Computers and Geotechnics 33 (2006) 355–370

www.elsevier.com/locate/compgeo

Nonlinear three-dimensional analysis of pile group supported

columns considering pile cap flexibility
Jinoh Won, Sang-Yong Ahn, Sangseom Jeong *, Jinhyung Lee, Seo-Yong Jang
School of Civil and Environmental Engineering, Yonsei University, Seoul 120-749, Republic of Korea

Received 17 May 2005; received in revised form 26 June 2006; accepted 5 July 2006
Available online 20 September 2006

Abstract

A numerical method that takes into account the coupling between the rigidities of the piles, the cap, and the column has been devel-
oped for analyzing the response of pile group supported columns. Special attention is given to consideration of pile cap flexibility. A load
transfer approach using t–z/q–z and p–y curves is used for the analysis of single piles. The finite element technique is used to combine the
pile stiffness with the stiffness of the cap and column. The numerical method developed has been verified by comparing the results with
other numerical methods for pile groups. Through comparative studies, it has been found that the maximum load on the individual piles
in a group is highly influenced by pile cap flexibility. The prediction of the lateral loads and bending moments in the pile cap is much
more conservative in the present analysis than in FBPier 3.0 and shows a definitely larger lateral load and bending moment for various
cap thicknesses.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Numerical method; Coupling; Pile group supported columns; Pile cap flexibility

1. Introduction to uniformly spaced pile groups no larger than approxi-

Piles are often used in groups for the support of bridge
structures. There are numerous analytical and numerical
methods for designing pile groups. These methods can gen-
erally be classified into three different types: (1) the equiv-
alent single pile method [1–5], (2) the elasticity method
[6–9], and (3) the general three-dimensional load transfer
method [10–14].
The first type of method is based on the analysis of a sin-
gle pile combined with group efficiency factors or group
amplification factors representing both the pile–soil–pile
interaction and the pile–cap interaction. The simplicity of
this method, which uses only a spreadsheet to calculate
the design parameters, is its main advantage. However, this
method has some restrictions in that it can be applied only
*
Corresponding author. Tel.: +82 2 2123 2807; fax: +82 2 364 5300.
E-mail addresses: wjo20@yonsei.ac.kr (J. Won), koen@dweng.co.kr
(S.-Y. Ahn), soj9081@yonsei.ac.kr (S. Jeong), icdij@chollian.net (J. Lee),
syjang@moiza.chonnam.ac.kr (S.-Y. Jang).
mately 5 · 5 arrangement and does not provide the load
distributions along the pile depths in a group [5]. The sec-
ond type of method is a continuum method that models the
soil around the piles as a three-dimensional linearly elastic
continuum. The interaction factor approach [6] using
Mindlin’s solutions, and direct continuum models [8] are
typically included in this category. It should be noted that
this method was originally an elastic method, although
numerous variations have been developed that permit the
modeling of soil layering, pile–soil slip, three-dimensional
geometry, and so on. The third type of method of analysis
of pile groups models all individual piles using beam–col-
umn elements attached to nonlinear soil springs (load
transfer curves) and the pile cap using rigid body [10–12]
or finite element [13,14] methods. This method is popular
because the modeling of randomly arranged piles is rela-
tively simple and the design parameters can be easily
obtained; as a result of many load tests performed during
the past two decades, many reliable load transfer curves
(t–z/q–z and p–y curves) have been suggested. The

0266-352X/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2006.07.007
356 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370
pile–soil–pile interaction that occurs in closely spaced piles
can be taken into account by introducing reduction factors
for the soil reactions in the p–y curves for single piles.
The general three-dimensional load transfer method was
used in the study described in this paper. To date, com-
pared with the pile–soil–pile interaction, the pile–cap inter-
action has not received the attention it deserves, since many
complex geometric and loading conditions are involved in
the structural and geotechnical analyses. In recent years,
however, McVay et al. [15] have reported the significance
of the fixity of the pile head, which is one of the factors
influencing the pile–cap interaction. According to the US
Army Manual [16], knowing the correct relationship
between the stiffnesses of the pile and cap is extremely
important for accurately designing pile groups for use in
flexible base structures. Two kinds of methods are available
for calculating the pile–cap interaction: the stiffness method
and the finite element method.
The stiffness method was suggested by Hrennikoff [17],
and has been extended into a general three-dimensional
method by Saul [18] and Reese et al. [10]. Group 6.0 [19],
a commercial package which has been used for practical
design, is based on this stiffness method. This method can
consider the pile–cap interaction, the nonlinear behavior
of individual piles, and the pile–soil–pile interaction, but
has limitations in that it cannot consider the coupling of
the rigidities of the pile cap to each other and to the col-
umn, since this method assumes a pile cap to be a rigid
body. To consider these coupled pile cap rigidities, a pile
cap needs to be modeled by finite elements such as beams,
frames, plates, and flat-shell elements.
In this study, a new numerical method that considers
coupled cap rigidities has been developed to overcome
the restrictions associated with the conventional stiffness
method. The conceptual methodology of the present
V V
M H
H
Pier
Pile cap
y (Beam-Column
Soil Pile
Fig. 1. Modeling of a pile group.
method is completely different from not only that of the
general structural models (FBPier 3.0 [13] and SAP2000)
but also that of the stiffness method.
2. New numerical method for pile group supported columns
2.1. Structural system
Fig. 1 illustrates a schematic diagram of a typical pile
group supported column (a piled pier) and the numerical
model of it used in the present study. The pile group con-
sists of four piles and a flexible pile cap, which is not in
direct contact with the soil. In the numerical method, the
individual piles are modeled using beam–column elements.
The soil around the individual piles is represented by a set
of load transfer curves, and the interaction between piles is
represented by a p-multiplier. For the modeling of the flex-
ible pile cap and the column, three-dimensional finite ele-
ments, such as four-node flat shell elements for the pile
cap and three-dimensional beam elements for the column,
were used.
In this study, unlike the case of a general three-
dimensional structural analysis (FBPier 3.0, SAP2000,
etc.), the individual piles were analyzed one by one to keep
a total of 10 load–displacement relationships (axial direc-
tion = 1; lateral direction = 8; torsional direction = 1) for
each pile head in the first instance. The stiffness matrices
½K pE  of the individual piles which were incorporated into
the structural analysis were derived from the load–displace-
ment curves of the individual piles.
2.2. Modeling of piles
Fig. 2 shows the coordinate system (X, Y, Z) of the gen-
eral structure and the local coordinate system (u, v, w) of a
M
Pier (Beam Element)
x
Pile Cap
(Flat-Shell Element)
Pile
Element)
z
p-y curve
t-z curve
q-z curve
 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370 357

