195
The resulting profiles of water saturated on the top surface of the aquifer are
compared in Figure 5.14 with the results of Safai and Pinder (1979) in broken
curves, showing relatively good agreement. It has to be mentioned that the model of
Safai and Pinder (1979) neglects the fluid accumulation due to changes in the
degree of saturation, which explains at least part of the differences in the early
stage of the pumping phase. The profiles of water saturation on the upper surface
and vertical displacements for time stations of 10 min, 3h, 10h, 30h and 28 days are
compared in Figure 5.15 with the results of Meroi (1993) in broken curves. In
Meroi (1993), only one phase flow is considered with the gas pressure fvted at
atmospheric pressure in the partially saturated zone, which could explain the
difference in time transident behaviours. The final values at the centre matched
very well.
5.5
With the same geometry as in the column drainage test example of Section 5.4, air
storage in an aquifer is simulated by the partially saturated model (Schrefler and
Zhan (1993)). Meiri and Karadi (1982) simulated this problem by a one-dimensional
finite element model, with a rigid soil skeleton. In this simulation the skeleton is
considered deformable, with Young's modulus E = 692kN/m2. The porous medium
system was assumed initially fully saturated with a reference permeability of
5 x lop"m2, a porosity of 0.2 and initial aquifer pressure of 5.066 M N / or
~ 50
~
Patm, under isothermal conditions at 149C (or 300F). Initial conditions are no
displacement and a constant water pressure of 5.066 M N / at
~ ~
all points. The
boundary conditions are as follows:
lateral surface, q, = 0, q, = 0, uh = 0
bottom surface, q, = O,p, = p,,f, u h 7 u, = 0
top surface, q, = 2.44 x 10p4kg/s/m- (air injection), q,
= 0.
The water and air viscosities are selected as 0.3mNs/m2 and 2 4 p ~ s / m 2respect,
ively. The water density for standard conditions is 100kg/m3 and the air density,
1.22kg/m3.
The following relationships between the relative permeabilities of water and air,
the capillary pressure and water saturation, proposed by Brooks and Corey (1966),
are employed in the simulation:
2lli
k,,
ST
196
---Hrs0010
20
-t30
---*-40
--m--
50
60
--+-.70
-+-80
-0.2
4.1
0.0
0. I
0.2
49.8
49.9
50.0
50.1
50.2
Saturation of water
(Dotted lines. results from Meiri & Karadi)
Figure 5.16 Profiles of vertical displacement, water pressure, air pressure and saturation of
air storage modelling in an aquifer. Reproduced with permission from Schrefler and Zhan
(1 993) American Geophysical Union
197
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Saturation
Figure 5.17 Profile of saturation of air storage modelling in an aquifer with double Young's
modulus for Figure 5.16. Reproduced with permission from Schrefler and Zhan (1993)
American Geophysical Union
(c
where S, = (Sw- Swc)/(l- Swc)is the effective saturation, Swcis the irreducible
saturation, X is the pore size distribution index and p b is the bubbling pressure. The
values for Swc,
X and p b are given as 0.2, 3.0, and 1 . 6 8 k ~ / m respectively,
~,
corresponding to sand with an intrinsic permeability of 5 x 10-I3m2.
The vertical displacement, water pressure, air pressure, and saturation versus
depth of the column are shown in Figure 5.16 for different time instances. Water
and air pressure have a minimum at the interface of the two fluids. The column was
first contracting and then finally expaning vertically with continuing air inflow from
the top. The saturations are compared with the result reported in Meiri and Karadi
(1982), indicated with dashed lines. The effects of a deformable skeleton assumption
can be seen clearly. In general, at the onset of air flow the desaturation is greater in
the rigid skeleton model while later on, the situation is reversed, i.e., the deformable
skeleton model has greater desaturations.
In order to point out the influence of the solid skeleton deformation, a second
simulation was performed with a double Young's modulus. The new saturation
profiles are shown in Figure 5.17. While the propfile of saturation after 80 hours of
pumping is roughly the same with both Young's moduli, the profiles of shorter time
spans are different. These simulations confirmed that the model proposed can
reproduce this quasi-static case and that the main features of the model, i.e., air
flow and solid deformation, have their importance.
198
5.6
In this example (Schrefler and Zhan (1993)), a flexible footing resting on a partially
saturated soil is modelled. The discretized cross-section (19 x 12 m) is shown in
Figure 5.18. The boundary conditions are as follow: for t < 0, lateral surface, q, = 0,
q, = 0, u h = 0; bottom surface, p, = 0, (pa = p w =p,,r = 101.325 kPa), i.e., S, = 1
from S , = S,(p,) and u h = u, = 0; capillary pressure increasing versus top surface
such that on the top surface S, = 0.75; for t 2 0, the top surface outside the footing,
S, = 1 and the top surface under the footing f,,,, = 405.3 kPa (4 p,,, acting downward): all other conditions are as for t < 0.
The wetting process is here simulated for a time span of 10's. This problem was first
solved by Lloret et 01. (1987), using two separate computer programs, one for the twophase flow field and one for deformation analysis. From a computational point of
view, this represents some sort of matrix partitioning and was applied to the original
fully coupled system of equations. Since no iterations were performed between the two
programs during a time step, it is not clear how far the coupling was preserved through
this procedure. Furthermore, the approach used in Lloret rt LEI.(1987) is based on state
surfaces while here the capillary pressure relationship by Brooks and Corey (1966) is
used, together with a modified effective-stress principle and the fully coupled system of
equations is solved as such. Hence the results of the two approaches are not expected to
be the same. The deformation after 100000s is shown in Figure 5.19.
The contour lines of the water degree of saturation at the same time is shown in
Figure 5.20 and water pressure and gas pressure contour lines, respectively, in
Figures 5.21 and 5.22.
Finally, the change in the vertical displacement with time of three surface points A,
B and C of Figure 5.18 is shown in Figure 5.23 and, for comparison, the results for
fully saturated conditions are also indicated.
Figure 5.18 Discretized cross-section for flexible footing resting on a partially saturated soil.
Reproduced with permission from Schrefler and Zhan (1993) ci,"~American Geophysical Union
199
First the soil is compacting, with much larger vertical displacements below the
footing, and with increasing time the cross-section swells. The constitutive model
used in Schrefler and Zhan (1993) required some refinement in order to simulate
collapse upon first wetting and rewetting of the soil, in the presence of a mean
compressive total stress, as is shown have happened a t point A.
200
Figure 5.21 Water pressure (atmospheres) at t = 100000s. Reproduced with permission from
Schrefler and Zhan (1993) American Geophysical Union
/c~
Figure 5.22 Air pressure (atmospheres) at t = 100000s. Reproduced with permission from
Schrefler and Zhan (1993) 0American Geophysical Union
201
I o4
I o5
I o6
108
I o7
Time (seconds)
Figure 5.23 Change in vertical displacement with time at three different points as indicated in
Figure 5.18. Reproduced with permission from Schrefler and Zhan (1993) 0American
Geophysical Union
5.7
202
5.7.1
This example is concerned with the consolidation of a one- dimensional ten metre
deep ground, fully saturated by water, infinitely extended in the horizontal direction,
and subjected to a step load applied at the top level, with drainage allowed only
through the top surface. The problem, which is one-dimensional since each vertical
section can be considered as a plane of symmetry, is modelled as a saturated soil
column under plane strain condition. The boundary conditions for the displacement
field are such that: all nodes are horizontally constrained and the bottom level is fixed
with no vertical movement. Atmospheric pressure is assumed at the top level and
impermeable boundaries are imposed at the lateral and bottom surfaces. For comparison with Prevost's (1981, 1982 and 1984) results of one-dimensional elastic
consolidation, different load levels were considered, reaching a maximum intensity
for the uniformly distributed load equal to the Young's modulus of the ground, E.
An initial porosity of 0.3, a specific permeability of 0.01 mls, an elastic modulus of
lGPa and a zero Poisson ratio are adopted, in accordance with Prevost (1982). A
backward difference integrator was used because of its superior efficiency in consolidation analyses reported in Prevost (1982).
With the finite-deformation formulation, the theoretical relationship between the
applied load and the final displacement is logarithmic for this example and it is
represented in Figure 5.24 by the dot-dash curve. The computational results (curve a)
are in very good agreement with the theoretical behaviour (curve c). The dashed
curve (curve d) represents the results of linear analysis while the dotted curve (curve
e) the solution obtained by Prevost (1982).
At high load levels, due to the assumption of constant elastic modulus, a near-zero
void ratio may occur computationally, and at this stage the soil behaviour should be
described by the elastic characteristics of the compacted grain itself: a relationship
between the elastic modulus of soil and its void ratio has been given in Monte and
Krizek (1976). It is assumed here, for the sake of simplicity, that the soil elastic
modulus becomes ten times as large as the initial one when the porosity approaches
zero. The results obtained under this assumption are represented by curve b.
203
Figure 5.24 Vertical settlement versus load level in one-dimensional elastic consolidation
problem with fully saturated conditions: curves a and b, computational results with the finite
deformation approach, for constant and variable elastic behaviour. respectively; curve c.
theoretical solution for the finite deformation regime, curve d: linear analysis response; curve e.
Prevost solution. Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
In Figure 5.25, the maximum vertical displacement versus the normalized time T,
( T , = c,t/h2 = c,t, where c, is the coefficient of consolidation and c, the time factor
(Lewis and Schrefler, 1987)) for different load levels: curves a, b and c refer to load
levels equal to 0.2,0.4 and 0.6 times the elastic modulus of the soil matrix, respectively.
A further improvement to the finite deformation model consists of the introduction of the dependence of the absolute permeability on the void ratio. In the present
analysis, permeability is assumed to be a linear function of the void ratio, varying
from the initial value to zero when porosity becomes zero. Figure 5.26, in which time
is given in the logarithmic scale, shows the influence on computational results of such
a relationship. Consolidation with constant and variable permeability is described,
respectively, by curves a and c for the load level equal to 0.3E and by curves b and d
for a load level of 0.5E. The significant increase in time (in logarithmic scale on the
figure) necessary to reach full consolidation can clearly be seen.
5.7.2
This example refers to the work by Advani et ul. (1993) and Kim et at. (1993). but
partial saturation conditions were also considered. A one-dimensional column of 7
metres high was modelled in accordance with Kim et al. (1993) using plane strain
204
Figure 5.25 Vertical settlement versus normalized time in the one-dimensional consolidation
problem with constant permeability and elastic modulus under fully saturated conditions and
finite-deformation assumptions. Curves a, b and c refer to load levels equal to 0.2, 0.4 and
0.6E. Reproduced from Meroi (1995) by permission of John Wiley & Sons Ltd
Figure 5.26 Vertical settlement vs. normalized time in the one-dimensional elastic
consolidation problem with fully saturated conditions. Curves a and c refer to a load level of
0.3E while curves b and d to 0.5E with c and d having variable permeability. Reproduced from
Meroi (1995) by permission of John Wiley & Sons Ltd
205
with the ones presented in Kim et al. (1993).The large deformation analysis was also
performed for the case of permeability linearly dependent on the void ratio (curve c).
In the same figure, the large deformation results for three different homogeneous
partially saturated initial conditions were also plotted. These partially saturated
initial conditions were imposed by assigning the initial capillary pressure distribution
corresponding to the degree of saturation via the capillary pressure relationship in
Figure 5.28. While the small deformation analysis with air pressure equal to the
atmospheric pressure gives the final results in one time step (curve d), in the large
deformation analysis a real consolidation process has taken place.
Figure 5.27 For the model drawn, vertical settlement, normalized with respect to the
corresponding values of Terzaghi's theory, vs. normalized time: fully (curve a, smalldeformation analysis; curves b and c, large-deformation analyses with constant and variable
permeability, respectively) and partially saturated conditions (curves d for small deformation
analysis). Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
Figure 5.28 Saturation and relative permeability vs. hydraulic head. Reproduced from Meroi
(1995) by permission of John Wiley & Sons Limited
206
Figure 5.29 Time transient water pressure. normalized with respect to applied load. for the
fully saturated case at the top. Curve a, small-deformation approach; curves b and c, largedeformation cases with constant and variable permeability, respectively. Reproduced from
Meroi (1995) by permission of John Wiley & Sons Limited
The dissipation of water pressure for the fully saturated case in the top element is
shown in Figure 5.29 for small deformation as curve a, and for large deformation
with constant and variable permeability as curves b and c
5.7.3
This example is also taken from Kim et ul. (1993) and consists of a two-dimensional,
plane strain analysis of an 8m deep soil layer loaded by a uniform pressure of 6GPa
by a 16m wide foundation (=2b). Because of symmetry, only half of the model is
considered, which extends for 48m in the horizontal direction.
All the material data are the same as in the previous section, apart from the
Poisson ratio which is assumed to be zero in accordance with Kim et al. (1993), the
resulting time factor is c, = 0.375 x
The boundary conditions for displacements are taken as: the bottom level is fixed, while laterally only vertical displacements are allowed. Both rigid and flexible footings are considered. In order to model
the rigid footing, a multi-point constraint technique, i.e. tied-nodes technique, is
employed.
Pressure is assumed to be zero at the top surface including the underside of the
foundation in accordance with Kim et al. (1993). Since a net flow is allowed through
the foundation, it is assumed to be permeable. Besides the fully saturated case, a
partially saturated model is also investigated with an initial saturation of 92%).The
initial condition is obtained by imposing a suction in accordance with the relationship given in Figure 5.28. In the case of a rigid foundation under partially saturated
conditions, the foundation is now considered as impervious.
207
w, = Ew/2bq are shown in Figure 5.30 for the rigid foundation and in Figure 5.31
for the flexible one. In both figures, curves a refer to small-deformation analysis
introduced in Chapters 2 and 3, for the fully saturated case, and in perfect agreement
with results proposed by Kim et 01. (1993) and with the analytical solution of Booker
(1974).
Curves c refer to large-deformation analysis. In the case of a rigid foundation, the
result is also given by Kim et 01. (1993) and is in good agreement with the one
reported here. Consolidation with the linear dependence of the permeability with the
void ratio is indicated by curve d.
Curves e describe the behaviour in the case of an initially uniform partial saturation of 92% and large deformations. For permeable flexible foundation, the behaviour is almost time independent, while for a rigid impermeable foundation, some
consolidation takes place and the time for transient behaviour is longer than the fully
saturated case, because the foundation impermeability forced the water not only to
follow a longer path, but also one with a smaller relative permeability.
Figure 5.32 represents the deformed mesh in the case of a flexible foundation.
giving consolidation patterns at T, equal to 0.01, 0.1 and 0.55. For the same
dimensionless times, Figures 5.33 and 5.34 give the results for a rigid foundation
with full and partial saturation, respectively. The comparison of the different
deformed shapes allows one to appreciate the influence of the different fluid pressures. In particular, a swelling close to the foundation during the first stage of the
analysis can be observed under fully saturated conditions.
Figure 5.30 Model description and normalized settlement of the top node at the centre line
versus normalized time for the rigid footing. Curves a and b, small-deformation regime for
fully and partially saturated initial corlditions, respectively; curves c and d, finite deformation
analysis from initial fully saturated conditions with constant and variable permeability,
respectively; curve e, finite deformation result from initial partially saturated conditions.
Reproduced from Meroi (1995) by permission of John Wiley & Sons Ltd
208
Figure 5.31 Normalized settlement of the top node at the centre line for the flexible footing
case versus normalized time. Curves a and b, small-deformation analysis for fully and partially
saturated initial conditions; curves c and d, finite deformation analysis from initial fully
saturated conditions with constant and variable permeability, respectively; curve e, finite
deformation results from initial partially saturated conditions. Reproduced from Meroi (1995)
by permission of John Wiley & Sons Ltd
Figure 5.32 Deformed mesh for flexible footing, the fully saturated case: the consolidation
pattern is given at dimensionless time, T, equal to 0.01, 0.1 and 0.55. Reproduced from Meroi
(1995) by permission of John Wiley & Sons Ltd
209
Figure 5.33 Deformed mesh for rigid footing, the fully saturated case: the consolidation
pattern is given at dimensionless time, T, equal to 0.01, 0.1 and 0.55. Reproduced from Meroi
(1995) by permission of John Wiley & Sons Limited
Figure 5.34 Deformed mesh for rigid footing, initial partial saturation of 92%: the
consolidation pattern is given at dimensionless time, T, equal to 0.01, 0.1 and 0.55.
Reproduced from Meroi (1995) by permission of John Wiley & Sons Limited
5.7.4
This example is used to test the code's capability of modelling large deformation with
plasticity. This example of liquefaction performed by Zienkiewicz et al. (1990b) and
Xie (1990) was used for this testing purpose and the Pastor-Zienkiewicz (1986) model
as described in Chapter 4 is adopted for the sand. It can be seen that the results were
not significantly affected by the use of geometric non-linearity because the fluid
pressure can rise even without the large deformation during a cycle of earthquake
210
loading. The first 10 seconds of the N-S component of the El-Centro 1940 earthquake is taken as the horizontal base acceleration input during the consolidating
phase of the sandy soil under a uniformly applied load of 600kPa. The geometry and
mechanical characteristics of the model are the same as those given by Zienkiewicz et
01. (1990). and, in particular, a n initial elastic modulus of 4.5 MPa is assumed.
Figure 5.35 shows the increase in pore pressure, both for large deformation (upper of
the twin curves) and for the small- deformation approach during the first 15 seconds for
Figure 5.35 Pore pressure versus time in the generation phase at three reference points of the
drawn model, both for large-deformation (upper one of the twin curves) and for the smalldeformation approach. Reproduced from Meroi (1995) by permission of John Wiley & Sons
Limited
Figure 5.36 Horizontal displacements versus time at points A and D for both largedeformation (upper of the twin curves) and for the small-deformation approach. Reproduced
from Meroi (1995) by permission of John Wiley & Sons Limited
211
5.7.5
t (s)
Figure 5.37 Pore pressure versus time in the dissipation phase for the given points. both for
the large (the one with the smallest final value of the twin curves) and for the smalldeformation approach. Reproduced from Meroi (1995) by permission of John Wiley & Sons
Limited
212
Figure 5.38 Vertical displacement versus time in the consolidation phase at the given points,
both for the large (the one with smallest final value of the twin curves) and for the smalldeformation approach. Reproduced from Meroi (1995) by permission of John Wiley & Sons
Limited
PRESSURE
TIME: 2570.0
Figure 5.39 Final pressure distribution over the deformed mesh. Reproduced from Meroi
(1995) by permission of John Wiley & Sons Limited
At the left and right vertical sides of the domain, horizontal displacements are fixed
and a hydrostatic pressure distribution with atmospheric value at the two corresponding top nodes is assigned. The material characteristics of soil-saturation relationships
included, are assumed with reference to the Pastor-Zienkiewicz (1986) model of the
clay core of the San Fernando dam, in accordanced with Zienkiewicz rt al. (1990).
214
10
100
15
20
25
1000
time (s)
time (s)
Figure 5.41 Vertical (v) and horizontal (h) displacements versus time during the seismic and
consolidation phases of the top point of the slope. Reproduced from Meroi (1995) by
permission of John Wiley & Sons Limited
215
CONCLUSIONS
5.8 CONCLUSIONS
The formulation developed in Chapter 2 and discretized in Chapter 3 is used to
analyse various one-dimensional and two-dimensional problems with fully saturated
and partially saturated soil respectively. The results obtained from static and consolidation analysis are highly satisfactory and compared well with available analytical and experimental solutions. However, it would be useful for practical purposes
for the formulation to be validated using model experiments and this will be introduced in the next chapter.
REFERENCES
Advani S. H., Lee T. S., Lee J. H . W. and Kim C. S. (1993) Hygrothermomechanical
evaluation of porous media under finite deformation. Part I-Finite element formulations.
hlt. J. N u r ~ Meth.
.
Eng., 36, 147-160.
Bathe K . J. and Ozdemir H. (1976) Elastic-plastic large deformation static and dynamic
analysis, Coinp. Struct., 6, 81-92.
Bathe K. J., Ramm E. and Wilson E. L. (1975) Finite element formulations for large-deformation dynamic analysis, Int. J. Nzm~.Meth. Eng.. 9. 353-386.
Booker J. R. (1974) The consolidation of a finite layer subject to surface loading, Int. J. Solids
& Structures, 10, 1053-1065.
Brooks R. N. and Corey A. T. (1966) Properties of porous media affecting fluid flow. ASCE
IR, 92. No. IR2, 61-68.
Carter J. P., Booker J. R. and Small J. C. (1979) The analysis of finite elasto-plastic consolidation, Int. J. Num. Anal. Geomech., 3, 107-129.
Chen W. F. and Mizuno E. (1990) Nonlinecrr Ancrlysis in Soil Mechcinics-Tlzeory and Implement~ztion, Elsevier, Amsterdam.
Drucker D. C. and Prager W., (1952) Soil Mechanics and Plastic Analysis or Limit Design.
Quart. Appl. Math. 10, 157-165.
Gawin D. and Schrefler B. A. (1996) Thermo-hydro-mechanical analysis of partially saturated
porous materials, Eng. Cornput., 13, No. 7, 113-143.
Gawin D., Baggio P. and Schrefler B. A. (1995) Coupled heat, water and gas flow in deformable porous media, Int. J. Nurn. Meth. Fluids, 20, 969-987.
Heyliger P. R. and Reddy J. N. (1988) On a mixed finite element model for large deformation
analysis of elastic solids, Int. J. Non-linenr Mrch. 23, 131-145.
Kim C. S., Lee T. S., Advani S. H. and Lee J. H. W. (1993) Hygrothermomechanical
evaluation of porous media under finite defromation: part I1 model validations and
field simulations, Int. J. Nurn. Meth. Eng., 36, 161-179.
Lewis R. W. and Schrefler B. A. (1987) The Finite Elenzent Method in the Drfbrmation crnd
Consolidution of Porous Mrdiu, John Wiley & Sons, Chichester.
Liakopoulos A. C. (1965) Transient flow through unsaturated porous media. D. Eng. Dissertation, University of California, Berkeley, USA.
-
216
CONSOLIDATION A N D P A R T I A L L Y S A T U R A T E D D Y N A M I C P R O B L E M S
Lloret A., Gens A., Batlle F. and Alonso E. E. (1987) Flow and deformation analysis of
partially saturated soils in Grouncln~rterejfkcts in Geotrc~liniculEngineering. A. A. Balkema,
Rotterdam, Netherlands.
Meijer K. L. (1984) Comparison of finite and infinitesimal strain consolidation by numerical
experiments, Int. J. Nurn. A n d . Geoniecll., 8, 53 1-548.
Meiri D. and Karadi G . M. (1982) Sin~ulationof air storage aquifer by finite element model,
Int. J. NZUII.
A n d . G e m ~ ~ r h6,. .339-351.
Meroi E. (1993) Comportamento non lineare per geometria di mezzi porosi parzialmente
saturi. Ph.D. Thesis, Istituto di Scienza e Tecnica delle Costruzioni, Universiti di Padova.
Meroi E. A , , Schrefler B. A. and Zienkiewicz 0 . C. (1995) Large strain static and dynamic
semisaturated soil behaviour, bit. J. Nuni. Anrrl. Geotnecli., 19, 81-106.
Monte J. L. and Krizek R. J. (1976) One-dimensional mathematical model for large-strain
consolidation. GPotechnique. 26, 495-5 10.
Narasimhan T. N. and Witherspoon P. A. (1978) Numerical model for saturated-unsaturated
flow in deformable porous media 3. Applications, Wuter Rrsources Rex., 14, 1017-1034.
Pastor M. and Zienkiewicz 0 . C. (1986) A generalised plasticity hierarchical model for sand
under monotonic and cyclic loading, NUMOG II, Ghent, April. 13 1-1 50.
Prandtl L. (1921) ~ b e die
r Eindringungsfestigkeit plastisher Baustoffe und die Festigkeit von
Schneiden. Zeitschr~ft,fCr Ange~c~rmclte
M(rthernntik und Mechnrlik, 1, No. 1 , 15-20.
Prevost J. H. (1981) Consolidation of anelastic porous media, A S C E E M , Vol. 107, 169-186.
Prevost J. H. (1982) Nonlinear transient phenomena in saturated porous media, Conzp. Merh.
Appl. Mech. Eng., 30, 3-18.
of'EnginrrrPrevost J. H. (1984) Non-linear transient phenomena in soil media, in Mec~l~crnics
ing Materirr1.s. Wiley, Chichester, Chapter 26.
Safai N. M. and Pinder G. F. (1979) Vertical and horizontal land deformation in a desaturating porous medium, A h . Wtrter Resourc,es., 2, 19-25.
Schrefler B. A. and Simoni L. (1988) A Unified approach to the analysis of saturated unsaturated elastoplastic porous media in Nurnericcrl Met1iod.s in Geon~echcinicr,Innsbruck,
Balkema, Rotterdam.
Schrefler B. A. and Zhan X. (1993) A fully coupled model for waterflow and airflow in
deformable porous media, Wuter Resources R e x , 29, No. 1, 155-167.
Schrefler B. A,, Zhan X. Y. and Simoni L. (1995) A coupled model for water flow, airflow and
heat flow in deformable porous media, Int. J. Nurw Metl~.Heut Fluid Flolc., 5, 53 1-547.
Shantaram D., Owen D. R. J. and Zienkiewicz 0 . C. (1976) Dynamic transient behaviour of
two-and three-dimensional structures including plasticity, large deformation effects and
fluid interaction, Errrtl~qunkeEngitic~ering& Structurcil Djworvics. 4, 561-578.
Soil Mecl~tmics,John Wiley and Sons, Inc.. New York.
Terzaghi K. (1943) Tl~c~oreticcil
Xie Y. M. (1990) Finite element solution and adaptive analysis for static and dynamic
problems of saturated-unsaturated porous media, Ph.D. Dissertation, Univ. Coll. of Swansea. Wales.
Zienkiewicz 0. C. and Pande G . N. (1977) Time-dependent multi- laminate model of rocks a
.
Geornech., 1,
numerical study of deformation and failure of rock masses, Int. J. N u t ~ Anrrl.
219-247.
