You are on page 1of 23

UPLIFTING RESPONSE OF UNANCnORED

LIQUID-STORAGE TANKS

By P. K. MalhotraI and A. S. Veletsos, 2 Members, ASCE


Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

ABSTRACT; A study of the principal effects of base uplifting on the seismic re-
sponse of laterally excited, unanchored, cylindrical liquid-storage tanks is made.
The analysis is implemented by a procedure that is highly efficient and believed to
adequately represent essential aspects of the problem. The method takes due ac-
count of the continuously varying area of base contact and the nonlinearities as-
sociated with membrane action and plastic yielding in the base plate. Critical re-
sponsesof representativetanks are evaluatedfor increasingintensitiesof an earthquake
ground motion, and the effects and relative importance of the problem parameters
are highlighted. The response quantities examined include the pseudoacceleration
of the system, which controls the magnitude of hydrodynamic pressures, the uplift
of the plate at points along its junction with the shell, the plastic rotation at the
plate boundary, and the maximum compressive stress at critical points in the tank
wall. It is shown that the base uplifting may significantlyreduce the hydrodynamic
pressures, but this reduction may still be associated with increased axial stresses in
the tank wall and large plastic rotations at the junction of the shell and base plate.
These actions are particularly prominent for tall, slender tanks.

INTRODUCTION

A sequel to two companion papers ( M a l h o t r a and Veletsos, 1994) this


paper deals with the response to earthquakes of liquid-containing, vertical
cylindrical tanks that are u n a n c h o r e d at the base and may, therefore, uplift
under intense horizontal ground shaking. W h e r e a s the c o m p a n i o n studies
were concerned exclusively with the uplifting resistance of the base plate in
the tank, this study examines the response of the entire tank-liquid system.
The dynamic response of uplifting tanks has been the subject of several
experimental and analytical studies during the past 2 decades. E x p e r i m e n t a l
studies involving both static-tilt tests and dynamic shaking table tests have
been reported by Clough (1977), Niwa (1978), C a m b r a (1982), Manos and
Cough (1982), and Shih (1981); analytical studies of varying scope, effi-
ciency, and sophistication have been r e p o r t e d by Auli et al. (1985), Natsiavas
(1987), Barton and P a r k e r (1987), and Ishida and Kobayashi (1988), among
others. These studies have d e m o n s t r a t e d that base uplifting m a y significantly
influence the dynamic response of tanks and lead to axial stresses in the
tank wall, which are substantially higher than those induced in similarly
excited fixed-base systems.
Although they have been of great help in identifying the effects of some
parameters involved, the analytical studies were generally based on highly
idealized representations of the uplifting action of the base plate and of the
dynamic interaction of the tank with the liquid. T h e r e has been a need for
a more refined yet efficient m e t h o d of analysis, and for a m o r e compre-

IEarthquake Engr., California Strong Motion Instrumentation Program, 801 K


St., Mail Stop 13-35, Sacramento, CA 95814-3531.
2Brown and Root Prof., Dept. of Civ. Engrg., Rice Univ., P.O. Box 1892, Hous-
ton, TX 77251.
Note. Discussion open until May 1, 1995. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on October 25,
1993. This paper is part of the Journal of Structural Engineering, Vol. 120, No. 12,
December, 1994. 9 ISSN 0733-9445/94/0012-3525/$2.00 + $.25 per page.
Paper No. 7215.

3525

J. Struct. Eng. 1994.120:3525-3547.


hensive and critical examination of the effects of all major problem param-
eters. The study reported here is intended to respond to this need. Its
objectives are to: (1) Present a reasonably accurate and efficient method
of analysis for tanks that may uplift under horizontal base shaking; (2)
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

highlight the principal effects of such uplifting on the seismic response of


tanks; and (3) identify the critical responses of such systems and factors that
control them.

SYSTEM CONSIDERED

The system examined is a vertical cylindrical tank of radius R, which is


filled to a height H with a liquid of mass density Pt. The tank is presumed
to rest on a rigid base and to be excited by a horizontal component of ground
motion, the intensity of which is sufficiently high to induce partial uplifting
of its flexible base plate and rocking of the wall, as shown in Fig. 1. Points
in the tank and liquid are specified by the cylindrical coordinate system (r,
dOand z), the origin of which is taken at the center of the base plate.

