Japanese Geotechnical Society Special Publication The 15th Asian Regional Conference on

Soil Mechanics and Geotechnical Engineering

Nonlinear analysis of seismic response of a base isolated building on a piled raft foundation with
grid-form ground improvement

Yoshimasa Shigeno i)

i) Chief Researcher, Research & Development Institute, Takenaka Corporation, 1-5-1, Otsuka, Inzai, Chiba 270-1395, Japan.

ABSTRACT

A seismic observation has being carrying out at a building on a piled raft foundation with grid-form ground
improvement. The building is located in Tokyo, and the observation records have been successfully obtained during
the 2011 off the Pacific Coast of Tohoku Earthquake. The observed earthquake at this site is ranked in a middle scale.
A seismic response analysis using a nonlinear ground model is conducted for this record. An elasto-plastic model
based on a subloading Mohr-Coulomb model is used for the ground. In this model, the G-γ and h-γ relations are
directly used to change the status of the subloading surface. A 3D fine finite element mesh model is used with two
directional input motions. The simulation results agree well with the observations. The validity of the nonlinear
model for the middle scale earthquake is confirmed.

Keywords: piled raft foundation, seismic response analysis, elasto-plastic model, the 2011 off the Pacific Coast of
Tohoku Earthquake

1. INTRODUCTION the analysis model for a middle scale earthquake was

A seismic observation has being carrying out at the
residential building in Tokyo, Japan. The building has a
piled raft foundation with grid-form ground
improvement. The grid-form ground improvement is to
cope with the liquefiable sand as well as to improve the
bearing capacity of the raft foundation (Yamashita et al.
(2011)).
Many studies on the behavior of a piled raft
foundation during earthquakes have been conducted.
However only a few case histories exist on the
monitoring of the soil-pile-structure interaction
behavior during earthquakes such as Mendoza (2000).
Therefore our observation is important to investigate
the seismic performance of piled raft foundations.
The observation records during the 2011 off the
Pacific Coast of Tohoku Earthquake have been
successfully obtained at this building (Hamada et al.
(2012), Yamashita et al. (2012)). Axial force and
bending moment of two piles, earth pressure and
pore-water pressure beneath the raft, and accelerations
of the ground and structure were measured. The
observation records are ranked in the middle scale
earthquake at this site.
In the previous study, the seismic response analysis
was done using an equivalent linear model for the
ground. The analytical results agreed well with these
observation records as shown in Onimaru et al. (2012)
and Hamada et al. (2014). From this result, validity of
confirmed.
Currently, we are planning to investigate the
response for a large scale earthquake based on this
model. However the equivalent linear model cannot
handle large strain. A nonlinear model is needed for the
ground. By applying the appropriate non-linear model
for the ground, a seismic response analysis is conducted
for the 2011 off the Pacific Coast of Tohoku
Earthquake in this study. The analysed seismic
response is compared with the observation records and
the validity of the nonlinear model is shown.
2 OVERVIEW OF THE BUILDING, GROUND
AND EARTHQUAKE OBSERBATION
The building analysed is a residential building in
Tokyo. It is a reinforced concrete structure with a base
isolation system. The plane section is 30.05 m by 33.25
m, the height is 38.7 m, and the number of stories is 12.
It has a piled raft foundation consisting of a raft, piles
and grid-form ground improvement. The piles are
sixteen precast concrete piles with diameters of 0.8 m
to 1.2 m and a length of 45 m.
Fig. 1 shows a schematic view of the building and
its foundation with the soil profile. The soil profile
down to a depth of 7 m is fill and loose silty sand. And
from 7 m to 44 m, the profile is very soft to medium
silty clay. The ground water table appears
approximately 1.8 m below the ground surface. The

http://doi.org/10.3208/jgssp.JPN-006 1372
positions of seismic monitors are also shown in the Lateral : Cyclic boundary

figure. Z Lateral : Cyclic boundary

Y (UD) X
(NS) (EW)
◆12FL

38.7m
Triaxial
accelerometer

SPT 75m
N-Value Borehole-type
triaxial
0 50 accelerometer
◆1FL
◆ Lateral : Cyclic boundary
◆ 1.5m
4.8m ●
5.8m ●
6.0m Lateral : Cyclic boundary
BFL
60m
60m
◆
16.0m ●
16.0m 16.0m
●

