Bulletin of Earthquake Engineering

https://doi.org/10.1007/s10518-023-01711-0

ORIGINAL ARTICLE

Centrifuge model tests on liquefaction mitigation effect

of soil‑cement grids under large earthquake loadings

Yuan Cao1 · Yuhei Kurimoto2 · Yan‑Guo Zhou3 · Akira Ishikawa2 ·

Yunmin Chen3

Received: 17 August 2022 / Accepted: 12 May 2023

© The Author(s), under exclusive licence to Springer Nature B.V. 2023

Abstract
In this study, the seismic response and liquefaction mitigation effect of the soil-cement
grid improved ground subjected to large earthquake loadings are studied through
dynamic centrifuge tests. The model tests included soil-cement grid improved and unim-
proved model grounds, both of which have a 15-m-thick liquefiable layer underlain by a
2.5-m-thick coarse sand layer. The pre-cast soil-cement grid adopted in this study enables
the dynamic responses of the model closer to the real improved ground. The recorded
responses of accelerations, excess pore pressures and the deformation of the enclosed soil
in the improved ground were carefully analysed with the comparison of the ground without
improvement. It shows that the soil liquefaction and post-shaking settlements were effec-
tively mitigated by the soil-cement grid even under very strong shakings. And the restric-
tion effect of the soil-cement grid on dynamic shear strain of the enclosed soil was the most
prominent in the middle height, regardless of the intensities of the shaking events. Such
mitigating “waist effect” could mainly be attributed to the dynamic soil-grid interaction
during shaking. However, the underlain soil layer may experience larger shear strain due
to the increasing inertial force of the overlying ground improved by the soil-cement grid.

Keywords Centrifuge modelling · Ground improvement · Liquefaction mitigation · Soil-

cement grid

List of symbols
ρs Soil particle density
Cv Coefficient of consolidation
D50 50% Diameter on the grain size diagram
Fc Fine fraction content of the soil

* Yan‑Guo Zhou
[email protected]
1
MOE Key Laboratory of Soft Soils and Geoenvironmental Engineering, Institute of Geotechnical
Engineering, Zhejiang University, Hangzhou 310058, People’s Republic of China
2
Institute of Technology, Shimizu Corporation, Tokyo, Japan
3
MOE Key Laboratory of Soft Soils and Geoenvironmental Engineering, Center for Hypergravity
Experiment and Interdisciplinary Research, Institute of Geotechnical Engineering, Zhejiang
University, Hangzhou 310058, People’s Republic of China

13
Vol.:(0123456789)
 Bulletin of Earthquake Engineering

Mw Moment magnitude of earthquake

Uc Uniformity coefficient of the soil
emax Maximum void ratio of the sand
emin Maximum void ratio of the sand
k Permeability coefficient of the soil
ru Excess pore pressure ratio

1 Introduction

In recent years, devastating earthquakes occurred frequently with the moment magnitudes
­(Mw) of 7.9 for the 2008 Wenchuan earthquake in China (Zhou et al. 2018a), 7.1 for the
2010 Canterbury earthquake in New Zealand (Chen and Faccioli 2013) and 9.1 for the
2011 Tohoku earthquake in Japan (Goda et al. 2013) respectively. A large number of build-
ings were tilted and collapsed due to damage of the foundation structure caused by earth-
quake loadings beyond the seismic design. The Recommendations for Design of Building
Foundations in Japan was revised in 2019, and a secondary design system of the founda-
tion structure was proposed based on the importance of buildings (e.g., Architectural Insti-
tute of Japan 2019). And it is recognized that soil liquefaction is always a major concern
in earthquake disasters (e.g., Bray and Dashti 2014; Zhou et al. 2020b). Thus, not only the
foundation structures but also the liquefaction countermeasures are required to be designed
in response to very large earthquakes.
During the past decades, several conventional methods were widely used for liquefac-
tion mitigation around the world, such as densification, drainage or stone columns (e.g.,
Garcia-Torres and Madabhushi 2019; Fallahzadeh et al. 2019; Zhou et al. 2021). Cement-
ing soil is a novel and well-recognized way of preventing liquefaction by mixing chemical
binders with liquefiable soil (Kitazume 2013; Mitrani et al. 2010, 2012). On this basis,
soil-cement grid has been developed and gradually applied in engineering practices as a
promising alternative liquefaction countermeasure (Sakai and Tazaki 2003), especially
for high-rise buildings (Japanese Geotechnical Society 1998). It is usually composed of
repeated cells, which are made up of overlapped soil columns stabilized by the deep mix-
ing method. The performance of soil-cement grids on liquefaction mitigation was well con-
firmed in the Hyogo-ken Nanbu earthquake in 1995 (Suzuki et al.1996; Tokimatsu et al.
1996) and the 2011 Tohoku earthquake (Kitazume 2013). Taya et al. (2008) proposed a
design method for the soil-cement grid that considered intervals, rigidity and depth of
cells. The seismic response of the improved ground was also calculated by the homogeni-
zation model and expressed by the liquefaction index (Wakai et al. 2004; Sato et al. 2005).
In consideration of the shear deformation of the enclosed soil, Ishikawa et al. (2012, 2016)
proposed a simple liquefaction evaluation method based on the homogenization theory
(Terada and Kikuchi 2003) and the so-called “initial shear strain” of the improved ground.
Nguyen et al. (2013) also evaluate the distribution of shear strain within enclosed soil by
linear elastic finite-element analyses.
However, the aforementioned design methods were all based on the linear elastic mod-
els, which could hardly reflect the elastoplastic responses of soil-cement grid improved
ground especially under large earthquake loadings. Several advanced elastoplastic mod-
els were applied in numerical simulations (Namikawa et al. 2007; Koseki and Namikawa
2010; Bradley et al. 2013) to acquire more realistic results. On the other hand, dynamic
centrifuge model tests have been used as an effective way to investigate the seismic

