You are on page 1of 19

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2001; 30:14191437 (DOI: 10.1002/eqe.70)


Non-stationary seismic response of tanks with soil interaction
by wavelets
Pranesh Chatterjee and Biswajit Basu

Department of Civil Engineering, Jadavpur University, Calcutta 700032, India


SUMMARY
A wavelet-based random vibration theory has been developed for the non-stationary seismic response
of liquid storage tanks including soil interaction. The ground motion process has been characterized
via estimates of statistical functionals of wavelet coecients obtained from a single time history of
ground accelerations. The tankliquidsoil system has been modelled as a two-degree-of-freedom (2-
DOF) system. The wavelet domain equations have been formulated and the wavelet coecients of the
required response state are obtained by solving two linear simultaneous algebraic equations. The explicit
expression for the instantaneous power spectral density function (PSDF) in terms of the functionals of
the input wavelet coecients has been obtained. The moments of this PSDF are used to estimate the
expected pseudo-spectral acceleration (PSA) response of the tank. Parametric variations are carried out
to study the eects of tank height, foundation natural frequency, shear wave velocity of soil and ratio of
the mass of tank (including liquid) to the mass of foundation on the PSA responses of tanks. Copyright
? 2001 John Wiley & Sons, Ltd.
KEY WORDS: wavelet; non-stationarity; tanks; seismic excitation; soil interaction
1. INTRODUCTION
The response of uid storage tanks subjected to seismic ground motions has been the subject of
research interest in the recent years. Such storage tanks are quite common e.g. in oil and petro-
chemical industries, water treatment plants and in nuclear industries as spent-fuel tanks. These
tanks are often unanchored at their bases and supported on exible soils. In such cases, the
displacement of the tankfoundation system and the soil beneath is generally dierent from the
free eld seismic ground displacement due to lateral exibility of the supporting soil. The rel-
ative displacement between the tankfoundation and the free eld ground induces a base shear
at the soilfoundation interface due to lateral soil impedance and aects the hydrodynamic
pressure on the tank wall when the tank is subjected to strong ground accelerations. This may

