SOILS AND FOUNDATIONS Vol. 44, No. 4, 53-65, Aug.

2004
Japanese Geo technical Society

PREDICTION METHOD FOR LIQUEFACTION-INDUCED SETTLEMENT OF

EMBANKMENT WITH REMEDIAL MEASURE BY DEEP MIXING METHOD

MITSU OKAMURAi) and KEIICHI T AMURAii)

ABSTRACT
In order to mitigate liquefaction-induced embankment settlement, remedial measures are often implemented for
existing embankments. This paper describes a practical prediction method for embankment settlement due to founda-
tion liquefaction, which are remedied with the deep mixing method. Based on a number of centrifuge shaking tests of
embankment, an empirical relation between crest settlement and displacement of the remedied zone is established.
Then, a calculation method for displacement of the remedied zone is developed. In the method, a macroscopic failure
envelope and a plastic displacement potential in the general load space are considered to evaluate the subgrade reaction
forces on the base of the remedied zone. The method is capable of calculating not only horizontal, vertical or rotation-
al displacement alone, but their combined effect. The method is validated through comparisons of calculated results
with centrifuge test observations. The calculated displacement components compare quite well with those measured.
This method can be a useful tool for design of countermeasures based on a specified limit settlement.

Key words: deep mixing method, deformation, embankment, liquefaction (IGC: E2/E8)

dikes have been remedied by early 2000s and another

INTRODUCTION hundreds of kilometers are supposed to be done in the
Although river dikes in Japan have often been near future.
damaged during past large earthquakes, any earthquake Countermeasures in the remediation program are
effects were not taken into account in the design practice designed and constructed so as to provide containment
of river dikes, because river dikes were usually easy to for deformation of liquefied foundation soils away from
restore in a short term. The Hyogoken-nambu earth- embankment centerline toward free field by forming
quake of 1995 hit 6 m high dikes of the Yodo river remedied stiff zones under embankment toes (Adalier
causing them to subside by as much as three meters. The et al., 1998; Matsuo and Shimazu, 1998). This counter-
highly urbanized hinterlands were in real danger of an measure philosophy does not intend to prevent occur-
overflow due to the dike settlement. After this earth- rence of foundation liquefaction and crest settlement but
quake, the Ministry of Construction started a remedia- to prevent occurrence of excessive settlement associated
tion program against liquefaction-induced failures of with a large lateral deformation of liquefied foundation
vulnerable dikes. soils. A ground improvement technique by the deep
It has been reported that significant settlement of dikes mixing method has often been used for this purpose.
during past earthquakes was more or less associated with In the current practice, countermeasures are designed
liquefaction of foundation soils (TCCRFE, 1996). so that factors of safety for bearing capacity, sliding and
Through the thorough review of river dike damages rotation failure modes as well as for failure of the
during past large earthquakes in Japan, the authors are remedied zone itself are higher than certain values. This
aware that the dikes supported by foundation soils pseudo static design procedure may assure that displace-
without liquefied soil layers were seldom damaged, with ment of remedial countermeasures and deformation of
the maximum crest settlement being lower than about countermeasures itself are limited, but does not directly
0.5 m or 15% of the dike height. On the other hand, dikes assess crest settlement. Since allowable crest settlement
were often damaged and the crest settlement sometimes due to an earthquake differs considerably for different
exceeded 50% when foundation soils or the dike itself dikes and the minimum requirement for river dikes is that
liquefied. In the current remediation program, therefore, the crest of dikes remains higher than a river water level,
only dikes with potential of large settlement due to the crest settlement should be directly considered in the
foundation liquefaction and with potential overflow of design procedures. A practical method is needed to
river water are considered. Hundreds of kilometers of estimate settlement of embankments with the remedial
i) Ehime University, Japan ([email protected]).
ii) National Institute for Land and Infrastructure Management, Japan.
Manuscript was received for review on August 19, 2003.
Written discussions on this paper should be submitted before March 1, 2005 to the Japanese Geotechnical Society, 4-38-2, Sengoku, Bunkyo-
ku, Tokyo 112-0011, Japan. Upon request the closing date may be extended one month.

53

This is an Open Access article under the CC-BY-NC-ND license.

NII-Electronic Library Service

54 OKAMURA AND TAMURA

Table 1. Summary of test conditions of solidification model

Embankment
1:2 slope
Acrylic block
Acrylic block
Soil type of Peak input
Test Embedment
Width, Height, bearing acceleration,
code depth,
B (m) H(m) stratum Amax (g)
D (m)
Dense sand
Sl 6.0 8.5 0.5
-E--23. o ~
~--------?5.0--------- S2 6.0 10.5 2.5
◄111111---........
~ Shaking direction Dense sand
S3 10.0 8.5 0.5
0.42
Fig. 1. Centrifuge model embankment with solidified zone at both S4 10.0 10.5 2.5
sides of toes (after Okamura and Matsuo, 2002a) S5 6.0 8.5 0.5
Soft clay
S6 10.0 8.5 0.5

