Professional Documents
Culture Documents
Pie Tru SZC Zak 1987
Pie Tru SZC Zak 1987
SUMMARY
In this paper, liquefaction potential of a loose sand deposit subjected to an earthquake loading is evaluated.
The analysis is performed by using a finite element technique incorporating the equations of dynamics of
saturated porous elastoplastic media. The soil response is modelled by an anisotropic hardening rule,
similar to that as proposed by Poorooshasb and Pietruszczak.' The concept is based on the theory
of bounding surface plasticity incorporating a non-associated flow rule and the idea of reflected plastic
potential. The present paper provides a modified formulation to that discussed in Reference 1. Modifications
are aimed at simplifying the concept for numerical implementations.
INTRODUCTION
The liquefaction phenomenon in soils subjected to dynamic loading has received considerable
attention since the experiences in Japan and Alaska during the mid 1960s. At that time, loss of
life and considerable damage were sustained owing to foundation related failures resulting from
seismically induced excess pore pressures. Subsequently, extensive research evolved, addressing
the question of fundamental understanding of liquefaction phenomenon and appropriate design
considerations. More recently, motivation for studying liquefaction is due to increased safety
requirements for important engineering structures such as offshore structures, onshore marine
structures, nuclear reactors, etc.
Investigation into the liquefaction phenomenon has taken many forms, including experimental
studies to evaluate soil response to cyclic loading and theoretical investigations combined with
numerical implementation of complex nonlinear dynamic formulations. Experimental studies on
liquefaction are usually carried out on saturated samples subjected to fluctuating load under
undrained constraint. These tests have provided considerable insight into liquefaction and cyclic
mobility mechanisms. They have also provided guidelines for the development of useful empirical
relationships which may be used to assess the liquefaction potential of sandy sites, thereby
minimizing the potential impact of earthquakes by prudent design.
The in situ response of soil deposits to cyclic loading not only depends on soil properties but
also on drainage characteristics of the material and the boundary conditions. Thus, application
of experimental findings from undrained cyclic loading tests to establish the in situ extent of
liquefaction may be misleading. For this reason, much research has been recently directed toward
development of nonlinear dynamic formulations along with their implementation in numerical
codes. In this approach, a realistic and meaningful solution to a liquefaction problem depends
to a large extent on a choice of an appropriate constitutive law. Such a law must be capable of
proper simulation of several fundamental aspects of soil response to complex loading histories
0363-906 1/87/030221-20$10.00
0 1987 by John Wiley & Sons, Ltd. Received 2 January 1986
222 S. PIETRUSZCZAK AND D. F. E. STOLLE
which have been identified through laboratory investigations; for example, densification, progres-
sive generation of excess pore pressures under undrained conditions (loose sand), hysteresis, etc.
This paper presents a constitutive concept which, together with an appropriate dynamic
formulation for porous saturated materials, can predict the extent of liquefaction in sandy soil
deposits. The concept is formulated within the framework of bounding surface plasticity in-
corporating an anisotropic hardening rule, similar to that recently proposed by Poorooshasb
and Pietruszczak.’ A modified formulation to that discussed in Reference 1 is provided. The
proposed modifications significantly reduce the difliculty of transferring the concept into
numerical code without sacrificing the quality of its performance. The boundary value problem
of a sand layer subjected to an earthquake shock is solved, by using the finite element technique,
to demonstrate the ability and appropriateness of the constitutive concept for modelling the
liquefaction phenomenon. Results of a parametric study are presented, showing the influence of
various factors such as permeability, elastic properties of soil skeleton, etc., on liquefaction
potential. Owing to the one-dimensional nature of the problem, the generalized equations for
the dynamics of saturated porous media2 are condensed to reduce the number of unknowns. A
brief summary on the condensation procedure of the governing equations is also presented.
is governed by the yield surface which is allowed to move within the domain enclosed by the
bounding surface. If the stress reversal continues until the stress point reaches the bounding
surface once again, and attempts to move outside it, the entire stress reversal history is erased
from the material memory. It should be emphasized that during stress reversal programs the
bounding surface expands or contracts depending on the mode of accumulation of plastic
distortions.
