Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

HOSTED BY Available online at www.sciencedirect.

com

ScienceDirect
Soils and Foundations 58 (2018) 819–837
www.elsevier.com/locate/sandf

Evaluation of ultimate behavior of actual large-scale pile


group foundation by in-situ lateral loading tests and
numerical analysis
Shuntaro Teramoto a,⇑,1, Tomonari Niimura b, Tomihiro Akutsu c, Makoto Kimura d
a
Setsunan University, Osaka 572-8508, Japan
b
Osaka Gas Co., Ltd, Japan
c
Obayashi Co., Ltd, Japan
d
Kyoto University Graduate School of Engineering, Japan

Received 27 January 2017; received in revised form 12 February 2018; accepted 28 March 2018
Available online 27 August 2018

Abstract

Structures resting on a liquefied natural gas (LNG) base must be completely stable during an earthquake in order to ensure a steady
gas supply. Thus, the stability of the foundations on which these large-scale structures rest is one of the most significant factors for real-
izing the safety and ease of maintenance in the event of an earthquake. Problems can potentially occur when the actual bearing capacity
of an in-service practical LNG tank foundation has not been verified, and when the operative design method for the pile-group effect is
inadequate, especially for large-scale pile groups. To date, no prediction method capable of appropriately simulating the mechanical
behavior of large-scale pile groups has been established. Therefore, in this study, in-situ lateral loading tests were conducted on a
real-life practical tank foundation upon the demolition of an LNG tank after 40 years of service. A 3D elasto-plastic finite element
method (FEM) analysis was also conducted. From the results, the bearing capacity and the failure level of the practical LNG tank foun-
dation were investigated and its soundness was confirmed. In addition, the behavior of a large-scale pile group foundation for load share
distribution was observed and compared with the operative design.
Ó 2018 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.

Keywords: LNG tank; Pile group; In-situ test; Lateral loading test; FEM; Pile group effect (IGC: E04)

1. Introduction recovery takes years longer than that for oil, natural gas
deposits can be found all over the world. Imported natural
The Tohoku earthquake of March 11, 2011 provided an gas arrives at LNG bases as 162 °C liquefied natural gas
opportunity to diversify energy sources and to address glo- (henceforth, LNG). The devolatilization of natural gas
bal environmental problems. Natural gas discharges lower compresses its volume to 1/600 of its original volume and
amounts of CO2, NO2, and SO2 during combustion than improves the efficiency of its transport. It is supplied
other fossil fuels. It is also said to be capable of providing through pipelines after storage, re-evaporation, and odor-
a huge amount of stable clean energy and, even though ization. In earthquake-prone countries, like Japan, the
safety and maintenance of an LNG-receiving terminal are
critical when servicing an LNG base. The LNG tank is
Peer review under responsibility of The Japanese Geotechnical Society. an extremely important facility whose foundation is com-
⇑ Corresponding author.
posed of multiple piles to support the superstructure. The
E-mail address: shuntaro.teramoto@civ.setsunan.ac.jp (S. Teramoto).
1
Former affiliation: Doctoral student, Kyoto University Graduate
superstructure must be capable of supporting the vertical
School of Engineering, Japan. load of its weight during normal operations, as well as

https://doi.org/10.1016/j.sandf.2018.03.011
0038-0806/Ó 2018 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.
820 Unit horizontal subgrade S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837

pHU
reaction force

pHU = p p pU

tan-1kHE kHE = k k kH

p: 1.0 0.5 0.5 0.5 0.5


0 Horizontal displacement H (m) k: 2/3 2/3 2/3 2/3 2/3 Loading
(1) (2) (3) (4) (100) direction

kHE : Coefficient of horizontal subgrade reaction (kN/m3) k : Correction factor of coefficient of horizontal subgrade reaction at single pile
pHU : Maximum unit horizontal subgrade reaction force (kN/m3) p : Correction factor of maximum unit horizontal subgrade reaction force at single pile
kH : Coefficient of horizontal subgrade reaction during : Correction factor of coefficient of horizontal subgrade reaction at group pile
earthquakes (kN/m2) k

pU : Earth pressure strength during earthquakes (kN/m2) p : Correction factor of maximum unit horizontal subgrade reaction force at group pile

Fig. 1. 2-D framework analysis design in Specification for Highway Bridges.

the lateral load and the bending moment caused by the pipe pile foundations, supporting integrated columns with
inertial forces of earthquakes. multiple steel pipe piles (Shinohara et al., 2013) have been
Problems related to pile groups subjected to lateral load- carefully studied. In addition, the mechanical behavior of
ing have been investigated by many researchers, and the seismic reinforced foundations for pile groups in an
pile-group effect has been found to be a significant factor improved ground (Bao et al., 2013), as well as reinforced
in the design of pile foundations. From the 1960s to existing foundations with steel pipe sheet piles (Isobe
1980s, an elastic theoretical analysis of the pile-group effect et al., 2014) and additional piles (Teramoto et al., 2016)
(Poulos, 1964), small-scale model tests (Tamaki et al., have been examined. In terms of the pile-group effect con-
1971), and an elastic numerical analysis using the theory sidered in the current 2D framework analysis design in the
developed by Poulos (Randolph, 1981) were conducted. Specification for Highway Bridges, shown in Fig. 1, the
In these studies, algebraic equations were suggested for back piles are assumed to have half the maximum subgrade
the pile-group effect which considered the spacing between reaction of the front-most pile (without considering the pile
piles, the matrix of a pile group, the fixation of a pile head, number). From past studies, however, it is clear that the
and the relative stiffness between a pile and the ground. number of piles increases the pile-group effect, and that
However, these equations were shown to have limitations there is a risk of the miscalculation in large-scale pile
in terms of the elastic ranges they could cover. In the groups due to the evaluation method in the design stage.
1980s and 1990s, along with the development of mechani- Considering the fact that pile group foundations are usu-
cal devices and measurement systems, centrifuge model ally built to be large for the sake of seismic reinforcement,
tests (Adachi et al., 1994; McVay et al., 1998) and real- it is important that the actual pile-group effect be clarified.
scale tests (O’Neil et al., 1985; Pedro et al., 1997) were Little progress has been made in the research on the lateral
performed. In the real-scale tests, in-situ lateral loading behavior of the large-scale pile groups typically used as
tests on a 9-pile group, including the ultimate behavior foundations for LNG tanks. This is due to the difficulty
(Kimura et al., 1994), and 3-D elastic perfectly-plastic of conducting in-situ loading tests with large-scale practical
FEM analyses of the tests (Kimura et al., 1995) were pile groups, and of finding agreement between the numer-
conducted, and the results became the basis for an evalua- ical analysis results and the in-situ loading test results.
tion of the pile-group effect in the current design for Spec- For these reasons, the actual bearing capacity of an in-
ification for Highway Bridges (Japan Road Association, service practical LNG tank foundation has yet to be
2012). determined.
In recent years, large-scale seismic loading model tests The challenges of conducting research on pile-group
(Shirato et al., 2008) and non-linear seismic numerical foundations are summarized as follows: (1) the actual bear-
analyses (Zhang et al., 2002a) have been conducted to ing capacity of an in-service practical LNG tank founda-
investigate the dynamic mechanical behavior of pile tion has not been determined, (2) the operative design
groups. To simulate the dynamic behavior of RC pile method for the pile-group effect is inadequate, especially
groups, a new beam model (AFD model), capable of con- for large-scale pile groups, due to the difficulty of conduct-
sidering the axial-force dependency in the nonlinear ing in-situ loading tests, and (3) there is no predication
moment-curvature relation, was introduced. In the devel- method capable of simulating the mechanical behavior of
opment of rational foundation structures, jacket-type foun- large-scale pile groups that can sufficiently consider the
dations using batter piles (Zhang et al., 2002b) and steel pile-group effect.
S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837 821