MODELING OF SYSTEM

It is desirable to begin with a brief review of the response of a fully


anchored tank to a combination of a horizontal and a rocking component
of base motion, xg(t) and ~(t), respectively, as shown in Fig. 2. Such base
motions also develop in laterally excited anchored tanks that are supported

H! ~/////~~//-- - - _- - - -
/
x~ ~ Q~

2 1

(b)

FIG. 1. Tank-Liquid System Considered

3526

J. Struct. Eng. 1994.120:3525-3547.


I

1
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

I
I i
I
I

FIG. 2. Anchored Tank Subjected to Base Translation and Rotation

through a rigid mat on a flexible medium; their effects have already been
investigated (Veletsos and Tang 1990).
The hydrodynamic pressures and forces in such a tank can be expressed
by the sum of two components: (1) An impulsive component, which rep-
resents the effect of the part of the liquid that may be considered to move
in synchronism with the tank wall as a rigid mass; and (2) a convective
component, which represents the effect of the part of contained liquid
undergoing a sloshing motion. The sloshing effects are characterized by long
period oscillations, whereas the impulsive effects are dominated by oscil-
lations of a much shorter period. Because of this difference in their dominant
periods, in most cases these two actions may be considered to be uncoupled.
Further, the contribution of the convective component of the response is
usually small and can be neglected. Finally, for relatively broad tanks, with
height-to-radius ratio of less than 1.5, the contribution of higher impulsive
modes of vibration to the response can also be ignored (Veletsos et al.
1992), and the tank-liquid system may be considered to respond as a single-
degree-of-freedom (SDOF) system in its fixed-base condition.
Hydrodynamic wall pressures for the tank, under these conditions, may
be expressed as (Veletsos and Tang 1990)
p(z, ~b, t) = eL(z)piRA(t)cos ~ (la)
where e~(z) = a dimensionless function, which defines the heightwise vari-
ation of pressure; and A(t) = instantaneous pseudoacceleration of a sim-
ilarly excited SDOF system. The computation of A(t) is described later.
Similarly, the pressures on the tank base may be approximately expressed
as

p(r, +, t) ~ ~(r)plRA(t)cos (b (lb)


where c~(r) = a dimensionless function of the radial position coordinate.
3527
J. Struct. Eng. 1994.120:3525-3547.
The "exact" .expression for base pressures also includes a term that is pro-
portional to t~. However, this term, which represents the effects of the part
of the liquid considered to rotate rigidly with the tank base, is generally
quite small and is neglected in the following development.
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

With the hydrodynamic forces established, the base shear and moment
may be determined by integration. The base shear, V(t), and the moment
at a section immediately above the base, M(t), may be expressed as

V(t) = mA(t) (2a)

and

M(t) = mhA(t) (2b)


where m = effective or modal mass of the liquid; and h = height at which
the resultant of wall pressures acts. Similarly, the overturning moment at a
section immediately below the tank base, the so-called foundation moment,
M'(t), may be expressed as
M'(t) = m(h + 8h)A(t) (2c)
where the term involving the ~h factor represents the overturning effect of
hydrodynamic base pressures, and h + ~h represents the height at which
the mass m or force V(t) must be concentrated to yield the correct M'(t).
The pseudoacceleration A(t) in (1) and (2) is defined by

A(t) = to/I~/1%/T-Z-~-
~2 [5~g(-r)+ (h + 8h)~('r)]

exp[-~to(t - x)lsin[&(t - "r)l dr (3)

in which to = undamped circular n a t u ~ u e n c y of the tank-liquid system


in its fixed-base condition; & = coV1 - ~2 = corresponding damped fre-
quency; and ~ = modal damping factor. The frequency to is given by

to = 2~rf = ~ (4)

where f = frequency in hertz; C1 = a dimensionless frequency coefficient;


and E and p = Young's modulus of elasticity and mass density of the tank
material, respectively:
The quantities m, h, ~h and C1 are functions of the aspect ratio for the
tank, H/R; the ratio of the effective uniform tank wall thickness to the tank's
radius, he~R; Poisson's ratio for the tank material, v; and ratio of the densities
of the liquid to the tank wall material, Pt/P- Selected values of these quantities
for different H/R and he/R ratios are reproduced from Veletsos and Tang
(1990) in Table 1. The effective mass m in this table is expresse_d in terms
of the total mass of the liquid in the tank, mr; and the heights h and/~ +
~h are expressed in terms of liquid height, H. These values are for steel
tanks filled with water, for which v = 0.3 and P t / P = 0.127. The pressure
distribution functions cx(z) and et(r) for steel tanks with h JR = 0.001 are
shown in Fig. 3, for three different values of H/R.
When used in conjunction with (1) and (2), (3) completely defines the
magnitudes of hydrodynamic pressures and associated tank forces. The first
3528

