You are on page 1of 11

Applicability of Conventional p-y Relations to the

Analysis of Piles in Laterally Spreading Soil


Christopher R. McGann1; Pedro Arduino, M.ASCE2; and Peter Mackenzie-Helnwein, M.ASCE3

Abstract: This paper presents a kinematic analysis of a single pile embedded in a laterally spreading layered soil profile and discusses the
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

relevancy of conventional analysis models to this load case. The research encompasses the creation of three-dimensional (3D) finite-element
(FE) models using the OpenSees FE analysis platform. These models consider a single pile embedded in a layered soil continuum. Three
reinforced concrete pile designs are considered. The piles are modeled using beam-column elements and fiber-section models. The soil
continuum is modeled using brick elements and a Drucker-Prager constitutive model. The soil-pile interface is modeled using beam-solid
contact elements. The FE models are used to evaluate the response of the soil-pile system to lateral spreading and two alternative lateral load
cases. Through the computation of force density-displacement (p-y) curves representative of the soil response, the FE analysis (FEA) results
are used to evaluate the adequacy of conventional p-y curve relationships in modeling lateral spreading. It is determined that traditional p-y
curves are unsuitable for use in analyses where large pile deformations occur at depth. DOI: 10.1061/(ASCE)GT.1943-5606.0000468.
© 2011 American Society of Civil Engineers.
CE Database subject headings: Soil-pile interactions; Lateral loads.
Author keywords: Lateral spreading; Soil-pile interaction; 3D FEA; p-y curves; Laterally loaded piles.

Introduction applications and has been extended for used in layered soils by
Georgiadis (1983), as well as liquefied and laterally spreading soils
Piles are used extensively to provide support for bridges and wharf by Wang and Reese (1998). Finn and Fujita (2002) compared the
facilities, very often in seismically active regions. In such areas, soil results of dynamic 1D BNWF analyses with centrifuge test data
profiles can exist in which a relatively loose saturated sand layer is and dynamic 3D FEA and found good agreement at low acceler-
located between two layers of more dense or unsaturated sand. Dur- ation levels and with certain curve parameters. Brandenberg et al.
ing or slightly after a seismic event, a liquefied condition can de- (2007) identified design guidelines for the lateral spreading prob-
velop in the loose liquefiable layer with the potential for lateral lem in the context of a BNWF approach that produce results that are
displacement of the upper nonliquefiable layer relative to the bot- reasonably similar to centrifuge test data. Therefore, despite known
tom soil layer and a corresponding kinematic demand on an em- shortcomings inherent to the method, the p-y approach can be
bedded pile. Because of the complexity of this lateral spreading used successfully in many types of lateral pile analysis. To this pur-
problem, it is difficult to obtain a reasonable estimate of the pose, and to obtain sensible results, the force density (unit soil
pile-bending moment and shear demands. As a result, the conserva- resistance)–displacement behavior of the p-y curves must be chosen
tism inherent in current simplified analysis methods often lead to carefully.
overdesigned and overly expensive solutions. The p-y curves most commonly used for cohesionless soils,
The most common method employed for the numerical analysis
those recommended by the American Petroleum Institute (API)
of piles under lateral loads is the p-y method (McClelland
(1987), are largely based upon the results of lateral load tests (Reese
and Focht 1958; Matlock and Reese 1960; Reese and Van Impe
et al. 1974). The resulting p-y curves are based upon the interpre-
2001). In this method, the three-dimensional (3D) laterally
tation of this experimental evidence using a simplified analytical
loaded pile problem is analyzed by using a beam on nonlinear
model (BNWF) that may or may not be applicable. In the test used,
Winkler foundation (BNWF) approach in which uncoupled
large pile deflections occur at shallow depths while little or no pile
one-dimensional (1D) springs are used to described the soil-pile
deflections occur at increased depths. As a result, the response of
interaction. This approach is used in many lateral pile design
the soil near the ground surface is captured fairly accurately,
1
Ph.D. Candidate, Dept. of Civil and Environmental Engineering,
whereas any response at depth cannot be characterized. Lack of
Univ. of Washington, Box 352700, Seattle, WA 98195-2700. E-mail: information at depth is relatively inconsequential to the analysis
mcganncr@uw.edu of piles with lateral loads applied at or above the ground surface
2 but becomes problematic for load cases in which large pile deflec-
Associate Professor, Dept. of Civil and Environmental Engineering,
Univ. of Washington, Box 352700, Seattle, WA 98195-2700 (correspond- tions occur at depth, such as the cases of lateral spreading and land-
ing author). E-mail: parduino@u.washington.edu slide stabilization.
3
Research Assistant Professor, Dept. of Civil and Environmental Engi- This paper evaluates the applicability of the API (1987) recom-
neering, Univ. of Washington, Box 352700, Seattle, WA 98195-2700. mended p-y curves for cohesionless soil to the lateral spreading
Note. This manuscript was submitted on November 11, 2009; approved
problem. This is accomplished by using 3D FE models that sim-
on October 20, 2010; published online on October 27, 2010. Discussion
period open until November 1, 2011; separate discussions must be sub- ulate a liquefaction-induced lateral spread. This approach elimi-
mitted for individual papers. This paper is part of the Journal of Geotech- nates the simplifying decoupling assumption used in the BNWF
nical and Geoenvironmental Engineering, Vol. 137, No. 6, June 1, 2011. approach and inherently includes potentially overlooked nonlocal
©ASCE, ISSN 1090-0241/2011/6-557–567/$25.00. coupling effects. Representative p-y curves are computed from 3D

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011 / 557

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


FEA using elastic pile elements and elastoplastic soil constitutive solid-to-liquefied layer interface at a higher resolution. The soil
behavior. profile modeled by this mesh is intended to represent an extreme
This paper also demonstrates that the choice of kinematics is of case which allows deep soil-pile interaction to be observed.
crucial importance to the identification of p-y curve characteristics Beam-solid contact elements (Petek 2006) are used to model the
suitable for a simplified, i.e., decoupled BNWF, analysis. The ways soil-pile interface. The contact elements create a link between the
in which differing pile kinematics affect the resulting p-y curves are line elements (pile) and brick elements (soil), enabling the use of
explored by using two alternative pile kinematic cases. The effect standard beam-column elements to model the piles. This allows for
of the liquefied layer interface on the lateral response of the unli- the use of fiber section models to describe the pile constitutive
quefied layers is considered through comparisons of results ob- behavior and enables the results to be interpreted in the context
tained using homogenous and liquefied soil profiles. Once a of traditional beam theory. The contact elements have the capacity
suitable process has been established for computing p-y curves for pile-soil separation and add a frictional interface allowing
by using the 3D FEA, these curves are compared with the API for sticking and frictional slip following a regularized Coulomb
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

