Seismic Lateral Response of Piles in Liquefying Soil

D. S. Liyanapathirana, M.ASCE,1 and H. G. Poulos, F.ASCE2

Abstract: Soil liquefaction is one of the major factors affecting the behavior of piles founded in seismically active areas. Although
methods are available for seismic analysis of pile foundations, in many of them, the supporting soil is assumed to be an elastic material.
Here a numerical model is presented which takes into account the reduction of soil stiffness and strength due to pore pressure generation
Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

and subsequent soil liquefaction, in addition to the material nonlinearity. Results obtained from the new method are compared with
centrifuge test data and show excellent agreement with the observed pile behavior during these tests. To investigate the effects of soil
liquefaction on the internal pile response, a parametric study is carried out for a range of material and geometric properties of the pile and
surrounding soil. The effect of the nature of the earthquake on pile performance has been studied using 25 earthquake records scaled to
different acceleration levels. It is shown that the “Arias intensity” and the natural frequency of the earthquake ground motion have a
significant influence on the pile performance in liquefying soil. Existing elastic analysis methods for kinematic pile loading in layered soil
deposits with soft and stiff layers are applied to the soil deposits with liquefying and nonliquefying layers. It is found that these simple
design methods, which were derived assuming that the soil is a linear elastic material, do not predict bending moments accurately when
nonlinear behavior of soil and effects of pore pressure generation are significant. Also a simplified limit equilibrium method proposed for
the evaluation of bending response of single pile foundations subjected to lateral spreading is compared with the bending response
obtained from the proposed numerical model. It is found that the limit equilibrium method, which is developed based on the centrifuge
test results, does not give accurate results when the pile diameter and the thickness of the liquefied soil layer deviates from the values used
for the centrifuge tests.
DOI: 10.1061/共ASCE兲1090-0241共2005兲131:12共1466兲
CE Database subject headings: Pile foundations; Earthquake loads; Liquefaction; Seismic effects.

Introduction centrifuge tests 共e.g., Dobry et al. 1995; Abdoun et al. 1997;
Horikoshi et al. 1998; Wilson et al. 1999; Wilson et al. 2000兲
The performance of piles in liquefying ground under earthquake provide an insight into the mechanism of pile–soil-structure inter-
loading is a complex problem due to the effects of progressive action in liquefying soil.
buildup of pore water pressures in the saturated soil. The loss of Numerical procedures for the analysis of piles founded in liq-
soil strength and stiffness due to liquefaction may develop large uefying soil have large uncertainties due to lack of understanding
bending moments and shear forces in piles founded in liquefying of the mechanisms involved in soil–pile interaction in the lique-
soil, leading to pile damage. The significance of liquefaction- fying soil. Although numerical models based on two-dimensional
related damage to pile foundations has been clearly demonstrated and three-dimensional finite element analyses 共e.g., Hamada et al.
by the major earthquakes that have occurred during past years 1994; Sakajo et al. 1995; Zheng et al. 1996; Shahrour and Ousta
such as the 1964 Niigata, 1964 Alaska, 1989 Loma-Prieta, and 1998; Finn et al. 2001兲 provide better insights into this interac-
1995 Hyogoken-Nambu. tion, they are computationally complex and time-consuming.
There remain many uncertainties in the mechanisms involved Therefore in recent years, one-dimensional Winkler models based
in pile–soil-structure interaction in liquefying soil, although the on finite element and finite difference methods for the seismic
data recorded during the 1995 Hyogoken-Nambu earthquake,
analysis of pile foundations have become popular amongst de-
shake table tests 共e.g., Ohtomo 1996; Tamura et al. 2000; Yasuda
signers. In Winkler models, the pile is modeled as a beam and the
et al. 2000; Mizuno et al. 2000; Nakamura et al. 2000兲, and
lateral soil pressure acting on the pile is modeled using a nonlin-
1
ear spring-dashpot model. These methods are computationally
Senior Lecturer, Dept. of Civil, Mining and Environmental Engineer- very efficient and give results in a very short time.
ing, Univ. of Wollongong, Northfields Ave., Wollongong, NSW 2522,
In Winkler models, a p-y curve is used to define the behavior
Australia.
2
Senior Principal, Coffey Geosciences Pty. Ltd. and Emeritus of the nonlinear spring at any depth, where p = soil resistance per
Professor of Civil Engineering, Dept. of Civil Engineering 共J05兲, Univ. of unit length of the pile and y = pile lateral displacement. These
Sydney, Sydney, NSW 2006, Australia. p-y curves should be back-figured from the field or model tests.
Note. Discussion open until May 1, 2006. Separate discussions must However, for liquefying soil, available case histories and experi-
be submitted for individual papers. To extend the closing date by one mental data are limited. Therefore this paper presents a method
month, a written request must be filed with the ASCE Managing Editor. based on Mindlin’s equation to determine the nonlinear spring
The manuscript for this paper was submitted for review and possible
publication on November 12, 2002; approved on August 16, 2004. This
constants of the Winkler model. Depending on the amount of pore
paper is part of the Journal of Geotechnical and Geoenvironmental pressure development, spring coefficients in the spring-dashpot
Engineering, Vol. 131, No. 12, December 1, 2005. ©ASCE, ISSN 1090- model are degraded. The effect of radiation damping is taken into
0241/2005/12-1466–1479/$25.00. account separately. This is an extension of the method developed

1466 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

 by Poulos 共1982兲 for piles subjected to lateral soil movements
under static conditions.

Numerical Model

The one-dimensional numerical model developed for the analysis

of piles founded in liquefying soil is based on the finite element
method and involves two stages. First, an effective stress based
free-field ground response analysis is carried out to determine the
external soil movement and the degradation of soil stiffness and
strength due to pore pressure generation. Second, a seismic analy-
sis of the pile is carried out by applying computed ground dis-
Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

