Soils and Foundations 2013;53(2):199214

The Japanese Geotechnical Society

Soils and Foundations

www.sciencedirect.com
journal homepage: www.elsevier.com/locate/sandf

A multidirectional py model for lateral sandpile interactions

D. Sua,n, W.M. Yanb
a
College of Civil Engineering, Shenzhen University, Nanshan District, Shenzhen 518060, China
Department of Civil Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong, China

Received 8 May 2012; received in revised form 12 September 2012; accepted 15 October 2012
Available online 12 March 2013

Abstract
The lateral loads applied to pile foundations, as induced by winds or earthquakes, are usually multidirectional. Experimental studies
have indicated that the lateral resistance of the pile under multidirectional paths is generally lower than that under a unidirectional path
and the degree of reduction depends on the characteristics of the loading paths. On the other hand, most currently used py models can
take the soilpile interaction under unidirectional lateral loading into account, but it cannot be applied directly to analyze the response
of piles under multidirectional lateral loading. A multidirectional py model is proposed in this study, which is formulated within the
framework of the bounding-surface elastoplastic theory and consists of two loading mechanisms: the parallel loading and the orthogonal
loading. The model has ve parameters, which are readily available or calibrated. To demonstrate its ability to model soilpile
interactions under both unidirectional and multidirectional lateral loadings, the proposed model is incorporated into a nite-element
program to analyze laterally loaded piles. The responses of piles with different embedment lengths subject to various loading paths are
investigated. The non-coaxial relationship between the force increment and the displacement increment vectors at the pile head under
the multidirectional loading, and the impact of the multidirectional loading on the lateral resistance are well captured in the analyses.
& 2013 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. All rights reserved.
Keywords: py Model; Multidirectional; Sandpile interaction; Elastoplastic; Nonlinear

1. Introduction
Pile foundations, a type of deep foundation, have been
widely used to support high-rise buildings, bridges and
offshore structures. These foundations must be designed for
lateral loads because the structures they support are subjected
to such loads as the result of wind, waves and earthquakes.
In the past several decades, a large number of in-situ and
n
Corresponding author. Tel.: 86 755 2653 5204;
fax: 86 755 2673 2850.
E-mail addresses: sudong@szu.edu.cn (D. Su),
ryanyan@hku.hk (W.M. Yan).
Peer review under responsibility of The Japanese Geotechnical Society.

physical model tests (e.g., McVay et al., 1998; Dyson and

Randolph, 2001; Kim et al., 2004; Lee et al., 2011; Tamura
et al., 2012) have been performed to investigate the response
of piles to lateral loads. Meanwhile, extensive analytical and
numerical approaches (e.g., Ashour et al., 1998; Zhang et al.,
1999; Hsiung, 2003) for investigating the behavior of laterally
loaded piles have also been developed.
The load-transfer approach, often referred to as the
py method, is one of the most popular approaches to
determine the behavior of laterally loaded piles. In this
approach, the pile is modeled as an elastic member while
the soil is idealized as a discrete set of independent nonlinear springs that describe the relationship between the
local lateral soilpile resistance and the lateral relative
displacement. A number of py relationships have been
proposed for sands (e.g., Reese et al., 1974; Wesselink
et al., 1988), silts (e.g., Reese and Van Impe, 2001) and

0038-0806 & 2013 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.sandf.2013.02.002

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

200

Nomenclature
p
y
p
y
dye
dyp
ke
dp
dp0
dp00
n
0
dyp
kp1
m
pu
00
dyp
kp2
dl1
dl2
pm

soil resistance per area

lateral displacement of the pile
vector of soil resistance per area
vector of lateral displacement of the pile
elastic component of the displacement
increment
plastic component of the displacement
increment
elastic resistance coefcient
resistance increment
part of resistance increment in parallel with the
current resistance p
part of resistance increment orthogonal to the
current resistance p
unit vector in parallel with p
plastic displacement increment induced by dp0
plastic resistance coefcient associated with the
parallel loading
direction of the plastic displacement increment
induced by dp00
ultimate soil lateral resistance
plastic displacement increment induced by dp00
plastic resistance coefcient associated with the
orthogonal loading
loading index associated with the parallel
loading
loading index associated with the orthogonal
loading
maximum soil resistance in the history

clays (e.g., Matlock, 1970; Reese and Welch, 1975).

However, most of these models were developed based on
the observed behavior of piles subject to monotonic lateral
loading. These models are generally semi-empirical and
can only take into account the soilpile interaction under
unidirectional lateral loading.
In reality, the lateral loads and motions applied to pile
foundations are often multidirectional, and in some cases
the amplitudes of the two horizontal orthogonal components are comparable (Su and Li, 2008). Mayoral et al.
(2005) conducted a series of model pile tests in clay under
various multidirectional displacement paths. The results
showed that the shape of the py curves is strongly affected
by the loading path. Su (2011, 2012a) performed a number
of laboratory-scale model tests on a pilesand system
under both unidirectional and multidirectional loading
paths. The results indicate that the lateral resistance
of the pile under the multidirectional path is generally
lower than that under the unidirectional path. The degree
of reduction depends on the characteristics of the
loading paths.
As Su (2005) noted, the principle of superposition does
not hold when the interaction between the pile and the

