See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/282709594

Nonlinear analyses of laterally loaded piles - A semi-analytical approach

Article  in  Computers and Geotechnics · October 2015

DOI: 10.1016/j.compgeo.2015.07.009

CITATIONS READS

7 402

3 authors:

Fei Han Rodrigo Salgado

Purdue University Purdue University
14 PUBLICATIONS   70 CITATIONS    209 PUBLICATIONS   4,565 CITATIONS   

SEE PROFILE SEE PROFILE

Monica Prezzi
Purdue University
120 PUBLICATIONS   1,887 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Experimental Determination of Displacement and Strain Fields around Piles using Digital Image Correlation (DIC) View project

Experimental Study of the Load Response of Large Diameter Closed-ended and Open-ended Pipe Piles Installed in Alluvial Soil View project

All content following this page was uploaded by Fei Han on 07 February 2019.

The user has requested enhancement of the downloaded file.

 Computers and Geotechnics 70 (2015) 116–129

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

Nonlinear analyses of laterally loaded piles – A semi-analytical approach

Fei Han ⇑, Rodrigo Salgado, Monica Prezzi
Lyles School of Civil Engineering, Purdue University, West Lafayette, IN 47907-1284, USA

f
a r t i c l e i n f o a b s t r a c t

oo
Article history: This paper describes a semi-analytical, continuum mechanics-based method of calculating the non-linear
Received 28 February 2015 response of pile foundations subjected to lateral loads. The displacement field in the soil is obtained as
Received in revised form 17 June 2015 the product of a pile deflection function and displacement decay functions. Based on the incremental
Accepted 20 July 2015
form of the virtual work, a system of coupled differential equations for the pile deflection and decay func-
Available online 25 August 2015
tions is derived and then solved by using the finite difference method, producing pile deflections, shear

Keywords:
Laterally loaded piles
Nonlinear analyses
Elasto-plastic model Pr
forces and bending moments, as well as displacements within the soil domain. The soil is modeled as an
elasto-plastic material. The analysis is computationally efficient and produces results that are in good
agreement with those from 3D finite element modeling with the same constitutive model for the soil.
The method models laterally loaded pile response more realistically than the traditional p—y method
at a comparable computational cost. The method is substantially more efficient than 3D finite element
simulations. The proposed method offers a good alternative to the p—y method in contexts in which fast
calculation of lateral pile response is required, such as in the design of piles for offshore wind farms.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction p—y curves, and site calibration is almost always needed if realistic
predictions are desired.
Design of laterally loaded piles often relies on the p—y method Three-dimensional numerical analyses, such as the finite ele-
f

in routine engineering practice due to its versatility and speed. The ment method or the finite difference method [3–12] would pro-
oo

pile is assumed to be an Euler–Bernoulli beam acted upon by a
point load (and/or a bending moment) at one end (the pile head)
and soil reactions along the pile length. The soil reactions are
assumed applied at discrete locations and are modeled as a series
of independent, non-linear, one-dimensional springs. Considering
one spring per unit length of pile, the force p per unit length of pile
Pr
duce the most realistic solutions of the laterally loaded pile
problem when appropriate soil models are used and the correct
soil state is considered in the soil around the pile. Yet, these meth-
ods are not often used in routine geotechnical engineering practice
due to the modeling knowledge required to properly use these
techniques and the relatively long problem set up and computation

transferred to each spring is assumed to be a non-linear function of
the lateral deflection y of the pile at the location of the spring.
The p—y method has certain limitations that result from its
rather simple modeling of the pile–soil interaction problem. For
example, the diameter (or width) of the pile is the only input
parameter used in current p—y curves that takes into account the
effect of the pile. However, it has been found that the pile geometry
(both shape and dimensions) and bending stiffness EI affect the
shape and magnitude of p—y curves [1]. Some of the parameters
(for instance, the dimensionless empirical constant J in the API soft
clay criterion [2]) in the p—y method are arbitrarily determined by
the users based on their experience rather than by fundamental
considerations. Thus, significant judgment enters the definition of
⇑ Corresponding author.
E-mail addresses: hanfei@purdue.edu (F. Han), rodrigo@ecn.purdue.edu
(R. Salgado), mprezzi@ecn.purdue.edu (M. Prezzi).
time. This means that semi-analytical solutions that can be shown
to produce accurate, realistic results in computation times compa-
rable to those of the p—y method have a role in the practice of
foundation engineering. This paper presents one such method that
can be used for piles in elasto-plastic soil.
By linking the soil displacement directly to the pile deflection
and describing the soil–pile system using energy principles, analyt-
ical or semi-analytical energy-based methods have been developed
and improved through the years; the applicability of these meth-
ods has been progressively extended from piles in uniform soil
[13,14] to piles in layered soil profiles [15–18] and from single
piles [19] to pile groups [20]. These analytical or semi-analytical
solutions have proven to be reliable and computationally efficient,
but all treated the soil as an elastic material, making it difficult to
use them in practice without considerable judgment.
A new formulation is proposed in this paper to accommodate
the implementation of an elasto-plastic constitutive model for
the soil while maintaining the efficiency of the previous

http://dx.doi.org/10.1016/j.compgeo.2015.07.009
0266-352X/Ó 2015 Elsevier Ltd. All rights reserved.
 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129 117

formulations. The new formulation enables us to make more real- 2.2. Principle of virtual work
istic predictions of the non-linear, three-dimensional response of
the soil–pile system. Energy is input into the pile–soil system when an external lat-
eral force F and a moment M are applied at the pile head, inducing
2. Formulation deformation in the pile and the soil surrounding it. The principle of
virtual work for the stated problem can be written in the incre-
2.1. Displacement formulation for the pile–soil domain mental form as:
Z L 2 ! ! Z  
2
d Dw d w dw 
EI d dz þ Drkl dekl dXs ¼ DFdwjz¼0  DMd
In this paper, we deal with a single, circular pile with diameter B 0 dz
2
dz
2
Xs dz z¼0
and length L embedded in an elasto-plastic soil deposit and sub-
ð4Þ
jected to a lateral load and/or moment at the head. Fig. 1 shows
a Cartesian coordinate system that is used to describe the pile–soil where E is the Young’s modulus of the pile; I is the second moment
domain. The origin of the coordinate system is placed at the center of area of the pile’s cross section; Dw is the increment in the pile
of the pile head, such that the x and y axes are along and perpen- displacement; the stresses rkl and strains ekl are written using indi-
dicular to the loading direction, respectively, and the z axis points cial notation, according to which repetition of any index (such as k
downward along the pile axis. or l) in a term implies summation over that index; DF and DM are
The pile is modeled as a vertical Euler–Bernoulli beam with the increments in the lateral force and moment applied at the pile
pile deflection w in the x-axis direction expressed as a function head, respectively; and Xs denotes the soil domain, which extends

f
of depth z: from 1 to 1 in both the x and y directions and from 0 to 1 in

oo
w ¼ wðzÞ ð1Þ the z direction except for the volume occupied by the pile. The left
hand side of Eq. (4) is the incremental form of the internal virtual
The displacements in the soil result from the pile–soil interac-
work stored in the pile and soil, and the right hand side is the incre-
tion as the pile is loaded laterally; the greater the pile deflection,
mental form of the external virtual work done by the external con-
the greater the displacement in the soil is. The displacement in
centrated force and the bending moment. In this incremental
the soil at any depth z depends on the position being considered
formulation of the principle of virtual work, an increment in exter-
with respect to the pile axis. In general, soil will undergo greater
displacements near the pile than far from it. Thus, we can express
the soil displacement in each direction as the product of the pile
deflection function wðzÞ and a corresponding decay function:
ux ¼ wðzÞf ðx; yÞ
uy ¼ wðzÞgðx; yÞ
Pr
ð2Þ
nal forces and internal stresses are used as the equilibrium set, and
the current displacements and strains are used as the compatible
set. Replacing the incremental terms by the differences between
their initial values (denoted by the subscript zero) and final values
at each increment, Eq. (4) becomes:
Z L 2 !
d w
2
d w
! Z  
dw 
uz ¼ 0 EI d dz þ D r de dX  Fdwj þ Md
dz z¼0
2 2 kl kl s z¼0
0 dz dz Xs
where ux ; uy and uz are the soil displacements in the x; y, and z Z L 2 ! 2
!  
d w0 d w dw 
directions, and f ðx; yÞ and gðx; yÞ are the decay functions for the soil ¼ EI d dz  F 0 dwjz¼0 þ M0 d ð5Þ
0 dz
2
dz
2 dz z¼0
displacements in the x—y plane. When a pile is subjected to a lateral
load, the displacement in the soil will be mainly in the horizontal
direction. The displacement in the vertical direction is negligible 2.3. Governing differential equations
f

when compared to those in the two other directions, and this is

2.3.1. Deflection function w(z)
oo

why it is disregarded. Differentiation of the displacements gives

us the strain components (which are positive in compression): The term Drkl dekl in Eq. (5) can be expanded to:
8 @ux
9 8 9
> > ðx;yÞ
wðzÞ @f @x Drkl dekl ¼ Drxx dexx þ Dryy deyy þ 2Drxy dexy þ 2Drxz dexz þ 2Dryz deyz
8
exx 9 >
> >
> > >
>
@x >
> > >
> >
> >
> >
>
>
> >
> >
>
@uy >
> >
> >
> ð6Þ
>
>
> e >
yy >
>
>
>
>
@y >
>
>
>
>
> wðzÞ @gðx;yÞ
@y
>
>
>
>
> >
> >
> @uz >
> >
> >
>
< = < = < = Since we aim to formulate the differential equations for the pile
ezz h 0
Pr