13
Bulletin of Earthquake Engineering

response and performance of the improved ground (Suzuki et al. 1991; Ishikawa and Asaka
2006; Takahashi et al. 2006; Taya et al. 2008; Rayamajhi et al. 2015; Khosravi et al. 2016,
2017; Hamada et al. 2016; Tamura et al. 2018; Mitrani and Madabhushi 2011, 2013),
which could also validate the numerical simulations. Suzuki et al. (1991) and Taya et al.
(2008) combined the cells with different spacing into a grid and established the relation-
ship between the seismic responses, such as the excess pore water pressures and geometric
parameters of the soil-cement grid. Ishikawa and Asaka (2006) and Rayamajhi et al. (2015)
studied the reinforcing effect of soil-cement columns in liquefiable sand, and concluded
that soil-cement columns were ineffective in reducing the potential for liquefaction trigger-
ing. Khosravi et al. (2016, 2017) focused on the seismic responses of the deep soft soil pro-
file reinforced by the soil-cement grid. In addition, Mitrani and Madabhushi (2011, 2013)
investigate the performance of soil-cement grid in a layered or partial-depth improved
ground as a liquefaction remediation method for a single-degree of freedom frame struc-
ture. Hamada et al. (2016) reproduced the failure behaviour of soil-cement grid with single
cell beneath the superstructure model made of steel plates. Nevertheless, the soil-cement
grid, as a liquefaction countermeasure, is always composed of multiple cells with the same
internal spacing in practical engineering. The existing centrifuge model tests have not cov-
ered this category and it is very necessary to address the case under very large earthquake
loadings.
In this study, two dynamic centrifuge model tests with soil-cement grid and another one
of free field ground without any improvement were performed to investigate the seismic
responses of the improved ground. The ground model adopted the pre-cast soil-cement grid
with assumed thickness, interval and area replacement ratio corresponding to the proto-
type. And the long-lasting shaking events with different amplitudes were used in the exper-
iments to meet the requirements of the secondary design of AIJ (2019). The experiment
results provide insights into the dynamic responses of the liquefiable ground improved by
the soil-cement grid and a well archived datasets for evaluating seismic design procedures
and numerical analysis methods.

2 Centrifuge model test

2.1 Model preparation and configurations

Three centrifuge model tests, including one unimproved ground model of Case FF-L1 and
two improved ground models of Case SC-L1 and Case SC-L2, were performed at a cen-
trifugal acceleration of 50 g by using the 4.5-m-radius centrifuge with an in-flight uniaxial
shaker at Zhejiang University, and the details of this centrifuge facility could be referred
to Zhou et al. (2018b, 2020a). Both types of models have a 15-m-thick liquefiable layer
underlain by a 2.5-m-thick dense coarse sand layer in prototype unit. The unimproved
ground model of Case FF-L1 was prepared in a laminar container with internal dimensions
of 595 mm (length) × 350 mm (width) × 500 mm (height), which allows the permanent
shear strains and behaves more like the free field condition. The improved models of Case
SC-L1 and Case SC-L2 with soil-cement grid were prepared in a rigid container with inter-
nal dimensions of 770 mm (length) × 400 mm (width) × 500 mm (height). The configura-
tions of both models are shown in Fig. 1, where the liquefiable soil (i.e., Fujian sand with
10% Kaolin clay) was enclosed by the pre-cast soil-cement grid with an area replacement
ratio of 25% in Case SC-L1 and Case SC-L2.

13
 Bulletin of Earthquake Engineering

Fig. 1  Model configurations: a soil-cement grid improved ground in Case SC-L1 and Case SC-L2 and b
unimproved ground in Case FF-L1

The soil-cement grid, in practice, always consists of multiple repeating cells, which is
difficult to be fully reproduced in a model test. In this study, the soil-cement grid having
six square cells in a two-by-three pattern was adopted to represent the central area of an
improved ground, as shown in Fig. 2. Thus, the boundary condition should be different
from a free-field condition and considered in the model preparation. Rigid container was
adopted for the improved models and the gap between the rigid container and the soil-
cement grid was filled with the coarse sand, then the stiffness of surrounding soil at both
transverse all sides would be higher and remained uniform even if any excess pore pressure
was generated during strong shaking. As for the narrow gap between the rigid container
and the longitudinal wall of soil-cement grid, the Teflon sheet and silicon grease were
applied to the inner sidewalls of the container to relieve the side friction effects, to mimic
the lower stiffness boundary condition compared to that in the shaking direction.

Fig. 2  Target region of soil-cement grid improved ground

13
Bulletin of Earthquake Engineering

The models were instrumented with accelerometers (ACCs), pore pressure transduc-
ers (PPTs), bender elements (BEs), and laser displacement transducers (LDTs) to record
dynamic responses. In both models, the arrays of ACCs (A1-A5) and PPTs (P1-P5 or
P1-P6) were placed in the centre of the model at different depths, and AX was attached
to the base plate of the container to record the input motion. Also, another two PPTs (P6,
P7 or P7, P8) were placed close to the sidewall of the container to evaluate the effect of
boundary conditions and LDTs were mounted above the model monitoring the settlements
of the soil surface. In the improved model, two ACCs (A6, A7) were glued on the top of
the soil-cement grid at the centre and the intersection of transverse and longitudinal walls.
The longitudinal wall mentioned herein is the wall parallel to the shaking direction, and the
transverse wall is that perpendicular to the shaking direction.