Correspondence to: Biswajit Basu, Department of Civil Engineeering, Jadavpur University, Calcutta 700032, India.
Received 18 May 2000
Revised 20 November 2000
Copyright
?
2001 John Wiley & Sons, Ltd. Accepted 15 February 2001
1420 P. CHATTERJEE AND B. BASU
cause an increase in stresses under certain conditions which may seriously damage the tank
wall. For these reasons the uid storage tanks supported on exible soil require critical safety
analysis. Further, the responses of a soiltank system to seismic excitations become complex
as the ground motions are non-stationary in nature with both amplitude and frequency non-
stationarities [1].
Several researchers have studied the responses of uid storage tanks both theoretically and
experimentally. Housner [2] proposed the values for equivalent masses with their locations
to represent the forces and moments exerted by a liquid on the tank. Veletsos and Yang [3]
found out the natural frequencies of the tankliquid system by RayleighRitz method. In
another study, Haroun [4] modelled the elastic tank shell by nite elements and treated the
uid region as a continuum to nd out the natural frequencies of vibration and the associated
mode shapes. The studies performed by Peek [5] and Peek and Jennings [6] concentrated on
tank behaviour due to lateral loads. However, none of these studies accounted for the soil-
structure interaction eects on uid storage tanks. Toki and Miura [7] and Fischer and Seeber
[8] have made signicant contributions in studying the eects of soilstructure interaction
on liquid storage tanks under seismic ground excitations. Veletsos and Tang [9] presented
a method which establishes the eects of impulsive and convective actions of the liquid
on the tank responses and considered the horizontal and rocking motions of the foundation.
In all these studies, the loadings were either considered to be harmonic or represented by
deterministic time histories. However, due to the random nature of the seismic excitation the
responses are also stochastic in nature. Further, the responses should also include the eects
of the amplitude and frequency non-stationarities of the ground motions which have not been
considered in earlier studies. Several studies have shown that the amplitude and the frequency
non-stationarities signicantly aect the structural responses (see [10, 1, 11]).
In the present study, the non-stationary response of unanchored liquid storage tanks sub-
jected to lateral component of seismic base accelerations has been formulated. The random
vibration analysis carried out here is based on the wavelet analytic formulation as proposed by
Basu and Gupta [1]. Wavelet transform enables one to obtain the frequency content of a non-
stationary process locally in time. Thus, the timefrequency localization property of wavelets
allows to study features of the signal locally, the broad and ne features being studied, respec-
tively, on large and small scales which is not possible either in the case of Fourier domain
analysis or even in the case of short-time (windowed) Fourier transform-based analysis. Seis-
mic ground motion processes show local concentrations of frequency components due to the
dispersive phenomenon associated with propagating waves. The eect of this non-stationary
frequency content on the stochastic response of liquid storage tanks with soil interaction eects
could be well formulated and studied by proposing a wavelet-based formulation.
In formulating the soil interaction eects, the foundationsoil system has been considered
to be laterally exible. The translational motion of the tank has been considered. The tank
liquidfoundation is modelled as a two-degree-of-freedom (2-DOF) system where the lateral
resisting forces of the soilfoundation medium are considered by using appropriate impedance
functions. The tankliquid structural system has been modelled by a linear spring and a viscous
damper. The ground motion is characterized in terms of certain statistical functionals of the
wavelet coecients of the ground acceleration time history (as in [1]). The functionals of
the wavelet coecients are obtained from a single realization of a ground motion process by
applying statistical techniques. The expected response spectra computed using these processed
wavelet coecients is shown to be very close to the average response spectra obtained by
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
NON-STATIONARY SEISMIC RESPONSE OF TANKS 1421
time-history simulations of an ensemble of a synthetically generated ground motion process.
A parametric study has been carried out to observe the eects of liquid height in the tank,
shear wave velocity in the foundation soil, ratio of impulsive mass of the liquid to the mass
of the foundation and foundations natural frequency on the tank responses.
2. THEORETICAL FORMULATION
2.1. System considered
The tankuid system consists of a right circular cylindrical, rigid thin-walled tank of radius,
R lled with an incompressible liquid to a height, H. The tank wall is assumed to be of
uniform thickness, h. The Youngs modulus of elasticity and the density of the tank material
are denoted by E and j, respectively. This system has a xed-base natural frequency of
vibration, c
n
and a damping ratio, . The liquid in the tank is assumed to be vibrating in its
impulsive mode. The tank is rigidly clamped to a thick and rigid circular base mat of radius,
R same as that of the tank. The tank with the mat is directly resting on the surface of a
homogeneous elastic half space which is assumed to be laterally exible. The shear modulus
of elasticity, the Poissons ratio and the velocity of shear wave propagation of the supporting
medium are denoted by G
s
, v
s
and J
s
, respectively. The tankliquid system is represented
by a massspringdashpot model with impulsive mass of the liquid, m
i
, a linear spring with
stiness, K and a linear viscous damper with coecient, C. The foundation system of the
tank is of mass, m
f
with a complex valued translational impedance function, K
x
due to the
surrounding soil. The rocking motion of the system considered has been ignored in the present
analysis. Studies have shown (see [12, 13]) that for increased values of base mat thickness
the rotation of the base and the rocking component are generally not very signicant. Further,
it is expected that the eect of rocking modes of vibration of the tank may not be usually
predominant if the height of the tank is not too large for tanks of same proportion (i.e. same
H}R). The tankliquidfoundation system and the corresponding schematic diagram of the
dynamical massspringdashpot model adopted are shown in Figures 1 and 2, respectively.
Figure 1. Tankliquidfoundation system.
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
1422 P. CHATTERJEE AND B. BASU
Figure 2. 2-DOF model of the system.
2.2. Equations of motion
Let the tankuidfoundation system as described earlier be subjected to a horizontal free eld
seismic ground displacement, x
g
(t). The absolute displacement of the foundation is denoted
by x(t) and the displacement of the liquid in the tank relative to the foundation is denoted
by x
i
(t). The equation of motion of the tankuidfoundation system can be written as
m
i
( x
i
(t) + x(t)) +m
f
x(t) +Q
s
=0 (1)
where the overdot represents the dierentiation with respect to time and Q
s
is the base shear at
the soilfoundation interface. When there is no slipping between the soil and the foundation,
the induced base shear is proportional to the relative deformation between the tankfoundation
system and the free eld ground surface and can be expressed as
Q
s
=K
x
(x(t) x
g
(t)) (2)
In Equation (2), K
x
represents the complex valued lateral impedance function for a circular
footing on a exible soil and is given by [9]
K
x
=
8G
s
R
2 v
s
(:
x
+ ia
0
[
x
) (3)
The term a
0
, in Equation (3) is a dimensionless frequency parameter dened by
a
0
=
cR
J
s
(4)
where c is the frequency of vibration of the foundation and :
x
and [
x
are dimensionless
functions of a
0
and v
s
. Closed-form solutions of these parameters are given by Veletsos and
Verbic [14]. In this paper, the value of v
s
has been chosen to be
1
3
for which :
x
=1.0 and
[
x
=0.65. It may be noted here that, when c=0, the complex valued impedance function
becomes real and it represents the static stiness function. On substituting Equation (2) in
Equation (1) we get
m
i
[ x
i
(t) + x(t)] +m
f
x(t) +K
x
(x(t) x
g
(t)) =0 (5)
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
NON-STATIONARY SEISMIC RESPONSE OF TANKS 1423
Also, the equation of motion of the tankliquid system considering the dynamic equilibrium
can be written as
m
i
x
i
(t) +C x
i
(t) +Kx
i
(t) = m
i
x(t) (6)
On dividing both sides of Equation (6) by m
i
and on using the relations, K}m
i
=c
2
n
and
C}m
i
=2c
n
, we obtain
x
i
(t) +c
2
n
x
i
(t) + 2c
n
x
i
(t) = x(t) (7)
The expression for c
n
is given by [9]
c
n
=
C
i
H