15.0m
Grid-form D.M.W. 33.25m 30.05m
Bottom : Viscous boundary
27.1m 60m 60m

Fig. 2 FE mesh of the ground-structure interaction model

● ●
46.5m Table 1 Material characteristics of piles
48.0m ◆
50.0m 50.0m

Pile diameter (mm) 800 1000 1200

60.0m
Settlement gauges Young's modulus (MN/m2) 40000 40000 40000
Damping (%) 2 2 2
Ae of SC pile (m2) 0.3268 0.4649 0.6714
Fig. 1 Schematic view of building and foundation with soil Ie of SC pile (m4) 0.02199 0.04899 0.10316
profile Ae of PHC pile 2
0.2441 0.3633 0.5054
(m )
Ie of PHC pile (m4) 0.01455 0.03437 0.06958
3 ANALYSIS MODEL Ae : Equivalent cross-sectional area
Ie : Equivalent moment of inertia of area
3.1 Analysis condition
Fig. 2 shows the finite element (FE) mesh. The 6.02 m 8.4 m 8.4 m 6.03 m
number of elements is 213,622 and the number of
degrees-of-freedom (DOFs) is 656,543. The building is
5.8 m

C B B C Grid-form ground
modeled by elastic bars and shells. The piles are also improvement
(W = 0.8m)
modeled by elastic bars. Table 1 shows the material of
8.85 m

the piles. B A A B
Fig. 3 shows the top view of the FE mesh under the Pile
raft. To consider the shape and volume of the piles, Pile 5B Pile 7B
9.3 m

cavities in the shape of the piles are made in the FE

B A A B
model. The nodes of the piles and the adjacent ground Pile
A:φ = 1,200
nodes at the same depth are bound by rigid bar
8.1 m

B:φ = 1,000
elements. The base isolation layer is modeled by C B B C C:φ = 800
Y (N) L = 45 m
tri-linear spring elements.
X (E)
The lateral boundaries are periodic boundaries.
They are positioned at 60 m outside of the building to Fig. 3 Magnified top view of FE mesh under the raft
minimize the boundary effect. The bottom is a viscous
boundary. 100
max = (112.09, 66.083)
EW direction
The input ground motion is based on the observation 50
Acc. (gal)

0
record of the 2011 off the Pacific Coast of Tohoku -50
min = (110.31, -78.369)
Earthquake. Two directional input motions that are the -100
Time (sec)
100 max = (110.42, 86.559)
EW wave and the NS wave are applied simultaneously. 50
NS direction
Acc. (gal)

The time interval is 0.01 s and the analysis time is 300 s. 0

Fig. 4 shows the input accelerations in the EW and NS -50
min = (110.82, -89.172)
-100
directions. 0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300
The analysis code is an in-house program called Time (sec)
MuDIAN developed by Shiomi et al. (1998). This code
is parallelized by the hybrid parallel method and is able Fig. 4 Input accelerations at GL-75 m. The waves are 2E.
to calculate a large-DOF model with high speed as
described in Shigeno et al. (2014).