placements dynamically to the pile. For the pile analysis, the soil Fig. 2. Beam on Winkler foundation model for dynamic pile analysis
shear modulus and ultimate lateral strength are calculated based
on the effective stress level of the soil.
Davis and Berril 共1998兲 reported that the shear wave velocity of
Ground Response Analysis the liquefied region is about 25 m / s. Ishihara and Towhata 共1982兲
also suggested that since shear stress application during earth-
The ground response analysis is carried out by dividing the soil quakes is multidirectional, even when shear stresses are reduced
deposit into a number of layers. One-half of the mass of each to zero in one direction, there will always be some shear stress
layer is concentrated at the layer boundaries as shown in Fig. 1. left in the soil. This was demonstrated in the rotational simple
The lumped masses in the layered system are connected by non- shear test performed by Ishihara and Yamasaki 共1980兲. Therefore
linear springs with a hyperbolic stress-strain relationship, which in the numerical studies performed in the following sections, a
reflects nonlinear, strain dependent and hysteretic behavior of lower limit has been set at 2% of initial effective vertical stress,
soil. Input motions are applied to the lumped mass at the bound- below which effective vertical stress is not allowed to decrease
ary between the soil deposit and the bedrock through a viscous and pore pressures are not allowed to build up.
dashpot with damping, ␳BRVBR, as shown in Fig. 1, where ␳BR and
VBR denote the density and the shear wave velocity of the bedrock
material, respectively. This viscous damping is included in the Pile Analysis
analysis to take into account the effect of energy loss due to the In the dynamic analysis of piles, moving soil interacts with the
dispersion of wave energy 共Joyner and Chen 1975兲. pile and in the vicinity of the pile, soil displacement is different
Pore pressure generation is calculated based on the effective from the displacement of the soil if there were no piles. Therefore
stress method developed by Liyanapathirana and Poulos 共2002兲. in the pile analysis, it is assumed that the displacement of the soil
During the analysis, soil stiffness and strength are degraded, de- away from the pile can be represented by the displacements ob-
pendent on the effective stress level of the soil. The time history tained from the free field ground response analysis. Soil–pile in-
of these values together with soil displacements are stored in a teraction is modeled using the analysis method for a dynamically
separate file and subsequently used to obtain the pile response in loaded beam on a nonlinear Winkler foundation, where the pile is
liquefying soil. modeled as a beam and the lateral pressure acting on the pile is
Although it is assumed that the liquefied soil does not have modeled using a spring-dashpot model with a plastic slider to
any stiffness, computationally it is difficult to obtain a stable so- limit the ultimate lateral pressure at the pile–soil interface as
lution with a near-zero shear modulus. By analyzing field data shown in Fig. 2. In this model, displacement of the soil adjacent
recorded at the Port Island site during the Kobe 1995 earthquake, to the pile wall is represented by the displacement of the plastic
slider, which is different from the displacement of the soil away
from the pile.
The partial differential equation for a beam on a Winkler foun-
dation is given by

E PI p 冉 冊 冉 冊
⳵ 4U p
⳵z 4 + Mp
⳵ 2U
⳵t2
= Kx共U f f − U p兲 + Cx 冉
⳵U f f ⳵U p
⳵t
−
⳵t
冊
共1兲
where E p = Young’s modulus of the pile material, I p = inertia of the
pile, U p = pile displacement, U f f = free field lateral soil displace-
ment, M p = mass of the pile, and Kx and Cx = spring and dashpot
coefficients of the Winkler model. A solution to the problem can
be obtained by solving Eq. 共1兲 using the finite element method.
The spring coefficients of the Winkler model are obtained by
integrating the Mindlin’s equation over a rectangular area in the
y-z plane extending from y = −d / 2 to +d / 2 and z = c2 to c1 given
by Douglas and Davis 共1964兲, where d = diameter of the pile.
Then the displacement at any point in the rectangular area is
Fig. 1. Lumped mass representation of the discretized system given by

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005 / 1467

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

 Table 1. Material Properties for Nevada Sand 共Popescu and Prevost 1993, Reprinted with Permission from Elsevier兲
Nevada sand

Property Dr = 40% Dr = 55% Dr = 80%

Mass density—solid 共kg/ m3兲 2,670.0 2,670.0 2,670.0
Porosity 0.424 0.406 0.373
Low-strain shear modulus, G0 共MPa兲 25.0 28.0 41.46
Reference mean effective normal stress 共kPa兲 100.0 100.0 100.0
Friction angle, ␾⬘ 33° 34.15° 39.5°
Permeability 共m/s兲 6.6⫻ 10−5 6.05⫻ 10−5 3.7⫻ 10−5
Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

p Py = Np · Pp 共6兲
Ux=0 = f共c1,c2,b,y,z,␯兲 共2兲
G where N p = factor ranging between 3 and 5, and P p = Rankine
where p = uniformly distributed load over the rectangular area, passive pressure given by
共c1 − c2兲d, ␯ = Poisson’s ratio of the soil, and G = shear modulus of
the soil, which varies with time along the depth of the soil deposit P p = ␴⬘v tan2共45 + ␾⬘/2兲 共7兲
due to pore pressure generation. If Uij represents the displacement where ␴⬘v = effective vertical stress and ␾⬘ = effective angle of in-
at the centroid of the ith rectangle due to a uniform pressure p j ternal friction of the soil.
distributed over the jth rectangle, then we can write The amount of radiation damping during soil yielding is still
not known but several researchers have shown that it is far less
Uij = p jFij 共3兲 than the value obtained from the elastic assumption. Nogami
共1987兲 compared numerical results obtained from his model with
where Fij is the influence coefficient determined from Eq. 共2兲.
field tests and concluded that neglecting the gaps due to soil
Since the analysis is based on the finite element method, nodal
yielding at the pile–soil interface results in an overestimation of
displacements should be calculated along the pile. The displace-
damping and an underestimation of pile deflection. Using centri-
ment at node i due to loading over the pile is given by
fuge tests, Chako 共1995兲 showed that the elastic formulation of
radiation damping is valid only during small amplitude shaking
Ui = p1Fi1 + p2Fi2 + ¯ + p jFij + ¯ + pnFin 共4兲 and is excessive when large displacements due to soil yielding
This can be written in matrix form as occurs. Therefore in the present analysis the effect of radiation
damping is neglected during soil yielding.
兵U其 = 关F兴兵p其 共5兲
where F = flexibility matrix of influence coefficients, which is cal- Comparison with Centrifuge Data
culated at y = −d / 2, 0, and +d / 2 and the average value is obtained
at the level of each pile node. The spring coefficients of the Win- For the method described in the previous section, a computer
kler model which represent the interaction between the pile and program based on the C language has been developed and the
soil are obtained by inverting 关F兴. ability of the method to simulate pile behavior in liquefying soil
When using Mindlin’s method, interaction between the spring has been demonstrated by comparing with centrifuge tests
coefficients along the pile is taken into account. Therefore the performed by Wilson et al. 共1999兲 and Abdoun et al. 共1997兲.
stiffness matrix which represents the interaction between the For the centrifuge test performed by Wilson et al. 共1999兲, the
pile and soil is a nonsymmetric matrix of size n ⫻ n, where soil profile consisted of two horizontal layers of saturated, fine,
n = number of nodes along the pile. Due to the gradual increase in uniformly graded Nevada sand. The 11.4 m thick dense lower
pore pressure level of the soil, the shear modulus of the soil layer had a relative density 共Dr兲 of 80% and the 9.1 m thick
changes with time. Therefore these spring coefficients should be medium dense upper layer had a relative density of 55%. The
calculated at the beginning of each time step during the analysis. structural model consisted of a single pile supported structure and
The Mindlin hypothesis does not automatically satisfy the soil was equivalent to a steel pipe pile with a diameter of 0.67 m and
radiation damping and this should be incorporated into the analy- wall thickness of 19 mm. The pile extended 3.8 m above the
sis separately. Here, the value of 5␳sVs proposed by Kaynia in ground surface and carried a superstructure load of 480 kN. The
1988 共Tabesh and Poulos 2000兲 is used for the dashpot coefficient depth of pile embedment was about 15 m.
where ␳s = density of the soil and Vs = shear wave velocity of the Properties of Nevada sand with Dr = 55 and 80% are given in
soil. This dashpot takes into account the radiation damping of the Table 1, and the variation of shear modulus along the depth of the
shear waves traveling away from the pile. soil deposit is given by 共Popescu and Prevost 1993兲
Time integration of Eq. 共1兲 is performed using the constant

冢 冣
0.7
average acceleration method. At each time step, the lateral pres- 1 + 2K0
sure at the soil–pile interface is monitored and an iterative proce- ␴⬘v
3
dure is used to keep it at or below the ultimate lateral presssure at Gs = G0 MPa 共8兲
p0
the pile–soil interface. When the lateral pressure at the pile–soil
interface reaches the ultimate value, soil yielding occurs. where K0 = coefficient of lateral earth pressure at rest,
For piles in sand, Broms 共1964兲 has suggested that the ultimate p0 = reference normal stress which is 100 kPa for sand 共Popescu
lateral pressure can be given by and Prevost 1993兲, and G0 = low strain shear modulus of the soil.