E
j
(8)
Equations (1) and (7) are the dierential equations of motions of the tankliquidfoundation
system and the tankliquid system, respectively. This set of coupled dierential equations
can be solved to obtain the non-stationary response of the system based on wavelet analytic
techniques for a seismic base acceleration, x
g
(t) with known non-stationary statistical char-
acteristics. In the following section, the non-stationary seismic response of the tankliquid
foundation system has been obtained by extending the wavelet based formulation of Basu and
Gupta [1] for dynamical systems.
2.3. Wavelet-based-stochastic-response
Let the seismic ground acceleration process, x
g
(t) be assumed to have zero mean, non-
stationary Gaussian characteristics. The wavelet transform of this function, x
g
(t) with respect
to a wavelet basis function, (t) and the inverse relationship are given as (see [15] for details)
W

x
g
(a, b) =
1
|a|
1}2

x
g
(t)
a, b
(t) dt (9)
and
x
g
(t) =
1
2C

1
a
2
W

x
g
(a, b)
a, b
(t) da db (10)
with
C

(c)|
2
|c|
dc (11)
In Equations (9) and (10), the function (.) is the basic or the mother wavelet with the
Fourier transform

(c) =
1

(t)e
ict
dt (12)
and Equation (11) is the admissibility condition for (t) to be a basic or mother wavelet.
The wavelet functions,
a, b
(t) have been constructed by the dilated and translated versions of
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
1424 P. CHATTERJEE AND B. BASU
the basic wavelet [1] using two parameters a and b as

a, b
(t) =
1
|a|
1}2

t b
a

; a, b R
+
(13)
The parameter, b localizes or centres the basis function at t =b and its neighbourhood by
windowing over a certain temporal stretch depending on the parameter, a. Equation (10) can
be discretized according to a discretization scheme [1] by assuming a
)
=o
)
and b
)
=()1)b
into the following form:
x
g
(t) =

)
K

b
a
)
W

x
g
(a
)
, b
i
)
a
)
, b
i
(t) (14)
where
K

=
2
4C

o
1
o

(15)
The non-stationary Gaussian seismic base acceleration process may be characterized by the
wavelet coecients, i.e. E[W
2

x
g
(a
)
, b
i
)]. Although the wavelet representation for the base
acceleration function, x
g
(t) has been discussed above, these expressions hold good for the
ground displacement and the structural response function as well. It may be noted that seismic
base excitations appear in Equation (1) in the form of ground displacement, x
g
(t). However,
usually the input seismic excitation is described in terms of base acceleration, x
g
(t). Thus,
it is necessary to relate the wavelet coecient of ground displacement with that of ground
acceleration, as follows. The ground displacement in terms of wavelet coecients may be
written as
x
g
(t) =

)
K

b
a
)
W

x
g
(a
)
, b
i
)
a
)
, b
i
(t) (16)
On dierentiating Equation (16) twice and then on Fourier transforming both sides, we obtain

x
g
(c) =

)
K

b
a
)
W

x
g
(a
)
, b
i
)

a
)
, b
i
(c)(c
2
) (17)
Also, Fourier transform of Equation (14) gives

x
g
(c) =

)
K

b
a
)
W

x
g
(a
)
, b
i
)

a
)
, b
i
(c) (18)
Comparison of Equations (17) and (18) yields

)
K

b
a
)
W

x
g
(a
)
, b
i
)

a
)
, b
i
(c)
=

)
K

b
a
)
W

x
g
(a
)
, b
i
)

a
)
, b
i
(c)

1
c
2

(19)
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
NON-STATIONARY SEISMIC RESPONSE OF TANKS 1425
relating the wavelet coecients of ground displacement in terms of that of ground accelera-
tion. Now, the absolute displacement of the foundation and the relative displacement of the
impulsive mass of the liquid with respect to the foundation may also be expanded in terms
of wavelet coecients as
x(t) =

)
K

b
a
)
W

x(a
)
, b
i
)
a
)
, b
i
(t) (20)
and
x
i
(t) =

)
K

b
a
)
W

x
i
(a
)
, b
i
)
a
)
, b
i
(t) (21)
On substituting the expressions for x(t), x
g
(t) and x
i
(t) in Equations (5) and (7), respectively,
we obtain
(m
i
+m
f
)

