Engineering Structures 33 (2011) 2186–2200

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

The effect of earthquake frequency content on the seismic behavior of concrete

rectangular liquid tanks using the finite element method incorporating
soil–structure interaction
M.R. Kianoush ∗ , A.R. Ghaemmaghami
Civil Engineering Department, Ryerson University, Toronto, ON, Canada

article info abstract

Article history: A three-dimensional soil–structure–liquid interaction is numerically simulated using the finite element
Received 15 February 2010 method in order to analyze the seismic behavior of partially filled concrete rectangular tanks subjected to
Received in revised form different ground motions. In this paper, the effect of earthquake frequency content on the seismic behavior
13 February 2011
of fluid rectangular tank system is investigated using four different seismic motions. A simple model with
Accepted 18 March 2011
Available online 17 April 2011
viscous boundary is used to include deformable foundation effects as a linear elastic medium. This method
is capable of considering both impulsive and convective responses of liquid-tank system. Six different soil
Keywords:
types defined in the well-recognized seismic codes are considered. The sloshing behavior is simulated
Sloshing using linear free surface boundary condition. Two different finite element models corresponding with
Rectangular tank flexible shallow and tall tank configurations are studied under the effects of longitudinal, transversal and
Dynamic analysis vertical ground motions. By means of changing the soil properties, comparisons are made on base shear,
Tank flexibility base moment and sloshing responses under different ground motions. It is concluded that the dynamic
Finite element behavior of the fluid–tank–soil system is highly sensitive to frequency characteristics of the earthquake
Liquid containing record.
© 2011 Elsevier Ltd. All rights reserved.

1. Introduction in a flexible tank can be significantly higher than the correspond-

The dynamic interaction between fluid, structure and soil is a
significant concern in many engineering problems. These problems
include systems as diverse as off-shore and submerged structures,
biomechanical systems, suspension bridges and storage tanks. The
interaction can drastically change the dynamic characteristics of
the structure and consequently its response to transient and cyclic
excitation. Therefore, it is desired to accurately model these diverse
systems with the inclusion of fluid–structure interaction (FSI).
One of the critical lifeline structures which has become
widespread during the recent decades is liquid storage tank. These
structures are extensively used in water supply facilities, oil and
gas industries and nuclear plants for storage of a variety of liquid
or liquid-like materials such as oil, liquefied natural gas (LNG),
chemical fluids and wastes of different forms.
Problems associated with liquid tanks involve many fundamen-
tal parameters. In fact, the dynamic behavior of liquid tanks is
governed by earthquake characteristics, the interaction between
fluid and structure as well as soil and structure along their bound-
aries. For example, it has been found that hydrodynamic pressure
∗ Corresponding author. Tel.: +416 979 5000x6455.
E-mail addresses: kianoush@ryerson.ca (M.R. Kianoush),
aghaemma@ryerson.ca (A.R. Ghaemmaghami).
0141-0296/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.03.009
ing rigid container due to the interaction effects between flexible
structure and contained liquid.
Even though there have been numerous studies done on the
fluid–structure interaction effects in liquid containers, most of
them are concerned with cylindrical tanks and the studies on
seismic response of rectangular tanks are quite rare. Housner [1]
developed the most commonly used analytical model for rigid
tanks in which hydrodynamic pressure is separated into impulsive
and convective components using lumped mass approximation.
This model has been adopted with some modifications in most of
the current codes and standards.
Later, Yang [2], Veletos and Yang [3] studied the effects of wall
flexibility on the pressure distribution in liquid and correspond-
ing forces in the tank structure through an analytical model. Mi-
nowa [4,5] investigated the effect of flexibility of tank walls and hy-
drodynamic pressure acting on the wall using both analytical and
experimental methods.
Studies on the three-dimensional fluid–structure interaction in
a domain of a more general geometry, other than the cylindrical
shape, can be found in the literature on the seismic design of
concrete dams.
Chopra and Gupta [6] investigated the effects of fluid–structure
and soil–structure interaction on the frequency function re-
sponses of water–dam–foundation system. They concluded that
 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200
fluid–structure interaction has a significant effect on resonant fre-
quencies of the dam under seismic ground motions.
Haroun [7] presented a very detailed analytical method in the
typical system of loading in rigid rectangular tanks. In addition,
Haroun [8] carried out a series of experiments including ambient
and forced vibration tests. Also, Haroun and Tayel [9] used the
finite element method (FEM) for analyzing dynamic response of
liquid tanks subjected to vertical seismic ground motions.
Veletsos and Tang [10] analyzed the liquid storage tanks
subjected to vertical ground motion on both rigid and flexible sup-
porting media. Haroun and Abou-Izzeddine [11] conducted a para-
metric study of numerous factors affecting the seismic soil–tank
interaction under vertical excitations.
Veletso et al. [12] found that the convective components of
response are insensitive to the flexibilities of the tank wall and
supporting soils, and may be computed considering both the tank
and the supporting medium to be rigid. Kim et al. [13] further
developed analytical solution methods and presented the response
of filled flexible rectangular tanks using both two- and three-
dimensional modeling. Park et al. [14] performed research studies
on dynamic response of the rectangular tanks using both finite
element method (FEM) and boundary element method (BEM).
Subhash and Bhattacharyya [15] developed a numerical scheme
using the finite element technique to calculate the sloshing
displacement of liquid and pressure developed to such sloshing.
Koh et al. [16] presented a coupled BEM–FEM, including free
sloshing motion, to analyze three-dimensional rectangular storage
tanks subjected to horizontal ground motion. In addition, they
conducted a shaking table test on a three-dimensional rectangular
tank under horizontal El-Centro ground motion record to verify
their numerical method.
Dogangun et al. [17] investigated the seismic response of liquid-
filled rectangular storage tanks using the finite element method
implemented in the general purpose structural analysis computer
code SAPIV.
Chen and Kianoush [18] used the sequential method to calcu-
late impulsive hydrodynamic pressure in two-dimensional rect-
angular tanks including wall flexibility effects. Also, Kianoush
and Chen [19] investigated the dynamic behavior of rectangular
tanks subjected to vertical seismic vibrations in a two-dimensional
space. In addition, Kianoush et al. [20] introduced a new method
for seismic analysis of rectangular containers in two-dimensional
space in which the effects of both impulsive and convective com-
ponents are accounted for in time domain.
Livaoglu [21] evaluated the dynamic behavior of fluid–
rectangular tank foundation system with a simple seismic analysis
procedure. In this procedure, interaction effects were presented by
Housner’s two mass approximations for fluid and the cone model
for soil foundation system. Two different earthquake records
were used to investigate the effect of earthquake frequency on
soil–structure–fluid interaction.
In the field of seismic behavior of structural buildings, many
studies have investigated the effect of vibration frequency on
dynamic responses. Some of the major findings are reported by
Jennings and Kuroiwa [22], Foutch et al. [23], Mcverry [24] and
Chopra [25]. It is clear that peak responses of the structure occur
when the fundamental frequency of the vibration reaches near the
corresponding natural frequencies of the structure.
Considering the previous studies done on this topic, it is
clear that very limited research has been done on the soil–
structure–fluid interaction (SSFI) effect on seismic behavior of
concrete rectangular tanks. Wolf and Song [26,27] simplified foun-
dation as an isotropic homogeneous elastic medium. The near field
was modeled using finite elements, and the far field was treated
by adding some special boundaries such as springs and dampers.
The soil treated in most cases as a semi-infinite medium, and
2187
this unbounded domain was large enough to include the effect of
soil–structure interaction. This approach has been used by Clough
and Penzien [28] and Wilson [29].
When analyzing a soil–structure–liquid system using FEM, a
foundation model that extends one tank length in the downstream,
upstream and downward directions usually suffices in most cases.
This approach permits different soil properties to be assigned to
different elements, so that the variation of soil characteristics with
depth can be considered.
There are different boundary models available in frequency
or time domains. Earlier, Lysmer and Kuhlmeyer [30] developed
a viscous boundary model using one-dimensional beam theory.
This theory has been commonly used with the FE method. Later,
more complex boundary types were used and developed such
as damping-solvent extraction [31,27], doubly asymptotic multi-
directional transmitting boundary [26,27] and paraxial boundary
methods [32].
This study comprehensively investigates the dynamic behavior
of concrete rectangular tanks using the finite element method
(FEM) in three-dimensional space in which the coupled fluid–stru-
cture equations are solved using the direct integral method in time
domain. Contrary to the common method of analysis in this subject
which is based on frequency domain and lumped mass approach,
the governing equation of fluid domain in this study is discretized
using the FEM which accounts for both impulsive and convective
behavior.
Four different earthquake ground motions including 1994
Northridge, 1940 El-Centro, 1971 San-Fernando and 1957 San-
Francisco with different frequency contents are applied in FE anal-
yses to account for the effect of frequency content. All records are
scaled in such a way that the horizontal peak ground accelera-
tion reaches 0.4 g. A good indicator of the frequency content of
the ground motion is the ratio of peak ground acceleration (PGA)
which is expressed in units of g to peak ground velocity (PGV) ex-
pressed in units of m/s. Earthquake records may be classified ac-
cording to the frequency content ratio into three categories, high
PGA/PGV ratio when PGA/PGV > 1.2, intermediate PGA/PGV ra-
tio when 1.2 > PGA/PGV > 0.8 and low PGA/PGV ratio when
PGA/PGV < 0.8. The Northridge record has low frequency con-
tent, the El-Centro and San-Fernando earthquakes have intermedi-
ate frequency contents and the San-Francisco record has the high
frequency content.
To account for the effect of soil–structure interaction, six
different soil types which are adopted from common design codes
are used in the FE modeling. It should be noted that the main
objective of this study is to investigate the effect of earthquake
frequency content on the dynamic behavior of liquid tanks. As a
result, this study is mainly focused towards this part and a brief
summary is given on SSI effect.
This study has led to some new findings which are presented
with the aid of two different tank configurations that are ana-
lyzed under time history excitations incorporating soil–structure
interaction. One of the major advantages of proposed method is in
accounting for soil–structure interaction (SSI), fluid–structure in-
teraction (SFI), fluid damping properties and considering impul-
sive and convective components separately which have not been
considered in previous studies. It should be noted that this study
is based on analyzing a linear system of fluid–structure–soil and
therefore effect of material and wave nonlinearity is not consid-
ered.
2. Analysis method for fluid–tank–soil system
In liquid domain, the hydrodynamic pressure distribution is
governed by the pressure wave equation. Owing to the small
volume of containers, the velocity of pressure wave assumed to
2188 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200

