Behavior_of_Pile_Foundations_in_Laterally_Spreadin
Behavior_of_Pile_Foundations_in_Laterally_Spreadin
net/publication/238179918
CITATIONS READS
223 690
4 authors, including:
All content following this page was uploaded by Bruce L. Kutter on 18 May 2015.
Peer Reviewed
Title:
Behavior of Pile Foundations in Laterally Spreading Ground during Centrifuge Tests
Author:
Brandenberg, Scott J, University of California - Los Angeles
Boulanger, R W, University of California - Davis
Kutter, Bruce L, University of California - Davis
Chang, Dongdong, University of California - Davis
Publication Date:
11-01-2005
Publication Info:
Postprints, Multi-Campus
Permalink:
http://escholarship.org/uc/item/1mt502cj
Original Citation:
Brandenberg, S.J., Boulanger, R.W., Kutter, B.L., and Chang, D. (2005). "Behavior of pile
foundations in laterally spreading ground during centrifuge tests." J. Geotech. Geoenviron. Eng.
131(11). 1378-1391.
Published Web Location:
http://ascelibrary.aip.org/vsearch/servlet/VerityServlet?
KEY=JGGEFK&smode=strresults&sort=rel&maxdisp=25&threshold=0&pjournals=IJGNAI
%2CJAEEEZ%2CJAEIED%2CJBENF2%2CJCRGEI
%2CJCCOF2%2CJCCEE5%2CJCEMD4%2CJLEED9%2CJENMDT%2CJOEEDU
%2CJGGEFK%2CJHEND8%2CJHYEFF%2CJITSE4%2CJIDEDH%2CJMENEA
%2CJMCEE7%2CJPCFEV%2CJPEPE3%2CJSENDH%2CJSUED2%2CJTPEDI%2CJUPDDM
%2CJWRMD5%2CJWPED5%2CLMEEAZ%2CNHREFO%2CPPHMF8%2CPPSCFX
%2CASCECP&possible1=brandenberg&possible1zone=article&OUTLOG=NO&viewabs=JGGEFK&key=DISPL
Keywords:
Pile foundations, Pile groups, Earthquakes, Liquefaction, Lateral loads, Soil deformation,
Centrifuge
Abstract:
Eight dynamic model tests were performed on a 9 m radius centrifuge to study the behavior of
single piles and pile groups in liquefiable and laterally spreading ground. Pile diameters ranged
from 0.36 to 1.45 m for single piles, and from 0.73 to 1.17 m for pile groups. The soil profile
consisted of a gently sloping nonliquefied crust over liquefiable loose sand over dense sand.
Each model was tested with a series of realistic earthquake motions with peak base accelerations
ranging from 0.13 g to 1.00 g. Representative data that characterize the important aspects of soil–
pile interaction in liquefiable ground are presented. Dynamic soil–pile and soil–pile cap forces are
backcalculated. Directions of lateral loading from the different soil layers are shown to depend
on the mode of pile deflection relative to the soil, which depends on the deformed shape of the
Abstract: Eight dynamic model tests were performed on a 9 m radius centrifuge to study the behavior of single piles and pile groups in
liquefiable and laterally spreading ground. Pile diameters ranged from 0.36 to 1.45 m for single piles, and from 0.73 to 1.17 m for pile
groups. The soil profile consisted of a gently sloping nonliquefied crust over liquefiable loose sand over dense sand. Each model was
tested with a series of realistic earthquake motions with peak base accelerations ranging from 0.13 g to 1.00 g. Representative data that
characterize the important aspects of soil–pile interaction in liquefiable ground are presented. Dynamic soil–pile and soil–pile cap forces
are backcalculated. Directions of lateral loading from the different soil layers are shown to depend on the mode of pile deflection relative
to the soil, which depends on the deformed shape of the soil profile, the pile foundation stiffness, and the magnitude of loads imposed by
the nonliquefied crust. Procedures for estimating the total horizontal loads on embedded piles and pile caps 共i.e., passive loads plus friction
along the base and sides兲 are evaluated. Due to liquefaction of the sand layer beneath the crust, the relative displacement between the pile
cap and free-field crust required to mobilize the peak horizontal loads is much larger than expected based on static pile cap load tests in
nonliquefied soils.