@z
¼  1 @ux @uy  ¼ i ð3Þ
@f ðx;yÞ deflection, all the terms in Eq. (6) need to be expressed in terms of
>
> exy >> >
> þ @x >
> >
>  1
2
wðzÞ þ @gðx;yÞ >
>
> > > 2 @y > > @y @x > the displacements. This is accomplished by replacing the stress
>
>
> e
>
>
>
>
>
> @ux @uz  >
>
>
>
>
>
>
>
>
>
> xz >
> >
> 1
þ >
> >
>  1 dwðzÞ
f ðx; yÞ >
> increments Dr by the product of the tangential stiffness matrix
: ; > 2 @z @x > > 2 dz >
eyz >
>   >
> >
> >
>
: 1 @uy @uz >
> ; : 1 dwðzÞ ; C ij and the strain increments and writing the strains in terms of
2 @z
þ @y
2 dz
gðx; yÞ
the displacements [by using Eq. (3)]. For example, the first term
Drxx dexx can be expanded as:

Drxx dexx ¼ C 11 Dexx þ C 12 Deyy þ C 13 Dezz þ C 14 Dexy þ C 15 Dexz þ C 16 Deyz dexx

     
@f @f @g @f @g @f
¼ w C 11 þ C 12 dw þ ww C 11 þ C 12 d
@x @x @y @x @y @x
     
1 @f @g @f 1 @f @g @f
þ C 14 w þ dw þ C 14 ww þ d
2 @y @x @x 2 @y @x @x
 
1 dw @f 1 dw @f
þ ½C 15 f þ C 16 g dw þ w½C 15 f þ C 16 g d
2 dz @x 2 dz @x
     
@f @f @g @f @g @f
 w0 C 11 0 þ C 12 0 dw þ w0 w C 11 0 þ C 12 0 d
@x @x @y @x @y @x
     
1 @f 0 @g 0 @f 1 @f 0 @g 0 @f
þ C 14 w0 dw þ þ C 14 w0 w þ d
2 @y @x @x 2 @y @x @x
 
1 dw0 @f 1 dw0 @f
þ ½C 15 f 0 þ C 16 g 0 dw þ w½C 15 f 0 þ C 16 g 0 d
2 dz @x 2 dz @x
Fig. 1. The Cartesian coordinate system used for the soil–pile domain.
ð7Þ
118 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129

After all the other terms in Eq. (6) are expanded in this manner, Eq. (8) contains the variation of both the pile deflection function
Eq. (5) becomes an equation with all of its terms containing the wðzÞ and its derivatives. The variation of the derivatives can be
variations of wðzÞ; f ðx; yÞ; gðx; yÞ or their derivatives. Collecting rewritten in terms of the variation of wðzÞ by doing integration
the terms associated with the variation of wðzÞ and its derivatives: by parts:
8" # " 3 #z¼L Z 9
WðwÞ  Wðw0 Þ ¼ 0 ð8Þ < d2 w dw z¼L d w L 4
d w =
EI 2
d  3
dw þ 4
dwdz
where: : dz dz dz 0 dz ;
z¼0 z¼0
! ! Z L !
Z 2 2 Z 2 1z 3z
1z d w dt dw 3z dw dt
L
d w d w @f @f @g @f þ t  2z
þ ðt  t Þ 
z
w þ k w dwdz
WðwÞ ¼ EI d dzþ C 11 wdw þC 12 wdw 2 dz dz dz dz
0 dz
2
dz
2
Xs @x @x @y @x 0 dz
  Z þ1 2 1z 3z
!
1 @f @g @f 1 dw @f 1 dw @f 1z d w dt dw 2z 3z dw dt z
þ C 14 wdw þ þ C 15 dwf þ C 16 dwg þ t 2
 þ ðt  t Þ  w þ k w dwdz
2 @y @x @x 2 dz @x 2 dz @x L dz dz dz dz dz
    z¼L   z!þ1
@f @g @g @g 1 @f @g @g dw 3z dw 3z
þC 21 wdw þC 22 wdw þ C 24 wdw þ þ dw t 1z þt w þ dw t 1z þt w
@x @y @y @y 2 @y @x @y dz dz
    z¼0 z¼L
1 dw @g 1 dw @g @f @f @g dw 
þ C 25 dwf þ C 26 dwg þC 41 wdw þ  Fdwjz¼0 þ Md
2 dz @y 2 dz @y @x @y @x dz z¼0
     8" 9
@g @f @g 1 @f @g @f @g < d2 w  #z¼L " 3 #z¼L Z
L 4 =
þC 42 wdw þ C 44 wdw

f
þ þ þ 0 dw d w0 d w0
@y @y @x 2 @y @x @y @x ¼ EI d  dw þ dwdz
    : dz2 dz dz
3
dz
4 ;

oo
0
1 dw @f @g 1 dw @f @g z¼0 z¼0
!
þ C 45 dwf þ þ C 46 dwg þ Z L 2 1z 3z
2 dz @y @x 2 dz @y @x 1z d w0 dt0 dw0 2z 3z dw0 dt 0 z
    þ t 0 2
 þ ðt 0  t0 Þ  w0 þ k0 w0 dwdz
dw @f dw @g 0 dz dz dz dz dz
þC 51 wd f þC 52 wd f !
dz @x dz @y Z þ1 2 1z 3z
     d w0 dt0 dw0 3z dw0 dt 0 z
1 dw @f @g 1 dw dw þ t 1z
0 2
 þ ðt 2z
0  t 0 Þ  w 0 þ k w
0 0 dwdz
þ C 54 wd f þ C 55 ff dz dz dz dz dz
2
1
þ C 56
2
1
dw dw
dz

þ C 64 wj d
2
1
d
dz @y @x
 

 

dw dw
dz
dw @f @g
dz @y @x
  
þ

gf þC 61 wd

þ

2
 
dw @f

1
dz

dz @x

g þ C 65
2
d

dw dw
dz
d
dz

g þC 62 wd
 

dz
 
fg

dw 
 
dw @g
dz @y
g
Pr L
 
þ dw t 1z 0
dw0 3z
dz

 F 0 dwjz¼0 þ M0 d
þ t 0 w 0
  z¼0
dw 
z¼L  

dz z¼0
þ dw t 1z dw0
0
dz
þ t 3z
0

where the k and t terms are integrals of the f ðx; yÞ and gðx; yÞ func-
w 0
z!þ1

z¼L

ð11Þ

þ C 66 d gg dXs Fdwjz¼0 þMd ð9Þ tions over the horizontal plane at depth z:
2 dz dz dz z¼0 ZZ 
1 1 1 1
and t1z ðzÞ ¼ C 55 ff þ C 56 gf þ C 65 fg þ C 66 gg dx dy
! ! 2 2 2 2
Z L 2 2 Z XðzÞ
d w0 d w @f @f
Wðw0 Þ ¼ EI d dz þ C 11 w0 dw 0 ZZ
dz
2
dz
2 @x @x 1 @f 1 @f 1 @g 1 @g
0 Xs t2z ðzÞ ¼ C 15 f þ C 16 g þ C 25 f þ C 26 g
  2 @x 2 @x 2 @y 2 @y
f

@g 0 @f 1 @f 0 @g 0 @f XðzÞ
þ C 12 w0 dw þ C 14 w0 dw þ    
@y @x 2 @y @x @x
oo

1 @f @g 1 @f @g
1 dw0 @f 1 dw0 @f @f @g þ C 45 f þ þ C 46 g þ dx dy
þ C 15 dwf0 þ C 16 dwg 0 þ C 21 w0 dw 0 2 @y @x 2 @y @x
2 dz @x 2 dz @x @x @y ZZ  
  @f @g 1 @f @g @g
@g 0 @g 1 @f 0 @g 0 @g t3z ðzÞ ¼ C 51 f þ C 52 f þ C 54 þ f þ C 62 g
þ C 22 w0 dw þ C 24 w0 dw þ @x @y 2 @y @x @y
@y @y 2 @y @x @y XðzÞ
  
Pr

1 dw0 @g 1 dw0 @g @f 1 @f @g
þ C 25 dwf0 þ C 26 dwg 0 þ C 61
g þ C 64 þ g dx dy ð12Þ
2 dz @y 2 dz @y @x 2 @y @x
   
@f 0 @f @g @g @f @g
þ C 41 w0 dw þ þ C 42 w0 dw 0 þ ZZ  
@x @y @x @y @y @x z @f @f @g @f 1 @f @g @f
     k ðzÞ ¼ C 11 þ C 12 þ C 14 þ
1 @f 0 @g 0 @f @g 1 dw0 @f @g @x @x @y @x 2 @y @x @x
þ C 44 w0 dw þ þ þ C 45 dwf0 þ XðzÞ
2 @y @x @y @x 2 dz @y @x  
    @f @g @g @g 1 @f @g @g
1 dw0 @f @g dw @f 0 þ C 21 þ C 22 þ C 24 þ
þ C 46 dwg 0 þ þ C 51 w0 d f @x @y @y @y 2 @y @x @y
2 dz @y @x dz @x    
     @f @f @g @g @f @g
dw @g 0 1 dw @f 0 @g 0 þ C 41 þ þ C 42 þ
þ C 52 w0 d f þ C 54 w0 d þ f @x @y @x @y @y @x
dz @y 2 dz @y @x   
    1 @f @g @f @g
1 dw0 dw 1 dw0 dw þ C 44 þ þ dx dy
þ C 55 d f f þ C 56 d g f 2 @y @x @y @x
2 dz dz 0 2 dz dz 0 ZZ 
    1 1 1 1
dw @f 0 dw @g 0 t1z C 55 f 0 f þ C 56 g 0 f þ C 65 f 0 g þ C 66 g 0 g dx dy
þ C 61 w0 d g þ C 62 w0 d g 0 ðzÞ ¼
dz @x dz @y 2 2 2 2
     XðzÞ
1 dw @f 0 @g 0 1 dw0 dw ZZ
þ C 64 w0 d þ g þ C 65 d f g 1 @f 1 @f 1 @g 1 @g
2 dz @y @x 2 dz dz 0 t2z
0 ðzÞ ¼ C 15 f 0 þ C 16 g 0 þ C 25 f 0 þ C 26 g 0
     2 @x 2 @x 2 @y 2 @y
1 dw0 dw dw  XðzÞ
þ C 66 d g g dXs  F 0 dwjz¼0 þM 0 d    
2 dz dz 0 dz 
z¼0 1 @f @g 1 @f @g
þ C 45 f 0 þ þ C 46 g 0 þ dx dy
ð10Þ 2 @y @x 2 @y @x
 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129 119