(1987) curves through a series of 1D BNWF analyses that consider law (e.g., Wriggers 2002).
both linear elastic and elastoplastic pile section behavior. The re- The soil and pile nodes on the base of the model are fixed
sults of these analyses are compared with similar results obtained against displacements in the vertical direction (parallel to the pile
with 3D FEA. axis). All of the nodes on the symmetry plane are fixed against
translation normal to this plane. Torsional and out-of-plane rota-
tions of the pile are prevented to enforce symmetry, but the pile
Finite-Element Model is allowed to rotate and curve in the plane of loading. Two pile head
support conditions are considered: a free condition and a fixed sup-
A 3D soil-pile interaction model is created for the case of a single port condition, representative of a pile cap or other structural com-
circular pile embedded in a soil continuum. Fig. 1 shows the FE ponent. All of the soil nodes lying on the outer surfaces of the
mesh used for the analysis in its deformed configuration for the model are held fixed against horizontal in-plane translations and
lateral spreading case. Symmetry conditions are used as shown. translations normal to their surface to enhance stability and to allow
Displacement-based beam-column elements are used to model for the kinematic loading of the lateral spreading case.
the pile, eight-node brick elements are used to model the soil, Gravity loads are applied using a soil unit weight of
and beam-solid contact elements are used to define the soil-pile γ ¼ 17 kN=m3 . This procedure creates an appropriate distribution
interface (Petek 2006; Lam et al. 2009). of confining pressure in the model, critical to achieving depth-
Three pile designs are considered in this study. These piles have appropriate shear strength in the soil. The self-weight of the pile
diameters of 0.6096 m (24 in.), 1.3716 m (54 in.), and 2.5 m is neglected. The kinematic loading of the lateral spreading event
(98.4252 in.). For simplicity, these piles will be referred to as is achieved in the model by gradually imposing a set displacement
the small, medium, and large-diameter piles, respectively. The di- profile to the soil nodes on all outer surfaces of the mesh excluding
mensions of the FE mesh are on the basis of the diameter of the the symmetry plane. The imposed displacement profile represents
embedded pile. The typical mesh, shown in Fig. 1, has a height the far-field kinematic demands on the soil system.
of 21 pile diameters. In the cases in which a liquefied layer is con- This far-field deformation is applied to the boundary of the
sidered, the liquefied layer is one pile diameter thick and located at model to simulate a lateral spread of large areal extent. The im-
the center of the mesh. A diameter-dependent mesh was chosen so posed displacement profile is depicted in Figs. 1 and 2(a). The pro-
each pile is subject to a dimensionally consistent deformation file is constant over the entire height of the upper unliquefied layer
pattern. Past work with 3D FEA of laterally loaded piles (Yang and is linearly increasing over the height of the liquefied layer. The
and Jeremic 2005; Brown and Shie 1990) used lateral extents of boundary surfaces of the lower unliquefied layer (excluding sym-
13 and 11 pile diameters, respectively. Using this information metry plane) are held fixed. The maximum magnitude of the
and past modeling experience, the models include a horizontal far-field horizontal displacement is taken to be one pile radius.
extent of 10 pile diameters from the pile centerline. The mesh is All numerical simulations are conducted using the open-source
refined at the center to capture the soil-pile interaction near the FE framework that OpenSees (http://opensees.berkeley.edu)

∆ ∆ ∆

(a) (b) (c)

Fig. 2. Three considered pile kinematic cases: (a) lateral spreading;


Fig. 1. Typical 3D FE mesh in deformed configuration (b) top pushover; (c) rigid pile

558 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


Table 1. Material Properties and Input Parameters Used in Soil Constitutive Models
Soil layer E o (kPa) ν ϕ (°) σY (kPa) γ (kN=m3 ) ρ ρ
Solid sand 25,000 0.35 36 5 17 0.398 0.150
Liquefied sand 2,500 0.485 0 5 17 0.000 0.000

pffiffiffi
developed through the Pacific Earthquake Engineering Research 2 2 sin ϕ
(PEER) Center. The commercial program, GiD (CIMNE 2008), ρ ¼ pffiffiffi ð3Þ
3ð3  sin ϕÞ
is used as a graphical pre- and postprocessor for OpenSees.
6c cos ϕ
σY ¼ pffiffiffi ð4Þ
Constitutive Modeling 2ð3  sin ϕÞ
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

Material nonlinearity in the soil is considered through a Drucker- A small amount of cohesion, c ¼ 3:5 kPa, is assigned to the soil for
Prager constitutive model, which includes pressure-dependent numerical stability. Internal friction angles of ϕ ¼ 36° and ϕ ¼ 0°
strength, tension cutoff, and nonassociative plasticity. In this are used in the unliquefied and liquefied layers to represent a
model, isotropic elastic behavior is controlled by the elastic modu- medium-dense sand and a liquefied sand, respectively. The param-
lus, E, and Poisson’s ratio, ν (or alternatively bulk modulus K b , and eter ρ controls the evolution of plastic volume change. For this
shear modulus G). Written in stress-type variables, the yield con- analysis, the unliquefied layers are considered to be drained. On
dition for the Drucker-Prager model can be expressed as this basis and the assumption of a medium initial density, an inter-
rffiffiffi mediate case where ρ ¼ 0:150 is used so there will be some dilation
pffiffiffiffiffi 2
f ðσÞ ¼ s∶s þ ρ trσ  σ ≤0 ð1Þ of the soil.
3 Y It is recognized that this soil constitutive model represents a sim-
where s ¼ devσ ¼ σ  ðtr σ=3Þ1 = deviatoric stress tensor; and plification of the true behavior of the soil, particularly for the lique-
ρ and σY are positive material parameters that define the yield sur- fied layer. The behavior of liquefied sand is too complex to be
face. A nonassociative plastic flow rule is defined by a parameter ρ, captured with this model, and some of the critical state behavior
which ranges from 0 ≤ ρ ≤ ρ. of the unliquefied sand is lost. The chosen model does capture
The upper and lower solid layers are defined with matching in- the frictional strength of the sand and the dependence of the
put parameters that are assumed to reflect drained conditions. The strength on the mean stress. Given the limitations of computing
liquefied layer, when used, has a separate set of parameters that power for a 3D analysis, this constitutive model was deemed
reflect a liquefied (undrained) condition. Variation in the elastic acceptable for the purposes of this work.
modulus, E, with confining pressure is accounted for using Pile models are created based upon three commonly used pile
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi designs. The template-reinforced concrete pile designs with dimen-
jσ j sions and reinforcement details are shown in Fig. 3. Linear elastic
E ¼ Eo 1 þ m ð2Þ
jpa j and nonlinear elastoplastic pile models are created using these tem-
plate designs. For elastoplastic behavior, fiber cross-section models
where Eo = elastic modulus at atmospheric pressure, pa ; and σm ¼ are used to represent the pile by a bundle of nonlinear fibers. Each
trσ=3 = mean stress. The selected value of Eo is established on the fiber represents an individual reinforcement bar or a concrete por-
guidelines of Kulhawy and Mayne (1990) considering a modulus tion of the pile by assigning to it the appropriate uniaxial nonlinear
reduction appropriate for the expected strain levels for the lateral material behavior. Limited tensile strength and cracking is included
spreading deformation. Poisson’s ratio for the unliquefied layers is in the concrete model and the increased compressive strength effect
chosen to be representative of a medium dense sand based on the of confinement is considered. The reinforcing steel has linear elas-
guidelines of Kulhawy and Mayne (1990). The material properties tic-plastic behavior with linear strain hardening. Kinematic com-
for all layers are summarized in Table 1. patibility of the cross section ensures that the resulting coupled
The material constants ρ and σY can be computed from the force and moment versus strain and curvature response of the pile
Mohr-Coulomb cohesion, c, and internal friction angle, ϕ, by using is properly captured. The results of this paper are presented pri-
Eqs. (3) and (4) after Chen and Saleeb (1994) marily for the medium-diameter pile. The OpenSees constitutive

