This article was downloaded by: [Technical University of Crete]

On: 21 July 2010

Access details: Access Details: [subscription number 912914330]
Publisher Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 3741 Mortimer Street, London W1T 3JH, UK

Structure and Infrastructure Engineering

Publication details, including instructions for authors and subscription information:

http://www.informaworld.com/smpp/title~content=t713683556

Effects of soil non-linearity on the seismic response of restrained retaining

walls
Prodromos N. Psarropoulosa; Yiannis Tsompanakisb; George Papazafeiropoulosb
a
Department of Infrastructure Engineering, Hellenic Air-Force Academy, Greece b Division of
Mechanics, Department of Applied Sciences, Technical University of Crete, University Campus,
Chania, Greece
First published on: 21 December 2009

To cite this Article Psarropoulos, Prodromos N. , Tsompanakis, Yiannis and Papazafeiropoulos, George(2009) 'Effects of

soil non-linearity on the seismic response of restrained retaining walls', Structure and Infrastructure Engineering,, First
published on: 21 December 2009 (iFirst)
To link to this Article: DOI: 10.1080/15732470903419677
URL: http://dx.doi.org/10.1080/15732470903419677

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf
This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.

Structure and Infrastructure Engineering

2009, 112, iFirst article

Eects of soil non-linearity on the seismic response of restrained retaining walls

Prodromos N. Psarropoulosa, Yiannis Tsompanakisb* and George Papazafeiropoulosb
a

Department of Infrastructure Engineering, Hellenic Air-Force Academy, Greece; bDivision of Mechanics, Department of Applied
Sciences, Technical University of Crete, University Campus, Chania 73100, Greece

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

(Received 5 May 2009; nal version received 16 October 2009)

Retaining structures characterised by high rigidity and various kinematic constraints, such as bridge abutments and
basement walls, do not permit limit-equilibrium conditions to be developed. Therefore, according to contemporary
norms and geotechnical design practice worldwide, they are most frequently designed utilising the elasticity-based
methods, which usually lead to substantially increased dynamic earth pressures. The present paper aims to examine
how and to what extent the potentially developed non-linearity of the retained soil may aect: (a) the dynamic
distress of a rigid xed-base retaining wall, and (b) the seismic response of the retained soil layer. For this purpose, a
parametric study is conducted which is based on two-dimensional dynamic nite-element analyses using various
idealised or real seismic excitations scaled to several intensity levels. Soil non-linearity is realistically taken into
account via the commonly used equivalent-linear approach. The results of the present study demonstrate that
potential non-linearity of a soil layer retained by a rigid xed-base wall alters the soil amplication pattern behind
the wall and leads to dynamic earth pressures usually lower than those proposed by seismic norms.
Keywords: retaining walls; dynamic wall-soil interaction; soil non-linearity; amplication; earthquake-induced
pressures

1.

Introduction

Bridge abutment walls, basement walls, or harbour

quay walls are characteristic cases in which a rigid
gravity retaining wall or a exible cantilever retaining
wall is constructed. More complicated retaining wall
structures are reinforced soil walls, anchored bulkheads, or tieback walls. The compliance of each
retaining wall type depends on its structural exibility,
and the kinematic constraints imposed by other
structural elements on various locations along its
height. Experience accumulated from failures occurred
in recent earthquakes has shown that in many cases the
seismic performance of retaining walls has not met the
requirements of an operational retaining structure.
There are many examples of retaining walls (primarily
harbour quay walls) that have failed during a seismic
event (Kobe, Japan 1995 and Lefkada, Greece 2003).
Retaining wall failures can cause great economical
losses for a local region, especially when they are used
to support vital structures and infrastructures. For
instance, severe damages took place in the Kobe port
area during the 1995 earthquake. In addition, many
cases of bridge failures have been reported during
recent earthquakes due to excessive abutment displacement. Seismic vulnerability of retaining systems is
usually related to the strength degradation of saturated

*Corresponding author. Email: jt@science.tuc.gr

ISSN 1573-2479 print/ISSN 1744-8980 online
2009 Taylor & Francis
DOI: 10.1080/15732470903419677
http://www.informaworld.com

cohesionless soils in the backll and the foundation.

A very limited number of cases of high earth pressures
have been reported. Nevertheless, constrained rigid
retaining systems seem to be designed rather conservatively. In any case, as it will be shown in the sequence,
a more accurate evaluation of dynamic earth pressures
can reduce either the cost or the risk, depending on the
circumstances of each problem and the performance
objectives of the design.
Despite its structural simplicity, the seismic response of a wall (retaining even a single soil layer) is a
rather complicated problem. The dynamic interaction
between the wall and the retained soil is the primary
factor that increases the complexity of the problem,
especially when material and/or geometry non-linearities are considered (Kramer 1996, Wu and Finn 1999,
Al-Homoud and Whitman 1999, PIANC 2001, Green
and Ebeling 2002, Moayyedian et al. 2008). Seismic
response of various types of walls that support a single
soil layer has been examined by a number of
researchers in the past either experimentally, analytically, or numerically (e.g. Veletsos and Younan 1997,
Iai 1998, Wu 1999, Psarropoulos et al. 2005).
Depending on the material behaviour of the retained
soil and the possible mode of the wall displacement,
there exist two main categories of analytical methods
used in the seismic design of retaining walls:

P.N. Psarropoulos et al.