be infinity. Assuming that water is incompressible and neglecting In the above formulation, matrices [H ] and [G] are constants
its viscosity, the small-amplitude irrotational motion of water is during the analysis while the force vector {F }, pressure vector
governed by the three-dimensional wave equation: {P } and its derivatives are the variable quantities. In the coupling
system of liquid–structure the pressures are applied to the
∇ 2 P (x, y, z , t ) = 0 (1) structure surface as the loads on the container walls. The general
where P (x, y, z , t ) is the hydrodynamic pressure in excess of equation of fluid–structure neglecting the soil interaction can be
hydrostatic pressure. written in the following form:
The hydrodynamic pressure in Eq. (1) is due to the horizontal [M ]{Ü } + [C ]{U̇ } + [K ]{U } = {f1 } − [M ]{Üg } + [Q ]{P }
and vertical seismic excitations of the walls and bottom of the = {F1 } + [Q ]{P }
container. For earthquake excitation, the appropriate boundary
condition at the interface of liquid and tank is governed by [G]{P̈ } + [C ′ ]{Ṗ } + [H ]{P } = {F } − ρ[Q ]T ({Ü } + {Üg })
= {F2 } − ρ[Q ]T {Ü } (7)
∂ P (x, y, z , t )
= −ρ an (x, y, z , t ) (2)
∂n where [M ], [C ] and [K ] are mass, damping and stiffness matrices
of structure. {Ü } is the acceleration vector of nodes in the structure
where ρ is the density of liquid and an is the component of
domain, {Üg } is the ground acceleration vector applied to the
acceleration on the boundary along the direction outward normal
system and [Q ] is the coupling matrix. The term [C ′ ] is the matrix
n. No wave absorption is considered in the interface boundary
representing the damping of liquid which is dependent on the
condition.
viscosity of liquid and wave absorption in liquid domain and
Accounting to the small-amplitude gravity waves on the free
boundaries and is rigorously determined. As previously discussed,
surface of the liquid, the resulting boundary condition is given as
the matrix [Q ] transfers the liquid pressure to the structure as well
1 ∂ 2P ∂P as structural acceleration to the liquid domain.
+ = 0. (3) The direct integration scheme is used to find the displacement
g ∂ t2 ∂z
and hydrodynamic pressure at the end of time increment i+1 given
In which z is the vertical direction and g is the gravitational the displacement and hydrodynamic pressure at i.
acceleration. Descriptions regarding the coupling matrix and direct inte-
However, due to the large amplitude of sloshing under the gration method can be found in the studies done by Kianoush
strong seismic excitations and turbulence effects in liquid tanks, et al. [20].
more complicated boundary conditions on the surface of liquid are Under free oscillations, the motion of free liquid surface de-
needed to accurately model the convective motions. In a recent cays due to damping forces created by viscous boundary layers.
study done by Virella et al. [33], the influence of nonlinear wave Basically, the damping factor depends on the liquid height, liquid
theory on the sloshing natural periods and their modal pressure kinematic velocity and tank dimensions. From this point of view,
distribution for rectangular tanks with HL /Lx ratios ranging from evaluation of damping characteristic for a fluid–tank system needs
0.4 to 1.65 was investigated. They concluded that the nonlinearity more considerations. However, due to lack of sufficient data in this
of the surface wave does not have a major effect on the pressure field, the classical damping scheme is used in the finite element
distribution on the walls and on natural sloshing frequencies. In model. Considering impulsive and convective parts of liquid do-
the present study, two different tank configurations namely a main, damping matrix can be given as
shallow and a tall tank are analyzed as will be discussed later.
The ratios of HL /Lx are 0.37 and 1.26 for the shallow and tall [Cf ] = a[G] + b[H ]. (8)
tanks, respectively. In this case, the linearized boundary condition In which a and b are computed by Rayleigh damping method.
is appropriate particularly for practical applications. Neglecting the In this equation, coefficient a is calculated based on fundamental
gravity wave effects leads to the free surface boundary condition frequency of liquid sloshing to present the convective part of the
which is appropriate for impulsive motion of liquid. The related response while, coefficient b is computed based on impulsive fun-
governing equation is given as damental frequency of the tank–liquid system which is obtained
P (x, y, Hl , t ) = 0 (4) by modal analysis. The fundamental frequencies of tank–liquid
system are calculated using both finite element method and an-
where Hl is the height of liquid in the container. alytical equations adopted from ACI 350.3-06 [34]. The convective
Using finite element discretization and discretized formulation frequencies are exactly the same but a significant difference is seen
of Eq. (1), the wave equation can be written as the following matrix for impulsive values. In this study, the values corresponding to fi-
form: nite element method is used to calculate the damping coefficients
which will be discussed later.
[G]{P̈ } + [H ]{P } = {F }. (5)
∑ e ∑ e ∑ e Sloshing in a tank without any anti-sloshing device is usually
In which Gi,j = Gi,j , Hi,j = Hi,j and Fi = Fi . The coef- damped by viscous forces. One more complete investigations
ficients Gei,j , Hie,j and Fie for an individual element are determined which have been done by Mikishev and Dorozhkin [35] showed
using the following expressions: that in a storage tank with rational dimensions viscous damping is
∫ less than 0.5%.
1 In the proposed FE procedure, Rayleigh damping as mentioned
Gei,j = Ni Nj dA
previously is used in the direct step-by-step integration method.
∫g 
Ae
∂ Ni ∂ Nj ∂ Ni ∂ Nj ∂ Ni ∂ Nj

