You are on page 1of 13

ARTICLE IN PRESS

Computers and Geotechnics xxx (2006) xxx–xxx


www.elsevier.com/locate/compgeo

A numerical model to investigate the effects of propagating


liquefied soils on structures
Sami Montassar, Patrick de Buhan *

Laboratoire des Matériaux et Structures du Génie Civil (LCPC, ENPC, CNRS UMR 113), Ecole Nationale des Ponts et Chaussees, 6 et 8 av.
Blaise Pascal, Cite Descartes, Champs/Marne, 77455 Marne-la-Vallée Cedex 02, France

Received 23 May 2005; received in revised form 3 February 2006; accepted 8 February 2006

Abstract

This contribution is concerned with the development of a computational method for simulating the propagation of a liquefied soil,
modelled as a Bingham material, while evaluating its effects on adjacent structures. The method is based upon the existence of a min-
imum principle for the velocity field governing the instantaneous evolution of the liquefied soil mass, which is solved by resorting to
a specific numerical algorithm implemented within the context of a finite element formulation. Making use of simplifying assumptions,
the calculation of the drag force induced by the moving soil mass on structures schematized as rigid obstacles, can be performed inde-
pendently, exploiting the same numerical tool. The whole procedure is illustrated on the typical example of a liquefied soil embankment
interacting with a single row of regularly spaced foundation piles placed across the propagating soil mass.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Liquefied soil; Residual shear strength; Viscosity; Bingham model; Drag force; Minimum principle; Finite element method; Decomposition–
co-ordination method

1. Introduction extensive measurements performed on the site of the 1964


Niigita earthquake and 1983 Nihonkai-Chubu earthquake:
Initial flow failure and subsequent propagation of lique- pffiffiffiffip
3
ffiffiffi
D ¼ 0:75 H h ð1Þ
fied soil masses triggered by the well-known seismically
induced liquefaction phenomenon of saturated granular where D (m) is the maximum horizontal displacement, H
soils, may have very serious consequences on the stability (m) the depth of the liquefied soil layer and h (%) the great-
of civil engineering structures, such as bridge piers, founda- er of the slope of ground surface and lower boundary face
tion piles or lifelines facilities. It is therefore essential to of liquefied soil. Such a formula also incorporates the
devise rational design methods capable not only to predict observations made on the San Fernando earthquake [9].
how such liquefied masses actually propagate, but also to Likewise, Bartlett and Youd [4] have derived the same kind
quantitatively assess the potential damage they may cause relationship based upon data gathered on six earthquakes
to structures. in the United States, including the one having occurred
A relatively abundant literature has been devoted to deal- in Alaska in 1964. Even though they remain empirical, that
ing with the first aspect of this question. In an attempt to esti- is based upon the fitting of observed field data, such rela-
mate the maximum permanent ground displacement due to tionships provide a first attempt to devise predictive meth-
an earthquake, the following empirical formula has been ods for estimating the amount of permanent ground
for instance proposed by Hamada et al. [17] from exploiting displacement induced by seismic liquefaction.
In order to avoid the use of complex finite element non
*
Corresponding author. linear simulations, simplified analytical models and related
E-mail address: debuhan@lmsgc.enpc.fr (P. de Buhan). calculation methods incorporating the rheological charac-

0266-352X/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2006.02.003
ARTICLE IN PRESS

2 S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx

teristics of the liquefied soil, have been developed by Towh- equation of a ‘‘yield stress’’ fluid or Bingham material,
ata et al. [38–40], in the case of a liquefied soil layer topped adopted by Uzuoka et al. [42] for simulating the post lique-
by a non liquefiable soil, in which lateral flow is predomi- faction behaviour of liquefied ground, or for dealing with
nant. The method, which is based on the application of closely related problems, such as flowslides [29], slurry
the minimum principle for the potential energy, is focused flows [6], or mudflows [7,18].
on the evaluation of the maximum displacements experi- Several simulation techniques have been developed
enced by the liquefied ground under the action of gravity. using such a constitutive equation, namely the so-called
It relies upon the following assumptions, some of them ‘‘shallow water approximation’’ [22] or the ‘‘depth inte-
being supported by experimental evidence. grated method’’ [6,29], based upon the simplifying assump-
tion that the flow is predominantly parallel to the slope,
 The lateral displacement is supposed to vary as a sinu- making it possible to account for very large displacements
soidal function of depth, as suggested for instance by of the laterally spreading soil mass, inducing considerable
the observations of centrifuge tests, leading to a much changes of geometry. In the more general situation where
easier numerical treatment of the problem than in the such a simplification cannot be made, specific simulation
case of finite element simulations. This makes it possi- techniques should be developed such as fluid dynamics
ble to generalize the method to three dimensional situ- based numerical methods [42], finite element [37] or cubic
ations, as it has been done by Orense and Towhata interpolated pseudoparticle (CIP) method [16].
[28]. It should be kept in mind however, that the adoption of
 The liquefied soil behaves as a Bingham or Newtonian a viscoplastic Bingham model, expressed by Eq. (3), is only
incompressible fluid, which may exhibit viscous as well justified in the situation when the consolidation time which
as residual shear strength properties. A constitutive rela- governs the pore pressure dissipation, may be considered
tionship of the following form is notably proposed by large enough in comparison with the time of propagation.
the authors [38,39] Indeed, in the case when these two times are of the same
ou order, the excess pore pressure which is at the origin of
s¼G þ sr ð2Þ the liquefaction phenomenon, may significantly dissipate
oz
during the propagation, thus allowing the granular soil to
where u is the lateral permanent displacement, function recover its frictional properties, bringing the moving soil
of the depth z, s (respectively sr) is the shear stress mass to stabilize again. The fundamental assumption upon
(respectively residual shear strength) and G the liquefied which all the previously described approaches, as well as
soil shear modulus. the numerical scheme developed in this paper, are based,
is that the rheological properties of the liquefied soil, which
While the first assumption seems quite legitimate, the may be conveniently captured by means of a Bingham
adoption of a constitutive law such as (2) for modelling the model, remain constant throughout the whole propagation
constitutive behaviour of liquefied soil may appear some- process.
what questionable (apart from the particular situation where Analysing the potentially damaging effects of a propa-
G = sr = 0), unless the displacement u is replaced by the gating liquefied soil on the neighbouring structures, requires
velocity (and the corresponding shear strain by the shear being able to evaluate the corresponding additional loads
strain rate), whereas G would denote the viscosity coefficient applied to these structures by the moving soil mass.
instead of the elastic shear modulus (see Eq. (3) below). Although being essential for engineering design purposes,
It is widely acknowledged today that, in accordance this second aspect has almost exclusively been treated from
with the terminology used, a ‘‘liquefied’’ soil behaves as an experimental point of view. One may quote for instance
an incompressible fluid, displaying both viscous and resid- the experimental work by Towhata and Hussaini [38], who
ual strength properties. Ample experimental evidence of performed tests aimed at evaluating the lateral loading
the viscosity of a liquefied soil, evaluated by different mea- exerted by a submarine mudflow on offshore structures, as
surement techniques (viscometer, lateral spreading, load also the contributions of Yasuda et al. [44] and Towhata
acting on a sphere, etc.) has been provided, the synthesis et al. [41] concerning shaking table tests carried out on bur-
of which may be found in Uzuoka et al. [42], Towhata ied pipelines. Several experimental studies have also been
et al. [38] or Hadush et al. [15,16]. The same kind of exper- undertaken on reduced scale centrifuged models, in order
iments makes it possible to evaluate the undrained residual to investigate the lateral resistance of foundation piles sub-
shear strength of a liquefied soil [35,21,11]. This clearly ject to liquefaction [5,1,19,43]. To the authors’ knowledge,
indicates that the rheology of a liquefied soil under pure apart from the works of Dobry et al. [8] and Ishihara and
shear conditions, may be appropriately described by the Cubrinovski [20], who have proposed simplified design
following constitutive equation: analyses of piles subject to liquefaction induced lateral load-
s ¼ l_c þ k ð3Þ ing, no other significant contributions can be quoted as
regards the numerical modelling of this kind of problem.
where k is the residual shear strength, c_ the shear strain rate The present contribution is concerned with the presenta-
and l the viscosity coefficient. This is the characteristic tion of a fully integrated numerical tool allowing to simul-
ARTICLE IN PRESS