X where ½K pE i is the stiffness matrix of the pile head, {d}i is the

displacement vector, and {F}i is the vector of the force at
the ith pile head. The pile head stiffness matrix ½K pE i is of
v
order 6 · 6, representing three displacement constants,
Mv αv three rotational constraints, and four couplings between
1 Fv the displacement and rotational constraints.
3
In this study, a nonsymmetric stiffness matrix ½K pE i , i.e.
Fu one where k26 6¼ k62 and k35 6¼ k53, was used to consider
Fw
the nonlinear problem, because in nonlinear analysis the
Mw
Y Mu entries of ½K pE i vary at each iteration and the coupling con-
αw stants are not required to be equal, even though in linear
w analysis the coupling constants are identical, by Maxwell’s
Z 2 reciprocal theorem. Each entry of ½K pE i was obtained from
the four principal modes of pile head movement (Fig. 3);
αu
u this procedure was suggested by Reese et al. [10]. In the
present method, the p-direction in Fig. 3 agrees with the
Fig. 2. Pile coordinates.
v-direction in the u–v plane and corresponds to the w-direc-
tion in the u–w plane. The Fp–dp and Mq–dp curves (Mode
pile. An arbitrary pile is connected to the pile cap at point
I) were obtained by repeated single-pile analysis, based on
1. Point 2 indicates the pile toe, and point 3 is a reference
the load transfer method using p–y curves with increasing
point, used for transformation between local and global
dp, with the boundary condition aq = 0. The Fp–aq and
coordinates. The equation for equilibrium at the pile head
Mq–aq curves (Mode II) were obtained by a similar lateral
(point 1) in the local coordinate system is as follows:
8 9 analysis with increasing aq, with the boundary condition
> Fu > dp = 0. The relationships (Modes I and II) between lateral
>
> >
>
>
> Fv > > load (moment) and displacement (rotation) in u–v plane are
>
> >
>
>
< >
Fw = not always identical to those in u–w plane. Thus, c1, c2, c3,
fF gi ¼ and c4 should be computed in both u–v plane and u–w
>
> Mu > >
>
> >
> plane. In the case of Mode III, for increasing settlement
>
> M >
>
>
> v >
> of the pile head, represented by du, the Fu–du curve was
: ;
Mw i obtained by continuous load transfer analysis using t–z/
2 38 9 q–z curves. Mode IV, representing the torsional behavior,
k 11 0 0 0 0 0 > du > ð1Þ
>
> >
> was not considered in this study.
6 0 k 7 >
k 26 7 > >
6 22 0 0 0 > dv >
>
6 7>> > > Here, all of the load–deformation functions which were
6 0 0 k 33 0 k 35 0 < dw =7 used for estimating ½K pE i generally showed nonlinear char-
¼66 7
6 0 0 0 k 44 0 0 7 7> >
> au >
> >
acteristics (Fig. 3). In the present method, a mixed incre-
6 7>> >
4 0 0 k 53 0 k 55 0 5> >
> av >
>
> ment–iteration method, with iteration within each load
: > ; increment, was used to adjust the deflections and reactions
0 k 62 0 0 0 k 66 i aw i
to ensure compatibility and was compared with other non-
¼ ½K pE i fdgi ; linear techniques such as purely incremental solutions [12]

Fig. 3. Pile head movement [10].

358 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370

Axial Load at the pier top (kN) element, having six degrees of freedom per node, permits
0 500 1000 1500 2000 2500 3000 3500 an easy connection to other elements such as beams and
0 folded
 s  elements. The stiffness matrix of a flat-shell element
K E , which is of order 24 · 24 per element, is con-
structed by combination of the stiffness matrices of a plate
Settlement at the pier bottom (mm)

Axial Load, V
element ðK eplate Þ and of a membrane element ðK emembrane Þ in a
A local coordinate system as follows:
10  e 
 s K plate 0
KE ¼ : ð2Þ
0 K emembrane
Purely increment method
(N=increment number)
N=1 Here, the stiffness matrix of a plate element K eplate is ex-
N=2
N=4 pressed in the following matrix form:
20 N=8
N=16
Z Z
e
Mixed increment & iteration method K plate ¼ Bb Db Bb dV þ BeT
eT e e e e
s Ds Bs dV ; ð3Þ
N=1 (secant method) V V
N=2
N=4
N=8
where Beb
is the bending-strain matrix and Bes is the shear-
N=16 strain matrix. For an isotropic material, Deb and Des are as
30
follows:
Fig. 4. Comparison of pile group behavior in nonlinear analysis methods. 2 3
3
1 m 0
Et 6 7
Deb ¼ 4m 1 0 5; ð4aÞ
and the secant modulus method (an iteration method) [10]. 12ð1  m2 Þ
0 0 ð1  mÞ=2
Fig. 4, which represents load–settlement curves at the bot-  
tom of the column (point A), shows that for a purely incre- e WEt 1 0 5
Ds ¼ ; W¼ ; ð4bÞ
mental method the curves are significantly dependent on 2ð1 þ mÞ 0 1 6
the increment number N, whereas for a mixed increment–
where E is the Young’s modulus, m is the Poisson’s ratio, t
iteration method identical load–settlement curves are esti-
is the constant thickness of the plate, and W is the shear
mated regardless of N. On the basis of this result, only
correction factor. Next, the stiffness matrix of a membrane
one increment (N = 1) with iteration, which is identical to
element K emembrane is as follows:
the secant method, was used for further analysis.
c
K emembrane ¼ K emembrane þ e he heT ; ð5Þ
2.3. Modeling of a flexible pile cap Z X
K emembrane ¼ ½Bem Ge Re T  C  ½Bem Ge Re  dX; ð6Þ
The stiffness method [17] can consider neither coupling Z Xe
between the pile group and the column nor flexibility of e
h ¼
T
hbe ; ge ; re i dX; ð7Þ
the pile cap because the pile cap is assumed to be a rigid Xe
body in the stiffness method. Much work has been done where C is the constitutive modulus and c is taken as the
to analyze this coupling and the effect of pile cap flexibility. shear modulus. Bem , Ge , and Re are the strain matrices rep-
Typically, a plate element has been used as a pile cap, in resenting the relationship between the displacements (the
several numerical methods [14,20,21]. These methods, how-
ever, have some limitations in that the horizontal behavior
of the pile cap cannot be considered, because horizontal
degrees of freedom (in the x- and y-directions) are
excluded. These limitations can be overcome by using a
flat-shell element [22]. In the present study, a four node
flat-shell element was adopted; this was developed by com-
bining a Mindlin plate element and a membrane element
with torsional degrees of freedom, as shown in Fig. 5. This

