Professional Documents
Culture Documents
s12205-024-2675-y
s12205-024-2675-y
a
Member, Dept. of Civil Engineering, Wonkwang University, Iksan 54538, Korea
b
Soosung Engineering, Seoul 05836, Korea
c
Member, University of California, Davis, CA 95616, USA
Received 21 December 2023 Liquefaction-induced settlement of shallow foundations is the result of bearing capacity failure
Accepted 14 March 2024 in undrained conditions and sedimentary settlement during the post-liquefaction process. The
Published Online 24 May 2024 bearing capacity of a shallow foundation is highly dependent on the size and dimensions of its
footprint. In addition, the reduction in shear strength in liquefiable soil, a key parameter for
KEYWORDS estimating bearing capacity, depends on the excess pore water pressure generated during an
earthquake. This study aims to investigate the impact of earthquake motion on the extent of
Liquefaction liquefaction-induced settlement in shallow foundations. A parametric study was conducted by
Foundation settlement varying the input earthquake motions in a three-dimensional response history analysis to
Numerical modelling directly consider the interaction between the soil and superstructures. The numerical analysis
Intensity measures model constructed for the parametric study was rigorously calibrated using a reference
Centrifuge test dynamic centrifuge test in a prototype scale. The effects of the horizontal boundary and
drainage conditions in the numerical model were closely examined during calibration. The
parametric study results indicate that the intensity measures of an earthquake, which quantify
the energy associated with the number of reversals, exhibit a close correlation with the
resulting liquefaction-induced settlement as opposed to other conventional earthquake
motion parameters, such as peak acceleration, magnitude, and frequency.
CORRESPONDENCE Jin-sun Lee blueguy@wku.ac.kr Dept. of Civil Engineering, Wonkwang University, Iksan 54538, Korea
ⓒ 2024 Korean Society of Civil Engineers
2 J. Lee et al.
Fig. 1. Prototype Layout of Centrifuge Test T3-30 Model (reproduced from Dashti et al., 2010b)
during the strong input motion. experimental configurations in a prototype scale with dimensions of
90.8 m, 42.9 m, and 26.2 m for length, width, and height,
3. Verification of Numerical Model respectively. The buildings were modeled as elastic solids with
the same weight and dimensions as the physical model to reduce
3.1 Numerical Model the computational costs associated with the use of equivalent
This study utilized the three-dimensional finite-difference analysis structural elements. The flexible shear beam container was
software FLAC3D v9.0 (Itasca Consulting Group, 2023) to simulated using two distinct numerical boundary conditions to
simulate the reference dynamic centrifuge test and conduct the reflect the influence of the horizontal boundary on the numerical
parametric study. This software employs an explicit method and analysis results. This is discussed in detail in the following
direct time integration to capture the cyclic nonlinear behavior of section. Within the soil layers, a 3 m thick layer of liquefiable
soil, incorporating pore water pressure changes under strong Nevada sand was simulated using a liquefaction model; the other
motions. The numerical model was constructed to replicate the soil layers were represented by a nonlinear hysteretic damping
Fig. 2. Prototype Dimension of the Numerical Model and Subsoil Layers: (a) Free-Field Boundary Condition Model, (b) Tied Boundary Condition