S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx 3

taneously analyse the post liquefaction behaviour of a (a) The liquefied soil, whose strength has been consider-
water-saturated soil mass and its consequences in terms ably reduced as a consequence of the liquefaction
of additional loadings exerted on the surrounding struc- process, is first set into motion under the action of
tures. The whole procedure, which relies on the application gravity, undergoing very large displacements as it
of a minimum principle for the velocities, analogous to that propagates downhill. It is therefore essential to devise
employed by Towhata et al. [38–40], is developed in the appropriate methods for tracking the geometry
context of a liquefied soil modelled as a viscoplastic Bing- changes of the moving soil mass as a function of time:
ham material. It is worth emphasizing that the ability of this is the evolution problem.
such a numerical method to produce reliable predictions, (b) Assuming that the evolution problem is being solved,
is subject to the condition that the extent and depth of which means that the successive configurations of the
the liquefied zone, along with its viscous and residual liquefied soil mass are known at any time along with
strength characteristics, are being known quantities, incor- the attached velocity field, it is essential to evaluate
porated as input data in the analysis. the efforts (drag forces) exerted by the flowing soil
mass on neighbouring civil engineering structures,
2. Problem statement such as bridge piers or deep foundations piles.

Analysing the post-liquefaction behaviour of a saturated In the general case, the comprehensive treatment to this
soil mass entails two different aspects, which are sketched twofold problem requires a fully coupled complex analysis,
in Fig. 1. which would lead us far beyond the scope of engineering
design methods. The present contribution is devoted to
the development of a simplified evaluation procedure
aimed at dealing with the particular configuration of a sin-
gle row of vertical cylindrical obstacles placed across the
flow of a liquefied soil, as shown in Fig. 2.
The following simplifying assumption are made
throughout the paper.

 The obstacles are fixed, vertical and remain undeformed


under the action of the liquefied soil. This hypothesis
should be reconsidered in the case of obstacles such as
for example flexible foundation piles, the deflection of
which would of the same order of magnitude as the liq-
uefied soil displacements.
 The presence of such obstacles does not significantly
Fig. 1. (a) Evolution problem of a liquefied soil mass; and (b) evaluation affect the propagation of the liquefied soil. This means
of its action on structures. from a practical point of view that the spacing 2H

Fig. 2. Different views of a row of fixed cylindrical obstacles placed across a propagating liquefied soil mass.
ARTICLE IN PRESS

4 S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx

between two adjacent obstacles may be considered as sig- uefied soil mass, subject to the sole action of gravity, is
nificantly larger than their diameter D = 2R. Such an characterised by the velocity field U(x,t) defined at any
hypothesis implies that the problem of predicting the point x on its current configuration Xt at time t P 0.
evolution of the liquefied ground can be solved indepen- As sketched in Fig. 3 the external boundary oXt of Xt is
dently from the calculation of the efforts on the obstacles. subdivided into two complementary surfaces oXrt and oXUt .
 Furthermore, let F(y) denote the density per unit vertical oXrt remains stress-free, that is:
length of the drag force horizontal component exerted 8t; 8x 2 @Xrt ; rðx; tÞ  nðx; tÞ ¼ 0 ð7Þ
by the liquefied soil on each individual obstacle. This
density may be evaluated from the solution to the two- where n(x,t) is the external unit normal to oXrt ,
while slip-
dimensional stationary flow problem sketched in page boundary conditions proposed by Fortin et al. [13],
Fig. 2c, where uniform velocity boundary conditions are imposed on Rt which denotes the interface between
are imposed at sufficient upstream and downstream dis- the liquefied soil and the rigid substratum, located within
tances from the obstacle. Its value, considered as an the liquefied soil mass along the boundary surface oXUt
input parameter for the stationary flow problem, is (Fig. 3).
equal to the horizontal component Ux(y) of the velocity The velocity V of any point belonging to the liquefied
of the liquefied ground at the point where the obstacle is soil mass at the interface Rt, which is equal to the velocity
located, which has been previously calculated from the jump across Rt following the unit normal m, since the veloc-
evolution problem. ity of any point of the rigid substratum is zero, must
remain tangential
Solving the evolution problem of the liquefied soil mass 8t; 8x 2 Rt ; V ðx; tÞ  mðx; tÞ ¼ 0 ð8Þ
as well as the stationary flow problem around the obstacle
in order to calculate the drag force, requires to specify the and the following slip boundary conditions are applied on
rheological behaviour of the liquefied soil. As previously Rt (Fig. 4):
discussed in the introduction, the Bingham model appears V
to provide a realistic description of such a post-liquefaction s ¼ l V þ k  if V 6¼ 0 ð9Þ
jV j
behaviour, both in terms of residual shear strength and vis-
cosity exhibited by such kind of material. Expressed in full V ¼ 0 if jsj 6 k  ð10Þ
tensorial form as a stress–strain rate relationship, such a where s = T  (T Æ m) m is the shear stress vector exerted by
constitutive behaviour may be written as (see [25], for more the liquefied soil on the substratum, k* is the interface shear
details): strength below which no slippage occurs, and l* is a viscos-
ity coefficient. The particular case of the perfect bonding
pffiffiffi d
r ¼ pI þ 2ld þ k 2 if d 6¼ 0 ð4Þ condition is recovered when taking k* ! 1. The similarity
kdk between the above boundary conditions and the constitu-
ksk tive Eqs. (4)–(6) of the Bingham model is to be noticed, be-
d¼0 if f ðrÞ ¼ pffiffiffi  k 6 0 ð5Þ
2
div U ¼ tr d ¼ 0 ð6Þ

