HOSTED BY Available online at www.sciencedirect.

com

ScienceDirect
Soils and Foundations 57 (2017) 92–110
www.elsevier.com/locate/sandf

Analysis of laterally loaded short and long piles in

multilayered heterogeneous elastic soil
Bipin K. Gupta, Dipanjan Basu ⇑
Dept. of Civil and Environmental Engineering, Univ. of Waterloo, Waterloo, ON N2L 3G1, Canada

Received 5 November 2015; received in revised form 7 August 2016; accepted 28 October 2016
Available online 10 February 2017

Abstract

A continuum-based method is developed for the analysis of laterally loaded piles in multilayered, heterogeneous elastic soil. The anal-
ysis considers the soil as a layered elastic continuum in which the modulus varies linearly or non-linearly with depth within each layer.
Rational soil displacement fields are assumed and differential equations describing the pile and soil displacements are obtained using the
principle of minimum potential energy. The differential equations describing the pile and soil displacements are solved using the Ritz
method and the finite difference method, respectively, following an iterative numerical scheme. The analysis is used to study different
pile geometries embedded in layered soil deposits with heterogeneity in each layer. The pile displacement, rotation, and maximum bend-
ing moment obtained from the analysis were found to be in good agreement with those obtained from an equivalent three-dimensional
finite element analysis and from other studies available in the literature. The analysis can be used to obtain the pile head displacement,
rotation, and maximum bending moment that can then be used in design.
Ó 2017 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Pile; Elastic analysis; Lateral load; Minimum potential energy; Ritz method

1. Introduction for pile displacements because it does not rigorously cap-

The analysis of laterally loaded piles is generally done
using the p-y method (Matlock, 1970; Reese et al., 1974,
1975; Cox et al., 1974; Briaud et al., 1984; Gabr et al.,
1994; Ashour and Norris, 2000). In this method, each pile
is assumed to behave as a beam supported by a series of lat-
eral soil springs, and the lateral load displacement response
of the pile is captured by characterizing the soil springs
with p-y curves at different soil depths (p is the horizontal
soil reaction acting on the pile per unit length and y is
the corresponding lateral pile displacement). However,
the p-y method sometimes fails to give accurate results
Peer review under responsibility of The Japanese Geotechnical Society.
⇑ Corresponding author.
E-mail addresses: bkgupta@uwaterloo.ca (B.K. Gupta), dipanjan.
basu@uwaterloo.ca (D. Basu).
ture the three-dimensional pile-soil interaction necessary
to predict the lateral pile behavior (Yan and Byrne, 1992;
Anderson et al., 2003; Kim et al., 2004; Higgins et al.,
2010; Haldar and Basu, 2014). Consequently, research on
laterally loaded piles has continued to be unabated and sev-
eral studies have been conducted in which the soil is mod-
elled as a continuum. In these studies, different solution
methods, such as (i) the finite difference (FD) method
(Poulos, 1971a,b; Verruijt and Kooijman, 1989; Zhang
et al., 2000), (ii) the finite element (FE) method
(Randolph, 1981; Trochanis et al., 1991; Filho et al.,
2005; Higgins et al., 2013), (iii) the boundary element
(BE) method (Banerjee and Davies, 1978; Budhu and
Davies, 1988; Ai et al., 2013), and (iv) the variational
method (Sun, 1994; Shen and Teh, 2002; Yang and
Liang, 2006; Basu et al., 2009; Salgado et al., 2014), have
been used to obtain solutions for laterally loaded piles

http://dx.doi.org/10.1016/j.sandf.2017.01.007
0038-0806/Ó 2017 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110
embedded in continua. A few of these studies account for
the effect of multiple soil layers (Randolph, 1981; Chow,
1987; Verruijit and Kooijman, 1989; Basu et al., 2009; Ai
et al., 2013; Salgado et al., 2014) and soil heterogeneity,
such as linear or nonlinear variations of the soil modulus
with depth (Poulos, 1973; Randolph, 1981; Banerjee &
Davies, 1978; Budhu and Davies, 1988; Zhang et al., 2000).
In actual field conditions, soils are stratified with layers
of sandy, silty, and clayey deposits; and thus, it is necessary
to consider the multiple discrete soil layers in the analyses.
However, the properties within each soil layer often vary
with depth. For example, in overconsolidated clay layers,
a parabolic variation in the soil modulus with depth is
often assumed (Scott, 1981). In the case of normally con-
solidated clays or sandy deposits, a linear variation in the
soil modulus with depth is often assumed (Reddy and
Valsangkar, 1971). Due to the wide variability of soil types
and the heterogeneity of the soil properties within each
layer, an analysis method needs to be developed for later-
ally loaded piles in which multiple soil layers are considered
and variations in the soil modulus with depth within each
layer can be easily taken into account.
In this paper, a method is developed for the analysis of
laterally loaded piles in multilayered elastic soil with the
soil modulus varying with depth within each layer. The
analysis assumes each pile to behave as an Euler-
Bernoulli beam supported by a three-dimensional (3-D)
soil continuum surrounding it. The behavior of the 3-D soil
continuum is simplified by assuming the soil displacements
as products of separable variables. Using the principle of
minimum potential energy and variational calculus, the
problem is posed as a beam on an elastic foundation for
which the foundation parameters are rigorously related
to the elastic constants of the soil continuum. The advan-
tage of this method is that the simplicity of the beam-on-
foundation approach is maintained, yet the rigor of the
continuum approach is incorporated. The developed differ-
ential equations for pile and soil displacements are solved
analytically and numerically in an iterative numerical
scheme. The time required for the analysis is significantly
less than that required by the 3-D numerical analysis
because the 3-D problem is represented by a set of one-
dimensional differential equations that can be solved
quickly.
2. Analysis
2.1. Problem definition
A pile with a circular cross-section having a radius rp,
length Lp, Young’s modulus Ep, and moment of inertia
Ip, embedded in a soil deposit with n + 1 layers, is consid-
ered (Fig. 1). The thickness of any generic soil layer (except
for the bottom layer) is Hi
Hi
1, where Hi is the depth to
the bottom of the ith layer from the ground surface (with
H0 = 0), and the bottom (n + 1)th layer has infinite thick-
ness (i.e., Hn+1 = 1). All the layers have infinite horizontal
93
extents and follow the laws of linear elasticity with Lame’s
constants ksi and Gsi (subscript i represents the ith layer).
The shear modulus Gsi within each layer varies with depth,
while the Poisson’s ratio msi within each layer remains spa-
tially constant. Thus, ksi [=2msiGsi/(1
2msi)] also varies
with depth within each layer. Mathematically, the variation
in Gsi with depth z within the ith layer is given by
xi
Gsi ¼ f i Gs0 þ si ðz
H i
1 Þ ð1Þ
where Gs0 is a reference shear modulus (=100 MPa), fi is a
scalar coefficient such that fiGs0 gives the value of the shear
modulus at the top of the ith layer (i.e., at z = Hi
1), si is a
constant for the ith layer that has the physical meaning of
the rate of change in shear modulus with depth if Gsi varies
linearly with depth, xi is an exponent that determines how
Gsi varies with depth, and z is the depth measured from the
ground surface. For physical problems, the typical range in
exponent x is 0.5–2 (Scott, 1981), where x = 0.5 represents
a parabolic variation (i.e., the slope of Gs versus depth
curve decreases with depth), x = 1 represents a linear vari-
ation (i.e., the slope of Gs versus depth curve remains con-
stant with depth), and x = 2 represents a hyperbolic
variation (i.e., the slope of Gs versus depth curve increases
with depth) in the modulus with depth. For simplicity, the
variation in shear modulus Gs with depth below the pile
base (i.e., in the (n + 1)th layer) is always assumed to be lin-
ear (with x = 1). This is because it was found in several
simulations that the actual variation (linear or nonlinear)
in Gs in the (n + 1)th layer does not affect the pile response
to any significant extent. Note that, for any layer, a con-
stant shear modulus with depth can also be assumed in
the analysis by setting si = 0.
The pile head is at the ground surface and subjected to
horizontal force Fa and moment Ma, as shown in Fig. 1.
The pile base rests on top of the (n + 1)th (bottom) layer
(i.e., Lp = Hn). No slippage or separation between the pile
and the surrounding soil or between the soil layers is
allowed. For analysis, a right-handed cylindrical (r-h-z)
coordinate system is chosen such that its origin lies at the
centre of the pile head, the z axis coincides with the pile axis
and points downward, the reference radial direction r0
coincides with the direction of applied force Fa, and angu-
lar distance h, measured from r0, is clockwise positive when
looking downward from the top of the pile.
2.2. Soil displacement, strain, and strain energy density
The horizontal soil displacement field, generated by the
pile displacement, is described as the product of three func-
tions each of which varies with one of the three dimensions.
The vertical soil displacement, uz, is assumed to be negligi-
ble. Mathematically, the radial (ur) and the tangential (uh)
soil displacements are assumed as follows (Basu, 2006):
ur ¼ wðzÞ/r ðrÞ cos h ð2aÞ
uh ¼
wðzÞ/h ðrÞ sin h ð2bÞ
94 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110

Fig. 1. Schematic of laterally loaded pile in multilayered elastic soil deposit.