2.2 Test materials and input motions

Based on the previous research (Ishikawa et al. 2015), the clayey sand (i.e., Fujian sand
with 10% of Kaolin clay) with the relative density of 50% was adopted as the liquefiable
layer in this study, which was dry pluviated into each square cell. And the coarse sand was
tamped to a dense condition of the relative density of 80% as the base layer. It contributes
to the uniform saturation of the upper layer and serves as the underlain non-liquefiable
layer corresponding to engineering practices. The grain size distribution curves of the
clayey sand and coarse sand are shown in Fig. 3 and the corresponding physical proper-
ties are given in Table 1. The silicon oil with a density of 0.95 g/cm3 (25 °C) and viscosity
specified to be 50 times that of water was adopted as pore fluid to satisfy the scaling law.
The centrifuge model ground was de-aired in a vacuum container and saturated slowly by
silicon oil through the holes in the bottom of the container for a week.
In order to reproduce the typically deep mixing or jet grouting condition in situ, a
soil-cement mixture of water, cement, Fujian sand and Kaolin clay with the mass ratio
of 1.0:0.572:1.64:0.29 was used to fabricate the soil-cement grid. After being stirred by
the mixing machine for around half an hour, the mixture in a uniform slurry state was
poured into the dedicated mould, as shown in Fig. 4. Several cylindrical specimens with a
diameter of 5 cm and a height of 10 cm were made at the same time and cured under the

Fig.3  Grain size distribution curves of test materials

13
 Bulletin of Earthquake Engineering

Table 1  Physical properties of test materials

Property Symbol Unit Clayey sand Coarse sand

Soil particle density ρs g/cm3 2.65 2.65

Maximum void ratio emax – 1.14 0.80
Minimum void ratio emin – 0.60 0.57
Fine fraction content Fc % 15 0.2
Uniformity coefficient Uc – 40 1.4
50% diameter on the grain size D50 mm 0.16 1.31
diagram
Permeability coefficient k m/s 8.04 × ­10–6 7.14 × ­10–3

Fig. 4  Fabrication procedures of soil-cement grid

Table 2  Test arrangement Model Dr (%) Improvement Replace- Target PBA(g)

ment ratio
(%)

Case FF-L1 50 Free field 0 0.15

Case SC-L1 50 soil-cement grid 25 0.15
Case SC-L2 50 soil-cement grid 25 0.40

same conditions as the soil-cement grid to evaluate the strength of the cemented soil. The
unconfined compressive strengths of the soil-cement samples were around 4 MPa (curing
after 35 days), which was consistent with that in engineering practice (Suzuki et al. 1996).
The centrifuge models were subjected to sine waves with the peak base accelerations
(PBA) of 0.15 g (Case FF-L1 and Case SC-L1) and 0.4 g (Case SC-L2) in prototype unit
(see Table 2). The motion consists of the first 5 cycles with increasing amplitude, the
middle 50 cycles with constant amplitude at peak acceleration and the last 5 cycles with
decreasing amplitude. The frequency of the input motion was 1.2 Hz in prototype unit. The
long-lasting shaking event adopted in this study is used to examine some extreme cases

13
Bulletin of Earthquake Engineering

corresponding to mega-earthquakes which were set based on previous research (Ishikawa

et al. 2015). For example, the significant duration of earthquakes recorded on KiK-net sta-
tions (surface and borehole) in the 2011 Tohoku Earthquake could be up to 150 s (Gho-
frani and Atkinson 2015). As recommended by Idriss (1997), the equivalent number of
uniform shear stress cycles for real earthquakes increases dramatically when the magni-
tude is beyond ­Mw = 7.5. In the meantime, the dynamic response of the improved ground
under shorter shaking durations corresponding to M ­ w less than 7.5, such as 10 cycles or 15
cycles, could also be obtained in the time history of a long duration record. In addition, it
indeed requires a strong enough and long-lasting shaking to cause liquefaction in the soil-
cement grid improved ground so that the most unfavarable condition for soil-grid interac-
tion could be investigated in model tests.

3 Test results

The test results of each model are discussed in this section. The dynamic responses of
Case FF-L1 and Case SC-L1 were presented together to illustrate the effectiveness of soil-
cement grids on liquefaction mitigation. And Case SC-L2 was used to further study the
performance of the soil-cement grid when subjected to larger earthquake loadings. All data
were presented in prototype.

3.1 Acceleration responses

The acceleration time histories at different depths of the unimproved ground (Case FF-L1)
and the improved ground (Case SC-L1) are shown in Fig. 5a, b for the 0.15 g shaking
event. Each figure displays the motions recorded at different depths of the model (i.e., A1,
A2, A3, A4, A5, A6) and the input motion (AX). In all cases, the acceleration responses
of AX and A1 are identical even if the excess pore pressure generated in the coarse sand
layer. It ensures that the desired shakings were properly applied to the upper liquefiable
layer. The accelerations of the unimproved ground show great attenuation in all depths and
significant dilation spikes were observed at depth of 10 m. Moreover, the vibration could
hardly propagate upward after the shaking lasted around 20 s due to the severe liquefaction
in all depths.
In the improved ground, the accelerations of the enclosed soil at different depths were
almost same as the input motions in the first ten cycles of the shaking, after which the
attenuation of the acceleration was also observed similar to the unimproved ground. How-
ever, the vibration kept propagating upward and even gradually enhanced as the shaking
continued, indicating the partial recovery of soil stiffness in such a long-lasting shaking
event. The high similarity of acceleration responses between soil-cement grid surface
(A6) and the enclosed soil surface (A5) also indicates that the acceleration responses of
improved ground were highly consistent even if there was a stiffness difference between
the cemented and uncemented soil. Thus, the assumption of strain compatibility adopted
in many design methods for soil-cement grid improved ground might be effective under
earthquake loadings with an amplitude of 0.15 g. When subjected to a stronger shaking
event (e.g., PBA = 0.40 g), the acceleration responses of the enclosed soil show a similar
attenuation and recovery process with higher amplitudes, as shown in Fig. 5c. However,
the soil-cement grid and the enclosed soil could not deform the same in Case SC-L2. The

13
 Bulletin of Earthquake Engineering

Fig. 5  Acceleration time histories in a Case FF-L1, b Case SC-L1 and c Case SC-L2