countermeasure. S7 6.0 8.5 0.5

Dense sand 0.25
In this study a practical method is developed for S8 6.0 10.5 2.5
predicting crest settlement of embankments remedied
with the deep mixing method, aiming at providing a tool
for design of countermeasures based on a specified limit
settlement. The proposed method is validated through
comparisons of calculated results with centrifuge test
observations. In this paper, the improved zones by the
deep mixing method are assumed as a rigid block and
deformation of the improved zones itself, which may be
termed as the internal stability, is out of the scope.
Fig. 2. Deformation of model Sl after shaking event (after Okamura
and Matsuo, 2002a)
DOMINANT FACTORS OF CREST SETTLEMENT
In this section, the experimental work done by
Okamura and Matsuo (2002a) is briefly reviewed first. deformation. After the model was fully saturated,
From observations of a series of centrifuge model tests on ground water level was lowered 1. 75 m below the soil
embankments underlain by loose saturated sand with and surface and the 5 m high model embankment with the
without countermeasures, they clearly demonstrated the slope angle of 1:2 and the crest width of 3 m was
major factors that contributed the crest settlement. Those constructed. Two additional models of which the bearing
centrifuge tests will also be used in the later part of this stratum was soft clay instead of dense sand were also
paper to verify the validity of the settlement prediction constructed.
method proposed in this study. Horizontal base shaking was imparted to the models in
Figure 1 depicts the configuration of the model a centrifuge in flight at 50 g, where g = the gravitational
embankment resting on the liquefaction prone sand layer acceleration. The base shaking was strong enough to
with solidified zones at the embankment toes. All scales liquefy the loose sand below the ground water table.
regarding the centrifuge models are represented in the A total of eight tests were carried out as summarized in
prototype scale throughout this paper, unless otherwise Table 1. These experiments were similar in all respects,
mentioned. The model foundation soil consisted of three except that three main testing parameters were varied
layers; a 1 m deep dense gravel layer on the bottom of the between tests: size (width and height) of the acrylic
container served as a drainage layer, a 4.5 m deep dense blocks, the peak input acceleration, Amax, of the base
sand layer with relative density Dr= 90% served as a shaking and the soil type of the bearing stratum. Detailed
bearing stratum of the solidified zones, and a 8 m deep information on the tests is reported by Okamura and
loose sand layer with a liquefaction resistance of 0.14 in Matsuo (2002a).
terms of the cyclic shear stress ratio to produce a 5% Shown in Fig. 2 is a photograph of one of the models
double amplitude axial strain in 20 cycles in triaxial tests. after shaking, in which lines in the foundation soil are
Acrylic blocks with roughened surfaces representing the dyed Silica sand put near the side glass of the container
solidified zones, which had the same unit weight as that of during model preparation so that ground deformation
the loose sand layer, were placed in the foundation soil. can be easily detected. From observations of the model
The acrylic blocks did not accurately model the treated deformation, it appeared that the following three major
soil by the deep mixing method but rather they were factors contributed the crest settlement; (i) shear
intended to be stiff and strong as compared with deformation of the embankment caused by horizontal
surrounding soils. This modeling corresponds to such a deformation of the underlying liquefied sand, (ii) lateral
field situation that treated zones are strong and stiff deformation of the liquefied loose sand layer and (iii)
enough to cause neither internal failure nor significant contractive volume change of the liquefied loose sand

NII-Electronic Library Service

 LIQUEFACTION COUNTERMEASURE FOR EMBANKMENT 55

Factor (i)
Shear deformation
of embankment

Fig. 3. Three major factors contributing crest settlement

~o
:g
2
(a)
ljt.9 C: s 2
(b)

·-'-"
c:~
0 ..c
~c:
~-a
0--
0
1-o

1j
1L.c
~..c::
-0 C:
-0 0

3~
~'+-<
Oo
i1s
O.,c
::E
::EB 0 Solidification model
0
C: 0 Solidification model
0
(.)
~ Rigid wall model

0 2 0 2
Settlement of embankment base, S 1 (m) Decrease in embankment height, S2 (m)
(calculated from factors (ii)+ (iii)) (calculated from factors (i))

Fig. 4. Relationship between measured and calculated settlement

below the embankment. These factors are schematically change in the area of the loose sand layer below the
illustrated in Fig. 3. The factors (ii) and (iii) are associ- embankment (volume change), and the horizontal
ated with settlement of the embankment base, while deformation at the toes were read off. Settlement S 1 and
factor (i) is associated with a decrease in the embankment S2 obtained by substituting these values into Eqs. (2) and
height. Generally, a change in volume of dikes during an (3) are plotted in Fig. 4 against those measured at the
earthquake is expected to be small. Also slopes of dikes center of the embankment base in the tests. In Fig. 4(a),
are stable even during earthquakes due partly to their there is a good correlation of the embankment base
gentle slope angles, typically 1:2 to 1:3, and partly to the settlement. The measured settlement is proportional to
apparent cohesion of soil materials at a natural water that estimated from factors (ii) and (iii), irrespective of
content. the size of the solidified zone, the soil type of the bearing
Assuming uniform settlement along the embankment stratum and the peak input acceleration. This is also the
base as well as uniform shear deformation of the case for the change in embankment height as shown in
embankment, the crest settlement, Sc, due to the embank- Fig. 4(b). The slopes of the least-square-fit straight lines
ment base settlement, S1, and the change in embankment provided in these figures depend on unevenness of the
height, S2, can be written as, settlement profile along the embankment base and
heterogeneity of the shear deformation of the embank-
Sc=S1+S2 (1)
ments.
a, Okamura et al. (2003) also conducted a series of
S1 =2-+evHT (2)
BE centrifuge shaking tests on embankments with vertical
rigid walls at toes fixed to the soil container as illustrated
S _2drHE
2- (3) in Fig. 5. This model corresponds to an extreme field
BE condition that remedied zones are stiff, wide and deep
where a,= area of the lateral deformation of the block, enough not to cause any deformation and displacement.
ev = volumetric strain of the liquefied soil beneath the The foundation soil was a 9.5 m deep liquefaction prone
embankment, HT= thickness of the liquefied soil layer, loose sand overlain by a 2.5 m deep unsaturated sand.
dt = lateral deformation at embankment toe, BE= width Twelve models were constructed and subjected to a
of the embankment base and HE= embankment height. horizontal base shaking to liquefy the foundation soil.
From photographs taken in-flight, just before and after The main testing parameters varied between these tests
the shaking, the area of the lateral deformation, the were size and shape of the embankments, that is, height,

NII-Electronic Library Service

56 OKAMURA AND TAMURA

Testing parameters
BE(m): 15,20,25,30,45, 50
Block ( mass, m)
e Acelerometer
HE(m): 0, 2.5, 5, 10 / / Embankment
• Pore pressure cell n : l, 2, 4 -- - ---- - ----- -- -- -- --