Fig. 5. Flat shell element. Fig. 6. Estimation method of stiffness at ith load increment.
 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370 359

membrane displacement, the rotation, and the midside 2.4. Computational procedure
incompatible displacement, respectively) and the strains;
be, ge, and re are the strain matrices for the infinitesimal For given conditions, such as the geometry, the load,
rotation fields. and the properties of the structures (piles, cap, and

Start

Input geometry, boundary conditions, loads, increment number

(N), properties of pile, cap, pier, and soil

Beam-column analyses
Perform Mode III analyses in each pile
in axial loads
Estimate load-displacement curves in each pile
Fu~δ
~ u

Perform Mode I, Mode II analyses of Beam-column analyses

each v-u, w-u plane in each pile in lateral loads

Estimate load-displacement curves of each v-u, w-u in each pile

( w, α v)~(Fw,Mv), (δ
(δ ( v, α w)~(Fv,Mw)

Construct stiffness matrix of flat shell and beam elements

Flat shell: [K]s, Beam: [K]b

Initialize load increment number

i=1

Back up displacements
[ D]p
[Δ

Initialize iteration number

j=1

No j>1 Yes

Estimate initial tangential slope at 10 Cubic Spline Estimate secant slope at 10

load-displacement curves in each pile analyses load-displacement curves in each pile

Estimate pile head stiffness

(c1~c4)v, (c1~c4)w, c5, c6

Construct 12 × 12 pile stiffness matrix

[K]p

Assemble stiffness matrices of flat shell, beam and piles

[K] = [K]s+[K]b+[K]p

Adjust stiffness matrix and load vector by boundary conditions

[K]’, [Δ
[ F]’

Calculate displacement increment vector

[ D]=[k]’-1 [Δ
[Δ [ F]’

No j=j+1
Δ uassumed-Δ
- ucomputed < ε
Δ uassumed ←Δuucomputed

ui = ui-1 + Δ u

No
i=i+1 i=N?

Output displacement ui and

Output strain and stress of flat shell and beam

Output displacement, bending moment, and shear Beam-column analyses

along depth of each pile in axial and lateral direction

End

Fig. 7. Flowchart of the present method (YSGroup).

360 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370

column) and of the soil layers, the deformations and the 3.1. Linear analyses
forces on members (the stress in the cap, and the bend-
ing moments and shear forces in the piles and column) A schematic diagram of a 2 · 2 pile group structure is
can be calculated by the proposed method (YSGroup). shown in Fig. 8. This structure consists of a pier, a pile
The properties of the structures here include the pile cap, and four identical vertical piles, which are spaced by
cap flexibility. Before constructing the stiffness matrices 3 m (=6D, where D is the pile diameter). A p-multiplier
of the flat-shell elements (for the cap) and of the beam of unity was used because the interaction between piles in
elements (for the column), our method calculates 10 this configuration (s/D = 6) seems not to be significant
load–displacement curves for each pile head by repeated and because this study focuses on the pile–cap interaction
load transfer analyses using t–z/q–z and p–y curves and rather than on interactions between piles. The four piles
saves them. In this step, reduction factors, typically p- have an embedded length of 10 m, a diameter of 0.5 m,
multipliers for the lateral response, are incorporated into and a flexural rigidity (EI) of 147,264 kN m2. The thickness
the load transfer curves. In this study, the p-multipliers of the pile cap is 0.75 m, and the pile head conditions are
which have been experimentally derived from full-scale fixed. The pier is 10 m in length and 1 m in diameter, and
load tests or from tests in centrifuge are used. These ini-
tially calculated load–displacement curves are not chan-
ged during the iterative procedure. For a nonlinear
analysis, all external forces are first divided by the incre-
ment number (N), and then the current stiffness matrices
of the individual piles ½K pE  are calculated from the stored
load–deflection curves using the tangential slope in the
first iteration (j = 1) and the secant modulus for j > 1
(Fig. 6). Here, the tangential slope (df(u)/du) and the
load (f(u)) are estimated using a cubic-spline method.
p
Combining  s½K E  with the stiffness matrix of the flat-shell
element K E and beam element ½K bE , the coupled stiff-
ness matrix of the pile group supported column [KE] is
calculated as follows:
 
½K E  ¼ K sE þ ½K bE  þ ½K pE : ð8Þ
From the global stiffness matrix [KG] and the global load
increment vector [DFG] obtained by combining [KE] and
[DFE], the total displacement increment vector [DDG] is
given by
1
½DDG  ¼ ½K G  ½DF G  ð9Þ
The procedure described above is iterated until the error
between the assumed and calculated displacements falls
within a tolerance limit. The displacement increments at
the target step are obtained by adding the calculated re-
sults to those from the previous step. The internal forces
in the pile cap and columns and in the individual piles are Fig. 8. Schematic diagram of a piled pier.
then calculated. Fig. 7 shows a flowchart of the present
analysis. Table 1
Material properties for pile groups
3. Validation by comparison with other numerical studies Elements Properties Values
Pier Elastic modulus (E) (MPa) 40,000
A series of linear pile group analyses were performed to Moment of inertia (Ix, Iy) (m4) 0.04909
verify the structural-analysis routine used in the present Area (A) (m2) 0.7864
method by comparison with other numerical methods Polar moment of inertia (J) (m4) 0.09817
which have been used in the design of pile groups. The Shear modulus (G) (MPa) 17,391.3
results of a nonlinear analysis of a typical pile group sup- Pile cap Elastic modulus (E) (MPa) 40,000
ported column (a piled pier) were then compared with Poisson’s ratio (m) 0.18
results obtained from Group 6.0 and FBPier 3.0. This sec- Thickness (t) (m) 0.75
tion focuses on the verification of the present method. In Pile Elastic modulus (E) (MPa) 48,000
addition, the load–deformation relationship of a piled pier Moment of inertia (Ix, Iy) (m4) 0.003068
Area (A) (m2) 0.19635
was investigated.
 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370 361

