t00 058

Color profile: Generic CMYK printer profile
Composite Default screen
1166
Dynamic analysis for laterally loaded piles and

dynamic py curves
M. Hesham El Naggar and Kevin J. Bentley
Abstract: Pile foundations are often subjected to lateral dynamic loading due to forces on the supported structure. In
this study, a simple two-dimensional analysis was developed to accurately model the pile response to dynamic loads.
The proposed model incorporates the static py curve approach (where p is the static soil reaction and y is the pile deflection) and the plane strain assumptions to represent the soil reactions within the frame of a Winkler model. The py
curves are used to relate pile deflections to the nonlinear soil reactions. Wave propagation and energy dissipation are
also accounted for along with discontinuity conditions at the pilesoil interface. The inclusion of damping with the
static unit transfer curves results in increased soil resistance, thus producing dynamic py curves. The dynamic py
curves are a function of the static py curve and velocity of the soil particles at a given depth and frequency of loading. The proposed model was used to analyze the pile response to the lateral Statnamic load test, and the predicted response compared well with the measured response. Closed-form solutions for dynamic py curves were established by
curve fitting the dynamic soil reactions for a range of soil types and loading frequencies. These solutions can be used
to model soil reactions for pile vibration problems in readily available finite element analysis (FEA) and dynamic
structural analysis packages. A simple spring and dashpot model was also proposed to be used in equivalent linear
analyses of transient pile response. The proposed models were incorporated into an FEA program (ANSYS) which was
used to compute the response of a laterally loaded pile. The computed responses compared well with the predictions of
the two-dimensional analysis.
Key words: dynamic, transient, lateral, piles, py curves, inertial interaction.
Rsum : Les fondations sur pieux sont souvent soumises un chargement dynamique latral d des forces
appliques sur la structure porte. Dans cette tude, une analyse unidimensionnelle simple a t dveloppe pour
modliser avec prcision la raction du pieu aux charges dynamiques. Le modle propos incorpore lapproche de la
courbe statique py et les hypothses de dformation plane pour reprsenter les ractions du sol dans le cadre dun
modle de Winkler. Les courbes py sont utilises pour faire une corrlation entre les dflexions du pieu et les
ractions non linaires du sol. La propagation dondes et la dissipation dnergie sont galement prises en compte en
mme temps que les conditions de discontinuit linterface pieu-sol. Linclusion de lamortissement avec les courbes
de transfert dunit statique rsulte en un accroissement de la rsistance du sol, produisant ainsi des courbes py
dynamiques . Les courbes py dynamiques sont une fonction de la courbe statique py et de la vlocit des particules
de sol une profondeur et une frquence de chargement donnes. Le modle propos a t utilis pour analyser la
rponse du pieu lessai de chargement statnamique, et la rponse prdite se comparat bien la rponse mesure. Les
solutions exactes pour les courbes py dynamiques ont t tablies par lissage des courbes des ractions dynamiques
du sol pour une plage de types de sol et de frquences de chargement. Ces solutions peuvent tre utilises pour
modliser les ractions du sol pour les problmes de vibration de pieux dans les progiciels FEA et danalyse
structurale dynamique couramment disponibles. Un modle comprenant un simple ressort avec pot amortisseur a aussi
t propos pour tre utilis dans des analyses linaires quivalentes de la rponse transitoire dun pieu. Les modles
proposs ont t incorpors dans un programme FEA (ANSYS) qui a t utilis pour valuer la rponse dun pieu
charg latralement. Les rponses calcules se comparaent bien avec les prdictions de lanalyse bidimensionnelle.
Mots cls : dynamique, transitoire, latral, pieux, courbes py, interaction inertielle.
[Traduit par la Rdaction]
El Naggar and Bentley
Introduction
Pile foundations are often subjected to lateral loading due
to forces on the supported structure. The horizontal loads at
the pile head can be the governing design constraint for sinReceived May 5, 1999. Accepted April 28, 2000. Published
on the NRC Research Press website on December 5, 2000.
M.H. El Naggar and K.J. Bentley. Geotechnical Research
Centre, The University of Western Ontario, London, ON
N6A 5B9, Canada.
Can. Geotech. J. 37: 11661183 (2000)
I:\cgj\Cgj37\Cgj06\T00-058.vp
Thursday, November 30, 2000 11:48:21 AM
1183
gle piles and pile groups supporting different types of structures in many situations including environmental loading
(wind, water, and earthquakes) and machine loading on
structures such as buildings, bridges, and offshore platforms.
Most building and bridge codes use factored static loads
to account for the dynamic effects of pile foundations. Although very low frequency vibrations may be accurately
modeled using factored loads, the introduction of nonlinearity, damping, and pilesoil interaction during transient
loading may significantly alter the response. The typical frequency ranges of interest are 010 Hz for earthquakes,
2000 NRC Canada

01 Hz for offshore environmental loading, and 5200 Hz

for machine foundations. The emphasis in the current study
is on the seismic loading case.
Novak et al. (1978) developed a frequency-dependent
pilesoil interaction model; however, it assumes strictly linear or equivalent linear soil properties. Gazetas and Dobry
(1984) introduced a simplified linear method to predict
fixed-head pile response accounting for both material and radiation damping and using available static stiffness (derived
from a finite element or any other accepted method). This
method is not suitable for the seismic response analysis because of the linearity assumptions. In general, there is much
controversy over advanced linear solutions (frequency domain), as they do not account for permanent deformation or
gapping at the pilesoil interface.
Nogami et al. (1992) developed a time-domain analysis
method for single piles and pile groups by integrating plane
strain solutions with a nonlinear zone around each pile using
py curves, (where p is the static soil reaction and y is the
pile deflection). El Naggar and Novak (1995, 1996) also developed a computationally efficient model for evaluating the
lateral response of piles and pile groups based on the
Winkler hypothesis, accounting for nonlinearity using a hyperbolic stressstrain relationship and slippage and gapping
at the pilesoil interface. The model also accounts for the
propagation of waves away from the pile and energy dissipation through both material and geometric damping.
The py curves (unit load transfer curves) approach is a
widely accepted method for predicting pile response under
static loads because of its simplicity and practical accuracy.
In the present study, the model proposed by El Naggar and
Novak (1996) is modified to utilize existing or developed cyclic or static py curves to represent the nonlinear behaviour
of the soil adjacent to the pile. The model uses unit load
transfer curves in the time domain to model nonlinearity and
incorporates both material and radiation damping to generate
dynamic py curves.
1167
Fig. 1. Element representation of the proposed model. m1, m2
represent the mass of the inner field lumped at two nodes; half
(m1) at the node adjacent to the pile, and the other half (m2) at
the node adjacent to the outer field.
the radiation damping in the soil medium. The soil reactions

