You are on page 1of 13

F. G.

Golzar
R. Shabani1
e-mail: r.shabani@urmia.ac.ir
Department of Mechanical Engineering,
Faculty of Engineering,
Urmia University,
Urmia 15311-57561, Iran

S. Tariverdilo
Department of Civil Engineering,
Faculty of Engineering,
Urmia University,
Urmia 15311-57561, Iran

G. Rezazadeh
Department of Mechanical Engineering,
Faculty of Engineering,
Urmia University,
Urmia 15311-57561, Iran

Sloshing Response of Floating


Roofed Liquid Storage Tanks
Subjected to Earthquakes
of Different Types
Using extended Hamiltonian variational principle, the governing equations for sloshing
response of floating roofed storage tanks are derived. The response of the floating roofed
storage tanks is evaluated for different types of ground motions, including near-source
and long-period far-field records. Besides comparing the response of the roofed and
unroofed tanks, the effect of different ground motions on the wave elevation, lateral
forces, and overturning moments induced on the tank is investigated. It is concluded that
the dimensionless sloshing heights for the roofed tanks are solely a function of their first
natural period. Also it is shown that while long-period far-field ground motions control
the free board height, near-source records give higher values for lateral forces and overturning moments induced on the tank. This means that same design spectrum could not be
used to evaluate the free board and lateral forces in the seismic design of storage tanks.
Finally, two cases are studied to reveal the stress patterns caused by different earthquakes. [DOI: 10.1115/1.4006858]
Keywords: liquid storage tank, sloshing response, earthquake, floating roof

Introduction

Floating roofs are used in the petroleum industries for storage


of liquid hydrocarbons in atmospheric storage tanks. Reducing
the evaporation of storage materials, floating roofs provide better
protection against possible ignition of the vapors by sparks generated by different sources, such as static electricity, earthquake,
cigarette smoking, etc. [1]. In addition to economic benefit by
preventing the evaporation of valuable products, these roofs are
also helpful in reduction of the environmental pollution caused
by evaporation. The conventional floating roofs can be categorized into two types: single deck and double deck. Single deck
roofs are composed of a circumferential buoyant ring (pontoon)
with relatively large stiffness surrounding an inner circular plate
(deck) with small thickness and stiffness. Double deck roofs, on
the other hand consisting of upper and lower decks, are heavier
and more rigid.
Serious damages of floating roofs in past earthquakes could be
attributed to the sloshing of the contained liquid [2]. Therefore,
it seems essential to take into account the interaction between
floating roof and contained liquid in any analysis of the system
subject to ground motion excitation. As a pioneering approach,
Sakai et al. [3] derived the equations of motion for a cylindrical
tank with floating roof using variational principle to evaluate
natural frequencies and pressure distribution upon the roof. They
validated the results experimentally for single deck and double
deck models. Isshiki and Nagata [4] used Hamiltons principle
for plate and Kelvins principle for water to derive the extended
Hamiltonian principle for fully coupled problem. Robinson and
Palmer [5] carried out modal analysis and derived transfer function for low amplitude oscillations of a floating plate. Amabili

1
Corresponding author.
Contributed by the Pressure Vessel and Piping Division of ASME for publication
in the JOURNAL OF PRESSURE VESSEL TECHNOLOGY. Manuscript received July 18, 2011;
final manuscript received April 22, 2012; published online September 10, 2012.
Assoc. Editor: Spyros A. Karamanos.