The stiffness proportional damping equivalent to 5% of critical
Hie,j = + + dV (6) damping is assumed as structural damping. For sloshing behavior
V
∫ e ∂ x ∂ x ∂ y ∂ y ∂z ∂z
e ∂P of liquid 0.5% of critical damping is applied. Damping values
Fi = Ni dA associated with horizontal and vertical motions are dependent on
∂n
Ae
earthquake shear wave velocity and the ratio of HL /R in which R is
where Ni is the shape function of the ith node in the liquid element. the equivalent radius of a cylindrical tank with the same plan area
Ae and Ve are the integration over side and area of the element, re- as the rectangular tank. These values can be drawn based on studies
spectively. done by Veletsos and Tang [10] and Veletsos and Shivakumar [36]
 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200 2189

Fig. 1. Viscous boundary condition in the three-dimensional finite element model [37].

for each tank configuration. In this study, the value of 5% of critical
damping is applied for liquid domain of shallow and tall tank model
as a conservative damping ratio.
When modeling a dynamic problem involving soil–structure in-
teraction, particular attention must be given to the soil bound-
ary conditions. Ideally, infinite boundary condition should be
surrounding the excited zone. Propagation of energy will occur
from the interior to the exterior region. Since the exterior region
is nonreflecting, it absorbs all of the incoming energy. Yet, in a
finite element analysis, we are constrained into applying finite
size boundaries for the foundations. Those boundaries in turn will
reflect the elastic waves which is contrary to the physics of the
problem.
In this study, a viscous boundary method is used in three-
dimensional space which was successfully employed in the FE
modeling of the elevated liquid tanks done by Livaoglu and
Dogangun [37]. The detailed formulation can be found in the work
done by Lysmer and Kuhlmeyer [30].
These viscous boundaries can be used with the FE mesh as
shown in Fig. 1 for a generalized three-dimensional model [37].
In this figure, An , At1 and At2 are the fields controlling the viscous
dampers, σ and τ are the normal and shear stresses occurring
in the boundaries of the medium, and the subscripts n and t
represent normal and tangent directions in the boundary. When
the viscous boundary is taken into consideration, an additional
damping matrix is applied to the system as follows:
An ρvp 0 0
 
∗
[Ci ] = 0 At1 ρvs 0 (9)
0 0 At2 ρvs
and is easy to use within an FE method for analyzing a linear system
in time domain. On this basis, to account for the soil–structure
interaction, it is necessary to apply the inertial loads only to the
structure not to the soil foundation. The far-field boundaries are
treated using viscous dampers to absorb the wave energy reflected
by the system.
Most computer commercial programs automatically apply
the seismic loading to all mass degrees of freedom within the
computer model and cannot solve the SSFI problem. This lack
of capability has motivated the development of the massless
foundation model in which the inertia forces within the foundation
material are neglected. Most of the studies done on the behavior of
liquid storage tanks including soil–structure interaction are based
on this approximation.
A parametric study was carried out to determine the dimen-
sions of the soil foundation in such a way that the displacements of
nodes on the lateral boundaries are almost zero. It should be noted
that the stress and strain due to liquid–tank weight are considered
in the soil foundation.
3. Finite element implementation
In this study, two different model configurations associated
with shallow and tall tanks are investigated in three-dimensional
space. A schematic view of rectangular tank model is shown
in Fig. 2. These tanks have also been used in some previous
investigations by Kianoush and Chen [19], Chen and Kianoush [18]
and Kim et al. [13]. The dimensions and properties of shallow and
tall tank are as follows:
Shallow Tank:

where vp and vs are dilatational and shear wave velocity of the ρw = 2300 kg/m3 ρl = 1000 kg/m3 Ec = 26.44 GPa
considered medium. It should be noticed that in two-dimensional ν = 0.17
analysis, the third row and column are eliminated. Lx = 15 m Lz = 30 m Hw = 6.0 m Hl = 5.5 m
It is assumed that the tank structure is anchored to the foun- tw = 0.6 m.
dation. Considering the effects of soil–structure interaction, Eq. (7)
can be re-written as Tall Tank:

([M ] + [Mi∗ ]){Ü } + ([C ] + [Ci∗ ]){U̇ } + ([K ] + [Ki∗ ]){U } ρw = 2300 kg/m3 ρl = 1000 kg/m3 Ec = 20.77 GPa
= {F1 } + [Q ]{P } (10) ν = 0.17
[G]{P̈ } + [C ′ ]{Ṗ } + [H ]{P } = {F2 } − ρ[Q ]T {Ü } Lx = 9.8 m Lz = 28 m Hw = 12.3 m Hl = 11.2 m

where [Ci∗ ] and [Ki∗ ] are the damping and stiffness matrices asso-
ciated with the foundation–structure interaction.
The most common soil–structure interaction (SSI) approach,
used for three-dimensional soil–structure systems, is based on
the ‘‘added motion’’ formulation by Clough and Penzien [28]. This
approximation is mathematically simple, theoretically accurate,
tw = 1.2 m.
It is assumed that the tank is anchored at its base and the effects of
uplift pressure are not considered.
In this study, an eight-node isoparametric element with three
translational degrees of freedom in each node is used in the finite
element procedure to model the tank walls and the base slab.
2190 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200

Fig. 2. Schematic configuration of a rectangular liquid tank.