where w(z) is the pile displacement, /r and /h are dimen- Further, using the elastic stress-strain relationship, the
sionless functions of the radial coordinate that are both strain energy density rpqepq/2 (rpq and epq are soil stress
assumed to be equal to one at r = rp, which ensures perfect and strain tensors and the summation is implied by the rep-
contact between the pile and the soil, and both assumed to etition of indices p and q) of soil is given by
be equal to zero at r = 1, which ensures that soil displace- "  2
ments due to pile movement decrease with an increase in 1 2 d/r d/ ð/
/h Þ
UD ¼ ðks þ 2Gs Þw þ 2ks w2 r r
radial distance from the pile and eventually become zero. 2 dr dr r
The sine and cosine functions ensure that the variation in 2   2
ð/
/ Þ d/h
the soil displacements in the tangential direction is compat- þðks þ 3Gs Þw2 r 2 h þ Gs w2
r dr
ible with the horizontal pile displacement.  2  2 #
Using the soil displacement field, the infinitesimal soil ð/
/ Þ d/ dw dw
þ2Gs w2 r h h
þ Gs /r þ Gs
2
/2h
strains (with contractive strains assumed as positive) can r dr dz dz
be expressed as
ð4Þ
2 3
2 3
@u r
err @r
6 7 2.3. Principle of minimum potential energy
6e 7 6
urr
1r @u h 7
6 hh 7 6 @h 7
6 7 6 7
6 ezz 7 6
@u z
7 The total potential energy of the pile-soil system is given
6 7¼6 @z 7
6 e 7 6
1 1 @ur þ @uh
uh  7 by
6 rh 7 6 2 r @h @r r 7
6 7 6 7
4 erz 5 6
1 @uz þ @ur  7 Z Lp  2 2 Z 
dw
4 2 @r @z 5 1 d w
  P ¼ Ep I p dz þ U D dX
F a wjz¼0 þ M a 
ehz
12 1r @u z
þ @uh 2 0 dz2 X0 dz z¼0
@h @z
2 3 ð3Þ ð5Þ

wðzÞ d/drr ðrÞ cos h
6 7
6
wðzÞ /r ðrÞ
/ h ðrÞ
cos h 7 where X0 is the appropriate soil domain that participates in
6 r 7
6 7 the pile-soil interaction. A mechanical system in equilib-
6 n 0 o 7
6
¼ 61 7 rium has minimum potential energy. Therefore, minimizing
wðzÞ /r ðrÞ
/h ðrÞ
þ d/h ðrÞ
sin h 7
62 r dr 7 the potential energy of the pile-soil system (i.e., setting the
6 7
6
2 dz /r ðrÞ cos h
1 dwðzÞ 7 first variation in potential energy dP equal to 0) will lead to
4 5
the required differential equations governing the equilib-
1 dwðzÞ
2 dz
/h ðrÞ sin h rium of the pile-soil system. Considering a layered system,
 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110 95

setting dP = 0 in Eq. (5) after substituting UD from Eq. (4)

results in

(   "  2
n Z
X Xn Z Hi Z 1
Hi
d 2 wi d 2 wi d/r / d/r
Ep I p 2 d 2
dz þ p ðk si þ 2G si Þwi dwi þ 2ksi wi dwi r
i¼1 H i
1 dz dz i¼1 H i
1 rp dr r dr
/h d/r /2 /2

2ksi wi dwi þ ðksi þ 3Gsi Þwi dwi 2r þ ðksi þ 3Gsi Þwi dwi 2h
r dr r r
 2
/r /h d/h /r d/h

2ðksi þ 3Gsi Þwi dwi 2 þ Gsi wi dwi þ 2Gsi wi dwi
r dr r dr
    
/h d/h dwi dwi 2 dwi dwi 2

2Gsi wi dwi þ Gsi d /r þ Gsi d /h rdrdz
r dr dz dz dz dz
Z 1    
dwi dwi dw1 
þ pr2p Gsðnþ1Þ d dz
F a dw1 jz¼0 þ M a d
Lp dz dz dz z¼0
(
nþ1 Z H i Z 1 
X      
d/r d/r 1 d/ / d/r
þ p ðksi þ 2Gsi Þw2i d þ ksi w2i d/r r þ ksi w2i r d ð6Þ
i¼1 H i
1 rp dr dr r dr r dr
 
/ d/r / /

ksi w2i h d þ ðksi þ 3Gsi Þw2i 2r d/r
ðksi þ 3Gsi Þw2i d/r 2h
r dr r r
 2 # )
1 d/ dwi
þGsi w2i d/r h þ Gsi /r d/r rdrdz
r dr dz
(
nþ1 Z H i Z 1
X 1 d/ / /
þ p
ksi w2i d/h r þ ðksi þ 3Gsi Þw2i 2h d/h
ðksi þ 3Gsi Þw2i 2r d/h
i¼1 H i
1 rp r dr r r
   
d/h d/h / d/h 1 d/
þ Gsi w2i d þ Gsi w2i r d
Gsi w2i d/h h
dr dr r dr r dr
   2 )# )
/ d/h dwi

Gsi w2i h d þ Gsi /h d/h rdrdz ¼ 0
r dr dz

n Z
X Hi  Z 1
Since the variations dw, d/r, and d/h in Eq. (6) are inde-
pendent, the terms associated with them must individually ms2 ¼ Gsi w2i dz þ Gsðnþ1Þ w2nþ1 dz ð8bÞ
i¼1 H i
1 Lp
be equal to zero in order to satisfy Eq. (6).
n Z
X Hi  Z 1
ms3 ¼ ksi w2i dz þ ksðnþ1Þ w2nþ1 dz ð8cÞ
2.4. Equations describing /r and /h i¼1 H i
1 Lp

n Z
X Hi  Z 1
Considering the variation in /r in Eq. (6) and equating ms4 ¼ ðksi þ 3Gsi Þw2i dz þ ðksðnþ1Þ
the terms associated with d/r results in i¼1 H i
1 Lp
Z   2 
1
d / d/ d/ / þ 3Gsðnþ1Þ Þw2nþ1 dz ð8dÞ

ms1 r 2 r þ r þ ðms2 þ ms3 Þ h þ ms4 r
rp dr dr dr r "Z  2 #
   Xn Hi
dwi
/h d/r ns ¼

ms4 þ ns r/r d/r þ ms1 r þ ms3 /r
ms3 /h d/r j1
rp ¼ 0
Gsi
dz
dz
r dr i¼1 H i
1

ð7Þ Z 1  2
dwnþ1
þ Gsðnþ1Þ dz ð8eÞ
where coefficients ms1, ms2, ms3, ms4, and ns are given by Lp dz

Xn Z H i  Z 1
The last term on the left-hand side of Eq. (7) becomes
ms1 ¼ ðksi þ 2Gsi Þwi dz þ
2
ðksðnþ1Þ zero because d/r = 0 at r = rp and r = 1 as function /r
i¼1 H i
1 Lp
is known at both points (/r = 0 at r = 1 and /r = 1 at
þ 2Gsðnþ1Þ Þw2nþ1 dz ð8aÞ r = rp). For the first term on the left-hand side of Eq. (7),
96 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the integrand must be zero because d/r – 0 over the inter- ms2 þ ms3
c6 ¼
val rp < r < 1 as it is not known a priori. This gives the ms2
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
governing differential equation for /r over rp < r < 1 as uPn
R
R
u Hi ksðnþ1Þ 1
u i¼1 Gsi þ 1 H i
1 Gsi wi dz þ Gsðnþ1Þ þ 1 Lp Gsðnþ1Þ wnþ1 dz
ksi 2 2
" #
d 2 /r 1 d/r
c 2 c 2 c2 d/h
c1 2 ¼t Pn R H i R1
i¼1 H i
1 Gsi wi dz þ Lp Gsðnþ1Þ wnþ1 dz
2 2
2
þ
1
þ 2
/r ¼ 3
/h
dr r dr r rp r dr r
ð12cÞ
ð9Þ
2.5. Solutions for /r and /h
where the dimensionless constants c1, c2, and c3 are given
by
rffiffiffiffiffiffiffi
Differential Eqs. (9) and (11) for /r and /h are interde-
ms4 pendent and are solved simultaneously using a one-
c1 ¼
ms1 dimensional finite difference scheme. The discretized forms
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
uPn
ksi R
R
1 of the differential equations are given by
u i¼1 þ 3
Hi 2
þ
ksðnþ1Þ
þ 3 2
u Gsi H
G si w i dz G Lp
G sðnþ1Þ w nþ1 dz "   2 #
¼ tP
 R i
1
Hi

sðnþ1Þ
ksðnþ1Þ
R
1 /lþ1
2/lr þ /l
1 1 /lþ1
/l
1 c1
2
c
n
i¼1 Gsi þ 2 Gsi wi dz þ Gsðnþ1Þ þ 2 Lp Gsðnþ1Þ w2nþ1 dz
ksi
H i
1
2 r
2
r
þ r r

þ 2 /lr
Dr rl 2Dr rl rp
ð10aÞ  2
rffiffiffiffiffiffiffi c2 /lþ1
/l
1 c
ns ¼ 3 h h