Journal of Pressure Vessel Technology

and Kwak [6] used RayleighRitz method and Hankel transformation to analyze the free vibrations of circular plates with uniform boundary conditions. Matsui [7] employed FourierBessel
series to derive the linear equations of motion for a floating roof
with relatively high rigidity and compared the results with those
of unroofed tank. Developing the frequency content of the wave
elevation and roof pressure for a tank under seismic excitation,
his results revealed the suppression of higher modes in wave elevation and magnification of higher modes in the pressure
response. His results were verified by shaking table tests carried
out by Nagaya et al. [8]. Their experiments showed the strong
dependency of damping ratio on roof type. Later, Matsui used
the same method to analyze the motion of single deck floating
roof with pontoon [9]. Virella et al. [10] used linear and nonlinear finite element methods to analyze the response of harmonically base excited rectangular tanks and used nondimensional
parameters to compare the results with those of the cylindrical
tanks. They showed that the wall pressures were not much
influenced by the nonlinearities. In a similar way, Mitra and Sinhamahapatra [11] investigated the response for seismic excitations. Response of base isolated cylindrical tanks, with and
without roof, to earthquake excitations was subject of a comprehensive experimental study by De Angelis et al. [12]. They verified the results by simple numerical methods and showed the
favorable effects of base isolators and floating roofs on reducing
the wall pressures and oscillation amplitudes, respectively.
Efforts were also put into mere nonlinear analysis. Utsumi and
Ishida [13,14] accounting for nonlinear sloshing investigated the
possibility of internal resonance under seismic and harmonic
excitations.
The above mentioned papers have considered elemental
sloshing characteristics like coupled frequencies and mode
shapes or/and case studies of roofed tanks subjected to certain
earthquake excitations while a need for a more inclusive deduction is felt. In this paper, by applying the Hamiltons variational
principle on the floating roof and contained liquid, the sloshing
response of the floating roof storage tanks is investigated for

C 2012 by ASME
Copyright V

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

OCTOBER 2012, Vol. 134 / 051801-1

different ground motion records. The records include longperiod far-field and near-source ground motions. Numerical simulations are used to evaluate the wave elevation, and lateral
force and overturning moment imposed on the tank wall.
Effects of the roof and different types of the ground motions on
the results are investigated thoroughly. Computations are carried
out for a range of commonly used dimensions for storage tanks.
The results could have practical implications for design
purposes.

Mathematical Model

Figure 1 depicts a typical liquid storage tank with a floating


roof and the cylindrical coordinate used in the analysis. The roof
is considered of double deck type with radius R, thickness h, uniform bending rigidity D, and areal mass density M. Peripheral
wall of the tank is reasonably considered rigid as the natural frequencies of shell vibrations are much higher than the sloshing frequencies. The tank is subjected to a horizontal unidirectional
seismic excitation xg .
Fluid movement in the domain, (0  r  R; 0  h
< 2p; 0  z  H) for an incompressible and inviscid liquid,
is described by the velocity potential U which satisfies the Laplacian equation (r2 U 0) as well as the following boundary
conditions:

@U 
0
@z z0

(1)


@U 
x_ g cos h
@r rR

(2)


@U 
@g

@z zH
@t

e r 
e z
k
k
cosh
R
 R  cos h
Ak
Ur; z; h r x_g cos h
ek H
J1 ek
k1
cosh
R
(4)
r x_g cos h U0
1
X

J1

Where J1 denotes the Bessel function of the first kind of order 1.


ek are eigenvalues of the velocity potential and are found by solving the equation J 01 ek 0. Floating plate motion can be considered as a weighted summation of its mode shapes in air

gr; h; t

1
X

Bi tXi r cos h

(5)

i1

where Bi and Xi denote the generalized coordinates and radial


mode shapes, respectively. The radial mode shapes consist of rigid
and elastic modes in the following form [15]:
r
; i1
R
h
 r
 r i
Xi r ai J1 ki
bi I1 ki
;
R
R
Xi r

(6)
i 2; 3   

where ji are the radial eigenvalues of the mode shapes and are
found by satisfying zero shear and zero moment conditions on the
free edge. J1 denotes the Bessel function of the first kind of order
1 and I1 denotes the modified Bessel function of the first kind of
order 1. The mode shapes are normalized in order to have
R 2p
MR
r0

h0

h
 r
 r i
2
b i I1 j i
cos h rdrdh 1 (7)
ai J1 ji
R
R

(3)

which suggest the impermeability condition at the bottom and peripheral wall and kinematic condition at the interface of plate and
liquid, respectively. Imposing these conditions, the velocity potential is obtained to be in the following form:

where MR denotes the areal mass density of the roof. Bearing in


mind that the roof and liquid stay in complete contact, the Lagrangian of the coupled system, accounting for the kinetic and potential energies, is written in form of the sum of roof Lagrangian LR
[15] and liquid Lagrangian Li [3]:

L LR Ll

R 2p 
r0 h0

1
MR g_ 2
2

2
!2
 2
1 4 @2g
1 @g 1 @ 2 g

 DR

2
@r 2
r @r r 2 @h2

2

 2 

@ g 1 @g 1 @ 2 g
rdrdh

@r 2
r @r r 2 @h2




1 @U0
@g
1
qL 
U0 U  gg2 rdrdh
2 @z
@t
2
h0
(8)

R 2p
r0

Note that the twisting effects on the strain energy of the roof have
not been accounted for. DR denotes the bending rigidity of the
roof and is calculated by
DR

Fig. 1 Cylindrical liquid storage tank with a double deck floating roof

EI
1  t2

(9)

where E is the Youngs modulus of elasticity and I is the second


moment of inertia of the roof cross section. Substituting Eq. (5)
into Eq. (8), we will have

051801-2 / Vol. 134, OCTOBER 2012

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Transactions of the ASME

 R

1
X

1
X

1
MR rdr
2
r0 i1
j1
!
!

R
1
1
X
X
1
Bi Xi00
Bj tXj00 r
 DR rdr

2
r0 i1
j1
!
R
1
1
1X
1X
Bi tXt0  2
Bi Xi

r i1
r0 r i1


 X
1
1 1
1X
1
 DR rdr
Bj Xj0  2
Bj Xj

r j1
r j1
2
!
!
R
1
1
1
X
1X
1X
00
0
Bi Xi
Bj Xj  2
Bj Xj DR rdr

r j1
r j1
r0 i1

#
R "X
1
ek
J1 ej r=R
H
Aj
tanh ej

J1 ej
R
r0 j1 R
"
#

1
X J1 el r=R
1
Al

 qL rdr
J
2
e

1 l
l1
#
#"
R "X
1
1
X
J
e
r=R
1
K

Aj
qL rdr
B_ i Xi r x_ g
J1 eK
r0 i1
j1
"
#"
#
R
1
1
2p
X
X
1
qL  g
Bi Xi
Bj Xj rdr 
cos2 h d h

2
r0
h0
i1
j1
B_ i Xi

B_ j Xj

Fig. 2 Sketch of radial and vertical pressure distributions on


the tank wall
Table 1
Input ground motion
Kobe (1995)
Imperial Valley (1940)
Tokachi-oki (2003)
Tohoku (2011)

Seismic data used in this study


Duration (s) PGA (g) Epicentral distance (km)
40
40
290
300

0.4862
0.3130
0.0800
0.1397

8.7
13
227
364

(10)
Referring to the principle of variation, time integral of variation of
Lagrangian vanishes in any desired time interval
t2

dLdt

t1

t2

CB dBj CA dAj dt 0

(11)

t1

where coefficients CB and CA are defined as


!
!
R
R
1
1
X
X
00

CB
Bi Xi Xj00 DR rdr
Bi Xi Xj MR rdr
r0

r0

i1

!

i1


1
1
1 0 1
Xj  2 Xj DR rdr
Bi Xi0  2
Bi Xi
r i1
r
r
r0 r i1
!
!
!
R " X
1
1
1 1
1X
1X

Bi Xi0 Xj00
Bi Xi00 Xj0  2
Bi Xi Xj00
r i1
r i1
r i1
r0
! #
1
1X
 2
Bi Xi00 Xj DR rdr
r i1
#
R "
1
X
J1 ej r=R
_

r
xg
Xj qL rdr
Aj
J1 ej
r0
j1
!
R
1
X

Bi Xi Xj qL grdr
(12)

r0

CA

1
X

1
X

i1

#

"X

1
ek J1 ek r=R
H
J1 el r=R
1
Al
tanh ek
 qL rdr
R
J1 el
2
r0 R J1 ek
l1

 #


R "X
1
el J1 el r=R
H J1 ek r=R
1

Al
tanh el
 qL rdr
J1 el
R
J1 ek
2
r0 l1 R
#


R "X
1
J1 ek r=R

qL rdr
B_ i Xi
J1 ek
r0 i1

Fig. 3 Time histories of different ground motions: (a) Kobe,


(b) Imperial Valley, (c) Tokachi-oki, (d) Tohoku



MR P B DR Q qL gUfBg qL T A_ qL xg fFg 0
(13)