9.5 mm ties 9.5 mm ties 22.2 mm ties


@ 76.2 mm @ 76.2 mm @ 200 mm

0.208 m 0.313 m 0.586 m 0.686 m 1.06 m 1.25 m


64 mm
casing

16, 15.2 mm dia 36, 15.2 mm dia 30, 57.3 mm dia


longitudinal bars longitudinal bars longitudinal bars
(a) (b) (c)

Fig. 3. Template-reinforced concrete pile designs: (a) small-diameter (0.6096 m) pile; (b) medium-diameter (1.3716 m) pile; (c) large-diameter
(2.5 m) pile

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011 / 559

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


7000
Conducting 3D FEA using a sand-clay-sand profile, Yang and Jere-
6000
mic (2005) demonstrated how the presence of a weaker soil layer
between two stronger layers reduces the available lateral resistance
of the stronger layers. The presence of a liquefied layer in a soil
moment M (kNm)
5000
profile affects the unliquefied layers in a similar manner. This
4000 reduction in the available lateral resistance must be reflected in
the p-y curves used in a lateral spreading analysis. To evaluate
3000 the effects of differing pile kinematics and layer strengths on
p-y curves, the 3D model is used to compute p-y curves for various
2000 pile-soil-loading combinations. These curves are also used as a di-
rect comparison to the API (1987) recommended curves.
1000
In contrast to back-calculating the pile forces from the moment
diagram as done in the experiment by Reese et al. (1974), the beam-
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

0
0 0.01 0.02 0.03 0.04 solid contact elements (Petek 2006) directly provide the forces ex-
−1
curvature φ (m ) erted on the pile. The contact elements resolve the interface forces
applied by the soil to the outer surface of the pile into a vector with
Fig. 4. Moment-curvature behavior for a medium-diameter (1.3716 m) three orthogonal components acting at each pile node (two lateral,
pile one axial component), though only the lateral component in the
direction of loading is needed. The process of computing this force
model, Concrete02, is used for the concrete portion of this pile with component is depicted schematically for the rigid pile case in Fig. 5.
the following input parameters: f 0c ¼ 43;170 kPa, εc ¼ 0:003, Elastic pile models are used to compute p-y curves from the 3D
f 0cu ¼ 8;270 kPa, εcu ¼ 0:011, f t ¼ 4;000 kPa, and Et ¼ model so only the nonlinear response of the soil is captured.
The forces obtained from the contact elements are distributed
1;400 MPa. The Steel01 constitutive model is used for the
over the tributary length of the pile associated with each respective
reinforcement in pile with the following input parameters: σY ¼
pile node to obtain mesh-independent curves. The resulting force
1;860 MPa, E ¼ 200 GPa, and b ¼ 0:001. Fig. 4 shows the com-
densities (units of FL1 ) are used to define the resistance portion
puted moment-curvature response for a medium-diameter pile to
(p) of the computed p-y curves. To obtain the displacement portion
demonstrate the realistic nonlinear behavior captured by the fiber
(y), the displacements of the pile relative to the surrounding soil are
model.
determined. Comparisons between sets of p-y curves are made us-
Linear elastic fiber models are also used for the piles in the 3D
ing two characteristic curve parameters: the initial tangent stiffness,
FE analyses to extract suitable p-y soil-response curves and as a
kT , and ultimate lateral resistance, pu .
control case for the elastoplastic models in the 1D BNWF analyses.
Force density and displacement values from the 3D models
These elastic pile models are assigned cross-sectional properties
are only obtained at a series of discrete points during the analyses.
established on their respective diameters and are assigned elastic
To determine the characteristic parameters necessary for compari-
moduli, E, such that their bending stiffness, EI, matches the initial
son, smooth continuous functions are fit to the data using least
bending stiffness for the fiber models. Linear elastic pile material squares. Two functional forms are used for this purpose. For direct
and section properties are summarized in Table 2. comparison to the API (1987) recommendations for p-y curves, a
hyperbolic tangent function is fit to the data to define a smooth
representative curve and a suitable value of pu . The fitted function
Computation of p-y Curves from the 3D has the form
Finite-Element Model  
kz
pðyÞ ¼ pu tanh y ð5Þ
The original experiments by Reese et al. (1974) to obtain p-y curves pu
for cohesionless soils used a single pile kinematic case. In the 35 where z = depth; and k = parameter that affects the slope of the
years since these tests, these results have been applied to many lat- function. The initial tangent stiffness, k T , in Eq. (5) is obtained
eral pile analyses, sometimes with differing pile kinematics. The as k T ¼ dpðyÞ=dyjy¼0 ¼ kz. An example of a computed p-y data
Winkler foundation analytical model used to interpret the experi- set with a corresponding fitted hyperbolic tangent function is pro-
mental results does not consider coupling between the springs. To vided in Fig. 6. As shown, the initial tangent stiffness of the fitted
include coupling, higher-order terms must be included in the gov- curve tends to be less than the initial tangent stiffness indicated by
erning differential equation. Because of the lack of coupling be- the data.
tween springs, certain effects of the pile kinematics on the p-y To obtain a better fit for the initial tangent stiffness of the com-
curves are lost, especially where pile displacements are relatively puted curves, a cubic polynomial function is fit to the first four data
small. This can lead to significant errors in an analysis with pile points. The chosen polynomial function has the form
kinematics that do not match the original experiments.
pðyÞ ¼ a1 y þ a3 y3 ð6Þ
Additionally, in the context of the liquefaction-induced lateral
spreading problem, there is a resistance loss mechanism near the from which the initial stiffness is obtained as kT ¼ dpðyÞ=
layer interface that must be accounted for in a numerical analysis. dyjy¼0 ¼ a1 . As shown in Fig. 6, the initial tangent stiffness