1468 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

 In the numerical model, dilative behavior of the soil is not
included. At present, it can model only pore pressure generation
due to cyclic loading and dissipation within the soil due to
vertical drainage. Therefore the sharp pore pressure decreases
observed during the experiment cannot be modeled using the
numerical model but the maximum pore pressure generated
during the earthquake loading can be modeled and is in general
agreement with the centrifuge test.
According to the superstructure acceleration distribution given
in Fig. 4, results obtained from the numerical model agree well
with the centrifuge test except at the point where pore pressure
decreased to a negative value due to dilation of the soil. As dis-
Fig. 3. Acceleration record of Kobe 共1995兲 earthquake scaled to cussed before, during the centrifuge test, pore pressure decrease
Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

0.22g generated for the centrifuge test by Wilson et al. 共1992兲 due to dilation has increased the soil shear modulus. As a result,
acceleration of the superstructure recorded during the centrifuge
test is higher than that obtained from the numerical model, which
This model was shaken with the acceleration record scaled to does not take into account the dilative behavior of the soil.
0.22g shown in Fig. 3, which is similar to the Kobe 1995 earth- In the design of pile foundations subjected to earthquake load-
quake but with slight modifications to the frequency content. Fig. ing, the maximum bending moment developed is the most impor-
4 shows the measured and computed pore pressure distribution tant design parameter. According to the bending moment distribu-
close to the surface of the soil deposit, the superstructure accel- tion at 2.3 m depth given in Fig. 4, the maximum pile bending
eration, and the bending moment at 2.3 m depth. The recorded moment is corresponding to the sudden increase in pore pressure
pore pressure distribution shows sharp pore pressure decreases observed in the upper layer about 3.5 s after the application of
due to soil dilation. In the acceleration record of the superstruc- earthquake loading. Since the present numerical model can simu-
ture, sharp acceleration spikes can be seen corresponding to sharp late pore pressure generation reasonably well, the maximum
pore pressure decreases. That means the pore pressure decrease bending moment obtained from the numerical model agrees well
due to soil dilation has increased the shear modulus of the soil with the maximum bending moment recorded during the centri-
and the corresponding increase in stiffness of the soil is sufficient fuge test. Therefore it can be concluded that the soil dilation does
to transmit large acceleration spikes through the ground to the not have a significant influence on the pile bending moment in
superstructure during the seismic event. liquefying soil because the maximum bending moment generated
in the pile is corresponding to the softening of the soil due to pore
pressure increase. Pore pressure decrease due to soil dilation has a
tendency to reduce the maximum pile bending moment. Softening
of the soil caused by pore pressure increase can be seen in the
bending moment and acceleration time histories as an increase in
the fundamental period of the structure.
Fig. 5 shows the bending moment and pile displacement ob-
tained about 11 s after commencement of the earthquake loading.
It can be seen that the results obtained from the numerical model
agree reasonably well with the values recorded during the centri-
fuge test despite the simplicity of the new method.
Next the centrifuge test carried out by Abdoun et al. 共1997兲
has been simulated using the numerical model presented in a
previous section. The model used for their centrifuge test con-
sisted of three sand layers. The middle layer was a 6 m thick
Nevada sand layer with relative density of 40%. Properties for
this sand are given in Table 1. The two cemented sand layers at
the top and bottom of the model were 2 m thick and pervious.
The input acceleration record generated for the test was a sine
wave with an amplitude of 0.25g and frequency of 2 Hz over a
period of 20 s. In the numerical analysis, it was assumed that the
pore pressures were generated only in the middle Nevada sand
layer.
The cemented sand used for the centrifuge test has a cohesion
of 0.65 MPa but other properties are not available. Therefore it is
assumed that the shear modulus of the cemented sand is 10 times
as high as that of the Nevada sand layer at the middle and the
density of the cemented sand is 2,500 kg/ m3.
During the centrifuge test, 11 s after the commencement of the
test, the sine wave was applied. Fig. 6 shows the pore pressure
Fig. 4. Comparison of time histories of pore pressure in upper sand ratio distributions at 7.8 and 6 m below the surface. Early lique-
layer, superstructure acceleration and bending moment at 2.3 m depth faction is observed in the numerical simulation as also found in
with centrifuge test by Wilson et al. 共1999兲 the experiment, but the pore pressure ratios agree only during the

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005 / 1469

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Displacement and bending moment along the pile 11.2 s after application of the earthquake loading

period of shaking. In the numerical model, at the end of shaking, tained from the numerical simulation matches well with the cen-
permeability of the soil was increased to a value four times higher trifuge data. However, at 24 s, which is after the end of shaking,
than that used during the shaking, as proposed by Popescu and displacements given by the numerical model are larger than those
Prevost 共1993兲, but the centrifuge test shows an even more rapid observed during the centrifuge test. If the pore pressure ratios are
rate of decrease in pore pressure after the end of shaking. A simi- compared at 6 m below the surface 24 s after shaking, numerical
lar behavior has been observed during the VELACS 共verification model and centrifuge test give pore pressure ratios of 0.9 and 0.6,
of liquefaction analysis by centrifuge studies兲 project. If the pore respectively. Therefore, 24 s after shaking, the soil behavior
pressure decrease occurs only due to pore pressure dissipation, it simulated using the numerical model shows more softening than
cannot happen at a fast rate similar to that shown in Fig. 6. The the centrifuge test. Hence the numerical model gives higher dis-
reason may be leakage of water along the walls of the laminar placements than the centrifuge model after the end of shaking, but
box and/or the instrumentation wires in the experiment. during the period of shaking, displacements obtained from the
Fig. 7 shows the lateral displacement along the depth of the numerical model agree well with the centrifuge test.
soil deposit at 4, 14, and 24 s after the commencement of appli- Figs. 8 and 9 show the pile bending moments obtained from
cation of load. At 4 and 14 s, displacement along the depth ob- the centrifuge test and the numerical model. Due to the discrep-
ancy between pore pressure distributions obtained from the nu-
merical model and the centrifuge test after the end of shaking, in
Fig. 8, bending moment distributions at different depths were
plotted only during the period of shaking. Again an excellent
agreement can be seen between the measured and computed data.
Therefore it can be concluded that the numerical model has the
ability to simulate pile behavior reasonably well in liquefying
soil.