J. Struct. Eng. 1994.120:3525-3547.


TABLE 1. Dynamic Properties of Fundamental Impulsive Mode of Vibration for
Steel Tanks Filled with Water (Veletsos and Tang 1990)
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

he I"
~-
(1)
~ = 0.5

(2) (3) (4) (5)


(a) Values of m/mt
0.0005 0.300 0.549 0.694 0.695
0.001 0.304 0.554 0.700 0.703
0.002 0.312 0.565 0.711 0.713
(b) Values of h/H
0.0005 0.385 0.415 0.491 0.547
0.001 0.387 0.417 0.492 0.548
0.002 0.392 0.420 0.494 0.549
(c) Values of (h + ~h)/H
0.0005 1.455 0.709 0.543 0.562
0.001 1.428 0.706 0.543 0.563
0.002 1.384 0.700 0.544 0.564
(d) Values of Cx
0.0005 0.0506 0.0620 0.0637 0.0563
0.001 0.0719 0.0875 0.0896 0.0792
0.002 0.1019 0.1231 0.1254 0.1108

H
~-=3

z_.
H
0.5
o
-'1
/ o.5
i

i
o i o , , , , , , , , , 1

o 0.5 0,5 0
(a) a,(:.) (b)

FIG. 3. Functions (~(z) and (~(r) for Impulsive Components of Hydrodynamic Pres-
sures (Veletsos and Tang 1990)

3529
J. Struct. Eng. 1994.120:3525-3547.
term in (3) involving the acceleration Yg represents the effect of base trans-
lation, whereas the term involving the acceleration ~ represents the effect
of base rocking. The significance of the two components of the latter term
may be understood by considering the base plate to be separated at its_..
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

boundary from the tank wall. The component involving the product ht~
represents the effect of the rocking motion of the tank wall when the base
plate remains in its horizontal position, whereas the component involving
the product ~h~ represents the effect of the rigid body rotation of the base
plate when the wall remains in its vertical position. It should be recalled
that the pressures induced by each of these actions is considered to deform
the tank wall in its fundamental impulsive mode.

Unanchored Tank Model


An important difference between the system just described and the unan-
chored system shown in Fig. 1 concerns the motion experienced by the base
plate. In the former system, the entire base plate rotates as a rigid body,
whereas in the latter, only a small portion participates in such motion. From
the results presented later in this paper, it is seen that, for practical cases,
the uplifting area of the base plate seldom exceeds 10% of the total base
area. In the evaluation of hydrodynamic action it must, therefore, be rea-
sonably accurate to consider the effects of both the translational and ro-
tational components of the tank wall motion, and to ignore the effects of
the rotation of the base plate itself. This is tantamount to deleting the term
~h in (3). The direct evaluation of A(t) from (3) is, of course, not possible
since the quantity ~(t) is not known. Its evaluation is described in a later
section.
Within the framework of the aforementioned approximations, the uplift-
ing tank-liquid system may be represented by the model shown in Fig. 4,
in which the rotational base spring represents the rocking resistance of the
base plate, the damping coefficient c refers to the fixed-base condition of
the system, and the mass m is considered to be concentrated at a height
/7 (not h + ~h). The relationship between the base moment M and the
spring rotation + is established by the method of analysis reported in Mal-
hotra and Veletsos (1994b). In this approach the tank's base plate is modeled
by uniformly loaded, semi-infinite, prismatic beams that are connected at
their ends to the cylindircal tank wall. For the results presented in this
paper, only eight beams, each of uniform width b = 2~rR/8,were used to
model the base plate. This number is considered to be adequate, as the use

c h

w ~u
/~/i/i/i i/i/i/L/// /
(b)
I~ ~I ~ ~I
R R
FIG. 4. Model of Unanchored Tank-Liquid System

3530

J. Struct. Eng. 1994.120:3525-3547.


of twice as many beams has been shown to imperceptibly change the uplifting
moment-rotation relationship of the base plate (Malhotra and Veletsos 1994b).
Both the temporal and spatial variations of the base pressures are accounted
for in the model by appropriately changing the intensity of uniform pressure
over the uplifted portion of each beam.
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

In addition to the validity of the aforementioned approximation, the


analysis presented here assumes that, in its fixed-base condition, the tank-
liquid system responds in its fundamental impulsive mode, and the convec-
tive component of response is negligible.