Table 2. Material and Section Property Values Used in Linear Elastic Pile Models
Pile diameter A (m2 ) E (GPa) G (GPa) I y (m4 ) I z (m4 ) J (m4 )
0.6096 m 0.154 31.3 12.52 0.0038 0.0038 0.0075
1.3716 m 0.739 28.7 11.48 0.0869 0.0869 0.1738
2.5 m 2.454 102.4 40.96 0.9587 0.9587 1.9174

560 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Extraction of p-y curves from the 3D FE model

obtained from the fitted polynomial function fits the computed data The rigid pile case is used as a benchmark to evaluate the effects
more closely than the initial tangent stiffness indicated by the hy- of pile kinematics on p-y curves. This case eliminates the influence
perbolic tangent function, although Eq. (5) provides a fit for pu. of pile deformation on the computed curves and activates the soil
Hence, the polynomial initial tangent is used for all work related response consistently at all depths. The lateral spreading and top
to initial curve stiffness. pushover cases are compared to the benchmark case with respect
to initial tangent stiffness, kT , and a lateral resistance ratio (LRR).
The LRR compares the resistance provided at a particular depth
Effects of Pile Kinematics on Computed p-y Curves and specified level of lateral displacement to that provided at
the same depth and level of displacement in the benchmark case
Three pile deformation patterns are considered to evaluate the ef- (Yang and Jeremic 2005). The LRR is evaluated at lateral displace-
fects of varying pile kinematics on the p-y curves resulting ments equal to 0.5%, 1%, 2%, and 4% of the pile diameter in each
from the FEA. These cases include: (1) a lateral spreading case respective case.
[Fig. 2(a)], which is the primary focus of this research; (2) a top The lateral spreading analyses use a layered soil profile with a
pushover case [Fig. 2(b)], matching the type of lateral load test liquefied layer in the center, with material properties as given in
used to empirically derive p-y curves; and (3) a rigid pile case Table 1. The top pushover analyses use a homogenous soil profile,
without a liquefied layer, having the properties of the solid layers in
[Fig. 2(c)], in which the pile is pushed a uniform lateral distance
Table 1. Each case is compared to a rigid pile case with a corre-
into the soil. All kinematic cases are modeled by using the FE mesh
sponding soil profile (i.e., a rigid pile case with a liquefied center
used in the lateral spreading simulations.
layer and a rigid pile case with a homogenous soil profile).
The top pushover case simulates a displacement-controlled lat-
The computed kT ratios and LRRs are presented in Figs. 7 and 8
eral load test. The top node of the pile is displaced laterally in the for the lateral spreading and top pushover kinematic cases, respec-
symmetry plane while all other pile nodes are free to translate and tively, for a medium-diameter pile. These results are typical of all
rotate in this plane. The pile head is left to rotate freely. This kin- considered piles. The evolution of the pile’s displacement relative
ematic mode results in large deformations near the surface and to the soil is provided alongside the ratios. As shown, there are
small deformations at greater depths. For the rigid pile case, all pile numerous locations in each load case in which the displacements
nodes are translated laterally into the soil without rotation. The re- are small and little soil resistance is mobilized. Because of this,
maining boundary conditions on the soil are unchanged. ultimate lateral resistances and LRRs for displacements greater

00 00 00
OpenSees data
8000 −2
−2 −2
−2 −2
−2
Fitted tanh curve
Fitted initial tangent −4
−4 −4
−4 −4
−4
force density p (kN/m)

normalized depth

6000 −6
−6 −6
−6 −6
−6

−8
−8 −8
−8 −8
−8

−10
−10 −10
−10 −10
−10
4000
−12
−12 −12
−12 −12
−12

−14
−14 −14
−14 −14
−14

2000 −16
−16 −16
−16 −16
−16 y/D = 0.5%
y/D = 1.0%
−18
−18 −18
−18 −18
−18 y/D = 2.0%
y/D = 4.0%
−20
−20
−0.2 0 0.2
−20
−20 −20
−20
0 −0.2 0 0.2 00 1
1 22 00 1
1 22
0 0.1 0.2 0.3 kT ratio LRR
kinematic normalized
displacement y (m) case displacement

Fig. 6. Example of computed p-y data with fitted initial tangent stiff- Fig. 7. Comparison of the lateral spreading kinematic case to the rigid
ness and hyperbolic tangent curve for a medium-diameter (1.3716 m) pile case for a medium-diameter (1.3716 m) pile; both analyses use a
pile three-layer liquefied soil system

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011 / 561

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


00 00 00 0
−2
−2 −2
−2 −2
−2

−4
−4 −4
−4 −4
−4 −5

normalized depth
−6
−6 −6
−6 −6
−6
Liquefied Layer
−8
−8 −8
−8 −8
−8 −10

depth (m)
−10
−10 −10
−10 −10
−10

−12
−12 −12
−12 −12
−12 −15
−14
−14 −14
−14 −14
−14

−16 −16 −16 y/D = 0.5% −20


−16 −16 −16
y/D = 1.0%
−18
−18 −18
−18 −18
−18 y/D = 2.0%
y/D = 4.0% −25
−20
−20 −20
−20 −20
−20 Homogenous Profile
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

−0.2
00 0
0.5
0.5 0.211 00 1
1 22 00 1
1 22
k T ratio LRR Liquefied Layer Included
kinematic normalized
case displacement 0 1000 2000 3000 4000 5000 6000 7000
ultimate lateral resistance (kN/m)
Fig. 8. Comparison of the top pushover kinematic case to the rigid pile
case for a medium-diameter (1.3716 m) pile; both analyses use homo- Fig. 9. Variation in ultimate lateral resistance, pu , for homogenous and
genous soil profiles liquefied soil profiles for a medium-diameter (1.3716 m) pile