ZZ   8 3    
@f 0 @g 0 1 @f 0 @g 0 @g > EI ddzw3 þ t 1z dw þ t3z w u  t 1z dw þ t 3z w l
t 3z ðzÞ ¼ C 51 f þ C 52 f þ C 54 þ f þ C 62 0 g >
> dz dz
0
@x @y 2 @y @x @y >
>    
< 3
XðzÞ ¼ EI ddzw30 þ t1z dw0 3z
0 dz þ t 0 w0  t 1z dw0 3z
0 dz þ t 0 w0
   u l ð18Þ
@f 0 1 @f 0 @g 0 >
>
þ C 61 g þ C 64 þ g dx dy ð13Þ >
> and
@x 2 @y @x >
: d2 w 2
EI dz2  EI ddzw20 ¼ 0
ZZ  
z @f @f @g @f 1 @f 0 @g 0 @f
k0 ðzÞ ¼ C 11 0 þ C 12 0 þ C 14 þ where the subscripts u and l denote values in the parenthesis eval-
@x @x @y @x 2 @y @x @x
XðzÞ uated infinitesimally above and below the pile base. Finally, at infi-
  nite depth, when z ! þ1; w ¼ 0.
@f 0 @g @g @g 1 @f 0 @g 0 @g
þ C 21 þ C 22 0 þ C 24 þ
@x @y @y @y 2 @y @x @y
    2.3.2. Decay functions f(x, y) and g(x, y)
@f 0 @f @g @g 0 @f @g In a similar manner, collecting all the terms containing varia-
þ C 41 þ þ C 42 þ
@x @y @x @y @y @x tions of f ðx; yÞ as well as the variations of its derivatives from Eq.
  
1 @f 0 @g 0 @f @g (5), integrating by parts, and setting the collection of the coeffi-
þ C 44 þ þ dx dy
2 @y @x @y @x cients associated with the first variation of f ðx; yÞ to be zero, the
Euler–Lagrange differential equations for the decay function
where XðzÞ refers to the whole x—y plane except for the area of the
f ðx; yÞ results. Repeating these same steps for gðx; yÞ, we obtain
pile’s cross section when z 6 L and the whole area of the x—y plane
the Euler–Lagrange differential equations for gðx; yÞ. The resulting

f
when z > L. These coefficients, which are evaluated at the depths
equations are:
being considered, are functions of the tangential stiffness matrix

oo
C ij , the values of the decay functions f ðx; yÞ and gðx; yÞ and their @2f @2f @2f @f @f @2g @2g
derivatives. Eq. (11) consists of terms evaluated at the boundaries
a1 2
þ a2 2 þ a3 þ a4 þ a5 þ a6 f þ a7 2 þ a8 2
@x @y @x@y @x @y @x @y
(at z = 0, L and +1) and the integrals of the product of dw and its @2g @g @g @2f 0 @2f 0 @2f 0
associated coefficients. The boundary terms will be satisfied by þ a9 þ a10 þ a11 ~1
þ a12 g ¼ a þa~2 ~3
þa
@x@y @x @y @x2 @y2 @x@y
suitable boundary conditions. Owing to the arbitrariness of the vari- @f @f @2 g0 @2 g0 @2g0 @g
ation dw within both intervals (from z = 0 to L and from z ¼ L to
+1), Eq. (11) is valid only when the term multiplying dw is equal
to zero throughout these two intervals. This leads to the governing
differential equations for the deflection function wðzÞ above and
below the pile base in the laterally loaded pile problem. These
governing differential equations are known as the Euler–Lagrange
Pr and
þa

@2f
~4 0 þ a

~
þ a11
@x
@g 0
@y

@2f
~5 0 þ a

~
@y
þ a12 g 0

@2f
~7
~6 f 0 þ a
@x

@f
2
þa~8

@f
@y 2
þa~9
@x@y

@2g
~10 0
þa

@2g
@x
ð19Þ

differential equations, which are expressed as: b1 2

þ b2 2 þ b3 þ b4 þ b5 þ b6 f þ b7 2 þ b8 2
4 2 1z
! 3z
! @x @y @x@y @x @y @x @y
d w 1z d w dt 2z 3z dw z dt @2g @g @g @2f 0 ~ @2f 0 ~ @2f 0
EI 4  t 2
þ  þt t þ k  w þ b9 þ b10 þ b11 ~
þ b12 g ¼ b1 2 þ b2 2 þ b3
dz dz dz dz dz @x@y @x @y @x @y @x@y
! !
4
d w0
2 1z 3z
~ @f 0 ~ @f 0 ~ ~ @ 2 g 0 ~ @ 2 g 0 ~ @ 2 g 0 ~ @g 0
1z d w0 dt 0 2z 3z dw0 z dt 0 þ b4 þ b5 þ b 6 f 0 þ b7 þ b8 þ b9 þ b10
¼ EI 4
 t0 2
þ  þ t0  t0 þ k0  w0 @x @y @x2 @y2 @x@y @x
dz dz dz dz dz
~ @g 0 ~
þ b11 þ b12 g 0 ð20Þ
f

ð14Þ @y
oo

for z < L, and where

! !
2 1z 3z Z
1z d w dt dw dt
z 1
2z 3z
t 2
þ  þt t þ k  w a1 ¼ C 11 t 1xy dz;
dz dz dz dz
! ! Zz¼0
1
2 1z 3z 1
1z d w0 dt 0 2z 3z dw0 z dt 0 a2 ¼ C 44 t 1xy dz
¼ t0 þ  þ t0  t0 þ k0  w0 ð15Þ 2
Pr

dz
2 dz dz dz Zz¼0
1  
1
a3 ¼ C 14 t1xy þ C 41 t 1xy dz
for z > L. 2
Zz¼0
1 
The boundary conditions for Eqs. (14) and (15) are determined @C 11 1xy 1 @C 41 1xy
from the boundary terms in Eq. (11). At z = 0: a4 ¼ t þ C 15 t2xy þ t  C 51 t 3xy dz
@x 2 @y
8 Zz¼0
1 
> w ¼ w 1 @C 14 1xy 1 @C 44 1xy 1 1
< a5 ¼ t þ t þ C 45 t 2xy  C 54 t3xy dz
or ð16Þ 2 @x 2 @y 2 2
  Zz¼0
1 
: EI d3 w  t1z dw þ t 3z w ¼ EI d3 w0  t 1z dw0 þ t3z w þ F   F
> 1 @C 15 2xy 1 @C 45 2xy 1 xy
dz3 dz dz3 0 dz 0 0 0 a6 ¼ t þ t  C 55 k dz
2 @x 2 @y 2
which indicates that either a displacement w or an external force Zz¼0
1 ð21Þ
1 1xy
F  can be specified at the pile head, and a7 ¼ C 14 t dz
8 dw  Zz¼0 2 1
< dz ¼ h
> a8 ¼ C 42 t 1xy
dz
or ð17Þ Zz¼0
> 1 
: d2 w 2 1
EI dz2 ¼ EI ddzw20 þ M   M 0 a9 ¼ C 12 t 1xy þ C 44 t 1xy dz
2
which means that either a rotation h or a bending moment M can Zz¼01  
1 @C 14 1xy 1 1 @C 44 1xy 1
be specified at the pile head. Due to the mathematical nature of Eq. a10 ¼ t þ C 16 t2xy þ t  C 54 t 3xy dz
2 @x 2 2 @y 2
(14), which is a fourth-order, ordinary differential equation, two Zz¼01 
@C 12 1xy @C 42 1xy 1
boundary conditions are required at z = 0. This includes one bound- a11 ¼ t þ t þ C 46 t  C 52 t 3xy dz
2xy
@x @y 2
ary condition from the two alternatives in Eq. (16) and one from the Zz¼01 
1 @C 16 2xy 1 @C 46 2xy 1 xy
two alternatives in Eq. (17). Similarly, two more boundary condi- a12 ¼ t þ t  C 56 k dz
z¼0 2 @x 2 @y 2
tions must be applied at the pile base (z ¼ L):
120 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129

Z 1 Z 1  
1 @C 25 2xy 1 @C 45 2xy 1
~1 ¼
a C 11 t 1xy
0 dz;
~6 ¼
b t0 þ
xy
t 0  C 65 k0 dz
2 @y 2 @x 2
Zz¼0
1 Z
z¼0
1
1 1
~2 ¼
a C 44 t 1xy
0 dz
~7 ¼
b 1xy
C 44 t 0 dz
z¼0 2 2
Z 1  Zz¼0
1
1 ~ ¼
~3 ¼
a C 14 t1xy
0 þ C 41 t 0
1xy
dz b8 C 22 t 1xy
0 dz
z¼0 2
Z 1  Zz¼0
1 
@C 11 1xy 1 @C 41 1xy ~9 ¼ 1
~4 ¼
a t 0 þ C 15 t 2xy 0 þ t  C t
51 0
3xy
dz b C 24 t1xy þ C t
42 0
1xy
dz
z¼0 @x 2 @y 0 z¼0 2
0
Z 1  Z 1 
1 @C 14 1xy 1 @C 44 1xy 1 1 ~ ¼ 1 @C 24 1xy 1 @C 44 1xy 1 1
~5 ¼
a t0 þ t0 þ C 45 t 2xy
0  C t
54 0
3xy
dz b10 t 0 þ t þ C t
46 0
2xy
 C t
64 0
3xy
dz
z¼0 2 @x 2 @y 2 2 z¼0 2 @y 2 @x 0 2 2
Z 1  Z 1 
1 @C 15 2xy 1 @C 45 2xy 1 xy ~11 ¼ @C 22 1xy 1 @C 42 1xy
~6 ¼
a t0 þ t  C 55 k0 dz b t þ C 26 t2xy 0 þ t  C 62 t 3xy dz
z¼0 2 @x 2 @y 0 2 z¼0 @y 0 2 @x 0 0
Z 1 ð22Þ Z 1 
1 ~12 ¼ 1 @C 26 2xy 1 @C 46 2xy 1 xy
~7 ¼
a C 14 t 1xy
0 dz b t0 þ t  C 66 k0 dz ð24Þ
z¼0 2 z¼0 2 @y 2 @x 0 2
Z 1
~8 ¼
a C 42 t 1xy
0 dz and
Zz¼0
1 
dw dw dw dw