EQUATIONS OF MOTION
The equations of motion for the model shown in Fig. 4 are formulated
by considering the dynamic equilibrium of the forces acting on the mass and
of the moments about the base. The equilibrium of the forces on the mass
requires that
miio + C(Uo - +ft) + k(uo - Oh) = -ms (5a)
whereas the equilibrium of base moments requires that
[c(Cto - ;-I,h) + k(uo - +h)]h = M(~) (5b)
where Uo = overall horizontal displacement of mass relative to the moving
base; + = rotation of the base; a dot superscript denotes differentiation
with respect to time t; c = 2~oJm = damping coefficient for the tank in its
fixed-base condition; k = m~o2 = corresponding stiffness of the structure;
and M = moment in the base spring, which is a function of the time-
dependent rotation t~.
Because of the nonlinear nature of the M-+ relationship, (5a) and (5b)
are solved incrementally, assuming linear variations within each moment
increment, AM, and each rotation increment Ate, i.e.
AM = K , At~ (6)
where K, = instantaneous value of rotational spring stiffness. With the
prefix A used to represent a small increment for each remaining response
quantity as well, (5a) and (5b) may be written as

-c ta+hj
[ k -k ] ~Auo~ = _ { m A
+
-k k + K,//~ 2 [ a , h J j(t)
}hE
(7)
where it should be noted that the mass matrix is singular (has a zero de-
terminant).
To avoid the problems of accuracy and convergence normally encountered
in the numerical solution of such a system of equations, it is desirable to
reduce it to a single differential equation. This is accomplished by the use
of the following approximate relationship between Auo and AQ, which is
obtained from the second part of equation (7) by assuming that c = 0:

{mbl~ {keh2lmblo (8)

3531

J. Struct. Eng. 1994.120:3525-3547.


In the latter expression ke = effective stiffness of the uplifting system, and
is given by
1 1 /~2
-- = - + -- (9)
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

ke k K,
On substituting (8) into (7) and premultiplying the resulting expression by
the transpose of vector on the right side of (8), one obtains the single
differential equation
mAiio + ceAlto + keAU o = -mA2g(t) (10)
in which Ce = effective damping of the uplifting system, and is given by
ce = c ( k J k ) 2 (11)
Inasmuch as ks is smaller than k, the damping for the unanchored, uplifting
system may be substantially smaller than that for the corresponding an-
chored system. The decrease is associated with the fact that, while it in-
creases the overall motion of the mass m, base rocking reduces the defor-
mation of the structure relative to its rotating base, as well as the rate of
change of this deformation.
In addition to the value of uo, of interest in the solution of (10) are the
values of the base rotation ~, and of the structural deformation u (Fig. 4).
The evaluation of these quantities is described in the following section. In
the interim, the displacements Uo, u, and ~ are interrelated by
uo = u + t~ft (12)

METHOD OF SOLUTION
The solution of (10) is obtained by the linear acceleration method (Clough
and Penzien 1975), in which the acceleration//o is assumed to vary linearly
within each time increment of integration, At. Let the values of//o, lto, Uo
and the rotational spring stiffness K, (and hence of k e and Ce) be known at
time t., and let it be desired to determine the corresponding values at time
t.+a = t. + At. The solution is obtained through the following steps:

1. For the linear variation in ao considered, the expressions for the ve-
locity increment, hlto, and the displacement increment, Auo, are given by
At
Alto = iio(t.)At + Alto--~ (13a)

and
At2 At2 (13b)
Auo = ao(t.)At + ao(t.) ~ + Aao ~ -

These expressions are substituted into (10) to obtain the following algebraic
equation in A//o:

m + c~-~ + ke Aiio

= - r n h 2 ~ - Ceao(t,,)zXt - ke uo(tn)At + a o ( t . ) - (14)

3532

J. Struct. Eng. 1994.120:3525-3547.


2. Using the value of A//o determined from (14), Auo is first computed
from (13a), and the acceleration of velocity at time tn+l are then obtained
from
ao(tn+~) = //o(tn) + Aao (15a)
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

Uo(t. + 1) = ao(t ) + Aao (15b)


3. The moment in the spring at time t.+l is obtained from
M(tn+l) = - [ a o ( t , + t ) + Yg(tn+l)]mh (16)
4. The base pressure for the uplifted region of the plate is computed by
superimposing, on the hydrostatic effect, the corresponding hydrodynamic
effect. The hydrodynamic component of the pressure is determined from
(lb), assuming that its radial variation within the uplifted region is negligible
and using the pressure corresponding to the plate boundary (r = R). Fur-
ther, the pseudoacceleration in this step is determined from A(tn+l)
-//o (t, + ~) - Yg(t.+ 1), which is reasonable for the small values of damping
involved. The resulting expression is
p ( R , +, tn+l) = p,gH - a(R)p,R[iio(tn+l) + • + (17)
5. By applying the method of analysis for the uplifting plate, referred to
in the previous section, the spring rotation t~ is computed from the base
moment and base pressure determined in steps 3 and 4. In the process of
computing +, one also evaluates the uplifting forces and plastic rotations at
the junction of the base plate and tank wall.
6. The displacements u and Uo at t. + 1 are finally determined from
U(/~+I) = -- {[tJo(tn+l) + JCg(tn+l)]m + CeUo(tn+l)}/k (18a)
Uo (t. + ~) = u(t,+ a) + , ( t , + 1)h (18b)
Additionally, the pseudoacceleration in (la) and (2), for the hydrodynamic
wall pressures and associated tank forces, is computed from
A(tn+l) = (D2U(/n+I) (19)
It should be noted that A (t) in this approach is determined without reference
to (3).