1 /lh ð13Þ
c2 ¼ r p rl 2Dr rl
ms1
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi "   2 #
u Pn R H i dwi 2 R1  2 /lþ1
2/lh þ /l
1 1 /lþ1
/l
1 c4
2
c
u dz þ Lp Gsðnþ1Þ dwdznþ1 dz
u i¼1 H i
1 Gsi dz
h h
þ h h

þ 5 /lh
¼ r p tP
R
Hi

k
R
1 Dr 2 rl 2Dr rl rp
n
i¼1 Gsi þ 2 Gsi w2i dz þ Gsðnþ1Þ þ 2 Lp Gsðnþ1Þ w2nþ1 dz
ksi
H i
1 sðnþ1Þ  2
c26 /lþ1
/l
1 c
ð10bÞ ¼
r r

4 /lr ð14Þ
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rl 2Dr rl
ms2 þ ms3
c3 ¼ where superscript l represents the node number at radial
ms1
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

R ffi distance rl from the pile edge and Dr is the discretization
u Pn ksi R
u i¼1 ð þ 1Þ H i Gsi w2i dz þ ksðnþ1Þ þ 1 1 Gsðnþ1Þ w2nþ1 dz length (i.e., the distance between two consecutive nodes).
u Gsi H G Lp
¼ tP
 R i
1
Hi

sðnþ1Þ
ksðnþ1Þ
R
1 The finite difference mesh is chosen to be sufficiently long
n
i¼1 Gsi þ 2 Gsi wi dz þ Gsðnþ1Þ þ 2 Lp Gsðnþ1Þ w2nþ1 dz
ksi 2
H i
1
and dense so as to allow for the proper attenuation of
ð10cÞ the displacement functions for accurate results (i.e., the
total number of finite difference nodes m is taken to be suf-
Considering the variation in /h in Eq. (6) and following ficiently large so that mDr ? 1).
a similar procedure as was done for /r, the differential Eqs. (13) and (14), excluding the 1st and the mth nodes
equation of /h over rp < r < 1 is obtained as at which the values for /r and /h are known (/r = 1 and
" #
d 2 /h 1 d/h
c 2 c 2 c26 d/r
c4 2 /h = 1 at r = rp, and /r = 0 and /h = 0 at r = 1), generate
þ
4
þ 5
/ h ¼

/r two sets of simultaneous equations each of which contains
dr2 r dr r rp r dr r
m
2 equations, which can be represented in matrix
ð11Þ form as
with boundary conditions /h = 0 at r = 1 and /h = 1 at ½X /r f/r g ¼ fY /h g ð15Þ
r = rp where dimensionless constants c4, c5, and c6 are given
½X /h f/h g ¼ fY /r g ð16Þ
by
rffiffiffiffiffiffiffi [X/r](m
2)(m
2) and [X/h](m
2)(m
2) are the tri-diagonal
ms4
c4 ¼ matrices with the finite difference coefficients of
ms2
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi unknown vectors {/r}(m
2)1 and {/h}(m
2)1, respec-
uPn
R
R
u ksi
þ 3
Hi 2
þ
ksðnþ1Þ
þ 3
1 2 tively. {Y/h}(m
2)1 and {Y/r}(m
2)1 are the right-hand
u i¼1 Gsi H i
1
G si w i dz Gsðnþ1Þ Lp
G sðnþ1Þ w nþ1 dz
¼t Pn R H i R1 side vectors containing unknowns /h and /r, respectively,
i¼1 H i
1 G w
si i
2
dz þ G sðnþ1Þ nþ1w 2
dz
Lp for the two sets of equations.
ð12aÞ The non-zero elements of the left-hand-side matrix [X/r]
rffiffiffiffiffiffiffi in Eq. (15) are given by
ns
c5 ¼ r p 1 1
ms2 ½X /r j;j
1 ¼
ð17aÞ
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2  Dr2 2rl Dr
uPn R H dw 2 R1
u dwnþ1 "   2 #
u i¼1 H i
1 Gsi dz dz þ Lp Gsðnþ1Þ dz2 dz
i i
2
¼ r p t Pn R H i R1 ð12bÞ 2
½X j;j ¼
2

/r c1 c
þ 2 ð17bÞ
i¼1 H i
1 Gsi wi dz þ Lp Gsðnþ1Þ wnþ1 dz
2 2
Dr rl rp
 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110 97

current
1 1
/lh j 6 10
6 , where /lprevious and /lcurrent are the val-
½X /r j;jþ1 ¼ 2
þ ð17cÞ
Dr 2rl Dr ues for function / (/r or /h) in the previous and current
iterations, respectively.
where j = l
1. The jth row of the corresponding {Y/h}
vector is given by 2.6. Equations describing w
 2
c23 /lþ1
/l
1 c Considering the variation in w in Eq. (6), collecting the
fY gj ¼
/h h h

1 /lh ð18aÞ
rl 2Dr rl terms associated with dw and d(dw/dz), and performing the
integration by parts results in
The boundary conditions (/r = 1 at node 1 and /r = 0 "Z  H i
X
n Hi
d 4 wi d 2 wi dwi 
at node m) modify the 1st and (m
2)th rows of {Y/h}, Ep I p dwi dz þ Ep I p 2 d
which are given by i¼1 H i
1 dz 4
dz dz H i
1
H i # n Z H i
 2 d 2 wi  X
1 1
ð2Þ
c23 /h
1 c ð2Þ
Ep I p 2 dwi  þ ff i Gs0 þ si ðz
H i
1 Þxi gk i wi dz
fY g1 ¼
2 þ
/h
þ
1 /h ð18bÞ dz H i
1 i¼1 H i
1
Dr 2r2 Dr r2 2Dr r2 
d 2 wi xi
1 dwi

2tff i Gs0 þ si ðz
H i
1 Þxi g
2ts x ðz
H Þ dwi
 2 dz2
i i i
1
dz
c23
/m
2 c1 H i
fY gm
2
/h
¼ h

/hm
1 ð18cÞ dwi 
rm
1 2Dr rm
1 þ2tff i Gs0 þ si ðz
H i
1 Þxi g dwi 
dz H i
1
ð21Þ
The non-zero elements of the left-hand-side matrix [X/h] Z 1
  d 2 wnþ1

dwnþ1
in Eq. (16) are given by
2tnþ1 f nþ1 Gs0 þsnþ1 ðz
Lp Þ 2
þ snþ1 dwnþ1
Lp dz dz
1
1 1 dwnþ1 
½X /h j;j
1 ¼
ð19aÞ þ2tnþ1 ff nþ1 Gs0 þ snþ1 ðz
Lp Þg dwnþ1 
Dr2 2rl Dr dz Lp

"  Z 1
 2 #
2 c4
2
c þ ff nþ1 Gs0 þ snþ1 ðz
Lp Þgk nþ1 wnþ1 dwnþ1 dz
F a dw1 jz¼0
½X j;j ¼
2

/h
þ 5 ð19bÞ Lp
Dr rl rp  
dw1 
þM a d ¼0
1 1 dz z¼0
½X /h j;jþ1 ¼ 2
þ ð19cÞ
Dr 2rl Dr where
The jth, 1st, and (m
2)th rows of the vector {Y } in /r
" Z 1  2 Z 1  2
Eq. (16) are given by ksi d/r d/h
ki ¼ p þ2 r dr þ r dr
Gsi rp dr rp dr
 2
c26 /lþ1
/l
1 c Z 1 Z 1
fY gj ¼

/r r r

4 /lr ð20aÞ ksi d/ d/
rl 2Dr rl þ2 ð/r
/h Þ r drþ2 ð/r
/h Þ h dr
Gsi rp dr rp dr
 2  Z 1 #
1 1 c26 /ð3Þ
r
1 c4 ksi 1
/rð2Þ
2
fY /r g1 ¼
þ

ð20bÞ þ þ3 ð/
/h Þ dr i ¼ 1; 2; . . . ; n þ 1
Dr2 2r2 Dr r2 2Dr r2 Gsi rp r r
 2
c2
/rm
2 c4 ð22aÞ
fY /r gm
2 ¼
3
/m
1
r ð20cÞ
rm
1 2Dr rm
1 "Z #
1
p
As the right-hand side vectors in Eqs. (15) and (16) con- t¼ ð/2r þ /2h Þrdr ð22bÞ
2 rp
tain unknowns /h and /r, these equations are interdepen-
"Z #
dent and iterations are necessary to obtain their values. p 1
An initial estimate of /r is made and given as input to tnþ1 ¼ ð/2r þ /2h Þrdr þ r2p ð22cÞ
2
{Y/r}, and /h is determined by solving Eq. (16). The calcu- rp

lated values for /h are then given as input to {Y/h} to Referring to Eq. (21) and considering the terms associ-
obtain /r from Eq. (16). The newly obtained values for ated with dwn+1 within the domain below the pile (i.e.,
/r are again used to obtain new values for /h, and the iter- the integral over Lp 6 z < 1), the integrand must be zero
ations are continued until the convergence for both /r and because dwn+1 – 0 (as wn+1 is not known a priori within
/h are reached. The criteria  set for the convergence are
Pm  l previous l current 
6 Pm  l previous the domain), which leads to the differential equation of w
1
m l¼1 /r
/r  6 10 and m l¼1 /h
1
(z) below the pile.
98 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110

d 2 wnþ1 dwnþ1 At the pile base (i.e., at z = Lp), the boundary conditions
2tnþ1 ff nþ1 Gs0 þ snþ1 ðz
Lp Þg 2
þ 2tnþ1 snþ1 are
dz dz


k nþ1 f nþ1 Gs0
snþ1 ðz
Lp Þ wnþ1 ¼ 0 ð23Þ wn ¼ constant ð27aÞ

The boundary conditions for Eq. (23) are wn+1 = 0 at or

z = 1 and wn = wn+1 at z = Lp. These boundary condi- d 3 wn dwn
x
tions make the boundary terms at z = Lp and z = 1 in Ep I p 3