than those shown cannot be computed at all depths, thus, no respec- application of a lateral load, effectively reducing the available lat-
tive values are presented. Depth and displacement scales have been eral resistance of the unliquefied layers in the vicinity of the lique-
normalized by pile diameter. fied layer. This mechanism is explored in values of kT and pu for p-y
If the force density depended only on relative soil-pile displace- curves obtained from 3D FEA. Comparisons are made among the
ment, as assumed in the Winkler model, then both the k T ratio and values of these parameters resulting from cases that consider
LRR would be equal to 1.0. Figs. 7 and 8 show that the k T ratio and homogenous and liquefied soil profiles.
LRR deviate significantly from 1.0; hence, there are significant Figs. 9 and 10 present the pu and kT distributions, respectively,
differences in the p-y curves computed from the different pile
resulting from homogenous and liquefied cases for a medium-
kinematics. Areas in which the displacement is large show good
diameter pile. As shown, the reduction in pu in the unliquefied
agreement between the respective p-y curves. However, significant
layers is more significant than the reduction in kT . For both param-
deviations occur in the areas in which the displacements are rela-
tively small. At a particular depth, the p-y curves for the lateral eters, the reduction lessens in magnitude with increasing distance
spreading and top pushover cases vary with respect to the rigid pile from the layer interface. The spatial extent of the reduction is lim-
results on the basis of the magnitude of the pile displacement and ited, and near the top and bottom of the soil profile, the homog-
how this displacement changes during the analysis. enous and liquefied cases display nearly identical results.
Differing results for different kinematic cases are observed for The magnitude and spatial extent of the lateral resistance reduc-
all three pile diameters. The areas of greatest divergence do not tion in the unliquefied layers depends on several factors. Among
occur at the same depth for different piles; therefore, depth or mesh- these factors are the depth and thickness of the liquefied layer
ing effects can be ruled out as the causes of the difference. For all and the magnitude of the effective stress at the layer interface. Only
cases, the only difference between the compared analyses is the pile the effects of increasing the effective stress at the layer interface are
kinematics; thus, the observed differences must be attributable to examined in this work. Without altering the depth of the liquefied
this factor. This finding suggests that the kinematics of the pile layer in the FE mesh, increased effective stress is obtained by using
strongly influence the resulting p-y curves. This must be considered a series of five surcharge pressures having magnitudes established
when attempting to define a set of curves suitable for use in the on hypothetical soil fills with vertical thicknesses equal to 0, 5, 10,
analysis of a general pile deformation. The selected p-y curves
15, and 20 diameters of the medium-diameter pile. This results in
should reflect only the response of the soil, not a combination
of the soil and pile responses. The pile response effect should
0
be captured by the numerical model of choice.
In the rigid pile case, the pile does not deform. Therefore, this
case produces p-y curves which are not influenced by pile kinemat- −5
ics. For this reason, unless otherwise indicated explicitly, the rigid Liquefied Layer
pile case is used to extract p-y curves from the 3D model in the −10
(m)
depth (m)

remainder of this paper. This is the benefit of the 3D FEA in com-


depth

parison with full-scale field experiments. It will be shown that p-y −15
curves obtained from this pile kinematic produce reasonable results
when applied to lateral spreading. −20

−25
Effects of Liquefied Layer on Observed Soil Homogenous Profile
Liquefied Layer Included
Response
0 0.5 20001 30001.54000 25000 2.5
1000 6000 3
7000
ultimate lateral
initial resistance
stiffness (kPa)(kN/m) x 10
4
In a liquefaction-induced lateral spread, once the middle layer has
liquefied, its shear strength is reduced significantly and its volumet-
Fig. 10. Variation in initial stiffness, k T , for homogenous and liquefied
ric stiffness is much less affected. The unliquefied soil at the layer
soil profiles for a medium-diameter (1.3716 m) pile
interface is now able to move into the liquefied layer during the

562 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


10 0
API (1987)

distance from center of liq. layer (m)


Broms (1964)
−5
Fleming et al. (1985)
5
Brinch Hansen (1961)
−10 3D FEA 1.3716 m Pile

depth (m)
0 Liquefied Layer
−15

247 kPa Eff. stresses at −20


−5 363.6 kPa center of liq. layer
480.2 kPa
596.8 kPa −25
713.4 kPa
−10
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

0.2 0.4 0.6 0.8 1 1.2 0 1 2 3


4
p ratio ultimate lateral resistance (kN/m) x 10
u

Fig. 11. Ratio of ultimate lateral resistance pu in soil with a liquefied Fig. 13. Variation of ultimate lateral resistance, pu , with depth for a
layer to homogenous soil for increasing effective stress for a medium- medium-diameter (1.3716 m) pile
diameter (1.3716 m) pile
analysis of a lateral spreading deformation. The method recom-
10
mended by the API (1987) is selected for defining the conventional
distance from center of liq. layer (m)

p-y curves. The rigid pile kinematic shown in Fig. 2(c) is used for
the computation of p-y curves from the 3D model for both homog-
5
enous and liquefied soil profiles. Comparisons are made with re-
spect to the characteristic curve parameters of initial tangent
stiffness, kT , and ultimate lateral resistance, pu .
0 Liquefied Layer
Comparison of Ultimate Lateral Resistance
Eff. stresses at The API recommends a hyperbolic tangent function similar to
247 kPa center of liq. layer
−5 363.6 kPa
Eq. (5) to describe the shape of p-y curves for cohesionless soils.
480.2 kPa In this method, the ultimate lateral resistance, pu , is taken as the
596.8 kPa lesser of the following expressions (Reese and Van Impe 2001):
713.4 kPa
−10 
0.2 0.4 0.6 0.8 1 1.2 K o z tan ϕ sin ϕ tan β
pu ¼ Aγz
 þ ðB þ z tan β tan αÞ
k T ratio tanðβ  ϕÞ cos α tanðβ  ϕÞ

Fig. 12. Ratio of initial stiffness kT in soil with a liquefied layer to þ K o z tan βðtan ϕ sin β  tan αÞ  K a B
homogenous soil for increasing effective stress for a medium-diameter
(1.3716 m) pile ð7a Þ

surcharge pressures of 0, 116.6, 233.2, 349.8, and 466.4 kPa, re- pu ¼ AγzB½K
 a ðtan β  1Þ þ K o tan ϕtan β
8 4
ð7b Þ
spectively, for a soil unit weight of γ ¼ 17 kN=m3 . where Eq. (7a) considers a wedge failure mechanism and Eq. (7b)
To evaluate the effect of increasing the effective stress at the considers a plane strain flow-around failure condition. In Eqs. (7a)
liquefied layer interface, ratios of pu and kT in the liquefied case and (7b), ϕ = angle of internal friction, α ¼ ϕ=2; β ¼ π=4 þ α;
are taken with respect to corresponding values in the homogenous K a ¼ tan2 ðπ=4  αÞ; A  is determined from the ratio of depth, z,
case. These ratios are plotted for a medium-diameter pile in Figs. 11 to pile diameter B; and K o is a constant recommended by API
and 12 for the five considered surcharge pressures. For larger ef- (1987) to be taken as 0.4.
fective stresses, there is a reduction in the size of the liquefied The distribution of pu recommended by the API (1987) is plot-
layer’s zone of influence with respect to ultimate lateral resistance. ted alongside the computed pu distribution for the medium-size pile
Increasing the effective stress increases shear strength in the upper in Fig. 13. Several alternative and commonly used methods for pre-
and lower solid layers, which leads to a smaller plastic zone in the dicting pu in cohesionless soils (Broms 1964; Fleming et al. 1985;
unliquefied layers during the lateral spreading deformation. The Brinch Hansen 1961) are included for reference. All of the distri-
size of the liquefied layer’s zone of influence is reduced accord- butions show an increase in pu with depth; however, the shape of
ingly. The zone of influence differs only slightly with respect to the computed distribution does not directly correlate to any method
initial stiffness as effective stress increases. For the cases analyzed, found in the literature. At shallow depths, all methods show a fair
it appears that the zone of influence on the initial stiffness of the amount of agreement, however, with increasing depth there is sig-
unliquefied layers is unchanged by increasing effective stress. nificant variation in the predicted ultimate resistance. At depth, the
computed values of pu have the same approximate magnitude as the
smaller estimated distributions (Broms 1964 ; Fleming et al. 1985)
Comparison and Evaluation of Computed and whereas the ultimate resistance recommended by the API (1987) is
Conventional p-y Curves significantly higher than any other method.
Only the results for the medium-diameter pile are presented
Soil response curves computed from the 3D FE models are used here, but on the basis of data from the three analyzed piles, it is
to investigate the applicability of conventional p-y curves to the seen that the length over which the computed results are similar