13
Bulletin of Earthquake Engineering

peak acceleration of the soil-cement grid surface (A6) was about 1.5 times that of the soil
surface.
Acceleration response spectra with 5% damping ratio for input motions (AX) recorded
in Case FF-L1, Case SC-L1 and Case SC-L2 are shown in Fig. 6a. The dominant frequency
of the achieved input motion was 1.2 Hz as designed. And the input motion also contained
some high-frequency components enriching the frequency band of the designed motion.
The spectral acceleration of Case SC-L1 was almost the same as that of Case FF-L1 in the
whole range of frequencies. It shows that identical earthquake loadings for Case SC-L1
and Case FF-L1 were performed, which also ensures the comparability of these two cases.
With the higher PBA (0.40 g) of the input motion at the base, the spectral acceleration of
Case SC-L2 was approximately twice that of the other two cases and with a similar distri-
bution in the frequency domain.
The spectral amplification ratios of the liquefiable layer of Case FF-L1, Case SC-L1 and
Case SC-L2 are shown in Fig. 6b, c and d respectively. The spectral amplification ratios
are the spectra for motions recorded in the liquefiable soil layer divided by those in the
coarse sand layer (i.e., A1). For the moderate shaking event (PBA = 0.15 g), the low-fre-
quency components, around 0.93 Hz, were mostly exaggerated at the shallow layer in the
unimproved ground (Case FF-L1) due to the soil softening. And the significantly ampli-
fied high-frequency components at the depth of 10 m were primarily induced by the dila-
tion spikes under high confining pressures. However, in the improved ground, the spectral
amplification ratios were more consistent along the depth due to the restriction of the soil-
cement grid. And the spectral amplification ratios of the enclosed soil at middle depth were
closest to that of the soil-cement grid. That means the dynamic responses of the enclosed
soil at middle depth might be more representative than that of the surface for the soil-
cement grids improved ground. In addition, spectral accelerations in the improved ground
were amplified not only at low frequencies but also at relatively higher frequencies around
5 Hz under the moderate shaking event. It shows that the predominant frequency might be

Fig. 6  Spectral accelerations of a input motions, b amplification ratios in Case FF-L1, c Case SC-L1 and d
Case SC-L2

13
 Bulletin of Earthquake Engineering

increased in the soil-cement grids improved ground, which should be taken into consid-
eration for the design of superstructures. For the strong shaking event (i.e., PBA = 0.40 g),
only the low-frequency components around 0.9 Hz were amplified in the improved ground
due to the more severe soil softening.

3.2 Excess pore pressures

The excess pore pressure ratios (ru) measured in the liquefiable layer of Case FF-L1,
Case SC-L1 and Case SC-L2 are presented in Fig. 7. In the unimproved ground, when
subjected to the shaking event with PBA = 0.15 g, the excess pore pressure ratios
reached unity in all depths, implying the severe liquefaction occurred throughout the
liquefiable layer, and the peak values of ru stayed at unity for a long duration after the
shaking ceased. Under the same shaking event, the excess pore pressure generated much
slower in the improved ground and the peak values of ru only ranged from 0.7 to 0.8. It
is also worth noting that long-lasting shaking events were adopted in this model test,
which implies that the peak values of ru inside the soil-cement grid might be lower
in the conventional shaking events with a shorter duration (e.g., several to a dozen of
seconds). In addition, the excess pore pressures in the enclosed soil dissipated much
faster than that in the unimproved ground. All these phenomena confirmed the effective-
ness of the soil-cement grid on liquefaction mitigation. With higher shaking intensity

Fig. 7  Time histories of ru in Case SC-L1, Case SC-L2 and Case FF-L1

13
Bulletin of Earthquake Engineering

(PBA = 0.40 g) in Case SC-L2, the generation rate and the peak values of excess pore
pressure ratios in all depths were significantly higher than that in Case FF-L1, and soil
liquefaction occurred in the shallow layer and the bottom half of the enclosed soil. There
were more intense oscillations of excess pore pressures in Case SC-L2 due to the strong
interaction of the enclosed soil and the soil-cement grid. The total amount of excess
pore pressures was less generated in SC-L1, and the liquefaction durations in other two
cases were much longer. Besides the higher peak value of ru, the longer dissipation in
Case SC-L2 and Case FF-L1 might be caused by the upward migration of fine particles
(i.e., Kaolin clay) and more significant densification during shaking, which may reduce
the permeability of upper layer and inhibit the dissipation of excess pore pressure. This
kind of fines migration due to the seepage force has also been characterized in element
tests (Indraratna et al. 2020; Zhao and Fauzi 2022). It was found that the average coeffi-
cient of consolidation ­(Cv) at different depths in Case SC-L2 was around 30% of that in
Case SC-L1, according to Terzaghi’s one-dimensional consolidation theory.
In all three cases, the high excess pore pressure generated in the base layer (P1),
which was supposed to be the same as the bottom of the liquefiable layer. In the unim-
proved ground, the generation rate of excess pore pressure in the base layer was slower
than that of the upper enclosed soil, which is common in free field. It is because the
underlying layer originally had higher effective stress, and in this experiment, the coarse
sand also had higher liquefaction resistance than clayey sand. However, the condi-
tion went to the opposite in the soil-cement grid improved ground. Especially in Case
SC-L1, the excess pore pressure in the base layer generated much faster and reached
higher peak value than that at a depth of 10 m. Figure 8 gives the isochrones of the
excess pore pressure ratio during shaking in Case FF-L1, Case SC-L1 and Case SC-L2,
and a distinct "waist" shape distribution of excess pore pressure that shrank at the mid-
dle depth could be found in both improved model ground. Therefore, more attention
should to be paid to the underlain soil layer of the soil-cement grid improved ground,
which may be originally considered as non-liquefiable layer while experienced higher
excess pore pressures during the first strong shaking. The similar recommendations
were also given by Mitrani and Madabhushi (2011) based on centrifuge test results. In
summary, the excess pore pressures distribution of enclosed soil along with the depth
could be described in such a pattern that it will be the least in the middle and reached
higher values at both the top and the bottom of the soil-cement grid, and it is termed as
the "waist effect" in this study hereafter.