and the pilesoil interface conditions are modeled separately
on both sides of the pile to account for slippage, gapping
and state of stress as the load direction changes.
Inner field element
The inner field is modeled with a nonlinear spring to represent the stiffness and a dashpot to simulate material
(hysteretic) damping. The stiffness is calculated assuming
plane strain conditions, the inner field is a homogeneous isotropic viscoelastic massless medium, the pile is rigid and circular, there is no separation at the pilesoil interface, and
displacements are small. Novak and Sheta (1980) obtained
the stiffness kNL under these conditions as
[1]
Model description
Pile model
The pile is assumed to be vertical and flexible with circular cross section. Noncylindrical piles are represented by cylindrical piles with equivalent radius to accommodate any
pierpile configurations. The pile and surrounding soil are
subdivided into n segments, with pile nodes corresponding
to soil nodes at the same elevation. The standard bending
stiffness matrix of beam elements models the structural stiffness matrix for each pile element. The pile global stiffness
matrix is then assembled from the element stiffness matrices
and is condensed to give horizontal translations at each layer
and the rotational degree of freedom at the pile head.
Soil model: hyperbolic stressstrain relationship
The soil is divided into n layers with different soil properties assigned to each layer according to the soil profile considered. Within each layer, the soil medium is divided into
two annular regions as shown in Fig. 1. The first region is an
inner zone adjacent to the pile and accounts for the soil
nonlinearity. The second region is the outer zone that allows
for wave propagation away from the pile and provides for
kNL
r
8 Gm (1 v)(3 4v) o + 1
r1
=
2
2

r o + (3 4v) 2 r o + 1 ln r 1 1
r
r1
r o
1
where ro is the pile radius, r1 is the outer radius of the inner

zone, and is Poissons ratio of the soil stratum. The ratio
r1/ro depends on the extent of nonlinearity, which depends
on the level of loading, and on the size of the pile. A parametric study showed that a ratio of 1.12.0 yielded good
agreement between the stiffness of a composite medium (inner zone and outer zone) and that of a homogeneous medium (no inner zone) under small strain (linear) conditions.
Gm is the modified shear modulus of the soil and is approximated, according to the strain level, by a hyperbolic law as
[2]
1
G m = Gmax
1 +
where Gmax is the maximum shear modulus (small strain

modulus) of the soil according to laboratory or field tests. In
the absence of actual measurements, maximum shear modulus for any soil layer can be calculated in this model by
(Hardin and Black 1968)
2000 NRC Canada

1168
Can. Geotech. J. Vol. 37, 2000
Fig. 2. Envelope of variations of horizontal stiffness and damping stiffness parameters between = 0.25 and = 0.4 (after Novak et
al. 1978).
[3]
Gmax =
3230(2.97 e) 2 0.5
o
1+ e
where e is the void ratio, and o is the mean principal effective stress (in kN/m2) at the soil layer. The parameter =
P/Pu is the ratio of the horizontal soil reaction in the soil
spring, P, to the ultimate resistance of the soil element, Pu.
The ultimate resistance of the soil element is calculated using standard relations given by the American Petroleum Institute (1991). For clay, the ultimate resistance is given as a
force per unit length of soil by
[4]
Pu = 3cud + Xd + JcuX
[5]
Pu = 9cud
where Pu is the minimum of the resistances calculated by

eqs. [4] and [5], cu is the undrained shear strength, d is the
diameter of the pile, is the effective unit weight of the soil,
X is the depth below the surface, and J is an empirical coefficient dependent on the shear strength. A value of J = 0.5
was used for soft clays (Matlock 1970) and J = 1.5 for stiff
clays (Bhushan et al. 1979).
The corresponding criteria for the ultimate lateral resistance of sands at shallow depths Pu1 or at large depths Pu2
are as follows (American Petroleum Institute 1991):
[6]
tan
K X tan sin
Pu1 = A X 0
+
tan( )
tan( ) cos
[7]
Far-field element
The outer field is modeled with a linear spring in parallel
with a dashpot to represent the linear stiffness and damping
(mainly radiation damping). The outer zone allows for the
propagation of waves to infinity. Novak et al. (1978) developed explicit solutions for the soil reactions expressed in
terms of complex stiffness, K, of a unit length of a cylinder
embedded in a linear viscoelastic medium given by
[8]
Pu 2 = A Xd [ Ka (tan8 1) + K0 tan tan4 ]
where A is an empirical adjustment factor dependent on the

depth from the soil surface, K0 is the earth pressure coefficient at rest, is the effective friction angle of the sand, =
/2 + 45, = /2, and Ka is the Rankine minimum active
earth pressure coefficient defined as Ka = tan2(45 /2).
K = Gmax[Su1(ao, , D) + iSu2(ao, , D)]
where ao = r1/Vs is the dimensionless frequency, is the

frequency of loading, Vs is the shear wave velocity of the
soil layer, i is the imaginary unit (= 1 ), and D is the material damping constant of the soil layer. Figure 2 shows the
general variations of Su1 and Su2 with Poissons ratio and material damping. Rewriting eq. [8], the complex stiffness, K,
can be represented by a spring coefficient, kL, and a damping
coefficient, cL, as
[9]
(d + X tan tan ) + K0 X tan (tan sin
tan ) Ka d
In the derivation of eq. [1], the inner field was assumed to

be massless (Novak and Sheta 1980). Therefore, the mass of
the inner field is lumped equally at two nodes on each side
of the pile: node 1 adjacent to the pile, and node 2 adjacent
to the outer field, as shown in Fig. 1.
K = kL + iaocL
Figure 2 shows that for the dimensionless frequency range

between 0.05 and 1.5, Su1 maintains a constant value and Su2
increases linearly with an increase in ao. The majority of dynamic loading on foundations falls within this frequency
range, including destructive earthquake loading. Therefore,
for the purpose of the time-domain analysis, the spring and
dashpot constants, Su1 and Su2, respectively, are considered
frequency independent and depend only on Poissons ratio
and are given as follows:
[10]
kL = GmaxSu1()
[11]
cL =
2Gmaxr1
Vs
Su2 (a o = 0.5, v)
2000 NRC Canada