Fig. 3. Finite element model of a three-dimensional rectangular tank: (a) tall tank model, (b) shallow tank model, and (c) viscous boundary condition.

The liquid domain is modeled using eight-node isoparametric fluid
elements with pressure degree of freedom in each node. The finite
element (FE) model configurations for both shallow and tall tanks
are shown in Fig. 3.
The longitudinal, transversal and vertical components recorded
for 1994 Northridge, 1940 El-Centro, 1971 San-Fernando and
1957 San-Francisco earthquakes are used as excitations of the
tank–liquid system. The components are scaled in such a way
that the peak ground acceleration in the longitudinal direction
reach 0.4 g, as shown in Fig. 4. Based on the ratio of peak ground
acceleration (PGA) to peak ground velocity (PGV), the Northridge
record is considered as low frequency ground motion; El-
Centro and San-Fernando records are categorized as intermediate
frequency excitations and San-Francisco record is considered as
high frequency content excitation.
4. Mesh sensitivity and error estimation
The first step in a finite element analysis is to estimate the
possible error by selecting a mathematical model to represent the
object being analyzed. The mathematical target for this study was
selected as the maximum pressure at the bottom of rigid tank using
equations derived by Haroun [7].
The analysis goal is to compute the exact solution of bottom
pressure named PEX and then calculate its value using the FEM
which is PFE . The term PEX depends only on the definition of
the mathematical model and not on the method used for finding
an approximate solution. Therefore, it does not depend on mesh
quality, type, and size of elements. The difference between PEX and
the physical property it represents is called the modeling error. As
a result, discretization error is defined by
 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200 2191

Fig. 4. Scaled longitudinal components of earthquake records: (a) 1994 Northridge (b) 1940 El-Centro (c) 1971 San-Fernando (d) 1957 San-Francisco.

Fig. 5. Finite element discretization error: (a) Rigid shallow tank model (b) Rigid tall tank model.

(PEX ) − (PFE ) A comparison between these results is shown in Table 1

e= . (11)
(PEX )
The variations of ratio of FE bottom pressure to the analytical
pressure with the number of mesh divisions are shown in Fig. 5(a)
and (b) for both two-dimensional rigid shallow and tall tank
models, respectively. It is found that the discretization errors are
equal to 1.66% and 0.03% for proposed shallow and tall tank models
respectively when subjected to an earthquake ground motion. On
this basis, the selected mesh pattern for this analysis is appropriate
to investigate the fluid behavior. Since both structure and soil
medium are linear, a very refined mesh is not necessary for the
analysis.
5. Spectrum analysis
Prior to conducting the time history analyses, the fundamental
periods of impulsive and convective behaviors as well as their
related mass ratios are calculated using both finite element and
analytical method for shallow and tall tank models. The analytical
fundamental frequencies and related mass ratios are obtained in
accordance with ACI 350.3-06 [34].
which indicates that the FE results are in agreement with
analytical values. Further details on the calculation of sloshing
frequencies and masses are given by Patkas and Karamanos [38]
and Karamanos et al. [39]. Using the design spectral accelerations,
the values of base shear and base moment can be easily calculated
for unit length of wall. A comparison between spectral results with
those calculated in accordance with ACI 350.3-06 [34] and the FE
time history will be discussed later.
6. Time history analysis
Two rectangular concrete liquid container models given in Fig. 2
are used basically for the example analyses in time domain. It
should be noted that both longitudinal and transversal compo-
nents of earthquake perpendicular to longer and shorter tank
walls (X and Y directions) respectively are applied simultaneously
in three-dimensional modeling and are referred to as horizontal
excitation.
Since the sloshing height variation is a matter of concern in
design of rectangular tanks, its value is measured at three different
locations (points A, B and C shown in Fig. 3) of three-dimensional
model. Points A, B and C are located at the fluid surface at the
2192 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200

Fig. 6. Time history of base shear response of a shallow tank model under longitudinal excitation: (a) Northridge (b) El-Centro (c) San-Fernando (d) San-Francisco.

Fig. 7. Time history of sloshing height due to all components of earthquake for a shallow tank model: (a) Northridge (b) El-Centro (c) San-Fernando (d) San-Francisco.

middle of longer length, middle of shorter length and the corner
of tank wall, respectively.
Finally, seismic analyses are performed using both horizontal
and vertical components of ground acceleration and the results are
compared with those calculated by spectral analyses.
The results of this study are presented in two parts. In the first
part, a detailed discussion on the effect of earthquake frequency
content on the seismic behavior of liquid tanks supported on a rigid
foundation is described.
In the second part the combined effects of SSI and earthquake
frequency content on the dynamic behavior of liquid tanks is
discussed.
6.1. Effect of earthquake frequency content on the dynamic behavior
of liquid tanks
Seismic behavior of shallow tank model with rigid base
The transient base shear and base moment for flexible shallow
tank model due to horizontal and vertical excitations are calculated
 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200
Fig. 8. Time history of base moment response of a tall tank model under longitudinal excitation: (a) Northridge (b) El-Centro (c) San-Fernando (d) San-Francisco.
Table 1
Modal analysis results.
Tank type FEM
Fundamental period (s) Effective mass ratio
Ti Tc mi /m
Shallow tank 0.15 8.58 0.18
Tall tank 0.33 5.30 0.60
by the proposed method. Four different earthquake records are
applied to investigate the effect of frequency content on the
response of the tank–liquid system. The base shear response of the
tank due to longitudinal ground motion measured at the middle
cross-section of the longer wall is presented in Fig. 6.
The absolute maximum values of resulting base shear and
base moment due to all ground excitations are presented in
Table 2. In addition, the results are presented in brackets in
terms of normalized values with respect to those of the El-
Centro record. These results show that the responses due to
high frequency content earthquake of San-Francisco are highly
magnified because of the similarity between impulsive dynamic
characteristics of the tank–liquid system and earthquake record.
Also, the least response values are obtained under low frequency
Northridge earthquake which are almost one-third of the values
due to San-Francisco earthquake. Unlike the impulsive behavior,
the convective response values are amplified due to intermediate
frequency content earthquake of El-Centro. In this case, the
absolute peak values of convective base shear and base moment are
presented for all records in Table 2. It should be noticed that when
the impulsive response reaches its peak, the convective response is
at the beginning stage and has not yet fully developed. As a result,
the convective component does not have a major effect on the total
structural response in time-domain analysis.
It is found that the effect of earthquake frequency content is
significant on the structural response of liquid–tank models and
may cause a considerable increase in time-domain peak response
values.
Considering the combined effect of vertical and horizontal
ground motions, the impulsive response almost remains un-
changed for all earthquakes. The convective part increases due to
2193
Analytical
Fundamental period (s) Effective mass ratio
mc /m Ti Tc mi /m mc /m
0.72 0.10 8.56 0.21 0.75
0.33 0.22 5.50 0.59 0.44
both Northridge and El-Centro vertical components but remains
unchanged under applying vertical components of San-Fernando
and San-Francisco ground motions.
In addition to the structural response, the fluid dynamic
behavior is thoroughly investigated. The time history diagrams of
surface sloshing height are shown in Fig. 7.
As previously mentioned, these values are measured at points
A, B and C. The maximum sloshing height which occurs at point
C have values of 325 mm, 838 mm, 311 mm and 138 mm
under Northridge, El-Centro, San-Fernando and San-Francisco
earthquakes, respectively.
As presented in Table 2, the vertical acceleration has the most
significant effect on sloshing behavior of fluid domain under the El-
Centro earthquake and causes an increase of about 14% in sloshing
height whereas, the sloshing height remains unchanged after
applying vertical motions of the Northridge and San-Fernando
earthquakes.
Unlike the convective behavior, the impulsive behavior is less
sensitive to vertical motions as seen for all earthquake records.
Since the impulsive response values are so much higher than
those of convective, it can be concluded that the effect of vertical
excitation is insignificant on overall seismic behavior of shallow
tank model.
7. Seismic behavior of tall tank model with rigid base
Fig. 8 presents the diagrams of impulsive structural response
in terms of base moment time history for tall tank model.
The absolute maximum values of the resulting base shear and
base moment of the impulsive and convective components are
presented in Table 3. As previously seen in shallow tank model, the
2194 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200