Zienkiewicz 0 . C. and Taylor R. L. (1991) The Fitiite E/(wer~tMcthorl-Volutne 2 Solid rrnti
Fluid M e c I ~ ~ n i cDynuniic~
.~,
~ n Nun-line(rritj'
d
(4th erln), McGrew-Hill Book Con~pany,
London.
Zienkiewicz 0 . C., Humpheson C. and Lewis R. W. (1975) Associated and non-associated
viscoplasticity and plasticity in soil mechanics, GPotechniyue, 25, 671-689.
Zienkiewicz 0. C., Xie Y. M., Schrefler B. A,, Ledesma A. and Bicanic N. (1990) Static and
Dynamic behaviour of soils: a rational approach to quantitative solutions, Part 11: Semisaturated problems, Proc. Roj*.Soc. Lond., A429, 310-323.
-
Validation of Prediction by
Centrifuge
6.1 INTRODUCTION
In the previous chapter (Chapter 5 ) , we presented several examples of the application of the full formulation for various static and consolidation problems. This
effectively tested the limit behaviour of various constitutive models and also the
interaction of slow drainage with deformation during the consolidation process. The
problem did not, however, stretch predictive capacity: in the first case of limit
behaviour it gave answers which were quite well known in general, and in the second
case, the departure from linear, elastic, behaviour during consolidation was small.
To test fully the possibilities offered by the formulation and the models presented we
should seek examples where: (1) the repeated loading generates substantial pore
pressures and possibly liquefaction; (2) the problem is such that non-linear, nearfailure, stresses are present and at least partial inelastic permanent deformation
results.
Clearly, the study of earthquake response presents the greatest challenge here but it
is hardly possible to measure it on a site where both conditions 1 and 2 above will
occur within a reasonable time span, or indeed ever.
Further, it would be almost impossible to achieve the so called 'Class A' prediction.
so beloved by soil mechanicians, in which the computation precedes the actual event.
Even if an earthquake of the desired magnitude with the desired effect did happen, its
precise detail of input would not be available before it happened.
What is often possible is to reconstruct catatrophic events, particularly if some idea
of the input motion is available through measurements within reasonable proximity.
In the next chapter, we shall show such reconstructions known frequently as backanalysis. These in the soil mechanics problems are, of course, Class C predictions and
therefore mistrusted by some as of course the soil parameters could be adjusted to
achieve the already known measured results.
For this reason it is desirable to attempt scale model tests of earthquake events for
which both Class A and Class B predictions are possible (The last one being computed
218
simultaneously with the measurement on the model). Two possible scale models of
environmental conditions exist: (a) the shaking table; and (b) the centrifuge.
The shaking table has been used with great success in modelling the dynamic effect
of structures but, unfortunately, it is less successful for soil mechanics problems. The
reason here is that for typical soil problems gravity is the most important external
force and this is obviously not modelled correctly in a scaled model in which densities
are kept constant and the linear din~ensionsreduced. For this reason, the centrifuge
was invented and this device permits a very considerable increase in gravitational
acceleration. The scale model in the centrifuge is usually rather small and thus the
whole frame of the test has to be rotated at high speed, producing a fairly uniform
field of up to 100g in the model area. For this reason, we shall draw our comparisons
in the following sections entirely from the centrifuge and here, as we mentioned, both
Class A and B type predictions will be possible.
In Section 6.2, the basic theory of centrifuge modelling in geotechnical applications
is introduced with particular attention to the use of an alternative fluid in order to
achieve dynamic compatibility of diffusion and inertial behaviour. It is precisely this
substitution of pore pressure which rendered centrifuge testing unsuitable for the
modelling of prototype dynamic events under partially saturated conditions and
here lies one of the limitations of the procedure. For the interest of the reader,
we recommend a study of various publications of Professor Schofield and others
concerning the physics of centrifuge modelling.
In the first section describing the centrifuge test, we shall concentrate on a model of
a dyke (Venter 1985, 1987) performed on the Cambridge geotechnical centrifuge.
Here comparisons of computations are done simultaneously with models and perhaps this section should be classified under Class B verifications. In this section and
indeed in the later section (6.5) where we describe a somewhat similar test done on an
embankment wall at Colorado by Dewoolkar, the same remarks apply. In both cases,
we did not use scaling as the centrifuge itself and the artificial earthquake itself, were
considered to be the 'prototype'. Section 6.4 represents results of a very major study
undertaken in USA under the name of VELACS (VERification of Liquefaction
Analysis by Centrifuge Studies - Arulanandan and Scott, 1993). This study was
funded by the National Science Foundation, USA and involved some twenty
laboratories in various parts of the world performing numerical predictions which
later were to be compared to several centrifuge studies in the USA and Cambridge
Geotechnical centrifuge, UK. In this example, all the predictions are of Class A
type as results had to be presented to the organisers before the centrifuge tests were
attempted. Only in one or two cases were the results obtained later and these
are specially marked as Class B. During this study, very many alternative situations
were investigated and it will be seen later that excellent comparisons were
obtained.
It is necessary, however, to remark that to date in no case has it been possible to
perform centrifuge models with a free fluid surface and such structures as dams,
retaining embankments with different levels of water at different sides, etc., cannot be
modelled because of the restriction on the partially saturated conditions. For this
particular case, the only results which are available will be presented by back-analysis
in the following chapter.
219
where the superscript M denotes the model scale and P denotes the prototype it
intends to model. We can write the mixture equilibrium equation (see (2.11) neglecting only convective terms) for model and prototype respectively.
If we assume that the density of the mixture and fluid, together with the stress statet
are maintained the same in both the model and the prototype, this would require the
acceleration to be scaled by 1/N times with:
and
Comparing with the linear dimension in (6.1), one would conclude that time is also
scaled by N times:
t The maintanence of the same stress level is important for soil behaviour as the stress: strain behaviour
of soils is highly stress level and strain-history dependent.
220
Therefore, in the model, a dynamic event will happen N times as fast as its corresponding prototype, thus pushing up the frequency of the dynamic events by N
times. However, for most practical earthquake events, the frequency and wavelengths
of the wave within the pores should still be within the range of laminar Darcy's flow.
The scaling relationship is then applied for the fluid mass conservation equation
(see (2.16))
which implies that pore pressure is the same in the model and the prototype if the
compressibilities and the void ratio remain the same. Also for any source term its rate
must also be the same. If any substitute fluid is used, its compressibility should not be
too different from that of water, though the bulk modulus of water does vary because
of the amount of air dissolved in it.
So far, there is no problem in the scaling, however, a problem arises when the fluid
flow equation (see (2.13) and (2.14) again neglecting only the convective terms) is
considered:
and
As the velocity is required to be the same in the model and the prototype if the scaling
of the displacement, acceleration and time is as described in Equations (6.1, 6.4, 6.5
and 6.6) we require the permeability to be scaled by I/N:
This cannot be readily achieved if the same solid material, pore fluid and porosity are
used. One of the solutions to this problem is therefore to use a different fluid. For
instance, in the Cambridge test of the dyke, silicon oil with the same density as water
is used. The viscosity is chosen to be N centi-stokes or cs (Dow Corning Limited,
1985) because water has a viscosity of 1 cs. This reduces the permeability by N times
22 1
and the above relation is retained. However, there is now a possibility that the bulk
modulus of the fluid and its damping characteristics are different from those of
water. In the VELACS exercise, in order to avoid problems in the interpretation of
the centrifuge tests, water was chosen as the pore fluid and consolidation was therefore found to be applied more rapidly than the corresponding prototype. Other
substitute fluids used include Metolose (Dewoolkar, 1996).
However, for semi-saturated material, the use of another liquid will in all probability lead to a different value for surface tension as well as different drying and
wetting characteristics. Although the use of the same fluid in experimental modelling
of pollutant transport in a semi-saturated environment has been reported (Cooke.
1991 and 1993; Cooke and Mitchell, 1991a and b) and the scaling of the capillary was
reported to be modelled correctly (Hellawell, 1994; Culligan-Hensley and Savvidou,
1995), it would be difficult at this stage to extend such tests to dynamic events.
6.3
The test represents a fully dynamic analysis with transient behaviour. The physical
model is a centrifuge experiment performed by Venter (1985) on the Cambridge
Geotechnical Centrifuge. The principle of the centrifuge has been briefly explained
in Section 6.2. Simply, the centrifuge reproduces similar stresses and strain history as
experienced by the prototype on the scaled model. If the behaviour of soil is controlled
mainly by the stress state and its strain history then the centrifuge model is able to
predict the generic behaviour of the prototype under earthquake conditions.
The layout of the experiment is given in Figure 6.1. The model is built in a strong box
with a dyke lying on a sand bed. An oil reservoir is created behind the dyke to provide
seepage conditions. Silicon oil is used so that the diffusion equation and the dynamic
equation can have the same timescale under the centrifuge condition. This is done by
using silicon oil (Dow Corning Limited, 1985) of viscosity of 80 centistokes (80 times
the viscosity of the water). Also shown in Figure 6.1 are the measurement devices,
which include 11 PPTs - pore pressure transducers, 1 LVDT - Linear Voltage Displacement Transducer, 7 ACCs accelerometers and 3 TSTs-total stress transducers.
With the triaxial test results (Venter, 1986) on the Leighton-Buzzard sand used in
the centrifuge experiment, soil parameters for the Pastor - Zienkiewicz Mark-111
model are identified. They are listed in Table 6.1. The Finite Element idealization and
the boundary conditions are given in Figure 6.2. Also shown in the figure are the
positions of 10 pore pressure transducers, 4 accelerometers and the LVDT presented
for displacement comparison purposes. The comparison is done for as many points
as possible so that an overall picture, together with appropriate mechanisms can be
obtained. The test was done at 78g and the material data for the Finite Element
analysis are given in Table 6.2. The first study is performed with 4-4 element, i.e. 4noded for soil displacement (u) and 4-noded for pore pressure (p).The computer code
used was SWANDYNE-I (Chan, 1988) using SSpj time stepping scheme. For each
pair of graphs presented in this section, the left-hand side one is a measured value
from experiment, while the right hand side is the computed value.
-
222
Table 6.1
Test
14
Predicted
Adjusted
Po(kPa)
Kevo
KevoIP,
Keso
KesoIP,
Mg
Mf
1986)
223
117
I Sand nuxturc
2. Concrete dyke
3. Retaming wall
4. 0 1 1 overflow
5 Oil sea
6. Coarse sand
drains
R i g d boundry
@
PPT2626
x
Figure 6.1
PPT2628
x(69)
Acceleratmn measurement
PPTZR4h
~(90)
- -
Water level
k = 2 . I x lo-
k = 2 . 1 x 10/
/ n=nuh
Impervious
u=v=O
ii
Shaking at bottom
0 Measured
Location
Computed location
224
0.00
0.02
0.04
0.06
0.08
Seconds
0.10
0.12
0.14
0.08
0 10
Seconds
Measured
0.12
0.14
Input motion
(ii)
0.00
(iii)
0.0002
0.0000
-.
0.02
0.04
DOF: 2
0.06
OF NODE: 26
-g -0.0004 -E
-0.0006 .--0.0002
-5 -0.0008 -
-0.OOlO
-0.0012
-0.0014
0.0b
0.02
0.04
0.06
0.08
0.10
0.12
Computed
Vertical d~splacementof the dyke
0.14
Figure 6.3(a) Comparison with centrifuge results; (4-4, Gamdm=2) (tip) input motion
(iii) vertical displacement of the dyke
(ii) &
225
The results are given in Figures 6.3(a)-6.3(h). The input motion was taken from
the accelerometer ACC1244 attached to the box. Also shown in Fig. 6.3(a) is the
vertical displacement at the crest of the dyke. The comparison is excellent. Besides
predicting correctly the final displacement, the rising time and the shape of the rise
are also predicted. The reader is reminded that this was the first set of soil parameters
obtained directly from the soil model tester, no parametric study has been performed
i.e. a prediction is stated as one of Class B-1.
The pore pressure transducer A gives good agreement too as shown in Figure
6.3(b) although the value is slightly lower. More oscillation is seen in B; however. the
predicted trend is still correct. The agreement of C is remarkable although the size of
oscillation is larger, nevertheless, the result is very good. The other graph in Figure
6.3(c) gives the comparison of pore pressure transducer D. The mean value is almost
the same as the measured value, although oscillation is more pronounced. E predicts
lower pore pressure rise and F is quite satisfactory, so does G. H, I and J are slightly
worse but the overall prediction on the pore pressure rise is very good.
This oscillation found in the excess pore pressure could be due to the proximity of
the rigid boundary condition either at the bottom of the container or near the
underside of the dyke. Let us consider the transducers in two groups:
(1) Far-field where the influence of the structure is less at locations A, C, E, F, G and H.
DOb I
OFNODE 198
OF NODE: 14
40000
40000 30000
30000 6 20000 h\~~;~~+~+d@+ ~0000
10000
g loo00 0
5
0-1 0000
;-loo00 -20000
Q 20000 -30000
-30000 I
0
0.00 0.02' 0.04 0.06 0.08 0.1'0 0.i2 0 . K
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
Seconds
(ii)
Mrusr~red
Cornputrd
Figure 6.3(b) Comparison with centrifuge results: (4-4, Gamdm=2) (top) excess pore
pressure at point A (bottom) excess pore pressure at point B
226
70000
700001 De\ I
Xi1000
Doll l
of node: 98
Doll l
of node 138
70000)
C type:
~
Figure 6.3(c) Comparison with centrifuge results: (top) excess pore pressure at point C;
(bottom) excess pore pressure at point D
(2) Near-structure where the influence of the structure is more pronounced at locations B, D, I and J.
It can be observed that comparatively less oscillations are found at locations which
are far-field. Even if there is substantial oscillation such as location H, the dominant
frequency is more akin to the input frequency (approximately 120Hz). However, for
the near-structure locations, the oscillations are much more pronounced and the
frequency is more akin to the frequency of the structure and of a higher frequency
nature (approximately 240Hz).
One possible reason for this oscillation is the proximity of the impermeable
solid boundary and the fact the average relative fluid acceleration has been neglected
in the u-p formulation. As the average relative fluid acceleration is neglected, any
volume change near an impermeable boundary will behave in an undrained
manner and lead to a large rate of change in pore water pressure via the fluid continuity
equation. Similar behaviour has been found in the proximity of a retaining wall in
(Dewoolkar, 1996). When the same analysis is repeated using the fully explicit u-w
formulation (Chan et al, 1995),significantly less oscillation is observed in, for instance,
point I-PPT2628 (Figure 6.4) when compared with Figure 6.3(f).
Initially, all the accelerometers show good agreement with experimental results. As
the soil weakens, the value on the higher level (L and M) departs from the experimental value. The predictions of N and 0 are reasonable. This may be due to the
10000
227
100000
Dof I
of node 140
of node: 196
0 00
0 02
0 04
0.06
0 08
0.10
0 12
0.14
Sccondc
.hlemreed
(11,
Figure 6.3(d) Comparison with centrifuge results: (top) excess pore pressure at point E;
(bottom) excess pore pressure at point F
0 !I0
0 O?
0 01
0 00
(ii)
0 ox
0 10
Seconds
0 1?
0 11
Aleom~d
Figure 6.3(e) Comparison with centrifuge results: (top) excess pore pressure at point G;
(bottom) excess pore pressure at point H
228
40000
10000
of node- 23
30000
$ 20000
loo00
0
-
000
of node: 17
2626
-: loo00 0
Dof I
40000 -1
Oil!
001
000
.ifcnnemd
(ii)
008
010
Sccmds
01:
10000
011
O(i0
002
001
000
r y y
I
008
'
010
I
012
011
i-ornpu~d
Figure 6.3(f) Comparison with centrifuge results: (top) excess pore pressure at point I;
(bottom) excess pore pressure at point J
0'00
0 02
0.04
0.06
0 08
0.10
0 I?
300
Dof I
of node. I39
Duf I
o f node: 142
0.14
SECONDS
(i)
wed
-'y
300
zoo
3004
002
(ii)
004
0.06
0.08
0.10
012
014
SECONDS
.&!LUSW~
Figure 6.3(g) Comparison with centrifuge results: (top) acceleration at point L; (bottom)
acceleration at point M
229
p I00
; 0
;-100
5
zoo
-300
0.00
(102
ii)
(104
006 008
SCCONDS
0.10
012
300
O.oO
0.14
0.02
0.04
0 Oh
0 OX
0 10
O 12
0.11
(b,,mrrrd
Wwwwd
Dc\lcu
tvne
Dof I
of node. 103
Figure 6.3(h) Comparison with centrifuge results: (top) acceleration at point N; (bottom)
acceleration at point 0
Excess pl-essure (kPe)
-20.0
0.00
0.04
0.08
Time (seconds)
0.12
0.16
40.0
T m e (seconds)
ib)
Figure 6.4 Numerical results of excess pore pressure at point I (PPT2628) using fully explicit
GLADYS-2E with 3-3 element: (a) with 44 elements; (b) with 96 elements
230
amount of shear wave energy being transmitted from the bottom to the top. As the
soil weakens, less shear wave should be transmitted. However, due to the oscillatory
nature of the pore water pressure, the mean effective stress is not reduced as much as
in the physical experiment. Therefore the shear modulus is not reduced adequately
leading to the excessive transfer of shear wave energy.
Despite a number of shortcomings, the results represent an excellent comparison
with the experimental results accounting for possible experimental errors. Nevertheless, these results represent a Class B-1 prediction and a set of more convincing
Class A predictions are given in the next and subsequent sections.
6.4
T H E VELACS PROJECT
Although many verification exercises have been performed by the authors (Chan,
1988) and (Zienkiewicz et NI., 1990) and others using reported centrifuge results-a
more systematic study became recently possible through the project VELACS
(VErification of Liquefaction Analysis by Centrifuge Studies-Arulanandan
and
Scott, 1993) funded by the National Science Foundation, USA. A numerical prediction of several postulated tests was requested from 'predicting participants' before the
experiments are performed and results obtained for 'centrifuge experiment participants'. The numerical predictions were kept in sealed condition by a
third party (Thompson and Lambe, 1994) and these were not made available to the
'centrifuge experiment participants'. This double blind policy was introduced to
minimize possible 'cheating' and thus enhance the credibility of the results. Such
'Class A' predictions were submitted by twenty 'predicting participants' by
30th September, 1992 (Table 6.3) when apparently the centrifuge tests were first
commenced by seven universities (University of California, Davis; California
Institute of Technology; Cambridge University; University of Colorado, Boulder;
Massachusetts Institute of Technology; Princeton University; and Rensselaer
Polytechnic Institute). However, some of the specified centrifuge tests could not be
carried out-and additional computations ('Class B') mostly because of the difference
in the prescribed and actual input earthquake motion were requested-without
however supplying other experimental results. It is instructive for the readers to
note that, except for the simpliest model No. 1 which represents a level soil layer, all
numerical predictors used computer codes based directly or indirectly on the Biot
theory and approximation form introduced in this book (Smith, 1994).
Nine centrifugal models (see Figure 6.5 taken from Arulanandan and Scott (1993))
were selected for the verification:
Model No.1-Horizontally
Model No.2-Sloping
Model No.3-A
0
Model No.4a-Stratified
Model No.4b-Stratified
231
I Model No.
7
Laminar box
Laminar box
E n d Dr = 40%
Sand: Dr = 40%
I(
~4
2"
1 Model No. .
~ o d eNo.
l 4
Laminar box
II
Laminar box
1 Sand: Dr = 60%]
1 Model No. t
~ o d e No.
l 4
Rigid Box
1 Sand: Dr
= 60%
R i g ~ dBox
;,
3r
I
1 Model No.
'
Rigid Box
~ o d eNo.
l 1
Rigid Box
7.6m
silt
1 ~ o d eNo.
l I
Project
232
m
Q x
<
X PI
x x x
x x x x
x x x
--
233
Model No.6-A
Model No.7-A
Model No.11-A
Model No.12-A
Most of the tests have been performed at more than one centrifuge for centrifuge
validation purpose and considerable scatter of results can be found between the
centrifuge results. Three of the authors led three different groups of predictions
using computer codes SWANDYNE-4 (implicit u-p with partial saturation), SWANDYNE-I1 (implicit u-p with full saturation) and MuDIAN (implicit u-U with fullsaturation, to be introduced in Chapter 7) based on their work in University College
of Swansea.
The SWANSEA group led by the senior author of this book, Zienkiewicz, together
with one of the authors, Pastor, presented eight predictions for model Nos. 1,2,3.4a.
4b, 6, 7 and 11 (Zienkiewicz et d . , 1993a-h). The second author, Chan, presented
seven predictions for model Nos. 1 , 2, 3, 4a, 4b, 7 and 11 (Chan et ul., 1993a-g) and
another of the authors, Shiomi, together with the senior author, presented four
predictions for model Nos 1, 2, 11 and 12 (Shiomi et (I/., 1993a-d). Lastly, Chan
(Chan rt ul., 1994) and Shiomi (Sture ct (I/.,1994) were also involved in the process of
overviewing numerical predictions.
Most of the predictive results can be classified as good or excellent. A selection
of these will be presented in the subsequent sections in this chapter with a brief
description of the experimental set-up and conlparisons with numerical results on the
same scale. But before going into the detail of the prediction, the following sections
are devoted to the analysing procedure using SWANDYNE-11.
234
235
(xv) Other post-processing e.g. excess pore water pressure ratio, spectral analysis
and response spectrum are calculated for reporting purposes.
6.4.2
236
mean effective stress p' or pore pressure versus axial strain plot in an
undrained test. Its value can be adjusted so that a better match of the curve
of pore pressure versus axial strain can be obtained. In the VELACS exercise,
this was done so that the end point in the predicted curves stayed close to the
experimental data.
(vi) KesOc (dimensions of stress): represents the value of three times the shear
modulus at the mean effective stress p'O. It can be obtained by matching the
initial slope of deviatoric stress q versus axial strain plot in an undrained
test. Its value can be adjusted so that a better match of the curve of q
versus axial strain can be obtained. In the VELACS exercise, this was done
so that the end point in the predicted curves stayed close to the experimental
data.
(vii) $o (dimensionless): is usually taken as 4.2 and this value is taken here.
(viii) fiI (dimensionless): is usually taken as 0.2 and this value is taken here.
(ix) Ho (dimensionless): is determined by fitting the curves in p' or q versus the
axial strain plot. It can be found by matching the shape of the q versus p' plot
for undrained tests also.
(x) Huo (dimensionless): is determined by matching the initial slope of the first
unloading curve.
(xi)
- y (dimensionless):
~ ~
is determined by matching the rate of change of the
slope of the first unloading curve or by matching the number of cycles in
a series of loading and unloading. The second method is used in this exercise.
(xii)
(xiii) pb (dimensions of stress): is the initial mean effective stress of the undrained
triaxial test.
6.4.4
237
There are quite a number of parameters in the Pastor-Zienkiewicz mark 111 model
which require definition. Three CIUC (Isotropically Consolidated followed by
Undrained Compression test) experimental results starting from 40kPa were chosen
to identify the parameters. The experimental results were taken from (Arulmoli et al.,
1992) which provided the standard soil model test results for the numerical predictors. The 40kPa ones were chosen because they are close to the mean effective stress
value at the middle of the centrifuge model. The permeability of silt is also calculated
for this level of mean effective stress. The set used in the Class A prediction for (Chan
et a1 1993a - is given below:
(9 Loose sand: (Dr = 40%) Experiment CIUC4051 was used. The parameters
obtained are as follows: Mg = 1.15 Mf = 1.03 af = cug = 0.45 KevOc = 770kPa
KesOc = 1155kPa, the elastic modulus is proportional to the mean effective stress,
Bo = 4.2 PI = 0.2 p:, = 4kPa HO = 600 HUO= 4000 kPa Y H =
~ 2yDM= 0.
(ii) Dense sand: (Dr = 60%) Experiment CIUC6012 was used. The parameters
obtained are as follows: Mg = 1.32, Mf = 1.30, a f = a g = 0.45.
KevOc = 2000kPa, KesOc = 2600kPa, the elastic modulus is proportional
to the mean effective stress Po = 4.2, P1 = 0.2 pk = 4kPa, Ho = 750,
Huo = 40000kPa, YH,, = 2, ?DM = 4
(iii) Silt: Experiment CIUCBS13 was used. The parameters obtained are as follows:
Mg = 1.15, Mf = 0.50, af = ag= 0.45, KevOc = 400kPa, KesOc = 1520kPa,
the elastic modulus is proportional to the mean effective stress, Po = 4.2.
pl = 0.2 pk = 4kPa, Ho = 900, Huo = 100000kPa, TH, = 2,yDM= 8.