responds to an SPT-N value of 40, and the tension part was

Fig. 9. Soil conditions for linear analysis of a single pile.
has a flexural rigidity of 1,963,600 kN m2. Table 1 shows
the material properties used in this study.
The results of linear analysis by the present method have
been compared with results from three different numerical
methods to verify the present structural analysis routine for
a piled pier. Fig. 9 shows the soil conditions used for the
linear analyses. The same axial soil spring constants were
used along the pile depth, with a constant value of
2000 kN/m2, which includes the pile perimeter, corre-
sponding to loose sand with an SPT-N value of 9. The
end-bearing spring constant was 10,000 kN/m2, which cor-
neglected. The constants of the horizontal soil springs were
increased from 0 to 100,000 kN/m2 along the pile depth.
The pile group was subjected only to a lateral load, of
1000 kN at the pier top. The pile head stiffness (c1  c5) cal-
culated by the load transfer method using the present
method, shown in Fig. 3, is described in Table 2 (case 1),
where the results are identical to those obtained from
FBPier 3.0.
Table 3 shows the estimated displacements and distrib-
uted forces when a lateral load of 1000 kN in the x-direc-
tion is applied to the pier top. The displacement method
and Group 6.0 cannot model pier elements; therefore an
equivalent moment was applied to the pile cap, equal to
the lateral load multiplied by the length of the pier.
As shown in Table 3, the axial and lateral displacements
at the pile head estimated by the present method, by Group
6.0, and by the displacement method are approximately the
same, while the values obtained from FBPier 3.0 are
greater by about 22%. This suggests that FBPier 3.0 gives
more flexible results for pile groups in the case of lateral
loading, because of the difference in the modeling methods,
whereby the former methods model piles using stiffness
matrices and the latter method does so using three-
dimensional elements. The axial forces and moments at

Table 2
Pile head stiffness of a single pile
c1 (kN/m) c2 (kN m/m) c3(kN/rad) c4 (kN m/rad) c5 (kN/m)
Case 1 31,435 49,971 49,971 128,492 19,610
Case 2 84,105 287,989 287,989 1,850,861 557,403

Table 3
Comparison of numerical methods on displacements and forces in pile groups
Check point Displacement method Group 6.0 FBPier 3.0 Present method
Displacements Lateral displ. at pier top (mm) – – 553.4 510.9
Axial displ. at pier top (mm) – – 9.2 8.5
Lateral displ. of 1, 3 pile head (mm) 57.4 53.4 68.0 52.9
Axial displ. of 1, 3 pile head (mm) 46.6 51.2 56.1 51.4
Rotation angle of 1, 3 pile head () 0.031 0.028 0.030 0.028
Forces Lateral force at 1, 3 pile head (kN) 250 250 250 250
Axial force at 1, 3 pile head (kN) 915 1000 1080 1008
Moment at 1, 3 pile head (kN m) 1128 968 830 993

Table 4
Properties used for estimating load transfer curves
N value Contents Loose sand Medium sand Dense sand
7 20 40
t–z, q–z curves [25] Ultimate skin friction, sf (kPa) 14 40 80
Initial shear modulus, Gi (kPa) 6000 8000 10,000
Poisson’s ratio, m 0.35 0.32 0.30
Ultimate bearing capacity, Qf (kN) 549.8 1570.8 3141.6
p–y curves (API, O’Neill) Internal friction angle () 26.6 31.8 38.0
Unit weights (kN/m3) 14.8 16.6 20.0
p–y modulus, k (kN/m3) 6800 24,400 61,000
362 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370

the pile head predicted by each method show some discrep- [24], respectively. It was found that the predictions were
ancies, whereas the same lateral forces are distributed to almost identical.
each pile head in all the numerical methods. The present After these single pile analyses, a series of group pile
method predicts smaller axial forces and larger moments analyses were performed on the pile group configuration
at the pile head than does FBPier 3.0 and estimates results shown in Fig. 8. The individual piles were assumed to be
similar to those of Group 6.0. For all the linear analyses, it PHC driven piles installed in a sandy soil. The soil around
is found that the present method gives results similar to the individual piles was modeled with nonlinear load trans-
those obtained from other numerical methods, especially fer curves. The types of soil considered were three sandy
Group 6.0. soils, which were a loose, a medium, and a dense sand
according to density, and had SPT-N values of 7, 20, and
3.2. Nonlinear analyses 40, respectively. The axial load transfer curves (t–z and q–
z curves) were estimated using an equation developed by
The behavior of axially and laterally loaded single piles McVay et al. [25] suitable for driven piles; the lateral load
was predicted by the present method (YSGroup) by consid- transfer curves (p–y curves) were used as an API model
ering nonlinear load transfer curves and then comparing [26]. The properties used for estimating the axial and lateral
them with the results of TZPile [23] and LPile 4.0 Plus load transfer curves are shown in Table 4. The ultimate skin

0 0
Present method
Lateral displacement at the pile head, A (mm)
Lateral displacement at the pier top (mm)

Lateral Load, H FBPier 3.0

Pier top
Group 6.0
50 5

100 10

Dense sand
150 Loose sand Dense sand 15
Lateral Load, H

Medium sand A
200 20
Present method Loose sand
FBPier 3.0 Medium sand

250 25
0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
a Lateral Load at the pier top (kN) b Lateral Load at the pier top (kN)

0 0
Present method
FBPier 3.0
Settlement at the pile head, A (mm)

Settlement at the pile head, B (mm)

