You are on page 1of 11

ARTICLE IN PRESS

Soil Dynamics and Earthquake Engineering 29 (2009) 300–310

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering


journal homepage: www.elsevier.com/locate/soildyn

Nonlinear seismic soil–pile–structure interactions:


Shaking table tests and FEM analyses
K.T. Chau a,, C.Y. Shen b, X. Guo c
a
Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hong Kong, China
b
Earthquake Engineering Research Test Center, Guangzhou University, Guangzhou, China
c
Institute of Engineering Mechanics, Harbin, China

a r t i c l e in f o a b s t r a c t

Article history: In this paper, a soil–pile–structure model is tested on a shaking table subject to both a sinusoidal wave
Received 7 November 2007 and the acceleration time history of the scaled 1940 El Centro earthquake. A medium-size river sand is
Received in revised form compacted into a 1.7-m-high laminar rectangular tank to form a loose fill with a relative density of 15%.
15 February 2008
A single-storey steel structure of 2.54 ton is placed on a concrete pile cap, which is connected to the four
Accepted 26 February 2008
end-bearing piles. A very distinct pounding phenomenon between soil and pile is observed; and, the
acceleration response of the pile cap can be three times larger than that of the structural response. The
Keywords: pounding is due to the development of a gap separation between soil and pile, and the extraordinary
Soil–pile–structure interaction large inertia force suffered at the top of the pile also induces cracking in the pile. To explain this
Shake table
observed phenomenon, nonlinear finite element method (FEM) analyses with a nonlinear gap element
Finite element analyses
have been carried out. The spikes in the acceleration response of the pile cap caused by pounding can be
modeled adequately by the FEM analyses. The present results suggest that one of the probable causes of
pile damages is due to seismic pounding between the laterally compressed soil and the pile near the pile
cap level.
& 2008 Elsevier Ltd. All rights reserved.

1. Introduction pile damages need to be further examined. For comprehensive


reviews on theoretical soil–pile–structure interaction (SPSI), we
Many big cities in the world are built on flat lands containing a refer to Meymand [2] and Novak [7]. The SPSI problem has also
thick layer of sediment, such as basins, river deltas, or valleys. Tall been investigated by using the shaking table test [2,8] and the
buildings or important structures in these cities have to be centrifuge test [9,10]. The present study continues the line of
founded on piles to avoid excessive ground settlements. In works on shaking table tests on the soil–pile–structure system.
addition to static load transferred from the dead load of the The main focus of the present study is to report a newly
structures, piles are also subject to dynamic loads. The most observed phenomenon in our shaking table tests—pounding
commonly encountered dynamic loads on a pile–soil–structure between soil and pile when a soil–pile–structure model is subject
system are those due to earthquakes. Past earthquake events to seismic excitations. When the soil–pile–structure model is
demonstrate that damages in piles are commonly induced during subject to seismic excitations, the soil surrounding the pile may be
moderate to strong earthquakes. Mizuno [1] compiled the earth- compressed laterally such that a soil–pile gap separation may
quake-induced damages of piles reported in Japan from 1923 to develop. Consequently, pounding may appear between soils and
1983, including those of the great Kanto earthquake. Damages in piles due to the different dynamic responses of the pile–structure
pile have been observed during the 1964 Niigata earthquake, the system and the soil. We will show that this pounding may lead to
1964 Alaska earthquake, the 1985 Mexico City earthquake, and a very large inertia force at the pile cap level, which may lead
the 1989 Loma Prieta earthquake [2]. More recently, severe to cracking in the foundation piles. Finite element analysis is
damages in piles were also reported during the 1995 Kobe used to explain the unusual large acceleration suffered at the pile
earthquake [3–6]. cap level.
The remedial works needed for damaged piles can be very Although soil–pile gaps have been observed in the field after
costly. Thus, pile–soil–structure interaction and mechanism for earthquakes and in shaking table tests after soil–pile–structure
models are subject to seismic excitations, the pounding between
soil and pile has not been recognized and examined. Photographs
 Corresponding author. Tel.: +852 27666015; fax: +852 23346389. in Fig. 1 show soil–pile gap separations observed in the field and
E-mail address: cektchau@polyu.edu.hk (K.T. Chau). in the laboratory. For soil–pile gaps observed in the field, Figs. 1(a)

0267-7261/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2008.02.004
ARTICLE IN PRESS

K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310 301

and (b) show two photographs of soil–pile gaps observed on cause of pile damages observed in the field especially when no
the reclaimed Port Island after the 1995 Kobe earthquake, while liquefaction is observed around the damaged piles [6].
Figs. 1(c) and (d) show two photographs of soil–pile gaps
developed along the Struve Slough Crossing during the 1989
Loma Prieta earthquake. For soil–pile gaps observed in shaking 2. Experimental set-up
table tests, Figs. 1(e) and (f) are reproduced from Fig. 8 of Wei et
al. [11] and Fig. 8.31 of Meymand [2], respectively. There has been Experiments on a model of a soil–pile–structure system were
no previous attempt to investigate the possibility of pounding performed on an MTS uniaxial seismic shaking table of size
between soil and pile induced by these gaps. Therefore, shaking 3 m  3 m. Fig. 2(a) shows a photograph of the experimental set-
table tests demonstrating the pounding between soil and pile will up, while schematic diagrams of the laminated soil tank and the
be presented here. cross-sections of the members used in constructing the pile, the
To simulate the free field response of soil, various soil tank columns of the structure, and the soil tank are shown in Figs. 2(b)
designs have been proposed to minimize the boundary effect of and (c), respectively. A maximum horizontal acceleration of 1g can
the finite soil tank. They include rigid tank with sufficiently large be applied at the full load of 10 ton. The working frequency of the
size [12,13], rigid tank packed with foam at the sides of the tank table ranges from 1 to 50 Hz. The shaking table can simulate
[14–17], laminated soil tank [18–27], soil tank with walls having a motions with displacement, velocity or acceleration control. The
hinged-base [28], and flexible circular container [2,29,30]. There is displacement control is primarily for low frequency range,
no conclusion on which particular soil tank system is better than velocity control for middle frequency range, and acceleration
others. For this study, a rectangular laminated tank system was control for high frequency range. The maximum overturning
selected. The results of the present study provide a new potential moment that can be restrained by the bearing of the table is

Fig. 1. Photographs of soil–pile gap observed in the field and laboratory: (a,b) piles in the reclaimed island of Kobe during the 1995 Kobe earthquake (after Refs. [2,5]); (c,d)
piles for the Struve Slough Crossing during the 1989 Loma Prieta earthquake ((c) after [2] and (d) photographed by H. G. Wilshire); (e) gap observed in shaking table test
(after [11]); and (f) gap observed in shaking table test (after [2]).

