Vous êtes sur la page 1sur 14

1268

A numerical study into lateral cyclic nonlinear


soil–pile response
Nii Allotey and M. Hesham El Naggar

Abstract: Pile foundations are generally designed to resist both axial and lateral loads. Under lateral cyclic loading, the re-
sponse of the pile foundation is affected by factors such as soil and pile yielding, gapping, and soil cave-in. These factors
directly influence the effective lateral stiffness and strength of the foundation and can govern the design. In this paper,
two case studies of single piles, one in clay and one in sand, are used to examine the influence of the aforementioned fac-
tors on nonlinear cyclic response of piles. The numerical study is conducted using a recently developed beam on a nonlin-
ear Winkler foundation (BNWF) model. The results of the study point to the important role soil cave-in and
recompression play in the cyclic soil–pile response, and elucidate how this could particularly be beneficial to piles that de-
velop plastic hinges below ground level.
Key words: cyclic, seismic, nonlinear, soil–pile–structure interaction, Winkler model, degradation, soil cave-in, plastic
hinge.
Résumé : Les fondations de pieux sont généralement conçues pour résister aux chargements tant latéraux qu’axiaux. Sous
un chargement latéral cyclique, la réaction de la fondation de pieux est affectée par des facteurs tels que la déformation du
pieu et du sol, les trous et les affaissements. Ces facteurs influencent directement la rigidité latérale effective et la résis-
tance de la fondation et peuvent en gouverner la conception. Dans cet article, deux études de cas de pieu simple dans
l’argile et le sable sont utilisées pour examiner l’influence des facteurs mentionnés ci-haut sur la réaction cyclique non li-
néaire des pieux. L’étude numérique a été conduite en utilisant un modèle de Winkler (« BNWF ») d’une poutre sur une
fondation non linéaire récemment développé. Les résultats de l’étude pointent le rôle important que les affaissements du
sol et la recompression jouent dans la réaction cyclique sol-pieu, et élucident comment ceci pourrait être particulièrement
bénéfique pour les pieux qui développent des charnières plastiques sous le niveau du sol.
Mots-clés : cyclique, séismique, non linéaire, interaction sol-pieu-structure, modèle Winkler, dégradation, affaissement du
sol, charnières plastiques.
[Traduit par la Rédaction]

Introduction ance with recommendations given in the overview article by


Liam Finn (2005), who suggests that the BNWF approach is
Dynamic soil–pile–structure interaction (SPSI) under ex- here to stay and must be improved. The main reason for its
treme loading conditions such as earthquakes involves many popularity is that, with limited computational effort, the
factors including soil and structural yielding, pile–soil gap BNWF model can satisfactorily account for soil nonlinearity
formation with or without soil cave-in and recompression, while still allowing for detailed structural modeling, a fea-
cyclic degradation of soil stiffness and strength, and radia- ture that is generally desired by structural engineers. Exam-
tion damping. Approaches used in modeling SPSI problems ples of enhanced BNWF models include those by Boulanger
include finite element (FE) formulations, semi-analytical and et al. (1999), El Naggar and Bentley (2000), Brown et al.
boundary element (BE) formulations, the extended Tajimi (2001), Gerolymos and Gazetas (2005), and Allotey and El
formulation, and the beam on a nonlinear Winkler founda- Naggar (2008). The major differences between these models
tion (BNWF) method (Gazetas and Mylonakis 1998). have been highlighted in Allotey (2006). Noting the signifi-
The advent of performance-based design, which relies cant importance of the BNWF approach in design,
heavily on nonlinear analysis procedures such as pushover Gerolymos and Gazetas (2005) have gone further to extend
and incremental dynamic analysis, has resulted in the adap- their model to include caisson applications.
tation and extension of the BNWF model. This is in accord-
Objective of the study
Received 6 December 2006. Accepted 23 April 2008. Published Under lateral cyclic soil–pile action there exists a
on the NRC Research Press Web site at cgj.nrc.ca on 29 August
2008.
repeated mechanism of soil flowability and recompression
in the near-field that influences the cyclic response of the
N. Allotey. Department of Civil Engineering, The University of pile (Long and Vanneste 1994). Depending on soil type
British Columbia, Vancouver, BC V6T 1Z4, Canada. and depth, this generally results in the formation of various
M.H. El Naggar.1 Department of Civil and Environmental types of hysteresis loops ranging from S-shaped loops to
Engineering, Faculty of Engineering, The University of Western oval-shaped loops (Fig. 1). Most cyclic pile–soil studies
Ontario, London, ON N6A 5B9, Canada.
generally focus on the effect nonlinear soil response has on
1Corresponding author (e-mail: naggar@uwo.ca). elastic pile behaviour (e.g., Swane and Poulos 1984; Pender

Can. Geotech. J. 45: 1268–1281 (2008) doi:10.1139/T08-050 # 2008 NRC Canada


Allotey and El Naggar 1269

Fig. 1. Typical hysteresis loops from different cyclic and seismic pile load tests: (a) modified from Meymand (1998); (b) modified from
Dou and Byrne (1996).