)
1
a
)
W

x(a
)
, b
i
)

a
)
, b
i
(t) +K
x

)
1
a
)
W

x(a
)
, b
i
)
a
)
, b
i
(t)
+m
i

)
1
a
)
W

x
i
(a
)
, b
i
)

a
)
, b
i
(t) =K
x

)
1
a
)
W

x
g
(a
)
, b
i
)
a
)
, b
i
(t) (22)
and

)
K

b
a
)
W

x
i
(a
)
, b
i
)

a
)
, b
i
(t) +c
2
n

)
K

b
a
)
W

x
i
(a
)
, b
i
)
a
)
, b
i
(t)
+2c
n

)
K

b
a
)
W

x
i
(a
)
, b
i
)

a
)
, b
i
(t)
=

)
K

b
a
)
W

x(a
)
, b
i
)

a
)
, b
i
(t) (23)
Further, on taking Fourier transform of both sides of the Equations (22) and (23) and using
Equation (19) we obtain
(K
x
(m
i
+m
f
)c
2
)

)
1
a
)
W

x(a
)
, b
i
)

a
)
, b
i
(c)
(c
2
m
i
)

)
1
a
)
W

x
i
(a
)
, b
i
)

a
)
, b
i
(c)
=

K
x
c
2

)
1
a
)
W

x
g
(a
)
, b
i
)

a
)
, b
i
(c) (24)
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
1426 P. CHATTERJEE AND B. BASU
and

)
1
a
)
W

x
i
(a
)
, b
i
)(c
2
)

a
)
, b
i
(c) +c
2
n

)
1
a
)
W

x
i
(a
)
, b
i
)

a
)
, b
i
(c)
+2c
n

)
1
a
)
W

x
i
(a
)
, b
i
)(ic)

a
)
, b
i
(c)
=

)
1
a
)
W

x(a
)
, b
i
)(c
2
)

a
)
, b
i
(c) (25)
Equations (24) and (25) can be solved to obtain the wavelet coecients of the tank displace-
ment in terms of that of ground acceleration. In order to solve these two equations, a suitable
wavelet basis function is to be chosen. In the present study, the LittlewoodPaley (LP) basis
has been chosen for further analysis. The Fourier transform of the LP basis function is given
by (see [1])

(c) =
1

2( 1)
6|c|6
=0 otherwise (26)
On taking the inverse Fourier transform of Equation (26) with o =, the wavelet basis
function is obtained as
(t) =
1

o 1
sin ot sin t
t
(27)
The value for o in Equation (27) has been chosen as 2
1}4
[1], as this value is reasonable
for representing most of the ground motions. Values smaller than this may also be used;
however, that would lead to increased computational eorts due to lesser number of energy
bands. Now, multiplying both sides of Equations (24) and (25) by


a
k
, b
!
(c) and on using the
relation

a
)
, b
i
(c)


a
k
, b
!
(c) =o
)k

a
)
, b
i


a
k
, b
!
(c) (28)
we get

i
W

x(a
)
, b
i
)

a
)
, b
i
(c)

K
x
m
f
(1 +)c
2

i
W

x
i
(a
)
, b
i
)

a
)
, b
i
(c)[c
2
]
=

i
W

x
g
(a
)
, b
i
)

a
)
, b
i
(c)

K
x
c
2
m
f

(29)
and

i
W

x
i
(a
)
, b
i
)

a
)
, b
i
(c)(c
2
) +c
2
n

i
W

x
i
(a
)
, b
i
)

a
)
, b
i
(c)
+2c
n

i
W

x
i
(a
)
, b
i
)(ic)

a
)
, b
i
(c) =

i
W

x(a
)
, b
i
)

a
)
, b
i
(c)(c
2
) (30)
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
NON-STATIONARY SEISMIC RESPONSE OF TANKS 1427
where represents the ratio m
i
}m
f
. On solving the linear simultaneous Equations (29) and
(30) in W

x
i
and W

x we obtain

i
W

x
i
(a
)
, b
i
)

a
)
, b
i
(c) =t(c)

i
W

x
g
(a
)
, b
i
)

a
)
, b
i
(c) (31)
where t(c) is the frequency domain transfer function of the relative displacement of the
tankliquid system and is expressed as
t(c) =
K
x
}m
f
[K
x
}m
f
c
2
(1 +)][c
2
n
c
2
+ 2icc
n
] c
4
(32)
where
K
x
m
f
=c
2
f