2tn ff n Gs0 þ sn ðz
H n
1 Þ n g
dz dz
Eq. (21) zero and satisfy Eq. (21). Note that, in this analy-
dwnþ1
sis, the variation in shear modulus with depth below the ¼
2tnþ1 ff nþ1 Gs0 þ snþ1 ðz
H n Þg ð27bÞ
pile is assumed to be linear (i.e., x = 1 for Gs(n+1) in Eq. dz
(1)). and
Referring back to Eq. (21) once again, and considering dwn
function w for the domain 0 6 z < Lp, collecting all the ¼ constant ð27cÞ
dz
terms associated with dw and d(dw/dz) and equating them
to zero so that w(z) within 0 6 z < Lp contributes to or
dP = 0 (because w(z) is not known a priori within any layer
d 2 wn
i, dwi – 0 within Hi and Hi+1) gives the differential equa- ¼0 ð27dÞ
tion of w(z) for the ith layer as dz2

d 4 wi 2
x d wi
Ep I p 4

2ti ff i Gs0 þ si ðz
H i
1 Þ i g 2 2.7. Solution for w
dz dz
xi
1 dwi

2ti si xi ðz
H i
1 Þ The solution for the differential Eq. (23) is assumed to be
dz the product of two functions (Polyanin and Zaitsev, 2002).
x
þ k i ff i Gs0 þ si ðz
H i
1 Þ i gwi ¼ 0 ð24Þ qffiffiffiffiffiffiffi
k nþ1

2tnþ1 ðz
Lp Þ
along with the following boundary conditions. At the wnþ1 ðzÞ ¼ W ðZÞe ð28Þ
ground surface (i.e., at z = 0)
where W(Z) is a displacement function (described later)
w1 ¼ constant ð25aÞ and the exponential term partly takes into account the
attenuation of horizontal soil displacement with depth
or
below the pile. Differentiating wn+1 in Eq. (28), with respect
d 3 w1 dw1 to z, and substituting it into Eq. (23) results in
x
Ep I p 3

2t1 ff 1 Gs0 þ s1 ðz
H 0 Þ 1 g
Fa ¼ 0 ð25bÞ  2
dz dz f nþ1 Gs0 d W
þ ðz
Lp Þ
and snþ1 dz2
" sffiffiffiffiffiffiffiffiffiffi #
dw1 k nþ1 f nþ1 Gs0 dW
¼ constant ð25cÞ þ 1
2 þ ðz
Lp Þ
dz 2tnþ1 snþ1 dz
sffiffiffiffiffiffiffiffiffiffi
or k nþ1

W ¼0 ð29Þ
2tnþ1
d 2 w1
Ep I p
Ma ¼ 0 ð25dÞ
dz2 Denoting
At the interface between the ith and (i + 1)th layer (i.e., sffiffiffiffiffiffiffiffiffiffi
k nþ1 f nþ1 Gs0
at z = Hi), the boundary conditions are Z¼2 þ ðz
Lp Þ ð30Þ
2tnþ1 snþ1
wi ¼ wiþ1 ð26aÞ
dwi dwiþ1 and expressing W and its derivatives in terms of Z, Eq. (29)
¼ ð26bÞ reduces to
dz dz
d 3 wi x dwi
d 2W dW 1
Ep I p
2ti ff i Gs0 þ si ðz
H i
1 Þ i g Z 2
þ f1
Zg
W ¼0 ð31Þ
dz3 dz dZ dZ 2
3
d wiþ1 x dwiþ1
¼ Ep I p
2tiþ1 ff iþ1 Gs0 þ siþ1 ðz
H i Þ iþ1 g which is a confluent hypergeometric differential equation
dz3 dz (Arfken and Weber, 2005). The solution for Eq. (31) is
ð26cÞ given by
   
d 2 wi d 2 wiþ1 1 1
¼ ð26dÞ W ðZÞ ¼ C 1 M ; 1; Z þ C 2 U ; 1; Z ð32Þ
dz2 dz2 2 2
 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110 99

where C1 and C2 are integration constants, and M(1/2,1,Z) It is possible for the analytical solution for the differen-
and U(1/2,1,Z) are the confluent hypergeometric functions tial Eq. (24) to be developed for the boundary conditions,
of the first and second kinds (Arfken and Weber, 2005), given by Eqs. (25)–(27), using the power series method, and
respectively, given by this was done. However, the power series solution did not
Z 1 produce sufficiently correct results with a reasonable num-
CðcÞ ber of terms, except for the cases of short stubby piles that
Mða; c; ZÞ ¼ eZx xa
1 ð1
xÞc
a
1 dx
CðaÞCðc
aÞ 0 exhibited almost rigid behavior. Since the power series
for RðcÞ > RðaÞ > 0 ð33aÞ solution, that considers the different soil layers and the
Z 1 variation in the soil modulus in each layer, was cumber-
1 some to handle and did not produce good results for long
e
Zx xa
1 ð1 þ xÞ
c
a
1
U ða; c; ZÞ ¼ dx
CðaÞ 0 piles, even after considering fifty terms, it was decided that
for RðZÞ > 0; RðaÞ > 0 ð33bÞ the Ritz method would be used to obtain the pile
displacements.
in which a = ½, c = 1, C is the gamma function, and R In the Ritz method, trial functions with unknown coef-
stands for a real number. ficients are assumed for the primary variable satisfying the
The soil displacement below the pile decreases with an geometric (essential) boundary conditions a priori (Shames
increase in depth and wn+1 = 0 at z = 1, in which W(Z) and Dym, 1985). These trial functions are then substituted
must be finite. This is only possible if C1 = 0 because M into a functional form (i.e., an integral equation) describing
(1/2,1,Z) ? 1 as Z ? 1. Therefore, substituting C1 = 0 the physical problem. The functional forms for mechanics
into Eq. (32) and using the boundary condition wn+1(z) problems are often the potential energy of the system.
= wn|z=Lp at z = Lp to determine C2, wn+1 is obtained as Next, the functional is minimized with respect to all the
qffiffiffiffiffiffiffi


k nþ1
ðz
Lp Þ
unknown coefficients associated with the trial functions in
wnþ1 ðzÞ ¼ wn jz¼Lp e
2t
nþ1
order to obtain a set of algebraic equations containing
h qffiffiffiffiffiffiffin oi the unknown coefficients. These equations are then solved
f nþ1 Gs0
U 12 ; 1; 2 2tknþ1 snþ1
þ ðz
L p Þ to obtain the unknown coefficients at which point the pri-
 h qffiffiffiffiffiffiffin oi ð34Þ
nþ1

k nþ1 f nþ1 Gs0
U 2 ; 1; 2 2tnþ1 snþ1
1
In all the foregoing equations involving wn+1 and
(dwn+1/dz), these two terms are evaluated using Eq. (33b).
Eq. (33b) involves the hypergeometric function of the sec-
ond kind which is obtained by writing a script in
MATLAB R2013a in which the built-in ‘kummerU’ func-
tion is used. It should be noted that if Gs(n+1) is spatially
constant, then sn+1 ? 0 in Eq. (34) for the equation to be
valid.
mary variable is completely determined. Since the laterally
loaded pile problem considered in this study is based on the
minimization of the potential energy (which is the func-
tional) of the pile-soil system, it was natural to use the Ritz
method to obtain the pile displacements.
Four different combinations of pile head and base
boundary conditions, commonly encountered in practice,
are considered in this study (Fig. 2). The different cases
considered are: (a) pile head free to rotate and translate
with pile base completely restrained against rotation and

Fig. 2. Different pile boundary conditions. (a) Pile head free to rotate and translate and base completely restrained against rotation and translation
(typical free-head long pile), (b) both pile and base free to rotate and translate (typical free head short pile), (c) pile head free to translate but completely
restrained against rotation, and pile base completely restrained against rotation and translation (typical long pile with pile cap), and (d) pile head
completely restrained against rotation (typical free head short pile with pile cap).
100 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110

translation (i.e., free-head, long pile), (b) both pile head eral representation is shown as per Eq. (38)) can be written
and base free to rotate and translate (i.e., free-head short as
2 38 9 8 9
pile), (c) pile head free to translate, but completely a11 a12 . . . . . . a1y > c1 > > b1 >
> > > >
restrained against rotation with pile base completely 6 a21 a22       a2y 7> > c2 >
> > > b2 >>
6 7>> >
> > >
> > >
> >
>
restrained against rotation and translation (i.e., long pile 6 . .. . . 7 <
.. 7 .. = < .. =
6 .
with a pile cap), and (d) pile head free to translate, but 6 . . . . 7 . ¼ . ð39Þ
6 7> > > > > > >
completely restrained against rotation with pile base free 6 .. .. .. .. 7> > . >
> .. > > .
> .. >>
4 . . . . 5> >
>
>
> > > > >
to rotate and translate (i.e., short pile with pile cap). : > ; > : > ;
To use the Ritz method, the total potential energy of the ax1 ax2       axy cx bx
soil-pile system (i.e., the functional form) within the where
domain 0 6 z < Lp is rewritten as follows: Z Lp

