Sunteți pe pagina 1din 33

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING

Int. J. Numer. Meth. Engng 2011; 87:457489


Published online 14 April 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/nme.3192
Computational geomechanics: The heritage of Olek Zienkiewicz
M. Pastor
1,2, ,
, A. H. C. Chan
3
, P. Mira
2, 4
, D. Manzanal
1, 5
, J. A. Fernndez Merodo
6
and T. Blanc
1
1
Department of Applied Mathematics and Informatics, ETS de Ingenieros de Caminos, Universidad Politcnica de
Madrid, Spain
2
Centro de Estudios y Experimentacin de Obras Pblicas, Cedex, Madrid, Spain
3
University of Birmingham, U.K.
4
Department of Civil Engineering and Construction Technology EU de Ingeniera Tcnica de Obras Pblicas
Universidad Politecnica de Madrid, Spain
5
Engineering Faculty, National Patagonian University, Comodoro Rivadavia, Argentina
6
rea de Investigacin en Peligrosidad y Riesgos Geolgicos, IGME, Madrid, Spain
SUMMARY
Geotechnical engineering design is based on mathematical, constitutive and numerical models implemented
in computer codes which are able to predict the behaviour of geostructures and foundations under a wide
variety of loading conditions, sometimes demanding or extreme. Prediction is fundamental in areas such
as seismic engineering, marine foundations or landslide hazard analysis. No computer code is better than
the mathematical, constitutive and numerical models which it implements. This paper is devoted to (i)
reviewing the decisive contributions of Professor Olek Zienkiewicz to the development of mathematical,
constitutive and numerical models in geotechnical engineering and (ii) presenting some extensions based
on his work which have been proposed by the authors. The extensions to be described here are the
following: (i) a particularization of the up
w
Zienkiewicz model to the case of fast catastrophic landslides
(ii) an extension of the generalized plasticity model proposed by him and (iii) an SPH numerical model
for the propagation of fast catastrophic landslides. Copyright 2011 John Wiley & Sons, Ltd.
Received 4 February 2010; Revised 12 January 2011; Accepted 4 February 2011
KEY WORDS: mathematical modelling; constitutive models; nite elements; SPH; fast landslides; marine
foundations; dry liquefaction
1. INTRODUCTION
Geomaterials is a relatively newconcept that embraces a wide range of materials of geological origin
such as soils, rocks and concrete. There exist geomaterials of granular type such as sands, or of
cohesive type, such as clays, concrete and rocks. In some cases the failure is ductile, while in others
it is fragile. Geomaterials are, in general, porous materials with voids lled by interstitial uids
such as air or water. They are, therefore, multiphase materials exhibiting a mechanical behaviour
governed by coupling between all the phases. The total stress acting on a geomaterial is decomposed
on the parts acting on each component of the mixture. Deformation of the solid skeleton is governed
by the effective stress tensor. This is the rst fundamental characteristic of geomaterials.
A second fundamental property of geomaterials is the dependence of their behaviour on the
effective conning pressure. This fact has been known since the work of Coulomb in 1773 [1], when
he proposed a failure criterion where the maximum shear stress depended on the normal stress.

Correspondence to: M. Pastor, Department of Applied Mathematics and Informatics, ETS de Ingenieros de Caminos,
Universidad Politcnica de Madrid, Spain.

E-mail: manuel.pastor@upm.es
Copyright 2011 John Wiley & Sons, Ltd.
458 M. PASTOR ET AL.
A third property of geomaterials is their dilatancy, a change in volume associated with the shear
strain. The rst study of dilatancy is due to Reynolds [2], who observed it in sands. Finally, in the
case of soils, the residual states after failure has occurred take place at what is called critical state
conditions, independent of the initial conditions or the type of laboratory test.
These four properties of soils have to be taken into account when developing mathematical
and constitutive models to describe their behaviour and are fundamental ingredients of modern
geomechanics.
Geotechnical engineers use models to predict the behaviour of foundations and geostructures,
such as earth dams or slopes, to study their performance under the design loads, and the conditions
under which failure will take place, i.e. the failure load and the failure mechanism.
Many of the ingredients of mathematical, constitutive and numerical models being currently
used by engineers have been introduced by Professor O. C. Zienkiewicz. The purpose of this paper
is to describe the evolution of computational geomechanicsfrom the authors point of viewand
the role of Prof. Zienkiewicz and his group.
The paper is structured as follows. We rst analyze the mathematical models which describe
the coupling between the different solid and uid phases of the soils. In the second section we
deal with the constitutive models that describe the behaviour of soil skeleton, where we describe
the generalized plasticity theory developed by Zienkiewicz and Mroz and review some models
developed based on their original idea.
The third section of the paper is devoted to numerical modelling. One important characteristic
of Professor Zienkiewicz and his group was to deal with problems in different areas of mechanics
and engineering, such as solid dynamics, uid dynamics and structural, hydraulics, harbour and
geotechnical engineering. The solution of some problems from one specic area was obtained in
many cases in another.
We devote the nal section to consider the failure of geostructures. There is a special case, the
diffuse failure, where the material can uidize. This is the case of liquefaction of loose granular
soils due to earthquake loading. When occurring in a slope, fast catastrophic landslides can be
triggered. The propagation of these uidized soils can be studied using tools from computational
uid dynamics.
2. MATHEMATICAL MODELS FOR GEOMATERIALS
2.1. Terzaghi, Bishop and the concept of effective stress in geomaterials
As geomaterials are mixtures of solids and uids, the mathematical models describing their
behaviour have to account for the distribution of the total stress on the components. Partial stresses
are therefore responsible for the behaviour of each phase.
The simplest model describing the distribution of the total stress in saturated soils was proposed
by Terzaghi [3], who introduced effective stress as the difference between the total stress and the
stress acting on the pore water. This denition is widely accepted today in practical geotechnics,
but alternative denitions have been proposed by Biot [4], Taylor [5] and Skempton [6]. The
concept of effective stress in saturated soils provides a way to both understand soil behaviour and
to design experiments where it is obtained from the total stress and the pore pressure.
In the case of unsaturated soils the situation is different. They present an apparent cohesion due
to the negative pressure in the pore water. The term suction or matrix suction is introduced to
dene the difference between air and water pressure:
s = p
a
p
w
In many engineering applications, the pore air will be at atmospheric pressure, i.e. p
a
=0 but
in laboratory tests researchers conduct tests with non-zero p
a
. To illustrate the effect of suction,
consider that building a sand castle requires sand with an apparent cohesion which does not exist
either on fully dry or on fully saturated sands.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 459
After the early attempts of Haines [7] and Fisher to understand the role of capillary forces,
Bishop [8] proposed a generalization of Terzaghi effective stress for unsaturated soils as
o

=o,
s
I +p
a
I
where , is a parameter varying between zero and one, often referred to as Bishop effective stress
parameter. The above expression can be rewritten as
o

=(o+p
a
I ),
s
I
where the stress tensor is decomposed into a net stress tensor o=o+u
a
I and the suction term.
Please note the use of negative values for compression components. Jennings and Burland [9]
showed that the effective stress proposed by Bishop cannot reproduce collapse in wetting paths
where suction decreases and the hydrostatic component of the net stress increases in such a way
that Bishop effective stress is constant. This led Bishop and Blight [10] to state that The principle
of effective stress can be applied to saturated soils only if account is taken of the effective stress
path. In the case of partly saturated soils it is not the effective stress path only, but the paths of the
two components op
a
I and s = p
a
p
w
which have to be taken into account, introducing the
so-called bi-tensorial formulations based on the net stress and the matric suction tensors widely
used later on.
In the decade of the 1980s most researchers had adopted the bi-tensorial framework to both
analyze and design experiments in unsaturated soils. However, some researchers continued working
with alternative denitions of the effective stress. Schreer [11] and Lewis and Schreer [12]
derived a simple and rational form. A straightforward use of this effective stress measure was
not enough, and the solution was found by Houlsby [13] who obtained the work dissipated in
unsaturated soils.
2.2. BiotZienkiewicz equations
Because of space limitation, only a brief outline is given in this section. Further details can be
found in Zienkiewicz et al. [1416] and Chan [17]. The BiotZienkiewicz formulation has been
developed for the static, consolidating and dynamic analysis of soils. It is based on the fully
implicit up approximation of the Biot [18] formulation for a saturated porous medium possibly
interacting with a solid component e.g. a structure.
Using the continuity, momentum and generalized Darcy equations, the complete Biot equation
governing deformable porous media can be expressed in three-dimensional Cartesian co-ordinates
using indices notation as
o
ij, j
+jb
i
j u
i
j
f
w
i
=0 (1)
where o
ij
is the total stress tensor (tensile positive), u
i
and w
i
are the displacement of the solid
skeleton and average (Darcy) uid velocity, respectively, b
i
is the body force per unit mass, j
s
, j
f
and j are the densities of the solid grain, uid and mixture, respectively, with j=(1n)j
s
+nj
f
and n being the porosity of the porous media. By including porosity n in the denition of the
averaged relative uid velocity
w
i
=n(

U
i
u
i
) (2)
their value conforms with that used in the original Darcy equation (see e.g. [19]) with U
i
being
the displacement of the uid. The generalized Darcy equation can be expressed as
w
i
=
v
k
v
f
j
k
g
k
k
ij
_
p,
j
+j
f
g
j
j
f
u
j

j
f
w
j
n
_
(3)
where k
ij
is the permeability tensor, Biot : is dened as :=1(K
T
/K
s
) and K
T
is the bulk
modulus of the porous media. For material with isotropic permeability k (unit = length/time),
k
ij
=ko
ij
, where o
ij
is the Kronecker delta. Furthermore, v
k
, j
k
and g
k
are the viscosity of the
uid, density of uid and gravitational acceleration at which the permeability is measured and v
f
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
460 M. PASTOR ET AL.
is the viscosity of the uid actually being used. It is common to use a substitute uid with higher
viscosity in centrifuge tests in order to maintain the compatibility of the dynamic behaviour and
the consolidation behaviour when the actual prototype soil is used in a centrifuge experiment.
Finally, the continuity equation for the uid phase can be expressed as
p
Q
+ u
i,i
+ w
i,i
=0 (4)
where Q is the averaged bulk modulus dened with K
s
being the bulk modulus of the soil grains
and K
f
the bulk modulus of the pore uid as:
1
Q
=
:n
K
s
+
n
K
f
(5)
By eliminating pore pressure p between Equations (3) and (4), the uw formulation (so-called
because u and w are the primary variables) results. On the other hand, if the averaged relative uid
acceleration is neglected from Equations (1) and (3), and are eliminated between Equation (1) and
(4), the up formulation is recovered. By neglecting further the solid acceleration in Equations
(1) and (3), the standard consolidation equation is obtained. Finally, if both the solid and uid
velocities are neglected, the resulting equations are now decoupled and they are the static solid
equation and the steady-state seepage equation.
The formulation can be easily extended in order to account for partially saturated conditions as
shown in Zienkiewicz et al. [15]. If the air pressure is assumed to be constant and equal to zero
and the weight of air is negligible, then the pore pressure becomes
p=,p
w
+(1,) p
a
,p
w
(6)
The average density is, neglecting the weight of air
j=nS
w
j
w
+(1n)j
s
(7)
and the generalized Darcys equation becomes
w
i
=
v
k
v
f
j
k
g
k
k
ij
(S
w
)
_
p
w, j
+j
f
g
j
j
f
u
j

j
f
w
j
n
_
(8)
where j
f
=S
w
j
w
.
The continuity equation for the water phase is now
p
w
Q