Pseudo-static methods incorporating the limit-equilibrium concept (Mononobe-Okabe type solutions),
which assume yielding walls and rigid-perfectly plastic
behaviour of the retained soil (Okabe 1926, Mononobe
and Matsuo 1929, Seed and Whitman 1970).
Elasticity-based solutions that regard the retained soil as
a linear (visco-) elastic continuum (Scott 1973, Wood
1975, Veletsos and Younan 1997).

According to an ecient simplication of the

Mononobe-Okabe (M-O) method, developed by Seed
and Whitman (1970), the maximum dynamic active
earth force imposed on the wall is given by:
DPAE  0:4ArH2

methods are adopted by the majority of seismic norms,

whereas in cases of walls constrained against deformation and movement, elasticity-based solutions are
most frequently adopted. The above are shown in
more detail in Figure 1, where provisions of the Greek
Regulatory Guide E39/99 (1999) for the seismic
analysis of bridge abutments are given. The dynamic
pressure distribution depends on the ratio of the
horizontal displacement of the top of the wall to the
wall height u/H. For the case of a very exible wall
(u/H  0.10%), the Greek Regulatory Guide adopts
the Seed & Whitman method and suggests that the
maximum dynamic earth force is equal to:

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

DPAE 0:375ArH2  0:4ArH2

where A is the peak base acceleration, r is the soil mass
density, and H is the wall height. In contrast, the
elastic solutions developed by Scott (1973) or Wood
(1975) suggest that for the case of low-frequency
(quasi-static) base motions, which refers to many
practical problems, the dynamic active earth force
developed on a rigid xed-base wall is equal to:
DPAE  ArH2

Dynamic response of rigid bridge abutments is

considered as a very important design problem in
seismic regions, therefore many studies have been
focused on this topic over the last years (Siddharthan
et al. 1994, Fishman and Richards 1995, 1997, Imbsen
et al. 1997, Al-Homoud and Whitman 1999, MCEER
2001, Munaf et al. 2003, Basha and Babu 2009, among
others). Seismic norms usually correlate the intensity
and the distribution of the dynamic earth pressures
with the ability of the wall to move and/or deform.
In the case of a deformable wall, pseudo-static

Figure 1.

and acts at a height of 0.6H above the wall base.

For a moderate exibility of the wall (0.05%  u/
H  0.10%) then the maximum dynamic earth force is
derived by:
DPAE 0:75ArH2

acting at a height of 0.5H above the wall base. Finally,

for a rather rigid wall (u/H  0.05%) the Greek
Regulatory Guide, following the Woods solution,
proposes that the maximum force is given by Equation
(2) and acts at a height of 0.58H above the wall base.
According to contemporary seismic provisions
(NAVFAC DM7.02 1986, ATC32-1 1996, EC8 2003,
FEMA450 2003), the method applied for the determination of the proper seismic coecient for the
seismic design of retaining systems is the well-known
Mononobe-Okabe method, modied accordingly to
incorporate the behaviour of the retaining wall. In
Eurocode 8 (EC8 Part 5) three basic parameters are

Provisions of the Greek Regulatory Guide E39/99 for the seismic analysis of bridge abutments.

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

Structure and Infrastructure Engineering

considered in order to determine the seismic coecient:
the design acceleration, the soil factor (which takes
into account the ground type) and the reduction factor
to account for the expected or acceptable displacement
of the wall during earthquake loading. In any case,
seismic norms propose that the seismic earth pressures
acting upon a rigid retaining wall are more than twice
those acting upon a deformable retaining wall, given
that the wall dimensions, soil properties and peak
ground acceleration remain constant. As will be shown
in the sequence, this is not always true. The two
aforementioned modes of wall-soil system behaviour
are rather extreme, and in many cases fail to be
realistic due to their exaggerated assumptions. The
limit-equilibrium solutions imply the capability of the
system to develop relatively large displacements (geometric non-linearity) together with the formation of
plastic zones (material non-linearity). In contrast, the
elasticity-based solutions may take into account only
the wall exibility and/or the wall foundation compliance (Veletsos and Younan 1997).
There exist many cases, such as bridge abutments,
basement walls, or braced excavations, in which the
existence of kinematic constraints on the wall movement is incompatible with the limit-equilibrium concept, while on the other hand, the available elasticitybased solutions overlook the potential non-linear
behaviour of the retained soil, leading thus to either
unsafe or over-conservative solutions. It has to be
stressed that restrained basement walls, bridge abutments and other rigid retaining walls subjected to
dynamic loading are frequently encountered in geotechnical engineering practice. A basement wall or a
bridge abutment (such as those shown in Figure 2)
is generally a non-compliant retaining structure.
Dierent reasons in each case prevent the retaining

wall from yielding. In the case of a restrained basement

wall, the rigidity is induced by the existence of concrete
slabs at its top and its bottom which conne its
displacements. Similarly, a bridge abutment, the structure is conned against displacement at its top by the
bridge deck and usually at its bottom by deep piles.
Apart from the dynamic earth pressures developed
on the non-compliant retaining walls considered in this
study, emphasis is also given on the soil amplication
of the base acceleration. Note that seismic norms
(such as the Eurocode 8 (EC8 2003), or the Greek
Seismic Code (EAK 2000)), being based on the limitequilibrium methods, underestimate the role of the
potential soil amplication behind a retaining wall.
For this reason, the objective of the present study is
to examine thoroughly the inuence of material nonlinearity not only on the dynamic distress of bridge
abutments and similar types of non-deformable
retaining walls in general, but also on the soil amplication of the base acceleration. For this purpose, twodimensional numerical simulations are performed,
utilising the nite-element method in order to investigate some of the most important aspects of the
complex phenomenon of dynamic non-linear wall-soil
interaction.
A parametric study has been performed in order to
examine how the level of applied acceleration may
aect: (a) the dynamic earth pressures induced on such
retaining structures, and (b) the soil amplication
behind a retaining wall. Dynamic response of any
system depends on the seismic excitation characteristics (both in the time and in the frequency domain).
In order to examine more thoroughly the aforementioned complex phenomena the excitations utilised in
this study, include harmonic and simple pulses and real
excitations. Material non-linearity is taken into

