You are on page 1of 15

HOSTED BY Available online at www.sciencedirect.

com

ScienceDirect
Soils and Foundations xxx (2018) xxx–xxx
www.elsevier.com/locate/sandf

Seismic motion response and fragility analyses of cantilever


retaining walls with cohesive backfill
Siavash Zamiran a, Abdolreza Osouli b,⇑
a
Marino Engineering Associates, Inc., 1370 McCausland Ave, St. Louis, MO 63117, USA
b
Department of Civil Engineering, Southern Illinois University Edwardsville, 61 Circle Dr., Edwardsville, IL 62026-1800, USA

Received 18 May 2017; received in revised form 14 December 2017; accepted 17 December 2017

Abstract

The seismic motion response of a cantilever retaining wall with cohesive and cohesionless backfill materials was evaluated using fully
dynamic analysis based on finite difference method. The dynamic analysis was validated based on experimental test results and then com-
pared to analytical and empirical correlations based on Newmark sliding block method. Seven different earthquake events and the back-
fills with low to high levels of cohesion were considered. Nonlinear regression analyses were carried out to provide correlations between
free-field peak ground acceleration (PGA) and maximum relative displacement of the retaining wall. These results were compared to
results from empirical and analytical methods. Furthermore, fragility analyses were conducted to determine the probability of damage
to the retaining wall for different free-field PGAs and backfill cohesions. It is demonstrated to what extent a small amount of cohesion in
backfill material can influence displacement of the retaining wall and probability of damage in seismic conditions.
Ó 2018 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.

Keywords: Retaining wall; Dynamic analysis; Fragility analysis; FLAC; Backfill cohesion; Newmark sliding block method

1. Introduction ate maximum retaining wall displacement in seismic condi-


tions. The Richards and Elms empirical correlation (R&E)
Several retaining wall deformations and failures have has been suggested in different design guidelines including
been reported during historical earthquakes (Fang et al., Army Corps (Whitman and Liao, 1985) and AASHTO
2003; Ko et al., 2017; Lai, 1998; Ling et al., 2001; LRFD Bridge Design Specifications (AASHTO, 2007). In
Shakya, 1987). The most well-known method for predict- a more recent study conducted by Anderson et al. (2008)
ing the seismic deformation of retaining wall is known as and as part of The National Cooperative Highway
Newmark sliding block method (Newmark, 1965). The Research Program (NCHRP) study, an updated correla-
Newmark sliding block method requires the acceleration tion was provided based on various Newmark analyses.
time history of an earthquake in the free-field. However, The updated NCHRP equation has been embedded in
as the acceleration time history might not be available for the recent guidelines including Caltrans (Ertugrul 2013).
a practical design, some investigators including Richards More advanced Newmark based pseudo-static methods
and Elms (1979) developed empirical correlations to evalu- have also been developed to evaluate the sliding deforma-
tion of the retaining walls (e.g., Biondi et al., 2014; Conti
et al., 2013). However, it is noteworthy that the above-
Peer review under responsibility of The Japanese Geotechnical Society. mentioned studies only consider the sliding deformation
⇑ Corresponding author.
of the retaining wall and the rotational and tilting deforma-
E-mail addresses: szamiran@meacorporation.com (S. Zamiran),
aosouli@siue.edu (A. Osouli).
tion are neglected. Some investigators including Nadim

https://doi.org/10.1016/j.sandf.2018.02.010
0038-0806/Ó 2018 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
2 S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx

and Whitman (1984), Rafnsson (1991), Prakash et al. explored using fully dynamic analysis (FDA). The numer-
(1995), and Wu and Prakash (2001) adopted analytical ical modeling procedure will be validated based on cen-
approaches to predict the seismic deformation of the trifuge tests conducted by Agusti and Sitar (2013). To
retaining walls considering tilt and rotation of the retaining evaluate the performance of the retaining wall, seven differ-
wall. However, due to the complexity of these analytical ent input motions with different acceleration intensities will
procedures, the approaches have not been adopted in be considered. The displacements for various free-field
design guidelines for practical purposes. peak ground accelerations (PGA) will be analyzed. Fragi-
The deformational behavior of gravity and cantilever lity analyses will be conducted to evaluate the probability
retaining walls in seismic conditions, have been studied of damage based on different earthquake accelerations.
numerically (Agusti and Sitar, 2013; Corigliano et al.,
2011; Green et al., 2008; Green and Ebeling, 2003; 2. Methodology
Stamatopoulos et al., 2006; Wilson and Elgamal, 2010;
Wu and Prakash, 2001) and experimentally (Green and A series of FDA analyses based on finite difference
Ebeling 2003; Huang et al., 2009; Nakamura 2006; method in FLAC (Itasca 2011) is used to identify the dis-
Richards et al., 1996; Stamatopoulos et al., 2006; Wilson placement characteristics of the retaining walls. In the first
and Elgamal 2010, 2015; Zeng and Steedman 2000). The step, the centrifuge study conducted by Agusti and Sitar
main scope of most of these studies was evaluating sliding (2013) on retaining walls with cohesive backfill material
and rotational displacement of retaining wall systems dur- was used to verify the numerical modeling methodology
ing earthquake events. In a few studies (Agusti and Sitar (i.e., Analysis Group A in Table 1). The free-field motion
2013; Nakamura 2006; Wilson and Elgamal 2010, 2015), responses of the backfill and the wall displacement during
seismic earth pressure and retaining wall motion responses FDA were compared with centrifuge results of Agusti
were also investigated. In addition, Huang et al. (2009) and and Sitar (2013).
Wu and Prakash (2001) proposed a wall displacement cri- The validated model was utilized to identify the effect of
terion for identifying the level of damage for seismic per- 0, 10, and 30 kPa cohesion in backfills to cover a common
formance of retaining walls. range of cohesions reported for retaining wall backfills by
There are some limitations with these studies. For exam- Caltrans (Kapuskar 2005). Backfill materials were modeled
ple, most of the mentioned studies focused on retaining in unsaturated conditions. For these scenarios, labeled as
walls with cohesionless backfill materials. A few studies Analysis Group B in Table 1, the Imperial Valley earth-
(Agusti and Sitar 2013; Latifi et al., 2016; Mikola et al., quake (1979) was introduced to the model as the earth-
2014; Osouli and Zamiran 2017; Wilson and Elgamal quake input motion. The horizontal relative wall
2015; Zamiran and Osouli 2014, 2015) considered backfill displacement (RWD) time history was monitored at the
cohesion, however, their main focus was not displacement top and bottom of the wall. The RWD time history was
behavior of retaining walls. Moreover, specific backfill determined by subtracting the horizontal ground displace-
cohesion was selected in these studies. Therefore, the effects ment variation in free-field condition from the horizontal
of cohesion variation on seismic response of the walls were total wall displacement at the desired location (Green
not considered. In addition, only limited number of seismic et al., 2008). The maximum RWD during the earthquake
events and shaking intensities were used in these studies. is recorded and is compared with the results of analytical
There is limited information about the seismic deforma- solutions including Newmark sliding block method,
tional response of retaining walls with cohesive backfills. R&E, and NCHRP correlations (Anderson et al., 2008;
However, field inspections by Kapuskar (2005) show low Newmark, 1965; Richards and Elms, 1979).
to high level of cohesiveness in backfill materials. The study In order to evaluate seismic deformation of the retaining
conducted by Caltrans (Kapuskar 2005) investigated 20 walls based on different input motions, Analysis Group C
different bridge sites in the State of California. It was found shown in Table 1 was conducted. Seven different earth-
that in 18 cases, the backfill material of bridge abutments quake loadings including Imperial Valley (1979), Loma
contains some level of cohesiveness. In 9 cases, backfill Prieta (1989), Chi-Chi (1999), Kobe (1995), Northridge
materials with up to 95 kPa cohesion were observed. (1994), Hollister (1961), and Friuli (1976) were considered.
The other critical factor is the stochastic nature of earth- For each event, different input acceleration intensities were
quake and its related damages, which is often characterized applied to the model to correlate the maximum RWD
by the probability of occurrence or failures. For example, based on free-field PGA variations. The Amplification Fac-
fragility analyses have been used to evaluate the probability tor (AF) was used as a representative of input acceleration
of failures of different structures (Baker 2015) and caisson intensity. The AF of 100% shows an earthquake with the
quay walls (Ichii 2004; Jafarian et al., 2014). However, input PGA of 0.25 g. In these series of analyses, backfills
there has been a lack of knowledge in determining the fra- with three different cohesions including 0, 10, and 30 kPa
gility functions of cantilever retaining wall structures espe- were considered.
cially with cohesive backfills. Finally, the motion response from the mentioned seven
In this study, the seismic motion behavior of cantilever different earthquakes with various input acceleration inten-
retaining walls with different backfill cohesions will be sities and backfill cohesions were used to evaluate the fail-