Z Lp  2 2 (  2 ) axy ¼ Ep I p U00y U00x dz

Xn Z Hi
1 d w ki 2 dw 0
 
I ¼ Ep I p dz þ Gsi w þt dz n Z
X Hi
2 dz 2
2 dz ki
0
 i¼1 H i
1 þ 2Gsi Uy Ux þ tU0y U0x dz ð40aÞ
dw i¼1 H i
1

2

F a wjz¼0 þ M a  ð35Þ
dz z¼0
by ¼ F a Ux jz¼0
M a U0x z¼0 x; y ¼ 1; 2; . . . ; a ð40bÞ

Pile displacement w is assumed as which can be solved to obtain the unknown cm coefficients.
The coefficients are substituted back into Eq. (36) to obtain
X
a
the pile displacement. The elements of the left-hand side
wðzÞ ¼ cm Um ðzÞ ð36Þ
matrix and the right-hand side vector in Eq. (39) for each
m¼1
of the pile boundary conditions (Eqs. (37a–d)) are shown
where cm are the unknown coefficients, Um are the trial in the Appendix.
functions, chosen such that they satisfy the geometric
boundary conditions, and a is the total number of terms 2.8. Solution algorithm
considered (a should be large enough to obtain a proper
solution). For the pile boundary conditions shown in To solve for the pile displacement using the Ritz method
Fig. 2, the trial functions assumed are given by for any of the four cases of pile boundary conditions
X    (Fig. 2), soil parameters ki, t, and tn+1 must be known.
a
ð2m
1Þp These parameters depend on functions /r and /h which,
wðzÞ ¼ cm 1
cos ðLp
zÞ ð37aÞ
m¼1
2Lp in turn, depend on w through the six dimensionless con-
  stants, c1-c6. Thus, an iterative algorithm is required to
ðLp
zÞ X a
fm
2gp obtain w.
wðzÞ ¼ c1 þ c2 þ cm sin ðLp
zÞ
Lp m¼3
2Lp Initial guesses on c1-c6 are made using which /r and /h
ð37bÞ are determined by solving Eqs. (9) and (11) iteratively using
   the one-dimensional finite difference scheme described by
Xa
mp Eqs. (13) and (14). From the calculated values for /r and
wðzÞ ¼ cm 1
cos ðLp
zÞ ð37cÞ
m¼1
Lp /h, ki, t, and tn+1, are obtained using Eqs. (22a)–(22c),
  respectively, using numerical integrations following the
Xa
f2m
3gp trapezoidal rule (the same grid as that used for the finite
wðzÞ ¼ c1 þ cm sin ðLp
zÞ ð37dÞ
m¼2
2Lp difference solution for /r and /h used for the numerical
integrations). The calculated values for ki, t, and tn+1 are
for the four cases of pile boundary conditions, (a), (b), (c), used to solve for the constants in Eq. (36) using which w
and (d), respectively (Fig. 2). is calculated. Using the calculated values for w, c1-c6 are
Minimizing I (i.e., setting oI/ocm = 0) leads to the fol- calculated using Eqs. (10a)–(10c) and (12a)–(12c). The cal-
lowing equation: culated values for c1-c6 are compared with the assumed ini-
Z Lp tial values and if the differences are more than the tolerable
@I @w00
¼ Ep I p w00 dz limit of 0.001, the same set of calculations is repeated with
@cm @cm
X n Z Hi
0
 the calculated values for c1-c6 as the new initial guesses.
k i @w @w0 Iterations are continued until the values for c1-c6 between
þ 2Gsi w þ tw0 dz
i¼1 H i
1
2 @cm @cm successive iterations fall below 0.001. Details of the solu-
 
@w  @w0  tion steps are given in the form of a flowchart in Fig. 3.

Fa þ M ¼0 ð38Þ
@cm z¼0 @cm z¼0
a

2.9. Modification of shear modulus of soil

where w0 = dw/dz and w00 = d2w/dz2.
Substituting Eqs. (37a)–(37d) along with the appropriate For soil displacement fields similar to those described in
derivatives in Eq. (38), an equation in matrix form (a gen- Eqs. (2a) and (2b), Guo and Lee (2001) found that the pile
 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110
Fig. 3. Solution flowchart.
response is excessively stiff for Poisson’s ratio close to 0.5.
To avoid this artificial stiffness, Guo and Lee (2001) recom-
mended setting ms = 0 (which is the same as setting ks = 0)
and taking into account the effect of ms indirectly through
Eq. (41) given by Randolph (1981).
Gs ¼ Gs ð1 þ 0:75ms Þ ð41Þ
The other advantage of using Eq. (41) is that the pile
response can be investigated in terms of a single soil param-
eter because it takes into account the effect of both elastic
101
constants. In this paper, the results are mostly generated
without using Gs* because the difference in the results
between using and not using Gs* for the problems investi-
gated here were negligible. However, for illustration pur-
poses, two problems are solved using Gs*, as will be
described later (see Sections 3.3.2 and 3.3.4).
3. Results
Laterally loaded piles with different geometries (long
and short) embedded in multilayered soil deposits, with lin-
early (x = 1) and nonlinearly (x = 0.5 and x = 2) varying
shear moduli with depth within each layer, are analysed
in this study. Four different cases of boundary conditions
are considered, as shown in Fig. 2. The pile displacement
function (Eqs. (37a)–(37d)) is selected following the bound-
ary conditions corresponding to each case.
The framework of the present analysis gives accurate
results for the profiles of pile displacement and rotation.
However, the pile bending moment profiles obtained from
this analysis are not always accurate. Nevertheless, the
location and the magnitude of the maximum bending
moment, as obtained from this analysis, were found to be
reasonably accurate.
The accuracy of the results obtained from this analysis is
verified by comparing pile displacement and rotation pro-
files obtained from an equivalent 3-D FE analysis (using
ABAQUS 6.10). The accuracy of the analysis is further ver-
ified by comparing the location and the magnitude of the
maximum bending moment with those obtained from the
3-D FE analysis. Further verification of the analysis is per-
formed by comparing the pile displacements obtained from
this analysis with results available in the literature.
In ABAQUS, the pile and the soil are modeled as a sin-
gle cylindrical part with appropriate partitioning to sepa-
rately represent the pile and the soil. This also ensures
that there is no slippage or separation between the soil
and the pile. The top soil surface is flush with the pile head,
and the bottom soil surface extends to a finite depth of
more than 10 m below the pile base, for each of the cases
analysed. The horizontal radial extent of the soil domain
(i.e., the vertical curved boundary of the FE domain) for
each case is considered at a distance of 20 m from the pile
axis. Eight-node continuum reduced integration (C3D8R)
elements are used in each of the models, with an approxi-
mate global seed size of 0.8. The optimal domains and
meshes described are obtained by performing convergence
checks ensuring that there are no boundary effects. Differ-
ent boundary conditions are prescribed at the soil bound-
aries of the FE domain
all displacements at the bottom
and the vertical boundary of the soil domain are made to
be zero. Concentrated force and moment are applied to a
reference point at the centroid of the pile head, to which
all the nodes at the pile top are connected using a rigid
body constraint for both free head and fixed-sway head
(i.e., piles with pile cap) piles. For the case of fixed head
102 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110
Fig. 4a. Variation in Young’s modulus with pile embedment depth in four-layer soil for free head long pile.
Fig. 4b. Pile displacement versus pile embedment depth in four-layer soil
with linearly varying modulus in each layer, for free head long pile.
piles, appropriate boundary conditions are imposed to
maintain a zero slope at the head.
In order to implement the linear and nonlinear variation
in the soil shear modulus with the depth within each layer
in ABAQUS, the ‘‘field variable” and the ‘‘user defined
field” option is used for the material definition. In the sim-
ulations, the field variable is assigned as the Young’s mod-
ulus and the user-defined field (USDFLD) FORTRAN
subroutine is written in Microsoft Visual studio where
the Young’s modulus is made to vary with the vertical
coordinate either linearly or non-linearly. The user-
defined field is linked to the model developed in the ABA-
QUS Create job option and the analysis is performed in
Intel Fortran environment.
3.1. Comparison of pile response for linear variation in soil
modulus with 3-D FE analysis
3.1.1. Free head long pile
A concrete pile with Young’s modulus Ep = 28 MPa,
length Lp = 25 m, and radius rp = 0.3 m, subjected to an
applied force Fa = 275 kN and moment Ma = 250 kNm
at the head, is assumed to be embedded in a four-layer soil
profile. The Poisson’s ratio ms for all the soil layers is 0.2.
The Young’s modulus Es of the soil at the surface of the
first (top) layer is 15 MPa. The Young’s modulus decreases
linearly with depth in the first layer at the rate of 1.25 MPa/
m. Therefore, the shear modulus at the surface Gs = [Es/{2
(1 + ms)}] = {15/2  (1 + 0.2)} = 6.25 MPa, and f1 = Gs/
Gs0 = 6.25/100 = 0.0625, and s1 =
dGs1/dz =
(dEs1/dz)/
{2(1 + ms1)} =
0.54 MPa/m. For the second layer, the
modulus remains constant with depth so that s2 = 0
MPa/m and f2 = 0.043. For the third layer, f3 = 0.043,
and the rate of increase in shear modulus is s3 = 1.09
MPa/m. For the fourth layer, f4 = 0.13, and the shear mod-
ulus is assumed to be constant with depth so that s4 ?
0 MPa/m. This problem is also simulated in ABAQUS. A
contour plot of the variation in Young’s modulus with
depth is shown in Fig. 4a.
Eq. (37a) is chosen to obtain the pile displacement pro-
file using the Ritz method, for this case. Since the displace-
ment at the base of the pile is zero, wn+1 = 0 which implies
that the definite integral within the limits Lp and 1 in the
calculation of c1-c6 are zero. Figs. 4b and 4c show the plot
of pile displacement w and pile rotation dw/dz versus pile
depth z obtained from the present and FE analyses. Table 1
shows the location zmax and magnitude Mmax of the maxi-
mum pile bending moment as obtained from the present
and FE analyses. A good agreement between the results
from the proposed analysis method using eight terms in
Eq. (37a) (i.e., a = 8) and those from the FE analysis is
obtained.
 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110 103