Figure 2. Two cases of restrained retaining walls: (a) a basement wall conned against displacement at its top and its bottom
by concrete slabs, and (b) a bridge abutment conned against displacement at its top by the bridge deck and at its bottom by
piles.

P.N. Psarropoulos et al.

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

account in a simplistic, yet ecient way, through

the use of an iterative equivalent-linear procedure in
which strain-compatible shear modulus, G, and critical damping ratio, x, are used to describe the soil
behaviour within each iteration.
Results provide a clear indication of the degree to
which both dynamic response of restrained walls and
soil amplication of ground acceleration are aected
by soil non-linearity, a crucial parameter which has
been ignored by the seismic norms as far as rigid
retaining walls are concerned. Soil non-linearity seems
to increase the degree of complexity, being either
benecial or detrimental for the distress of a rigid
retaining system, depending on the circumstances. This
fact justies the necessity for a serious consideration of
potential soil non-linearity by the seismic norms
applied worldwide.
2. Numerical modelling
In order to examine the non-linear dynamic distress of
restrained walls and the non-linear soil amplication,
two-dimensional (2D) numerical simulations of the
retaining system depicted in Figure 3 were conducted.
The simulations were performed utilising the popular
QUAD4M nite-element code developed by Hudson
et al. (1993), which performs dynamic non-linear
analyses incorporating the well-known iterative
equivalent-linear procedure. Each iteration includes:
. A linear direct-integration dynamic analysis of
the model.
. The calculation of the maximum eective shear
strain, ge, for each element (calculated as a
percentage of the maximum strain).
. The calculation of the strain-compatible shear
modulus G(ge) and critical damping ratio x(ge)
to be used in the next iteration, by means of
G/Gmax 7 g and x 7 g curves.
The procedure is terminated when convergence
in the values of G and x occurs. The G/Gmax 7 g and
x 7 g curves used (see Figure 4) are characteristic of

Figure 3. The retaining system examined in the present

study: a rigid xed-base wall retaining a single soil layer with
strain-dependent material behavior, both excited by an
acceleration time-history A(t).

sandy soil material (Seed and Idriss 1970, Idriss 1990),

modied accordingly to provide critical damping ratio
equal to 5% in the low shear strain range.
As the wall exibility is examined in relation to soil
stiness and the earth pressures are normalised with
respect to r and H, the soil material properties and the
wall height values do not aect the dynamic pressures
on the wall (Veletsos and Younan 1997). This was also
veried in a recent numerical study by Psarropoulos
et al. (2005) where linear analyses of the examined
model were performed utilising ABAQUS software.
With respect to the inhomogeneity of the soil, it has
also been shown in the aforementioned paper that it is
benecial for the walls in the elastic range, since it was
proven that the dynamic earth thrust is substantially
lower compared to the one developed in the case
of homogeneous soil. Therefore, a single homogenous
soil layer and the parameters used in that study
have also been applied in the current non-linear
investigation.
All equivalent-linear analyses in QUAD4M were
performed considering an 8-m-high wall and the
retained soil layer is characterised by a relatively low
small-strain shear-wave velocity VS equal to 100 m/s
and a mass density of 1.8 tn/m3. The discretisation
of the retained soil was performed by four-noded
plane-strain quadrilateral elements. The model was
adequately elongated so as to reproduce adequately
the free-eld conditions at its right-hand side. In order
to simulate the dynamic response of restrained walls,
the wall is considered to be rigid and xed at its base.
The rigid wall was simulated by an extremely sti
column with linear elastic behaviour. The simplifying
assumption of no de-bonding or relative slip at the
wall-soil interface was used.
The base of the wall and the soil stratum were
considered to be excited by a horizontal motion,
assuming an equivalent force-excited system. Initially,
the model was subjected to harmonic and Ricker
pulses which allow for a better understanding and
interpretation of the results. Furthermore, the results
of harmonic excitations can be easily generalised for
any real earthquake excitations via Fourier transformations. Subsequently, two characteristic real earthquake records were used in the numerical analysis:
the Shin-Kobe record from the 1995 Kobe (Japan)
earthquake and the record from the 1995 Aegion
(Greece) earthquake. Both real records used in this
study dier substantially with respect to their frequency content, while they are characterised by peak
ground acceleration (PGA) values close to the maximum intensity level considered in this work (0.50 g).
More specically, the record from Aegion earthquake
has PGA equal to 0.54 g, while Shinkobe record has
PGA equal to 0.52 g. The acceleration time-history of