T t  B_  Sf Ag 0

The significance of compatibility condition between roof and liquid is noticed by Eq. (14) as it, evidently, is a restatement of Eq.
(3). The shape factor A can be rolled out by substituting its equivalent form from Eq. (14) into Eq. (13) so as to have


In order to satisfy Eq. (11), both coefficients are equated to zero


yielding the following matrix equations:

(14)

 
MR P qL T  S1 T t  B DR Q qL gUfBg
qL xg fFg 0

Journal of Pressure Vessel Technology

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

(15)

OCTOBER 2012, Vol. 134 / 051801-3

Fig. 4 Velocity response spectra of different ground motions: (a) Kobe, (b) Imperial Valley, (c) Tokachi-oki,
(d)Tohoku

Table 2

The data used in this study

Contained liquid height, H (model A)


Contained liquid height, H (model B)
Roof areal mass density, MR
Roof bending rigidity, DR (model A)
Roof bending rigidity, DR (model B)
Roof thickness, h
Liquid density, qL
Damping ratio, unroofed tank
Damping ratio, roofed tank
Effective cross-sectional coefficient (model A)
Effective cross-sectional coefficient (model B)
Poissons ratio, 
Earth acceleration of gravity, g
Number of modes adopted for liquid sloshing
Number of modes adopted for roof vibration

15 m
20 m
120 kg=m2
1:612  105 kNm
3:022  105 kNm
800 mm
800 kg=m3
0:001
0:01
1050 mm2
1200 mm2
0:3
9:81 m=s2
29
5

Fig. 5 Lowest two natural periods (T1 , T2 ) of assumed models: (a) model A: H 5 15, (b) model B: H 5 20

051801-4 / Vol. 134, OCTOBER 2012

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Transactions of the ASME

As the equation suggests, there exists both dynamic and static


couplings between the roof and oscillating fluid, resulting in a
coupled set of equations. The number of equations is the same as
the number of modes considered in evaluating roof displacements.
Elements of matrices P; Q; U; F; S, and T are specified as follows:
Pij

R
Xi Xj rdr



J1 ej r=R
  rdr
Tij
Xi
J1 ej
r0
R
rXi rdr
Fi
r0
 

R
ej H
1
1 2
Sjj ej tanh
2
R
ej
R



R
1 0 1
1 0 1
00 00
Xi  2 Xi
Xj  2 Xj rdr
Xi Xj rdr
r
r
r
0
0 r



R

1 0 1
1 0 1
00
00

X j  2 Xj Xj
Xi  2 Xi rdr
Xi
r
r
r
r
0
R
Uij Xi Xj rdr
Qij

(16)

Fig. 6 Response of model A to Kobe earthquake: (a) maximum


sloshing height, (b) maximum lateral force, (c) maximum overturning moment

In Eq. (15), the mass matrix is composed of two parts: Matrix P


accounting for the presence of floating plate and matrix TS1 T t , representing the added inertial effect of the liquid. Likewise, stiffness
matrix is composed of matrix Q accounting for the plate stiffness and
matrix U representing the added stiffness due to fluid covibration.
Linearized Bernoullis equation is used to evaluate the pressure
acting on the wall

Fig. 7 Response of model B to Kobe earthquake: (a) maximum


sloshing height, (b) maximum lateral force, (c) maximum overturning moment

Journal of Pressure Vessel Technology

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

OCTOBER 2012, Vol. 134 / 051801-5

Fig. 9 Response of model B to Imperial Valley earthquake: (a)


maximum sloshing height, (b) maximum lateral force, (c) maximum overturning moment
Fig. 8 Response of model A to Imperial Valley earthquake: (a)
maximum sloshing height, (b) maximum lateral force, (c) maximum overturning moment

p qL

@U
 qL gg
@t

(17)