and Pranjoto 1996; Yang and Jeremic 2002) and do not as- p–y curves have been derived and can be found in Allotey
sess its effect on nonlinear pile response. The primary goal (2006) and Allotey and El Naggar (2008). The initial hori-
of this study is therefore to assess the influence gapping, zontal earth pressure is modeled as a prestraining effect that
and soil flowability and recompression have on the non- shifts the backbone curve leftwards; elements on both sides
linear cyclic response of soil–pile systems. This is achieved of the pile are therefore preloaded at zero pile displacement.
by using the BNWF model developed by Allotey and El A similar approach has been used by Pender and Pranjoto
Naggar (2008), which is a generalization of the ‘‘pure (1996) and Allotey and Foschi (2005).
gap’’ and ‘‘no gap’’ methods (El Naggar and Bentley The SRC (i.e., segments 7, 8, 9, and 10 in Fig. 2a) and
2000; Brown et al. 2001), to study the response of two sin- GUC (i.e., segments 5 and 6 in Fig. 2a) are scaled versions
gle pile case studies available in the literature. This BNWF of the backbone curve. They are derived using procedures
model has been incorporated into the large displacement similar to the well-established approach of extended Masing
nonlinear structural analysis program, SeismoStruct (Seis- rules (Vucetic 1990), with the scaling factor developed as a
moSoft 2003), which allows for detailed modeling of non- function of the reload or unload point (in accordance with
linear cyclic pile behaviour. Pyke (1979)) and the current cyclic degradation factor
(Allotey and El Naggar 2008). For monotonic backbone
BNWF model description curves, the SRC comprises four segments, 7, 8, 9, and 10.
For backbone curves exhibiting post-peak behaviour, the
The dynamic BNWF model by Allotey and El Naggar SRC comprises three segments, 7, 8, and 9 (i.e., segments 9
(2008) can be classified as a degrading polygonal hysteretic and 10 are merged). The SRC represents the reloading curve
model with defined rules for loading, reloading, and unload- when reloading is not in the slack zone.
ing. The model is compression-dominant, requiring two ele-
ments at each depth for the modeling of pile–soil interaction Direct reload curve
and is made up of four main parts: the backbone curve, The DRC simulates soil reactions when the pile moves in
standard reload curve (SRC), general unload curve (GUC), the slack zone. It commences immediately after movement
and direct reload curve (DRC). These different parts are de- at the minimum force level in the negative direction ends
scribed briefly below. The model is capable of accounting (i.e., after segment 11). The DRC is designed as a convex
for cyclic soil degradation and reduced radiation damping strain-hardening curve that is controlled by a limiting-force
due to increased soil nonlinearity. These capabilities are, parameter, lf , (0 £ lf £ 1) referenced to the past maximum
however, not discussed in this report, and for a detailed force. In addition, a curve-shape parameter, ls, (0 £ ls £ 1)
overview of the model the reader is referred to Allotey can also be used to control the shape of the DRC (Fig. 2b).
(2006), El Naggar and Allotey (2007), and Allotey and El For fully unconfined response, or a pure gap (e.g., piles in
Naggar (2008). stiff clay), lf = 0, and for fully confined response (e.g.,
lower portions of piles in dry sand), lf = ls = 1. These
Backbone curve, standard reload curve, and general parameters are similar in function to those available in the
unload curve effective stress analysis program, CYCLIC 1D, developed
The backbone curve is a four segment adaptable multi- by Elgamal and coworkers (Elgamal et al. 2002; Yang et al.
linear curve (i.e., segments 1, 2, 3, and 4 in Fig. 2a) that 2003), which efficiently models cyclic soil mobility.
can either be monotonically increasing (as shown in Figure 2c shows two standard reload curves: curve A rep-
Fig. 2a) or exhibit post-peak residual behaviour. (In Fig. 2, resents the SRC corresponding to stable gap formation (e.g.,
segments 1 and 2 model the monotonically increasing curve as occurs in some clays), and curve B, which is offset from
up to the peak force, after which segments 3 and 4 model curve A to the left, represents a foundation moving back to
the post-peak behaviour). The various nodes and slopes of the point where it last separated from the soil (the approach
the curve are established by fitting the curves to prescribed used in PYLAT (El Naggar and Bentley 2000) and FLPIER-D
force–displacement curves. Parameters for different standard (Brown et al. 2001)). The offset of curve B (termed the base
# 2008 NRC Canada
1270 Can. Geotech. J. Vol. 45, 2008

Fig. 2. Details of Allotey and El Naggar (2008) beam on a non- fit with the mean estimates of Long and Vanneste (1994),
linear Winkler foundation model: (a) backbone, standard reload, and also satisfactorily models fully confined behaviour. For
and general unload curves; (b) direct reload curve; and (c) empiri- L = 0, no soil cave-in occurs and curve B becomes curve A.
cal hyperbolic curve for estimating origin of base standard reload Based on the above description, the three cyclic-curve
curve. parameters, lf, ls, and L have a physical significance and
vary with soil type and depth, and their depth-wise distribu-
tion is linked to the expected soil failure mechanism. Exam-
ple distributions that have been investigated (Allotey 2006)
include a linear distribution representing a planar failure
mechanism, and a step-function distribution representing
caved-in soil occupying the lower portions of the gap depth.

Case study I: Elastic pile in medium-stiff clay


This case study was initially investigated by Pender and
Pranjoto (1996) and later by Pranjoto and Pender (2003).
The problem involves a 12 m long reinforced concrete pile
with diameter, d, of 600 mm. The pile was embedded in
uniform medium-stiff clay with undrained strength, cu, of
50 kPa, and unit weight, gt, of 19 kN/m2. It was loaded lat-
erally at its head (a distance of one diameter aboveground)
with 25 uniform two-way loading cycles with an amplitude
of 213 kN. The pile-yield moment, My, was 636 kNm, and
the pile was assumed to stay fully elastic below the yield
moment.
Pender and Pranjoto (1996) and Pranjoto and Pender
(2003) (hereafter the two references will be referred to as
Pender–Pranjoto) modeled the problem using an extended
version of an in-house BNWF program developed by Carter
(1984). The soil at each depth was represented by two
springs, one on the left and the other on the right. Gapping
was modeled using a ‘‘pure gapping’’ procedure in which
the full gap distance had to be traversed before reloading
commenced. Figure 3 is a schematic of their approach and
is taken from the Pranjoto and Pender (2003) reference.
Their BNWF formulation is similar to the model used in
this study, and their results serve as a baseline for compari-
son purposes.

Model description and input parameters estimation


The pile was modeled using 40 elastic beam–column ele-
ments, and the soil was modeled using two springs at each
pile node. The p–y curve used by Pender–Pranjoto was the
generalized hyperbolic function
n
p pf hy
½1 y¼
K0 ðpnf hy  pnhy Þ
where p is soil reaction, y is the pile displacement, K0 is the
initial stiffness, pf is the ultimate force, and nhy is the curve
shape parameter. For clays, nhy varies between 0.2 and 0.3,
and a nhy of 0.2 was used in the study. Table 1 lists the
SRC) increases with the extent of soil cave-in, and the finite model parameters for a four segment multilinear curve fit
volume that the compressed soil occupies. From a large to the hyperbolic curve. In Table 1, pi and aj are the ith
dataset of cyclic load tests on piles compiled by Long and endpoint and jth segment of the multilinear curve.
Vanneste (1994), Allotey (2006) proposed a hyperbolic Pender–Pranjoto assumed pf to increase linearly from 5cu
curve (shown in Fig. 2c) that could be used to estimate the at the ground surface to 12cu at a depth of 3.5d, and then
origin of the base SRC. In Fig. 2c, fh, is the cyclic loading remain constant below this depth. This was based on the
ratio (where an fh of –1 represents two-way loading, and an results of lateral load pile tests in clay conducted by Carter
fh of 0 represents one-way loading) and L is the soil cave- (1984). The initial p–y curve stiffness, K0, was derived from
in parameter. A soil cave-in parameter, L = 5 gave the best the distributed stiffness, k0, which was estimated using a
# 2008 NRC Canada
Allotey and El Naggar 1271

Fig. 3. (a) Nonlinear Winkler p–y relationship; (b) detaching and reattachment of springs on the front and rear of the pile shaft (after
Pranjoto and Pender 2003).