1 + i
0.65cR
J
s

(33)
In Equation (33), c
f
represents the natural frequency of vibration of the foundation. Substitu-
tion of Equation (33) into Equation (32) and subsequent simplication transforms the transfer
function to
t(c) =
A + iB
C + iD
(34)
where
A =c
2
f
(35a)
B =
0.65AcR
J
s
(35b)
C =(c
2
n
c
2
)[A c
2
(1 +)] 2Bcc
n
c
4
(35c)
and
D=2cc
n
[A c
2
(1 +)] +B(c
2
n
c
2
) (35d)
Equation (31) gives the relationship between the wavelet coecients of the tank responses
and that of the ground accelerations. On taking expectation of the square of the amplitude of
both sides of Equation (31), integrating over c and using the orthogonality relationship for
the proposed basis function [1], we get

i
E[|W

x
i
(a
)
, b
i
)|
2
]
=

i
E[|W

x
g
(a
)
, b
i
)|
2
]

|t(c)|
2
|

a
)
, b
i
(c)|
2
dc (36)
On using Parsevals identity and the wavelet analysis relationship (see [15])

x
i
2
(t) dt =

)
K

b
a
)
W
2

x
i
(a
)
, b
i
) (37)
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
1428 P. CHATTERJEE AND B. BASU
we thus obtain from Equation (36) the contribution to the energy or expected (mean) square
response for the frequency band corresponding to the dilation factor, a
)
as

E[|X
)
(c)|
2
] dc
=

i
K

b
a
)
E[|W

x
g
(a
)
, b
i
)|
2
]

|t(c)|
2
|

a
)
, b
i
(c)|
2
dc (38)
From Equation (38), on using the time-localization property of the wavelet coecients, the
instantaneous mean square response at t =b
i
, corresponding to the band with dilation factor,
a
)
is obtained as

E[|X
)
i
(c)|
2
] dc=
K

b
a
)
E[|W

x
g
(a
)
, b
i
)|
2
]

|t(c)|
2
|

a
)
, b
i
(c)|
2
dc (39)
For the value of o assumed in this paper (close to one) the dierent energy bands become quite
narrow and the above integral has to be performed within a very short interval over which
the integrand may be assumed to be almost constant. Thus, a pointwise relation following
Equation (39) may be written as
E[|X
)
i
(c)|
2
] =
K

b
a
)
E[|W

x
g
(a
)
, b
i
)|
2
]|t(c)|
2
|

a
)
, b
i
(c)|
2
(40)
Thus, from Equation (40), the total energy distribution over dierent frequencies at the instant
t =b
i
may be obtained by summing up contributions from all energy bands since the bands
are non-overlapping. This energy distribution then averaged over an interval of b gives the
instantaneous power spectral density function (PSDF) of the tankliquid response, x
i
as
S
x
i
(c) =

)
E[|X
)
i
(c)|
2
]
b
=

)
K

a
)
E[|W

x
g
(a
)
, b
i
)|
2
]|t(c)|
2
|

a
)
, b
i
(c)|
2
(41)
In order to obtain more information about the statistics of the response process, x
i
(t), the mo-
ments of instantaneous PSDF of x
i
(t) must be evaluated. These moments of the instantaneous
output PSDF will be used here to obtain some statistical parameters like rate of crossings
of the zero level, bandwidth parameter, etc. which may further be used to obtain the largest
peak amplitude of the process x
i
(t). The expressions for the zeroth, rst and second moments
of the PSDF are given as
m
0
|
t =b
i
=

)
K

E[W

x
g
(a
)
, b
i
)
2
]I
0, )
(c
n
, ) (42a)
m
1
|
t =b
i
=

)
K

E[W

x
g
(a
)
, b
i
)
2
]I
1, )
(c
n
, ) (42b)
and
m
2
|
t =b
i
=

)
K

E[W

x
g
(a
)
, b
i
)
2
]I
2,)
(c
n
, ) (42c)
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
NON-STATIONARY SEISMIC RESPONSE OF TANKS 1429
with
I
0, )
=

o}a
)
}a
)
A
2
+B
2
C
2
+D
2
dc (43a)
I
1, )
=

o}a
)
}a
)
A
2
+B
2
C
2
+D
2
cdc (43b)
I
2, )
=

o}a
)
}a
)
A
2
+B
2
C
2
+D
2
c
2
dc (43c)
and
K

=
K

(o 1)
(44)
The instantaneous rate of the zero crossings,
i
and the band-width parameter, z
i
are given
as

i
=

m
2
|
t =b
i
m
0
|
t =b
i
(45)
and
z
i
=

1
m
2
1
|
t =b
i
m
0
|
t =b
i
m
2
|
t =b
i
(46)
Also, the expected peak value of the relevant process can be calculated using the non-stationary
peak factors. For calculating the non-stationary peak factors, the expression for the probability
P
1
(x) (that the process |x
i
(t)| remains below the level x within the time interval 0, 1) given
by [16]
P
1
(x) = exp

1
0
:(t) dt

(47)
After being discretized Equation (47) takes the form
P
1
(x) = exp

i
:(t)|
t =b
i
b

(48)
where :(t)|
t =b
i
is the time dependent rate of the Poisson process given by
:(t)|
t =b
i
=

i

exp
x
2
}2o
2
i
1 exp(

}2[z
i
(t)]
1.2
x}o
i
)
1 exp(x
2
}2o
2
i
)
(49)
The expected value of the largest peak can be obtained from
E(x
(1)
) =