In these equations, r is the Cauchy stress tensor, s its devia-


toric part, p =  1/3tr r is the mean pressure, which can be
interpreted as the Lagrange multiplier associated with the
incompressibility condition (6), where d is the strain rate ten-
sor, and f(r) is the von Mises function which involves the
pffiffiffiffiffiffiffiffi
norm of the stress deviator ksk ¼ s : s. It is worth noting
that the purely viscous Newtonian model is recovered from Fig. 3. Initial and current configurations of a liquefied soil mass subject to
the Bingham model when the shear strength k is reduced to gravitational forces.
zero, while the so-called ‘‘rigid-plastic’’ model is obtained
when the viscosity coefficient l tends to zero.

3. A velocity minimum principle for solving the evolution


problem

The liquefied soil mass is described within the frame-


work of three-dimensional continuum mechanics as occu-
pying a geometrical domain X0 in its initial configuration
at time t = 0, immediately after the liquefaction process Fig. 4. Imposition of a non-linear friction boundary condition allowing
has been triggered. The subsequent propagation of this liq- for possible slippage at the liquefied soil/substratum interface.
ARTICLE IN PRESS

S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx 5

cause it will be helpful in the numerical treatment of the rate field and velocity jump distribution across the soil/sub-
problem. stratum interface, as minimization variables denoted by d 0
Let us now introduce the sets S(t) of statically admissible and v 0 , respectively, linked to the velocity field by the fol-
stress fields r 0 , and K(t) of kinematically admissible incom- lowing constraints:
pressible velocity fields U 0 defined, respectively as: d0 ¼ dðU 0 Þ ¼ 1=2ðgrad U 0 þT grad U 0 Þ on Xt ð15Þ
(
div r0 þ qg ¼ 0 8x 2 Xt
r0 2 SðtÞ () ð11Þ and
r0  n ¼ 0 8x 2 @Xrt v0 ¼ V 0 ; V 0  m ¼ 0; on Rt ð16Þ
 0 0
0
div U 0 ¼ 0 on Xt Denoting by k and w the Lagrange multipliers associated
U 2 KðtÞ () ð12Þ with the above constraints (15) and (16), the minimisation
V 0  m ¼ 0 on Rt
problem defined by (10) and (11) leads to a generalised sad-
A velocity field U 2 K(t) is a solution for the evolution dle-point problem of the form:
problem if it is associated with a stress field r 2 S(t)
Lr ðU  ; p ; d ; k ; v ; w Þ
through the constitutive Eqs. (4), (5), (9) and (10). It has
been proved in [24,25] that such a solution U realises the ¼ min
0
max
0
min
0
max
0
min
0
max
0
Lr ðU 0 ; p0 ; d0 ; k0 ; v0 ; w0 Þ
U p d k v w
minimum of the following functional defined on K(t):
ð17Þ
F ðU Þ ¼ MinfF ðU 0 Þ; U 0 2 KðtÞg ð13Þ
where
with Z Z pffiffiffi
Z Z  Z Lr ðU 0 ; p0 ;d0 ;k0 ; v0 ; w0 Þ ¼
2
lkd0 k dXt þ k 2kd0 kdXt
2 l 2
F ðU 0 Þ ¼ lkd 0 k dXt þ jV 0 j dRt  qg  U 0 dXt Zt
X Xt
Z
Xt Rt 2 Xt
Z pffiffiffi Z  0
qg  U dXt p0 divU 0 dXt
þ k 2kd 0 kdXt þ k  jV 0 jdRt ð14Þ X
Z t
Xt
Z 
Xt Rt
0 0 0 l 02
þ k : ðd  d ÞdXt þ jv j dRt
The two first integrals appearing on the upper line of (14) Xt Rt 2
Z Z
are equal to half the viscous dissipation, while the third
integral corresponds to the work of gravity, which is the þ k  jv0 jdRt þ w0  ðV 0  v0 ÞdRt
R Rt
unique external force. The lower line of integrals is equal Z t
r1 2
to the plastic dissipation. This minimum principle is analo- þ ðdivU 0 Þ dXt
Xt 2
gous to the minimum of potential energy invoked by Z
Towhata et al. [38–40], which has been briefly described r2 0 2
þ kd  d0 k dXt
in Section 1. In the particular case when the residual shear Xt 2
Z
strength is equal to zero, so that the liquefied soil is mod- r3 0 2
þ jV  v0 j dRt ð18Þ
elled as a Newtonian fluid, F reduces to a quadratic func- R 2
tional, and the principle is analogous to the minimum of is the augmented Lagrangian functional defined for all
potential energy established in the context of linear elastic- r1 > 0, r2 > 0 and r3 > 0, whereas p 0 is the Lagrange multi-
ity (see for instance [33]), except that the velocity field is plier associated with the incompressibility condition. It is
substituted to the displacement field. to be noticed that, for any choice of parameters
In the general case, this functional is non-differentiable (r1, r2, r3), the solution ðU~ ; ~p; ~d; ~k; ~v; w~ Þ to the above sad-
due to the presence of the plastic dissipation terms. The res- dle-point problem (17) verifies in particular:
olution of the minimisation problem (13) then requires a ~ ¼U
U ð19Þ
special treatment. The decomposition–co-ordination
method by augmented Lagrangian, proposed by Fortin Such a solution can be calculated using the following
and Glowinski [12], has proved its efficiency for dealing generalised Uzawa’s algorithm proposed by Fortin and
with such a numerical problem (see [24,25] for a compari- Glowinski [12] and adapted to our problem [24,25]
son with other numerical methods).
 Step 1: Initialisation at iteration i = 0; p(0), k(0), w(0) and
4. Numerical treatment: the decomposition–coordination U(0) are arbitrarily specified.
method by augmented Lagrangian  Step 2: for any i P 1 :
– Calculate d(i) as a function of k(i1) and U(i1) using
The present section is devoted to a general outline of the the following relationship:
numerical method adopted for performing the above min- * pffiffiffi +
ðiÞ 1 k 2
imization procedure, without going into details of any d ¼ 1  ði1Þ bði1Þ with
mathematical and theoretical justification, which may be 2l þ r2 kb k
found for instance in [12] Glowinski and Le Tallec [14], bði1Þ ¼ kði1Þ þ r2 dðU ði1Þ Þ and hi ¼ supf; 0g
Fortin et al. [13] or more recently Roquet and Saramito
[31]. Its principle is based on the introduction of the strain ð20Þ
ARTICLE IN PRESS

