Engineering Structures 25 (2003) 303–311

www.elsevier.com/locate/engstruct

Behavior of axially loaded pile group under lateral cyclic loading

S. Küçükarslan a,∗, P.K. Banerjee b
a
Civil Engineering Department, Celal Bayar University, PK 44, 45000 Manisa, Turkey
b
Civil Engineering Department, State University of New York at Buffalo, Buffalo, NY, USA

Received 22 April 2002; received in revised form 10 September 2002; accepted 10 September 2002

Abstract

In this paper, a hybrid boundary element technique is implemented to analyze behavior of axially loaded pile group under lateral
cycling loading. Nonlinear material behavior of soil is introduced by a rational approximation to continuum with nonlinear interface
springs along the piles. Linear beam column finite elements are used to model the piles. By enforcing displacement and equilibrium
conditions at each increment, a system of equations is generated which yields the solution. A numerical study to verify the proposed
model is performed. To investigate the cyclic behavior three groups are loaded (1×2, 2×3, and 3×3 groups) initially half of the
ultimate axial load, then a lateral loading is applied for cyclic behavior of piles, is done to investigate the behavior of pile groups.
 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Pile–soil interaction; Inelastic behavior; Lateral cyclic loading; Pile groups; Boundary and finite element methods

1. Introduction on observations of past experiment results. This

approach does not have ability to incorporate the pile
Numerical methods for analyses of piles can be done soil pile interaction effects. The use of this method to
in two ways: First one is continuum based method like predict the pile group behavior is empirical and depen-
BEM or FEM, in which continuity of the soil domain is dent on the group configuration, though attempts made
inherent in formulations; second method is a load trans- to blend continuum theory [16] and reach a unified cri-
fer method which models the soil through a set of inde- teria for different soil types [17]. It is clear that load
pendent springs attached to the piles. transfer method based on nonlinear springs is limited in
In the first approach, the use of Mindlin’s solution [1] its capability due to absence of continuum. On the other
was used in a linear boundary element formulation [2,3]. hand, continuum based approaches are highly expensive
After that, Banerjee and Davies [4] developed linear sol- if nonlinear effects are incorporated in them. An econ-
utions for piles and pile groups in nonhomogenous soils. omically reliable solution can be done by hybrid
Some finite element method approaches were attempted approach which blends two approaches into one solution.
in [5–7]. Besides them, Pressley and Poulos [8] used an It was first introduced by Chow [18] and Leung and
elastic perfectly plastic soil model in an axisymmetric Chow [19] in which they characterize the pile through
finite element method to approximately analyze pile the use of ‘p–y’ curves and incorporate group interaction
groups. Three dimensional analysis of vertically loaded effects by making use of Mindlin’s solution.
pile groups was studied in [9–11]. In practice, after a foundation is constructed, it is
In the second approach, load displacement curves of
quasi-statically loaded axially due to gravity loads of the
springs are assumed to be known a priori. This approach
superstructure. Lateral loads are transmitted through
is also known as ‘p–y’ and ‘t–z’ methods [13–15]. The
base due to lateral wind or earthquake type of loads.
nonlinear curves used in analysis are empirical and based
Therefore, it is necessary to examine the effects of slow
cyclic loads on the foundation behavior. For this pur-
∗
Corresponding author. Tel.: +90-546-2364993; fax: +90-236- pose, a hybrid type numerical formulation is presented
2412143. to investigate pile behavior of axially loaded pile group
E-mail address: [email protected] (S.
Küçükarslan).
under lateral cycling loading.