+ u
i,i
+ w
i,i
=0 (9)
with
1
Q

=
(:n),
K
s
+
nS
w
K
f
+n
dS
w
dp
w
If the air pressure cannot be neglected, then separate continuity and ow equations have to be
written for the water and air phases, respectively, and the equations can be found in Zienkiewicz
et al. [15].
2.3. Beyond failure: depth-integrated models for uidized soils
The numerical simulation of localized failure has attracted the attention of many researchers in
the past. However, there exists another failure mechanism, which has been described as diffuse,
which appears in soils of very low density. This is the case of liquefaction, where the soil fails at
effective stress states inside the MohrCoulomb surface. This model of failure has been studied
by Darve and coworkers [2022] and Nova [23]. Liquefaction can be triggered by both monotonic
loads, as in the case of static liquefaction processes, and by cyclic loads such as sea waves or
earthquakes. In some cases, the material is conned by the surrounding soil, while in others it
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 461
can ow freely. This is the case of some landslides, referred to as owslides, where the uidized
material can ow reaching distances of the order of hundreds of meters. The initial pore pressure
is dissipated along the propagation process.
To model such phenomena it is necessary to use models where the coupling between soil skeleton
and pore uids is taken into account. The problem can be modelled directly with the approach
described in the preceding sections, although due to the shape and geometrical properties of the
landslides some simplications can be done.
The rst simplication is to assume that the dissipation of pore pressure takes place mainly
along the normal to the surface, as the landslide depth is much smaller than its length. This
model is a propagationconsolidation model, which can be derived assuming that the velocity
and pore pressure elds can be split into two components, i.e. propagation and consolidation as
v =v
0
+v
1
and p
w
= p
w0
+p
w1
, where the subindices 0 and 1 refer, respectively, to propagation
and consolidation.
The equations of the propagationconsolidation model are
j
Dv
0
Dt
=jb+divo with divv
0
=0 (10)
and
Dp
w
Dt
=
*
*x
3
_
c
v
*p
w
*x
3
_
(11)
where j is the mixture density and c
v
the coefcient of consolidation. They can be obtained from
the Swansea displacementpore pressure form of BiotZienkiewicz equations.
The second simplication is also based on assuming that landslides have average depths which
are small in comparison with their length or width. In this case, it is possible to simplify the 3D
propagationconsolidation model described above by integrating its equations along the vertical
axis. The resulting 2D depth-integrated model presents an excellent combination of accuracy
and simplicity providing important information such as velocity of propagation, time to reach a
particular place and depth of the ow at a certain location.
Depth-integrated models have been frequently used in the past to model ow-like landslides. It
is worth mentioning the pioneering work of Hutter and coworkers [24, 25], and those of Laigle and
Coussot [26], McDougall and Hungr [27] and the authors [2830]. As regards coupling between
soil skeleton and pore water, it is curious to note that its development has taken place later than that
of soils. We can mention here the work of Hutchinson [31], who proposed a sliding consolidation
model to predict runout of landslides, Iverson and Denlinger [32], Pastor et al. [28, 30] and Quecedo
et al. [29].
We will use the reference system given in Figure 1, where we have depicted some magnitudes
of interest which will be used in this section.
The depth-integrated equations are obtained by integration along depth and using Leibnitz rule.
We will describe here their lagrangian form. First of all, we will dene the material derivative as

d
dt
=
*
*t
+ v
j
*
*x
j
(12)
where the overbar denotes depth-integrated magnitudes, for instance, v
j
=
_
v
j
dx
3
/h.
From here, we obtain the lagrangian forms of the balance of mass, depth-integrated equation:

dh
dt
+h
* v
j
*x
j
=e
R
(13a)
where e
R
is the erosion rate [L
1
].
The balance of momentum equation is
h

d
dt
v
i

*
*x
i
_
1
2
b
3
h
2
_
=
1
j
*
*x
j
(h o

ij
)+b
i
h+
1
j
|N
B
|t
B
i
e
R
v
i
(13b)
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
462 M. PASTOR ET AL.
Figure 1. Reference system and notation used in the analysis.
where b
i
is the body force acting along X
i
. In the above equation, we have introduced the
decomposition
o
ij
= po
ij
+o

ij
with p=
1
2
jb
3
h and o

ij
= o
ij
+ po
ij
(14)
The term t
B
i
is the i th component of the normal stress acting on the basal surface, and |N
B
| is
|N
B
| =((*Z
2
/*x
1
)+(*Z
2
/*x
2
)+1)
1/2
, where Z is the height of the basal surface. Finally, after
integrating the vertical consolidation equation in depth we arrive at:

d
dt
( p
w
h)=c
v
*p
w
*x
3

Z+h
Z
(15)
Next, we will assume that the pore pressure can be approximated as
p
w
(x
1
, x
2
, x
3
, t ) =
Np
w

k=1
P
k
(x
1
, x
2
, t )N
k
(x
3
) (16)
Choosing
N
k
(x
3
) =cos
(2k 1)
2h
(x
3
Z), k =1, Np
w
(17)
and keeping only the rst term, we have
p
w
(x
1
, x
2
, x
3
, t ) =P
1
(x
1
, x
2
, t ) cos

2h
(x
3
Z) (18)
from where we obtain the depth-integrated equation

dP
1
dt
=

2
4h
2
c
v
P
1
(19)
which is the lagrangian form of the vertically integrated 1D consolidation equation. It is important
to note that the results obtained above depend on the rheological model chosen, from which we
will obtain the basal friction and the depth-integrated stress tensor o