6 S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx

– Calculate v(i) as a function of w(i1) and U(i1) using : terms have to be taken into account, notably in the initial
  phase of propagation, since such a principle still applies,
ðiÞ 1 k
v ¼  1  ði1Þ xði1Þ with provided that the acceleration field is being known. In such
l þ r3 jx j
a case the linear system (26) becomes:
xði1Þ ¼ wði1Þ þ r3 V ði1Þ ð21Þ
½DfU g ¼ fF g  fAg ð27Þ
– Update the Lagrange multipliers associated with con-
where {A} represents the acceleration terms, which may be
straints (15) and (16) :
approximated as follows at current time t:
kðiÞ ¼ kði1Þ þ r2 ðdðU ði1Þ Þ  dðiÞ Þ ð22Þ fU gðtÞ  fU gðt  DtÞ
ðiÞ ði1Þ ði1Þ ðiÞ
fAgðtÞ ¼ ½M ð28Þ
w ¼w þ r3 ðV v Þ ð23Þ Dt
– Find U(i) as a function of p(i1), d(i), k(i), v(i) and w(i) by where [M] is the matrix of masses, and {U}(t) (respectively
minimising the following functional, which is qua- {U}(t  Dt)) is the vector of nodal velocities at time t
dratic with respect to U 0 : (respectively at the previous time step t  Dt), so that the
linear system to be solved becomes:
GðU ðiÞ ; pði1Þ ; dðiÞ ;kðiÞ ; vðiÞ ; wðiÞ Þ
8R R ði1Þ 9 ½M ½M
> r1 0 2 0 > ½D þ fU gðtÞ ¼ fF g þ fU gðt  DtÞ ð29Þ
> Xt h2ðdivU Þ dXt  Xt p divU dX
>
< i
t >
>
= Dt Dt
R r2 0 ðiÞ ðiÞ 0
¼ Min ðdðU Þ  2d Þ þ k : dðU Þ dXt Finally, the simulation of the evolution problem is car-
U0 > >
Xt 2 >
>
: þ R
r3 ðV 0  2vðiÞ Þ þ wðiÞ  V 0 dR  R qg  U 0 dX
> >
; ried out in an incremental way, by updating the geometry
Rt 2 t Xt t
of the propagating soil mass from time t to time t + Dt,
ð24Þ
according to the following explicit relationship:
– Update the Lagrange multiplier associated with the fxgðt þ DtÞ ffi fxgðtÞ þ DtfU gðtÞ ð30Þ
incompressibility condition:
where {x} denotes the set of co-ordinates of all the nodes of
pðiÞ ¼ pði1Þ  r1 div U ðiÞ ð25Þ the finite element mesh, considered as material points.
The whole above described numerical procedure for
 Step 3: Application of convergence tests. simulating the lateral flow of liquefied soil has been favour-
ably compared [24] with the reference analytical solution
According to Fortin and Glowinski [12], the explicit for- obtained in the particular situation of a uniform layer of
mula (20) (respectively (21)) represents the solution to the liquefied soil moving on an inclined substratum [24,26].
minimisation of Lr with respect to d (respectively v), with
U, p, k, v and w (resp. U, p, d, k and w) being kept fixed. 5. Evaluation of the efforts exerted on structures
Owing to the decomposition procedure, the above described
algorithm transforms the initial non-differentiable problem The previously elaborated numerical method, based on
(13) into a family of classical problems (explicit computa- the minimum principle for the velocities, is now applied to
tions for (20) and (21) and standard minimisation of a qua- calculate the drag force exerted by the propagating liquefied
dratic functional for (24)), co-ordinated via the Lagrange soil on structures, in the particular situation shown in Fig. 2c,
multipliers, updated at each iteration through Eqs. (22), where it reduces to solving a quasistatic steady flow problem
(23) and (25). of a Bingham material around a circular obstacle in the
The minimisation procedure of the quadratic form (24) (x, z)-plane. On account of the different symmetries exhib-
is carried out numerically in the context of the finite ele- ited by such a problem, it may be easily shown that the flow
ment method. The evolution problem being treated as a domain to be considered in the analysis may be restricted to
plane-strain two-dimensional problem, the volume of liq- the region of the plane defined by the following conditions:
uefied soil is discretized into a finite element mesh made pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
of four-noded quadrilateral elements with linearly varying 0 6 x 6 L; 0 6 z 6 H ; r ¼ x2 þ z2 P R ð31Þ
velocities, so that performing the minimization (24) reduces where the horizontal extension L is significantly larger than
to solving a linear system of the general form: H. The corresponding boundary conditions are (Fig. 5):
½DfU g ¼ fF g ð26Þ
where {U} is the column matrix formed by all the nodal  Uniform velocity imposed on the right-hand side x = L:
velocities, and {F} represents the external forces (grav- Ux ¼ U; Uz ¼ 0 ð32Þ
ity).The choice of this kind of elements is primarily justified
 Left-hand side x = 0:
by the fact that it lends itself very easily to the simulation of
the very large deformations experienced by the liquefied T x ¼ 0; Uz ¼ 0 ð33Þ
soil during it propagation [24].  Upper and lower sides z = 0, H:
The minimum principle and related numerical algorithm
may be readily extended to the case when acceleration T x ¼ 0; Uz ¼ 0 ð34Þ
ARTICLE IN PRESS

S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx 7

100

80
present simulations

60

F (N ) Mitsoulis (2004)
40

20
Fig. 5. Boundary conditions for analysing the stationary flow problem
relative to the evaluation of the drag force.
0
 Contact surface with the obstacle r = R: 0 5 10 15 20 25 30