0141-0296/03/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0141-0296(02)00152-9
304 S. Küçükarslan, P.K. Banerjee / Engineering Structures 25 (2003) 303–311

2. Analytical formulations K̄ ⫽ FaGs / pD (4b)

where D is diameter of pile,Es is Young’s modulus of
A hybrid boundary element formulation is used to rep- soil, Gs is shear modulus of soil. Fa and Ft are factors
resent boundary integral representation of soil domain determining the level of nonlinearity existing at the pile
and pile equations represented by linear finite element soil interface [20,21]
structural components. Nonlinear behavior of continuum
is modeled with nonlinear springs that have a material 2.3. A single pile formulation
constitutive law and hysteretic rule. These springs are
distributed along the pile and soil interface. Pile is modeled as a linear two dimensional beams
characterized by: Ep Young’s modulus of pile material;
2.1. Continuum equations Ip second moment of inertia of pile; Ap the cross-sec-
tional area of pile; D diameter of the pile ux,and uz lateral
The soil domain is modeled as a semi-infinite con- and axial displacements. The general solution to these
tinuum. By using indirect boundary element formulation equations rewritten in incremental matrix form:
of the half space, the elastic response of the soil is {u̇p} ⫽ [D]{ṫp} ⫹ [bp]{u̇c} (5)
obtained. The static Green’s functions are used in solv-
where [D] is a coefficient matrix, [bp] is a boundary con-
ing the governing differential equation of a semi-infinite
dition matrix for a unit pile head displacements and
elastic half-space. By discretizing the pile-soil interface
rotations, {u̇c}is the vector of pile head displacements
into cylindrical elements, and integrating them along the
and rotations, {ṫp} is the incremental tractions, and the
surface, the matrix form of soil equations in an
subscript ‘p’ represents the displacements and tractions
incremental form can be represented as [20,21]:
obtained by pile domain only.
[G]{ṫs} ⫽ {u̇s} (1)
2.4. Assembly of pile and soil equations
where is the incremental displacements of the soil; [G]
is the Green’s function for the half space, {ṫs} is the By imposing a constraint on the traction vector {ṫp},
incremental tractions at the pile soil interface. the algebraic equations for soil domain and pile domain
can be coupled as following form:
2.2. Nonlinear springs along the pile surfaces [B]{ṫp} ⫽ {ḟc} (6)
A one dimensional simple method was developed by where [B] is matrix dependent on pile geometry, material
Özdemir [22]. Due to its simplicity and requiring only properties of pile and soil; {ḟc}is the vector of externally
a few parameters, it is very attractive to use it. Besides applied loads on pile cap.
its simplicity, it provides hysteretic rule that is needed
2.5. Transformations for raked pile
to model the cyclic behavior [20,21]. This model is used
to relate displacements and tractions. The modified The equations presented up to here assume that pile
Özdemir’s model is written as: is vertically in the ground (Fig. 1). For inclined piles,

冋 冉 冊册
ṫ ⫽ K ẇ⫺兩ẇ兩
t⫺b
Y
(2)

冉 冊
ḃ ⫽ aK兩ẇ兩
t⫺b
Y
(3)

where t=traction and ṫ ⫽ traction rates ( increment),

K=elastic modulus, ẇ ⫽ displacement rates (increment),
b=back stress with ḃ ⫽ incremental back stress, Y=yield
stress, a=constant controlling the slope of t–w curve.
Parameters a, Y mainly controls the behavior of the
model. The constant ‘a‘ defines the post yield rate of
the t–w curve. The elastic modulus, K in Eq. (2) is a
function of soil modulus at a particular depth. In this
study following spring stiffnesses are used.
For lateral coefficients given by:
K̄ ⫽ FtEs / D (4a)
and for axial coefficients given by: Fig. 1. Vertical pile group.
 S. Küçükarslan, P.K. Banerjee / Engineering Structures 25 (2003) 303–311 305

called as raked piles (Fig. 2), the formulations are similar this study, pile soil interface is modeled with springs to
to work presented by Banerjee and Driscoll [23]. If lij account the nonlinearity in the real physical behavior.
is a matrix defining the direction cosines of the angles Equilibrium at pile soil interface is satisfied by:
between the local and global axes,
{ṫs} ⫽ ⫺{ṫp} (12)
{u̇∗s } ⫽ [G∗]{ṫ∗s } (7)
Using Eqs. (11) and (12) and vanishing {u̇s} from Eqs.
where {u̇∗s }, {ṫ∗s } are components of incremental dis- (2) and (5) results in:

冋 册再 冎 再 冎
placements and tractions in the local system respectively.
G ⫹ D ⫹ K⫺1
ozd bp ṫp 0
[G ] ⫽ [la] [G][lb]
∗ T
(8) ⫽ (13)
B 0 u̇c ḟc
[la] is the direction cosine of the local axis of the pile
where Kozd is the equivalent nonlinear stiffness given by
at which displacements are calculated and [lb] is that
Özdemir’s model and can be represented as:
for the load application point; superscript ‘T’ denoting
transpose of matrix.
The [D] matrix in the pile equation does not change
because it is written in the local axes of pile itself. The
冋
Kozd ⫽ K 1⫺sgn(ẇ) 冉 冊册
t⫺b
Y
(14)

coefficients of the [bp] matrix are transformed by the 2.7. Formulation for a Pile Group
direction cosines of the pile since pile cap displacements
vector {u̇c}is in global coordinate system. The modified Extension of the formulation to a group is best under-
equation is rewritten as: stood by rewriting the equations presented for a single
{u̇∗p } ⫽ [D]{ṫ∗p } ⫹ [ḃp∗]{u̇c} (9) pile. The soil flexibility Eq. (2) can be rearranged as:
and
[B∗]{ṫp∗} ⫽ {ḟc} (10)
再冎 冋u̇s1
u̇s2
⫽ 册再 冎
G11 G12 ṫs1
G21 G22 ṫs2
(15)

The pile equation can be rewritten as:

2.6. Assembly of pile soil interface

By allowing the differential displacements in pile and

soil to be related by:
再 冎 冋
u̇p1
u̇p2
⫽
D11 0
0
册再 冎 冋 册
D22 ṫp2
ṫp1
⫹
b p1
bp2
{u̇c} (16)

[Kozd]({u̇s}⫺{u̇p}) ⫽ {ṫp} (11) The flexibility coefficients of the matrix [D] are
uncoupled. Whereas the coupling between the piles is
where Kozd is a nonlinear artificial stiffness which embedded in the matrix bp, that is the result of the rigid
depends on the history of displacements, stresses and cap assumption. For a two piles group, the equilibrium
material properties as shown in Eqs. (2) and (3). equation is given by:

再冎
The incremental traction vectors {ṫp}, {ṫs} and
incremental displacement vectors {u̇p}, {u̇s}, {u̇c} are ṫp1
[B1 B2] ⫽ {ḟc} (17)
unknowns. In order to couple these above equations, dis- ṫp2
placement compatibility relations have to be ensured. In
By using same procedure applied in single pile formu-
lation, one can write the following:

冤 冥冦 冧 冦 冧
G11 ⫹ D11 ⫹ K⫺1
11ozd G12 bp1 ṫp1 0
G21 G22 ⫹ D22 ⫹ K ⫺1
22ozd bp2 ṫp2 ⫽ 0 (18)

B1 B2 0 u̇c1 ḟc

In Eq. (18), the indices ‘1’ and ‘2’ resembles to piles

in the group.

3. Verification of the model

A study was done (by Throchanis et al.) in [11,12],

to analyze pile foundations under axial and lateral loads
by using a three-dimensional finite element method code
Fig. 2. Raked pile group. ABAQUS. The piles and soil were modeled by solid
306 S. Küçükarslan, P.K. Banerjee / Engineering Structures 25 (2003) 303–311