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011 / 563

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


0
2.5 m Pile distribution is not the linear variation recommended by the API
1.3716 m Pile (1987). The computed initial stiffness appears to have an essentially
−2 0.6096 m Pile bilinear distribution with depth, where one slope corresponds to
shallow depths and the other to deeper locations. A curved transi-
tion zone links the two linear sections.
depth (m)

−4
At shallow depths the computed initial stiffness distributions
have linear slopes which fall within the range of kpy values recom-
−6 mended by the API (1987) for loose to medium dense sands. These
slopes are shown as the dashed lines in Fig. 15. This behavior is
likely that which was captured by the original experiments of Reese
−8
et al. (1974) and is reflected in API (1987). However, it is noted that
there is a diameter dependence on these slopes that is not inherently
captured by the API recommendations. These results also suggest
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

−10
0 2000 4000 6000 8000
ultimate lateral resistance (kN/m)
that extrapolation of this shallow slope to deeper depths signifi-
cantly overestimates the initial stiffness of the soil beyond the
Fig. 14. Variation of computed ultimate lateral resistance, pu , over the near-surface zone. The observed slope at deeper locations is less
first 10 m below ground surface for three pile diameters than the lowest value of k py recommended by the API (1987).
The form of the computed distribution of initial stiffness correlates
well with findings by Lam and Cheang (1995), who showed that a
to the estimated results is approximately two pile diameters. The parabolic variation of initial stiffness, when applied to a p-y curve
wedge failure mechanism assumed in the API formulation at shal- analysis, produces pile moment distributions at depths that match
low depths is observed in the 3D FEA. The vertical extents of the experimental results more closely than curves defined with a linear
observed failure zones do not correlate particularly well with those stiffness variation.
computed by using Eq. (7a), however, the presence of the wedge- A case using a linear elastic soil constitutive model is analyzed
type failure in the 3D analysis further verifies the observation that to further assess the variation in initial stiffness observed in the
p-y curves obtained from the top pushover kinematic provide a computed p-y curves. The linear elastic model uses the elastic
good representation of the soil response at shallow depths. parameters summarized in Table 1. Fig. 15 shows the distribution
As expected, the magnitude of pu depends on the pile diameter of k T from this elastic soil analysis. All three piles produce identical
in the computed results. Fig. 14 shows the variation in pu over the results for elastic soil. In the elastoplastic cases, the initial stiffness
first 10 m below the surface for the three pile diameters. A larger near the surface is less than that in the elastic cases. In this region,
zone of soil is affected by a larger pile and consequently more re- lack of overburden pressure leads to the early failure of the soil and
sistance is provided prior to yielding of this zone. This effect be- a corresponding reduction in the initial stiffness. This near-surface
comes more pronounced with increasing depth. As depth and behavior shows a diameter dependence. The larger-diameter piles
confining pressure increase, the shear strength of the soil also in- cause a larger extent of soil to yield sooner in the deformation, lead-
creases, leading to a larger ultimate resistance. ing to a reduced initial stiffness. Below this region, the value of kT
observed in the elastoplastic model matches that in the elastic
Comparison of Initial Stiffness model. Here, sufficient overburden pressure exists such that the in-
In the method recommended by the API (1987), the initial tangent itial response of the soil is primarily elastic, thus, kT is equal in the
stiffness is defined to be the product of the coefficient of subgrade elastoplastic and elastic cases. This value is not dependent on pile
reaction, kpy (units of FL3 ), and the depth, z. The coefficient of diameter for these circular piles.
subgrade reaction is taken as constant for a given angle of internal
friction, ϕ, and is determined from an empirical correlation. Fig. 15 Comparison of p-y Curves
shows the variation of initial stiffness, kT , with depth for the com- Fig. 16 shows representative computed p-y data and fitted hyper-
puted p-y curves for the three analyzed piles. The computed bolic tangent curves obtained for a medium-diameter pile. API
(1987) recommended p-y curves are shown for comparison. As
0 is expected based upon the differences in kT and pu between the
Elastic Soil computed and API curves, the two sets of curves are comparable
at shallow depths while differing substantially at greater depths. For
−2
analysis of top-loaded piles, these differences are relatively incon-
sequential, however, for analysis of a lateral spreading deformation
−4 where large pile deformation can occur at depth, the problems cre-
depth (m)

ated by these differences become apparent.


Elastoplastic Soil
−6 Comparison of Pile-Bending Demands through 1D
Elastic Soil (3 diameters) BNWF Analysis
−8 2.5 m Pile
A final evaluation of the computed and conventional p-y curves in
1.3716 m Pile
the context of the lateral spreading problem is conducted using a
0.6096 m Pile
−10 series of 1D BNWF models. These models represent the pile with
0 1 2 3 4 the same beam-column elements used in the 3D model. Linear elas-
4
initial stiffness (kPa) x 10 tic and elastoplastic fiber models are used to represent the pile cross
sections. In lieu of a soil continuum, the response of the soil is mod-
Fig. 15. Variation of computed initial stiffness, k T , over the first 10 m
eled using nonlinear springs attached to each pile node. There are
below ground surface for three pile diameters using both elastoplastic
two series of 1D models. In the first, the springs have behavior
and elastic soil formulations
defined using the p-y curves computed from the 3D simulations.