ij
. Here, we will use a simple
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 463
frictional uid model, where the basal friction depends on the effective normal stress acting at the
bottom and the friction coefcient between the sliding mass and the bottom material.
3. CONSTITUTIVE MODELLING IN GEOTECHNICAL ENGINEERING
3.1. Historic remarks
The rst relevant model describing the constitutive behaviour of materials after Robert Hookes
lawut tension sic visis the Coulomb failure criterion which applies to geomaterials [1]. Almost
one century later, Tresca [33] proposed his failure criterion, which was generalized by Levy
to three-dimensional conditions [34]. The plasticity theory continued its development with the
contributions of Huber [35], Von Mises [36], Prandtl [37], Reuss [38] and Melan [39]. Finally,
Drucker established in 1949 [40] the principles of what we know today as the classical plasticity
theory, introducing the concepts of load surface, loadingunloading and neutral loading directions,
consistency and unicity.
Tresca and von Mises criteria were not suitable for geomaterials, whose behaviour depends on
the conning pressure. Drucker and Prager [41] proposed an elastoplastic model where the yield
surface depended on the conning pressure. Zienkiewicz et al. [42] made a substantial contribution
in the implementation of constitutive models such as classical elastoplasticity and viscoplasticity
within the nite element method, and the application of non-linear numerical procedures including
the unication of various different classical material models into a single convenient model, still
used in many nite element software packages. Nevertheless, one of the important limitations of
the classical models was that the yield surface was open and plastic deformations occurring along
both isotropic consolidation and oedometric paths were not reproduced. In addition to this, the
associative rule predicted dilative behaviour in failure, which is not observed in soft clays.
Most of the models currently used in geotechnical engineering are based on the critical
state theory developed at Cambridge University in the 1950s and 1960s [4345]. The main
ingredients of the theory were the following:
(i) There exists a critical state where the soil, once it has reached the residual conditions
after failure ows at constant deviatoric and conning stress at constant volume.
(ii) Soil behaviour can be reproduced within the framework of the classical plasticity
theory, using closed yield and plastic potential surfaces that coincide, i.e. ow rule is
associated.
(iii) The size of the yield surface depends on the volumetric plastic strain.
This model reproduced with good accuracy the behaviour of normally consolidated clays,
but presented important limitations when applied to overconsolidated and granular soils.
For the former, the problem consists of the model being unable to reproduce plastic defor-
mations inside the yield surface. For the latter, there were problems both for very dense
sands in drained conditions and for very loose sands under undrained conditions exhibiting
liquefaction. The solution consisted of including a deviatoric strain-dependent deviatoric
hardening together with a non-associative ow rule, and was proposed by Nova [46]
and Wilde [47].
At the beginning of the 1980s, although great advances had been made in constitutive modelling
of soils, there was no model able to reproduce the soil behaviour under cyclic loading conditions.
The problem was important in the design of road pavements, railroad platforms, foundation of
marine structures and seismic engineering.
Laboratory tests have been performed since 1943 [48], and the rst cyclic triaxial was built
by Buchanan and Kuri in 1956 [49]; the contributions by Seed and coworkers [50, 51] are worth
mentioning. One key result obtained by Seed and McNeill [52] is that the cyclic behaviour of soils
exhibiting similar properties under monotonic loading can be very much different, hence cyclic
testing is crucial.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
464 M. PASTOR ET AL.
At this time, testing machines were applied to explore the behaviour of soils under seismic
loading, including liquefaction, and it is not surprising that the rst tests were performed by Prof.
Seed and coworkers [5355]. It is also worth mentioning the contribution of Castro [56] concerning
the liquefaction of sands.
In Japan, Ishihara and coworkers performed a systematic research aiming to fully characterize
the behaviour of both cohesive and granular soils under seismic loading, special attention being
paid to liquefaction and cyclic mobility phenomena [57, 58].
The third area, in addition to trafc and seismic loading, is marine geotechnics. Oil production
structures existed by the end of the 19th century in California, and were used in lake Maracaibo
in 1922. The size of offshore structures and the depth where they operated have been increasing
since then. Indeed, the crisis of 1973, together with the discovery of the oil elds in the North
Sea, provided an important impulsion to marine engineering technology, with existing platforms
such as Ekosk built in 1973 on 70 m of depth water, and the Troll Condeep built in 1995 with
303 m depth. The environmental actionswind, currents and waves-induced high cyclic loads in the
foundation. The problem was different from seismic engineering, where the number of cycles is
much smaller, and required further laboratory investigation. The Norwegian Geotechnical Institute
of Oslo, among other centres, contributed both to improving laboratory testing techniques and the
development of simplied design methods [59]. Another demanding problem was the closure of
Dutch dykes, such as the Oosterschelde, with important cyclic loading due to tides [60].
At the end of the 1970s, there existed a large number of tests describing the behaviour of soils
under the cyclic loading induced by trafc, earthquakes and ocean waves, however, no complete
constitutive model able to describe the phenomena observed in laboratory tests has been devel-
oped yet.
3.2. Generalized plasticity: the basic model
The generalized plasticity theory was introduced by Zienkiewicz and Mroz [61] and elaborated
by Pastor and Zienkiewicz [62]. It provides a framework within which accurate models can
be developed to describe softening and liquefaction under monotonic and cyclic loading. The
basic model has been recently extended by Tamagnini and Pastor [63], Fernndez Merodo
et al. [64] and Manzanal [65, 66] to non-saturated and collapsible soils.
The starting point of the generalized plasticity theory is to assume that the incremental
response of the material to an increment of stress depends on its direction. The dependence is
based on the denition of two tensorial zones, loading and unloading, by introducing a unit
tensor n such that
d =C
L
: do for n: do>0 (loading)
d =C
U
: do n: do<0 (unloading)
(20)
In the above equation, C
L
and C
U
are the constitutive tensors for loading and unloading
situations. Neutral loading corresponds to the limit case for which
n: do=0
Continuity between loading and unloading states requires that constitutive tensors for loading
and unloading are of the form
C
U
=C
e
+
1
H
U
n
gU
n and C
U
=C
e
+
1
H
U
n
gU
n (21)
where n
gL
and n
gU
are the arbitrary tensors of unit norm and H
L/U
two scalar functions
dened as loading and unloading plastic moduli. It can be very easily veried that both laws
predict the same strain increment under neutral loading where both expressions are valid and
hence non-uniqueness is avoided.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 465
The strain increment can be decomposed into two parts, elastic and plastic, as:
de = de
e
+de
p
de
e
= C
e
: dr
de
p
=
1
H
L/U
n
gL/U
n
(22)
The main advantage of the generalized plasticity theory is that all ingredients can be postulated
without introducing any yield or plastic potential surface. Moreover, it can be seen that both
classical plasticity and bounding surface plasticity models are special cases of the GPT.
Pastor, Zienkiewicz and Chan [62] proposed a simple model for sands and clays under
monotonic and cyclic loading.
The main features of sand behaviour under monotonic and cyclic loading are the following:
(i) Volumetric deformations depend mainly on the stress ratio p=q/p. There is a characteristic
value p=M
g
at which the behaviour changes from contractive to dilative. Failure at constant
volume takes place also at this line, referred to as characteristic state line, and it can be
interpreted as a critical state Line for granular soils. The basic idea behind is that the soil,
before failure, crosses a state at which there is no volume change, and comes back to it at
residual conditions.
(ii) Very loose sands exhibit compaction under shearing, which results in an increase of pore
pressures when the loading process is not fully drained. In this limit, liquefaction can occur.
(iii) Dense sands exhibit dilation once the characteristic state line has been crossed. Dilation
causes softening, and the strength decreases after a peak has been reached. Here, localization
of strain in shear bands obscures the experimental results as the specimen is not homogeneous.
(iv) Under cyclic loading we observe the same compaction and dilation patterns. Plastic deforma-
tion occurs and the soil compacts progressively or the pore pressure increases. Liquefaction
under cyclic loading is just the result of the increase of the pore pressure and the mechanism
that is observed in monotonic loading.
(v) Medium dense sands under undrained cyclic loading develop a special type of behaviour
which is referred to as cyclic mobility. The difference with liquefaction consists of dilation
that causes the pore pressure to decrease, hardening the soil in turn.
Taking into account all the experimental facts described above, it is possible to develop a model
within the generalized plasticity theory as follows:
First of all, the direction of plastic ow in the ( p, q) plane is postulated as
n
T
g
= (n
gv
, n
gs
)
n
gv
= d
g
/(1+d
2
g
)
1/2
n
gs
= 1/(1+d
2
g
)
1/2
(23)
where the dilatancy d
g
, which depends on two material parameters : and Mg is given by
d
g
=(1+:)(M
g
p)
The loadingunloading discriminating relation n is obtained in a similar way
n
T
= (n
v
, n
s
)
n
v
= d
f
/(1+d
2
f
)
1/2
n
s
= 1/(1+d
2
f
)
1/2
d
f
= (1+:)(M
f
p)
(24)
where M
f
is a material parameter.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
466 M. PASTOR ET AL.
The third ingredient is the plastic modulus, which is dened both for loading and unloading.
During loading, we will assume
H
L
=H
0
p

H
f
(H
v
+H
s
)H
DM
(25)
where H
0
is a constitutive parameter. In the above equation, H
f
is given by
H
f
=
_
1
p
p
f
_
4
and
p
f
=
_
1+
1
:
_
M
f
This factor varies between 1 at q =0 and 0 at the straight line tangent to the yield surface at the
origin.
The terms H
v
, H
s
and H
DM
refer, respectively, to the volumetric and deviatoric strain hardening
and the discrete memory. They are given by:
H
v
=
_
1
p
M
g
_
H
s
= [
0
[
1
exp([
0
)
H
DM
=
_

max

(26)
Let us now consider each term. The volumetric term is zero at CSL, and therefore, failure
would take place there if H
s
were zero. It can be observed in triaxial tests that both in drained and
undrained processes, the stress paths are able to cross this line. The role of H
s
is to prevent failure
at this stage, but to allow it at residual conditions. This is achieved by making H
s
to depend on
the accumulated deviatoric strain dened from d=(de
p
: de
p
)
1/2
, where de
p
is the increment
of the plastic deviatoric strain tensor. In the above equation, [
0
, [
1
and are model parameters.
Finally, the variable is a measure of the stress intensity, which has been taken as
=p

_
1
_
1+:
:
_
p
M
_
1/:
(27)
As an example, we depict in Figure 2 the experimental results and model predictions for the
liquefaction of very loose sand under undrained loading in a triaxial apparatus.
3.3. Further developments: state parameter-based generalized plasticity for sands
The above basic model presented the limitation of treating sands of initial different densities or
conning pressures as different materials with different constitutive parameters. In some situations,
a granular material is subjected to large changes in conning pressures, such that occur in large
earth dams, for instance, where the material placed at the bottom has a conning pressure increasing
from zero to values of the order of MPa. In fact, the loose or dense type of behaviour depends
both on conning pressure and on density. One effective way of accounting for this effect is to
introduce a state parameter in the model.
Uriel [67] proposed a simple constitutive model able to describe with a single set of parameters
the behaviour of dense and loose sands, and the effect of the conning pressure. The most widely
accepted state parameter today is that proposed by Been and Jefferies [68]. It is dened as the
difference in the current voids ratio and the voids ratio at critical state under the same conning
pressure
=ee
c
(28)
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 467
Figure 2. Liquefaction of very loose banding sand (data from [56] and predictions from [62]).
The critical state line is dened as [69]:
e=e
atm
+z
_
p

atm
_
c
(29)
In the above expression, z and
c
are two material parameters, and e
atm
is a reference voids
ratio at pressure p

atm
. Of course, other alternative descriptions of the CSL can be chosen. We have
used for dilatancy the law [70]
d
g
=
d
0g
M
g
(M
g
exp(m)p) with d
0g
=(1+:)M
g
(30)
where m is a new constitutive parameter which can be obtained from experiments. With this
information it is possible to dene the unit tensor n
g
that characterizes the plastic ow direction.
The second ingredient of the model is the loadingunloading unit vector n
f
which is obtained
using from d
f
(see Equation (24))
d
f
=
d
0 f
M
f
(M
f
exp(m)p) (31)
where d
0 f
is a material parameter. Manzanal et al. [71] has proposed the following expression
for M
f
M
f
M
g
=h
1
h
2

q
(32)
which depends on the parameters h
1
and h
2
.
Finally, the third ingredient of the model, which is the plastic modulus is
H =H
0
_
p

0
H
DM
H
f
(H
v
+H
s
) (33)
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
468 M. PASTOR ET AL.
po= 980kPa
0
200
400
600
800
1000
1200
1400
0 200 400 600 800 1000 1200
p [kPa]
q

[
k
P
a
]
Dr = 53%
Dr = 45%
Dr = 42%
Dr = 35%
po= 980kPa
0
200
400
600
800
1000
1200
1400
0 0.02 0.04 0.06 0.08
Axial strain
q

[
k
P
a
]
Dr = 53%
Dr = 45%
Dr = 42%
Dr = 35%
po= 980kPa
0
200
400
600
800
1000
1200
1400
0 0.02 0.04 0.06 0.08
Axial strain
P
o
r
e

p
r
e
s
s
u
r
e

[
k
P
a
]
Dr = 53%
Dr = 45%
Dr = 42%
Dr = 35%
Figure 3. Experimental results and model predictions for banding sand at different relative densities [56].
where
H
0
=H

0
exp([

q
)
H
v
=H
v0
(p
P
p) with p
P
=M
g
exp([
v
)
(34)
In the above expression, we have used the peak stress ratio p
P
proposed by Li and Dafalias [70]
and introduced the new parameters H

