Beruflich Dokumente
Kultur Dokumente
Purdue e-Pubs
Lyles School of Civil Engineering Faculty
Lyles School of Civil Engineering
Publications
1-1-2009
Rodrigo Salgado
Purdue University, rodrigo@ecn.purdue.edu
Monica Prezzi
Purdue University, mprezzi@purdue.edu
Basu, D; Salgado, Rodrigo; and Prezzi, Monica, "A Continuum-Based Model for Analysis of Laterally Loaded Piles in Layered Soils"
(2009). Lyles School of Civil Engineering Faculty Publications. Paper 3.
http://docs.lib.purdue.edu/civeng/3
This document has been made available through Purdue e-Pubs, a service of the Purdue University Libraries. Please contact epubs@purdue.edu for
additional information.
Basu, D., Salgado, R. & Prezzi, M. (2009). Geotechnique 59, No. 2, 127140 [doi: 10.1680/geot.2007.00011
An analysis is developed to calculate the response of Dans la presente communication, on concoit une analyse
laterally loaded piles in multilayered elastic media. The permettant de calculer la reaction de pieux a charge laterale
displacement fields in the analysis are taken to be the dans des milieux elastiques multicouches. Les champs de
products of independent functions that vary in the verti- deplacement dans cette analyse sont censes etre les produits
cal, radial and circumferential directions. The governing de fonctions independantes variant dans les sens vertical,
differential equations for the pile deflections in different radial et circonferentiel. Les equations differentielles deter-
soil layers are obtained using the principle of minimum minantes pour les fleches des pieux dans differentes couches
potential energy. Solutions for pile deflection are obtained de terrain sobtiennent en appliquant le principe de lener-
analytically, whereas those for soil displacements are gie potentielle minimale. Les solutions sobtiennent de dif-
obtained using the one-dimensional finite difference ferentes facons : de facon analytique pour la fleche des
method. The input parameters needed for the analysis pieux, et avec une methode des differences finies unidimen-
are the pile geometry, the soil profile, and the elastic sionnelles pour les deplacements du sol. Les parametres
constants of the soil and pile. The method produces dentree necessaires pour lanalyse sont la geometrie des
results with accuracy comparable with that of a three- pieux, le profil du terrain, et les constantes elastiques du sol
dimensional finite element analysis but requires much less et des pieux. Cette methode permet dobtenir des resultats
computation time. The analysis can be extended to dont la precision est comparable a celle des analyses tridi-
account for soil non-linearity. mensionnelles aux elements finis, mais qui necessite des
temps de calcul beaucoup plus courts. En outre, il est
possible de renforcer cette analyse pour tenir compte de la
KEYWORDS: elasticity; piles; theoretical analysis non linearite du terrain.
INTRODUCTION et al., 1989; Verruijt & Kooijman, 1989; Trochanis et al., 1991;
Piles subjected to lateral forces and moments at the head are Bransby, 1999; Ng & Zhang, 2001; Klar & Frydman, 2002).
analysed in practice with the py method (Reese & Cox, The 3D FE or FD method can capture the most important
1969; Matlock, 1970; Reese et al., 1974, 1975; Reese & Van features of the complex pilesoil interaction, but three-dimen-
Impe, 2001). In the py method the pile is assumed to behave sional analyses are computationally expensive for routine prac-
as an EulerBernoulli beam with the soil modelled as a series tice. The BE method accounts for the pilesoil interaction by
of discretely spaced springs, each connected to one of the pile discretising the pile into small strips and modelling the inter-
segments into which the pile is discretised. The springs model action between these strips with the soil continuum through
the soil response to loading through py curves ( p is the unit numerical integration of Mindlins solution (Mindlin, 1936) for
resistance per unit pile length offered by the springs, and y is a point force within a continuum.
the pile deflection), which are developed empirically by Considering the soil surrounding the pile as a continuum,
adjusting the curves until they match actual loaddisplace- Sun (1994), Zhang et al. (2000) and Guo & Lee (2001)
ment results (Cox et al., 1974; Briaud et al., 1984; Yan & developed closed-form solutions based on linear elasticity
Byrne, 1992; Brown et al., 1994; Gabr et al., 1994; Briaud, that can be used to obtain lateral pile deflection with depth.
1997; Wu et al., 1998; Bransby, 1999; Ashour & Norris, Their analyses capture the three-dimensional aspects of the
2000). However, the py method often fails to predict pile interaction of the pilesoil system and produce results
response (Anderson et al., 2003; Kim et al., 2004), for it is quickly, which is advantageous in practice. However, these
not capable of capturing the complex three-dimensional inter- authors made an assumption that the variation of displace-
action between the pile and the soil. ments within the soil mass depends on the same displacement
The continuum approach is conceptually more appealing; function for both the radial and the circumferential directions.
however, in order to model the soil as a continuum, the use of This leads to a soil response that is stiffer than it is in reality.
numerical techniques such as the three-dimensional (3D) finite Most continuum analyses of laterally loaded piles do not
element (FE) method, finite elements with Fourier analysis, the consider soil layering. Soil heterogeneity with depth has been
boundary element (BE) method or the finite difference (FD) approximately taken into account in the BE and FE analyses by
method is often required (Poulos, 1971a, 1971b; Banerjee & assuming (typically) a linear variation of soil modulus with
Davis, 1978; Randolph, 1981; Budhu & Davies, 1988; Brown depth (Poulos, 1973; Randolph, 1981; Budhu & Davies, 1988).
The BE analysis has also been used to analyse two-layer
systems (Banerjee & Davies, 1978; Pise, 1982). However, BE
Manuscript received 18 February 2007; revised manuscript accepted analysis of laterally loaded piles is strictly not applicable to
24 September 2008. Published online ahead of print 3 December layered systems, because Mindlins solution used in BE analy-
2008. sis is valid only for homogeneous continuums. Verruijt &
Discussion on this paper closes on 1 August 2009, for further
details see p. ii.