f
1 xy
~9 ¼
a C 12 t 1xy þ C t
44 0
1xy
dz t1xy ðzÞ ¼ ww; t2xy ðzÞ ¼ w; t 3xy ðzÞ ¼ w ; k ðzÞ ¼
0
2 dz dz dz dz

oo
z¼0
Z 1  ð25Þ
1 @C 14 1xy 1 1 @C 44 1xy 1
~10 ¼
a t 0 þ C 16 t2xy 0 þ t  C t
54 0
3xy
dz
z¼0 2 @x 2 2 @y 0 2
Z 1  dw0 dw xy dw0 dw
@C 12 1xy @C 42 1xy 1 t1xy 2xy
0 ðzÞ ¼ w0 w; t 0 ðzÞ ¼ w; t 3xy
0 ðzÞ ¼ w0 ; k0 ðzÞ ¼
~11 ¼
a t þ t þ C 46 t2xy0  C 52 t 0
3xy
dz dz dz dz dz
z¼0 @x 0 @y 0 2
Z 1  ð26Þ
~12 ¼
a

b1 ¼

b2 ¼
Zz¼0
1
1 @C 16 2xy 1 @C 46 2xy 1
z¼0 2 @x

Z 1
C 41 t

1
t0 þ

1xy
dz

C 24 t 1xy dz
2 @y 0 2
xy
t  C 56 k0 dz

Pr The boundary conditions for the decay functions are:

f ðx; yÞ ¼

1 within the pile cross section
0 as x ! 1 or y ! 1

0 within the pile cross section

ð27Þ

2 gðx; yÞ ¼ ð28Þ
Zz¼0
1  0 as x ! 1 or y ! 1
1
b3 ¼ C 21 t 1xy þ C 44 t1xy dz
z¼0 2 These boundary conditions mean that the soil displacement ux
Z 1 
along the loading direction is the same as the pile deflection w
@C 21 1xy @C 41 1xy 1
b4 ¼ t þ t þ C 45 t2xy  C 61 t3xy dz right at the pile–soil interface, while the soil displacement uy per-
z¼0 @y @x 2
Z 1  pendicular to the loading direction is equal to zero at the pile–soil
1 @C 24 1xy 1 2xy 1 @C 44 1xy 1
b5 ¼ t þ C 25 t þ t  C 64 t3xy dz interface. Both the displacements in the x and y directions tend
f

z¼0 2 @y 2 2 @x 2
Z 1  towards zero at an infinite distance from the pile.
1 @C 25 2xy 1 @C 45 2xy 1
oo

xy
b6 ¼ t þ t  C 65 k dz
z¼0 2 @y 2 @x 2
2.4. Constitutive model
Z 1 ð23Þ
1
b7 ¼ C 44 t 1xy dz
z¼0 2 In previous energy-based, semi-analytical studies [13–16,19–
Z 1
24], soil was assumed to behave as a linear elastic material. This
b8 ¼ C 22 t1xy dz
Pr

assumption resulted in overestimation of the soil stiffness and pro-

Zz¼0
1 
1 duced greater reaction forces or bending moments at the pile head
b9 ¼ C 24 t 1xy þ C 42 t1xy dz than expected in reality [25]. One of the advantages of the analysis
z¼0 2
Z 1  presented in this paper is that it allows the implementation of
1 @C 24 1xy 1 @C 44 1xy 1 1
b10 ¼ t þ t þ C 46 t2xy  C 64 t 3xy dz more realistic constitutive relationships for the soil. In this paper,
z¼0 2 @y 2 @x 2 2
Z 1  the soil is assumed to behave as an elasto-plastic material.
@C 22 1xy 1 @C 42 1xy
b11 ¼ t þ C 26 t 2xy þ t  C 62 t 3xy dz Calculations in later sections will be done for both linear elastic,
z¼0 @y 2 @x perfectly plastic and non-linear elastic, perfectly plastic soil.
Z 1 
1 @C 26 2xy 1 @C 46 2xy 1 xy
b12 ¼ t þ t  C 66 k dz
z¼0 2 @y 2 @x 2 2.4.1. Degradation of elastic moduli
Z Simple non-linear elastic relationships have been used in
1
~1 ¼
b C 41 t1xy geotechnical engineering for decades [26–29]. They are based on
0 dz
Zz¼0 degrading the tangential shear modulus G, which takes a maxi-
1
~ ¼ 1 mum value at very small strain levels (below 106) [26], with
b 2 C 24 t1xy
0 dz
2 increasing strain level. A simple hyperbolic degradation formula-
Zz¼0
1  
1 tion for the tangential elastic shear modulus G is given by [26]:
~3
b ¼ C 21 t 1xy þ C t
44 0
1xy
dz
0
z¼0 2 pffiffiffiffi !2
Z 1  G J2
~ @C 21 1xy @C 41 1xy 1 ¼ 1  pffiffiffiffiffiffiffiffiffiffiffi ð29Þ
b 4 ¼ t0 þ t0 þ C 45 t 2xy
0  C 61 t 3xy
0 dz G0 J 2;max
z¼0 @y @x 2
Z 1 
~5 1 @C 24 1xy 1 1 @C 44 1xy 1 where J2 is the second invariant of the deviatoric stress tensor, and
b ¼ t0 þ C 25 t 2xy0 þ t  C 64 t 3xy
dz
z¼0 2 @y 2 2 @x 0 2 0
J2;max is related to the soil shear strength. For example, if a yield
 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129 121

pffiffiffiffi
criterion, expressed as F ¼ J 2  aI1 ¼ 0, is used, then J 2;max is @F 1
e_ pij ¼ k_ ¼ k_ pffiffiffiffi sij ð37Þ
determined by the first invariant I1 of the stress tensor and the yield @ rij 2 J2
criterion constant a as J2;max = a2 I21 . A minimum value of G is
normally set as a percentage of G0 [26,29], i.e., Gmin = a1 G0 . The 2.4.4. Formulation of the tangential stiffness matrix
tangential bulk modulus K is related to G by the Poisson’s ratio m: The consistency condition requires the stress state to remain on
2ð1 þ tÞ the yield surface during yield:
K¼ G ð30Þ
3ð1  2tÞ @F
F_ ¼ r_ ij ¼ 0 ð38Þ
@ rij
2.4.2. Smoothed Drucker–Prager yield criterion for sand where r_ ij is the stress rate that can be related to the strain rate by
The Drucker–Prager yield surface is shown in Fig. 2 in the Hooke’s law:
meridional plane with the apex rounded using a hyperbolic  
approximation, as suggested by [30], in order to eliminate the r_ ij ¼ C ijkl e_ ekl ¼ C ijkl e_ kl  e_ pkl ð39Þ
gradient singularity at the tip of the cone yield surface. The corre-
The plastic multiplier k_ can be determined by substituting the
sponding yield criterion is:
flow rule [Eq. (34) or Eq. (37)] into Hooke’s law [Eq. (39)] and then
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 substituting the stress rate r_ ij into the consistency condition [Eq.
F¼ J 2 þ b a2  ðaI1 þ jÞ ¼ 0 ð31Þ
(38)]. The flow rule [Eq. (34) or Eq. (37)], now with the value of

f
where b is the distance between the apex of the original yield sur- the plastic multiplier k_ known, is then substituted into Hooke’s

oo
face and that of the smoothed yield surface (see Fig. 2), and the law to give us the relationship between the stress rates or incre-
parameters a and j [31] can be calculated from the soil friction ments Dr and the strain rates or increments De, which in matrix
angle / and the cohesive intercept c of a Mohr–Coulomb material form is expressed as:
using these relationships: 2 3 2 3
Drxx Dexx
2 sin / 6 7 6 7
6 Dryy 7
a ¼ pffiffiffi

j ¼ pffiffiffi
3ð3  sin /Þ

6c cos /
3ð3  sin /Þ
When b is taken as 0, the expression becomes that of the exact
Pr
ð32Þ

ð33Þ
6
6
6
6 Dr 7
6
6
7
7
7¼ C 6
xy 7
6 Dr 7
4
7
xz 5

Dryz
6 Deyy 7
6
6
6 De 7
6 xy 7
6
6 De 7
4 xz 5
7
7
6 Drzz 7 ep
6 Dezz 7
7
7

Deyz
ð40Þ

Drucker–Prager yield surface. An associated flow rule is used: ep

0 1 where C is the tangential elasto-plastic stiffness matrix, which is

@F 1 given by:
B C
e_ pij ¼ k_ ¼ k_ @ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sij  adij A ð34Þ 2 3
@ rij 2 2 Cep ep ep ep ep ep
2 J2 þ b a 11 C12 C13 C14 C15 C16
6 Cep Cep ep ep ep ep 7
6 21 22 C23 C24 C25 C26 7
where e_ pij is the plastic strain rate tensor; rij is the stress tensor; sij is 6 7
ep
6 Cep C32 C33 C34 C35 Cep
ep ep ep ep
7
C ¼6 36 7
f

31
the deviatoric stress tensor; dij is the Kronecker delta, which is equal 6 Cep ep ep ep ep ep 7
6 41 C42 C43 C44 C45 C46 7
oo

to 1 if i ¼ j, and 0 otherwise; and k_ is the plastic multiplier. 6 ep ep 7

4 C51 Cep ep ep ep
52 C53 C54 C55 C56
5
Cep ep ep ep ep ep
61 C62 C63 C64 C65 C66
2.4.3. Von Mises yield criterion for clay 2 3
The behavior of clay can be modeled using the von Mises yield K þ 43 G K  23 G K  23 G 0 0 0
criterion: 6K  2G K þ 4G K  2G 0 0 7
0
6 3 3 3 7
pffiffiffiffi 6 7
6 K  23 G K  23 G K þ 43 G 0 0 0 7
Pr