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx 3

Table 1
Numerical modeling scenarios conducted in this study.
Analysis Scope Earthquake Backfill Cohesion AF (%)
Group (kPa)
A Centrifuge test simulation Loma Prieta 15 170
Kobe 15 60, 100, 170, 235
B Comparison of FDA with analytical correlations Imperial 0 335
Valley 10 335
30 335
C Evaluating displacement response based on FDA and fragility Imperial 0 60, 100, 170, 235, 270, 310
analysis Valley 10 60, 100, 170, 235, 310, 335
30 60, 100, 170, 235, 270, 310, 335,
370
Loma Prieta 0 60, 100, 170, 235, 270
10 60, 100, 170, 235, 335
30 60, 100, 170, 235, 270, 335, 370
Chi-Chi 0 60, 100, 170, 235, 335
10 60, 100, 170, 235, 335
30 60, 100, 170, 235, 335
Kobe 0 60, 100, 170, 235, 270, 310
10 60, 100, 170, 235, 335
30 60, 100, 170, 235, 270, 335
Northridge 0 60, 100, 170, 235, 270, 310
10 60, 100, 170, 235, 335
30 60, 100, 170, 235, 270, 335, 370
Hollister 0 60, 100, 170, 235, 270, 310
10 60, 100, 170, 235, 335
30 60, 100, 170, 235, 270, 335, 370
Friuli 0 60, 100, 170, 235, 270, 310
10 60, 100, 170, 235, 335
30 60, 100, 170, 235, 270, 335, 370

ure probability of the retaining walls using fragility analy- represent the soil used by Agusti and Sitar (2013) (i.e.,
sis. The failure criterion for the damage of retaining wall low plasticity lean clay). The unit weight, strength and elas-
was selected based on previous literature (Huang et al., tic properties of the soil obtained by Agusti and Sitar
2009; Wu and Prakash, 2001). (2013) using laboratory tests were used for Analysis Group
A. Agusti and Sitar (2013) conducted triaxial tests for
3. Numerical modeling procedure determining strength properties and they used shear wave
velocity tests to determine the small strain shear modulus
3.1. Model geometry on the samples of the studied soil. For Analysis Group B
and C, other cohesion values (i.e., 10 kPa and 30 kPa) for
For centrifuge test simulations (i.e., Analysis Group A) backfill and foundation were also considered according to
of this study, the geometry of the prototype model consid- observations made by Kapuskar (2005).
ered by Agusti and Sitar (2013) is used. The prototype It should be noted that the foundation and backfill
model is a 6-m high cantilever retaining wall with the stem assumed to be the same material as in some field experi-
and base thickness of 35 cm. The thickness of the retaining ences backfills are from adjacent borrow pits. In this case,
wall in the Group B and C was reduced slightly to repre- one might argue that by increasing the cohesion of the soil,
sent a wall with critical factor of safety (FS) of 1.5 for slid- the dynamic thrust of backfill reduces and at the same time
ing. Therefore, for these analyses, the retaining wall the strength of foundation increases. In order to distinguish
geometry with the height of 6 m was used but with the stem the pure effect of increase in backfill cohesion, a set of pre-
and base thickness of 30 cm (see Fig. 1a). The mesh size of liminary analyses was initially conducted to explore this
each finite difference element was considered as 50 cm to effect. It was concluded that in case of having cohesionless
avoid wave distortion in the numerical modeling during versus cohesive foundation with the same backfill, the max-
dynamic analyses (Kuhlemeyer and Lysmer, 1973). imum wall displacements magnitudes would be minimally
affected. The specific trial models were two models with
foundation cohesions of 0 and 30 kPa. For both models
3.2. Constitutive model and material properties the backfill cohesion was 30 kPa. The trial models were
subjected to Imperial Valley earthquake ground motion
3.2.1. Soil with an input PGA of 0.58 g as shown in Fig. 2a. Based
The backfill and foundation materials considered in this on the considered trial models, the total retaining wall
study had dilation angle and tensile strength of zero to

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
4 S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx
!n
displacements at top of the wall are shown in Fig. 2b. It is g
observed that the maximum total wall displacement at the G ¼ Gmax 1  1 Rf  Mod 1 ð1Þ
g1f
top of the wall for the case with cohesionless foundation is
relatively close to the one with cohesive foundation. More-
In this equation named UBCHYST equation, n and Rf
over, in most cases, the backfill material is from the same
are constants and calibration parameters to define the
site that the wall is constructed at. Due to this negligible
shear modulus degradation curve. The Mod1, is a reduc-
effect, for analysis presented in this paper, the foundation
tion factor for first-time or virgin loading. The default val-
and backfill soil assumed to be same material.
ues of n, Rf, and Mod1 are suggested to be 2, 1, and 0.6 to
To represent the hysteretic behavior of soil during the
0.8, respectively by Naesgaard (2011).
earthquake loading, the UBCHYST (Naesgaard 2011),
Cyclic simple shear test models can be simulated in
an elastic perfectly plastic constitutive model with a
FLAC to identify the shear modulus and damping curves
Mohr-Coulomb (MC) yield criterion, was utilized. This
using UBCHYST model. The required model parameters
constitutive model has also been used by Mikola and
include (Rf), n, the reduction factor (Mod1), and MC
Sitar (2015). Based on UBCHYST, the secant shear modu-
strength properties of the soil (i.e., cohesion and internal
lus of a soil element is equal to the initial maximum shear
friction angle, tensile strength), elastic properties, and dila-
modulus times a reduction factor which is the function of
tion angle.
soil stress tensors. As shown in Fig. 3a, the secant shear
Consequently, the shear modulus reduction and damp-
modulus decreases within each loading cycle. In this figure,
ing curves can be compared with empirical correlations,
s is the shear stress of a soil element, cs is the shear strain,
laboratory or field test results to assure that the reduction
Gmax is the peak initial shear modulus, and G is the secant
of shear modulus is reasonable. For this study, the default
shear modulus in a given cyclic loop. According to Fig. 3b,
values of the constants and the reduction factor were cho-
for each loop, the stress ratio (g) can be defined as the ratio
sen for a series of conducted cyclic simple shear tests in
of shear stress to vertical stress (s/rv). Based on Naesgaard
FLAC to identify shear modulus reduction and damping
(2011) and Mikola and Sitar (2015), g1 is the change in
curves for three types of soils with the cohesion of 0, 10,
stress ratio since last reversal (g  gmax), gmax is the max-
and 30 kPa. The cyclic simple shear test was modeled in
imum stress ration at last reversal, and g1f is the change in
FLAC using a single element subjected to cyclic load on
stress ratio to reach failure envelope in direction of loading.
its upper and lower boundaries. The soil parameters for
The secant shear modulus is defined as Eq. (1) (Naesgaard
the UBCHYST model are provided in Table 2. Shear mod-
2011):
ulus reduction and damping curves obtained from cyclic

56 m
(c) zmin
35 m
(a)
Stem N kn
6m
Base ks
S
1.5 m
3m
13 m

(b) (d)
D=0 m force
6m Fsmax (Fnmax)

ks (kn)
1
D=6 m
13 m relative shear (normal)
displacement

100 m Fsmax(Fnmax)

Fig. 1. Details of (a) the retaining wall in this study, (b) the free-field model, (c) soil-wall interaction system, and (d) material behavior of shear (and
normal) springs (figures are not to the scale).