The objectives of the present study are (1) to determine in Fig. 3. Photo 1 shows the externals of the specimen and
and then verify the lateral behavior of an in-service practi- the reaction piles for loading. The tank is a double-walled
cal LNG tank foundation, (2) to confirm the pile-group metal tank with a maximum capacity of 45,000 m3. The
effect of a large-scale pile group and to evaluate an opera- foundation consists of a slab and 496 steel pipe piles.
tive design method, and (3) to develop and evaluate a pre- Details of the piles are shown in Fig. 4. The slab is
diction method capable of simulating the mechanical 48 m in diameter, 0.8 m in thickness, and is made of rein-
behavior of large-scale pile groups that gives appropriate forced concrete. The piles are 23.5 m in length, 406.4 mm
consideration to the pile-group effect. In-situ lateral load- in diameter, 12.7 mm in thickness in the upper part, and
ing tests were conducted with a 7  9 pile group (m  n pile 9.5 mm in thickness in the lower part. There is a distance
group: arranged in n lines orthogonal to the loading direc- of 0.8 m between the bottom of the slab and the ground
tion, and m lines parallel to the loading direction), and with surface. The ground around the edge of the tank is
a 3  1 pile group used as part of the practical tank foun- improved to an N-value of about 15 by means of sand
dation. This was in accordance with the demolition of an compaction piles.
LNG tank after 40 years of service at Osaka Gas Senboku Each pile is an open-toe type of end-supported pile and
LNG Receiving Terminal 1. An investigation of the lateral the N-value of the bearing ground located at the pile tip is
bearing capacity of the LNG tank foundation was carried around 50. The upper part of the pile is filled with rein-
out. It was possible to observe the behavior of the tank forced concrete to a depth of 6 m. The upper part of the
from the yielding mode to the ultimate mode to the pile, 0.8 m above the ground surface and 0.3 m under the
destruction mode since the LNG tank foundation was ground, is covered with protective concrete, is rectangular
completely demolished after the tests. Simultaneously, a in shape, and has a thickness of 150 mm.
simulation was also conducted by means of a 3D elasto-
plastic FEM analysis. In the analysis, the soil model for 2.2. Preparation of specimens
the subloading tij model (Nakai et al., 2004) and the pile
model for the hybrid model (Zhang et al., 2000) were cap- The positions of the 7  9 and the 3  1 pile group spec-
able of appropriately simulating the pile-group effect of imens in the slab are shown in Fig. 3. The specimens were
large-scale pile groups. However, the ordinary simplified loaded laterally using the right side of the slab as a reaction
models that were used in the past, such as the 2D model, pile, as shown in Photo 1. The specimens were prepared by
the no volume beam alone model, and the contracted pile removing other parts of the slab, so that only the specimens
group model, have been shown to be inadequate for simu- remained. However, the piles around the specimens
lating the pile-group effect. The pile-group effect and the remained in the ground. The protective concrete over the
mechanism of the large deformation of large-scale pile ground surface was removed so as to enable the direct mea-
groups are discussed considering both sets of results. surement of the pile shaft, while the concrete under the
In this study, the ultimate bearing capacity of a practi- ground was left intact.
cal foundation after its service is determined, the yield
load to the ultimate load, and the foundation deformation
(including the yielding of the pile head and the residual 2.3. Measurement
deformation of the pulled-out piles) are observed, and
the lateral mechanical behavior of a large-scale pile group, A contact-type displacement transducer was employed
especially in terms of the pile-group effect, is clarified. A to measure the pile and slab specimens by using the refer-
comparison between the load share distribution of a prac- ence piles located in front of each specimen (Figs. 3 and
tical pile group and the design in the Specification for 5). However, because these reference piles were likely to
Highway Bridges validates the design method. Finally, a be affected by the pile group displacement at loading, the
simulation method which is applicable to large-scale pile displacement of these piles was measured with the total sta-
groups is developed. In this analysis, the model pile sim- tion set up a distance of 25 m from the specimen (orthogo-
ulates the pile-group effect properly and the soil model nal to loading direction).
simulates the ground around the cyclically-loaded large- The shear strain and bending strain of the piles were
scale pile group. With this method, the hard-to-predict measured with the strain gauge, the inclination of the piles
behavior of large-scale pile groups can also be simulated was measured with the inclinometer, and the applied load
before a loading test. was measured as shown in Fig. 5. The pile-group effect in
the loading direction was investigated by measuring seven
2. Outline of lateral loading tests on 7  9 pile group piles in the center line, while those orthogonal to the load-
foundation ing direction were investigated by measuring the five piles
in the front line. The strain, the inclination, and the
2.1. Structure of demolished LNG tank displacement of the part of the piles in the ground were
not measured due to the difficulty of measuring piles filled
The demolished LNG tank used for the loading tests is with concrete. This point was complemented by a numeri-
shown in Fig. 2, and a top view of the foundation is given cal analysis.
822 S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837

3×1pile group Reaction piles


Loading
6000 direction
7×9 pile group

14000 2000

Double-walled metal tank


Photo.1
45,000 [kL]
18000

Colu-
N
48 m
mnar
sec- value
tion 0 20 40

23.5 m
Bearing ground (N = 46)

single pile Fig. 3. Elevated view of LNG tank.


2×1pile group
3×1 pile group
2000 19@2000=38000 2000
750 2250 48000 2250 750
7×9 pile group
7×9 pile group Reaction piles
5,000 kN jack
(two series)
Loading plate
Base (H-steel)
Displacement gauge SDP-300D, Strain gauge
Displacement gauge CDP-100, Strain gauge
Fixed inclinometer 5,000 kN jack Reference pile Reaction piles

Fig. 2. Specimen and measurement arrangement. Photo 1. Externals of specimen and reaction piles.

2.4. Loading stages and determination of maximum load 3. FEM analysis conditions

The loading stages of the 7  9 and the 3  1 pile groups 3.1. FEM code and FEM mesh
are shown in Figs. 6 and 7, respectively. The unidirectional
multiple-cycle, multiple-step loading method was selected The FEM code DBLEAVES (Ye et al., 2007) was used
as the lateral loading test method. There were five cycles for the FEM analysis. The analysis area and the FEM
and ten loading steps, amounting to a total loading time mesh are shown in Figs. 8 and 9. In this analysis, the
of about 5 h, as the lateral load was applied with six 7  9 pile group and the piles remaining around them were
hydraulic jacks of 5000 kN, as seen in Fig. 3 and Photo simulated precisely.
1. The outputs of the six jacks were controlled by an oil
pressure device which also simultaneously measured the 3.2. Modeling of pile foundation
pressure.
In order to observe the ultimate behavior of the 7  9 As shown in Fig. 10, each pile was modeled by a hybrid
pile group, it was necessary to predict the deformation element with the same volume as the pile. This was neces-
while determining the maximum load (capacity of the sary to simulate the pile-group effect properly because the
jack). Firstly, the yield load per pile, 247 kN, was calcu- pile-group effect is the interaction between one pile and
lated by Y. L. Chang’s formula (fixed pile head condition). the other piles and depends on the pile spacing. The pile-
A yield load of 15,561 kN (247 kN was multiplied by 63) group effect can be simulated more precisely with a hybrid
was assumed for the 7  9 pile group. Secondly, an analy- element than with a beam element alone. The parameters
sis using a 3-dimensional elasto-plastic FEM was con- of the pile foundation were determined through the use
ducted before the tests on the 7  9 pile group. The of the design values of the specimen according to the calcu-
results were used to establish the maximum load in the lation flow shown in Fig. 10.
loading tests; it was set at 30,000 kN (with the six The piles around the specimens that remained in the
5000 kN jacks). ground were also modeled to simulate the actual condi-
S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837 823

Fig. 8. Analysis area and FEM mesh.