Kooijman (1989) solved a layered elastic system by discretis-
* Department of Civil and Environmental Engineering, University ing the soil layers using FE and the pile by FD methods.
of Connecticut, Storrs, USA In this paper, an advanced continuum-based method of
School of Civil Engineering, Purdue University, West Lafayette, analysis of laterally loaded piles is proposed by assuming
USA. the soil displacement field to have a shape that is consistent
127
128 BASU, SALGADO AND PREZZI
with the drop in displacement expected as distance from the The soil medium is assumed to be an elastic and isotropic
pile increases, and with the fact that the displacement is continuum, homogeneous within each layer, with Lames
expected to depend on the direction of the load with respect constants s and Gs . There is no slippage or separation
to the point considered in the soil. The analysis considers a between the pile and the surrounding soil, or between the
pile embedded in a multilayered elastic soil (continuum), soil layers. The pile behaves as an EulerBernoulli beam
and rigorously takes into account the three-dimensional with a constant flexural rigidity Ep Ip .
pilesoil interaction. The governing differential equations
for the pile and soil displacements are developed using
variational principles. Closed-form solutions are obtained for Potential energy
pile deflection, and soil displacements are obtained using the The total potential energy of the pilesoil system, includ-
one-dimensional (1D) FD method. Pile response obtained ing both the internal and external potential energies, is given
using this method compares favourably with 3D FE analysis, by
although the computation effort required by this method is Lp 2 2 1 2 1
small. Because soil displacements and strains can be calcu- 1 d w 1
lated alongside pile deflection using this method, the analysis Ep I p dz pq pq rdrddz
2 0 dz 2 0 0 rp 2
forms the basis for future analysis that can model the inter-
action of piles in a group, and can account for soil non- 1 2 rp
1
linearity by relating the progressive degradation of soil pq pq rdrddz
Lp 0 0 2
stiffness to induced soil strains.
dw
Fa w M a
z0 dz z0
ANALYSIS (1)
Problem definition
We consider a pile with a circular cross-section of radius where w is the lateral pile deflection, and pq and pq are the
rp and length Lp embedded in a soil deposit that has n layers stress and strain tensors (see Fig. 2) in the soil (summation is
(Fig. 1). Each layer extends to infinity in all radial direc- implied by the repetition of the indices p and q in the product
tions, and the bottom (nth) layer extends to infinity in the of corresponding stress and strain components). The first
downward direction. The vertical depth to the base of any integral represents the internal potential energy of the pile.
intermediate layer i is Hi , which implies that the thickness The second and third integrals represent the internal potential
of the ith layer is Hi Hi1 with H0 0 and Hn 1. The
pile head is at the ground surface, and the base is embedded r0
in the nth layer. The pile is subjected to a horizontal force Fa
Fa and a moment Ma at the pile head such that Fa and Ma
are orthogonal vectors. In the analysis, we choose a cylind- ur
rical (rz) coordinate system with its origin coinciding u
uz
with the centre of the pile head and the positive z-axis
(coinciding with the pile axis) pointing downwards. The goal
of the analysis is to obtain pile deflection as a function of
depth caused by the action of Fa and/or Ma at the pile head.
z
Ma (a)
Pile
r0 r0
Fa
r
2rp
zz
Ma Fa r0
z
Layer 1 H1
H2 rz
Layer 2
Hi1
Hi
rr
Hn2 r
Lp r
Hn1
Layer i z zr
Layer n 1
Layer n Pile
z z
(b)
Fig. 1. Laterally loaded pile in layered elastic medium Fig. 2. (a) Displacements and (b) stresses within soil mass
ANALYSIS OF LATERALLY LOADED PILES IN LAYERED SOILS 129
energy of the continuum (note that the third integral repre- The strains in equation (3) are related to stresses using the
sents the energy of the column of soil with radius rp starting elastic stressstrain relationships, which allow expression of
at the pile base and extending to infinity downward, while the the soil potential energy density 12 pq pq in terms of the
second integral represents the energy of the soil surrounding displacement functions w(z), r (r) and (r), and the soil
both the pile and this column of soil). The remaining two elastic constants s and Gs (see Appendix 1). Substituting
terms represent the external potential energy. this expression for the potential energy density into equation
(1), we obtain
"
Lp 2 2 2
Soil displacement 1 d w 1 1 2 d r
We assume the following displacement fields (Fig. 2) in Ep I p dz s 2G s w
2 0 dz 2 2 0 rp dr
the soil:
u r w z r r cos d r r 2
(2a) 2s w 2 s 3Gs w 2 r 2
dr r r
u w z r sin (2b) 2 2
uz 0 (2c) 2 d 2 r d dw
Gs w 2Gs w Gs 2r
dr r dr dz
where w(z) is a displacement function (with a dimension of 2 # 2
length), varying with depth z, representing the deflection of dw 2 1 dw
the pile axis; r (r) and (r) are dimensionless displace- Gs rdrdz rp Gs
2
dz
dz 2 Lp dz
ment functions varying with the radial coordinate r; and is
the angle measured clockwise from a vertical reference dw
Fa w M a (4)
section (r r0 ) that contains the applied force vector Fa. dz z0
z0
Equation (2c) is based on the assumption that the vertical
displacement of the pile caused by the lateral load and Principle of minimum potential energy
moment applied at the pile head is negligible. A system in equilibrium exists with its potential energy at
The functions r (r) and (r) describe how the displace- a minimum. Hence minimising the potential energy of the
ments within the soil mass (due to pile deflection) decrease pilesoil system (i.e. setting the first variation of the poten-
with increasing radial distance from the pile axis. We set tial energy equal to 0) produces the equilibrium equa-
r (r) 1 and (r) 1 at r rp (this ensures compat- tions. We apply 0 to obtain an equation of the form
ibility at the pile/soil interface) and r (r) 0 and (r) (see Appendix 1)
0 at r 1 (this ensures that displacements in the soil
decrease with increasing radial distance from the pile). Thus dw
A ww B w
r and vary between 1 at the pile/soil interface and 0 at dz (5)
infinite radial distance from the pile.
C r r D 0
Since the variations w(z), (dw/dz), r (r) and (r) of
Stressstraindisplacement relationships the functions w(z) (and its derivative), r (r) and (r) are
The straindisplacement relationship, considering equation independent, the terms associated with each of these varia-
(2), leads to tions must individually be equal to zero (i.e. A(w)w 0,
2 3
@u r B(w)(dw/dz) 0, C( r ) r 0 and D( ) 0) in
6 7 order to satisfy the condition 0. The resulting equa-
6 @r 7
6 7 tions produce the optimal functions wopt (z), r,opt (r) and
2 3 6 u r 1 @u 7
rr 6 7 ,opt (r) that describe the equilibrium configuration of the
6 r r @ 7 pilesoil system.