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx 5

high-frequency noises during the earthquake loadings. A


0.4 (a)
low amount of critical damping ratio of 0.2% within the
Aff (g)

0.2
predominant frequency of the earthquakes is sufficient for
0.0
-0.2
the noise filtering (Agusti and Sitar 2013).
-0.4
3.2.2. Retaining wall
Displacement (cm)

20 (b) CFoundation= 0 kPa


An elastic beam was used to model the cantilever retain-
CFoundation= 30 kPa
10 ing wall. Table 2 shows the material properties of the
Total

retaining wall used in Analysis Group A (centrifuge test


0
simulation), and Analysis Groups B and C. The material
-10 properties for the wall in Analysis Group A were selected
0 5 10 15 20 25 30 35 based on Agusti and Sitar (2013), in which they used a
Time (sec) laboratory-constructed wall. However, the properties of
the wall used in Analysis Group B and C were based on
Fig. 2. Time history of (a) Imperial Valley earthquake with free-field PGA Noguchi et al. (2009) to represent a conventional rein-
of 0.58 g, and (b) total displacement at top of the wall for backfill cohesion
of 30 kPa and foundation cohesions of 0 and 30 kPa.
forced concrete retaining wall in practice.

simple shear tests in FLAC are shown in Fig. 4a and b, 3.2.3. Soil-wall interaction
respectively. In terms of shear modulus reduction and The soil-wall interaction was simulated based on shear
damping curves, a close agreement is generally noted and normal springs (see Fig. 1c) representing the behavior
between the results of the cyclic simple shear test conducted of the sliding and compression/separation between soil and
in this study using FLAC and Seed and Idriss (1970) limits wall elements, respectively. The yielding criteria of the
for cohesionless soil and Vucetic and Dobry (1991) results shear and normal springs were bilinear and based on MC
for cohesive soils. However, for shear strains larger than strength properties. The moduli of elasticity for shear and
0.05%, the damping based on cyclic simple shear tests in normal springs were defined as shear stiffness (ks) and nor-
FLAC overestimates the ones obtained from lab tests. mal stiffness (kn), respectively. As shown in Fig. 1c, the
Mikola and Sitar (2015) conducted same numerical proce- strength properties of shear and normal wall-soil interac-
dure to obtain damping curves for UBCHYST model and tion springs were defined as S and N, respectively. The
observed similar overestimation in large strains. The over- interaction element behaves linearly according to the stiff-
estimation is likely attributed to the larger width of hys- ness of ks and kn for shear and normal interactions, respec-
teretic loops developed by the UBCHYST constitutive tively. However, in a case that an interaction force reaches
model (Mikola and Sitar 2015). to an ultimate capacity according to the MC strength prop-
The main damping feature of the soil material is consid- erties, the interaction spring fails. This behavior is shown in
ered by the hysteretic behavior of the UBCHYST constitu- Fig. 1d for the unit length of an interaction element. In this
tive model. However, high-frequency noises cannot be figure, the bilinear behavior is shown for both normal and
dissipated by a hysteretic damping model (Han and Hart shear interactions. The normal interaction characteristics
2006). Therefore, Rayleigh damping was used to diminish are shown inside parentheses. In the figure, Fmax n and Fmax
s

(b)
(a) Gmax G
f

max
(previous reversal)

1f 1

s v

max
(last reversal)

Fig. 3. Schematic figure of (a) shear stress degradation during cyclic simple shear loadings and (b) stress ratio (g) in shear stress versus vertical stress plot
(modified after Naesgaard (2011)).

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
6 S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx

Table 2
Soil, retaining wall, and interaction properties for Analysis Groups A, B, and C.
Category Parameter Analysis Group A Analysis Group B & C
Primary parameters of soil Bulk modulus, K (MPa) 76 76
Shear modulus, G (MPa) 38 38
Unit weight, c (KN/m3) 20 20
Internal friction angle, u (°) 30 35
Cohesion, C (kPa) 15 0, 10, 30
Dilation angle, w (°) 0 0
Tensile strength, T (kPa) 0 0
UBCHYST equation constants of soil Coefficient component in UBCHYST equation, n 2 2
Exponential component in UBCHYST equation, Rf 1 1
Reduction factor, Mod1 0.7 0.7
Rayleigh damping parameters of soil Minimum critical damping ratio (%) 0.2 0.2
Predominant frequency (Hz) 2.8 See Table 3
Retaining wall properties Bulk modulus, K (MPa) 43,000 19,400
Shear modulus, G (MPa) 32,000 14,600
Unit weight, c (KN/m3) 31.3 23.5
Soil-wall interaction properties Shear stiffness, ks (MPa) 100 2500
Normal stiffness, kn (MPa) 100 2500
Friction angle, u (°) 10 23.3
Cohesion, C (kPa) 0 0, 7, 20
Dilation angle, w (°) 0 0
Tensile strength, T (kPa) 0 0

are maximum normal and shear forces tolerated by the quiet boundary was utilized. The viscous quiet boundary
interaction element per unit length, respectively. at the bottom of the model should have full freedom of
In the centrifuge simulation modeling (i.e., Analysis motion to absorb reflected waves. Therefore, the accelera-
Group A), the shear and normal stiffness, cohesion, and tion dynamic loading input which dictates the motion of
internal friction angle values for interaction springs were nodes at the bottom of the boundary cannot be used
selected based on Agusti and Sitar (2013). They conducted (Mejia and Dawson 2006). The alternative loading scenario
back analysis using FLAC to identify the appropriate inter- is to apply the shear stress history of the earthquake to the
action properties which represent the interface characteris- bottom of the model. The shear stress history of an earth-
tics of their experimental models. These values were quake is determined based on Eq. (3) and is a function of
assigned to Analysis Group A. However, for the numerical soil density (q), shear wave velocity of the medium (Cs),
studies conducted in Analysis Groups B and C, the shear and earthquake horizontal velocity time history (vs).
and normal stiffness were assigned using Eq. (2) (Marino Although shear stress history was used as the input loading
et al., 2016): type at the bottom boundary, the equivalent PGA of the
input load is provided for each case.
K þ 1:3G
k n ¼ k s ¼ 10 ð2Þ
Dzmin
s ¼ 2qC s vs ð3Þ
where K is bulk and G is the shear modulus of the adjacent
soil, and Dzmin is the smallest size of the adjacent soil ele- It is important to note that for numerical simulation of
ment as shown in Fig. 1c. Two-third of cohesion and inter- centrifuge tests, the loading and bottom boundary condi-
nal friction angle of backfill materials was used for strength tion mentioned above provide a more representative wall
properties of interface elements (Zamiran et al., 2016). behavior than displacement-controlled loading with a rigid
Table 2 provides the mentioned stiffness and strength prop- boundary at the bottom (Ilankatharan and Kutter, 2010).
erties of interface springs for all analyses of this study. Furthermore, as the prototype model of the centrifuge test
is simulated in the numerical analysis, the utilized loading
3.3. Dynamic considerations and earthquake loading and bottom conditions are more applicable than
displacement-controlled loading with the rigid bottom
To avoid unwanted reflection of waves into the model boundary.
from the boundaries, ‘‘free-field” lateral boundaries were It is also noteworthy that before calculating shear stress
used for the numerical study. The bottom boundary, which history using an acceleration input time history, it is neces-
the dynamic input motion is introduced to, should be able sary that raw earthquake accelerations be baseline cor-
to absorb the reflected waves from the surface. To fulfill rected and frequency filtered. The baseline correction
these requirements, a viscous boundary setting named should be conducted to assure that integration of accelera-

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx 7

1.0 farthest distance from lateral boundaries to avoid turbu-


(a) lences due to the boundary conditions.
0.8
3.4.2. Acceleration
0.6 The free-field PGA of an earthquake represents the
G/Gmax

input motion intensity. Therefore, the free-field model


0.4 shown in Fig. 1b was used to evaluate free-field accelera-
C=30 kPa
C=10 kPa tion time history and free-field PGA. Based on preliminary
0.2 C=0 kPa analyses, the acceleration output from the free-field model
Seed & Idriss Limits still had some noises due to refraction and boundary effects
Vucetic & Dobry Limits
0.0 in contrast to the velocity output (Boore and Bommer,
0.001 0.01 0.1 1 10 2005). The noise level of acceleration time history is espe-
Shear Strain (%)
30
cially high for a more severe input motion with high AF.
(b) Therefore, the velocity time history was used instead of
25 acceleration per Anderson et al. (2008). The velocity time
history of the seismic load was extracted at the top of the
Damping (%)