F ¼ J2  n ¼ 0 ð35Þ 6
¼6 7
6 0 0 0 2G 0 0 77
where n is determined from the undrained shear strength su of the 6 7
4 0 0 0 0 2G 0 5
clay using:
0 0 0 0 0 2G
2 2  3
n ¼ pffiffiffi su ð36Þ H11 H11 H11 H22 H11 H33 2H11 H12 2H11 H13 2H11 H23
3 6 H H H  H H H 2H22 H12 2H22 H13 2H22 H23 7
6 22 11 22 22 22 33 7
6  7
An associated flow rule results in the following plastic strain 16 H33 H11 H33 H22 H33 H33 2H33 H12 2H33 H13 2H33 H23 7
 66
7
rate expression: H 6 H12 H11 H12 H22 H12 H33 2H12 H12 2H12 H13 2H12 H23 7
  
7
6  7
4 H13 H11 H13 H22 H13 H33 2H13 H12 2H13 H13 2H13 H23 5
H23 H11 H23 H22 H23 H33 2H23 H12 2H23 H13 2H23 H23
ð41Þ
where K and G are the tangential bulk and shear moduli of the soil.
The first term on the right hand side is the elastic stiffness
matrix, and the second term on the right hand side depends on
the yield criterion selected. For the rounded Drucker–Prager yield
criterion, we get:

GJ2
H¼ 2
þ 9a2 K ð42Þ
J 2 þ b a2
Fig. 2. Smoothed Drucker–Prager yield surface.
122 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129

 2 1z   
G Dz t 4Dz2 t 1z
Hij ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sij  3aKdij ð43Þ  3wkþ4 þ 14wkþ3 þ  24 wkþ2 þ 18  wkþ1
J 2 þ b a2
2 EI EI
 
3Dz2 t 1z 2Dz3 t 3z
þ 5 wk
G EI EI
Hkl ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi skl  3aKdkl ð44Þ  2 1z   
2 Dz t 0 4Dz2 t 1z
J 2 þ b a2 ¼ 3wkþ4
0 þ 14wkþ30 þ  24 wkþ2 0 þ 18  0
wkþ1
0
EI EI
 
For the von Mises yield criterion: 1z
3 Dz 2 t 0 3z
2 Dz 3 t 0 2 Dz 3
þ 5 wk0 þ ðF   F 0 Þ
H¼G ð45Þ EI EI EI
ð50Þ
G
Hij ¼ pffiffiffiffi sij ð46Þ The coefficients appearing in these equations are integrals that
J2 are evaluated at the depth of node k. These finite difference
approximations for the governing differential equations, as well
G as the boundary conditions, form a system of linear equations,
Hkl ¼ pffiffiffiffi skl ð47Þ
J2 whose solution results in the deflection w at each and every node
along the z axis.

f
3. Finite difference approximations 3.2. Decay functions f(x, y) and g(x, y)

oo
3.1. Deflection function w(z) The Euler–Lagrange differential equations, Eqs. (19) and (20),
for the decay functions are two coupled, linear, second-order, par-
The Euler–Lagrange differential equations for the deflection tial differential equations that must be solved simultaneously.
function wðzÞ [Eq. (14) and (15)] are fourth-order and Similarly, finite difference approximations can be used to solve
second-order, linear, inhomogeneous ordinary differential equa- these differential equations.
tions, respectively. These equations can be solved by the finite dif-
ference method. The soil–pile domain is discretized in the z
direction by uniformly spaced nodes with spacing Dz and a total
number of nodes equal to nz . The finite difference approximation
for the governing differential equations at a node k can be
expressed as:
" ! #
Pr 4. Iterative solution algorithm

The governing differential equations for the deflection function

wðzÞ and the decay functions f ðx; yÞ and gðx; yÞ form a coupled sys-
tem of differential equations in which the coefficients in the gov-
1z erning differential equations [Eqs. (19) and (20)] for the decay
kþ2 3 dt 2z 3z functions are integrals of the deflection function and its derivatives
2EIw þ Dz  þt t  2t Dz  8EI wkþ1
1z 2
dz and, in turn, the coefficients in the differential equations [Eqs. (14)
" ! # and (15)] for the deflection function are integrals of the decay func-
3z
4 z dt tions and their derivatives. These coupled differential equations
þ 2 Dz k  þ 4t Dz þ 12EI wk
1z 2
dz need to be solved iteratively. When an elasto-plastic model is used
" ! # for the soil, the analysis is done in an incremental manner. Thus, an
1z
f

dt iterative, sub-stepping algorithm, shown in Fig. 3, is used in the

3
 Dz  þt t2z 3z
þ 2t Dz þ 8EI wk1 þ 2EIwk2
1z 2
dz
oo

solution scheme.
" ! # ð48Þ Fig. 3(a) shows the steps of the global adaptive sub-stepping
1z
kþ2 3 dt 0 solution algorithm used in this paper. A pseudo time T that ranges
¼ 2EIw0 þ Dz  þ t0  t 0  2t 0 Dz  8EI wkþ1
2z 3z 1z 2
0
dz from 0 to 1 is used; the analysis starts with T = 0 and finishes when
" ! # T reaches 1. An increment of load DT (= fraction of the total load),
3z
4 z dt 0 the size of which is adaptively adjusted based on error estimation,
þ 2 Dz k 0  þ 4t 0 Dz þ 12EI wk0
1z 2
Pr

dz is applied in steps. The analysis starts by calculating the initial geo-

" 1z
! # static stress states for each and every point in the soil domain.
dt z
Initial values of k and t1z can be set as the elastic bulk modulus
 Dz3  0 þ t 2z 0  t 3z
0 þ 2t 1z
0 D z 2
þ 8EI wk1
0 þ 2EIw0k2
dz 2
K and GB (G being the elastic shear modulus and B being the pile
diameter), respectively, at depth z whereas t 2z and t3z can be set to
for z < L, and
" # " ! # zero. The first trial loading step DT is assumed to be 0.1, indicating
1z 3z
dt1z kþ1 2 z dt 10% of the total load, and tentatively applied at the pile head. Too
 2t þ Dz w þ 2 Dz k  þ 4t wk
1z
large of a loading step DT may result in considerable error, which
dz dz
" ! # must be prevented. The error jEr_sj for each loading step is esti-
1z
dt 2z 3z mated by comparing the solution (deflection w or shear force S
 Dz  þt t þ 2t wk1
1z
dz along the pile) obtained by applying the trial loading step as a sin-
" # " ! # ð49Þ gle increment of size DT with the solution obtained by applying the
1z 3z
1z dt 0 kþ1 2 z dt 0 loading in two consecutive sub-increments of size 0.5 DT:
¼  2t0 þ Dz w0 þ 2Dz k0  þ 4t 0 wk0
1z
dz dz 
" ! # kfw2 g  fw1 gk kfS2 g  fS1 gk
1z
dt 0 jEr sj ¼ max ; ð51Þ
 Dz  þ t0  t 0 þ 2t 0 wk1
2z 3z 1z kfw2 gk kfS2 gk
0
dz
where the curly brackets {} denote vectors composed of values of w
for z > L. or S at each node along the pile length and jjjj refers to the opera-
The boundary conditions are also expressed by the finite differ- tion of taking the L2 norm of a vector. If the estimated error is smal-
ence approximation. For example, the external concentrated load ler than the tolerance (jEr_sj 6 TOL), for which a value of 104 is
boundary condition, expressed in Eq. (18), is written as: considered sufficient, the trial loading step is accepted and a larger
 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129 123

z
(a) and k appearing in Eqs. (14) and (15) are used to solve for the pile
deflection function w ¼ wðzÞ, which is then used to calculate the
xy
coefficients txy and k in Eqs. (19) and (20) needed to obtain values
for the decay functions f ðx; yÞ and gðx; yÞ. The intermediate solu-
tions for f ðx; yÞ and gðx; yÞ are substituted back into the expressions
for the coefficients t z and kz to start the second iteration. The elas-
tic stiffness matrix is used in the calculations of the coefficients
appearing in the governing equations in the first loading step.
The iteration continues until the solution converges, when the rel-
ative error is smaller than the tolerance (jEr_ij 6 TOL), for which a
value of 104 is considered sufficient. The relative error for the iter-
ation is evaluated as the maximum of the relative difference in the
deflection w or the internal shear force S on the pile between two
consecutive iterations j  1 and j:

kfwj g  fwj1 gk kfSj g  fSj1 gk
jEr ij ¼ max ; ð52Þ
kfwj gk kfSj gk

f
The elasto-plastic stiffness matrix is updated after each loading
step is accepted. This is accomplished by updating the stresses at

oo
each and every node in the soil domain using the adaptive
sub-stepping integration algorithm described in [32,33]. In this
algorithm, the stresses are calculated from the strains, which are
calculated from the displacements in the soil using Eq. (3).
Integration of the constitutive relations is then done by automati-

Pr cally dividing the strain increment into a number of sub-steps

and integrating over each sub-step. The local error of each of these
sub-steps is estimated by taking the difference between an Euler
solution and a modified Euler solution; the size of the subsequent
sub-step is then adjusted based on this error estimation. The
updated stress values are then used to compute the stiffness matrix
following Eq. (41). The analysis is completed when the applied dis-
placement or external load at the pile head is reached (T = 1).

5. Illustration of application of analysis

(b) 5.1. Problem

f

A 10-m-long circular pile with a diameter of 0.5 m is embedded

oo

in a uniform soil with / = 30° and cohesive intercept of 3 kPa. The

unit weight of the soil is 20 kN/m3. The soil is modeled as a linear
elastic (with E = 30 MPa and m ¼ 0:2), perfectly-plastic material fol-
lowing a Drucker–Prager yield criterion with a rounded apex. The
pile is modeled as an elastic material with Young’s modulus
Ep ¼ 30 GPa. A displacement of 10 mm is applied at the pile head,
Pr

where rotation is constrained.