Table 1. Parameters of the p-y curve for Pender-Pranjoto case study.

Backbone curve parameters*


p1 0.15
p2 0.60
a2 0.28
a3 0.03
Cyclic curve shape parameters
Case lf L Remarks
A 0 0 Static p-y curves; lf, L constant along depth
B 0 0 Cyclic p-y curves; lf, L constant along depth
C 1 5 Static p-y curves; lf, L constant along depth
D 1 0 Static p-y curves; lf, L constant along depth
E 0–1<6.5d 0–5<6.5d Static p-y curves; linear variation of lf, L up to 6.5d
1>6.5d 5>6.5d
F 0<5d 0<5d Static p-y curves; stepwise variation of lf, L up to 5d
1>5d 5>5d
Note: lf, limiting force parameter; L, soil cave parameter.
*Curve parameters defined for a unit p-y curve.

modified version of the relationship proposed by Vesic Fig. 4. Comparison of Matlock p–y curves with adjustments by
(1961), i.e., Bhushan et al. (1979) for stiff clay and Pender–Pranjoto p–y curves
sffiffiffiffiffiffiffiffiffiffiffiffi at selected pile depths.
1:3 12 Es d 4 d
½2 k0 ¼
ð1  s Þ Epl Ipl dref

where ys is the soil Poisson’s ratio, Es and Epl are the


Young’s modulus of the soil and pile, respectively, Ipl is
the moment of inertia of the pile cross-section, and dref is a
reference pile diameter taken to be 1.0 m. The soil Young’s
modulus, Es, was taken to be 600cu based on the results of a
back-analysis study conducted by Ling (1988). The coeffi-
cient of lateral earth pressure was also taken to be 1.0.
Table 1 also shows the six different p–y model parameter
combinations that were investigated for this case study. The
base case (case A) is representative of the ‘‘pure gapping’’
approach used by Pender–Pranjoto. As previously noted,
this is the approach used in PYLAT and FLPIER-D. Case B
is similar to case A, but the p–y curves were modified using
the pseudostatic procedure recommended by Matlock (1970)
to account for cyclic degradation. This modification pro-
duced a third segment with a negative descending branch
for p–y curves within the top 3.5d of the pile (Fig. 4). Cases
# 2008 NRC Canada
1272 Can. Geotech. J. Vol. 45, 2008

Fig. 5. (a) Variation of ground line displacement ratio and maximum moment ratio with number of loading cycles; (b) variation of gap
depth ratio with number of loading cycles.

C through F differ from the base case in terms of the values tio (Mmax/My) for the base case (case A), with those obtained by
of the cyclic curve shape parameters, i.e., L and lf (ls = 1), Pender–Pranjoto. As expected, both sets of results are gen-
and represent different assumed soil cave-in and recompres- erally in good agreement for the entire loading history.
sion behaviours. Case C assumes no slack zone develops, si- Figure 5b also shows computed gapping depths slightly
milar to cyclic p–y models that are based on Masing rules larger than those obtained by Pender–Pranjoto, but similar
(e.g., Kagawa and Kraft 1980; Thavaraj 2001). Cases E and in trend. Such differences could most likely be attributed
F differ in terms of the value L and lf within the top third to slight differences in the loading curves used in both
of the pile. In both these cases, gapping is assumed to occur studies.
to a depth of 6.5d, however, in case F, the caved-in soil is As shown in Fig. 5, maximum ground line displacement
assumed to fill the created gap to a depth of 1.5d. In case ratio, maximum bending moment ratio, and gap depth ratio
E, L, and lf vary linearly with depth, and this case is as- all increase with the number of cycles. This is because per-
sumed to represent a planar soil face failure scenario. Case manent displacements increase during each cycle, resulting
D is an intermediary case between that of cases A and C, in larger lateral pile movements in the following cycle, and
and is helpful in understanding the results of the other cases. consequently, the tensile forces in springs along the lower
portion of the pile overcome the initial confining pressure,
Results and discussion leading to an increase in the gap depth. This subsequently
results in larger pile loads and increased bending moments.
Base cyclic response behaviour
Figure 5a compares the predicted maximum ground line dis- Effect of cyclic p–y parameter variation
placement ratio (yGL_max/d) and maximum bending moment ra- A summary of the predicted cyclic responses for model
# 2008 NRC Canada
Allotey and El Naggar 1273

Fig. 6. Computed cyclic p–y loops for six curve shape parameter cases.

Fig. 7. Maximum moment ratio depth profile for six curve shape parameter cases A through F is presented in Figs. 6 and 7.
parameter cases. Figure 6 shows the cyclic loops computed for the six differ-
ent cases, whereas Fig. 7 shows the corresponding maximum
moment ratio versus depth profiles. As expected, case C dis-
played an oval shape and slight increase in displacement
with cycling. In Fig. 7, the bending moment profile for this
case was close to the N = 1 profile, demonstrating that load
cycling had a limited effect on response. The loops for case
D were similar to those for case C, but with a larger dis-
placement increase. The bending moment profile for case D
was larger than case C, and its maximum moment occurred
at a lower depth. The only difference between the two cases
was the value of parameter L, which is noted to signifi-
cantly influence the gapping depth, and, consequently, the
pile bending moment. The same can be observed for case
E, where L increases linearly with depth. The displacements
for case E were smaller than those for case D, and this re-
sulted in a smaller maximum bending moment occurring at
a higher depth. The response for case F, on the other hand,
# 2008 NRC Canada
1274 Can. Geotech. J. Vol. 45, 2008