NUMERICAL SOLUTIONS
Two tanks, one broad and one slender, with the following characteristics
are examined:

Broad Tank
Designed in accordance with the American Petroleum Institute (API)
Standard 650 ("Welded" 1988), this is a large capacity, steel tank of 80 ft
(24.4 m) radius and 64 ft (19.5 m) height, which is filled with water to a
height of 48 ft (14.6 m). The tank wall consists of eight courses, each 8 ft
(2.44 m) high with thicknesses ranging from 1.14 in. (2.90 cm) for the bottom
course to 0.31 in. (0.79 cm) for the top. A uniform base plate thickness of
h = 0.57 in. (1.45 cm) is assumed. The material properties for the tank are
as follows: Young's modulus of elasticity E = 29 x 103 ksi (200 GPa); yield
stress ~ry = 36 ksi (248 MPa); and Poisson's ratio v = 0.3. The unit weights
of water and the tank material are ",// = Ptg = 62.4 pcf (9.81 kN/m 3) and
3533

J. Struct. Eng. 1994.120:3525-3547.


y = pg = 490 pcf (77 kN/m3), respectively. The total weight of the tank
wall, W = 846 kips (3.76 MN). System parameters for the first impulsive
mode of vibration are obtained from Table 1 for a value of H/R = 48/80
= 0.6, and a uniform effective shell thickness of h e 0.80 in. (2.03 cm).
=
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

These parameters are: mg = 21,272 kips (94.7 MN);/~ = 226 in. (574 cm);
and f = ~o/2~r = 3.94 Hz. Modal damping for the system in its fixed-base
condition is assumed to be 2% of its critical value. The elastic boundary
constraints for the base plate were evaluated for the base course thickness
of the shell, i.e. hs = 1.14 in. (2.90 cm).

Tall Slender Tank


The second tank, also examined in an Electric Power Research Institute
(EPRI) study (1991), is a relatively tall, slender tank of 20 ft (6.10 m) radius
filled with water to its full height of 37 ft (11.3 m). The shell thickness varies
from 0.38 in. (0.96 cm) for the bottom course to 0.19 in. (0.48 cm) for the
top. The thickness of the base plate is taken as h -- 0.19 in. (0.48 cm). The
material properties of this tank are the same as those for the broad tank,
and the total weight of the tank wall, W = 44 kips (196 kN). Determined
from Table 1 for H/R = 1.85 and an effective uniform shell thickness of he
= 0.23 in. (0.58 cm), the system parameters for the first impulsive mode
of vibration are mg = 2,012 kips (8.95 MN), h = 214 in. (544 cm), and f
= c0/2~r = 6.43 Hz. Modal damping for the tank, in its fixed-base condition,
is again considered to be 2% of its critical value. The elastic boundary
constraints for the base plate were evaluated for the base course shell thick-
ness of hs = 0.38 in. (0.96 cm).

Ground Motions
The ground motions considered are scaled versions of the first 6.3 s of
the N-S component of the 1940 E1 Centro, California earthquake ground
motion record; the acceleration, velocity and displacement traces for which
are shown in Fig. 5. The pseudoacceleration response spectra for systems
with 2% and 5% damping subjected to this ground shaking are shown in
Fig. 6.