H /R
U ¼0 ð35Þ
Fig. 7. Evaluations of the non dimensional drag force as a function of the
Note that in conditions (33) and (34), Tx represents the
relative spacing between obstacles: case of Newtonian fluid.
horizontal component of the stress vector acting upon the
corresponding boundary. may be fitted by a curve, drawn in the same figure (solid
curve), defined by the following equation:
The numerical simulations, have been carried out for 4p
F ðN Þ ¼
L = 6H, varying the ratio H/R between 2 and 50. lnðH =RÞ þ a1 þ a2 ðR=H Þ þ a3 ðR=H Þ4 þ a4 ðR=H Þ6 þ a5 ðR=H Þ8
2

Fig. 6 shows a typical finite element mesh of the ð38Þ


flow domain (H/R = 3). The drag force per unit transver-
sal length along the y-direction may be calculated as where the coefficients ai, i = 1, . . ., 5, involved in the above
follows: formula, have been calculated using the least square
Z p=2 method:
F ¼4 ex  rðr ¼ R; hÞ  er R dh ð36Þ
0 a1 ¼ 0:661; a2 ¼ 1:985; a3 ¼ 25:332;
since r(R,h) Æer is the stress vector applied by the liquefied a4 ¼ 193:823; a5 ¼ 443:505. ð39Þ
soil in each point (R, h) of the obstacle.
Referring to the slightly different problem of the flow
around a cylinder kept between parallel plates, Faxén
5.1. Newtonian model of liquefied soil (k = 0) [10], and quite recently Mitsoulis [23], have produced the
same kind of evaluation reported in the same figure, where
It may be easily shown from dimensional analysis argu- the squares correspond to the numerical simulations per-
ments that, in the case of a purely viscous Newtonian fluid, formed by Mitsoulis, while the dashed curve is obtained
characterized by its sole viscosity coefficient l, the drag from fitting formula (38). As could be expected, the numer-
force density may be put in the following form: ical simulation of the latter problem leads to slightly over-
estimating the non dimensional drag force, both estimates
F ðN Þ ¼ lUF ðN Þ ½H =R ð37Þ converging to zero as the relative spacing tends to infinity.

where F ðN Þ is a dimensionless factor which is a function of


the sole relative spacing H/R. 5.2. Viscoplastic Bingham model
Fig. 7 displays the evaluation of this factor (triangles)
obtained from the finite element simulations. Following This kind of problem has already been investigated in
the suggestion made by Faxén [10] these numerical points the case of a non-viscous soil (‘‘rigid-plastic’’ model:

Fig. 6. Finite element mesh of the flow domain adopted for H/R = 3.
ARTICLE IN PRESS

8 S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx

l = 0) by several authors such as Randolph and Houlsby Table 1


[30], Murff et al. [27], or more recently Salençon [34] in Coefficients a and b involved in Eq. (42) evaluated for different relative
spacings of the obstacles
the context of the limit analysis or yield design theory. It
has been extended to the case of a Bingham material by H/R F ðNÞ a b
Adachi and Yoshioka [2] making use of variational 2/1 52.059 0.531 0.963
principles. 3/1 23.117 1.197 0.958
4/1 15.886 2.127 0.881
Let F(B) denote the drag force per unit length of the 10/1 7.58 7.345 0.801
obstacle exerted by the liquefied soil modelled as a visco- 15/1 6.318 9.835 0.855
plastic Bingham medium, characterized by its residual 20/1 5.701 15.021 0.796
shear strength k along with its viscosity coefficient l. 25/1 5.437 18.88 0.788
Resorting once again to dimensional analysis, such a drag 30/1 5.212 19.457 0.801
35/1 5.056 25.89 0.762
force density may be written as: 40/1 4.936 25.23 0.784
F ðBÞ ¼ lUF ðBÞ ½H =R; Bn  ð40Þ 50/1 4.774 29.486 0.764

where F ðBÞ is the non dimensional drag force coefficient All the results derived from these numerical simulations
depending not only of the relative spacing H/R, but also are summed up in Table 1 below. This table, combined
on the dimensionless Bingham number defined as: with formula (38) along with (39), makes it possible to
kR compute in a very straightforward manner, the drag force
Bn ¼ ð41Þ per unit length exerted by a propagating liquefied soil on
lU
a row of evenly spaced cylindrical obstacles.
which compares the contribution of the shear strength to
that of the viscosity of the Bingham material. The case of 6. An illustrative example
Newtonian fluid is obviously recovered when Bn ! 0. The
numerical simulations where carried out for different values The whole numerical procedure will now be illustrated
of the ratio H/R, varying the residual shear strength k, on the example sketched in Fig. 9. The liquefied soil mass
while keeping all the other parameters of the problem, under consideration occupies in its initial configuration at
namely R, l and U, constant and equal to unity. The re- time t = 0 a quadrilateral domain of height 1 m and hori-
sults of such simulations are shown in Fig. 8, in the form zontal extension equal to 3 m along its bottom, where it
of a series of curves giving the non dimensional drag force relies upon a horizontal substratum. Its left hand side also
coefficient F ðBÞ as a function of H/R and Bn. As suggested remains in contact with such a substratum, while no sur-
by Zisis and Mitsoulis [45] who carried out the same kind charge is applied to the upper and right hand surfaces.
of FEM simulations using a regularized Bingham model, The only driving force is the liquefied soil specific weight
the numerical points reported in this figure have been fitted equal to 18 kN/m2. A row of regularly spaced vertical
by solid curves obeying the following analytical expression: cylindrical obstacles is placed at a horizontal distance of
b
F ðBÞ ðH =R; Bn Þ ¼ F ðN Þ ðH =RÞ½1 þ aBn  ð42Þ x = 2 m from the lower left hand corner of the embank-
ment, taken as the origin of the axes. Their diameter is
where F ðBÞ
is given by Eq. (38), whereas coefficients a and b
are both functions of H/R only.