relocatable projection center on the bounding

surface
Euclidian distances
r, r
z
depth below soil surface
b
pile diameter
Zh
subgrade modulus
h1
scaling factor for the modulus kp1
sp
passive earth pressure of soil
cp
model constant associated with pu
k
model constant associated with evolution of m
yr
referenced displacement (yr pu/kini)
h2
scaling factor for the modulus kp2
E
Youngs modulus of the pile
I
moment of inertia of area of the pile
dy1, dy2 pile displacement increments along the X1
direction and the X2 direction respectively
dp1, dp2 soil resistance increments along the X1 direction and the X2 direction respectively
l
element length
dyij , dfij lateral displacement and rotation at the ith
node (i 1, 2, 3) along the Xj direction (j 1,
2) respectively
dFji , dMji external lateral force and moment at the ith
node (i 1, 2, 3) along the Xj direction (j 1, 2)
respectively
ksijk
components of the stiffness matrix to take into
account the multidirectional soilpile interaction at the ith node
kjk
components of the stiffness matrix of py
model
a

surrounding soil is nonlinear. Therefore, most currently

used py models cannot be directly applied to analyze the
responses of pile foundations under multidirectional lateral
loadings. This study proposes a multidirectional py model
for laterally loaded piles in sand, formulated within the
framework of the bounding-surface elastoplastic theory.
The model is incorporated into the nite-element program
Code for Analyzing Multidirectionally Loaded Piles
(CAMLP) that was developed to analyze the response of
piles under lateral loadings (Su, 2012b). This ability of this
model to capture the key features of the soilpile interactions under both unidirectional and multidirectional lateral
loadings is demonstrated by comparing numerical simulation results to experimental observations.
2. Formulation of the multidirectional py model
2.1. Framework
In the model, y represents the lateral displacements of
the pile at a given depth and p represents the soil resistance
per area at the same depth. For a pile under multidirectional lateral loading, both p and y are vectors in a plane.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

201

where pu is the ultimate soil lateral resistance at the given

depth and k is a model constant. Eqs. (6) and (7) show that
(1) when 9p9=0, f(p)=1 and m=dp00 /9dp00 9; (2) when
9p9=pu, f(p)=0 and m=p/9p9=n; and (3) in between, the
direction of the plastic displacement increment induced by

dp'
dp
dp"

p
1
p

=1.0

0.8

=0.5

0.6

f(p)

Fig. 1. Decomposition of the resistance increment dp.

=5.0
0.4

=3.0

The displacement increment dy induced by the resistance

increment dp can be decomposed into two parts:
dy dye dyp

where dye is the elastic component of the displacement

increment and dyp is the plastic component. For the elastic
component, the relationship with dp is
dp ke dye

=2.0
0

0.2

0.4

0.6

0.8

Fig. 2. Effect of parameter k on the relationship between 9p9/pu and f(p).

Bounding surface p p = 0

p dp

and the analytical expression of the orthogonal part is

dp00 dpdp0 dpdpT Unn

in Eqs. (3) and (4), n is the unit vector in parallel with p,

calculated by n=p/9p9.
Generally, both dp0 and dp00 can generate plastic
displacement increments. For convenience, dp0 and dp00
are referred to as parallel and orthogonal loading, respectively. The plastic displacement increment induced by dp0 is
coaxial with dp0 , as encountered in cases of unidirectional
loading, and is determined by
0

dyp dp0 =kp1

Bounding surface p pm = 0

where dyp is the plastic displacement increment induced by

dp0 and kp1 is the plastic resistance coefcient associated
with the parallel loading.
The direction of the plastic displacement increment
induced by dp00 depends on both the direction of dp00 and
the direction of p, which is assumed to be determined by
dp00
p
 f p   1f p
 p
dp00 

m 

|p|/pu

where ke is the elastic resistance coefcient.

To obtain the plastic component dyp, the resistance
increment dp is decomposed into two parts. One part is
parallel to the current resistance p and the other is
orthogonal to it, as illustrated in Fig. 1. The analytical
expression of the parallel part is
dp0 dpT Unn

0.2

dp
p

in Eq. (6), f(p) is a function of p, which takes the form:

   
p p
k
f p sin
1
7
pu 2

Fig. 3. Mapping rule for p: (a) p  dp 40; (b) p  dpo0.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

202

the value of f(p) decreases faster with the increasing 9p9/pu

when k becomes larger. Therefore, the direction of the
plastic displacement increment induced by the orthogonal
loading approaches the direction of p at a faster rate with
larger values of k.
Dening the direction vector m allows the plastic
displacement increment associated with the orthogonal
loading to be determined by

the orthogonal loading gradually moves away from the

direction of dp00 toward the direction of p as 9p9 increases.
Assumptions about the evolution of the direction
described in Eq. (6) are inspired by soil behavior under
stress rotation in which the plastic strain increment
gradually moves away from the direction of the stress
increment and toward the direction of stress as the stress
ratio increases (Gutierrez et al., 1991). As shown in Fig. 2,
1.2

dyp 00
h =3.0

h =0.5

k /k

h =0.1
0.6

h =0.3

dyp dyp 0 dyp 00

0.4

Hyperbolic

dp0 dp00 T Um

m
kp2
kp1
dpT Un
dpdpT UnnT Um

n
m
kp1
kp2
dl1 n dl2 m

0.2
0
0.001

where kp2 is the plastic resistance coefcient associated

with the orthogonal loading.
According to Eqs. (3)(5) and (8)

h =1.0

0.8

dp00 T Um
m
kp2

0.01

0.1

10

y/y

Soil resistance along X2 direction (kPa)

Displacement along X2 direction (mm)

Fig. 4. Coefcient reduction as a function of the parameter h1.