was initially similar to case A, but differed as cycling pro- Table 2. Soil and pile details for 2d and 6d aboveground bridge
gressed because of the reduction in gapping depth due to bent experiments.
soil cave-in and recompression.
Adjusting the original p–y curves to account for cyclic 2d 6d
degradation as explained for case B resulted in only an Soil and structural properties
increase in the pile maximum displacement and bending Pile diameter, d (mm) 406 406
moment; the general form of the response, however, re- Embedded pile length 13.5d 13.5d
mained similar to case A. Also, the results of the various Soil relative density, Dr (%) 53 84
cases show that lf has a limited influence on the pile bend- Soil friction angle, f’ 378 428
ing moment, but a considerable effect on hysteretic energy Soil bulk density, gt (kN/m3) 17 18.2
dissipated. Reference shear wave velocity, Vsr 171 261
(m/s)
Discussion Reference depth, zVs 3d 7.3d
Overall, the results obtained show that soil cave-in and re- Concrete
compression affect the response of SPSI systems in three 28 day target strength (MPa) 34.5 34.5
main ways: (i) it reduces the pile maximum bending mo- Actual compressive strength, fc0 41.3 47.1
ment, (ii) it moves the point of maximum bending moment (MPa)
closer to the ground surface, and (iii) it increases the hy- Longitudinal reinforcement
steretic damping of the soil–pile system. Yield strength, fy (MPa) 421 421
These effects were noted to be more pronounced when Ultimate strength, fu (MPa) 634 634
soil cave-in occurs within the topmost part of the pile. This Young’s modulus, Est (GPa) 190.5 190.5
could be because soil cave-in increases the effective con- Reinforcement ratio, rl (%) 2.1 2.1
finement of the pile, thereby enhancing the performance of Transverse reinforcement
the SPSI system. This could be particularly beneficial in sit- Yield strength, fy (MPa) 710 605
uations where plastic hinges develop below ground level, Ultimate strength, fu (MPa) 793 679
which will be further investigated in the following case Young’s modulus, Est (GPa) 207.2 182
study. Confining steel ratio, rs (%) 0.57 1.06
The different responses obtained for the various cases
Concrete fibre: Constant concrete confinement model*
suggest the need for caution in using BNWF models based
Tensilepstrength,
ffiffiffiffi ft, where ft = 4.82 5.15
only on either a pure gap formation (e.g., FLPIER-D) or on
0.75 fc0 (MPa)
Masing rules (e.g., Kagawa and Kraft 1980) to model the
Strain at peak stress, 3cp 0.002 0.002
cyclic response of soil–pile systems.
Confinement factor, kc 1.32 1.42
Steel fibre: Enhanced Menegotto–Pinto steel model{
Case study II: Full-scale reinforced concrete Post-yield strain-hardening ratio, aps 0.009 0.009
piles in sand experiments Initial transition curve shape factor, 20 20
rmp
Full-scale SPSI experiments were conducted at the Uni-
Transverse curve calibrating coef- 18.5, 0.15 18.5, 0.15
versity of California at Davis for the California Department
ficients (a1, a2)§
of Transportation (Caltrans) in 1997 and 1998. The experi-
Isotropic hardening calibrating coef- 0.25, 2 0.25, 2
ments involved reversed cyclic quasistatic testing of four re- ficients (a3, a4)§
inforced concrete (RC) piles, each with a diameter, d, of
406 mm, and embedded in loose and dense sand beds to a *Based on Mander et al. (1988) with enhancements by Martinez-Rueda
and Elnashai (1997).
depth of 13.5d. The surrounding soil was contained in a {
Confinement factor, kc estimated from transverse reinforcement
cylindrical container with a diameter of 6.7 m and a depth, using equation by Mander et al. (1988).
{
5.5 m. The distance between the pile and container wall was Menegotto–Pinto model (Menegotta and Pinto 1973) with modifica-
set greater than 6d to minimize the influence of boundary tions by Filippou et al. (1983) and SeismoSoft (2003).
§
Based on recommended values by SeismoSoft (2003).
effects on the measured pile response (Park et al. 1987).
The pile head was 2d above ground surface for two piles,
and 6d for the other two. The sand used was classified as case and 18.2 kN/m3 for the 6d case. A summary of relevant
clean, poorly graded sand with 3% fines, with a coefficient soil information is provided in Table 2.
of uniformity, Cu= 4.4, and a median grain size distribution, Table 2 also summarizes the relevant RC pile concrete
D50, & 0.5–0.6 mm. In this report, only two cases were con- and reinforcement details. The 2d and 6d piles comprised
sidered, one with 2d free pile length in loose sand, and an- seven grade A706 No. 22 longitudinal reinforcement bars
other with 6d free pile length in dense sand; these two cases embedded in concrete. Concrete compressive strengths were
will be referred to as 2d and 6d experiments, respectively. 41 and 47.5 MPa for the 2d and 6d piles, respectively. The
The in situ relative density, Dr, was estimated from cone concrete cover for both piles was 50 mm, and the transverse
penetrometer tests (CPT) to be approximately 53% for the reinforcement consisted of a 50 mm pitch continuous spiral
2d experiment and 84% for the 6d experiment. The bulk of MW25 (spiral diameter dsp = 5.4 mm) smooth steel wire
soil density for the two tests was fairly uniform over the for the 2d pile, and the same in MW45 (dsp = 7.3 mm)
depth of the container, and averaged 17 kN/m3 for the 2d smooth steel wire for the 6d pile (Fig. 8a).
# 2008 NRC Canada
Allotey and El Naggar 1275

Fig. 8. Test pile cross section showing reinforcement arrangement: (a) actual experiment; (b) model used.

Fig. 9. Cyclic stress–strain responses at a depth of 1.5d for 6d pile example: (a) concrete fibres C145 and C201; (b) steel bar No. L5.