Fig. 6. 共a兲 Comparison of pore pressure distribution at 7.8 m below

the surface with the centrifuge test by Abdoun et al. 共1997兲. 共b兲
Comparison of pore pressure distribution at 6 m below the surface Fig. 7. Comparison of lateral displacement along depth at different
with the centrifuge test by Abdoun et al. 共1997兲. times with centrifuge test by Abdoun et al. 共1997兲

1470 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Comparison of bending moment profiles with centrifuge test by Abdoun et al. 共1997兲

Parametric Study cap-mass, to simplify the analysis. When a pile carrying a

cap-mass is subjected to seismic loading, the maximum
In order to study the effects of soil liquefaction on pile perfor- bending moment generated in the pile has a contribution from the
mance, a parametric study has been carried out by varying geo- lateral displacements and curvatures imposed on a pile due to
metric and material properties of the pile and soil, and the nature ground motion, which is known as kinematic bending and the
of the earthquake. The influence of pore pressure generation on inertial forces acting on the cap-mass, which represents the
pile performance is studied by comparing results obtained from a superstructure.
total stress analysis neglecting any pore pressure effects, with In this section, the significance of the bending due to inertia
the results given by the effective stress analysis described in a forces on piles founded in liquefying soil has been studied by
previous section. comparing results obtained from a pile carrying a cap-mass and a
For all analyses presented, it is assumed that the soil follows pile without a cap-mass. In the actual situation, piles are often
the Mohr-Coulomb failure criterion, and the shear modulus of the connected to the superstructure at the pile head. Therefore it is
soil is given by assumed that the pile head is restrained against rotation but free to
move in the horizontal direction.

冢 冣
0.5
1 + 2K0
␴⬘v The cap-mass carried by a particular pile is determined based
3 on the ultimate load carrying capacity of piles in sand. The input
Gs = G0 共9兲
p0 acceleration record used for the analysis is the 1995 Kobe earth-
quake record scaled to 0.15g. The analysis has been repeated for
With a change in relative density, the shear modulus of the soil
and friction angle have been changed. Results are presented for pile lengths of 15, 25, 35, and 45 m and pile diameters of 0.3, 0.6,
relative densities of 50, 60, 70, 80, and 90%. For these relative 0.9, and 1.2 m. The concrete piles used for the analysis have a
densities, G0 in Eq. 共9兲 has been changed linearly from density of 2,400 kg/ m3 and a Young’s modulus of 3 ⫻ 104 MPa.
30 to 50 MPa and the effective friction angle of the soil has been For the piles with and without cap-mass, the same free field
changed linearly from 30 to 45°. ground displacements and degraded soil stiffness and strength
In all cases, the length of the pile is set to be equal to the depth have been applied during the pile analysis. The depths of the
of the soil deposit which overlays the bedrock. The soil has a liquefied region obtained from the ground response analysis for
density of 1,900 kg/ m3 and permeability of 10−4 m / s. It is as- each case are given in Fig. 10.
sumed that the bedrock has a shear modulus of 3,500 MN/ m2 and Fig. 11 shows the ratio of bending moments obtained for piles
a density of 2,445 kg/ m3. with and without cap-mass for different pile diameters, lengths,
and soil relative densities. Since the pile head is restrained against
rotation, in all cases, the maximum bending moment is developed
Effect of Cap-Mass on Pile Performance at the pile head. According to Fig. 11, cap-mass, which represents
Generally the superstructures supported by pile foundations are the mass of the superstructure, has increased the maximum bend-
multi-degree of freedom systems. In the design of pile founda- ing moment developed in the pile due to seismic loading. In some
tions, these multi-degree of freedom systems are commonly re- cases it is increased more than fivefold. Since the kinematic bend-
duced to a single mass at the pile head, which is termed here the ing depends on the lateral displacements and curvatures imposed

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005 / 1471

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Comparison of time histories of bending moments along pile with centrifuge data 共Abdoun et al. 1997兲

on the pile due to ground motion, it can be assumed that the same density will increase the cap-mass carried by a particular pile and
kinematic bending has been developed in piles with and without reduce the degree of soil liquefaction, enabling large accelerations
cap-mass. Therefore the increase in maximum bending moment to be transmitted through the ground to the superstructure. There-
indicates the significance of inertia force applied on piles carrying fore the significance of cap-mass increases with the increase in
a cap-mass. relative density of the soil.
The inertia force at the pile head depends on the cap-mass and According to the parametric study, when the thickness of the
the acceleration of the superstructure. The increase in relative liquefied region exceeds 30% of the thickness of the soil deposit,
the bending moment ratio is less than 1.5, irrespective of the
diameter and length of the pile. Therefore it can be concluded that
when the thickness of the liquefied region exceeds 30% of the
total thickness of the soil deposit, an approximate value for the
maximum pile bending moment induced in a pile carrying a cap-
mass can be obtained by considering only the kinematic bending
of the pile. However, for an accurate assessment of the maximum
pile bending moment, the combined effect of inertial and kine-
matic bending should be taken into account.
Results obtained from the numerical study can be validated
using the maximum bending moments recorded during the centri-
fuge test carried out by Abdoun et al. 共2003兲 for 0.475 m diam-
Fig. 10. Liquefied depth from the ground response analysis eter piles founded in two layer soil deposits. During these tests

1472 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

Fig. 11. Comparison of maximum pile bending moment with and without pile cap

the top soil layer with relative density of 40% liquefied and the Fig. 13 shows the ratio between the effective and total stress
thickness of the liquefied region is about 75% of the total thick- analyses for maximum pile bending moment, relative displace-
ness of the soil deposit. These tests have been carried out for ment, pile head acceleration, and shear force. Piles carrying a
free-head piles, where the maximum bending moment is recorded cap-mass have been used for the analysis to incorporate the su-
at the interface between the liquefied and nonliquefied soil layers. perstructure load transferred to the piles. As before, it is assumed
For piles with and without cap-mass the maximum bending mo- that the pile head is restrained against rotation but free to move in
ments recorded during the centrifuge tests are, respectively, 170 the horizontal direction.
and 113 kN m, which results in a bending moment ratio of 1.5. In all cases, inclusion of pore pressure generation has in-
creased the maximum relative displacement and the bending mo-
Effect of Pore Pressure Generation on Pile ment developed at the pile head. In many cases, the maximum
Performance pile head acceleration has decreased when pore pressure effects
are included. This happens due to softening of the soil, i.e., soft-
Although numerical methods have been developed to obtain pile
ening of the soil will increase the lateral ground displacements
performance under earthquake loading, in many of them, the ef-
imposed on the pile and it smoothes high frequency peaks and
fect of soil strength and stiffness degradation due to pore pressure
troughs in the pile head acceleration. Therefore it can be con-
generation and subsequent soil liquefaction has not been incorpo-
rated 共e.g., Kavvadas and Gazetas 1993; Norris 1994; El Naggar
and Novak 1996; and Tabesh and Poulos 2001兲. Therefore, in
many instances, pile design is based on the maximum bending
moments and shear forces calculated assuming soil as a linear
elastic material or a nonlinear material neglecting strength and
stiffness degradation due to pore pressure generation. Therefore,
in this section, results obtained from the effective stress analysis
described in a previous section have been compared with a total
stress analysis neglecting any pore pressure effects but incorpo-
rating the nonlinear behavior of the soil.
Before going into the details of pile behavior, it is important to
study the influence of pore pressure generation on ground motion,
which controls the kinematic bending of the pile. Fig. 12 shows
the ratio of maximum ground displacement, which occurs at the
ground surface, obtained from effective and total stress analyses.
It can be seen that in some cases, the maximum ground displace-
ment from the effective stress analysis is more than four times Fig. 12. Comparison of maximum ground displacement obtained
that given by the total stress analysis. from effective and total stress analyses