6 7
6 7 6 6 7
6 7 6 @u z 7 While considering the terms of the variation of the
6 7 6 7
6 zz 7 6 @z 7 potential energy related to w, we do so for the following
6 76 7
6 7 6 1 @u r @u u 7 7
sub-domains: 0 < z < H1 , H1 < z < H2 , . . ., Hn1 < z <
6 r 7 6 7 Lp , and Lp < z , 1. Accordingly, w is forced to satisfy
6 7 6
4 rz 5 6 r @ @r r 7 7 equilibrium within each of these sub-domains, and hence
6 7 over the entire domain. For r and the domain over
z 6 @u z @u r 7
6 7 which the potential energy and its variation are calculated is
6 @ r @z 7
6 7 rp < r , 1.
4 1 @u z @u 5
2 r @ @z 3 (3)
d r r Soil displacement profiles
w z cos
6
6 dr 7
7 We first consider the variation of r (r). Referring back to
6
6
7
7 the equation 0, represented by equation (5), we first
6 r r r 7 collect all the terms associated with r and collectively set
6 w z cos 7
6 r 7 them equal to zero to obtain
6 7
6
6 0 7
7 1
d2 r d r d
6
6
7
7 ms1 r 2 ms2 ms3
6 w z r r d r 7 dr dr dr
6
r
sin 7
rp
6 r dr 7
6 7
6
6 dw z 7
7 ms4 r ms4 ns r r r dr (6)
6 r r cos 7 r r
6 dz 7 1
6 7 d r
4 5
dw z
r sin ms1 r ms3 r ms3 r 0
dz dr rp
130 BASU, SALGADO AND PREZZI
n Hi
where X d4 w i d2 w i
Ep I p 4 2t i 2 k i w i w i dz
1 i1 H i1
dz dz
X
n1 Hi
ms1 s 2Gs w 2 dz si 2Gsi w2i dz
0 H i1 1
i1 d2 w n1
(7) 2t n1 k n w n1 w n1 dz
Lp dz 2
1 X
n1 Hi
ms2 Gs w 2 dz Gsi w2i dz (8)
d3 w1 dw1
0 i1 H i1
Ep I p 2t F w
1 Hi dz 3 1
dz
a 1
X
n1 z0
ms3 s w 2 dz si w2i dz (9)
0 H i1
i1
d2 w1 dw1
1 Hi Ep I p Ma
X dz z0
n1
dz 2
ms4 s 3Gs w 2 dz si 3Gsi w2i dz
0 H i1
i1 "
(10) d3 w1 dw1
Ep I p 2t w
1 2 X
n1 H i 2 dz 3 1
dz
1
dw dw i z H1
n s Gs dz Gsi dz (11)
0 dz H dz #
i1 i1
d3 w2 dw2
Ep I p 2t2
w2
The subscript i in the above equations refers to the ith layer dz 3 dz z H1
of the multilayered continuum (Fig. 1); w i represents the
" (14)
function w(z) in the ith layer with w i j z H i w i1 j z H i . Note
that the nth (bottom) layer is split into two parts, with the d2 w1 dw1
Ep I p
part below the pile denoted by the subscript n+1; therefore, dz 2 dz z H1
in the analysis, Hn Lp and Hn1 ! 1.
The last term on the left-hand side of equation (6) is a #
multiple of the subtraction of the value of r at r rp d2 w2 dw2
Ep I p ... :
from the value of r at r 1, and is therefore identically dz 2 dz z H1
zero for the boundary conditions of our problem ( r 0 at
r 1 and r 1 at r rp ) because a known (or "
prescribed) r implies that its variation r 0. After this d3 w n dw n
term is made equal to zero, what is left is an equation of Ep I p 3
2t n w n
dz dz z Lp
form C( r ) r 0. The function r (r) has a non-zero
variation (i.e. r 6 0) for rp , r , 1 because r is not #
known a priori in this interval, so C( r ) 0, which means dw n1
the integrand in equation (6) must be set to zero, leading to 2t n1 w n1
dz z Lp
the differential equation
" # 2
d2 r 1 d r 1 2 2 2 23 d 1 d2 w n dw n
Ep I p
dr 2 r dr r rp r
r dr r dz 2 dz z Lp
(12)
dw n1
2t n1 w n1 0
where the s are dimensionless constants given by 21 dz z1
ms4 =ms1 , (2 =rp )2 ns =ms1 and 23 (ms2 ms3 )=ms1 .
When solved, equation (12) yields r,opt . where
We now consider the variation of (r). We collect the 8 " #
1
terms containing in the equation 0 (equation (5)) >
>
and, proceeding similarly as for r , we get the following
>
> Gsi r rdr ;
2 2
i 1, 2, . . ., n
< 2 rp
governing differential equation for : ti "
1
#
>
>
" # 2 >
> G
: 2 sn r 2
r 2
rdr r 2
p ; i n1
d2 1 d 4 2 5 2 26 d r 4 p
r
dr 2 r dr r rp r dr r (15)
" 1 2 1 2
(13) d r d
k i si 2Gsi r dr Gsi r dr
dr dr
with the following boundary conditions: 0 at r 1 rp rp
Infinite soil
w~1 constant (21a) B columns of
infinitesimal
or thickness
providing
.