Fig. 10. Comparison of American Petroleum Institute (API 1993) Table 3. Model parameters on p-y curve for 2d and 6d above-
and Yan and Byrne (1992) p–y curves for sand at selected pile ground bridge bent experiments.
depths for 2d example.
Backbone curve parameters*
American Yan and Byrne Yan and Byrne
Petroleum Institute (2d) Dr = 53% (6d) Dr = 84%
p1 0.55 0.20 0.15
p2 0.86 0.55 0.51
a2 0.54 0.32 0.28
a3 0.13 0.13 0.095
Cyclic curve shape parameters
Case I Case II Case III
L 0 0–5<4.5d 0–5<9d
5>4.5d 5>9d
lf (ls = 1) 0 0–1<4.5d 0–1<9d
1>4.5d 1>9d
Note: L, soil cave in; lf, limiting force; ls=1, curve shape parameter
*Curve parameters defined for a unit p-y curve.

Model description and input parameter estimations


The piles were instrumented to measure longitudinal bar The pile was modeled using 31 beam–column elements,
strain, pile curvatures, axial load, lateral load, and dis- each 0.5d in length. The beam–column elements were mod-
placement. Both piles were subjected to an initial axial eled using the fibre element approach, which is the default
compression of 445 kN. The lateral loading started with beam–column model in SeismoStruct. For a circular RC sec-
load-controlled cycles, and was later changed to tion, only symmetrical arrangements of longitudinal bars
displacement-controlled cycles after substantial softening of around the vertical and horizontal axes of the cross-section
the soil–pile system had occurred. Further information on are allowed. Therefore, as shown in Fig. 8, eight bars were
the tests can be found in Chai and Hutchinson (1999). considered in the analysis to have the same total area as the
# 2008 NRC Canada
1276 Can. Geotech. J. Vol. 45, 2008

Fig. 11. Four segment multilinear fit to: (a) unit American Petroleum Institute (API) p–y curve for sand; (b) Yan and Byrne curve for 2d
example.

Fig. 12. Assumed variations of the shear wave velocity with depth: (a) 2d example; (b) 6d example.


seven steel bars actually used in the experiment. The rele- pus ¼ ðC1 h þ C2 dÞ 0 h if pus  pud
½3b pu ¼
vant steel and concrete fibre parameters are shown in Ta- pud ¼ C3 d 0 h if pud < pus
ble 2; typical concrete and steel cyclic stress–strain
responses for fibres defined in Fig. 8b are also presented in where pu is the ultimate resistance, g’ the effective soil unit
Fig. 9. weight, h the depth, and k the subgrade modulus. Para-
Two types of p–y curves were used in the analysis: the meters C1 through C3 are constants that can be found in the
American Petroleum Institute (API) recommended p–y curve reference, and A is 0.9 for cyclic loading.
for sand (API 1993), and the Yan and Byrne (1992) p–y Yan and Byrne (1992):
curve. Unlike the API curve, Yan and Byrne’s p–y curve is
represented by an ever-increasing power function, with no
defined ultimate pressure. The formulation for the two 
Emax y if y  2 d
curves are presented in eqs. [3] and [4]. ½4 p¼
Emax dðy=dÞ0:5 if y > 2 d
API (1993):
  where a depends on the soil relative density, Dr, and is gi-
khy ven as a = 0.5(Dr)0.8.
½3a p ¼ Apu tanh
Apu Figure 10 shows comparisons between API and Yan and
Byrne p–y curves for the 2d example. Table 3 shows the
backbone curve parameters for the four segment multilinear
fit to both curves. The parameters provided for the Yan and
# 2008 NRC Canada
Allotey and El Naggar 1277

Table 4. Summary of global results for 2d and 6d aboveground bridge bent experiments.

First yield First yield Peak Peak Post-peak


force,* displacement,* force, displacement, Residual force, stiffness, Displacement
Pyi (kN) yi (mm) Pu (kN) u (mm) Pres (kN) Kre (kN/m) ductility, m
2d
Measured 97.0 68.2 124.1 144.1 94.9 142.6 4.1
API (base case) 67.6 51.5 85.2 109.3 36.4 201.9 5.3
Yan linear (base case) 99.8 49.8 124.3 89.4 72.5 198.6 5.5
Yan parabolic (base case) 97.2 57.5 116.4 98.8 70.1 184.3 5.1
Yan parabolic (case II) 97.4 57.9 116.5 99.8 75.4 164.3 4.85
Yan parabolic (case III) 97.2 57.5 116.4 99.2 75.2 164.2 4.88
6d
Measured 43.3 78.3 52.5 137.5 5.5 166.4 4.6
API (base case) 38.7 80.3 45.3 140.6 9.1 130.6 4.5
Yan linear (base case) 44.5 81.2 53.8 139.8 14.5 140.4 4.3
Yan parabolic (base case) 44.1 81.4 54.1 140.1 13.8 143.6 4.3
Yan parabolic (case II) 44.7 82.3 54.0 140.1 14.0 142.8 4.3
Yan parabolic (case III) 44.5 81.8 54.1 140.1 13.8 143.6 4.3
Note: API, American Petroleum Institute
*Chai and Hutchinson (1999)

Table 5. Summary of local results for 2d and 6d aboveground bridge bent experiments.

2d 6d
Maximum Normalized depth Curvature Maximum Normalized depth Curvature
curvature, fmax to maximum ductility, curvature, fmax to maximum ductility,
(mrad/m)* damage, z/d{ mf{ (mrad/m)§ damage, z/d{ mf||
Measured 167 3.4 12.7 114 2.1 8.6
API (base case) 172 4–4.5 13.1 170 2.5–3 12.8
Yan linear (base case) 155 2–2.5 11.8 140 1.5–2 10.5
Yan parabolic base case) 162 3–3.5 12.4 130 1.5–2 9.7
Yan parabolic (case II) 110 1–2 8.4 95 1–1.5 7.1
Yan parabolic (case III) 130 1.5–2 9.9 115 1–1.5 8.6
*Curvature at a displacement ductility of  = 2.8.
{
Results presented in 0.5d ranges since model nodes spaced at 0.5d.
{
Based on the elasto-plastic yield curvature, y = 13.1 mrad/m and moment capacity, Mu = 205 kNm from M– response.
§
Curvature at a displacement ductility of  = 3.1.
||
Based on the elasto-plastic yield curvature, y = 13.6 mrad/m and moment capacity, Mu = 213 kNm from M– response.

