Computers and Geotechnics xxx (xxxx) xxx

Contents lists available at ScienceDirect

Computers and Geotechnics

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

Research Paper

Performance-based analysis of cantilever retaining walls subjected to

near-fault ground shakings
Milad Aghamolaei , Alireza Saeedi Azizkandi *, Mohammad Hassan Baziar , Sadegh Ghavami
School of Civil Engineering, Iran University of Science and Technology, Tehran, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: Considering the devastating damages of ‘forward-directivity’ on structures, a series of finite element models were
Near-fault conducted to evaluate the seismic performance of cantilever retaining walls under near-fault excitations. The
Performance-based analysis wavelet approach was used to extract the velocity pulse of near-source motions, and a semi-artificial records
Wavelet approach
reagent far-field earthquake was produced. Both were then imposed on the model, separately. The results
Frequency content
indicated a vivid difference in lateral displacement, in which some cases up to differences of experienced 85%
and forces along the walls were approximately equal. In view of this finding, a wide range of PGAs was applied to
the near-fault scenarios of the models. The captured movements were compared with the recommended criteria
for performance-based aseismic design of retaining structures. According to the numerical analysis, in most
earthquakes, for accelerations exceeding 0.4 g, lateral displacement of the wall had a higher value than the
permissible proposed limits. Also, accelerations exceeding 0.6 g for both near and far-field records resulted in
wall failure (>5% H). The final section of this research presents a comprehensive parametric study on the effects
of ground motion characteristics and soil mechanical properties on system performance.

strike with strike-slip mechanism or in the fault-normal direction for

dip-slip faults may be affected by a permanent static movement called
1. Introduction the ‘fling-step’ (Somerville et al., 1997). The FD effect is a dynamic
phenomenon that does not leave permanent ground movements; as
Earthquake ground motions recorded close to a fault plane recog­ observed in the time history, FD produces two sided velocity pulses,
nized as near-fault ground motions can be extremely different from while the fling-step, caused by permanent earth displacements, makes
motions captured far from the ruptured source. While there are different one-sided velocity pulses (Bray and Rodriguez-Marek, 2004).
opinions regarding the near-fault zone location, the most common is All of the above explanations reveal that near-fault records are
restricted within a 20 km distance of the ruptured fault (Bray and inherently different from larger site-to-source recorded shakings and,
Rodriguez-Marek, 2004). Baziar and Rostami (2017) mentioned that the therefore, require special consideration when designing geotechnical
near-fault region is limited by the magnitude of a seismic event that is and structural systems. Bertero was the first to report the devastating
equal to Rt (km) = 0.3M2w . Wave propagation effects or the so-called failure capacities of near-source earthquakes (Bertero et al., 1978). The
‘forward-directivity’ (FD) affect near-fault sites. The fault rupture catastrophic earthquakes of the 1990s such as the Northridge (1994),
propagation toward a site at a speed that nears the shear wave velocity, Kobe (1994) and Chi-Chi (1999) earthquakes led to a wide range of
oriented perpendicularly to the fault plane, form the FD effect (Somer­ research efforts, aimed at assessing the performance and damage po­
ville, 2003). Most seismic energy demands in FD pulses accumulate at tential of various geotechnical and structural systems subjected to near-
the beginning of the record, which is evident in the large-period pulses fault pulse shakings (Hall et al., 1995; Alavi and Krawinkler, 2000;
in the velocity time history. It is worth noting that the ratio of seismic Garini and Gazetas, 2013; Davoodi et al., 2013).
energy of directivity pulses to the energy of the whole earthquake in Gazetas et al. (2009) presented a numerical study on a rigid block
near-fault records can be up to 80%. This indicates the importance of that was supported by a frictional contact surface and charged by mo­
near-fault excitations (Mukhopadhyay and Gupta, 2013). tions having forward-directivity or fling-step effects. They concluded
Also, ground shaking in the near-fault zone, parallel to the fault

* Corresponding author.
E-mail addresses: [email protected] (M. Aghamolaei), [email protected] (A. Saeedi Azizkandi), [email protected] (M.H. Baziar), s_
[email protected] (S. Ghavami).

https://doi.org/10.1016/j.compgeo.2020.103924
Received 12 March 2020; Received in revised form 10 November 2020; Accepted 13 November 2020
0266-352X/© 2020 Elsevier Ltd. All rights reserved.

Please cite this article as: Milad Aghamolaei, Computers and Geotechnics, https://doi.org/10.1016/j.compgeo.2020.103924
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Notations ρ Bulk mass density

PPV Peak to peak velocity
PGA Peak ground acceleration σ Stress
FD Forward directivity ε Shear strain
FE Finite element G Shear modulus
Mw Magnitude of earthquake f Frequency
M-O Mononobe–Okabe υ Poisson ratio
PGV Peak ground velocity ф Friction angle
ω Natural frequencies C Soil cohesion
μ Frictional coefficient Ψ Dilation angle of the soil
vs Shear wave velocity γ Unit weight
vp Body wave velocity Tp Predominant period