A
d3 w~1 d w~1 resistance to pile
2 ~t1 F~a 0 (21b) movement
z3
d~ d~z
and
Soil columns get
d w~1 compressed (or
constant (21c)
d~z extended) owing to pile
movement from, say,
or point A to point B
(accounted for by ki)
d2 w~1 ~a 0
M (21d) Shear resistances between
z2
d~
soil columns below the
At the interface between any two layers (z Hi or z~ H~ i ), pile produce pile base shear
(occurs only if pile base
w~i w~i1 (22a) deflects laterally)
d w~i d w~i1 Fig. 3. Illustration of the two sources of soil resistance: soil
(22b)
d~z d~
z compression and shear
132 BASU, SALGADO AND PREZZI
v
s
parameter t i (with dimension of F) determines the fraction u H~ i
uX n ~t n1 2
of the soil resistance due to the shear stresses that develop u
u si 3Gsi w~ i d~z s n 3Gs n
2
~ n j~z1
w
between soil layers of infinitesimal thickness displaced dif- u i1 H~ i1 2 k~ n
4 u u s
ferentially in the lateral direction (Vlasov & Leontev, 1966). H~ i
The boundary conditions (equations (22a)(22d)) at the u X n ~t n1 2
t Gsi w 2
~ i d~z Gs n w~ n j~z1
interface of any two layers (~ ~
z H i ) ensure the continuity of ~ i1
H 2 k~ n
i1
pile deflection, slope of the deflection curve ( dw i /dz),
bending moment ( Ep I p (d2 w i =dz 2 )) and shear force (27)
3 3 v
s
( Ep I p (d w i =dz ) 2t i (dw i =dz)) (the shear force has con- u H~ i 2
u X n
tributions from both pile flexure and soil deformation). At u dw ~i k~ n 2
z 0) the shear force equals the applied force u G d~z Gs n w~ j
the pile head (~ u 1 i1 si H~ i1 d~z 8 ~t n1 n ~z1
(equation (21b)), and either the slope of the pile deflection u
5 u 2 s (28)
u X n H~ i
curve is a known constant (equation (21c)) (this boundary t 2
~t n1 2
condition is generally used with the value of slope taken Gsi w~ i d~z Gs n ~ n j~z1
w
i1 ~ i1
H 2 k~ n
equal to zero when fixed-head conditions are used to idealise
v
s
the case of a pile that is part of a group of piles joined at u H~ i
uX n ~t n1 2
the head by a cap) or the pile bending moment is equal to u
the applied moment (equation (21d)) (free-head case). Equa- u si Gsi ~ i d~z s n Gs n
w 2
~ n j~z1
w
u i1 ~ i1
H 2 k~ n
tion (21a) must be used in the analysis instead of equation 6 u u s
u X n H~ i
(21b) to estimate the magnitude of an applied force required t 2
~t n1 2
to produce a given (known) head deflection. Similarly, if, for Gsi w~ i d~z Gs n ~ n j~z1
w
i1 H~ i1 2 k~ n
both the free- and fixed-head cases, it becomes necessary to
estimate the magnitude of an applied moment that produces (29)
a given (known) slope at the head, equation (21c) must be
used. where Lp /rp. These expressions can be directly used in
At the pile base (~ z 1), either the pile deflection is set the computations.
(equation (23a)) (used for the ideal fixed-base case, for
which the deflection is taken equal to zero, which may be
used with satisfactory results if the pile is socketed into a ANALYTICAL SOLUTION FOR PILE DEFLECTION
very firm layer, like hard rock) or the shear force at a The general solution of equation (20) is given by
section infinitesimally above the pile base is equal to that
infinitesimally below (equation (23b9)) (free-base case). The z) C (i)
w~i (~ (i)
1 1 C 2 2 C 3 3 C 4 4
(i) (i)
(30)
other boundary condition active at the pile base is that either
the slope is a constant (equation (23c)) (assumed to be zero where C (i) (i) (i) (i)
1 , C 2 , C 3 and C 4 are integration constants (for the
for the fixed-base case) or the pile bending moment is zero ith layer), and 1 , 2 , 3 and 4 are individual solutions
(equation (23d)) (free-base case). (functions of z~) of the differential equation. The functions 1 ,
2 , 3 and 4 are standard trigonometric or hyperbolic func-
tions that arise in the solution of the linear ordinary differential
Expressions for the s in terms of dimensionless deflections equations (Table 1). The integration constants for each layer
The s appearing in equations (12) and (13) are expressed can be determined using the boundary conditions. The bound-
in terms of the dimensionless pile deflection and slope as ary conditions given in equations (21)(23) lead to a system
follows: of linear algebraic equations (see Appendix 2) of the form
v
s C F (31)
u H~ i
uX n ~t n1 2
u where []4 n34 n is a matrix containing the functions 1 , 2 ,
u si 3Gsi w~ i d~z s n 3Gs n
2
~ n j~z1
w
u i1 ~ i1
H 2 k~ n 3 and 4 calculated at the boundaries of the soil layers,
u
1 u s [C]4 n31 is the vector of unknown integration constants of all
uX n H~ i
t si 2Gsi ~t n1 2 the layers, and [F]4 n31 is the right-hand side vector contain-
w~ i d~z s n 2Gs n
2
~ n j~z1
w
i1 ~ i1
H 2 k~ n ing the applied forces and/or displacements (the subscript 4n
denotes the number of equations, which is four times the
(24) number of soil layers). Simultaneous solution of the system
2 of equations represented by equation (31) produces the values
v of the integration constants C (i) (i) (i) (i)
1 , C 2 , C 3 and C 4 , which,
u s
u X n H~ i 2 when substituted in equation (30), produce the particular
u dw ~i k~ n 2 solution of pile deflection (i.e. the pile deflection profile) for
u Gsi d~z Gs n w~ j
u1 ~ i1
H d~z 8~t n1 n ~z1 a given set of boundary conditions and applied loads. The
u i1
s
u2 n H~ i slope of the deflected pile axis, and the bending moment and
u X ~t n1 2
t si 2Gsi ~ i d~z s n 2Gs n
w 2
w~ n j~z1 shear force in the pile, can be obtained as a function of depth
i1 ~
H i1 2 k~ n by successively differentiating equation (30) and using the
values of the integration constants.