Model
4 J. Lee et al.
model incorporating the Mohr-Coulomb failure criteria. The Table 2. Basic Engineering Parameters of Soil
nonlinear behavior of the non-liquefiable layer followed the Nevada sand
Monterey Nevada sand
mean curve proposed by Seed and Idriss (1970) for sandy soil. Properties (Liquefiable
sand (Dense Layer)
All the observed data from the dynamic centrifuge test, including Layer)
building settlements, acceleration records, and pore water pressure Relative density 85% 30% 90%
responses, were captured in the numerical analysis. The Specific gravity 2.59 2.63 2.65
constructed numerical model is shown in Fig. 2. Void ratio 0.56 0.67 0.54
Dry density (kg/m3) 1,660 1,580 1,720
3.2 Liquefaction Model Friction angle (cvo˚) 33 27 33
Over the past few years, numerous liquefaction models have Cohesion 0 0 0
been developed to account for undrained cyclic responses of Maximum shear modulus (Pa) *3.27 × 107 2.7 × 107 *7.12 × 107
saturated sands. In an earlier model (Byrne, 1991), generation of Poisson ratio 0.35 0.35 0.35
excess pore water pressure was expressed as a function of the Permeability (m/s)** 3.31 × 104 4.69 × 105 1.41 × 105
number of cycles, treating the mechanical response of sand as a *Estimated using Eq. (1)
distinct phenomenon. Advanced constitutive liquefaction models **Dashti and Bray (2013)
rooted in the critical state behavior of sand proposed by Bolton
(1986) have emerged since and gained widespread acceptance. test results or estimated using the primary parameters with typical
Recently, Boulanger and Ziotopoulou (2017) introduced a values given the initial stress condition of the numerical model. The
liquefaction model, PM4SAND, that encompasses post liquefaction basic engineering properties of the soil used in the dynamic
behavior. Based on classical elastoplastic theory, alternative centrifuge test are summarized in Table 2, as provided by Dashti and
perspectives have also been suggested: the nested yield surface- Bray (Dashti and Bray, 2013). Among the auxiliary parameters, this
based model (Yang et al., 2003) and the model employing study directly employs the maximum and minimum void ratios (emax
energy-based theories (Alavi and Gandomi, 2012), among and emin) provided by Dashti et al. (2010b), whereas the remaining
others. Each liquefaction model exhibits distinct merits and parameters are estimated based on relative density. To account for the
limitations concerning parameter selection, post liquefaction effect of the confining pressure on the shear modulus of non-
behavior, and applicability to three-dimensional modeling. liquefiable sandy soil, the maximum shear modulus was updated by
Incorporating a three-dimensional numerical model requires a considering the mean confining stress after static initialization using
liquefaction model that can operate effectively in the three- Eq. (1). The reference shear modulus (Gref) in Eq. (1) is derived from
dimensional space because defining stress reversal when the Eq. (2) (Dashti and Bray, 2013) and the SPT N value, (N1)60 is derived
soil experiences excitation forces in three directions is a challenging from the relative density using Eq. (3) (Tokimatsu and Seed, 1987).
task. This study adopted the P2PSAND model proposed by n
′
Gmax = G × P × ⎛ -----m-⎞ ,
ref
Cheng (2018), which is based on the bounding surface plasticity (1)
⎝P⎠
model SANISAND (Dafalias and Manzari, 2004), in 3D stress
space (Cheng and Detournay, 2021). The P2PSAND model was where m′ is the mean effective stress; P is atmospheric pressure;
implemented using the FLAC 3D software. n is an exponential constant (approximately 0.5 (Beaty and
Byrne, 2011)); and Gref is the reference shear modulus determined
3.3 Geotechnical Parameters using Eq. (2)
The accuracy of numerical analyses in geotechnical engineering ref
G = 325.5 × ( ( N1 )60 ) 0.333 , (2)
depends heavily on the engineering properties of the soil, which
serve as constitutive parameters in soil models. Dashti et al. 1⁄2
Dr = 15 ( N1 )60 , (3)
(2010b) conducted a dynamic centrifuge test and published the
experimental results and engineering parameters for the soil used where (N1)60 is the SPT N value, corrected for energy and
in the test. Dashti and Bray (2013) utilized the UBCSAND overburden pressure.
model under plane-strain conditions for their numerical analyses. The P2PSAND model adjusts the maximum shear modulus
There are distinctions in the model parameters between the of the liquefiable layers using an effective confining pressure
liquefaction models UBCSAND and P2PSAND used in this (Cheng, 2018). As a result, a contour of the bulk modulus of the
study, stemming from the different approaches that characterize numerical soil model, which takes into account the confining
the state of the sandy soil in the stress space and along the stress pressure before shaking, is presented in Fig. 3.