DOI: 10.1061/共ASCE兲1090-0241共2005兲131:11共1378兲
CE Database subject headings: Pile foundations; Pile groups; Earthquakes; Liquefaction; Lateral loads; Soil deformation;
Centrifuge.
Introduction ior in liquefiable level ground from centrifuge model tests. Ash-
ford and Rollins 共2002兲 developed cyclic p – y relations from lat-
Extensive damage to pile-supported bridges and other structures eral load tests of piles in blast-induced liquefied soil. Tokimatsu et
in areas of liquefaction and lateral spreading has been observed in al. 共2004兲 characterized p – y relations in liquefiable soil during
many earthquakes around the world 关e.g., Japanese Geotechnical full-scale shaking table tests. Peak subgrade reaction values in
Society 共JGS兲 1996, 1998兴. Many important lessons and insights liquefiable sand were estimated from centrifuge tests by Abdoun
have been learned from case histories, physical model tests, and et al. 共2003兲 and Dobry et al. 共2003兲. Differences in the subgrade
numerical studies in recent years, but numerous questions still reaction behavior observed in the above studies are consistent
remain regarding the basic mechanisms of soil–pile interaction in with the effects of relative density, pile stiffness, dynamic shaking
liquefiable soil and laterally spreading ground. Model studies can characteristics, and site response. For example, relatively small
begin to clarify some of these questions, and can aid in the subgrade reaction loads were observed in loose sand, while larger
development of reliable design methods. loads were observed in medium dense sand.
A review of physical model studies and lessons from case
Past model tests of piles in liquefiable ground have resulted in
histories showed a need for characterization of soil–pile interac-
varying observations of foundation response. Wilson et al. 共1998,
tion in liquefiable ground in which lateral spreading occurred.
2000兲 presented the first dynamic characterization of p – y behav-
Hence, a series of dynamic centrifuge model tests were conducted
1
on the 9 m radius centrifuge to study the behavior of pile foun-
Graduate Student, Dept. of Civil and Environmental Engineering, dations in liquefiable and laterally spreading ground, and obser-
Univ. of California, Davis, CA 95616 共corresponding author兲. E-mail:
vations and results of the tests are presented in this paper. Single
[email protected]
2
Professor, Dept. of Civil and Environmental Engineering, Univ. of piles and pile groups were embedded in a soil profile consisting of
California, Davis, CA 95616. a gently sloping nonliquefiable crust over liquefiable loose sand
3
Professor, Dept. of Civil and Environmental Engineering, Univ. of over dense sand. A series of realistic earthquake motions were
California, Davis, CA 95616. applied to the base of each of the models. Important quantities
4
Graduate Student, Dept. of Civil and Environmental Engineering, obtained by processing the raw recorded data, including backcal-
Univ. of California, Davis, CA 95616. culated subgrade reaction behavior, are presented. Phasing of the
Note. Discussion open until April 1, 2006. Separate discussions must subgrade reaction loads from the nonliquefied crust and liquefi-
be submitted for individual papers. To extend the closing date by one able sand for different pile foundations is discussed, and factors
month, a written request must be filed with the ASCE Managing Editor.