three-dimensional quadratic isoparametric 27 node 3.1. Behavior of pile under axial loading
elements (nine nodes per face) included in the finite
element library of ABAQUS, the commercial finite Pile soil slippage and yielding at the base primarily
element method package used. The interface elements govern the nonlinear axial deflections. The large amount
were 18-node elements comprising two nine-node sur- of the axial load is resisted through skin friction and the
faces compatible with the adjacent solid elements; the pile base contributes less than 10% of the total load.
two surfaces coincide initially. The pile elements are Load carrying capacity at the level of each depth is dic-
assumed to remain elastic at all times, while the soil was tated by the limiting friction value at that level. Above
idealized as a Ducker–Prager elastoplastic material. This this value, the pile and soil slip locally. Since any extra
model was selected from the library of ABAQUS, since additional loading is transferred to the base, the limiting
it can be represent soil dilatancy and its parameters were value capacity of shaft is very close to the ultimate load
related to the physical soil properties (cohesion and fric- carrying capacity.
tion angle) in a rather straightforward way. The interface The FEM model and BEM load settlement results
elements can transfer only shear forces across their sur- agree quite well as shown plotted in Fig. 3. The side
faces when a compressive normal pressure acts on them; strength in three dimensional FEM model varies with
otherwise a gap opens between them. When contact depth up to level of 84 kN/m at the pile tip. All the other
exists, the relationship between shear force and normal soil properties used in BEM were identical to those used
pressure is governed by modified Columb’s friction in the FEM.
theory. As a result, these elements are completely In the physical problem, the strength of bond is affec-
defined by their geometry. A friction coefficient and an ted by the type of soil and pile installation procedure.
elastic stiffness was used to provide convergence. The deflection of pile drags the soil along with it before
In order to validate the results of the FEM model, the slipping occurs. In Fig. 4, axial load distribution is
predicted pile response was compared to actual experi- shown.
mental results from a test conducted in Mexico City. A It is also observed that, more axial force is transferred
15 m long and 0.3 m width square pile tested by apply- from pile to soil at greater depths indicated by greater
ing a vertical load to its head. The average effective curvature of nonlinear axial force curve as opposed to
properties of the soil at site were: friction angle f=16.7°, the transfer at smaller depths.
and cohesion c=34 kPa. Other estimated soil parameters
were: submerged unit weight gs=11.8 kN/ m3 , shear 3.2. Behavior of pile under lateral loading
modulus Gs=6800 kPa and Poisson’s ratio ns=0.45. The
experimental load settlement curve can be found in Fig. Comparison of overall load deflection and deflected
3 along with the FEM and BEM results. The FEM sol- shape curves at peak load are illustrated in Figs. 5 and
ution was obtained by modeling with a Young’s modulus 6, respectively. The model is compared in the case of
Ep=2×107 kPa, and using the Drucker–Prager model for lateral loading by allowing separation at pile soil inter-
the soil. face. The nonlinear model posses a bilinear lateral
strength distribution with depth. The present analysis is
showed that agreements are fairly good with 3D FEM

Fig. 3. Comparison of three dimensional model with test in Mexico

City. Fig. 4. Comparison of lateral load displacement at the pile head.
 S. Küçükarslan, P.K. Banerjee / Engineering Structures 25 (2003) 303–311 307

Fig. 5. Comparison of lateral displacement along the pile. Fig. 7. Comparison of shaft normal load distribution along the pile.

Fig. 6. Comparison of shaft normal load distribution along the pile.

Fig. 8. Lateral cyclic behavior of a single pile.

results. Fig. 7 shows the comparison of pile shaft normal
pressure distribution obtained from FEM and BEM. In
the FEM study, it was shown that the length over which
a gap has opened up can be identified by zero traction
on the trailing face. For normal traction values in the
lower quarter of the pile are due to the geostatic stresses
considered in the FEM model. A slight difference in the
pressure distribution on the leading face is resulted the
effect of higher stress prediction [11,12].

4. Cyclic behavior of a single pile

In this section, the salient features of the cyclic

behavior are studied. For this purpose, a single pile is
subjected to axial and lateral cyclic loads and results of
load displacements are plotted in Figs. 8 and 9. The
behavior under axial and lateral cyclic loads is different
and it is influenced by development of the mechanisms
of the nonlinearity. Fig. 9. Axial cyclic behavior of a single pile.
308 S. Küçükarslan, P.K. Banerjee / Engineering Structures 25 (2003) 303–311