In order t o provide a general mathematical formulation, let us introduce the following stress
invariants:
I = - Oii/J3; 6 = (+s,.s..)'/2.
L J ij 3 - 13 s i j s j k s k i
J3 - (1)
where sij= aij- 4 8 i j g k k and sij denotes the stress deviator. The strain rate invariants which are
compatible with I and 6 are given by
E = - d../ J 3 ; 6 = (p..,j..)1/2
IJ IJ (2)
where tij= iij- $$jd,& represents the strain rate deviator.
The equation of the bounding surface is postulated to be of the form
F = 5- r]g(e)r= o (3)
In the principal stress space, equation (3) represents an irregular cone having its apex at the
origin and its axis coinciding with the diagonal of the principal stress space. The parameter r]
depends on the history of plastic distortions, that is r] = r ] ( g P ) where cp is defined as
where
g
(a>
-
9
g
( E)
-- =
3 - sin 4
3+sin4=k
where r]/ represents the ultimate value of r] (at failure). For q -,qJ, equation (6) defining g( - n/6)
224 S. PIETRUSZCZAK AND D. F. E. STOLLE
This restriction on g( - 4 6 ) ensures that the form of g(8) is not in conflict with the conditions
dictated by the Mohr-Coulomb failure criterion. On the other hand, the case q + O results in
which implies that, for small q, the surface represented by equation (3) may be approximated
by a regular cone.
It is well known that virgin soils exhibit irreversible deformations almost immediately after
application of an external load. This allows speculation that the elastic domain can, in fact, be
reduced to almost an infinitesimal size. Therefore, equation (10) (valid for 7 +0) will be used
later when defining the shape of the yield surface. It should be noted that the concept of an
evanescent elastic domain has already been proposed by other investigator^.^-'
An example of function g(8) satisfying equation (6) is the expression introduced by William
and Warnke,’ namely
( 1 -&’)a + (2k - 1)[(2 + b)(l - k z ) + 5kZ -4&]”*
=
(2 + b)(1 - &’) + ( 1 - 2k)*
where
a = J ( 3 ) cos 8 - sin 8
b = cos 28 - J(3) sin 28
Other relationships which may be adopted in place of equation (1 1) are discussed e1~ewhere.l.~
In the present study, the function 9 = yl(cP)is postulated in a simple hyperbolic form
where A is a positive constant. This type of function describes strain hardening only. Other more
sophisticated functions, permitting strain softening as well, are available in the
Now that all primary definitions have been introduced, we can derive the corresponding
constitutive relations. Let us first consider the case when the stress point is located on the
bounding surface. In what follows, a non-associated flow rule is assumed:
The parameters qc and I, (equation 14) are assumed as constant; I, has the dimension of stress
and qc defines the value of O / l for which 2; = 0 (by analogy to the critical state concept).
Satisfying the consistency condition
SAND UNDER EARTHQUAKE EXCITATION 225
Thus, for the present case, the constitutive relation will assume its usual form
where DFjklrepresents the elastic constitutive matrix and H p is defined by equation (1 7).
Consider now the stress reversal process. As explained previously, the plastic flow is now
described by the evolution of the yield surface which is created inside the bounding surface. Upon
stress reversal, the yield surface is initially tangential to the bounding surface at the stress reversal
point. For subsequent loading, if the stress point remains inside the yield surface the response of the
material is elastic; beyond this range irreversible deformations take place and the yield surface
moves within the domain enclosed by the bounding surface.