that the upper-bound sliding displacements from near-source excitations the structural loads, thereby, subjecting the retaining walls to significant
may substantially exceed the values obtained from some of the currently permanent displacements. Cakir (2013) analyzed the effect of earth­
available design charts. Song and Rodriguez-Marek (2014) developed a quake frequency content on the seismic response of retaining structures
coupled method for analyzing the sliding-blocks of slopes under near- and reported that wall responses are highly dependent on thePGV/PGA
fault pulse-like and nonpulse-like ground motions. The authors found ratio and can cause a spiked increase or decrease in system displacement
that the slope is expected to experience larger displacements when near- by the frequency content variation. Bakr and Ahmad (2018) developed
fault ground motions have pulse-like characteristics. Zou et al. (2017) charts and correlated between seismic earth pressure and wall move­
conducted a numerical analysis and found that the seismic response of ment. The authors reported that accelerations greater than 0.4g enabled
concrete face rockfill dams increased with an increasing ratio of the peak the retaining wall to continue moving without enhancing the dynamic
ground velocity to the peak ground acceleration (PGV/PGA). Higher passive earth pressure forces. Mikola et al. (2016) recorded distribution
values of crest displacement as well as intense damages to the concrete of the seismic earth pressures on cantilever retaining structures using
face were among the consequences of near-fault shakings. centrifuge tests. Salem et al. (2020) performed a series of two-
It is evident from the literature that near-fault ground motions are dimensional finite element methods for analyzing the seismic response
susceptible to inducing large displacements on different types of of cantilever retaining walls. The sensibility of the system response to
geotechnical structures. Hence, the role of these types of excitations is the soil constitutive model was studied. A Rigid perfectly plastic (M-C)
crucial to consider when designing by performance-based procedures. and an advanced nonlinear elastoplastic model (HSSMALL) were used.
Retaining walls are an example of a one such system and are widely used The results of the analysis showed that in the M-C model, a larger force
for stabilizing excavations in roads and highways, especially for urban than HSSMALL was captured. Furthermore, a higher value of lateral
areas. Extensive applications lead to constructing retaining structures in displacement for the 1989 Loma Prieta-UCSC earthquake was recorded
seismic regions and areas that are close to active faults. The seismic in the M-C model. Conti and Caputo (2019) investigated the dynamic
response of retaining walls is a complicated problem because it involves response and phase shift between soil and the inertia forces under a real
dynamic soil-structure interactions. Seismically induced lateral dis­ earthquake. Jadhav and Prashant (2020) proposed displacement-based
placements, dynamic bending moments and pressures behind the design procedures for cantilever retaining walls. The authors reported
retaining structures are multi-dimensional problems that depend on wall that using shear key placed at the heel of cantilever retaining wall was
foundation and backfill soil, the inertial and rigidity of the wall itself, reduced the transitional displacement by 40%. Santhoshkumar et al.
and the nature of input excitations. (2019) investigated the earth pressure behind cantilever retaining walls
The classic methods proposed by Okabe (1924) and Mononobe and using a pseudo-dynamic approach. Zamiran and Osouli (2018) corre­
Matsuo (1929), known as the Mononobe–Okabe (M-O) as later devel­ lated the free-filed PGA to the relative displacement of the wall under
oped by Seed (1970), are still the main approaches for the design of real earthquakes. They reported that 50% of walls experienced failure
retaining walls. This method recruits the pseudo-static equilibrium by state when input PGA reached to 0.47g for cohesionless backfill.
simplifying earthquake loading as an inertial force, without considering Reviewing the literature shows that most dynamic studies on
the dynamic characteristics of input earthquake loads and retaining retaining walls are limited to the earth pressures and forces that act
walls. Since then, various researches have been conducted to assess the along with the structures. The number of displacement-based studies of
seismic performance of retaining walls by means of experimental retaining walls is rare. Also, the performance of cantilever retaining
(Nakamura, 2006; Kloukinas et al., 2014; Jo et al., 2017; Candia et al., walls under near-fault excitations is not yet well understood. The con­
2016); numerical and analytical approaches (Veletsos and Younan, formity of seismic wall movements in real earthquake scenarios with
1997; Psarropoulos et al., 2005; Nimbalkar and Choudhury, 2007; di failures and permissible states are also unknown. So, further research
Santolo and Scotto and Aldo Evangelista., 2011; Brandenberg et al., about response and seismic forces behind retaining walls that are
2017; Bakr et al., 2019). motivated by near source motions are needed. Qualitative insight into
Gazetas et al. (2004) used finite-element modeling to explore the the performance analysis of retaining walls under near-fault strong
magnitude and distribution of dynamic earth pressure forces on several ground motions will emphasis on the importance of displacement-based
types of flexible retaining systems. By using dynamic centrifuge exper­ designs.
iments performed on cantilever walls and following two-dimensional In this regard, the present research evaluated the results of a series of
nonlinear finite-element analysis, Atik and Sitar (2010) concluded that dynamic 2D finite element (FE) numerical models based on the perfor­
the current design methods based on the M-O theory significantly mance of cantilever retaining structures under near-fault excitations
overestimated the captured dynamic earth pressure forces and moments with a focus on seismically induced lateral displacements. Due to the
and mentioned that seismic earth pressures along with cantilever higher damage potential of FD over fling step (Bray and Rodriguez-
retaining walls can be neglected at accelerations below 0.4 g. By Marek, 2004; Kalkan and Kunnath, 2006), this research was mainly
focusing on displacements, Conti et al. (2012) showed that maximum focused on FD shakings and used fully dynamic time-domain analyses in
accelerations smaller than the critical limit equilibrium value increase the process. In the first section and to illustrate the importance of the

2
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

discrepancies in wall responses under near and far-field ground motions, record from the Landers earthquake in 1992 at the Lucerne station. A
FD pulses were extracted from the velocity time series of near-fault, comprehensive database of different magnitudes, durations and peak
main, and residual records and imposed on the verified model. Then, value of velocities was considered for assessing the effect of the direc­
near-fault strong ground motions with a wide range of PGAs from 0.1 g tivity pulse on the seismic response of cantilever retaining walls
to 0.6 g were applied for different shaking scenarios. The captured (Table1).
movements were compared with the recommended criteria for the As mentioned above, it is obvious that all characteristics of the two
performance-based aseismic design of soil retaining structures in the records are the same, except in the pulse acting domain. A comparison of
literature. A comprehensive parametric study was applied to assess the Fig. 2a and b shows that eliminating the directivity pulse decreased the
effect of different parameters. The effect of the mechanical properties of peak ground velocity by about 50%; whereas decreased the maximum
backfill/foundation soil as well as the frequency content of the ground magnitude of acceleration by only 10%. To further elaborate, the ratio of
motion was investigated. peak ground velocity to peak ground acceleration (PGV/PGA) went from
0.18 s in the main record to 0.1 s in the generated motion, indicating the
2. Methodology importance of the directivity pulse.