100000 F(B∗ )
2m 0.2m F (y)
10000

1000

y
100 H /R=2 2m

3
10 10 ρg = 18 kN/m3
50 F (y)
1m
1
Bn 10
0.001 00.1 0.1 1 10 100 1000 x

Fig. 8. Numerical evaluations of the non dimensional drag force 1


3m
coefficient as a function of the Bingham number for different values of
the relative spacing between obstacles. Fig. 9. Liquefied soil embankment in its initial configuration.
ARTICLE IN PRESS

S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx 9

equal to 0.2 m, while the interval between two neighbour- firmed by the observation of the instantaneous velocity
ing obstacles is taken equal to 2 m, so that H/R = 10. field represented by a distribution of arrows attached to
The different simulations are performed for a coefficient the nodal points of the mesh: the volume of liquefied soil
of viscosity equal to l = 1 kPa s, the undrained shear involved in the flow failure of the embankment, that is
strength being progressively increased from 0 to 4 kPa. experiencing large deformations, is becoming less extensive
Referring to the values reported in Table 1, it follows that as the liquefied soil strength increases, almost disappearing
the drag force density applied by the propagating liquefied for k = 4 kPa. Furthermore, it should be noted that, con-
soil characterized locally by its horizontal velocity Ux trary to what is observed, either in centrifuge experiments
(x = 2 m, y) at depth y is: or by means of the present numerical procedure [24], in
 0:801 the case of an infinite layer of liquefied soil flowing down
0:1k
F ðyÞ ¼ 7:58U x ð2; yÞ 1 þ 7:345 ð43Þ an inclined slope, the point of maximum velocity (and
U x ð2; yÞ hence displacement) is not located on the ground surface
where the value of F(y) is expressed in kN/m, while the un- but, as shown in Fig. 10 tends to approach the toe of the
drained shear strength k and the velocity Ux(y) are ex- embankment, as the residual shear strength increases. This
pressed in kPa and m/s, respectively. A relatively coarse is of course to be attributed to the particular configuration
and regular mesh comprising 10 · 30 quadrilateral ele- of the liquefied soil mass, where the vertical facing remains
ments is used throughout the different simulations. stress free.
This is corroborated by referring to a limit analysis (or
6.1. Influence of the undrained shear strength yield design) point of view. Indeed, it may be shown
[18,26], that viscoplastic flow failure of any liquefied soil
A first series of simulations is performed, assuming that structure will occur as soon as the latter is proved to be
the liquefied soil mass is perfectly bonded to the substratum unstable in the sense of limit analysis. In the present case,
along the lower and left-hand sides, which means that no a classical result states (see for instance [32]) that the stabil-
velocity discontinuity is allowed along these boundaries. ity of a vertical embankment of height h, as that considered
Adopting a time increment equal to Dt = 0.02 s, the evolu- in our simulations, is ensured as far as its stability factor
tion of the liquefied embankment is simulated up to t = 1 s. remains lower than a critical value, which is close to 3.8:
The corresponding deformed configurations of the qgh
embankment for increasing values of the undrained shear 6 K ffi 3:8 ð44Þ
k
strength are pictured in Fig. 10.
This figure clearly shows that the adoption of a Bingham thus indicating that the liquefied soil embankment will not
instead of a Newton model for describing the rheological propagate as soon as the undrained shear strength exceeds
behaviour of the liquefied soil, that is the introduction of the following value:
an undrained shear strength in addition to the viscosity,
results in considerably limiting the amplitude of the geom- qgh 18  1
kP ffi ffi 4:73 kPa ð45Þ
etry changes experienced by the liquefied soil. This is con- K 3:8

Fig. 10. Deformed configurations at time t = 1 s of the liquefied soil embankment for different values of the undrained shear strength, and corresponding
drag force profiles on the obstacles.
ARTICLE IN PRESS

10 S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx

The distributions of the drag force F(y) exerted by the l ¼ 1 kPa s; k ¼ 2 kPa; l ¼ 1 kPa s m;
propagating soil on the row of vertical piles are pictured k  ¼ 1 kPa ð46Þ
in the same figure. They are derived from the profiles of
horizontal velocities of the liquefied soil along the vertical It should be noted in particular, that the selected value of
line where the obstacles are located, previously deduced the soil-substratum strength parameter k* should be lower
from the evolution problem, by using formula (43). These than that of the liquefied soil strength k, otherwise the
profiles are decreasing from a maximum value reached at introduction of such a friction boundary condition will
the top of the embankment to zero at the bottom. More- have no influence of the liquefied soil propagation.
over, it can be observed that, every other parameter being Fig. 11 pictures a selected set of successive configurations
held constant, this profile increases in a first stage with the of the embankment at regular time intervals between t = 0 s
undrained shear strength, passing through a maximum for and t = 1.5 s. Comparing the configuration obtained for
k = 1–2 kPa, then decreases again down to almost vanish- t = 1 s with that of Fig. 10 for k = 2 kPa, gives clear evi-
ing for k = 4 kPa. This could be tentatively explained from dence of the influence of the friction boundary condition
examining formula (43), which shows that the drag force is both on the amplitude of the liquefied soil propagation
an increasing function of k, while at the same time the hor- and on the drag force profile. As regards the first aspect,
izontal velocity component involved in this equation is a Fig. 11 shows that the liquefied soil is no more adherent
decreasing function of k. to the substratum, the slippage condition being activated
along the bottom horizontal substratum. This leads to
6.2. Introduction of a friction boundary condition much larger values for the velocities, and hence for the
cumulated displacements, notably near the toe of the
The perfect bonding condition imposed at the interface embankment, while the drag force profile is significantly dif-
between the liquefied soil and the substratum in the previ- ferent from that observed in the case of the perfect bonding
ous set of simulations, may lead to somewhat unrealistic condition. Furthermore, the extension of the volume of
results as regards for instance the deformations experi- propagating liquefied soil is larger in this second simulation.
enced by the liquefied soil in the vicinity of the toe of the It is to be noted that selecting appropriate values of
embankment, where excessive mesh distortion appears as parameters l* and k* to be incorporated into the analysis,
time passes, leading to an ill-conditioned problem. This is remains a difficult task. This is due to the complete lack of
why such a condition should be relaxed by the introduction available experimental data in the field of geotechnical engi-
of a friction boundary condition, allowing for possible slip- neering, whereas the introduction of such a slip boundary
page beyond a shear stress value. Therefore the same kind condition is of primary importance for the simulation of
of simulation has been performed for the following set of non-Newtonian fluids in the material process industries
rheological parameters: [13]. It could provide however, a relevant description for a

Fig. 11. Successive configurations of the liquefied soil embankment and associated drag force profiles as a function of time, taking a friction boundary
condition into account.
ARTICLE IN PRESS

S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx 11

40

quasi - static evolution


30

R (k N ) 20 dynamic evolution

10

0 t(s)
0 0.5 1 1.5 2 2.5 3

Fig. 12. Evolution of the resultant drag force as a function of time: quasi-static vs. dynamic simulations.