path. In addition to the common parameters used in both models, The internal friction angle at constant volume (cv) is also
the remaining parameters are estimated based on geotechnical derived from the SPT (N1)60 value using Eqs. (4) and (5).
evidence. The P2PSAND model requires two types of model
( N1 )60 ( N1 )60 – 15
parameters: primary and auxiliary. The primary parameters are cv = f – -------------
- = max 0, ------------------------ ,
10 5
estimated based on the relative density of the soil, whereas the
(Beaty and Byrne, 2011) (4)
auxiliary parameters can be determined either directly from the
KSCE Journal of Civil Engineering 5
Fig. 4. P2PSAND Predicted Response of Nevada Sand (Dr = 30) by a Cyclic Simple Shear Test: Stress Path; Shear Stress-Shear Strain Relationship;
Excess Pore Water Pressure Ratio-Shear Strain Relationship; Excess Pore Water Pressure Ratio Time-History
6 J. Lee et al.
The test results and measured acceleration signal at the bottom of test.
the test box are publicly available in the NEES database. This
Δu
study used the “Large Port Island” record for test number T3-30 Ru = -----′- , (6)
m
to input the motion parameters for numerical analyses. This input
was applied at the bottom of the numerical model as shear stress, where Δu denotes excess pore water pressure, and ′m is the
considering a compliant base condition to help reduce numerical mean effective stress under static conditions.
errors during the analysis (Mejia and Dawson, 2006). Dashti et
al. (2010b) conducted dynamic centrifuge tests using a flexible 4.1 Mechanism of Liquefaction-Induced Settlement
shear-beam container box. The container box had a structure Liquefaction-induced settlement is predominantly caused by
similar to that of the equivalent shear beam box but was designed bearing capacity failures when the bearing pressure exceeds the
to account for the behavior of the soil not just in terms of the capacity, thereby decreasing according to the pore water pressure
initial fundamental frequency but also to lower the natural buildup during liquefaction. This study comprehensively reviews
frequency of the soil undergoing large strains under strong excitation. this mechanism by leveraging detailed insights from numerical
This design is suitable for testing liquefiable soil layers. However, analysis results. To simulate liquefaction-induced problems,
the reference centrifuge test model included both liquefiable and numerical analyses must incorporate changes in pore water pressure
non-liquefiable layers. Thus, the flexible shear-beam container and the resulting mechanical soil response, as alterations in pore
box is not ideal for addressing this issue. In fact, none of the water pressure leads to seepage. However, the drainage conditions
available soil containers can perfectly replicate the stress-strain under which the soil is liquefied remain unclear. Thus, the
field of real soil strata under seismic conditions (Esmaeilpour et groundwater flow conditions were also reviewed for the drained
al., 2023). and undrained conditions, along with the horizontal boundary
In addition to physical model testing, the modeling of the conditions.
horizontal boundary conditions representing an infinite soil layer The distribution of the maximum excess pore water pressure
or soil containers is another challenge in numerical analysis. To ratio during the analysis and final deformed shapes, shear strain
address this challenge, two types of numerical horizontal boundary increments, and displacement vectors are shown in Fig. 6. It is
conditions were employed in this study to determine suitable evident from the observations that liquefaction occurred extensively
representations of the flexible shear beam container, as shown in within the liquefiable layers and extended to the upper strata of
Fig. 2. The first is the free-field element, originally proposed by the dense Nevada sandy layer under drained conditions. Notably,
Cundall et al. (1980) and implemented in the FLAC 3D software. differences stemming from the horizontal boundary conditions
The free-field element emulates the mechanical behavior of the are negligible. The shear strain increment and displacement vectors
adjacent soil model along with an attached viscous dashpot clearly reveal plastic deformation, which triggers liquefaction-
designed to absorb horizontally reflected waves. This approach induced settlement arising from bearing capacity failure leading
is useful for representing infinite horizontal soil layers; however, to punching failure. Under free-field boundary conditions, lateral
it does not exactly match the seismic response of a flexible shear spreading toward the horizontal boundaries occurred within the
beam container. Another approach involves the use of tied boundary liquefied layers. This phenomenon is attributed to volume change
conditions, which impose constraints between two opposite- allowances facilitated by this boundary type. Such volume changes
side elements to ensure compatibility in terms of displacement. are not observed in tied boundary conditions and flexible shear
This numerical boundary condition is identical to that of the beams in physical tests.