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005 / 1473

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Comparison of results obtained from the effective stress and total stress analyses

cluded that when pore pressure generation is taken into account Effect of Nature of Earthquake
the bending due to kinematic forces is more significant than the
The nature of the earthquake ground motions on the significance
bending due to inertia forces.
of pore pressure generation has been studied by comparing the
Although maximum pile bending moment develops at the pile ratio of maximum pile bending moments obtained from an effec-
head, maximum shear force shown in Fig. 13共d兲 does not develop tive stress and total stress analyses. Here the nature of the earth-
at the pile head in all cases. However, it develops within the top quake is characterized by the “Arias Intensity” and the natural
25% of the pile length. A close examination of Fig. 13共d兲 shows 共predominant兲 frequency of the earthquake record.
that in some cases inclusion of pore pressure generation increases The earthquake records used for the analysis are given in Table
the maximum shear force developed in the pile while in others, it 2 and all these records have been scaled to 0.1g, 0.15g, and 0.2g.
lessens the maximum shear force. Each earthquake record has been applied to a 25 m long pile with
If the maximum shear force is developed at the pile head, the 0.9 m diameter founded in soil with relative density of 50%.
inertia force at the pile head 共cap-mass⫻ pile head acceleration兲 Fig. 14 shows the variation of bending moment ratio with the
controls the shear force. Therefore the maximum shear force is Arias Intensity 共Arias 1970兲 given by
reduced when the pore pressure effects are included. For other

冕
Td
cases, kinematic forces due to large soil displacements caused by ␲
pore pressure generation and subsequent softening of the soil Arias Intensity = 共acc兲2dt 共10兲
2g 0
have a more significant influence than the inertia force at the pile
head. Therefore when pore pressure generation is taken into ac-
count, the maximum shear force developed in the pile increases. where acc= input acceleration at the level of bedrock at any time
t and Td = duration of the earthquake.
In the total stress analysis, although pore pressure generation
According to Fig. 14, there is a critical value of Arias Intensity
is not incorporated, softening of the soil will occur due to soil
and in this case it is about 0.06 m / s. When the Arias Intensity is
yielding. Therefore, when there is no liquefaction, the maximum
less than the critical value, the maximum bending moments ob-
bending moment and shear force obtained from effective and total tained from the effective and total stress analyses are nearly the
stress analyses are nearly the same. This can be seen in Figs. 13共a same. Therefore a total stress analysis can be carried out to de-
and d兲 when the relative density of the soil is 90%. Therefore, termine the maximum bending moment developed in the pile.
when there is no liquefaction, maximum bending moment and Beyond the critical value, the scatter between results is very large
shear force developed in a pile can be obtained from a total stress and bending moment ratio varies between one and four. Therefore
analysis, which takes into account the nonlinear behavior of the it can be concluded that there is no unique relationship between
soil. However, if there is liquefaction, an effective stress based the Arias Intensity and the bending moment ratio when the Arias
ground response analysis should be carried out to determine the intensity of the earthquake ground motion exceeds the critical
maximum bending moment and shear force developed in the pile. value.

1474 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

 Table 2. Arias Intensity 共0.1g兲 and Natural Frequency of Different Earthquake Records and Bending Moment Ratio Obtained from the Numerical Model
at 0.1g
Arias Intensity Natural Bending
共m/s兲 frequency moment ratio
Earthquake Maximum acceleration= 0.1g 共cycles/s兲 共effective/total兲
Kobe–1995 0.106 3.333 1.37
Taft 0.170 3.350 1.28
Loma-Prieta 0.117 3.150 1.36
Pasadena 0.243 2.067 1.74
El-Centro 0.183 3.117 1.87
Niigata 0.128 1.958 1.6
Whittier Narrows 0.055 5.792 1.03
Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

Melendy Ranche 0.016 3.650 1.05

Meckering 0.016 13.714 1.03
Cadoux 0.004 14.111 1.0
Superstition Hill 0.126 8.667 1.1
Northridge 0.108 2.140 1.63
Newcastle–1989 0.087 1.472 1.45
Newcastle–1994 0.077 2.175 1.31
Oolong 0.013 14.917 1.0
Tenant Creek 0.008 19.500 1.02
Gunjung 0.068 3.875 1.28
Saitama 0.255 1.367 3.18
New Zealand–1973 0.113 4.950 1.11
Iran 0.145 4.732 0.92
San Fernando 0.132 4.620 0.98
Tomako 0.085 7.267 0.89
Miyagi 0.163 3.360 1.56
Tangshan 0.355 3.043 1.73
Elmore Ranche 0.149 5.917 1.1

In Fig. 14, the maximum value of bending moment ratio is Fig. 15 shows the variation of bending moment ratio with the
four and it corresponds to Arias intensity of 0.5 m / s. Beyond this natural 共predominant兲 frequency of the earthquake record. Al-
value, bending moment ratio starts to decrease with increasing though the bending moment ratios have been plotted for different
Arias intensity. Although the total stress analysis does not take acceleration levels, the natural 共predominant兲 frequency of the
into account any pore pressure effects, with the increase in inten- earthquake does not change with the acceleration level of the
sity of the earthquake motion, softening of the soil due to yielding earthquake. According to Fig. 15, when the natural frequency of
occurs. Therefore an increase in bending moment is observed in the earthquake exceeds 3.5 cycles/ s, maximum bending moment
the total stress analysis. As a result, the bending moment ratio has obtained from the effective and total stress analyses are nearly the
decreased for cases with large Arias Intensity.

Fig. 15. Variation of maximum bending moment ratio obtained from

Fig. 14. Variation of maximum bending moment ratio obtained from effective and total stress analyses with the natural frequency of the
effective and total stress analyses with Arias Intensity earthquake

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005 / 1475

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

 same. Also, beyond 3.5 cycles/ s, the maximum bending moment
ratio does not depend on the maximum acceleration level of the
earthquake. It varies only with the natural 共predominant兲 fre-
quency of the earthquake. However, when the natural frequency
of the earthquake is less than 3.5 cycles/ s, the increase in maxi-
mum bending moment developed in a pile due to stiffness and
strength degradation is substantial.
From Fig. 15 and Table 2, it can be seen that when the natural
frequency is greater than 3.5 cycles/ s, in all cases the bending
moment ratio is approximately 1. However, with the Arias Inten-
sity, bending moment ratio does not show such a unique relation-
ship beyond the critical value. Although the bending moment
ratio is approximately 1 for all cases with Arias Intensity less than
Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

0.06 m / s, Elmore Ranche with Arias Intensity of 0.149 m / s also

gives a bending moment ratio of 1.1. Therefore the natural fre-
quency of an earthquake can be considered as an important pa-
rameter in characterizing the nature of an earthquake on the pile
behavior.