6.5
238
Figure 6.7(a). Although the model contained sand deposited at both 40% and 70%
relative densities, laboratory data were available only for 40% and 60% relative
1-
LVOTI
LVOTS
Concentrated masses
(aluminum alloy rectangular rings)
F.E mesh
RPI Test- 2
(Sccond)
Time
ID1
0.6
0.6
0.4
0.4
0.2
5
k 0.0
E
-
00
2 -0.2
3 -0.2
-0 4
-0.4
0.2
-0 6
(Second)
Tune
101
-0.6
0
12 14 16 18 20
0
2
4
6
(Second)
T~me
Expcrirnental and predicted horizontal ~ c e l c r a t i o nA H 3
10
(Second)
Time
10 12
(Second)
Time
(Second)
Time
14
16
18
20
239
0.4
.sc2 0.2
+ 0.0
.g
g -0.2
0.2
+E
0.0
-0.4
-0.2
-0.4
-0.6
0
(Second)
Time
8 10 12 14 16 18 20
(Second)
Time Expennental and Predicted horizontal acceleration AH5
(KNM:)
(KNM')
2 1003
Z 80'2
a 60E
40203 0
0
2 100
$ 802
60-
RPI-Test 2
P5
k
g
r/]
Prediction
P5
4020-
<
0 10 20 30 40 50 60 70 80 90 '00
l b 20 30 40 50 60 70 80 90 100
(Second)
(Second)
Ti]%
Ti'"e Experimentel and predicted excess pore-pressure
(KNM~I
(KNM')
2 1100
2 100
2 80 Prediction
RPI-Test 2
802
60 60 2
P6
P6
a 40 40
m
k
g
20
(KNM')
2 100
80 I
2
60 -
i
L
0 10 20 30 40 50 60 70 80 90 100
10 20 30 40 50 60 70 80 90 100
(Second)
(Second)
Time
Time
Experimental and predicted excess pore-pressure P6
(KNM')
Prediction
RPI-Test 2
P7
m
g
3
20
0
X
I
W
;
I
I
I
I
l
I
1
I
0 10 20 30 40 50 60 70 80 90 I00
(Second)
Time
20
O
10 20 30 40 50 60 70 80 90 100
(Second)
Time
240
RPI-Tcst 1
Prediction
-p
-0.1
C;
-0.2
1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 0
I
l O P 3 0 4 1 3 W W 7 C I ~ W l 0 0
(second)
(second)
Time
Experiment$ and predicted vertid displaccmmt LVDTl
Figure 6.6 VELACS Centrifuge Model No. 1: instrumentation, finite element mesh and
comparison of experimental and predicted results
241
(a) Although the predicted horizontal acceleration on the surface of the soil layer is
much higher than the experimental results, it is obvious that the experimental results
of the surface acceleration is not necessarily reliable after soil liquefaction. We also
have recognized that the numerical result of the surface acceleration obtained by
using mixed 8-4 noded elements, probably gives a much higher peak value of
acceleration. This can be improved by using 4 4 noded elements. There is no large
difference in the acceleration at the middle points between prediction and experiment. The prediction also showed that the peak acceleration on the surface decreased
F. E, mesh
2
(KNIM')
(KNIM
(second)
TlME
(second)
TIME
50
100
150
200
250
200
250
"h
(second)
Tme
(KNIM )
Prediction
M 20
Y
0
0
50
100
I50
(second)
Tme
Figure 6.7 VELACS Centrifuge Model No. 3: instrumentation, finite element mesh and
comparison of experimental and predicted results
242
-&- - f "
!-lTI
CIT-Test
k20
i:
3
0
0
50
100
I50
200
250
(second)
Time
(KNlM )
0
0
50
100
I50
200
250
(second)
Time
Expenmental and predicted excess pore-pressure PB (loose sand)
50
100
150
200
I50
200
250
(second)
Ttme
"
(second)
T~me
I00
50
I50
200
250
(sccond)
Tme
Ex~nmentaland predtctcd cxcess pare-prcssurc P9 (dense sand)
250
(second)
Time
(m)
I00
":TI
9
SO
IKNIM )
ti 20
(KN M
(second)
Time
lKN,h4-)
g
3
50
100
150
200
250
(second)
Time
(m)
(second)
Time
Prediction
4.3
1
44
0
50
100
150
200
250
(second)
Time
Experimental and predicted vertical displacement 1 5 (loose sand)
50
100
I50
200
250
(second)
Time
Exper~mentaland pred~ctedvert~caldisplacement L6 (dense sand)
Figure 6.7 (cont.) VELACS Centrifuge Model No. 3: instrumentation, finite element mesh
and comparison of experimental and predicted results
243
(b) The predicted time histories of pore-pressure agreed closely with those recorded
in the centrifuge tests. In model No. 1, the experimental results showed slightly faster
pressure dissipation; however, the differences were not large. The experimental
10
-Initial
boundary
- - . - - Uoundaty after sarthquate t s s ~
(m)
0.5
localion
0.4
0.3
2 0.2
LVDT 1
Prsdiction
0.1
0.0
0
10
15
20
25
30
35
40
Figure 6.8 VELACS Centrifuge Model No. 11: instrumentation, finite element mesh and
comparison of experimental and predicted results
244
results of model No. 3 by RPI and CIT gave obviously different peak values of porepressure and made comparison very difficult. The predicted post-earthquake consolidation was consistent with the values measured in the centrifuge test.
(c) The prediction of displacement showed similarities to those measured during
centrifuge experiment. The horizontal displacement of the retaining wall in model
No. 11 was well predicted. The predicted values of surface settlement in model No. 1
were lower than those recorded in the centrifuge test, but the difference was not too
marked. In model No. 3, although the prediction of surface settlement in the loose
sand agreed closely with that measured in the centrifuge tests, the numerical analysis
0.75"
AC4
ELEVATION
Figure 6.9 (a)Schematic model configuration of test M M D l (Dewoolkar 1996); (b) Finite
element mesh with slip element, in heavy line (Dewoolkar, 1996)
245
CENTRIFUGE T E S T OF A RETAINING W A L L
did not predict more settlement in dense sand than in the loose sand which was
recorded in all experiments for this model, despite a later repeat of the experiments
which showed that different results could also be obtained (Dobry, 1996). Almost all
numerical predictors failed to achieve this qualitative difference. The mechanism
involves a failed dense sand wedge moving into the liquefied loose sand due to the
lack of support. It is of interest to carry out further research into this phenomenon.
More research is also needed on the modelling and calibration of data for the
behaviour of dense and and silt under cyclic loading.
-0.1
0.1
0.2
0.3
0.4
Deflection (inch)
0.5
0.6
(a)
-.
. Numerical
-10
0
10
20
Total lateral earth pressure (psi)
('J)
---- ---
: Theoretical
+ : Experimental
Figure 6.10 (a) Static deflection of the wall (Dewoolkar, 1996); (b) Total lateral earth
pressure on the wall (Dewoolkar, 1996)
246
(a)
MMDI SWANDYNE 11
. . . . . . ., . . . - .; . . . . . . . . . . . . . .:..AC9
.....
'
..
-40
0.1
0.2
0.3
Time (s)
0.4
0.5
Time (s)
MMDI SWANDYNE 11
(b)
........:........ ;..............
1
....
4 0
01
02
03
04
05
..)
. . . . . . . . . . . . . . . . :.. . . . . . .:...... .
0.1
T ~ m e(s)
0"
Ol
0.2
03
T ~ m e(s)
0.3
0.4
0.5
MMDI SWANDYNE I1
;-500
0
0.2
Time (s)
(c)
2 -500
: AC8
...........
04
05
0"
0.1
0.2
0.3
Time (s)
0.4
0.5
Figure 6.11 Comparison with centrifuge test MMDl (Dewoolkar, 1996): (a) Horizontal wall
accelerations at the top; (b) Horizontal soil accelerations; (c) Dynamic bending strains of the
wall
different stages of the centrifuge, all the material parameters used are derived from
the VELACS exercise as given in Section 6.4.4. Only the amount of damping has
been varied to investigate the effect of different levels of damping.
247
REFERENCES
(a)
MMDI (DI)
-0.2
(model scale)
MMDI SWANDYNE I I
1
0
0.1
0.2
0.3
Time (s)
0.4
0.5
T ~ m (s)
e
(a)
MMDl (Dl)
(b)
= 15
MMDI SWANDYNE I I
(model scale)
-a
.A
PP6
2 I0
2
2
a 5
3 0
0
-5
05
15 2 2 5
Tune (s)
35
0.5
1.5 2 2.5
Time (s)
3.5
Figure 6.12 Comparison with centrifuge test MMDI (Dewoolkar, 1996): (a) Dynamic wall
deflections; (b) Long-term excess pore pressure
and the VELACS project. Very good and excellent agreements have been obtained
thus validating the formulation and the computer code for various types of analysis
under saturated condition. In the next chapter, we are going to show numerical
predictions of practical examples and the use of the numerical procedure in design.
REFERENCES
Arulanandan K. and Scott R. F. Eds. (1993) Proceedings of VELACS symposium, A. A.
Balkema, Rotterdam.
Arulanandan K. and Zeng X. ( I 993a) Experimental results of Model No. 1 in Verificution of
Numerical Proceclures,for the Ana1~vsi.sqf Soil Liquefimion Problems, (eds.) Arulanandan K .
and Scott R. F., UC Davis, 17-20 Oct, A. A. Balkema, Rotterdam, 19-24.
Arulanandan K . and Zeng X. (1993a) Experimental results of Model No. 1 in Verification o f
Numerical Procerlures,for the Ana1ysi.r ?fSoil LiqutIfaction Problems, (eds.) Arulanandan K .
and Scott R. F., U C Davis, 17-20 Oct, A. A. Balkema, Rotterdam, 295-300.
ArulmoIi K., Muraleetharan M. M. H. and Fruth L. S. (1992) VELACS Laboratory testing
program - Soil data report, Earth Technology Corporation, project No. 90-0562, Irvine.
California.
Chan A. H. C. (1988) A unified finite element solution to static and dynamic geomechanics
problems, Ph.D. Dissertation, University College of Swansea, Wales.
248
Chan A. H. C., Famiyesin 0 . 0. R. and Muir Wood D . (1992a) Report No. CE-GE92-23-0:
Numerical Simulation Report for the VELACS Project - General Description, Department
of Civil Engineering, Glasgow University, Glasgow.
Chan A. H. C.. Famiyesin 0 . 0 . R. and Muir Wood D. (1992b) Report No. CE-GE92-23-1:
Numerical Simulation Report for the VELACS Project-Class A prediction of RPI model,
Department of Civil Engineering. Glasgow University, Glasgow.
Chan A. H. C., Famiyesin 0 . 0 . and Muir Wood D . (1993a) Numerical Prediction for Model
No. 1 in Verification qf Numericrrl Procetlure.s~forthe Anulysis if Soil Liqurfirction Problems,
(eds.) Arulanandan K. and Scott R. F., UC Davis, 17-20 Oct, 1, A. A. Balkema, Rotterdam. 87-108.
Chan A. H. C.. Famiyesin 0. 0. and Muir Wood D . (1993b) Numerical Prediction for Model
No. 3 in firificution of Nun~ericrilProcec/~rrc~.s,fi,r
the Anulysi.~of Soil Liqu+ction Problems,
(eds.) Arulanandan K. and Scott R. F., U C Davis, 17-20 Oct, 1, A. A. Balkema, Rotterdam. 343-362.
Chan A. H. C., Famiyesin 0 . 0 . and Muir Wood D. ( 1 9 9 3 ~Numerical
)
Prediction for Model
No. 3 in Verifi'cntion ofArun~ericcrlProceduresfor tile Anrrl~vsisoJ'Soil Liquejkction Problems,
(eds.) Arulanandan K. and Scott R. F., U C Davis, 17-20 Oct, 1, A. A. Balkema, Rotterdam. 489-510.
Chan A. H. C., Famiyesin 0 . 0 . and Muir Wood D. (1993d) Numerical Prediction for Model
procedure.^ ,fbr the Ancr1j~si.sof' Soil Liqwfirction ProNo. 4a in Vmficution of' N~in~eric,uI
blenu, (eds.) Arulanandan K . and Scott R. F., U C Davis, 17-20 Oct, 1, A. A. Balkema,
Rotterdam, 623-630.
Chan A. H. C., Famiyesin 0 . 0 . and Muir Wood D. (1993e) Numerical Prediction for Model
No. 4b in Ver~ficutionof' Nurnericnl Proceclures f i ~ rtlw Ancrlysis q f Soil Liquclfirction Problenis. (eds.) Arulanandan K. and Scott R . F., U C Davis, 17-20 Oct, 1, A. A. Balkema,
Rotterdam, 71 1-720.
Chan A. H. C., Famiyesin 0. 0 . and Muir Wood D. (19930 Numerical Prediction for Model
No. 7. in Verification of Nun~eric,ulProcedures ,for the Analj,.r.is @Soil LiquLlfirction Problems, (eds.) Arulanandan K. and Scott R. F., U C Davis, 17-20 Oct, 1, A. A. Balkema,
Rotterdam, 835-850.
Chan A. H. C., Famiyesin 0 . 0 . and Muir Wood D. (19938) Numerical Prediction for
Model No. 11, In Ver~ficntion of Nun~c~riccrlProcedures ,for the Anrrl~~sisof Soil
Liquefnction Prohlenis, (eds.) Arulanandan K . and Scott R. F., U C Davis, 17-20 Oct.. 1,
909-93 1.
Chan A . H. C., Siddharthan R. and Ito K. (1994) Overview of the Numerical Predictions for
VELACS Model No. 3 in Ver~ficcirionof Nutnericul Procedures ,for the unulysis of soil
liquefuction yrobletns, (eds.) Arulanandan K. and Scott R. F., U C Davis, 17-20 Oct, A.
A. Balkema, Rotterdam, 1443-1456.
Chan A. H. C.. Famiyesin 0. 0. and Muir Wood D. (1995) User Manual for GLADYS-2E,
School of Civil Engineering, University of Birmingham, December, Birmingham.
Cooke B. (1991) Centrifuge modelling of flow and contaiminant transport through partially
saturated soils, Ph.D. Dissertation, Queen's University, Kingston, Ontario, Canada.
Cooke B. (1993) Physical modelling of contaminant transport in the unsaturated zone in
Geotecl~niculMunugement of Waste and Contaminution, Balkema, Rotterdam.
Cooke B. and Mitchell R . J. (1991a) Evuluution of' Contcrn~incrntTrunsport in Purtiully Srrtur n r d Soil in Centrifuge '91, Balkema. Rotterdam.
Cooke B. and Mitchell R. J. (1991b) Physical Modelling of dissolved contaiminant transport in
an unsaturated sand, Cunudiun Geotecl~nicalJournal, 28, No. 6, 829-833.
Culligan-Hensley P. J. and Savvidou C. (1995) Environmental geomechanics and transport
process in Geotechniccrl Centrifuge Technology. (ed.) Taylor R. N. Blackie Academic &
Professional, London. Chapter 8, 196263.
REFERENCES
249
250
Stadler A. T.. KO Hon-Y. and Sture S. (1993) Experimental results of Model No. 1 in
Ver~fi'cutionqf Numerical Procedures for the Ancr1~~si.sqf' Soil Liquefuction Problems,
(eds.) Arulanandan K . and Scott R. F., U C Davis, 17-20 Oct, A. A. Balkema, Rotterdam,
25-44.
Steedman R. S. and Zeng X. (1995) Dynamics in Geotechnicirl Cmtr~f'ugeTecl~nology,(ed.)
Taylor R. N., Blackie Academic & Professional, London Chapter 7, 168-195.
Sture S., Law H. K., Shiomi T. and Iai S. (1994) VELACS: Overview of numerical predictions
for model No. 12, in Veriji'cation c?f'NumericcrlProceclure.s,fbr the Analysis qf'Soil Liquefuction Problems, (eds.) Arulanandan K. and Scott R . F., U C Davis, 17-20 Oct., 2, A. A.
Balkema, Rotterdam, 1635-1646.
Taboada V. M. and Dobry R. (1993) Experimental results of Model No. 1 at RPI in
Verification of Nurnericd Procedures,fbr rlze Ancrlysis qf' Soil Liqugfuction Problen~s,(eds.)
Arulanandan K . and Scott R. F., UC Davis. 17-20 Oct, 1, A. A. Balkema, Rotterdam,
3-18.
Thompson P. Y. and Lambe P. C. (1994) Project VELACS management and coordination in
Verif~cationof Numerical Procedures ,fbr the Andysis of Soil Liquefirction Problems, (eds.)
Arulanandan K. and Scott R. F., U C Davis, 17-20 Oct, 2, A. A. Balkema, Rotterdam,
1267-1280.
Towhata I. (1994) Review of prediction 'A' on model I1 in Ver;fi'crrtionqfNun~ericcr1Proceduresfor the Anabsis o f s o i l Liyuefuction Problems (eds.) Arulanandan K. and Scott R . F.,
UC Davis, 17-20 Oct. 1, A. A. Balkema, Rotterdam, 1607-1612.
Venter K. V. (1985) KVV03 data report: Revised data report of a centrifuge model test and
two triaxial tests, Andrew N. Schofield and Associates, Cambridge, England.
Venter K. V. (1986) Triaxial data report: Report on seven triaxial tests, Andrew N. Schofield
and Associates, Cambridge, England.
Venter K. V. (1987) Modelling the response of sand to cyclic loads, Ph.D. Dissertation,
Cambridge University Engineering Department.
Zienkiewicz 0. C., Chan A. H. C., Pastor M., Paul D. K. and Shiomi T. (1990) Static and
Dynamic Behaviour of Geomaterials - A rational approach to quantitative solutions, Part I
- Fully Saturated Problems. Proc. Roy. Soc. Lond.. A429, 285-309.
Zienkiewicz 0 . C., Huang M. and Pastor M. (1993a) Numerical Prediction for Model No. 1, in
Verrfication of Numerical Procedures for the Analysis qf' Soil Liquefuction Problen~s,(eds.)
Arulanandan K . and Scott R. F., U C Davis, 17-20 Oct., 1, 259-276.
Zienkiewicz 0. C., Huang M. and Pastor M. (1993b) Numerical Prediction for Model No. 2, in
Verifi'cation of Numericcrl Procedures for the Anal~vsiso j Soil Liquejuction Problems, (eds.)
Arulanandan K . and Scott R . F., U C Davis, 17-20 Oct., 1,423-434.
Zienkiewicz 0 . C., Huang M. and Pastor M. ( 1 9 9 3 ~Numerical
)
Prediction for Model No. 3, in
Verification of Numerical Procedures,for the Analysis qfSoi1 Liquefirction Problems, (eds.)
Arulanandan K. and Scott R. F., U C Davis, 17-20 Oct., 1, 583-591.
Zienkiewicz 0. C., Huang M. and Pastor M. (1993d) Numerical Prediction for Model No. 4a,
in Veriji'cation ofNumerica1 Procedures,for the Analysis qf'Soil Liyuefuction Problems, (eds.)
Arulanandan K. and Scott R. F., U C Davis, 17-20 Oct., 1, 675-680.
Zienkiewicz 0 . C., Huang M. and Pastor M. (1993e) Numerical Prediction for Model No. 4b.
in Verrfication of Numericrrl Procedures jbr the Anu1,vsis of' Soil Liquejuction Proble~ns,
(eds.), Arulanandan K. and Scott R. F., UC Davis, 17-20 Oct., 1, 731-736.
Zienkiewicz 0 . C., Huang M. and Pastor M. (1993f) Numerical Prediction for Model No. 6, in
Verification of Numerical Procedures,for the Ancrlysis qj'Soil Liqwfuction Problems, (eds.)
Arulanandan K. and Scott R. F., U C Davis, 17-20 Oct., 1, 777-782.
Zienkiewicz 0 . C., Huang M. and Pastor M. (19938) Numerical Prediction for Model No. 7, in
Verif~cationof' Numerical Proceclures,fbr the Anulysis of'Soil Liyuefuction Problems, (eds.)
Arulanandan K. and Scott R. F., U C Davis, 17-20 Oct., 1, 873-880.
REFERENCES
251
Zienkiewicz 0 . C., Huang M. and Pastor M. (1993h) Numerical Prediction for Model No. 11,
in Verification of Numerical Proceduresfor the Analysis of Soil Liquefaction Problems, (eds.)
Arulanandan K . and Scott R. F., UC Davis, 17-20 Oct., 1, 997-1006.
7.1 INTRODUCTION
In this chapter we will introduce case histories of the two-phase liquefaction analysis
(effective stress dynamic analysis) for real engineering problems. Seven examples are
introduced.
Soil layer liquefaction problems I: simulation of recorded data at Kobe Port
Island, to illustrate the effect of multi-directional loading (1995).
Soil layer liquefaction problems 11: simulation of liquefaction behaviour
during Niigata earthquake (1964), to illustrate the effect of initial shear
stress.
Quay wall failure in Kobe City (1995).
Dam failure I: Lower San Fernando Dam, to illustrate the effect of pore water
migration on stability of dams (1971).
Dam failure 11: investigation of liquefaction failure mechanism for an earth dam
at Hokkaido (1993).
Soil structure interaction problem: building tilted due to Niigata earthquake
(1964).
Effect of counter measures using the deep soil mixing method-a
prediction.
design
The most difficult part of liquefaction analysis is the determination of the soil
parameters. It is important that soil parameters are determined from the site investigation data including laboratory test data. For example, the cohesion and friction
angle can be determined directly from the drained tri- axial test data. But parameters
of liquefaction are not so easy for simple constitutive models. For example, dilatancy
254
parameters are dependent on the shape of the potential surface, yield surface,
difference of failure line from phase transformation line, hardening parameters of
shear behaviour, and so on. For this reason, the densification model mentioned in
Chapter 4 is a useful model for design engineers. In this chapter the densification
model is used in most of the examples. As the original densification is not capable of
representing cyclic mobility phenomena, a modification for the original model is
presented in Section 4.5 of Chapter 4.
Examples (1)-(7) are all back analyses of liquefaction induced by various earthquakes. In most of these cases, only records of permanent deformations are observed
and no measurements are available for acceleration or pore water pressures. Example
(1) is not only typical of soil layer problems but is the only case in which accelerations
were recorded at four positions in depth during the Hyogoken-Nanbu Earthquake in
1995. The earthquake heavily damaged Kobe City and over five thousand people
died. The example is also a very rare case in which the input motion is available at
the deep base support layer and this was recorded at the damage site. Thus the
input motion at the site can be estimated from the measurements in the vicinity.
This contrasts with the Niigata Kawagishi-cho problem in example (2), where
the input motion was estimated from the data recorded at a location over 100 km
distant.
Example (3) is a quay-wall problem, typical of liquefaction. The quay wall was
heavily damaged during the same earthquake as in example (1) i.e. the HyogokenNanbu Earthquake of 1995. The foundation behind the quay wall was not damaged
because of the counter measure installed.
Dams are another type of structure for which liquefaction is important. In this
chapter, two dams are analysed and reported in examples (4) a rock-fill dam and (5)
an earth-fill dam. Both show a similar failure mechanism but they are included here
as they showed different failure patterns. The lower San Fernando dam example, (4),
is included to illustrate the effect of pore water migration on the stability of dams,
and the earth dam at Hokkaido is included to show an investigation of the liquefaction failure mechanism using a numerical method.
Example (6) is included for its interest in three-dimensional analysis. The example
is taken from an apartment block heavily damaged during the Niigata Earthquake of
1964.
Example (7) is included as a case of dynamic earth pressure action and shows how
safety can be increased. Here, dynamic earth pressure acted on the embedded
structure when liquefaction occurred and this also presents important problems.
255
256
studied the effect of MDL on a soil layer problem with their proposed numerical
method using the fully coupled Biot's equation with u-U formulation (Zienkiewicz
and Shiomi, 1984) for the dynamic equation and a non-linear material model. The
material model was based on the hyperbolic stress-strain relation for shear (Kondner
and Zelasko, 1963), and the effective stress-path approach for dilatancy (Ishihara et
al.. 1975). In this model, the decrease of effective stress is a function of both
components of the horizontal shear strains. A hypothetical, horizontally layered
ground, subjected to the El Centro Earthquake was solved as a case study. Twodimensional analysis showed a marked difference in the build-up behaviour of pore
water pressure and some differences in the surface velocity spectra. Even when the
amplitude of the acceleration is increased to the resultant peak accelerations from the
two directions, the results were different from the results of the two-dimensional
analysis. As an alternative, Fukutake et al. (1995) suggested the use of input motions
that are 1.3 times larger than the earthquake and of a stronger direction between
north-south (NS) or east-west (EW), although this conclusion is considered
premature.
Therefore there is a need to survey the three-dimensional behaviour of the level
ground subjected to MDL, together with other variations such as soil properties and
input motions. This section investigates the effects of MDL on a real site. The site is a
typical soil 'column' on the Port Island in Kobe City, and earthquake motions for the
studies were recorded during the Hyogoken-Nanbu earthquake on January 17th in
1995, where the liquefaction phenomenon has been observed throughout the island
and settlement was about 20 cm (estimated from the relative gap between buildings
supported by piles and the ground surface after the earthquake). At first, back
analysis of the observed data is explained and then the parameter study for M D L
is reported.
7.2.2 Multi-directional loading observed and its numerical modellingsimulation of liquefaction phenomena observed at Povt Island
During the Hyogoken-Nanbu Earthquake, liquefaction took place along most of the
seaside of Kobe City. Sand boiling and flushing water due to liquefaction occurred in
many places including Port Island where an array of seismometers was set at four
depths (Kansai-Kyogikai 1995). Figure 7.1 shows the orbit of the records at Port
Island for the Hyogoken-Nanbu earthquake in 1995. Two-to-four very large amplitudes can be seen from the figures. The maximum acceleration at the surface (GL
O.Om) was 314gal (3.14msp2)for the NS and 288gal for the EW direction. They are
about half of the value record at GL-83.8m. The diagram at the bottom left of Figure
7.1 shows the orbit at GL-83.8m during zero to five seconds. Several large amplitudes
were clearly seen. The direction, when the maximum amplitude occurred, was about
20 degrees from north to west. This direction is considered as the principal axis of the
earthquake components. The time history of the direction was the North-South
direction (Figure 7.2). The diagram at the bottom right of Figure 7.1 shows the
orbit of the NS (north-south)-UD (up down) motion. At GL-83.8m. the UD component was not considered significant.
257
G.L. 0.0m
-observed
G.L. -83.8m
.--
-800
400
400
EW Ace. (gal)
-observed
800
-800
G.L. -83.81~1
-u
800
G.L. -83.8m
800
--m
"
400
400
9
in
0
400
EW Ace. (gal)
-observed
800
--w
-400
Y
n
0 -
400
1100
-800
-800
Figure 7.1
400
0
400
EW Acc. (gal)
800
-800
-800
400
0
400
NS Ace. (gal)
800
258
--g
-
Time (Sec)
800
.4-
-2
'
. - - - - - - - - - - - - - - - - - - - - - - - - - - -
-800
0
10
15
Time (Sec)
20
25
30
Principal direction
0
.---.-----------------------
-800
10
15
Time (Sec)
20
25
30
Table 7.1
Cases studied
Case
Analysis type
Input motion
1
2
3
4
5
3 D analysis
3 D analysis
2 D analysis
2 D analysis
2 D analysis
NS, EW and U D
NS and EW only
NS and UD only
EW and U D only
Principal axis and U D only
Results of simulation
Figure 7.4 shows the time history of response accelerations in the NS direction overlaying the observed accelerations. These agreed well initially, however, some difference
in the response is found at the surface after 5 seconds. The observed acceleration shows
a very long period wave, while the calculated acceleration has a component of higher
frequencies. Figure 7.5 shows the time history of the excess pore pressure in Cases 1
(NS+EW+UD) and 2 (NS+EW). The layer between GL -12.6m and -14.0m was
fully liquefied at about 6 seconds and kept liquefied. But the built-up speed
259
Soil type
V, (mls)
V, (mls)
Average
N-value
Density
(kNim3)
Man-made fill
Man-made fill
Man-made fill
Man-made fill
Silty c!ay
Layers of
gravelly
sand and silt
Silty clay
Diluvium sand
of the excess pore pressure above the layer was slowed down after 6 seconds. This is
because the input motion transferred from the bottom to the surface was significantly
reduced by the sudden loss of the strength at the layer between GL - 12.6m and - 14.0m.