p Liquefied sand

~f ---------
r-i
X, x: Relative displacement
P: Resultant force of earth pressure on embankment
Fig. 5. Centrifuge model embankment with vertical rigid walls at side and free field side
embankment toes (after Okamura et al., 2003)
Fig. 6. Single degree-of-freedom problem; model representing a block
sliding on rigid bearing stratum during base shaking
slope angle and crest width of the embankments. In this
model condition, the embankment settlement due to the
factors (i) and (ii) is zero. The volume change of the of the block and of horizontal acceleration of the rigid
foundation soils were read off from photographs. The bearing stratum, the motion of the block is described by
calculated settlement due to the factor (iii) is plotted an equation of motion,
against that measured in Fig. 4(a).
mx=P-H-mX (5)
It is clearly seen that results of the rigid wall models lie
on the same best-fit line despite the different testing where X = horizontal displacement of the rigid bearing
conditions from those of the acrylic block models stratum, x = relative displacement of the block with
including depth of the liquefied layer and the non-lique- regard to the bearing stratum, H = frictional force at the
fied unsaturated sand layer at the surface, the imparted interface between the block and the bearing stratum and
acceleration time histories of base shaking and the size, m = mass of the block. Provided that the force His given,
the shape of the embankment and the fixity of the wall sliding displacement of the block, x, can be calculated by
and the block. integrating Eq. (5) twice, in much the same way as the
From these observations it can be concluded that the sliding block analysis originally proposed by Newmark
crest settlement is reasonably estimated by Eq. (4) and the (1965).
following equations, The sliding block analogy as well as the model shown in
Fig. 6 is only available to calculate sliding displacement
(4)
alone. But, generally, movements of a remedied zone on
where c 1 ( = 1.9) and c2( = 1.2) are constants. Crest settle- a bearing stratum in reality cannot result from purely
ment is represented as a sum of the settlement at the sliding, subsidence or rotation but their combined effect.
embankment base (c 1 Si) and the change in the embank- Not only sliding displacement but also rotation is of
ment height (c2S2). In other words, crest settlement can primary importance, since horizontal displacement at the
be estimated with a good accuracy provided that the top of the remedied zones is one of the dominant factors
displacement of solidified zones and the volume change of the embankment settlement. In addition, the frictional
of liquefied sand beneath embankments are obtained. force His not easy to be determined. In the calculations
With regard to the volumetric change of liquefied sand, of sliding block method, the maximum frictional force is
an empirical chart proposed by Ishihara and Y oshimine often assumed conventionally to be the weight of the
(1992) is ready to be used in practice. In the following block times the frictional coefficient for a cohesionless
sections of this paper, we concentrate on the displace- bearing stratum or to be the cohesion times the block
ment of the solidified zone, that is, the area of lateral width for a frictionless bearing stratum. This convention-
displacement and horizontal displacement at toes. al assumption is true only when the weight of a block is
much smaller than the bearing load capacity of the
bearing stratum and the moment load is negligible
DISPLACEMENT PREDICTION OF SOLIDIFIED (Okamura and Matsuo, 2002b). Existence of moment
ZONE load degrades the maximum frictional force H. Thus, the
Single Degree-of-Freedom Problem conventional assumption tends to overestimate the
Consider a rigid block resting on ideal rigid level frictional forces, resulting in the estimation of sliding
ground as shown in Fig. 6, with an admissible movement displacement on the unsafe side.
of the block to be sliding only, which may be a simplest In the following sections, a method which is capable of
model for displacement calculation. For a given time calculating fully coupled displacement, that is vertical,
history of external horizontal force, P, acting on the side horizontal and rotational displacement, of a block under

NII-Electronic Library Service

 LIQUEFACTION COUNTERMEASURE FOR EMBANKMENT 57

;
;
;
;
;

Block ~B~ ,,//Embankment

Tlille
(a) example external load time history

E: external load vector

::i:: R.: subgrade reaction force vector
"g
..9
]
C:
Fig. 7. Forces acting on improved zone for coupled displacement ~o~--------~-
model
.§ Vertical load V t load MIB
::i::

(b) external load is within the failure envelope (F~ 0)

and the block is not moving
earthquake loadings is developed. In this method, a
macroscopic constitutive law for the entire foundation
soil-block system is incorporated in which horizontal,
vertical and moment forces on the base of the block are
considered.

Fully Coupled Displacement Prediction

Figure 7 provides a schematic illustration of a block
(c) external load violates the failure condition (F>O)
beneath the embankment toe resting on a bearing stratum
of an ideal rigid-perfectly plastic body. Assuming that the
direction of the input acceleration to the bearing stratum
as well as the earth pressures acting on the side of the
block from the liquefied soil are horizontal alone, the
o~---=------~-
motion of the block relative to the bearing stratum can be v, 0V MIB,Bt50
expressed by the equations of motion as follows:
· (d) external load comes down within the failure envelope (F~O)
-V+mg=my (vertical) (6) but the block is still moving
P-H-mX=mx (horizontal) (7)
Fig. 8. External load, subgrade reaction forces in the general load
P(yp-yo)+ H·yo-M=Io0 (rotational) (8) space and superimposed incremental displacement vectors

in which 0 = rotation of the block, Io= polar moment of

inertia of the block about the center of gravity, Yo= the forces (V, H, M/B) and their work conjugate
height of the center of gravity from the base, P= earth displacements (y, x, B0) take the place of element stress-
thrust from liquefied sand, yP = height of the acting point es and strains in the formulation of the macroscopic
of P, and y and x are vertical and horizontal displacement constitutive law for the entire bearing stratum-block
at the center of gravity relative to the bearing stratum. system. The failure envelope and the plastic displacement
The subgrade reaction force acting on the base of the potential as well as the earth thrust, P, used in the
block can be divided into three components of vertical calculation will be discussed later in this paper.
and horizontal forces, V and H, and moment with regard
to the center of the base, M. In Eqs. (6)-(8), subgrade Computational Procedure
reaction forces, V, H and MI B and earth thrust, P, are Computational procedure for displacement of the
unknown values and have to be properly determined to block resting on the bearing stratum having the failure
obtain displacement of the block x, y, and 0, under the envelope, F=F(V, H, M/B) =0, subjected to an external
action of a given base acceleration X. To preserve load, E<1) = { Ve<t>, He<t), Me<t) / B} at a time tis schematical-
dimensional homogeneity the moment, M, is divided by ly illustrated in Fig. 8. The external load, E, includes all
the block width, B. the loads acting on the block except for the subgrade
In order to determine sub grade reaction forces ( V, H, reaction force on the base, R<t) = { Vit), H<1), M<t) I B}. Also
MIB), a macroscopic constitutive law for the entire shown in Fig. 8(a) is an example time history of the
bearing stratum-block system is introduced in this study. external load. The time period is divided into n equal
The behavior of the block resting on an ideal rigid- segments, ti, t2, · · ·, ti,· · ·, tn, starting from t = 0 when the
perfectly plastic body can be described by a failure block stops under the equilibrium condition of E=R, as
envelope and a plastic displacement potential, in a shown in Fig. 8(b). The block does not move relative to
manner consistent with the theory of plasticity (Butter- the bearing stratum and the equilibrium holds (E=R)
field, 1985; Martin and Houlsby, 2000). In other words, while the external load vector E stays within the failure

NII-Electronic Library Service

58 OKAMURA AND TAMURA

envelope. When the external load becomes larger at t = ti M/B

and violates the failure condition, the subgrade reaction

force is no longer the same as the external load and the
difference in the forces (E<r,) - R<r,)) accelerates the block
(Fig. 8(c)). The work-conjugate displacement of the
block, s= {x, y, B0}, can be obtained by integrating the
equations of motion. Knowing the relative velocity s\r,-,l
and displacement S(t,-,l at the preceding time step, the rela-
tive velocity and the displacement of the block at this i-th (a) perspective view
time step can be derived by integrating the relative
acceleration s<t,) = { .Y(l,), .x<t,), B0<t,l};