that influence the different observations are explained. Theoretical
The manuscript for this paper was submitted for review and possible
publication on March 1, 2004; approved on March 18, 2005. This paper is predictions of the lateral loads imposed on the pile foundation by
part of the Journal of Geotechnical and Geoenvironmental Engineer- the laterally spreading crust 共including passive loads on the
ing, Vol. 131, No. 11, November 1, 2005. ©ASCE, ISSN 1090-0241/ upslope face, and friction between the crust and the sides and
2005/11-1378–1391/$25.00. base of the cap兲 are compared with measured ultimate loads
Soil Properties
Fig. 1 shows schematic model layouts for the first centrifuge test
PDS01 and for test SJB03, which is similar to all seven of the
tests after PDS01. The soil properties for four of the tests are
summarized in Table 1. The soil profile for all of the models
consisted of a nonliquefiable crust overlying loose sand
共Dr ⬇ 21– 35% 兲 overlying dense sand 共Dr ⬇ 69– 83% 兲. All of the
layers sloped gently toward a river channel carved in the crust at
one end of the model. The nonliquefiable crust consisted of re-
constituted Bay mud 共liquid limit⬇ 88, plasticity index⬇ 48兲 that
was mechanically consolidated with a large hydraulic press, and
Fig. 1. Centrifuge model layouts with most of approximately 100 subsequently carved to the desired slope. The sand layers beneath
transducers per model omitted for clarity the crust consisted of uniformly graded Nevada Sand 共Cu = 1.5,
D50 = 0.15 mm兲. A thin layer of coarse Monterey sand was placed
on the surface of the Bay mud for some of the models. Water was
共Fcrust,ult兲. Finally, the relative displacements between the free- used as a pore fluid for all of the models.
field soil and the pile cap required to mobilize the ultimate loads The undrained shear strength of the clay, su, for each test was
are compared with those for static pile cap load tests in nonlique- measured using a T-bar 共Stewart and Randolph 1991兲. The su
fied soil profiles, and a load transfer relationship representing profile from the T-bar tests was reasonably consistent with the su
soil–pile cap interaction for a crust overlying liquefiable sand is profile estimated using a normalized strength ratio 共e.g., Ladd and
presented. Foott 1974兲 according to
for every test except PDS03 and the cap inertia was therefore
estimated for PDS03, as described later.
The calculated crust load includes the passive resistance on the
Fig. 2. Free-body diagram of pile cap and pile segments above shear upslope face of the pile cap, friction on the sides and base of the
gages cap, and loads on the pile segments between the shear gauges and
the bottom of the cap as shown in Fig. 2
Fcrust = F p + F2 + F4 + 2 · Fs + 2 · Fc + 2 · Fn 共5兲
noise levels in the low-frequency portion of the signals containing
the permanent displacement data. The crust displacement was These components of horizontal soil loading could not be calcu-
therefore calculated by adding together the low frequency com- lated separately based on the test data, therefore estimates of the
ponent from the displacement transducers and the high frequency various components were made analytically as described later in
component from the double-integrated accelerations. The perma- the paper. Active loading on the downslope face of the cap was
nent component of the deeper sand displacements was not mea- excluded from the free-body diagram because large gaps formed
sured by transducers, but was estimated based on measurements along this face when the soil spread laterally downslope, and are
of the soil deformation profile taken when the model was dis- believed to have remained open during the large earthquake mo-
sected after all of the shaking events. All displacement values tions. However, during the small earthquake motions, the gap was
were calculated relative to the base. small and may have closed during upslope displacement cycles.
The lateral loading mechanism in Fig. 2 assumes that the clay
Relative Displacement between Piles and Soil beneath the pile cap flows around the piles, thereby mobilizing
lateral loads on the pile segments and friction loads at any contact
The subgrade reaction is related to the relative displacement be- between the crust and the base of the cap. However, there is
tween the piles and the soil 共y = y pile-y soil兲. The calculated time another possible mechanism in which the clay crust beneath the
histories of y were difficult to accurately determine because they pile caps becomes trapped between the piles, thereby acting as an
contained measurement errors that were attributed primarily to equivalent block. In such cases, the passive force and side friction
assumptions regarding the boundary conditions required to calcu- forces would act along the entire thickness of the nonliquefied
late y pile and the low frequency part of y soil. The results do, how- crust layer, while the lateral loads on the pile segments and base
ever, permit qualitative observations of the relative displacement of the pile cap would be considered internal forces 共i.e., not ex-
profiles at snapshots in time. Hence, time histories are presented ternal to a free body of the equivalent block兲. The controlling
for p, but not for y, while snapshots of the displaced shapes of the mechanism is that which produces the smaller total lateral load.