-2

-4
-2

150
100
50
h2=0.08
0
-50
h2=0.1
-100
h2=1.0
-150
-2

Displacement along X 1 direction (mm)

150
100
50
0

h2=0.08

-50
h2=0.1

-100

h2=1.0
-150
-150 -100

-50

50

100

150

Soil resistance along X1 direction (kPa)

Soil resistance along X2 direction (kPa)

Soil resistance along X1 direction (kPa)

Displacement along X1 direction (mm)

150
100
50
0

h2=0.08

-50
h2=0.1
-100
h2=1.0
-150
-4

-2

Displacement along X2 direction (mm)

Fig. 5. Effect of parameter h2 on the py relationships under the circular path: (a) the displacement path; (b) the trajectory of reaction pressure; (c) the
py relationship along the X1 direction; and (d) the py relationship along the X2 direction.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

where dl1 and dl2 are the loading indices associated with
the parallel and orthogonal loadings, respectively.
To consider the effect of loading history on the lateral
soilpile interactions under cyclic loadings, a bounding
surface with the expression
 
ppm 0
10
is introduced, where pm is the maximum historic soil
resistance at the given depth. To map the current resistance
p on the bounding surface, a relocatable projection
center a on the bounding surface is chosen, such that the
vectors p and a are collinear and (p  a)  dp 40, as
illustrated in Fig. 3(a) and (b). Correspondingly, two
Euclidian distances r and r, which are necessary for the
evaluation of kp1, are dened. For the case of p  dp 4 0,
r pm 9p9 and r 2pm , as shown in Fig. 3(a). For the
case of p  dp o 0, however, r pm  9p9 and r 2pm , as
shown in Fig. 3(b).
2.2. Elastic resistance coefficient ke

203

for such a unidirectional loading, both p and y can be

considered to be scalars. From Eqs. (1), (2), (5) and (14),
one has
!


1
1
1
1

 dp
15

dp
dy
ke kp1
ke h1 ke pu =p1
the integration of both sides of Eq. (15) yields

1
p
1
1
ke 1 
ln1r
y
h1 h1 r

16

where r p/pu. Eq. (16) can be converted into a normalized

resistance coefcient reduction curve with h1 as a variable,
as shown in Fig. 4, where the resistance coefcient is
normalized to the initial coefcient while the displacement
is normalized by a reference displacement yr (yr pu/kini).
Curves with different h1 values are compared to the curve
evaluated from the hyperbolic py relationship proposed

It is assumed that the elastic resistance coefcient is

independent of the displacement and always equals the
initial resistance coefcient kini. Furthermore, the initial
resistance coefcient is assumed to increase linearly with
depth as follows:
Z
ke kini h z
11
b
where z is the depth below the ground surface, b is the pile
diameter and Zh is the subgrade modulus. The values of Zh
proposed by Terzaghi (1955) are 11003300 kN/m3 for
loose sands, 330011,000 kN/m3 for medium sands and
11,00023,400 kN/m3 for dense sands. According to Kim
et al. (2004), the values of Zh also depend on factors such as
the pile installation method.
2.3. Plastic resistance coefficient kp1
The plastic resistance coefcient associated with the
parallel loading takes the following form:


pu r
12
kp1 h1 ke
U 1
pm r
where h1 is a scaling factor for the modulus.
For simplicity, the following expression, which is similar
to that proposed by Ismael (1990), is adopted in this
research:
pu cp sp

13

where sp is the passive earth pressure of soil around the

pile and cp is a model constant.
To gain insight into the inuence of parameter h1 on the
models behavior, a monotonic and unidirectional loading
is considered. In this case, dp00 0, r r, and therefore,


pu
kp1 h1 ke
1
14
pm

Fig. 6. Soilpile system for the analysis of a multidirectional laterally

loaded pile.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

204

by Kondner (1963) in the form of

y
p
1=kini y=pu

interdependent and coupled, which are determined by the

multidirectional py model described above.
Due to the nonlinear nature of the soilpile system, the
governing equations of the laterally loaded pile are
expressed in the incremental form for the X1 direction
and the X2 direction as follows:

17

it can be seen that the kp1 used here is capable of modeling

the py relationships with different degrees of nonlinearity,
and that the model provides more exibility than the
hyperbolic function.

EI

@4 dy1
dp1 0
@z4

19

EI

@4 dy2
dp2 0
@z4

20

2.4. Plastic resistance coefficient kp2

The plastic resistance coefcient associated with the
orthogonal loading takes the following form:


pu
1
18
kp2 h2 ke
p

where E is Youngs modulus of the pile; I is the second

moment of area of the pile cross sections; y1, p1 and y2, p2
are the pile displacement and soil resistance in the X1 and
X2 directions, respectively; and d denotes the increment.
To solve Eqs. (19) and (20) numerically, the nite element
program CAMLP was written in Matlab (Su, 2012b). The
pile was rst divided into a number of elements with each
node connected to two perpendicular springs, as shown in
Fig. 6. There are two degrees-of-freedom (displacement and
rotation) along each direction, i.e. four degrees-of-freedom at
each node. The Hermite interpolation function was adopted
to construct the element stiffness matrices and the global
stiffness matrix (Fung and Tong, 2001). Given that the two
springs at each node are coupled, the matrices along the two
directions should be assembled into a system of simultaneous
equations for simultaneously solving the lateral displacements along the two directions. The following is an example
of the assembled algebraic equations after introducing the
stiffness matrices of the py model (see the appendix for the
derivation of the stiffness matrix) into the global stiffness
matrix for a case of three elements with an equal element
length of l:

where h2 is a scaling factor for the modulus. Similar to kp1,

kp2 approaches zero as 9p9 approaches pu. To illustrate the
effect of h2 on the models behavior, a multidirectional
loading case with a circular pile displacement is simulated
and the results are presented in Fig. 5. In the numerical
simulations, all parameters other than h2 remain the same.
It can be seen from the gure that as h2 increases, the
resistance of soil to pile movement increases. Yet, the
characteristics of the py curves with different h2 values are
similar in both directions.
3. Finite element formulation for analyzing a
multidirectional laterally loaded pile
In analyzing a unidirectional laterally loaded pile, the
pile is typically modeled as an elastic beam with a series
of discrete and independent springs connecting the pile to
the surrounding soil. To analyze a pile under multidirec0