1
0
P
1
1
(u) du (50)
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
1430 P. CHATTERJEE AND B. BASU
3. CHARACTERIZATION OF GROUND MOTION
In order to characterize the ground motions, the wavelet coecients are calculated from a
single realization of the synthetic accelerogram process generated by using the SYNACC
program [1721]. This accelerogram simulates the motion recorded at the Dumbarton bridge
site near Coyote Hills during the 1989 Loma Prieta earthquake. The parameters considered for
generating the synthetic accelerogram for Loma Prieta earthquake are same as those considered
by Gupta and Trifunac [22]. The discretization parameter, b has been taken as 0.02. The
wavelet coecients (totaling 45034 in number) are found out for i varying from 1 to 2047
and ) varying from 17 to 4. For a particular frequency band, all the corresponding wavelet
coecients over the duration, 1, i.e., with i =1 to 2047 are considered. The total duration, 1
is divided into a number of stretches each of duration, 1
)
equal to the period corresponding
to the central frequency, i.e., (}a
)
+o}a
)
)}2 of the band considered. The duration of each
stretch is given by the expression
1
)
=
4a
)
1 +o
(51)
In each stretch, the wavelet coecients are squared and then averaged over the stretch of du-
ration, 1
)
. The average value thus obtained represents the value of E[W