20 free-field model at the surface (see Fig. 1b) and the acceler-
ation time history was obtained by calculating the deriva-
15
tive of the velocity output and is called modified
10 acceleration.
As an example, Fig. 5 shows the free-field response of
5 the numerical model for Kobe earthquake with AF of
290%. In this figure, PGVff is the free-field Peak Ground
0 Velocity (PGV), PGAff-d is the direct free-field PGA
0.001 0.01 0.1
Shear Strain (%) obtained from the history monitoring of the acceleration
at the ground surface, and PGAff is the modified free-
Fig. 4. Cyclic simple shear test results for three types of soil used in this field PGA based on the aforementioned methodology here-
study and the comparison with empirical correlations for (a) shear
degradation curves and (b) damping curves.
after called free-field PGA. As it can be seen in the figure,
direct free-field PGA fluctuates intensely along the surface
of the free-field model. The level of fluctuation is increased
tion values through the time reaches to zero (Zamiran adjacent to the boundary of the model. However, the mod-
et al., 2012). Moreover, the frequency filtration should be ified free-field PGA approximately remains constant along
carried out to cutoff low and high frequency noises the model surface. At the middle of the free-field model, the
(Douglas and Boore, 2011). The Seismosignal program PGAff-d is about 1.3–2 times of the PGAff showing the level
was used to conduct baseline correction and frequency fil- of the noises in the direct acceleration output. At the
tering (Seismosoft, 2016). For frequency filteration, the fre- boundary of the model, this ratio increases to 2.3 demon-
quency contents of input accelerations lower than 0.25 Hz strating the severe raise on the noise level due to the bound-
and higher than 15 Hz were filtered using Seismosignal ary condition.
(Agusti and Sitar, 2013).
The earthquake events used for different Analysis 4. Numerical modeling verification
Groups of this study are provided in Table 3 with their his-
torical information and motion characteristics. The signif- The centrifuge model consisted of a 17-cm high retain-
icant duration of the earthquake shown in this table is the ing wall representing a prototype wall with a height of 6
time of the earthquake with 5% Arias Intensity (Is) to the m (Agusti and Sitar 2013). The geometry of the prototype
time of the earthquake with 95% of Arias Intensity model is shown in Fig. 1a. The results of site response anal-
(Trifunac and Brady, 1975). yses including free-field displacement, velocity, and modi-
fied acceleration for fully dynamic analysis of prototype
3.4. Free-field analyses model are compared with centrifuge test results in Fig. 6
for Loma Prieta 1989 earthquake. As shown in Fig. 6, a
3.4.1. Displacement satisfactory agreement for free-field displacement, velocity,
In order to evaluate the RWD response of the wall, the and accelerations time history is observed between numer-
ground displacement, when no wall is present, is needed. ical results and centrifuge output at the model surface.
Therefore, a site response analysis was conducted to deter- Acceleration and velocity are approximately in the same
mine the ground displacement of the system in the free-field phase for both numerical and centrifuge results. Also, the
condition. The geometry of the free-field model in this displacement time histories for FDA and centrifuge analy-
study is shown in Fig. 1b. The ground displacements were ses have less than one-second phase difference. The differ-
extracted at the middle section of the model which has the ence between FDA and centrifuge results for the majority

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
8 S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx

of the free-field peak values of acceleration and velocity is quake was selected with the AF of 335%. The higher level
less than 5% except in about four to five peaks where the of AF was chosen to identify the prediction of each method
difference reaches up to 20%. This can be attributed to in the worst-case scenario with the most severe ground
the level of noise in the numerical model and/or inaccuracy shaking. In weaker earthquakes, the deformations
in the data acquisition in the centrifuge tests. obtained from empirical correlations are negligible. The
The maximum RWD at the top of the wall was also results of numerical analyses based on Imperial Valley
reported in Agusti and Sitar (2013) using Kobe input earthquake with AF of 335% in the cohesionless backfill
motion with different AFs. The maximum RWD in various are shown in Fig. 8. Fig. 8a shows the free-field accelera-
AFs obtained from numerical and centrifuge tests are tion time history recorded in the free-field model during
shown in Fig. 7.The results indicate a reasonable agree- the earthquake excitation. Fig. 8b shows the total wall dis-
ment between the two methods specifically for all free- placement at the top and bottom of the stem section as well
field PGA except between 0.42 and 0.46 g. This can be as the free-field displacement time history at depths of zero
due to the inaccuracy of the data acquisition or boundary and 6 m. The RWD at the top and bottom of the stem sec-
condition problems or the higher level of energy damping tion was calculated based on the mentioned method in Sec-
in the experimental study compared to numerical simula- tion 3, and the results are shown in Fig. 8c. The major
tions for free-field PGAs of 0.42 g and 0.46 g. It is also deformation of the system occurred approximately about
observed that the correlation of maximum RWD at the the 10th second during the most intense shaking.
top and the free-field PGA is nonlinear. For the same acceleration time history in the free-field
model, the estimated wall deformation was computed using
Newmark sliding block method (Newmark 1965). Based on
5. Comparison of numerical analysis with analytical and
the Newmark method, the RWD can be calculated by dou-
empirical correlations
ble integration of the free-field acceleration values above
the yielding acceleration (ky). The yielding acceleration is
5.1. Comparison with Newmark sliding block method
defined as the acceleration which the retaining wall is trig-
gered to experience movement (Newmark 1965). Based on
For comparison between the numerical results using
Richards and Elms (1979), this can be assumed as an accel-
FDA and analytical correlations, Imperial Valley earth-

Table 3
Earthquake events and their characteristics used in different Analysis Groups of this study.
Analysis Group A B, C C C A, C C C C
Earthquake Loma Imperial Loma Chi-Chi Kobe Northridge Hollister Friuli
Prieta Valley Prieta
Year 1989 1979 1989 1999 1995 1994 1961 1976
Country USA USA USA Taiwan Japan USA USA Italy
Station SC-1 USGS 5115 CDMG TCU 045 CUE 90 090 CDMG USGS TOLMEZZO
24278 1028 000
Predominant frequency 2.8 7.1 3.3 2.9 3.1 3.8 2.5 3.8
(Hz)
Total time of the event 25.0 39.5 39.9 52.8 40.9 39.9 39.9 36.3
(s)
Significant duration, (s) 12.0 9.1 12.0 13.5 13.2 9.3 17.1 4.1
Arias Intensity, Ia (cm/s) 149.3* 80.6** 60.7** 42.4** 143.7** 46.3** 56.1** 89.1**
*
Arias Intensity is reported for the input motion with AF of 170%.
**
Arias Intensity is reported for the input motion with AF of 100%.

1.5 1.5
A B
PGA
1.2 ff-d 1.2
PGV
PGV (m/sec)
PGA (g)

ff
0.9 0.9

0.6 0.6

0.3 PGA 0.3


ff
A B
0.0 0.0
0 5 10 15 20 25 30 35 40 45 50
Distance (m)

Fig. 5. Free-field PGA values for direct and modified methodology and free-field PGV, (modified after Osouli and Zamiran (2017)).

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx 9

mate RWD based on Newmark analysis is approximately


6 (a)
Displacement

50 cm which is nearly 40% and 60% smaller than the max-


3
imum RWD obtained from FDA for top and bottom of
(cm)

0
the stem section, respectively. It is noteworthy that New-
-3
mark sliding block method and empirical correlations
-6
developed based on this method only consider the sliding
40 (b) deformation of the retaining wall and they neglect the rota-
(cm/sec)
Velocity

20
0 0.50 (a) ky=0.18g ky=0.13g
-20 0.25

Aff (g)
-40 0.00
0.4 (c)
-0.25
Acceleration

0.2
-0.50
(g)

Total Displacement (cm)


0.0
-0.2 Centrifuge (b) Wall, Top
100
-0.4 FDA
0 5 10 15 20 25 50 Wall, Bottom
Time (sec)
0
Fig. 6. Ground motion response based on the centrifuge and FDA during Free-field Top & Bottom
Loma Prieta 1989 earthquake with AF = 170% (a) Free-field displace- 150
ment, (b) Free-field velocity, and (c) Free-field acceleration. (c) ky=0.18g
RWD (cm)
100
Max. RWD at Wall Top (cm)

20 FDA 50
Centrifuge
16 0
150
12 (d) ky=0.13g
RWD (cm)

8 100
4
50 FDA, Top NCHRP
FDA, Bottom NCHRPx2
0
Newmark
0.1 0.2 0.3 0.4 0.5 0.6
0
PGAff (g) 0 10 20 30 40
Time (sec)
Fig. 7. Comparison of maximum RWD at the top of the wall between the
centrifuge test and FDA during Kobe 1995 earthquake. Fig. 8. Motion response of the retaining wall with cohesionless backfill
during Imperial Valley earthquake with free-field PGA of 0.58 g (a) Free-
eration at which the FS against the global stability of the field acceleration, (b) Ground motion and total displacement history at
retaining wall is 1.0 (Richards and Elms 1979). A top and bottom of the wall, (c) Comparison between FDA and empirical
pseudo-static approach can be used to determine the yield correlations with ky of 0.18 g, and (d) Comparison between FDA and
empirical correlations with ky of 0.13 g.
acceleration (Anderson et al., 2008; Masini et al., 2015). In
this study, the FS analysis was performed in FLAC using
strength reduction method per Dawson et al. (1999) and 0.7
Griffiths and Lane (1999). The strength reduction method ky, C=30 kPa=0.65g
0.6
is based on obtaining FS of the system by reducing the
strength of the soil in iterative steps of the analyses until 0.5
the system ‘‘fails”. Fig. 9 shows the results of various
kh (g)