Fig. 4c. Pile rotation versus pile embedment depth in four-layer soil with Fig. 5a. Pile displacement versus pile embedment depth in two-layer soil
linearly varying modulus in each layer, for free head long pile. with linearly varying modulus in each layer, free head short pile.

3.1.2. Free head short pile

A concrete pile with Young’s modulus of elasticity
Ep = 28 MPa, length Lp = 8 m, and radius rp = 0.5 m with
an applied force Fa = 200 kN and moment Ma = 300 kNm
at the head is assumed to be embedded in a two-layer soil
profile. The Young’s modulus Es of the soil at the surface
of the first layer is 10 MPa, and the Poisson’s ratio for all
the soil layers is ms = 0.2. The Young’s modulus increases
linearly with depth in both layers. The corresponding sca-
lars, f1 = 0.04167, f2 = 0.125, and f3 = 0.132, and the rate
of change in modulus s1 = 0.347 MPa/m, s2 = s3 = 0.347 -
MPa/m, are used for the analysis.
Eq. (37b) is chosen to obtain the pile displacement pro-
file using the Ritz method, for this case. Since the displace-
ment at the base of the pile is non-zero, it implies that the
definite integral within limits Lp-1 in the calculation of c1-
c6 are also non-zero. Figs. 5a and 5b show the plot of pile
displacement w and pile rotation dw/dz versus pile depth z.
Table 1 shows the location and the magnitude of the max- Fig. 5b. Pile rotation versus pile embedment depth in two-layer soil with
imum pile bending moment. A good agreement between linearly varying modulus in each layer, free head short pile.

Table 1
Location zmax and magnitude Mmax of maximum bending moment from finite element analysis (FEA) and present analysis (PA).
Modulus variation Pile boundary conditions zmax (m) zmax (m) Mmax (kN-m) Mmax (kN-m)
with depth FEA PA FEA PA
Linear Free head long pile 0.99 2.00 300.2 310.0
Free head short pile 2.02 2.25 423.7 436.1
Long pile with pile cap 0.00 0.00 223.3 230.6
Nonlinear Free head long pile 2.00 2.55 631.1 689.5
Short pile with pile cap 0.00 0.00 696.5 700.9
Long pile with pile cap 0.00 0.00 593.2 601.6
104 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110
the results from the proposed analysis method using eight
terms in Eq. (37b) (i.e., a = 10) and those from the FE
analysis is obtained.
3.1.3. Long pile with pile cap
A concrete pile with Young’s modulus of elasticity
Ep = 28 MPa, length Lp = 20 m, and radius rp = 0.3 m with
an applied force Fa = 450 kN at the head is assumed to be
embedded in a three-layer soil profile. The Young’s modu-
lus Es of the soil at the surface of the first layer is 25 MPa,
and the Poisson’s ratio for all the soil layers is ms = 0.25.
Fig. 6a. Pile displacement versus pile embedment depth in three-layer soil
with linearly varying modulus in each layer, for long with pile cap.
Fig. 6b. Pile displacement versus pile embedment depth in three-layer soil
with linearly varying modulus in each layer, for long pile with pile cap.
The Young’s modulus remains constant in the first and
third layers and increases linearly with depth in the second
layer. The corresponding scalars, f1 = 0.12, f2 = 0.12, and
f3 = 0.2, and the rate of change in the modulus,
s1 = 0.0 MPa/m, s2 = 0.5 MPa/m, and s3 ? 0.0 MPa/m,
are used for the analysis.
Eq. (37c) is chosen to obtain the pile displacement pro-
file using the Ritz method, for this case. The displacement
at the base of the pile is zero; therefore, the definite integral
within limits Lp-1 in the calculation of c1-c6 are also zero.
Figs. 6a and 6b show the plot of pile displacement w and
pile rotation dw/dz versus pile depth z. Table 1 shows the
location and the magnitude of the maximum pile bending
moment. A good agreement between the results of the pro-
posed analysis method using eight terms in Eq. (37c) (i.e.,
a = 8) and those of the results from the FE analysis is
obtained.
3.2. Comparison of pile response for nonlinear variation in
soil modulus with 3-D FE analysis
3.2.1. Free head long pile
A concrete pile with Young’s modulus of elasticity
Ep = 28 MPa, length Lp = 28 m, and radius rp = 0.45 m
with applied force Fa = 375 kN and moment Ma = 400 -
kNm at the head is assumed to be embedded in a single-
layer soil profile with shear modulus varying nonlinearly
(x = 0.5 in Eq. (1)) with depth. This non-linear variation
in Young’s modulus with depth is shown by the contour
plot along with a plot of the variation in Young’s modulus
with depth in Fig. 7a. The function chosen to obtain the
pile displacement profile using the Ritz method is the same
as Eq. (37a). The pile displacement at the base of the pile is
zero. Figs. 7b and 7c show the plot of pile displacement w
and pile rotation dw/dz versus pile depth z with all the soil
properties evaluated for the computation. Table 1 shows
the location and the magnitude of the maximum pile bend-
ing moment. A good agreement between the results from
the proposed analysis method using eight terms in Eq.
(37a) (i.e., a = 8) and those from the FE analysis is
obtained.
3.2.2. Short pile with pile cap
A concrete pile with Young’s modulus of elasticity
Ep = 28 MPa, length Lp = 10 m, and radius rp = 0.6 m with
an applied force Fa = 475 kN at the head is assumed to be
embedded in a two-layer soil profile. The shear modulus of
the first layer varies nonlinearly (x = 0.5 in Eq. (1)) with
depth, whereas the shear modulus of the second layer var-
ies linearly (x = 1 in Eq. (1)) with depth. The function cho-
sen to obtain the pile displacement profile is the same as
Eq. (37d). The displacement at the base of the pile is
non-zero. Figs. 8a and 8b show the plot of pile displace-
ment w and pile rotation dw/dz versus pile depth z. Table 1
shows the location and the magnitude of the maximum pile
bending moment. A good agreement is obtained between
the results from the proposed analysis method using nine
 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110
Fig. 7a. Variation in Young’s modulus with pile embedment depth in single-layer soil for free head long pile.
Fig. 7b. Pile displacement versus pile embedment depth in single-layer soil
with non-linearly varying modulus, with free head long pile.
terms in Eq. (37d) (i.e., a = 10) and the results from the FE
analysis.
3.2.3. Long pile with pile cap
A concrete pile with Young’s modulus of elasticity
Ep = 28 MPa, length Lp = 22 m, and radius rp = 0.45 m
with an applied force of Fa = 500 kN at the head is
assumed to be embedded in a single-layer soil profile with
non-linearly varying (x = 2 in Eq. (1)) shear modulus with
depth. The function chosen to obtain the pile displacement
profile is the same as Eq. (37c). The pile displacement at the
tip of the pile is zero. Figs. 9a and 9b show the plot of pile
displacement w and pile rotation dw/dz versus pile depth z.
Table 1 shows the location and the magnitude of the max-
imum pile bending moment. A good agreement is obtained
105
Fig. 7c. Pile rotation versus pile embedment depth in single-layer soil with
non-linearly varying modulus, with free head long pile.
between the results from the proposed analysis method
using eight terms in Eq. (37c) (i.e. a = 8) and the results
from the FE analysis.
3.3. Comparison of pile response with results available in the
literature
For further validation of the developed analysis, pile
responses are obtained from the literature and compared
with the responses obtained from the present analysis
described below.
3.3.1. Free head long pile
A field example of a load test, performed by McClelland
and Focht (1958), on a free head long pile of length
106 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110

Fig. 8a. Pile displacement versus pile embedment depth in two-layer soil Fig. 9a. Pile displacement versus pile embedment depth in single-layer soil
with non-linearly varying modulus in first layer and linearly varying with non-linearly varying modulus, for long pile with pile cap.
modulus in second layer, for short pile with pile cap.