In what follows, the size of the yield surface, measured by parameter qI,will be assumed small
compared with 1 ', as defined by equation (8).+As a consequence, the equation of the yield surface
will be approximated as being independent of 8 (i.e. g(8) = 1). This equation takes the form
V' a..)
f (g.. V = - qI1(l)= 0 (19)
where
1") = .
i j 0 i.j , &I) = (+sijsij)'/2;sij= gij- aijl(l); qI<< q, = const. (20)
In the principal stress space, equation (18) represents a regular cone (Drucker-Prager type of
surface)having its apex at the origin and its axis directed along the unit tensor aij(sothat aijaij= I).
Let us note that if aij is directed along the stress space diagonal, i.e.
a.. = - - 81 . .
IJ 43 'I
The kinematics of the yield surface is guided by a particular stress tensor known as the conjugate
stress tensor located on the bounding surface. Referring to Figure 2, which presents a'-plane
section (with normal aij) of both the yield and bounding surface, if nij is the current state of
'A finite size of the yield surface is required in order to avoid a ratchetting dect.
226 S. PIETRUSZCZAK AND D. F. E. STOLLE
where A is a constant which can be determined from the equation of the bounding surface
F(afi,E P ) = 0, equation (3). Solution to this equation also provides the location of a so-called
datum stress tensor o!j (see Figure 2) which is located on the bounding surface and is coplanar
(in the stress space) with both aij and o : ~ .
In order to formulate the translation rule in such a way as to avoid intersection between the
yield and bounding surface, it is convenient to define the so-called conjugate yield surface, f,= 0
(Figure 2). This surface is tangential to the bounding surface at the conjugate point afi and
its location is specified by the unit tensor a:j:
Let us still note that since qr << qf is assumed in equation (20), a sufficient degree of accuracy
may be preserved when afj is replaced by a unit tensor along o:j in translation rule (26). This
avoids the necessity of determining the location of conjugate yield surface, namely equation (25),
and is thus advantageous for numerical implementation.
SAND UNDER EARTHQUAKE EXCITATION 227
The simple mathematical constraint on the kinematics of the yield surface, equation (26), is
the result of equation (19) being independent of 0. In Reference 1 the yield surface was assumed
to be geometrically similar to the bounding surface. Consequently, additional rotation of the
yield surface in the plane perpendicular to its axis was required in order to ensure a smooth
transition with the bounding surface. This fact significantly complicated the previous formulation.
To complete the present description, the rule for the variation of plastic hardening modulus
in the course of deformation must be specified. For the considered reverse loading process, the
non-associated flow rule can be conveniently expressed by
.?J
- hp(nkl
&.. &kI)&j (28)
Here, nij and fiij represent unit tensors normal to the yield and plastic potential surface,
respectively, i.e.
The magnitude of the parameter h, is assumed to depend on the position of the stress tensor
relative to the bounding surface, i.e. its relation to the conjugate and the datum stress tensors.
Let the spatial angle between the stress and the conjugate stress tensor be 6:
and denote the analogous angle between the datum and conjugate stress tensor by 6,. The
proposed expression for h, is
h, = hB( 1 -$)'
where y is a constant and h, is defined as
[
h, = K , I'):(
+KO - (33)
where K , = h i l , K O = const. and K O >> K,. The interpolation rules (31) and (33) are similar in
nature to that originally suggested by Dafalias and P O P O V . ~
Finally, the local plastic potential $ = 0 is postulated in the form of a surface of revolution
about aij axis having one of its generating curves in common with the global plastic potential
surface, equation (14).
The equation of the surface $= 0 is assumed in the form
228 S. PIETRUSZCZAK AND D. F. E. STOLLE
where
The value of /I3 = const. is determined from the condition that equation (34) is satisfied by the
components of the current stress point a i j .
Equation (34)is based on the generalization of equation (54) which is discussed in the following
paragraph. Further details are provided in Appendix I.