Table 2
Summary of maximum dynamic responses of a shallow tank model.
Impulsive response Convective response
H H +V (H + V )/H H H +V (H + V )/H
Base shear (kN/m) 102 (0.53) 104 (0.54) 1.02 16 (0.44) 17 (0.42) 1.10
Northridge Base moment (kN m/m) 273 (0.57) 284 (0.59) 1.04 36 (0.61) 37 (0.58) 1.03
Sloshing (mm) – – – 326 (0.44) 325 (0.39) 1.00
Base shear (kN/m) 191 (1.00)a 193 (1.00) 1.01 36 (1.00) 40 (1.00) 1.11
El-Centro Base moment (kN m/m) 479 (1.00) 478 (1.00) 1.00 59 (1.00) 64 (1.00) 1.08
Sloshing (mm) – – – 734 (1.00) 838 (1.00) 1.14
Base shear (kN/m) 236 (1.24) 236 (1.22) 1.00 12 (0.33) 12 (0.30) 1.00
San-Fernando Base moment (kN m/m) 657 (1.37) 645 (1.35) 0.98 30 (0.51) 30 (0.47) 1.00
Sloshing (mm) – – – 311 (0.42) 311 (0.37) 1.00
Base shear (kN/m) 323 (1.69) 337 (1.75) 1.04 5 (0.14) 5 (0.13) 1.00
San-Francisco Base moment (kN m/m) 821 (1.71) 860 (1.80) 1.05 13 (0.22) 13 (0.20) 1.00
Sloshing (mm) – – – 128 (0.17) 138 (0.16) 1.08
a
The values in the bracket are normalized responses with respect to El-Centro results.

Fig. 9. Time history of sloshing height due to all components of earthquake for a tall tank model: (a) Northridge (b) El-Centro (c) San-Fernando (d) San-Francisco.

Table 3
Summary of maximum dynamic responses of a tall tank model.
Impulsive response Convective response
H H +V (H + V )/H H H +V (H + V )/H
Base shear (kN/m) 700 (0.83) 655 (0.78) 0.94 279 (5.58) 266 (4.43) 0.95
Northridge Base moment (kN m/m) 3295 (0.77) 2994 (0.70) 0.91 1131 (5.36) 1087 (4.71) 0.96
Sloshing (mm) – – – 1968 (2.14) 1957 (1.73) 0.99
Base shear (kN/m) 846 (1.00)a 839 (1.00) 0.99 50 (1.00) 60 (1.00) 1.2
El-Centro Base moment (kN m/m) 4261 (1.00) 4251 (1.00) 0.99 211 (1.00) 231 (1.00) 1.09
Sloshing (mm) – – – 920 (1.00) 1134 (1.00) 1.23
Base shear (kN/m) 814 (0.96) 908 (1.08) 1.11 48 (0.96) 49 (0.82) 1.02
San-Fernando Base moment (kN m/m) 4059 (0.95) 4119 (0.97) 0.98 245 (1.16) 244 (1.06) 1.00
Sloshing (mm) – – – 973 (1.06) 965 (0.85) 0.99
Base shear (kN/m) 429 (0.51) 483 (0.58) 1.13 10 (0.2) 11 (0.18) 1.10
San-Francisco Base moment (kN m/m) 2369 (0.56) 2606 (0.61) 1.10 56 (0.26) 60 (0.26) 1.08
Sloshing (mm) – – – 211 (0.23) 254 (0.22) 1.20
a
The values in the bracket are normalized responses with respect to El-Centro results.

effect of frequency content is significant on the structural response the El-Centro earthquake which has the nearest frequency to the
of the tank. In this case, the structural response is amplified under impulsive fundamental frequency of the tank–liquid system.
 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200
Fig. 10. Impulsive pressure distribution along height of a three-dimensional tall
tank model measured at the middle section of longer wall under longitudinal
excitations for different earthquake records.
Considering the free surface motion, the absolute maximum
values of convective base shear and base moment are calculated
and shown in Table 3. It is obvious that applying Northridge ground
motion results in the highest convective responses. Under applying
three components of ground motion, the structural response of the
tank is affected by the frequency of the earthquake. In comparison
with the El-Centro record, the Northridge vertical component leads
to a decrease in peak values of base shear and moment while the
San-Francisco earthquake results in an increase in these values.
In comparison with shallow tank, the tall tank is more sensitive
to vertical component of ground motion and its response varies
from one earthquake to another.
The time history diagrams of sloshing height due to all
components of earthquake are shown in Fig. 9 for all records.
Similar to shallow tank, the maximum sloshing height occurs at
point C which is almost 20% higher than its value measured at
point A. Applying vertical acceleration may lead to an increase
or decrease in sloshing height or in some cases may not affect
the sloshing height as seen for Northridge and San-Fernando
Fig. 11. Comparisons of peak base shear responses of shallow tank models for different soil types: (a) Northridge (b) El-Centro (c) San-Fernando (d) San-Francisco.
2195
earthquakes. This variation is affected by the characteristics of the
ground motion and tank configuration.
The impulsive pressure distribution along tank wall measured
at the middle section of longer wall is presented in Fig. 10 for
all records under longitudinal ground motions. Higher values
are obtained due to the El-Centro earthquake as compared to
other earthquake records. This is consistent with other structural
response values as presented in Table 3.
7.1. Effect of soil–structure interaction on the dynamic behavior of
liquid tank
Response of shallow tank with flexible foundation
As mentioned before, in order to consider the effects of
deformability of tank foundation on impulsive and convective
response of tank structure, additional FE models with flexible
foundation boundary condition are investigated in this study.
To evaluate the dynamic response of liquid tank supported on
flexible foundation, six soil types recommended in the literature
and design codes are considered. The soil properties are shown in
Table 4. In this table, S1 , S2 , S3 , S4 , S5 and S6 soil types represent
hard rock, rock, very dense soil and soft rock, stiff soil, soft clay soil
and soils vulnerable to potential failure or collapse under seismic
loading, such as liquefiable soils and quick and highly sensitive
clays, respectively.
A comparison among peak values of impulsive base shear of
shallow tank supported on different soil types is presented in
Fig. 11 under different ground motions.
It is clear that the effect of soil–structure interaction (SSI) on
the structural response of liquid tanks is highly dependent on
the earthquake frequency content. SSI has led to an increase of
about 47%, 64% and 52% in peak base shear response in comparison
with rigid foundation condition under Northridge, El-Centro and
San-Fernando records, respectively. However, a decrease of about
2196 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200