Selecting near-fault ground shakings beside the far-field records, as 3. Numerical modelling and calibration
subdivided into pulse-like and non-pulse ground motions, and
comparing the structure responses under these types of seismic loads are The two–dimensional plane strain dynamic implicit method is
common procedures widely applied in the literature. This study employs applied for the numerical analyses of a cantilever retaining wall using
a novel approach called the wavelet analysis. The wavelet approach is a the ABAQUS finite element based software (2014). The shaking table
signal processing procedure that decomposes signals such as seismic test results obtained by Kloukinas et al. (Kloukinas et al., 2014, 2015)
ground motions (Baker, 2007). Many researchers have evaluated the were simulated in the prototype scale to calibrate the results of the
wavelet analysis for the characterization of near-source earthquakes numerical modeling.
(Mavroeidis and Papageorgiou, 2003; Mollaioli and Bosi, 2012; Vassi­
liou and Makris, 2011). 3.1. Model geometry
Baker (2007) developed a wavelet based transform for records with
PGV > 30 cm/sec which are classified as pulse-like ground motions. In The experimental tests conducted by Kloukinas et al. (2014, 2015)
this research, near-fault records containing forward directivity were and used for numerical validation were performed in a large Equivalent
compiled from a database developed by Baker (2007) and Hayden et al. Shear Beam container of EQUALS, with inner dimensions of 4.80 m long,
(2014), which was obtained from the Pacific Earthquake Engineering 1 m wide and 1.15 m deep. A maximum soil height of 1 m was selected,
Research Center website (PEER). Table 1 presents a brief summary of the corresponding to a wall height of 0.6 m and a foundation soil layer of
near-fault ground motion properties used in this research. 0.4 m. A frequency scale magnitude equal to 5 was considered for the
After selecting the near-fault earthquake ground motions, the model/prototype of this research. The dynamic time and frequency for 1
extracted pulses from the velocity–time history were subtracted from the g shaking table tests should be scaled in the prototype to satisfy the
main records, and new excitations indicating far-field earthquakes were compatibility between gravity and inertia forces (Iai et al., 2005). The
created. An example of the extracted pulse from the primary record is simulation law for 1 g shaking table tests, proposed by Iai et al. (2005),
depicted in Fig. 1. was applied and resulted in a prototype to model the scale factor of 8.5
Contrary to the common procedure used in the literature, these semi- (N = 8.5). An 8.5 m height was, therefore, modeled for the prototype
artificial earthquakes act precisely like the main ground motion, except soil layer. A backfill of 5.1 m and a foundation layer of 3.4 m were
that they occur in the range of pulse time, manifesting just the effect of simulated in the prototype scale. The length of the wall was equal to the
the forward directivity pulse. Fig. 2 shows the acceleration, velocity and backfill layer. Hence, a wall height of 5.1 m and a thickness of 0.25 m
displacement time series of the main record and the generated far-field were modeled in the numerical simulation.
The scaling factors for different parameters used in the numerical
Table 1 model are presented in Table 2.
near-fault ground motions database considered in this study investigation. The soil body was modeled with structured triangle six-noded ele­
ments (CPE6) due to the geometrical nonlinearity of the system. The
Event Station PGV Rrup PGV/ Arias Tp
(cm/s) (km) PGA intensity (s)
wall was modeled with the same element and a refined size. The sen­
(s) (m/s) sibility analysis of different mesh sizes was used to select the best mesh
dimensions, satisfying both accuracy and time of analysis. A dimension
Imperial El Centro - 92.6 0.07 0.302 1.105 0.56
Valley-06 Meloland of 0.2 m * 0.3 m was selected for the region next to the wall, and the
Geot. Array mesh sizes were larger at the soil boundaries. Finer element size will
Kobe, Japan Takarazuka 86.3 0.27 0.143 3.935 0.48 increase runtime with no significant change in results.
Morgan Hill Coyote Lake 78.3 0.53 0.061 3.855 0.3
Dam
Parkfield- Parkfield- 63.9 3.01 0.104 1.643 0.66 3.2. Soil properties and the constitutive model
02, CA Cholame
2WA The elastic perfectly plastic soil model with a Mohr-Coulomb (M-C)
San Pacoima Dam 114.47 1.81 0.096 8.948 0.38 plastic failure criterion and isotropic softening was applied as a consti­
Fernando (upper left
abut)
tutive model for the soil body. Softening is implemented by the reduc­
Tabas, Iran Tabas 123 2.05 0.146 11.808 0.2 tion of friction and dilatation angle referred to (Anastasopoulos et al.,
Loma Prieta LGPC 96 3.8 0.172 7.194 0.7 2007).
Landers Lucerne 133.5 2.19 0.188 6.972 0.08 Despite the simplicity, the M-C model is the most common model
Northridge- Pacoima Dam 103.6 7.01 0.082 8.596 0.16
used in numerical programs for modeling the soil behavior (Salem et al.,
01 (upper left)
Chi-Chi, TCU102 91.65 1.5 0.307 2.025 0.72 2020). This model was widely utilized for modeling the response of
Taiwan cantilever retaining walls subjected to seismic motions in previous
Landers Yermo Fire 151.33 23.6 0.213 8.089 0.68 contributions (Green et al., 2008; Madabhushi and Zeng, 2007). This soil
Station model was used even with software that was capable to apply advance
Bam, Iran Bam 124 1.7 0.157 8.019 0.2
models like FLAC software (Conti and Caputo, 2019). Salem et al. (2020)

3
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Fig. 1. Morgan hill velocity time history recorded during shaking at Coyote Lake Dam: (a) Main near-fault record; (b) Extracted pulse.

Fig. 2. Landers time histories recorded during shaking at Lucerne station: (a) Main near-fault record; (b) Generated far field record.

used Plaxis software and M-C and HSSMALL models. Based on (Salem for future interests.
et al., 2020) results, it can be concluded that although M-C constitutive Both the backfill and foundation soil layers were dry sand with
model is a simple model but predicts the performance of cantilever different compaction levels. An un-cemented sand (yellow Leighton
retaining wall with an acceptable degree of accuracy especially in case Buzzard 14–25 (Fraction B) silica sand) with 60% relative density of a
of displacement studies. Furthermore, the finding of authors revealed foundation layer and 22% for a backfill with dense and medium dense
that using the M-C model may be conservative due to the prediction of layers, respectively, were used in the experimental tests. The peak values
higher values of forces and displacement. Base on the mentioned of friction angle for backfill and foundation are reported as 33.5◦ and
statement using M-C is an appropriate model for addressing seismic 42.5◦ by Kloukinas et al (Kloukinas et al., 2014; di Santolo and Aldo,
lateral displacement of cantilever retaining wall under real earthquake 2011), respectively. The numerical results are calibrated by using
scenarios and initial investigation for addressing current study. sensitivity analysis of the friction angles of backfill and foundation soils
It should be noted that despite the wide application of the M-C to be matched with the shaking table test results. The best compatibility
model, many aspects of soil may not be modeled as real. However, the in the case of lateral displacement between numerical modeling and the
fundamental findings and basic concepts of this research can be useful experimental test is achieved when the peak friction angles of

4
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Table 2 boundary. In order to capture the effects of box and wave reflection into
Scaling factor for 1 g shaking table tests. the model, two vertical free-field soil columns were defined and con­
Item Sign Prototype/model This research nected each side of the model to the main part using normal and shear
viscous dashpots, representing the viscous boundary condition. Co­
P/M
efficients of the dashpots were ρvs and ρvp per unit area; whereρ is the
Density ρ 1 1 density of the material andvs and vp are the s-wave and p-wave speed,
Length L N 8.5
respectively. Dashpots absorb energy in a procedure similar to that used
Stress σ N 8.5
in the application of quiet boundaries. Dashpots make viscous normal
Shear strain ε N0.5 2.9
and shear tractions. The mesh generated pattern as well as artificial soil
Displacement d N1.5 24.7
columns and viscous boundaries are depicted in Fig. 3.
Shear modulus G N0.5 2.9
Acceleration a 1 1 3.4. Calibration
Frequency f N− 0.75 0.2
Dynamic time t N0.75 5 Numerical modeling was applied to simulate shaking table experi­
ments based on a prototype wall. Seismic sinusoidal excitation was,
therefore, used in the experimental program and scaled to the time and
foundation and backfill in the FEM model are reduced to 39◦ and 30◦ ,
frequency domain, with the same amplitude of 0.23 g and rough inter­
respectively. The calibrated soil properties used in the numerical model
face configuration of the experimental program. The top wall seismic
are summarized in Table 3.
displacement, accelerations in both the bottom and top of the wall, and
The damping ratio was defined in the model via two Rayleigh
dynamic bending moment along the wall were all validated. Fig. 4 shows
damping coefficients, α and β. The coefficients determine the damping
the numerical and experimental results of FE and the experimental
matrix C, which is a function of the mass and stiffness matrices, see Eq.
models. Based on the results, it can be concluded that the proposed
(1).
numerical model matched the physical modeling results with a reason­
C = αM + βK (1) able degree of accuracy. The difference found in the range of lateral
displacement in the numerical and physical model is mainly due to the
The two Rayleigh damping coefficients α and β were determined fact that in the FE model, relative displacement between the wall and
using Eqs. (2a) and (2b), as developed by Ju and Ni (2007): displacement time history of loading is captured and is not a pure
α = 2ω1 ω2 (D1 ω2 − D2 ω1 )/(ω2 2 − ω1 2 ) (2a) movement of the soil and structure system. In this condition, both the
wall and motion displacement time series have a similar oscillation and
( )
β = 2(D2 ω2 − D1 ω1 )/[π ω2 2 − ω1 2 ] (2b) different absolute values. Using relative displacement eliminates the
oscillation domain and captures the residual displacement in the wall.
where D1 and D2 are fractions of critical damping at two different cir­ Based on the aforementioned explanations and despite the in­
cular natural frequencies (ω1 and ω2 ). compatibility found in oscillations of the upper and lower bands of
The retaining wall was modeled as a non-yielding element in displacements, it can be concluded that the total residual movement is
experimental procedures by using aluminum material. Because the wall fitted with a good degree of accuracy as marked in Fig. 4 a.. Garini et al.,
behavior is fully linear elastic and perfectly rigid, damage potential and on the other hand, reported a similar trend of bending moments along
flexibility were not defined in the numerical model. The aluminum wall the wall, as modeled by the FE method (Garini et al., 2016).
in the experimental tests was modeled using the reinforced concrete Following model validation through imposing earthquake motions
material in the prototype simulating the real type of cantilever retaining (near-fault and semi-artificial far-field records) to the base of the model
walls. Also, a rough interface between wall and soil was created in the in X direction, the soil-structure response was captured and reported in
experimental tests by pasting rough sandpaper on the footing surface the present research.
(Kloukinas et al., 2014, 2015). This rough contact simulated the real
concrete-soil interface, and the assumption made regarding the concrete 4. Response of the cantilever retaining wall to near-fault
retaining wall was logical. motions