Soil Liquefaction in Layered Soil Deposits

In previous sections, uniform soil deposits with the same relative
density have been studied using piles carrying a cap-mass with
rotational restraint at the pile head. In all cases, the maximum
bending moment occurred at the pile head. To study the pile per-
formance in layered soil deposits, a two-layer soil deposit with a
liquefying and a nonliquefying soil layer is considered. The dense
nonliquefying soil layer has a relative density of 90% and is over- Fig. 16. Bending moment envelopes for a 0.9 m diameter pile in
lain by a soft liquefying soil layer with a relative density of 50%. nonhomogeneous soil
By varying the thickness of the liquefying soil layer, for a 25 m
deep soil deposit, the maximum positive and negative bending
moment envelopes are examined for piles with diameters 0.3 and M = 1.86共EI兲3/4共G1兲1/4␥1F共G2/G1兲 共11兲
0.9 m for three different boundary conditions at the pile head: 共1兲
where E = Young’s modulus of pile material, I = moment of inertia
free-head pile; 共2兲 fixed-head pile; and 共3兲 pile carrying a cap- of pile, G1 and G2 = shear moduli of soil on either side of the
mass with rotational restraint at the pile head. In all cases, the interface with G1 being the shear modulus of the weaker soil,
thickness of the nonliquefying layer equals the depth of soil ␥1 = free field shear strain corresponding to the weaker soil
deposit 共H兲 minus thickness of the liquefying layer 共hL兲. stratum, and F共G2 / G1兲 = function depending on G1 and G2.
Fig. 16 shows maximum positive and negative bending mo- Although Eq. 共11兲 is derived for nonliquefying soil layers with
ment envelopes for the 0.9 m diameter pile when the thickness of stiffness contrast, here it is applied for a two-layer soil deposit
the upper liquefying soil layer is 4, 8, 12, and 16 m. When the when the weaker soil layer liquefies. Table 3 summarizes the
thickness of the liquefying layer exceeds one-third of the total results obtained from Eq. 共11兲 and the finite element model de-
thickness of the soil deposit, the bending moment at the interface scribed in a previous section. The shear modulus of the liquefied
between liquefying and nonliquefying layers given by all three soil is calculated by limiting the effective stress level of the liq-
cases are the same. When the pile head is free, irrespective of the uefied soil to 2% of the initial effective overburden pressure. It
thickness of the liquefying layer, the maximum bending moment can be seen that this model overpredicts the bending moment
develops at the interface. For a fixed-head pile, maximum bend- developed at the interface when it is applied to a two-layer soil
ing moment develops at the interface if the thickness of the liq- deposit with a liquefying soil layer. By applying the Dobry–
uefying layer is greater than one-third and less than two-thirds of O’Rourke 共1983兲 model for nonliquefying soil layers with stiff-
the total thickness of the soil deposit. When the thickness of the ness contrast, Nikolaou et al. 共2001兲 also showed that it overpre-
liquefying layer lies outside this range, the maximum bending dicts bending moment at the interface.
moment develops at the pile head. When the pile carries a cap- Since the Dobry–O’Rourke 共1983兲 model does not take into
mass, which is based on the ultimate load carrying capacity of the account the thickness of soil layers, natural frequency of the
pile, and with rotational restraint at the pile head, the maximum earthquake, and number of cycles of the earthquake, Nikolaou et
bending moment always develops at the pile head. al. 共2001兲 proposed another model by taking into account those
Dobry and O’Rourke 共1983兲 and Nikolaou et al. 共2001兲 pro- factors, as shown below
posed simple models to obtain kinematic pile bending moments,
which neglects the contribution of inertia forces from the super-
structure towards pile bending moments, at the interface between
M = 0.042␶cd3␩ 冉冊 冉 冊 冉 冊
L
d
0.30
Ep
E1
0.65
V2
V1
0.50
共12兲

soil layers with stiffness contrast. Dobry and O’Rourke 共1983兲 where d = pile diameter, L = pile length, E1 and V1 = respectively,
modeled the pile as a beam on a Winkler foundation and derived Young’s modulus and shear wave velocity of the upper soil layer,
an expression for the pile bending moment at the interface as V2 = shear wave velocity of the lower soil layer, ␩ = reduction

1476 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

 Table 3. Comparison of Numerical Model with Simple Design Methods Available for Layered Soil Deposits with Stiffness Contrast
Maximum BM from nonlinear
effective stress based
dynamic FE analysis Pile bending moment at the interface from
共MN m兲 simple design methods 共MN m兲

Nikolaou et al.
共2001兲
Free- Fixed- Pile Dobry and
D hL head head with O’Rourke Frequency Time Abdoun et al.
共m兲 共m兲 pile pile cap 共1983兲 domain domain 共2003兲
0.3 4.0 Head 0.101 0.209
Interface 0.133 0.103 0.137 0.363 0.045 0.028 0.025
Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

8.0 Head 0.115 0.205

Interface 0.092 0.106 0.092 0.195 0.111 0.070 0.1
12.0 Head 0.127 0.226
Interface 0.157 0.157 0.157 0.159 0.085 0.054 0.222
16.0 Head 0.117 0.208
Interface 0.081 0.081 0.081 0.155 0.079 0.050 0.4
0.9 4.0 Head 0.930 2.810
Interface 1.150 0.319 1.180 9.793 0.876 0.552 0.076
8.0 Head 2.330 3.780
Interface 2.460 2.320 2.930 5.253 2.163 1.363 0.297
12.0 Head 2.370 3.530
Interface 2.540 2.910 3.070 4.291 1.643 1.035 0.667
16.0 Head 2.920 3.890
Interface 2.110 2.110 2.110 4.202 1.552 0.978 1.187

factor applied to the maximum steady-state pile bending moment liquefying soil subjected to lateral spreading based on the bending
in the frequency domain to arrive at the corresponding peak value moments recorded during centrifuge tests carried out for concrete
in the time domain, and ␶c = shear stress that is likely to develop at piles in two- and three-layer soil deposits. Tests were carried out
the interface. for free-head piles with and without cap-mass and maximum
The Kobe earthquake record used for the analysis is similar to bending moment was developed at the boundary between the liq-
the earthquake record given in Fig. 3. For the four soil deposits uefied and nonliquefied layers. By calibrating centrifuge test data,
considered for the analysis given in Table 3, the natural periods they found that the lateral pressure applied over a unit area of pile
vary from about 1.05 to 0.89 s and lie within the range of the by the liquefying soil is a constant over the depth. This constant is
predominant periods of the excitation. According to Fig. 3, the given as 10.3 kPa. If the depth of the liquefied region is hL in a
Kobe earthquake record has about 10 effective excitation cycles two-layer soil deposit with liquefied and nonliquefied layers, the
共Nc兲. Therefore the frequency-to-time conversion factor ␩ is maximum bending moment at the interface is given by