On the other hand, at a deeper level of GL -15.4m, pore pressure continued to
build up and almost reached full liquefaction. Therefore, contrary to popular belief, a
tendency for liquefaction at a deeper layer was found in this large impact-type
earthquake. In order to investigate the effect of vertical input motion, Case (2)
(NS+EW only) was then analysed. The excess pore pressure response using the updown component (UD) has a very high frequency shown by the thin solid line in
Figure 7.5. Nevertheless, excess pore pressure response without UD motion (bold
260
600
G.L. 0 . 0 m
600
- - - - simulation
-observed
10
Time(s)
20
15
G.L. 1 6 . 8 m
- - - - simulation
-observed
...
10
Time(s)
15
20
G.L. -32.8m
- - - - simulation
10
Time(s)
15
20
G.L. 8 3 . 8 m
....simulation
o b s e r v e d
Figure 7.4
10
Time(s)
15
20
solid line) passes through nearly the mean value of the excess pore pressure obtained
with UD motion. This implied that vertical input motion induced only compressible
wave in the water but did not affect the response of the soil skeleton. Therefore the
result of Case 1 is found to be very similar to Case 2 when the high frequency is
filtered out.
Next the results of unidirectional loading (Figure 7.6 and Figure 7.7) and multidirectional loading (Figure 7.5) were compared. The response obtained with unidirectional EW direction loading showed a different response while the response
obtained with unidirectional NS direction loading was close to the one obtained with
multidirectional loading. This is because the NS direction was very close to the
principal direction of the earthquake motion. The build-up of pore water pressure
could have been dominated by the NS component of the earthquake. In order to find
out the effect of the direction of input motion, response with unidirectional loading
261
EFFECT OF M U L T t - D I M E N S I O N A L LOADING
10
I5
Time (See)
20
25
30
GL-12.6111-14.0m
Time (See)
G.L.-15.4m -1 6.8m
Time (Sec)
Figure 7.5
+ UD)
in the principal direction was then calculated and shown in Figure 7.8. In this case.
the results were very similar to those obtained using the NS component alone.
262
I0
15
Time (Sec)
20
25
G.L.-9.0m I I .Om
30
10
15
Time (Sec)
20
25
G.L.-12.6m 1 4 . 0 m
30
10
15
Time (Sec)
20
30
25
T H E EFFECT OF INITIAL ( S H E A R ) S T R E S S
263
10
15
Time (Sec)
20
25
30
10
15
Time (Sec)
20
25
30
.0
44
1.0
-- -
---------
G.L.-12.6m 1 4 . 0 m
-:
- -
---------
44
10
10
15
Time (Sec)
20
15
Time (Sec)
20
25
30
G.L.-15.4m 1 6 . 8 m
25
30
straightforward, since not only shear strength but also the dilatancy characteristic is
affected by ISS.
A layered soil ground was analysed with two different initial conditions, i.e. 'with
ISS' and 'without ISS' conditions. The earth pressure at rest KOis 1.O in the case of the
condition 'without ISS' and 0.5 for the case 'with ISS'. The mean stresses were kept
the same at the same depth since the liquefaction strength has long been assumed to
be the same for the same mean stress in the soil column problem (Yoshimi, 1991). This
means that the vertical stress changes to maintain the mean stress. This assumption was
made to simulate the ordinary calculation procedure for the liquefaction safety factor.
The analysis of these examples was very significant. The same results were obtained for
most of the constitutive models in which criteria depend on the maximum shear stress
in three-dimensional space. However, there are some analysis codes which do not
detect any significance for this problem. This difference in response is obviously caused
by the different constitutive equation used in different codes.
264
0
1.0
10
15
Timc (Sec)
20
25
G.L.-9.0111-11 .Om
30
10
15
Time (Sec)
20
25
G.L.-12.6m 1 4 . 0 m
30
10
15
Time (Sec)
20
25
G.L.-15.4m -16.8111
30
10
15
Time (Sec)
20
30
25
265
T H E EFFECT OF INITIAL ( S H E A R ) S T R E S S
( N S + EW + UD)
-NS
.--o-.NS i , m.j
EW (NS + EW + UD)
---A-.EW.i, d.j
N S (observed)
-4-
200
400
EW (observed)
600
800
+UD
(NS + EW + UD)
( p r i n c i p a l axis + UD)
UD (observed)
200
400
600
Max. Acc. (gal)
(a)
(b)
ROO
Figure 7.9 Maximum response acceleration (a) horizontal component; (b) vertical
component
266
[I;.,.:.,
--,.
- . - .--
1
2
Max. disp. (cm)
40
80
120
Max. acc. (gal)
....
-KO = 0.5
K O = 1.0
Time (s)
10
267
by Vaid and Finn (1979) and Hyodo et ul. (1988) except for the final pore water
pressure. The liquefaction ratio is 75% for the 'with ISS' and 40%)for the 'without
ISS' at 3.5 seconds.
Figure 7.1 1 may give a clue as to why the liquefaction strength curve gives the same
value as for the same mean stress. The liquefaction strength curves were determined
by the number of cycles taken for the samples to reach final liquefaction. The criteria
for the final liquefaction is either 5% shear strain in deformation or pore pressure
reaching 95% of the initial vertical stress. The pore water pressure ratios reached 1.0
at almost the same time, at about 8 seconds.
Theoretical considerations
ISS can therefore be classified into two types. Type I, ISS is seen in the case where an
external force is applied in a perpendicular direction to ISS. In this case, the incremental shear stress has mainly the effect of rotating the principal stress with the
increment in the equivalent stress being small. For example, the soil beneath a
structure has almost no shear stress in the horizontal direction; however, deviatoric
shear stress (vertical stress minus horizontal stress) is relatively large. The maximum
shear stress acts for the direction of approximately 45 degrees. An external force due
to an earthquake produces a large horizontal shear stress and the principal stress
direction is rotated. Type 11, ISS is the case where an external force is applied to the
direction parallel to the maximum shear direction of ISS. In this case the incremental
equivalent shear stress is equal to the external incremental shear stress. For example.
a slope such as that of a dam has ISS close to the horizontal direction. These two
types of ISS might work differently. There is no substantial evidence, however, that
ISS should affect liquefaction differently. A soil-column-type dynamic effective
analysis can obviously be classified into Type I.
It was found that the existence of ISS creates a slower build-up of pore water
pressure, as indicated by the experiments. In the example problem, the upper soil
layers with ISS were weakened more than in the case without ISS. It should be
noticed that ground layer analyses frequently neglect ISS. Constitutive models based
on a typical elasto-plastic theory, which uses the flow rule, produce different results
for the conditions 'with ISS' and 'without ISS.' The constitutive models developed as
an extension of a one-dimensional shear soil column model often ignored ISS
(deviatoric stress due to the difference of vertical stress and horizontal stress) since
most of the models use the hyperbolic stress-strain model for the shear behaviour and
their stresses always start at zero. That means that ISS is not involved in the
formulation.
PREDICTION APPLXCATIONS A N D B A C K A N A L Y S I S
268
Figure 7.12
Damaged quay wall along the building (after Suzuki ct rrl., (1995))
countermeasure for the foundations. The countermeasure was lattice walls, which are
made of the stiffened ground by mixing cement to reduce the shear movement of the
ground (Suzuki et al, 1995). The lattice-shaped stiffened ground walls (LSSGW) were
built surrounding the piles. At the design stage, prediction analysis for the structural
design stage was conducted to confirm the effects of the LSSGW.
7.4.1
Configuration
Two-dimensional analysis was conducted. The numerical model of the foundation
and building is shown in Figure 7.13. The building was built on a pier shaped ground,
269
34m
a
-10 3m
-13
4m
-
-15
8m
-
-26 1m
-32 7m
(a) Sectmn
plle
.1
.I
.I.
J,
J,
J,
( b ) Pldne
which projected into the sea. The foundation was surrounded by quay walls. Deformation due to the earthquake in the shorter section of the building was anticipated
larger than that in the longer section so that the numerical study was made in the
shorter section. The foundation of the building was made using a repeated pattern of
piles and LSSGW shown in Figure 7.13. LSSGW and ground with piles were
modelled into separate groups. These two groups were connected at a corner of the
lattice. The building was considered as elastic and Mohr-Coulomb criteria was used
for the LSSGW.
270
Porosity
Depth
-(m)
Friction
angle
(deg)
Shear
modulus
(MW
34.00
34.00
59.00
79.00
38.00
141.00
92.00
RIS
2
3
4
5
0.1534
0.1495
0.1574
0.1997
5.0
5.0
5.0
5.0
0.02100
0.02100
0.02000
0.01200
1 .OO
1.OO
1.OO
1 .OO
Table 7.3) were obtained by a site investigation conducted during the design stage.
The liquefaction strength R I 5 is the only data available for the liquefaction parameter. This is largely because safety against liquefaction is normally judged by the
liquefaction strength. R l s or RZ0 is the shear stress ratio against the confining
pressure when liquefaction takes place after 15 or 20 cycles of the undrain triaxial
test. Parameters r, A and B in Table 7.4 are for the densification model and are
determined to satisfy R 1 5 .Figure 7.14 shows an example of a fitted curve for the
liquefaction strength.
log N
271
Time (s)
Figure 7.15
Input motion
Input motions for design were E l Centro of 1940, EW component, Taft of 1952, NS
component and some other records of large earthquakes which occurred in Japan.
An artificial earthquake generated from a design spectrum is also often used. Here an
earthquake predicted at the site from an earthquake recorded at the Kobe Ocean
Weather Station during the Hyogoken-Nanbu Earthquake of 1995, was used.
272
Figure 7.19 shows the maximum shear stresses which occurred in the soil layers
and in the LSSGW itself. These stresses indicated that most of the shear forces acting
in the ground were shared by the LSSGW. Therefore, soil strains of the inner zone of
the LSSGW were 0.19% at most of those which took place at the bottom of the
LSSGW zone. These small strains were due to the stiff LSSGW and could produce
no liquefaction. The strains of the outside of the LSSGW, i.e. back-fill of the quay
273
-6
-8
-I0
-12
-1 4
JA
-
'
?
W
Pore Pressure
-5
5 10 15 20 25
Pore Pressure
5 10 15 20
Pore Pressure
Pore Pressure
<Line 4 > ( x I OkPa)
Figure 7.18 Pore water pressure and initial effective stress (see Figure 7.13 for lines 1 to 4)
Figure 7.19 Shear stress of lattice shaped stiffened ground wall and ground
wall, were very large (more than 10%) since the quay would be able to move freely
towards the sea, so that the liquefaction took place and then more, larger, strains
were induced.
274
7.5
The failure of the lower San Fernando earth dam in 1971 with nearly catastrophic
consequences, is typical of what can occur in a poorly consolidated soil structure
affected by shaking, resulting from an earthquake. Zienkiewicz and Xie (1991)
have reported some results of numerical simulation of the failure. Here, the details
are presented. In particular, the effect of cohesion resulting from negative pore
pressure and the influence of such parameters as permeability and relative density
on the dynamic response of the dam, are illustrated.
Although full comparative measurements are not available, the reconstruction of
the event by Seed et ul. (1975) and Seed (1979) is remarkable in attempting to explain
why the failure occurred apparently some 60-90 seconds after the start of the earthquake, which was recorded to last some 14 seconds. The actual collapsed dam and a
'reconstructed' cross-section are shown in Figure 7.20 following Seed (1979). The
hypothesis made here was that the important pressure build-up occurring as a result
of cyclic loading which manifested itself first in the central portions of the dam,
'migrates' in the post earthquake period to regions closer to the 'heel' of the dam
where it triggers the failure.
Reconstructed cross-section
Figure 7.20 Failure and reconstruction of original conditions of the Lower San Fernando
dam (Seed. 1979)
275
Rigid boundary
Base excitation
)le boundary
Impermeable boundary
Figure 7.21 Idealization of San Fernando dam for analysis: (a) material zones (see Table 7.5):
(b) displacement discretization and boundary conditions; and (c) pore pressure discretization
and boundary conditions
We show in Figure 7.21 the material idealization, finite element meshes and
boundary conditions used in the present computations. First an initial, elastic static
analysis is carried out by the full program considering a semi-saturated condition and
assuming the gravity and external water pressure to be applied without dynamic
effects. Figure 7.22 shows such an initial steady-state solution for the saturation and
the pore pressure distribution, indicating clearly the 'phreatic' line and the suction
pressures developing above.
Figure 7.22 Initial steady-state solution: (a) Pressure (kPa); and (b) Saturation contours
276
Table 7.5
K f (Pa)
1
2a
2b
3
2
2
2
2
2090.0
2020.0
2020.0
2020.0
980.0,
980.0
980.0
980.0
10''
10"
lo1'
lo1'
x 109.2857
x 10'
0.2857
x lo9 0.2857
x 10'
0.2287
Porosity k (mis)
0.375
0.375
0.375
0.375
lop3
lo-'
lo-'
lo-'
Starting with the above computed effective stress and pressure distribution, a full
non-linear dynamic computation is carried out for the period of the earthquake and
continued for a further time of 200 seconds. The material properties assumed to
describe the various zones of the dam using the constitutive model described in
Chapter 4 and also found in Pastor and Zienkiewicz (1986) and Pastor er al. (1988,
1990) are summarized in Table 7.5.
Figure 7.23 shows the displaced form of the dam at various times. The displacements at some characteristic points and the development and decay of excess pore
pressures are shown in Figure 7.24 and Figure 7.25. It is noted that deformations are
increasing for a considerable period after the end of the earthquake. This undoubtedly is aided by the redistribution of pore pressures. Near the upstream surface the
pore pressures continue to rise well after the passage of the earthquake. This indeed
was conjectured by Seed (1979). It is also noted that the pattern of deformation is
very similar to that which occurred in the actual case showing large movements near
the upstream base and indicating the motion along the failure plane.
The suction pressures which developed above the phreatic line give a substantial
cohesion there. Indeed, preliminary computation indicates that, without such
cohesion, an almost immediate local failure develops in the dry material upon shaking.
The amount of cohesion depends on the S,- pw (or the S,-h,)
curve in the
following equation. Since p, is assumed as zero, it is seen that the cohesion is of
the value Swp, in the otherwise cohensionless granular soil since the effective stress is
defined as
When we reduce the parameter b of Van Genuchten's formula by a factor of 100, the
pore pressure distribution of static solution will be almost the same as in Figure
7.22(b) but the saturation in the semi-saturated zones will be close to 1.
277
90s
(iii)
Figure 7.23 Deformed shapes of the dam at various times: ( i ) 15 s (end of earthquake): (ii) 30
s; (iii) 90 s: (iv) 200 s.
For sand'
For San Fernando
I
,?(cm-')
u(cm-I)
c1
0.0689
0.0842
0.01740
0.00700
2.5
2.0
0.0667
0.0500
5.0
4.0
1 .OO
0.90
The higher value of S, results in stronger cohesion in the upper part of the dam. The
dotted lines in Figure 7.24a are the results of computation occurred because of the
now increased cohesion in the upper regions of the dam.
If the permeability of the dam material is sufficiently high, it may be impossible for
an earthquake to cause any build-up of pore pressures in the embankment, since the
278
Figure 7.24 Horizontal (left) and vertical (right) displacements: (a) at the crest (dashed line
represents the result of computation with increased 'cohesion'); (b) at point E; (c) at point H;
(d) at point I (see Figure 7.21a).
L O W E R S A N FERNANDO D A M FAILURE
279
pore pressure can dissipate by drainage as rapidly as the earthquake can generate
them by shaking. Figure 7.26 shows the results which indicate a rapid dissipation of
pore pressures and much reduced permanent deformations.
280
Figure 7.26 Results of analysis with increased permeabilities: (a) deformed shape of the dam
after 15 s; (b) deformed shape of the dam at 200 s; (c) horizontal displacement on the crest; (d)
~~ertical
displacement on the crest; (e) excess pore pressure at point A; excess pore pressure at
point D (see Figure 7.21a)
In an additional analysis the relative density of the dam material is assumed lower,
which implies in the present constitutive model that the ratio Mf/M, is considered
equivalent to the relative density Dr [Pastor et a1 19851. With M, values in Table 7.5
fixed. Mf values are now reduced to 1.24,0.453,0.604 and 0.906 for material zones 1,
Za, 2b and 3, respectively. In this case significantly larger displacements are recorded
at the early stages of the earthquake shaking as shown in Figure 7.27.
28 1
(ii)
10s
(iii)
Figure 7.27 Results of analysis with softer materials, showing deformed shapes at: (i) 5s; (ii)
10s; (iii) 15s; (iv) 200s.
282
The modelling of the constitutive relationship and the initial stress condition plays an
important role in the analysis and is reported here in detail. Also in this study, the
failure pattern of the earth dam is evaluated.
In the dam analysis, the treatment of the initial shear stress condition is a difficult
issue. At first a static analysis was conducted for the self-weight of the dam. Then
liquefaction analysis was conducted with the densification model, which was proposed by Zienkiewicz et a1 (1978). The liquefaction parameters were estimated from a
liquefaction strength curve, which was only one test data set, for its liquefaction
properties. The dam quickly liquefied and failed using those parameters. To find the
reason for this we surveyed the influence of mesh, soil parameters and initial stress
conditions. Finally, a reasonable simulation result was obtained by controlling the
influence of the initial shear stress condition. There we found that we could not alter
the initial shear stress itself with respect to the shear failure criterion, because of the
presence of the initial shear stress, due to the sloping of the dam side. This can neither
be neglected nor reduced. However, its influence on the dilatancy then came into
question. Therefore, this relation was further investigated.
0.1
Time (Sec)
283
-1501
'
'
'
'
'
'
'
13
Time (Sec)
'
'
'
26
Soil properties
The soil parameters used in the analysis are shown in Table 7.7. The various values of
the shear moduli G were evaluated through the N values of the standard penetration
test. Cohesion c and friction angle 4 were obtained from the drained tests. Densities y
'
Sedimentation
(a)
Figure 7.30 Configuration of the dam: (a) distribution of soil properties; (b) FEM mesh and
boundary condition
284
were obtained by the physical tests. Permeability was determined through 20%
particle size D20 by Creager's approach (Creager et al., (1944)).
Name
Core I
Core I1
Random I
Random I1
Random 111
Random IV
Random V
Sand
Sand & gravel
Sand & gravel
Sandy rock I
Sandy rock I1
Sandy rock 111
Number of cycles N
Figure 7.31
Liquefaction strength
4 (deg.)
285
M E C H A N I S M OF LIQUEFACTION FAILURE
N-value
F C ('%I)
RIzo
r.
Random I1
Random IV
Gravel
3.4
4.7
20.6
32.2
29.3
0
0.28
0.17
2
0.1692
0.2239
0.2998
0.01 1
0.007
0.004
8
6
3
-50.65
-86.18
-59.78
1
1
Initial stress
Initial stress was calculated for self-weight assuming that Poisson's ratio was 113.
Figure 7.32 shows the initial maximum shear stress.
The initial mean effective stress and shear stress are important, since both influence
the build-up of excess pore water pressure. With the densification model, the initial
shear stress causes a rapid build-up of pore water pressure, which does not agree
with ordinary engineering experience. Therefore, the initial shear stress was
ignored in the densification model modifying (4.162) to the following equation in
this analysis.
where Oo is the initial shear stress ratio. The dilatancy of the sand may not depend on
the initial shear stress if the material pass a long period after the construction
of the dam.
Figure 7.32
286
dam could be said to be about 200gal. Figure 7.33 shows the damage to the dam,
where three cracks were observed. The largest depth of the crack C1 was 1.Om-1.2m.
The cracks C2 and C3 are shown on the surface of the upper stream.
Figure 7.34 shows the progress of the deformation of the dam in the case of 200gal.
At about 10 seconds, the surface of the upper stream showed movement toward the
toe of the dam. This deformation might have caused the observed cracks C2 and C3
DISP. SCALE :1 - 1
1.9117
At 10 seconds
At 20 seconds
At 26 seconds
287
at the surface of the upper stream as shown in Figure 7.33. During the first 20
seconds the deformation of the upper stream surface became enlarged continuously.
Then the top part of the upper stream began to settle 0.84m due to the liquefaction
near the dam core. This might have caused crack C 1. Excess pore water pressure was
very high at 26 seconds as shown in Figure 7.35. This caused the overall land slide
and the largest crack, C1.
7.6.5 Remavks
The liquefaction analysis made clear the failure mechanism of the dam. Since sand
region has liquefied and dyke sank under self-weight no significant slip line developed. This mechanism was explained by a rather simple densification model with a
simple modification.
288
Figure 7.36
Analytical model
The input motion used is the recorded data at Akita-Kencho for the Niigata
Earthquake of 1964. The time history is shown in Figure 7.37.
Table 7.9 shows the soil-layer model and material parameters. The liquefaction
strength calculated by the soil parameters is shown in Figure 7.38. All layers have the
possibility of liquefaction but layers 7-9 are stronger than the upper layers. Figure
7.39 shows the result of the undrained tri-axial test by changing the applied
stress ratio. The N-value is the number of loading cycles for the strain amplitude,
5% is calculated. Figure 7.40 shows the initial stress of ground under and far
from the building. The stress contour under the building is higher due to the overburden load.
7.7.1 Results
Figure 7.41(a) shows the deformation due to earthquake. Layers 5 and 6 show
permanent horizontal deformation for both sections A and B and the ground surface
settled. Figure 7.41(b) shows the bird's-eye view of the deformation. Parts of the
ground element are eliminated to show the inside. The excess pore pressure beneath
the building is large due to the over-burden load. This can be seen in Figure 7.42 by
comparing the results at sections A and B.
289
.""
Time (Sec)
Figure 7.37 Recorded data at Akita-Kencho for the Niigata Earthquake of 1964
- --------
10
Number of cycles N
10
290
20
40
60
Mean stress (kPa)
N= 5
80
100
20
40
60
Mean stress (kPa)
N=20
80
100
N=5
- I4.Orn)
porosity
Poisson's
ratio
G
(kPa)
Permeability
(rnls)
v,(~P.,)
Cohesion
(kPa)
(deg.)
291
MI
tr,
M,
ru,
8"
3,
Ho
Hue
7"
YDM
GL-2.0
-5.0
GL-5.0
-8.0
GL-8.0-14.0
0.39
0.50
0.58
0.46
0.54
0.62
1.1 1
1.11
1.11
0.35
0.35
0.35
1.0
1.0
1.0
0.12
0.02
0.0062
2.8
2.7
6.5
2.8
5.7
2.5
2.6
3.6
5.7
2.6
3.6
5.7
--
(a) Section A
(b) Section A
292
(a) Section A
Figure 7.41
DISP. SCALE :
5.47E-02
293
294
( a ) Section A
( b )Section B
Section A
Gl
E
F
C
H
0.00
10.0
20.0
30.0
40.0
50.0
83.0
70.0 &Pa)
Figure 7.42 Vertical stress immediately after the earthquake stops (Pore Pressure)
In Figure 7.43 numerical models for a cross-section of a power plant is shown. The
building is about 320m long and 65m wide and modelled as a simple shear beam with
stiffness equivalent to the whole structure like some form of super-element. The dark
295
Figure 7.43 Configuration of the numerical model and mesh (shallow layer model)
hatched zone is the improved soil which is supported by a strong support layer with
shear wave velocity of over 400mIs in which condition the soil layer will not liquefy.
The surrounding area is the sandy soil layer where liquefaction is anticipated if a
major earthquake strikes. The liquefaction analysis is conducted for a probable
major earthquake artificially generated for the Tokyo Bay area. The maximum
velocity is 50 mls.
The material properties of the building are summarized in Table 7.10. The soil
property is determined by soil sampling at a real site as shown in Table 7.1 1. Density
and porosity are obtained by the physical property test. The N-value is obtained by
the standard penetration test, the shear velocities ( V s )by the wave velocity investigation at the field, drained cohesion and friction angles ( c and 4) and liquefaction
strength, by laboratory tests; permeability from particle size DZo(diameter at which
20'51 of the soil is finer) according to Creager's experimental data (Creager et al. 1944).
In this analysis, all necessary soil tests mentioned above have been completed but in
most cases, only limited data will be available for the prediction analyses, since they
will be carried out at a very early stage of construction. At this stage, the details of the
project will not be finalized. Engineers will be forced to interpolate between the
limited amount of data, occasionally they need to extrapolate.
Table 7.10 Equivalent beam model of a building: (a) stiffness; (b) lumped mass at node
(a)
Element No.
Height (m)
Long dir.
Short dir.
29.1(34.4
92.4
3.53
24.0(29.1
57.45
8.51
18.5(24.0
165.
9.54
13.0(18.5
74.1
11.2
7.0(13.0
108.
17.4
0.0(7.0
153.
25.8
Node. No.
Height (m)
Long dir.
Short dir.
34.5
42.6
7.82
29.1
187.
34.34
24.0
88.0
16.1
18.5
46.56
8.54
13.0
241.
44.3
7.0
223.
40.8
0.0
1910.
363.
-5.9
2680.
529.
(unit: Mglm)
Fine sand
SFs
Silty sand
Finesand
Fine sand
Silty clay
Fine sand
SFs
As1
As2
Dcl
Dsl
-1.40
2.70
4.40
5.30
-6.80
-8.80
-1 1.90
-13.05
-14.60
-19.80
11
11
7
3
3
3
27
density
Share
( ~ ~ / m
wave
~ )
velocity
(mlsec)
Poisson's Friction
ratio
Angle
(degree)
Cohesion
(kPa)
Permeability
(cmisec)
Rl2O
ud/2uo1
Soil
particle
density
Porosity
297
298
400 1
Time (Sec)
(b) Acceleration spectra
Figure 7.45
motion at a depth of bearing layer such as the Edogawa Layer (V, = 300-500 mls).