0
and [

0
.
Using this state parameter-based model, it is possible to reproduce the set of tests on Banding
Sand reported by Castro [56] with a single set of parameters. Figure 3 shows the model predictions
together with the experimental results.
3.4. Generalized plasticity for unsaturated soils
The rst generalized plasticity model for unsaturated soils was proposed by Bolzon, Schreer and
Zienkiewicz in 1996 [72]. It was based on the use of the denition of the effective stress tensor
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 469
proposed by Schreer [11], but presented a limitation when reproducing collapse tests where a
sample of soil is inundated. The second extension of the basic model was proposed by Tamagnini
and Pastor [63] using the same denition of effective stress as Bolzon et al. [72], but taking into
account the pairs of work-coupled variables proposed by Houlsby [13], who obtained the increment
of work as
oW =(o+S
r
s I p
a
I )o+s.(noS
r
) (35)
where S
r
is the degree of saturation, s the matrix suction, s = p
a
p
w
, p
a
and p
w
the pore pressures
of air and water, respectively.
This model was rened by Santagiuliana and Schreer [73]. Both models were able to describe
collapse, but presented some limitations that have been addressed by Manzanal [65, 66] and which
are described next.
The model incorporates the state parameter dened in the previous section, which requires a
suitable denition of the critical state line for unsaturated soils. The problem is that there is no
unique CSL because it depends on suction. In order to normalize it, we have followed the work
of Gallipoli et al. [74] who introduced a cementation parameter dened as
= f (s)(1S
r
) (36)
where the function f (s) is the relationship between stabilizing pressure at a given suction s and
at zero suction proposed by Haines [7] and Fisher [75]. Based on these ideas, we have proposed
a relation between the effective pressure at a given suction and saturated at the same void ratio
p

s
=exp(g()) (37)
where g() =a{exp(b)1} (38)
In order to illustrate this normalization, we provide in Figure 4 the experimental results obtained
by Sivakumar [76] on a cohesive soil (Speswhite kaolin) without and with the proposed normal-
ization. In Figure 4(b), we have plotted the relationship between the void ratio and the effective
conning pressure at saturation, obtained using (37) and (38), i.e.
p

s
= p

exp(g())
Concerning the projection of the critical state line on the p

q plane, we have introduced a


second normalization which reduces some dispersion observed in the experiments. It is based on
the denition of a residual degree of saturation S
r0
:
S
re
=
S
r
S
r0
1S
r0
(39)
This modied degree of saturation is used in the denition of effective stress, and the results can
be seen in Figure 5, which display the CSL with and without this modication for a silty soil. The
experimental results have been taken from Matouk et al. [77].
We have assumed, following the original idea of Tamagnini and Pastor [63] that the increment
of strain can be decomposed as
d
ij
=d
e
ij
+d
p
ij
o+d
p
ijs
(40)
where the last term is given by:
d
p
s
=
1
H
b
n
g
ds (41)
The plastic modulus H
b
is given by
H
b
=w()H
0
_
p

p
atm
H
DM
H
v
(42)
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
470 M. PASTOR ET AL.
Speswhite kaolin
(Sivakumar, [76] )
0.8
0.9
1.0
1.1
1.2
0 0 0 1 0 0 1 0 1
p [kPa]
V
o
i
d

r
a
t
i
o
,

e
s = 0
s = 100
s = 200
s = 300
(a)
0.8
0.9
1.0
1.1
1.2
0 0 0 1 0 0 1 0 1
V
o
i
d

r
a
t
i
o
,

e

s = 0
s = 100
s = 200
s = 300
kPa
s
p

(b)
Figure 4. (a) Critical states at different suctions (from [76]) and (b) normal-
ization of CSL (experiments from [76]).
where
H
DM
=
_

max
J
s

(43)
is a modied discrete memory function incorporating the effect of the suction
J
s
=exp(c.g()) (44)
w() incorporates the effect of the cementing parameter dened above
w() =
_
{1exp[g()]
2
}
2
(wetting)
1 (drying)
(45)
and H
0
, H
f
, H
v
and H
s
are the same functions dened for saturated soils.
The model is completed with a suitable water retention curve, including hysteretic effects [65].
In order to show the predictive capability of the model, we will present a comparison between
model predictions and experimental results for Kurnell sand, which is a ne granular soil. The
experiments have been described by Rusell and Khalili [78]. It is important to note that all the
model predictions were done using a single set of parameters which characterizes the behaviour of
Kurnell sand at different conning pressures, densities, suctions and stress paths. Figure 6 shows
the experimental results and the model predictions for a drained triaxial on Kurnell sand, under
two initial net conning pressures of 50 and 100 kPa. Suction was kept constant during both cases
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 471
= Sra
0
200
400
600
800
1000
1200
1400
1600
0 200 400 600 800 1000
p[kPa]
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

q

[
k
P
a
]
s = 0
s = 80
s = 150
s = 400
s = 600
Silty soil (Matouk et al.,[77] )
= Sre
0
200
400
600
800
1000
1200
1400
1600
0 200 400 600 800 1000
p [kPa]
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

q

[
k
P
a
]
s = 0
s = 80
s = 150
s = 400
s = 600
Figure 5. CSL on p

q plane: without correction (top) and using the


effective degree of Saturation (bottom).
at a value of s =50, 100, 200 and 400 kPa. The model predictions agree well with the experimental
results.
4. NUMERICAL MODELLING: FINITE ELEMENTS
4.1. The origins: rst applications in soil mechanics
The roots of the nite element method can be found in the works of Hrennikoff [79], McHenry [80]
and Courant [81], who studied the application of variational methods to solve a torsion problem.
The idea of a discretization based on a partition of the continuum in elements and nodes that
connect them was proposed by Turner [82], Argyris [83, 84] and Clough [85, 86]. All these works
are based on a structural approach, where the concepts of stiffness matrix and assembling are found.
Owing to the coupling that exists between soil skeleton and pore uids in geomaterials, the
mathematical models incorporate an equation describing the ow of pore water within the soil.
Here we nd the rst crucial contribution by Zienkiewicz and Cheung [87].
The rst coupled model for the analysis of soil behaviour is due to Sandhu and Wilson [88],
who analyzed the consolidation of a soil layer under a footing using quadratic polynomials to
approximate the displacements and linear for the pore pressure. This formulation was extended
to more complex problems by Yokoo et al. [89, 90], Smith and Hobbs [91] and Valliappan
et al. [92]. Smith and Hobbs [93] proposed a model for consolidation comparing the computed
results to experiments obtained in a Cambridge centrifuge machine. Again, different orders of
interpolation were used for displacements and pressures. It is worth mentioning the contributions
of Lewis et al. [94] where changes in permeability during consolidation were considered for the
rst time.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
472 M. PASTOR ET AL.
0
100
200
300
400
500
600
700
800
0.00 0.10 0.20 0.30 0.40
Deviatoric strain
D
e
v
i
a
t
o
r
i
c
s
t
r
e
s
s
,

q

[
k
P
a
]
10050D-DTest 5050L-DTest
10050D-DModel 5050L-DModel
-0.14
-0.12
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.00 0.10 0.20 0.30 0.40
Deviatoric strain
V
o
l
u
m
e
t
r
i
c
s
t
r
a
i
n
0
100
200
300
400
500
600
700
0.00 0.10 0.20 0.30 0.40
Deviatoric strain
D
e
v
i
a
t
o
r
i
c
s
t
r
e
s
s
,

q

[
k
P
a
]
50100L-DTest 100100D-DTest
50100L-DModel 100100D-DModel
-0.14
-0.12
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.00 0.10 0.20 0.30 0.40
Deviatoric strain
V
o
l
u
m
e
t
r
i
c
s
t
r
a
i
n
0
200
400
600
800
0.00 0.10 0.20 0.30 0.40
Deviatoric strain
D
e
v
i
a
t
o
r
i
c
s
t
r
e
s
s
,

q

[
k
P
a
]
50200L-DTest 100200D-DTest
50200L-DModel 100200D-DModel
-0.12
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0
Deviatoric strain
V
o
l
u
m
e
t
r
i
c
s
t
r
a
i
n
0
100
200
300
400
500
600
700
800
0.00 0.10 0.20 0.30 0.40
Deviatoric strain
D
e
v
i
a
t
o
r
i
c
s
t
r
e
s
s
,

q

[
k
P
a
]
50400L-DTest 100400D-DTest
50400L-DModel 100400D-DModel
-0.14
-0.12
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.00 0.10 0.20 0.30 0.40
Deviatoric strain
V
o
l
u
m
e
t
r
i
c
s
t
r
a
i
n
0.1 0.2 0.3 0.4
(a) s = 50kPa
(b) s = 100kPa
(c) s = 200kPa
(d) s = 400kPa
(h)
Figure 6. Comparison between model predictions and experimental data of drained triaxial compression
tests at constant suction on Kurnell sand (experimental data from [78]).
4.2. Swansea models for the coupled behaviour of soils
Using the nite element method [95, 96] for spatial discretization, the equations of the up
formulation are obtained as follows:
The dynamic form:
M u+P(u)Qp=f
u
G u+Q
T
u+S p+Hp=f
p
(46)
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 473
A simplication is obtained when the accelerations are neglected, which is the consolidation form:
P(u)Qp = f
u
Q
T
u+S p+Hp = f
p
(47a)
Finally, the static and steady state form is obtained if the variations with respect to time are
assumed to be very small:
P(u)Qp=f
u
Hp=f
p
(47b)
In the above equations, M is the mass matrix, G is the dynamic coupling matrix, S is the compress-
ibility matrix, H is the permeability matrix (sometimes a viscous damping term is added to the
rst equation in (46), u and p are the vectors for the nodal value of u and p, respectively, Q is
the coupling matrix and P(u) is non-linear internal force vector which is given by
P(u) =
_

B
T
o

d (48)
where B is the usual displacementstrain transformation matrix for the nite element method,
o

is the vector form of the effective stress tensor at the integration (Gauss) points and is
the domain concerned. The above equations are systems of ordinary differential equations which
can be integrated along time using a suitable scheme, such as the generalized Newmark method
presented by Katona and Zienkiewicz [97].
Coupled models have been applied to a great variety of problems, such as subsidence. Here we
can mention the work of Schreer et al. [98], Morgan et al. [99], and the classical textbook by
Lewis and Schreer [12].
4.3. SPH models for uidized soils
As in Computational Fluid Mechanics problems, uidized soils can be modelled using two alter-
native approaches. The rst is Eulerian, based on either a non-structured or structured mesh within
which the material ows. In the case of uidized landslides, one of the main problems is the need
for a very ne computational mesh for both the terrain information and for the uidized soil.
Lagrangian methods allow the separation of both meshes, with an important economy of compu-
tational effort. If we combine a Lagrangian method with a mesh based discretization technique,
we will nd problems as soon as the mesh deforms, making it necessary to use mesh renement.
As an alternative, meshless methods that do not rely on meshes avoid distortion problems in an
elegant way. Among the different meshless methods that have been used for uidized soils, we
can mention the material point model [100104], and the SPH [27, 30, 105107].
We will next describe the SPH method, which has been extended by the authors to depth-
integrated coupled problems.
Smoothed particle hydrodynamics (SPH) is a meshless method introduced independently by
Lucy [108] and Gingold and Monaghan [109] and rst applied to astrophysical modelling, a
domain where SPH presents important advantages over other methods. SPH is well suited for
hydrodynamics, and researchers have applied it to a variety of problems, like those described in
Gingold and Monaghan [110], Monaghan and Gingold [111], Monaghan et al. [112], Bonet and
S. Kulasegaram [113] and Monaghan et al. [114], just to mention a few.
In the case of the depth integrated, coupled equations describing the movement of uidized
landslides, we will introduce a set of nodes {x
K
} with K =1, . . . , N and the nodal variables:
h
I
height of the landslide at node I
v
I
depth averaged, 2D velocity
t
b
I
surface force vector at the bottom
o