2
x
g
(a
)
, b
i
)] uniformly
over that stretch and can be written as
E[W
2

x
g
(a
)
, b
i
)] =
b
1
)
k
2
=4a
)
(m+1)}(1+o)b

k
1
=4a
)
m}(1+o)b
W
2

x
g
(a
)
, b
k
), m=0, 1, 2, . . . (52)
where k
1
6i6k
2
. To see that E[W

2
x
g
(a
)
, b
i
)] really characterizes the ground motion, we
obtain the wavelet based response spectra using the stochastic formulation for the response of
SDOF systems by Basu and Gupta [1]. Also, an ensemble of 50 accelerograms is generated
using the SYNACC program. From these 50 accelerograms the responses of SDOF systems
are simulated by direct time history integration [23]. The response spectra for pseudo-spectral
acceleration (PSA) thus generated are then averaged to obtain the expected PSA spectra from
digitally simulated accelerogram records. Comparison of these two response spectra, i.e., one
obtained by ensemble averaging and the other by stochastic formulation shows good agreement
as seen from Figure 3.
4. NUMERICAL RESULTS
A parametric study on the stochastic responses of the tankuidfoundation system is carried
out by the proposed formulation by varying the tank height, H, shear wave velocity of soil
medium, J
s
, ratio of impulsive mass of liquid (water being considered for the present study)
to the mass of the foundation, (mass ratio) and the natural frequency of foundation, c
f
.
The values of some of the parameters assumed constant throughout the numerical study are
as follows. The tankuid super-structure is assumed to have a damping ratio, of 5%. The
values of Youngs modulus of elasticity and density of the tank material are 2.1 10
6
kg}cm
2
,
and 7850 kg}m
3
, respectively. The tank wall thickness to tank radius ratio, h}R is assumed
to be 0.001. Correspondingly, the value of C
i
in Equation (8) can be chosen from Table II
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
NON-STATIONARY SEISMIC RESPONSE OF TANKS 1431
Figure 3. Comparison of PSA spectra obtained by stochastic-wavelet
formulation and ensemble averaging.
Figure 4. Variation in PSA with tank height for dierent mass ratios in case of H}R=0.5.
of Veletsos and Tang [9]. For the purpose of numerical study, the ground motion process in
the case of the 1989 Loma Prieta earthquake ground accelerations at Dumbarton bridge site
(Coyote Hills) as characterized in Section 3 has been considered.
Figures 47, show the variation of PSA response of broad tanks (H/R=0.5) with dierent
heights of the tank varying from 2 m to 20 m. Figure 4 represents the variation for dierent
values of mass ratio, (0.5, 2.0 and 6.0) with J
s
=600 m}s and c
f
=12.57 rad}s. It can be
seen from the gure that when the mass ratio is low, say 0.5, the response is more or less
uniform for tanks of height 4 m to 12 m. When the tank is heavier than the foundation, say
with mass ratio 2 or 6, the response is highest for very small tanks (of height 2 m) and then
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
1432 P. CHATTERJEE AND B. BASU
Figure 5. Variation in PSA with tank height for dierent shear wave velocities.
Figure 6. Variation in PSA with tank height for dierent natural
frequencies of foundation at shear wave velocity 600 m}s.
gradually decreases as tank height increases. The graph also shows that the variation in the
eect of soiltank interaction for tanks of dierent height is more prominent in the case of
tanks with higher mass ratio. The curves in Figure 5 are plotted for dierent values of J
s
varying from 100 to 2000 m}s, in the case of =6.0 and c
f
=12.57 rad}s. It can be seen
from Figure 5, that the response at a given tank height increases due to increase in J
s
. It is
also observed that the response gradually decreases asymptotically as the tank height increases.
Further, variation in the eect of the soil interaction with tank height is more prominent in
the case of stier soils. The graphs in Figures 6 and 7 are for ve dierent values of c
f
, viz.,
3.14, 7.85, 31.42, 62.83 and 157.08 rad}s with as 6.0 and J
s
=600 m}s and 2000 m}s,
respectively. Both the graphs are similar in trend with the values of PSA greater in magnitude
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
NON-STATIONARY SEISMIC RESPONSE OF TANKS 1433
Figure 7. Variation in PSA with tank height for dierent natural frequencies
of foundation at shear wave velocity 2000 m}s.
in the case of J
s
=2000m}s as compared to those of J
s
=600m}s. It is observed from Figures
6 and 7 that for very sti foundations resting on sti soil the response of the tank approaches
the xed base response. The PSA response of the tank in such case is maximum for a tank
height of about 15 m height (having time period of 0.25 s) which corresponds to the peak of
the xed base PSA response in Figure 3 at a period of about 0.25 s. From both the graphs
it is found that the variation in the soil interaction eects with tank height are negligible for
tanks resting on exible foundation systems.
Figure 8 represents the variation of PSA with shear wave velocity for dierent values of
tank heights with H}R=0.5, c
f
=12.57 rad}s and =6.0. As the shear wave velocity in-
creases, the soilfoundation medium becomes stier and the eects of soilstructure interaction
diminish which is shown by the atter trends of the curves for higher shear wave velocities.
Further, it is seen that for very low values of J
s
the PSA responses of the tanks are almost
independent of the tank height.
Figure 9 shows the variation of PSA with the natural frequency of foundation, c
f
, for
dierent values of tank height, H in the case of H}R=0.5, J
s
=600 m}s and =6.0. It is
observed that up to the foundation frequency range of about 8090 rad}s the PSA response
decreases with the increase in tank height whereas the trend is reversed beyond the mentioned
range.
To study the variation in the responses with the slenderness of the tanks, Figure 10 has
been plotted to show the variation in the values of PSA responses of tanks with shear wave
velocities for three dierent H}R ratios, viz. 0.5, 1 and 3. The tank height, mass ratio and
lateral frequency of the foundation are assumed to be 10 m, 6 and 12.57 rad}s, respectively.
It can be observed from Figure 10 that, the values of PSA increase for more slender tanks.
Further, it is seen that the responses increase with the increase in the shear wave velocity. The
change in the response is however predominant for soft soils with low shear wave velocity.
For very soft soils the responses of all types of tanks, broad or slender, are found to be very
close.
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
1434 P. CHATTERJEE AND B. BASU
Figure 8. Variation in PSA with shear wave velocity for dierent tank heights.
Figure 9. Variation in PSA with natural frequency of foundation for dierent tank heights.