To determine the seismic intensity in the Tokyo Bay area, several earthquakes were
used such as El Centro, 1940, NS, Taft, 1952, EW, and artificial earthquakes
recommended by Japan Architectural Centre. Figure 7.45 shows the time history of
the Rinkai 1992, an artificial earthquake. A flat spectrum between 0.2 and 0.6 Hz at
maximum is assumed.
Figure 7.46 Pressure at the side boundary of the improved soil ground: (a) maximum
distribution; (b) comparison with Westergaard's
299
side boundary of the improved soil ground was not well known in the case when the
surrounding soil layers were liquefied. Therefore, a liquefaction analysis was made.
Figure 7.46 (a) shows the pressure of the skeleton and the water at each level from left
to right. The maximum total pressure from both sides is shown in Figure 7.46(b). The
results obtained by the effective stress FE analysis agreed well with Westergaard's
formula for dynamic water pressure.
-x
5
-2
V]
--
x
.U
a
0
-t:
-0
Maximum
contact
pressure
(kPa)
Subgrade
reaction
2?
- m0
Subgrade
reaction
for ISG
(kPa)
Slip
[safety factor]
V)
(kPa)
Horizontal
shera force
(kPa)
Vertical
shera force
(kPa)
Ground
kind
Building
type
Side
Analysis
Design
Analysis
Design
Shallow
layer
Lighter
Long
Short
Short
Short
Long
Short
Short
Short
Long
Short
Short
Short
Long
Short
Short
Short
Long
Short
Short
Short
Long
Short
Short
Short
221
239
293
513
1.69
1.91
2.01
1.99
221
239
290
646
247
310
330
310
81
78
92
186
147
105
1 16
142
224
269
326
688
1.57
1.43
1.44
2.10
224
269
326
688
0.98
0.89
0.90
0.75
0.99
0.89
0.89
0.94
Heavier
Deep layer Heavier
Shallow
Lighter
layer
Heavier
Deep layer Heavier
Shallow
Lighter
layer
Heavier
Deep layer Heavier
Shallow
Lighter
layer
Heavier
Deep layer Heavier
Shallow
Lighter
layer
Heavier
Deep layer Heavier
Shallow
Lighter
layer
Heavier
Deep layer Heavier
176
205
24 1
253
0.83
0.51
0.48
0.56
300
3 0
2 0
1 0
0
10
Horizontal distance from centre (m)
20
30
but further research is necessary for the limitation taking into account the dependency of the confining pressure for the strength of the improved soil ground. The
possibility of overturning was very small since the side ground supported the
improved soil.
Internal safety
In order to check the safety of the improved soil ground, the maximum stresses at the
local point were checked against the strength for normal stress. The maximum stress
naturally occurs at the bottom edge of the building or of the improved soil ground as
shown in Figure 7.48. There, very little moment was seen.
Contour lines for the maximum value at a time of the shear stress (T,,,.) under the
building shown in Figure 7.48 is almost horizontal. This means that the failure line
can be developed along the line. Distributions of the maximum value of the shear
stress along the assumed horizontal slip line for the depth from GL-6.35m to G L
- 12.0m. are shown in Figure 7.49. The shear stress close to the foundation has sharp
stress concentration at the edge of the building and low stress at the centre of the
building. Shear stresses at the side area of the improved soil ground are very low due
to liquefaction.
40
301
Various types of possible shear failure mechanisms were also checked. The footing
type of failure was not triggered by the seismic force in this problem. The failure line
may be horizontal near the bottom and/or vertical near the edge. They were both
examined using the shear stress (r,.) calculated by the liquefaction analysis. Figure
7.50 (a) shows the distribution along the horizontal section for the maximum value at
the time of the average shear stress. As in the ordinary slip-line analysis approach,
the accumulated shear stress per unit length along the vertical or horizontal
section was compared with the shear strength of the material. In the vertical
section, the values were comparable and less than the failure strength. In the
horizontal direction, the result of the calculation was much lower than the material
strength.
From this study it is concluded that the failure mechanism for the safety check is
adequate for horizontal section but overestimated the vertical one.
7.8.3 Remarks
The safety of the improved soil ground by the deep-soil mixing method for very large
structures was examined by dynamic effective stress FE analyses. The analyses
concluded that the following mechanisms should be checked and are sufficient for
302
the safety evaluation of the improved soil ground: (1) sub-grade reaction of the
improved soil ground to the bearing layer; (2) slip failure at the bottom; (3) maximum
contact pressure at the bottom of the improved soil ground; and (4) average shear
stress along the vertical line for internal safety. Following these conclusions a design
procedure based on the seismic intensity method was proposed. The results of the
analyses also helped to determine the seismic intensity and other external forces such
as the earth pressure to the side boundary of the improved soil ground.
REFERENCES
Bazant Z.P. and Krizek R.J. (1976). Endochronic constitutive law for liquefaction of sand,
Proc. ASCE. J. of Mech. Div., 102 (EM4), 701-722.
Creager W.P., Justin J.D. and Hinds J. (1944) Engineering for dams, Vol. 111: Earth, Rockfill,
Steel and Timber Dams, John Wiley and Sons, 649.
Fukutake K. and Ohtsuki A. and Fujikawa S. (1995), Applicability of 2D analysis and merit of
3D analysis in liquefaction phenomena, Proc. Symp. On Three-dimensionul Evaluation o f
Ground Failure, Tokyo, 229-236 (in Japanese).
Ghaboussi J. and Dikmen S.U. (1981), Liquefaction analysis for multidirectional shaking, J.
Geotechnical Engineering Division, Proc. qf ASCE, 197, GT5, 605-627.
Hatanaka M. and Uchida A. (1995). Simple method for the determination of the KO value in
sandy soil, First In t. Conference on Eurthquake Geotechnical Engineering, I, 309-3 14.
Hyodo M., Murata N., Yasufuku and Fuji (1988), Effect of initial shear stress on development
of undrained cyclic residual shear strain in saturated sand, Proc. Symp. on the use of'
laboratory test for undrained cyclic behaviour of'soil and in-situ test, 199-204.
Ishihara K. (1989). Dynamic behaviour of ground and earth structure Numerical method
and problem specification, Proc. Symp. on hehaviour of ground and earth structure during
earthquake held by Japanese Societj, of Soil and Foundation Engineering in Tokyo, 50-63, (in
Japanese)
Ishihara K. and Tatsuoka F., Yasuda S. (1975), Undrained deformation and liquefaction of
sand under cyclic stresses, Soils and Founclations, 15(1), 29-44.
Ishihara K. and Yamazaki F. (1980), Cyclic simple shear tests on saturated sand in multidirectional loading, Soil and Foundations, 20(1), 45-59.
Ishihara K. and Nagase H. (1988) Multi-directional irregular loading tests on sand, Soil
Dynamics and Earthquake Engineering, 7 , 4 Oct 1988, 201-212.
Kansai - Kyogikan, Kansai - Kyogikan report, 1995
Kishino Y., Narikawa M., Masuda A,, Nakamura N., Shiomi T. and Suzuki Y. (1998), A
study on building foundation with deep-soil cement mixing method for thermal-power plat,
Part 1: design procedure and its verification by effective stress analysis, 6th US National
Conference on Earthquake Engineering No. 488.
Kondner R.L. and Zelasko J.S. (1963) A hyperbolic stress-strain formulation for sands, Proc,
2"d Pan American Conference on Soil Mechanics and Foundations Engineering, 289-324.
Nagase H. and Ishihara K. (1988) Liquefaction-induced compaction and settlement of sand
during earthquake, soil.^ and Foundutions 28 1 Mar 1988, 65-76.
Pastor M. and Zienkiewicz O.C. (1986), A Generalised Plasticitv, Hierarchical Model for Sand
under Monotonic and Cyclic Lor~ding,(eds. G.N. Pande and W.F. van Impe), 131-150,
Jackson and Son, London.
Pastor M., Zienkiewicz O.C. and Chan A.H.C. (1988), Simple models for soil behaviour and
applications to problems of soil liquefaction, in Numerical Methods in Geon~eclzr~nics,
(ed.
G. Swaboda) 169-180, A.A. Balkhema. geomechanics
-
REFERENCES
303
Pastor M., Zienkiewicz O.C. and Chan A.H.C. (1990), Generalised plasticity and the modelling of soil behaviour, Inr. J. NLIIII.
AIINI.Mrtli. Geolllech., 14, 151-1 90.
Pastor M., Zienkiewicz O.C. and Leung K.H. (1985). Simple model for transient soil loading in
earthquake analysis, I 1 Non-associative models for sands, h t . J. NUIIW.Anrrl. Met11.
Geotech., 9, 477-498.
Pyke R.M. (1973) Settlement and liquefaction of sands under multi-directional loading, thesis
presented to the University of California, Berkeley. California, in partial fulfilment of the
requirements for the degree of Ph.D.
Seed H.B. (1979) Consideration in the earthquake resistant design of earth and rockfill dams.
Geotacl~nique,29, 21 5-263.
Seed H.B., Lee K.L., Idriss I.M. and Makdisi F.I. (1975) Analysis of slides of the San
Fernando dams during the earthquake of February 9, (1971). J. Geotecl~ Eng. Dir.
ASCE, 101. GT7, 651-688.
Seed H.B., Pyke R.M. and Martin G.R. (1978). Effect of multidirectional shaking on pore
water pressure development on sands, J. Gpotecl~.Big. Div. ASCE, 104(GTI), 27-44.
Shiomi T., Shigeno, Y., Sugimoto, M. and Suzuki, Y. (1991) Influence of liquefaction to pilesoil-structure interaction, Second Int. Con$ Geomec/~at~ics/Mi.~.~ouri
Rollu, pp. 465-72.
Shiomi T. and Shigeno Y. (1993) Consideration of initial shear stress on ground liquefaction.
Second Asian-Pasific conference on Computational MechanicslSydney, 1077-1082.
Shiomi T., Shigeno Y. and Zienkiewicz O.C. (1993) Numerical prediction for model No. 1.
Proceedings q f ' Ve~rijicr~tion
of' Nu~?~erical
Procwlures for the Antrlysis of Soil Liquefirction
ProblemslDavis, California11 7-20 October, 2 13-220.
Shiomi T., Muromoto T. and Tani S. (1996) Dynamic behaviours of fill dam during earthquake on liquefaction analysis, Proceedings of the Thirty-First Japan National Conference
on Geotechnical Engineering, 1263-4.
Shiomi T . (1995) U . w M(IIIN(I/
O ~ M I I D I A Takenaka
N.
Corp..
Shiomi T., Shigeno Y. and Zienkiewicz O.C. (1993) Numerical prediction for Model No. 1.
Verificution c?f'Nutnc,ric~rl
Proccdurt,.~,fort11c.Ancr/issisof'Soil Liquclfirction Problems, (eds. K .
Arulanandan and R.F. Scott), 1. pp. 213-219.
Suzuki K., Saito S., Kimuira T., Kibayashi M. and Hosonii H. (1995), Investigation report on
the building foundation made of the lattice wall soil improvement for liquefaction prevention, Kim - Kon, 10, 5 4 5 8 (in Japanese)
Tokiniatsu K. and Yoshimi Y. (1983). Empirical correlation of soil liquefaction based on SPT
N-value and tines content, Soil unci Fou~~tlutior~,
23, No. 4, p.p. 56-74.
Van Genuchten M.Th., Pinder G . F . and Saukin W.P. (1977) Modeling of leachate and soil
interactions in an aquifer, Proceedings, 3'-d Annual Municipal Solid Waste Res. Symp.
EPA- 60019-77-026, 95-1 03.
Vaid Y.P. and Finn W.D.L. (1979), Effect of static shear on liquefaction potential. J. Geotech.
Eng. Div., ASCE, 105, (GTIO), 1233-1246.
Yoshimi. Y. ( 199 1 ), hzfluer~ceq f Confnit~gPre.s.sure. Liquc~firctiotiqf' Sund (2nd Ed.), Gihodo
publisher, Section 3.3., 32-34, (in Japanese)
Zienkiewicz O.C., Chang C.T., and Hinton E. (1978). Nonlinear seismic response and liqueFaction. Int. J. Nurw. rind A n d . Method in Geonwcl~.,2(4), 381404.
Zienkiewicz O.C. and Xie Y.M. (1991) Analysis of Lower San Fernando dam failure under
earthquake, DNIUEtzginrering. 2, Issue 4, 307-322.
Zienkiewicz O.C. and Shiomi T . (1984), Dynamic behaviour of saturated porous media. The
generalized Biot formula and its numerical solution. It~t.J. Nuw. ( r t d AI?N/.M ~ t h .in
Geoniech., 8. pp. 71-96.
8.1 INTRODUCTION
In the presentation of the essential theory and the finite element discretization
procedures we have deliberately omitted some 'finer points' which on occasion
might be essential to obtain more accurate or more generally applicable solutions
to realistic engineering problems. We shall introduce these 'finer points' in the
present chapter in sufficient detail to allow the reader to follow the current literature
and to devise his or her own programme modifications.
The chapter will be divided into three sections corresponding to the topics
discussed and each section can be studied independently. These sections are:
$8.2 Input for earthquake analysis and the radiation boundary.
48.3 Adaptive refinement for improved accuracy and the capture of localized
phenomena.
48.4 Stabilization of computation for nearly incompressible behaviour with equal
interpolation.
306
The input for earthquake analysis is based on measured recorded data of actual
earthquakes and is generally presented as the values of the displacement u and or of
the acceleration ii at the time interval of 0.02 seconds given for the duration of the
earthquake.*
If the time history of the input can be specified, we can proceed as outlined in this
book to obtain the solution by time integration of the discretized form of the
equations of motion such as those given by equations (2.1 1 ) and (2.13) of Chapter 2.
The simplest case for the specification of input is illustrated in Figure 8.1 which
attempts to model a structure resting on a stratified soil foundation of unlimited
extent, by specifying the input motion at some arbitrary internal boundary shown.
Such a model corresponds well with such physical models as those of the shaking
table or centrifuge where the specified boundary represents a 'box' into which the
model is fitted and which moves in a specified manner. In Chapters 6 and 7 we have
shown several calculations which correspond to such physical experiments and which
model the real phenomena of practice reasonably well.
With all displacements or tractions at the boundaries specified, we can use the
discretization of Chapter 3 and proceed with the solution of any transient problem. It
should, however, be remarked that, if only uniform motion is specified on the
boundaries, it is sometimes convenient to recast the equations of motion in terms
of the relative displacement u~ which we define as
Figure 8.1 Specified motion on the boundaries of a 'shaking table box' modelling of an
infinite foundation
* In the USA such records can be obtained from The Earthquake Research Institute at the University of
California at Berkeley and similar sources are available in other countries.
307
where UE = uE(t) is the prescribed earthquake motion which does not depend on the
position. The governing equations (2.20) and (2.21) of Chapter 2 become now
(neglecting the source terms and putting cy = 1).
which is replaced by
If the relative velocity is used in the finite element discretization of the problem, the
numerical computations are identical to those of the absolute displacement if the
same initial conditions (e.g. u = 0) are assumed. However, the input is now the
acceleration iiE giving a prescribed body force and this is often more accurately
known.
In a more realistic treatment of the foundation problem, we shall impose somewhat different boundary conditions recognizing the fact that, in the input, only the
incoming wave motion is specified and that outgoing waves must leave the problem
domain unimpeded. Figure 8.2 again shows the problem initially suggested in Figure
8.1 indicating the position of the same limiting boundary but on which the motion
will not now be directly specified. We shall discuss this problem in the next section in
more detail and suggest how such problems can be dealt with. First, however, a few
words about the way knowledge of the seismic input wave is obtained.
Figure 8.2 A more realistic model of an 'infinitie' foundation with a specified incoming
wave
308
The seismic signal is usually measured at or near the free surface, and it represents the
modification of the original seismic wave which is initiated at the earthquake source,
caused by passing through different material zones and involving a number of
internal reflections and refractions at the interfaces between layers of different
material.
The geological conditions at the site will very often be such that the so called
'bedrock' exists as a zone of a significantly more rigid material underneath softer
soil layers. Any incoming seismic wave passing from the bedrock to the softer soil
layers will amplify depending on the material properties of both bedrock and the soil
layers. The significant consequence of the presence of the bedrock lies in the fact that
all of the reflected waves are practically trapped inside the soft soil layers, as only a
small fraction of these can be transmitted back to the bedrock through the interface
with the softer soil. If the bedrock is significantly more rigid, the transmitted wave is
smaller than the reflected one back towards the soil surface. In such a case, the simple
fixed-base approach is valid and no transmitting boundary conditions need be
imposed on the bedrock level as practically no waves get transmitted into the bedrock.
The need for an arbitrary model truncation emerges in the cases where no distinct
base rock exists, or when the extent of the softer soil layers is so great that it would be
prohibitive to include the whole zone in a mathematical model. Such a situation may
also arise when the non-linear material behaviour can be expected only near the
surface and deeper layers (with material properties still far from bedrock-like characteristics) are expected to remain elastic.
To model such a case correctly it is necessary to reconstruct the incoming seismic
wave at the model truncation boundary. In the simplest case of a one-dimensional
elastic, homogeneous, isotropic wave propagation problem involving the free surface, it is very well known that the free surface displacement wave equals the double
of the incoming displacement wave. Here the incoming signal can be easily extracted
from the recorded total signal on the undisturbed surface. Even in the case of the
elastic non-homogeneous domain the incoming signal can again be extracted from
the total signal recorded on the surface. Therefore, in the following, it will be
assumed that the incoming wave (displacement, velocity or acceleration) is known
at a position corresponding to the model truncation boundary, and that outside of
this homogeneous elastic conditions pertain.
8.2.2
309
I N P U T FOR EA R T H O U A K E ANALYSIS
"An
tttttt
Thus, all the derivatives with respect to the x axis are made identically zero.
T o demonstrate the wave nature of the problem we shall assume that in the vicinity of
the arbitrary, 'input' surface CD (and below this surface) the conditions are such that:
(i) only isotropic elastic behaviour exists;
(ii) the body forces have been taken into account independently; and finally
that
(iii) the dynamic phenomena are sufficiently rapid so that the flow in the porous
medium can be neglected and k = 0 is assumed.
Writing the total relative velocity in terms of its components
310
where only total stresses are considered. The elastic constitutive relation under
isotropic undrained conditions gives
and
where
is the restrained axial modulus. E and u are Young's modulus and Poisson's ratio.
Equation (8.6a) becomes on insertion of the above
d2uY
dy2
--
p..
TU
p=
(8.9b)
Each of the above equations corresponds to the well known scalar wave equation
a2Ch
1d2d
in which c is the wave velocity and 4,and 4, represent two waves travelling in
the positive and negative directions of y respectively (incoming and outgoing
waves).
Thus, 4 = u, represents shear waves travelling with velocity
311
We observe that c, tends to infinity for fully incompressible solid and fluid
situations.
To obtain the radiation condition we observe the solution sought at the 'cut-off'
line C D should represent only an outgoing wave, i.e.
where
to ensure the existence of outgoing waves alone. Using the relationships (8.7) and
noting the definitions of (8.12) and (8.13) we will observe that on the boundary C D
the tangential traction becomes
312
suggested almost simultaneously by Zienkiewicz and Newton (1969) and Lysmer and
Kuhlmeyer (1969).
In the one-dimensional case presented here the radiation condition is exact. However, on many occasion it has been used effectively on two or three-dimensional
boundary shapes where the conditions of equations 8.17(a) and (b) imply dashpots in
the tangential and normal direction at any position of the boundary. The numerical
tests of the effectiveness of such a radiation boundary condition are presented by
Zienkiewicz et a/. (1987) where it was shown that, for a given wave input form,
identical results are obtained independently of the arbitrary cut-off position.
Many alternative forms of radiation boundary conditions have been developed.
Here the early work of Smith (1973) which was recently generalised by Zienkiewicz et
a1 (1987a) is one possibility. Alternative methods are discussed by Wolf and Song in
their recent text (1996), White et a/. (1977), Kunar and Marti (1981) and also by
Zienkiewicz and Taylor (199 1).
It is usual to conduct an analysis in terms of the relative displacement UR defined
by (8.1) and to apply the radiation condition to this relative displacement only (see
Zienkiewicz et al., 1987a, Clough and Penzien, 1993).
8.2.3
A more general situation of engineering interest is the one illustrated in Figure 8.4
where a structure 'perturbs' the simple one dimensional solution of the layered
foundation. Once again the horizontal boundary on which the vertically propagating
waves enter the problem domain is treated identically to that of the one dimensional
~ e r t i k awave
~ prbpakation
Figure 8.4 Foundation of Figure 8.3 perturbed by imposition of a structure (a) and the ID
problem (b)
313
t t t t t t t t t t t t t t
Figure 8.5 Repeatable boundary conditions. Displacement at A=displacement at B.
case. Indeed, identical 'dampers' are placed on this boundary to ensure transmission
of the exiting waves (but now these are approximate only as the transmission
conditions d o not apply exactly to waves exiting obliquely to the boundary).
More serious difficulties are, however, posed on the two vertical boundaries AA and
BB and the boundary condition which needs to be imposed on these. Clearly, at
points which are far away from the superposed structure the solution must be
asymptotic to the previously discussed one-dimensional one. A possible way of
dealing with the boundary conditions on these sections is therefore to impose the
radiation damper between the interior region and the one-dimensional, free field.
solution. Such a treatment is suggested by Zienkiewicz r t 01. (1988) but a simpler
alternative is that of repeatable boundary conditions which is also given there. The
latter achieves identical results more simply.
In the repeatable boundary condition, which is illustrated in Figure 8.5, it is
assumed that a sequence of structures is placed on the foundation at regular intervals
B. Treatment of such repeatable conditions is simple in the finite element context (see
Zienkiewicz and Taylor, 1989) as clearly the values of displacement, stresses etc. are
identical on such a section as A or B due to periodicity and the assembly of nodal
values at these boundaries is ensured by suitable node numbering.
In Figure 8.6 we illustrate a test problem where different depths and widths of the
analysed domain are used. A homogeneous elastic material is here assumed throughout the space and Figure 8.7 shows the time histories of displacement, acceleration
and stress for a typical point at the base of the structure and with different domains
of computation. It is surprising to note how little the results are affected by the extent
of the domain assumed.
8.2.4
We have limited our discussion so far to that of the behaviour of the two-dimensional
foundation layer problems with a vertically propagating wave input. Extension of
314
\\
the problem to three dimensions for the same wave input is trivial but of course three
dampers will now be necessary on the radiation boundaries. Greater difficulties are
presented by problems in which the earthquake (or shock) waves enter the boundary
obliquely or indeed horizontally. Here, of course, the input motion history will be
dependent on the position and the determination of this in itself is a major problem.
However, once such motion is established it is possible to apply radiation boundary
conditions throughout. We shall not discuss this difficult problem further as it is not
frequently encountered in practice.
316
Figure 8.8 First adaptive solution of a purely plastic deformation problem. A perforated bar
(a) initial mesh (b) final adapted mesh with elongation DOF 1039 (c) initial material
ADAPTIVE REFINEMENT
317
For such problems we can separate the process of error determination and of
refinement of the mesh. Thus the latter can be guided, for instance, by requiring that
such a quantity (or indicator) as
where C is constant between all the elements, hnli, is their minimum size and s the
direction of the maximum gradient of 4, the function of interest. This quantity can be
interpreted as the maximum value of the first term of the Taylor expansion defining
the local error of the scalar quantity 4. By ensuring that the mesh is generated so that
the quantity C is constant throughout all elements, we achieve a solution which
captures well all local discontinuities and which is efficient in achieving the progression which gives overall accuracy. At any stage of refinement, estimates of error are
possible by using various recovery procedures, (see Zienkiewicz and Zhu, 1992), but
alternatively convergence to exact solution can be studied by simply reducing the
constant C in refinement.
The use of the indicator defined by (8.18) allows element elongation to be included
in the refinement, as on many occasions the feature occurring at high gradients is
almost one-dimensional. Indeed, in a truly one-dimensional feature the maximum
sizes of an element along its direction would be arbitrary and any reasonable value
would be fixed on the maximum element size h,,,,. However, if the contours of the
function 4 diverge or are curved, the upper limit h,,,, can be specified more closely.
Thus, for instance, if the contours separated by the value of h,,,,, diverge by an angle 0
then the limit on /I,~, could be replaced as
it is often specified where cw is circa 0.1 and R is the radius of curvature. These type of
procedures are discussed in detail by Zienkiewicz and Wu (1994) in the context of
fluid mechanics.
An alternative refinement indicator has been used for the longer time in fluid
mechanics. This is a requirement that
specifies the minimum size of elements. This specification was first formulated by
Peraire et ul. (1987) and is very effective in the capture of shocks. Here, the elongation of elements can be computed directly in terms of principal curvatures.
318
It appears that the first indicator (i.e. that of (8.18)) is most efficient in the capture of
narrow discontinuities, but both provide a remeshing which gives a rapid convergence and reduction of both local and global errors. Figure 8.9 shows how an
adaptive analysis based on the first indicator can model discontinuity developed
during the failure of the foundation under an eccentric load. Here a von Mises type
of yield surface is used with ideal plasticity assumptions (Zienkiewicz et a/., 1995a
and b).
In Chapter 3 we have already mentioned that special conditions have to be satisfied
by mixed finite element forms for incompressible, or nearly incompressible, behaviour such as is encountered under undrained conditions. Indeed, such behaviour will
Figure 8.9 Adaptive solution of the problem of foundation collapse with an ideally plasticelastic material (a) ecentrically loaded footing (b) final adapted mesh and deformed
configuration showing displacement discontinuity.
ADAPTIVE REFINEMENT
319
occur in many applications of plasticity using von Mises or Tresca yield surfaces. For
adequate solution it is always necessary to use here special mixed forms of elements
which are outlined in Chapter 3. In the two examples quoted already we used the
T6CI3C triangle where six nodes define a quadratic variation of continuous displacement and three nodes interpolate pressures in a continuous manner.