The first term in the above equation stems from the liquid movement and the second is the hydrostatic pressure imposed by the
elevated liquid. Figure 2 shows the pressure profiles on front and
top cross sections. The net lateral force and overturning moment
are evaluated by adding up the projection of all radial forces along
the excitation direction using the following formulae:

Fx

H 2p
z0

Mx

pRdhdz

(18)

zpRdhdz

(19)

h0

H 2p
z0 h0

Numerical Results

To investigate the sloshing response of the tank, following four


sets of ground motion data were utilized as base excitation: 090
component of 1995 Kobe earthquake at Japan recorded at NishiAkashi station (KOBE/NIS000), 180 component of 1940 Imperial
Valley earthquake recorded at El Centro station (IMPVALL/IELC180), EW component of 2003 Tokachi-oki earthquake
recorded at Tomakomai station (HKD1290309260450EW), and
EW component of 2011 Tohoku earthquake (eastern Japan earthquake) recorded at Chiba station (CHB0091103111446). Table 1
shows the duration, peak round acceleration, and epicentral distance of earthquake records. Time histories and velocity response
spectra (VRS) of the records are shown in Figs. 3 and 4.
Kobe and Imperial Valley ground motions are classified as nearsource earthquakes and have rather similar frequency content containing spectral peaks at low periods. On the other hand, Tokachioki and Tohoku are classified as long-period, far-field records, with
broad band frequency content. Near-source earthquakes act in a
shorter time interval but impose larger accelerations to the

051801-6 / Vol. 134, OCTOBER 2012

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Transactions of the ASME

structures compared to long-period, far-field earthquakes. In this


study, recordings of the aforementioned earthquakes are used as
base excitation for two ranges of tank dimensions with the geometrical and material properties shown in Table 2.
In numerical simulations, Newmarks method is employed to
solve Eq. (15) where the equations can also be modified for the
case of unroofed tank by omitting plate related parameters. Even
though the equations were derived for an ideally undamped system, the actual structures show a rate of energy dissipation [8]
which may be attributed to various sources like viscous interaction
between liquid and tank wall(and roof) and also frictional contact
between roof and wall. In order to account for these terms, damping matrix C is added to Eq. (15) to yield the following more practical equation:

Fig. 10 Response of model A to Tokachi-oki earthquake: (a)


maximum sloshing height, (b) maximum lateral force, (c) maximum overturning moment

 

MR P qL T  S1 T t  B C B_
DR Q qL gUfBg qL xg fFg 0

(20)

The stiffness proportional damping is adopted with the modal


damping ratios shown in Table 2. The selected values for the
roofed (0.01) and unroofed (0.001) cases are in accordance with
the experimental findings of Nagaya et al. [8] and theoretical
assumptions of Matsui [7] where it was also stated that the stiffness proportional damping was appropriate for double deck roofs
while Rayleigh damping was more appropriate for single deck
roofs.
Figure 5 shows the variation of the first and second natural periods with respect to radius for the two models. It is observed that
while the fundamental natural period of the tank is not much

Fig. 11 Response of model B to Tokachi-oki earthquake: (a)


maximum sloshing height, (b) maximum lateral force, (c) maximum overturning moment

Journal of Pressure Vessel Technology

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

OCTOBER 2012, Vol. 134 / 051801-7

altered by the presence of the roof, second natural period of the


system is lowered substantially. This observation is in accordance
with the results of Sakai et al. [3] and Matsui [7]. Higher modes
periods also decrease noticeably though not shown in the figure.
Figures 6 and 7 show the responses of the models to Kobe excitation. Presence of the roof has suppressed the sloshing of liquid
throughout whole radius range, though the absolute value of the
sloshing height is quite negligible relative to the radius (Figs. 6(a)
and 7(a)). However, the values of the force and moment show an
ascending trend for increase in the tank radius for both models
(Figs. 6(b), 6(c), 7(b), and 7(c)). On the other hand, presence of
the roof has hardly affected the resulting lateral force and
moments on the tank wall.
Figures 8 and 9 demonstrate how the models respond to the Imperial Valley excitation. Descriptions given for Kobe results are