Fig. 9b. Pile displacement versus pile embedment depth in single-layer soil
Fig. 8b. Pile displacement versus pile embedment depth in two-layer soil with non-linearly varying modulus, for long pile with pile cap.
with non-linearly varying modulus in first layer and linearly varying
modulus in second layer, for short pile with pile cap.

sent analysis with s1 = s2 = 0.8 MPa/m, f1 = 0 and

Lp = 23 m and radius rp = 0.305 m embedded into a nor-
f2 = 0.184. A comparison of the pile displacement profile
mally consolidated clay deposit is chosen. The pile was sub-
(Fig. 10) obtained from the present analysis with those of
jected to a lateral force of Fa = 300 kN and a moment of
the field test and the FE study by Randolph (1981) show
Ma =
265 kNm at the head. For comparison with the
that the present analysis works well.
FE study, Randolph (1981) back calculated the pile modu-
lus Ep = 68420 MPa from the reported pile flexural rigidity
and assumed that the shear modulus Gs of the soil deposit 3.3.2. Free head short pile
at the test site varied linearly as 0.8z MPa/m with a Pois- In their Fourier FE study on laterally loaded piles
son’s ratio of ms = 0.3. This problem is simulated in the pre- embedded in linearly varying soil profiles, Higgins et al.
 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110 107

Fig. 10. Displacement profile of free head long pile in single-layer soil with Fig. 12. Displacement profile in three-layer soil with linearly varying
linearly varying modulus. modulus, for long pile with pile cap.

!  ! 

1:99
2:93
Fa Lp Ma Lp
H ¼ 0:29  3 þ 0:33  4 ð42bÞ
s rp rp s rp rp

Fig. 11. Displacement profile in single-layer soil with linearly varying
modulus, for short pile.
Higgins et al. (2013) obtained the pile head slope and the
displacement for a concrete pile with Ep = 28 GPa,
Lp = 8 m, and rp = 1 m embedded in a linear varying soil
profile with s = 0.5 MPa/m and ms = 0.25 to an applied
loading of Fa = 500 kN and Ma = 400 kN-m. For this
problem, Lp/rp = 8/1 and 2{Ep/(s*rp)} 0.22 = 2  [28  103/
{0.5  (1 + 0.75  0.25)  1}]0.22 = 21.34. Therefore, Lp/
rp < 2{Ep/(s*rp)}0.22. Hence, the pile is rigid, and pile head
displacement wh = 0.0322 m and pile head rotation
H = 0.0044 rad are calculated using Eqs. (42a) and (42b).
As the pile is rigid, it can be assumed to have a constant
slope along its shaft so that the pile base displacement is
wb = 0.0322–0.0044  8 =
0.003 m. The problem is also
simulated using the present analysis and the results match
well, as seen in Fig. 11. Details of the inputs for the pile,
considering the pile to be flexible, as well as the soil are
shown in this figure.

(2013) recommended that, for free head piles with Lp/rp < 2
{Ep/(s*rp)}0.22 where s* = s(1 + 0.75ms) (according to 3.3.3. Long pile with pile cap
Randolph (1981)), the pile behavior is rigid and is indepen- Salgado et al. (2014) analysed a laterally loaded long pile
dent of the relative pile-soil stiffness. Based on the study, with a cap such that Ep = 25 MPa, Lp = 15 m, rp = 0.25 m,
closed-form dimensionless fitted equations were given for and Fa = 370 kN. The pile was embedded in a three-layer
calculating the pile head displacement (wh) and the rotation soil deposit with constant shear modulus within each layer.
(H) for free head rigid piles in linearly varying soil profiles. The shear modulus for the first, second, and third layers are
3.7, 12, and 26.1 MPa, respectively. This problem is simu-
!  !  lated using the framework of the present analysis and a

1:14
1:99
Fa Lp Ma Lp
wh ¼ 0:37  2 þ 0:29  3 ð42aÞ very good match is obtained for the pile displacement pro-
s rp rp s rp rp file (Fig. 12).
108 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110

3.3.4. Short pile with pile cap parametric study closed-form dimensionless equations
Higgins et al. (2013) proposed closed-form dimension- can be developed to estimate the pile head displacement,
less fitted equations for calculating the pile head displace- the rotation, and the bending moment for use in practice.
ment (wh) of fixed head rigid piles in linearly varying soil
profiles subjected to applied forces at the pile head, as
follows: Appendix A
! 

1:50
2a 8 9 8 9
wh ¼ 0:14  2
Fa Lp
ð43Þ 11 a12 . . . . . . a1y 3> c1 > > b1 >
>
> > > >
s rp rp 6 a21       a2y 7 > c2 >> >
> b2 >
>
6 a22 7>>
>
>
> >
> >
>
6 . .. .. .. 7 < . = < . >
> > =
They further solved the problem of a concrete pile with 6 . 7 . .
6 . . . . 7 . ¼ .
Ep = 28 GPa, Lp = 6 m, and rp = 1 m embedded in a linear 6 7> > > > > . >
6 .. .. .. . 7> > ... >
> > > >
>
.. >
>
varying soil profile with s = 0.15 MPa/m, ms = 0.40, and 4 . . . .. 5> >
>
>
> >
> >
>
applied force of Fa = 350 kN. For this problem, Lp/ : > ; > : > ;
ax1 an2       axy cx bx
rp = 6/1 = 6 and 2{Ep/(s*rp)} 0.22 = 2  [28  103/{0.15 
(1 + 0.75  0.49)  1}]0.22 = 26.96. Hence, the pile is rigid, Free head long pile (Eq. (37a))
and the pile head displacement of wh = 0.14  [350/
Z
{0.15  (1 + 0.75  0.49)  1000  12}]  (6/1)
1.50 = 0.016 m axy ¼ Ep I p
Lp
½fð2y
1ÞKg2 cosfð2y
1ÞWgfð2x
1ÞKg2 cosfð2x
1ÞWgdz
is obtained using Eq. (43). The problem is also simulated using 0

X n Z Hi
the present analysis with the following inputs: s*1 = s*2 = 0.205 - ki
þ 2Gsi ð1
cosfð2y
1ÞWgÞð1
cosfð2x
1ÞWgÞ
2
MPa/m, f1 = 0, and f2 = 0.009. The pile head displacement i¼1 H i
1

obtained from the present analysis is 0.019 m. þ tðfð2y
1ÞKg sinfð2y
1ÞWgÞðfð2x
1ÞKg sinfð2x
1ÞWgÞgdz

4. Conclusions bx ¼ F a f1
cos½ð2x
1ÞKLp g

M a f½ð2x
1ÞK sinðð2x
1ÞXLp Þg x; y ¼ 1; 2; . . . ; a
A continuum-based method has been presented to anal-
yse laterally loaded piles in multilayered heterogeneous Free head short pile (Eq. (37b))
elastic soil deposits. The method is based on rational
Xn Z Hi  
assumptions of displacement fields for the soil and the ki
piles. Differential equations were obtained to solve the pile a11 ¼ 2Gsi dz
i¼1 H i
1
2
and soil displacements using the minimum potential energy
principle. The differential equations describing the soil dis- n Z
X  
Hi
ki Lp
z
placement were solved numerically. The differential equa- a12 ¼ 2Gsi dz
H i
1 2 Lp
tion describing the pile displacement was solved using the i¼1

Ritz method by substituting appropriate trail functions n Z

X Hi  
ki
for the primary variable into the functional form of the dif- a1y ¼ 2Gsi sinfyWgdz
ferential equation. i¼1 H i
1 2
For verification of the developed semi-analytical proce- n Z
X  
dure, different pile geometries were considered in a multi-
Hi
ki Lp
z
a21 ¼ 2Gsi dz
layered soil by linearly and nonlinearly varying the shear i¼1 H i
1 2 Lp
modulus with depth in each layer for four different cases "  #
n Z
X 2
of pile boundary conditions that are commonly found in Hi
ki Lp
z 1
practice. The pile displacement and rotation profiles a22 ¼ 2Gsi þ t 2 dz
i¼1 H i
1 2 Lp Lp
obtained from the analysis method in each of the cases
are compared to those obtained from a 3-D FE analysis n Z
X   
Hi
ki Lp
z
and were found to compare fairly well. Although accurate a2y ¼ 2Gsi sinðyWÞ
H i
1 2 Lp
pile displacement and rotation profiles were obtained with i¼1

the Ritz method, the bending moment profiles were not 1

þðyKÞ cosðyWÞ dz
very accurate because of the choice of trial functions. How- Lp
ever, reasonable results for the location and the magnitude  
n Z
X Hi
of the maximum bending moment were obtained from the ki
ax1 ¼ 2Gsi sinfðx
2ÞWgdz
analysis. The accuracy of the analysis method was further i¼1 H i
1 2
verified with results available from the literature.
n Z
X  
The developed analysis method has the advantage that it k i Lp
z
Hi

can be used to conduct an exhaustive parametric study for ax2 ¼ f g sinfðx
2ÞWg
2Gsi
i¼1 H i
1
2 Lp
different cases of pile geometry and soil profiles without 
resorting to a 3-D FE analysis, which is time-consuming 1
þt fðx
2ÞKg cosfðx
2ÞWg dz
and requires the expertise of FE software. From the Lp
 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110 109