824 S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837

tions, as shown in Fig. 10. To investigate the influence of


the remaining piles, a comparison analysis was conducted
before the tests in which the remaining piles were removed.
Consequently, it was confirmed that the 7  9 pile group
without any remaining piles had a 10% larger displacement
than the 7  9 pile group with remaining piles.

3.3. Constitutive models of soils


Loading direction
The ground was modeled by the subloading tij model
(Nakai et al., 2004). This model is regarded as one of the
most useful elastoplastic models because it can consider
the intermediate principal stress by adapting the concept
of modified stress tij. A comparison of the tensors and sca-
lars related to the stress and strain increments (between the
ordinary concept and the tij concept) is shown in Table 1.
The influence of the stress-path dependency of the density Fig. 9. Details of FEM mesh around 7  9 pile group.
and/or the confining pressure on the deformation and
strength characteristics was considered.

Column element
(elastic model)
Beam element
Bending moment (kN*m)

(bilinear model) Mp
EI EI column EI beam

EA EA column EA beam
Fully plastic moment Mp = 416 (kN m)
(Design value of actual pile)
Beam element
EI beam EA beam Degradation factor of stiffness after fully plastic
tan tan = 1/10
Column element
(1/1000 for pile head)
EI column EA column

Actual pile Pile model in FEM Curvature (1/m)

Actual pile
Esp = 2.0×108[kN/m2]
Outer diameter

Steel pipe Asp = 0.0157[m2]


0.406 m

Isp = 3.04×10-4[m4] (EA)pile = Esp × Asp + Esc × Asc = 6.35 ×106[kN]


Esc = 2.82×107[kN/m2] (EI)pile = Esp × Isp + Esc × Isc = 8.98 ×104[kN*m2]
Concrete Asc = 0.114[m2]
Thickness 12.7 mm Isc = 1.03×10-3[m4]

0.1 ( EI ) pile 0.9 ( EI ) pile

Column element Beam element

( EI ) column 0.1 EI pile 8.98 103 ( EI ) beam 0.9 E pile 8.08 10 4 [kN * m 2 ]

Abeam Asp 0.0157 m 2


0.4 m

0 .4 4 3 4
I column 2.1 10 m EA
12 Pile heart Ebeam
pile
4.04 108 kN / m 2
Abeam
EI column
0.4 m Ecolumn 4.21 10 6 kN / m 2 EI beam 4
I column I beam 2.0 10 m4
Ebeam

Fig. 10. Modeling of pile and calculation flow of parameters.


S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837 825

Table 1 Top view 800

600
Tensors and scalars related to stress and strain increment in ordinary
concept and tij concept.
Jack
Ordinary concept tij concept

4200
3000
Tensor normal to dij ðunittensorÞ aij
reference plane
Stress tensor rij tij ¼ rik akj
Mean stress p ¼ rij dij =3 tN ¼ tij aij

600
Deviatoric stress tensor sij ¼ rij  pdij t0ij ¼ tij  tN aij
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffi 7000
Elevated
Deviatric stress q ¼ ð3=2Þsij sij ts ¼ t0ij t0ij view 1200
1200
Footing
Jack
Stress ratio tensor gij ¼ sij =p xij ¼ t0ij =tN
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1200
400
pffiffiffiffiffiffiffiffiffiffi
Stress ratio g ¼ q=p ¼ ð3=2Þgij gij X ¼ ts =tN ¼ xij xij
dev ¼ deij dij deSMP ¼ deij aij

100
Strain increment

50
normal to reference
Beam
plane
Deviatoric strain deij ¼ deij  13 dev dij de0ij ¼ deij  dev SMP aij 800
increment tensor qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Strain increment ded ¼ 23 deij deij dcSMP ¼ de0ij de0ij Unit : mm
parallel to reference
plane Fig. 11. Outline of loading tests on 1  2 pile group prior to 7  9 tests.

Bending moment [×10 2 kN m]


2000 -16 -12 -8 -4 0
Test 0.0 [Test]
1500 FEM 500 kN
-1.0
Lateral load (kN)

1000 kN
-2.0 1500 kN
1000
Depth (m)

1900 kN
-3.0
[FEM]
500 500 kN
-4.0
1000 kN
0 -5.0 1500 kN
0 20 40 60 80 100
1900 kN
Lateral displacement of pile top (mm) -6.0

Fig. 12. Results of simulation analysis on 1  2 tests prior to 7  9 tests.

The main superiority of the subloading tij model in this Fig. 11, while the test and analysis results (load–displacement
analysis is shown below. In simulating the cyclic lateral and bending moment distributions) are shown in Fig. 12.
loaded pile group, the soil around the piles must be modeled FEM parameters k and j were determined with sufficient
such that it can consider the stress path during unloading accuracy with the 1  2 tests, which were conducted before
and reloading, and it is simulated by the subloading surface the 7  9 tests. The calculation flows for all the input param-
concept (Hashiguchi, 1980). The specifics of this model eters of the soil are shown in Figs. 13 and 14 and Table 2. The
enable it to simulate not only clay soil behavior, but also improved area given in Fig. 13 shows that the ground has been
sand behavior without any distinction between normally improved (N = 15) by the sand compaction pile method dur-
consolidated and overconsolidated soil. On the other hand, ing the construction of the LNG tank. The area is distributed
the Cam-clay model cannot simulate sand behavior. And, in around the circumference of the LNG tank (Figs. 8 and 13). In
general, to simulate laterally loaded piles, an interface ele- the analysis of the 7  9 and 3  1 tests, the j determined by
ment is typically arranged between the piles and the soil the 1  2 tests was converted to consider the difference in the
behind them to demonstrate the delamination behavior of N-value distribution according to the flow shown in Fig. 13.
the pile surface. At this point, the model can simulate
delamination behavior without the interface element by 3.4. Analysis procedure
decreasing the deformation modulus of the soil according
to the decrease in confined pressure. Firstly, the initial stresses of the ground were generated
The input parameters of the soil, except for k and j, were by a self-weight analysis under an elastic ground condition.
determined by the results of soil tests and standard penetra- Secondly, the elastic ground was replaced by subloading tij
tion tests. Compression index k and swelling index j were model soil and hybrid element piles. However, the initial
determined with the results of parallel 1  2 lateral loading stresses remained. Finally, footing nodes were applied at
tests conducted near the LNG tank foundation (prior to the the actual loading positions and under the same condition
7  9 tests). An outline of the 1  2 tests is shown in as that of the tests (Figs. 6 and 7).
826 S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837