Structure and Infrastructure Engineering

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

the Ricker pulse together with its Fourier spectrum are

given in Figure 5, while the acceleration time-histories
and the corresponding response spectra (for 5%
damping) of the earthquake records of Aegion and
Shin-Kobe are shown in Figure 6. Note that plotted
time-histories are scaled to 0.50 g, while their spectra
are scaled to 0.01 g.
Various characteristic levels of peak base acceleration (PGA) were used in this study, aiming at the
development of dierent degrees of material nonlinearity. The desired intensity levels were obtained by

scaling the aforementioned pulses and seismic excitations to specic peak base acceleration values. As
the frequency content of the imposed acceleration timehistories varied substantially, their scaling was performed in terms of PGA values to maintain the
frequency content of each input motion almost unchanged and use only its PGA as the varying parameter.
It has to be noted at this point that, as it will be
further discussed in the numerical results, the use of
base instead of surface peak values for scaling purposes
is reasonable for the examined problems, since the
surface acceleration time-histories and their corresponding peak values vary substantially depending on
the distance of the surface backll point from the rigid
wall. Moreover, due to non-linearities (which are
related to base PGA levels) the eigenfrequency of the
soil layer is not only a function of its height as in the
linear case, thus, surface response and the resulting
peak values are dierent for each intensity level. In
addition, having a unique reference acceleration value
for each analysis is also more suitable for the normalisation of the results. Nevertheless, regardless of the
choice of the scaling parameter, the rather minor
scaling of the records that has been used in this
investigation does not induce any signicant bias on the
non-linear response of the system (Shome and Cornell
1999).

3.

Soil amplication of the retained backll

3.1. Linear response

3.1.1. Harmonic excitation
Figure 4. The G/Gmax g and x g curves (modied by the
authors based on Seed and Idriss 1970; Idriss 1990) used to
describe the non-linear behavior of the retained soil.

The dynamic linear response of a single soil layer with

horizontal stratigraphy, that actually resembles 1D
conditions, has been studied by many researchers, and

Figure 5. (a) Acceleration time-history (A is scaled to peak acceleration of 1m/s2), and (b) Fourier spectrum of the Ricker pulse
excitation (with central frequency fR 2 Hz).

P.N. Psarropoulos et al.

analytical solutions for harmonic excitation can be

found in the literature (Roesset 1977, Kramer 1996). In
the case of harmonic excitation the response is
controlled by the ratio f/fo, where f is the dominant
period of the excitation, and fo the fundamental period
of the soil layer. In the model examined in this study
the fundamental eigenfrequency of the soil layer fo is
almost 3.1 Hz (or equivalently, the fundamental
eigenperiod of the soil layer To is 0.32s). The duration
of the sinusoidal pulse was such that steady state
conditions were reached. In that case the maximum
amplication factor (AF) for linear response is given by
the simple expression:
2
1
px 2n 1

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

AF

where x is the critical damping ratio and n the

eigenmode number. For rst mode (n 0) and
x 5%, AF is approximately equal to 12.7.
In this study the response of the soil layer under 1D
conditions is compared with the corresponding response of the model shown in Figure 3. The presence
of a rigid retaining wall essentially imposes a vertical
boundary condition, leading thus to a 2D model. In
addition, this model has a fundamental low-strain
eigenfrequency slightly lower than the corresponding
1D model, due to the fact that the existence of the rigid
wall makes the model stier. However, the dierence in
the two values of eigenfrequency is considered
negligible as it is lower than 0.03 Hz. In addition, the
results of the equivalent linear analysis for very low
intensity levels (A 0.0001 g) are identical with the
results of linear analysis.
The distribution of the amplication factor (AF) on
the surface of the backll in the case of harmonic
excitation at resonance (f fo) is plotted in the upper
curve (for A 0.0001 g) in Figure 7. It is evident that,
for the rigid xed-base wall examined, the motion in
the vicinity of the wall is practically induced by the
wall itself, and therefore, no amplication is observed
(AF 1). The amplication factor converges to its
maximum value (AF  12.7) at a distance of 10H from
the wall, since at that distance 1D conditions are
present (free-eld motion). Note that this distance was
also calculated by Wood (1975), as the minimum
distance needed to eliminate the eects of the wall on
the retained soil.
3.1.2.

Ricker pulse excitation

As previously mentioned, apart from harmonic excitations a Ricker pulse with central frequency fR 2Hz
has also been used in the present study (Ricker 1960).
Despite the simplicity of its waveform, this wavelet
covers smoothly a broad range of frequencies up to
nearly 3fR ( 6Hz). The acceleration time-history and
the corresponding Fourier spectrum of this pulse are
given in Figure 5. Figure 8 depicts the waveforms of
the acceleration time-histories on the surface of the
retained soil layer for the rigid wall excited with the
Ricker pulse. It is obvious that in the vicinity of
the rigid wall the amplication is approximately equal
to unity, while in a similar distance ( 10H) as in the
case of the harmonic excitation, the response of the soil
converges to free-eld conditions.
Figure 6. The acceleration time-histories of the records
from the 1995 Aegion, Greece, earthquake (top) and the 1995
Kobe, Japan, earthquake (bottom), both scaled to peak
ground acceleration equal to 0.50 g. The corresponding
response spectra, both scaled to 0.01 g, are also shown.

3.2. Non-linear response

The aforementioned results referring to the case of
linear soil behaviour are valid for very low levels of
base input acceleration, when the induced strains

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

Structure and Infrastructure Engineering

Figure 7. Non-linear analysis results: distribution of the soil

amplication factor (AF) along the surface of the backll in
the case of harmonic excitation with frequency f equal to the
low-strain rst Eigen frequency fo.