1373
3.2 Constitutive model for ground c = G0 γ 50 (3)
A multi-hardening model proposed by Shiomi and
Fujiwara (2014) is used as the constitutive model for The grid-form ground improvement is a liquefaction
the ground. This model is based on the countermeasure. Liquefaction is expected to occur in
multi-dimensional Yoshida model (Ishihara et al. the silty sand form GL-5.0 m to GL-8.0 m. However,
(1985), Yoshida and Tsujino (1993)). The Yoshida liquefaction was not actually observed during the 2011
model is characterized using the G-γ and h-γ off the Pacific Coast of Tohoku Earthquake at this site.
characteristics directly as input data. The Yoshida From this fact, liquefaction is not considered in the
model uses a non-linear elastic model for the analysis.
stress-strain relation. This causes overestimation of the Even if the ground is not liquefied, the pore water
strain under multi directional input-motion. A pressure slightly rises and the shear modulus is reduced.
multi-hardening model uses an elasto-plastic model for However, the thickness of the silty sand layer is not too
the stress-strain relation. Therefore the direction of large compared with the model. The effect of the shear
softening can be considered. modulus reduction is assumed to be limited.
Nonlinearity before the stress reaches yield state is Ground Shear velocity, Vs (m/s)
0 200 400 600 800
expressed by a hardening coefficient H’ defined by 0
Ground Layer 2
equation (1). GT is the tangential shear modulus that is Optimized initial Vs Ground Layer 1

evaluated from the τ-γ curve. The τ-γ curve under the 10 PS logging Vs

simple shear condition is obtained from the G-γ Ground Layer 2

20
characteristics. The hardening coefficient H’ appears in
30
Depth (m)
the denominator of the plastic multiplier in equation
(2). 40 Ground Layer 3

GT (1)
H' = 50
G
1− T 60
G0 Ground Layer 4

70
∂f
T
De dε
dλ = ∂σ (2) 80

∂f ∂f
T
H' + De
∂σ ∂σ Fig. 5 Optimized initial shear velocity and PS logging shear
where De is an elastic modulus tensor, and f is a yield velocity
function.
1 25
3.3 Modeling of ground and ground improvement 0.9
The shear wave velocity distribution in the ground 0.8 20
layer is obtained by an optimization method. The 0.7
error in the transfer function is adapted as a cost Ground Layer 1
0.6 15
Ground Layer 2
function. Observation records of small earthquakes that h（％）
G/G0

0.5 Ground Layer 3

occurred before the 2011 off the Pacific Coast of 0.4 Ground Layer 4 10
Tohoku Earthquake are used in the optimization. The 0.3 Ground Improvement
PS logging results are applied as the initial VS 0.2 5
distribution of the optimization. The initial shear 0.1
modulus distribution for the analysis is created from the 0 0
optimized VS distribution. Fig. 5 shows the distribution 0.0001 0.001 0.01 0.1 1 10
of the optimized VS. γ (%)
The nonlinear characteristics of the ground were
obtained from the cyclic triaxial tests of the samples. Fig. 6 Strain dependence characteristics of the ground and the
ground improvement
Fig. 6 shows the G-γ and h-γ relations of each soil layer
at this site.
The subloading Mohr-Coulomb model is used for 3.4 Modeling of ground improvement
The modulus and the strength of the ground
the stress and strain relation of the multi-hardening
improvement are set according to the 28-day strength of
model. The parameters related to shear strength are
the samples. The initial shear modulus is set at 700
needed. From the previous study (Hamada et al. (2014)),
MPa (ρ = 2.0 t/m3, VS = 592 m/s) and the uniaxial
the effect of the confined stress is negligible. Therefore,
strength is set at 1.8 MPa referred to Saito and Konishi
the shear strength is given as cohesion. The cohesion of
(2010).
layers except for being investigated is assumed to
The nonlinearity of the ground improvement is
follow equation (3).