250 m

B-1(5) Previous 1×2 test area N0, 0

Unimproved ground area N1, 1

Estimated ground surface strength Improved ground area N2, 2

Tank used for Strong Weak


this test series (N ,0Nm~, N8 are
0 1 average N-value of
2
m in ground surface
(
SCP ground improved area
150 m

(N = 15) N0 = N-value of B-1 = 5

A-4(5) N1 = average N-value of A-1~A-4 = 7.5


A-3(11)
N2 = 15
Previous loading test
of 1×2 pile group
2 1
3×1 1 0 , 2 0
3 3
Assumption
7×9
2×1 3(1 2 0 )(1 e0 ) p
E0
Single 48 m
E0 2800 N
1 ( 0, e0, p are taken the
A-1(9) A-2(5) Position of SPT N same value in all cases)
( ) : AVG N-value in surface 8 m area

N-value N-value N-value N-value N-value


0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70

AVG N-value = 9 AVG N-value = 5 AVG N-value = 11 AVG N-value = 5 AVG N-value = 5
5
Depth of max. bending
moment distribution in
FEM analysis Filling soil
8

10
Depth (m)

15

20

Depth of pile tip

25
A-1 A-2 A-3 A-4 B-1

Fig. 13. Conversion method for j and SPT data.


S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837 827

~ Parameters except and ~ ~ Parameters and ~


[Reference of past study] [Assumption]
a = 500 = 1.5 Filling layer ~ 8 m 8m~

[Assumption of ordinal value]


K0 a, , , , e0, Rf
Self weight

1
analysis

decide as to
[soil test, standard penetration test] past 2×1 test result are simulated
e0 N

6 sin
m (= p) Rf decide as right section
(3 sin ) degree

Input parameters of prediction analysis of 7×9 group pile


Fig. 14. Calculation flow of input soil parameters for prediction analysis of 7  9 pile.

Table 2
Input parameters of soil.
Depth of Soil classify- Failure N- Poisson’s Unit weight Coefficient of earth Void Compression Swelling index j
layer (m) cation stress ratio value ratio m c (kN/m3) pressure at rest K0 ratio index k
Unimproved Improved
RCS N e0
ground j1 ground j2
0.175 Filling soil 2.8 17 0.3 17.9 0.43 0.80 0.0018 0.0008 0.0004
0.525 (Gravel) 2.8 17 0.3 17.9 0.43 0.80 0.0018 0.0008 0.0004
0.875 2.8 17 0.3 17.9 0.43 0.80 0.0018 0.0008 0.0004
1.225 2.3 13 0.3 8.1 0.43 0.95 0.0018 0.0008 0.0004
1.65 2.3 13 0.3 8.1 0.43 0.95 0.0018 0.0008 0.0004
2.2 Filling soil 2.5 6 0.3 8.1 0.43 0.77 0.0018 0.0008 0.0004
2.85 (Silty-sand) 2.5 6 0.3 8.1 0.43 0.77 0.0018 0.0008 0.0004
3.6 2.4 5 0.3 8.1 0.43 0.77 0.0018 0.0008 0.0004
4.5 3.1 4 0.3 8.1 0.43 0.75 0.0018 0.0008 0.0004
5.5 Filling soil 1.9 3 0.3 8.1 0.43 0.95 0.0018 0.0008 0.0004
6.5 (Silty-fine 1.9 3 0.3 8.1 0.43 0.95 0.0018 0.0008 0.0004
8 clay) 3.5 9 0.3 8.1 0.43 1.70 0.0018 0.0008 0.0004
10.5 Gravel 3.4 8 0.3 8.1 0.43 0.57 0.0009 0.0004 0.0004
14 Gravel sand 3.5 36 0.3 8.1 0.43 1.00 0.0009 0.0004 0.0004
18 Silty-clay 2.8 13 0.3 8.1 0.43 0.72 0.0009 0.0004 0.0004
22.5 Gravel 4.6 46 0.3 8.1 0.43 0.45 0.0009 0.0004 0.0004

4. Test and FEM analysis results the test results, given in Fig. 15, was actually seen to have
decreased. However, the ductile deformation progressed
4.1. Load displacement results and the load gradually increased without the occurrence
of brittle failure.
The load displacement results for the 7  9 pile group at From a comparison of the results of the prediction
the loading position (average displacement at both sides of analysis before the tests and the test results, the analysis
the slab surface) are presented in Fig. 15. Fig. 15 is shown was shown to have predicted the test results with high
in Fig. 16 as the logarithmic axis for reading the yielding accuracy until around 40 mm (=10%  d). However, there
point. The maximum load of the tests was 25,400 kN was little decrement in the lateral stiffness in the analysis.
and the displacement at the time was 240.5 mm This is because the characteristic of the hybrid element
(=60%  d). The residual displacement after the tests was simulates the piles. At the pile head, as shown by the blue
158 mm. Compared to the design yield of 15,561 kN, cal- circle in Fig. 17, the solid elements arranged in the hybrid
culated with the Chang’s equation, the lateral stiffness in element have high stress due to the fixation of the pile
828 S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837

36000

Displacement revised by
32000 the reference pile

Jacks capacity 30,000 kN Cycle 6


28000 Continue Maximum load 25,408 kN
loading Displacement 240.5 mm
Cycle 5
121.0 mm +39 mm
24000
+19 mm Unload
Cycle 4
20000 65.7 mm
+10 mm
Load (kN)

16000
Cycle 3 Yield load of pile head
32.1 mm in design 15,561 kN
12000 (Y. L. Chang s formula)
+9 mm

Test
8000

4000

0
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280
Lateral displacement at loading point mm

Fig. 15. Load displacement results of 7  9 tests, prediction analysis, and analysis after the tests.

head, and the solid elements share higher shear force. ulates the initial stiffness of the test results as lower than
Therefore, this increase in shear force in the solid ele- that of the 7  9 pile group. However, the accuracy was
ments results in a decrease in the plastic deformation of higher at around 60 mm (=15%  d). These results indi-
the beam element. cate the possibility that FEM underestimated the soil
The results indicate that the model for the pile head is parameter since the lateral stiffness depends more on
improved after the tests, as presented in Fig. 17. The plastic the ground stiffness than on the pile stiffness in the initial
deformation in the large deformation phase, shown by the phase.
blue line in Fig. 15, was able to simulate the test results
more appropriately due to the improvement. However, 4.2. Deformation mode of large-scale pile group
the deformation was still underestimated at the maximum
load in the analysis. The main reason is that the analysis The deformation modes for the inclination of the piles
was not able to simulate the falling of the pile head from during the tests and FEM are shown in Fig. 19. The
the slab (shown in Photo 2) because the pile head and pulled-out piles (front-most and back-most piles) after
the slab share the same nodes in the FEM analysis. If an the tests are shown in Photo 3.
interface element is used to simulate the falling for the next Both the experimental results and those of the analysis
improvement, some tests have to be conducted to deter- indicate that the front-most pile inclined more than the
mine the parameters of the interface element; that issue back-most pile at an applied phase of 8000 kN kN, while
can be addressed in the future. On the other hand, given the inclination of the front-most pile increased exponen-
the shape of the unloading and reloading cycles, FEM tially at an applied phase of 25,400 kN. As is to be expected
was able to simulate the test results with high accuracy. from the large inclination of the pile, the pile and the slab
This is because the subloading tij model introduces a did not cross orthogonally in the tests. This is because the
subloading surface concept, which ordinary elasto-plastic pile head and the slab surface both fail because of the large
models cannot simulate. deformation, as shown in Photo 2. The falling of the pile
The load displacement results for the 3  1 pile group head from the slab and the falling away of the slab skin
in the loading position are shown in Fig. 18. FEM sim- reduce the pile head connectivity.
S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837 829