Response of Broad Tank


The top two curves in Fig. 7 represent the pseudoacceleration histories
for the anchored and unanchored tank-liquid systems subjected to the E1
Centro earthquake record. (Unless otherwise noted, the responses in Figs.
7-10 and 13 are for the unanchored condition.) It is observed that, compared
to the response history of the anchored tank, the history of the unanchored
tank (1) Has longer periods of oscillation and smaller amplitudes; (2) exhibits
nearly uniform amplitudes for a major part of the response; and (3) decays
at a much slower rate after the end of the excitation.
The first trend may be appreciated by noting that the unanchored system
is the "softer" of the two, with the longer effective natural period. In fact,
if the unanchored system were linear, its natural period 7" would be related
to that of the anchored system, T, by the equation

- x/-;~ (20)

in which Uo and u are the instantaneous values of the deformations identified


in Fig. 4. For the example considered, the histories of these deformations
3534

J. Struct. Eng. 1994.120:3525-3547.


/~ 0.31g
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

~ v ^ ":
~ 15I 1 ~ ~ 14~ in/see/ ' ~

~-1 / - - - 8.28 in

l I I I I I I I

0 1 2 3 4 5 6 7
t, sec
FIG. 5. N-S Component of 1949 El Centro Earthquake Ground Motion Record

are displayed at the bottom of Fig. 7. If one uses the peak values of Uo and
u, which are 1.09 and 0.44 in. (2.77 and 1.12 cm), along with T = 1 / f ~-
0.25 s, one obtains a value of T = 0.4 s; this value is very close to the
average period of the dominant large-amplitude oscillations in the response
history of the unanchored system. The increased effective period of the
uplifting system also explains the reduced value of the peak pseudoacce-
leration. The reduction, however, is even greater than would be expected
from the response spectra plots in Fig. 6 for an increase in period from 0.25
to 0.40 s (a decrease in frequency from 4 to 2.5 Hz).
The nearly constant pseudoacceleration amplitudes in the response history
of the uplifting system imply nearly uniform moment amplitudes, as the two
quantities are proportional to each other. This nearly uniform response is
attributed to the nature of the M-+ diagram, (shown in Fig. 12) which shows
only a slight increase in moment resistance for values of ~ > 0.05 ~
Finally, the reduced rate of decay in the free vibrational segment of the
response of the uplifting system implies reduced damping for this system.
This reduction, as already noted, is due to the reduced deformation and
associated velocity of the superstructure. Similar reductions in overall system
damping have been found in studies of soil-structure interaction (Veletsos
and Nair 1975), for which the rocking action of the foundation is due to the
3535

J. Struct. Eng. 1994.120:3525-3547.


51
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

= 2%
.2F 4-

o 3-

t~
=~ 2-
~-( = 5%

11 I I I I I I I I I I

0 2 4 6 8 10
Frequency, Hz
FIG. 6. Pseudoacceleration Response Spectra for Systems Subjected to El Cen-
tro Record

flexibility of the underlying soil. Further information on the overall damping


of the uplifting tank is given in a later section.

Response of Tall Tank


This tank's response is evaluated for a scaled down version of the El
Centro earthquake record, taking the peak ground acceleration as • =
O.15g. Considering that the tank's height is nearly the same as that of the
tank examined earlier but that its radius is only one-fourth as large, and
recalling that the stiffness of the rotational base constraint K, is proportional
to R 3 (Malhotra and Veletsos 1994b), one would expect significantly larger
base rotations and longer periods of oscillations for this tank. That this is
indeed the case is demonstrated by the pseudoacceleration and deformation
response histories shown in Fig. 8 .
The ratio of natural periods T/T computed from (20) is 5.3 for the tall
tank and 1.6 for the broad tank. The significant increase in time period, for
the tall tank, explains the corresponding reduction in pseudoacceleration
response amplitudes.
The deformation histories shown in the lowermost plots of Fig. 8 reveal
that, after uplifting, nearly the entire deformation of the system is concen-
trated in the base plate, with the superstructure (tank wall-liquid system)
experiencing a nearly rigid body rocking motion. The component of system
3536

J. Struct. Eng. 1994.120:3525-3547.


Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

^.A^A^A A AA AA^ ^^_.


11 -vvvvvv
~ -21
.~ -3 j

!- -vvvvvv ARKAn
VVVVV Uo

0 __^^A.AAAAAAA
AAi
vvvvvvvvvvvvv
VVV
~-1
! I I I I I I I

0 1 2 3 4 5 6 7
t, sec
FIG. 7. PseudoacceleraUon and Deformation Histories for Broad Tank Subjected
to El Centro Record

damping due to the structural deformation is, therefore, practically negli-


gible in this case.