laminar box employed in the physical liquefaction tests and is
consistent with the approach adopted by Dashti and Bray (2013). 4.2 Excess Pore Water Pressure
In this study, an additional modification of the tied boundary A detailed comparison between the reference test and numerical
condition was implemented, wherein the stiffness of both lateral analysis is presented in Fig. 7 which depicts the excess pore
elements is modeled using an elastic material with properties water pressure ratio and settlement histories. The settlements of
identical to those of the adjacent soils. This modification closely the buildings were recorded at each corner of the buildings using
simulates the behavior of flexible shear beams. The analysis an LVDT, and the pore water pressures were recorded beneath
results of both horizontal boundary conditions were compared the buildings. The peak increment and build-up pattern of excess
with the experimental data in detail and are presented in the next pore water pressure ratio are reasonably similar between the
section. reference physical test and numerical analyses. However, the
numerical analyses fail to precisely capture the dissipation patterns
4. Validation of the Numerical Analysis of excess pore water pressure. This discrepancy may be attributed to
the inadequate representation of the redistribution of excess pore
The numerical analysis results were validated for building settlement, water pressure in the liquefaction model, which is frequently
considering the excess pore-water pressure ratio beneath the observed in physical tests (Adamidis and Madabhushi, 2018;
buildings using Eq. (6). All the numerical analysis results were Kassas et al., 2021). Excess pore water pressure accumulates
recorded at the same transducer position as that in the reference rapidly during the initial and subsequent strong tremors, ultimately
KSCE Journal of Civil Engineering 7
Fig. 6. The Maximum Excess Pore Water Pressure Contour, Shear Stain Increment and Displacement Vector (Deformation shape is magnified in 2
times): (a) Free-Field Boundary with Drained Condition, (b) Free-Field Boundary with Undrained Condition, (c) Tied Boundary with
Drained Condition, (d) Tied Boundary with Undrained Condition
leading to bearing capacity failure of the shallow foundation. As exhibit contractive behavior, leading to a buildup of excess pore
discussed in the preceding section, excess pore water pressure water pressure. At this stage, the dissipation of excess pore water
build-up patterns may depend on drainage conditions. The pressure through the seepage process is not possible, as there is
maximum increase in excess pore water pressure ratio is more no pathway for flow out to the adjacent soil. Moreover, changes
similar to the results of the numerical analyses for undrained in the volume of the liquefiable layer may result in a decrease in
conditions than drained ones, which is in line with the reference pore water pressure without necessitating seepage in undrained
physical test results. However, regarding the dissipation pattern, conditions. Such variations in excess pore water pressure are
the drained condition yields more reasonable outcomes than the depicted in Fig. 7(d).
undrained condition. This closely aligns with the observations
made by Kassas et al. (2020), suggesting that current numerical 4.3 Liquefaction-Induced Settlement
analysis techniques cannot comprehensively capture the intricate Although the overall pattern of the liquefaction-induced settlement
patterns associated with the dissipation of excess pore water accumulation remained consistent, the amount of settlement and
pressure during liquefaction. tilting behavior varied depending on the analysis conditions.
The drainage conditions in the numerical model are controlled Notably, the free-field boundary condition exhibits a larger
by allowing the seepage of pore water, with the permeability liquefaction-induced settlement than the tied-boundary condition.