Kramer 1996). In the current study the only case in

which the induced seismic shear strains increase
substantially (approaching values in the range of 1%)
is the case of the highest acceleration level and high
mean period earthquake (i.e. the Kobe excitation with
PGA 0.50 g). Therefore, the results of the performed dynamic nite-element analyses can be regarded as accurate.
The wall-soil system behaviour strongly depends,
not only on the level of applied acceleration, but also
on the f/fo ratio, as it is justied by the subsequent
results. The distribution of the amplication factor on
the surface of the backll in the case of the harmonic
excitation at resonance (f fo) is plotted in Figure 7,
for ve levels of peak base acceleration: 0.0001 g
(corresponding practically to a linear soil behaviour),
0.12 g, 0.24 g, 0.36 g, and 0.50 g, covering a broad
range of the induced dynamic strains. Note that in the
range of small shear strains the critical damping ratio,
x, was set equal to 5%, instead of the much lower
values of the curves proposed by Seed and Idriss
(1970), in order to ensure that the theoretical
amplication (AF  12.7) for linear conditions is also
numerically achieved for the lowest peak base acceleration case (0.0001 g). As it was expected, increasing
the degree of material non-linearity makes the system
more exible, thus decreases its fundamental Eigen
frequency and leads to detuning. This phenomenon
can be easily observed in Figure 7 by noticing the
substantially reduced values of AF for all levels (0.12 g
to 0.50 g) of non-linear behaviour.

4.
4.1.

Figure 8. Linear analysis results: waveforms of the

acceleration time-histories along the surface of the retained
soil layer excited with the Ricker pulse. It is evident that in
the vicinity of the rigid wall there exists no amplication.

remain small (g 5 0.005%). However, when the

maximum acceleration acting on the soil mass takes
more realistic values, the induced strains are substantially greater, and thus, the impact of material nonlinearity (expressed by the G/Gmax 7 g and x 7 g
curves) becomes more evident. The equivalent linear
approach (in which as previously described a few linear
analyses are performed in each cycle), is considered
accurate enough for shear strains lower than 1% (see

Dynamic distress of restrained walls

Harmonic excitation

4.1.1. Linear response

In order to calculate the linear elastic dynamic earth
pressures which develop behind a rigidly restrained
retaining wall (bridge abutment or basement wall), two
harmonic excitations were employed, which had two
characteristic frequencies: the rst was set equal to the
low-strain fundamental eigenfrequency of the soil layer
(f fo), and the second pulse had frequency six times
lower (f fo/6), approximating a quasi-static excitation. The curves for A 0.0001 g in Figure 9 and
Figure 10 presents the height-wise distribution of the
normalised induced linear elastic dynamic earth
pressures for the two harmonic excitations examined.
According to common practice (Veletsos and
Younan 1997, Psarropoulos et al. 2005), all dynamic
distress results presented here are normalised with
respect to basic mechanical and geometrical parameters of the model. In particular, the dynamic earth
pressures are normalised with respect to rHA and the

P.N. Psarropoulos et al.

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

Figure 9. Non-linear analysis results: height-wise

distribution of the normalized induced dynamic earth
pressures for the harmonic excitation with frequency f
equal to the low-strain rst eigenfrequency fo.

Figure 10. Non-linear analysis results: height-wise

distribution of the normalised induced dynamic earth
pressures for the low-frequency harmonic excitation
(f fo/6).

dynamic earth thrust is normalised in terms of rH2A.

Therefore, if any of the three basic parameters, namely,
the soil density r, the soil layer height H and
acceleration amplitude A change, new estimates of
the dynamic distress measures can be calculated from
their normalised values. For instance, a change in H
will aect only the eigenfrequencies of the retaining
system (the higher the soil layer thickness, the lower
the eigenfrequencies). An increase in the soil density r
will make the system stier, thus, the degree of nonlinearity will be reduced.

values ranging from 0.60 to 0.90 (corresponding to

A 0.50 g and A 0.12 g levels, respectively) compared to the previously calculated value of 3.00 for
the linear soil behaviour case. It is evident that for
excitations having dominant frequency close to the
low-strain fundamental eigenfrequency of the retained
soil layer, the material non-linearity seems to act in a
benecial way.
The case of quasi-static excitation is of greater
interest. In Figure 10 the height-wise distribution of
the normalised induced dynamic earth pressures is
plotted for the case of the low-frequency harmonic
excitation (f fo/6) for the ve levels of peak base
acceleration examined. Apparently, for higher levels of
peak base acceleration, which incur non-linear response
of the retained system, the normalised dynamic earth
pressures are always higher than those corresponding
to the linear elastic response. The two envelope curves
for the harmonic excitations shown in Figure 11 (which
depicts the cumulative results for all the examined
excitations) stem from the results shown in Figures 9
and 10, as it presents DPAE as a function of peak base
acceleration A. Taking also into consideration the
values obtained using Equations (1) and (2) (i.e. 0.4 and
1.0, respectively), one could easily notice that when the
fundamental frequency of the input motion f approaches that of the retained soil layer fo, the dynamic
earth force DPAE is almost three times greater in the
case of resonance, compared with the corresponding
value in the linear (A 0.0001 g) case of quasi-static
excitation, which is almost equal to unity.
This seeming discrepancy can be attributed to the
fact that at certain increased levels of induced strain,
the shear modulus of the retained soil becomes lower,
according to the G/Gmax 7 g curves proposed by Seed
and Idriss (1970). Therefore, this softening of the wallsoil system shifts its fundamental Eigen frequency to
lower values, closer to resonance with the applied input
excitation, and thus, leads to an increased distress.
For higher levels of induced strain (A  0.50 g), the
resonance is avoided since the excitation frequency
exceeds the fundamental low-strain eigenfrequency of
the retained system. Therefore, for the case of lowfrequency harmonic excitation, the eects of material
non-linearity may be benecial or detrimental depending on the circumstances.