12a ks111
B
B
6la
B
B ks1
B
21
B
B
B
B 12a
B
B
B
6la
B
B
B
B
B
B
B
B
B
B
B
B
@

6la
4l 2 a

ks112

12a
6la

6la
2l 2 a

12a ks122

6l a

12a

6la

6l a

4l 2 a

6la

2l 2 a

24a ks211

6la
2

ks212
2

2l a

8l a
12a

6la

6la

2l 2 a

ks221

12a

6la

6la

2l 2 a

24a ks222

12a
8l 2 a

12a
6la

6la
12a ks311
6la

6la
2l 2 a
12a

6la

6la

tional lateral loadings, two orthogonal nonlinear springs

representing the multidirectional soilpile interactions are
necessary to connect the pile to the surrounding soil at a
given depth, as shown in Fig. 6. Furthermore, the py
relationships of the two springs at the same depth are

2l a

ks321

6la
4l 2 a

ks312
12a ks322
6la

1
10 1 1 0
dy1
dF 1
C B 11 C
1
CB
df1 C B dM1 C
CB
C B
C
CB
CB
dy12 C
dF21 C
C B
C
CB
C B
1
B
C
CB
df2 C
B dM21 C
CB
B
C
C
CB 2 C B
CB dy1 C B dF 2 C
B 1 C
CB
B
C
CB df2 C
B dM12 C
CB 1 C
B
C
CB 2 C

C
dy C B dF22 C
6la C
C
CB
B 22 C B
B
C
B df2 C B dM 2 C
C
2l 2 a C
2
B
C
C
CB 3 C B
CB dy1 C B dF 3 C
B 1 C
CB
B
C
CB 3 C
CB df1 C
dM13 C
C B
B
C
CB
C
dy3 C B dF 3 C
6la C
AB
@ 2A @ 2 A
dM23
df32
4l 2 a

21

where a EI/l3, dyij and dfij are the lateral displacement

and rotation at the ith node (i 1, 2, 3) along the Xj
direction (j 1, 2), dFji and dMji are the external lateral
force and moment at the ith node (i 1, 2, 3) along the Xj
direction (j 1, 2) and ksijk (j 1, 2 and k 1, 2) are the

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

components of the stiffness matrix to take into account the

multidirectional soilpile interactions at the ith node (i 1,
2, 3). Their relationships with the components of the
stiffness matrix for the py model are as follows:
ksjk bUl 0 Ukjk

205

of medium dense sand with embedment lengths of 350 mm,

400 mm, 450 mm or 500 mm. The loading point on the pile
was 200 mm above the soil surface, at which only lateral
forces and no moments were applied and the pile head was

22
10
Displacement along X2 direction (mm)

where b is the diameter or width of the pile, l0 l/2 for end

nodes and l0 l for other nodes and kjk are the components
of the stiffness matrix for the py model.
With the imposed force and/or displacement boundary
conditions on the pile, algebraic equations such as Eq. (21)
can be solved and the piles response to multidirectional
lateral loading can be obtained.
4. Simulation of pile responses under unidirectional and
multidirectional lateral loadings
To validate the capacity of the multidirectional py
model, the results of model tests with different embedment
lengths for a pile under various loading paths (Su, 2011,
2012a) were reproduced using the program described
above. The model pile in these tests was a stainless steel
tube with a 38 mm outer diameter and a 0.8 mm wall
thickness. The properties of the pile are presented in
Table 1. The pile was driven into soil models consisting
Table 1
Properties of the steel pile.

A
5

-5
C

-10

0
5
10
Displacement along X1 direction (mm)

15

Fig. 8. Displacement paths at the pile head for unidirectional and

multidirectional lateral loading tests.

Table 2
Summary of the model parameters.

Property

Value

Parameter

Value

Youngs modulus, E (GPa)

Outer diameter, b (mm)
Wall thickness, t (mm)
Moment of inertia, I (m4)
Flexural rigidity, EI (N  m2)

210
38
0.8
1.6  10  8
3360

Zh
h1
h2
cp
k

22,000 kN/m3
0.5
0.1
10
3

Servo motor
motor
Servo

Ball screw
screw
Ball

Slide unit

Ball screw
screw
Ball

Motionguide
guide
Motion
Motionguide
guide
Motion

Fig. 7. The biaxial motion platform.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

206

Moment (Nm)

300

20

40

60

80

100

200
0.1

100
0

Depth (m)

Force along X1 direction (N)

Numerical result
-100

Experimental result

Experimental result
Numerical result

0.2

0.3

-200
0.4

-300
-5

0
5
10
Displacement along X1 direction (mm)

15
0.5

300

200

200

100
0
elliptical
-100

circular

-200
-300
-5

Force along X2 direction (N)

Force along X1 direction (N)

300

0
5
10
Displacement along X 1 direction (mm)

100
0
-100

300

300

200

200

100
0
-100
elliptical
-200
-300
-10

circular
-5

Displacement along X2 direction (mm)

10

elliptical

circular

-200
-300
-5

15

Force along X2 direction (N)

Force along X1 direction (N)

Fig. 9. Comparison of results for the unidirectional loading test: (a) forcedisplacement relationship at the pile head and (b) moment distribution along
the depth at X1 10 mm.