Moreover, the analysis results indicate that the maxi- 30000


mum point of the bending moment of the front-most pile Test results After pile yielding
was shallower than that of the back-most pile. As shown 20000
in Photo 3, the pulled-out front pile actually deformed Pile yielding area
more than the back pile, and the maximum point of the 55 mm, 15000 kN
12000
bending moment was shallower again. Due to the failure
10000
of the front-most pile head, there was a distinct change in 9000
8000 After ground yielding

Load (kN)
the supporting structure of the pile group, and the position 7000 before pile yielding
of the maximum bending moment of the front-most pile 6000
was shallower, resulting in a dramatic incline in the 5000 8 mm, 6000 kN
front-most pile over the ground. 4000
The inclination of the slab is shown in Fig. 20. This incli-
3000
nation was calculated from the vertical displacement of the
slab surface, as measured by the contact-type displacement Initial stiffness
2000
transducer. The inclination of the slab increased with the
load increments. However, from around 16,000 kN, the
inclination was seen to decrease in the test results. This
decrement was attributed to the failure of the front-most 1000
pile head. However, FEM cannot simulate this decrement 1 10 100
Lateral displacement at loading point (mm)
in inclination, because the pile head of the FEM mesh is
modeled as a continuum and is not capable of simulating 30000
the pile head desorption shown in Photo 2. In addition, it FEM results After pile yielding
is also the reason why the displacement at the maximum 20000
load from the FEM analysis was still underestimated, as Pile yielding area
is shown in Fig. 15.
12000
55 mm, 14000 kN
10000
4.3. Difference in bending moment distribution between 9000
scales of pile group 8000
Load (kN)

7000
6000
The bending moment distributions of the 7  9 and the 5000 After ground yielding
3  1 pile groups at a displacement of 30 mm (7.5%  d) before pile yielding
4000
are presented in Fig. 21. Only the FEM analysis results
3000 7 mm, 4000 kN
are shown because no measurements were taken in the
ground. For comparison, Fig. 22 shows the results of past
in-situ lateral loading tests on a 3  3 pile group at a dis- 2000
placement of 7.5%  d (Kimura et al., 1994).
The value beside the maximum bending moment repre- Initial stiffness
sents the bending moment of each pile normalized by the 1000
front-most pile in the center line. 1 10 100
In all cases, the front-most pile had the largest bending Lateral displacement at loading point (mm)
moment and its position in the ground was the shallowest. Fig. 16. Load displacement results shown as common logarithmic axis.
A strict comparison of these cases is difficult due to the dif-
ference in test conditions. The bending moment distribu-
80 cm

tions for the 7  9 and the 3  1 pile groups were similar Slab (elastic) Slab (elastic)
to those of the past 3  3 tests.
From a comparison of the middle pile (pile 4 in the 7  9
Beam
90 cm

pile group case) and the back-most pile, the distributions Ground
surface
were nearly the same as those in the 3  1 pile group case.
However, the bending moment of the middle pile in the
Hybrid element

Hybrid element

7  9 pile group case was smaller. The maximum bending


moments of the middle pile and the back-most pile were
deeper in the 7  9 pile group case due to the larger pile-
group effect generated in the 7  9 pile group. Considering
that the depth of the ground has an influence on the lateral
bearing capacity, equal to 1/b (b: characteristic value Prediction analysis Analysis after the tests
of the pile), and that b is proportional to kH1/4(kH:
coefficient of horizontal subgrade reaction), the kH of the Fig. 17. Improvement in modeling of pile.
830 S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837

1400 Falling out of the pile head


Loading direction from the slab
1200
Yield load of pile head
1000
in design 741 kN
(Y. L. Chang s formula)
Load (kN)

800

600 Falling away


of slab skin
400

200
Front-most pile
0
0 10 20 30 40 50 60 70 80 90 100 Photo 2. Desorption of pile head and slab surface.
Lateral displacement (mm)

Fig. 18. Load displacement results of 3  1 tests and simulation.

Front Loading direction Back


Inclina
-tion
FEM 0.30° 0.27° 0.25° 0.24° 0.24° 0.25° 0.26°
Test 0.22° 0.16° 0.19°

-2.1 m 40 cm 40 cm
-2.5 m
-2.8 m -3.0 m -3.1 m -3.2 m -3.2 m

Pile head Pile head


Position of the maximum bending moment from FEM results
Front-most pile Back-most pile

Photo 3. Deformation of pulled-out piles after tests.


Front Loading direction Back
10000
Rotational angle of footing θ (μrad)

FEM 2.2° 1.8° 1.7° 1.5° 1.5° 1.5° 1.6° FEM


Test 4.1° 3.0° 3.0°
1000
Test

-1.4 m
-2.4 m
-2.9 m 100 Measured
Measured displacement
-3.4 m -3.6 m displacement
-3.8 m -3.9 m θ Loading

Position of the maximum bending moment from FEM results Front Back
10
0 4000 8000 12000 16000 20000 24000 28000
Load (kN)
Fig. 19. Pile deformation in ground (FEM) and inclination of piles (Test,
FEM). Fig. 20. Inclination of slab.

middle pile in the 7  9 pile group case is lower and has From the initial phase to around 12,000 kN, the posi-
deeper ground effects. This decrement in kH is the pile- tion of the maximum bending moment becomes deeper
group effect. with incremental increases in the load for all piles. After
12,000 kN, the position was deeper in the back-most pile,
4.4. Bending moment distribution according to loading remained constant in the middle pile, and was shallower
in the front-most pile. This is because of a decrease in kH
Fig. 23 shows the evolution of the bending moment dis- (it becomes deeper) and the yielding of the pile head (it
tribution for the underground 7  9 pile group in the FEM become shallower). As with the back-most pile, if the pile
analysis. As was stated previously, the front-most pile had head did not yield, the position in the front-most pile was
the largest bending moment and the shallowest position. deeper.
S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837 831

Bending moment (kN*m) Bending moment (kN*m)


-200 -100 0 100 200 300 -200 -100 0 100 200 300
2 2
1 Slab Slab
1
0 0
Ground surface Ground surface
-1 -1
Front-most
-2 center pile (1) -2
Middle Front-most pile
-3 center pile (4) -3
Middle pile
-4 1.13 Back-most -4
center pile (7) Back-most pile

Position (m)
Position (m)

1.00 1.00
-5 Front-most -5
outer pile (8) 0.89
-6
0.80
-6
0.63 Lateral load = 600 kN
-7 -7 (7.5%×d)
Lateral load (8) outer
-8 = 9,000 kN -8
(7.5%×d) Loading direction
-9 (7) (4) (1) center -9
-10 Loading
direction -10 Back-most Middle Front-most
-11 -11
pile pile pile
outer
back front
-12 -12
-13 -13
7×9 pile group 3×1 pile group

Fig. 21. Results of bending moment distribution in FEM analysis.