1200
1260

100

Pile

90
Laminated
H=1700

Column
90

soil tank

50
50

Soil tank

Unit: mm
1500
Unit: mm

Fig. 2. (a) A photograph of the soil–pile–structure system used; (b) an elevated side view of the system showing the sizes and the laminated soil tank; and (c) cross-sections
used for piles and columns, and for constructing the laminar frame of the tank.
ARTICLE IN PRESS

302 K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310

10 ton-m. In our experiment, the total weight of our soil–pile– two 1-mm-thick layers of Teflon (or polytetrafluoroethylene) are
structure system is close to the limit of 10 ton. glued to the top and bottom of each steel frame section. The
To simulate the shaking of soil in the free field, a rectangular frictional coefficient between two Teflon surfaces is found equal to
laminated soil tank is constructed by stacking up 32 laminar 0.126. The height of the soil tank is about 1.7 m. To ensure the
rectangular steel frames made by welding four rectangular hollow overall sliding stability, the bottom frame section is welded to a
sections of 50 mm  50 mm  2.8 mm together (see Fig. 2(c)). The bottom steel plate of 12 mm thickness, on which another three-
laminar rectangular frame has an internal size of 1.4 m  0.9 m. To dimensional (3-D) steel box frame with cross bracing is con-
reduce the friction and allow sliding between adjacent frames, structed to limit the maximum translation of the laminated soil

M
as
1040

structure

ap
a2
200

204
209
Sand D

213
800
1700

1700

213
a1
Pile

367
φ 100

437
board a0

Shaking table
Laminated tank
Ground

Shaking table
100

Soil tank
3# 4#
150

Pile cap C5 C6 C7 C8
S5 S6 S7 S8
1000
500

Pile
1# 2#
C1 C2 C3 C4
S1 S2 S3 S4
150

sand
100

150 200 800 200 150


1500

: displacement transducer : pile

: accelerometer : sand
: strain gauge

Fig. 3. A vertical cross-section showing the piles and soil within the laminated tank with a horizontal cross-section cut at the bottom level of the pile cap. Locations of the
strain gauges for steel bars (S1–S8) and for concrete surfaces (C1–C8), the nine displacement transducers, and the five accelerometers are also showed.
ARTICLE IN PRESS

K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310 303

tank, as shown in Fig. 2(a). This box frame also provides a The soil used is poorly graded river sand imported from the
reference for measuring the horizontal distance of the soil tank at Mainland China. The D10, D30, D50, and D60 of the sand are about
various levels (see Fig. 3). 0.22, 0.34, 0.41, and 0.42 mm, respectively, with all particles
Four concrete piles are cast independently using a template of smaller than 1 mm. The specific gravity of the sand is 2.662. The
PVC pipe of 100 mm in diameter. The piles are 1.7 m long, which fine content (i.e. particles smaller than 0.063 mm) is less than
gives a slenderness ratio of the piles as 17. The Young’s modulus, 0.1%. Therefore, it can be considered as pure sand, 35% coarse
Poisson ratio and the 28-day cube strength of the concrete are (0.6–2 mm), 61% medium (0.2–0.6 mm) and 4% fine sand
21.8 GPa, 0.18, and 19.55 MPa, respectively. The reinforcement in (0.06–0.2 mm). From the results of seven triaxial tests, the Young’s
the pile consists of eight vertical mild steel bars of diameter 6 mm modulus of the soil is estimated to range from 0.375 to about
(or a steel ratio of 2.8%). Circular stirrups made of 9 mm mild steel 5 MPa depending on the confining stress and the strain level.
with a spacing of 20 mm are fixed to the vertical bars. Eight strain Before the sand is packed into the laminated soil tank, an
gauges were attached to the vertical steel bars near the top of the expansible water-resistant nylon bag is custom made to fit the
piles, shown as S1–S8 in Fig. 3, with a pair of strain gauges is size of the soil tank to prevent the loss of soil particles through the
installed on the surfaces of each pile along the shaking direction. joints of the laminated tank. A total of 3622 kg of sand is used to
After the concrete is cast, eight more strain gauges were installed fill the laminated tank in 11 layers. When each sub-layer is filled,
to the surface of the concrete piles, again on the surface of each an electric hammer of 9.55 kg is used to compact the soil to a
pile along the shaking direction (shown as C1–C8 in Fig. 3). The specific thickness of about 154 mm. The hammer is the Kango
spacing of the piles along the shaking direction is 800 mm Type 628 Light Demolition Hammer, and the base of the hammer
while the spacing along the transverse direction is 500 mm. is a flat disk of 145 mm diameter. The overall density of the soil is
Therefore, since the spacing for in-line shaking piles is larger 1.459 Mg/m3. The fill is relatively loose and should resemble the
than six diameters of the pile, the pile–pile interaction can be condition of loose hydraulic fill found in the reclaimed areas of
neglected [16]. The piles are fixed to the bottom steel plate by a Hong Kong. The water content of the sand in the laminated shear
wooden template of 10 mm thickness when soil is put into the tank is about 4%, therefore, it can be basically considered dry.
laminated soil tank. Thus, the piles can be considered as hinged Thus, no capillary stress needs to be considered, and the sandy soil
end-bearing piles. can be considered as cohesionless.
After the soil is filled, a concrete pile cap for all four piles is
cast. The size of the pile cap is 1200 mm (length)  800 mm
(width)  200 mm (thickness). Mild steel bars of 4 mm diameter
Table 1
Test program of the shaking table tests, where Amax is the peak acceleration.
are placed at 20 mm spacing along both the shaking direction and
its orthogonal direction, and at both the top and bottom of the
Experiment number Test type Wave form Amax (g) Frequency (Hz) cap. The pile cap can, therefore, be considered as rigid.
A single-storey structure made of steel frame is then attached
E1 Sweep Sinusoidal 0.02 1–10
to the pile cap by four tie-down bolts. Two steel plates of
E2 Random Random 0.02 1–10
E3 (61) Spectrum Sinusoidal 0.02 1–10 1200 mm (length)  800 mm (width)  20 mm (thickness) are
E4 Small EQ Sinusoidal 0.04 4.7 used as the base and top of the frame structure. Four columns
E5 Random Random 0.02 1–10 of hollow square section of 90 mm  90 mm  5.5 mm are welded
E6 (29) Spectrum Sinusoidal 0.02 1–10
to both the upper and lower plates (see Fig. 2(c)). The Young’s
E7 Small EQ Sinusoidal 0.05 4.4
E8 (27) Spectrum Sinusoidal 0.02 1–10 modulus, Poisson ratio and yield strength of the mild steel are
E9 Small EQ Sinusoidal 0.06 4.3 206 GPa, 0.28, and 215 MPa, respectively. Additional mass of 2 ton
E10 Random Random 0.02 1–10 is added to the top plate of the structures (see Fig. 2(a)) and the
E11 Moderate EQ Sinusoidal 0.116 4.3 total mass of the structure is 2358 kg.
E12 Random Random 0.02 1–10
Five accelerometers are installed at various locations of the
E13 Moderate EQ El Centro 0.135
E14 Random Random 0.02 1–10 soil–pile–structure systems, shown as triangles in Fig. 3. The
E15 Random Random 0.02 1–15 accelerometers a0, a1, a2, ap and as are installed at the surface of
E16 Moderate EQ Sinusoidal 0.134 3.2 the shaking table, at the soil tank at an elevation of 804 mm above
E17 Random Random 0.02 1–15
the table, at the top level of the soil tank, on the pile cap, and at
E18 Large EQ Sinusoidal 0.2 3.2
E19 Random Random 0.02 1–15
the top steel plate of the structure. Another nine displacement
transducers are installed at various levels, as shown in Fig. 3.