Fig. 8  Isochrones of ru during shaking in a Case FF-L1, b Case SC-L1 and c Case SC-L2

13
 Bulletin of Earthquake Engineering

3.3 Dynamic shear stress–strain responses of soil

The soil-cement grids could largely share the earthquake-induced shear stresses and restrict
the deformation of the enclosed soil, thereby effectively inhibiting the generation of the
excess pore water pressure. The dynamic shear stress and shear strain loops were com-
puted at different depths using the recorded data from vertical array of accelerometers. The
calculation procedures were proposed in Zeghal et al. (1999) with the assumption of one-
dimensional shear beam condition. It is worth noting that the shear stresses of the enclosed
soil mentioned here are actually the average shear stresses for this composite soil-grid sys-
tem, as the lateral earth pressure acting on the enclosed soil are unknown during soil-grid
interaction and the assumption of one-dimensional shear beam condition might not hold
true for the enclosed soil. Although they do not distinguish how the shear stresses are dis-
tributed between the soil-cement grid and the enclosed soil, it could provide a basis for
evaluating the overall system stiffness. Figure 9a presents the shear stress–strain loops at
different depths in Case SC-L1. The “waist effect” of dynamic shear strains, similar to the
depth profiles of excess pore pressure ratios, was also observed, which shows that the shear
strain at the middle height was vastly smaller than that in shallower or deeper layers. In
addition, the overall stiffness at a depth of 1.25 m and the stiffness of the underlain coarse
sand layer were estimated by the slope of the hysteretic loop, as shown in Fig. 9b, c. It was
found that the overall stiffness at a depth of 1.25 m decreased rapidly from the 5th to the
10th loading cycles and finally stayed around 320 kPa. However, the stiffness of the coarse
sand layer was gradually recovered after the drastic reduction at the beginning of the shak-
ing. This is probably the reason why the vibration of the improved soil layer will strengthen
in the later stage of a long duration shaking.
Figure 10 summarizes the profile of the maximum shear strain of the enclosed soil in
all three models, where the positions of the accelerometers are also given for better under-
standing. The maximum shear strains of the enclosed soil maintained at around 0.08% in
Case SC-L1 (PBA = 0.15 g) within a depth range of 7.5 m to 12.5 m, and it increased
almost linearly in the upper half as the depth decreased and reached to 0.97% at a depth of
1.25 m. When being subjected to the stronger shaking event in Case SC-L2 (PBA = 0.40 g),

Fig. 9  Shear stress–strain loops a at different depths in Case SC-L1, b at depth of 1.25 m with calculated
stiffness, c at depth of 16.25 m with calculated stiffness

13
Bulletin of Earthquake Engineering

Fig. 10  Maximum shear strain

of the soil in Case SC-L1, Case
SC-L2 and Case FF-L1

the maximum shear strains of the upper half of the enclosed soil increased by 50% com-
pared with that in Case SC-L1, and the maximum shear strains in the lower half increased
as the depth increased. It shows the maximum shear strain was least at a depth of 7.5 m
in Case SC-L2, the value of which was almost the same as that in Case SC-L1. The peak
value of the maximum shear strains appeared in the base layer in both cases reflecting the
relative movement between the whole improved liquefiable layer and the coarse sand base
layer. Overall, the limiting effect of the soil-cement grid on the shear strain of the enclosed
soil was most effective in the middle parts, which was not sensitive to the intensity of the
shaking event.
In the unimproved ground (Case FF-L1), the maximum shear strains in the middle and
lower parts were much higher than those in the improved ground (Case SC-L1) under the
same shaking event. And the values of maximum shear strains were at the level of 0.5%
and relatively more uniform in Case FF-L1. Nevertheless, it was found that the existence
of the soil-cement grid vastly increases the maximum shear strains in the shallow layer
above 5 m and the underlying layer. For the shallow layer, the soil-cement grid enabled the
upward propagating of the vibration and applied the vibration to the enclosed soil through
the intense soil-grid interaction, thus enlarging the corresponding shear strains, which was
not observed in the unimproved ground due to the soil softening. And the cemented soil
grids largely increased the inertial forces of the improved layer during the shaking, also
enlarging the shear strains in the underlying layer.

3.4 Dynamic soil‑grid interaction

Different from the free field, the soil-cement grid tends to diffract the upward propagat-
ing S-waves, thereby modifying soil deformations. This type interaction between soil
and grid is called kinematic interaction, the intensity of which could be portrayed by the
relative displacement between them. Figure 11 summarized the relative displacements
between the enclosed soil and soil-cement grid at different depth in Case SC-L1 and

13
 Bulletin of Earthquake Engineering

Fig. 11  Relative displacement between the enclosed soil and the soil-cement grid at different depths: a Case
SC-L1 and b Case SC-L2

Case SC-L2, where the horizontal displacements of the enclosed soil at different depths
and soil-cement grid were computed by double integrating the corresponding accelera-
tions. In view of the high stiffness, the horizontal displacement of soil-cement grid was
supposed to be the same at different depths and represented by that of grid surface.
As shown in Fig. 11, the relative displacements at ground surface were the most sig-
nificant and those at the middle height were the least in both cases. The development
of relative displacement went to several distinct stages as the shaking continued: at
the first several cycles, the rapid generation of excess pore water pressure softened the
enclosed soil significantly and then the relative displacement enlarged; then the under-
lain coarse sand layer liquefied around 10–12 cycles in Case SC-L1 and around 7–8
cycles in Case SC-L2 respectively, which led to the sudden drop of the capability to
transfer the upward shear stresses for the soil-grid system, and the relative displacement
decreased drastically at this short moment; and then the stiffness of the underlain coarse
sand layer recovered to some extent due to the upward dissipation of excess pore water
to the overlying enclosed soil during such a long shaking, the soil-grid interaction will
recover to some extent simultaneously. It could also be found in Fig. 11 that, the rela-
tive displacements in Case SC-L2 were generally larger than those at the same depths in
Case SC-L1, which could be attributed to the additional softening of the enclosed soil
under stronger shaking in Case SC-L2 in view of the higher excess pore water pressure
as shown in Figs. 7 and 8. Figure 12 gives the time histories of the displacements, rela-
tive displacement and phase difference of the enclosed soil and soil-cement grid at the
depth of 1.25 m, to explore this issue a bit further. It could be found in Fig. 12 that at
the beginning, the development of relative displacement corresponds to the increase of
phase difference between the enclosed soil and the soil-cement grid despite that their
amplitude were almost the same, which resulted from the gradual loss of soil stiffness
inside the grid cell; however, the phase difference (i.e., the stiffness contrast) stayed
almost constant in the remaining loading cycles, the gradual increase of relative dis-
placement resulted from the recovery of the soil stiffness in the underlain coarse sand
layer. In addition, this soil-grid interaction continuously provided the inertial force
required for soil vibration, so that the acceleration of the enclosed soil did not attenuate

13
Bulletin of Earthquake Engineering

Fig. 12  Relative displacement and phase difference between the enclosed soil and soil-cement grid: a Case
SC-L1 and b Case SC-L2

to a level of almost zero as that of the unimproved ground (Fig. 5), even if excess pore
pressure had reached a high level (Fig. 7).