L/D=20 and S/D=3 are used in analysis L=10 m and

D=0.5 m. Estimated soil parameters are: shear modulus
Gs=6800 kPa and Poisson’s ratio ns=0.45. Pile Young’s
modulus is Ep=2×107 kPa. An axial load of 400 kN per
pile is applied and then five cycles of a lateral load of
300 kN per pile are applied to the groups.
The load deflection curves are plotted in Figs. 12, 15
and 18. For all groups, the axial behavior shows a mildly
nonlinear response at half the ultimate vertical load. The
deflections are plotted against the per pile load for com-
paring the groups.
After subsequent application of lateral load to the
groups, the group settles more in every lateral cyclic
load. Amount of additional settlement is appreciable for
the level of lateral loading considered and its gradually
decreases with each cycle of load. This type of behavior
Fig. 10. Axial load settlement curve for ultimate load capacity of is observed in all groups.
pile. The sources of additional settlement are: (1) the axial
stiffness of individual piles decreases thereby causing
In the axial mode, pullout force causes a higher dis- downward movement due to the opening of the gap near
placement due to low resistance of base. By increasing the ground surface; (2) the applied lateral load is resisted
the number of cycles, this causes a steady degradation. by the push pull action among the piles in the group.
The change in slope towards each compressive loading
corresponds to the pile making a contact with the pile
base. However, the lateral cyclic behavior of pile is
strongly affected by pile soil separation phenomenon.
This indicates a pinching characteristic in the hyster-
etic response.

5. Behavior of axially loaded pile group under

lateral cyclic loading

Pile group behavior strongly depends on the geometry

and material characteristics of soil and piles also pile
configuration. In this section, three groups are loaded
(1×2, 2×3, and 3×3 groups) initially half of the ultimate
axial load (Fig. 10), then lateral loading is applied as a
cyclic load.
Group configurations are shown in Fig. 11 in which Fig. 12. Axial load settlement curve for 1×2 group.

Fig. 11. Pile group configurations used in analysis.

 S. Küçükarslan, P.K. Banerjee / Engineering Structures 25 (2003) 303–311 309

Fig. 13. Individual pile behavior in group 1×2.

Fig. 15. Axial load settlement curve for 2×3 group.

Fig. 14. Lateral cyclic behavior of 1×2 group.

Fig. 16. Individual pile behavior in group 2×3.
The lateral load induces uneven levels of nonlinearity at
the interface due to the nonlinear stress history at the
pile soil interface.
Under the vertical load all piles behave identically due
to the symmetric geometry. By applying lateral loads to
the groups, pile 1 in 1×2 group; piles 1 and 4 in 2×3
group; and piles 1,4 and 7 are pulled upward in Figs.
13, 16 and 19 respectively. Whereas pile 2 in 1×2 group;
piles 3 and 6 in 2×3 group; and piles 3, 6 and 9 are
pushed downward.
This means that these downward moving piles are
subjected to additional load taking them to a higher non-
linear regime while the load in piles moving upwards is
partially released. In lateral unloading, behavior of these
piles are reversed. Subsequent cycles of lateral load
accentuate this case and at the end of all cycles piles
going upwards carries a higher amount of axial load than
those going downwards.
Fig. 17. Lateral cyclic behavior of 2×3 group.
The lateral load deflection curves are plotted in Figs.
14, 17 and 20 for three types of groups. Subsequent
310 S. Küçükarslan, P.K. Banerjee / Engineering Structures 25 (2003) 303–311

increase of lateral load reveal a constant increase in cap

movement, which occurs due to the increase in pile soil
separation and increase in nonlinearity.

6. Conclusions

A hybrid type of numerical algorithm was presented

to analyze axially loaded pile group under lateral cyclic
loading. Nonlinear effects were taking into account by
Özdemir’s model. Coupled systems of equations were
formulated for a single pile and pile groups. This algor-
ithm benefits from the advantages of continuum and ‘p–
y’, ‘t–z’ type methods. To present the efficiency of pro-
posed algorithm, a numerical example was studied for
1×2, 2×3, and 3×3 groups which were loaded initially
Fig. 18. Axial load settlement curve for 3×3 group. half of the ultimate axial load, then a lateral loading is
applied for cyclic behavior of piles. The results indicated
that the presence of the continuum in the formulation
allows geometrical effects to be captured whereas
localized nonlinear effects are adequately modeled by
interface springs.