冉 冊
approximately
hL2 D
M max = 10.3 + A ch c 共14兲
␩ ⬇ 0.04 ⫻ Nc + 0.23 = 0.63 共13兲 2
Bending moments at the interface computed from the where D = diameter of the pile, Ac = area of pile cap subjected to
Nikolaou et al. 共2001兲 model are given in Table 3. It can be seen the pressure of the liquefied soil, and hc = distance from the
that for some cases the bending moment at the interface calcu- boundary between the liquefied and nonliquefied regions and the
lated in the frequency domain, which gives the harmonic steady- center of the area of pile cap in touch with the liquefied soil. They
state results, agrees with the results obtained from the numerical concluded that the maximum bending moment predicted with this
model described in a previous section, but for others, it underpre- value is within 15% of the maximum bending moment recorded
dicts the bending moment at the interface. When the harmonic during the centrifuge tests.
steady-state bending moments are converted to the time domain, Table 3 shows the maximum bending moment calculated from
for all cases, the Nikolaou et al. 共2001兲 model underpredicts the this method when the top and bottom layers of the soil deposit
bending moment at the interface. Although Eq. 共12兲 was derived have relative densities of 50 and 90%, respectively. When piles
based on the results obtained from a beam on a Winkler founda- carry a cap-mass, it is assumed that the pile cap is not in contact
tion model, they performed dynamic finite element analyses ne- with the liquefied soil, i.e., Ac in Eq. 共14兲 is zero. The soil layer at
glecting the nonlinear behavior of soil and the pore pressure gen- the top has been completely liquefied for all cases listed. The limit
eration due to earthquake loading. Therefore it can be concluded equilibrium method proposed by Abdoun et al. 共2003兲 is based on
that when the nonlinearity of the soil and pore pressure generation the results obtained from centrifuge tests carried out for piles with
are significant, the bending moments are underpredicted by the diameter 0.27 and 0.35 m and when the thickness of the liquefied
Nikolaou et al. 共2001兲 model. soil layer is 6 m. When comparing the results obtained from the
Recently Abdoun et al. 共2003兲 proposed a method to evaluate numerical model and the limit equilibrium method for different
the maximum bending moment developed in piles founded in pile diameters, D, and depths of liquefied soil, hL, only the case

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005 / 1477

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

 with D and hL closest to the centrifuge test, i.e., D = 0.3 m and soil deposit, bending moment at the interface does not depend on
hL = 8 m, agrees well with the limit equilibrium method. For the the boundary condition at the pile head.
other cases the maximum bending moment at the interface ob- Simple models available to obtain kinematic pile bending mo-
tained from the present numerical model and the limit equilibrium ments at the interface between two layers with stiffness contrast
method do not agree well. For example, when D = 0.9 m and have been applied to the case of a nonhomogeneous soil deposit
hL = 12 m, the maximum bending moment at the interface ob- with a liquefying soil layer underlain by a nonliquefying soil
tained from the numerical model is five times larger than that layer. It is found that when the nonlinear behavior of soil and pore
obtained from the Abdoun et al. 共2003兲 limit equilibrium method. pressure generation are taken into account, these simple models
cannot be applied to obtain consistently accurate values of the
maximum bending moment at the interface.
Conclusions
Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

A numerical procedure has been described which can be used to Acknowledgments

simulate pile performance in liquefying soil. First an effective
stress based ground response analysis is carried out. Second the This work is part of a project on “Design of Pile Foundations for
resulting ground displacements and degraded soil stiffness are Seismically Active Areas” funded by the Australian Research
applied to the pile dynamically to obtain the pile performance in Council and this support is gratefully acknowledged. Also the
liquefying soil. The spring coefficients of the Winkler model are writers would like to thank Dr. Daniel W. Wilson for providing
derived from the Mindlin’s equation. Results obtained from two data files of the centrifuge test and the reviewers for their valuable
centrifuge tests are simulated from the new method and the close comments.
agreement between results given by the numerical model and the
data recorded during the centrifuge tests demonstrate the ability
of the new method to simulate pile behavior in liquefying soil. References
Results are also presented from a parametric study by varying
geometric and material properties of the pile and soil, and the Abdoun, T., Dobry, R., and O’Rouke, T. D. 共1997兲. “Centrifuge and
nature of the earthquake. It is shown that the inclusion of a cap- numerical modelling of soil-pile interaction during earthquake
mass can significantly increase the bending moment developed in induced soil liquefaction and lateral spreading.” Observation and
a pile by increasing the inertia forces acting on the pile. The Modelling in Numerical Analysis and Model Tests in Dynamic Soil-
inclusion of pore pressure effects increases the bending moments Structure Interaction Problems—Proc., Sessions held in Conjunction
and relative displacements of the pile but it reduces the maximum with Geo-Logan ’97, Logan, Utah, 76–90.
Abdoun, T., Dobry, R., O’Rouke, T. D., and Goh, S. H. 共2003兲. “Pile
cap-mass acceleration. In some cases, the peak shear force devel-
response to lateral spreads: Centrifuge modelling.” J. Geotech.
oped in the pile increases due to the inclusion of pore pressure Geoenviron. Eng., 129共10兲, 869–878.
effects. In others, the inclusion of pore pressure reduces the peak Arias, A. 共1970兲. “A measure of earthquake intensity.” Seismic design for
shear force developed in the pile, due to the dependency of shear nuclear power plants, R. J. Hansen, ed., MIT Press, Cambridge, Mass.
force on the cap-mass acceleration. Broms, B. B. 共1964兲. “Lateral resistance of piles in cohesionless soils.”
It is found that the Arias intensity and the natural frequency of J. Soil Mech. Found. Div., 90共3兲, 123–156.
an earthquake are important parameters in characterizing the in- Chako, M. J. 共1995兲. “Analysis of dynamic soil-pile structure interac-
fluence of the nature of an earthquake on pile behavior. If the tion.” masters thesis, Department of Civil and Environmental
Arias intensity of the earthquake ground motion is less than Engineering, University of California, Davis.
0.06 m / s, maximum bending moments obtained from total and Davis, R. O., and Berrill, J. B. 共1998兲. “Energy dissipation and liquefac-
effective stress analyses are nearly the same. However, beyond tion at Port Island, Kobe.” Bull. New Zealand Nat. Soc. Earthquake
Eng., 31, 31–50.
this critical value, the scatter between results is very large. If the
Dobry, R., and O’Rourke, M. J. 共1983兲. “Discussion on ‘Seismic
natural frequency of the earthquake is greater than 3.5 cycles/ s, a response of end-bearing piles,’ by R. Flores-Berrones and
total stress analysis can be carried out to determine the maximum R. V. Whitman.” J. Geotech. Eng., 109共5兲, 778–781.
bending moment developed in the pile. If the natural frequency of Dobry, R., Taboada, V., and Liu, L. 共1995兲. “Centrifuge modelling of
the earthquake is less than 3.5 cycles/ s, the maximum bending liquefaction effects during earthquakes.” Proc., First Conf. on Earth-
moment given by an effective stress analysis can be as high as quake Geotechnical Engineering, Tokyo, 1291–1324.
four times that given by a total stress analysis. Douglas, D. J., and Davis, E. H. 共1964兲. “The movement of buried foot-
If the soil deposit is nonhomogeneous and the liquefying soil ings due to moment and horizontal load and the movement of anchor
layer is underlain by a nonliquefying soil layer, for free-head plates.” Geotechnique, 14, 115–132.
piles, irrespective of the thickness of the liquefying soil layer, the El Naggar, M. H., and Novak, M. 共1996兲. “Non-linear analysis for
maximum bending moment is developed at the interface between dynamic lateral pile response.” Soil Dyn. Earthquake Eng., 15共4兲,
233–244.
two layers. When a pile carries a cap-mass, which is based on the
Finn, W. D. L., Thavaraj, T., and Fujita, N. 共2001兲. “Piles in liquefiable
ultimate load carrying capacity of the pile, and restrained against
soils: Seismic analysis and design issues.” Proc., 10th Int. Conf. on
rotational motion at the pile head, the maximum bending moment
Soil Dynamics and Earthquake Engineering, Philadelphia, 48.
always develops at the pile head. For a fixed-head pile, the maxi- Hamada, M., Sato, H., and Nakamura, T. 共1994兲. “An experimental and
mum bending moment develops at the interface if the thickness of numerical study on liquefaction-induced ground displacement.” Proc.,
the liquefying layer is greater than one-third, and less than two- 5th US National Conf. on Earthquake Engineering, 4, 169–178.
thirds, of the total thickness of the soil deposit. When the thick- Horikoshi, K., Tateishi, A., and Fujiwara, T. 共1998兲. “Centrifuge model-
ness of the liquefying layer lies outside this range, the maximum ling of a single pile subjected to liquefaction-induced lateral spread-
bending moment develops at the pile head. When the thickness of ing.” Soils Found., 共Special Issue No. 2兲, 193–208.
the liquefying layer exceeds one-third of the total thickness of the Ishihara, K., and Towhata, I. 共1982兲. “Dynamic response analysis of level