The influence of the yielding pile head in the front-most


Loading direction
pile, at 14,640 kN, became larger than the influence of the
decrease in kH. This influence of the yielding pile head is
that the position of the maximum bending moment was (3) (2) (1)

9m
9m

shallower, while the decrease in kH results in a deeper max-


imum bending moment. Therefore, the position of the
maximum bending moment of the front-most pile was
19.4 m 9m
shallower. Reaction piles 3×3 pile group
Subsequently, once the front-most pile head yields, the
Pile diameter : 1.2 m Lateral load : 12,000 kN
foundation leans forward. The more forward the pile, the Pile spacing : 2.5d (7×9 test is 5d) (7.5%×d)
deeper the position of the maximum bending moment. Bending moment (kN*m)
Therefore, it is understood that the influence of the decre- -2000 -1000 0 1000 2000 3000
ment in kH (the pile-group effect) and the influence of the 0
Ground surface
yielding of the pile head are nearly equal for the middle
pile. These variations in the depth of the maximum bending Sand gravel (filling)
moments and other factors are summarized in Fig. 24. 5 N-value = 5~10

4.5. Load share of large-scale pile group 10


Position (m)

The load share of each pile normalized by the front- 13


most pile is shown in Fig. 25. This shared load is calculated 15
from the shear force measured by the strain gauge attached
to the pile surface above the ground.
20
Diluvium

In all loading stages, the front-most pile shares the lar-


gest load, and at around the fourth pile, it shares the small-
Front-most pile (1)
est load. These results are the same as those seen in past
25 Middle pile (2)
model tests (Shibata et al., 1989). The load share of the
back-most piles decreases according to the increase in Back-most pile (3)
applied load; FEM can simulate this tendency in the evolu- 30
tion of the load share distribution.
The load share distribution of the framework analysis Fig. 22. Outline and bending moment distribution results of past in-situ
design in the Specification for Highway Bridges is shown lateral loading tests on 3  3 pile group (Kimura et al., 1994).
832 S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837

in Fig. 25. In the results for a load of 16,000 kN (assuming


4,000 kN
the ground reaches the maximum subgrade reaction), the
load share averages of the back piles ((2)–(7)) are 0.56 in 12,000 kN
(7) (4) (1)
the test and 0.48 in FEM. While the averages are approxi- 20,000 kN
mately equal, the subgrade reaction in the design is 25,400 kN
back front
overestimated in the center and underestimated in both
Bending moment (kN*m)
the second and the back-most piles.
-500 -400 -300 -200 -100 0 100 200 300 400 500
These results indicate that the design is at risk of overes- 2
timating the maximum subgrade reaction in cases where Slab
the pile number increases to more than 7, as the number 0
Ground surface
of overestimated center piles does partly increase. -2
Moreover, the results indicate that the reduction rate for Procedure of depth of
the maximum subgrade reaction at the back piles (=0.5 -4 maximum bending moment

Position (m)

Fully-plastic moment
Fully-plastic moment
currently) must be properly determined when designing -6
each pile.

at 14,640 kN
at 21,840 kN
-8

4.6. Ground subsidence around 7  9 and 3  1 pile groups -10

-12
The ground behind the front-most and the back-most
piles after testing is shown in Photos 4 and 5, respectively; -14
the loading system at the maximum load is shown in Photo Front-most pile (1)
Bending moment (kN*m)
6. The crack generated at the backside of the front-most
-500 -400 -300 -200 -100 0 100 200 300 400 500
pile presumably was caused by the separation of the rectan- 2
gular protective concrete of the pile and the ground. Slab
Behind the back-most pile, large ground subsidence was 0
generated, and this exacerbated the unstable condition of -2
the jacks. A subsidence of 200–350 mm was measured, indi-

Fully-plastic moment
cating a slide down of the ground between the back-most -4
Position (m)

pile and the remaining piles underground. The prediction

at 16,840 kN
-6
fully-plastic moment

analysis indicates that this subsidence would be around


170 mm. The quantity evaluated by FEM was smaller -8
Less than

due to the difficulty of simulating the slide down. However,


-10
this analysis is a useful prediction method for the ground
subsidence around a pile group. -12
To compare the scale of the pile group, an additional
-14
analysis of the 3  1 pile group (loaded at the same dis- Middle pile (4)
placement as the 7  9 pile group) was conducted. The Bending moment (kN*m)
results indicate that the ground subsidence of the 3  1 pile -500 -400 -300 -200 -100 0 100 200 300 400 500
group was 40 mm, which is smaller than the 170 mm subsi- 2
dence of the 7  9 pile group. Therefore, the larger subsi- Slab
0
dence of the 7  9 pile group can be assumed to be
specific to large-scale pile groups. -2

-4
Fully-plastic moment
Position (m)

4.7. Lateral deformation of ground


at 17,280 kN

-6
fully-plastic moment

The followability of the ground with adjacent piles, the -8


range of influence of the pile group, and the evolution of
Less than

-10
the ground deformation form are discussed in this section.
Distributions of the lateral displacement of the ground at -12
4000 kN and 25,400 kN are shown in Fig. 26. The displace-
ment was normalized by the average lateral displacement -14
Back-most pile (7)
of the pile at the ground surface.
According to the increase in loading, the inner ground Fig. 23. Evolution of bending moment distribution of 7  9 pile group in
of the pile group began to follow the pile group as a clod. FEM analysis.
S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837 833

In particular, the large ground deformation of the inner Bending moment (kN*m)
backside of the ground apparently affected the decrease -400 -200 0 -400 -200 0 -400 -200 0
0
in the load share, as shown in Fig. 25. Ground surface
From the representative contour line for the 10% dis- -2
placement, the range of influence of the 7  9 pile group
was wide in all directions, especially in the loading direc- [Main factor]-4
tion, compared with the 3  1 pile group. In the case of Influence of Influence of

Position (m)
the 7  9 pile group, the range increased for the vertical -6 the yielding the decrease
pile head in kH and
direction (from 13d to 15.5d) and for the loading direction -8 the yielding [Main factor]
(from 15d to 25d) according to the increase in load. How- [Main factor] pile head Influence of
Influence of are equal the decrease in kH
ever, the range decreased for the orthogonal to the loading -10 the decrease in kH
direction. This reduction can be attributed to the large [Main factor]
-12 Influence of
deformation of the pile group caused by the shear failure the decrease in kH
of the ground beside the pile group, and the decrease in
-14
the followability of the ground. Front-most pile (1) Middle pile (4) Back-most pile (7)
From the representative contour line of 40% near the
distribution of maximum bending moments, the contour Fig. 24. Evolution of depth of maximum bending moment.
has a symmetrical shape in the 4000 kN phase. However,
4,000 kN 8,000 kN 12,000 kN 16,000 kN
according to the increase in loading, the contour starts to
FEM
have a downward-sloping shape. This change is due to
the position of the maximum bending moment in the Test
front-most piles, which do not become deeper due to the 1
yielding of the pile head. Thus, the deformation of the pile
group affects the shallow forward ground around the front-
Load share of each pile normalized by front-most pile

most piles and the deep ground. 0.9


(1) (7)
According to the above results, the loading influence on
the ground around a pile group is expanded horizontally front back
0.8
and vertically by expanding the scale.