0
5
10
Displacement along X1 direction (mm)

15

100
0
-100
-200
-300
-10

elliptical
circular

-5
0
5
Displacement along X2 direction (mm)

10

Fig. 10. Comparison of results for the multidirectional loading tests: (a) experimental results along the X1 direction; (b) numerical results along the X1
direction; (c) experimental results along the X2 direction; and (d) numerical results along the X2 direction.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

allowed to rotate freely. The testing device was a computer

controlled biaxial motion platform developed at the
Geotechnical Center of Shenzhen University, as shown in
Fig. 7. The main components of the device include a slide
unit, two ball screws, two linear motion guides, two servo
motors and two motor drivers. Through a program written

in a high-level programming language, the two servo

motors can be controlled independently and simultaneously, such that the slide unit can move along the two
horizontal axes independently, alternatively or simultaneously. This allows various displacement paths, including
regular and irregular, to be achieved at the pile head. In
300
Force along X2 direction (N)

300
Force along X2 direction (N)

207

200
100
0
-100
elliptical

-200

200
100
0
-100
elliptical

-200

circular

circular

5
0

20

40N

-5

5
10
Displacement along X1 direction (mm)

5
0

20

40N

-5

-300
-300 -200 -100 0
100 200
Force along X1 direction (N)

Displacement along X2 direction (mm)

300

5
10
Displacement along X1 direction (mm)

Displacement along X2 direction (mm)

Displacement along X2 direction (mm)

Displacement along X2 direction (mm)

-300
-300 -200 -100 0
100 200
Force along X1 direction (N)

300

5
0

20

40N

-5

5
10
Displacement along X1 direction (mm)

5
0

20

40N

-5

5
10
Displacement along X1 direction (mm)

Fig. 11. Comparison of forces at the pile head: (a) measured force trajectories; (b) calculated force trajectories; (c) measured force increment vectors for
the circular loading; (d) calculated force increment vectors for the circular loading; (e) measured force increment vectors for the elliptical loading; and
(f) calculated force increment vectors for the elliptical loading.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

208

90

30
elliptical
unidirectional

15
0
-15
-30
-45
-45

circular

Soil resistance along X2 direction (kPa)

Soil resistance along X2 direction (kPa)

45

elliptical

0
-30

unidirectional
circular

-60

-60
-30
0
30
60
90
Soil resistance along X1 direction (kPa)

180

100
50
unidirectional

0
elliptical
circular

-100
-150
-150 -100 -50
0
50
100 150
Soil resistance along X1 direction (kPa)

Soil resistance along X2 direction (kPa)

Soil resistance along X2 direction (kPa)

30

-90
-90

-30
-15
0
15
30
45
Soil resistance along X1 direction (kPa)

150

-50

60

120
60

elliptical

0
-60

circular unidirectional

-120
-180
-180 -120 -60
0
60
120 180
Soil resistance along X1 direction (kPa)

Fig. 12. Evolution of soil resistances under different loading paths: (a) z 100 mm; (b) z 200 mm; (c) z 300 mm; and (d) z 400 mm.

the tests, the moving speed of the slide unit was controlled
to be a constant of 0.1 mm/s.
4.1. Response of piles under regular displacement paths
In this section, three tests with regular displacement
paths at the pile head were simulated. These tests included
a unidirectional, an elliptical and a circular path as shown
in Fig. 8. In all of these tests, the free and embedment
lengths of the pile were 200 mm and 500 mm, respectively.
The details of these tests have been presented in Su (2011).
In the numerical simulations, the 700 mm long pile was
evenly divided into 70 elements, i.e., 10 mm in length for
each element. The loading point was at the pile head (the
1st node). Two coupled springs representing the multidirectional py model were connected to each node starting
from the 21st node. As in the model tests, the numerical
tests were displacement-controlled with the three paths
presented in Fig. 8 evenly divided into 1000 increments.
The values of the py model parameters are summarized in
Table 2. As the model pile was driven into the sand during
the tests, the sand around the pile was potentially densied, such that the value of Zh for dense sands, instead of

that for medium dense sands as suggested by Terzaghi

(1955), was chosen. The values of h1 and cp were determined by a trial and error approach to obtain a reasonable
agreement between the measured and calculated force
displacement relationships at the pile head for the unidirectional test, in terms of both the initial stiffness and the
maximum force, as shown in Fig. 9(a). The value of h2 was
chosen such that the calculated maximum force along the
X1 direction in the circular test matched the measured
value, as shown in Fig. 10(a) and (b).
Fig. 9(a) compares the measured and computed force
displacement relationships at the pile head for the unidirectional test while Fig. 9(b) compares the moment distributions along the pile when the displacement at the pile
head was 10 mm. While the results shown in Fig. 9(a) are
in good agreement with the results from the virgin loading
stage, they are not well correlated with those from the
unloading phase. During the model test, a cavity around
the pile near the soil surface developed due to the collapse
of the sand behind the pile. Clearly, the numerical modeling was not able to take this phenomenon into account.
This was probably what caused the discrepancy in the
unloading stage. Fig. 9(b) shows that the numerical

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

300

300
L=0.5 m

L=0.45 m
Force along X1 direction (N)

Force along X1 direction (N)

L=0.5 m
200

L=0.4 m
100

L=0.35 m

0
-100
-200
-300
-5

209

200

L=0.45 m
L=0.4 m

100

L=0.35 m

0
-100
-200

0
5
10
Displacement along X1 direction (mm)

-300
-5

15

0
5
10
Displacement along X 1 direction (mm)

15

Fig. 13. Comparison of results for unidirectional loading tests with different embedded pile depths: (a) experimental results and (b) calculated results.