To better understand the impact of near-fault earthquakes, the re­

3.3. Soil wall interaction and boundary condition sults of two famous events with medium and high PGA values, specif­
ically the Landers earthquake, recorded at the Lucerne station, with a
The contact type interaction was used to appropriately model the PGA equal to 0.725 g and the Chi-Chi event, recorded at the TCU102
interface between soil and wall. The tension-less but frictional behavior station, with a PGA equal to 0.3 g, are compared and presented in this
was simulated by the defining μ coefficient for tangential behavior. An section. The acceleration and velocity time histories of the TCU102 re­
interface friction angle of 28.5◦ was obtained from the static pull out test cord are depicted in Fig. 5.
by Kloukinas et al. (2015). The normal behavior and the hard contact
was applied to allow both separation and sliding, 4.1. Top wall displacement and settlement
To reduce the radiation damping effect, Lysmer and Kuhlemeyer
(1969) and Kuhlemeyer and Lysmer (1973) proposed applying viscous The displacement and settlement time histories of the Landers
boundary conditions to absorb reflected waves along the artificial earthquake at the Lucerne station and the Chi-Chi earthquake at the
TCU102 station along the highest point of the wall are depicted in
Table 3 Fig. 6a and 6b, respectively.
Soil properties used in validation. It can be concluded from the results that about 47% and 39% of the
Item G Poisson φp φres ψp ψ res γ final residual displacements and settlements caused by earthquake
(MPa) ratio (υ) (◦ ) (◦ ) (◦ ) (◦ ) (KN/ loading occurred in a narrow band of 2 s, which is mainly due to the
m3)
directivity pulse of the Landers and Chi-Chi earthquakes, respectively.
Backfill soil 58 0.3 30 30 0 0 15.1 This finding is in complete agreement with the fact that the main part of
Foundation 116 0.3 39 36 9 6 16.1 seismic demand in pulse-like motions is accumulated in the directivity
soil
pulse (Mukhopadhyay and Gupta, 2013). Eliminating the pulse from the

5
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Fig. 3. Cantilever retaining system a) Numerical model b) experimental configuration (Kloukinas et al., 2015).

Fig. 4. Comparison between the numerical prediction and experimental results.

main records helped reduce the final residual displacements imposed by slightly reduced, prominent differences captured in the displacement
these earthquakes. As previously mentioned, while the maximum ac­ showed the vital role of the velocity time history. The effect of forward
celeration of the generated far-field motion in the main records was only directivity on the performance-based design of retaining structures, as

6
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Fig. 5. Chi-Chi time histories recorded during shaking at TCU102 station: (a) Acceleration; (b) Velocity.

Fig. 6. Top wall displacement and settlement time histories of main and generated far field record for earthquake (a) Landers; (b) Chi-Chi.

Fig. 7. Bending moments of main and generated far field record for earthquakes: (a) Landers; (b) Chi-Chi.

7
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

illustrated by the results of the movements mentioned herein, is also fault type of motion imposes a larger movement than the permissible
important to consider. and ultimate values of the displacements mentioned in the literature.
Hence, a comprehensive study was employed by imposing different
4.2. Bending moment along the wall near-fault earthquake scenarios with PGAs ranging from 0.1 g to 0.6 g, to
the verified model, and the results were compared with the criteria
The effect of near-fault ground motions on the distribution of forces found in the literature and codes. Some of these criteria are reported in
along the wall and bending moment in the height of the wall should be Table 4.
considered. Results of the bending moment induced by the Landers The results of the numerical model are depicted in Fig. 9 and re­
(Lucerne) and Chi-Chi (TCU102 station) excitations are illustrated in ported in Table 5.
Fig. 7. Table 5 and Fig. 9 show that the walls that fell into the range of low
It can be concluded that having approximately the same PGA but to moderate earthquakes (0.1–0.3 g) experienced lower horizontal dis­
different PGV can still provide equal distribution of bending moment placements than permissible and the failure states have been noted in
along the wall. Besides the apparent discrepancy in the lateral dis­ the literature. However, acceleration of 0.4 g induced a lateral
placements of the wall, these results illustrated the inherent differences displacement to the wall that was larger than the permissible state but
of the earthquakes recorded in the vicinity of the ruptured faults. These failure condition did not occurred, yet. For base accelerations of more
observations are compatible with the findings of Bakr and Ahmad than 0.5 g the lateral movement of wall exceeded from the failure state
(2018). The authors noted that seismic active earth pressures were not reported by Huang et al. (2009). Fig. 10 illustrates the normalization of
dependent on wall lateral displacements during a shaking. As mentioned the data by dividing the Commission of the European Communities (The
before, eliminating the pulse from the main shocking slightly decreases Commission of the European Communities European prestandard,
the PGA of input motion, causing the bending moment of the wall in the 1994) values into the residual displacements captured from different
near-fault and far-field scenarios to experience almost the same seismic events. As seen in Fig. 10; strong motions caused the retaining wall to
forces. The accuracy of the findings can be justified with these facts. experience a displacement of up to 5 times the recommended value.
Under the assumptions made in this research, an acceleration of 0.4 g for
near-fault records was the critical value that imposed devastating dis­
4.3. Effect of velocity and acceleration placements onto the system and caused failure state. For validation of
this fact, all scenarios with the seismic properties reported in Table 1
Fig. 8 compares the magnitude of a peak to peak velocity (PPV) and were compared with the permissible and failure states reported in
the peak ground acceleration of records reported in Table 3 with the Table 6. The data showed that 83% of ground motions in this study had a
captured responses of the numerical modeling. PGA greater than 0.4 g. It is noticeable that all the records sensed a
As seen in Fig. 8a, the absence of a clear relationship between the displacement that was larger than the permissible state. According to the
magnitude of acceleration and induced lateral displacement led to a vast Huang et al. failure state protocol (Lysmer and Kuhlemeyer, 1969); 50%
data scatter. However, as seen in Fig. 8b, using PPV instead of PGA made of motions experienced movement greater than 5% of the wall height.
the chart more regular and trend-line has been manifested. Davoodi Based on the Wu and Prakash et al. failure limit (Wu and Prakash, 1996),
et al. (2013) reported a similar trend for an embankment dam that was 33% of cases experienced a larger lateral displacement than 10% of the
excited by near fault records. Zamiran and Osouli (2018) developed an wall height. The wall failure is reported when PGA of input motion
equation and connected the maximum displacement of the cantilever reaches 0.47 g in 50% of cases (Zamiran and Osouli, 2018) which is close
wall to the PGA of free field. Findings of the current study showed that
this equation is not accurate enough for the near-fault earthquake
The report on the close connection between velocity and displace­ Table 4
ment found velocity to be the main factor in the seismic analysis of Criteria for performance-based analysis of retaining walls.
retaining structures excited by pulse-like ground motions, a factor that Eurocode 8 The Wu and AASHTO Huang
had not been considered considerably in traditional approaches. While Commission of the Prakash (2002) et al.
the same process was observed for PGV, the PPV trend-line reported in European Communities (1996) (2009)
(1994)
this research was found to be more accurate.
Permissible 300amax 0.02*H 250amax 0.02*H
state(mm)
5. Performance-based analysis Failure state – 0.1*H – 0.05*H
(mm)
A comparison of the results shows that the lateral displacement of the
amax = maximumearthquakeacceleration(g)
wall was remarkably higher under near-fault ground motions than far-
H = Wallheight(mm)
field ground motions. This brings up the concern that perhaps near-