In Figure 8.10 we show again an analysis of an ideal elasto- plastic problem in
which a strong localization occurs. Here two forms of regular mesh are comparedone named 'lucky' mesh in which the triangle sub-division lines follow approximately
the slip surface, and the other, the 'bad' mesh in which these lines are orthogonal to
the slip surface. It is clearly noted that for the same sub-division the 'bad' mesh gives
answers which are always inferior to those of the 'lucky' mesh. However, the adaptive
solution starting from either refinement shows nearly exact values of the collapse
load.
A serious problem with adaptive analysis of non-linear problems of plasticity in
which the results are path dependent is that of data transfer between the various
stages of analysis. In principle, the control of the error should be achieved at each
load increment separately and this, of course, necessitates the transfer of history
dependent data such as stresses, strains, etc. from the mesh of the previous step to
that used in the next increment. To avoid difficulties we have re-analysed the
problems in each of the previous cases from the start of loading for every new
mesh developed. Indeed such a procedure has also been used quite effectively in
transient analysis of the San Fernando dam by Zienkiewicz and Xie (1991) and
Zienkiewicz et ul. (1995a, and b) with results shown in Figure 8.11. However,
currently new procedures of transferring data have been developed and it is now
possible to change the mesh at each load increment, thus ensuring a constant degree
of accuracy (Zienkiewicz et al., 1998).
320
element size h to that of the total length .t a progressively steeper unloading branch of
the load deformation plot will occur. This can reach negative slope values and allows
only elastic unloading which, obviously, is not correct. But certainly the steepening of
this slope will increase to infinity for a finite value of / I and will imply a displacement
discontinuity or full localization.
ting
Figure 8.10 Failure of a rigid footing on a vertical cut. Ideal, von Mises. plasticity and
triangular T6Cl3C element (quadratic, continuous displacements, linear continuous pressure)
( a ) geometry data; (b) Mesh 2 (fine 'lucky'); (c) Mesh 3 (coarse. 'bad'); (d) Mesh 6 (adaptive
solution obtained from Mesh 3); (e) displacen~entvectors; (f) Effective strain contour for Mesh
1 (fine. bad) (g) Load, displacement results for various meshes.
ADAPTIVE REFINEMENT
Figure 8.11 Earthquake analysis of lower San Fernando Dam (a) initial mesh
refinement at t-75 seconds (c) adaptive refinement at t = 30 seconds.
321
(b) adaptive
In the example quoted, the localization was caused by a small weakness due to the
statistical nature of the material strength behaviour. However, in other geometrically
more complex problems the stress concentration, etc., will act in precisely the same
manner, always causing a localization with softening material behaviour.
However, the example discussed shows up another feature of the problem, i.e. that
of numerical non-uniqueness as the slope of the unloading portion of the displacement load curve depends largely on the size of the element used.
This non-uniqueness of the problem becomes most serious in multi-dimensional
behaviour in many structural problems. In Figure 8.13(a) we illustrate the fairly large
discrepancies which occur in the estimate of the maximum load for the problem
illustrated in Figure 8.10 for which now a softening modulus has been assumed and
different mesh subdivisions used in the solution.
Even discounting the results obtained by the use of the coarse, bad, mesh as being
very unreliable, we note a difference of about 20% in the estimate of the maximum
load capacity when the simulation is achieved by meshes which for ideal plasticity
give almost identical answers. While the reason for this has been hinted at it in the
simple example of Figure (8.12), the manner in which the problem can be overcome
has supplied many researchers with material for exercising their ingenuity. De Borst
er al. (1993), Ortiz et al. (1987), Bazant and Lin (1988), Belytschko er al. (1988) and
Belytschko and Tabarrok (1993) describe some of the possible procedures which
range from the consideration of material as a Cosserat medium, through so called
gradient plasticity, to a simple failure energy consideration introduced in the last of
322
323
ADAPTIVE REFINEMENT
these references. We shall only refer here to that last procedure which, in the opinion
of the authors, deals adequately and in a simple manner with the difficulties encountered.
The procedure considers, in the manner common to that of early theories of
fracture, namely Griffiths (1921), the constancy of work required for failing the
material and requires the energy to be independent of the discretization used and
therefore to be a pure material property.
In Figure 8.14 we show a typical stress-strain relation with strain softening in
which failure is reached. The area under the full triangle is the work required to cause
this failure and, in a unit volume of material, becomes
--
where C is a constant.
It appears therefore necessary to reduce the softening modulus in the manner of
(8.24) as the size of the elements modelling the localization decreases. This indeed was
done in the case of a problem illustrated in Figure 8.10(a) and the results are shown in
Figure 8.13(a,b) which gives an almost identical failure load obtained by two very
different meshes.
It will be observed that the above discussion leads to two conclusions:
324
(i) that with strain softening, localization will always occur in the failure zone and
this will show a continuously decrease in size with the element size h, and
(ii) that the softening modulus cannot remain a material constant but must tend to
zero (i.e. giving no softening) as the size of the element tends also to zero to
present a consistent work estimation. This idea can be incorporated in a
material model with concentrated localization singularity and has been introduced by Simo et ul. (1993) and Oliver (1995). It is clear that such a model will,
in the limit, tend to give identical results to the adaptive refinement if equation
(8.24) is used.
The adaptive refinements of the type here discussed have been introduced by Zienkiewicz et ul. (1995a and b) from which the examples and previous figures have been
quoted. This and other papers in the field, indicate that adaptive refinement is a
feature which can improve the results of analysis significantly although with experience reasonable engineering estimates can be obtained without this feature.
325
S T A B I L I Z A T I O N OF COMPUTATION
Hafez and Soliman (1991) and Sampaio (1991) suggests many alternatives. Some of
these were shown by Zienkiewicz and Wu (1991) to derive very simply from the same
roots of time-stepping analysis.
The motivation for most of this work lies in problems of fluid mechanics and their
numerical solution and it was shown recently by Zienkiewicz and Codina (1995) that
an algorithm using the operator split procedure suggested by Chorin in 1965 automatically provides the desired stabilization.
The use of such stabilization in the context of geomechanics was first made by
Zienkiewicz and Wu (1994) and extended by Pastor et al. (1999). In this chapter we
shall discuss only this last process of stabilization as it appears to be the most simple
and efficient.
8.4.2
In this section we shall outline the semi-explicit time-stepping, operator split procedure which is effective in dealing with the incompressibility problems arising in
geomechanics and which follows the methodology originally suggested by Chorin
(1967 and 1968) and extended by Zienkiewicz and Codina (1995).
It is convenient to introduce the velocity, v, as the basic variable and to compute
the displacement increment by subsequent integration. Thus we have the definition
sT/
Vp + pb and
noting that
a
=a
da'
m p and
= Dds
(8.27~)
326
is the constitutive relation and that (3.21) needs to be used for strain calculation.
The operator split algorithm solves (8.27a) in two steps. In the first the quantity v* is
calculated explicitly from
where the RHS is computed at t = t,,. In the second part the velocity v* is corrected
implicitly in terms of known pressures using
and
All these matrices are defined in Chapter 3 in (3.23-3.26) and need not be repeated
here. We must, however, mention that the evaluation of both V* and Vn+' is fully
explicit if the mass matrix M is diagonalized. This can be done in a variety of ways by
well known procedures discussed in finite element texts (see, for example, Zienkiewicz and Taylor, 1989).
The determination of the pressure increment A p and hence of p"+l requires the
solution of (8.27b). We now write the implicit time approximation as
S T A B I L I Z A T I O N OF COMPUTATION
327
from which A p can be established after discretization. This again proceeds in the
manner previously described and we now have
In the above the matrices S, H and Q are defined in Chapter 3 by (3.29-3.32). The
only new matrix occurring now is H* which is the approximation to the Laplacian
operator.
We shall delay the explanation of the reasons why the split operator procedure
permits the use of arbitrary interpolations for u and p (N and Np respectively) and
shall first illustrate its effectiveness in examples.
8.4.3
Two examples are here quoted. The first of these is the soil layer subject to a periodic
surface load. Indeed this problem is identical to the one used in Section 2.2.3 where
the limits of applicability of various formulations are tested and for which exact
solutions are readily available. Here we shall only use the u-p formulation and shall
demonstrate how the very oscillatory results obtained by an equal interpolation can
be improved by the use of the stabilization just described.
In Figure 8.15 we show the details of the problem and in Figure 8.16 we show
solutions obtained by the use of 2D elements. The first uses 20 Q4P4 elements and
shows oscillations which are very pronounced. The second one shows the very close
approximation and suppression of oscillation obtained using the Q9P4 element as
well as the new stabilizing algorithm.
In Figures 8.17 and 8.18 a fully two-dimensional problem of a foundation load is
solved again showing similar results.
328
--H Im
-1
q= 100 exp (-lwt)
t- = 0
11, =
apw
--
a.~
k,
n
E
p,
10-'mn/s
0.333
X
7.492 10 (pa)
0.2
3
2.0 x I0 (N/m3)
PW
1.0 x 10'(N/m3)
8.4.4
In Chapter 3 we have indicated the main reasons for the difficulties encountered in
solving the problem where incompressibility is approached. We first made a comment on these difficulties when discussing the Jacobian matrix used in the solution of
an iterative step by the Newton-Raphson procedure where the matrices of (3.31) and
(3.30) tend to zero, i.e. when
and H
- I,,
which occurs when the compressibility and the permeability of both tend to zero.
This zero limit leads to a zero diagonal which occurs also in steady state equations
of Section 3.2.5 giving a linear form
As we mentioned there, satisfactory solutions can still be obtained but these require
that the number of parameters describing the variable u must be greater than these
describing the variable p i.e.
329
S T A B I L I Z A T I O N OF COMPUTATION
Medium compressibility
Small compressibility
Q * = 10 MPa
"."
Q*= I0 MPa
0.0
1.5
~ 1 4
(a) Solution with standard column with 20 Q4P4 elements
330
(a)
30 MPa
(b)
Example 2: a saturated soil foundation under transient load; (a) the problem
domain; (b) transient load applied data
Figure 8.17
Figure 8.18 Example 2: two dimensional foundation pressure contours computed for small
permeability and compressibility Q* = lo9 MPa, k = lo-' mis; (a) direct use of implicit
algorithm with Q4P4 elements; (b) direct use of implicit algorithm with Q8P4 elements;
(c) Q4iP4 elements with stabilized procedure
This is a necessary condition for avoiding singularities and can be readily achieved with
certain interpolations. However, if the problem is recast in the manner given in Section
8.4.2. we shall find that even in the limiting case (i.e. with zero compressibility and
permeability) a non-zero diagonal will be obtained a n d stability can always be achieved.
As we have recast the problem in terms of velocities we shall linearize using these
variables and write
REFERENCES
331
and a non-zero diagonal is found to exist in its finite time steps. This seems to achieve
complete stabilization and any interpolation of the V / u and p variables can be used
with equal interpolation, of course, being the obvious choice.
The procedure outlined unfortunately results only in conditional stability,
although the time-step length is now given by the speed of the shear wave and
hence is not too restrictive.
We find that the integration of the new stabilization procedure into the computer
code is reasonably economic and can well be made use of in many programmes,
especially those in which nearly explicit solution is going to be used.
REFERENCES
Bazant Z. P. and Lin F. B. (1988) Non-local yield limit degradation. Int. J. Nunz. Metl~.Eng..
26. 1805-1 823.
Belytschko T. and Tabarrok M. ( 1 993) H-adaptive finite element methods for dynamic problems
with emphasis on localization, Int. J. Num. Metll. Eng., 36, 42454265.
Belytschko T.. Fish J. and Englemann B. E. (1988) A finite element with embedded localization
zones, Comjl. Metl~.Appl. Mec11. Eng., 79, 59-89.
Brezzi F. and Pitkaranta J. (1984) On the stabilization of finite element approximations of the
Stokes problem in Efi'cient solutions uf elliptic prohlenn. Notes on Nzrmerical Fluid
Mechanics, Vieweg, Wiesbaden.
332
REFERENCES
333
Wolf J. P. and Song C. (1996) Finite Element Modelling o f Unbounded Media, John Wiley &
Sons, Chichester.
Zienkiewicz 0.C. and Codina R. (1995) A general algorithm for compressible and incompressible flow. Part I: The split characteristic based scheme, Int. J. Nurn. Metk. in Fluids, 20.
869-885.
Zienkiewicz 0. C. and Huang G . C. (1990) A note on localization phenomena and adaptive
finite element analysis in forming processes, Comm. Appl. Num. Meth., 6, 71-76.
Zienkiewicz 0.C. and Newton R. E. (1969) Coupled Vibrations of a Structure Submerged in a
Compressible Fluid, Proc. Int. Symp. on Finite Element Techniques, Stuttgart, pl-15, May.
Zienkiewicz 0. C. and Taylor R. L. (1989) The Finite Element Method - Volunze I : Basic
Formulation and Linear Problems (4th edn), McGraw-Hill Book Company, London.
Zienkiewicz 0. C. and Taylor R. L. (1991) The Finite Element Method - Volume 2: Solid and
Fluid Mcd7cmics, Dynamics und Non-linearity (4th edn), McGraw-Hill Book Company,
London.
Zienkiewicz 0.C. and Wu J. (1991) Incompressibility without tears! How to avoid restrictions
of mixed formulations, Int. J. Nunz. Meth. Eng., 32, 1184-1203.
Zienkiewicz 0. C. and Zhu J. Z. (1992) The superconvergent patch recovery (SPR) and
adaptive finite element refinement, Comp. Meth. Appl. Mech. Eng., 101, 207-224.
Zienkiewicz 0. C., Bicanic N. and Shen F. Q. (1987a) Single step averaging generalized Smith
boundary - (a) transmitting boundary for computational dynamics in Proc. Int. Conf Nurn.
Mrth. Eng.: Theor,y crnd Applications ( N U M E T A 87) (eds. Pande G. N. and Middleton J.)
Vol. 11, paper T4911, Swansea, 6-10 July 1987, Martinus Nijhoff Publishers, Dordrecht.
Zienkiewicz 0. C., Clough R. W. and Seed H. B. (1987b) Earthquake analysis procedures for
dams, CIGB ICOLD Bulletin 52.
Zienkiewicz 0. C.. Bicanic N. and Shen F. Q. (1988) Earthquake input definition and the
transmitting boundary condition in Advtrnces in Conl~~urutioncrl
Nonlinear. Mechanics, I . St.
Doltsinis (ed.), Springer-Verlag, 109-1 38.
Zienkiewicz 0.C., Huang M., Wu J. and Wu S. (1992) A new algorithm for coupled soil-pore
fluid problem, Shock and Vibration, 1, 215-233.
Zienkiewicz 0. C., Huang M., Wu J. and Wu S. (1993) A new algorithm for the coupled soilpore fluid problem, Shock und Vibration, 1, 3-14.
Zienkiewicz 0. C., Huang M. and Pastor M. (1995a) Localisation problems in Plasticity using
finite elements with adaptive remeshing, Int. J. Num. Anal. Geomech., 19, 127-148.
Zienkiewicz 0. C., Pastor M . and Huang M . (1995b) Softening, localisation and adaptive
remeshing. Capture of discontinuous solutions, Con~p.Mech., 17, 98-109.
Zienkiewicz 0. C., Boroomand B. and Zhu J. Z. (1998) Error Estimate and adaptivity using
recovery procedure. Comm. Aj~pl.Nurn. Meth.
Zienkiewicz 0.C. and Xie Y. M. (1991) Analysis of the lower San Fernando dam failure under
earthquake, Dum Engineering, 2, 307-322.
Zienkiewicz 0. C. and Wu J. (1994) Automatic directional refinement in adaptive analysis of
compressible flows, Int. J. Nurn. Meth. Eng., 37, No. 13, 2189.
9.1 INTRODUCTION
This volume on computational geomechanics would not be complete without a
description of the finite element implementation. However, due to space limitation,
only a brief introduction to the finite element program DIANA-SWANDYNE 11, or
simply, SWANDYNE I1 could be described. A limited (not more than 50 elements)
executable version of the program together with
1. pre-processing program: DYNMESH
2.
3.
4.
have been made available on the World Wide Web at URL: http://www.bham.ac.uk/
CivEngIswandyne as well as the associated example data and result files. Also
available is a limited version of the executable of GLADYS-2E which is an explicit
implementation of the u-w formulation as described in Chapter 3.
9.2
336
337
Plotting can be done with all the nodal values of the mesh. This includes displacement, pressure, velocity, acceleration and rate of change of pressure. Effective
stress state, incremental strain and pore pressure can be given at the gauss
points. Acceleration, absolute and relative displacement, total and excess pore
pressure can also be plotted from anywhere in the spatial domain. Deformed
shape, pore pressure contour, stress distribution and plastic development can
also be traced. Basic mesh plotting facilities are available and the program is
accompanied with a post-processing program to handle result processing and plotting.
There are quite a number of constitutive models currently available in the program
including:
Elastic family: Linear Elastic model, Anisotropic Elastic model (Graham and
Houlsby 1983), Elastic model with Moduli varying with mean effective confining pressure and with a Coulomb friction envelope.
Classical Elasto-plastic family: von Mises, Tresca, Non- associative elastoperfectly plastic Mohr Coulomb and Drucker Prager model (Owen and Hinton
1980), CAP model (Chen and Mizuno 1990; Sander and Rubin 1979)
Cam-Clay family: Original Cam-Clay and modified Cam-Clay
Kinematic hardening family: Al-Tabbaa and Wood model (Al-Tabbaa and
Wood 1989, Al-Tabbaa 1987), Two surface kinematic hardening plane strain
model for sand (Hamilton 1997, Hamilton et a1 1998).
Generalised plasticity model: Pastor-Zienkiewicz Mark-I11 model (Pastor and
Zienkiewicz 1986)
9.3.1
The top two level of the program has been given in Figure 9.1.
DYNE11 is the dummy program unit. It gives the version number and copyright message
before calling the main subroutine DYNMAIN.
338
- Figure 9.2
- Figure 9.3
- Figure 9.4
DYNMAIN is the main calling subroutine in the program. As shown in Figure 9.1, it calls
four major subroutines to execute the various functions of the program in the
following order.
INTIAL checks the hardware and software platform the system is running on. It finds
out the machine constants for the software platform. It also checks if the system
stores arrays in columnwise format as assumed in the program. The names of
files for input and output purposes are also determined. It reads from the
DATA file the title and user information about the run up to the line starting with
EXEC.
INDATA inputs the main control data and material information for the program
from the DATA file such as the type of earthquake input and whether or not the
global solution matrix is symmetric. Also keywords such as STATIC, CONSOLIDATION or DYNAMIC are used to specify the analysis intended. The major subroutines called by INDATA are given in Figure 9.2 and are described in Section 9.3.2.
INMESH inputs the finite element mesh information from DATA and MESH file. The
major subroutines called by INMESH given in Figure 9.3 and are described in
Section 9.3.3.
TMSTEP is the main analysis routine which controls the whole time stepping and
iteration process. The major subroutines called by TMSTEP are given in Figure 9.4
and described in Section 9.3.4.
9.3.2
MODNAM
DAMPNM
339
MODNMI
ICONSTI
- DAMPMD
- Figure 9.8
DAMPDF
IDAMPEL
PERMFN
DAMPNM reads in the name of the damping model to be used and finds out its
corresponding internal model number. It then calls subroutine DAMPMD with
option 1 to read in the damping model data.
DAMPMD is the standard interface for damping models. It is called and calls to various
damping models with the same set of argument. Four functions are expected from the
damping models:
2 . k cc e3/l
340
- GNSIDl
7: GENNOD
GENFND
- NFIND1
CHECK9
CHECK4
GETPRE
CHECK6
GETPLT
WRITEP
FDIMEN
CHECK5
Figure 9.3
EQNORD
CKCONN
MINEQN
CKCON2
SLOAN l
-CKCONl
CHECK7
- CHKPLT
CHECK1 deduces the phase of each node using the element connectivity and the number
of nodes in each phase for each type of elements. There are five phases used in
the program: 1 for solid with 2 dofs, 2 for fluid with 1 dof, 3 for temperature with
1 dof, 4 for rigid block with 3 dofs and 5 for nodes not referenced therefore with 0
dof.
CHKTIE checks if the tied nodes are of the same phase.
GNSIDE generates mid-side node. It can also generate higher order elements from a
lower order mesh e.g. 8-noded quadrilaterals from 4-noded ones.
GNSIDl keeps track on the mid-side node number generated for each element side.
GETNOD gets nodal coordinates from the MESH file.
GETRGP gets material properties of the rigid block and generates the area of the rigid
block if necessary.
CHKNOD generates nodal coordinate of mid-side nodes and other phases of the same
element. For example, if only the coordinates of the solid corner nodes of a 8-4
element are specified, the coordinate of the mid-side nodes of the solid phase and of
341
all the nodes of the fluid phase will be generated. The order of precedence is solid,
then fluid and lastly temperature.
GENNOD generates coordinates for mid-side nodes which are not specified.
GENFND generates coordinates for nodes of other phases if they are not specified.
CHKRIG checks if the nodes connected to the rigid block are of solid phase.
NFINDl returns the other tied node of a pair given one of them, zero if the node is not
tied.
CHECK2 checks which of the phases are present in the analysis.
GETBOU gets boundary condition code from data file.
CHECK3 determines equation number for each dof in the global equation system.
CHECKS performs profile minimisation
CHECK9 finds the minimum profile length using either the various Cuthill-McKee
methods or the Sloan method (Sloan 1989, Sloan and Ng 1989).
EQNORD rearranges the equation numbers so that the solid nodes are eliminated before
the fluid nodes of the same element. This is done to avoid excessive numerical error.
CKCONN prepares the connectivity list for various Cuthill- McKee methods in
MINEQN.
CKCONl helps subroutine CKCONN to create the connectivity list.
MINEQN minimises the profile length using one of the following methods:
1. Cuthill-McKee method
2. Reversed Cuthill-McKee method
3. Modified reversed Cuthill-McKee method
CKCON2 prepares the connectivity list for the Sloan's method in subroutine SLOANl.
SLOANl (originally called GRAPH) minimises profile length by minimising the diameter
of the linked list as given in Sloan (1989) and Sloan and Ng (1989). The Fortran 77
coding has been obtained from the author.'
CHECK7 forms the profile index array for each active equations.
CHECK4 forms the element index array from the global profile.
GETPRE gets time-independent and time-dependent prescribed values from the DATA
file. If the dof involved is fixed, the value is used as prescribed and if the dof is free,
the value is used as (generalised) force.
'
The subroutines are free for academic usage although a nominal charge will be made for commercial
usage. Please contact Dr. Scott Sloan, Department of Civil Engineering, University of Newcastle, NSW
2308, Australia for more information. His email at the time of writing is cesws @clod.newcastle.edu.au.
342
CHECK6 checks the boundary code, time-independent and time- dependent prescribed
values for the tied nodes to see if they are consistent.
GETPLT gets information for the plotting output file DISP which records the time
history at various user-specified locations.
CHKPLT checks the plotting informations, calculates and stores the relevant information
such as the local coordinate of time history points that do not coincide with a node.
WRTEP writes the mesh and other relevant information into the plotting output file
PLOT.
FDIMEN generates the location of major arrays on the global integer array and double
precision array.
CHECKS checks the boundary code, time-independent and time- dependent prescribed
values for the all nodes to see if they are consistent.
343
PRCONV
FMCART
GETINT
CALINT
7: CONSTI
- CMPROP
- WRTMES
- WRTPLT
- TMCOEF
- FORMGL
FMELOl
FORMEL
ASSEMB
DATRIA
- CALINTl
- Figure 9.8
- Figure 9.5
- Figure 9.6
- INFLIG
- TFBOUN - TMVALU
- PREACC
- UPDISP
- CMPINC
- RESIDU
- TRNFRC
- CHKREl
- DASOLN
- INCRDP
- CONVER
- EARTHQ
- FDERIV
- Figure 9.7
- RFNORM
- CHKRE2
- QNBFGS
- OUTCRK
- RESIDl
- UPPROP
- UPCORD - CHKNOD
- ADVALI - ADVALC
-
NADVAL
ADVALP
ADVARE
ADVLN2
SAVFIN
- RESID2
DATRIA performs triangular decomposition for a matrix stored in a profile form. This is
adopted from Zienkiewicz and Taylor (1989).
INFLIG opens, reads then closes an in-flight command data file. For some multitasking
operating system, the user would be able to supply a command in the data file leading
such as to a soft termination for the analyses.
TFBOUN calculates the multplicative coefficient for time- dependent prescribed values.
TMVALU calculates piece-wise linear multiplicate coefficient for subroutine TFBOUN.
PREACC obtains the prescribed solid acceleration for dynamic analysis.
EARTHQ calculates the current acceleration for each direction
UPDISP updates the current displacement.
344
CMPINC calculates currentlincremental quantities such as gradient of current displacement and incremental temperature.
FDERIV calculates Cartesian derivatives for the shape functions from the derivatives in
local coordinates.
RESIDU see Section 9.3.5.3
TRNFRC transforms residual force vector from containing all dofs to just the active, nonrestraint, dofs. For tied nodes, only the master dofs are used. Round-off errors are
removed if the residual force vector is close to zero.
RFNORM calculates and outputs the error norm for each phase.
CHKREl finds the maximum residual error at each dof for each phase
CHKRE2 performs insert sort for subroutine CHKREl.
DASOLN finds solution to a system of simultaneous equations stored in profile form and
the coefficient matrix already decomposed into triangular form using DATRIA. This
is adopted from Zienkiewicz and Taylor (1989).
INCRDP calculates the incremental displacement (for STATIC), velocity (for CONSOLIDATION) or acceleration (for DYNAMIC) for the current time step depending
on the type of analysis performed.
CONVER checks if the convergence criteria is met within the time step.
QNBFGS calculates the forward and backward transformation using the quasi-Newton
BFGS iterative method. The code has been adapted from Matthies and Strang (1979).
OUTCRK outputs information of cracked Gauss points.
OUTCRZ outputs cracked Gauss point information on the screen.
RESIDl outputs error norm, incremental norm and residual force norm for each phase.