values listed in Table 2. Due to this seepage process, the dissipation This divergence can be attributed to the potential of horizontal
of excess pore water pressure is facilitated in drained conditions. residual displacements toward the boundaries inducing vertical
However, the accumulation of excess pore water pressure can settlement, even without alterations in the total volume. This
occur even in drained conditions if all the liquefiable layers phenomenon is apparent even for undrained conditions. The
8 J. Lee et al.
Fig. 7. Comparison of Settlement and Excess Pore Water Pressure Build Up Patterns: (a) Free-Field Boundary with Drained Condition, (b) Free-
Field Boundary with Undrained Condition, (c) Tied Boundary with Drained Condition, (d) Tied Boundary with Undrained Condition
KSCE Journal of Civil Engineering 9
Fig. 7. (continued)
disparities in settlement due to drainage conditions are generally the liquefaction-induced settlement of shallow foundations.
insignificant, except in the case of BLDG_C. More liquefaction- Video clips of the graphs in Fig. 7, representing free-field boundary
induced settlement is observed in the drained analysis than in the conditions in the drained analysis, are presented as supplementary
undrained analysis, particularly for BLDG_C. Thus, it is inferred material.
that an overall residual displacement occurs in the direction of
the right side, which is adjacent to BLDG_C. Consequently, the Supplementary Material - Video
vertical settlement compensates for the discharged volume, as a Deformed shape, Excess pore water pressure ratio contour with Set-
sizable region near the boundary experiences liquefaction and is tlement of structures, Excess pore water pressure ratio time histories
discharged towards the right side. The differences in settlement
accumulation at each corner correspond to the tilting of the 4.4 Additional Findings in Validation
buildings. BLDG_B, which is characterized by the largest base- Initially, a tied boundary with a fundamental natural period
contacting area and a low center of gravity, exhibited minimal identical to that of the model soil layers was considered in the
tilting behavior in the physical tests, contrary to numerical numerical analysis. The fundamental natural period of the model
simulations results revealing a tilt for BLDG_B. The rocking soil layers during liquefaction was estimated from free-field
effect on liquefaction-induced settlement, as discussed by Adamidis acceleration records in the reference physical test. Subsequently,
and Madabhushi (2022), does not manifest in the physical test the shear wave velocity corresponding to the fundamental period
outcomes but is observed in the numerical analysis when reviewing was used to compute the shear modulus of the lateral elastic
the case of BLDG_C. These differences may arise from the element adjacent to the fixed boundary. This process parallels the
simplified superstructures used in numerical modeling. To increase procedure used in designing a flexible shear beam. However, this
computational efficiency, the superstructures were modeled as approach yields unrealistic responses. Therefore, the lateral elastic
elastic solids. However, the physical test model exhibited a different elements were designed to match the initial modulus of the
mode of dynamic response possibly because it is composed of adjacent soil. Second, the direction of the excitation load was
plates, even though the two structural models have identical considered because the NEES database does not specify the
weights and centers of gravity. loading direction as right or left, in indicating the positive
After thoroughly reviewing the analysis results encompassing acceleration motion. Consequently, the impact of the loading
various conditions, it was deduced that the physical model test direction is insignificant. However, residual lateral spreading
within finite boundaries cannot precisely replicate the actual occurs in the opposite direction. Obtaining accurate acceleration
liquefaction-induced settlement in the field. The liquefaction- histories of fully liquefied soil poses a significant challenge due
induced settlement observed in both the physical tests and to the difficulty in maintaining the reference position and direction
numerical analyses comprise settlements induced by bearing of the acceleration transducers during physical tests. As a result,
capacity failure as well as lateral spreading. However, the horizontal there are only a few acceleration records that can be compared
boundary conditions influence the extent of lateral spreading. between the centrifuge test and numerical analysis. Fig. 8 presents a
Despite certain limitations in the validation of the numerical comparison of one of the highest-quality acceleration records
analyses, the model employed in this study is valuable for obtained from the centrifuge test. The acceleration records from
investigating the effects of nonstationary earthquake motion on the numerical analysis align well with those from the centrifuge