3.5 Post‑shaking observations

After the completion of the shaking event, there was almost no visual damage to the soil-
cement grid in Case SC-L1, as shown in Fig. 13a. However, several cracks at the end walls
were observed on one side of the soil-cement grid in Case SC-L2 (Fig. 13b). And the after-
shaking excavation shows that there were no other cracks on the soil-cement walls and
most cells remained intact, especially the inner cell where the sensors were embedded. It
implies that the soil-cement grid is not prone to be damaged under such a long-lasting
shaking event with moderate amplitude (i.e., PBA = 0.15 g for 50 cycles) despite the large
earthquake magnitude at prototype, and only limited damage will occur under very high
ground motion amplitude, which is PBA = 0.40 g in this case. Furthermore, it could be
found that there was a mass of fine white sands (glued with Kaolin clay) ejected to the
surface in Case SC-L2 due to the severe liquefaction of the enclosed soil, which echoes the
aforementioned excess pore pressure responses in Sect. 3.2.

Fig. 13  Post-test condition of soil-cement grid in: a Case SC-L1 and b Case SC-L2

13
 Bulletin of Earthquake Engineering

1.0
Case FF-L1
Case SC-L1
Case SC-L2
0.8

Settlement (m)
0.6

0.4

0.2

0.0
Saturation Spin-up Shaking event
Event sequence

Fig. 14  Soil settlements in different events

Table 3  Relative densities of the Model Initial condi- After satura- After spin After
centrifuge models tion (%) tion (%) up (%) shaking
(%)

Case FF-L1 41 58 68 70
Case SC-L1 40 48 63 64
Case SC-L2 40 48 64 70

The settlements of the enclosed soil in the centre were monitored by laser displace-
ment transducers in flight and also manually measured before and after the shaking
event. As shown in Fig. 14, the settlements after saturation and spin-up processes were
almost the same in Case SC-L1 and Case SC-L2, which both reached around 0.40 and
0.65 m at prototype scale respectively. In Case FF-L1, larger settlement occurred dur-
ing the saturation process compared with that during the spin-up process. However, the
total settlements before the shaking event were close to each other among all three mod-
els and the settlement in Case FF-L1 was only 10% higher than those of the other two
improved models. The corresponding relative densities of the models were also sum-
marized in Table 3. It is noted that the particle structure between the fines and sands
would result in a mixture with high volumetric compressibility under relatively low den-
sities (Lade and Yamamuro, 1997). In general, the relative densities of the enclosed
soil in the three models were close to each other and around 65% before the shaking
event. Under the shaking event with PBA = 0.15 g, the settlement of the improved model
ground was only 40% of that of the unimproved ground. It shows the soil-cement grid is
highly effective in reducing the earthquake shaking-induced settlements of the liquefi-
able sands. However, much more settlement of the enclosed soil was observed in Case
SC-L2 under stronger shaking intensity even if the reinforcement of the soil-cement
grid had been adopted.

13
Bulletin of Earthquake Engineering

4 Conclusions

Three centrifuge tests were used to characterize the seismic responses of the coupled soil-
grid system when subjected to strong earthquake loadings. The 15-m-thick liquefiable layer
in two model tests was improved by the soil-cement grid and another one of free-field with-
out any liquefaction countermeasure played as a reference case. Generally, the soil-cement
grids have a great liquefaction mitigation effect on the enclosed soil even under very strong
shaking loadings. Some interesting conclusions could be drawn as follows:

1. The test results show that the soil-cement grid would significantly reduce the liquefac-
tion potential and post-shaking settlements of the improved ground by restricting the
dynamic shear strain of the enclosed clayey sands, especially in the middle height of
the grid cell. Such interesting mitigation effect is named as “waist effect” in this study,
where the dynamic shear strain of the enclosed soil could be described in such a depth
distribution pattern that it will be the least in the middle and reach higher values at both
the top and the bottom of the soil-cement grid. It is worth noting that the underlying
soil layer may experience larger shear strain due to the increasing inertial force of the
overlain improved ground, compared with a free field condition of the unimproved
ground.
2. The soil-cement grid considerably increased the acceleration of the ground surface
compared with that of the unimproved ground, where severe liquefaction occurred in
the whole depth. The acceleration responses of the soil surface and soil-cement grid
were almost the same in Case SC-L1 during the moderately strong shaking event, while
bigger difference was observed between them in Case SC-L2 when being subjected
to the stronger shaking event. And the relatively high-frequency components of the
earthquakes would also be amplified due to the higher predominant frequency of the
improved ground.
3. The relative displacement between the enclosed soil and the soil-cement grid along
depths shows similar “waist effect” to that occurs to excess pore pressure or shear strain
of the enclosed soil, which demonstrates the dominant role of the soil-grid interaction
in dynamic response and liquefaction process of the enclosed soil. The actual stress
state and dynamic stress path of the enclosed soil are worthy of further research by the
combination of physical modelling and numerical simulation to characterize the soil-
grid interaction in a more comprehensive way.

Acknowledgements The authors would like to acknowledge the National Natural Science Foundation of
China (Nos. 51988101, 51978613 and 52278374) and the Chinese Program of Introducing Talents of Disci-
pline to University (the 111 Project, No. B18047), which collectively funded this project. Sincere thanks to
Dr. Yasuhiro Shamoto, Dr. Kiyoshi Fukutake, Dr. Yoshiharu Asaka and Dr. Youhao Zhou, Institute of Tech-
nology, Shimizu Corporation, for their valuable comments, and to Mr. Toshiyuki Iwai, Mr. Takumi Hirai
and Mr. Katsumi Yoshinari, Institute of Technology, Shimizu Corporation, Dr. Junchao Li and Mr. Zizhuang
Yan, Zhejiang University, for their supports in jointly conducting the centrifuge model tests.