soil and piles are presented for critical loading cycles. The pile groups in the centrifuge tests all exhibited the failure
mechanism in Fig. 2 in which the clay crust flowed around the
Lateral Loads on Pile Caps pile segments, which was verified both by photos taken after the
tests during excavation of the models and by comparison of the-
Strain gages oriented in Wheatstone full bridges to measure shear oretical predictions of the total lateral loads for each mechanism.
共hereafter called shear gages兲 located on the piles near the inter-
face between the loose sand and clay provided a measure of the
shear forces, Vs, Vn, and Vc, as shown in the free-body diagram in Accuracy of Numerical Differentiation
Fig. 2. These shear forces had contributions from loads imposed The shear gages on the piles provide a means of verifying the
on the foundation by the clay crust and from pile cap inertia. The accuracy of the shear and p histories obtained from differentiation
loading imposed on the pile group by the crust, Fcrust, was calcu- of the moment distributions. Fig. 3 shows two shear 共V兲 histories,
lated as the difference between the total shear and the cap inertia one measured directly by the shear gauges and the other from
force differentiation of the moment distributions 共V = dM / dz兲. The two
time histories exhibit similar dynamic characteristics, similar
Fcrust = 共2 · Vs + 2 · Vn + 2 · Vc兲 + ah · mcap 共4兲
peak loads, and both show nonzero initial offsets that represent
where the cap inertia force⫽−ah · mcap. A simple superstructure residual loads remaining from previous earthquakes. The reason-
attached to the pile caps for some of the tests contributed an able agreement between the time histories indicates that the data
additional inertia component that was also subtracted out to ob- processing methods and differentiation technique are sufficient to
tain Fcrust for those tests. The pile cap acceleration was measured produce reasonably accurate first derivatives of moment with
depth. Most of the p histories are believed to be accurate with the 2. The pile cap inertia force was 5,790 kN, which was the maxi-
exception of zones near soil layer interfaces or the pile head or mum for the test.
tip, though an independent check of the accuracy of the p histo- 3. The subgrade reaction, p, 6.7 m below the ground surface
ries 共p = d2M / dz2兲 could not be obtained from the measured data. was −370 kN/ m, which was a local minimum. Note that the
subgrade reaction was negative; the loose sand restrained the
pile from moving downslope.
Observations from SJB03 4. The excess pore pressure ratio was 0.5, near a local mini-
mum, in spite of having been close to 1.0 earlier in the mo-
Observations from centrifuge Test SJB03 are presented in detail tion, and returning to near 1.0 later in the motion.
in this section and comparisons with other centrifuge tests are 5. The displacement of the pile cap was 0.5 m, and was ap-
made in later sections. A series of four simulated earthquake proaching a local maximum that was larger than the perma-
events were applied to Model SJB03. The first was a small Santa nent pile cap displacement of about 0.3 m at the end of
Cruz motion with a peak acceleration of 0.13g. Soil deformation, shaking.
excess pore pressure ratio, pile cap displacement, and moments 6. The displacement of the clay crust was 2.3 m, which was less
were all relatively small during the small Santa Cruz motion, than the permanent crust displacement of nearly 3.5 m.
as summarized in Table 2. The subsequent motions were a The relative displacement between the crust and the pile cap was
medium Santa Cruz 共amax,base = 0.35g兲, a large Santa Cruz
共amax,base = 0.67g兲, and a large Kobe 共amax,base = 0.67g兲. The loading
mechanics are most clearly illustrated during the large motions, so
observations will be presented in reverse chronological order be-
ginning with the large Kobe motion, then the large Santa Cruz
motion, and finally the medium Santa Cruz motion.