I
depth averaged modied stress tensor
P
1I
pore pressure at the basal surface
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
474 M. PASTOR ET AL.
If the 2D area associated with node I is
I
, we will introduce for convenience:
(i) a ctitious mass m
I
moving with this node m
I
=
I
h
I
,
(ii) and, an averaged pressure term p
I
, given by p
I
=
1
2
b
3
h
2
I
It is important to note that m
I
has no physical meaning, as when node I moves, the material
contained in a column of base
I
has entered it or will leave it as the column moves with an
averaged velocity which is not the same for all particles in it.
There are several possible alternatives for the equations, according to the discretized form chosen
for the differential operators results. We will show those obtained with the three symmetrized
forms:

dh
I
dt
=h
I

J
m
J
h
J
v
IJ
gradW
IJ
where we have introduced v
IJ
=v
I
v
J
Alternatively, the height can be obtained once the position of the nodes is known as:
h
I
=h(x
I
) =

J
h
J

J
W
IJ
=

J
m
J
W
IJ
The discretized balance of linear momentum equation is:

d
dt
v
I
=

J
m
J
_
p
I
h
2
I
+
p
J
h
2
J
_
grad W
IJ
+
1
j

J
m
J
_
o
I
h
2
I
+
o
J
h
2
J
_
grad W
IJ
+b+
1
jh
I
|N
B
|t
B
I
Finally, the SPH discretized form of the basal pore pressure dissipation is:

d
dt
P
1I
=

2
c
v
4h
I
P
1I
So far, we have discretized the equations of balance of mass, balance of momentum and pore
pressure dissipation. The resulting equations are ODEs which can be integrated in time using a
scheme such as Leap Frog or RungeKutta (second or fourth order).
5. SELECTED APPLICATIONS
5.1. Waves over a seabed without a structure [115]
Teh [116] carried out a series of wave ume experiments to investigate the nature of liquefaction
around a pipeline. In additon to measuring the pore pressure in the vicinity of the pipeline, he also
measured the pore pressure build-up in the far eld. Figure 7 shows a schematic of the ume and
the silt box. The properties of the silt have been listed in two different references [116] and are
summarized in Tables I and II.
Figure 7. Wave ume layout.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 475
Table I. Silt physical properties1 [116].
Property Value Units Method of measurement
D
10
10 m AccuSizer 780 machine
D
50
33 m AccuSizer 780 machine
D
60
47 m AccuSizer 780 machine
D
90
120 m AccuSizer 780 machine
Uniformity coefcient 4.7
Specic gravity (G
s
) 2.71 BS 1377:1975 Clause 2.6.2
Minimum void ratio 0.39 ASTM Standards: D4253 (04.08)
Maximum void ratio 1.18 ASTM Standards: D4254 (04.08)
Friction angle at critical state 36

Triaxial test
Table II. Silt physical properties2 [116].
Parameter Value Method of measurement
D
10
(m) 3 BS 1377:1990 Part 2
D
60
(m) 45 BS 1377:1990 Part 2
Uniformity coefcient 14.1
Specic gravity 2.7 BS 1377:1990 Part 2
Liquid limit (%) 21 BS 1377:1990 Part 2
Plastic limit (%) 16.6 BS 1377:1990 Part 2
Plasticity index (%) 4.4 BS 1377:1990 Part 2
Minimum void ratio 0.46 BS 1377:1990 Part 4
Maximum void ratio 1.67 BS 1377:1990 Part 4
Optimum moisture content (%) 11.5 BS 1377:1990 Part 4
Maximum dry density (kg/m
3
) 1800 BS 1377:1990 Part 4
Permeability (m/s) 210
7
to 610
7
BS 1377:1990 Part 6
Coefcient of consolidation (m
2
/year) 2001200
Table III. Summary of test parameters for model validation (after [116]).
Test No. Soil depth (m) I
D
(%) H (m) H (m) T (s)
TA04 0.25 48 0.45 0.095 1.25
TA03 0.25 40 0.45 0.15 1.25
The experiments were carried out on silt with a range of relative densities. When the bed
was in a loose to medium dense state (i.e. 30%<I
D
<50%) it liqueed quite quickly under
moderate wave conditions (i.e. H<0.1m). Two test cases, TA04 and TA03, were used to assess the
ability of the model to predict the progressive build-up of pore pressure under cyclic loading (see
Table III).
The PastorZienkiewicz [117] mark-III model (PZ3) requires a large number of material param-
eters to describe the loading and unloading behaviour of the silt. Usually, these are obtained from
the detailed laboratory testing such as cyclic triaxial tests and a method of parametric identication
for the PZ3 model has been proposed in [17]. However, these results were not available for this
exercise. Instead, one set of test results, TA04, was used to calibrate the model parameters and
these were then applied to a separate test, TA03. The nite element mesh used in this study is
shown in Figure 8.
The rst step in calibrating the model was to analyze the variation in the pore pressure amplitude
with depth prior to liquefaction. This enables a calibration of the two elastic parameters, the bulk
modulus of the soil and the water (or the degree of saturation). Figure 9 shows a plot of the
experimental results versus the numerical model. It was found that in order to obtain the rapid
decay in pore pressure amplitude with depth, it was necessary to decrease the degree of saturation
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
476 M. PASTOR ET AL.
Figure 8. Typical nite element mesh for the silt box.
Figure 9. Calibration of numerical model against ume data (test TA04). Nodes represent the experimental
data and lines represent the numerical solutions.
from 1 to 0.95. Clearly, the response is highly dependent on this parameter. During the experiments
it was noted that small gas bubbles escaped from the bed. The source of these bubbles is not
certain, although some gases may have been trapped as the ume lled with water or the release
of air dissolved in water by a water pressure decreases as explained in Gawin and Sanavia [118].
Alternatively, it is possible that the bubbles may have been generated during the regular ushing
of the bed between runs to restore the bed to a loose state. The pressure variation was not greatly
inuenced by the choice of either the shear modulus, G, or the permeability, k.
Having obtained a reasonable t to the elastic parameters, the next task was to determine the
value of the parameters that determine the plastic behaviour, such as the loading and unloading
modulus, H
0
and H
u0
, and the slope of the critical state lines, M
g
and M
f
. For all the other material
constants, such as :
g
, :
f
, [
0
and [
1
, the recommended values were used (see Table IV).
Although it was not possible to t each parameter independently, typical values for loose silt
were adopted and then slightly adjusted in order to t the observed data. Figures 10 and 11 show
a comparison of the observed against predicted pore pressure response at two different depths
for test TA04. It is noticeable that the model tends to underestimate the pore pressure amplitude
post-liquefaction. This suggests that the model predicts response in the liqueed state than what
was observed in the experiment. However, it was not possible to reduce the stiffness any further as
it introduced numerical instabilities. It should be noted that in order to properly model the liqueed
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 477
Table IV. Silt properties required for FE model.
Parameter Description Value Unit
Kev0c Bulk modulus of the soil 3 MPa
Kes0c Three times the shear modulus of the soil 1.5 MPa
M
g
Slope of the critical state line for determination
of loading vector
1.42
M
f
Slope of the critical state line for determination
of plastic strain vector
0.90
:
g
Parameter which determines relationship of dila-
tancy with stress ratio for loading vector
0.45
:
f
Parameter which determines relationship of dila-
tancy with stress ratio for plastic strain vector
0.45
[
0
Constant for shear hardening 4.2
H
1
Constant for shear hardening 0.2
H
0
Constant for the loading plastic modulus 1050
H
u0
Constant for the unloading plastic modulus 4 MPa
v Void ratio 0.3
k Darcy coefcient of permeability 2.510
6
m/s
p

o
Reference mean conning stress at which the
moduli are evaluated
10 kPa
S
r
Degree of saturation of the pore uid 0.95
Figure 10. Pore pressure variation at z =0.04m for test TA04.
Figure 11. Pore pressure variation at z =0.085m for test TA04.
state, it would be necessary to modify or include new constitutive relations. However, the primary
purpose of this study is not to investigate the post-liquefaction behaviour, but rather to model the
behaviour of the bed and structures leading up to the point of liquefaction.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
478 M. PASTOR ET AL.
Figure 12. Stress path at z =0.045m.
Figure 13. Pore pressure variation at z =0.185m for test TA04.
Figure 12 shows the stress path at z =0.045m for test TA04. As the pore pressure increases
the mean effective stress decreases and cycles in towards zero.
In general, the model, with the material properties as per Table IV, was able to adequately
reproduce the observed behaviour of the bed and showed the commonly observed trend that the
bed liquees from the top down.
To further support the validity of the model, a separate comparison was undertaken with the
results from test TA03. The primary difference between tests TA04 and TA03 is that the wave
height in TA03 is signicantly larger (H =0.15m for TA03 compared with H =0.095m for TA04).
Although TA03 had a slightly lower initial density of 1920kg/m
3
compared with 1950kg/m
3
for
TA04, they are sufciently close to warrant using the same material properties (i.e. Table IV).
Figure 13 shows a comparison between the predicted and observed pore pressure at z =0.185m.
First, it should be noted that due to the larger wave height and hence larger cyclic shear stress
ratio, the mean excess pore pressure rises more quickly. For example, at z =0.185m (i.e. 0.185 m
below the midline) it took 33 cycles to liquefy in TA04, whereas it took only 23 cycles in TA03.
For this case the model also displayed the same tendency as in test TA04 of under predicting the
pore pressure uctuations in the liqueed state.
The comparisons with both analytical solutions and experimental data indicate that the model
is able to reproduce both the elastic and inelastic behaviour of a non-cohesive bed under cyclic
wave loading. In the following sections, the problems of liquefaction around structures will be
discussed.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 479
Figure 14. Mesh used for analysis of Las Colinas landslide.
Table V. Parameters of the constitutive model for Las Colinas landslide.
M
g
M
f
: [
0
[
1
p
s0
p
m0
p
r0
[
p
j
m
j
t