UPPROP calculates the updated value of quantities such as average density due to
change in void ratio during the finite deformation analysis.
UPCORD updates the coordinate for Updated Lagrangian analysis.
CHKNOD see Section 9.3.3
ADVALl interface routine for subroutine ADVALC
ADVALC calculates the minimum and maximum alpha values (a measure of mesh
distortion suggested by Cheung et al., 1996) for the triangular elements in a mesh.
For six-noded elements, the maximum angle of deviation from centre line is also calculated.
NADVAL finds the nodes of the required phase within the connectivity of one element
ADVALP returns the value of alpha measure of quality of triangle
ADVARE returns the area of the triangle.
ADVLN2 returns the square of the distance between two points.
SAVFIN saves the current state in FINL output file. The format is compatible with the
input file INIT. It can be used to obtain the initial state for dynamic analysis from a
static analysis or restarting an analysis.
345
RESID2 outputs a summary of the current analysis such as the average number of
iterations per time step for easy reference.
9.3.5.1 Subroutines for the analytical formation of element matrix for 3-noded
elements (Figure 9.5)
FMELOl forms element matrices for 3-noded element
CONSTI see Section 9.5
DAMPMD see Section 9.3.2
FMASOl forms element mass matrix M for 3-noded element
PERMFN see Section 9.3.2
FMPMOl forms element permeability matrix H for 3-noded element
FMCMOl forms mass part of the element damping matrix C for 3-noded element
FMQMOl forms element coupling matrix Q for 3-noded element
FMGMOl forms element mass coupling matrix G for 3-noded element
CONSTI
FMSTO1
DAMPMD
FMASO 1
PERMFN
FMPMO 1
FMCMO1
FMQMO 1
FMGMO 1
Figure 9.8
346
347
=y
+ ax.
348
DCOPY is the BLAS 3.0 routine for double precision vector operation of y = x.
DDOT is the BLAS 3.0 routine for double precision vector operation of d = x.y.
DMACH obtains the machine constants such as the smallest difference between a double
precision number and unity.
DOTPRD is the former function name for vector dot product. It is now an interface for
DDOT.
FRICTM calculates the friction angle required for the current stress state and the given
cohesion.
GIVALU gives a double precision array a particular value.
GIVINT gives an integer array a particular value.
HlSHAP returns one-dimensional finite element shape function.
H2CORD returns the local coordinates of the nodes for a two-dimensional finite
element.
HZSHAP returns two-dimensional finite element shape functions for triangular and
quadrilateral elements.
HANDLE handles an error condition in the input data and sets a flag so that the
execution would stops after the data has been read in.
IALLOC allocates a section of the global integer or double precision array to
be used.
INVARl calculates stress invariants for plane strain condition.
IOFMTl ensures the same format statement is used in read and write operation for
plotting output file PLOT.
PRINT prints an integer matrix.
IVECTC prints only one digit, i.e. in compact form, for each element of an integer array.
IVECTP prints an integer array.
LOADVC copies one double precision vector into another. This is now only an interface
to DCOPY.
LOWERC converts a character string to lower case.
LSAMEl checks if two character strings are the same under case insensitive condition.
MBTMUL calls BLAS 3.0 routine DGEMM for double precision matrix multiplication.
MPRINT prints a double precision matrix.
MROWPR prints a row of a matrix stored in profile form.
MSTORE stores message summary for one unified output at the end of the execution.
MVECTP prints a double precision vector.
NEOLIN finds the end of text within a character string.
NFACTL returns the factoral function in integer form.
OPENFL is the subroutine to unify the open statements on different computer systems
POLYOZ calculates the various properties of a n-sided polygon including the centroid,
area and the second moment of inertia
349
9.5
Quite a large number of constitutive models have been made available to the
computer program DIANA-SWANDYNE 11. They are all linked to the computer
program via a standard constitutive model interface CONSTI which is described
in Section 9.5.1. Constitutive models available for general dissemination are
described in Section 9.5.2. Other models available are described in Section 9.5.3.
Their dissemination is restricted both by licensing conditions and difficulties of
implementation. Some of them, although connected, have not been thoroughly
tested for general application. Due to limited space, the subroutines called by the
main constitutive model subroutines listed in Sections 9.5.2 and 9.5.3 will not be
shown.
350
Four other functions are prepared but at the time of writing, not, fully implemented:
6.
7.
8.
9.
9.5.2
351
ADJCN4
ADJJIM
ADJMH4
ALTER0
BRICK1
CAPMOD
CJHMOD
CSMOOl
DEPOIN
ELAS3D
ELASGM
ELASTA
EXPERI
HASHIS
MCOULS
MODCAM
NCRIS2
SARAH
SLIP03
STATE2
TABBA2
TSMODl
UNSATS
VONBAC
VONMIS
1. The bulk modulus and the shear modulus can vary with mean effective confining
stress. The variation can be linear, square root or generally nonlinear
2. Cohesion and the Mohr Coulumb friction envelope is also available
Cam-Clay model
352
TABBA2-Two
1989).
NCRIS1-Cristescu saturated sand model (Cristescu 1989 and 1991, Roatesi and Chan
1994, Roatesi 1995, Chan and Roatesi 1998, Roatesi and Chan 1998)
SARAH-Three-surface
TSMOD1-Pietruszczak
TSMAIN.
REFERENCES
353
SYSFUN returns values for various system functions such as system time, CPU time
elapsed, page faults and date.
SETSUF sets the name of file extension.
COMLIN returns the text given on the command line.
ZPSYMB draws a symbol on the screen.
GETKEY gets one key from the keyboard.
DSORTX sorts a double precision array.
TIMEXX returns time and date for the current time.
XPLOT maintains a set of basic plotting primitives such as pen up, pen down and
opening a graphic device.
9.7 REFERENCES
Alonso E. E., Gens A. and Josa A. (1990) A Constitutive Model for Partially Saturated Soils,
GPotechnique, 40, No. 3, 405430.
Al-Tabbaa A. (1987) Permeability and Stress Strain Response of Speswhite Kaolin, Ph.D.
Dissertation, Cambridge University Engineering Department.
Al-Tabbaa A. (1995) Excess Pore Pressure During Consolidation and Swelling with Radial
Drainage, Giotechnique, 45, No. 4, 701-707.
Al-Tabbaa A. and Muir Wood D. (1989) An experimentally based 'bubble' model for clay,
NUMOG 111, Niagara Falls, 91-99.
Britto A. M. and Gunn M. J. (1987) Critical State Soil Mechanics via Finite Elements, Ellis
Horwood Ltd, Chichester.
Chan A. H. C. (1988) A Unified Finite Element Solution to Static and Dynamic Geomechanics
Problems, Ph.D. Thesis, University College of Swansea, Wales.
Chan A. H. C. (1995) User Manual for SM2D - Soil Model Tester for 2-Dimensional Application, School of Civil Engineering, University of Birmingham, December, Birmingham.
Chan A. H. C. and Roatesi S. (1998) Finite Element Approach in Viscoplasticity for Cristescu
Saturated Sand Model, Rev. Roum. Sci. Techn. Mec. Appl., No. 1-2.
Chen W. F. and Mizuno E. (1990) Nonlinear Analysis in Soil Mechanics - Theory and
Implementation, Elsevier, Amsterdam.
Cheung Y. K., Lo S. H. and Leung A. Y. T. (1996) Finite Element Implementation, Blackwell
Science, Cambridge, MA 02142.
Cristescu N. (1989) Rock Rheology, Kluwer, Dordrecht.
Cristescu N. (1991) Nonassociated elastic/viscoplastic constitutive equations for sand, Int. J.
Plasticity, 7 , 41-64.
Gibb (1994) Nonlinear Concrete Model-Theory
and Validation Report, Gibb Ltd,
Reading, UK.
Gibb (1995a) Concrete Construction Joint Interface Model and Validation Report, Gibb Ltd,
Reading, UK.
Gibb (1995b) Nonlinear Soil Model-Theory Report, Gibb Ltd, Reading, UK.
Gibb (1995~)Nonlinear Soil Model-Validation Report, Gibb Ltd, Reading, UK.
Graham J. and Houlsby G. T. (1983) Anisotropic Elasticity of a Natural Clay, GCotechnique,
33, NO. 2, 165-180.
Hamilton C. J. (1997) A Plane Strain Constitutive Model for Sands under Non-monotonic
Loading, Ph.D. Thesis, University of Birmingham, UK.
354
Hamilton C. J., Chan A. H. C. and Muir Wood D. (1998) Dynamic Finite Element Analyses
of Sand: Structure Interaction Using a New Kinematic Hardening Model in Seismic
Design Practice into the Next Century, Booth E. (ed.), A. A. Balkema, Rotterdam,
179-186.
Hashiguchi K. (1988) A Mathematical Modification of Two Surface Model Formulation in
Plasticity, Int. J. Solids & Structures, 24, No. 10, 987-1 001.
Josa A. (1988) An Elastoplastic Model for Partially Saturated Soils, Ph.D. Thesis, ETSICCP,
Barcelona (in Spanish).
Katona M. G. (1985) A General Family of Single-step Methods for Numerical Time Integration of Structural Dynamic Equations, NUMETA 85, l , 213-225.
Katona M. G. and Zienkiewicz 0 . C. (1985) A Unified Set of Single-step Algorithms Part 3:
The Beta-m Method, a Generalisation of the Newmark Scheme, Znt. J. Num. Meth. Engrg.,
21, 1345-1359.
Matthies H. and Strang G. (1979) The Solution of Nonlinear Finite Element Equations, Int. J.
Num. Meth. Engrg., 14, 1613-1626.
Molenkamp F. (1982) Kinematic Model for Alternating Loading ALTERNAT, LGM Report
Co-218598, Delft, NL.
Molenkamp F. (1987) Elasto Plastic Model for Analysis of Liquefaction Under Alternating
Loading in Workshop on Constitutive Laws for the Analysis of Fill Retention Structures,
Ottawa.
Molenkamp F. (1990) Reformulation of ALTERNAT to Minimise Numerical Drift Due to
Cyclic Loading, University of Manchester Internal Report, Manchester, UK.
Molenkamp F. (1992) Application for Non-linear Elastic Model, Int. J. Num. Anal. Geomech.,
16, 131-150.
Muir Wood D., Belkheir K. and Liu D. F. (1994) Technical Note - Strain Softening and state
parameter for sand modelling, Gotechnique, 44, No. 2, 335-339.
Newrnark N. M. (1959) A Method of Computation for Structural Dynamics, Proc. ASCE, 8,
67-94.
Owen D. R. J. and Hinton E. (1980) Finite Elements in Plasticity-Theory and Practice,
Pineridge Press, Swansea, Wales.
Pastor M. and Zienkiewicz 0. C. (1986) A Generalised Plasticity Hierarchical model for Sand
Under Monotonic and Cyclic Loading, NUMOG 11, Ghent, April, 131-150.
Roatesi S. (1995) Finite Element Approach in Viscoplasticity, Technical Report, No. 887B,
Technology and Research Ministry, Bucharest, Romania.
Roatesi S. and Chan A. H. C. (1994) Numerical integration of a viscoplastic constitutive
equation for geomaterials. Comparison with the analytical solution for step creep in The
XVIII National Conference of Solid Mechanics, Brasov, Romania, 31-38.
Roatesi S. and Chan A. H. C. (1998) Comparison of Finite Element Analysis and Analytical
Solution for Underground Openings Problems in Viscoplastic Rock Mass, Rev. Roum. Sci.
Techn. Mec. Appl., No. 1-2.
Sandler I. S. and Rubin D. (1979) An Algorithm and a Modular Subroutine for the Cap
Model, Znt. J. Num. Anal. Geomech., 3, 173-186.
Simpson B. (1992a) Development and Application of a New Soil Model for Prediction of Ground
Movements, The Wroth Memorial Symposium-Predictive Soil Mechanics', St. Catherine's
College, Oxford, 27-29 July, 628-643.
Simpson B. (1992b) Retaining Structures: Displacement and Design, GPotechnique, 42, No. 4,
541-576.
Sloan S. W. (1989) A FORTRAN Program for Profile and Wavefront Reduction, Int. J. Num.
Meth. Engrg., 28, 2651-2679.
Sloan S. W. and Ng W. S. (1989) A Direct Comparison of Three Algorithms for Reducing
Profile and Wavefront, Comp. Struct., 33,411-419.
355
Stallebrass S. E. (1990) Modelling the Effect of Recent Stress History on the Deformation of
Overconsolidated Soils, Ph.D. Dissertation, City University, London, England.
Taylor D. W. (1948) Fundamentals of Soil Mechanics, John Wiley and Sons, Inc., New York.
Whitman R. V. (1953) After Marcuson (1995): An Example of Professional Modesty, The
Earth. Engineers and Education, MIT, 200-202.
Zienkiewicz 0 . C. and Taylor R. L. (1989) The Finite Element Method-Volume I: Basic
Formulation and Linear Problems (4th edn), McGraw-Hill Book Company, London.
ICDAT,ICOUT,DSTRE
l,DSTAN,DMATX,ESTRE,PARAM,LPARA, IELEM, I G A U S , V O I D R , I C P R T
2,KTEST,NTEST)
C....
C
ELASTICCONSTITUTIVEMODELFORVERIFICATION
L....
INTEGER NSTRE
PARAMETER ( N S T R E = 4 )
The subroutine name EXPERD should be used for material model development. Using such
would not require the change in the subroutines CONSTI and MODNMl until the model is
ready for production.
9A.1.1
Variable NSTRE
NSTRE is the number of stress components in the analysis. For the two-dimensional analysis,
d, and T,,. In the case
the number of stress components is four. They are ordered as: d,,, dyy,
of three-dimensions, two more stress components are required: T,, and T,,. The introduction of
parameter NSTRE is to ease the changeover from two-dimensional to three-dimensional stress
state.
356
9A.2.1
INPUT VARIABLES
INTEGER IELEM
INTEGER I C P R T
INTEGER IGAUS
INTEGER ISWDP
INTEGER LPRPD
INTEGER ICDAT
INTEGER ICOUT
INTEGER NTEST
I N T E G E R LPARA
d o u b l e p r e c i s i o n D S T A N (NSTRE)
DOUBLE P R E C I S I O N V O I D R
Variable IELEM
IELEM is the current element number. This is just for information. If it is desirable to trace the
progress of a particular Gauss point, this could be useful.
9A.2.2
Variable ICPR T
ICPRT is the unit number for debugging or information output that can be massive. Unit
ICOUT should not be used for this purpose.
9A.2.3
Variable IGA US
IGAUS is the current Gauss point number. This is just for information. If it is desirable to
trace the progress of a particular Gauss point, it could be useful. The Gauss point is counted
for each element i.e. it is the local Gauss point number.
9A.2.4
Variable ZSWDP
ISWDP is the controlling switch and its function will be fully described later. However, here is
a brief overview of its capabilities
1.
INPUT O N L Y VARIABLES
357
5.
Initialise internal variables, check current stress state. For initial stress state calculated
according to the Gauss point coordinate GPCOD, it can be implemented with a flag in the
propd array, see e.g. subroutine TABBA2.
6. Return values for optimisation
7. Output internal values to unit icout
9A.2.5
Variable LPRPD
The dimensions of the material property array. All of the elements from PROPD(1) to
PROPD(LPRPD) are available in each material.
9A.2.6
Variable ICDA T
9A.2.7
Variable ICOUT
The unit number for echoing the material data input. Minor output (less than two lines) per
entry is acceptable. Monitoring of a Gauss point can be done with PRINT statement.
9A.2.8
Variable NTEST
9A.2.9
Variable LPARA
The dimension of the soil model internal parameter array. All elements from PARAM(1) to
PARAM(LPARA) are available at each Gauss point.
9A.2.11
Variable VOIDR
358
9A.3
Variables and arrays described in this section will carry information into the subroutine or out
of the subroutine.
C.
...
INPUT/OUTPUTVARIABLES:
INTEGERKTEST (NTEST)
d o u b l e p r e c i s i o n PROPD ( L P R P D )
d o u b l e p r e c i s i o n E S T R E (NSTRE)
d o u b l e p r e c i s i o n P A R A M (LPARA)
This is the soil model information array. They are defined as follow:
1. KTEST(1) Valid only for ISWDP = 2. 0 indicates this is a trial incremental, maybe using
it to determine the number of sub-increments. 1 indicates this increment may be taken as
the accepted increment
2. KTEST(2) The type of subdivision that is active (variable NSUBV)
3. KTEST(3) The number of times MDIVT is exceeded
4.
KTEST(4) Number of warnings from soil models. If there is a warning state in the
current model, this can be incremented by one. For a serious error, STOP statement
can be used.
KTEST(7) Soil model error counter B: Usually used to denote allowable stress ratio has
been exceeded and remedial action is needed. If such a case occurs, increment KTEST(7)
by one.
8. KTEST(8) Soil model error counter C: Usually used to denote CONSl = nTD, . ng is less
than zero and remedial action is needed. If such a case occurs, increment KTEST(8) by
one.
9.
KTEST(9) Soil model error counter D: Usually used to denote the plastic modulus
CONSl is less than zero and remedial action is needed. If such a case occurs, increment
KTEST(9) by one.
The programmer of new soil model is free to redefine any or all of these soil model error
counters. Please use PRINT statement to indicate your definitions.
359
For options ISWDP = 2, 3,4, 5. this contains the current effective stress state (tension positive)
which enters. For option ISWDP = 2, it should contain the incremented stress state which
exits. The value of ESTRE should not be altered for iswdp = 3 and 4. For iswdp = 2, 3 and 4.
the ESTRE should either be the output value of iswdp = 5 or inputloutput value of iswdp = 2.
The array for internal parameters. Different PARAM(LPARA) will be available for each
Gauss point. The subroutine should initialise the values within this array for iswdp = 5 and
check if the array is long enough for its purpose. The value of PARAM should not be altered
for iswdp = 3 and 4. For iswdp = 2, 3 and 4, the value in PARAM should either be the output
value of iswdp = 5 or inputloutput value of iswdp = 2. A check should be used within the
routine so that iswdp = 5 will be called automatically if the values in PARAM have not be
initialised.
OUTPUTVARIABLES:
d o u b l e p r e c i s i o n D S T R E (NSTRE)
d o u b l e p r e c i s i o n D M A T X (NSTRE,NSTRE)
When ISWDP = 2, DSTRE(NSTRE) should contain the incremental stress i.e. ESTRE(final)-ESTRE(initia1) when exits.
360
9A.4.2
END DO
END DO
9A.4.3
Local Variables
Any number of local variables can be defined in the subroutine, although large arrays are not
recommended. Except for very good reasons, COMMON should not be used. If it were used, it
must not retain values that are specific to a Gauss point. Note, value of all the local variables is
not necessarily retained for the next entry of the soil model nor retained for different Gauss
points in the same material region. All real constants should be given in double precision also
to preserve accuracy.
C....
C..
..
9A.5
LOCAL VARIABLES
LOCALDEBUG
doubleprecisionYOUNG
doubleprecisionPOISS
doubleprecisionRLMDA
double precision SHEAR
INTEGER ICOUN
double precision REAL1, REALO, SVOLV, DVOLV
PARAMETER (REAL1=1.OdO,REALO=O.OdO)
doubleprecisionREAL2
PARAMETER (REAL2=2.OdO)
double precision REAP5
PARAMETER (REAP5=0.5DO)
CHECK
IF (DEBUG)WRITE (ICOUT,") 'VOIDRATIO:',VOIDR
ISWDP BRANCHING
The simpliest way of branching for ISWDP is to use a computed G O TO. It is not
advisable, however, to jump from one option to another within the code unless it is a very
IS W D P BRANCHING
361
straight forward jump. The computed GO TO is used in the same way as the CASE statement
in Fortran 95.
An error should be given if ISWDP is outside the range and CALL TERMIN will terminate
the program after issuing a warning
C. ..
ERROR CONDITION
WRITE (ICOUT,*) 'ISWDP:', ISWDP
CALLTERMIN ('ELASTI-1:UNKNOWNOPTION:')
GOT0 2999
2001
READ INMATERIALDATA
CONTINUE
READ(ICDAT,*,ERR=3100,END=3200) YOUNG,POISS
RLMDA=YOUNG*POISS/( G REAL^-REAL~*POISS)*(REAL~+POISS))
SHEAR=YOUNG*REAP5/(REALl+POISS)
WRITE (ICOUT,2001) YOUNG,POISS,RLMDA, SHEAR
FORMAT ( ' YOUNG:', E11.3, 'POISS:', E11.3
,/, 'RLMDA:', E11.3, 'SHEAR:', E11.3)
1
PROPD(1) =RLMDA
PROPD (2) = SHEAR
GOT0 2999
Note that the input data is echoed. Calculated values are stored in the array
362
C....
SYMMETRICVERSIONOFDMATX
CONTINUE
C....
= 4.
COMMON EXIT
363
PARAM(1COUN) = R E A L 0
END DO
GOT0 2 9 9 9
9A.6
COMMON EXIT
It is advisable to use a common exit for all options. This is especially for debugging purpose.
C . . ..
2999
COMMONEXIT
CONTINUE
RETURN
3200
ERRORCONDITION
CONTINUE
WRITE ( I C O U T , * ) ' I C D A T : ' , I C D A T
C A L L H A N D L E ('ELASTI-2:ERRORINCHANNEL1)
RETURN
CONTINUE
WRITE ( I C O U T , * ) ' I C D A T : ' , ICDAT
CALLHANDLE ( ' E L A S T I - 3 : E N D O F F I L E I N C H A N N E L : ' )
RETURN
END
1. porep
2. dpore - change in pore pressure, incremental pore pressure applied with dstan
3. gpcod Gauss point coordinate
-
364
9A.8.2
Output to screen
A common variable ICHPC is used for the output to the screenlfile determined by the calling
program.
INTEGER ICHPC
COMMON /PCOMOl/ ICHPC
The value ICHPC should not be changed. Direct output to the screen using PRINT statement
or WRITE(6,fmt) should be avoided except for debugging purpose.
9A.8.4
A common block was introduced to relay time stepping, iteration and linear elastic step
information to the soil model routine:
integeritime, i i t e r , k e l a s
common / t m s t p l / i t i m e , i i t e r , k e l a s
C.. . .
C
C.. . .
VONMISESELASTOPLASTICMODELWITHLINEARHARDENING
365
updatedon27/3/1991for amistakeinepstn
..
updatedon15/10/1992toconformwith~alford~ortrancompilercheckingstandard
C.. . .
C....
C.. .
C....
C....
C....
C....
C....
C....
IMPLICIT NONE
INTEGER NSTRE
PARAMETER (NSTRE = 4)
INPUT VARIABLES
INTEGER IELEM, ICPRT
INTEGER IGAUS
INTEGER ISWDP
INTEGER LPRPD
INTEGER ICDAT
INTEGER ICOUT, NTEST
INTEGERLPARA, KTEST(NTEST)
DOUBLEPRECISIONDSTAN(NSTRE),VOIDR
INPUT/OUTPUTVARIABLES:
DOUBLEPRECISIONPROPD(LPRPD),ESTRE(NSTRE),PARAM(LPARA)
OUTPUTVARIABLES:
DOUBLEPRECISIONDSTRE(NSTRE),DMATX(NSTRE,NSTRE)
LOCALVARIABLES
LOGICAL DEBUG
PARAMETER (DEBUG= .FALSE.)
! YOUNG'S MODULUS
DOUBLE PRECISIONYOUNG
! POISSON'S RATIO
DOUBLEPRECISIONPOISS
! LAME ' S CONSTANT
DOUBLE PRECISIONRLMDA
! SHEAR MODULUS
DOUBLE PRECISION SHEAR
DOUBLEPRECISIONFSTRE (NSTRE)
DOUBLEPRECISIONDEVIA (NSTRE)
DOUBLEPRECISIONYIELD
DOUBLEPRECISIONCURJ2
DOUBLE PRECISIONREQJ2
INTEGER ISTRE, JSTRE , IPARA
INTEGER ISTEP, MSTEP
DOUBLE PRECISIONRATIO
DOUBLE PRECISIONHMODU
DOUBLEPRECISIONCONSl, CONS2
DOUBLE PRECISIONPMEAN
DOUBLE PRECISIONSMALL
PARAMETER ( SMALL = 1. OD 10)
DOUBLE PRECISIONvsmal
PARAMETER (vsmal= 1. Od-30)
INTEGER ICOUN
DOUBLEPRECISIONACOEF,BCOEF, CCOEF
DOUBLEPRECISIONEPSTN (NSTRE)
DOUBLE PRECISIONREALl,REALO, SVOLV, DVOLV
PARAMETER (REAL1= 1.ODO, REAL0 = 0. ODO)
DOUBLE PRECISIONREAL2
PARAMETER (REAL2= 2. OD0 )
DOUBLE PRECISIONREAP5
PARAMETER (REAP5= 0. 5DO)
CHECK
IF (DEBUG)WRITE (ICOUT,* ) 'VOIDRATIO:',VOIDR
366
. . . . ERROR C O N D I T I O N
p r i n t * , ' e x a m p l e e x p e r i .f o r '
i f ( p o i s s . l t . O . O d 0 ) then
poiss = -poiss
p r i n t *, ' E l a s t i c D - m a t r i x i s u s e d f o r i t e r a t i o n '
w r i t e ( 6 , * ) ' E l a s t i c D - m a t r i x is used f o r i t e r a t i o n '
propd ( 1 1 ) = l . O d O
else
propd ( 1 1 ) = O . O d O
end i f
RLMDA = YOUNG * P O I S S / ( ( R E A L 1 REAL^ * P O I S S ) * REAL^ + POISS)
-
S H E A R = YOUNG * R E A P 5 / ( R E A L 1 + P O I S S )
C U R J 2 = Y I E L D * Y I E L D / 3.ODO
WRITE ( I C O U T , 2 0 0 1 ) YOUNG,POISS,HMODU,YIELD,RLMDA, SHEAR
2 0 0 1 FORMAT ( ' Y O U N G : ' , E 1 1 . 3 ,
'POISS:',E11.3
1
,/, 'HMODU: ' , E 1 1 . 3 , ' Y I E L D : ' , E 1 1 . 3
2
,/, ' R L M D A : ' , E 1 1 . 3 , ' S H E A R : ' , E 1 1 . 3 )
\
P R O P D ( 1 ) = RLMDA
PROPD ( 2 ) = SHEAR
P R O P D ( 3 ) =HMODU
PROPD ( 4 ) = Y I E L D
PROPD ( 5 ) = CURJ2
GOT0 2 9 9 9
C . . . . FORM T H E I N C R E M E N T A L S T R E S S
1200
CONTINUE
C
P R I N T * , I E L E M , I G A U S , DSTAN
C . . . . E L A S T I C INCREMENT
RLMDA=PROPD ( 1 )
SHEAR=PROPD ( 2 )
HMODU=PROPD ( 3 )
C U R J 2 = PARAM ( 1)
C . . . . VOLUMETRIC P A R T O F THE I N C R E M E N T , I N VON M I S E S , T H I S I S ALWAYS
C
ELASTIC
DVOLV=DSTAN ( 1 ) +DSTAN ( 2 ) +DSTAN ( 3 )
S V O L V = R L M D A * DVOLV
C....