Fig. 3. Flow chart of the solution scheme: (a) the global sub-stepping solution
algorithm and (b) the local iterative algorithm for each loading step.
step size (1.1 DT) can be used in the next loading step. Otherwise, if
the current step fails to pass the tolerance criterion, then the step
size is halved and the calculation for the new trial loading step is
repeated.
The steps followed within each of the trial loading increments
are illustrated in Fig. 3(b). The initial values for the coefficients t z
We solved the problem using two methods: the analysis pro-
posed in this paper (denoted by SAM, for semi-analytical method,
in the figures that follow) and the finite element method (denoted
by FEM in the same figures). A Visual C# code was written to per-
form the semi-analytical analysis. The FEM analyses are 3D simu-
lations using ABAQUS 6.12-1 with the smoothed Drucker–Prager
yield criterion and linear elasticity within the yield surface. The
pile–soil system is modeled by a domain with dimensions equal
to 10  10  15 m. The entire domain consists of a total number
of about 11,000 8-noded, linear brick, hexahedral elements. At
the four vertical surrounding boundaries, only the displacement
in the direction perpendicular to the corresponding surface was
confined, whereas all degrees of freedom were confined at the bot-
tom boundary. A horizontal displacement of 10 mm, along with
rotation constraint, was applied at the pile head.
5.2. Deflection of the pile
The pile deflection profiles obtained by both methods, plotted in
Fig. 4, are quite similar. The deflection of the pile is most
124 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129

Pile deflection wx (mm) Shear force S (kN)

-2 0 2 4 6 8 10 12 -100 0 100 200 300 400 500 600
0 0

2 2

4 x 4
Depth (m)

Depth (m)
Loading

6 6
y

FEM
8 FEM 8
SAM
SAM

f
10 10

oo
Fig. 4. Pile deflection calculated by SAM and FEM. Fig. 6. Shear force at cross sections along the pile.

pronounced near the pile head, becoming negligible below a depth

of approximately 4 m (8B; B being the pile diameter).

5.3. Bending moment and shear force profiles

According to beam theory, the internal shear force and bending

moment can be obtained from sequential differentiations of the
deflection function wðzÞ. Figs. 5 and 6 show the bending moment
and shear force at each cross section along the pile length, calcu-
Pr
lated by both the current analysis and the finite element analysis.
Again, very good agreement was obtained between the results
from the two methods.
(a)
5.4. Displacements in the soil
Loading direction Pile
Fig. 7 shows the solution for the decay function f ðx; yÞ. Since 0.8
f

0
f ðx; yÞ is symmetric with respect to the x axis, only one-half of 0.6 0.4 1.0
oo

the domain in the x—y plane is plotted in the figure. The values 1
0.2
0.8
of the decay function f ðx; yÞ at each location ðx; yÞ in the soil
domain are multiplied by the value of the deflection function at
Y (m)

2 0.6
each depth to obtain the soil displacement ux . At any given depth
z, ux at the pile–soil interface is exactly the same as the lateral 3 0.4
Pr

4 0.2

Bending moment M (kN.m) 0.0

5
-500 -400 -300 -200 -100 0 100 200 -5 -4 -3 -2 -1 0 1 2 3 4 5
0
X (m)
(b)
2 Fig. 7. Decay function f ðx; yÞ plotted as: (a) 3D surface over the x—y plane and (b)
contour lines in the x–y plane.

4
Depth (m)

deflection of the pile and decreases to zero at large distances from

the pile. As shown in Fig. 7(a), the decay rate for f ðx; yÞ decreases
6 with increasing distance from the pile. Note also that the decay
rate for the f ðx; yÞ function is different behind and ahead of the pile.
This contrasts with the symmetry exhibited by linear elastic
FEM solutions.
8
SAM Similarly, the solution for the decay function gðx; yÞ, which is
skew-symmetric with respect to the x axis, is shown in Fig. 8 in half
of the x—y plane. The decay function gðx; yÞ, multiplied by the
10
deflection function at each depth, results in the displacement uy
Fig. 5. Bending moment at cross sections along the pile. in the soil, which is zero right at the pile–soil interface; uy peaks,
 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129 125

and 6B). The displacement in the soil decreases significantly with

increasing depth, becoming almost negligible at depth z = 6B.
Fig. 10(a) is the vector plot of the top view of the displacement
profile at the ground surface (z = 0). The contour lines represent
soil displacements with the magnitude of 20%, 40% and 60% of
the deflection at the pile head. Greater displacements in the soil
are observed ahead of the pile than behind it. We can define an
area of influence around the pile at the ground surface within
which the magnitude of the soil displacement is greater than 20%
of the deflection at the pile head. As shown in Fig. 10(a), the influ-
ence area extends to 5.5B (B being the pile diameter) ahead of the
pile, but only to 2B behind the pile and 3B to the two sides of the
(a) pile. The different responses of the soil at different locations are
due to the different stress states at the locations with respect to
X (m) the pile: soil is mostly in compression ahead of the pile, mostly
-5 -4 -3 -2 -1 0 1 2 3 4 5 in extension behind the pile and in a state approximating simple
5
0.11 shear on the sides of the pile. The pattern of the profile for the soil
0.00 displacement at the ground surface matches the DIC observations

f
4 0.09
by [34]. In contrast, when the soil is modeled as an elastic material,

oo
3 0.06 the displacement profile at the ground surface, as plotted in
Y (m)

Fig. 10(b), is symmetric with respect to both the x and y axes.

2 0.03 0.03
0.09 0.00 5.5. Load vs. deflection at the pile head
1 0.06
-0.03
Fig. 11 shows the bending moment and shear force versus pile
0
Loading direction Pile
(b)
Fig. 8. Solution for the decay function gðx; yÞ represented as: (a) 3D surface and (b)
contour plots in the x—y plane.
Pr head deflection for both elastic and elasto-plastic soil. The slope
of the curves represents the stiffness of the pile–soil system when
the pile is loaded laterally. When the soil is modeled as an
elasto-plastic material, the load–deflection curve (the solid line
in Fig. 11) starts with the same slope as that of an elastic analysis
with the same Young’s modulus and Poisson’s ratio (the dashed
with a magnitude of about 11% of the deflection at the pile head, at line in Fig. 11), but, as the load increases, the stiffness of the
a distance of 2B from the pile center in a direction making an angle pile–soil response degrades as a result of the progressive yielding
of approximately 45° with the x axis. Values of uy behind the pile and plastic deformation occurring in the soil surrounding the pile.
are negligible, with a maximum value of only 3% of the deflection
at the pile head. 5.6. Computational efficiency
The displacements ux and uy at each and every point in the soil
SAM is computationally efficient when compared to full
f

are calculated from Eq. (2). The magnitude of the resultant dis-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi three-dimensional finite element analyses. Running on the same
oo

placement u can be calculated as u2x þ u2y . Fig. 9 shows sectional laptop equipped with IntelÒ Core™ i7-2760QM CPU and 6 GB of
horizontal views of u in the soil at several depths z (0B, 2B, 4B RAM, the 3D finite element analysis using ABAQUS to solve this

Loading direction Pile Loading direction Pile

8
0 (mm) 0 (mm)
Pr

6 6
4 10 4 10
1 1 2
2 8 8
Y (m)
Y (m)

2 6 2 6
3 4 3 4
2 2
4 (a) z = 0B 4 (b) z = 2B
0 0
5 5
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
X (m) X (m)

Loading direction Loading direction Pile

Pile
0 (mm) 0 (mm)
4 2 2
10 10
1 1
8 8
Y (m)
Y (m)

2 6
2 6

3 4 3 4
2 2
4 z = 4B 4 z = 6B
(c) 0
(d) 0
5 5
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
X (m) X (m)

Fig. 9. Sectional plots in horizontal planes at depths (a) z = 0B, (b) z = 2B, (c) z = 4B and (d) z = 6B of the magnitude of displacement in the soil.
126 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129

5 5
4 4
3 3
2 2
1 1
40%

y (m)

y (m)
0 0
60% 60% 40%
-1 20% -1
20%
-2 -2
-3 -3
-4 -4
-5 -5
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
x (m) x (m)
(a) (b)
Fig. 10. Soil displacement at the ground surface: (a) solution using elasto-plastic model and (b) solution using elastic model.

f
oo
600 800
(a) (b)
Bending moment M (kN.m)

600
Shear force S (kN)

400

200

Elastic
Elasto-plastic
Pr 400

200
Elastic
Elasto-plastic

0 0
0 2 4 6 8 10 0 2 4 6 8 10
Deflection at pile head w (mm) Deflection at pile head w (mm)

Fig. 11. Reaction forces at the pile head vs. deflection at the pile head: (a) bending moment and (b) shear force.
f
oo

example problem requires 16,220 s, while the semi-analytical thickness of more than 10 times the pile diameter) was backfilled
analysis takes only 530 s for the same level of accuracy. after the pile was driven into the clay layer and the lateral response
of the pile is mainly governed by the soil near the ground surface
[35], this pile can be treated as a non-displacement pile when per-
6. Case study forming an analysis to determine its lateral loading capacity. In
order to simulate offshore conditions, the site was flooded and
Pr

A lateral load test was performed on the campus of the the water level was kept 0.3 m (1 foot) above the ground surface
University of Houston, Texas [35–37]. A 4.9-m-long, grouted, steel during the test [37]. The saturated unit weight, as reported in
pipe pile with a diameter of 273 mm was embedded in the soil pro- [37], is 18.9 kN/m3 and 20.1 kN/m3 for the sand and clay layers,
file shown in Fig. 12. The soil profile consists of a medium dense, respectively.
poorly graded sand (SP) layer underlain by a layer of stiff clay with Cone penetration and standard penetration tests were per-
high plasticity (classified as CH) [35]. Since the sand layer (with formed in the sand and clay layers, respectively [35,37]. Results
of these tests are plotted in Fig. 13. The coefficient of earth pressure
K 0 ¼ 0:6, as reported in [37], was used to calculate the initial stress
state in the compacted sand layer, with relative density DR ¼ 50%
according to [35,37]. These values have been confirmed as reason-
able by a check using the correlation (between DR ; qc and the
stress state) proposed in [38].
A peak friction angle of 38.5° determined by direct shear tests
performed on reconstituted samples with the same density as
the sand in the field is reported in [35]. The value is consistent with
DR ¼ 50% for the level of confining stress for the middle of the sand
layer. With this relative density and assuming a critical-state fric-
tion angle /c in the 30–32° range, a correlation especially devel-
oped to accommodate low confining stresses [39] would produce
a triaxial compression peak friction angle /p in the 37.4–39.3°
Fig. 12. Soil profile at the load test site in Houston, Texas. range at the midheight of the sand layer (z = 1.45 m). Although
 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129 127