Author contributions All authors contributed to the study conception and design. Material preparation, data
collection and analysis were performed by YC, YK and YGZ. The first draft of the manuscript was written
by YC and YGZ, and all co-authors commented on previous versions of the manuscript. All authors read
and approved the final manuscript.

Funding This work was supported by the National Natural Science Foundation of China (Nos. 51988101,
51978613 and 52278374). Author Yan-Guo Zhou has received research support from the National Natural
Science Foundation of China. This work was supported by the Chinese Program of Introducing Talents of

13
 Bulletin of Earthquake Engineering

Discipline to University (the 111 Project, No. B18047). Author Yunmin Chen has received research support
from the Ministry of Education and the State Administration of Foreign Experts Bureau jointly.

Data availability The datasets generated during and/or analysed during the current study are available from
the corresponding author on reasonable request.

Declarations
Conflict of interest The authors have no relevant financial or non-financial interests to disclose.

References
AIJ (Architectural Institute of Japan) (2019) Recommendations for design of building foundations. Minato-
ku, Tokyo, Japan
Bradley BA, Araki K, Ishii T et al (2013) Effect of lattice-shaped ground improvement geometry on seismic
response of liquefiable soil deposits via 3-D seismic effective stress analysis. Soil Dyn Earthq Eng
48:35–47. https://​doi.​org/​10.​1016/j.​soild​yn.​2013.​01.​027
Bray JD, Dashti S (2014) Liquefaction-induced building movements. Bull Earthq Eng 12(3):1129–1156.
https://​doi.​org/​10.​1007/​s10518-​014-​9619-8
Chen L, Faccioli E (2013) Single-station standard deviation analysis of 2010–2012 strong-motion data
from the Canterbury region New Zealand. Bull Earthq Eng 11(5):1617–1632. https://​doi.​org/​10.​1007/​
s10518-​013-​9454-3
Fallahzadeh M, Haddad A, Jafarian Y, Lee CJ (2019) Seismic performance of end-bearing piled raft
with countermeasure strategy against liquefaction using centrifuge model tests. Bull Earthq Eng
17(11):5929–5961. https://​doi.​org/​10.​1007/​s10518-​019-​00696-z
García-Torres S, Madabhushi G (2019) Performance of vertical drains in liquefaction mitigation under
structures. Bull Earthq Eng 17(11):5849–5866. https://​doi.​org/​10.​1007/​s10518-​019-​00717-x
Ghofrani H, Atkinson GM (2015) Duration of the 2011 Tohoku earthquake ground motions. J Seismol
19(1):9–25. https://​doi.​org/​10.​1007/​s10950-​014-​9447-y
Goda K, Pomonis A, Chian SC et al (2013) Ground motion characteristics and shaking damage of the 11th
March 2011 Mw9.0 great East Japan earthquake. Bull Earthq Eng 11(1):141–170. https://​doi.​org/​10.​
1007/​s10518-​012-​9371-x
Hamada J, Honda T and Nakatsu N (2016) Dynamic centrifuge model tests on failure behaviour of grid-
form DMWs supporting a tall building. Takenaka Research and Development Institute, Koto-ku,
Tokyo, Japan, Takenaka Technical Research Report 72, pp. 1-13
Idriss IM (1997) Evaluation of liquefaction potential and consequences: historical perspective and updated
procedure. San Francisco: Presentation Notes, 3rd Short Course on Evaluation and Mitigation of
Earthquake Induced Liquefaction Hazards
Indraratna B, Singh M, Nguyen TT et al (2020) Laboratory study on subgrade fluidization under undrained
cyclic triaxial loading. Can Geotech J 57:1767–1779
Ishikawa A, Shamoto Y, Jang G et al (2012) A simplified method for evaluating the liquefaction of sandy
soil confined by a lattice-type deep mixing wall. J Jpn Soc Civ Eng Ser C (geosphere Engineering)
68(2):297–304. https://​doi.​org/​10.​2208/​jscej​ge.​68.​297
Ishikawa A, Zhou YG, Shamoto Y et al (2015) Observation of post-liquefaction progressive failure of shal-
low foundation in centrifuge model tests. Soils Found 55(6):1501–1511. https://​doi.​org/​10.​1016/j.​
sandf.​2015.​10.​014
Ishikawa A and Asaka Y (2006) Seismic responses of column and grid-type improved grounds. In Proceed-
ings of the 6th International Conference on Physical Modelling in Geotechnics, Hong Kong, R. P.
China, pp. 521–526
Ishikawa A, Shamoto Y and Kimura T (2016) A simplified method for evaluating the liquefaction of sandy
soil confined by a lattice-type deep mixing wall. In Proceedings of the 15th Asian Regional Confer-
ence on Soil Mechanics and Geotechnical Engineering (Japanese Geotechnical Society Special Publi-
cation), Fukuoka, Japan, 2 (63): 2153–2158. http://doi.org/https://​doi.​org/​10.​3208/​jgssp.​JPN-​017
Japanese Geotechnical Society (1998) Remedial Measures Against Soil Liquefaction: from Investigation
and Design to Implementation. Routledge, Bunkyo-ku, Tokyo, Japan
Khosravi M, Boulanger RW, Wilson DW et al (2017) Dynamic centrifuge tests of structures with shallow
foundations on soft clay improved by soil-cement grids. Soils Found 57(4):501–513. https://​doi.​org/​10.​
1016/j.​sandf.​2017.​06.​002