Fig. 8. Correlation between lateral displacement of wall and (a) PGA; (b) PPV.

8
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

displacements and have lower failure states in all cases, except at an

acceleration of 0.6 g. These findings again confirm the importance of
near-fault excitations and show that basing a decision on only the base
acceleration may cause catastrophic consequences.
On the other hand, certain instructions, including the EN 1998–5
procedure, tend to underestimate the residual displacements and cannot
predict failure (Deyanova et al., 2016). The above statements show that
the displacement analysis of retaining walls is a complicated problem
that needs precise approaches in high seismic hazard sites.

6. Parametric study

The verified FE model was used on a prototype scale to examine the

effect of various parameters, such as the magnitude of input accelera­
tion, geotechnical properties of backfill and foundation soil as well as
Fig. 9. Seimic displacement of wall under near-fault motions with
the frequency content of the main shocking on seismic responses of the
performance-based criteria.
retaining structure.

Table 5 6.1. Magnitude of input acceleration

Displacement of wall in different scenarios.
Acceleration (g) 0.1 0.2 0.3 0.4 0.5 0.6
As illustrated in Fig. 12, the different absolute values of the input
Event acceleration of the Loma Prieta earthquake at the LGPC station, which
ranged between 0.1 g and 0.57 g (the PGA of main records), were
Landers(Lucerne) 9.8 29.7 74.8 134.2 217.1 437.9
Imperial Valley-06(El 4.8 16.8 92.8 351.4 606.5 1028.7 applied.
Centro - Meloland Geot. As seen in Fig. 12, all parameters were kept at a constant value, and
Array) the effects of the acceleration magnitude were recorded. Fig. 12a and
Kobe(Takarazuka) 9.7 16.2 37.6 78.8 129.4 198.6 12b show that increasing the acceleration strongly changes the lateral
Loma Prieta(LGPC) Landers 12.9 36.8 78.3 165.3 318.6 528.4
Tabas, Iran 20.3 55.5 123.7 223.2 366.0 523.8
and vertical displacements of the top wall. The figures also show the
Chi-Chi(TCU102) 19.8 73.2 182.1 476.1 908.1 1593.2 effect of directivity pulses on wall displacement. It is worth noting that
The Commission of the 30 60 90 120 150 180 scaling the acceleration magnitude kept the PGV/PGA ratio the same in
European Communities all models. The results show that increasing the acceleration magnitude
(1994)
also increases the displacement and, thereby, worsens the condition of
Wu and Prakash (1996) 108 108 108 108 108 108
AASHTO (2002) 25 50 75 100 125 150 the settlements. The bending moment along the wall is shown in
Huang et al. (2009) 108 108 108 108 108 108 Fig. 12c. The more the acceleration values of the wall, the greater the
Huang et al. (2009) 270 270 270 270 270 270 magnitude of the bending moments. Fig. 13 shows that aligning the
(failure) lateral displacement and bending moment magnitudes with the accel­
eration values identified a distinctive trend, whereby, enhancing the
input values led to further refining of the bending moment.
Lateral displacement of the acceleration value had a trend-line with a
polynomial equation (order of 2). This is due to the seismic demand
imposed on the models. The seismic energy of an earthquake, as widely
discussed in the literature, is proportional to the square of velocity
∫
( v2 dt). A trend-line with an equation of order 2 and a downward
concavity was observed for the moment-acceleration chart, which in­
dicates that increasing the acceleration reduces the rate of the moment
enhancing. This is contrary to the results found in the displacement-
acceleration charts. At an acceleration level greater than 0.4 g, the
retaining wall was observed to move without increments in the seismic
forces (Bakr and Ahmad, 2018). This finding is in good agreement with
the results.
Another parameter measured against the acceleration magnitudes
was the amplification patterns along the soil profile (Fig. 12d).
Increasing the input motion value changes the value and amplification
Fig. 10. Normalized Seimic displacement of wall under near-fault motions with factor of the soil . The amplification factor was seen to decrease under
Eurocode 1994. higher level of excitations . This finding can be justified by the theory
that suggests higher amplification values can be captured in lower levels
to the results of current numerical models of acceleration.
These findings indicate that the absence of proper solutions in the
design process of retaining structures for reducing displacements in­ 6.2. Properties of backfill soil
creases the failure states under near-fault ground shakings. Semi artifi­
cial motions generated in Section 2 were scaled with the same values of The response and performance of the retaining wall was examined
near-fault records, and responses of the wall were recorded to compare against the geotechnical properties of the backfill layer by a compre­
the results of near and far field records. For better comparison, only hensive sensitivity analysis. The Loma Prieta earthquake at the LGPC
records with a single obvious pulse in the velocity time series were station with PGA equal to 0.57 g is used as input motion in Sections 6.2
selected. As seen in Fig. 11, far field scenarios cause smaller and 6.3. Fig. 14 illustrates the cohesion and friction angle.
The mechanical properties of the backfill soil had a considerable

9
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Table 6
Displacement of wall at different scenarios for site acceleration.
Event Station Residual displacement (mm) Permissible state Huang et al. (2009) Wu and Prakash (1996)