2 Group 6.0

-10 Dense sand Loose sand

4

Medium sand
-20 Medium sand 6

Loose sand Lateral Load, H

8
Lateral Load, H
-30 B
A
Present method 10
FBPier 3.0 Dense sand
Group 6.0
-40 12
0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
c Lateral Load at the pier top (kN) d Lateral Load at the pier top (kN)

Fig. 10. The displacement of a piled pier subjected to lateral loads (fixed head condition). (a) Lateral displacement at pier top, (b) lateral displacement at
pile head A, (c) tensile displacement at pile head A and (d) settlement at pile head B.
 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370 363

friction (sf) and ultimate end-bearing capacity (qp) were esti- 0

mated from Meyerhof’s equations, to obtain sf = 2N (kPa)

Lateral displacement at the pier top (mm)

and qp = 400N (kPa). The end bearing shear modulus Gi
was applied together with the values of the modulus for
loose, medium, and dense sand used in the analysis of Lateral Load, H Pier top
200
Zhang and Small [21]. The internal friction angles for the
p–y curves were estimated from the meanpvalueffiffiffiffiffiffiffiffiffi of the pre-
dictions of Dunham’s equation, / ¼ 12N þ 15 , and
Peck’s equation, / = 0.3N + 27, using SPT-N values.
Dense sand
The values of the p–y modulus k used were the proposed 400 Fixed head
values in the LPile 4.0 Plus manual. Present method
Fig. 10 shows the results of a nonlinear analysis of a FBPier 3.0

piled pier, subjected to only a lateral load at the pier top. Pinned head
Present method
Very large lateral displacements are estimated at the pier FBPier 3.0
top in Fig. 10(a), which can be predicted only by a coupled 600
analysis method such as FBPier 3.0 and the present 0 100 200 300 400 500 600 700
method. Note that the lateral displacement at the pier a Lateral Load at the pier top (kN)
top obtained by the present method is almost the same as
that obtained from FBPier 3.0. Fig. 10(b)–(d) represents 0

Lateral displacement at the pile head, A (mm)

the lateral displacement, the settlement of the pile head
A, and the settlement of the pile head B, respectively.
The lateral displacements of the pile heads obtained by 4
Lateral Load, H
the present method are a little smaller than the values
obtained from FBPier 3.0, but are quite similar to those A
obtained from Group 6.0. The reason for the different lat- 8
eral displacements in these cases of lateral loading is
thought to be the difference in the modeling method for Dense sand
12
individual piles, as previously mentioned. Sudden changes Fixed head
Present method
in both the tensile displacement in pile A and the settlement FBPier 3.0
in pile B occurred at a lateral load of 100 kN in loose sand 16
Group 6.0

and at 300 kN in medium sand, as shown in Fig. 10(c) and Pinned head
Present method
(d). The reason is that the tensile displacement increases FBPier 3.0
Group 6.0
suddenly as the skin friction reaches its ultimate level. 20
The results of an analysis of a piled pier under pinned 0 100 200 300 400 500 600 700
conditions are shown in Fig. 11 together with the results b Lateral Load at the pier top (kN)
for fixed head conditions in dense sand. At the pier top,
the lateral displacement obtained under pinned condi- 0
tions is somewhat larger than that for fixed head condi-
Pullout displacement at the pile head, A (mm)

tion. At a lateral load of 550 kN, yielding occurs and Lateral Load, H
-20
then the displacement increases suddenly. This is caused
by a sudden increase in the tensile displacement of pile A
A, as shown in Fig. 11(c). The lateral displacement of -40
pile head A is estimated to be smaller under pinned con-
ditions than under fixed head conditions. From these -60
piled pier analyses, it was concluded that the overall Dense sand

Fixed head
behavior of a piled pier is affected significantly by the Present method
pile head condition. -80 FBPier 3.0
Group 6.0
Fig. 12 shows the profiles of the bending moment and the
shear force on a piled pier in dense sand, subjected to com- -100 Pinned head
Present method
bined external loads (V = 3000 kN and H = 500 kN) at the FBPier 3.0
Group 6.0
pier top. The bending moments and shear forces calculated -120
by the present method and by FBPier 3.0 show differences
0 100 200 300 400 500 600 700
between the leading row and the trailing row near the pile
head. The differences in the bending moment and shear c Lateral Load at the pier top (kN)

force near the pile head compared with the results from Fig. 11. The effect of pile head conditions. (a) Lateral displacement at
Group 6.0 are caused by the effect of the local bending stress pier top, (b) lateral displacement of pile A and (c) tensile displacement of
and membrane horizontal stress in the pile cap. pile A.
364 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370

4
Depth (m)

6 Combined Load (V=3000kN, H=500kN)

Dense sand
Present method (Trail row)
Present method (Lead row)
8 FBPier 3.0 (Trail row)
FBPier 3.0 (Lead row)
Group 6.0 (Trail row)
Group 6.0 (Lead row)

10
-100 0 100 200 300 400
a Bending moment (kN-m)

Fig. 13. Schematic diagram of a piled pier.

2
4.1. Problem description

4 A total of four piles are arranged in a 2 · 2 pattern and

Depth (m)

the center-to-center spacing of the piles is three times the

pile diameter D (=1.5 m). One pier is located at the center
6
of the pile cap. External forces are applied at the top of the
pier, namely a vertical load V0 = 17,585 kN and a lateral
load H0 = 938 kN. No torsional load is considered. The
8
thickness of the pile cap is 2 m and the pile cap is located
above the ground surface, so that the soil reaction beneath
the cap can be ignored. The connection between the pile
10
and the cap is a fixed condition. The SPT-N value of the
-200 -100 0 100 200
homogeneous soil is 6, and the modulus of the lateral
b Shear force (kN)

Fig. 12. Comparison of individual pile behavior on bending moment and

shear force. 2

present method
4. The effect of pile cap flexibility displacement method
4

For the design of flexible base structures in which the

relative difference between the stiffnesses of the pile cap
Settlement (mm)