1374
assumed to follow the modified HD model expressed Acceleration (cm/s2) Displacement (cm)
by equation (4). -200-150-100-50 0 50 100 150 200 -10 -8 -6 -4 -2 0 2 4 6 8 10
40 40
G0  G  30 30
τ= γ h = hmax 1 −  (4) 20 20
1 + (γ γ 0.5 )
a
 G0  10 10
0 0
where, α is a parameter. The parameter is set by

Depth (m)
Depth (m)
-10 -10
-20 -20
simulating the uniaxial compression test. The maximum -30 -30
damping hmax is assumed to be 20% equivalent to the -40 -40
ground. As a result of the simulation, α is set at 1.06. -50 -50
Fig. 6 shows the G/G0-γ and h-γ relations for the ground -60
-70
-60
-70
improvement. -80 -80
Nonlinear Equiv. linear Nonlinear Equiv. linear
Observation Observation Observation Observation
4 RESULTS
Fig. 7 shows the distribution of the peak acceleration (a) Acceleration (b) Relative displacement
and relative displacement in the EW direction. The Fig. 7 Distribution of peak acceleration and relative displacement
results of the equivalent linear analysis are also shown. (EW direction)
The analytical acceleration agrees well with the
observation except for the ground at GL-1.5 m. 200 max = (110.6, 192)
Acceleration (cm/s2)

GL-1.5m
However as shown in Fig. 8, the analytical and 100
0
observed values show no significant difference. -100
The peak distribution of the relative displacement -200 min = (110.7, -151)
agrees well with the observations. Especially, the 100 110 120 130 140

values of the ground agree well. The max displacement Time (sec)
Displacement (cm)

of the building is slightly small. The small sift toward 4

max = (109.0, 3.18) GL-1.5m
2
the minus direction around the max value causes this
0
distribution. -2
Fig. 9 shows the distribution of the peak acceleration min = (117.4, 3.58)
-4
and relative displacement in the NS direction. The 100 110 120 130 140
Time (sec)
analytical results also agree well with the observations.
Observation Analysis
In comparison with the equivalent linear analysis,
the results of nonlinear analysis are almost the same. Fig. 8 Time history of the seismic response of the ground away
The peak shear strain of the ground in the EW direction from the building at GL-1.5 m (EW direction). The response
is 0.16%. The minimum value of G/G0 derived from the shown is from 100 s to 140 s. The time frame includes the main
shock.
G-γ curve shown in Fig. 6 is about 0.5. This indicates
that the equivalent linear analysis can satisfactorily
Acceleration (cm/s2) Displacement (cm)
evaluate the response in this degree of nonlinearity. -200-150-100-50 0 50 100 150 200 -10 -8 -6 -4 -2 0 2 4 6 8 10
Fig. 10 shows the distribution of the peak bending 40 40
moment of the piles in the EW direction. According to 30 30
20 20
the observation, the bending moment of the pile head of 10 10
5B is larger than that of 7B. And the peak value of the 0 0
Depth (m)
Depth (m)

pile head of 5B is larger than the value at the lower end -10 -10
-20 -20
of the ground improvement (GL-16 m). The analytical -30 -30
results also show these tendencies. -40 -40
Fig. 11 shows the time histories of the incremental -50 -50
-60 -60
bending moment of pile 5B from 100 s to 150 s. The
-70 -70
observed values indicate the same phase at the lower -80 -80
end of the ground improvement (GL-16 m) and near the Nonlinear Eqiv. linear Nonlinear Equiv. linear
Observation Observation Observation Observation
pile head (GL-6 m). The analytical results also exhibit
(a) Acceleration (b) Relative displacement
the same phase between the two points.
Fig. 9 Distribution of peak acceleration and relative displacement
(NS direction)

1375
 5B Bending Moment (kNm) 7B Bending Moment (kNm)
REFERENCES
-600 -400 -200 0 200 400 600 -600 -400 -200 0 200 400 600 1) Hamada, J., Tanikawa, T., Onimaru, S. and Yamashita, K.
0
(2012): Seismic observations on piled raft foundation with
ground improvement supporting a base-isolated building,
-10 Proceeding of 15thWCEE
2) Hamada, J., Shigeno, Y., Onimaru, S., Tanikawa, T.,
Depth (m)