The strain parameters 8, and E, equation (2), which are compatible with the stress parameters
P, Q,are now defined through the set of relations
=Q + J(2)qCPg(8)ln - =0
K O )
For the active loading process which satisfies the equation of the bounding surface, we have
aF . aF . + -4
aF
- P + -Q =O
ap aQ att
SAND UNDER EARTHQUAKE EXCITATION 229
which, in view of equations (38) and (40) and the non-associated flow rule:
(42)
yields
A* = Q - J(2)tlde)P where Hp= ,/(2)Pg(O)-all
(43)
HP asp
For the stress reversal programs, equation (20), which defines the stress invariants P,8'),
simplifies to
Z ( ' ) = O ~ U ~+ 2 a , ~ , = P ' ; =Q'/J2 (44)
so that
+
P = PUP QuQ; Q = Q u -PuQ
~ (45)
where
-(uI +2u3)/J3; uQ=/(i)(U3-U1) (46)
In the above equations, {P',Q') represents a set of local stress parameters as measured with
respect to the axis of the current yield surface and up,tlQdenote the direction cosines of this axis.
The equation of the yield surface, equation (19), simplifies to
f = Q' - ,/(2)q1P' = 0 (for Q' > 0) (47)
and the consistency condition takes the form
The kinematics of the yield surface is completely defined by satisfying the condition
~ p & p+ uQ&Q= 0 (49)
The solution to equation (52), for g(0) specified by equation (39), yields two values of the parameter
A defining, according to equation (51), the corresponding location of both conjugate and datum
points.
The interpolation rule for the plastic hardening modulus can be assumed in either of the forms
discussed in the previous section, i.e. equations (31) or (33). The spatial angle between the stress
and the conjugate stress points, equation (30), is now defined by
6 = c0s-l ( ( P P c+ QQ,)/C(P2 + QZ)(P4 + Q:)11’2} (53)
Finally, the local plastic potential surface $= 0 cuts the P,Q stress space along two curves.
The first curve satisfies equation (40) and the second curve, referred to as the reflected plastic
potential, I)” = 0, is simply a mirror image of I)= 0 about the u axis (see Figure 3).
According to equation (34) and in view of relations (44) and (45), the equation of local plastic
potential takes the form
where
“Q
B1 = rl,g(@ and B2 = --
J2 up
The identity 6 = I)”= 0 is obtained from equation (54) through the transformation Q’ = - Q‘;
P=P.
More detailed discussion of the form of local plastic potential $ = 0 is provided in
Appendix I.
is capable of simulating both the progressive generation of pore pressures leading to liquefaction
(strain controlled regime) and cyclic mobility effects (stress controlled regime).
It seems that one of the major drawbacks of the present concept is its inability to predict
irreversible deformations for stress path satisfying q = consts. (including q = 0). This deficiency
may be removed by incorporating, for example, the effect of volumetric hardening. The present
authors, however, are of the opinion that q = const. paths are unlikely to arise in the context of
a boundary value problem and are, therefore, reluctant to introduce further complications to
this concept.
w2.2 + u2.2 + Pn = 0
- (mass balance) (57)
K,
a!I j -
- bij + PSij (effective stress principle) (58)
where aij,aijrepresent total and effective stresses, respectively, p is the pore pressure, n is soil
porosity, K , represents fluid compressibility, k is soil permeability, p,p, and y,yf represent the
total and fluid densities and unit weights, respectively. Of course, to define the problem of
fluid-solid dynamic interaction completely, the above set of equations must be supplemented
by strain-displacement and constitutive relationships.
By neglecting compressibility of fluid (compared to that of soil skeleton), equation (57) can
232 S. PIETRUSZCZAK AND D. F. E. STOLLE
be simplified to
u2 + w 2 = const. (Kf-+ 00) (59)
If no drainage is permitted at the lower boundary, the above equation reduces to u2 = - w 2 .
After substituting this identity into equilibrium statement ( 5 5 ) and utilizing equation (56), we
obtain
cJ12.2 =Pi, (604
g22.2 +Ys =P i 2 + u 2
where
Yr = Y -Y/ (buoyant unit weight)
P= P +P / ( ; - 2)
The above system (60)contains only two primary unknowns, ul and u2. It should be noted that the
influence of seepage is introduced via the damping term in equation (60b).