(9)
~v
0 Vmd2 Vmax
scr,i +s<t,-,) dt+s
I
t,
s = (b) projection on MIB- V plane
(t,) 2 (t,_,) (10)
li-1

This calculation is repeated until the velocity of the block

drops to zero, even after the external load E returns H-H~
within the failure envelope (Fig. 8(d)). Here, all what we
know about the subgrade reaction forces is that the 0
- )~v
Vmd2 Vmax
bearing stratum is in the failure condition and thus, the (c) projection on H-Vplane
subgrade reaction force vector locates somewhere on the MIB
failure envelope. In order to identify the location of R on
the failure envelope, the flow rule is employed. In a
manner consistent with the theory of plasticity, a plastic
displacement potential, Q, is defined in which the ratio of
Hmax = c,,B H
incremental displacement is related to the state of the
load (Butterfield, 1985). The vector R is determined so
that R is on the failure envelope and the direction of the
incremental displacement vector is normal to the plastic
displacement potential. More detailed information on the
calculation procedure is given by Okamura and Matsuo (d) Section perpendicular to V axis at V<Vmd2
(2002b).
Fig. 9. Failure envelope of clay in general V-H-M/B load space

FAILURE ENVELOPE AND PLASTIC

DISPLACEMENT POTENTIAL thus the subgrade reaction forces. Hmax is the ultimate
horizontal load when M=0 and V=0, Mmax is the
Clay Bearing Stratum ultimate moment when V = 0 and H = 0 and Vmax is the
For a rigid block founded on a clay with a level surface maximum vertical load capacity for the central vertical
subjecting a combined load (V, H, M/B), the failure loading (H=M=0). The reference point of the moment
envelope in the V-H-M I B general load space has been is the center of the block base. Hmax, Vmax and Mmax may
analyzed and several researchers have proposed be taken as CuB, CuNcB and CuNcB /2, respectively, where
mathematical expressions (Murff, 1994; Georgiadis and Cu is undrained strength of clay and Ne is the bearing
Butterfield, 1988; Nova and Montrasio, 1991; Bransby capacity factor. Figures 9(b) and (c) show the cross
and Randolph, 1998; Ukritchon et al., 1998; Martin and sections of the failure envelope in the MI B- V plane ( H =
Houlsby, 2000). One of the simplest expressions of the 0) and in the H- V plane (M = 0) to be constant for
failure envelope for clay is of the form shown in Fig. 9 V ~ Vmax /2 and parabolas for V> Vmax /2. The cross
and Eqs. (11) and (12). section of the failure envelope in the MI B-H plane at a
for V~ Vmax /2 constant Vis an ellipse with an axis ratio of Ne /2.
2 2 2 2 In this study associated plasticity is assumed for clay so
H ) + ( -M- ) -16 ( - V ) ( 1 -V-) -0
F- ( ~ (11) that the failure envelope also describes the plastic
Hmax Mmax Vmax Vmax displacement potential defining the relative magnitudes
for V < Vmax /2 of the incremental plastic displacement during failure.
2 2 The assumption of normality to the failure envelope is
F- ~ H ) + ( -M- ) -1-0
( Hmax
(12) considered to be adequate for the first approximation for
Mmax the reason that the validity of the associated flow rule in
The symbols V, Hand M stand for the total vertical and the H-M / B plane is confirmed through model tests
horizontal forces and moment sustained by the ground, (Martin and Houlsby, 2000).

NII-Electronic Library Service

 LIQUEFACTION COUNTERMEASURE FOR EMBANKMENT 59

Sand Bearing Stratum the block of the centrifuge models in the range of B = 6 m
The failure envelope of a sand in the V-H-M I B general to 10 m and the embedment depth D = 0.5 m to 2.5 mare
load space has been analyzed by Georgiadis and determined to be (f) = 0.46, µ = 0.52. The constant p = 22°
Butterfield (1988), Nova and Montrasio (1991), Gottardi is invoked based on the footing test results (Saito et al.,
and Butterfield (1995), Butterfield and Gottardi (1994) 2002).
and Okamura et al. (2002) among others. One of the With regard to the plastic displacement potential for
simplest expressions of the failure envelope is of the form sand, it has been reported that incremental rotation and
shown in Fig. 10 and Eq. (13) (Butterfield and Gottardi,
1994), which has been termed as a 'cigar shaped enve- MIB
lope'.
H
2 2
H ) ( M ) 2CHM
F= (µ Vmax + (f)BVmax - µ(f)BV~ax
2 2
-
(
1-~
V ) (~ V ) -o (13)
Vmax Vmax
(a) perspective view
C= tan 2p (µ- (f))(µ + (f)) (14)
2µ(/J MIB

Figures lO(b) and (c) show the cross sections of the failure
envelope in the MI B-V plane (H = 0) and in the H-V
plane (M = 0) to be parabolas which intersect the Vaxis at ~ v
V = 0 and V = Vmax, with the tangential slopes of either (f)
0 vmax
(b) projection on MIB-V plane
or µ. The constant µ is the friction coefficient at the
foundation base. The cross section of the failure envelope
in the MI B-H plane at a constant Vis an ellipse with an
axis ratio of (f) Iµ, which is rotated at an angle p around l ~ V
the center of the ellipse as indicated in Fig. lO(d). The
constantsµ, (f) and p are reported to be insensitive to the 0 v,,,~
(c) projection on H-V plane
density of sands (Nova and Monstansio, 1991), but varies
with some factors including width, B, and embedment MIB
depth of the foundation. Figure 11 demonstrates
variations ofµ and (f) with the foundation width and the <pVma/4
embedment depth obtained from bearing capacity tests
(Saito et al., 2002; Okamura et al., 2002). It is inferred
that (f) increased with foundation width and approached
to 0.5 which is the maximum value of (f), since the
maximum possible eccentricity of any vertical load is
0.5B. Figure 1 l(b) exhibits the variation of the ratio µ/(f)
with the embedment ratio DI B which was obtained from
small scale bearing capacity tests at 1 g. It is assumed that (d) Section perpendicular to V axis at V= Vmd2

this relationship hold good for the wider footing width.