Z Lp h i
2 2
Anderson, J.B., Townsend, F.C., Grajales, B., 2003. Case history
axy ¼ Ep I p fðy
2ÞKg sinfðy
2ÞWg fðx
2ÞKg evaluation of laterally loaded piles. J. Geotech. Geoenviron. Eng.
0
Xn Z Hi  ASCE 129 (3), 187–196.
ki Arfken, G.B., Weber, H.J., 2005. Mathematical methods for Physicists,
 sinfðx
2ÞWgdz þ 2Gsi ½sinfðy
2ÞWg 6e.Academic Press An imprint of Elsiever; ISBN: 0-12-059876-0.
i¼1 H i
1
2
Ashour, M., Norris, G., 2000. Modeling lateral soil–pile response based
 sinfðx
2ÞWg þ t½fðy
2ÞKg cosfðy
2ÞWg on soil–pile interaction. J. Geotech. Geoenviron. Eng. ASCE 126 (5),
420–428.
fðx
2ÞKg cosfðx
2ÞWggdz Banerjee, P.K., Davies, T.G., 1978. The behaviour of axially and laterally
loaded single piles embedded in nonhomogeneous soils. Geotechnique
b1 ¼ F a 28 (3), 309–326.
Ma Basu, D., 2006. Analysis of Laterally Loaded Piles in Layered Soil. Ph.D.
b2 ¼ F a
Thesis; Purdue University.
Lp Basu, D., Salgado, R., Prezzi, M., 2009. A continuum-based model
bx ¼
F a ½sinfðx
2ÞWLp g
M a ½fðx
2ÞKg sinfðx
2ÞWLp g foranalysis of laterally loaded piles in layered soils. Geotechnique 59
(2), 127–140.
x; y ¼ 3; 4;... ;a Briaud, J.L., Smith, T., Meyer, B., 1984. Laterally loaded piles and the
pressuremeter: comparison of existing methods. In: Langer, J.A.,
Long pile with pile cap (Eq. (37c)) Mosley, E.T., Thompson, C.C. (Eds.), Laterally Loaded Deep
Z Lp Foundations. ASTM STP 835, pp. 97–111. Philadelphia, PA: ASTM.
2 2
axy ¼ Ep I p ½f2yKg cosf2yWgf2xKg cosf2xWgdz Budhu, M., Davies, T.G., 1988. Analysis of laterally loaded piles in soft
0 clays. J. Geotech. Eng. Div. ASCE 114 (1), 21–39.
X n Z Hi  Chow, Y.K., 1987. Iterative analysis of pile–soil–pile interaction.
ki
þ 2Gsi ð1
cosf2yWgÞð1
cosf2xWgÞ Geotechnique 37 (3), 321–333.
i¼1 H i
1
2 Cox, W.R., Reese, L.C., Grubbs, B.R., 1974. Field testing of laterally
loaded piles in sand. Proc. 6th Offshore Technol. Conf., Houston, TX
þ tðf2yKg sinf2yWgÞðf2xKg sinf2xWgÞgdz
2, pp. 459–472.
bx ¼ F a f1
cos½2xKLp g x; y ¼ 1; 2; . . . ; a Filho, R.M., Mendonca, A.V., Paiva, J.B., 2005. Static boundary element
analysis of pilessubmitted to horizontal and vertical loads. Eng. Anal.
Short pile with pile cap (Eq. (37d)) Boundary Elem. 29, 195–203.
Xn Z Hi   Gabr, M.A., Lunne, T., Powell, J.J., 1994. P–y analysis of laterally loaded
ki
a11 ¼ 2Gsi dz piles in clay using DMT. J. Geotech. Eng. ASCE 120 (5), 816–837.
i¼1 H i
1
2 Guo, W.D., Lee, F.H., 2001. Load transfer approach for laterally loaded

Xn Z Hi   piles. Int. J. Numer. Anal. Methods Geomech. 25 (11), 1101–1129.

ki Haldar, S., Basu, D., 2014. Resistance factors for laterally loaded piles in
a1y ¼ 2Gsi sinfð2y
3ÞWgdz clay. In Proceedings of Geo-Congress, Atlanta, Georgia, U.S.A., CD-
i¼1 H i
1
2
ROM, Geotechnical Special Publication, No. 234, ASCE.
Xn Z Hi   Higgins, W., Martindale, H., Chakraborty, T., Basu, D., 2010. Assessment
ki
ax1 ¼ 2Gsi sinfð2x
3ÞWgdz of the P-Y Method for Laterally Loaded Pile in Sand. Indian
i¼1 H i
1
2 Geotechnical Conference IGS Mumbai Chapter and IIT Bombay,
Z Lp pp. 833–836.
2 2
axy ¼ Ep I p ½fð2y
3ÞKg sinfð2y
3ÞWgfð2x
3ÞKg Higgins, W., Vasquez, C., Basu, D., Griffiths, D.V., 2013. Elastic solutions
0 of laterally loaded piles. J. Geotech. Geoenviron. Eng. 139 (7), 1096–
Xn Z Hi  1103.
ki
 sinfð2x
3ÞWgdz þ 2Gsi ½sinfð2y
3ÞWg Kim, B.T., Kim, N.K., Lee, W.J., Kim, Y.S., 2004. Experimental load-
i¼1 H i
1
2 transfer curves of laterally loaded piles in Nak-Dong river sand. J.
Geotech. Geoenviron. Eng. ASCE 130 (4), 416–425.
 sinfð2x
3ÞWg þ t½fð2y
3ÞKg cosfð2y
3ÞWg Matlock, H., 1970. Correlations for design of laterally loaded piles in soft
 fð2x
3ÞKg cosfð2x
3ÞWggdz clay. Proc. 2nd Offshore Technol. Conf., Houston, TX 1, pp. 577–594.
Polyanin, A.D., Zaitsev, V.F., 2002. Handbook of Exact Solutions for
b1 ¼ F a Ordinary Differential Equations, second ed. Chapman and Hall/CRC
bx ¼ F a sinfð2x
3ÞWLp g x; y ¼ 2; 3; . . . ; a Press.
Poulos, H.G., 1971a. Behavior of laterally loaded piles: I – Single piles. J.
where Soil Mech. Found. Div. ASCE 97 (SM5), 711–731.
Poulos, H.G., 1971b. Behavior of laterally loaded piles: III–Socketed piles.
p
K¼ J. Soil Mech. Found. Div. ASCE 98 (SM4), 341–360.
2Lp Poulos, H.G., 1973. Load deflexion prediction for laterally loaded piles.
p Aust. Geomechanics, J. G3, No. 1.
W¼ ðLp
zÞ Shames, I.H., Dym, C.L., 1985. Energy and Finite Element Methods in
2Lp Structural Mechanics. Taylor and Francis.
Randolph, M.F., 1981. The response of flexible piles to lateral loading.
References Geotechnique 31 (2), 247–259.
Reddy, A.S., Valsangkar, A.J., 1971. Analysis of laterally loaded pile in a
ABAQUS, 2010. User Manual. Dassault Systemes Simulia Corp, layered soil by energy method. Acta Technica, Acad. Scient. Hungar-
Providence. icae, Tomus 70 (1–2), 235–249.
Ai, Z.Y., Feng, D.L., Cheng, Y.C., 2013. BEM analysis of laterally loaded Reese, L.C., Cox, W.R., Koop, F.D., 1974. Analysis of laterally loaded
piles in multi-layered transversely isotropic soils. Eng. Anal. Boundary piles in sand. Proc. 6th Offshore Technol. Conf., Houston, TX 2, pp.
Elem. 37, 1095–1106. 473–483.
110 B.K. Gupta, D. Basu / Soils and Foundations 57 (2017) 92–110
Reese, L.C., Cox, W.R., Koop, F.D., 1975. Field testing and analysis of
laterally loaded piles in stiff clay. Proc. 7th Offshore Technol. Conf.,
Houston, TX 2, pp. 671–690.
Salgado, R., Tehrani, F.S., Prezzi, M., 2014. Analysis of laterally
loaded pile groups in multilayered elastic soil. Comput. Geotech. 62,
136–153.
Scott, R.F., 1981. Foundation analysis. Prentice-Hall Civil Engineering
and Engineering Mechanics series.
Shen, W.Y., Teh, C.I., 2002. Analysis of laterally loaded pile groups using
a variational approach. Geotechnique 52 (3), 201–208.
Sun, K., 1994. Laterally loaded piles in elastic media. J. Geotech. Eng.
ASCE 120 (8), 1324–1344.
Trochanis, A.M., Bielak, J., Christiano, P., 1991. Three-dimensional
nonlinear study of piles. J. Geotech. Eng. ASCE 117 (3), 429–447.
Verruijt, A., Kooijman, A.P., 1989. Laterally loaded piles in a layered
elastic medium. Geotechnique 39 (1), 39–46.
Yan, L., Byrne, P.M., 1992. Lateral pile response to monotonic pile head
loading. Can. Geotech. J. 29, 955–970.
Yang, K., Liang, R., 2006. Numerical solution for laterally loaded piles in
atwo-layer soil profile. J. Geotech. Geoenviron. Eng. ASCE 132 (11),
1436–1443.
Zhang, L., Ernst, H., Einstein, H.H., 2000. Nonlinear analysis of laterally
loaded rock-socketed shafts. J. Geotech. Geoenviron. Eng. ASCE 126
(11), 955–968.