The integrated equivalent for equations (60),suitable for reduction to a finite element equivalent,
is given by
and
where 6ui are weighting functions compatible with ui and F i ( t ) represent the time-dependent
loading. The numerical procedure adopted in this study incorporated a central difference time-
marching scheme and lumped mass matrix which helped minimize storage requirements. Owing
to the nature of the problems which may be solved using equations (61), only 4 degrees-of-freedom
per element are required for the linear interpolation. The reader is referred to Zienkiewicz” for
details on reducing equation (60) to a finite element form.
At this point it should be noted that the pore pressures are not explicitly required, since
equations (60) are given in terms of effective stresses. The effective stress rates are related to
strain rates by the constitutive law discussed in the previous sections. The strains c l = E~~ = 0
are known, a priori, based on the physical and mathematical restrictions of the model.
- 1
BASE MOTION
I 7hl a18 kN/d
7 . - l O kN/d
Figure 4. Geometry of the problem; finite element subdivision and specification of material parameters
2.00
1.50
1 .oo
0.50
-
E
=
0
0.00
I-
<
-0.50
-I
Y ’I I
U
-1.oc
-1.X
-2.a
-2.51 I I I I I I
1.00 2.00 3.00 4.00 5.00 6.00
TIME (I)
be noted that the applied load excitation induced in the material, cycles with relatively low stress
amplitudes. Thus, the assumption of g(0) = const. had little influence on the quantitative aspects,
yet it increased the efficiency of the computer program significantly.
The results of numerical analysis are presented in Figures 6-8. Material parameters (indicated
in Figure 4 and adapted after Reference 14) correspond to loose Fuji river sand as tested by
Ishihara and T a t ~ u 0 k a . INote
~ that nonlinear elastic properties were assumed, i.e. the bulk
modulus K varying linearly with confining pressure and the shear modulus G = const. Figure
234 S. PIETRUSZCZAK AND D. F. E. STOLLE
i_
Ly
Y
SAND UNDER EARTHQUAKE EXCITATION 235
x
236 S. PIETRUSZCZAK AND D. F. E. STOLLE
2 00-
1.50-
1.00-
3
Y
O’*
z
-c
0 0.00.
2
s -0.50.
V
V
<
V
Ly - 1.00.
<
Y
E
3
u) -1.50.
-2.00.
-2.50 b
1 I I I I
1.00 2.00 3.00 4.00 500 6.00
TIME (f)
6(a) shows the time history of effective pressure I, for elements 6, 7 and 9, whereas Figure 6(b)
presents the corresponding time history of shear stress. The results correspond to the coeflicient of
permeability k = m/s. This value of k is very low, implying that the response of the system is
close to that obtained under undrained constraint. In Figure 6(a), the values of effective pressures
are normalized with respect to initial confining pressures generated by the self-weight of soil.
During the earthquake excitation, a gradual build-up of pore pressure takes place throughout the
sand layer and after approximately 3.5 sec complete liquefaction occurs in element number 7
(effective pressure reduces to zero).
The effect of an increase in the value of coefficient of permeability to 10-’m/s is presented
in Figure 7. In this case, transient seepage has a significant impact on the results. The time
required for the deposit to liquefy increases to over 6sec and both the depth and the extent of
the liquefying zone also increase. Figure 7(c)shows the computed surface acceleration. Comparing
this result with the base input (Figure 5), it is evident that during early stages of loading history
the surface accelerations are significantly amplified, while frequency is similar to that at the base.
After the peak acceleration occuring at approximately 2.5sec, amplitudes at the base and the
surface remain almost the same.