Table 6. Total energy dissipated in 2d and 6d aboveground Two shear wave velocity profiles (Fig. 12) were assumed
bridge bent experiments. to represent the soil shear wave velocity distribution. The
first profile, termed the linear profile (Fig. 12a), represents
Energy dissipated (kNm) a linear variation of shear wave velocity following the CPT
2d 6d profile, whereas the second profile, termed the parabolic
Measured 325.5 244.5 profile (Fig. 12b), represents a parabolic variation of shear
API (base case) 237.5 187.9 modulus with depth (power exponent for shear wave veloc-
Yan linear (base case) 413.7 257.9 ity, ne, is 0.25).
Yan parabolic (base case) 362.6 252.8 Table 3 shows the different cases chosen for the cyclic
Yan parabolic (case II) 550.7 305.6 curve shape parameters. Case I, the base case, corresponds
Yan parabolic (case III) 447.1 268.5 to pure gapping. Case II and III correspond to a linear varia-
tion of L and lf with depth up to one and two thirds of the
pile length, respectively. Case II is assumed to represent sig-
Byrne p–y curve are for a unit curve ending at (p, y) = (1, 1) nificant soil cave-in and recompression, whereas case III is a
(Fig. 11); the actual curves were then derived from the unit reduced version of case II. Because the soil around the piles
curve based on a maximum displacement of 0.4d (Yan and was backfilled and compacted, a coefficient of lateral earth
Byrne 1992). In estimating the parameters of the backbone pressure, K0, of 0.8 was used in the analysis.
curve, the first segment is the same as the initial portion of The lateral cyclic loading was applied as a displacement
Yan and Byrne’s curve (i.e., pi and aj are the ith endpoint history, and was derived from the measured load–
and jth segment of the multilinear curve). deformation response. This history comprised three stepped
# 2008 NRC Canada
1278 Can. Geotech. J. Vol. 45, 2008

Fig. 13. Comparison of measured and base case American Petroleum Institute p–y curve lateral load–displacement hysteresis loops for: (a)
2d example; (b) 6d example.

Fig. 14. Comparison of measured and base case of Yan and Byrne p–y curve lateral load versus displacement hysteresis loops for: (a) 2d
example; (b) 6d example.

cycles of increasing displacement, which reached a maxi- Base cyclic response behaviour
mum of 350 and 400 mm for the 2d and 6d examples re- For both the 2d and 6d experiments, limited soil cave-in
pectively. was observed and gaps were noted to extend to significant
depths. For example, a soil gap of 75 mm extending to a
Results and discussion depth of 2.03 m was observed at a lateral force of 66.7 kN
Plastic hinges generally occur in extended pile shafts at (i.e., about half the peak force, Pu) in the 2d experiment.
some distance below the ground level. The ATC-32 (ATC Based on this, the cyclic curve parameter case with no soil
1996) bridge seismic design guideline, Improved seismic de- cave-in (i.e., case I or the base case) should best represent
sign criteria for California bridges: Provisional recommen- the condition of both tests. The results for this case are
dations, rates extended pile shafts as limited ductility therefore compared with the measured cyclic lateral force
structures and assigns them a displacement ductility factor, versus displacement results in Figs. 13, 14, and 15.
mD, of three. Both global and local responses of the SPSI Figure 13 shows that although the initial stiffness was rea-
system play an important role in their design, and are dis- sonably predicted, the capacity, Pu, of the hysteresis loops
cussed below. The results obtained for the 2d and 6d tests computed using the API p–y curves was underpredicted for
are presented in Tables 4–6 and Figs. 13, 14, 15, and 16, both 2d and 6d experiments. This can be attributed to soil’s
and are the averages of results obtained for both loading ability to sustain higher loads at larger displacements than
directions. predicted by the API curves (see Fig. 10). Figure 14, on the
# 2008 NRC Canada
Allotey and El Naggar 1279

Fig. 15. Comparison of measured and predicted longitudinal strain– 6d test, measured depth was 2.1d while predicted was
depth profiles for bar No. L1 for the parobolic case: (a) 2d exam- 1.5–2d.
ple; (b) 6d example.
Effect of cyclic p–y parameter variation
Based on the good agreement obtained between the meas-
ured and predicted results for the parabolic profile, the influ-
ence of soil cave-in and compression on the computed
response was assessed by comparing the results of the three
cases (i.e., cases I, II, and III; see Table 3) for this profile.
From Fig. 16, a considerable increase in the energy dissipa-
tion is noted for cases II and III although the results of the
global response show only a slight increase in the post-peak
residual load (Table 6). Also, the results of the local re-
sponse presented in Table 5 show a significant reduction in
the maximum curvature and corresponding curvature ductil-
ity. A study of the curvature profile showed curvatures
spread over a larger length of the pile, thus effectively de-
creasing the maximum curvature at a given displacement
ductility. In other words, soil cave-in helps increase the ef-
fective length of the plastic hinge.

Discussion
The results obtained underscore the beneficial effect soil
cave-in and recompression have on the SPSI system. It helps
increase the effective confinement of the pile and spreads
the curvature demand, thereby minimizing localized curva-
ture and increasing the effective length of the plastic hinge.
This assertion is partly corroborated by the results of full-
scale drilled shaft experiments by Wallace et al. (2001),
who noted that spalled-off concrete increased the effective
soil–pile frictional resistance.
This benefit may not be significant for undamaged piles;
however, for damaged piles, the extra confinement provided
to the damaged zone by the cave-in soil could contribute
significantly to the satisfactory performance of the SPSI sys-
tem. In the final preparation of this article, the authors’ at-
tention was drawn to two sets of large-scale cyclic lateral
load tests conducted in California by Juirnarongrit and
Ashford (2003) and Budek et al. (2004) on cast-in drilled
hole (CIDH) piles with diameters ranging from 0.4 to
1.8 m. Both studies report that the surrounding soil around
other hand, shows that the predictions obtained with the damaged portion of the pile was observed to provide sig-
Yan and Byrne’s p–y curves were generally satisfactory, nificant external confinement. In fact, Budek et al. (2004),
with the parabolic profile giving better results (Table 4). quote the following in their conclusions:
As further confirmation, Table 6 shows that the total
The results of these tests suggest that soil confinement
energy dissipated (given by the area enclosed by the can play a very significant role in pile shaft response.
loops) for the base case with a parabolic profile The confining pressure provided by the soil can signifi-
matched closely the measured results for both the 2d cantly increase the effective confinement of the section
and 6d examples. and retard development of high levels of localized plastic
Figure 15 shows a comparison between the predicted rotation, thus providing a sizable increase in ductility
longitudinal strains for bar No. L1 (refer to Fig. 8 for capacity through the formation of an elongated hinge
bar location) based on the parabolic profile and the meas- region.
ured longitudinal strains. For both tests, the predicted This conclusion validates the observations made from the
depth to maximum damage, z, compared favourably with current study, and further verifies the ability of the proposed
the measured value, i.e., for the 2d test, measured depth approach to simulate these phenomena. Currently, no ra-
was 3.3d while predicted was 3.0d; and for the 6d test, tional approach exists to account for the confining effect of
measured depth was 1.25d while predicted was 1.5d. soil in design, and more efforts are required to adequately
From Table 5, the predicted depth to maximum curvature quantify this effect in order to help facilitate its incorpora-
for both tests was also found to be in good agreement tion into design. This is particularly relevant considering re-
with the measured values, i.e., for the 2d test, measured cent changes in design philosophy to allow for foundation
depth was 3.4d while predicted was 3–3.5d; and for the yielding (Martin and Lam 2000).
# 2008 NRC Canada
1280 Can. Geotech. J. Vol. 45, 2008