To see how the PSA responses of slender tanks are aected with the variation in the height
and mass ratios of the tanks and to compare those with the corresponding variations in the
case of broad tanks Figure 11 has been plotted. Figure 11 represents the change in the values
of PSA response with tank height for varying mass ratios in case of slender tanks (H}R=3),
with J
s
and c
f
as 600m}s and 12.57rad}s, respectively. It is seen that the trend in variations
is similar to that in Figure 4, however, for H}R=3 the tank responses are generally greater
for a particular mass ratio as compared to those for H}R=0.5.
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
NON-STATIONARY SEISMIC RESPONSE OF TANKS 1435
Figure 10. Variation in PSA with shear wave velocity for dierent H}R ratios.
Figure 11. Variation in PSA with tank height for dierent mass ratios in the case of H}R=3.0.
5. CONCLUSIONS
A wavelet-based stochastic formulation for the non-stationary seismic response of liquid stor-
age tanks with soilfoundation interaction has been presented here. A single realization of
the ground motion process has been used to characterize the non-stationary ground excitation
via statistical estimates of the functionals of wavelet coecients of the process. The proposed
characterization scheme works appropriately as is seen from the comparison of the expected
PSA spectra obtained from the wavelet-based stochastic formulation for an SDOF system with
those obtained by ensemble averaging of digital simulation results. The following conclusions
can be drawn from the parametric variation results for broad tanks (with H}R=0.5) and
for the given ground acceleration process. It has been observed that the eects of soiltank
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
1436 P. CHATTERJEE AND B. BASU
interaction leading to the amplication of PSA responses are pronounced for smaller or heav-
ier tanks. Further, the variation in the interaction eects with tank height are predominant
for tanks with larger mass ratio or for those supported on stier foundations or stier soil
medium. For tanks supported on very exible soil, the response is, however, almost constant
irrespective of the height or the slenderness of the tank. Also, generally the slender tanks with
soil interaction have greater PSA responses as compared to the broad tanks and the responses
further increase as the soil becomes stier. However, chances of rocking motions aecting
the results in case of tanks of greater height demand further studies which would include this
eect in soiltank interaction. The proposed formulation in this paper may thus be utilized
to obtain the site-specic non-stationary lateral response of soiltank systems subjected to
ground accelerations characterized through statistical functionals of wavelet coecients. The
present formulation may be extended to include the phenomena of rocking of tanks and
also more complex liquid motions which will be addressed in a forthcoming paper by the
authors.
REFERENCES
1. Basu B, Gupta VK. Seismic response of SDOF systems by wavelet modeling of non-stationary processes.
Journal of Engineering Mechanics ASCE 1998; 124(10):11421150.
2. Housner GW. The dynamic behaviour of water tanks. Bulletin of the Seismological Society of America 1963;
53(2):381387.
3. Veletsos AS, Yang JY. Earthquake response of liquid storage tanks. In Advances in Civil Engineering through
Engineering Mechanics, Proceedings of the Engineering Mechanics Specialty Conference, 1977, Rayleigh, NC,
ASCE: New York, 124.
4. Haroun MA. Vibration studies and tests of liquid storage tanks. Earthquake Engineering and Structural
Dynamics 1983; 11(2):179206.
5. Peek R. Analysis of unanchored liquid storage tanks under lateral loads. Earthquake Engineering and Structural
Dynamics 1988; 16(7):10871100.
6. Peek R, Jennings PC. Simplied analysis of unanchored tanks. Earthquake Engineering and Structural
Dynamics 1988; 16(7):10731085.
7. Toki K, Miura F. Non-linear seismic response analysis of soilstructure interaction systems. Earthquake
Engineering and Structural Dynamics 1983; 11(1):7789.
8. Fischer FD, Seeber R. Dynamic response of vertically excited liquid storage tanks considering liquidsoil
interaction. Earthquake Engineering and Structural Dynamics 1988; 16:329342.
9. Veletsos AS, Tang Y. Soilstructure interaction eects for laterally excited liquid storage tanks. Earthquake
Engineering and Structural Dynamics 1990; 19:473496.
10. Basu B, Gupta VK. Nonstationary seismic response of MDOF systems by wavelet transform. Earthquake
Engineering and Structural Dynamics 1997; 26:12431258.
11. Yeh C-H, Wen YK. Modeling of non-stationary motion and analysis of inelastic structural response. Structural
Safety 1990; 8(14):281298.
12. Malhotra PK. Base uplifting analysis of exibly supported liquidstorage tanks. Earthquake Engineering and
Structural Dynamics 1995; 24:15911607.
13. Malhotra PK. Seismic response of soilsupported unanchored liquidstorage tanks. Journal of Structural
Engineering ASCE 1997; 123(4):440450.
14. Veletsos AS. Verbic B. Vibrations of visco-elastic foundations. Earthquake Engineering and Structural
Dynamics 1973; 2:87102.
15. Daubechies I. Ten Lectures on Wavelets, Society for Industrial & Applied Mathematics, Philadelphia,
Pennsylvania, 1992.
16. Vanmarcke EH. On the distribution of the rst-passage time for normal stationary random processes. Journal
of Applied Mechanics Transport Transactions of ASME 1975; 42:215220.
17. Wong HL, Trifunac MD. Generation of articial strong motion accelerograms. Earthquake Engineering and
Structural Dynamics 1979; 7:509527.
18. Lee VW, Trifunac MD. Torsional accelerograms. Soil Dynamics and Earthquake Engineering 1985; 4(3):
132139.
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437
NON-STATIONARY SEISMIC RESPONSE OF TANKS 1437
19. Lee VW, Trifunac MD. Rocking strong earthquake accelerograms. Soil Dynamics and Earthquake Engineering
1987; 6(2):7589.
20. Lee VW, Trifunac MD. A note on ltering strong motion accelerograms to produce response spectra of specied
shape and amplitude. European Earthquake Engineering 1989; III(2):3845.
21. Trifunac MD. Curvograms of strong ground motions. Journal of Engineering Mechanics ASCE 1990;
116(6):14261432.
22. Gupta VK, Trifunac MD. A note on the eects of ground rocking on the response of buildings during 1989
Loma Prieta earthquake. Earthquake Engineering and Engineering Vibration 1993; 13(2):1228.
23. Nigam NC, Jennings PC. Calculation of response spectra from strong motion earthquake records. Bulletin of
the Seismological Soceity of America 1969; 59(2):909922.
Copyright ? 2001 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2001; 30:14191437

You might also like