Fig. 3. Soil model: (a) composite medium and py curve, and
(b) inclusion of damping.
1169
[12]
1
1
1
=
+
k py
kL
k NL
The stiffness of the nonlinear strength is then calculated

as
[13]
Pilesoil interface
The pilesoil interface is modeled separately on each side
of the pile, thus allowing gapping and slippage to occur on
each side independently. The soil and pile nodes in each
layer are connected using a no-tension spring, that is, the
pile and soil will remain connected and will have equal displacement for compressive stresses. The spring is disconnected if tensile stress is detected in the soil spring to allow
a gap to develop. This separation or gapping results in permanent displacement of the soil node dependent on the magnitude of the load. The development of such gaps is often
observed in experiments, during offshore loading, and after
earthquake excitation in clays. These gaps eventually fill in
again over time until the next episode of lateral dynamic
loading. The pilesoil interface for sands does not allow for
gap formation, but instead the sand caves in, resulting in the
virtual backfilling of sand particles around the pile during
repeated dynamic loading. When the pile is unloaded, the
sand on the tension side of the pile follows the pile with zero
stiffness instead of remaining permanently displaced as in
the clay model. In the unloading phase, the stiffness of the
inner field spring is assumed to be linear in both the clay
and sand models.
Soil model: py curve approach
The soil reaction to transient loading comprises stiffness
and damping. The stiffness is established using the py
curve approach and the damping is established from analytical solutions that account for wave propagation. A similar
approach was suggested by Nogami et al. (1992) using py
curves.
Based on model tests, py curves relate pile deflections to
the corresponding soil reaction at any depth (element) below
the ground surface. The py curve represents the total soil
reaction to the pile motion (i.e., the reactions of the inner
and outer zones combined). The total stiffness, kpy, derived
from the py curve is equivalent to the true stiffness (real
part of the complex stiffness) of the soil medium. Thus, referring to the hyperbolic law model, the combined inner
zone stiffness (kNL) and outer zone stiffness (kL) can be replaced by a unified equivalent stiffness zone (kpy) as shown
in Fig. 3a. Hence, to ensure that the true stiffness is the
same for the two soil models, the flexibility of the two models is equated, i.e.,
k NL =
k py k L
k L k py
The constant of the linear elastic spring, kL, is established

from the plane strain solution (i.e., eq. [10]). The static soil
stiffness, kpy represents the relationship between the static
soil reaction, p, and the pile deflection, y, for a given py
curve at a specific load level. The py curves are established
using empirical equations (Matlock 1970; Reese and Welch
1975; Reese et al. 1975) or curve fit to measured strain data
using an accepted method such as the modified RambergOsgood model (Desai and Wu 1976). In the present study,
internally generated static py curves are established based
on commonly used empirical correlations for a range of soil
types.
Damping
The damping (imaginary part of the complex stiffness) is
incorporated into both the py approach and the hyperbolic
model to allow for energy dissipation throughout the soil.
The nonlinearity in the vicinity of the pile, however, drastically reduces the geometric damping in the inner field.
Therefore, both material and geometric (radiation) damping
are modeled in the outer field. A dashpot is connected in
parallel to the far-field spring, and its constant is derived
from eq. [11]. If the material damping in the inner zone is to
be considered, a parallel dashpot with a constant cNL, coefficient of material damping in the inner zone, to be suitably
chosen may be added as shown in Fig. 3b. The addition of
the damping resistance to static resistance represented by the
static unit load transfer (the py curve) tends to increase the
total resistance as shown in Fig. 4.
Static py curve generation for clay
The general procedure for computing py curves in clays
both above and below the groundwater table and corresponding parameters are recommended by Matlock (1970) and
Bhushan et al. (1979), respectively. The py relationship was
based on the following equation:
[14]
y
p
= 0.5
Pu
y 50
where p is the soil resistance, y is the deflection corresponding to p, Pu is the ultimate soil resistance from eqs. [4] and
[5], n is a constant relating soil resistance to pierpile deflection, and y50 is the corrected deflection at one-half the
ultimate soil reaction determined from laboratory tests. The
tangent stiffness constant, kpy, of any soil element at time
step t + t is given by the slope of the tangent to the py
curve at the specific load level as shown in Fig. 4. This slope
is established from the soil deflections at time steps t and t
t and the corresponding soil reactions calculated from
eq. [14], i.e.,
2000 NRC Canada

1170
Fig. 4. Determination of stiffness (kpy) from an internally generated static py curve to produce a dynamic py curve (including damping).
[15]
k py ( t + t) =
p t p t t
yt yt t
Therefore, eqs. [10] and [15] can be substituted into

eq. [13] to obtain the nonlinear stiffness representing the inner field element in the analysis. Thus, the linear and nonlinear qualities of the unit load transfer curves have been
logically incorporated into the outer and inner zones, respectively.
Static py curve generation for sand
Several methods have been used to experimentally obtain
py curves for sandy soils. Abendroth and Greimann (1990)
performed 11 scaled pile tests and used a modified
Ramberg-Osgood model to approximate the nonlinear soil
resistance and displacement behaviour for loose and dense
sand. The most commonly used criteria for development of
py curves for sand were proposed by Reese et al. (1974)
but tend to give very conservative results. Bhushan et al.
(1981) and Bhushan and Askari (1984) used a different procedure based on full-scale load test results to obtain nonlinear py curves for saturated and unsaturated sand. A stepby-step procedure for developing py curves in sands
(Bhushan and Haley 1980; Bhushan et al. 1981) was used to
estimate the static unit load transfer curves for different
sands below and above the water table. The procedure used
to generate py curves for sand differs from that suggested
for clays. The secant modulus approach is used to approximate soil reactions at specified lateral displacements. The
soil resistance in the static py curve model can be calculated using the following equation:
[16]
p = (k)(x)(y)(F1)(F2)
where k is a constant that depends on the lateral deflection y