T H I S ISTHEDEVIATORICCOMPONENTOFTHE I N C R E M E N T A L S T R E S S
DSTRE ( 1 ) = R E A L 2 * SHEAR*DSTAN ( 1 )
D S T R E ( 2 ) = R E A L 2 * SHEAR * DSTAN ( 2 )
DSTRE ( 3 ) = REAL2 * SHEAR * DSTAN ( 3 )
C . . . . S I N C E ENGINEERING SHEAR S T R A I N I S USED
DSTRE ( 4 ) =SHEAR*DSTAN ( 4 )
C . . . . CHECK CURRENT Y I E L D S T R E S S
PMEAN=-(ESTRE ( 1 ) +ESTRE ( 2 ) +ESTRE ( 3 ) ) /3.ODO
DEVIA ( 1 ) = E S T R E ( 1 ) +PMEAN
367
D E V I A ( 3 ) = E S T R E ( 3 ) +PMEAN
D E V I A ( 4 ) =ESTRE ( 4 )
PMEAN=PMEAN-SVOLV
C . . . . FORM THE C O E F F I C I E N T S
ACOEF=DSTRE ( 1 ) *DSTRE ( 1 ) +DSTRE ( 2 ) *DSTRE ( 2 )
1+2.ODO*DSTRE ( 4 ) "DSTRE ( 4 ) + D S T R E ( 3 ) *DSTRE ( 3 )
B C O E F = D E V I A ( 1 ) * DSTRE ( 1 ) + D E V I A ( 2 ) * D S T R E ( 2 )
1+2.ODO*DEVIA ( 4 ) "DSTRE ( 4 ) +DEVIA ( 3 ) *DSTRE ( 3 )
BCOEF = 2.ODO * BCOEF
C C O E F = D E V I A ( l ) * D E V I A ( l ) +DEVIA ( 2 ) * D E V I A ( 2 )
1 + 2 . O D O * D E V I A ( 4 ) * D E V I A ( 4 ) +DEVIA ( 3 ) * D E V I A ( 3 )
CCOEF = C C O E F - 2 . O D 0 * C U R J 2
I F ( a b s ( A C O E F ) . l e . v s m a l ) THEN
RATIO = 1.ODO
ELSE
RATIO= (-BCOEF+SQRT (BCOEF*BCOEF-~.ODO*ACOEF*CCOEF))
1
/ (2.ODO"ACOEF)
RATIO=MIN (RATIO, 1.ODO)
END I F
C . . . . HOW FAR OUT I S THE S T R E S S P O I N T ?
R E Q J 2 = 0 . 5 D O * ( ( D E V I A (1)+ D S T R E ( 1 ) ) * * 2
+ (DEVIA ( 2 ) +DSTRE ( 2 ) ) * * 2
1
+ (DEVIA ( 3 ) +DSTRE ( 3 ) ) * * 2
2
3
+ 2 . 0 D O k (DEVIA ( 4 ) + D S T R E ( 4 ) ) * * 2 )
C . . . . UPDATE THE S T R E S S S T A T E
DEVIA ( 1 ) = D E V I A ( 1 ) + R A T I O k D S T R E ( 1 )
DEVIA ( 2 ) = D E V I A ( 2 ) + R A T I O k D S T R E ( 2 )
DEVIA ( 3 ) = D E V I A ( 3 ) + R A T I O k D S T R E ( 3 )
DEVIA ( 4 ) = D E V I A ( 4 ) + R A T I O k D S T R E ( 4 )
EPSTN ( 1 ) = (1.ODO-RATIO)*DSTAN ( 1 )
EPSTN ( 2 ) = (1.ODO-RAT1OIXDSTAN ( 2 )
EPSTN ( 3 ) = (1.ODO-RATIO)*DSTAN ( 3 )
E P S T N ( 4 ) = (1.ODO-RATIO)*DSTAN ( 4 )
C . . . . THE NUMBER O F I N T E G R A T I O N S T E P S
RATIO=SQRT (REQJ2 /CURJ2)
MSTEP=MAX ( I N T (lO.ODO* ( R A T I O - O . 9 0 0 0 0 0 0 0 0 0 0 1 ~ O ) ) , 0 )
c ..
c o r r e c t i o n on 2 7 / 3 / 1 9 9 1
1 2 4 1 i s t r e = 1, n s t r e
do
epstn ( i s t r e ) =epstn ( i s t r e ) / d b l e (max ( 1 , m s t e p ) )
1241
continue
c . . . . e n d of c o r r e c t i o n
C
P R I N T *, 'MSTEP: ' ,MSTEP, R E Q J 2 , CURJ2
C . . . . L O O P OVER THE NUMBER O F S T E P S
..
DO
1250ISTEP=l,MSTEP
DSTRE ( 1 ) = 2 . O D O * SHEAR* EPSTN ( 1 )
DSTRE ( 2 ) = 2 . O D O * S H E A R * E P S T N ( 2 )
DSTRE ( 3 ) = 2 . O D O * S H E A R * E P S T N ( 3 )
SHEAR * E P S T N ( 4 )
DSTRE ( 4 ) =
C O N S 2 = D E V I A ( 1 ) *DSTRE ( 1 ) + D E V I A ( 2 ) "DSTRE ( 2 )
2 . 0 D O k D E V I A ( 4 ) " D S T R E ( 4 ) + D E V I A ( 3 ) "DSTRE ( 3 )
CONSl=CONS2*2.0DO*SHEAR/ ((HMODU+~.ODO*SHEAR)
*2.ODO*CURJZ)
I F ( D E B U G ) PRINT * , 2 . ODO * SHEAR * HMODU
"(DEVIA ( 1 ) *EPSTN ( 1 ) +DEVIA ( 2 ) *EPSTN ( 2 )
368
2
3
1
2
3
369
ANOTHER E X A M P L E IMPLEMENTATION
C..
DMATX ( 4 , 3 ) = R E A L 0
DMATX ( 4 , 4 ) = S H E A R
CHECK C U R R E N T Y I E L D S T R E S S
PMEAN=-(ESTRE ( 1 ) + E S T R E ( 2 ) + E S T R E ( 3 ) ) /3.ODO
DEVIA ( 1 ) = E S T R E ( 1 ) +PMEAN
DEVIA ( 2 ) = E S T R E ( 2 ) +PMEAN
DEVIA ( 3 ) = E S T R E ( 3 ) +PMEAN
DEVIA ( 4 ) =ESTRE ( 4 )
REQJ2=0.5DO* (DEVIA ( 1 ) * * 2 + D E V I A ( 2 ) * * 2
1
+2.ODO*DEVIA(4) **2+DEVIA ( 3 ) ""2)
I F (REQJ2/CURJ2.GT.1.ODO+SMALL) T H E N
P R I N T * , 'REQJ2 > CURJ2 I N V O N M I S : ' , R E Q J 2 , C U R J 2
STOP 'STOPPEDINVONMIS'
END I F
I F (REQJ2/CURJ2.GT.O.99DO.and.abs ( p r o p d ( l l ) ) . l e . v s r n a l ) T H E N
CONS1=2.ODO*SHEAR*SHEAR/
(HMODU+~.ODO*SHEAR)*REQJ~)
DO1420JSTRE=l,NSTRE
1 4 1 0 I S T R E = 1, N S T R E
DO
DMATX ( I S T R E , J S T R E ) =DMATX ( I S T R E , J S T R E )
1
-(CONSl*DEVIA (JSTRE)) *DEVIA (ISTRE)
CONTINUE
CONTINUE
END I F
GOT0 2 9 9 9
C . . . . STORE SOMETHINGINTOPARAM
1500 CONTINUE
D O 1 5 1 0 I P A R A = 1, L P A R A
PARAM ( I P A R A ) = O . O D O
CONTINUE
PARAM ( 1 ) = P R O P D ( 5 )
PARAM ( L P A R A ) = - I 9
GOT0 2 9 9 9
C . . . . COMMON E X I T
2 9 9 9 CONTINUE
RETURN
C . . . . ERROR C O N D I T I O N
3100 C O N T I N U E
WRITE ( I C O U T , * ) ' I C D A T : ' , I C D A T
CALLHANDLE ( ' E L A S T I - 2 : E R R O R I N C H A N N E L ' )
RETURN
CONTINUE
WRITE ( I C O U T , * ) ' I C D A T : ' , ICDAT
CALLHANDLE ( ' E L A S T I - 3 : E N D O F F I L E I N C H A N N E L : ' )
RETURN
END
Author Index
Adkin J. E. 18, 50
Advani S. H. 202, 203. 204, 206, 207, 215
Al-tabbaa A. 337, 350, 352,353
Alonso E. E. 14, 15, 198, 216, 352,353
Anandarajah A. 158, 171
Armero F. 324, 332
Arulanandan K. 5, 15, 218, 230, 231, 247
Arulmoli K. 237, 247
Atkinson J. H. 113, 135, 171
Aubry D. 131 , 159,171
Babuska I. 64, 70, 81
Bacchus D. R . 140, 175
Baggio P. 188, 189, 193, 215
Bahda F. 151, 157, 171
Bailey 184
Baker R. 158, 161, 171
Balakrishna J. 14, 15
Balasubramanian A. S. 134, 135, 138, 139,
171
Balestra M. 324, 332
Banerjee P. K. 158, 172
Bardet J. P. 131, 172
Bathe K. J. 202, 215
Batlle F. 198, 216
Bazant Z. P. 129, 172, 302, 321, 331
Bear J. 14, 15
Belkheir K. 352, 354
Bell 184
Bellman R. 81
Belytschko T. 321, 331
Bergan P. G. 66, 81
Bettess P. 27, 51
Bianchini G. 144, 147, 148, 150, 151, 175
Bicanic N. 18, 36, 52, 84, 188, 209, 212, 216,
312, 313,333
372
Creager W.P. 284, 295. 302
Crisfield M. A. 63, 82
Cristescu N. 352, 353
Crouse C. B. 236, 249
Cuellar V. 129, 172
Culligan-hensley P. J. 221, 248
Dafalias Y. F. 90, 131, 132, 133, 158, 171,
172, 173, 175
Darve F. 90, 133, 159, 172
De Boer R. 8, 15, 18, 39,50
De Borst R. 321. 332
de Saint-Venant 85, 175
Derski W. 18, 50
Desai C. S. 36, 50, 158, 161, 171
Desrues J. 133, 172
Dewoolkar M.M. 62,221, 227, 245-246,249
Di Prisco C. 158, 172
Dikmen S. U. 255, 261,302
Dobry R. 237, 245,249, 250
Drucker D. C. 86, 111, 120, 144, 172, 188,
215
Ehlers W. 39, 50
Englemann B. E. 32 1, 331
Esrig M. I. 131, 175
Famiyesin 0.0. R. 25,26,50, 55,78,81,226,
233,235,237,248
Felippa C. A. 64. 83
Fillunger P. 6, 7, 10, 15
Finn W. D. L. 129, 267, 303
Fish J. 321, 331
Flavigny E. 133, 172
Franca L. P. 324,332
Frossard E. 135, 172
Fruth L. S. 236, 247
Fuji 267, 302
Fujikawa S. 256, 302
Fukutake K. 256,302
Gantmacher F. R. 82
Gawin D. 40, 51, 188, 189, 193,215
Gens A. 14, 15, 198, 216, 352,353
van Genuchten M. T. 14, 16, 277,303
Ghaboussi J. 25, 50, 72, 82, 131, 158, 172,
255, 261,302
Gibson R. E. 8, 15, 111, 120, 172
Graham J. 337, 351, 353
Gray W.G. 39, 40,42, 50
Green A. E. 18, 39, 50
A UTHOR INDEX
AUTHOR INDEX
373
Makdisi F. I. 274, 303
Marti J. N. 312, 332
Martin G. R. 129, 255, 303
Masuda A. 293,302
Matsuoka H. 159, 173
Matthies H. 63, 82, 344, 354
Mchenry D. 8, 15
Meijer K. L. 202, 216
Meimon Y. 131, 159, 171
Meiri D. 195, 197. 216
Melan E. 86, 173
Meroi E. A. 193, 194, 195, 201, 201, 202,216
Mira P. 332
Mitchell R. J. 221, 248
von Mises R. 86, 175
Miura S. 163, 173
Mizuno E. 21 1, 337, 350, 353
Molenkamp F. 352, 354
Mollener 66, 81
Momen H. 131, 158. 172
Monte J. L. 202, 216
Morgan K. 317, 332
Morland L. W. 18, 39, 51
Mroz Z. 90, 92, 131, 158, 173, 174, 175
Muir Wood D. 25, 26, 50, 55, 78, 81, 128,
174, 226, 233, 235, 237, 248, 337, 351. 352.
353,354
Muller L. 1, 2, 15
Muraleetharan M. M. H. 236, 247
Murata H. 267, 302
Muromoto T. 28 1 , 303
Nagase H. 255,302
Nakamura N. 293,302
Nakazawa S. 64. 70. 84
Narasimhan T. N. 14, 15, 188,216
Narikawa M. 293,302
Needleman A. 321,332
Neuman S. P. 14, 15
Newmark N. M. 60, 82, 336, 354
Newton R. E. 312,333
Ng W. S. 341,354
Norris V. A. I3 1, 158, 173, 174
NovaR. 127, 128,143,158, 160, 161,172,
174
Ohtsuki A. 256, 302
Okada S. 151, 152, 153, 173
Oliver J. 324. 332
Ortiz M. 321, 332
Owen D. R. J. 202, 216, 337, 351,354
372
Creager W.P. 284, 295, 302
Crisfield M. A. 63, 82
Cristescu N. 352, 353
Crouse C. B. 236, 249
Cuellar V. 129, 172
Culligan-hensley P. J. 221, 248
Dafalias Y. F. 90, 131, 132, 133, 158, 171,
172, 173, 175
Darve F. 90, 133, 159, 172
De Boer R. 8, 15, 18, 39, 50
De Borst R. 321,332
de Saint-Venant 85, 175
Derski W. 18, 50
Desai C. S. 36, 50, 158, 161, 171
Desrues J. 133, 172
Dewoolkar M.M. 62, 221,227, 245-246, 249
Di Prisco C. 158, 172
Dikmen S. U. 255, 261,302
Dobry R. 237, 245, 249, 250
Drucker D. C. 86, 11 1, 120, 144, 172, 188,
215
Ehlers W. 39, 50
Englemann B. E. 321, 331
Esrig M. I. 131, 175
Famiyesin 0.0.R. 25,26,50,55, 78,81,226,
233,235,237,248
Felippa C. A. 64, 83
Fillunger P. 6, 7, 10, 15
Finn W. D. L. 129, 267,303
Fish J. 321, 331
Flavigny E. 133, 172
Franca L. P. 324, 332
Frossard E. 135, 172
Fruth L. S. 236, 247
Fuji 267, 302
Fujikawa S. 256, 302
Fukutake K. 256,302
Gantmacher F. R. 82
Gawin D. 40, 51, 188, 189, 193,215
Gens A. 14, 15, 198, 216, 352,353
van Genuchten M. T. 14, 16, 277,303
Ghaboussi J. 25, 50, 72, 82, 131, 158, 172,
255, 261,302
Gibson R. E. 8, 15, 111, 120, 172
Graham J. 337, 351,353
Gray W.G. 39,40,42,50
Green A. E. 18, 39, 50
AUTHOR INDEX
A UTHOR INDEX
Sugimoto M. 264,303
Suzuki Y. 264, 268, 293,302, 303
Tabarrok M. 321, 331
Taboada V. M. 237,250
Tanaka Y. 143, 175
Tani S. 281, 303
Tanimoto K. 143, 175
Tatsuoka F. 125, 142, 153, 1
302
Taylor D. W. 124. 144, 146, 147, 175, 339,
355
Taylor P. W. 140, 175
Taylor R. L. 21,51, 53, 58, 60, 63, 64, 66, 70,
83,84, 186,312,313,319,326,333,343,
344, 355
Terzaghi K. Von 16, 25,51
Terzaghi K. 181, 216
Thompson P. Y. 230, 250
Thurairajah A. 11 1. 120, 121, 175
Toki S. 163, 173
Tokimatsu K. 257, 303
Tresca H. 85, 99, 175
Uchida A. 262,302
Ueno M. 131. 173
Vahadati M. 317,332
Vaid Y. P. 267, 303
Valliappan S. 312. 332
Venter K. V. 21 8, 221, 250
Vilotte J. P. 332
Wang Z. L. 131, 133, 175
Whitaker S. 39, 51
White W. 312, 332
Whitman R . V. 25, 50, 184, 336, 355
Wilde P. 127, 128, 139, 175
Willis P. G. 8, 15, 18, 50
Wilson E. L. 25, 50, 51, 72, 82, 202, 215
375
Wineman A. S. 175
Witherspoon P. A. 14, 15, 188, 216
Wolf J. P. 312, 333
Wood D. M. see Muir Wood D.
Wood W. L. 60, 66, 83,84
Wroth C. P. 1 1 I, 175
Wu J. S. S. 78, 83
Wu J. 317, 325, 333
Wu S. 333
Wu W. 133, 173, 175
Subject Index
ABAQUS 350
Accoustic emission 143
Accumulated shear strain 255
Accuracy control 3 14
Adaptive refinement 305, 3 14-324
Analysing procedure 233-234
anisotropic material tensor 158
Anisotropy 151, 157-1 63
fabric tensor 159
initial or fabric 158, 159
loadistress induced 151, 158, 159
modified stress invariants - see stress
invariants
structure tensor 159
transverse isotropic material 160-1 6 1
Autogenous volumetric strain 129, 166
Babuska-Brezzi condition 64, 70, 324
Backward Euler integration scheme 352
Balance equation
of fluid and air mass for partially saturated
soil 38
of fluid mass for fully saturated soil 22-23
of fluid mass for partially saturated
soil 34-35
see ulso general field equation
Balance equation for mixture theory
macroscopic balance equations 4 2 4 3
microscopic balance equations 42
of energy 42-43
of entropy 4 2 4 3
of mass 4 2 4 3
of momentum 4 2 4 3
Banding sand 145, 154
Bangkok clay 138-1 39
Biot alpha - see effective stress
capillary pressure 13
Capilliary pressure 40
Centrifuge 6
dynamic compatibility 2 19-22 1
numerical modelling of loose and dense
sand bed test 237-245
numerical modelling of retaining dyke
test 221-230
numerical modelling of retaining wall
test 245-247
numerical modelling of submerged quay
wall test 240, 242, 243
numerical modelling of uniform loose sand
bed test 237-245
Scaling laws 21 9-22 1
Silicon oil as substitute fluid 221
use of substitute fluid 220-221
validation of numerical solutions 21 8
VELACS study 218, 230-245
Characteristic state line 125, 142, 144, 146.
148, 151
Co-continuous 64
Cohesion due to suction 276
Consistency condition 10&105
Consolidation equation 28, 66, 67
fully and partially saturated soil
column 203-206
378
Consolidation equation (contd.)
fully and partially saturated
two-dimensional soil layer 206-209
small strain and finite deformation
202-203
Terzaghi theory 204
Constitutive equation
for gas (Dalton's law) 4 4 4 5
for soil 19, 44
Constitutive models (see ah0 plasticity
framework)
Al-Tabbaa and Wood 337. 352
alternate model 352
anisotropic Elastic 337, 351
Barcelona unsaturated soil model 352
benchmark tests 1 10
brick model 352
calibration 1 11
Cam-Clay 120-123, 337, 351
CAP model 337. 350
computational aspects 105-1 10
concrete interface model 352
concrete model 352
constitutive model subroutines 349-352
Cristescu visco-elastoplastic model for
saturated sand 352
critical state models 120-124
extending to sands 12&129
for normally consolidation clays
111-115
densification models 130
liquefaction parameters 270, 282,
28&285, 296
modified for cyclic mobility 165-1 71,
254
determination of soil parameters 253
Drucker Prager 102-104,113,187-188,337
elastic model with varying moduli 337, 35 1
endochronic theory 129
generalized plasticity model
for normally consolidated clays 134-1 39
for overconsolidated clays 139-141
for sands 141-157
Pastor Zienkiewicz mark-111 model 132,
143, 287, 337, 351
parameters for Lower San Fernando
dam analysis 276
parameters for N-dam analysis 291;
parameters for VELACS study 236
parametric idenfication 235-236
true triaxial path 149-1 51, 164
SUBJECT INDEX
379
SUBJECT INDEX
viscous damping 71
Darcy's law, generalized
Dashpots 3 1 1
Degree of saturation
of air 37
of water 12, 33, 37
Del Monte sand 188
Density
fluid 21
solid 21
volume averaged 21
deviatoric stress tensor 95
DIANA-SWANDYNE 11-seeSWANDYNE
Dilatancy 122
dilatancy rule for clay 134-135
dilatancy rule for sand 128, 141-142, 155,
163-164
Effective stress path approach 256
negative dilatancy 256
Discontinuity in displacement 3 16
Discrete memory factor 153
Drained Analysis 67-69, 70
Drained behaviour 4
Drucker stability criteria 144
Earth pressure coefficient at rest (KO)
262-263, 265-266
in situ measurement 262
Earthquake
El-Centro (I 940) 2 10, 256, 27 1, 298
Hyogoken-Nanbu (1995) 255-256, 257,
258, 269, 273
Kobe (1995) see Hyogoken-Nanbu (1995)
Nihonkai Nanseibu (1993) 283
Niigata (1964) 253-254, 287-293
Akita-Kencho record 288-290
Rinkai (1992) artificially generated for
Tokyo Bay area 295. 297-298
Taft (1952) 271, 298
USA record 306
Earthquake damage Counter measures
deep soil mixing 253, 293-302
lattice-shaped stiffened ground walls
(LSSGW) 267-273
Earthquake loading examples
earth dam failure in Hokkaido 253-254,
28 1-287
effect of deep soil mixing counter
measure 253-254, 293-302
elasto-plastic large strain behaviour
21 1-215
SUBJECT INDEX
Failures (contd.)
Lower San Fernando dam 1-3 see also
earthquake loading examples
see ulso earthquake loading examples
static failures 1, 4
Vajont 1, 2
Finite Element
Quadrilateral Q41P4 324, 327, 330
Quadrilateral Q81P4 330
Quadrilateral Q9/P4 324
Triangular T6CI3C 3 19-320, 322
Finite element discretization 53-55
Finite rotation 20
Fluid compressibility 26
Flushing water 256
Free surface see phreatic surface
Fuji river sand 153, 164, 165
-
SUBJECT INDEX
Overconsolidation ratio 1 15
partial saturation
Partially saturated examples
air storage modelling in an aquifer
195-197
consolidation of soil column 203-206
consolidation of two-dimensional soil
layer 206-209
elasto-plastic large strain behaviour
21 1-215
flexible footing resting on a partially
saturated soil 198-201
one-dimensional column 36, 188-192
subsidence due to pumping from a phreatic
aquifer 193-195
Patch test 70, 324
Permeability 65
anisotropic permeability 78
Creager's approach 284, 295
Darcy's Law 22, 37-38
381
382
Quasi-Newton method 69
BFGS 344
Quay wall 267-273
Radiation boundary condition 305,
306-3 15
one-dimensional problem 308-3 12
two-dimensional problem 3 12-3 15
Rankine 85
Reid Sand 150
Relative fluid displacement 26
Repeated boundary condition 3 13
Representative elementary volume 43
Residual condition 143, 147
Roscoe surface 120
Rotation of principal stress axes 150-1 51
Sand boiling 256, 281
saturated, fully
Saturation-capillary pressure
relationship 188-189, 205
Secant update 63
Shape of yield surface in x-plane 98, 100,
137. 186-188
Shear wave 310
Shock wave 314, 317
Simple shear test apparatus 255
Skempton B soil parameter for pore water
pressure 24-25
SM2D 350
example new model subroutine 364-369
implementing new model subroutines
355-364
Softening behaviour 96-97
localization 3 19-324
of sand 147
Soil improvement - see earthquake damage
counter measures
St. Venant 85
Stability criteria 60, 62
conditional stability 33 1
unconditional stability 66
Stabilization of staggered scheme 64
Staggered procedure - see time stepping
Standard penetration test 257
Static analysis 64, 178-188
embankment 179-1 81
footing 179, 181-183
intermediate constraint on
deformation 181, I84
small constraint on deformation 179-1 8 1
SUBJECT INDEX
383
SUBJECT INDEX
GN32 scheme 61
implicit scheme 55, 66
Newmark method 60, 77. 336
SSpj 60
staggered procedure 64
trapezoidal scheme 62
Tresca 85
Two-phase flow 188
Uncoupled equation 68
Undrained Analysis 64, 67-69, 70, 182
Undrained behaviour 4, 24-25, 27-3 1, 182
.see crlso incompressible behaviour
Uniaxial behaviour 86-87
Unloading criteria 90-92, 94
Unloading plasticity 132, 152
u-p formulation 25-27, 55-57
explicit u and implicit p scheme 62
fully Implicit scheme
spatial discretization 58-59
structure of numerical equations 69-70
temporal discretization 60-65
tensorial form of the equations 78-81
Updated Lagrangian Formulation 46, 336