Cone resistance qc (MPa) SPT blow count (blows/ft) material. Table 1 summarizes the parameters used in the constitu-
0 1 2 3 4 0 5 10 15 20 25 tive models. As reported in [45–47], small values, ranging from
0 2 0.002 to 3 kPa, have been used for the cohesive intercept c for
Sand Sand sands in numerical analyses to improve numerical stability with-
3 out affecting analysis accuracy. A value of c = 0.1 kPa was used in
0.5 Clay
this paper since values lower than 0.1 kPa produce essentially
4
the same results. It was suggested [33] that values equal to or less
1
Depth (m)

Depth (m)
5 than 0.25 ccot/ = 0.03 kPa can be used for the cone smoothing fac-
tor b. Values in the range from 0.02 to 0.25 have been used for the
1.5 6 non-linear elasticity parameter a1 [26,28,29]. It is not possible to
determine a value for a1 for the sand in which the test pile was
7
installed based on the data that are available. The median of the
2
8 reported values of a1 (=0.135) [26,28,29] was used in the analyses
(a) (b) given that the results of the analyses are not very sensitive to a1 ;
2.5 9 using values for a1 within the [0.02, 0.25] range produce
predicted lateral loads at the pile head at the maximum applied
Fig. 13. In situ test results (after [35]): (a) cone resistance in the sand layer and (b)
displacement of 40 mm within 10% of that predicted by
SPT blow count in the clay layer.
taking a1 ¼ 0:135. The bending stiffness EI is estimated as

f
18,000 kN  m2 based on the geometry of the pile’s cross section
there would be a small difference between triaxial compression

oo
[35]. Since the pile was loaded at the pile head without a moment,
and direct shear test friction angles, these calculations largely con- a free-head boundary condition (which means that the pile head
firm the reasonableness of the reported friction angle for the con- can rotate freely) is used in the analyses.
ditions in situ. The p—y analyses were performed using the software Pygmy
The small-strain shear modulus G0 is not reported in the litera- [48]. The p—y curves for the clay layer were defined based on
ture and thus was estimated based on the widely used relationship relationships found in [49], while [50], based on testing on bored
[40,41]:
 
G0
PA
¼ Cg
1þe
 n
ðeg  eÞ2 r0m g
PA
where P A is the reference stress (=100 kPa); e is the void ratio, which
is estimated to be 0.69 given the sand dry unit weight
Pr
ð53Þ
and driven model piles, and [2], proposed for driven piles, were
used to define the p—y curves for the sand layer. The p—y
relationship proposed in [50] is:

p¼ 1
kini
y
þ pyL
ð56Þ

cd ¼ 15:4 kN/m3 [35] and assuming the specific gravity Gs ¼ 2:65;

C g ; eg and ng are constants for a given sand, and values of 650, where kini is the initial horizontal subgrade reaction modulus, given
2.17, and 0.45 can be used when specific information is not avail- by:
able [41]; r0m ¼ ðr0v þ 2r0h Þ=3 is the mean effective stress at the loca- kini ¼ gh z ð57Þ
tion being considered. G0 is then calculated as a function of depth z
using Eq. (53). The Poisson’s ratio m for the sand is taken as 0.15 and pL is the limit resistance, given by:
[29,42].
f

The undrained shear strength su of the clay is estimated based PL ¼ BK P c0 zn ð58Þ

oo

on the SPT blow count N SPT [43] using:

where gh is the subgrade modulus gradient, which is suggested to
su ¼ PA f 1 NSPT ¼ 72 kPa ð54Þ be equal to 4329 kN/m3 for medium dense sand and a free head
boundary condition (rotation is not constrained) [50]; B is the pile
where f 1 is a function of the plasticity index of the clay. Since the
diameter; c0 is the effective unit weight of the soil; n = 0.4 for a
clay is of high plasticity, f 1 ¼ 0:045 [41]. The average N SPT in the clay
free-head boundary condition. The coefficient of passive earth pres-
Pr

layer is 16 [see Fig. 13(b)]. The choice of this correlation is not

essential to the results of the analysis as the lateral response of
the pile is mainly governed by the sand layer, whose thickness is
greater than 10 times the pile diameter. The su calculated from
Eq. (54) is then used to estimate the small-strain shear modulus
G0 of the clay [44]:
G0 ¼ 477s0:928
u PA ¼ 25:8 GPa
A Poisson’s ratio of 0.49 was taken for the clay to simulate its
undrained behavior.
The test pile was analyzed with the p—y method and the
method proposed in this paper with the sand modeled as a
Drucker–Prager non-linear elastic, perfectly plastic material and
the clay modeled as a von Mises non-linear elastic, perfectly plastic
sure K P is related to the peak friction angle /p :
1 þ sin /p 1 þ sin 38:5
KP ¼ ¼ ¼ 4:30 ð59Þ
1  sin /p 1  sin 38:5
Fig. 14 shows the load vs. pile head deflection curves obtained
from the lateral load test, the present method and the p—y method
ð55Þ
using the API (1993) criterion [2] and the Kim et al. (2004) criterion
[50]. Both SAM and the API (1993) method provide reasonable pre-
dictions of the pile–soil response. The API (1993) method is more
conservative and produces a softer response than the measured
response. The Kim et al. (2004) method underestimates the reac-
tion load at the pile head. This is likely due to the fact that the
method is based on model pile tests that were likely affected by

Table 1
Constitutive model input parameters.

Sand Clay
G0 (MPa) m csat (kN/m3) /c (°) c (kPa) b (kPa) G0 (MPa) m su (kPa) csat (kN/m3)
a
0.15 18.9 32 0.1 0.03 25.8 0.49 72 20.1
a
G0 for the sand layer varies along depth.
128 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129

160 Table 2
Comparison of the maximum bending moments obtained by different methods.
SAM
Measured Measured SAM API (1993) Kim et al. (2004)
API (1993)
Max bending moment 101 112 150 67
120 Kim et al. (2004)
(kNm)
Load at pile head (kN)

Relative error – +10.9% +48.5% 33.7%

80 transfers the load to the inner pipe through the in-between grout,
which may not enable perfect composite action. Thus, the magni-
tude and the position of the peak bending moment measured from
the inner pipe may be different from the value for the composite
40 cross section.
These results illustrate the need for caution in selecting p—y
curves, since these are often developed based on a limited number
of load tests on model or prototype piles. If relying on p—y curves
0 developed from model pile load tests, scale effects must be taken
0 10 20 30 40
into account. If relying on p—y curves developed based on proto-
Deflection at pile head w (mm)

f
type pile load tests, it is essential for engineers to ascertain
whether the soil was properly characterized and whether the piles

oo
Fig. 14. Load at the pile head vs. deflection at the pile head obtained by
measurement, the present method and the p—y method with API (1993) and Kim were instrumented for every load test in the database on which the
et al. (2004) p—y curves. curve is based. The present method presents an alternative to the
use of p—y curves, requiring instead the determination of parame-
scale and boundary effects. It is worth mentioning that no explicit ters that describe the mechanical response of the soil itself.
guidance is offered on how the friction angle is to be determined
for use in most of the p—y equations used in practice.
Fig. 15 shows the bending moment profiles obtained from the
test measurements, the present method and the p—y methods with
API (1993) and Kim et al. (2004) p—y curves. The maximum bend-
ing moments obtained by these methods and their relative error
with respect to the measured value are compared in Table 2. The
Pr 7. Summary and conclusions

This paper presented a new semi-analytical method (SAM) to

analyze, using continuum mechanics, laterally loaded piles in elas-
tic–plastic soil. The method is based on the principle of virtual
work in incremental form and displacement formulations that link
proposed method produces the best estimate of maximum bending
the displacement in the soil to the pile deflection. The governing
moment. The location of the peak bending moment predicted by
differential equations for the deflection w along the pile axis and
both SAM and the p—y method are lower than the measured one.
the displacement decay functions f ðx; yÞ and gðx; yÞ within the soil
This is probably due to the local grout cracks, as explained in
domain were derived. These equations are solved using an iterative
[36], that occurred on the tensile side of the pile and to the method
algorithm in an incremental manner, which produces profiles of
of measurement of the bending moment. The pile consists of two
deflection, internal shear force and bending moment in the pile
coaxial steel pipes grouted together [35,37] and the bending
and displacement in the soil.
f

moment was measured by strain gages attached to the inner steel

In the proposed method, soil can be modeled as an
oo

pipe. The local grout cracks disrupt the basic measurement

elasto-plastic material. Taking full account of the 3D, non-linear
assumption (Euler–Bernoulli beam assumption) that plane cross
interaction between the pile and soil, the proposed method pro-
sections throughout the pile remain plane during bending. The
vides more realistic simulations of pile response than the p—y
response of the two steel pipes to the external load may be asyn-
method, which treats the soil as a series of non-linear 1D springs.
chronous; the outer pipe responds promptly to the load and then
The method is computationally efficient, matching the speed of a
Pr

p—y method analysis and significantly exceeding that of 3D finite

Bending moment M (kN.m) element analyses while having comparable accuracy.
0 20 40 60 80 100 120 140 160 180 The paper further illustrated that SAM produced reasonable
0
estimates of pile deflection for a well-documented case history;
in contrast, p—y analyses would be very dependent on the choice
of p—y curves, which are often empirically derived and may not
1
apply to a given site.