Fig. 16. Comparison of cases I, II, and III lateral load–displacement hysteresis loops for parabolic profile based on Yan and Byrne’s p–y
curve: (a) 2d example; (b) 6d example.

Conclusions Canadian Geotechnical Journal, 45(4): 560–573. doi:10.1139/


T07-106.
This paper employed the BNWF model developed by Allotey, N.K., and Foschi, R.O. 2005. Coupled p-y t-z analysis of
Allotey and El Naggar (2008) to conduct a numerical assess- single piles in cohesionless soil under vertical and/or horizontal
ment of the effects of gapping and soil cave-in and recom- ground motion. Journal of Earthquake Engineering, 9: 755–775.
pression on the lateral cyclic response of soil–pile systems. doi:10.1142/S1363246905002316.
Two case studies of cyclic lateral load tests for single piles API. 1993. Recommended practice for planning designing and con-
representing different levels of incurred pile damage were structing fixed offshore platforms – working stress design. API
examined. From the results of the study, it was noted that RP2A-WSD. American Petroleum Institute, Washington, D.C.
the effect of soil cave-in and recompression is to decrease ATC. 1996. ATC-32- Improved seismic design criteria for Califor-
pile maximum moment, move its point of occurrence closer nia bridges: Provisional recommendations. Applied Technology
to ground surface, and increase hysteretic energy dissipation. Council, Redwood City, Calif.
For damaged piles that develop plastic hinges at some depth Bhushan, K., Fong, P.T., and Haley, S.C. 1979. Lateral load tests
below ground, soil cave-in could be beneficial as it helps in- on drilled piers in stiff clay. Journal of Geotechnical Engineer-
crease the effective confinement of the plastic hinge. This in ing, 105(8): 969–985.
Boulanger, R.W., Curras, J.C., Kutter, B.L., Wilson, D.W., and
turn spreads the curvature demand, checking the develop-
Abghari, A. 1999. Seismic soil–pile–structure interaction experi-
ment of high levels of localized curvature and resulting in
ments and analyses. Journal of Geotechnical and Geoenviron-
an increase in the effective length of the plastic hinge. mental Engineering, 125(9): 750–759. doi:10.1061/(ASCE)
These observations have been corroborated by experimental 1090-0241(1999)125:9(750).
results and point to the need to account for the confining Brown, D.A., O’Neill, M.W., Hoit, M., McVay, M., El Naggar,
effect of soil in design. M.H., and Chakraborty, S. 2001. Static and dynamic lateral
loading of pile groups. NCHRP Report 461. National Coopera-
tive Highway Research Board, Transportation Research Board,
Acknowledgements National Research Council, Washington, D.C.
The authors would like to express their deepest gratitude Budek, A.M., Priestley, M.J.N., and Benzoni, G. 2004. The effect
to Dr. Stelios Antoniou of SeismoSoft for implementing the of external confinement on the flexural hinging in drilled pile
BNWF model in SeismoStruct, and to Dr. Rui Pinho of the shafts. Earthquake Spectra, 20: 1–24. doi:10.1193/1.1647579.
Rose School at University of Pavia, Italy, for reviewing the Carter, D.P. 1984. A nonlinear soil model for predicting lateral pile
BNWF model report. The authors would also like to express response. M.Sc. thesis, University of Auckland, Auckland, New
their appreciation to Professors Rob Chai and Tara Zealand.
Hutchinson of the University of California at Davis, and at Chai, Y.H., and Hutchinson, T.C. 1999. Flexural strength and
Irvine, respectively, for making the data on the extended ductility of reinforced concrete bridge piles. Report No. UCD/
bridge shaft experiments available. STR-99/02. University of California, Davis, Calif.
Dou, H., and Byrne, P.M. 1996. Dynamic response of single piles
and soil–pile interaction. Canadian Geotechnical Journal, 33(1):
References 80–96. doi:10.1139/t96-025.
Allotey, N.K. 2006. Nonlinear soil–structure interaction in Elgamal, A., Yang, Z., and Parra, E. 2002. Computational model-
performance-based design. Ph.D. thesis, University of Western ing of cyclic mobility and post-liquefaction site response. Soil
Ontario, London, Ont. Dynamics and Earthquake Engineering, 22: 259–271. doi:10.
Allotey, N.K., and El Naggar, M.H. 2008. Generalized dynamic 1016/S0267-7261(02)00022-2.
Winkler model for nonlinear soil–-structure interaction analysis. El Naggar, M.H., and Allotey, N.K. 2007. Seismic soil-structure in-