1478 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479

 ground based on the effective stress method.” Soil mechanics— Eng., 12, 73–90.
transient and cyclic loads, G. N. Pande and O. C. Zienkiewicz, eds., Poulos, H. G. 共1982兲. “Developments in the analysis of static and cyclic
Wiley, New York, 133–172. lateral response of piles.” Proc., 4th Int. Conf. on Numerical Methods
Ishihara, K., and Yamasaki, F. 共1980兲. “Cyclic simple shear tests on satu- in Geomechanics, Canada, 1117–1135.
rated sand in multi-directional loading.” Soils Found., 20共1兲, 45–59. Sakajo, S., Chai, J. C., Nakajima, K., and Maeda, M. 共1995兲. “Effect of
Joyner, W. B., and Chen, A. T. F. 共1975兲. “Calculation of nonlinear group pile on liquefaction resistance of sandy ground.” Proc., First
ground response in earthquakes.” Bull. Seismol. Soc. Am., 65共5兲,
Conf. on Earthquake Geotechnical Engineering, Tokyo, 755–760.
1315–1336.
Shahrour, I., and Ousta, R. 共1998兲. “Numerical analysis of the behaviour
Kavvadas, M., and Gazetas, G. 共1993兲. “Kinematic seismic response and
bending of free-head piles in layered soils.” Geotechnique, 43共2兲, of piles in saturated soils under seismic loading.” Proc., 11th Euro-
207–222. pean Conf. on Earthquake Engineering.
Liyanapathirana, D. S., and Poulos, H. G. 共2002兲. “A numerical model for Tabesh, A., and Poulos, H. G. 共2000兲. “A simple method for the seismic
dynamic soil liquefaction analysis.” Soil Dyn. Earthquake Eng., 22, analysis of piles and its comparison with the results of centrifuge
1007–1015. tests.” Proc., 12th World Conf. on Earthquake Engineering, Auckland,
Downloaded from ascelibrary.org by Kyoto U Kogaku Chikyu on 05/06/25. Copyright ASCE. For personal use only; all rights reserved.

Mizuno, H., Sugimoto, M., Mori, T., Iiba, M., and Hirade, T. 共2000兲. New Zealand, Paper No. 1203.
“Dynamic behaviour of pile foundation in liquefaction process— Tabesh, A., and Poulos, H. G. 共2001兲. “The effect of soil yielding on
Shaking table tests utilising big shear box.” Proc., 12th World Conf. seismic response of single piles.” Soils Found., 41共3兲, 1–16.
on Earthquake Engineering, Auckland, New Zealand, Paper No. Tamura, S., Suzuki, Y., Tsuchiya, T., Fujii, S., and Kagawa, T. 共2000兲.
1883. “Dynamic response and failure mechanisms of a pile foundation dur-
Nakamura, T., Sugano, T., Oikawa, K., and Mito, M. 共2000兲. “An ing soil liquefaction by shaking table test with a large scale laminar
experimental study on the pier damaged by 1995 Hyogoken-Nanbu shear box.” Proc., 12th World Conf. on Earthquake Engineering,
earthquake.” Proc., 12th World Conf. on Earthquake Engineering,
Auckland, New Zealand, Paper No. 0903.
Auckland, New Zealand, Paper No. 1038. Wilson, D. W., Boulanger, R. W., and Kutter, B. L. 共1999兲. “Lateral
Nikolaou, S., Mylonakis, G., Gazetas, G., and Tazoh, T. 共2001兲.
resistance of piles in liquefying sand.” Geotechnical Special Publica-
“Kinematic pile bending during earthquakes: Analysis and field
measurements.” Geotechnique, 51共5兲, 425–440. tion No. 88, Proc., Analysis, Design, Construction and Testing of
Nogami, T. 共1987兲. “Prediction of dynamic response of nonlinear single Deep Foundations, ASCE, Reston, Va., 165–179.
pile by using Winkler soil model.” J. Geotech. Eng., 114共9兲, Wilson, D. W., Boulanger, R. W., and Kutter, B. L. 共2000兲. “Observed
1512–1525. seismic lateral resistance of liquefying sand.” J. Geotech. Geoenviron.
Norris, G. M. 共1994兲. “Seismic bridge pile foundation behaviour.” Proc., Eng., 126共10兲, 898–906.
Int. Conf. on Design and Construction of Deep Foundations, 1, Yasuda, S., Ishihara, K., Morimoto, I., Orense, R., Ikeda, M., and
27–136. Tamura, S. 共2000兲. “Large-scale shaking table tests on pile founda-
Ohtomo, K. 共1996兲. “Effects of liquefaction induced lateral flow on a tions in liquefied ground.” Proc., 12th World Conf. on Earthquake
conduit with supporting piles.” Proc., 11th World Conf. on Earth- Engineering, Auckland, New Zealand, Paper No. 1474.
quake Engineering, Paper No. 386. Zheng, J., Susuki, K., and Ohbo, N. 共1996兲. “Evaluation of sheet pile-ring
Popescu, R., and Prevost, J. H. 共1993兲. “Centrifuge validation of a nu- countermeasure against liquefaction for oil tank site.” Soil Dyn.
merical model for dynamic soil liquefaction.” Soil Dyn. Earthquake Earthquake Eng., 15共6兲, 369–379.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / DECEMBER 2005 / 1479

J. Geotech. Geoenviron. Eng., 2005, 131(12): 1466-1479