Finally, the quantitative aspects of the response are significantly affected by the elastic properties
of the material. Figure 8 presents the comparison of two solutions admitting nonlinear and
linear elastic properties, respectively. In the latter case, a constant bulk modulus, K = 15,000 kPa,
SAND UNDER EARTHQUAKE EXCITATION 237
l5C
125
x"lo(
Y
w
Ly
3
v)
v)
w 0 75
0
w
L
c
LL
& osc
B
N-
az
w
0 025
Z
0 O(
TIME (5)
Figure 8. Comparison of predictions based on linear and nonlinear elastic properties ( k = m/s)
Figure 9. Plastic potential surface $ = 0 in compression (a) and extension (b) domain
238 S. PIETRUSZCZAK AND D. F. E. STOLLE
was assumed. It is evident from this figure that the linear elastic properties significantly increase
the rate of effective pressure reduction as the liquefaction conditions are approached. The results
corresponding to fully drained analysis (with linear elastic properties) in which no excess pore
pressure is generated, are also presented in Figure 8 for comparison.
CONCLUSIONS
The constitutive concept discussed in this paper is applicable to both loose and dense sands.
The overall effectiveness of the concept has been tested quite extensively elsewhere.1 , 1 4 The
results indicate that the model is capable of appropriate simulation of several fundamental
aspects of soil behaviour which have already been recognized through laboratory investigations.
In particular, liquefaction of loose sand deposits and the effect of cyclic mobility in dense sand can
be adequately simulated. l 4 Therefore, there is sufficient degree of confidence when applying this
model to boundary value problems. The numerical analysis presented in the last section is purely
qualitative in nature. It proves, however, the objective of the paper, i.e. it demonstrates the
capability of this concept to model liquefaction phenomenon under transient dynamic conditions.
Numerical analysis of liquefaction based on realistic material description cannot only predict
the phenomenon but can actually indicate the influence of several factors that can be altered,
like initial stress conditions, permeability etc. Thus, such an analysis, when combined with a
suitably defined experimental program, can provide a valuable insight into liquefaction
mechanisms and, in perspective, can help to answer the question as how to prevent liquefaction.
ICI, = Q + J(2)Pl p In (g) =0 (in compression domain, i.e. for Q > 0) (62a)
and
ICI, = - Q + J(2)PlPln($) =O (in extension domain, i.e. for Q c 0) (62b)
where
PI = %S(@
and
Assume now that the local plastic potential associated with the yield surface f = 0 intersects the
P, Q space along two curves (Figure 9). The first curve satisfies equation (62a) or (62b) and the
second one, shown by $” = 0 and called the reflected plastic potential, is postulated in the form
of a surface of revolution about the uij axis having one of its generating curves in common with
the global plastic potential surface (equation 14).
In order to derive the equation of the local plastic potential $ = 0, let us write equation (62a)
in a transformed reference frame P , Q ’ , such that
+
Q = Q’u, Pu,
P = P u p - Q‘uQ
SAND UNDER EARTHQUAKE EXCITATION 239
Dividing the above expression by a pwe can write the equation of plastic potential in an alternative
form:
where
uQ and 8 3 = P 0 / a p
8"J&
It is obvious that the same transformation applied to equation (62b) yields
Thus, the local plastic potential $ = 0 is defined as follows (see Figure 9):
(i) in compression domain (Q > 0)
$= *c for Q' > 0
$=$p forQ'<O
(ii) in extension domain (Q < 0)
$=*e forQ>O
$=$," forQ'<O
Note that = 0 (in equation 69) and $: = 0 (in equation 70) are obtained from equations
(67) and (68), respectively, through the transformation Q = - Q' and P' = P'.
The above derived equations for $= 0 can easily be generalized using relations (44)in
combination with equations (20).Thus, introducing (after equation 44)the following definitions
The above equation represents the local plastic potential as a surface of revolution about aij
axis. One of the generating curves of $ = 0 (i.e. that corresponding to the meridional section
8 = const, with 8 evaluated for the current state of stress a i j )is common with the global plastic
potential surface $ = 0 specified by equation (14).
In equation (62), the constants 8, and /I2 are generalized to
240 S. PIETRUSZCZAK AND D. F. E. STOLLE
where g(0) is specified by equation (1 1) and tLij denotes the deviatoric part of aij:
~..=a..--16
tj rj
.a
3 i j Irk (74)
The constant P3 is determined from the condition that the surface $ = 0 contains the current
stress point oij.