also valid here: suppressed sloshing heights (but with a different


trend) (Figs. 8(a) and 9(a)) and an ascending trend in force and
moment diagrams. Therefore, it could be concluded that the presence of the floating roof does not have appreciable effect on the
dynamics of the storage tanks in the case of excitation by nearsource ground motions.
Figures 10 and 11 show the responses of the models to
Tokachi-oki earthquake. Tanks with different radii exhibit different responses to the excitation. While for major parts of the range
the floating roof has a suppressing effect on the wave elevation,
there are also increases occurred for some subranges. For slender
tanks (tanks with smaller radius), roof has a negative (increasing)
effect on the force and moment but as the radius of the tank is
increased positive (reducing) effect of the roof emerges (Figs.
10(b), 10(c), 11(b), and 11(c)). As the fundamental period of the

Fig. 12 Response of model A to Tohoku earthquake: (a) maximum sloshing height, (b) maximum lateral force, (c) maximum
overturning moment

Fig. 13 Response of model B to Tohoku earthquake: (a) maximum sloshing height, (b) maximum lateral force, (c) maximum
overturning moment

051801-8 / Vol. 134, OCTOBER 2012

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Transactions of the ASME

Fig. 14 Dimensionless sloshing heightsinduced by (a) Kobe, (b) Imperial Vally, (c) Tokachi-oki, (d) Tohoku earthquakes (U, unroofed; R, roofed)

Fig. 15 Dimensionless lateral forces induced by (a) Kobe, (b) Imperial Vally, (c) Tokachi-oki, (d) Tohoku earthquakes (U, unroofed; R, roofed)

Journal of Pressure Vessel Technology

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

OCTOBER 2012, Vol. 134 / 051801-9

Fig. 16 Dimensionless overturning moments induced by (a) Kobe, (b) Imperial Vally, (c) Tokachi-oki, (d) Tohoku
earthquakes (U, unroofed; R, roofed)

system falls out of the rich area in the VRS diagram of the earthquake, the absolute value of the sloshing height for both roofed
and unroofed tanks decreases and the force and moment diagrams
lose their increasing trend and remain relatively constant.
The responses of models to Tohoku excitation are depicted in
Figs. 12 and 13. Surface sloshing has been suppressed throughout
the radius range especially at the local peaks (Figs. 12(a) and
13(a)). Force and moment graphs for the roofed and unroofed
cases do not differ significantly except for the dimensions that natural period and earthquake peak coincide (Figs. 12(b), 12(c),
13(b), and 13(c)).
Bearing in mind that the fundamental period of the tank does
not differ much by the presence of the roof, (Fig. 5), the change in
the values of sloshing height (g), lateral force (Fx ), and overturning moment (Mx ) may be associated with the drastic change in the
higher modes periods or the impulsive mode participation. Presence of the roof alters the contribution of higher modes to the
response by rearranging their periods at lower values, thus pushing them out of strong content area. It can be said that if the second mode participation in the response is in-phase with the
fundamental mode, its exclusion will decrease the response and
vice versa.
Responses of the two models are redrawn in terms of dimensionless ratios versus fundamental periods to provide a comparable graph. Figure 14 shows the dimensionless sloshing heights
(g=R) of the two models for all four excitations. It is observed that
while the diagrams of unroofed tanks differ with each other, those
of the roofed tanks overlap over the whole area. This indicates
that, regardless of excitation type (near-source or long-period, farfield), the dimensionless sloshing heights are solely a function of
fundamental period of the system (Fig. 15).
Response of the models in terms of dimensionless forces and
moments are depicted in Figs. 16 and 17, respectively. The differences between the diagrams of roofed tanks suggest that the first
natural period is not the only parameter influencing the dimen-

sionless values of force and moment. It is once again seen that the
roof does not have major effect on the relative force and moment
response of the tanks subjected to short period, near-source excitations Kobe and Imperial Valley.
Figures 17 and 18 compare the effects of different earthquakes
on the models. Short period earthquakes of Kobe and Imperial
Valley cause small sloshing heights due to both their short duration and weak frequency content around the fundamental natural
frequencies. On the contrary, sloshing heights produced by longperiod, far-field earthquakes Tokachi-oki and Tohoku are relatively larger as a result of their longer duration and stronger frequency content around the first and second natural frequencies.
Considering the lateral force and moment imposed on the tank
wall, near-source ground motions compared to long-period farfield records produce larger values. This may be attributed to the
larger acceleration and consequently larger contribution of impulsive mode for near-source ground motions.
In order to reveal the pattern and intensity of stresses occurred
within the roof, two certain tank dimensions (small tank,
H 15; R 25; large tank, H 20; R 40), one from each
type, were selected to be investigated. Radial and circumferential
moments are calculated using following formulae:
 2