10 10 10
as/a0
8 ap/a0 8 8
Acceleration ratio

6 6 6

4 4 4

2 2 2

0 0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Frequency (Hz) Frequency (Hz) Frequency (Hz)

Fig. 4. The acceleration response spectra at the pile cap and structure at various test stages: tests (a) E3, (b) E6, and (c) E8.
ARTICLE IN PRESS

304 K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310

3. Shaking table test program Therefore, a grand total of 133 shaking table experiments was
conducted on the soil–pile–structure system. For checking
Table 1 lists a total of 19 sets of shaking table tests carried out purposes, additional tests have also been carried out for the
on the soil–pile–structure system (i.e. E1–E19). The test type, the structures, the soil in the laminated tank and the pile–soil
waveform of the acceleration, the peak acceleration of the input, systems.
and the frequency range of these tests are given in Table 1. Both
the frequency sweep and random vibrations are for the determi-
nation of the fundamental frequency of the soil–pile–structure 4. General observations of the shaking table tests
system. The spectrum test is to find the acceleration response
spectrum of the system. The acceleration time history of the 1940 4.1. Cracking in the concrete piles
El Centro earthquake is also used as input (E13 in Table 1), but the
peak acceleration has been scaled to 0.135g to avoid excessive In general, strain gauge data at the surface of the concrete piles
overturning moment. Sinusoidal waves have been used for the indicate that cracking occurs at the top of the pile when Amax is
frequency sweep (E1), all spectrum tests (E3, E6, and E8) and all increased to 0.116g in test E11 (with a tensile strain of 0.06%).
earthquake tests with peak ground acceleration larger than 0.02g When Amax ¼ 0.134 (i.e. E16), the tensile strain on the concrete
(E4, E7, E9, E11, E13, E16, and E18). For the spectrum tests, the surface is up to 0.15%. And, when Amax ¼ 0.2, the tensile strain in
numbers in the brackets behind E3, E6, and E8 in Table 1 are concrete is about 0.35%.
the actual number of tests used in obtaining the spectrum. For the
earthquake tests, the magnitudes of the applied sinusoidal waves
increase gradually from 0.02 to 0.2g, and the applied frequency is 4.2. Deflections of the laminated soil tank
the current fundamental frequency of the system. The natural
frequency of the soil–pile–structure system is checked after each The maximum deflections at the top of the soil tank are 1, 2,
earthquake test. The durations for all sine wave inputs are 20 s and 2.5, and 4 mm when the acceleration a2 is 0.035, 0.064, 0.087, and
those for random tests are 90 s. 0.116g, respectively. The deflection profile of the laminated tank

2 2
acce. (m/s2)

acce. (m/s2)

1 1
0 0
8 8.5 9
-1 0 2 4 6 8 10
-1
-2 ap -2 ap
time (s) as time (s)
as

10 10
acce. (m/s2)

5
acce. (m/s2)

5
0 0
0 5 10 15 8 8.5 9
-5 -5
ap ap
-10 as -10 time (s) as
time (s)

10 10
acce. (m/s2)
acce. (m/s2)

5 5
0 0
0 5 10 15 10 10.5 11
-5 -5
-10 ap -10 ap
time (s) time (s)
as as

20 20
acce. (m/s2)

acce. (m/s2)