Imperial Valley-06 El Centro - Meloland Geot. Array 90.0 108 270 540
Kobe, Japan Takarazuka 198.8 108 270 540
Morgan Hill Coyote Lake Dam 144.2 108 270 540
Parkfield-02, CA Parkfield-Cholame 2WA 58.6 108 270 540
San Fernando Pacoima Dam (upper left abut) 595.9 108 270 540
Tabas, Iran Tabas 1340.3 108 270 540
Loma Prieta LGPC 466.9 108 270 540
Landers Lucerne 941.2 108 270 540
Northridge-01 Pacoima Dam (upper left) 333.5 108 270 540
Chi-Chi, Taiwan TCU102 122.6 108 270 540
Landers Yermo Fire Station 826.3 108 270 540
Bam, Iran Bam 1138.8 108 270 540

6.4. Effect of the frequency content

One of the important aspects of the dynamic analysis of soil-

retaining-structure systems is the frequency content of the input mo­
tion, which wasn’t considered in more ordinary methods, such as the
pseudo-static and M-O approaches. For assessing the effect of the fre­
quency content on the seismic performance of the cantilever retaining
walls, a series of input motions from distinct stations scaled in the same
PGA imposed on the models and wall responses were captured. A similar
method was employed by Cakir (2013), with the difference that the
author used different earthquakes scaled to the same amplitude. Using
different earthquakes for this purpose may not, however, only introduce
the frequency content because the records have different faulting
mechanisms and site conditions. For better comparison, motions of the
same earthquake events recorded at different stations were used for
Fig. 11. Seimic displacement of wall under far field motions with performance-
evaluating the frequency content. The input motion of “Imperial Valley-
based criteria. 06′′ was selected, and the results are reported herein. All ground motions
were in the vicinity of the fault rupture, representative of the same
condition and scaled to the PGA of 0.3 g. The main properties of the
effect on displacement-based performances of the retaining structures.
selected station are reported in Table 7, and the Fourier spectrum of this
Improvements in the mechanical resistance properties of the backfill
motion is illustrated in Fig. 16.
improved wall performance and reduced wall displacement. Enhancing
The residual displacement and bending moment results of the fre­
the cohesion and friction angle did not significantly impact the distri­
quency content effect are depicted in Fig. 17.
bution of the bending moment along the wall height, with variation of
As seen in the figure, the maximum bending moment value in the
friction angle showing the least effect
record by the record with RSN number 171 was at a minimum among
. Increasing the magnitude of soil cohesion reduced the maximum
three records, but this motion imposed the highest residual displace­
bending moment value. A similar conclusion was reported by Osouli and
ment value into the wall. As reported in Table 7, this record contains a
Zamiran (2017). Their results showed that compaction of soil with a
directivity pulse and PGV value. Given that the distribution of bending
more significant relative density without changing the fine content
moments along the wall and the PGA were the same in the other two
improved the responses of soil-retaining structure systems. Furthermore,
records, it can be concluded that increasing the PGV/PGA ratio, reported
Zamiran and Osouli (2018) reported that Maximum lateral displacement
in Table7, increased the values of the lateral displacements of the wall.
of the wall with the cohesion of 10 kPa was a quarter of the displacement
Cakir (2013), Bakr and Ahmad (2018) also discussed the effects of PGV/
of cases with 0 kPa of cohesion. This statement is compatible with the
PGA ratios on wall response. So, the effects and main role of the fre­
results of Fig. 14.
quency content of the ground motions are manifested in the response of
the retaining structures, herein.
6.3. Properties of the foundation soil

7. Conclusion
By using the findings of Section 6.2, the friction angle of foundation
soil was changed to assess its impact on the system’s response. Results of
A series of dynamic finite element numerical models were conducted
the displacements are depicted in Fig. 15. The lateral movement of the
in the present research to evaluate the performance of cantilever
wall had a direct correlation with the friction angle. Changing the fric­
retaining walls. Addressing seismically induced lateral displacements
tion angles from 39 to 33 and from 39 to 45 changed the magnitude of
under near-fault excitations were the main part of interest. Due to the
the residual displacement by 49% and − 26%, respectively.
fact that forward directivity effect is more destructive than other near-
It should be noted that the modulus of elasticity was constant in all
fault effects, this paper concentrated on FD motions using a fully dy­
models and the effect of the different magnitudes wasn’t investigated in
namic time-domain analysis. The wall bending moments, top horizontal
this work. Cakir (2014) performed a sensitivity analysis based on
displacement and settlements were measured as an index of the dynamic
modulus magnitudes. It can be concluded from the results that lower
response of cantilever retaining walls. The captured lateral movements
values of displacements were captured in case of higher modulus.
in a wide range of PGAs were compared with criteria that developed in
literature and codes, for the permissible and failure states of wall
displacement. A comprehensive parametric study was applied and ef­
fects of different parameters on the response retaining wall were

10
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Fig. 12. Effect of acceleration value on: (a) Lateral displacement of top wall; (b) Vertical displacement of top wall; (c) Bending moment; (d) Amplification pattern.

Fig. 13. Trend of parameters to acceleration: (a) Residual lateral displacement; (b) Bending moment.

recorded. The effects of input motion PGA, backfill and foundation time, due to the pulse effect. Eliminating this pulse through the
mechanical properties and frequency content of ground motion were wavelet method considerably reduced the maximum lateral
examined and reported. displacements.
The following conclusions are drawn based on the physical proper­ (2) The results showed that the near-fault ground motions and far-
ties of the materials and soil model used in the present study: field records imposed approximately identical moment values
to the wall. This fact illustrates that distribution of force behind
(1) Ground motions with the forward directivity pulse, induced very the wall is not only design parameter.
large residual displacements into the cantilever retaining walls. (3) The retaining wall under near-source earthquakes with PGAs 0.1
The main part of displacement occurred in a narrow bandwidth of g to 0.3 g, experienced horizontal displacements lower than the

11
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Fig. 14. Seismic response of retaining wall based on backfill properties (a) Cohesion; (b) friction angle.

Table 7
Main properties of “Imperial Valley-06′′ selected stations.
Station RSN PGV Rrup Arias Tp PGV/
number (cm/s) (km) intensity (s) PGA
(m/s) (s)

“El Centro - 171 92.7 0.07 1.105 0.56 0.302

Meloland
Geot. Array”
“El Centro 184 26.67 5.09 0.831 0.4 0.087
Differential
Array”
“El Centro Array 175 44.5 17.94 1.709 0.2 0.151
#12′′

Sliding was, therefore, considered as the main failure mechanism

Fig. 15. Seismic response of retaining wall based on foundation friction angle. in strong ground motions.
(6) The ratio of PGV/PGA had an undeniable effect on the response of
permissible and failure states based on criteria reported in table4 the system. The results showed that, with same strong ground
in particular, Huang et al. (2009). The lateral displacements motion properties including PGA, duration and faulting mecha­
under records with PGA = 0.4 g were almost larger than the nisms, system response depends strongly on PGV/PGA values. A
permissible state. The residual displacements exceeded from 5% higher value of displacement was observed in records with larger
of the wall height for earthquakes with PGAs more than 0.5 g and PGV/PGA magnitudes, even in cases where the bending moment
the failure condition happened. distribution was lowest.
(4) Lower displacements were recorded for generated far-field re­ (7) A comparison between the results and the literature shows that if
cords, although the failure state was observed for records with soil properties remain constant, wall movement will be affected
PGA = 0.6 g. by both the PGA and the velocity of records and not only PGA
(5) The rotation of wall had a reverse correlation with free field PGA. value.
Models under smaller PGAs had a larger magnitude of rotation.