Fig. 10. Failure envelope of sand in V-H-M I B general load space
The constants used in the following section to simulate

--- --- ---

Footing tests at lg and at 75g

centrifugal acceleration

2 4 6 8 10 0.0 0.1 0.2 0.3 0.4 0.5

Foundation width, B (m) Embedment ratio, DIB

(a) Variation of <pwith foundation width (b) Variation of µ/<p with embedment ratio DIB
(reproduced from Okamura et al., 2002) (after Saito et al., 2002)

Fig. 11. Variations of constants

NII-Electronic Library Service

60 OKAMURA AND TAMURA

MIB, B88

V, 8y

Fig. 12. Perspective view of plastic displacement potential

horizontal displacement at failure are fairly consistent

with the associated flow in the MI B-H plane but the
associated flow rule apparently contradicts the real
behavior and significantly overestimates vertical displace-
ment in the upward direction when the vertical load is
lower than Vmax/2 (Gottardi and Butterfield, 1995;
Bransby and Randolph, 1998; Martin and Houlsby,
2000; Okamura et al., 2002). This indicates the clear need
of a separate plastic displacement potential to describe
the block behavior. The plastic displacement potential
determined based on the model test results by Georgiadis
and Butterfield (1988) and Okamura et al. (2002) is shown
in Fig. 12 and Eq. (15). The constants, (f)' and µ', in Fig. 13. Typical responses of rigid wall model: embankment height= 5
m, 1:2 slope angle and crest width= 3 m. Locations of instrumenta-
Eq. (15) are 0.35 and 0.45, respectively (Okamura and
tion are given in Fig. 5.
Matsuo, 2002b).
2 2 2
Q= - - + (--
( H ) M~ ) +0.44 ( ~
V ) -0.44=0
µ' Vmax (f)' BVmax Vmax been investigated.
(15)
However, as far as authors are aware, information
It should be noted here that, as far as authors are about the effect of embankments existing in the vicinity
concerned, available information about the shape of of the walls on both the monotonic and the dynamic
plastic displacement potential is limited in the literature, component of the earth pressures is very limited.
especially for sand. The values of the constants in Okamura et al. (2003) conducted a series of centrifuge
Eq. (15) are tentatively derived from a limited number of tests to investigate the effects of embankments near the
bearing capacity test results. Accumulation of experimen- walls. In the test, as previously mentioned in Fig. 5, earth
tal data and the revision of the plastic displacement pressures on the rigid walls which were fixed to the soil
potential function may be needed. container were measured with twelve built-in earth
pressure cells with small accelerometers attached on
them. The earth pressure cells were capable of measuring
EARTH PRESSURES ACTING ON BLOCK both normal and shear stresses simultaneously.
Liquefaction Layer Presented in Fig. 13 is typical centrifuge model
Horizontal earth pressures on rigid vertical walls of responses including the earth pressures, the acceleration
liquefied soils with a level ground surface have extensively of the pressure cell and the excess pore pressure at the
been studied by Tsuchida (1968), Dewoolkar et al. (2001) depth from embankment base z= 8 m. The prototype
and Tamari and Towhata (2003) amongst others. It has input acceleration time history depicted in Fig. 13(a)
been recognized that time histories of earth pressures consisted of four cycles of sinusoidal wave with 1 Hz
during earthquakes consist of a monotonic increase and frequency and 0.5 g peak acceleration, followed by a
dynamic fluctuations. The monotonic component is series of sinusoidal waves with 2 Hz frequency and of
essentially equal to the total overburden pressure at any sinusoidal waves with 1 Hz frequency, with the accelera-
depths, while the dynamic component can be well tion amplitude of the both series of waves increasing
represented by the Westergaard's formula (Westergaard, from 0.1 g to 0.45 g. The frequency and the acceleration
1933) as long as liquefied soils behave as heavy fluid. amplitude of the first four cycles are probably not very
Effects on the dynamic component of influential factors important provided that it liquefies the foundation soil.
such as the rotational and translational movements and The following two series of waves were selected to see the
flexibility of walls (Matsuzawa et al., 1985; Dewoolkar et effects of the frequency and the acceleration amplitude on
al., 2001; Tamari and Towhata, 2003), and the possibility the earth pressures of liquefied soils.
of resonance (Werner and Sundquist, 1943) have also The broken line in Fig. 13(b) represents the initial

NII-Electronic Library Service

 LIQUEFACTION COUNTERMEASURE FOR EMBANKMENT 61

Table 2. Coefficients for calculating of residual earth pressure (after

Okamura et al., 2003)

Depth of the bottom of liquefaction layer

Coefficient
4m Sm 12m 16m
2
a110/m ) -0.0921 -0.0887 -0.0839 -0.080

a12 (1 /m) 0.694 0.187 0.0290 0.014

b 1 (I /m) -3.35 -0.959 -0.190 0

a 22 (m) -0.167 -0.161 -0.154 -0.147

b 2 (m) 1.42 3.63 5.83 8.04

-100 .___ _....___ ___.__ _ _..___ __, a31 2.11 4.41 6.71 9.01
-1 -0.5 0 0.5
Acceleration of wall (ac20) (g) a 32 (m) -7.14 -6.24 -5.34 -4.44

b 3 (m) 34.9 30.8 26.6 22.4

Fig. 14. Relationship between horizontal earth pressure on the wall
and wall acceleration after foundation soil liquefied Eq. (19) is applicable for embankment with HE= 2 m-10 m and n = 1-4

effective vertical stress, a~ 0 , calculated assuming the of earth pressures from liquefied soils acting on both the
embankment height as an uniform vertical surcharge. embankment side and the free field side of the block. The
Excess pore pressure increased and reached the initial earth thrust Pr can be obtained by integrating the residual
effective vertical stress to level off in a few cycles of input earth pressure at depth z (m) from ground surface, Pr(z),
motion, indicating that the foundation soil liquefied at which is given by Okamura et al. (2003) as follows,
about t = 3 s and continued to liquefy throughout the
shaking event.
The monotonic component of the horizontal earth
Pr= r 0
Pr<zl dz (17)

pressure represented by the broken line in Fig. 13(d)