@ g
1 @g 1 @ 2 g
(21)


Mr DR
@r 2
r @r r 2 @h2
Mh DR



1 @g 1 @ 2 g
@2g
2 2  2
r @r r @h
@r

(22)

Radial and circumferential stresses (rr ; rh ) are then obtained by


dividing the moments by the effective cross-sectional coefficients
given in Table 2 [9]. Figure 19 shows the radial and circumferential stress patterns in the roof of small tank. It is observed that
peak stresses caused by the near-source earthquakes Kobe and

051801-10 / Vol. 134, OCTOBER 2012

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Transactions of the ASME

Fig. 17 Comparison of responses of model A to different


earthquakes: (a) sloshing height, (b) lateral force, (c) overturning moment

Fig. 19
stress

Fig. 18 Comparison of responses of model B to different


earthquakes: (a) sloshing height, (b) lateral force, (c) overturning moment

Roof stress patterns in model A (H 5 15, R 5 25): (a) maximum radial stress, (b) maximum circumferential

Journal of Pressure Vessel Technology

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

OCTOBER 2012, Vol. 134 / 051801-11

Fig. 20 Roof stress patterns in a floating roofed tank (H 5 20, R 5 40): (a) maximum radial stress, (b) maximum
circumferential stress

Fig. 21 Frequency content of stress at roof (H 5 15, R 5 25): (a) Kobe excitation, (b) Tokachi-oki excitation

Fig. 22 Frequency content of stress at roof (H 5 20, R 5 40): (a) Kobe excitation, (b) Tokachi-oki excitation

Imperial Valley are far larger than the stresses caused by far-field
earthquakes Tokachi-oki and Tohoku. Both types of stress reach
their peak within the range r 0:35R  0:40R with the radial
stress vanishing at the plate edge as anticipated.
Figure 20 depicts the radial and circumferential stress patterns
in the roof of large tank. While the far-field earthquakes result in
stresses qualitatively similar to those occurred in the small tank,
near-source inputs show different stress patterns on the roof. In
this case, the peaks of the graphs have moved toward the inner
radii (r 0:25R). It is again observed that the near-source earthquakes result in more destructive stresses in the roof.

In order to determine the effect of different modes on the bending


stresses, the frequency content of the stress oscillation at the location
of peak stresses for two cases is presented in the following figures.
Figure 21 shows the FFTs of maximum stresses in small tank roof
produced by Kobe and Tokachi-oki earthquakes. It is seen that while
the fundamental mode plays a significant role in the stresses caused
by far-field earthquake Tokachi-oki, second mode is approximately
the sole controlling parameter of stress oscillation in case of Kobe
excitation. This may be attributed to the strong content of the nearsource earthquakes around the higher modes periods. The same explanation is also valid for the case of large tank (Fig. 22).

051801-12 / Vol. 134, OCTOBER 2012

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

Transactions of the ASME

Conclusion

Using Hamiltonian variational principle, the governing equations for sloshing response of floating roofed storage tanks are
derived. The results show that the presence of the floating roof
hardly affects the fundamental period of the system but noticeably
increases higher natural frequencies. Evaluating the response of
tanks with floating roofs to different types of ground motion
records, it is shown that the dimensionless vertical deflection of
roof is solely a function of the tank fundamental period. Also it is
shown that while long-period far-field ground motions give higher
roof deflection, the near-source records give higher value for lateral forces and overturning moments induced on the tank wall.
This indicates that the same design spectrum would not be a
proper reference to evaluate the free board and lateral forces in
the seismic design of storage tanks. Finally, a comparison of stress
patterns in two certain tanks demonstrated the more destructive
effects of near-source earthquakes in comparison to far-field,
long-period earthquakes.