4.1.2. Non-linear response

Figure 9 depicts the height-wise distribution of the
normalised induced dynamic earth pressures for the
ve levels of peak base acceleration examined in
the case of the harmonic excitation with f fo.
As the level of applied acceleration increases the
dynamic earth pressures decrease. In particular, the
normalised dynamic earth force on the wall reduces to

4.2.

Ricker pulse excitation

The height-wise distribution of the normalised induced

dynamic earth pressures in the case of the Ricker pulse
excitation is plotted in Figure 12, for all ve levels of
peak base acceleration examined. The pattern revealed
in Figure 9 for the harmonic excitation with f fo is
repeated in this case, due to the fact that the excitation

Structure and Infrastructure Engineering

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

Figure 11. Non-linear analysis results: maximum

normalised dynamic earth force as a function of peak base
acceleration A, for all the excitations examined. Graph also
includes the proposals made by Wood and Seed and
Whitman (which are also adopted by the Greek Regulatory
Guide E39/99).

Figure 12. Non-linear analysis results: height-wise

distribution of the normalised induced dynamic earth
pressures for the case of the Ricker pulse excitation.

pulse includes a broad range of frequencies close

to the fundamental frequency of the retained soil layer
(fo  3 Hz), as shown in the Fourier spectrum of the
pulse shown in Figure 5(b). As a result, despite the
lower levels of linear elastic earth pressures in this case,
the system response is quite similar to that caused by
the harmonic excitation at resonance.
As this Ricker pulse covers smoothly the range of
frequencies between 1 and 5 Hz, it provides an ecient
way to comprehend the eect of material non-linearity
on the wall distress in the frequency domain as well.
Figure 13 presents the variation of the pressure
amplication factor (PAF) as a function of frequency.
This amplication parameter is dened as:


FFT DPAE t
PAF
6
FFTAt


where FFT DPAE t is the Fourier spectrum of the
normalised induced dynamic earth force time-history

Figure
13. Non-linear
analysis
results:
pressure
amplication factor (PAF) values for the case of the Ricker
pulse excitation.

DPAE t, and FFT[A(t)] is the Fourier spectrum of the

acceleration time-history of a Ricker pulse excitation
with unit peak value (see Figure 5).
It is evident that in the case of linear soil behaviour,
PAF reaches its maximum value frequencies close to
the fundamental frequency of the retained soil layer.
This result matches the value calculated previously in
the case of linear harmonic response at resonance.
Additionally, for low-frequency excitations, the value
of PAF converges to that proposed by Scott (1973) and
Wood (1975) as previously calculated. By examining
Figure 13 it is obvious that for increased levels of peak
base acceleration, the development of material nonlinearity not only aects the maximum value of PAF,
but also shifts the range of its maximum values
towards lower frequencies. This phenomenon can be
either benecial or detrimental, depending on the
predominant frequency of the input motion.
4.3.

Seismic excitations

As it was previously mentioned, two real earthquake

records are used in the present study. The rst record
was derived from the Shin-Kobe station during the
earthquake of Kobe in Japan in 1995, and the second is
taken from the OTE Building in Aegion, during the
1995 Aegion earthquake in Greece. Figure 6 shows the
acceleration time-histories of these earthquake records,
both scaled to peak acceleration of 0.50 g. The heightwise distribution of the normalised induced dynamic
earth pressures in the cases of the Aegion and the ShinKobe records is plotted in Figure 14, for ve levels of
peak base acceleration. The maximum pressure distribution occurs for very low acceleration amplitude
imposed by the seismic excitation, for both earthquakes. Furthermore, the pressure distributions in the
case of linear elastic response (0.0001 g) are nearly the

10

P.N. Psarropoulos et al.

frequency content of the excitation. More specically,
DPAE varies between the values of 1.0 and 3.0, being
thus always higher than the standard bounding values
adopted by the seismic norms (noted as Wood 1975
and Seed and Whitman 1970 (M-O)). It has to be
mentioned that, the divergence is maximum when the
fundamental frequency of the base excitation equals
that of the retained soil layer under 1D linear elastic
conditions, i.e.:

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

VS
fo

4H

Figure 14.
distribution
pressures in
Kobe record

Non-linear analysis results: height-wise

of the normalized induced dynamic earth
the cases of Aegion record (top) and Shin(bottom).

same for both earthquakes. Moreover, it is observed

that the pattern of the pressure distributions is the
same with that in the case of Ricker excitation, and it
diers essentially from the corresponding one for the
quasi-static harmonic excitation.
4.4.