Fig. 12. Comparisons of peak sloshing heights of shallow tank models for different soil types: (a) Northridge (b) El-Centro (c) San-Fernando (d) San-Francisco.

Table 4 In this case, the structural responses under moderate and high
Properties of the soil types considered in this study. frequency content earthquakes of El-Centro, San-Fernando and
Soil E (kN/m2 ) G(kN/m2 ) γ (kg/m3 ) ν vs (m/s) vp (m/s) San-Francisco decrease as the foundation soil becomes softer.
types It is clear that the response amplification or reduction pattern
S1 7000,000 2692,310 2000 0.30 1149.10 2149.89 due to deformable foundation is highly dependent on the nature of
S2 2000,000 769,230 2000 0.30 614.25 1149.16 earthquake.
S3 500,000 192,310 1900 0.35 309.22 643.68 To clarify the changes of structural response due to soil stiffness
S4 150,000 57,690 1900 0.35 169.36 352.56 variation, the deviations of the base shear forces in time due to the
S5 75,000 26,790 1800 0.40 120.82 295.95
El-Centro earthquake are illustrated and compared among S1, S5
S6 35,000 12,500 1800 0.40 82.54 202.18
and S6 soil types for tall tank models in Fig. 14. These time history
responses have different frequency characteristics and describe
13% is seen for base shear peak value due to the San-Francisco different behavior of the tanks.
earthquake. In addition, the power spectral density (PSD) functions for four
The response amplification has occurred in different soil types ground motions are shown in Fig. 15. The center of the mass of PSD
as the input earthquake changes. For moderate frequency content functions are located at 2.04 Hz, 3.20 Hz, 3.99 Hz and 6.21 Hz for
records, the highest increase is seen in S4 and S5 soil types Northridge, El-Centro, San-Fernando and San-Francisco records,
while under low frequency content record, the highest values are respectively. These values correspond with natural periods of
0.49 s, 0.31 s, 0.25 s and 0.16 s, respectively. According to the PSD
obtained for S6 soil type which represents a very soft soil.
concept, the significant amount of energy of the ground motion is
Under high frequency content record of San-Francisco, the peak
concentrated at the center of the mass of the PSD function. As a
structural responses decrease as the soil stiffness decreases. On the
result, the maximum response of the structure is expected to occur
other hand, the lowest peak values are obtained for S6 soil type. under the ground motion record which has the nearest center of
For all cases, it is found that the soil type has an insignificant mass location to the fundamental natural period or frequency of
effect on the convective behavior of liquid tank. A comparison the structure.
among maximum sloshing height measured at the right top corner The modal analysis results of tank–liquid–foundation system
of the liquid domain for different soil types and earthquake records are shown in Table 5 for both shallow and tall tanks. It is clear that
is shown in Fig. 12 under horizontal excitation. It is concluded the SSI significantly affects the impulsive fundamental periods. On
that the foundation deformability does not significantly affect the the other hand, as the soil stiffness decreases, the impulsive period
sloshing height of the liquid tank. increases. This phenomenon has also been reported by Larkin [40].
However, convective fundamental periods are almost independent
of the flexibility of the foundation. This justifies the insensitivity
8. Response of tall tank with flexible foundation
of the sloshing height to the variation of foundation properties.
This might be due to the long period nature of sloshing waves as
Fig. 13 shows the effect of different soil types under different presented in Table 5.
ground motions on base moment of tall tank models. Except It is clear that the variation in structural response is not
for Northridge earthquake, a significant reduction in structural only dependent on the soil properties but also dependent on the
response is seen for tanks supported on S6 soil type. These earthquake characteristics.
reduction values are about 11%, 26%, 28% and 47% for Northridge,
El-Centro, San-Fernando and San-Francisco records compared to 9. Comparison with other methods
rigid base, respectively. However, an amplification of structural
response occurs in S4 soil type under Northridge earthquakes in The seismic impulsive and convective responses of liquid–tank
comparison with rigid foundation condition. models are obtained in this study as discussed previously
 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200 2197

Fig. 13. Comparisons of peak base moment responses of tall tank models for different soil types: (a) Northridge (b) El-Centro (c) San-Fernando (d) San-Francisco.

Fig. 14. Time history of impulsive base shear for a tall tank model with flexible foundation under horizontal excitation of the El-Centro earthquake for different soil types:
(a) S1 (b) S5 (c) S6.

Table 5 are calculated. The mapped design spectral response acceleration

Modal analysis results of soil–structure–liquid system.
at short periods are 1.50 g, 2.20 g, 2.63 g and 2.75 g for El-
Impulsive Convective Centro, San-Francisco, San-Fernando and Northridge earthquakes,
fundamental period fundamental period respectively. The mapped MCE spectral response accelerations at
(s) (s)
a period of 1 s are 0.60 g, 0.90 g, 0.92 g and 1.0 g for the El-
S1 S2 S3 S4 S5 S6 S1 S2 S3 S4 S5 S6
Centro, San-Francisco, San-Fernando and Northridge earthquakes,
Shallow 0.15 0.14 0.16 0.22 0.29 0.41 8.58 8.58 8.58 8.57 8.55 8.49 respectively. A comparison between the FE and spectral results
tank
based on the approximation used by ACI 350.3-06 [34] are shown
Tall tank 0.33 0.34 0.38 0.52 0.63 0.87 5.13 5.13 5.13 5.13 5.12 5.09
in Figs. 16 and 17 for shallow and tall tank models under different
ground motions. It should be noted that in the ACI method, the
considering four earthquake records with different frequency hydrodynamic pressure is the product of the first fundamental
contents and accounting for foundation deformability, wall impulsive or the convective mass and the period-dependent
flexibility and fluid damping properties. spectral amplification. On the other hand, the equivalent forces
In order to verify the proposed FE method as well as to consider exerted by the fluid can be obtained using the following equation:
the effect of fluid damping properties, two different conditions Pi = Ci Wi (12)
with zero and non-zero fluid damping ratios were used for
rigid tank models and the results were compared with analytical Pc = Cc Wc . (13)
solutions derived by Haroun [7]. These results were presented in In which Pi and Pc are impulsive and convective forces, Ci and Cc
terms of impulsive hydrodynamic pressure over the tank height are the seismic impulsive and convective coefficients and Wi and
in previous research done by the authors [41]. The FE pressure Wc are impulsive and convective weights.
distribution is found to be in agreement with analytical results Although higher values are obtained using ACI method, the
when the fluid damping is ignored. same trend due to earthquake frequency content is seen. For both
Using the design earthquake response spectrum which is shallow and tall tank models the ACI method results in higher
commonly used in practice, the impulsive and convective base impulsive response. The maximum difference in results between
shear and base moment of both shallow and tall tank models the FE and design code method is about 22%. Overall, it is clear
2198 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200

Fig. 15. PSD function of: (a) Northridge (b) El-Centro (c) San-Fernando (d) San-Francisco.