-20 Nakamura, N., Yamashita, K. (2014): Numerical analysis on

seismic response of piled raft foundation with ground
-30 improvement based on seismic observation records,
Proceeding of 14thIACMAG
-40 3) Ishihara, K., Yoshida, N. and Tsujino, S. (1985): Modeling of
stress-strain relations of soils in cyclic loading, Proc. 5th
International Conference for Numerical Method in
-50
Nonlinear Equiv. linear Nonlinear Equiv. linear Geomechanics, Nagoya, Vol.1, pp.115-143
Observation Observation 4) Mendoza, M. J., Romo, M.P., Orozco, M. and Dominguez, L.
(2000): Static and seismic behavior of a friction pile-box
Fig. 10 Distribution of peak bending moment of the piles (EW
foundation in Mexico City clay, Soils & Foundations, Vol.40,
direction)
No.4, pp.143-154
5) Onimaru, S., Hamada, J., Nakamura, N. and Yamashita, K.
Bending moment (kNm)

600 (2012): Dynamic soil-structure interaction of a building

max = (109.0, 335) GL-6m
300
supported by piled raft and ground improvement during the
0
2011 Tohoku Earthquake, Proceeding of 15thWCEE.
-300
min= (117.5, -293) 6) Saito, S. and Konishi, K. (2010): Proceeding of the 9th
-600
100 110 120 130 140 national symposium on ground improvement, Elastic constant
Time (sec) for design of cement-improved soil, pp.287-292 (in
Japanese)
Bending moment (kNm)

600
max = (109.0, 231)
300
0 (1998): Development of structural analysis system
-300
-600
100 110 120
Time (sec)
Observation
Fig. 11 Time history of the bending moment of 5B (EW
direction). The response shown is from 100 s to 140 s. The time
frame includes the main shock.
5 CONCLUSIONS
A nonlinear earthquake response analysis with a
detailed 3D ground and structure interaction model has
been conducted. The input motion is based on the
observation records of the 2011 off the Pacific Coast of
Tohoku Earthquake. A multi-hardening model is used
as the constitutive model of the ground. The model
incorporates the elasto-plastic model into the Yoshida
model. The analysis results agree well with the
observation records of the 2011 off the Pacific Coast of
Tohoku Earthquake. The earthquake is ranked in the
middle scale at the observation site. Therefore, this
result indicates the validity of this analysis model to a
middle scale earthquake. This becomes a good
preparation for the evaluation of a large scale
earthquake.
GL-16m 7) Shiomi, T., Yoshizawa, M., Onimaru, S. and Tsukuni, S.
min= (117.5, -249) considering non-linear behaviors of soil and structure,
Takenaka technical research report, No.54, pp.1-8 (in
130 140 Japanese)
8) Shiomi, T. and Fujiwara, Y. (2014): Liquefaction analysis of
Analysis
Urayasu-site by the multi-hardening model, Proceeding of
the special symposium of JGS -Overcoming the Great East
Japan Earthquake-, pp.141-145 (in Japanese)
9) Shigeno, Y., Hamada, J., Nakamura, N. (2014): Hybrid
parallelization of earthquake response analysis using K
computer, Proceeding of 14thIACMAG
10) Yamashita, K., Hamada, J. and Yamada, T. (2011): Field
measurements on piled rafts with grid-form deep mixing
walls on soft ground, Geotechnical Engineering Journal of
the SEAGS & AGSSEA, Vol.42, No.2, 1-10.
11) Yamashita, K., Hamada, J., Onimaru, S. and Higashino, M.
(2012): Seismic behavior of piled raft with ground
improvement supporting a base-isolated building on soft
ground in Tokyo, Soils & Foundations, Special Issue on
Geotechnical Aspects of the 2011 off the Pacific Coast of
Tohoku Earthquake, Vol.52, No.5, 1000-1015.
12) Yoshida, N. and Tsujino, S. (1993): A simplified practical
stress-stain model for the multi-dimensional analysis under
repeated loading, The 28th Japan national conference on soil
mechanics and foundation engineering, pp.1221-1224 (in
Japanese)

1376