(i.e., k decreases as y increases) and relates the secant modulus of soil for a given value of y to depth (Es = kx); x is the
depth at which the py curve is being generated; and F1 and
F2 are density and groundwater (saturated or unsaturated)
factors, respectively, and can be determined from Meyer and
Reese (1979). The main factors affecting k are the relative
density of the sand (loose or dense) and the level of lateral
displacement. The secant modulus decreases with increasing
displacement and thus the nonlinearity of the sand can be

modeled accurately. This analysis assumes a linear increase
of the soil modulus with an increase in depth (but varies
nonlinearly with displacement at each depth) which is typical for many sands.
Equation [16] was used to establish the py curve at a
given depth. The tangent stiffness kpy (needed in the timedomain analysis), which represents the tangent to the py
curve at the specific load level, was then calculated using
eq. [15] based on calculated soil reactions from the corresponding pile displacements for two consecutive time steps
(using eq. [16]).
Degradation of soil stiffness
Transient loading, especially cyclic loading, may result in
a buildup of pore-water pressures and (or) a change of the
soil structure that causes the shear strain amplitudes of the
soil to increase with an increasing number of cycles (Idriss
et al. 1978). Idriss et al. (1978) reported that the shear stress
amplitude decreased with an increasing number of cycles for
harmonically loaded clay and saturated sand specimens under strain-controlled, undrained conditions. These studies
suggest that repeated cyclic loading results in the degradation of the soil stiffness. For cohesive soils, the value of the
shear modulus after N cycles, GN, can be related to its value
in the first cycle, Gmax, by
[17]
GN = Gmax
where the degradation index, , is given by = N t, and t is

the degradation parameter defined by Idriss et al. (1978).
This is incorporated into the proposed model by updating
the nonlinear stiffness, kNL, by an appropriate factor in each
loading cycle.
Time-domain analysis and equations of

motion
The time-domain analysis was used to include all aspects
of nonlinearity and examine the transient response logically
and realistically. The governing equation of motion is given
by
2000 NRC Canada

1171
Fig. 5. Soil modulus variation for profiles considered in the analysis. d, pile diameter; L, pile depth; z, depth below the surface.
Fig. 6. Calculated dynamic soil reactions at 1.0 m depth (for a prescribed harmonic displacement at the pile head with amplitude
0.03d = 0.015 m, L/d = 30).
[18]
[M ]{u&&} + [C ]{u&} + [K ]{u} = {F(t)}
where [M], [C], and [K] are the global mass, damping, and
stiffness matrices, respectively; and {u&&}, {u&}, {u}, and F(t) are
acceleration, velocity, displacement, and external load vectors, respectively. Referring to Fig. 1, the equations of motion at node 1 (adjacent to the inner field) and node 2
(adjacent to the outer field) are
[19]
m 1u&&1 + c NL (u& 1 u& 2 ) + k NL (u1 u 2 ) = F1
[20]
m 2u&& 2 c NL (u& 1 u& 2 ) k NL (u1 u 2 ) = F2
where u1 and u2 are displacements of nodes 1 and 2, respectively; F1 is the force in the nonlinear spring including the
confining pressure; and F2 is the soil resistance at node 2.
The equation of motion for the outer field is written as
[21]
cu&& 2 + k Lu 2 = F2
Assuming compatibility and equilibrium at the interface

between the inner and outer zones leads to the following
equation, which is valid for both sides of the pile:
[22]
F1
=
0
Am 1 + B(c L + cNL ) + kNL
kNL BcNL
k
k
Am
B
c
c
)
+
+
+
(
+
k
Bc
NL
2
L
NL
L
NL
NL
u1 F i 1
+ 1i 1
u 2 F 2
where F 1i 1 and F i21 are the sums of inertia forces and soil
reactions at nodes 1 and 2, respectively; and A and B are
constants of numerical integration for inertia and damping,
respectively.
The linear acceleration assumption was used and the
Newmark method was implemented for direct time integration of the equations of motion. The modified NewtonRaphson iteration scheme was used to solve the nonlinear
equilibrium equations.
2000 NRC Canada

1172
Fig. 7. Pile head response under applied harmonic load with single amplitude equal to 10 kN (L/d = 30, Ep /Es(L) = 1000, linear profile).
Verification of the analytical model

Verification of clay model
Different soil profiles were considered in the analysis.
Figure 5 shows the typical pilesoil system and the soil profiles considered including linear and parabolic soil profiles.
The py model was first verified against the hyperbolic
model (El Naggar and Novak 1996). Figures 6 and 7 compare the dynamic soil reaction and pile head response for
both the hyperbolic and py curve models for a single reinforced concrete pile in soft clay. A 0.5 m diameter, 15 m
long pile was used with an elastic modulus (Ep ) equal to 35
GPa. A parabolic soil profile with the ratio Ep /Es = 1000 at
the pile base was assumed. The undrained shear strength of
the clay was assumed to be 25 kPa. Figure 6 shows the calculated dynamic soil reactions for a prescribed harmonic
displacement of a single amplitude equal to 0.03d (0.015 m)
at a frequency of 2 Hz at the pile head. Figure 6 shows that
the soil reactions obtained from the two models are very
similar and approach stability after five cycles. The pattern
shown in Fig. 6 is also similar to that obtained by Nogami et
al. (1992) showing an increasing gap and stability after approximately five cycles. Figure 7 shows the displacement
time history of the pile head installed in the same soil profile. A harmonic load with single amplitude equal to 10 kN
was applied at the pile head. The hyperbolic and py curve
models show very similar responses at the pile head and
both stabilize after approximately five cycles.
The dynamic soil reactions are, in general, larger than the
static reactions because of the contribution from damping.
Employing the same definition used for static py curves,
dynamic py curves can be established to relate pile deflections to the corresponding dynamic soil reaction at any depth
below the ground surface. The proposed dynamic py curves
are frequency dependent. These dynamic py curves can be
used in other static analyses that are based on the py curve
approach to approximately account for the dynamic effects
on the soil reactions to transient loading.
Figures 8 and 9 show dynamic py curves established at
two different clay depths for a prescribed harmonic displacement at the pile head with single amplitude equal to 0.05d,
for a frequency range from 0 to 10 Hz. The shear modulus
of the soil was assumed to increase parabolically along the

pile length. A concrete pile 12.5 m in length and 0.5 m in diameter was considered in the analysis. The elastic modulus
of the pile material was assumed to be 35 GPa and the ratio
Ep /Es = 1000 (at the pile base). Both the py curve and hyperbolic models were used to analyze the pile response. The
dynamic soil reaction (normalized by the ultimate pile capacity, Pult) obtained from the py curve model compared
well with that obtained from the hyperbolic relationship
model, especially for lower frequencies, as shown in Figs. 8
and 9, which also show that the soil reaction increased as the
frequency increased. This increase was more evident in the
results obtained from the py curve model.
Verification of sand model
The py curve and hyperbolic models were used to analyze the response of piles installed in sand. The sand was assumed to be unsaturated and a linear soil modulus profile
was adopted. The pile used in the previous case was considered. Figure 10 shows the calculated dynamic soil reactions
at 1 m depth for a prescribed harmonic displacement with an
amplitude equal to 0.038d at the pile head with a frequency
of 2 Hz. The two models feature very similar dynamic soil
reactions. It should be noted that the soil reactions at both
sides of the pile are traced independently. The upper part of
the curve in Fig. 10 represents the reactions for the soil element adjacent to the right face of the pile, when it is loaded
from the left. The lower part represents the reactions of the
soil element adjacent to the left face of the pile as it is
loaded from the right. Both elements offer zero resistance to
the pile movement when tensile stresses are detected in the
nonlinear soil spring during unloading of the soil element on
either side. However, the soil nodes remain attached to the
pile node at the same level, allowing the sand to cave in
and fill the gap. Observations from field and laboratory pile
testing confirmed that, unlike clays, sands usually do not
experience gapping during harmonic loading. Thus both
analyses model the physical behaviour of the soil realistically and logically.
The pile head displacementtime histories obtained from
the py curve and hyperbolic models for a pile installed in a
sand with linearly varying elastic modulus due to an applied
2000 NRC Canada