13
Bulletin of Earthquake Engineering

Khosravi M, Tamura S, Boulanger RW et al. (2016) Dynamic centrifuge tests on soft clay improved by
soil-cement grids. In Proceedings of the International Foundations Congress and Equipment Expo,
Texas, CA, United States of America, pp. 2349–2358. https://​doi.​org/​10.​1061/​97807​84479​087.​218
Kitazume M, Terashi M (2013) The deep mixing method. CRC Press, Florida, the United States of
America
Koseki J, Namikawa T (2010) Behaviour of lattice-type ground improvement by cement-mixing for liq-
uefaction mitigation. In Proceeding of the International Symposium on Ground Improvement Tech-
nologies and Case Histories, Singapore, pp. 257–262
Lade PV, Yamamuro JA (1997) Effects of nonplastic fines on static liquefaction of sands. Can Geotech J
34(6):918–928
Mitrani H, Madabhushi SPG (2010) Cementation liquefaction remediation for existing building. Ground
Improv J 163(G12):81–94
Mitrani H, Madabhushi SPG (2011) Rigid containment walls for liquefaction remediation. J Earthq Tsu-
nami 6(3):1250017
Mitrani H, Madabhushi SPG (2013) Geomembrane containment walls for liquefaction remediation.
Ground Improv J 166(GI1):9–20
Mitrani H, Madabhushi SPG, Villalobos FA (2012) Discussion on cementation liquefaction remediation
for existing buildings. Ground Improv J 165(G13):183–184
Namikawa T, Koseki J, Suzuki Y (2007) Finite element analysis of lattice-shaped ground improvement
by cement-mixing for liquefaction mitigation. Soils Found 47(3):559–576. https://​doi.​org/​10.​3208/​
sandf.​47.​559
Nguyen TV, Rayamajhi D, Boulanger RW et al (2013) Design of DSM grids for liquefaction remedia-
tion. J Geotech Geoenviron Eng 139(11):1923–1933. https://​doi.​org/​10.​1061/​(ASCE)​GT.​1943-​
5606.​00009​21
Rayamajhi D, Tamura S, Khosravi M et al (2015) Dynamic centrifuge tests to evaluate reinforcing mech-
anisms of soil-cement columns in liquefiable sand. J Geotechn Geoenviron Eng. https://​doi.​org/​10.​
1061/​(ASCE)​GT.​1943-​5606.​00012​98
Sakai K, Tazaki K (2003) Development and applications of diaphragm walling with special section
steel–NS-Box. Tunn Undergr Space Technol 18(2–3):283–289. https://​doi.​org/​10.​1016/​S0886-​
7798(03)​00037-3
Sato Y, Maeda K, Hayashi K et al (2005) A study on overall properties of seismic response and inves-
tigation method of counter-measured ground against liquefaction. J Soc Mater Sci 54(11):1141–
1146. https://​doi.​org/​10.​2472/​jsms.​54.​1141
Suzuki Y, Saitoh S, Onimaru S et al (1996) Grid-shaped stabilized ground improved by deep cement mixing
method against liquefaction for abuilding foundation. J Jpn Soc Soil Mech Found Eng 44(3):46–48
Suzuki K, Babasaki R and Suzuki Y (1991) Centrifuge Tests on liquefaction-proof foundation. In Pro-
ceedings of the International Conference Centrifuge 1991, Boulder, Colorado, United States of
America, pp. 1461–1464
Takahashi H, Yamawaki S, Kitazume M et al (2006) Effects of deep mixing method on liquefaction preven-
tion and proposal on new arrangement of grid-type improvement. Report Port Airport Res Inst 45(2):
135–167
Tamura S, Khosravi M, Wilson DW et al (2018) A simple method for detecting cracks in soil-cement
reinforcement for centrifuge modelling. Int J Phys Modell Geotech 18(6):281–289. https://​doi.​org/​
10.​1680/​jphmg.​17.​00036
Taya Y, Uchida A, Yoshizawa M et al (2008) Simple method for determining lattice intervals in grid-
form ground improvement. Jpn Geotechn J 3(3):203–212
Terada K, Kikuchi N (2003) Introduction of homogenization method. Maruzen Publishing, Tokyo, Japan
Tokimatsu K, Mizuno H, Kakurai M (1996) Building damage associated with geotechnical problems.
Soils Found Special Issue 1:219–234
Wakai A, Ugai K, Hagiwara T (2004) Homogeneous modelling of composite ground with columnar
improvement based on its seismic response characteristics. J Appl Mech 7:899–908. https://​doi.​org/​
10.​2208/​journ​alam.7.​899
Zeghal M, Elgamal AW, Zeng X (1999) Mechanism of liquefaction response in sand–silt dynamic cen-
trifuge tests. Soil Dyn Earthq Eng 18(1):71–85
Zhao C, Fauzi UJ (2022) Visualized liquefaction behavior of sandy soil deposited in water under und-
rained cyclic shearing. Acta Geotech 17(8):3143–3160
Zhou YG, Liu K, Ling DS et al (2018a) Threshold seismic energy and liquefaction distance limit dur-
ing the 2008 Wenchuan earthquake. Bull Earthq Eng 16(11):5151–5170. https://​doi.​org/​10.​1007/​
s10518-​018-​0396-7

13
 Bulletin of Earthquake Engineering

Zhou YG, Sun ZB, Chen YM (2018b) Zhejiang University benchmark centrifuge test for LEAP-
GWU-2015 and liquefaction responses of a sloping ground. Soil Dyn Earthq Eng 113:698–713.
https://​doi.​org/​10.​1016/j.​soild​yn.​2017.​03.​010
Zhou YG, Meng D, Ma Q et al (2020a) Frequency response function and shaking control of the ZJU-400
geotechnical centrifuge shaker. Int J Phys Modell Geotech 20(2):97–117. https://​doi.​org/​10.​1680/​
jphmg.​19.​00029
Zhou YG, Xia P, Ling DS, Chen YM (2020b) A liquefaction case study of gently sloping gravelly soil
deposits in the near-fault region of the 2008 M(w)7.9 Wenchuan earthquake. Bull Earthq Eng
18:6181–6201. https://​doi.​org/​10.​1007/​s10518-​020-​00939-4
Zhou YG, Liu K, Sun ZB, Chen YM (2021) Liquefaction mitigation mechanisms of stone column-improved
ground by dynamic centrifuge model tests. Soil Dyn Earthq Eng 150:106946. https://​doi.​org/​10.​1016/j.​
soild​yn.​2021.​106946

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under
a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted
manuscript version of this article is solely governed by the terms of such publishing agreement and applicable
law.

13