thin layer of liquefied soil, modelled as a Bingham fluid, using the same numerical approach (decomposition–co-
across which the velocity increases from zero to a finite value. ordination method), making it thus possible for instance
Finally, in order to assess the role played by the acceler- to evaluate the drag force applied by the liquefied soil on
ation forces in the propagation of the liquefied soil mass a pile foundation modelled as a fixed obstacle. A fully inte-
and consequently on the evaluation of the drag force grated computational tool dealing with both aspects simul-
exerted on the structures, curves giving the resultant drag taneously has been set up in the simplified situation when
force defined as: the evolution problem can be solved independently from
Z y¼lðtÞ the evaluation of the drag forces applied to the structures.
RðtÞ ¼ F ðy; tÞ dy ð47Þ Its feasibility has been demonstrated on an illustrative
y¼0 example, and preliminary comments have been made on
the relative influence of some key parameters such as the
where l(t) denotes the length of the obstacle remaining in liquefied soil or soil-substratum shear strengths.
contact with the liquefied soil embankment in its current Looking forward to devising an engineering design-ori-
configuration (l(t = 0) = 1 m), have been drawn in Fig. 12 ented numerical code, further improvements may be con-
as a function of time up to 3 s. The curve exhibiting a max- sidered in the near future:
imum value of R(t @ 0.4s) @ 21.6 kN corresponds to the dy-
namic simulation, where inertia forces have been accounted  The generalization of the method to three dimensional
for, while the curve which monotonically decreases from an situations, which would probably require significant
initial maximum value of R(t = 0) @ 42 kN is associated progress as regards finite element discretization tech-
with a quasi-static simulation, where the acceleration forces niques, although decisive simplifications, such as the
have been neglected. The fact that, in the latter simulation, ‘‘shallow water approximation’’, could be envisaged in
the initial value of the drag force is not equal to zero, comes the frequently encountered situation when the liquefied
from the quasi-static approximation, since in such a case, soil flow is predominantly parallel to a plane.
the velocity field attached to the initial configuration (and  The use of remeshing techniques such as those devel-
hence the drag force distribution: see Fig. 11) is also non- oped in the quite different context of the simulation
zero. Fig. 12 shows that, even though both curves become metal-forming processes (see for instance [3]) where
almost coincident after an elapsed time of 1.5 s, they are large strains are also involved, which may induce exces-
quite significantly different in the first stage of propagation. sive mesh distortion and thus affect the numerical accu-
It turns out in particular that, owing to such acceleration racy of the computations.
forces, the maximum peak value of the resultant drag force  Extending the approach to the case when the evolution
is significantly reduced (almost cut in half in the present problem can no longer be disconnected from the evalu-
example). ation of the forces exerted on structures. Such a situa-
tion arises when the characteristic size of the latter is
7. Concluding remarks not sufficiently small with respect to the overall dimen-
sions of the propagating liquefied soil mass.
A general procedure based upon the numerical imple-
mentation of a variational principle for the velocities, com- It should however be kept in mind that the validity of the
bined with a time-step integration algorithm, has been whole procedure relies upon the assumption that both the
proposed in order to predict the propagation of a liquefied viscosity and shear strength parameters governing the lique-
soil mass, modelled as a Bingham material, even in the case fied soil, which are captured through the adoption of a
when large displacements are involved. The effect of such a Bingham model, remain constant and unaffected by the
moving liquefied soil on structures is then investigated propagation. A procedure could however be imagined in
ARTICLE IN PRESS