1173
Fig. 8. Calculated dynamic py curves at 1.5 m depth (for a prescribed harmonic displacement at the pile head with amplitude of
0.05d) using (a) hyperbolic model, and (b) py curve model. p , pile mass density; s , soil mass density; Pult , lateral ultimate load of
the pile.
Fig. 9. Calculated dynamic py curves at 3.0 m depth (for a prescribed harmonic displacement at the pile head with an amplitude of
0.05d) using (a) hyperbolic model, and (b) py curve model.
harmonic load with a single amplitude of 15 kN are shown

in Fig. 11, which shows good agreement between the results
from the py curve model and the hyperbolic model.
Figures 12 and 13 show dynamic py curves established at

two different depths for a prescribed harmonic displacement
equal to 0.05d at the pile head for a steel pile driven in sand
2000 NRC Canada

1174
Fig. 10. Calculated dynamic soil reactions at 1.0 m depth (for a prescribed harmonic displacement at the pile head with an amplitude
of 0.038d, L/d = 25, Ep /Es(L) = 1000).
Fig. 11. Pile head response to applied harmonic load with single amplitude of 15 kN of the pile lateral capacity (L/d = 25, Ep /Es(L) =
1000, linear profile).
for a frequency range from 0 to 10 Hz. The results from

both the py curve and hyperbolic models displayed the
same trend, as shown in Figs. 12 and 13.
Validation of dynamic model with lateral Statnamic tests
To verify that the py curve model can accurately predict
dynamic response, it was employed to analyze a lateral
Statnamic load test, and the computed response was compared with measured values.
The test site was located north of the New River at the
Kiwi maneuvers area of Camp Johnson in Jacksonville,
North Carolina. The soil profile is shown in Fig. 14 and consists of medium dense sand extending to the water table, underlain by a very weak, gray, silty clay. There was a layer of
2000 NRC Canada

1175
gray sand at a depth of 7 m and a calcified sand stratum below the gray sand. The pile tested at this site was a cast-inplace reinforced concrete shaft with a steel casing having an
outer diameter of 0.61 m and a casing wall thickness of
13 mm. More details on the soil and pile properties and the
loading procedure can be found in El Naggar (1998).
Statnamic testing was conducted on the pile 2 weeks after

lateral static testing was performed. Statnamic loading tests
were performed by M. Janes and P. Bermingham, both of
Berminghammer Foundation Equipment, Hamilton, Ontario.
The computed lateral response of the pile head is compared with the measured response in Fig. 15 for two separate
2000 NRC Canada

1176
Table 1. Description of parameters used for each test case.

Test case no.
Case I (clays)
C1
C2
C3
Case II (sands)
S4
S5
S6
S8
S9
Soil type
cu
(kPa)
Soft clay
Medium clay
Stiff clay
<50
80
>100
Loose sand, saturated

Medium sand, saturated
Medium sand, saturated
Dense sand, saturated
Dense sand, unsaturated
Dr
(%)
35
50
50
90
90
()
32
34
34
38
38
d
(m)
L/d
Ep /Es
Gmax
(kPa)
Vs
(m/s)
0.45
0.45
0.45
0.25
0.25
0.25
40
40
40
10 000
4 500
1 600
6.6106
1.6107
8.3107
70
150
200
0.3
0.3
0.3
0.3
0.3
0.25
0.25
0.50
0.25
0.25
40
40
20
40
40
6 300
3 800
3 800
1 580
790
1.2107
2.0107
2.0107
4.7107
9.7107
70
100
100
150
220
Note: All values represent those calculated at a depth of 1.5 m.
Table 2. Dynamic py curve parameter constants for a range of soil types (d = 0.25; L/d = 40; 0.015 < ao < 0.225, where ao = ro /Vs).
Soil type
Description
ao < 0.025
ao > 0.025
Soft clay
Medium clay
Stiff clay
Medium dense sand, saturated
Medium dense sand, unsaturated
Dense sand, saturated
cu < 50 kPa; Vs < 125 m/s

50 < cu < 100 kPa; 125 < Vs < 175 m/s
cu > 100 kPa; Vs > 175 m/s
50 < Dr < 85%; 125 < Vs < 175 m/s
50 < Dr < 85%; 125 < Vs < 175 m/s
Dr > 85%; Vs > 175 m/s
1
1
1
1
1
1
180
120
2900
3320
1960
6000
200
360
828
1640
960
1876
80
84
100
100
20
100
0.18
0.19
0.19
0.1
0.1
0.15
Note: See text and eq. [23].
tests with peak load amplitudes of 350 and 470 kN. The
agreement between the measured and computed values was
excellent, especially for the first load test. The initial displacement was slightly adjusted for the computer-generated
model to accommodate initial gapping that occurred due to
the previous static test performed on the pile. The static py
curve for the top soil layer was reduced significantly to
model the loss of resistance due to the permanent gap developed near the surface.
Fig. 14. Soil profile and Statnamic pile test setup at Camp Johnson, Jacksonville. SPT, standard penetration test.
Dynamic py curve generation

The dynamic py curves presented in Figs. 810, 12, and
13 show that a typical family of curves exists related to
depth, much like the static py curve relationships. Thus, dynamic py curves could be established at any depth and be
representative of the soil resistance at this specific depth. In
this study, they were obtained at a depth of 1.5 m which was
found to illustrate the characteristics of the dynamic py
curves.
More dynamic py curves were generated using prescribed harmonic displacements applied at the pile head
which allowed for the development of plastic deformation in
the soil along the top quarter of the pile length. Steel pipe
piles were considered in the analysis. It was assumed that
sand had a linear soil profile and the clay had a parabolic
profile to match the soil profile employed to derive the static
py curves used in the analysis. The soil shear wave velocity
profiles and the pile properties are given in Fig. 16. The tests
were divided into two separate cases involving clays (case I)
and sandy soils (case II). Table 1 summarizes the characteristics of each case and relevant pile and soil parameters. The
dynamic py curves were generated over a frequency range
2000 NRC Canada