References

[1] Mindlin HD. Force at a point in the interior of a semi infinite

solid. J. Applied Physics 1936;7:195–202.
[2] Poulos HC. Analysis of the settlement of pile group. Geotech-
nique 1968;18:449–71.
[3] Banerjee PK. Analysis of axially and laterally loaded pile groups.
Developments in Soil Mechanics. pp. 317–346 London: Applied
Science Publishers, 1978.
[4] Banerjee PK, Davies TC. The behavior of axially and laterally
loaded piles embedded in nonhomogeneous soils. Geotechnique
1978;28(3):309–26.
[5] Desai CS. Numerical design and analysis for piles in sand. J.
Fig. 19. Individual pile behavior in group 3×3. Geotech. Engng. Div., ASCE 1974;100(6):613–35.
[6] Faruque MO, Desai CS. 3D material and geometric nonlinear
analysis of piles. In: Proc Second Int Conf on Numerical Methods
in Offshore Piling, Texas, 1982. p. 553–7.
[7] Randolph MF. The response of flexible piles to lateral loading.
Geotechnique 1981;31(2):247–59.
[8] Pressley JS, Poulus HG. Technical notes on practical appli-
cations: finite element analysis of mechanisms of pile group. Int.
J. Numer. Anal. Methods Geomech. 1987;10:213–21.
[9] Ottaviani M. Three-dimensional finite element analysis of verti-
cally loaded pile groups. Geotechnique 1975;25(2):159–74.
[10] Muqtadir A, Desai CS. Three dimensional analysis of a pile group
foundation. Int. J.Numer. Anal. Methods Geomech.
1986;10(1):41–55.
[11] Trochanis AM, Bielak J, Christiano P. A three-dimensional non-
linear study of piles. J. Geotech. Engng., ASCE
1991;117(3):429–47.
[12] Trochanis AM, Bielak J, Christiano P. Simplified model for
analysis of one or two piles. J. Geotech. Engng., ASCE
1991;117(3):448–66.
[13] Coyle HM, Reese LC. Load transfer for axially loaded piles in
clay. J. Soil Mech. Found. Engng. Div. 1966;92(2):1–26.
[14] Matlock H. Correlations for design of laterally loaded piles in
soft clay. Offshore Technology Conference, OTC 1208, Houston,
Fig. 20. Lateral cyclic behavior of 3×3 group. TX, 1970.
 S. Küçükarslan, P.K. Banerjee / Engineering Structures 25 (2003) 303–311 311

[15] Reese LC, Cox WR, Koop FD. Analysis of laterally loaded piles [20] Guin J. Advanced soil pile structure interaction and nonlinear
in sand. In: Proc 6th Offshore Technology Conference, OTC behavior. PhD thesis, State University of New York, Buffalo,
2080, 1974. NY, 1997.
[16] O’Neill MW, Hawkins FA, Mahar LJ. Load transfer mechanisms [21] Küçükarslan S. Linear and nonlinear soil–pile–structure interac-
in piles and pile groups. J. Geotech. Eng., ASCE tion under static and transient impact loading. PhD thesis, State
1982;108(12):1605–23. University of New York, Buffalo, NY, 1999.
[17] Gazioglu SM, O’Neill M. Evaluation of p-y relationships in [22] Özdemir H. Nonlinear transient dynamic analysis of yielding
cohesive soils. In: Meyer JF, editor. Analysis and design of pile structures. PhD thesis, University of California, Berkeley, CA,
foundations. New York: ASCE; 1984. p. 192–213. 1976.
[18] Chow YK. Axial and lateral response of pile groups embedded [23] Banerjee PK, Driscoll RC. Three-dimensional analysis of raked
in nonhomogeneous soils. J. Num. Anal. Methods Geomech. pile groups. In: Proc Institute Civil Engrs, vol. 61. London, 1976.
1987;11(3):621–38. p. 653–71.
[19] Leung CF, Chow YK. Response of pile groups subjected to lat-
eral loads. Int. J. Numer. Anal. Methods Geomech.
1987;11(1):307–14.