1.47 0.3 0.45 1. 0.2 9.0 120 kPa 240 kPa 24 kPa 0.06 8333 1000 0.1
5.2. Dry liquefaction: Las Colinas landslide in El Salvador (2001)
This section presents an analysis of a landslide in dry volcanic soil that took place in Las Colinas,
in the city of San Salvador, during the 13 January earthquake that shook El Salvador in 2001.
An analysis of the problem is based on the eld geometry consisting of a 150 m high slope of
volcanic soil saturated with air subjected to a horizontal seismic action. The mesh used to model
this problem is dened in such terms as can be seen in Figure 14.
A dynamic up
a
formulation is used to analyze this case, where u is displacement of the soil
skeleton and p
a
is the pressure of interstitial air. The analysis requires in the rst place the selection
of adequate constitutive models to account for the constitutive behaviour of the two phases that
constitute the soil: the soil skeleton and the air.
To reproduce the constitutive behaviour of the volcanic soil, a generalized plasticity model
has been used including the improvement for bonded materials. The information available in the
literature about the properties of Las Colinas soil and about tests performed on it is insufcient and
does not have the format required to calibrate the models dened in sections. However, following
Konagai et al. [119], it is possible to estimate some of the parameters of the model. For the rest of
them it is possible to heuristically assign values, based on the qualitative information also available
from these reports.
Based on the test results from Konagai et al. [119], the mass density, void ratio and related
parameters were taken as j
s
=2410kg/m
3
, e
max
=3.044, e
min
=2.297, e
0
=3.0, which correspond
to very loose material. The elastic component of soil behaviour is frequently modelled in its most
general form as non-linearly dependent on effective pressure taking a reference pressure p
0
. The
present work is a highly non-linear type of analysis with a large amount of plastic dissipation
taking place along a signicant lapse of time in a non-trivial geometrical domain with complex
boundary conditions. In this work therefore, aiming only to obtain a rst a approximation of a
complex phenomenon and for the sake of simplicity, the elastic component has been assumed to
behave in a linear fashion.
The elastic constants have been chosen as K =55555kPa, G=41667kPa, and following Konagai
et al. [119], we have chosen [=36.1

and zero cohesion. The generalized plasticity model param-


eters are given in Table V.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
480 M. PASTOR ET AL.
Figure 15. Ground acceleration in NS direction from the El Salvador earthquake in the 13 January 2001.
Coupling between the soil skeleton and the pore air has been assumed. There are many available
references in the literature to obtain a value of water permeability for a specic soil. However, this
is not the case with air permeability. In order to compute air permeability a general expression to
compute the permeability of a specic soil with respect to a given uid will be used
k
f
=
k
intr
j
f
j
f
g
where k
intr
is the intrinsic permeability solely dependent on soil skeleton features, j
f
is the uid
viscosity, j
f
is the uid density, g is the gravity acceleration.
Solving for k
intr
in the previous expression and assuming the intrinsic permeability obtained
with water parameters is the same as that obtained with air parameters it is possible to relate air
permeability with water permeability:
k
a
=k
w
j
a
j
w
j
w
j
a
Assuming that:
k
w
= 1.5e4 m/s as that of a relatively permeable soil
j
w
= 1.0e3 Pa s
j
a
= 1.8e5 Pa s
j
w
= 1.0e3kg/m
3
j
a
= 1.293kg/m
3
the following value for air permeability is obtained: k
a
=1.07e5m/s.
The seismic action is modelled through a base acceleration input obtained from the NS compo-
nent of the Santa Tecla recording of the El Salvador earthquake. The peak ground acceleration
(PGA) was 0.5 g. The acceleration graph is shown in Figure 15.
The analysis is performed in two stages.
In the rst stage, a ground base acceleration is applied to each of the columns that limit the
mesh. In each of these soil columns a tied node boundary condition is applied, combined with
the ground base acceleration condition. The tied node condition means that the nodal variables
(displacements and pore pressure) at every node in the left boundary of a column are enforced to
have the same value as the nodal variables of the corresponding node at the right boundary, that
is, the one with the same height. The result of this computation is a time history for displacements
and pore pressures at each of the nodes throughout the depth of each of the columns. These time
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 481
Figure 16. Evolution in time of p
t
parameter for coupled formulation (up
a
).
histories are input as boundary conditions in the second stage of the analysis. This is equivalent to
assuming that the geometry does not vary for a signicant length at either side of the domain, in
the horizontal direction. In this rst stage of the analysis the central part of the analysis domain,
that is, the slope itself, is not considered.
In the second stage, the base ground acceleration is applied to the central part of the analysis
domain. The time histories for displacements and pore pressures obtained in the rst stage of the
analysis are applied as boundary conditions for the second stage.
For both stages a preliminary static analysis under gravity loading is necessary to obtain the
initial stress conditions for the dynamic analysis.
From the constitutive point of view the response of this type of soil model under earthquake
loading exhibits three distinct and successive phases:
(1) Soil bonds are broken due to plastic volumetric strain. From the constitutive modelling point
of view, the consequence is that parameters p
t
and p
m
reduce to zero.
(2) Once bonds are broken and p
t
= p
m
the soil behaves as sand. Subject to cyclic loading in
undrained conditions a pore pressure build-up takes place due to volumetric plastic strain.
(3) Pore pressure progressively attracts the major portion of total volumetric stresses and effective
stresses evolve towards and nally reach the origin in the pq plane, that is p=q. This
stress state is associated with a quasi zero stiffness and causes the soil to exhibit a uid-like
behaviour. The process dened by the previous phase followed by the present one is usually
referred to as liquefaction.
Some of the most representative results of the analysis are presented in Figures 1518 The
aim of these graphs is to present the evolution in time, as earthquake motion takes place, and
throughout the slope geometry of a series of constitutive parameters ( p
t
), stress variables ( p

that
is effective pressure) and pore pressure.
Figure 16 shows how the p
t
parameter evolves in time towards zero as earthquake motion takes
place. This process starts at the points closest to the surface but gradually propagates to deeper
points. Figure 17 presents the evolution of p
a
. Except for the points strictly at the surface where
a zero pore pressure is prescribed, the points closest to the surface exhibit the characteristic pore
pressure build-up mentioned previously. Finally, Figure 18, presents the evolution of effective
pressure p

. It is also possible to observe how in the points closest to the surface, p progresses
towards a zero value. The propagation in depth of the liqueed state of the soil causes the slope
to be unstable. This instability in turn causes the accelerated movement of the slope surface as
reected by the horizontal displacement plot shown in Figure 19, of node 385 located at midheight
at the slope surface, and ultimately its failure. All these graphs conrm the constitutive response
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
482 M. PASTOR ET AL.
Figure 17. Evolution in time of excess pore pressure p
a
for coupled formulation (up
a
).
Figure 18. Evolution in time of the effective pressure p

for coupled formulation (up


a
).
described in the three-phase sequence previously presented. In order to assess whether these results
are compatible with time evolution of the landslide it would be necessary to establish a comparison
between them and the available in situ information regarding this aspect. According to Evans
and Bent [120], the duration of the landslide was 45 s. Assuming that this lapse of time includes
initiation as well as propagation time evolution, as shown in Figure 19, is compatible with the in
situ information.
5.3. Propagation of fast catastrophic landslides
We present here the case of a owslide where the coupling between pore pressures and the solid
skeleton was crucial. Dawson et al. [121] reported three cases of owslides of coal mine waste
dumps in the Western Canadian Rocky Mountains which they selected among some 50 owslides
which occurred there in the years 19721997.
The owslides propagated distances up to 3500 m, with a modal value of 5001000 m in the
histogram representing the number of owslide events against runout distances.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 483
Figure 19. Horizontal displacement at midslope (node 385) for coupled formulation (up
a
).
Figure 20. Topography of the with isolines of terrain elevation.
Figure 21. Position of SPH nodes at 5, 10, 20 and 30 s.
The case we have selected is Fording Greenhills, where in May 1992 the Cougar 7 dump failed.
The mobilized volume of debris consisted of approximately 200 000 m
3
which slided off the 100 m
high dump. It is important to remark that wet ne-grained layers were found at the foundation
contact, near the crest, and in the debris.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
484 M. PASTOR ET AL.
Figure 22. Depth contours of the owslide at 5, 10, 20 and 30 s.
Figure 23. Contours of normalized pore water pressure.
According to Dawson et al., these ne-grained layers played a crucial role both in the initiation
and the propagation phases. Indeed, the owslide is thought to have been triggered by liquefaction
of the ne-grained layers.
Dawson et al. performed laboratory tests, from which we have obtained a density of 1900kg/m
3
an effective friction angle [

=37

and a characteristic consolidation time of 68 s.


Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 485
We have analyzed this case using the depth averaged coupled, SPH model described in the
preceding sections.
Figure 20 depicts the terrain model used in the computations. The results of the simulations are
given in Figures 2123, where we give the perspectives of the owslide extension and isolines of
the debris depth and pore pressures.
6. CONCLUSIONS
Professor Olek Zienkiewicz contributed in a decisive manner to the development of Computational
Geomechanics by proposing mathematical, constitutive and numerical models able to reproduce the
behaviour of geostructures under a wide variety of loading conditions. We have described here in
a succinct way what in our opinion are his three more decisive contributions, i.e. the mathematical
model describing the coupling between soil skeleton and pore uid for general dynamic conditions,
the generalized plasticity which provides a framework within which complex phenomena such as
soil liquefaction can be modelled, and a nite element scheme for coupled problems. These three
ingredients have been the key for modelling liquefaction of dams under seismic loading and the
behaviour of marine structures.
He also opened research lines, which are active today, from which we have presented a selection,
describing (i) how the original generalized plasticity model can be extended to unsaturated soils
incorporating a state parameter, (ii) how the up
w
model can be extended to fast catastrophic
landslides, and (iii) how uidized soils can be modelled using meshless methods.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the nancial support granted by (i) the Spanish Ministry of Science
and Innovation (MCINN) and (ii) the European Commission (Project SafeLand).
REFERENCES
1. Coulomb CA. Essai sur une application des rgles de maximis et minimis quelques problmes de statique
relatifs larchitecture. Mmoires de Mathmatique et de Physique 1773; 7:343.
2. Reynolds O. On the dilatancy of media composed of rigid particles in contact. Philosophical Magazine 1885;
20:469481.
3. Terzaghi K. The shearing resistance of saturated soils. Proceedings of the First International Conference Soil
Mechanics, Cambridge, 1936; 5456.
4. Biot MA. General theory of three-dimensional consolidation. Journal of Applied Physics 1941; 12(2):155164.
5. Taylor DW. Fundamentals of Soil Mechanics. Wiley: New York, 1948.
6. Skempton AW. Effective stress in soils, concrete, rocks. Proceedings of the Conference on Pore Pressure and
Suction in Soils. Butterworths: London, 1960; 416.
7. Haines WB. Studies in the physical properties of soils II. A note on the cohesion developed by capillary forces
in an ideal soil. Journal of Agricultural Science 1925; 15:529535.
8. Bishop AW. The principle of effective stress. Tek Ukeblas 1959; 39:859863.
9. Jennings JEB, Burland JB. Limitations to the use of effective stresses in partly saturated soils. Geotechnique
1962; 12(2):125144.
10. Bishop AW, Blight GE. Some aspects of effective stress in saturated and partly saturated soils. Geotechnique
1963; 13(3):177197.
11. Schreer BA. The nite element method in soil consolidation (with applications to surface subsidence). Ph.D.
Thesis, University College of Swansea, Swansea, 1984.
12. Lewis RW, Schreer BA. The Finite Element Method in the Static and Dynamics Deformation and Consolidation
of Porous Media (2nd edn). Wiley & Blackwell, 1998.
13. Houlsby GT. The work input to an unsaturated granular material. Geotechnique 1997; 47(1):193196.
14. Zienkiewicz OC, Chan AHC, Pastor M, Paul DK, Shiomi T. Static and dynamic behavior of soilsa rational
approach to quantitative solutions. 1. Fully saturated problems. Proceedings of the Royal Society of London
Series aMathematical Physical and Engineering Sciences 1990; 429:285309.
15. Zienkiewicz OC, Chan AHC, Pastor M, Schreer BA, Shiomi T. Computational Geomechanics with Special
Reference to Earthquake Engineering. Wiley: Chichester, 1999.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
486 M. PASTOR ET AL.
16. Zienkiewicz OC, Pastor M, Chan AHC, Xie YM. Computational approaches to the dynamics and statics of
saturated and unsaturated soils. In Advanced Geotecnical Analysis, Banerjee PK, Buttereld R (eds). Elsevier
Applied Science: London, 1991; 146.
17. Chan AHC. A unied nite element solution to static and dynamic geomechanics problems. Ph.D. Thesis,
University College of Swansea, Wales, Swansea, 1988.
18. Biot MA. Theory of propagation of elastic waves in a uid-saturated porous solid. 1. Low-frequency range.
Journal of the Acoustical Society of America 1956; 28(2):168178.
19. Craig RF. Soil Mechanics (5th edn). Chapman & Hall: London, 1992.
20. Darve F, Laouafa F. Instabilities in granular materials and application to landslides. Mechanics of Cohesive-
Frictional Materials 2000; 5(8):627652.
21. Khoa HDV, Georgopoulos IO, Darve F, Laouafa F. Diffuse failure in geomaterials: experiments and modelling.
Computers and Geotechnics 2006; 33(1):114.
22. Laouafa F, Darve F. Modelling of slope failure by a material instability mechanism. Computers and Geotechnics
2002; 29(4):301325.
23. Nova R. Controllability of the incremental response of soil specimens subjected to arbitrary loading programmes.
Journal of the Mechanical Behavior of Materials 1994; 5(2):193201.
24. Hutter K, Koch T. Motion of a antigranulocytes avalanche in an exponentially curved chuteexperiments and
theoretical predictions. Philosophical Transactions of the Royal Society of London Series AMathematical
Physical and Engineering Sciences 1991; 334(1633):93138.
25. Savage SB, Hutter K. The dynamics of avalanches of antigranulocytes materials from initiation to runout. 1.
Analysis. Acta Mechanica 1991; 86(14):201223.
26. Laigle D, Coussot P. Numerical modeling of mudows. Journal of Hydraulic Engineering (ASCE) 1997;
123(7):617623.
27. McDougall S, Hungr O. A model for the analysis of rapid landslide motion across three-dimensional terrain.
Canadian Geotechnical Journal 2004; 41(6):10841097.
28. Pastor M, Quecedo M, Merodo JAF, Herreros MI, Gonzalez E, Mira P. Modelling tailings dams and mine waste
dumps failures. Geotechnique 2002; 52(8):579591.
29. Quecedo M, Pastor M, Herreros MI, Merodo JAF. Numerical modelling of the propagation of fast landslides using
the nite element method. International Journal for Numerical Methods in Engineering 2004; 59(6):755794.
30. Pastor M, Haddad B, Sorbino G, Cuomo S, Drempetic V. A depth-integrated, coupled SPH model for ow-like
landslides and related phenomena. International Journal for Numerical and Analytical Methods in Geomechanics
2009; 33(2):143172.
31. Hutchinson JN. A sliding consolidation model for ow slides. Canadian Geotechnical Journal 1986; 23(2):
115126.
32. Iverson RM, Denlinger RP. Flow of variably uidized granular masses across three-dimensional terrain 1.
Coulomb mixture theory. Journal of Geophysical ResearchSolid Earth 2001; 106(B1):537552.
33. Tresca H. Sur Icoulement des corps solides soumis des fortes pressions. Comptes Rendus Academie des
Sciences Paris 1864; 59:754.
34. Levy M. Extrait du mmoire sur les equations gnrales des mouvements intrieures des corps solides ductiles
au del des limites o llasticit pourrait les ramener leur premir tat. Journal de Mathmatiques Pures et
Appliques 1871; 16:369372.
35. Huber MT. Die spezische formnderungsarbeit als mader anstrengung eines materials. Czasopismo Techniczne
1904; 20:8183.
36. Von Mises R. Mechanik der festen korper im plastisch deformablen zustand. Gttinger Nachrichten Mathematics
Physics 1913; 1:582592.
37. Prandtl W. Spannungsverteilung in plastischen krpern. Proceedings of the First International Congress on
Applied Mechanics, Delft, 1924; 4354.
38. Reuss A. Consideration of the elastic form variation in the plasticity theory. Zeitschrift fr Angewandte
Mathematik Und Mechanik 1930; 10:266274.
39. Melan E. Zur Plastizitt des rumlichen kontinuums. Ingenieur-Archiv 1938; 8:116126.
40. Drucker DC. Relation of experiments to mathematical theories of plasticity. Journal of Applied Mechanics
Transactions of the (ASME) 1949; 16(4):349357.
41. Drucker DC, Prager W. Soil mechanics and plastic analysis or limit design. Quarterly of Applied Mathematics
1952; 10(2):157165.
42. Zienkiewicz OC, Humpheson C, Lewis RW. Associated and non-associated visco-plasticity and plasticity in soil
mechanics. Geotechnique 1975; 25(4):671689.
43. Roscoe KH, Schoeld AN, Wroth CP. On the yielding soils. Geotechnique 1958; 8:2253.
44. Roscoe KH, Schoeld AN, Wroth CP. Yielding of clays in states wetter than critical. Geotechnique 1963;
13:211240.
45. Schoeld AN, Wroth CP. Critical State Soil Mechanics. McGraw-Hill: New York, 1968.
46. Nova R. Hardening of Soils. Archives of Mechanics 1977; 29(3):445458.
47. Wilde P. 2 Invariants-dependent models of granular media. Archives of Mechanics 1977; 29(6):799809.
48. Kersten MS, Miles S. Repeated load tests on highway subgrade soil and bases. Report, 1943.
49. Buchanan SJ, Khuri FI. Elastic and plastic properties of soils and their inuence on the continuous support of
rigid pavements. Report, 1954.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 487
50. Seed HB, Chan CK, Lee CE. Resilience characteristics of subgrade soils and their relation to fatigue failures in
asphalt pavements. Proceedings of the International Conference on the Structural Design of Asphalt Pavements,
1962; 611636.
51. Seed HB, Fead JWM. Apparatus for repeated loading tests on soils. American Society for Testing Materials,
STP 1959; 254:7887.
52. Seed HB, McNeil RL. Soil deformation in normal compression and repeated loading tests. Highway Research
Board Bulletin 1956; 141:4453.
53. Seed HB, Chan CK. Clay strength under earthquake loading conditions. Journal of the Soil Mechanics and
Foundations Division 1966; 92:5378.
54. Seed HB, Lee KL. Liquefaction of saturated sands during cyclic loading. Journal of the Soil Mechanics and
Foundations Division 1966; 92(SM6):105134.
55. Seed HB, Peacock WH. Test procedures for measuring soil liquefaction potential. Journal of the Soil Mechanics
and Foundations Division 1971; 97(SM8):10991119.
56. Castro G. Liquefaction of sands. Ph.D. Thesis, Harvard University, 1969.
57. Ishihara K, Tatsuoka F, Yasuda S. Undrained deformation and liquefaction of sand under cyclic stress. Soils
and Foundations 1975; 15:2944.
58. Tatsuoka F, Ishihara K. Drained deformation of consolidation beneath embankments. Soils and Foundations
1974; 14:5165.
59. Andersen KH. Behavior of clay subjected to undrained cyclic loading. Proceedings of the International Conference
Boss76, Norwegian Institute of Technology, 1976.
60. Marr WA, Christian JT. Permanent displacements due to cyclic wave loading. Journal of the Geotechnical
Engineering Division (ASCE) 1981; 107(8):11291149.