Discussion of the results

As mentioned previously, Figure 11 presents the

maximum normalised dynamic active earth force as a
function of peak base acceleration A for the ve
excitations examined in this study. In the same plot,
the values of DPAE proposed by Wood (1975) and
by Seed and Whitman (1970) are also included for
reference. As it has been mentioned in the introductory
section of this study, Woods solution is identical to the
guidelines of the Greek Regulatory Guide E 39/99 for
rigid bridge abutments, whereas the dynamic earth
force proposed by Seed and Whitman is identical to
the provisions for exible walls.
It can be observed that in the case of linear
response the wall distress is dominated by the

p
Gmax r
4H

where VS is the shear-wave velocity of the soil, and

Gmax is the corresponding small-strain shear modulus.
Nevertheless, as the degree of non-linearity increases
the induced dynamic active earth force decreases
substantially, ranging between the aforementioned
Wood and M-O bounds in the cases of the harmonic
resonant motion, the Ricker pulse and the two records
from Kobe and Aegion earthquakes.
In contrast, the distress in the case of lowfrequency (quasi-static) harmonic excitation and nonlinear response is always higher than the upper bound
(Woods solution) for all levels of peak base acceleration, and is approximately 50% greater than the value
of Woods solution. An important conclusion resulting
from Figure 11 is that, for high values of the imposed
base acceleration, the resulting force approximates
in general the proposal of Seed and Whitman, even
though the limit-equilibrium conditions (imposed
by Coulombs static theory, or its pseudo-static
extension of M-O) are not valid in the specic retaining
system. In other words, the force acting on the back of
a yielding retaining wall (resulting from the weight of a
rigid wedge of soil above a planar failure surface,
according to M-O theory) coincides with the force that
acts on the back of a restrained wall (resulting from the
earth pressures of a yielding soil material).
In current engineering practice, the dynamic earth
pressures depend on the PGA and the compliance of
the wall and/or its foundation. The current study
has shown that the potential soil non-linearity may
aect substantially the distribution of earth pressures,
usually leading to a decrease in the wall distress.
Therefore, the following equation that contains a
reduction factor l could be adopted in seismic norms:
DPAE  lArH2

According to the cumulative results and the related

discussion of Figure 11 the reduction factor l in
Equation (8) will be a function of PGA levels. More
specically, for small intensity levels (A 5 0.1 g),
no reduction factor is required, thus l 1; for

Structure and Infrastructure Engineering

medium intensity levels (A 0.1 g to 0.3 g), it will be
equal to l 0.75; nally for greater intensity levels
(A 4 0.3 g) then it will be set as l 0.5, actually
converging to M-O solution.

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

5.

Conclusions

The present study has examined how and to what

extent the potential soil non-linearity that a retained
soil layer exhibits under moderate or severe seismic
excitations can possibly aect: (a) the seismic response
of the retained soil layer itself, and (b) the dynamic
distress of restrained walls. It was found that soil nonlinearity reduces in general the soil amplication of the
retained soil and the dynamic earth pressures, leading
thus to lower wall distress. However, as soil nonlinearity alters the eigenfrequencies of the wall-soil
system, there is (under certain circumstances) the
possibility that increased non-linearity may lead to
an amplied response. This phenomenon is more
probable to occur when the frequency content of the
excitation is narrow and concentrated around a
fundamental frequency that is lower than the linear
eigenfrequency of the soil layer.
A major conclusion of this study is the fact that
the seismic active earth force induced on restrained
walls is not necessarily much larger than the corresponding force induced on yielding retaining walls.
It depends on the potential non-linearity of the
retained soil and the relationship between the fundamental frequency of the imposed excitation and the
fundamental large-strain eigenfrequency of the retaining system. Consequently, the existence of a rigid
retaining wall, such as a bridge abutment or a
basement wall, does not necessarily entail that its
distress will be much greater than the distress induced
on the corresponding yielding wall, as it is stated
in the majority of seismic provisions. Therefore,
these provisions have to be modied accordingly, as
suggested by the ndings of this study, to include the
eects of soil non-linearity on the seismic distress of
restrained walls.
Acknowledgements
The authors would like to acknowledge the assistance of V.
Zania and S. Tsimpourakis and to thank the anonymous
reviewers for their constructive comments.

References
Al-Homoud, A.S. and Whitman, R.V., 1999. Seismic
analysis and design of rigid bridge abutments considering
rotation and sliding incorporating non-linear soil behaviour. Soil Dynamics and Earthquake Engineering, 18,
247277.