200
100

300

L=0.5 m
L=0.45 m
L=0.4 m
L=0.35 m

0
-100
-200
-300
-5

0
5
10
Displacement along X1 direction (mm)

Force along X1 direction (N)

Force along X1 direction (N)

300

-200

-5
0
5
Displacement along X2 direction (mm)

10

Force along X2 direction (N)

Force along X2 direction (N)

L=0.5 m
L=0.45 m
L=0.4 m
L=0.35 m

-100

-300
-10

L=0.5 m
L=0.45 m
L=0.4 m
L=0.35 m

0
-100
-200

0
5
10
Displacement along X1 direction (mm)

15

300

200

100

-300
-5

15

300

100

200

200
100

L=0.5
L=0.45 m
L=0.4 m
L=0.35 m

0
-100
-200
-300
-10

-5
0
5
Displacement along X2 direction (mm)

10

Fig. 14. Comparison of results for the circular loading tests with different embedded pile depths: (a) experimental results along the X1 direction;
(b) calculated results along the X1 direction; (c) experimental results along the X2 direction; and (d) calculated results along the X2 direction.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

210

Displacement along X1 direction (mm)

modeling can effectively reproduce the moment response

of the free-head short pile, with respect to the distribution
along the pile and the maximum value.
Fig. 10 compares the displacementforce relationships
along the X1 and X2 directions for the elliptical and
circular tests. The major features of the curves observed
in these tests were successfully captured by the numerical
model: (1) the maximum force along the X1 direction in the
circular test was smaller than it was in the elliptical test,
and the latter was less than that in the unidirectional
test. (2) Conversely, the maximum force along the X2

direction in the circular test was larger than it was in the

elliptical test. (3) Unlike the curve in the unidirectional
test presented in Fig. 9(a) in which the maximum force
occurred at the maximum displacement (10 mm), the
maximum force occurred before the displacement reached
its maximum value along both the X1 and X2 directions in
the multidirectional tests, especially in the circular test.
Fig. 11 compares the measured and computed force
trajectories for the circular and elliptical tests. It can be
seen that the shapes of the latter resemble that of the
former, although the forces are somewhat overestimated.

15
10
5
0
-5
-10
-15

0.2

0.4
0.6
Number of cycles

0.8

0.2

0.4
0.6
Number of cycles

0.8

0.2

0.4
0.6
Number of cycles

0.8

Force along X1 direction (N)

400

200

-200

-400

Force along X1 direction (N)

400

200

-200

-400

Fig. 15. Comparison of results for the unidirectional irregular loading test: (a) time history of displacement; (b) measured force at the pile head; and (c)
calculated force at the pile head.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

ratio of the mobilized maximum soil resistance to pu

decreased with the increasing depth. (2) Similar to the
displacement paths at the pile head, the soil resistance
trajectories at different depths moved in a clockwise
direction in the elliptical and circular tests, but the shapes
of the latter were different than those of the former due to
the nonlinear characteristics of the soilpile interactions.
(3) The mobilized maximum soil resistance in the unidirectional test was greater than that in the elliptical and
circular tests at all of the depths other than 0.1 m, which
was consistent with the observed larger resistance at the
pile head in the unidirectional test.
4.2. Response of piles with different embedment lengths
The unidirectional and circular tests were conducted on
piles with embedment lengths of 450 mm, 400 mm and

15
10
5
0
-5
-10
0

0.2

0.4

0.6

0.8

Displacement along X direction (mm)

Displacement along X direction (mm)

In Fig. 11(c)(f), the displacement paths are divided evenly

into 20 segments, and the force increment vector in each
displacement segment is calculated and plotted. The
following observations from the tests were well reproduced
in the simulation: (1) in both the circular and elliptical
tests, the direction of the force increment vector varied
continuously and always deviated from the direction of the
displacement increment vector. (2) The force increment
vector often pointed forward along the moving direction
and tended to bend toward the inner side of the displacement loop. (3) In the elliptical test, variations in the
magnitude and direction of the force increment vector
were more signicant than those in the circular test.
Fig. 12 shows the evolution of the calculated soil
reaction pressure at depths of 100 mm, 200 mm, 300 mm
and 400 m in the three tests with the circles representing pu
at such depths. It can be seen that (1) in all of the tests, the

-15

15
10
5
0
-5
-10
-15

0.2

Number of cycles

Force along X direction (N)

Force along X direction (N)

0.8

-200

0.2

0.4

0.6

0.8

0.8

0.8

200

-200

-400

0.2

Number of cycles

0.4

0.6

Number of cycles

400
Force along X direction (N)

400
Force along X direction (N)

0.6

400

200

200

-200

-400

0.4

Number of cycles

400

-400

211

0.2

0.4

0.6

Number of cycles

0.8

200

-200

-400

0.2

0.4

0.6

Number of cycles

Fig. 16. Comparison of results for the multidirectional irregular loading test with different displacement amplitudes along the two directions: (a) time
history of displacement along the X1 direction; (b) time history of displacement along the X2 direction; (c) measured force at the pile head along the X1
direction; (d) calculated force at the pile head along the X1 direction; (e) measured force at the pile head along the X2 direction; and (f) calculated force at
the pile head along the X2 direction.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

212

calculated values of 120.4 N, 165.2 N, 215.6 N and

268.2 N. The maximum lateral force at the pile head
increased with an increase in the embedment length.
The comparison of the measured and calculated results for
the circular tests in Fig. 14 suggests that the effect of
embedment length on the forcedisplacement relationships
along both directions was generally well-captured. As the
embedment length increased, the maximum force along
both directions increased. The calculated resistance along
the X1 direction was 76.6 N, 117.5 N, 167.6 N and 201.0 N
for embedment lengths of 350 mm, 400 mm, 450 mm and
500 mm, respectivelyall of which were smaller than the
corresponding values calculated from the unidirectional tests
with the same embedment lengths. This was consistent with
the experimental results presented in Figs. 13(a) and 14(a).