0.4 ky, C=10 kPa=0.39g


pseudo-static trials for evaluating yield acceleration for a
retaining wall with different backfill cohesions. Based on 0.3
this figure, the yield accelerations for a retaining wall with C=30 kPa
0.2 C=10 kPa ky, C=0 kPa=0.18g
backfill cohesions of 0, 10, and 30 kPa are 0.18 g, 0.39 g,
C=0 kPa
and 0.65 g, respectively. 0.1
0.7 0.8 0.9 1.0
Newmark sliding block analysis was conducted for the
FS
free-field acceleration shown in Fig. 8a considering the
yield acceleration of 0.18 g to evaluate the RWD for a wall Fig. 9. Results of pseudo-static analyses to identify yield acceleration for
with zero backfill cohesion. As shown in Fig. 8c, the ulti- retaining walls with different backfill cohesion. *The maximum sliding
RWD drops to zero when yield acceleration is larger than free-field PGA.

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
10 S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx

tional displacement of the wall. However, the RWD at the wall might be due to the fact that the rotational displace-
top in FDA analysis includes both rotational and sliding ment of the wall is neglected in NCHRP method.
movements of the wall. Hence, according to Green The prediction of R&E (Richards and Elms 1979) in the
(2016), it is more appropriate to compare the RWD at wall studied case was computed as 196 cm which is much higher
bottom in FDA representing sliding deformation with the than FDA results at the bottom of the wall, Newmark
Newmark method results. As it can be seen, the Newmark results, and NCHRP method. This shows that R&E
method underestimates the seismic deformation, particu- method can overestimate the seismic wall displacement
larly the deformations at the top of the retaining wall. specifically in earthquakes with higher intensities, which
was also noted by Whitman and Liao (1985).
5.2. Comparison with empirical correlations developed based There are some studies that suggest the yield accelera-
on Newmark sliding method tion is determined when FS for global stability is slightly
above one (Green et al., 2008; Green and Ebeling, 2003).
There are also empirical correlations such as R&E It is noteworthy that Newmark sliding block method and
(Richards and Elms 1979) and Anderson et al. (2008) that related empirical methods are very sensitive to the selection
provide maximum RWD estimation based on Newmark of yielding criterion. Also, the FS may vary depending on
sliding method using PGA, PGV, and yield acceleration. the utilized FS calculation method. For example, Fig. 8d
It is worth noting that using these methods result in mini- shows the updated RWDs based on Newmark method
mal displacements for cases that backfill has cohesiveness; and NCHRP approach considering yield acceleration of
however, they are commonly used for cohesionless backfill 0.13 g which results in the FS of 1.1 in the pseudo-static
materials. As shown in Fig. 9, the reason for minimal dis- analysis. As it can be seen from Fig. 8c and d, the RWD
placement for the cohesive backfills from these Newmark- displacements calculated by Newmark or related empirical
based methods is their large yield acceleration. The large methods are sensitive to calculated yield acceleration.
yield acceleration leads to have limited duration of the Regardless of yield acceleration criterion, the mentioned
earthquake that acceleration is above the yield and can empirical methods underestimate the RWD predictions at
contribute in wall sliding per these methods. Similar to the top of the wall.
Newmark method, these empirical correlations can evalu- A sensitivity analysis was conducted to demonstrate the
ate maximum RWD due to sliding. Anderson et al. rate of variations of maximum sliding RWD resulting from
(2008) recommended Eq. (4) which is known as NCHRP different yield accelerations and in various backfill condi-
equation and Richards and Elms (1979) suggested Eq. (5). tions. The results of the sensitivity analysis are shown in
Fig. 10. The maximum RWD at the wall bottom obtained
   from FDA are compared to NCHRP and R&E predictions
ky ky of maximum RWD due to sliding based on different yield
log ðDr Þ ¼ 1:5  0:8 log þ 3:3 log 1 
PGA PGA accelerations. The prediction of the maximum RWD based
 0:8 log ðPGAÞ þ 1:6 logðPGV Þ ð4Þ on R&E and NCHRP in case of safety factor of one is
shown with circle symbols. As can be seen in Fig. 10c, this
 4
PGV 2 k y prediction reaches to zero in higher yield accelerations for
Dr ¼ 0:087 ð5Þ both NCHRP and R&E methods. The reason is that the
PGA PGA
yield acceleration for a safety factor of one for a model
where Dr is the maximum RWD due to sliding, and ky is with 30 kPa backfill cohesion is higher than the free-field
the yield acceleration. In Eq. (4), the parameters of dis- PGA. In such cases, the earthquake acceleration never
placement, velocity, and acceleration should be in inch, reaches to yield acceleration and hence, the displacement
inch/s, and inch/sec2, respectively. In Eq. (5), the parame- is zero. The same case occurred when yield acceleration
ters of velocity and acceleration should be in a consistent reaches to free-field PGA. In these cases, which is more vis-
unit system with the desired displacement unit. It should ible in Fig. 10b and c, the maximum RWD drops abruptly
be noted that Anderson et al. (2008) suggested that for a to zero.
higher confidence level or in a more conservative design, The variation of predicted displacements based on R&E
the resulted displacement from Eq. (4), i.e., NCHRP method is more sensitive to the changes of yield accelera-
approach, can be multiplied by a factor of two. tion in all cases compared to NCHRP. According to
Fig. 8c shows the predicted maximum RWD due to slid- Fig. 10a, for cohesionless backfill material for a yield accel-
ing based on NCHRP approach for yield acceleration of eration representing FS of 1.0, NCHRP and R&E methods
0.18 g. As can be seen in this figure, the NCHRP’s RWD provide underestimation and overestimation of maximum
prediction for sliding is less than the maximum RWD at RWD at the wall bottom, respectively. However, in the
the wall bottom obtained from FDA, which represents cases with backfill cohesion of 10 kPa and 30 kPa for yield
the sliding movement of the wall. There is a close agree- acceleration representing FS of 1.0, both empirical correla-
ment between NCHRP RWD predictions that are factored tions underestimated the maximum RWD compared to
by two and FDA results at wall bottom. The larger differ- FDA predictions at the wall bottom. The underestimation
ence between NCHRP and FDA results at the top of the for backfill cohesion of 30 kPa is more noticeable than the

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx 11

250 As can be seen in Fig. 11, the displacement outputs for


(a) C=0 kPa
R&E- Sliding different earthquakes in lower free-field PGAs (i.e., less
200
Max. RWD (cm)

NCHRP- Sliding
FDA-Bottom than 0.3 g) are less scattered than higher free-field PGAs.
150 Prediction for FS=1 Moreover, based on the regression trends shown in the fig-
100
ure, it can be concluded that maximum RWD of the cases
with the cohesion of 30 and 10 kPa are nearly 8 and 25 per-
50 * cent of the RWDs for cases with zero cohesion. The dis-
0 placement response of the wall in Chi-Chi earthquake
30 was relatively lower than other earthquakes. The reason
(b) C=10 kPa is that the Arias Intensity of Chi-Chi input motion was
Max. RWD (cm)

the lowest compared to other earthquakes (see Table 3).