10 10
0 0
0 5 10 15 10 10.5 11
-10 -10
ap ap
-20 time (s) as -20 time (s) as

Fig. 5. The acceleration response spectra at the pile cap (dotted lines) and structure (solid lines) at various levels of ground input (Amax): (a) E3 sinusoidal wave input,
f ¼ 4.8 Hz, Amax ¼ 0.02g; (b) E7 sinusoidal wave input, f ¼ 4.4 Hz, Amax ¼ 0.05g; (c) E11 sinusoidal wave input, f ¼ 4.3 Hz, Amax ¼ 0.116g; and (d) E18 sinusoidal wave input,
f ¼ 3.2 Hz, Amax ¼ 0.2g. The chosen frequencies at various stages correspond to the updated natural frequency of the soil–pile–structure system, which changes with time.
The enlargement is plotted to illustrate the differences in the acceleration response of the pile cap and the structure.
ARTICLE IN PRESS

K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310 305

recorded from the displacement transducers verifies that only laminated tank, displacement, velocity, and acceleration transdu-
mode one shaking is induced in the soil. cers are installed along the various levels of the tank such that the
natural period can be measured by inputting a random wave with a
peak acceleration of less than 0.02g. More specifically, it can be
4.3. Dynamic characteristics of different components identified by noting the frequency at the peak of the Fourier spectra
plot. The equivalent shear wave speed and dynamic shear modulus
Different components of the system are also tested indepen- of the sand are 78 m/s and 8.865 MPa, respectively. This equivalent
dently using the shaking table. In particular, the fundamental shear modulus appears to be larger than the static elastic modulus
frequencies of the structure with the additional mass, and the soil measured in triaxial tests reported in Section 2. This increase can
with laminated tank are, respectively, 7.6 and 11.5 Hz (or with be attributed to the fact that dynamic modulus is normally larger
natural periods of 0.13 and 0.087 s). For the case of only soil in than the static one, and that the friction between the laminates of

15 15

10 10
acceleration (m/s2)

5 5

0 0
-9 -6 -3 0 3 6 9 -9 -6 -3 0 3 6 9
-5 -5

-10 -10

-15 -15

15 15
acceleration (m/s2)

10 10

5 5

0 0
-9 -6 -3 0 3 6 9 -9 -6 -3 0 3 6 9
-5 -5

-10 -10

-15 -15
displacement (mm) displacement (mm)

Fig. 6. The phase diagrams of acceleration versus displacement at the pile cap for those enlargements shown in Fig. 5.

2 1940 El Centro earthquake


acce. (m/s2)

0
0 5 10 15 20
-2

3
acce. (m/s2)

-1 0 5 10 15 20

-3 ap

as
-5 time (s)

Fig. 7. (a) The acceleration time history of the 1940 El Centro earthquake, and (b) acceleration responses of the system at pile cap and at structure level (test E13 in Table 1).
The peak ground acceleration of the El Centro wave has been scaled down to 1.35 m/s2.
ARTICLE IN PRESS

306 K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310

the soil tank inevitably exists. Note that the predominant frequency subject to shaking of acceleration larger than 0.04g, out-of-phase
of the structure (7.6 Hz) is lower than that of the soil (11.5 Hz), and motions are observed between the structure and the soil tanks.
thus, radiation damping of the piles of the system is negligible [31]. More importantly, the fundamental frequency recorded at the top
In addition, accelerometers have also been attached to the edge of of the soil tank increases from 4.6 to 11.6 Hz, which agrees with
the structure along the direction perpendicular to the shaking the fundamental frequency of the soil and the laminated tank
direction, and the shaking table responses are again recorded. The alone (without pile and structure) reported earlier. Therefore, it is
acceleration records for these two channels are essentially the clear that the soil and laminated tank vibrate independently
same, and this shows that no torsional or rotational motion has from the pile–structure system. This again supports our earlier
been triggered in our shaking table tests. speculation that the soil has compressed laterally and a gap is
developed between soil and pile.

5. Experimental results and discussion


5.2. Model subject to the acceleration time history of the 1940 El
Centro earthquake
5.1. Response spectra of the system at various stages

All previous experiments are carried out for the soil–pile–


Figs. 4(a) to (c) plot the acceleration ratios at the pile cap and at
structure system subject to sinusoidal waves. Fig. 7 shows plots of
the structure (i.e. ap/a0 and as/a0) versus the excitation frequency of
the acceleration time histories of the 1940 El Centro earthquake
a sinusoidal wave of 0.02g in magnitude. Fig. 4(a) shows that the
and the acceleration responses of the pile cap and structure. Note
initial natural frequency of the system is about 4.9 Hz (from E3). As
that the peak acceleration of the 1940 El Centro earthquake has
expected, the acceleration response of the structure is larger than
been scaled down to 0.135g to avoid excess overturning moment
that of the pile cap. After a sinusoidal wave of magnitude 0.04g (E4)
exerted on the table. Fig. 7 clearly shows that the pile cap
is applied to the system, Fig. 4(b) shows that the natural frequency
response is again larger than the response at the structure.
drops to about 4.5 Hz. More importantly, the acceleration response
at pile cap is larger than that of the structure, which is somewhat
unexpected. After a sinusoidal wave of magnitude 0.05g (E7) was
applied to the system, Fig. 4(c) shows that the natural frequency 6. Finite element analyses
further drops to 4.3 Hz. Again, the acceleration response at pile cap
is larger than that of the structure. In order to demonstrate what has happened in the system that
To further examine this phenomenon, Figs. 5(a) to (d) plot the leads to a larger acceleration response at the pile cap than at the
acceleration at the pile cap ap (dotted lines) and at the structure as structure, the finite element method (FEM) has been applied to
(solid lines) versus time. To make a better comparison, enlarge-
ments of each of the acceleration responses are also given in
Figs. 5(a) to (d). For experiment E3 (or Amax ¼ 0.02g), the response
of the pile cap is clearly less than that of the structure. In addition, δ
the response of the structure resembles the input sinusoidal k
wave more closely than that of the pile cap. For experiment E7
(or Amax ¼ 0.05g), Fig. 5(b) shows clearly that a number of spikes Gap Element
is observed in the acceleration response of the pile cap, comparing Structure
to the roughly sinusoidal response of the structure. The maximum
acceleration at ap is about the double of that of as. For experiment
E11 (or Amax ¼ 0.116g), Fig. 5(c) shows that the acceleration at the
pile cap is about three times of that of the structure. From E7 to
E11, although the input acceleration is more than double, the Pile cap
structural response remains roughly constant at about 3 m/s2. For
the largest earthquake wave input used in E18 (or Amax ¼ 0.2g),
the conclusion is again similar to those obtained in Figs. 5(b) and
(c), except that the difference between the acceleration of the pile
L