(18)
started to increase at the very beginning of the shaking
and reached the plateau at about t = 3 s, showing a similar A =a11 ·HE+a12·n+b1
trend to the excess pore pressure. It should be noted that
the fluctuation of the earth pressure, that is the dynamic B= -2A(a22·n+b2)
component, is quite similar to the acceleration of the wall
(Fig. 13(c)) especially after the soil liquefied. Figure 13(e)
is the plot of the shear stress on the wall measured with
where HE= embankment height (m), y = unit weight of
the same earth pressure cell (ep8), indicating that there
the embankment soil (kN /m 3 ) and n = embankment slope
was essentially no friction mobilized between the wall and
angle. Coefficients a11, a 12, a22, a31, a32, b1, b2 and b3 are
the soil after the soil liquefied.
given in Table 2. It is often assumed that a residual earth
The relationship between the horizontal earth pressure
pressure of liquefied soil on a wall is the same as the
and the acceleration of the pressure cell, and thus the
overburden pressure at the depth for the case of level
acceleration of the rigid wall during the two series of
ground surface. The residual earth pressure is greater if
shaking waves, is given in Fig. 14. It appears that dynam-
an embankment exists on the ground surface. Equation
ic component of the earth pressure of liquefied sand
( 18) represents the residual earth pressure increase due to
was in proportion to the acceleration of the rigid wall
the existence of an embankment. It should be noted that
irrespective of amplitude and frequency of the input
Eq. (18) with the coefficients shown in Table 2 is provided
motion. Such a linear relationship remained true in-
for easy reference in practice and only applicable to the
dependently of the shape and size of the embankment.
embankment with a height and a slope angle in the ranges
The horizontal earth pressure for a certain wall accelera-
of HE= 2 m-10 m and n = 1-4. A method to estimate the
tion, that is the slope of the relation, was the function of
residual earth pressure for embankments with any height
the shape and size of the embankment, depth of the lique-
and slope angle is given by Okamura et al. (2003).
fied layer and the unit weight of liquefied soil. From these
For the dynamic component of the earth thrust Pct and
observations, Okamura et al. (2003) concluded that the
the added mass ma, Okamura et al. (2003) expressed as:
earth thrust from liquefied layer on the side of the wall,
P, has the relation,
(16)
P,-(2.0+4.98xJ0- 3 : )-JJ; y,,.,fthz)dz (19)

Pct
where Pr= residual earth thrust (monotonic component) ma=- (20)
and ma= added mass representing the dynamic compo- g
nent of the earth thrust. Note that Pis the resultant force in which HT= depth of the bottom of liquefied layer,

NII-Electronic Library Service

62 OKAMURA AND TAMURA

Ysat = saturated unit weight of liquefied soil, q = average

embankment load (weight of the embankment divided by VALIDATION OF PROPOSED METHOD
the base width) in kPa, q 0 = unit pressure in kPa and g = The proposed method is utilized to simulate the cen-
the gravitational acceleration ( = 9.8 m/s 2). Substituting trifuge models of embankments with the acrylic blocks as
Eq. (16) into Eqs. (7) and (8), acceleration of the block a countermeasure shown in Fig. 1 and Table 1, subjected
can be expressed as to the base shaking (Okamura and Matsuo, 2002a). The
method is validated through the comparison of computa-
.x= -Pr-H
- - -x.. (21) tional results with observations in the tests.
ma+m
Pr-H +Pr ) (yp-yo)+H·yo-M Input Parameters
( ma
0=-_,____m_a_+_m_ __,__________
The necessary input data for the proposed calculation
(22) method, except for the constants in the failure envelope
Io
and the plastic displacement potential for sand, includes
It is reasonable to consider that Pr and ma decrease with the maximum load capacity, Vmax, the width, B, the mass,
increasing settlement of embankments during an earth- m, and the polar moment of inertia, Io, of the block and
quake. This influence of the embankment settlement is the input base acceleration time history.
taken into account in the calculation by updating The undrained strength of the clay in the bearing
embankment height at each calculation time step. stratum for the models S5 and S6 was estimated to be
53 kPa based on the measured water contents after the
Non-Liquefaction Layer at Ground Surface centrifuge tests, the stress history of the clay in the
It was observed in the rigid wall model tests that the models, and the undrained strength-increasing ratio,
residual earth pressures of the non-liquefied layer at the Cu/ a~, obtained from triaxial tests. The vertical load
ground surface were very small as compared to that of the capacity of the clay is determined as, Vmax = cuBNc, where
liquefied layer. On the other hand, dynamic component Ne is the bearing capacity factor taken as 5.14.
of earth pressures of the non-liquefied layer was more or On the other hand, the angle of shear resistance of the
less similar to that of the liquefied layer. In this study, dense sand in the bearing stratum was determined to be
only dynamic component of earth pressures on the block ¢' =42°. This angle is based on plane strain compression
from the non-liquefied soil layers above the liquefied test results with effects of the stress level and the
layer are considered. Equation (19) is used to estimate the anisotropy (Okamura et al., 1993) taken into considera-
dynamic component for the non-liquefied soil layer. tion. The vertical load capacity of the dense sand is
obtained as Vmax = 0.5y' BNr, where Nr( = 156) is the bear-
Bearing Stratum ing capacity factor (Vesic, 1973). It should be mentioned
The bearing capacity characteristics described in the that displacement of the block calculated by the proposed
preceding section are the ultimate loads mobilized on the method was less sensitive to the angle ¢', because the
base of the block. In cases of the block being embedded vertical load Von the base of the block was much smaller
into the bearing stratum, the earth pressures on the side than Vmax of the beating stratum.
of the block in the bearing stratum have to be considered The volumetric strain of liquefied sand beneath the
separately. The earth pressures invoked in the calculation embankment used to calculate settlement due to the
are the Rankine's passive pressure on the free field side factor (iii) was 3%, which was derived from the empirical
and the active pressure in the embankment side. chart (Ishihara and Yoshimine, 1992). This strain value
was fairly consistent with the observed volumetric strain
Negative Excess Pore Pressure on the Base of 3.5% in the tests. The acceleration measured on the
The failure envelope for sand shown in Fig. 10 is base of the container was used as the input base accelera-
relevant to cohesionless soil in the drained condition that tion in the calculations.
is the usual case in the bearing capacity problem of sand
in the practice. However, sand in the undrained condition Comparison of Computational Results with Test
permits excess pore pressures to generate. Even for a Observations
foundation without vertical loads, moment loads can be The measured displacement time histories of the blocks
sustained by the negative excess pore pressure which may are compared with those calculated from the proposed
be generated when the foundation base moves upward. In method in Fig. 15, together with the imparted base
order to account for the effect of the negative pressure on acceleration. Figures 15(b) and (c) represent displace-
the sustainable moment load, for the first approximation, ments of the blocks resting on the dense sand stratum and
the negative excess pore pressure on the half width of the the clay stratum, respectively, with both the block having
base of the block, of which maximum value equals to the the width of 10 m and the embedment depth of 0.5 m.
hydrostatic pressure at the depth plus 98 kPa, is consid- Overall, both horizontal and rotational displacements
ered to resist moment loads with regards to the center of obtained from the calculation agree quite well with those
the base. The moment sustained by the negative pressure observed for the two models. The displacement of the
is subtracted from the external moment load at each time block increases during strong base shaking of about 0.4 g
step in the calculation. acceleration amplitude (t = 4-12 s), and neither monoton-