(25)
v
s
u H~ i
u X n ~t n1 2
u FINITE DIFFERENCE SOLUTION FOR SOIL
u si Gsi w~ 2i d~z s n Gs n ~ n j~z1
w
u i1 ~ i1
H 2 k~ n DISPLACEMENTS
u
3 u s
H~ i The differential equations (12) and (13) for r and are
uX n ~t n1 2 solved using the FD method. The equations are interdepen-
t si 2Gsi w~ i d~z s n 2Gs n
2
~ n j~z1
w
i1 ~ i1
H 2 k~ n dent and must, as a result, be solved simultaneously. Using
the central-difference scheme, equations (12) and (13) can
(26) be respectively written as
ANALYSIS OF LATERALLY LOADED PILES IN LAYERED SOILS 133
rj1 2 rj rj1 1 rj1 rj1
z sin b~
rj 3
z
j
(a2 b2 )4
a3 b1
2ab2
cosh b~z
rj rp rj 2r rj
b 4
b3 3
b3
4
sinh a~
2
(32)
j1 2j j1 1 j1 j1
r 2 rj 2r
" 2 # 2
4 2 5 2 j1 rj1 4
j 6 r rj
rj rp rj 2r rj
a(a2 3b2 )4 b(b2 3a2 )2
(33)
z sin b~
z
(a2 b2 )3
a4 b2
2ab1
b 3
b3 4
cosh a~
Individual solutions of equation (20)
r
rj
Pile
r r r r
z cos b~
z
(a2 b2 )2
a1 b3
3 j1
2ab4
Node number
cosh a~
j
z
1 2 m1 m
a2 2
a3 1
a1
2
cosh a~
Normalise input parameters: calculate Fa, Ma, Hi,
z cos b~
z
(a2 b2 )1
a2 b4
2ab3
sinh a~
z
a2 1
a3 2
a2
1
sinh a~
Table 1. Functions in equation (30) for piles crossing multiple soil layers
Calculate r and
Calculate ki and ti
derivatives
Functions
and their
90
90
9
0
9
0
dw i
Calculate w i,
dz
q
~t ( ~t 2 k~) ~t ( ~t 2 k~)
q r
k~ ~t)
q
Input
2(
old new
1 1 ,
1
No Check if
old
2 2 ,
new
| old new
1/2/.../6 1/2/.../6| 0001
r
....,
q
k~ ~t)
old new
6 6
p
Yes
a
2(
dw i
Store w i, as the final values
dz
magnitudes
of k~ and ~t
Relative
End
k~ . ~t 2
k~ , ~t 2
2
The non-zero elements of the left-hand side matrix 26 rj1 rj1 4
Fj rj
[K r ] m3 m in the above equation are given by rj 2r rj (43)
1 1 j 3, 4, . . . , m 2
K j, rj1 (35)
r 2 2r j r 2
" 1 1 26 (3)
r 1 4
2 #
F2 (2) (44)
2 1 2 2 r 2 2r2 r r2 2r r2 r
K j, rj 2 (36)
r rj rp
2 rm2 4 2 m1
F m1 6 r (45)
1 1 r m1 2r r m1
K j, rj1 (37)
r 2 2r j r
in which the subscript j is valid for nodes 2 to m 1, with
r
Since the right-hand side vectors fF r g m31 and
the exception that K 2,1r 0 and K m1, m 0 (as is evident fF g m31 contain the unknowns and r , iterations are
from equation (34)). necessary to obtain their values. An initial estimate of rj is
The elements of the right-hand side vector fF r g m31 in made and given as input to fF g, and j is determined by
equation (34) are given by solving equation (41). The j values are then given as input
j1 j1 2 to fF r g to obtain rj from equation (34). The newly
2 1 obtained values of rj are again used to obtain new values
Fj r 3 j (38)
rj 2r rj of j , and the iterations areP continued
until convergence
is
reached. The criteria m1 mj1 rj previous rj current < 106
P
where j represents nodes 3 through m2. The elements and m1 mj1 j previous j current < 106 are used (a strin-
corresponding to node 2 and m1 are given by: gent value of 106 is used because this iterative solution
(3) 2 scheme is central to another set of iterations described next)
1 1 2 1 1 to ensure that accurate values of r and are obtained.
F2 r 2 3 (2)
(39)
r 2r2 r r2 2r r2
r 2 m2 1 2 m1
F m1 3 (40)
r m1 2r r m1
SOLUTION ALGORITHM
In order to obtain pile deflections by solving equation
Using equation (33) and the boundary conditions that (20), the soil parameters k i and t i must be known. However,
(1) ( m)
1 (at r rp ) and 0 (at r 1), a matrix these soil parameters depend on r and , which are not
equation (similar to equation (34)) for can also be known a priori. Hence an iterative algorithm (separate from
formed for the discretised nodes: the iterations between r and described in the previous
K f g f F g (41) section) is necessary to solve the problem. First, initial
guesses for 1 to 6 are made, and for these assumed values
r and are determined using the iterative technique
The number and positioning of the non-zero elements of described in the previous section. Using the calculated
[K ] m3 m in equation (41) are exactly the same as that of values of r and , k i and t i are calculated by numerical
[K r ] m3 m of equation (34). The expressions of the off- integrations (with r of Fig. 4 as the step length). Using the
diagonal elements of [K ] m3 m and [K r ] m3 m are also the values of k i and t i , the pile deflection is calculated. From
same (i.e. K p,q
r K for p 6 q). The diagonal elements of
p,q the calculated values of pile deflection and slope of the
[K ] m3 m for j 2 to m 1 are given by deformed pile, 1 to 6 are obtained. The new values of 1
" 2 # to 6 are then used to recalculate r and , and so on. The
r 2 4 2 5 entire process is repeated until convergence on each of
K j, j 2 (42)
r rj rp the s is attained. The tolerance limit prescribed on
the s between the pth and ( p+1)th iteration is
The structure of fF g m31 in equation (41) is also similar j(1=2=...=6
p1)
(1=2=...=6
p)
j , 0:001. The details of the solution
to fF r g m31 of equation (34) with F 1 1 and F m 0. steps are given in the form of a flow chart in Fig. 5. We
The remaining elements of fF g m31 are given by chose an initial guess of one for all the s, but any other
ANALYSIS OF LATERALLY LOADED PILES IN LAYERED SOILS 135
choice would produce results with the same level of accu- Poissons ratio of soil is greater than 0.3. This artificially
racy at approximately the same computation time. stiff pile response is not observed in our analysis. To
illustrate this point, we present below two examples compar-
ing the results of our analysis, the analysis based on the
BENEFITS OF THE PRESENT ANALYSIS displacement assumption of Sun (1994), and 3D FE analysis.
This analysis is an improvement over the analysis of Sun We consider as an illustration of use of the analysis a
(1994) for laterally loaded piles in homogeneous soil on at 15 m long drilled shaft, with a diameter of 0.6 m and pile
least two accounts: (a) our assumption of the displacement modulus Ep 24 3 106 kN/m2 , embedded in a four-layer
field is more general and more realistic than that assumed soil deposit with H1 2.0 m, H2 5.0 m, and H3 8.3 m;
by Sun (1994), who chose r (r) (r) (r) for both Es1 20 MPa, Es2 35 MPa, Es3 50 MPa and Es4
the displacements u r and u (equation (2)); and (b) we 80 MPa; s1 0.35, s2 0.25, s3 0.2 and s4 0.15
obtained solutions for a multilayered soil, whereas the solu- (Es i and s i are the soil Youngs modulus and Poissons ratio
tion of Sun (1994) is valid only for a single soil layer. for the ith layer; Es i and s i are related to s i and Gs i by
The need for an improved form for the displacement field si Esi si =(1 si )(1 2si ) and Gsi Esi =2(1 si )). A
(equation (2)) arises from the fact that the displacement horizontal force Fa 300 kN acts on the pile. The pile head
assumption of Sun (1994) produces zero displacement in the and base are free to deflect and rotate. Fig. 7 shows the pile
soil mass perpendicular to the direction of the applied force deflection profile obtained using our analysis, the analysis
Fa (Fig. 6). Consequently, the resultant displacement vector based on the displacement assumption of Sun (1994), and a
u e r u r e u (e r and e are the unit basis vectors in the 3D finite element analysis (FEA). The pile response obtained
radial and tangential directions respectively) at any point from our analysis closely matches that of the 3D FEA (a
within the soil mass is forced to be parallel to the applied difference of 9.6% in the head deflection was observed
force Fa. Thus the displacement field in the soil mass between our analysis and the FEA); the analysis of Sun
(which, in general, has a component perpendicular to the (1994) produces a stiffer pile response.