1177
Fig. 15. Pile head displacement for Statnamic test with peak load of (a) 350 kN, and (b) 470 kN.
of 010 Hz (2 Hz intervals) for different classifications of

sand and clay based on standard laboratory and field measurements (standard penetration test value, relative density,
cu, etc.). All results were obtained after one or two cycles of
harmonic loading.
Results and discussion
The results from the computational model showed a general trend of increasing soil resistance with an increase in
the load frequency. The dynamic py curves obtained seem
to have three distinct stages or regions. The initial stage (at
small displacements) shows an increase in the soil resistance
(compared with the static py curve) that corresponds to increasing the velocity of the pile to a maximum. This increase in the soil resistance is larger for higher frequencies.
In the second stage, the dynamic py curves have almost the
same slope as the static py curve for the same displacement. This stage occurs when velocity is fairly constant and
consequently the damping contribution is also constant. The

third stage of the dynamic py curve is characterized by a
slope approaching zero as plastic deformations start to occur
(similar to the static py curve at the same displacement).
There is also a tendency for the dynamic curves to converge
at higher resistance levels approaching the ultimate lateral
resistance of the soil at depth x, Pu (determined from American Petroleum Institute 1991).
The overall relationship between the dynamic soil resistance and loading frequency for each test was established in
the form of a generic equation. The equation was developed
from an regression analysis relating the static py curve, frequency, and apparent velocity, y, so
[23]
y
Pd = Ps + a 2o + a o
,
d
Pd Pu at depth x
2000 NRC Canada

1178
Fig. 16. Description of soil and pile properties for cases I and II.
where Pd is the dynamic py curve at depth x (N/m); Ps is

the static soil reaction (obtained from the static py curve) at
depth x (N/m); ao is the dimensionless frequency = ro/Vs;
is the frequency of loading (rad/s); d is the pile diameter
(m); y is the lateral pile deflection at depth x when soil and
pile are in contact during loading (m); and , , , and n are
constants determined from curve fitting eq. [23] to the computed dynamic py curves from all cases considered in this
study. A summary of the best-fit values for the constants is
provided in Table 2. The constant is taken equal to unity to
ensure that Pd = Ps for = 0. For large frequencies or displacements, the maximum dynamic soil resistance is limited
to the ultimate lateral resistance of the soil, Pu.
Figures 17 and 18 show dynamic py curves established
using eq. [23] and the best-fit constants (as broken lines).
The approximate dynamic py curves established from
eq. [23] represented soft medium clays and loose medium
dense sands reasonably well. However, the accuracy is less
for stiffer soils (higher Vs values). The precision of the fitted
curves also increases with an increase in frequency ( 4 Hz)
where the dynamic effects are important. The low accuracy
at a lower frequency (ao < 0.02) may be attributed to the application of the plane strain assumption in the dynamic analysis. This assumption is suitable for higher frequencies, as
the dynamic stiffness of the outer field model vanishes for
ao < 0.02 due to the assumption of plane strain. Test case C1
was also used to obtain dynamic py curves at depths of 1.0
and 2.0 m to examine the validity of eq. [23] to describe the
dynamic soil reactions at other depths along the soil profile.
The results showed that eq. [23] (using the constants in
Table 2) predicted the dynamic soil reactions reasonably
well.
Development of a simplified model

For many structural dynamic programs, soilstructure interaction is modeled using static py curves to represent the
soil reactions along the pile length. However, the use of
static py curves for dynamic analysis does not include the
effects of velocity-dependent damping forces. The dynamic

py curves established using eq. [23] and the parameters
given in Table 2 allow for the generation of different dynamic py curves based on the frequency of loading and soil
profile. Substituting dynamic py curves in place of traditional
static py curves for analysis should result in better estimates of the response of structures to dynamic loading.
Alternatively, the dynamic soil reactions can be represented using a simple spring and dashpot model. This model
can still capture the important characteristics of the nonlinear dynamic soil reactions. A simplified dynamic model that
can be easily implemented into any general finite element
program is proposed herein.
Complex stiffness model
As discussed previously, eq. [23] can be used directly to
represent the dynamic relationship between a soil reaction
and a corresponding pile displacement. The total dynamic
soil reaction at any depth is represented by a nonlinear
spring whose stiffness is frequency dependent.
A more conventional and widely accepted method of calculating dynamic stiffness is through the development of the
complex stiffness. The complex stiffness has a real part K1
and an imaginary part K2, i.e.,
[24]
Pd = Ky = (K1 + iK2)y
The real part, K1 represents the true stiffness, k, and the

imaginary part of the complex stiffness, K2, describes the
out-of-phase component and represents the damping due to
the energy dissipation in the soil element. Because this
damping component generally grows with frequency (resembling viscous damping), it can also be defined in terms of
the constant of equivalent viscous damping (the dashpot
constant) given by c = K2 /.
Then the dynamic py curve relation can be described as
[25]
Pd = (k + i c) y = ky + cy&
in which both k and c are real and represent the spring and
dashpot constants, respectively; and y& = dy/dt is the velocity.
2000 NRC Canada

1179
Fig. 17. Dynamic py curves and static py curve for test case C1 (depth = 1.0 m).
Using eq. [23], the dynamic py curve can be written in the

form of eq. [24], i.e.,
[28]
[26]
Pd = (K 1 + iK 2 ) y =
n
Ps a 2o + a o
d
Ps
y
+i
y
y
The stiffness and damping constants are then calculated as
[27]
k = K1 =
Ps
y
y
Ps a 2o + a o
d
c= 2 =
The complex stiffness can be generated at any depth along

the pile using the static py curves and eqs. [27] and [28].
Complex stiffness constants: soft clay example
The complex stiffness constants were calculated for test
case C1 (see Table 2) using the method described in the previous section. The values of the true stiffness, k, were obtained for the range of displacements experienced by the pile
for the frequency range from 0 to 10 Hz. The stiffness parameter (S1)py was defined as
[29]
(S1)py =
k py
Gmax
2000 NRC Canada