NII-Electronic Library Service

 LIQUEFACTION COUNTERMEASURE FOR EMBANKMENT 63

g8~~::
:§

]
]
(a) Input

-0.2

0
oose acceleration

-0.4 .___...__ _.____

IO
__.___ _.__ __,___
20
Time (s)
__.__
30
.
___.__
~
___.40
E
a
(I)
4 ------------------------"""T"""-

0.
0

'E
~ 2
ti
~
0
u
"0

*
"3
..::!]
u"' Soil type of bearing stratum
0 sand
e clay

0 2 3 4
Observed crest settlement (m)

Fig. 16. Comparisons of calculated crest settlement with that observed

in tests

3
Calculated
§
2

I
11 Centrifugr.! test
D(m) bearing stratum Amax
~
u 0 0.5 sand 420g;:il
• 2.5 sand 420giil

0 5 IO
Width of block, B (m)

Fig. 15. Comparisons of calculated displacement of the block with

§3 Calculated.....__,
centrifuge test observations

I 2

ic increase nor fluctuation is seen during the succeeding 1

shaking with lower acceleration amplitude (t = 15-31 s). ~ 1 Centrifugr.! test
D(m) bearingstratum
u Amax
A significant difference between calculated and observed • 0.5 cla 420
0 5 IO
displacement time histories can be seen at the very begin- Width of block, B (m)
ning of the shaking event. In the calculation, the blocks
start to accumulate the monotonic displacement increase Fig. 17. Variation of crest settlement with width of block B
in the first cycle of the base motion, while the observed
block displacement was very small in the first cycle and
started accumulation in the second cycle as soon as the This may be arisen from decrease in the embankment
loose sand layer liquefied. This difference may be height as the shaking proceeded. The same trend is seen in
attributed to the fact that, in the calculation, the loose the displacement time histories calculated by the
sand layer is assumed to be liquefied and the earth proposed method which takes the effect of the decrease in
pressures on the side of the block are applied from the the embankment height into account by updating the
beginning ( t = 0) of the calculation. embankment height in each time step. Figure 16 is a plot
Since vertical load due to the self-weight of the block is of calculated crest settlement versus observed settlement
smaller than the half of the bearing load capacity ( Vmax/ for all eight tests. The proposed method estimated the
2) of the clay stratum and much smaller than that of the crest settlement reasonably well, with the calculated
dense sand stratum, the incremental displacement vector settlement mostly being on the safe side.
sis always almost perpendicular to the Jy axis in Figs. 9 Variations of crest settlement with the block width are
and 12, resulting in the calculated settlement of the block indicated in Fig. 17. It can be seen that the calculated
being essentially zero. This is consistent with the test settlement tends to be conservative for the cases of the
results that the observed settlement of the block is very blocks supported by the sand-bearing stratum. In the
small, of the order of 0.01 m. bearing stratum around the block, shear deformation of
It is also of interest to poirtt out that the accumulation dense sand imposed by the block displacement generates
of the observed displacement gradually slowed down with negative pore pressures, which causes the shear strength
time during strong shaking (t=4-12 s) although almost increase of the sand. The fact that this strength increase
the uniform amplitude of base motion was continued. being not taken into account in the calculation is

NII-Electronic Library Service

64 OKAMURA AND TAMURA

Centrif~ test condition that the walls were fixed to the model contain-
B(m) bearingstratum Amax
o 6 sand 420g;tl er. The model blocks, however, were translated and
• 10 clay 420~ rotated during the shaking events and the liquefied sand
A 6 sand 250!?Jil
§ 3 in the vicinity of the blocks deformed. When the displace-
ment of the block, and hence the deformation of the
5 r - - - - n = IO m, clay
liquefied sand becomes larger, the liquefied sand regains
~
j
2 - - -8 - \ - - - - - - -

• B = 6 m, sand,
04 Calculated
stiffness and strength in each cycle of shaking as a result
of the dilative behavior of sand in the undrained
u
iJ l 250 gt!
• condition at large shear strain as long as the sand is not
0 I 2 3 4 very loose to cause flow type failure (Ishihara, 1993). The
Embedment depth, D (m) earth pressures of liquefied sand employed in the
Centrif~ test calculation invokes that the liquefied sand always behaves
B(m) D (m) bearing stratum
O 6 0.5 sand as heavy fluid and the effect of the strength and stiffness
§3.._________
• 6 2.5
liiiii_ _ __.v
sand recovery is not considered. The calculated settlement may
be more conservative as the behavior of liquefied sand
] 2 0
becomes more dilative.
~ •
i 1
u CONCLUSIONS

500 600
In order to assess crest settlement of embankments
0 100 200 300 400
Maximum input acceleration, Amax (gal) remedied with the deep mixing method, a practical
prediction method is developed in this study. First of all,
Fig. 18. Variation of crest settlement with embedment depth of block dominant factors of crest settlement were discussed and
D empirical relationships were established which enable to
predict crest settlement from displacements of the im-
Centrif~ test proved zone and volumetric strain of liquefied soil layers.
B(m) D (m) bearing stratum
A new calculation method for displacement of the
§ o 6 ...,.
3..____ 0.5 sand_ _ _. .
___iiiiiiii
• 6 2.5 sand improved zone was then developed. In the method, a
macroscopic constitutive law for the entire improved
0
zone-bearing stratum system is considered to evaluate the
• subgrade reaction forces from the bearing stratum. The
method is capable of calculating not only horizontal,
vertical or rotational displacement alone, but also their
0 100 200 300 400 500 600
combined effect. The method is validated through
Maximum input acceleration, Amm (g;tl) comparison with centrifuge test results of embankments
resting on a loose saturated sand subjected to the strong
Fig. 19. Variation of crest settlement with maximum input accelera- base shaking. It appeared that the calculated displace-
tion Amax ment time history from the method compared quite well
with that observed in the tests after the foundation soil
considered as a reason for the conservative prediction for liquefied. The variations of calculated crest settlement
cases of blocks resting on the dense sand stratum. Shown with factors including width and embedment depth of
in Fig. 18 is the effect of the embedment depth of the improved zones and the peak input acceleration were
block on the crest settlement. The trend of the decrease in similar to those observed. In this method, however,
the calculated crest settlement with the embedment depth simplifications are made about the complicated nature of
is quite similar to that observed. Variations of crest soil behavior including; (1) liquefied soils are assumed as
settlement with the maximum input base acceleration, heavy fluid, and (2) the drained condition is basically
Amax, are depicted in Fig. 19. The proposed method assumed for the bearing capacity characteristics of the
simulated the test observations quite well that the ob- sand-bearing stratum. Although these simplifications
served settlement increased with Amax , with the settlement make the calculated crest settlement conservative, the
being larger for the block with smaller embedment depth. proposed method is an effective tool to assess displace-
However, the method tends to overestimate the test ment of solidified zones by the deep mixing method
observations as Amax and the settlement become larger. beneath embankment toes.
The similar tendency can be seen in Fig. 17 that the
method overestimates more marked when the block
REFERENCES
width is smaller and the settlement is larger. A possible
I) Adalier, K., Elgamal, A.-W. and Martin, G. R. (1998): Foundation
reason for this is that, in the proposed method, the earth liquefaction countermeasures for earth embankment, J. of
pressures on the side of the blocks from the liquefied Geo technical and Geoenvironmental Engrg. Div., ASCE, 124 (6),
layers were determined based on the observed pressures 500-517.
on the rigid walls in the loose liquefied sand, with a 2) Bransby, M. F. and Randolph, M. F. (1998): Combined loading of