6
and the piles is small, and thus bending of the cap is
expected, use of the correct stiffness relationship between
the cap and the piles is extremely important for accurately
8
designing a pile group [16]. For this reason, consideration
of the flexibility of the pile cap is needed in the design of
structures that include large diameter piles.
10
In this study, a pile group supported column (a piled
pier) with large diameter piles (D = 1.5 m), as shown in
Fig. 13, was analyzed by the present method. To investigate 12
the effect of pile cap flexibility, the results predicted by the
0 2.25 4.5 6.75 9
present method have been compared with results obtained
x-coordinate of a cap (m)
by the displacement method, which assumes that a pile cap
is a rigid body. Fig. 14. Displacement of a pile cap.
 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370 365

subgrade reaction kh is 0.974 kg/cm3, obtained by using an
empirical equation suggested by the Korea Specification
for Bridges [27]. The p-multiplier is set to one to allow com-
parison with the results of the displacement method and to
investigate the pile–cap interaction rather than the interac-
tion between piles, although in this configuration (s/D = 3)
a reduction of the soil reaction is expected. The pile head
stiffness (c1  c5) calculated by the load transfer method
using the present method is presented in Table 2 (case 2).
In the displacement method, the calculated pile head stiff-
ness was used.
4.2. Pile cap displacement
Fig. 14 shows the distribution of the settlement of the
pile cap. A larger settlement was estimated on the right-
hand side of the pile cap owing to the effect of the lateral
load. The settlements at both edges were similar for the
two methods, but a somewhat larger value was calculated
by the present method at the center of the cap. Fig. 15
shows the three-dimensional deformed shape of the pile
cap, calculated by the present method. It is drawn with a
200 times enlarged scale of the displacement for better vis-
ibility of the deformed shape. The center point, at which
the pier is located, has sunk, not only in the x–z plane
but also in the y–z plane. The displacement added by the
deformation of the pile cap is smaller than that caused by
the movement of the pile cap. However, it is important
to consider the effects of the flexibility of the pile cap for
correct estimation of the individual forces on the pile
heads.
4.3. Stress in the pile cap
In general, the magnitude and the distribution of stres-
ses are important parameters for the design of a pile cap.
The stiffness method (displacement method, Group 6.0,
etc.), in which the pile cap is assumed to be a rigid body,
cannot estimate the stress in a pile cap. However, in the
present method, using a flat-shell element, the stress in a
pile cap can be calculated at the integration points.
Fig. 16 shows the horizontal membrane stress and the
plate-bending stress in the pile cap along the x-direction.
This shows that a horizontal tensile stress occurs in the
pile cap elements to the left of the pier, and a horizontal
compressive stress occurs to the right. From the result
for the plate-bending stress in Fig. 16(b), it is found that
the cap shows a slightly convex curvature on the outside

Fig. 15. Deformed shape of a pile cap.

a b
Fig. 16. Stress distribution of a pile cap. (a) Membrane horizontal stress and (b) plate bending stress.
366 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370

a b
Fig. 17. Distributed forces and bending moments at the pile head. (a) Present method and (b) displacement method.

of the piles and a concave curvature between the piles. In
the pile cap elements on the right side of the pier, a sig-
nificant bending stress is estimated, which is caused by
the combined effect of the external axial and lateral
loads.
4.4. Pile head reactions
The axial and lateral loads and bending moments in the
individual pile heads were estimated and compared with
those obtained from the displacement method. The results
are shown in Fig. 17. From an equilibrium check, it was
found that the present method satisfied the equilibrium
condition, although the distributed loads were somewhat
different from those obtained from the displacement
method. Somewhat larger forces were distributed to the
leading row (piles 2 and 4) and somewhat smaller axial
forces to the trailing row (piles 1 and 3) in the present
method, as compared with the results obtained from the
displacement method. It is likely that the compressive hor-
izontal stress in the pile cap on the right side of the pier
increases the lateral forces in the leading row and the ten-
sile horizontal stress decreases the lateral forces in the trail-
ing row.
The bending moments at the pile heads calculated by the
present method were 152 kN m in the trailing row and
264 kN m in the leading row. The values obtained by the
displacement method were 23 kN m regardless of the pile
location. Significantly larger bending moments are gener-
ated at each pile head in the present method compared with
the results obtained by the displacement method. There-
fore, it is concluded that the maximum load on the individ-
ual piles in a group is highly influenced by the flexibility of
the pile cap.
Note that in the present method, the bending moment at
the pile head in the leading row is opposite to that in the
trailing row. The reason for this is that as a combined axial
and lateral load is acting on the center portion of the pile
cap, the pile heads in the leading row tend to rotate in a
counterclockwise direction and, simultaneously, the trail-
ing row tends to rotate in a clockwise direction, as shown
in Fig. 18.
In the present method, large lateral forces (30 and
34 kN) and bending moments (194 and 222 kN m) are
developed in the y–z plane, perpendicular to the loading
direction, even though no lateral loads are applied in this
plane (Fig. 17). However, the lateral forces and bending
moments in piles 1 and 3 and those in piles 2 4 are identical
in magnitude but opposite in direction, canceling all the
lateral loads and moments in the y–z plane out, and there-
fore have no effect on the overall equilibrium conditions of
the pile cap.
V
H
Moment
Moment Deformed
pile cap
Clockwise Counterclockwise
rotation rotation
Fig. 18. Direction of bending moments at the pile head connected to
flexible pile cap.
 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370 367

Increase at the pile head

0 0

4 4

Depth (m)
8
Depth (m)

12 Increase at the 12
trail row pile

Trail row (1 & 3 piles)

16 Trail row (1 & 3 piles) 16 Lead row (2 & 4 piles)
Lead row (2 & 4 piles) Displacement method
Displacement method

20 20
-800 -400 0 400 800 -100 0 100 200 300
Bending moment (kN m) Shear force (kN)

Fig. 19. Member forces along the pile depth.