direction of Fa ) is artificially constrained, the normal strain Next, we consider a large-diameter drilled shaft, 40 m
in the circumferential direction (equation (3)) becomes long, with a diameter of 1.7 m and Ep 25 3 106 kPa,
zero, and the pile response is stiffer than what it is in reality. embedded in a four-layer soil profile with H1 1.5 m, H2
In fact, Guo & Lee (2001) found that the Sun (1994) 3.5 m, and H3 8.5 m; Es1 20 MPa, Es2 25 MPa,
analysis produces unreliable pile response, particularly if the Es3 40 MPa and Es4 80 MPa; s1 0.35, s2 0.3, s3
0.25 and s4 0.2. A 3000 kN force acts at the pile head,
Displacement which is free to deflect and rotate. Fig. 8 shows the pile
vectors deflection profiles, as obtained from our analysis, the analy-
sis based on the displacement assumption of Sun (1994),
and 3D FEA. As before, our results match those of the FEA
Ma
more closely than the results based on the Sun (1994)
assumption; the difference in the head deflection obtained
Fa
from our analysis and FEA is 6.6%.
The 3D FE analyses were performed using ABAQUS. The
domain for these analyses can be visualised as a cylinder of
soil mass containing the pile at its centre as a concentric
cylinder. The top (horizontal) surface of the soil cylinder
was flush with the pile head, and the bottom (horizontal)
Pile
surface was located at a finite distance below the pile base
r0, x ux (thus the soil mass below the pile base participating in the
u
pilesoil interaction was incorporated in the analysis). The
uy ur horizontal force Fa (acting at the pile head) was applied as a
y r uniformly distributed shear stress (i.e. force per unit pile
(a) cross-section area) acting on the pile-head surface (the
Pile deflection: mm
Displacement
2 0 2 4 6 8 10
vectors 0
Ma 25
Fa
50
Depth, z: m
75
Pile ux 100
r0, x
u
uy ur Present analysis
125
y r Sun (1994)
Finite element analysis
(b)
150
Fig. 6. Displacement field in soil: (a) according to Sun (1994)
assumption; (b) according to assumption made in this paper Fig. 7. Deflection profile of a 15 m long pile
136 BASU, SALGADO AND PREZZI
Pile deflection: mm
5 0 5 10 15 20 25 pile load test performed by McClelland & Focht (1958). The
0 length (Lp ) and radius (rp ) of the pile are 23 m and 0.305 m,
and the pile was embedded in a normally consolidated clay.
The pile was acted upon by a lateral force Fa 300 kN and
a negative moment Ma 265 kNm at the head. Randolph
10
(1981) back-calculated the pile modulus Ep as 68.42 3
106 kN/m2 from the reported pile flexural rigidity. Randolph
(1981) further suggested, based on back-calculation of test
results to match his FEA (coupled with Fourier series), that
Depth, z: m
the soil shear modulus profile for this soil deposit can be
20 represented as Gs 0.8z 3 103 kN/m3 with s 0.3. We
divided the soil profile into four layers and calculated the
shear modulus at the middle of each layer, which were
considered the representative values for each layer (Table 2).
30 Present analysis
Using these values of soil modulus, we calculated the pile
deflection profile using both our analysis and that based on
Sun (1994)
the assumption of Sun (1994). Fig. 9 shows the pile
Finite element analysis
responses. Also plotted are the measured pile response and
that obtained by Randolph (1981). Our analysis produces a
40 pile deflection profile that closely matches the measured
profile.
Fig. 8. Deflection profile of a 40 m long drilled shaft
We now investigate how an explicit incorporation of soil
layering can be useful in obtaining proper pile response. For
distributed shear stress multiplied by the pile cross-section that purpose, we studied the response of two piles a short
area produced Fa ). The vertical plane passing through the stubby pile with Lp 10 m, rp 0.5 m and Ep 25 3
pile axis parallel to Fa is a plane of symmetry (the plane 106 kN/m2 and a long slender pile with Lp 20 m, rp
contains the Fa vector) and divides the cylindrical domain 0.25 m and Ep 25 3 106 kN/m2 for various soil profiles.
into two equal and symmetrical halves. Only one such half Both piles are subjected to a horizontal force Fa 1000 kN,
was used as the analysis domain. Different boundary condi- and both are assumed to be free at the head and base.
tions were prescribed at different boundaries of the FE For the short pile (Lp 10 m), we consider the following
domain: all components of displacements were assumed to cases:
be zero along the bottom (horizontal) surface and along the
outer, curved (vertical) surface of the soil domain; on the (a) a homogeneous soil layer with Gs 25 MPa
(vertical) boundary surface created by the plane of symme- (b) a two-layer system with H1 2 m, Gs1 25 MPa and
try, the displacement perpendicular to the boundary was Gs2 50 MPa
assumed to be zero. A perfect contact (with no slippage or
separation) between the pile and the surrounding soil was Table 2. Soil properties at the pile load test site of McClelland
assumed. The radial distances of the outer curved (vertical) & Focht (1958)
boundary of the soil domain from the pile axis were taken
as 20 m and 25 m for the 15 m and 40 m piles respectively; Depth: m Extent of soil layers: m Shear modulus, Gs : MPa
the corresponding vertical distances from the pile base to
2. 0 0 to 4.0 1.6
the bottom (horizontal) boundary of the soil domain were 6. 0 4.0 to 8.0 4.8
5 m and 20 m. Twenty-noded brick elements were used to 10.0 8.0 to 12.0 8.0
represent both the pile and the soil for both the problems. 17.5 12.0 to great depth 14.0
The element size in the pile and at the pile/soil interface
was approximately 0.1 m for both the examples, and was
increased gradually with increasing radial distance from the Pile deflection: mm
pile axis to 2.0 m (for the 15 m pile) and 3.8 m (for the 5 0 5 10 15 20 25
0
40 m pile) at the outer curved boundary of the soil domain.