1180
Fig. 18. Dynamic py curves and static py curve for test case S5 (depth = 1.5 m).
The constant of equivalent damping, c, was obtained by

averaging the value from eq. [28] for the range of velocities
experienced by the pile for each frequency of loading. Then
the equivalent damping parameter (S2)py was defined as
[30]
(S2)py =
cV s
Gmaxro
Figure 19 shows the true stiffness calculated from the

static py curve and this stiffness is identical at all loading
frequencies considered. There is a definite trend of decreasing stiffness with increased displacement due to the soil
nonlinearity. The constant of equivalent damping presented
in Fig. 20 decreases with a decrease in frequency which can
be attributed to separation at the pilesoil interface. The values from Figs. 19 and 20 can be directly input into a finite
element program as spring and dashpot constants to obtain
the approximate dynamic stiffness of a soil profile similar to
that of test case C1.
Implementing dynamic py curves in ANSYS

A pile and soil system similar to that of test case C1 was
modeled using the commercial finite element program
ANSYS (1996) to verify the applicability and accuracy of
the dynamic py curve model in a standard structural analysis program. A dynamic harmonic load with peak amplitude
of 100 kN at a frequency of 6 Hz was applied to the head of
the same steel pipe pile used in test case C1. The soil stiffness was modeled using three procedures: (1) static py
curves; (2) dynamic py curves using eq. [23]; and (3) a
complex stiffness method using equivalent damping constants.
The pile head response for each test was obtained and compared with the results from the two-dimensional (2D) py
curve model.
The pile was modeled using two-noded beam elements
and was discretized into 10 elements that increased in length
with an increase in depth. At each pile node, a spring or a
spring and a dashpot was attached to both sides of the pile to
represent the appropriate loading condition at the pilesoil
2000 NRC Canada

1181
Fig. 19. True stiffness parameter for test case C1 (soft clay).
Fig. 20. Equivalent damping parameter for test case C1 (soft clay) with dimensionless frequency, ao.
interface. The pile and soil remained connected and had

equal displacement for compressive stresses. The spring or
the spring and dashpot model disconnects if tensile stress is
detected in the soil, allowing a gap to develop.
The soil was first modeled using nonlinear springs with
forcedisplacement relationships calculated directly from
static py curves. The soil stiffness was then modeled using
the approximate dynamic py curve relationship calculated
for test case C1 using Table 2. The last test considered a
spring and a dashpot in parallel.
The pile head response for each test is shown in
Figs. 21 and 22 along with the calculated response from
the 2D analytical py model. Figure 21a shows that the
static py curves model computed larger displacements
with increasing amplitudes as the number of cycles increased. Figure 21b shows that the response computed using the dynamic py curve model was in good agreement
with the response computed using the 2D analytical

model. The results obtained using the complex stiffness
model are presented in Fig. 22a and show a decrease in
displacement amplitude. The overdamped response can be
attributed to using an average damping constant, which
overestimates the damping at higher frequencies and large
nonlinearity. Figure 22b shows the response of the 2D
model compared to the complex stiffness approach with
the average damping constant reduced by 50%. The results show that the response in this case is in good agreement with the response computed using the 2D analytical
model.
Conclusions
A simple 2D analysis was developed to model the response of piles to dynamic loads. The time domain was
2000 NRC Canada

1182
Fig. 21. Calculated pile head response using 2D analytical model compared with ANSYS using (a) static py curves, and (b) dynamic
py curves.
Fig. 22. Calculated pile head response using 2D analytical model compared with ANSYS using (a) complex stiffness, and (b) modified
complex stiffness.
chosen to efficiently model the transient nonlinear response

of the pilesoil system. Static py curves were used to generate the nonlinear soil stiffness in the frame of a Winkler
model. The piles were assumed to be vertical and circular
and were modeled using standard beam elements. A practically accurate and computationally efficient model was developed to represent the soil reactions. This model
accounted for soil nonlinearity, slippage and gapping at the

pilesoil interface, and viscous and material damping.
Dynamic soil reactions (dynamic py curves) were generated for a range of soil types and harmonic loading with
varying frequencies applied at the pile head. Closed-form
solutions were derived from regression analysis relating the
static py curve, dimensionless frequency, and apparent
2000 NRC Canada