References
[1] Changa, J. I., and Linb, C.-C., 2006, A Study of Storage Tank Accidents,
J. Loss Prev. Process Ind., 19(1), pp. 5159.
[2] Hatayama, K., Zama, S., Nishi, H., Yamada, M., Hirokawa, M., and Inoue, R.,
2005, The Damages of Oil Storage Tanks During the 2003 Tokachi-Oki Earthquake and the Long Period Ground Motions, Proceedings of the JSCE-AIJ
Joint Symposium on Huge Subduction EarthquakesWide Area Strong Ground
Motion Prediction, pp. 718.
[3] Sakai, F., Nishimura, M., and Ogawa, H., 1984, Sloshing Behavior of
Floating-Roof Oil Storage Tanks, Comput. Struct., 19(12), pp. 183192.

[4] Isshiki, H., and Nagata, S., 2001, Variational Principles Related to Motions of
an Elastic Plate Floating on a Water Surface, Proceeding of the 11th International Offshore and Polar Engineering Conference, Stavanger, Norway, pp.
190197.
[5] Robinson, N. J., and Palmer, S. C., 1990, Modal Analysis of a Rectangular
Plate Floating on an Incompressible Liquid, J. Sound Vib., 142(3), pp.
453460.
[6] Amabili, M., and Kwak, M. K., 1996, Free Vibration of Circular Plates
Coupled With Liquids: Revising the Lamb Problem, J. Fluids Struct., 10(7),
pp. 743761.
[7] Matsui, T., 2007, Sloshing in a Cylindrical Liquid Storage Tank With a Floating Roof Under Seismic Excitation, ASME J. Pressure Vessel Technol.,
129(4), pp. 557566.
[8] Nagaya, T., Matsui, T., and Wakasa, T., 2008, Model Tests on Sloshing of a
Floating Roof in a Cylindrical Liquid Storage Tank Under Seismic Excitation,
Proceedings of the ASME Pressure Vessels and Piping Division Conference,
Chicago, IL, Paper No. PVP2008-61675.
[9] Matsui, T., 2009, Sloshing in a Cylindrical Liquid Storage Tank With a Single
Deck Type Floating Roof Under Seismic Excitation, ASME J. Pressure Vessel
Technol., 131(2), pp. 557566.
[10] Virella, J. C., Prato, C. A., and Godoy, L. A., 2008, Linear and Nonlinear 2D
Finite Element Analysis of Sloshing Modes and Pressures in Rectangular Tanks
Subject to Horizontal Harmonic Motions, J. Sound Vib., 312, pp. 442460.
[11] Mitra, S., and Sinhamahapatra, K. P., 2007, Slosh Dynamics of Liquid-Filled
Containers With Submerged Components Using Pressure-Based Finite Element
Method, J. Sound Vib., 304, pp. 361381.
[12] De Angelis, M., Giannini, R., and Paolacci, F., 2009, Experimental Investigation on the Seismic Response of a Steel Liquid Storage Tank Equipped With
Floating Roof by Shaking Table Tests, Earthquake Eng. Struct. Dyn., 39, pp.
377396.
[13] Utsumi, M., and Ishida, K., 2008, Vibration Analysis of a Floating Roof Taking
Into Account Nonlinearity of Sloshing, ASME J. Appl. Mech., 75, p. 041008.
[14] Utsumi, M., and Ishida, K., 2010, Internal Resonance of a Floating Roof Subjected to Nonlinear Sloshing, ASME J. Appl. Mech., 77(1), p. 011016.
[15] Meirovitch, L., 1997, Principles and Techniques of Vibrations, Prentice-Hall,
Englewood Cliffs, NJ, Chap. VII.

Journal of Pressure Vessel Technology

Downloaded From: http://asmedigitalcollection.asme.org/ on 10/10/2013 Terms of Use: http://asme.org/terms

OCTOBER 2012, Vol. 134 / 051801-13

You might also like