Fig. 16. Impulsive and convective structural responses of a shallow tank model: (a) peak impulsive base shear (b) peak impulsive base moment (c) peak convective base
shear (d) peak convective base moment.

that the effect of earthquake frequency content can be predicted
accurately using both methods. However, the values calculated
based on the ACI code are more conservative.
Considering the SSI effect, different conclusions are reported
in the literature. Veletsos and Tang [42] and Veletsos et al. [12]
showed that soil–structure interaction may reduce significantly
the critical responses of shallow tanks, but may increase those of
tall, stiff tanks that have high fundamental natural frequencies.
Also, it was shown that the soil–structure interaction has a
negligible effect on the convective components.
 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200
Fig. 17. Impulsive and convective structural responses of a tall tank model (a) peak impulsive base shear (b) peak impulsive base moment (c) peak convective base shear
(d) peak convective base moment.
Moreover, Haroun and Abou-Izzeddine [11] showed that
interaction of the tank and foundation soil magnifies the tank
response, and that the magnification is a factor of both shear wave
velocity of the soil as well as the geometric properties of the tank.
Livaoglu [21] concluded that the displacements and base shear
forces generally decrease, with decreasing soil stiffness. However,
the sloshing response was not practically affected by soil–structure
interaction in frequency domain which is similar to current FE
results.
Also, Cho et al. [43] obtained the same results as Livaoglu [21]
showed that structural response values are reduced as soil stiffness
decreases. The liquid sloshing height and motion were not affected.
All of these studies show that the convective term of dynamic
response of liquid tank is not sensitive to soil properties. The same
results are obtained using the FEM in this study. However, for
both tank configurations, the soil–structure interaction effect is
highly dependent on the earthquake frequency content and may
amplify or reduce the structural response. On the other hand, a
unique pattern which is able to reflect the effect of soil–structure
interaction on tank behavior cannot be concluded from the results
of this study.
10. Conclusions
In this study, a finite element method is introduced that
can be used for the analysis of dynamic behavior of partially
filled rectangular fluid container under horizontal and vertical
ground excitations in three-dimensional space. The liquid sloshing
is modeled using an appropriate boundary condition and the
damping effects due to impulsive and convective components of
the stored liquid are modeled using the Rayleigh method. The
soil foundation is modeled as an elastic homogeneous medium
with viscous boundary condition applied on the truncated zone to
simulate the wave energy absorption. Two different configurations
including shallow and tall tank models are considered to
investigate the effect of geometry on the response of the
liquid–structure system in time domain. Effect of wall flexibility
on the dynamic response of system is taken into account. Four
different ground motions with the same peak ground acceleration
2199
are applied to investigate the effect of earthquake frequency
content on the seismic behavior of liquid–tank system.
The effect of foundation deformability on the overall dynamic
response of the system is investigated by comparing the results
among six different soil types.
The results are presented in terms of the maximum structural
base shear and base moment obtained from time history analysis of
the considered system as well as pressure distribution and surface
sloshing heights for different seismic excitations. It is obvious that
the records with frequency characteristics close to those of liquid
tanks highly magnify the responses of the system. Assuming a
rigid foundation, the high frequency earthquakes result in the
highest impulsive response in shallow tank model, whereas the
intermediate frequency earthquakes highly amplify the tall tank
response. Due to significant difference between impulsive and
convective fundamental frequencies, a different trend is observed
for the convective response under the same earthquakes. It should
be noticed that due to the high magnitude of impulsive response
and a significant time lag between peak impulsive and convective
response, the overall seismic behavior of the tank is governed by
the impulsive component.
In this study, the FE results are compared with those obtained
by spectral analysis based on design earthquake response spectra.
Although the spectral values are higher than the FE results, the
same trend due to earthquake frequency content is seen for both
methods.
Considering the effect of SSI, the results show that the maxi-
mum impulsive base shear and base moment obtained from the
time history analysis of the considered system may increase or de-
crease as the soil stiffness changes which is a result of dynamic
pressure variation in the middle of the wall due to the rocking mo-
tion of the foundation. This phenomenon is highly dependent on
earthquake frequency content and tank configuration.
A unique trend is seen under low frequency content earthquake
for both shallow and tall tank configurations. In this case, the
structural responses increase as the soil stiffness increases.
In addition, the convective response is almost independent of
variations of flexibility of the foundation and seems to be related
to geometric configurations of tank, earthquake characteristics and
liquid properties.
2200 M.R. Kianoush, A.R. Ghaemmaghami / Engineering Structures 33 (2011) 2186–2200
It is clear that the dynamic behavior of liquid concrete tanks
depends on a wide range of parameters such as seismic properties
of earthquake, tank dimensions and fluid–structure interaction
which should be considered in current codes of practice. This study
shows that the proposed FE method can be used in the time history
analysis of rectangular liquid tanks.
The present study is done based on the three-dimensional
analysis of rectangular tanks in time domain using four different
earthquake records and six different soil types. Although the
liquid–tank design procedures are based on simplified frequency-
based methods, the current FE time history results which are more
realistic can be used to verify these methods in future works.
References
[1] Housner GW. The dynamic behavior of water tanks. Bull Seismol Soc Amer
1963;53(2):381–7.
[2] Yang JY. Dynamic behavior of fluid–tank systems. Ph.D. thesis, Department of
Civil Engineering, Houston (TX): Rice University; 1976.
[3] Veletsos AS, Yang JY. Earthquake response of liquid storage tanks-advances