cap and the structure increase further.


The spikes observed in the acceleration responses of the pile
cap suggest that impact or pounding may occur between the pile
and the soil. To further verify this, Figs. 6(a) to (d) plot the phase Sand
H

h1

diagram of acceleration versus displacement for those ‘‘one-


second responses’’ of pile cap shown in Figs. 5(a) to (d). It is clear
that there is a jump of the acceleration of the pile cap at a fixed
h2

displacement; and this is a unique feature of pounding or impact.


Our speculation is that after strong ground motions are input,
repeated dynamic contacts between the soil and the piles lead to a
lateral compression of the soil. Thus, a gap between the soil and
the pile is developed. As remarked in the Introduction, gap
between soil and pile has been observed in other shaking table
tests and in the field after some major earthquakes (see Fig. 1). To Fig. 8. The two-dimensional finite element model used in SAP 2000 Nonlinear. The
further examine this phenomenon, detailed analyses of the soil and the additional masses on the top of the structure are modeled by four-
Fourier spectra at the top of the structure, pile cap, and the soil node plane strain elements, while all piles, pile cap, and structures are modeled by
frame elements. The gap element used between pile and soil from the top of the
tank were conducted. We found that when the system is subject
soil to L meter below the soil surface is enlarged. The gap element is characterized
to shaking of acceleration less than 0.02g, all displacements by a stiffness k and a gap distance d. Although it is not shown explicitly in the
(i.e. a2, ap as, and shown in Fig. 3) are in phase; when the system is diagram, gap elements are installed on both sides of the two piles.
ARTICLE IN PRESS

K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310 307

simulate the responses of the system and the results are divided into 15 layers. In order to model the depth of the gap L
compared to the experimental observation. Initially, the program more accurately, the top 12 layers are of smaller thickness h1
FLUSH was used, but the results were not satisfactory even when a while the bottom three layers are of larger thickness h2. In our
gap was artificially inserted between the soil and the pile. actual input, we have set h1 ¼ 0.08 m and h2 ¼ 0.24 m. Although it
Apparently, the pounding between the soil and the pile has not is not shown explicitly in Fig. 8, the gap elements have been
been modeled appropriately. Subsequently, ‘‘SAP 2000 Nonlinear’’ installed on both sides of each of the two piles.
is used to model the gap development [32]. In particular, the ‘‘Gap Fig. 9 plots the acceleration time histories at the pile cap and the
element’’ in the NLLink properties is selected to model the structure versus time for test E9, together with three FEM
pounding between the soil and the pile. The nonlinear force– simulations. The input frequency in the FEM analyses is 5 Hz which
deformation relation for the Gap element is given by is about the same as that applied in test E9 (i.e. 4.3 Hz). The actual
( stiffness of contact between the soil and the pile (the so-called
kðd þ dÞ if d þ do0; dynamic sub-grade reaction) has not been measured in the present
f ¼ (1)
0 otherwise; study. The shaking table tests by Yahata et al. [18] suggest that the
sub-grade reaction can be very complicated, in contrast to an
where k is the spring constant (or contact stiffness), d denotes increasing trend assumed in static case [33]. Therefore, the triangle,
deflection of the Gap element, and d is the initial gap separation rectangle, and inverse triangle shown in Figs. 9(b), (c) and
(dX0). Note that d40 for opening mode and do0 for closing (d) respectively, indicate that a linear increasing, a constant, or a
mode of the gap. linear decreasing trend of the contact stiffness k has been assumed
A sketch for the FEM modeling is shown in Fig. 8. An equivalent in the FEM analysis. The maximum stiffnesses k used in Figs. 9(b),
2-D model is proposed. The structure and pile foundation are (c) and (d) are 0.1, 0.05, and 0.08 GN/m, respectively. In these
modeled by beam elements, while the additional mass on the analyses, the depth of the gap is assumed as L/H ¼ 6/21, where H is
structure as well as the soil is modeled by four-node plane strain the total thickness of the soil shown in Fig. 8. This assumed value,
solid elements. As shown in the enlargement, the characteristics however, cannot be verified since no pressure transducers have
of the Gap element are controlled by two input parameters: the been installed along the piles. In the calculations of Fig. 9, the gap
contact stiffness k and the gap separation d. The 1.7 m soil is separation has been assumed to vary from 0.2 mm from the top of

12 12
Experiment E9
6 6
acce. (m/s2)

acce. (m/s2)

0 0
0 5 10 15 12.1 12.6 13.1
-6 -6
pilecap pilecap
-12 time (s) structure -12 time (s) structure

12 12
FEM:
6 6
acce. (m/s2)
acce. (m/s2)

0 0
0 5 10 15 12.1 12.6 13.1
-6 -6
pilecap
pilecap
-12 time (s) -12 time (s) structure
structure

12 12
FEM:
6 6
acce. (m/s2)

acce. (m/s2)

0 0
0 5 10 15 12.1 12.6 13.1
-6 -6
pilecap pilecap
-12 time (s) structure -12 time (s) structure