12 S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx

which for instance, due to the dissipation of the excess pore [16] Hadush S, Yashima A, Uzuoka R, Moriguchi S, Sawada K.
pressures as time passes, the soil may be recovering its Liquefaction induced lateral spread analysis using the CIP method.
Comput Geotech 2001;28:549–74.
strength, which could be modelled by a corresponding [17] Hamada M, Towhata I, Yasuda S, Isoyama R. Study on permanent
increase with time of parameter k and possibly l in the Bing- ground displacement induced by seismic liquefaction. Comput
ham model. Geotech 1987;4:197–220.
The precise evaluation of these parameters by appropriate [18] Hild P, Ionescu IR, Lachand-Robert T, Rosca I. The blocking of an
in situ or laboratory experimental tests remains a difficult inhomogeneous Bingham fluid. Applications to landslides. Math
Modell Numer Anal 2002;36(6):1013–26.
task for the geotechnical engineer. A possible line of [19] Imamura S, Hagiwara T, Tsukamoto Y, Ishihara K. Response of pile
research, which is currently in progress, consists in imple- groups against seismically induced lateral flow in centrifuge model
menting the numerical procedure developed in this paper tests. Soils Foundat 2004;44(3):39–55.
for simulating the response of liquefied sand in the free field, [20] Ishihara K, Cubrinovski M. Soil–pile interaction in liquefied deposits
and compare the velocity and displacement profiles so- undergoing lateral spreading. In: Maric B et al., editors. Geotechni-
cal hazards. Rotterdam: Balkema; 1998. p. 51–64.
obtained with experimental results drawn from centrifuge [21] Kawakami T, Suemasa N, Hamada M, Sato H, Takada T, Exper-
model tests on earthquake-induced lateral spreading using imental study on mechanical properties of liquefied sand. In:
a laminar box [36]. This would serve the double purpose of Proceedings of the 5th US–Japan workshop on earthquake resistant
validating the proposed numerical scheme and related com- design of lifeline facilities and countermeasures against soil liquefac-
puter code on the one hand, assessing realistic values to the tion, vol. 1, Technical Report, NCEER-94-0026, Salt Lake City,
USA; 1994. p. 99–285.
key constitutive parameters of liquefied soil (k and l) from [22] Liu KF, Mei CC. Slow spreading of a sheet of Bingham fluid on an
a back calculation procedure, on the other hand. inclined plane. J Fluid Mech 1989;207:505–29.
[23] Mitsoulis E. On creeping drag flow of a viscoplastic fluid past a
References circular cylinder: wall effects. Chem Eng Sci 2004;59:789–800.
[24] Montassar S, Contribution to the numerical simulation of liquefied
[1] Abdoun T, Dobry R, O’Rourke TD, Goh SH. Pile response to lateral ground flow and its effects on structures. PhD thesis (in French),
spreads: centrifuge modeling. J Geotech Geoenviron Eng ENPC, Paris; 2005.
2003;129(10):869–78. [25] Montassar S, de Buhan P. Minimum principle and related numerical
[2] Adachi K, Yoshioka N. On creeping flow of a visco-plastic fluid past scheme for simulating initial flow and subsequent propagation of
a circular cylinder. Chem Eng Sci 1973;28:215–26. liquefied ground. Int J Numer Anal Meth Geomech 2005;29:1065–86.
[3] Aymone JLF, Bittencourt E, Creus GJ. Simulation of 3D metal- [26] Montassar S, de Buhan P. Some general results on the stability and
forming using an arbitrary Lagrangian–Eulerian finite element flow failure of rigid viscoplastic structures. Mech Res Commun
method. J Mater Process Technol 2001;110:218–32. 2006;33:63–71.
[4] Bartlet SF, Youd TL, Empirical prediction of lateral displacement. [27] Murff JD, Wagner DA, Randolph MF. Pipe penetration in cohesive
In: Proceedings of the 4th US–Japan workshop on earthquake soil. Géotechnique 1989;39(2):213–29.
resistant design of lifeline facilities and countermeasures against soil [28] Orense RP, Towhatal I. Three dimensional analysis on lateral
liquefaction, vol. 1, Technical Report, NCEER-92-0019, Honolulu, displacement of liquefied subsoil. Soils Foundat 1998;38(4):1–15.
USA; 1992. p. 65–351. [29] Pastor M, Quecedo M, Fernandez Merodo JA, Herreros MI,
[5] Boulanger RW, Wilson WW, Kutter BL, Soil–pipe-superstructure Gonzalez E, Mira P. Modelling tailing dams and mine waste dumps
interaction in liquefiable sand, Transportation Research Record 1569, failures. Géotechnique 2002;52(8):579–91.
Transportation Research Board, Washington (DC); 1997. [30] Randolph MF, Houlsby GT. The limiting pressure on a circular pipe
[6] Chen H, Lee CF. Runout analysis of slurry flows with Bingham loaded laterally in cohesive soil. Géotechnique 1984;34(4):613–23.
model. J Geotech Geoenviron Eng 2002;128(12):1032–42. [31] Roquet N, Saramito P. An adaptive finite element method for
[7] Coussot PH. Steady, laminar flow of concentrated mud suspension in Bingham fluid flows around a cylinder. Comput Meth Appl Mech
open channel. J Hydraul Res 1994;32(4):535–59. Eng 2003;192:3317–41.
[8] Dobry R, Abdoun T, O’Rourke TD, Goh SH. Single piles in lateral [32] Salençon J. An introduction to the yield design theory and its
spreads: field bending moment evaluation. J Geotech Geoenviron Eng applications to soil mechanics. Eur J Mech, A/Solids 1990;9(5):
2003;129(10):879–89. 477–500.
[9] Fallgren RB, Smith JL , Ground displacement at San Fernando valley [33] Salençon J. Handbook of continuum mechanics. Berlin: Springer;
juvenile hall during San Fernando earthquake. In: San Fernando 2000.
California earthquake of February 9, 1971, vol. 3, US Dept. of [34] Salençon J, Indentation of a cohesive soil by a circular pipe. In:
Commerce, NOAA, Washington (DC); 1973. p. 189–196. Proceedings 15th international congress on soil mechanics and
[10] Faxèn OH. Forces exerted on a rigid cylinder in a viscous fluid foundation engineering, Istanbul; 2001. p. 1311–4.
between two parallel fixed planes. Proc Roy Swedish Acad Eng Sci [35] Seed HB, Harder J, SPT-based analysis of cyclic pore pressure
1946;187:1–13. generation and undrained residual strength. In: Duncan JM, editor.
[11] Finn WDL. State-of-the-art of geotechnical earthquake engineering Proceedings H. Bolton Seed memorial symposium, vol. 2, University
practice. Soil Dynam Earthquake Eng 2000;20:1–15. of California, Berkeley, May 9–11; 1990. p. 351–76.
[12] Fortin M, Glowinski R. Augmented Lagrangian methods. Amster- [36] Taboada V, Dobry R. Centrifuge modelling of earthquake-induced
dam: North-Holland; 1983. lateral spreading in sand. J Geotech Geoenviron Eng 1998;124(12):
[13] Fortin A, Côté D, Tanguy PA. On the imposition of friction 1195–206.
boundary conditions for the numerical simulations of Bingham fluid [37] Tamate S, Towhata I. Numerical simulation of ground flow caused by
flows. Comput Meth Appl Mech Eng 1991;88:97–109. seismic liquefaction. Soil Dynam Earthquake Eng 1999;18: 473–85.
[14] Glowinski R, Le Tallec P. Augmented Lagrangian and operator [38] Towhata I, Al-Hussaini TM. Lateral loads on offshore structures
splitting method in non-linear mechanics. SIAM Stud Appl Math 1989. exerted by submarine mudflows. Soils Foundat 1988;32:26–34.
[15] Hadush S, Yashima A, Uzuoka R. Importance of viscous fluid [39] Towhata I, Orense RP, Toyota H. Mathematical principles in
characteristics in liquefaction induced lateral spreading analysis. prediction of lateral ground displacement induced by seismic lique-
Comput Geotech 2000;27:199–224. faction. Soils Foundat 1999;39(2):1–19.
ARTICLE IN PRESS

S. Montassar, P. de Buhan / Computers and Geotechnics xxx (2006) xxx–xxx 13

[40] Towhata I, Sasaki Y, Tokida KI, Matsumoto H, Tamari Y, Yamada [43] Wilson DW, Boulanger RW, Kutter BL. Observed seismic lateral
K. Mechanism of permanent displacement of ground caused by resistance of liquefying sand. J Geotech Geoenviron Eng
seismic liquefaction. Soils Foundat 1992;32(3):79–96. 2000;126(12):898–906.
[41] Towhata I, Toyota H, Vargas-Monge W, Dynamics in lateral flow of [44] Yasuda S, Kiku S, Yosida T, Shaking table tests on the friction force
liquefied ground. In: Proceedings of the 10th Asian regional confer- between the buried pipelines and the ground during liquefaction. In:
ence on SFME, vol. 1, Beijing, 1995. p. 497–500. Proceedings of the 24th national conference JSSMFE (in Japanese);
[42] Uzuoka R, Yashima A, Kawakami T, Konrad J-M. Fluid dynamics 1989. p. 759–60.
based prediction of liquefaction induced lateral spreading. Comput [45] Zisis TH, Mitsoulis E. Viscoplastic flow around a cylinder kept
Geotech 1998;32(3–4):243–82. between parallel plates. J Non-Newtonian Fluid Mech 2002;105:1–20.

You might also like