500
Present method
FBPier 3.0
Lateral load at the pile head (kN)

400 Group 6.0

Lead row
300

200

Trail row
100

0
0.1 1 10
a E / Eo

1200
Present method
Bending moment at the pile head (kN-m)

800 FBPier 3.0

Group 6.0
Trail row
400

-400
Lead row
-800

-1200
0.1 1 10
b E / Eo

Fig. 20. Comparison of the results of elastic modulus of a pile cap. (a) Lateral load and (b) bending moment.
368 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370

4.5. Forces and displacements in pile members From this comparative study, it has therefore found that
the bending moments in individual piles are changed, as
Fig. 19 shows the bending moments and shear forces well as the pile head forces, when the flexibility of the pile
along the depth of the piles. The bending moments cap is considered. The bending moment and shear force
obtained by the present method show different shapes obtained when pile cap flexibility is considered are larger
along the pile depth between the trailing row (piles 1 and in one row than the results obtained by the displacement
3) and the leading row (piles 2 and 4): there is a negative method. Therefore, the flexibility of the pile cap must be
bending moment at the pile heads in the leading row and considered for the safe design of a piled pier.
a positive bending moment in the trailing row. The dotted
line, showing the results calculated by the displacement 5. Parametric study
method, which are identical for the two rows, follows
almost the mean of the values for the trailing and leading 5.1. Pile head forces
rows obtained by the present method.
The shear forces in the leading row are somewhat larger The lateral loads and the bending moments at the indi-
than those in the trailing row for the present method. The vidual pile heads predicted by the present method have
results obtained by the displacement method are identical been compared with results from FBPier 3.0 and Group
for both rows, and the values along the pile depth corre- 6.0, for various values of the elastic modulus (E/E0) of
spond to the mean of the values obtained by the present the pile cap (Fig. 20). Except for the elastic modulus of
method. the pile cap, all other conditions were identical to those

800
Present met hod
600 FBPier 3.0
Lateral load at the pile head (kN)

Lead row Group 6.0

400

200

Trail row
0

-200

-400
0 0.4 0.8 1.2 1.6 2
a T / To

4000
Bending moment at the pile head (kN-m)

Present method
FBPier 3.0
Group 6.0
2000
Trail row

Lead row
-2000

-4000
0 0.4 0.8 1.2 1.6 2
b T / To

Fig. 21. Comparison of the results of thickness of a pile cap. (a) Lateral load and (b) bending moment.
 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370 369

in Fig. 13. The basic elastic modulus E0 is 2.5 · 107 kN/m2. 0

Lead row Trail row
Fig. 20 shows well the significant effect of the elastic mod- a
ulus on the maximum load on the individual piles. In pres-
ent method, a very large lateral load is distributed to the 0.2
leading row when the elastic modulus of the pile cap is Ratio of Elastic modulus of Pile Cap
E / Eo =0.5
low, but when a very high elastic modulus (E/E0 = 10) is
E / Eo =1
used, identical lateral loads are generated in the two rows, 0.4 E / Eo =2
and these loads are almost equal to the results from the

z / zo
E / Eo =10
stiffness method (Group 6.0). FBPier 3.0 predicts smaller
lateral loads than both the present method and Group 0.6
6.0, since the model of a pile in FBPier 3.0 is different from
the model in the present method and in Group 6.0, as pre-
viously described. 0.8
As shown in Fig. 20(b), the bending moment in the lead-
ing row has a large negative value (264 kN m), and that
in the trailing row has a large positive value (152 kN m), 1
for E/E0 = 1 with the present method. The absolute magni- -800 -400 0 400 800
tudes of these two bending moments are larger than that Bending moment (kN-m)
calculated by the stiffness method (Group 6.0), which was
Trail row Lead row
very small, only 23 kN m. However, it should be noted that
0
these bending moments at the pile head are significantly b
smaller than the maximum bending moments developed
at a depth of 3D (=4.5 m) depth in this piled pier, as shown 0.2
in Fig. 19. The maximum bending moment calculated by
the displacement method is 519 kN m, and the bending
moments in the leading and trailing rows calculated by 0.4
the present method are 464 kN m and 593 kN m, respec-
z / zo

tively. The difference in the bending moments between

Ratio of Elastic modulus of Pile Cap
the trailing row and the leading row is more significant in 0.6 E / Eo =0.5
the present method than in FBPier 3.0. E / Eo =1
Similarly, the present method has been compared with E / Eo =2
E / Eo =10
FBPier 3.0 and Group 6.0, for various thicknesses (T/T0) 0.8
of the pile cap, as shown in Fig. 21. The difference in the
bending moments between the trailing row and the leading
row in the present method became more significant than in 1
FBPier 3.0 when the thickness of the pile cap was varied. -100 0 100 200 300
From this comparative study, it is concluded that the Shear force (kN)
present method predicts that the effect of flexibility of the
Fig. 22. Effects of elastic modulus of a pile cap.
pile cap is more significant than in FBPier 3.0, and can
be regarded as a conservative method.