Table 3. Summary of Measurements from Pile Foundations in PDS01 for First Large Kobe Motion
Peak Peak
incremental incremental Peak
Pile soil surface pile crust
Pile diameter displacement displacement Yielding of load pl
foundation 共m兲 共m兲a 共m兲a pile 共kN兲b 共kN/ m兲c
BP 1.45 0.99 0.29 None 550 90
GN 0.73 0.55 0.48 None 280 −60
MP 0.73 0.55 0.55 None 280 −190
SP 0.36 0.99 1.29 Extensive n.ad n.a.d
a
Displacements continued to accumulate through the earthquake sequence. Incremental displacements are relative to the displacements existing before
each earthquake event.
b
Estimated from p histories at the middle of the clay layer, and extrapolated to boundaries using theoretical models for variation of p with depth.
c
Subgrade reaction mobilized in middle of liquefied sand at time of peak moment.
d
Not available.
冉 冊
ated against the recorded ultimate values. The recorded lateral
loads on the pile foundations were measured using shear gages on v⬘ z
Np = 3 + +J for z 艋 zcr 共7兲
the piles a short distance beneath the bottom of the pile cap. c b
Hence, the peak measured lateral loads include pile cap inertia,
the passive force on the upslope pile cap face, the friction forces
Np = 9 for z ⬎ zcr 共8兲
on the sides and base of the pile cap, and the lateral forces on the
pile segments between the pile cap base and the shear gages 共Fig. where N p⫽bearing factor; J⫽empirical constant, assumed to be
2兲. For comparison purposes the results are presented in terms of 0.5 in this study; and z⫽depth below the ground surface. Eq. 共7兲
the peak lateral crust loads imposed on the pile group 共i.e., controls at shallow depths, and is based on a wedge-type failure
the peak crust load on the pile cap is the measured load in mechanism in which the failure mass exhibits vertical deforma-
the shear gages minus the inertia forces of the pile cap and any tion, while Eq. 共8兲 is based on a plane-strain solution in which the
superstructure兲. clay flows around the pile without vertical deformation.
Matlock 共1970兲 found that displacement-controlled cyclic
loading of piles in clay reduced the p – y resistance compared to
Passive Earth Pressure on Pile Cap
that for static monotonic loading. The p – y curves for cyclic load-
The peak passive load imposed on the upslope face of each pile ing had a peak resistance that was 72% of the ultimate resistance
cap was estimated using a Coulomb-based analysis. The non- for static loading, followed by further reductions in resistance that
liquefied crust in some tests consisted of Monterey sand overlying depended on the depth and magnitude of y. These “cyclic load-
clay, so the Coulomb method was modified to account for soil ing” p – y curves were derived for loading conditions that are
layering as shown in Fig. 12. For the surface Monterey sand layer, more representative of wave or wind loading, whereas the loading
the passive force, F p,sand, and the force on the failure wedge, R, conditions encountered during lateral spreading are very different.
were assumed to be the same as for a pile cap with height Hsand in Matlock’s “cyclic loading” p – y curves substantially underesti-
a uniform sand layer. For SJB03 the layer thicknesses were mate the peak lateral loads imposed by a laterally spreading crust
Hsand = 0.8 m and Hclay = 1.4 m 共Fig. 12兲. The properties for the 共Boulanger et al. 2003兲, as will be illustrated later for the single
clay are in Table 1, and properties for the Monterey sand were piles in PDS01.