12 12
FEM:
6 6
acce. (m/s2)

acce. (m/s2)

0 0
05 10 15 12.1 12.6 13.1
-6 -6
pilecap pilecap
-12 time (s) structure -12 time (s) structure

Fig. 9. (a) The experimentally observed acceleration time histories at ap and as for test E9 in Table 1 (f ¼ 4.3 Hz, Amax ¼ 0.06g); (b) FEM results by assuming a linear
variation of the stiffness k with depth; (c) FEM results by assuming a constant stiffness k with depth; and (d) FEM results by assuming a linear decreasing stiffness k with
depth. For all FEM results, we have assumed L/H ¼ 6/21, and that the gap distance varies linearly from 0.2 mm at the top to zero at the base of the gap. The maximum
stiffnesses kmax for (b), (c), and (d) are 0.1, 0.05, and 0.0857 GN/m, respectively.
ARTICLE IN PRESS

308 K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310

the gap to 0 at the bottom of the gap (i.e. L below the soil surface as for Figs. 10(b), (c), (d), (e), (f) and (g). In particular, the maximum
shown in Fig. 8). The acceleration time histories of the pile cap gap separations at the top dmax used are 0.08, 0.2 and 0.35 mm for
shown in Figs. 9(b) and (d) agree well with the experimental results cases of L/H ¼ 4/21, 6/21 and 9/21, respectively. For cases of
given in Fig. 9(a). More importantly, the FEM results are capable of increasing k, the maximum stiffnesses k used in Figs. 10(b), (c) and
capturing the essential feature that the pile cap acceleration (d) are 4, 0.0857 and 0.06 GN/m respectively; whereas, for cases of
response is larger than the structural response. In conclusion, the decreasing k, the maximum stiffnesses k used in Figs. 10(e), (f) and
abnormally large pile cap response is caused by pounding, which is (g) are 3, 0.1 and 0.08 GN/m, respectively. Fig. 10 shows that the
simulated by installing gap elements between soil and the pile in number of impacts between the soil and the pile per each cycle of
our FEM analyses. Thus, this verifies our speculation of pounding the structural response increases with the depth L for both cases
between the soil and pile. Although uncertainties exist in assigning of decreasing and increasing contact stiffnesses. Among these
the values of sub-grade reaction and the depth of the gap, the results figures, Fig. 10(c) appears to model the experiment better.
given in Fig. 9 clearly demonstrate that the spikes in acceleration are In short, although there is uncertainty regarding the assumed
the results of pounding between the soil and the pile. dynamic sub-grade reaction, our results of the FEM analyses are
Since the depth of the gap has not been observed, Fig. 10 shows not sensitive to the actual distribution of the stiffness k. Therefore,
the acceleration response at the pile cap and at the structure our analyses should reliably reflect what has happened in the
versus time for various depths of gap L (L/H ¼ 4/21, 6/21 and shaking table tests.
9/21). In Figs. 10(b), (c) and (d), a linear decreasing k with depth
was assumed, whereas, in Figs. 10(e), (f) and (g) a linear increasing
k with depth was assumed. Note that Figs. 10(c) and (f) are 7. Conclusion
identical to Figs. 9(d) and (b), respectively. Again, a linear gap
separation has been assumed, but because of the changing depth In this paper, a soil–pile–structure model is tested on a shaking
of the gap, different maximum gap separation has been assigned table subject both sinusoidal wave of various magnitudes and

Experimental result (E9)


12 (b) L/H=4/21
acce. (m/s )

6
2

0
12.112.613.1
12.1 12.6 13.1
-6 pilecap
pilecap

-12 time(s)
time (s) structure
structure

12 12
L/H = 4/21 L/H = 4/21
acce. (m/s )

6 6
acce. (m/s )
2

0 0
12.1 12.6 13.1 12.1 12.6 13.1
-6 -6
pilecap pilecap

-12 time (s) structure -12 time (s) structure

12 12
L/H = 6/21 L/H = 6/21
acce. (m/s2)

6 6
acce. (m/s2)

0 0
12.1 12.6 13.1 12.1 12.6 13.1
-6 -6
pilecap pilecap

-12 time (s) structure -12 time (s) structure

20 L/H = 9/21 12
L/H = 9/21
acce. (m/s2)

12 6
acce. (m/s )
2

4
0
-412.112.613.1 12.1 12.6 13.1
-12 -6
pilecap pilecap

-20 time (s) structure -12 time (s) structure

Fig. 10. FEM results by assuming the stiffness of the gap elements increases or decreases linearly with depth. (a) The experimentally observed acceleration time histories at
ap and as for test E9 in Table 1; (b) FEM results by assuming L/H ¼ 4/21, dmax ¼ 0.08 mm, and kmax ¼ 4 GN/m; (c) FEM results by assuming L/H ¼ 6/21, dmax ¼ 0.2 mm,
and kmax ¼ 0.0857 GN/m; (d) FEM results by assuming L/H ¼ 9/21, dmax ¼ 0.35 mm, and kmax ¼ 0.06 GN/m; (e) FEM results by assuming L/H ¼ 4/21, dmax ¼ 0.08 mm, and
kmax ¼ 3 GN/m; (f) FEM results by assuming L/H ¼ 6/21, dmax ¼ 0.2 mm, and kmax ¼ 0.1 GN/m; and (g) FEM results by assuming L/H ¼ 9/21, dmax ¼ 0.35 mm, and
kmax ¼ 0.08 GN/m. Linear decreasing and linear increasing are denoted by inverted triangle and triangle, respectively. The gap distance d is linearly decreasing with depth,
with zero value at the base of the gap.
ARTICLE IN PRESS