20 Moreover, for Chi-Chi earthquake, the accelerations which
are in the range of PGA values occur in short period. How-
10 ever, in other earthquakes studied herein, the high acceler-
* ation occurs in longer periods.
For each retaining wall with certain backfill cohesion,
0
nonlinear regression analysis was conducted to estimate
12 (c) C=30 kPa the correlation between the free-field PGA and maximum
Max. RWD (cm)

RWD. The regression curves for each case with different


9 backfill cohesions are shown in Fig. 11. The resulted equa-
tions for the regression curves are provided in Eq. (6).
6
*
RWDMax
3 ½% ¼ ð0:012C þ 0:363Þeð0:05Cþ6:82ÞPGAff ð6Þ
H
0 where RWDMax/H is the normalized maximum RWD at
0.1 0.2 0.3 0.4 0.5 0.6 0.7
ky (g) the top in percent, C is the backfill cohesion in kPa, and
PGAff is the free-field PGA in g (gravitational accelera-
Fig. 10. Sensitivity analysis of yield acceleration in empirical correlations
tion). According to Fig. 11, the amount of backfill cohesion
of sliding displacement of the wall for Imperial Valley earthquake with AF
of 310% for backfill cohesions of (a) 0 kPa, (b) 10 kPa, and (c) 30 kPa. can considerably decrease the wall deformation and reduce
*The maximum sliding RWD drops to zero when yield acceleration is the level of damage in moderate to severe earthquakes that
larger than free-field PGA. have free-field PGA of more than 0.3 g.
The regression lines shown in Fig. 11 are replotted to
correlate the maximum RWD at the wall top with the
case with 10 kPa backfill cohesion. For the backfill with the free-field PGA normalized by the yield acceleration of
higher cohesion of 30 kPa, the yield acceleration is higher. 0.18 g, 0.39 g, and 0.65 g for backfills with cohesion of 0,
Therefore, in very limited and shorter periods, the earth- 10, and 30 kPa (see Fig. 12). The data of maximum
quake acceleration exceeds the yield limit and causes wall RWD at the wall top for all three backfill types would be
displacement. This results in negligible wall deformation reduced into almost similar trends as shown in Fig. 12.
and underestimation of sliding maximum RWD. In this figure, the regression lines show an overlap and sim-
ilar trend for cohesive backfill materials (i.e., cohesion of 10
6. Displacement study based on seven different earthquakes and 30 kPa) and they are different than the one for cohe-
using FDA sionless material. Therefore, as an alternative to Eq. (6),
which is based on cohesion level of the backfill, Eq. (7)
Different earthquakes due to their different characteris- can be developed to correlate the maximum RWD to
tics including acceleration fluctuations, Arias Intensity, sig- PGAff/ky. In this equation, c1 and c2 are the constants
nificant duration, and frequency content can influence and equal to 0.064 and 3.124, respectively for cohesive
motion response of a retaining wall system. To evaluate backfills. For cohesionless backfill, c1 and c2 are 0.408
the dispersion of motion response of the retaining wall with and 1.206, respectively.
different backfill cohesions, the retaining wall models were
subjected to seven different events with various AFs (see
Table 3). The results of numerical analyses conducted RWDMax c2 PGAff

½% ¼ c1 e ky ð7Þ
based on these seismic events with different AFs for backfill H
cohesion of 0, 10, and 30 kPa are shown in Fig. 11. The
results are shown for the free-field PGA range of 0.05 g 7. Fragility analysis
to 0.65 g. As the most significant displacement occurs at
the top of the retaining wall, only the results of the maxi- As discussed in the previous section, the level of uncer-
mum RWD at the wall top are shown. tainty for evaluation of wall deformation during a seismic

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
12 S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx

80 (a) (b) (c) Loma Prieta


C=0 kPa C=10 kPa C=30 kPa Chi-Chi
70
Max. RWD at Wall Top (cm)

Kobe
60 Northridge
Imperial Valley
50 Hollister
40 Friuli

30

20

10

0
0.1 0.2 0.3 0.4 0.5 0.6 0.1 0.2 0.3 0.4 0.5 0.6 0.1 0.2 0.3 0.4 0.5 0.6
PGAff (g) PGAff (g) PGAff (g)

Fig. 11. Maximum RWD at the top of the wall based on FDA for different free-field PGAs for backfill cohesions of (a) 0 kPa, (b) 10 kPa, and (c) 30 kPa.

choose the appropriate level of damage considering the


Max. RWD at Wall Top (cm)

80 C=0 kPa
C=10 kPa importance of a retaining wall and its adjacent
60 C=30 kPa infrastructure.
The fragility analysis helps to identify the probability of
40 different levels of damage due to an induced earthquake
20
intensity. As mentioned before, the free-field PGA is
selected as the representation of the seismic event intensity.
0 The maximum RWD at the wall top normalized by the wall
0.0 0.5 1.0 1.5 2.0 2.5 3.0 height for free-field PGA values of zero to 0.85 g is deter-
PGAff/ky mined from displacement-acceleration plots shown in
Fig. 12. Maximum RWD at the top of the wall based on FDA versus free-
Fig. 11 for each earthquake. Therefore, for any selected
field PGA normalized by yield acceleration. free-field PGA, seven different normalized displacements
resulted from seven earthquake events were determined.
The normalized displacement variables can be considered
event is high. An effective approach for characterization of as a lognormal distribution data with the probability den-
motion responses with a high level of uncertainty for a sity function (PDF) shown as Eq. (8) (Argyroudis et al.,
retaining wall during seismic condition is to conduct fragi- 2013; Chiou et al., 2011).
lity analysis. The initial step to carry out the fragility anal-
ysis for the seismic response of a structure such as a "  2 #
retaining wall is to define the level of damages in the seis- 1 ln X  kÞ
f X ðX Þ ¼ pffiffiffiffiffiffi exp 0:5 06X <1
mic conditions. Different investigators have suggested crite- 2p1X 1
ria for permissible and seismic failure displacements of
retaining walls. Most of these criteria are based on engi- ð8Þ
neering judgment, historical retaining wall performances where fX(X) is probability density function, X is the ran-
during earthquakes, and experimental studies using shake dom variable of normalized maximum RWD at the wall
table or centrifuge tests (Huang et al., 2009). top, and k and 1 are calculated using Eqs. (9) and (10).
Conservative permissible displacement is considered as
0.02 of the wall height per Wu and Prakash (2001) and k ¼ ln l  0:512 ð9Þ
Huang et al. (2009). The failure displacement can vary "  2 #
from 0.05 to 0.1 of the wall height (Huang et al., 2009; r
1 ¼ ln 1 þ
2
ð10Þ
Wu and Prakash, 2001). Therefore, 0.08 of the wall height l
was considered as failure displacement, and the median of
permissible and failure displacements is assumed as the where l and r are the mean value and standard deviation
moderate level of wall deformation. Consequently, Zone of seven different normalized maximum RWD at the wall
I, II, and III shown in Fig. 13, represent the wall displace- top as a random variable for each free-field PGA data,
ments less than 0.02H, 0.05H, and 0.08H, respectively. respectively. The conditional probability that the normal-
These levels of displacements are interpreted as the level ized maximum RWD at the wall top (i.e. X) is higher than
of minor, moderate, and severe damage. Zone IV repre- the considered normalized displacement or level of damage
sents wall displacements of more than 0.08H and failure. (xi) in a given free-field PGA value is determined using Eq.
It is noteworthy that it is up to a designer’s decision to (11) (Chiou et al., 2011).

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx 13

Conservative Permissible
Damage Boundaries Moderate displacement Failure displacement
displacement
Wu and Prakash 0.02H 0.1H
Huang et al. 0.02H 0.05H
This Study 0.02H 0.05H 0.08H

Damage zones Minor Zone Failure Zone


Moderate Zone (II) Severe Zone (III)
(I) (IV)

Fig. 13. Criteria for level of damage of retaining walls based on literature and this study.
 
  ln ðxi Þ  k the free-field PGA of 0.75 g and 0.85 g in the case of a
P X > xi jPGAff ¼ 1  ; ð11Þ
1 backfill with 10 kPa and 30 kPa cohesion, respectively.
Moreover, for an earthquake with free-field PGA of 0.35
where P is probability, and ; is the cumulative distribution g, which is defined as an earthquake with moderate poten-
function (CDF) of the standard normal distribution of the tial damage (Evans 2003), the probability of moderate
variables. Using Eq. (11), the probability that the normal- damage for walls with backfill cohesion of 0, 10, and 30
ized maximum RWD at the wall top reaches to each level kPa are 30, 0, and 0%, respectively. However, for earth-
of displacement boundaries can be determined as shown quakes with free-field PGA of 0.65 g, defined as earth-
in Fig. 14 for three different backfill cohesions. As noted quakes with high potential of damage (Evans 2003), the
earlier, based on the site circumstances and the acceptable probability of moderate damage for walls with backfill
risk, the design damage zone can be selected in any zones cohesion of 0, 10, and 30 kPa is 100, 65, and 5 percent,
from Zone I to Zone III. respectively. The results in Fig. 14 are based on seven dif-
As can be seen in Fig. 14, the cohesiveness of the backfill ferent earthquakes; therefore, they can provide a better
material has a critical role in motion response of the understanding regarding the behavior of retaining walls
retaining wall. The probability of the failure would in various earthquakes.
decrease with increase in the cohesiveness of the backfill. In addition to normalization of the wall height, the nor-
For example, a retaining wall with the cohesionless backfill malization of free-field PGA by the yield acceleration of
has more than 50% likelihood of failure beyond the free- the cases were also considered. The fragility curves in
field PGA of 0.45 g while this likelihood is excepted at
1.0
(a) (d)
Probability of Damage