The gradient tensor (d&/daij) may be evaluated in a usual manner as
(76)
and
(77)
REFERENCES
1. H. D. Poorooshasb and S. Pietruszczak, ‘On yielding and flow of sand, a generalized two-surface model’, Comp.
Geotech., 1, 33-58 (1985).
2. 0.C. Zienkiewin and P. Bettess, ‘Soils and other saturated media under transient dynamic conditions: general
formulation and the validity of various simplifyingassumptions’, in Soil Mechanics-Transient and Cyclic Loads, G. N
Pande and 0. C. Zienkiewicz (eds), Wiley, 1982.
3. Y. F. Dafalias and E. P. Popov, ‘A model for nonlinearly hardening materials for cyclic loading’, Acta Mech., 21, 173-
192 (1975).
4. Z. Mroz, V. A. Norris and 0.C. Zienkiewin. ‘An anisotropic hardening model for soils and its application to cyclic
loading’, lnt. j . numer. anal. methods Geomech., 2, 203-221 (1978).
5. A. Mroz and S. Pietruszczak, ‘A constitutive model for sand with anisotropic hardening rule’, 1nt.j. numer. anal. methods
geomech.. 7, 305-320 (1983).
6. Y. F. Dafalias and L. R. Herrmann, ‘Bounding surface formulation of soil plasticity’, in Soil Mechanics- Transient and
Cyclic Loads, G. N. Pande and 0. C. Zienkiewin (eds), Wiley, pp. 253-282. 1982.
7. C. S. Desai, S. Somasundram and M. 0.Faruque, ‘Constitutive modelling of geological materials: a general procedure’,
in Developments in Soil Mechanics and Foundation Engineering-2, P. K. Banerjee and R. Butterfield (eds), Elsevier-
Applied Science, London, pp.43-67, 1985.
8. K. J. William and E. P. Warnke, ‘Constitutive model for the triaxial behaviour of concrete’, Seminar on Concrete
Structures Subjected to Triaxial Stresses, ISMES, Bergamo, 1972.
9. 0.C. Zienkiewicz and G. N. Pande, ‘Some useful forms of isotropic yield surface for soil and rock mechanics’. in
Numerical Methods in Geomechanics,G. Gudehus (ed), Wiley, 1977.
10. C. S. Desai and H. J. Siriwardane, Constitutive Lawsfor Engineering Materials, Prentice-Hall, Englewood Cliffs, N.J.,
1984.
11. J. Ghaboussi and H. Momen, ‘Plasticity model for cyclic behaviour of sands’, in Numerical Methods in Geomechanics,
3rd Int. Conf., Aachen, Balkema, vol. 1, pp.423-434, 1979.
12. J. H. Prevost and K. Hoeg, ‘Effectivestress-strain model for soils’, J . Geotech. Diu., ASCE, 101(GT3),257-278 (1975).
13. P. A. Vermeer, ‘A double hardening model for sand’, Geotech., 28, 413-433 (1978).
14. S. Pietruszczak and H. D. Poorooshasb, ‘Modelling of liquefaction and cyclic mobility effects in sand, in Numerical
Methods in Geomech., 5th Int. Conf., Nagoya, vol.4, 1867-1875, 1985.
15. 0.C. Zienkiewin, The Finite Element Method, 3rd edition, McGraw-Hill. London, 1977.
16. T. Shiomi, S. Pietruszczak and G. N. Pande, ‘Liquefaction study of sand layers using the reflecting surface model’, in
Numerical Models in Geomechanics, 1st. Int. Sympo., Zurich, Balkema, pp. 41 1-418, 1982.
17. K. Ishikara and F. Tatsuoka, ‘Undrained deformation and liquefaction of sand under cyclic stresses, Soils Found., IS,
29-44 (1975).