The number of degrees of freedom used for the 15 m pile
was 56 653, and that used for the 40 m pile was 90 564. The
optimal domains and meshes described above were obtained 5
by ensuring that there were no boundary effects and by
performing convergence checks.
The CPU run times of the 3D FE analyses (run in a 16-
core x86 server containing eight 2.6 MHz dual-core Opteron 10
Depth, z: m
The Poissons ratio was kept constant at 0.25 for all the 2
cases and for all the layers. Figs 10(a), 10(b) and 10(c) show
the pile deflection, bending moment and shear force profiles.
Next, we consider the long pile (Lp 20 m) and obtain 4
Depth, z: m
the pile response for the following cases:
(a) a homogeneous soil layer with Gs 10 MPa
(b) a four-layer system with H1 1 m, H2 3 m, H3 6 (1) Gs2/Gs1 10
5 m, Gs1 10 MPa, Gs2 20 MPa, Gs3 40 MPa and
Gs4 80 MPa (2) Gs2/Gs1 20
(c) a four-layer system with H1 1 m, H2 3 m, H3
(3) Gs2/Gs1 40
5 m, Gs1 10 MPa, Gs2 40 MPa, Gs3 40 MPa and 8
Gs4 80 MPa (4) Gs2/Gs1 05
The Poissons ratio was again assumed to be 0.25 for all the
cases and for all the layers. Figs 11(a) and 11(b) show the 10
pile deflection and bending moment profiles for the above (a)
cases respectively.
The effect of soil layering on lateral pile response is Bending moment: kN m
evident from Figs 10 and 11. The modulus and thickness of 200 0 200 400 600 800 1000
0
different soil layers (particularly those near the pile head)
have a definite effect on lateral pile response. The examples
show that proper characterisation of soil deposits and expli- H1/Lp 02
cit accounting for the different layers are necessary for 2
accurate prediction of pile response and optimal design of
laterally loaded piles. Using our analysis, the three-dimen-
sional interaction between pile and soil can be explicitly
4
accounted for with full consideration of soil layering. The
Depth, z: m
10
CONCLUSIONS (c)
An advanced method of analysis for a single, circular pile
embedded in a multilayered elastic medium and subjected to Fig. 10. Effect of layering in a two-layer soil: (a) pile deflection,
a horizontal force and a moment at the head was presented. (b) bending moment and (c) shear force of a 10 m long pile
138 BASU, SALGADO AND PREZZI
Pile deflection: mm
10 0 10 20 30
determine the rate at which the displacements in the soil
0 medium decay with increasing radial distance from the pile
axis. These parameters are not known a priori and must be
determined iteratively. Hence an iterative scheme was devel-
4 oped and coded to obtain solutions for a variety of boundary
conditions and soil profiles. Notwithstanding the iterations
on the s, the solutions are obtained in seconds.
8
Illustrations of use of the analysis for layered soils show
Depth, z: m
20
Bending moment: kN m
200 0 200 400 600
0
APPENDIX 1
The potential energy density of the soil can be expressed in terms
4 of the displacement functions and elastic constants as
" 2
1 1 d r
8 pq pq s 2Gs w 2 cos2
Depth, z: m
2 2 dr
d r r
2s w 2 cos2
12 dr r
r 2 2
H1/Lp 005 s Gs w 2 2
cos2 Gs w 2 r 2
r r
16 H1/H2 033 2
H1/H3 020 d d 2
Gs w 2 sin2 2Gs w 2 r sin
dr r dr
20 2 2 #
dw dw
Gs r cos Gs
2 2
sin
2 2
(46)
(1) Gs1 Gs2 Gs3 Gs4 dz dz
(48)
APPENDIX 2
The expanded form of the matrix [] in equation (31) is given by
2 3
(Head)
z~0
0234 0234
6 7
6 (1) (2)
z~ H~ 1 0434 0434 7
6 z~ H~1 7
6 (2)
z~ H~ (3)
z~ H~ 0434 0434 7
6 0 434 7
6 2 2 7
4 n34 n 6
6
..
. 0
..
.
..
.
..
.
..
.
7
7 (49)
6 434 7
6 .. .. .. .. .. 7
6 . . . . . 0 7
6 434 7
6 0 0434
( n1)
0434 z~ H~ z~ H~ n1 7
( n)
4 434 n1
5
0234 0234 0234 (Base)
z~1
where [](i) ~
~ l (i 1, 2, . . . n) is a 4 3 4 matrix valid for the ith soil layer. It contains the parameter t i , the functions
z~ H
1 , 2 , 3 , 4 and
their derivatives (Table 1) calculated at the layer interface z~ H ~ l (l i1 or l i). In its expanded form, [](i)
~ l can be expressed as
z~ H
2 3
1 2 3 4
6
6 91 92 93 94 7
7
(i) ~l 4 5 (50)
z~ H 10 20 30 40
1 2 t i -
~ 1 1 2 t i 2- 1 2 t i 3- 1 2 t i 4- z~ H~ l
~ ~ ~
in which the functions 1 , 2 , 3 , 4 and their derivatives are calculated at z~ 0 (pile head). The 2 3 4 matrix [](Base)
z~1 is given by
8
> 1 2 3 4
>
> ; fixed-base condition
>
> 91 92 93 94 z~1
>
>" q q
>
<
1- 2 ~t n 91 2 k~n ~t n1 1 2- 2 ~t n 92 2 k~n ~t n1 2
(Base)
z~1 (52)
> 10 20 #
>
> q q
>
>
>
> -3 2~t n 93 2 k~n ~t n1 3 - t n 94 2 k~n ~t n1 4
4 2~ ; free-base condition
>
:
30 40 z~1
in which the functions 1 , 2 , 3 , 4 and their derivatives are calculated at z~ 1 (pile base). The matrices [0]234 and [0]434 have
respectively two rows and four columns and four rows and four columns, and contain 0 as all the elements.
140 BASU, SALGADO AND PREZZI
The vector [C] (with a dimension of 4n) in equation (31) is given Nak-Dong river sand. J. Geotech. Geoenviron. Engng ASCE
by: 130, No. 4, 416425.
h Klar, A. & Frydman, S. (2002). Three-dimensional analysis of
C C (1)
1 C (1)
2 C (1)
3 C (1)
4 C (2)
1 C (2)
2 C (2)
3 C (2)
4 lateral pile response using two-dimensional explicit numerical
T scheme. J. Geotech. Geoenviron. Engng ASCE 128, No. 9, 775
C (1n) C (2n) C (3n) C (4n) 784.