C=0 kPa
0.8 C=0 kPa

0.6 I II
I II III III
IV
0.4 IV
0.2

0.0
1.0
(b) (e)
Probability of Damage Probability of Damage

C=10 kPa Maximum PGAff/ky


0.8 C=10 kPa considered in this

0.6 I II III study


I II III
0.4

0.2 IV IV
0.0
1.0 (f)
(c)
C=30 kPa
0.8 C=30 kPa
Permissible Displacement
0.6 Moderate Displacement I I Maximum PGAff/ky
Failure Displacement
0.4 II III considered in this study
II
0.2 III IV

0.0 IV
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
PGAff (g) PGAff/ky

Fig. 14. Results of fragility analyses for a retaining wall with backfill cohesions of (a) 0 kPa, (b) 10 kPa, and (c) 30 kPa versus free-field PGA; and (d) 0
kPa, (e) 10 kPa, and (f) 30 kPa versus normalized free-field PGA by yield accelartion.

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
14 S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx

respect to the PGAff/ky are shown in Fig. 14d–f for backfill the probability of damage levels for their designed retain-
cohesions of 0, 10, and 30 kPa, respectively. As shown in ing wall based on their design earthquake acceleration
the figure, this normalization reduces the data into approx- and acceptable risk. According to the assumption that
imately similar fragility curves for all three cases. It is the backfill acts as a rigid system during the earthquake,
worth noting that the PGAff ranges considered in this the results of the study can be used for retaining walls up
study, which covers most common ranges in practice, to the approximate height of 9 m.
would result in maximum PGAff/ky of approximately 2.2
and 1.35 for backfill cohesions of 10 kPa, and 30 kPa, Acknowledgements
respectively. However, the general trend of the normalized
fragility curves can be seen in Fig. 14d, which is for cohe- The authors would like to acknowledge the Itasca Con-
sionless backfill. The fragility curves in normalized format sulting Group, Inc. for providing the license of FLAC
(i.e., Fig. 14d–f) is useful tool for determining the probabil- based on Itasca Education Partnership program and Dr.
ity of damage to backfills with other cohesion values that Zorica Radaković-Guzina from Itasca Consulting Group
were not used in this study. for her technical supports. Also, authors would like to
Moreover, as the failure criterion selected for this study appreciate Dr. Russell Green from Virginia Polytechnics
was based on the normalized displacement of the wall Institute and State University, Dr. Abouzar Sadrekarimi
regarding the wall height, the provided results would be from University of Western Ontario, Dr. Prabir Kolay
extendable to retaining walls with other heights. This and Dr. Sanjeev Kumar from Southern Illinois University
extension is also reliable for the results obtained from and Dr. Shamsher Prakash from University of Missouri
Eq. (6). Extending the normalized results to other heights Science and Technology for providing technical supports.
for different cantilever retaining walls is based on the fact
that the backfill of the retaining wall acts as a rigid body References
during an earthquake excitation (Anderson et al., 2008).
However, according to Anderson et al. (2008), the rigid AASHTO, 2007. ASSHTO LRFD Bridge Design Specifications, 4th ed.
body assumption is only reliable for retaining walls up to The American Association of State Highway and Transportation
Officials (AASHTO).
the approximate height of 9 m (30 ft). For the retaining Agusti, G.C., Sitar, N., 2013. Seismic Earth Pressures on Retaining
walls with higher heights, this assumption might not be Structures in Cohesive Soils. California Department of
valid. Transportation.
Anderson, D.G., Martin, G.R., Lam, I. (Po), Wang, J. N. (Joe), 2008.
Seismic Analysis and Design of Retaining Walls, Buried Structures,
8. Conclusion
Slopes, and Embankments. National Cooperative Highway Research
Program (NCHRP), Transportation Research Board.
In this study, the seismic, horizontal displacement of a Argyroudis, S., Kaynia, A.M., Pitilakis, K., 2013. Development of
cantilever retaining wall was evaluated using fully dynamic fragility functions for geotechnical constructions: application to
analysis. Seven different earthquake events with different cantilever retaining walls. Soil Dyn. Earthquake Eng. 50, 106–116.
free-field peak ground accelerations (PGAs) and three Baker, J.W., 2015. Efficient analytical fragility function fitting using
dynamic structural analysis. Earthquake Spectra 31 (1), 579–599.
backfill cohesions of 0, 10, and 30 kPa were considered. Biondi, G., Cascone, E., Maugeri, M., 2014. Displacement versus pseudo-
A deformation prediction model based on the earthquake static evaluation of the seismic performance of sliding retaining walls.
input acceleration and backfill cohesions was presented. Bull Earthq. Eng. 12 (3), 1239–1267.
A fragility analysis was performed to determine the proba- Boore, D.M., Bommer, J.J., 2005. Processing of strong-motion accelero-
bility of damage to a retaining wall based on various free- grams: needs, options and consequences. Soil Dyn. Earthquake Eng.
25 (2), 93–115.
field PGA levels for different backfill cohesions. Chiou, J.-S., Chiang, C.-H., Yang, H.-H., Hsu, S.-Y., 2011. Developing
In comparison to FDA method, the empirical correla- fragility curves for a pile-supported wharf. Soil Dyn. Earthquake Eng.
tions underestimate the maximum RWD because they 31 (5–6), 830–840.
neglect rotation of the wall. Based on the regression analy- Conti, R., Viggiani, G.M.B., Cavallo, S., 2013. A two-rigid block model
for sliding gravity retaining walls. Soil Dyn. Earthquake Eng. 55, 33–
sis, the maximum RWD for models with backfill cohesions
43.
of 30 and 10 kPa were approximately 8 and 25 percent of Corigliano, M., Lai, C.G., Pasquali, R., 2011. Displacement-based seismic
the RWDs for cohesionless backfill, respectively. The design of gravity walls using double-support excitation. In: Proceed-
results of numerical modeling and fragility analyses ings of the 8th International Conference on Structural Dynamics.
showed that with an increase in backfill cohesion, the Dawson, E.M., Roth, W.H., Drescher, A., 1999. Slope stability analysis
deformation of the wall reduces and hence, the probability by strength reduction. Géotechnique 49 (6), 835–840.
Douglas, J., Boore, D.M., 2011. High-frequency filtering of strong-motion
of failure of wall decreases. The failure displacement never records. Bull Earthq. Eng. 9 (2), 395–409.
occurred for backfill cohesion of 30 kPa up to the free-field Ertugrul, T., 2013. Development of Improved Guidelines for Seismic
PGA of 0.85 g. However, with a probability of 50% the Analysis and Design of Earth Retaining Structures. California
wall will experience failure displacement when the free- Department of Transportation, University of California, Los Angeles.
Evans, J.R., 2003. The SideBar Computer Program, A Seismic-Shaking
field PGA reaches to 0.47 g and 0.75 g for backfill cohe-
Intensity Meter: Users’ Manual and Software Description. United
sions of 0 and 10 kPa, respectively. Presented fragility States Geological Survey.
curves can be a very helpful tool for designers to examine

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010
S. Zamiran, A. Osouli / Soils and Foundations xxx (2018) xxx–xxx 15