12
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Fig. 16. Imperial Valley-06 scaled acceleration time histories and Fourier spectrum recorded during shaking at station: (a) “El Centro - Meloland Geot. Array”; (b) “El
Centro Array #12′′ ; (c) ”El Centro Differential Array“

Fig. 17. Effect of frequency content on Seismic response of retaining wall (a) Lateral displacement; (b) Bending moment.

13
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

CRediT authorship contribution statement Green, Russell A., Guney Olgun, C., Cameron, Wanda I., 2008. Response and modeling of
cantilever retaining walls subjected to seismic motions. Comput.-Aided Civ.
Infrastruct. Eng. 23 (4), 309–322.
Milad Aghamolaei: Software, Validation, Formal analysis, Writing - Hall, J.F., Heaton, T.H., Halling, M.W., Wald, D.J., 1995. Near-source ground motion and
original draft. Alireza Saeedi Azizkandi: Conceptualization, Method­ its effects on flexible buildings. Earthq. Spectra 11 (4), 569–605.
ology, Supervision. Mohammad Hassan Baziar: Conceptualization, Hayden, Connor P., Bray, Jonathan D., Abrahamson, Norman A., 2014. Selection of near-
Fault Pulse Motions. J. Geotech. Geoenviron. Eng. 140 (7), 04014030.
Methodology, Supervision. Sadegh Ghavami: Writing - review & Huang, C.-C., Wu, S.-H., Wu, H.-J., 2009. Seismic displacement criterion for soil
editing. retaining walls based on soil strength mobilization. J. Geotech. Geoenviron. Eng. 135
(1), 74–83.
Iai, S., Tobita, T., Nakahara, T., 2005. Generalised scaling relations for dynamic
centrifuge tests. Geotechnique 55 (5), 355–362.
Declaration of Competing Interest
Jadhav, Prajakta R., Prashant, Amit, 2020. Computation of seismic translational and
rotational displacements of cantilever retaining wall with shear key. Soil Dyn.
The authors declare that they have no known competing financial Earthquake Eng. 130, 105966.
interests or personal relationships that could have appeared to influence Jo, Seong-Bae, Ha, Jeong-Gon, Lee, Jin-Sun, Kim, Dong-Soo, 2017. Evaluation of the
seismic earth pressure for inverted T-shape stiff retaining wall in cohesionless soils
the work reported in this paper. via dynamic centrifuge. Soil Dyn. Earthquake Eng. 92, 345–357.
Ju, Shen-Haw, Ni, Sheng-Huoo, 2007. Determining Rayleigh damping parameters of soils
for finite element analysis. Int. J. Numer. Anal. Meth. Geomech. 31 (10), 1239–1255.
Appendix A. Supplementary material
Kalkan, Erol, Kunnath, Sashi K., 2006. Effects of fling step and forward directivity on
seismic response of buildings. Earthq. Spectra 22 (2), 367–390.
Supplementary data to this article can be found online at https://doi. Kloukinas, Panos, Penna, Augusto, di Santolo, Anna Scotto, Bhattacharya, Subhamoy,
Dietz, Matt S., Dihoru, Luiza, Evangelista, Aldo, Simonelli, Armando L., Taylor, Colin
org/10.1016/j.compgeo.2020.103924.
A., Mylonakis, George, 2014. Experimental investigation of dynamic behavior of
cantilever retaining walls. In: Seismic Evaluation and Rehabilitation of Structures,
References 477-493: Springer, 2014.
Kloukinas, Panos, di Santolo, Anna Scotto, Penna, Augusto, Dietz, Matthew, Evangelista,
Aldo, Simonelli, Armando Lucio, Taylor, Colin, Mylonakis, George, 2015.
Alavi, Babak, Krawinkler, Helmut, 2000. Consideration of near-fault ground motion
Investigation of seismic response of cantilever retaining walls: limit analysis vs
effects in seismic design. In: Proceedings of the 12th World Conference on
shaking table testing. Soil Dyn. Earthq. Eng. 77 (2015) 432–445.
Earthquake Engineering, 8, 2000.
Kuhlemeyer, Roger L., Lysmer, John, 1973. Finite element method accuracy for wave
American Association of State Highway and Transportation Officials AASHTO, 2002.
propagation problems. J. Soil Mech. Foundations Div 99, no. Tech Rpt (1973).
Standard specifications for highway bridges, Secs. 3 and 7, Washington, D.C.
Lysmer, John, Kuhlemeyer, Roger L, 1969. Finite dynamic model for infinite media.
Anastasopoulos, I., Gazetas, G., Bransby, M.F., Davies, M.C.R., El Nahas, A., 2007. Fault
J. Eng. Mech. Division 95 (4), 859–878.
rupture propagation through sand: Finite-element analysis and validation through
Madabhushi, S.P.G., Zeng, X., 2007. Simulating seismic response of cantilever retaining
centrifuge experiments. J. Geotech. Geoenviron. Eng. 133 (8), 943–958.
walls. J. Geotech. Geoenviron. Eng. 133 (5), 539–549.
Al Atik, Linda, Sitar, Nicholas, 2010. Seismic earth pressures on cantilever retaining
Mavroeidis, George P., Papageorgiou, Apostolos S., 2003. A mathematical representation
structures. J. Geotech. Geoenviron. Eng. 136(10), 1324–1333.
of near-fault ground motions. Bull. Seismol. Soc. Am. 93 (3), 1099–1131.
Baker, Jack W., 2007. Quantitative classification of near-fault ground motions using
Mikola, Geraili, Roozbeh, Gabriel Candia, Sitar, Nicholas, 2016. Seismic earth pressures
wavelet analysis. Bull. Seismol. Soc. Am. 97 (5), 1486–1501.
on retaining structures and basement walls in cohesionless soils. J. Geotech.
Bakr, Junied, Ahmad, Syed Mohd, 2018. A Finite element performance-based approach
Geoenviron. Eng. 142 (10), 04016047.
to correlate movement of a rigid retaining wall with seismic earth pressure. Soil Dyn.
Mollaioli, Fabrizio, Bosi, Anna, 2012. Wavelet analysis for the characterization of
Earthquake Eng. 114, 460–479.
forward-directivity pulse-like ground motions on energy basis. Meccanica 47 (1),
Bakr, Junied, Ahmad, Syed Mohd, Lombardi, Domenico, 2019. Finite-element study for
203–219.
seismic structural and global stability of cantilever-type retaining walls. Int. J.
Mononobe, N., Matsuo, H., 1929. On the determination of earth pressures during
Geomech. 19 (10), 04019117.
earthquakes. Volume 9, Tokyo, 1929.“ In World Engineering Congress.
Baziar, M.H., Rostami, H., 2017. Earthquake demand energy attenuation model for
Mukhopadhyay, S., Gupta, V.K., 2013. Directivity pulses in near-fault ground