NII-Electronic Library Service

 LIQUEFACTION COUNTERMEASURE FOR EMBANKMENT 65

skirted foundations, Geotechnique, 48 (5), 637-655. pressures on vertical walls with adjacent embankments, Proc. 11th
3) Butterfield, R. (1985): Load-path dependent stability of shallow Int. Conj. Soil Dynamics and Earthquake Engrg., 429-434.
footings (Discussion contribution), Soils and Foundations, 25 (1), 17) Okamura, M. and Matsuo, 0. (2002a): Effects of remedial
150-154. measures for mitigating embankment settlement due to foundation
4) Butterfield, R. and Gottardi, G. (1994): A complete three-dimen- liquefaction, Int. J. Physical Modelling in Geotechnics, 2 (2), 1-12.
sional failure envelope for shallow footings on sand, Geotechnique, 18) Okamura, M. and Matsuo, 0. (2p02b): A displacement prediction
44 (1), 181-184. method for retaining walls under seismic loading, Soils and
5) Dewoolkar, M. M., Ko, H. Y. and Park, R. Y. S. (2001): Seismic Foundations, 42 (1), 131-138.
behavior of cantilever retaining walls with liquefiable backfills, J. of 19) Okamura, M., Mihara, A., Takemura, J. and Kuwano, J. (2002):
Geotechnical and Geoenvironmental Engrg. Div., ASCE, 127 (5), Effects of footing size and aspect ratio on the bearing capacity of
424-435. sand subjected to eccentric loading, Soils and Foundations, 42 (4),
6) Georgiadis, M. and Butterfield, R. (1988): Displacements of 43-56.
footings on sand under eccentric and inclined loads, Canadian 20) Okamura, M., Takemura, J. and Kimura, T. (1993): A study on
Geotechnical J., 25 (1), 199-212. bearing capacities of shallow footings on sand, Proc. JSEC,
7) Gottardi, G. and Butterfield, R. (1995): The displacement of a 463(IIl-22), 85-94 (in Japanese).
model rigid surface footing on dense sand under general planar 21) Saito, Y., Okamura, M., Tamura, K. and Tsubokawa, H. (2002):
loading, Soils and Foundations, 35 (3), 71-82. On the seismic displacement of gravity retaining walls-bearing
8) Ishihara, K. (1993): Liquefaction and flow failure during earth- capacity of foundation soil, Proc. 57th Annual Conj. JSCE, CD-
quakes, Geotechnique, 43 (3), 349-415. ROM.
9) Ishihara, K. and Yoshimine, M. (1992): Evaluation of settlements 22) Tamari, Y. and Towhata, I. (2003): Seismic soil-structure interac-
in sand deposits following liquefaction during earthquakes, Soils tion of cross sections of flexible underground structures subjected
and Foundations, 32 (1), 173-188. to soil liquefaction, Soils and Foundations, 43 (2), 69-87.
10) Martin, C. M. and Houlsby, G. T. (2000): Combined loading of 23) TCCRFE (Technical Committee on Countermeasures for River
spudcan foundations on clay: laboratory tests, Geotechnique, 50 Facilities against Earthquake) (1996): Committee Report, 39 (in
(4), 325-338. Japanese).
11) Matsuo, 0. and Shimazu, T. (1998): Design and construction 24) Tsuchida, H. (1968): Dynamic pressures on walls due to liquefied
manual for countermeasures against liquefaction-induced river dike sandy soil deposit, Tsuchi-to-Kiso, 526, 3-10, (in Japanese).
failure, Technical Memorandum of Public Works Research 25) Ukritchon, B., Whittle, A. J. and Sloan, S. W. (1998): Undrained
Institute, 3513 (in Japanese). limit analyses for combined loading of strip footings on clay, J. of
12) Matsuzawa, H., Ishibashi, I. and Kawamura, M. (1985): Dynamic Geotechnical and Geoenvironmental Engrg. Div., ASCE, 124 (3),
soil and water pressures of submerged soils, J. of Geotechnical 265-276.
Engrg. Div., ASCE, 111 (10), 1161-1176. 26) Vesic, A. S. (1973): Analysis of ultimate loads of shallow founda-
13) Murff, J. D. (1994): Limit analysis of multi-footings foundation tion, J. of Soil Mechanics and Foundation Engrg. Div., ASCE, 99
systems, Proc. 8th Int. Conj. Comput. Methods. Adv. Geomech. (1), 45-73.
Morgantown, l, 223-244. 27) Werner, P. W. and Sundquist, K. J. (1943): On hydrodynamic
14) Newmark, N. M. (1965): Effects of earthquakes on dams and earthquake effects, Transactions American Geophysical Union, 30
embankments, Geotechnique, 15 (2), 139-160. (5), 636-657.
15) Nova, R. and Montrasio, L. (1991): Settlements of shallow founda- 28) Westergaard, H. M. (1933): Water pressures on dams during
tions on sand, Geotechnique, 41 (2), 243-256. earthquakes, Transactions, ASCE, 1835, 418-472.
16) Okamura, M., Ishihara, M. and Tamura, K. (2003): Liquefied soil

NII-Electronic Library Service