4.8. Investigation of stress distribution of soil around 7  9 0.7


pile group and deformation of piles
0.6
The increment distributions of tN (p of the common
elasto-plastic model, Table 1) at 4000 kN and the lateral
0.5
displacement distribution of the piles are shown in
Fig. 27. On the front side of the piles in the ground, an
increase is generated from a depth of 1.5 m in the 0.4 Design 0.5
forward-depth direction, while the increment is larger for Ave. of (2)~(7) Ave. of (2)~(7)
0.48
the front-most piles. On the back side of the piles in the 0.56
0.3
ground, a decrease is generated in the back-depth direction, (1) (2) (3) (4) (5) (6) (7)
while the decrement is larger at the back-most piles.
Our earlier results showed that the front-most piles are Front Back
supported by the forward ground, and the back-most piles
Fig. 25. Load share distribution of 7  9 pile group.
are supported by both the forward and the backside ground.
Additionally, from the deformation distributions of the
piles, the shallow part of the front-most pile deformed con-
siderably (arrow A) due to a large subgrade reaction from
the shallow front side of the pile, while the deep part of the
back-most pile deformed largely (arrow B) due to a large
subgrade reaction from the deep backside of the pile.

4.9. Influence of subgrade reaction on load share mechanism The front-most pile Crack

The cause of the above load share distribution and its


behavior are discussed here using the reaction force of
the piles. As shown in Fig. 28, the subgrade reaction is cal-
culated as rx (stress of the loading direction) at the front Photo 4. Crack behind front-most pile.
834 S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837

Reaction piles 7×9 pile group

Gap of jack and base


The back-
most pile

Settlement
160 cm 35 cm of base
20 cm

Position of
remaining piles
under ground

Photo 5. Subsidence behind back-most pile. Photo 6. Situation of loading system at maximum load.

Ground surface Ground surface


2.5 m = 6.3d 1.7 m = 4.3d

10 20 40 60 80 10 20 40 60 80 90
6 m = 15d 10 m = 25d
0
80 90
6.2 m = 15.5d
5.2 m = 13d

1 60 80 170 mm
2
Depth (m)

40 60 subsidence
3
4 20 40
5 10 20
6
7 10
8
4,000 kN (6.4 mm = 1.6%×d) 25,400 kN (6.4 mm = 42%×d)

Ground surface
6080 Loading direction
40
20 Lateral displacement of ground normalized by
1.6 m = 4d front back
10 lateral displacement of pile at ground surface (%)
0 0 20 40 60 80 100
= 7.3d
2.9 m

1
2 section
Depth (m)

3
4
5 2.4 m Position of the maximum bending moment
= 6d front back
6
7
8
200 kN (6.4 mm = 1.6%×d)

Fig. 26. Distribution of displacement of ground around 7  9 pile group (FEM).

and rear faces of the hybrid element. Subsequently, the tribution of initial stress rx0 and the stress at 16,000 kN, rx,
reaction force of the pile is calculated by the vertical inte- are shown in Fig. 30.
gration of the subgrade reaction (from the ground surface In this figure, the compression in front of the pile results
to the pile tip). in an increase in the reaction force, but the compression
The reaction forces at the front face and the rear face of behind the pile results in a decrease in the reaction force.
the pile are shown in Fig. 29(a), while the total of these From the results of the reaction force at the front face, the
reaction forces is shown in Fig. 29(b). Additionally, the dis- front-most pile receives the largest reaction force and the
S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837 835

Deformation of pile (mm) tN increment (kPa) 4,000 kN


8 7 6 5 4 3 2 1 0 -20 0 20
1
0 -5 -5 -5 -5 -5
-1
A
-2 20
Depth (m)

-3 10 -2.85 m
-2.85 m B 10 10 5
5 5
-4 5 5
5 -5 -5

10 m
-5 5 5 -10
5 -5 -10
-6 Front-most pile
-7 -5 -5
Middle pile
-8
Back-most pile
-9 Remaining piles Front-most pile Middle pile Back-most pile
-10

Fig. 27. Relation between ground stress distribution around pile group and deformation of piles (FEM).

Loading direction Pile heart beam element that


has 9/10 EI of actual pile

Element for calculation of subgrade Element for calculation of subgrade


reaction at front face of pile reaction at rear face of pile

Column element solid element Not used for calculation due


that has 1/10 EI of actual pile to small shear force

Fig. 28. Elements using calculation of subgrade reaction.

back-most pile receives a negative reaction force. This is due In this study, the resisting forces of the in-service
to the deformation of the ground in front of the pile shown in practical foundation were checked for the first time with
Fig. 30. This negative reaction shows the decrement of the a simulation method capable of assessing the behavior of
initial stress and does not exceed the initial stress. laterally loaded large-scale pile groups.
The results of the reaction force at the rear face indicate
that the posterior pile has a large reaction force. The reac- (1) By using predictions from the FEM analysis of the
tion force at the rear face does not become greater than soil tests, and lateral loading tests performed before
that at the front face. This is because the reaction force the soil tests, the load displacement results for 7  9
at the rear face cannot exceed the initial stress. Therefore, and 3  1 pile groups were predicted with accuracy,
the reaction force at the rear face of the back-most pile is including the shape of the unloading and reloading
almost equal to the reaction force at the front face of the cycles. This has helped render an appropriate deter-
front-most pile in the initial phase. However, depending mination of the test conditions. An analysis per-
on the loader, it can be relatively small. formed after the tests that improved the pile
This mechanism of the reaction forces explains the modeling was able to simulate the behavior of the
almost symmetrical U-shape in the initial phase for the 7  9 pile group, after yielding, to some extent.
load share ratio distribution shown in Fig. 25. However, (2) The foundation yield around the load was calculated
the shape changes as the load-share ratio of the back- with Chang’s formula along with the increase in load
most piles decreases. to the ultimate load of 25,400 kN. At the ultimate
load, the falling of the pile head from the slab and
the falling away of the slab skin were predominantly
5. Conclusion observed at the front-most pile. However, brittle fail-
ure was not observed.
In-situ lateral loading tests on large-scale pile groups (3) The deformation of the piles in the FEM results was
and a 3D elasto-plastic FEM analysis of the tests were con- found to correspond to the deformation of the
ducted to investigate the bearing capacity of an in-service pulled-out piles after the tests. The difference between
practical LNG tank foundation and the mechanical behav- each line was found to be close to that seen in past in-
ior of the pile groups. situ lateral loading tests on a 3  3 pile group.
836 S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837

4,000 kN 8,000 kN 12,000 kN 16,000 kN


Front Front Front Front
face face face face
Total Total Total Total
Rear Rear Rear Rear
face face face face
500 500

450 450

400 400
(1) (7) 350 (1) (7)
350
300
Reaction force of pile (kN)

Reaction force of pile (kN)


300
250 250
front back front back
200 200

150 150

100 100

50 50

0 0

-50 -50
-100 -100
-150 -150
(1) (2) (3) (4) (5) (6) (7) (1) (2) (3) (4) (5) (6) (7)
(a) Reaction force at front and rear face (b) Total of reaction force

Fig. 29. Evolution of reaction force of each pile.