# 2008 NRC Canada


Allotey and El Naggar 1281

teraction in performance-based design. In Proceedings of the 6th and bending. In Proceedings of Symposium on the Resistance and
Alexandria Conference on Structural and Geotechnical Engi- Ultimate Deformability of Structures Acted on by Well-Defined
neering, Alexandria University, Alexandria, Egypt. 15-17 April Repeated Loads. International Association for Bridge and Struc-
2007. Invited keynote lecture. tural Engineering, Zurich, Switzerland. pp. 15–22.
El Naggar, M.H., and Bentley, K.J. 2000. Dynamic analysis for lat- Meymand, P.J. 1998. Shaking table scale model tests of nonlinear
erally loaded piles and dynamic p–y curves. Canadian Geotech- soil–pile–superstructure interaction in soft clay. Ph.D. thesis,
nical Journal, 37(6): 1166–1183. doi:10.1139/cgj-37-6-1166. University of California, Berkeley, Calif.
Filippou, F.C., Popov, E.P., and Bertero, V.V. 1983. Modeling of Park, R.J.T., Priestley, M.J.N., and Berrill, J.B. 1987. Seismic per-
R/C joints under cyclic excitations. Journal of Structural Engi- formance of steel-encased concrete piles. Research Report 87–5.
neering, 109(11): 2666–2684. Department of Civil Engineering, University of Canterbury,
Gazetas, G., and Mylonakis, G. 1998. Seismic soil-structure inter- Christchurch, New Zealand.
action: new evidence and emerging issues. In Geotechnical Pender, M.J., and Pranjoto, S. 1996. Gapping effects during cyclic
Earthquake Engineering and Soil Dynamics-III. Edited by P. lateral loading of piles in clay. In Proceedings of 11th World Con-
Dakoulas, M. Yegian, and R.D. Holtz. ASCE Geotechnical ference on Earthquake Engineering, Acapulco, Mexico, 23–28
Special Publication No. 75, ASCE publications, Reston, Va. June 1996. Paper No. 1007.
pp. 1119–1174. Pranjoto, S., and Pender, M.J. 2003. Gapping effects on the lateral
Gerolymos, N., and Gazetas, G. 2005. Phenomenological model ap- stiffness of piles in cohesive soil. In Proceedings of Pacific Con-
plied to inelastic response of soil–pile interaction systems. Soils ference on Earthquake Engineering, Christchurch, New Zealand,
and Foundations, 45: 119–132. 13–15 February 2003. Paper No. 096.
Juirnarongrit, T., and Ashford, S.A. 2003. Effect of soil confine- Pyke, R. 1979. Nonlinear soil models for irregular cyclic loadings.
ment in enhancing inelastic behaviour of cast-in-drilled-hole Journal of Geotechnical Engineering, 105(GT6): 715–726.
piles. In Proceedings of 16th ASCE Engineering Mechanics SeismoSoft. 2003. Seismostruct – A computer program for static
Conference, Seattle, Wash. 16–18 July 2003. Paper No. 718. and dynamic nonlinear analysis of framed structures. Available
Kagawa, T., and Kraft, L.M. Jr. 1980. Lateral load-deflection rela- from www.seismosoft.com.
tionships of piles subjected to dynamic loading. Soils and Foun- Swane, I.C., and Poulos, H.G. 1984. Shakedown analysis of a
dations, 20(4): 19–36. laterally loaded pile tested in stiff clay. In Proceedings of 4th
Liam Finn, W.D.L. 2005. A study of piles during earthquakes: Australia–New Zealand Conference on Geomechanics, Perth,
issues of design and analysis. Bulletin of Earthquake Engineer- Australia, 14–18 May 1984. Paper No. C1545. pp. 275–279.
ing, 3: 141–234. doi:10.1007/s10518-005-1241-3.Ling, L.F. Thavaraj, T. 2001. Seismic analysis of pile foundations for bridges.
1988. Back analysis of lateral load tests on piles. M.Sc. thesis, Ph.D. thesis, University of British Columbia, Vancouver, BC.
University of Auckland, Auckland, New Zealand. Vesic, A. 1961. Beams on elastic foundations. In Proceedings of
Long, J.H., and Vanneste, G. 1994. Effects of cyclic lateral loads 5th International Conference on Soil Mechanics and Foundation
on piles in sands. Journal of Geotechnical Engineering, 120(1): Engineering, Paris. Vol. 1, pp. 845–850.
225–243. doi:10.1061/(ASCE)0733-9410(1994)120:1(225). Vucetic, M. 1990. Normalized behavior of clay under irregular
Mander, J.B., Priestley, M.J.N., and Park, R. 1988. Theoretical loading. Canadian Geotechnical Journal, 27(1): 29–46. doi:10.
stress–strain model for confined concrete. Journal of Structural 1139/t90-004.
Engineering, 114: 1804–1826. Wallace, J.W., Fox, P., Stewart, J.P., Janoyan, K., Qiu, T., and
Martin, G.R., and Lam, I.G. 2000. Earthquake resistant design of Lermitte, S.P. 2001. Cyclic large deflection testing of shaft
foundations- retrofit of existing foundations. In Proceedings of bridges: Part I- Background and field test results. Research Re-
GeoEng 2000 – International Conference in Geotechnical and port to California Department of Transportation, University of
Geological Engineering, Melbourne, Australia. 19–24 November California, Los Angeles, Calif.
2000. Technomic Publishing Company, Lancaster, Penn. Yan, L., and Byrne, P.M. 1992. Lateral pile response to monotonic
pp. 1025–1047. loads. Canadian Geotechnical Journal, 29(2): 955–970. doi:10.
Martı́nez-Rueda, J.E., and Elnashai, A.S. 1997. Confined con- 1139/t92-106.
crete model under cyclic load. Materials and Structures, Yang, Z., and Jeremic, B. 2002. Numerical analysis of pile beha-
30(3): 139–147. doi:10.1007/BF02486385.
viour under lateral loads in layered elastic-plastic soils. Interna-
Matlock, H. 1970. Correlations for design of laterally loaded piles
tional Journal for Numerical and Analytical Methods in
in soft clay. In Proceedings of 2nd Offshore Technology Confer-
Geomechanics, 26: 1385–1406. doi:10.1002/nag.250.
ence (OTC), Houston, Texas, 22–24 April 1970. Paper
Yang, Z., Elgamal, A., and Parra, E. 2003. Computational model
No. 1204. pp. 577–594.
for cyclic mobility and associated shear deformation. Journal of
Menegotto, M., and Pinto, P.E. 1973. Method of analysis for cycli-
Geotechnical and Geoenvironmental Engineering, 129(12):
cally loaded R.C. plane frames including changes in geometry and
1119–1127. doi:10.1061/(ASCE)1090-0241(2003)129:12(1119).
non-elastic behaviour of elements under combined normal force

# 2008 NRC Canada

Vous aimerez peut-être aussi