Critical Responses
The performance of the unanchored system cannot be completely assessed
by pseudoacceleration histories. The critical responses for such systems are
the maximum base uplift and plastic rotation at the plate-shell junction, and
the maximum compressive axial stress in the tank wall. For the ground
excitation considered, these quantities occur at the extreme left and extreme
right of the base, represented by points 1 and 2 in Fig. 1. The base uplift
and plastic rotation at these points are, of course, equal to those for cor-
responding beams in the base plate model. The response histories of these
quantities for the two tanks examined are shown in Figs. 9 and 10. The
maximum base uplifts of 5.5 in. (14.0 cm) for the broad tank and 2.6 in.
(6.6 cm) for the tall tank call for appropriate flexibilities in the design of
any piping that may be attached to the tank wall. Further, because of its
cyclic nature, the plastic rotation is a potential source of fatigue damage at
3537

J. Struct. Eng. 1994.120:3525-3547.


'~1764S illlllAnA
llllnl
IAA
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

--vwuu U U vUV U U ' b '

AA ^k

~-1
I I I I I I I I

0 1 2 3 4 5 6 7
t, sec
FIG. 8. Pseudoacceleration and Deformation Histories for Tall Tank Subjected to
Scaled El Centro Record with ~g = 0.15g

the plate-shell junction. For the mild-steel plates used in these two tanks,
the maximum plastic angles of 8~ and 14 ~, computed for the broad and tall
tanks, are not deemed to be unduly high.
The maximum compressive axial stress in the tank wall is evaluated ap-
proximately, following a suggestion made on the basis of experimental evi-
dence by Cambra (1982). Specifically, it is assumed that the axial stresses
induced, after uplifting, by the concentrated force at the point of base
rocking extends over an arc suspended by a central angle of 40 ~ and the
stress distribution over this arc is triangular in shape, with its peak at the
center. This is admittedly a fundamental approximation, which requires
further study. If Q1 is the reaction at point 1 when the tank base is rocking
about that point (see Fig. 1), then the maximum compressive stress com-
puted by this approach is given by
9 Ol
~r - (21)
~r R h ,

Incidentally, the reaction Q1 is simply the force at the end of the beam
3538
J. Struct. Eng. 1994.120:3525-3547.
6t
~0
9

I r
A |'
A^^^^A^AA^^^.
~:,;,, "i " ~ ;I I ; , , ; ' Itj ,: ~, v
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

i ! t l '."v
,,H
9
iI
'; " i;

~
: 6
For point 1 . ~
For point 2 . . . . . . .
u

~ ,/]../1,
! ! 9 I I I

I c---,.7 ...... , ;-,.- .... ~: . . . . .


t! ! --
" d
a. 8J

..... "'I~-~;,, ,, ',t',:',P,~",l',;, I',I',",'II', '


, , , ,,,,,, il ~ ~j ~ ~j ~i is ii ii i~ ~j ~.,
~'! i!
1 ^ ^ ~~ ~ For anchored tank

0
^l -.~A^ AAII!I!II!!AAm~",
u F--~...~,,.v~,,,,,,,
i i,,,; ,,~, ,,, ,,7,2,,
^,',,,.-.-
-.-~ .... ,
,, ,I , ,, , ~ Ii 4 v
J ,i ,,: "i "; '~'"
I I ! I I I I I

0 1 2 3 4 5 6 7
t, sec
FIG. 9. Critical Responses of Broad Tank Subjected to El Centro Record

passing through point 1. For the tank, before uplifting, the compressive
stress at point 1 is computed from
MR W
cr = + - - (22)
I 2wRhs
where I = "rrR3hs = moment of inertia of the shell cross section.
The compressive stress histories computed from (21) and (22) are dis-
played in the bottom plots of Figs. 9 and 10. As expected, base uplifting
significantly increases the maximum compressive stresses in the tank wall.
The computed values of 8.7 ksi (60 MPa) for the broad tank and 8.15 ksi
(56.2 MPa) for the tall tank are, in fact, larger than those allowed by the
API Standard 650 ("Welded" 1988), in its guidelines to guard against buck-
ling. The allowable values of 7.0 ksi (48.3 MPa) for the broad tank and 7.5
ksi (51.7 MPa) for the tall tank were determined from
l 06h s
F.- 5R + 600N/H (23)

3539

J. Struct. Eng. 1994.120:3525-3547.


Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

, point
For t~ , i
" ,
For point 2 . . . . . . . ~,,

ol I I--~
"--~
I
'. . . .
I
'-----~
I
i'--~ ,L_,
I

LJ |I I I 1
I I
12 j

' f ..... lil


~ 4 iit~J', <'!i, ~ ,'i '
8J " t, ,; -' ', ',,,, \,,
. For anchored tank--~
4 \

[ .,, ,,. ~I:.,,,,,,. ~.,~,; ,.,~,,,, ~ ~ # < ,

I I I I I I I I

0 1 2 3 4 5 6 7
t, sec

FIG. 10. Critical Responses of Tall Tank Subjected to Scaled El Centro Record
with Peak Ground Acceleration ~ = 0,15g

in which F~ = maximum allowable longitudinal compressive stress (in psi);


H and R = liquid height and tank radius in ft; and hs = thickness of the
bottom shell course in inches.