K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310 309

frequencies and the acceleration time history of the 1940 El Acknowledgments


Centro earthquake. The soil is medium-graded river sand and
is embedded into a 1.7 m laminated tank made of steel This research was by a grant from the Research Grants Council
sections covered with Teflon to reduce the sliding friction. of Hong Kong Special Administrative Region, China (Project no.
A single-storey steel structure is placed on a concrete pile cap, PolyU5035/02E) and fully supported by the Hong Kong Poly-
which is supported by four end-bearing concrete piles. A very technic University through the ASD project of ‘‘Seismic and
distinct but unexpected phenomenon is observed. After the Landslide Hazards in Dense Urban Areas’’ (A226). The shaking
system has been subjected to a number of sinusoidal waves of tests were inspired by the discussions with Professors John Berrill
moderate magnitude (say peak acceleration larger than 0.05g), and Rob Davis when K.T. Chau visited the shaking table facility at
the acceleration response at the pile cap level may increase to University of Canterbury, New Zealand in 1996. Technical support
three times of that of the structural response (e.g. see Fig. 5). In from Mr. T.T. Wai and Mr. C.F. Cheung is acknowledged.
addition, the acceleration response of the pile cap shows spikes of
large acceleration response, which resemble that of nonlinear
pounding between two systems. This observation seems to References
contradict the prediction by structural dynamics. However, a
closer look of the model after the test reveals that gaps develop [1] Mizuno H. Pile damage during earthquake in Japan (1923–1983). In: Nogami
T, editor. Dynamic responses of pile foundations—experiment, analysis and
between soil and piles, probably due to the lateral soil compres- observation. Geotechnical Special Publication no. 11. ASCE; 1987. p. 53–78.
sion induced by the prolonged shaking of moderate earthquake [2] Meymand PJ. Shaking table scale model test of nonlinear soil–pile–super-
waves input to the system. Finite element analyses were structure interaction in soft clay. PhD dissertation, University of California,
Berkeley, 1998.
carried out using ‘‘SAP 2000 Nonlinear’’ incorporating a nonlinear [3] Matsui T, Oda K. Foundation damage of structures. Spec Issue Soils Found
gap element installed along the top one-third of the soil–pile 1996:189–200.
interface. Various contact stiffness and gap separations have [4] Tokimatsu K, Mizuno H, Kakurai M. Building damage associated with
geotechnical problems. Spec Issue Soils Found 1996:219–34.
been assumed and it turns out that as long as appropriate [5] Horikoshi K, Tateishi A, Ohtsu H. Detailed investigation of piles damaged by
magnitudes of the stiffness, depth of the gap, and initial gap Hyogoken Nambu earthquake. In: Proceedings of the 12th world congress on
separation are assumed, the spikes in the acceleration response of earthquake engineering 2000, Paper no. 2477 (in CD-ROM).
[6] Luo X, Murono Y. Seismic analysis of pile foundations damaged in the
the pile cap can be modeled quite adequately. The results appear January 17, 1995 South-Hyogo earthquake by using the seismic deformation
to be insensitive to the depth variation of stiffness of the gap method. In: Proceedings of the fourth international conference on recent
element. advances in geotechnical earthquake engineering and soil dynamics. 2001,
Paper no. 6.18.
In addition, when the magnitude of the input acceleration
[7] Novak M. Piles under dynamic loads. In: State of the art paper. Second
is 0.116g or above, strain gauge data show that cracking can international conference on recent advances in geotechnical earthquake
occur at the pile near the pile cap level. This cracking is clearly engineering and soil dynamics, vol. III. Missouri: University of Missouri-Rolla;
induced by the large inertia force experienced by the pile 1991. p.250–73.
[8] Normand P. Shaking table tests on model piles: a literature survey. Research
cap due to the pounding between soil and pile. However, it Report 95–3. Department of Civil Engineering, University of Canterbury, New
remains to be seen whether such a pounding between soil Zealand, 1995.
and pile did occur in the field during strong earthquakes. [9] Hushmand B, Scott RF, Crouse CB. Centrifuge liquefaction tests in a laminar
box. Geotechnique 1988;38:253–62.
If it does happen, those pile damages reported in the field may [10] Dorby R, Abdoun T. Recent studies on seismic centrifuge modelling of
be caused by soil–pile pounding, instead of liquefaction. The liquefaction and its effect on deep foundations. In: Proceedings of the fourth
shaking table tests by Lui and Chen [34] showed that the international conference on recent advances in geotechnical earthquake
engineering and soil dynamics, 2001, Paper no. SOAP-3.
maximum response of the pile foundation may appear before [11] Wei X, Fan L, Wu X. Shaking table tests of seismic pile–soil–pier–structure
the onset of liquefaction. In addition, Luo and Murono [6] showed interaction. In: Proceedings of the fourth international conference on recent
that severe pile damages did occur when no liquefaction was advances in geotechnical earthquake engineering and soil dynamics, 2001,
Paper no. 9.18.
observed. [12] Sato H, Kouda M, Yamashita T. Study on nonlinear dynamic analysis method
Since the pounding phenomenon between soil and pile was of pile subjected to ground motion, Part 2: comparison between theory and
observed in a scaled model, direct application to field situations experiment. In: Proceedings of the 11th world conference on earthquake
engineering, 1996, Paper no. 1289.
may not be straightforward. Therefore, a brief discussion of the
[13] Iiba M, Tamori S, Kitagawa Y. Shaking table test on effects of combination of
scaling model that we have adopted is given here. When we soil and building properties on seismic response of building. In: Proceedings
designed the shaking table model, we have assumed a 1:7 scale of the fourth international conference on recent advances in geotechnical
model. This scaling is based upon the scale laws discussed for earthquake engineering and soil dynamics, 2001, Paper no. 9.23.
[14] Mizuno H, Sugimoto M, Mori T, Iiba M, Hirade T. Dynamic behavior of pile
shaking table test for structures by Harris and Sabnis [35] and for foundation in liquefaction process-shaking table tests utilizing big shear box.
soil–structure models by Iai [36]. More specifically, our soil– In: Proceedings of the 12th world congress on earthquake engineering, 2000;
pile–structure system roughly corresponds to a three-storey Paper no. 1883 (in CD-ROM).
[15] Mizuno H, Iiba M, Kitagawa Y. Shaking table testing of seismic building–
building with a natural period of 0.4 s resting on end-bearing pile–two-layered–soil interaction. In: Eighth world conference on earthquake
piles of 0.7 m diameter and 12 m length. Therefore, we also expect engineering, San Francisco, vol. 3, 1984, p. 649–56.
that the same pounding phenomenon between soil and pile will [16] Finn WDL, Gohl WB. Response of model pile groups to strong shaking. In:
Prakash S, editor. Piles under dynamic loads. Geotechmical Special Publica-
occur in the field for a prototype of such scale. We should, tion No. 34. ASCE; 1992. p. 27–55.
however, emphasize that this is only an approximation since a [17] Nomura S, Tokimatsu K, Shamoto Y. Soil–pile–structure interaction during
complete set of scaling for soil–pile–structure model is still not liquefaction. In: Proceedings of the second international conference on recent
advances in geotechnical engineering and soil dynamics, 1991, Paper no. 5.16,
available in the literature. p. 743–50.
Nevertheless, the present study provides a new explanation [18] Yahata K, Suzuki Y, Funahara H, Yoshizawa M, Tamura S, Tokimatsu K. Pile
for the observed damages in piles during strong earthquakes. response characteristics of liquefied soil layers in shaking table tests of a large
scale laminar shear box. In: Proceedings of the fourth international
In particular, we speculate that one of the causes of damaging
conference on recent advances in geotechnical earthquake engineering and
the piles may be due to pounding between the laterally soil dynamics, 2001; Paper no. 6.34.
compressed soil and the pile near the pile cap level. However, [19] Adachi N, Suzuki Y, Tsugawa T. Experimental study on pile stress in liquefied
much work remains to be done on this seismic pounding and laterally spreading soils. In: Proceedings of the 12th world congress on
earthquake engineering, 2000, Paper no. 0799 (in CD-ROM).
phenomenon between soil and pile, before conclusive statements [20] Tokimatsu K, Suzuki H, Suzuki Y. Back-calculated p–y relation of liquefied
can be made. soils from large shaking table tests. In: Proceedings of the fourth international
ARTICLE IN PRESS