liquefaction potential assessment. Earthq. Spectra 33 (2), 757–780.
motions—I: Identification, extraction and modeling. Soil Dyn. Earthq. Eng. 50, 1–15.
Bertero, V.V., Mahin, S.A., Herrera, R.A., 1978. Aseismic design implications of near-
Nakamura, Shinya, 2006. Reexamination of mononobe-okabe theory of gravity retaining
fault San Fernando earthquake records. Earthq. Eng. Struct. Dyn. 6 (1), 31–42.
walls using centrifuge model tests. Soils Found. 46 (2), 135–146.
Brandenberg, Scott J., Mylonakis, George, Stewart, Jonathan P., 2017. Approximate
Nimbalkar, Sanjay, Choudhury, Deepankar, 2007. Sliding stability and seismic design of
solution for seismic earth pressures on rigid walls retaining inhomogeneous elastic
retaining wall by pseudo-dynamic method for passive case. Soil Dyn. Earthquake
soil. Soil Dyn. Earthquake Eng. 97, 468–477.
Eng. 27 (6), 497–505.
Bray, Jonathan D., Rodriguez-Marek, Adrian, 2004. Characterization of forward-
Okabe, Saburo, 1924. General Theory on Earth Pressure and Seismic Stability of
directivity ground motions in the near-fault region. Soil Dyn. Earthq. Eng. 24(11),
Retaining Wall and Dam. Proc. Civil Engrg. Soc., Japan 10 (6), 1277–1323.
815–828.
Osouli, Abdolreza, Zamiran, Siavash, 2017. The effect of backfill cohesion on seismic
Cakir, Tufan, 2013. Evaluation of the effect of earthquake frequency content on seismic
response of cantilever retaining walls using fully dynamic analysis. Comput.
behavior of cantilever retaining wall including soil-structure interaction. Soil Dyn.
Geotech. 89, 143–152.
Earthquake Eng. 45, 96–111.
Psarropoulos, P.N., Klonaris, G., Gazetas, G., 2005. Seismic earth pressures on rigid and
Cakir, Tufan, 2014. Backfill and subsoil interaction effects on seismic behavior of a
flexible retaining walls. Soil Dyn. Earthquake Eng. 25 (7–10), 795–809.
cantilever wall. Geomech. Eng. 6 (2), 117–138.
Salem, Abdelwahhab N., Ezzeldine, Omar Y., Amer, Mohamed I., 2020. Seismic loading
Candia, Gabriel, Mikola, Roozbeh Geraili, Sitar, Nicholas, 2016. Seismic response of
on cantilever retaining walls: Full-scale dynamic analysis. Soil Dyn. Earthquake Eng.
retaining walls with cohesive backfill: centrifuge model studies. Soil Dyn.
130, 105962.
Earthquake Eng. 90, 411–419.
Santhoshkumar, G., Ghosh, Priyanka, Murakami, Akira, 2019. Seismic active resistance
Conti, R., Caputo, G., 2019. A numerical and theoretical study on the seismic behaviour
of a tilted cantilever retaining wall considering adaptive failure mechanism. Int. J.
of yielding cantilever walls. Géotechnique 69 (5), 377–390.
Geomech. 19 (8), 04019086.
Conti, R., Madabhushi, G.S.P., Viggiani, G.M.B., 2012. On the behaviour of flexible
Seed, H., 1970. Design of earth retaining structures for dynamic loads. In: ASCE Specialty
retaining walls under seismic actions. Géotechnique 62 (12), 1081.
Conf.-Lateral Stress in the Ground and Design of Earth Retaining Structures, 1970,
Davoodi, M., Jafari, M.K., Hadiani, N., 2013. Seismic response of embankment dams
1970.
under near-fault and far-field ground motion excitation. Eng. Geol. 158, 66–76.
Somerville, P.G., 2003. Magnitude scaling of the near fault rupture directivity pulse.
Deyanova, Manya, Lai, Carlo G., Martinelli, Mario, 2016. Displacement–based
Phys. Earth Planet. Inter. 137 (1–4), 201–212.
parametric study on the seismic response of gravity earth-retaining walls. Soil Dyn.
Somerville, P.G., Smith, N.F., Graves, R.W., Abrahamson, N.A., 1997. Modification of
Earthquake Eng. 80, 210–224.
empirical strong ground motion attenuation relations to include the amplitude and
di Santolo, Anna Scotto, Evangelista, Aldo, 2011. Dynamic active earth pressure on
duration effects of rupture directivity. Seismol. Res. Lett. 68 (1), 199–222.
cantilever retaining walls. Comput. Geotech. 38(8) 1041–1051.
Song, Jian, Rodriguez-Marek, Adrian, 2014. Sliding Displacement of Flexible Earth
Garini, E., Tsantilas, L., Gazetas, G., 2016. Seismic response of cantilever retaining walls:
Slopes Subject to near-Fault Ground Motions. J. Geotech. Geoenviron. Eng. 141 (3),
verification of centrifuge experiments. Proceedings of the 1st ICONHIC (207).
04014110.
Garini, E., Gazetas, G., 2013. Damage Potential of near-Fault Records: Sliding
The Commission of the European Communities European prestandard, 1994. Eurocode 8:
Displacement against Conventional “Intensity Measures”. Bull. Earthq. Eng. 11 (2),
Design provisions for earthquake resistance of structures—Part 5: Foundations,
455–480.
retaining structures and geotechnical aspects. Brussels.
Gazetas, George, Garini, Evangelia, Anastasopoulos, I., Georgarakos, T., 2009. Effects of
Vassiliou, Michalis F, Makris, Nicos, 2011. Estimating time scales and length scales in
near-fault ground shaking on sliding systems. J. Geotech. Geoenviron. Eng. 135(12),
pulselike earthquake acceleration records with wavelet analysis. Bull. Seismol. Soc.
1906–1921.
Am. 101 (2), 596–618.
Gazetas, G., Psarropoulos, P.N., Anastasopoulos, I., Gerolymos, N., 2004. Seismic
Veletsos, Anestis S., Younan, Adel H., 1997. Dynamic response of cantilever retaining
behaviour of flexible retaining systems subjected to short-duration moderately
walls. J. Geotech. Geoenviron. Eng. 123 (2), 161–172.
strong excitation. Soil Dyn. Earthquake Eng. 24 (7), 537–550.

14
M. Aghamolaei et al. Computers and Geotechnics xxx (xxxx) xxx

Wu, Yingwei, Prakash, Shamsher, 1996. On seismic displacements of rigid retaining Zou, Degao, Han, Huichao, Liu, Jingmao, Yang, Dixiong, Kong, Xianjing, 2017. Seismic
walls. In: Analysis and Design of Retaining Structures against Earthquakes, 21–37. Failure Analysis for a High Concrete Face Rockfill Dam Subjected to near-Fault
Zamiran, Siavash, Osouli, Abdolreza, 2018. Seismic motion response and fragility Pulse-Like Ground Motions. Soil Dyn. Earthquake Eng. 98, 235–243.
analyses of cantilever retaining walls with cohesive backfill. Soils Found. 58 (2),
412–426.

15