11

ATC32-1, 1996. Improved seismic design criteria for California bridges: Resource document. USA: Applied Technology Council.
Basha, B.M. and Babu, G.L.S., 2009. Computation of sliding
displacements of bridge abutments by pseudo-dynamic
method. Soil Dynamics and Earthquake Engineering, 29,
103120.
E39/99, 1999. Greek regulatory guide for the seismic analysis
of bridges. Athens, Greece: Ministry of Public Works (in
Greek).
EAK, 2000. Greek seismic code. Athens, Greece: Greek
Ministry of Public Works (in Greek).
EC8, 2003. Eurocode 8: Design of structures for
earthquake resistance. European Standard CENENV-1998-1. Brussels: European Committee for
Standardisation.
FEMA 450, 2003. NEHRP2003: Recommended provisions for
seismic regulations for new buildings and other structures.
Part 2: Commentary Chapter 7: Commentary foundation
design requirements. USA: Building Seismic Safety
Council (BSSC).
Fishman, K.L. and Richards, R.J., 1995. Seismic vulnerability of existing bridge abutments [online]. NCEER
Bulletin, 9 (3). Available from: http://mceer.bualo.edu/
research/HighwayPrj/bullarticles/Jul95Vol9No3Pg8.asp
[Accessed November 2008].
Fishman, K.L. and Richards, R.J., 1997. Seismic analysis
and design of bridge abutments considering sliding and
rotation. Technical Report NCEER-97-0009. NCEER,
Bualo, NY.
Green, R.A. and Ebeling, R.M., 2002. Seismic analysis of
cantilever retaining walls. Phase I Report ERDC/ITL
TR-02-3. Washington, DC: US Army Corps of
Engineers.
Hudson, M., Idriss, I.M., and Beikae, M., 1994. Users
manual for QUAD4M. Davis, USA: Center for Geotechnical Modeling, Department of Civil and Environmental
Engineering, University of California.
Iai, S., 1998. Seismic analysis and performance of retaining structures. In: P. Dakoulas, M. Yegian, and
R.D. Holtz, eds. Proceedings of geotechnical earthquake engineering and soil dynamics III, Geotechnical
Special Publication No. 75. Reston, VA: ASCE, 1020
1044.
Idriss, I.M., 1990. Response of soft soil sites during
earthquakes. In: J.M. Duncan, ed. Proceedings of
H. Bolton Seed Memorial Symposium, Vol. 2. May,
Berkeley, CA. Vancouver, B.C., Canada: BiTech Publishers Ltd., 273289.
Imbsen, R.A., et al., 1997. Structural details to accommodate
seismic movements of highway bridges and retaining
walls. Technical Report NCEER-97-0007, NCEER,
Bualo, NY.
Kramer, S.L., 1996. Geotechnical earthquake engineering.
New Jersey: Prentice-Hall.
MCEER, 2001. Recommended LRFD guidelines for the
seismic design of highway bridges. NCHRP Project
12-49. Bualo, NY: MCEER.
Moayyedian, M., Moslem, K., and Shooshtari, A., 2008.
Proposed dynamic soil pressure diagram on rigid walls.
IJE Transactions A: Basics, 21 (3), 213224.
Mononobe, N. and Matsuo, H., 1929. On the determination
of earth pressures during earthquakes. In: Proceedings of
the World Engineering Congress, Vol. 9, Paper 388,
OctoberNovember, Tokyo, Japan. Tokyo: JSCE,
177185.

Downloaded By: [Technical University of Crete] At: 09:15 21 July 2010

12

P.N. Psarropoulos et al.

Munaf, Y., Prakash, S., and Fennessey, T., 2003. Geotechnical seismic evaluation of bridge abutments in southeast Missouri. In: Proceedings of the 2003 Pacic
Conference on Earthquake Engineering, 1315 February,
Christchurch, New Zealand. Christchurch, NZ:
Department of Civil Engineering, University of Canterbury. CD ROM Paper No 5.
NAVFAC DM7.02, 1986. Foundations and earth retaining
structures design manual. Alexandria, Virginia: Naval
Facilities Engineering Command.
Okabe, S., 1926. General theory of earth pressures. Journal of
the Japanese Society of Civil Engineering, 12 (1), 311.
PIANC, 2001. International Navigation Association
(PIANC): Seismic design guidelines for port structures.
Rotterdam, The Netherlands: Balkema Publishers.
Psarropoulos, P.N., Klonaris, G., and Gazetas, G., 2005.
Seismic earth pressures on rigid and exible retaining
walls. Soil Dynamics and Earthquake Engineering, 25 (7
10), 795809.
Ricker, N., 1960. The form and laws of propagation of
seismic wavelets. Geophysics, 18, 1040.
Roesset, J.M., 1977. Soil amplication of earthquakes. In:
C.S. Desai and J.T. Christian, eds. Numerical methods in
geotechnical engineering. New York, NY: McGraw Hill
Book Co., 639682.
Scott, R.F., 1973. Earthquake-induced pressures on retaining
walls. In: Proceedings of the 5th World Conference on
Earthquake Engineering, Vol. 2, 2529 June, Rome, Italy.
Rome: Ministry of Public Works, 16111620.

Seed, H.B. and Idriss, I.M., 1970. Soil moduli and damping
factors for dynamic response analyses. Report EERC 70
10, Earthquake Engineering Research Center, University
of California, Berkeley, CA.
Seed, H.B. and Whitman, R.V., 1970. Design of earth
retaining structures for dynamic loads. In: Proceedings of
the special conference on lateral stresses in the ground and
design of earth retaining structures, 2224 June, Ithaca,
NY, New York, NY: ASCE, 103147.
Shome, N., and Cornell, C.A., 1999. Probabilistic seismic
demand analysis of nonlinear structures. Report No.
RMS-35, RMS Program, Stanford University, Palo Alto,
CA.
Siddharthan, R., El-Gamai, M., and Maragakis, E.A., 1994.
Investigation of performance of bridge abutments in
seismic regions. ASCE Journal of Structural Engineering,
120 (4), 13271346.
Veletsos, A.S. and Younan, A.H., 1997. Dynamic response
of cantilever retaining walls. ASCE Journal of Geotechnical and Geoenvironmental Engineering, 123 (2), 161172.
Wood, J.H., 1975. Earthquake-induced pressures on a rigid
wall structure. Bulletin of New Zealand National Earthquake Engineering, 8, 175186.
Wu, G. and Finn, W.D.L., 1999. Seismic lateral pressures for
design of rigid walls. Canadian Geotechnical Journal, 36
(3), 509522.
Wu, Y., 1999. Displacement-based analysis and design of rigid
retaining walls during earthquake. PhD Dissertation.
University of Missouri-Rolla, USA.