Acknowledgements
2
Depth (m)

This material is based upon work supported by the National

Science Foundation under Grant No. 0969949. The authors are very
3 grateful for this support. The authors are also grateful for support
provided by Fugro Consultants, Inc.
SAM
4 References
Measured
API (1993)
[1] Ashour M, Norris G. Modeling lateral soil–pile response based on soil–pile
Kim et al. (2004) interaction. J Geotech Geoenvironmental Eng 2000;126:420–8. http://
5 dx.doi.org/10.1061/(ASCE)1090-0241(2000)126:5(420).
[2] Recommended practice for planning, designing and constructing fixed offshore
Fig. 15. Bending moment along depth obtained by measurement, the present platforms – working stress design. 20th ed., American Petroleum Institute;
method and the p—y method with API (1993) and Kim et al. (2004) p—y curves. 1993.
 F. Han et al. / Computers and Geotechnics 70 (2015) 116–129
[3] Brown D, Shie C. Three dimensional finite element model of laterally loaded
piles. Comput Geotech 1990;10:59–79.
[4] Yang Z, Jeremic B. Numerical study of group effects for pile groups in sands. Int
J Numer Anal Methods Geomech 2003;27:1255–76. http://dx.doi.org/
10.1002/nag.321.
[5] Yang Z, Jeremić B. Study of soil layering effects on lateral loading behavior of
piles. J Geotech Geoenvironmental Eng 2005;131:762–70. http://dx.doi.org/
10.1061/(ASCE)1090-0241(2005)131:6(762).
[6] Papadopoulou MC, Comodromos EM. On the response prediction of
horizontally loaded fixed-head pile groups in sands. Comput Geotech
2010;37:930–41. http://dx.doi.org/10.1016/j.compgeo.2010.07.011.
[7] Kim Y, Jeong S. Analysis of soil resistance on laterally loaded piles based on 3D
soil–pile interaction. Comput Geotech 2011;38:248–57. http://dx.doi.org/
10.1016/j.compgeo.2010.12.001.
[8] Wakai A, Gose S, Ugai K. 3-D elasto-plastic finite element analyses of pile
foundations subjected to lateral loading. Soils Found 1999;39:97–111.
[9] Fan C-C, Long JH. Assessment of existing methods for predicting soil response
of laterally loaded piles in sand. Comput Geotech 2005;32:274–89. http://
dx.doi.org/10.1016/j.compgeo.2005.02.004.
[10] Muqtadir A, Desai CS. Three-dimensional analysis of a pile-group foundation.
Int J Numer Anal Methods Geomech 1986;10:41–58. http://dx.doi.org/
10.1002/nag.1610100104.
[11] Chen L, Poulos HG. Analysis of pile–soil interaction under lateral loading using
infinite and finite elements. Comput Geotech 1993;15:189–220.
[12] Mylonakis G, Syngros C, Price TE. Elastodynamic fe analysis of pile response for
oo
combined horizontal and vertical soil heterogeneity. In: 5th GRACM int. congr.
comput. mech., 2005.
[13] Sun K. Laterally loaded piles in elastic media. J Geotech Eng
1994;120:1324–44. http://dx.doi.org/10.1061/(ASCE)0733-9410(1994)120:8
(1324).
[14] Sun K. A numerical method for laterally loaded piles. Comput Geotech
1994;16:263–89. http://dx.doi.org/10.1016/0266-352X(94)90011-6.
Geomech Geoengin 2007;2:183–96. http://dx.doi.org/10.1080/174860207014
01007.
[16] Basu D, Salgado R. Analysis of laterally loaded piles with rectangular cross
sections embedded in layered soil. Int J Numer Anal Methods Geomech
2008;32:721–44. http://dx.doi.org/10.1002/nag.639.
[17] Choi YS, Basu D, Prezzi M, Salgado R. Study on laterally loaded piles with
rectangular and circular cross sections. Geomech Geoengin 2014:1–14. http://
dx.doi.org/10.1080/17486025.2014.902119.
Pr
[15] Basu D, Salgado R. Elastic analysis of laterally loaded pile in multi-layered soil.
[18] Choi YS, Basu D, Salgado R, Prezzi M. Response of laterally loaded rectangular
and circular piles in soils with properties varying with depth. J Geotech
Geoenvironmental Eng 2014;140:04013049. http://dx.doi.org/10.1061/
(ASCE)GT.1943-5606.0001067.
[19] Basu D, Salgado R, Prezzi M. A continuum-based model for analysis of laterally
loaded piles in layered soils. Géotechnique 2009;59:127–40. http://dx.doi.org/
10.1680/geot.2007.00011.
[20] Salgado R, Tehrani FS, Prezzi M. Analysis of laterally loaded pile groups in
f
multilayered elastic soil. Comput Geotech 2014;62:136–53. http://dx.doi.org/
10.1016/j.compgeo.2014.07.005.
oo
[21] Yang K, Liang R. Numerical solution for laterally loaded piles in a two-layer soil
profile. J Geotech Geoenvironmental Eng 2006;132:1436–43. http://
dx.doi.org/10.1061/(ASCE)1090-0241(2006)132:11(1436).
[22] Teh CI, Shen WY. Analysis of laterally loaded pile groups using a variational
approach. Géotechnique 2002;52:201–8. http://dx.doi.org/10.1680/geot.2002.
52.3.201.
[23] Salgado R, Basu D, Prezzi M, Tehran FS. Semi-analytical solutions for laterally
Pr
loaded piles in multilayered soils. In: Proc. 18th int. conf. soil mech. geotech.
eng., 2013, p. No. 2125.
[24] Guo WD, Lee FH. Load transfer approach for laterally loaded piles. Int J Numer
Anal Methods Geomech 2001;25:1101–29. http://dx.doi.org/10.1002/nag.169.
View publication stats
129
[25] Basile F. Discussion: analysis of laterally loaded pile groups using a variational
approach. Géotechnique 2003;53:525–6. http://dx.doi.org/10.1680/geot.2003.
53.5.525.
[26] Fahey M, Carter JP. A finite element study of the pressuremeter test in sand
using a nonlinear elastic plastic model. Can Geotech J 1993;30:348–62. http://
dx.doi.org/10.1139/t93-029.
[27] Pestana JM, Whittle AJ. Formulation of a unified constitutive model for clays
and sands. Int J Numer Anal Methods Geomech 1999;23:1215–43. http://
dx.doi.org/10.1002/(SICI)1096-9853(199910)23:12<1215::AID-NAG29>3.0.
CO;2-F.
[28] Lee J, Salgado R. Analysis of calibration chamber plate load tests. Can Geotech J
2000;37:14–25. http://dx.doi.org/10.1139/t99-061.
[29] Loukidis D, Salgado R. Modeling sand response using two-surface plasticity.
Comput Geotech 2009;36:166–86.
[30] Abbo A, Sloan S. A smooth hyperbolic approximation to the Mohr-Coulomb
yield criterion. Comput Struct 1995;54:427–41.
[31] Yu H. Plasticity and geotechnics. Springer; 2006.
[32] Sloan SW. Substepping schemes for the numerical integration of elastoplastic
stress–strain relations. Int J Numer Methods Eng 1987;24:893–911. http://
dx.doi.org/10.1002/nme.1620240505.
[33] Abbo A. Finite element algorithms for elastoplasticity and consolidation. vol.
1997; 1997.
[34] Yuan B, Chen W, Jiang T, Wang Y, Chen K. Stereo particle image velocimetry
f
measurement of 3D soil deformation around laterally loaded pile in sand. J Cent
South Univ 2013;20:791–8. http://dx.doi.org/10.1007/s11771-013-1550-7.
[35] Brown D, Morrison C, Reese L. Lateral load behavior of pile group in sand. J
Geotech Eng 1988;114:1261–76.
[36] Brown DA, Reese LC. Behavior of a large-scale pile group subjected to cyclic
lateral loading. 1988.
[37] Ochoa M, O’Neill MW. lateral pile-group interaction factors for free-headed
pile groups in sand from full-scale experiments. 1988.
[38] Salgado R, Prezzi M. Computation of cavity expansion pressure and
penetration resistance in sands. Int J Geomech 2007;7:251–65.
[39] Chakraborty T, Salgado R. Dilatancy and shear strength of sand at low. J
Geotech Geoenvironmental Eng 2010;136:527–32. http://dx.doi.org/10.1061/
(ASCE)GT.1943-5606.0000237.
[40] Hardin B, Richart FE. Elastic wave velocities in granular soils. J Soil Mech Found
Div ASCE 1963;89:33–65.
[41] Salgado R. The engineering of foundations. McGraw-Hill; 2008.
[42] Robertson PK. Interpretation of cone penetration tests – a unified approach.
Can Geotech J 2009;46:1337–55. http://dx.doi.org/10.1139/T09-065.
[43] Stroud M. The standard penetration test in insensitive clays and soft rock. In:
Proc 1st Eur symp penetration test, vol. 2; 1974. p. 367–75.
[44] Hara A, Ohta T, Niwa M, Tanaka S, Banno T. Shear modulus and shear strength
of cohesive soils. Soil Found 1974;14:1–12. http://dx.doi.org/10.3208/
sandf1972.14.3_1.
[45] Huang B, Bathurst RJ, Hatami K. Numerical study of reinforced soil segmental
walls using three different constitutive soil models. J Geotech
Geoenvironmental Eng 2009;135:1486–98. http://dx.doi.org/10.1061/
(ASCE)GT.1943-5606.0000092.
[46] Numerical methods in geotechnical engineering, vol. 1. CRC Press; 2014.
[47] Loukidis D. Advanced constitutive modeling of sands and applications to
foundation engineering. 2006.
[48] Stewart DP. User manual PYGMY 2000. Perth: University of Western Australia;
2000.
[49] Dunnavant TW, O’Neill MW. Experimental model for submerged, Stiff Clay. J
Geotech Eng 1989;115:95–114. http://dx.doi.org/10.1061/(ASCE)0733-
9410(1989)115:1(95).
[50] Tak Kim B, Kim N-K, Jin Lee W, Su Kim Y. Experimental load–transfer curves of
laterally loaded piles in Nak-Dong River Sand. J Geotech Geoenvironmental
Eng 2004;130:416. http://dx.doi.org/10.1061/(ASCE)1090-0241(2004)130:
4(416).