564 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


2000
2000
1.0 m below surface
2000
2.4006
2.4 m belowsurface
m below surface ∆

force density p (kN/m)


OpenSees data

force density p (kN/m)


Fitted tanh curve
1500 1500
API (1987)
1000
1000 1000
force density p (kN/m)

Imposed
500 500 Equal DOF displacement
profile
000 0.1 0.2
0
0.3 0 01 02 03
9.9 m below
m belowsurface
displacement
9.9436 ysurface
(m) 14.7 m below
14.7446 m belowsurface
surface No springs in
12000
12000 12000 liquefied layer
force density p (kN/m)

force density p (kN/m)


10000 10000

8000
8000 8000 Pile node
Fixed
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

6000 6000

4000
4000 4000

2000 2000

000 0
0 0.1
0.1 0.2
0.2
displacement y (m)
0.3 0
0.3 0 0.1
0.1 0.2
0.2
displacement y (m)
0.3
0.3 Fig. 17. Schematic representation of 1D beam-spring model
displacement y (m)

Fig. 16. Comparison of computed and API-recommended p-y curves similar. This validates that the p-y curves computed using the rigid
for a medium-diameter (1.3716 m) pile pile case in 3D, which are free from the influence of pile kinemat-
ics, can be used successfully in a simplified lateral spreading analy-
sis which requires far less computational effort than 3D FEA. For
In the second, the springs are defined using curves established us- the elastic pile models, the 3D analysis took approximately 2.6e3
ing the API (1987) recommendations with some modifications to times longer than the corresponding 1D analysis; and for the non-
the initial stiffness and ultimate lateral resistance. The results of linear pile models, the 3D case took approximately 55 times longer.
these 1D simulations are compared with one another and to 3D The API-based curves, which are inherently tied to a top push-
FEA of lateral spreading with respect to the bending response over type of pile deformation, produce pile shear and moment de-
of the piles. mands which are much larger than the other analysis models, even
The computed p-y curves are obtained using a 3D rigid pile kin- with the modifications made to their initial stiffness and ultimate
ematic analysis with a liquefied layer in the center of the soil pro- resistance. The deflected shapes shown in Fig. 18 emphasize the
file. The values of pu in these curves are therefore reduced to effects of the overestimated stiffness of the API-based curves.
account for the presence of the weaker liquefied layer. The ultimate As shown, the higher stiffness of the API curves causes the base
lateral resistance of the API curves is defined per Eqs. (7a) and (7b) of the pile to be held essentially in place during the lateral spreading
by using material parameters corresponding to those summarized in deformation, generating an increased bending moment demand as
Table 1. Because the API curves have been shown to be too stiff compared with the other analyses in which the pile is able to push
with depth, the approximate lowest value of k py recommended by back into the soil.
the API (1987), k py ¼ 5;000 kN=m3 , is used to define the initial When the three analysis models are applied to elastoplastic pile
stiffness of the API p-y curves used in the BNWF analyses. This cross sections, another observation about their relative differences
is done to increase the applicability of the API curves to the lateral can be made. Fig. 19 shows the evolution of the extreme pile mo-
spreading case. Similarly, the ultimate resistance of the API curves ment located in both the top and bottom unliquefied soil layers re-
is reduced using the pu ratio obtained through comparison of 3D sulting from each analysis type. As a result of the higher stiffness at
analyses with liquefied and homogenous soil profiles. This modi- depth of the API-based curves, significantly larger curvature de-
fication, although not included in the API recommendations, cre- mands are placed upon the pile and large curvatures develop much
ates a more realistic distribution of pu with depth for the considered earlier in the lateral spreading deformation. The API-based analysis
soil profile, and also enhances the applicability of the API curves to indicates that the pile has reached its moment capacity when the
a lateral spreading analysis. upper layer has displaced approximately 0.24 m (≈0:18 diameters)
To simulate a lateral spreading event, an imposed displacement relative to the lower layer. In the computed p-y curve and 3D analy-
profile, matching that used in the 3D lateral spreading analysis, is ses, the moment capacity is reached at about 0.48 m (≈0:35 diam-
applied to the soil end of the nonlinear springs in a similar manner eters) of upper-layer displacement. This is a significant difference
to the BNWF analyses conducted by Brandenberg et al. (2007). A in the perceived pile-bending demand and pile ductility between the
schematic illustrating the 1D BNWF model is provided in Fig. 17. two cases, even with the assumed lateral resistance reduction in-
The soil stiffness is assumed to be negligible in the liquefied layer cluded in the API curves near the liquefied layer.
and thus is set to zero. It is recommended that p-y curves defined by using API (1987)
Fig. 18 presents the deflected shapes of the piles along with the or similar methods not be used in the analysis of piles for lateral
shear and moment diagrams for the 1D BNWF analyses and a 3D spreading. The 1D BNWF analyses have shown that the computed
FEA of the lateral spreading case, for the medium pile. The moment p-y curves produce similar results to the 3D model, and that there
and shear diagrams are computed at the end of the lateral soil de- are significant differences between the results obtained by using the
formation, and elastic pile sections are used to emphasize the rel- computed and API-based p-y curves, even with modifications to the
ative differences between the three analysis models. These results API curves that serve to increase their applicability to the lateral
are typical of all three piles. Because the BNWF analysis is a de- spreading case. The BNWF assumptions can still be used success-
coupled approximation of the 3D soil-pile interaction, it is not ex- fully in a simplified analysis of this problem; however, appropriate
pected that the computed p-y curves should reproduce the 3D considerations must be made with respect to the differences in the
results exactly; however, as shown in Fig. 18, the results are quite pile kinematics used to define the p-y curves and the pile kinematics

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011 / 565

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


Height above base (m)
Height above base (m)
Computed
25
API
20 3D FEA

15
10
Computed Computed
5 API API
3D FEA 3D FEA
0
0 0.5 1 −20000 −10000 0 −4 −2 0 2
4
pile deflection (m) Shear force V (kN) Moment M (kN) x 10
(a) (b) (c)

Fig. 18. Deflected shape, shear force, and moment diagrams for 1D BNWF and 3D FE analyses of lateral spreading with elastic pile model for a
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

medium-diameter (1.3716 m) pile

6000
one pile kinematic are not necessarily applicable to another type of
API pile deformation. It is recommended that an analysis should not
5000
make use of p-y curves that are influenced by pile kinematics that
differ greatly from those of the analysis.
3D FEA
moment M (kNm)