10 J. Lee et al.
Table 4. Intensity Measures and Earthquake Motion Parameters Used in This Study
No. Intensity measure parameter Definition Reference
t
1 Arias intensity (m/s) Ia = ------ ∫ a ( t ) 2 dt Arias (1970)
2g 0
3⁄2
2 Characteristic intensity (m1.5/s2.5) Ic = ( arms ) ttot Park et al. (1987)
where, arms is rms acceleration and ttot is strong motion duration
t
3 Cumulative absolute velocity (cm/s) CAV = ∫ a ( t ) dt Reed and Kassawara (1990)
0
t
CAV5 = ∫ 〈 X〉 a ( t ) dt
0
4 CAV5 (cm/s) 2 Kramer and Mitchell (2006)
⎧ 0 for a ( t ) < 5 cm ⁄ s
where, 〈 X〉 = ⎨
2
⎩ 1 for a ( t ) ≥ 5 cm ⁄ s
N i
CAVdp = ∑ ⎛ H ( PGAi – 0.025 ) ∫ a ( t ) dt⎞
⎝ i–1 ⎠
i=1
5 CAVdp (cm/s) Campbell and Bozorgnia (2011)
⎧ 0 for H ( X ) < 0
Where, H ( X ) = ⎨
⎩ 1 for H ( X ) ≥ 0
I
PD = ---a-2
6 Destructive potential factor (m/s) v0 Araya and Saragoni (1984)
where v0 is number of zero crossings in velocity history
2.5
HI = ∫ PSV ( = 0.05, T ) dt
7 Housner intensity (m) 0.1 Housner (1952)
where, PSV is Pseudo Spectral Velocity
2
∑ C i ⁄ f-i
Tm = ----------------
2
8 Mean period (s) ∑C i Rathje et al. (1998)
Acknowledgments
ORCID
References
Adamidis O, Madabhushi SPG (2018) Experimental investigation of
drainage during earthquake-induced liquefaction. Géotechnique 68(8):
655-665, DOI: 10.1680/jgeot.16.P.090
Adamidis O, Madabhushi SPG (2022) Rocking response of structures
with shallow foundations on thin liquefiable layers. Géotechnique
72(2):127-145, DOI: 10.1680/jgeot.19.P.077
Alavi AH, Gandomi AH (2012) Energy-based numerical models for
assessment of soil liquefaction. Geoscience Frontiers 3(4):541-555,
DOI: 10.1016/j.gsf.2011.12.008
Araya R, Saragoni GR (1984) Earthquake accelerogram destructiveness
potential factor. 8th World Conference on Earthquake Engineering,
SanFrancisco, U.S.A, 835-842, https://www.nicee.org/wcee/search8.php
Arias A (1970) Measure of earthquake intensity. Hansen RJ (ed) Seismic
design for nuclear power plants. MIT press, Cambridge MA, 438-483
Arulmoli K, Muraleetharan KK, Hossain MM, Fruth LS (1992) VELACS:
Verification of liquefaction analysis by centrifuge studies, Laboratory
Testing Program. Soil data Report, 90-0562
Atik LA, Abrahamson N (2010) An improved method for nonstationary
spectral matching. Earthquake Spectra 26(3):601-617, DOI: 10.1193/
1.3459159
Ayoubi P, Pak A (2017) Liquefaction-induced settlement of shallow
foundations on two-layered subsoil strata. Soil Dynamics and Earthquake
Engineering (94):35-46, DOI: 10.1016/j.soildyn.2017.01.004
Beaty MH, Byrne PM (2011) UBCSAND constitutive model: Version
904aR. Documentation Report: UBCSAND constitutive model on
Itasca UDM web site, https://www.itascacg.com/software/udm-library/
ubcsand
Bolton MD (1986) The strength and dilatancy of sands. Géotechnique
36(1):65-78, DOI: 10.1680/geot.1986.36.1.65
Boulanger RW, Ziotopoulou K (2017) PM4SAND (VERSION 3.1): A
sand plasticity model for earthquake engineering applications. Dept.
of Civil and Environmental Engineering, University of California
Davis, Report No. UCD/CGM-17/01, https://www.itascacg.com/
Fig. 10. Liquefaction Induced Settlement with the Intensity Measures software/udm-library/pm4sand
(IMs) Byrne PM (1991) A cyclic shear-volume coupling and pore-pressure
KSCE Journal of Civil Engineering 13