estimated to be ␥ = 17 kN/ m3, ⬘ = 36°, ␦ = 22°, and c⬘ = 0. The Matlock’s bearing equations were developed for single piles in
remaining forces in Fig. 12 are solved from force equilibrium. clay with a free ground surface such that a failure wedge on the
The inertia of the soil failure wedge was also included as Okabe passive side of the pile could exhibit some upward vertical dis-
placement at shallow depths. If a gap exists between the base of a Total Lateral Crust Loads on Single Piles
pile cap and the underlying clay, then failure wedges against the The peak measured p values against piles BP and GN near the
piles can move upwards. In this case, the depth z in Matlock’s center of the clay layer are summarized in Table 5, along with
equations would be best taken as relative to the base of the pile static pult values predicted using Matlock’s bearing factors. This
cap. If the base of a pile cap is in contact with the underlying clay, comparison of measured and calculated pult values is made for the
then the clay may be restrained to fail in plain strain against the center of the clay where the differentiation of the recorded bend-
piles, producing larger bearing factors 关i.e., Eq. 共8兲 versus Eq. 共7兲兴 ing moment distribution is most accurate 共i.e., away from inter-
Backcalculated values for the peak subgrade reaction loads on the faces and boundaries兲. Values are not compared for MP because
pile segments at mid-depth in the clay beneath the pile cap for electronic drift in bending moment recordings in the crust layer
SJB03 at the time of peak moment 共Fig. 7兲 ranged from about 200 induced errors in the second derivatives in the crust. The pre-
to about 450 kN/ m, which corresponds to quite reasonable bear- dicted values are close to the measured peak values, which indi-
ing factors of N p = 3.9– 8.7, respectively. The accuracy of these cate that Matlock’s static bearing factors are appropriate to use for
backcalculated values is limited by the small number of strain piles in laterally spreading clays at shallow depths. The p values
gages in the clay layer beneath the cap, and so the comparison predicted using Matlock’s 共1970兲 cyclic loading p – y curves are
with Matlock’s bearing equations really only represents a quali- less than 25% of the static pult values 共Table 5兲, and would have
tative check on the consistency of the data. Consequently, in the resulted in a significant underprediction of pile displacements and
absence of a suitable equation for bearing factors on piles beneath bending moments for these tests.
pile caps, the loads on the pile segments beneath the caps were
estimated using Matlock’s equations with the depth z being rela- Crust Load versus Relative Displacement between
tive to the adjacent ground surface, which gives N p values inter- Free-Field Soil and Pile Cap
mediate to those calculated for plane strain and for a gap beneath
The centrifuge test data provide a means of assessing the load
the cap.
transfer relation for a laterally spreading crust over liquefiable
soil. Fig. 13 shows crust load versus relative displacement at vir-
Total Lateral Crust Loads on Pile Caps gin loading peaks 共i.e., crust load peaks that exceed the maximum
past crust load兲 from tests with a six-pile group. The recorded
The total lateral load on the pile cap was calculated for each test, loads were normalized by the ultimate load measured for each
and Rbase was adjusted so the predictions agreed reasonably with test. Relative displacement, which was the soil displacement to
the measured ultimate crust loads. The total loads and the indi-
vidual components estimated from the analysis are presented in
Table 4. The passive force exerted on the upslope face of the pile Table 5. Predicted and Peak Measured Lateral Crust Loads on Single
cap ranged from 43 to 50% of the total crust load, with pile Piles
segment forces and side and base friction accounting for the re- Predicted pult Predicted pult
maining fraction. Neglecting friction forces in design would result Peak with static with cyclic
in a substantial underprediction of the peak loads that were mea- measured loading loading
sured during the test, which would be unconservative. The base p relationsa relationsa
friction reduction factor, Rbase, ranged from 0 to 1 / 2, with an Pile 共kN/ m兲 共kN/ m兲 共kN/ m兲
average value of 1 / 4. The range in Rbase is consistent with obser- BP 169 175 36
vations of the contact between the crust and base of the pile cap GN 82 90 22
during model excavation, with more gaps being observed for the a
Using Matlock’s 共1970兲 p – y relations for static or cyclic loading as
tests with low backcalculated Rbase values. appropriate.
Fcrust
Fcrust,ult
= 冋冉 冊 冉 冊 册
y
C·H
−0.33
+
16 · y
C·H
−1 −1
艋1 共9兲