61. Zienkiewicz OC, Mroz Z. Generalized plasticity formulation and application to geomechanics. In Mechanics of
Engineering Materials, Desai CS, Gallaher RH (eds). Wiley: New York, 1984; 655679.
62. Pastor M, Zienkiewicz OC, Chan AHC. Generalized plasticity and the modeling of soil behavior. International
Journal for Numerical and Analytical Methods in Geomechanics 1990; 14(3):151190.
63. Tamagnini R, Pastor M. A thermodynamically based model for unsaturated soils: a new framework for generalized
plasticity. Proceedings of the Second International Workshop on Unsaturated Soil, Naples, 2004; 114.
64. Fernndez Merodo JA, Tamagnini R, Pastor M, Mira P. Modelling damage with generalized plasticity. Rivista
Italiana di Geotecnica 2005; 4:3242.
65. Manzanal D. Constitutive model based on generalized plasticity incorporating state parameter for saturated and
unsaturated sand. Ph.D. Thesis, Universidad Politecnica de Madrid, Madrid, 2008.
66. Manzanal D, Pastor M, Fernndez Merodo JA. Generalized plasticity state parameter based model for saturated
and unsaturated soils. Part II: unsaturated soil modelling. International Journal for Numerical and Analytical
Methods in Geomechanics 2010; DOI: 10.1002/nag.983.
67. Uriel AO, Merino M. Harmonic response of sands in shear. Proceedings of the Third International Conference
on Numerical Methods in Geomechnics, Aachen, 1979.
68. Been K, Jefferies M. A state parameter for sand. Geotechnique 1985; 35(2):99112.
69. Li XL. Modeling of dilative shear failure. Journal of Geotechnical and Geoenvironmental Engineering 1977;
123:609616.
70. Li XS, Dafalias YF. Dilatancy for cohesionless soils. Geotechnique 2000; 50(4):449460.
71. Manzanal D, Fern andez Merodo JA, Pastor M. Generalized plasticity state parameter based model for saturated
and unsaturated soils. Part I: saturated state. International Journal for Numerical and Analytical Methods in
Geomechanics 2010; DOI: 10.1002/nag.961.
72. Bolzon G, Schreer BA, Zienkiewicz OC. Elastoplastic soil constitutive laws generalized to partially saturated
states. Geotechnique 1996; 46(2):279289.
73. Santagiuliana R, Schreer BA. Enhancing the BolzonSchreerZienkiewicz constitutive model for partially
saturated soil. Transport in Porous Media 2006; 65(1):130.
74. Gallipoli D, Gens A, Sharma R, Vaunat J. An elastoplastic model for unsaturated soil incorporating the effects
of suction and degree of saturation on mechanical behaviour. Geotechnique 2003; 53(1):123135.
75. Fisher RA. On the capillary forces in an ideal soil; correction of formulae given by W. B. Haines. Journal of
Agricultural Science 1926; 16:492505.
76. Sivakumar V. A critical state framework for unsaturated soil. Ph.D. Thesis, University of Shefeld, 1993.
77. Ma atouk A, Leroueil S, LaRochelle P. Yielding and critical state of a collapsible unsaturated silty soil.
Geotechnique 1995; 45(3):465477.
78. Russell AR, Khalili N. A unied bounding surface plasticity model for unsaturated soils. International Journal
for Numerical and Analytical Methods in Geomechanics 2006; 30(3):181212.
79. Hrennikoff A. Solution of problems of elasticity by the Grame-work method. Journal of Applied Mechanics
Transactions of the (ASME) 1941; 8:169175.
80. McHenry D. A lattice analogy for the solution of plane stress problems. Journal of the Institution of Civil
Engineers 1943; 21:5982.
81. Courant RL. Variational methods for the solution of problems of equilibrium and vibration. Bulletin of the
American Mathematical Society 1943; 49:123.
82. Turner MJ, Clough RW, Martin HC, Topp LJ. Stiffness and deection analysis of complex structures. Journal
of the Aeronautical Sciences 1956; 23(9):805.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
488 M. PASTOR ET AL.
83. Argyris JH. Energy theorems and structural analysis. Aircraft Engineering 1954; 26:347356 (October) and
383384, 394 (November).
84. Argyris JH. Energy theorems and structural analysis. Aircraft engineering 1955; 27:4258 (February), 8094
(March), 125134 (April) and 145158 (May).
85. Clough RW. The nite element method in plane stress analysis. Proceedings of the Second ASCE Conference
on Electronic Computation, Pittsburgh, 1960.
86. Clough RW. The nite element method in structural mechanics. In Stress Analysis, Zienkiewicz OC, Hollister G
(eds). Wiley: New York, 1965.
87. Zienkiewicz OCYK. Cheung nite elements in the solution of eld problems. The Engineer 1965; 220:507510.
88. Sandhu RS, Wilson EL. Finite-element analysis of seepage in elastic media. Journal of Engineering Mechanics
Division 1969; 95(EMR3):641652.
89. Yamagata Y, Nagaoka H. Finite element method applied to Biots consolidation theory. Soils and fundations
1971; 11(1):2946.
90. Yokoo Y, Yamagata K, Nagaoka H. Finite element analysis of consolidation following undrained deformation.
Soils and Foundations 1971; 11:3758.
91. Smith IM, Hobbs R. Finite-element analysis of centrifuged and built-up slopes. Geotechnique 1974; 24(4):
531559.
92. Valliappan S, Lee IK, Boonlualohr P. Finite element analysis of consolidation problem. Proceedings of the
International Symposium on Finite Element Methods in Flow Problems, Swansea, 1974; 741755.
93. Smith IM, Hobbs R. Biot analysis of consolidation beneath embankments. Geotechnique 1976; 26(1):149171.
94. Lewis RW, Roberts GK, Zienkiewicz OC. A non linear ow and deformation analysis of consolidation problems.
Numerical Methods in Geomechanics 1976; 2:11061118.
95. Zienkiewicz OC, Taylor RL. A Non-linear Flow and Deformation Analysis of Consolidation Problems. McGraw-
Hill Book Company: London, 1989.
96. Zienkiewicz OC, Taylor RL. The Finite Element MethodVolume 2: Solid and Fluid Mechanics, Dynamics and
Non-linearity. McGraw-Hill Book Company: London, 1991.
97. Katona MG, Zienkiewicz OC. A unied set of single step algorithms. 3. The beta-M method, a generalization
of the Newmark scheme. International Journal for Numerical Methods in Engineering 1985; 21(7):13451359.
98. Schreer BA, Lewis RW, Norris VA. A case study of surface subsidence of the Polesine area. International
Journal for Numerical and Analytical Methods in Geomechanics 1977; 1:377386.
99. Morgan K, Lewis RW, White IR. The mechanisms of ground surface subsidence above compacting multiphase
reservoirs and their analysis by the nite-element method. Applied Mathematical Modelling 1980; 4(3):217224.
100. Andersen S, Andersen L. Modelling of landslides with the material-point method. Computational Geosciences
2010; 14(1):137147.
101. Coetzee CJ, Vermeer PA, Basson AH. The modelling of anchors using the material point method. International
Journal for Numerical and Analytical Methods in Geomechanics 2005; 29(9):879895.
102. Idelsohn SR, Onate E, Del Pin F. A Lagrangian meshless nite element method applied to uidstructure
interaction problems. Computers and Structures 2003; 81(811):655671.
103. Onate E, Idelsohn SR, Celigueta MA, Rossi R. Advances in the particle nite element method for the analysis
of uidmultibody interaction and bed erosion in free surface ows. Computer Methods in Applied Mechanics
and Engineering 2008; 197(1920):17771800.
104. Zhang HW, Wang KP, Chen Z. Material point method for dynamic analysis of saturated porous media under
external contact/impact of solid bodies. Computer Methods in Applied Mechanics and Engineering 2009;
198(1720):14561472.
105. Bui HH, Fukagawa R, Sako K, Ohno S. Lagrangian meshfree particles method (SPH) for large deformation and
failure ows of geomaterial using elasticplastic soil constitutive model. International Journal for Numerical
and Analytical Methods in Geomechanics 2008; 32(12):15371570.
106. McDougall S. A new continuum dynamic model for the analysis of extremely rapid landslide motion across
complex 3D terra. Ph.D. Thesis, University of British Columbia, 2006.
107. Rodriguez-Paz M, Bonet J. A corrected smooth particle hydrodynamics formulation of the shallow-water
equations. Computers and Structures 2005; 83(1718):13961410.
108. Lucy LB. Numerical approach to testing of ssion hypothesis. Astronomical Journal 1977; 82(12):10131024.
109. Gingold RA, Monaghan JJ. Smoothed particle hydrodynamicstheory and application to non-spherical stars.
Monthly Notices of the Royal Astronomical Society 1977; 181(2):375389.
110. Gingold RA, Monaghan JJ. Kernel estimates as a basis for general particle methods in hydrodynamics. Journal
of Computational Physics 1982; 46(3):429453.
111. Monaghan JJ, Gingold RA. Shock simulation by the particle method SPH. Journal of Computational Physics
1983; 52(2):374389.
112. Monaghan JJ, Cas RAF, Kos AM, Hallworth M. Gravity currents descending a ramp in a stratied tank. Journal
of Fluid Mechanics 1999; 379:3970.
113. Bonet J, Kulasegaram S. Correction and stabilization of smooth particle hydrodynamics methods with applications
in metal forming simulations. International Journal for Numerical Methods in Engineering 2000; 47(6):
11891214.
114. Monaghan JJ, Kos A, Issa N. Fluid motion generated by impact. Journal of Waterway Port Coastal and Ocean
Engineering (ASCE) 2003; 129(6):250259.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme
COMPUTATIONAL GEOMECHANICS 489
115. Dunn SL, Vun PL, Chan AHC, Damgaard JS. Numerical modeling of wave-induced liquefaction around pipelines.
Journal of Waterway Port Coastal and Ocean Engineering (ASCE) 2006; 132(4):276288.
116. Teh TC. Stability of marine pipelines on unstable and liqueed seabed. Ph.D. Thesis, University of Cambridge,
2003.
117. Pastor M, Zienkiewicz OC. A generalized plasticity, hierarchical model for sand under monotonic and cyclic
loading. Proceedings of the Second International Symposium on Numerical Models in Geomechanics. M Jackson
and Son Publisher: Ghent, 1986; 131150.
118. Gawin D, Sanavia L. A unied approach to numerical modeling of fully and partially saturated porous
materials by considering air dissolved in water. CMESComputer Modeling in Engineering and Sciences 2009;
53(3):255302.
119. Konagai K, Johansson J, Mayorca P, Uzuoka R, Yamamoto T, Miyajima M, Pulido N, Sassa H, Fujuoka H,
Duran F. Las Colinas landslide, Santa Tecla: rapid and long-traveling soil ow caused by the January 13th,
2001, El Salvador earthquake. In Natural Hazards in El Salvador, Rose WI, Bommer JJ, Lopez DL, Carr Major
MJJJ (eds). The Geological Society of America, Special Paper No. 375, 2004; 3953.
120. Evans SG, Bent AL. The Las Colinas landslide, Santa Tecla: a highly destructive owslide triggered by the
January 13, 2001, El Salvador Earthquake. In Natural Hazards in El Salvador, Rose WI, Bommer JJ, Lopez
DL, Carr Major MJ (eds). The Geological Society of America, Special Paper No. 375, 2004; 2537.
121. Dawson RF, Morgenstern NR, Stokes AW. Liquefaction owslides in rocky mountain coal mine waste dumps.
Canadian Geotechnical Journal 1998; 35(2):328343.
Copyright 2011 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2011; 87:457489
DOI: 10.1002/nme

S-ar putea să vă placă și