Displacement along X2 direction (mm)

Displacement along X1 direction (mm)

350 mm. The measured forcedisplacement relationships at

the pile head, together with those from the tests with a pile
embedment length of 500 mm, are presented in Fig. 13(a)
for the unidirectional tests and in Fig. 14(a) and (c) for the
circular tests. The tests were simulated by the CAMLP
program (Su, 2012b). Note that the values of the multidirectional py model parameters were the same as those
used in the last section.
Fig. 13 compares the measured and computed results for
the unidirectional tests, which show that the effect of the
embedment length on the forcedisplacement relationship
and the lateral pile resistance was well captured. The measured maximum forces were 114.2 N, 159.8 N, 217.9 N
and 270.3 N for embedment lengths of 350 mm, 400 mm,
450 mm and 500 mm, respectively with corresponding

15
10
5
0
-5
-10
-15

0.2

0.4

0.6

0.8

15
10
5
0
-5
-10
-15

0.2

Number of cycles

Force along X1 direction (N)

Force along X1 direction (N)

-200

0.2

0.4

0.6

0.8

0.8

0.8

200

-200

-400

0.2

Number of cycles

0.4

0.6

Number of cycles

400
Force along X2 direction (N)

400
Force along X2 direction (N)

0.8

400

200

200

-200

-400

0.6

Number of cycles

400

-400

0.4

0.2

0.4

0.6

Number of cycles

0.8

200

-200

-400

0.2

0.4

0.6

Number of cycles

Fig. 17. Comparison of results for the multidirectional irregular loading test with equal displacement amplitudes along the two directions: (a) time history
of displacement along the X1 direction; (b) time history of displacement along the X2 direction; (c) measured force at the pile head along the X1 direction;
(d) calculated force at the pile head along the X1 direction; (e) measured force at the pile head along the X2 direction; and (f) calculated force at the pile
head along the X2 direction.

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

213

4.3. Response of piles under irregular paths

Appendix

In engineering practice, pile foundations are often

subjected to irregular loading, such as those induced by
ocean waves or earthquakes. In the model tests, irregular
displacement paths, which were obtained by doubly integrating the typical time histories of acceleration recorded
in an earthquake and ltering out the low-frequency
components, were applied to the pile head to investigate
its response under irregular loading (Su, 2012a). In this
section, three tests conducted on a pile with an embedment
length of 500 mm under different displacement paths
were simulated using the values of the model parameters
presented in Table 2.
Fig. 15 compares the measured and calculated force
response at the pile head when the irregular displacement was applied only along the X1 direction. Comparing
the time histories of the force presented in Fig. 15(b) and
(c) revealed the successful reproduction of both the evolution and the maximum value of force. Figs. 16 and 17
compare the responses under multidirectional irregular
loading. In the test presented in Fig. 16, the amplitude
of displacement along the X2 direction was about 0.2
times that along the X1 direction while the amplitudes
along the two directions were the same in the test
presented in Fig. 17. The variation of force against
the number of cycles was simulated reasonably well,
although the magnitudes of force were occasionally slightly
overestimated.

Derivation of the stiffness matrix: according to Eq. (9),

one has

5. Conclusions
A multidirectional py model formulated within the
framework of the bounding-surface elastoplastic theory is
proposed. This model consists of two loading mechanisms that decompose the soil resistance increment into
two parts: (1) parallel to and (2) orthogonal with the
current resistance vector. The model is incorporated into a
nite-element program for analyzing piles under lateral
loadings. Its capacity in modeling soilpile interactions
under unidirectional and multidirectional lateral loadings
has been validated through the successful simulation of
the response of piles with different embedment lengths
that have been subjected to various displacement paths.
Although the model presented in this paper was designed
specically for sandpile interactions, it can be easily
extended to analyze claypile interactions with appropriate
modications to the expressions for coefcients such
as ke and pu.
Acknowledgments
The authors acknowledge the nancial support provided
by the National Natural Science Foundation of
China under Grant nos. 50978165 and 51178272, and the
Shenzhen Municipal Science and Technology Plan under
Grant no. JC200903130268A.

dl1

dpT Un
dp1 n1 dp2 n2

kP1
kP1

23

and
dpdpT UnnT Um
kP2
dp1 n22 m1 n1 n2 m2 dp2 n21 m2 n1 n2 m1

kP2

dl2

24

substituting Eqs. (23) and (24) into Eq. (9), one has
dp1
dp1 dp1 n21 dp2 n1 n2
dl1 n1 dl2 m1