Loading direction
x: Stress of loading direction at 16,000 kN x0 : Initial stress of loading direction

Stress of loading direction x kPa Stress of loading direction x kPa


200 150 100 50 0 0 50 100 150 200 200 150 100 50 00 50 100 150 200

Stress decrease at rear face


(positive reaction force)
Stress increase
at front face
(positive reaction force)

Stress increase Stress decrease


Stress increase at rear face at front face Can t decrease
at front face (negative reaction force) (negative reaction force) under 0
20 m

20 m
(positive reaction force)

Stress decrease in total


(negative reaction force)

(a) Fro nt-mo st p ile (b) Back-mo st p ile

Fig. 30. Transition of ground stress distribution at front-most and back-most piles.

(4) The affected range of the ground (lateral and vertical (5) The maximum subgrade reaction of each pile for the
displacement) increased in all directions with the framework analysis design in the Specification for
7  9 pile group compared to the 3  1 pile group. Highway Bridges is different from the load share dis-
These results indicate the scale effect of the pile tribution of the 7  9 test and the FEM results; how-
group. ever, their averages are nearly equal. Therefore, the
S. Teramoto et al. / Soils and Foundations 58 (2018) 819–837 837

reduction rate of the maximum subgrade reaction at Isobe, K., Kimura, M., Ohtsuka, S., 2014. Design approach to a
the back piles (= 0.5 currently) must be properly method for reinforcing existing caisson foundations using steel
determined when designing each pile. pipe sheet piles. Soils and Foundations 54 (2), 141–154.
(6) According to the yield of the pile heads of the front- Japan Road Association, 2012. Specifications for Highway
Bridges, Part IV Substructures.
most piles, the deformation of the ground around
Kimura, M., Kosa, K., Morita, Y., 1994. Full scale failure tests
these piles changed modes. Thus, a different tendency
on lateral loading cast-in-place concrete piles. In: Proc. of the
was seen at the back-most piles. 5th Int. Conf. and Exhibition on Piling and Deep Founda-
(7) From the stress distribution results, the front-most tions, Bruges, 5.15.1–5.15.10.
piles were supported by the forward ground, while Kimura, M., Adachi, T., Kamei, H., Zhang, F., 1995. 3-D finite
the back-most piles were supported by the forward element analysis of the ultimate behavior of laterally loaded
and backside ground. This variation caused a differ- cast-in-place concrete pile. In: Pande, Pietruszczak (Eds.),
ence in the pile deformation at the pile positions. Proc. of 5th Int. Symp. Numerical Models in Geomechanics,
(8) The back-most pile received the reaction force as a pp. 589–594.
decrement in stress due to the initial stress, since that McVay, M., Zhang, L., Molnit, T., Lai, P., 1998. Centrifuge
quantity had limitations. Therefore, according to the testing of large laterally loaded pile groups in sands. J.
Geotech. Geoenviron. Eng. 124 (10), 1016–1026.
increase in load, the total reaction force at the back-
Nakai, T., Hinokio, M., 2004. A simple elasto-plastic model for
most pile was small compared to that at the front-
normally and over consolidated soils with unified material
most pile. This difference in reaction forces is why parameters. Soils Found. 44 (2), 53–70.
the load-share ratio distribution had an almost sym- O’Neil, M.W., Dunnavant, T.W., 1985. An evaluation of the
metrical U-shape in the initial phase. However, the behavior and analysis of laterally loaded pile groups. A Report
shape changed as the load-share ratio of the back- to the American Petroleum Institute, University of Houston-
most piles decreased. University Park, Research Report No. UHCE 85-11.
Pedro, F., Frank, C., 1997. Evaluation of laterally loaded pile
In the next study, the applicability of the above analysis group at Roosevelt bridge. J. Geotech. Geoenviron. Eng. 123
to large-scale pile groups will be shown under a one- (12), 1153–1161.
boundary condition in in-situ lateral loading tests. Chang- Poulos, H.G., 1964. Behavior of laterally loaded piles. II – Pile
groups. Proc. of ASCE, vol. 90, no. EM3, pp. 223–254.
ing the pile number by means of this analysis method, a
Randolph, M.F., 1981. The response of flexible piles to lateral
parametric analysis will be conducted and the influence
loading. Geotechnique 31 (2), 247–259.
of the pile number on the pile-group effect will be investi- Shibata, T., Yashima, A., Kimura, M., 1989. Model tests and
gated quantitatively. analyses of laterally loaded pile groups. Soils Found. 29 (1),
Subsequently, the mechanical behavior of large-scale 31–44.
pile groups subjected to seismic loads will be investigated Shinohara, M., Kanaji, H., Oniki, K., Kimura, M., 2013. A
with this method along with the pile number being applied proposal of a bridge column integrated by multiple steel pipes
to the design of pile groups. with directly-connected piles and a seismic response analysis. J.
JSCE, C 69 (3), 312–325.
Shirato, M., Nonomura, Y., Fukui, J., Nakatani, S., 2008. Large-
Acknowledgements scale shake table experiment and numerical simulation on the
nonlinear behavior of pile-groups subjected to large-scale
This was a collaborative research effort by Osaka Gas earthquakes. Soils Found. 48 (3), 375–396.
Co., Ltd, Kyoto University, and Obayashi Co., Ltd. Sin- Tamaki, O., Mitsuhashi, K., Imai, T., 1971. Horizontal resistance
cere thanks are given to T. Nishizaki of Osaka Gas Co., of a pile group subjected to lateral load. J. JSCE 192, 79–89.
Ltd and A. Inoue of Obayashi Co., Ltd, for their valuable Teramoto, S., Kimura, M., 2016. Investigation of pile group effect
suggestions. Thanks are also extended to the interested par- subjected to influence of pile arrangement and pile stiffness.
ties for their support and encouragement. Jpn. Geotech. Soc. Spec. Publ. 2 (38), 1362–1367.
Ye, B., Ye, G.L., Zhang, F., Yashima, A., 2007. Experiment and
numerical simulation of repeated liquefaction-consolidation of
References sand. Soils Found. 47 (3), 547–558.
Zhang, F., Kimura, M., Nakai, T., Hoshikawa, T., 2000.
Adachi, T., Kimura, M., Kobayashi, H., Morimoto, A., 1994. Mechanical behavior of pile foundations subjected to cyclic
Behavior of laterally loaded pile groups in dense sand. Proc. lateral loading up to the ultimate state. Soils Found. 40 (5), 1–
Int. Conf. Centrifuge 94, 509–514. 17.
Bao, X., Morikawa, Y., Kondo, Y., Nakamura, K., Zhang, F., Zhang, F., Kimura, M., 2002. Numerical prediction of the
2013. Shaking table test on reinforcement effect of partial dynamic behaviors of an RC pile group foundation. Soils
ground improvement for pile group foundation and its Found. Jpn. Geotech. Soc. 42 (3), 77–92.
numerical simulation. Soils Found. 52 (6), 1043–1061. Zhang, F., Yashima, A., Higuchi, M., Kimura, M., 2002. 3-D
Hashiguchi, K., 1980. Constitutive equation of elastoplastic analysis on dynamic behavior of jacket-type foundation with
materials with elasto-plastic transition. J. Appl. Mech., ASME inclined piles. In: Proc. of 1st Int. Workshop on New Frontiers
102 (2), 266–272. in Computational Geotechnics, Banff, pp. 123–128.

You might also like