in civil engineering through mechanics. In: ASCE proceedings of the second
engineering mechanics specially conference, Raleigh (NC). 1977. p. 1–24.
[4] Minowa C. Dynamic analysis for rectangular water tanks. Recent Adv Lifeline
Earthq Eng Japan 1980;135–42.
[5] Minowa C. Experimental studies of seismic properties of various type water
tanks. In: Proceedings of eighth WCEE. San Francisco. 1984. p. 945–52.
[6] Chopra AK, Gupta S. Hydraudynamic and foundation interaction effects in
earthquake response of a concrete gravity dam. J Struct Div 1981;107(8):
1399–412.
[7] Haroun MA. Stress analysis of rectangular walls under seismically induced
hydrodynamic loads. Bull Seismol Soc Amer 1984;74(3):1031–41;
Haroun MA, Ellaithy MH. Seismically induced fluid forced on elevated tanks. J
Tech Topics Civ Eng 1984;111:1–15.
[8] Haroun MA. Vibration studies and tests of liquid storage tanks. Earthq Eng
Struct Dyn 1983;11:190–206.
[9] Haroun MA, Tayel MA. Response of tanks to vertical seismic excitations. Earthq
Eng Struct Dyn 1985;13:583–95.
[10] Veletsos AS, Tang Y. Dynamics of vertically excited liquid storage tanks. ASCE
J Struct Eng 1986;112(6):1228–46.
[11] Haroun MA, Abou-Izzeddine W. Parametric study of seismic soil–tank
interaction. II: vertical excitation. ASCE J Struct Eng 1992;118(3):783–97.
[12] Veletsos AS, Tang Y, Tang HT. Dynamic response of flexibly supported liquid
storage tanks. ASCE J Struct Eng 1992;118:264–83.
[13] Kim JK, Koh HM, Kwahk IJ. Dynamic response of rectangular flexible fluid
containers. ASCE J Eng Mech 1996;122(9):807–17.
[14] Park JH, Koh HM, Kim J. Liquid–structure interaction analysis by coupled
boundary element — finite element method in time domain. In: Brebbia CA,
Ingber. MS, editors. Proceedings of the 7th international conference on bound-
ary element technology. Southampton, (UK): Computational Mechanics Pub-
lication BE-TECH/92, Computational Mechanics Publications; 1992. p. 89–92.
[15] Subhash S, Bhattachryya SK. Finite element analysis of fluid–structure
interaction effect on liquid retaining structures due to sloshing. Comput Struct
1996;59(6):1165–71.
[16] Koh HM, Kim JK, Park JH. Fluid–structure interaction analysis of 3D rectangular
tanks by a variationally coupled BEM–FEM and comparison with test results.
Earthq Eng Struct Dyn 1998;27:109–24.
[17] Doganun A, Durmus A, Ayvaz Y. Earthquake analysis of flexible rectangular
tanks using the lagrangian fluid finite element. Euro J Mech-A/Solids 1997;16:
165–82.
[18] Chen JZ, Kianoush MR. Seismic response of concrete rectangular tanks for
liquid containing structures. Canad J Civ Eng 2005;32:739–52.
[19] Kianoush MR, Chen JZ. Effect of vertical acceleration on response of concrete
rectangular liquid storage tanks. Eng Struct 2006;28(5):704–15.
[20] Kianoush MR, Mirzabozorg H, Ghaemian M. Dynamic analysis of rectangular
liquid containers in three-dimensional space. Canad J Civ Eng 2006;33:
501–7.
[21] Livaoglu R. Investigation of seismic behavior of fluid-rectangular tank–
soil/foundation systems in frequency domain. Soil Dyn Earthq Eng 2008;28(2):
132–46.
[22] Jennings PC, Kuroiwa JH. Vibration and soil–structure interaction tests of
a nine-story reinforced concrete building. Bull Seismol Soc Amer 1968;58:
891–916.
[23] Foutch DH, Housner GW, Jennings PC. Dynamic responses of six multistory
buildings during the San Fernando earthquake. Report No. EERL 75-02,
California Institute of Technology; 1975.
[24] Mcverry GH. Frequency domain identifications of structural models from
earthquake records. Report No. EERL 79-02, California Institute of Technology;
1979.
[25] Chopra AK. Dynamics of structures. Upper Saddle River (NJ): Prentice-Hall;
1995.
[26] Wolf JP, Song CH. Doubly asymptotic multi-directional transmitting boundary
for dynamic unbounded medium-structure-interaction analysis. Earthq Eng
Struct Dyn 1995;24(2):175–88.
[27] Wolf JP, Song CH. Finite element modeling of unbounded media. In: The 11th
world conference on earthquake engineering, 1996, 70, p. 1–9.
[28] Clough R, Penzien J. Dynamics of structures. 2nd ed. New York: McGraw-Hill,
Inc; 1993.
[29] Wilson EL. Three-dimensional static and dynamic analysis of structures—a
physical approach with emphasis on earthquake engineering. 3rd ed. Berkeley
(CA, USA): Computers and Structures, Inc.; 2002.
[30] Lysmer J, Kuhlmeyer RL. Finite dynamic model for infinite media. ASCE Eng
Mech Div J 1969;95:859–77.
[31] Song CH, Wolf JP. Dynamic stiffness of unbounded medium based on solvent
damping extraction method. Earthq Eng Struct Dyn 1994;23:1073–86.
[32] Anrade WP. Implementation of second order absorbing boundary condition in
frequency domain computations, Ph.D. thesis. Texas: The University of Texas
of Austin; 1999.
[33] Virella JC, Prato CA, Godoy LA. Linear and nonlinear 2D finite element analysis
of sloshing modes and pressure in rectangular tanks subjected to horizontal
harmonic motions. J Sound Vibration 2008;312(3):442–60.
[34] ACI 350.3-06. Seismic design of liquid-containing concrete structures (ACI
350.3-06) and commentary (350.3R-06). American Concrete Institute (ACI)
Committee 350, Environmental Engineering Concrete Structures, Farmington
Hills, (MI). 2006.
[35] Mikishev GN, Dorozhkin NY. An experimental investigation of free oscillations
of a liquid in containers. Izvestiya Akademii Nauk SSSR, Otdelenie Tekhnich-
eskikh Nauk, Mekhanika I Mashinostroenie 1961;4:48–83.
[36] Veletsos AS, Shivakumar P. Tanks containing liquid or solids. A handbook in
computer analysis and design of earthquake resistant structures. Computa-
tional Mechanical Publications 1997;725–74.
[37] Livaoglu R, Dogangun A. Effect of foundation embedment on seismic behavior
of elevated tanks considering fluid–structure-soil interaction. Soil Dyn Earthq
Eng 2007;27(9):855–63.
[38] Patkas LA, Karamanos SA. Variational solutions for externally induced sloshing
in horizontal-cylindrical and spherical vessels. ASCE J Eng Mech 2007;133(6):
641–55.
[39] Karamanos SA, Patkas LA, Platyrrachos MA. Sloshing effects on the seismic
design of horizontal-cylindrical and spherical industrial vessels. ASME J Press
Vessel Tech 2006;128(3):328–40.
[40] Larkin T. Seismic response of liquid storage tanks incorporating soil
structure interaction. ASCE J Geotech Geoenvironmental Eng 2008;134(12):
1804–14.
[41] Ghaemmaghami AR, Kianoush MR. Effect of wall flexibility on dynamic
response of concrete rectangular tanks under horizontal and vertical ground
motions. ASCE J Struct Eng 2010;136(4):441–51.
[42] Veletsos AS, Tang Y. Soil–structure interaction effects for laterally excited
liquid storage tanks. Earthq Eng Struct Dyn 1990;19:473–96.
[43] Cho KH, Kim MK, Lim MY. Seismic responses of base-isolated liquid storage
tanks considering fluid–structure-soil interaction in time domain. Soil Dyn
Earthq Eng 2004;24:839–52.