Fang, Y.S., Yang, Y.C., Chen, T.J., 2003. Retaining walls damaged in the Naesgaard, E., 2011. A Hybrid Effective Stress – Total Stress Procedure
Chi-Chi earthquake. Can. Geotech. J. 40 (6), 1142–1153. for Analyzing Soil Embankments Subjected to Potential Liquefaction
Green, R.A., 2016. Technical Correspondence with Dr. Russell Green. and Flow. The University of British Columbia, Vancouver, Canada.
Green, R.A., Ebeling, R.M., 2003. Modeling the Dynamic Response of Nakamura, S., 2006. Reexamination of Mononobe-Okabe theory of
Cantilever Earth-Retaining Walls Using FLAC. A.A. Balkema Pub- gravity retaining walls using centrifuge model tests. Soils Found. 46
lishers, Sudbury, ON, Canada, pp. 333–342. (2), 135–146.
Green, R.A., Olgun, C.G., Cameron, W.I., 2008. Response and modeling Newmark, N.M., 1965. Effects of earthquakes on dams and embank-
of cantilever retaining walls subjected to seismic motions. Comput.- ments. Geotechnique 15 (2), 139–160.
Aid. Civ. Infrastruct. Eng. 23 (4), 309–322. Noguchi, T., Tomosawa, F., Nemati, K.M., Chiaia, B.M., Fantilli, A.P.,
Griffiths, D.V., Lane, P.A., 1999. Slope stability analysis by finite 2009. A practical equation for elastic modulus of concrete. ACI Struct.
elements. Géotechnique 49 (3), 387–403. J. 106 (5), 690–696.
Han, Y., Hart, R., 2006. Application of a simple hysteretic damping Osouli, A., Zamiran, S., 2017. The effect of backfill cohesion on seismic
formulation in dynamic continuum simulations. Minneapolis, response of cantilever retaining walls using fully dynamic analysis.
Minnesota Comput. Geotech. 89, 143–152.
Huang, C.-C., Wu, S.-H., Wu, H.-J., 2009. Seismic displacement criterion Prakash, S., Wu, Y., Rafnsson, E.A., 1995. On seismic design displace-
for soil retaining walls based on soil strength mobilization. J. Geotech. ments of rigid retaining walls. In: Third International Conference on
Geoenviron. Eng. 135 (1), 74–83. Recent Advances in Geotechnical Engineering and Soil Dynamics, St.
Ichii, K., 2004. Fragility curves for gravity-type quay walls based on Louis, MO, pp. 1183–l192.
effective stress analyses. In: 13th World Conference on Earthquake Rafnsson, E.A., 1991. Displacement based design of rigid retaining walls
Engineering, Vancouver, B.C., Canada. subjected to dynamic loads considering soil nonlinearity. Ph.D.
Ilankatharan, M., Kutter, B., 2010. Modeling input motion boundary Dissertation. Univ. of Missouri Rolla.
conditions for simulations of geotechnical shaking table tests. Earth- Richards, R., Elms, D., 1979. Seismic behavior of gravity retaining walls.
quake Spectra 26 (2), 349–369. J. Geotech. Eng. Div., ASCE 105 (GT4), 449–464.
Itasca, 2011. FLAC 7.0. Itasca Consulting Group Inc, Minneapolis, MN. Richards, R., Fishman, K.L., Divito, R.C., 1996. Threshold accelerations
Jafarian, Y., Miraei, M., Lashgari, A., Shakeri, P., 2014. Probabilistic for rotation or sliding of bridge abutments. J. Geotech. Eng. 122 (9),
evaluation of dynamic response of caisson quay walls in soil improved 752–759.
by fiber: a numerical study. In: Hicks, M., Brinkgreve, R., Rohe, A. Seed, H.B., Idriss, I.M., 1970. Soil Moduli and Damping Factors for
(Eds.), Numerical Methods in Geotechnical Engineering. CRC Press, Dynamic Response Analyses. University of California, Berkeley.
pp. 909–914. Seismosoft, 2016. SeismoSignal. Seismosoft Ltd., Pavia, Italy.
Kapuskar, M., 2005. Field Investigation Report for Abutment Backfill Shakya, D.A., 1987. Engineering Solutions for the Displacement of Rigid
Characterization. Earth Mechanics Inc, prepared for California Retaining Walls Subjected to Earthquake Loads. University of
Department of Transportation (Caltrans) and University of California Missouri Rolla.
San Diego, Sacramento, CA. Stamatopoulos, C.A., Velgaki, E.G., Modaressi, A., Lopez-Caballero, F.,
Ko, Y.-Y., Han, J.-Y., Chou, J.-Y., 2017. Application of close-range 2006. Seismic displacement of gravity walls by a two-body model. Bull
photogrammetry for post-failure reconnaissance of a retaining wall. Earthq. Eng. 4 (3), 295–318.
In: Hazarika, H., Kazama, M., Lee, W.F. (Eds.), Geotechnical Trifunac, M.D., Brady, A.G., 1975. A study on the duration of strong
Hazards From Large Earthquakes and Heavy Rainfalls. Springer, earthquake ground motion. Bull. Seismol. Soc. Am. 65 (3), 581–626.
Japan, Tokyo, pp. 503–512. Vucetic, M., Dobry, R., 1991. Effect of soil plasticity on cyclic response. J.
Kuhlemeyer, R.L., Lysmer, J., 1973. Finite element method accuracy for Geotech. Eng. 117 (1), 89–107.
wave propagation problems. J. Soil Mech. Found. Div. 99 (5), 421– Whitman, R.V., Liao, S., 1985. Seismic Design of Gravity Retaining
427. Walls. US Army Corps of Engineers.
Lai, S., 1998. Rigid and flexible retaining walls during Kobe earthquake. Wilson, P., Elgamal, A., 2010. Large-scale passive earth pressure load-
Latifi, N., Marto, A., Eisazadeh, A., 2016. Experimental investigations on displacement tests and numerical simulation. J. Geotech. Geoenviron.
behaviour of strip footing placed on chemically stabilised backfills and Eng. 136 (12), 1634–1643.
flexible retaining walls. Arab. J. Sci. Eng. 41 (10), 4115–4126. Wilson, P., Elgamal, A., 2015. Shake table lateral earth pressure testing
Ling, H.I., Leshchinsky, D., Chou, N.N.S., 2001. Post-earthquake with dense c-/ backfill. Soil Dyn. Earthquake Eng. 71, 13–26.
investigation on several geosynthetic-reinforced soil retaining walls Wu, Y., Prakash, S., 2001. Seismic displacements of rigid retaining walls.
and slopes during the Ji-Ji earthquake of Taiwan. Soil Dyn. Earth- In: Fourth International Conference on Recent Advances in Geotech-
quake Eng. 21 (4), 297–313. nical Earthquake Engineering and Soil Dynamics and Symposium in
Marino, G., Osouli, A., Zamiran, S., Shafii, I., 2016. Performance of a pier Honor of Professor W.D. Liam Finn, San Diego, California.
group foundation in swelling rock. Geotech. Geol. Eng. Zamiran, S., Ghojavand, H., Saba, H., 2012. Numerical analysis of soil
Masini, L., Callisto, L., Rampello, S., 2015. An interpretation of the nail walls under seismic condition in 3D form excavations. Appl.
seismic behaviour of reinforced-earth retaining structures. Géotech- Mech. Mater. 204–208, 2671–2676.
nique 65 (5), 349–358. Zamiran, S., Osouli, A., 2014. Earth pressures of clayey backfilled
Mejia, L.H., Dawson, E.M., 2006. Earthquake deconvolution for FLAC. retaining walls under seismic loading. In: XV Danube – European
In: 4th International FLAC Symposium on Numerical Modeling in Conference on Geotechnical Engineering, Vienna, Austria.
Geomechanics. Zamiran, S., Osouli, A., 2015. Seismic performance of cantilever retaining
Mikola, R.G., Candia, G., Sitar, N., 2014. Seismic earth pressures on walls with clayey backfills. American Society of Civil Engineers, pp.
retaining structures and basement walls. In: Tenth U.S. National 1439–1448.
Conference on Earthquake Engineering, Frontiers of Earthquake Zamiran, S., Osouli, A., Marino, G.G., 2016. Modeling of swelling rocks
Engineering, Anchorage, Alaska. for group pier foundation applications. In: 50th U.S. Rock Mechanics/
Mikola, R.G., Sitar, N., 2015. UBCHYST – A Total Stress Hysteretic Geomechanics Symposium, American Rock Mechanics Association,
Model. Jacobs Associates. Houston, TX.
Nadim, F., Whitman, R.V., 1984. Coupled sliding and tilting of gravity Zeng, X., Steedman, R.S., 2000. Rotating block method for seismic
retaining walls during earthquakes. In: The Eighth World Conference displacement of gravity walls. J. Geotech. Geoenviron. Eng. 126 (8),
on Earthquake Engineering, San Francisco, CA, pp. 477–484. 709–717.

Please cite this article in press as: Zamiran, S., Osouli, A., Seismic motion response and fragility analyses of cantilever retaining walls with cohesive back-
fill, Soils Found. (2018), https://doi.org/10.1016/j.sandf.2018.02.010

You might also like