Matlock, H. (1970). Correlations for design of laterally loaded piles
(53)
in soft clay. Proc. 2nd Offshore Technol. Conf., Houston, TX 1,
The superscript T implies the transpose of a matrix. 577594.
The right-hand side vector of equation (31) is given by McClelland, B. & Focht, J. A. Jr (1958). Soil modulus for laterally
loaded piles. Trans. ASCE 123, 10491063.
F4 n31 F~a F2 0 0 T (54) Mindlin, R. D. (1936). Force at a point in the interior of a semi-
infinite solid. Physics 7, May, 195202.
~ a for the free-
with F2 0 for the fixed-head condition and F2 M Ng, C. W. W. & Zhang, L. M. (2001). Three-dimensional analysis
head condition. of performance of laterally loaded sleeved piles in sloping
ground. J. Geotech. Geoenviron. Engng ASCE 127, No. 6, 499
509.
Pise, P. J. (1982). Laterally loaded piles in a two-layer soil system.
REFERENCES J. Geotech. Engng Div. ASCE 108, No. GT9, 11771181.
Anderson, J. B., Townsend, F. C. & Grajales, B. (2003). Case Poulos, H. G. (1971a). Behavior of laterally loaded piles: I single
history evaluation of laterally loaded piles. J. Geotech. Geoen- piles. J. Soil Mech. Found. Div. ASCE 97, No. SM5, 711731.
viron. Engng ASCE 129, No. 3, 187196. Poulos, H. G. (1971b). Behavior of laterally loaded piles: III
Ashour, M. & Norris, G. (2000). Modeling lateral soilpile re- socketed piles. J. Soil Mech. Found. Div. ASCE 98, No. SM4,
sponse based on soilpile interaction. J. Geotech. Geoenviron. 341360.
Engng ASCE 126, No. 5, 420428. Poulos, H. G. (1973). Loaddeflection prediction for laterally
Banerjee, P. K. & Davies, T. G. (1978). The behaviour of axially loaded piles. Aust. Geomech. J. G3, No. 1, 18.
and laterally loaded single piles embedded in nonhomogeneous Randolph, M. F. (1981). The response of flexible piles to lateral
soils. Geotechnique 28, No. 3, 309326. loading. Geotechnique 31, No. 2, 247259.
Bransby, M. F. (1999). Selection of py curves for the design of Reese, L. C. & Cox, W. R. (1969). Soil behavior from analysis of
single laterally loaded piles. Int. J. Numer. Anal. Methods tests of uninstrumented piles under lateral loading. In Perform-
Geomech. 23, No. 15, 19091926. ance of deep foundations, ASTM STP 444, pp. 160176.
Briaud, J.-L. (1997). Sallop: simple approach for lateral loads on Philadelphia, PA: ASTM.
piles. J. Geotech. Geoenviron. Engng ASCE 123, No. 10, 958 Reese, L. C. & Van Impe, W. F. (2001). Single piles and pile
964. groups under lateral loading. Rotterdam: A. A. Balkema.
Briaud, J.-L., Smith, T. & Meyer, B. (1984). Laterally loaded piles Reese, L. C., Cox, W. R. & Koop, F. D. (1974). Analysis of
and the pressuremeter: comparison of existing methods. In laterally loaded piles in sand. Proc. 6th Offshore Technol. Conf.,
Laterally loaded deep foundations (eds J. A. Langer, E. T. Houston, TX 2, 473483.
Mosley and C. C. Thompson), ASTM STP 835, pp. 97111. Reese, L. C., Cox, W. R. & Koop, F. D. (1975). Field testing and
Philadelphia, PA: ASTM. analysis of laterally loaded piles in stiff clay. Proc. 7th Offshore
Brown, D. A., Shie, C. & Kumar, M. (1989). Py curves for Technol. Conf., Houston, TX 2, 671690.
laterally loaded piles derived from three-dimensional finite ele- Sun, K. (1994). Laterally loaded piles in elastic media. J. Geotech.
ment model. Proc. 3rd Int. Symp. on Numerical Models in Engng ASCE 120, No. 8, 13241344.
Geomechanics (NUMOG III), Niagara Falls, 683690. Trochanis, A. M., Bielak, J. & Christiano, P. (1991). Three-dimen-
Brown, D. A., Hidden, S. A. & Zhang, S. (1994). Determination of sional nonlinear study of piles. J. Geotech. Engng ASCE 117,
py curves using inclinometer data. Geotech. Test. J. 17, No. 2, No. 3, 429447.
150158. Verruijt, A. & Kooijman, A. P. (1989). Laterally loaded piles in a
Budhu, M. & Davies, T. G. (1988). Analysis of laterally loaded layered elastic medium. Geotechnique 39, No. 1, 3946.
piles in soft clays. J. Geotech. Engng Div. ASCE 114, No. 1, Vlasov, V. Z. & Leontev, N. N. (1966). Beams, plates and shells on
2139. elastic foundations. Jerusalem: Israel Program for Scientific
Cox, W. R., Reese, L. C. & Grubbs, B. R. (1974). Field testing of Translations.
laterally loaded piles in sand. Proc. 6th Offshore Technol. Conf., Winkler, E. (1867). Die Lehre von der Elasticitaet und Festigkeit.
Houston, TX 2, 459472. Prague: Dominicus.
Gabr, M. A., Lunne, T. & Powell, J. J. (1994). Py analysis of Wu, D., Broms, B. B. & Choa, V. (1998). Design of laterally loaded
laterally loaded piles in clay using DMT. J. Geotech. Engng piles in cohesive soils using py curves. Soils Found. 38, No. 2,
ASCE 120, No. 5, 816837. 1726.
Guo, W. D. & Lee, F. H. (2001). Load transfer approach for Yan, L. & Byrne, P. M. (1992). Lateral pile response to monotonic
laterally loaded piles. Int. J. Numer. Anal. Methods Geomech. pile head loading. Can. Geotech. J. 29, 955970.
25, No. 11, 11011129. Zhang, L., Ernst, H. & Einstein, H. H. (2000). Nonlinear analysis
Kim, B. T., Kim, N.-K., Lee, W. J. & Kim, Y. S. (2004). of laterally loaded rock-socketed shafts. J. Geotech. Geoenviron.
Experimental load-transfer curves of laterally loaded piles in Engng ASCE 126, No. 11, 955968.