The 3D FE analyses conducted by using liquefied soil profiles


4000
demonstrate the effects of a weak middle layer on the surrounding
solid layers. The ultimate lateral resistance, pu , of the soil adjacent
3000
to the liquefied layer is reduced by a factor of up to one-half for
Computed corresponding depths within a completely homogenous layer. The
2000
extents of the zone of strength reduction decrease with increasing
effective stress, although the magnitude of the reduction does not
1000 abs(max M) in top layer display a tangible decrease. Compared with the observed reduction
abs(max M) in bottom layer in pu , the initial tangent stiffness, k T , of the soil sees a negligible
0.1 0.2 0.3 0.4 0.5 0.6 reduction.
soil displacement (m) The API (1987) recommended p-y curves tend to have values of
initial stiffness and ultimate lateral resistance that are significantly
Fig. 19. Evolution of maximum moment demand in top and bottom overestimated at depth. For the lateral spreading case, where large
layers with increasing displacement of upper soil layer during 1D and pile displacements can occur at depths beyond the first few meters
3D lateral spreading analyses for a medium-diameter (1.3716 m) pile below the ground surface, use of these p-y curves in a 1D BNWF
analysis results in very large pile-bending demands. These de-
mands greatly exceed those obtained using p-y curves computed
modeled by the analysis. A careful approach to defining the initial
from 3D FE models and the results of 3D lateral spreading analy-
stiffness and ultimate resistance with depth and in the zone of in-
ses. This suggests that the application of the API curves to the de-
fluence of the liquefied layer greatly increases the similarity of the
sign of a pile for the lateral spreading case could result in an
results as compared with 3D FEA of lateral spreading.
unnecessarily large and expensive solution. Use of these curves
without modification for analysis of lateral spreading, or another
load case with deep-seated soil-pile interaction, is not recom-
Summary and Conclusions
mended.
3D FE analyses have been used to evaluate the problem of a single
pile subjected to lateral spreading through the extraction of
representative soil-response (p-y) curves. Both linear elastic and
Acknowledgments
elastoplastic pile and soil behavior was considered. Three pile kin- This work was supported by the Pacific Earthquake Engineering
ematic cases were compared to evaluate the effects of differing pile Research (PEER) Center under Subagreement No. 00006405,
deformation on the computed curves. The characteristic p-y curve which is gratefully acknowledged. The writers are also grateful
parameters of initial stiffness, k T , and ultimate lateral resistance, pu , for the support and ideas provided by Ignatius Po Lam.
were used for all comparisons. The curves obtained from 3D mod-
els with a homogenous soil profile and with a layered soil profile
with a liquefied layer at the center were compared to assess the References
effects of the liquefied layer on the surrounding dense layers.
The computed p-y curves were compared with a set of convention- American Petroleum Institute (API). (1987). “Recommended practice
ally defined p-y curves (API 1987) to evaluate the applicability of for planning, designing and constructing fixed offshore
the conventional curves to the lateral spreading load case. platforms.” API Recommended Practice 2A (RP-2A), 17th Ed.,
Pile kinematics plays an important role in defining the p-y Washington, DC.
Brandenberg, S. J., Boulanger, R. W., Kutter, B. L., and Chang, D. (2007).
curves computed from the 3D FE model. It is observed that there
“Static pushover analyses of pile groups in liquefied and laterally
is a tangible effect on the computed p-y curves which can only be spreading ground in centrifuge tests.” J. Geotech. Geoenviron. Eng.,
attributed to the kinematics of the piles. The generated curves differ 133(9), 1055–1066.
in quality and consistency on the basis of the magnitude of the dis- Brinch Hansen, J. (1961). “The ultimate resistance of rigid piles against
placement at their respective nodes and how these displacements transversal forces.” Bulletin No. 12, Geoteknisk Institute, Copenhagen,
evolve over the duration of the deformation. Curves obtained from 5–9.

566 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011

J. Geotech. Geoenviron. Eng. 2011.137:557-567.


Broms, B. B. (1964). “Lateral resistance of piles in cohesionless soils.” S. L. Kramer, and R. Siddharthan, eds., ASCE, New York, 110–135.
J. Soil Mech. Found. Div., 90(SM3), 123–156. Matlock, H., and Reese, L. C. (1960). “Generalized solutions for laterally
Brown, D. A., and Shie, C. F. (1990). “Three dimensional finite element loaded piles.” J. Soil Mech. Found. Div., 86(SM5), 63–91.
model of laterally loaded piles.” Comput. Geotech., 10(1), 59–79. McClelland, B., and Focht, J. (1958). “Soil modulus for laterally loaded
Chen, W. F., and Saleeb, A. F. (1994). Constitutive equations for engineer- piles.” Transactions, ASCE, 123, 1049–1086.
ing materials. Volume I: Elasticity and modeling, Elsevier Science B.V., OpenSees [Computer software]. “Open System for Earthquake Engineering
Amsterdam. Simulation.” Pacific Earthquake Engineering Research Center, Univ. of
Finn, W. D. L., and Fujita, N. (2002). “Piles in liquefiable soils: Seismic California, Berkeley, CA. 〈http://opensees.berkeley.edu〉 (May 2011).
analysis and design issues.” Soil Dyn. Earthquake Eng., 22(9-12), Petek, K. A. (2006). “Development and application of mixed beam–solid
731–742. models for analysis of soil-pile interaction problems.” Ph.D. thesis,
Fleming, W. G. K., Weltman, A. J., Randolph, M. F., and Elson, W. K. Univ. of Washington.
(1985). Piling engineering, Surrey University Press, London.
Reese, L. C., Cox, W. R., and Koop, F. D. (1974). “Analysis of laterally
Georgiadis, M. (1983). “Development of p-y curves for layered soils.”
loaded piles in sand.” Proc., 6th Offshore Technology Conf., Vol. 2,
Geotechnical practice in offshore engineering, ASCE, Reston, VA,
Downloaded from ascelibrary.org by University of Western Ontario on 09/29/14. Copyright ASCE. For personal use only; all rights reserved.

Houston. 473–483.
536–545.
International Center for Numerical Methods in Engineering (CIMNE). Reese, L. C., and Van Impe, W. F. (2001). “Single piles and pile groups
(2008). GiD User Manual, 〈http://gid.cimne.upc.es/〉 (May 2011). under lateral loading.” Balkema, Rotterdam, Netherlands.
Kulhawy, F. H., and Mayne, P. W. (1990). “Manual on estimating soil prop- Wang, S. T., and Reese, L. C. (1998). “Design of pile foundations in lique-
erties for foundation design.” EPRI EL-6800, Project 1493-6 Final fied soils.” Geotechnical Earthquake Engineering and Soil Dynamics
Rep., Electrical Power Research Institute. III (GSP 75): Proc., Specialty Conf., P. Dakoulas, M. K. Yegian,
Lam, I. P., Arduino, P., and Mackenzie-Helnwein, P. (2009). “Opensees and R. D. Holz, eds., 1331–1343, Seattle, WA.
soil-pile interaction study under lateral spread loading.” International Wriggers, P. (2002). Computational contact mechanics, Wiley, West
Foundation Congress and Equipment Expo, Orlando, FL. Sussex, UK.
Lam, I. P., and Cheang, L. (1995). “Dynamic soil-pile interaction behavior Yang, Z., and Jeremic, B. (2005). “Study of soil layering effects on lateral
in submerged sands.” Earthquake-induced movements and seismic loading behavior of piles.” J. Geotech. Geoenviron. Eng., 131(6),
remediation of existing foundations and abutments (GSP 55), 762–770.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JUNE 2011 / 567

J. Geotech. Geoenviron. Eng. 2011.137:557-567.

You might also like