velocity of the soil particles. A simple spring and dashpot

model was also proposed whose constants were established
by splitting the dynamic py curves into real (stiffness) and
imaginary (damping) components. This model could be used
in equivalent linear analysis for harmonic loading at the pile
head. The proposed dynamic py curves and the spring and
dashpot model were incorporated into a commercial finite
element program, ANSYS, that was used to compute the response of a laterally loaded pile. The computed responses
compared well with the predictions of the 2D analysis.
The following conclusions were drawn from the study:
(1) The developed pilesoil model is capable of analyzing
the response of piles to lateral transient loading accounting
for the nonlinear behaviour of the soil and energy dissipation.
(2) The soil resistance to the pile motion increases with
the frequency of the pile head loading (inertial loading) for
single piles.
(3) The developed dynamic py curve can represent the
dynamic soil reactions using a predefined static py curve.
(4) The pile head response under harmonic loading can be
approximately modeled using dynamic py curve functions
in most of the available structural analysis programs.
(5) The implementation of the dynamic py curves in
time-domain analyses to model the soil reactions is preferred
because it accounts for the variation in damping with the
displacement level. The damping constant in the complex
stiffness model must be chosen accounting for the level of
nonlinearity expected.
Acknowledgements
This research was supported by a grant from the Natural
Sciences and Engineering Research Council of Canada
(NSERC) to the first author and research contract 24-9 from
the National Cooperative Highway Research Program
(NCHRP). Both sources of support are greatly appreciated.
References
Abendroth, R.E., and Greimann, L.F. 1990. Pile behavior established from model tests. Journal of Geotechnical Engineering,
ASCE, 116(4): 571588.
American Petroleum Institute. 1991. Recommended practice for
planning, designing and constructing fixed offshore platforms.
API Recommended Practice 2A (RP 2A). 19th ed. American Petroleum Institute, Washington, D.C., pp. 4755.
ANSYS Inc. 1996. General finite element analysis program. Version 5.4. ANSYS Inc., Canonsburg, Pa.
Bhushan, K., and Askari, S. 1984. Lateral load tests on drilled pier
foundations for solar plant heliostats. In Laterally loaded piles.
Edited by J.A. Langer. American Society for Testing and Materials, Special Technical Publication STP 835, pp. 141155.
Bhushan, K., and Haley, S.C. 1980. Development of computer program using Py data from load test results for lateral load design of drilled piers. Woodward-Clyde Consultants Professional
Development Committee, San Francisco, Calif.
1183
Bhushan, K., Haley, S.C., and Fong, P.T. 1979. Lateral load tests
on drilled piers in stiff clays. Journal of the Geotechnical Engineering Division, ASCE, 105(GT8): 969985.
Bhushan, K., Lee, L.J., and Grime, D.B. 1981. Lateral load tests on
drilled piers in sand. In Proceedings of a Session on Drilled
Piers and Caissons, Geotechnical Engineering Division, American Society of Civil Engineers, National Convention, St. Louis,
Mo., pp. 131143.
Desai, C.S., and Wu, T.H. 1976. A general function for stress
strain curves. In Proceedings of the 2nd International Conference on Numerical Methods in Geomechanics, American Society of Civil Engineers, Blacksburg, Vol. 1, pp. 306318.
El Naggar, M.H. 1998. Interpretation of lateral statnamic load test
results. Geotechnical Testing Journal, 21(3): 169179.
El Naggar, M.H., and Novak, M. 1995. Nonlinear lateral interaction in pile dynamics. Journal of Soil Dynamics and Earthquake
Engineering, 14(3): 141157.
El Naggar, M.H., and Novak, M. 1996. Nonlinear analysis for dynamic lateral pile response. Journal of Soil Dynamics and Earthquake Engineering, 15(4): 233244.
Gazetas, G., and Dobry, R. 1984. Horizontal response of piles in
layered soils. Journal of Geotechnical Engineering, ASCE,
110(1): 2040.
Hardin, B.O., and Black, W.L. 1968. Vibration modulus of normally consolidated clay. Journal of the Soil Mechanics and
Foundations Division, ASCE, 94(SM2): 353369.
Idriss, I.M., Dobry, R., and Singh, R.D. 1978. Nonlinear behavior
of soft clays during cyclic loading. Journal of the Geotechnical
Engineering Division, ASCE, 104(GT12): 14271447.
Matlock, H. 1970. Correlations for design of laterally loaded piles
in soft clay. In Proceedings of the 2nd Offshore Technology
Conference, Houston, Tex., Vol. 1, pp. 577588.
Meyer, B.J., and Reese, L.C. 1979. Analysis of single piles under
lateral loading. Preliminary Review Copy, Research Report 2441, Center for Highway Research, The University of Texas at
Austin, Austin, Tex., pp. 1145.
Nogami, T., Konagai, K., Otani, J., and Chen, H.L. 1992. Nonlinear soilpile interaction model for dynamic lateral motion. Journal of Geotechnical Engineering, ASCE, 118(1): 106116.
Novak, M., and Sheta, M. 1980. Approximate approach to contact
effects of piles. In Proceedings of a Speciality Conference on
Dynamic Response of Pile Foundations: Analytical Aspects,
American Society of Civil Engineers, Hollywood, Fla., pp. 53
79.
Novak, M., Nogami, T., and Aboul-Ella, F. 1978. Dynamic soil reactions for plane strain case. Journal of the Engineering Mechanics Division, ASCE, 104: 953959.
Reese, L.C., and Welch, R.C. 1975. Lateral loading of deep foundations in stiff clay. Journal of the Geotechnical Engineering Division, ASCE, 101(GT7): 633649.
Reese, L.C., Cox, W.R., and Koop, F.D. 1974. Analysis of laterally
loaded piles in sand. In Proceedings of the 6th Annual Offshore
Technology Conference, Houston, Tex., Vol. 2, Paper OTC
2080, pp. 473483.
Reese, L.C., Cox, W.R., and Koop, F.D. 1975. Field testing and
analysis of laterally loaded piles in stiff clay. In Proceedings of
the 7th Annual Offshore Technology Conference, Houston, Tex.,
Vol. 2, pp. 671690.
2000 NRC Canada

t00 058

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

t00 058

Caricato da

Copyright:

Formati disponibili

Color profile: Generic CMYK printer profile

Composite Default screen

Dynamic analysis for laterally loaded piles and

El Naggar and Bentley

Color profile: Generic CMYK printer profile

El Naggar and Bentley

01 Hz for offshore environmental loading, and 5200 Hz

the radiation damping in the soil medium. The soil reactions

where ro is the pile radius, r1 is the outer radius of the inner

where Gmax is the maximum shear modulus (small strain

Color profile: Generic CMYK printer profile

Can. Geotech. J. Vol. 37, 2000

where Pu is the minimum of the resistances calculated by

Pu 2 = A Xd [ Ka (tan8 1) + K0 tan tan4 ]

where A is an empirical adjustment factor dependent on the

K = Gmax[Su1(ao, , D) + iSu2(ao, , D)]

where ao = r1/Vs is the dimensionless frequency, is the

(d + X tan tan ) + K0 X tan (tan sin

In the derivation of eq. [1], the inner field was assumed to

Figure 2 shows that for the dimensionless frequency range

Color profile: Generic CMYK printer profile

El Naggar and Bentley

The stiffness of the nonlinear strength is then calculated

The constant of the linear elastic spring, kL, is established

Color profile: Generic CMYK printer profile

Can. Geotech. J. Vol. 37, 2000

Therefore, eqs. [10] and [15] can be substituted into

where k is a constant that depends on the lateral deflection y

displacement and thus the nonlinearity of the sand can be

where the degradation index, , is given by = N t, and t is

Time-domain analysis and equations of

Color profile: Generic CMYK printer profile

El Naggar and Bentley

[M ]{u&&} + [C ]{u&} + [K ]{u} = {F(t)}

m 1u&&1 + c NL (u& 1 u& 2 ) + k NL (u1 u 2 ) = F1

m 2u&& 2 c NL (u& 1 u& 2 ) k NL (u1 u 2 ) = F2

Assuming compatibility and equilibrium at the interface

Am 1 + B(c L + cNL ) + kNL

Color profile: Generic CMYK printer profile

Can. Geotech. J. Vol. 37, 2000

Verification of the analytical model

of the soil was assumed to increase parabolically along the

Color profile: Generic CMYK printer profile

El Naggar and Bentley

harmonic load with a single amplitude of 15 kN are shown

Figures 12 and 13 show dynamic py curves established at

Color profile: Generic CMYK printer profile

Can. Geotech. J. Vol. 37, 2000

for a frequency range from 0 to 10 Hz. The results from

Color profile: Generic CMYK printer profile

El Naggar and Bentley

Statnamic testing was conducted on the pile 2 weeks after

Color profile: Generic CMYK printer profile

Can. Geotech. J. Vol. 37, 2000

Table 1. Description of parameters used for each test case.

Loose sand, saturated

Note: All values represent those calculated at a depth of 1.5 m.

cu < 50 kPa; Vs < 125 m/s

Note: See text and eq. [23].

Dynamic py curve generation

Color profile: Generic CMYK printer profile

El Naggar and Bentley

of 010 Hz (2 Hz intervals) for different classifications of