310 K.T. Chau et al. / Soil Dynamics and Earthquake Engineering 29 (2009) 300–310

conference on recent advances in geotechnical earthquake engineering and [26] Chambers AM, McManus KJ, Yang J, Berrill JB. The dynamic behaviour of
soil dynamics, 2001, Paper no. 6.24. drilled shaft micropiles. Civil Engineering Research Report, University of
[21] Tamura S, Miyazaki M, Fujii S, Tsuchiya T, Tokimatsu K. Earth pressure acting Canterbury, 1999, p. 99–2.
on embedded footing during soil liquefaction by large-scale shaking table [27] Yao S, Kobayashi K. Soil–pile–superstructure system in liquefaction. In:
test. In: Proceedings of the fourth international conference on recent Prakash S, editor. Piles under dynamic loads. Geotechnical Special Publication
advances in geotechnical earthquake engineering and soil dynamics, 2001, no. 34. ASCE; 1992. p. 241–55.
Paper no. 6.19. [28] Iwasaki T. Soil liquefaction studies in Japan: state of the art. Soil Dyn
[22] Funahara H, Fujii S, Tamura S. Numerical simulation of pile failure in liquefied Earthquake Eng 1986;5:28–33.
soil observed in large-scale shaking table test. In: Proceedings of the [29] Meymand PJ, Riemer M, Seed RB. Large scale shaking table tests of seismic
12th world congress on earthquake engineering, 2000; Paper no. 0927 soil–pile interaction in soft clay. In: Proceedings of the 12th world congress
(in CD-ROM). on earthquake engineering, 2000, Paper no. 0915 (in CD-ROM).
[23] Tamura S, Suzuki Y, Tsuchiya T, Fujii S, Kagawa T. Dynamic responses and [30] Lu XL, Chen Y, Chen B, Huang W, Zhao L. Shaking table testing of dynamic
failure mechanisms of a pile foundation during soil liquefaction by shaking soil–structure interaction system. Earthquake Eng Eng Vib 2000;4:20–9
table test with a large-scale laminar shear box. In: Proceedings of the (in Chinese with English abstract).
12th world congress on earthquake engineering, 2000, Paper no. 0903 [31] Gazetas G, Dorby R. Horizontal response of piles in layered soil. J Geotech Eng
(in CD-ROM). Div ASCE 1984;110:20–40.
[24] Nakagawa T, Tanaka M, Fujimori T, Hatori T, Yahata K. Experimental study on [32] SAP2000 Manual. Computers and Structures Inc., Berkeley, CA, USA.
the soil–pile–structure interaction by shaking table tests using large-scale [33] Poulos HG, Davis EH. Pile foundation analysis and design. New York: Wiley; 1980.
laminar box. In: Proceedings of the 12th world congress on earthquake [34] Liu HS, Chen KJ. Test on behaviour of pile foundation in liquefiable soils. Proc
engineering, 2000, Paper no. 1922 (in CD-ROM). second Int Conf Recent Adv Geotech Eng Soil Dyn 1991;1:233–5.
[25] Yasuda S, Ishihara K, Morimoto I, Orense R, Ikeda M, Tamura S. Large-scale [35] Harris HG, Sabnis GM. Structural modelling and experimental techniques.
shaking table tests on pile foundations in liquefied ground. In: Proceedings of 2nd ed. Boca Raton, FL: CRC Press; 1999.
the 12th world congress on earthquake engineering, 2000, Paper no. 1474 [36] Iai S. Similitude for shaking table tests on soil–structure–fluid model in 1g
(in CD-ROM). gravitational field. Soils Found 1989;29(1):105–18.

You might also like