ke
ke
kP1
dp1 n22 m21 n1 n2 m1 m2 dp2 n21 m1 m2 n1 n2 m21

kP2


1
n21
n22 m21 n1 n2 m1 m2

dp1
ke kP1
kP2


n1 n2 n21 m1 m2 n1 n2 m21

dp2 Adp1 Bdp2

kP1
kP2

dy1

25
where
1
n2
n2 m2 n1 n2 m1 m2
1 2 1
ke kP1
kP2
and
n1 n2 n21 m1 m2 n1 n2 m21
B

kP1
kP2
A

and
dp2
dp2 dp1 n1 n2 dp2 n22
dl1 n2 dl2 m2

ke
ke
kP1
dp1 n22 m1 m2 n1 n2 m22 dp2 n21 m22 n1 n2 m1 m2

kP2


n1 n2 n22 m1 m2 n1 n2 m22

dp1
kP1
kP2


1
n2
n2 m2 n1 n2 m1 m2

2 1 2
dp2 Cdp1 Ddp2
ke kP1
kP2
26

dy2

where
n1 n2 n22 m1 m2 n1 n2 m22

kP1
kP2
and
1
n2
n2 m2 n1 n2 m1 m2
D
2 1 2
ke kP1
kP2

from Eqs. (25) and (26), one has

0
1
D
B
!
!
dp1
B ADBC ADBC C dy1
B
C
@ C
A dy2
A
dp2
ADBC ADBC

27

214

D. Su, W.M. Yan / Soils and Foundations 53 (2013) 199214

therefore, the stiffness matrix can be expressed as

0
1
D
B
!
k11 k12
B ADBC ADBC C
C
B
@ C
A
A
k21 k22
ADBC ADBC

28

References
Ashour, M., Norris, G., Pilling, P., 1998. Lateral loading of a pile in
layered soil using the strain wedge model. Journal of Geotechnical and
Geoenvironmental Engineering 124 (4), 303315.
Dyson, G.J., Randolph, M.F., 2001. Monotonic lateral loading of piles in
calcareous sand. Journal of Geotechnical and Geoenvironmental
Engineering 127 (4), 346352.
Fung, Y.C., Tong, P., 2001. Classical and Computational Solid
Mechanics. World Scientic Printers, Singapore.
Gutierrez, M., Ishihara, K., Towhata, I., 1991. Flow theory for sand
during rotation of principal stress direction. Soils and Foundations 31
(4), 121132.
Hsiung, Y.M., 2003. Theoretical elasticplastic solution for laterally
loaded piles. Journal of Geotechnical and Geoenvironmental
Engineering 129 (5), 475480.
Ismael, N.F., 1990. Behavior of laterally loaded bored piles in cemented
sands. Journal of Geotechnical Engineering 116 (11), 16781699.
Kim, B.T., Kim, N.K., Lee, W.J., 2004. Experimental loadtransfer
curves of laterally loaded piles in Nak-Dong river sand. Journal of
Geotechnical and Geoenvironmental Engineering 130 (4), 416425.
Kondner, R.L., 1963. Hyperbolic stressstrain response: cohesive soils.
Journal of the Soil Mechanics and Foundation Engineering 89 (1),
115144.
Lee, J., Kyung, D., Hong, J., Kim, D., 2011. Experimental investigation
of laterally loaded piles in sand under multilayered conditions. Soils
and Foundations 51 (5), 915927.
Matlock, H., 1970. Correlations for design of laterally loaded piles in soft
clay. In: Proceedings of 2nd Annual Offshore Technology Conference.
Houston, pp. 577594.

Mayoral, J.M., Pestana, J.M., Seed, R.B., 2005. Determination of multidirectional py curves for soft clays. Geotechnical Testing Journal 28
(3), 111.
McVay, M., Zhang, L.M., Molnit, T., Lai, P., 1998. Centrifuge testing of
large laterally loaded pile groups in sands. Journal of Geotechnical
and Geoenvironmental Engineering 124 (10), 10161026.
Reese, L.C., Cox, W.R., Koop, F.D., 1974. Analysis of laterally loaded
piles in sand. In: Proceedings of 6th Annual Offshore Technology
Conference, vol. II, Paper no. 2080. Houston, pp. 473484.
Reese, L.C., Welch, R.C., 1975. Lateral loading of deep foundation in
stiff clay. Journal of the Geotechnical Engineering Division 101 (7),
633649.
Reese, L.C., Van Impe, W.F., 2001. Single Piles and Pile Group Under
Lateral Loading. A.A. Balkema, Rotterdam.
Su, D., 2005. Centrifuge Investigation on Responses of Sand Deposit and
SandPile System Under Multi-directional Earthquake Loading.
Ph.D. Dissertation. The Hong Kong University of Science and
Technology, Hong Kong.
Su, D., 2011. Experimental study on a single pile under horizontal elliptical/
circular loading. Chinese Journal of Geotechnical Engineering 33 (5),
738745 (in Chinese).
Su, D., 2012a. Resistance of short, stiff piles to multidirectional lateral
loadings. Geotechnical Testing Journal 35 (2), 313329.
Su, D., 2012b. CAMLPCode for Analyzing Multidirectionally Loaded
Piles. Shenzhen University, Shenzhen, China.
Su, D., Li, X.S., 2008. Impact of multidirectional shaking on liquefaction
potential of level sand deposits. Geotechnique 58 (4), 259268.
Tamura, S., Adachi, K., Sakamoto, T., Hida, T., Hayashi, Y., 2012.
Effects of existing piles on lateral resistance of new piles. Soils and
Foundations 52 (3), 381392.
Terzaghi, K., 1955. Evaluation of coefcients of subgrade reaction.
Geotechnique 5 (5), 297326.
Wesselink, B.D., Murff, J.D., Randolph, M.F., Nunez, I.L., Hyden,
A.M., 1988. Analysis of Centrifuge Model Test Data From Laterally
Loaded Piles in Calcareous Sand Engineering for Calcareous Sediments,
vol. 1. Balkema, Rotterdam, The Netherlands, pp. 261270.
Zhang, L.M., McVay, M.C., Lai, P., 1999. Numerical analysis of laterally
loaded 3  3 to 7  3 pile groups in sands. Journal of Geotechnical and
Geoenvironmental Engineering 125 (11), 936946.