5.2. Bending moments and shear forces in piles 6. Summary and conclusions

Fig. 22 shows profiles of the bending moment and the
shear force along the pile depth for various values of the
elastic modulus E of the pile cap. Four different elastic
modulus ratios (E/E0 = 0.5, 1, 2, and 10) were investigated.
Except for the elastic modulus of the pile cap, all other con-
ditions were identical to the conditions in Fig. 13.
As shown in Fig. 22, the difference in the bending
moment and shear force between the trailing row and the
leading row decreases with increasing elastic modulus of
the pile cap. This figure shows that there is an effect of pile
cap flexibility on the bending moments and shear forces in
individual piles and therefore represents profiles of bending
moment and shear force different from those obtained by
assuming a rigid pile cap.
The behavior of pile group supported columns (piled
piers) has been investigated by using a numerical study.
The emphasis was on developing a new numerical method
to overcome the restrictions associated with the conven-
tional stiffness method. A numerical method that takes into
account the coupling between the rigidities of the piles, the
cap, and the column has been developed for analyzing the
response of piled piers, taking account of soil nonlinearity,
group effects, and the flexible pile cap.
In this study, the pile cap was modeled with four-node
flat shell elements, the pier with three-dimensional beam
elements, the piles with beam-column elements, and the soil
with nonlinear load transfer curves. A nonlinear analysis
algorithm using a mixed incremental and iterative
370 J. Won et al. / Computers and Geotechnics 33 (2006) 355–370
technique was proposed. The proposed method for piled
piers subjected to both axial and lateral loads was verified
by comparing the results with other numerical methods.
The conceptual methodology of the present method is
completely different from that of general structural models
and the stiffness method. The present method predicts well
the general trend for pile groups, even though piles are
modeled in a simple way by using pile head stiffness matri-
ces. To investigate the effects of the flexibility of the pile cap
on the design parameters of piled piers subjected to lateral
loads, parametric studies have been conducted by the pres-
ent method. From the findings of this study, the following
conclusions have been drawn:
1. The present method, based on a combination of pile
head stiffnesses for individual piles and finite element
models for the cap–pier structure, is intermediate in
complexity and theoretical accuracy between general
three-dimensional structural models (FBPier 3.0,
SAP2000, etc.) and the conventional stiffness method
(Group 6.0). The lateral displacement obtained by the
present method is similar to that obtained by the dis-
placement method and Group 6.0, while it shows smal-
ler values than those obtained by FBPier 3.0 owing to its
different modeling methods; the overall agreement
appears to be satisfactory.
2. The flexibility of the pile cap affects individual pile head
forces significantly and affects the bending moments and
shear forces in individual piles as well, even though the
displacement of the pile cap does not vary much. From
parametric studies, it has been found that the present
method predicts well the changes in the bending
moments and shear forces in individual piles and the
changes in the stresses in the pile cap when the flexibility
of the pile cap is varied.
3. From comparative studies, it has been found that the
maximum load on the individual piles in a group is
highly influenced by the flexibility of the pile cap. The
prediction of the present method is much more conser-
vative than that of FBPier 3.0 and thus represents a def-
initely larger lateral load and bending moment for
various cap thicknesses.
4. The present method, when used in nonlinear analysis, is
useful for predicting the displacement of a superstruc-
ture such as a pier, taking account of soil nonlinearity,
the flexibility of the pile cap, and the pile arrangement,
since the pile cap and the pier are modeled as finite
elements.
References
[1] Bogard D, Matlock H. Procedures for analysis of laterally loaded pile
groups in soft clay. In: Proceedings, specialty conference of geotech-
nical engineering in offshore practice, ASCE, 1983. p. 499–535.
[2] Ooi PSK, Duncan JM. Lateral load analysis of groups of piles and
drilled shaft. J Geotechn Geoenviron Eng, ASCE 1994;120(6):
1034–50.
[3] Brown D, Morrison C, Reese LC. Lateral load behavior of pile group
in sand. J Geotech Eng, ASCE 1988;114(11):1261–76.
[4] Mokwa RL, Duncan JM. Investigation of the resistance of pile caps
and integral abutments to lateral loading. Research report submitted
to the Virginia Transportation Research Council, Report No. VTRC
00-CR4, Charlottesville, VA, 2000.
[5] Ooi PSK, Chang BKF, Wang S. Simplified lateral analyses of fixed-
head piles and pile groups. J Geotechn Geoenviron Eng, ASCE
2004;130(11):1140–51.
[6] Poulos HG. Behavior of laterally loaded piles. II: pile groups. J Soil
Mech Found Div, ASCE 1971;97(5):733–51.
[7] Randolph MF. PIGLET: a computer program for the analysis and
design of pile groups under general loading conditions. Soil Report
TR91, CUED/D, Cambridge University, England, 1980.
[8] Banerjee PK, Driscoll PM. Three-dimensional analysis of raked pile
groups. Proc Inst Civil Eng 1976;61:653–71.
[9] Kuhlemeyer RL. Static and dynamic laterally loaded floating piles.
J Geotech Eng Div, ASCE 1979;105(2):289–304.
[10] Reese LC, O’Neill MW, Smith RE. Generalized analysis of pile
foundations. J Soil Mech Found Div, ASCE 1970;96(1):235–50.
[11] Chow YK. Three-dimensional analysis of pile groups. J Geotech Eng,
ASCE 1987;113(6):637–51.
[12] O’Neill MW, Ghazzaly OI, Ha HB. Analysis of three-dimensional
pile groups with nonlinear soil response and pile–soil–pile interaction.
In: 9th Annual OTC in Houston, 1977.
[13] Hoit M, McVay MC, Hays C. Florida pier computer program for
bridge substructure analysis: model and methods. In: Proceedings of
the design of bridges for extreme, Atlanta, 1996.
[14] Kitiyodom P, Matsumoto T. A simplified analysis method for piled
raft and pile group foundations with batter piles. Int J Numer Anal
Meth Geomech 2002;26:1349–69.
[15] McVay MC, Shang T, Casper R. Centrifuge testing of fixed-head
laterally loaded battered and plumb pile groups in sand. Geotech Test
J, ASTM 1996;19(3):41–50.
[16] US Army Corps of Engineers. Engineering and design: design of pile
foundations. Engineer Manual, No. 1110-2-2906, 1991.
[17] Hrennikoff A. Analysis of pile foundations with batter piles. Trans.
ASCE 1950;115 (paper no. 2401).
[18] Saul WE. Static and dynamic analysis of pile foundation. J Struc Div,
ASCE 1968;94(5):1077–100.
[19] Reese LC, Wang ST. Group 6.0 for Windows, Analysis of a group of
piles subjected to axial and lateral loading, Ensoft, Inc., Austin, TX,
2004.
[20] Clancy P, Randolph MF. An approximate analysis procedure of piled
raft foundations. Int J Numer Anal Methods Geomech
1993;17:849–69.
[21] Zhang HH, Small JC. Analysis of capped pile groups subjected to
horizontal and vertical loads. Comput Geotech 2000;26:1–21.
[22] Choi CK, Lee WH. Versatile variable-node flat-shell element. J Eng
Mech, ASCE 1996;122(5):432–41.
[23] Reese LC, Wang ST. Analysis of load versus settlement for an axially
loaded deep foundation. TZpile version 1.0, Ensoft, Austin, 1997.
[24] Reese LC, Wang ST. A program for the analysis of piles and drilled
shafts under lateral loads. LPILE version 4.0, Ensoft, Austin, 2000.
[25] Zhang L, McVay MC, Lai P. Numerical analysis of laterally loaded
3 · 3 to 7 · 3 pile groups in sands. J Geotech Eng, ASCE
1999;125(11):936–46.
[26] O’Neill MW, Murchison JM. An evaluation of p–y relationship in
sands. A report to the American Petroleum Institute, PRAC 82-41-1.
University of Houston (TX), 1983.
[27] Korea Ministry of Construction and Transportation, The Korea
specification for bridges, 2000.