Uplifted Regions of Base Plate


The uplifted region of the base plate for the tall tank is shown in Fig. 11
for selected instants of time. The maximum uplifted area is attained at t =
5.2 s, and is nearly 8% of the total base area. For the broad tank, the
maximum uplifted area is only 6% of the total base area. A small annular
portion of the plate adjacent to the tank wall uplifts even before the ground
shaking starts. As already noted in a companion paper (Malhotra and Ve-
letsos 1994b), this uplifting is due to hydrostatic pressures acting on the tank
wall.

Damping in Unanchored Systems


The damping in an uplifting tank stems partly from the viscous action in
the superstructure and partly from the hysteretic action associated with
3540

J. Struct. Eng. 1994.120:3525-3547.


Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

Uplifted reg

FIG. 11. Uplifted Region of Base Plate at Selected Time Responses of Unan-
chored Tall Tank

yielding in the base plate. As already noted, the effectiveness of the first
source decreases with increasing uplifting. The effectiveness of the second
source may be appreciated from Fig. 12, which shows the M-~ diagrams for
the rotational base springs in the models of the two tanks examined. The
moment M in these diagrams is normalized with respect to W~R, where W~
= 'n',hR2H = total weight of liquid in the tank. Due to the "pinched" nature
of these diagrams the effect of hysteretic damping would not be expected
to be very significant.
An estimate of the overall damping in an unanchored system may be
obtained from the rate of decay of its free-vibrational response amplitudes.
Computed for an initial velocity of 20 in./s (51 cm/s), for which the resulting
maximum structural deformations are comparable to those induced by the
E1 Centro record, the free-vibrational response histories for the broad tank
are shown in Fig. 13. The maximum value of the effective viscous damping
in this case is approximately 3% of the critical value; the corresponding
value for the tall tank was determined to be 5%. These values refer to the
first cycle of vibration. Because of smaller hysteretic loops for subsequent
cycles, the overall damping for these cycles is substantially smaller. It follows
that, despite the hysteretic action of the base plate, the overall damping of
the uplifting system may be smaller than that of the corresponding fully
anchored system. An additional source of damping in unanchored systems
3541

J. Struct. Eng. 1994.120:3525-3547.


0"05S
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

]~ cycles

0.2
I

-0.2 ' -o'.1 ' /~ ' o.1' '


Later ~ I//llJ r degrees

(al For broad tank

0.1-

M
W1R

0.4
i I i I I i I i

-0.4 -0.2 0.2


r degrees

(b) For tall taak


FIG. 12. Base Moment-Rotation Diagrams for Unanchored Tanks

is the energy loss due to the tank wall periodically impacting on the rigid
foundation. This source of damping has not been considered.

Effects of Ground Motion Intensity


The sensitivity of the response to the intensity of ground motion was
investigated by considering scaled versions of the El Centro earthquake
record. Table 2 lists the maximum responses computed for the broad tank
using several values of the peak ground acceleration 2g, in the range between
0.14g and 0.50g. Additionally, selected histories of the normalized response
pseudoacceleration A(t)/2g are shown in Fig. 14. As expected, an increase
in the intensity of ground motion leads to a more compliant system, with a
longer effective natural period T, and to typically smaller values of the
normalized maximum pseudoacceleration A/2g. The absolute values of A
still increase with increasing 2g.
Response quantities, which are most sensitive to changes in the intensity
of ground shaking, are the maximum values of base uplift and plastic rotation
at the plate boundary, Wmax and A0max, respectively. In contrast, the absolute
values of the maximum pseudoacceleration of the system A, and of the
maximum axial stresses in the tank wall r . . . . are much less sensitive. For
example, an increase of over threefold in Yg, from 0.15g to 0.50g, increases
the values of A and ~max by less than 50%. This relative insensitivity to
3542
J. Struct. Eng. 1994.120:3525-3547.
10 ~A . r---For anchored tank
olIIIIIIIIIAAAAAA^^^/^^^o
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

. . . . . . . .

r~
oJVVVI/v
.~ -20-J-

~ AAAAAAAAAAAA ,AAA
:, tVVVVVVVVVVVVVVVV VVV V''
AAAAAA
'
1

AAAAAAAAAA IAA^^
A
.~ u

9 0 AAAAt
jVVVVVV'VVvvvvvvvvvv vv '
I ' 'l i ......'l ' i i I' ' 1
0 1 2 3 4 5 6 7
t, sec
FIG. 13. Free Vibration R e s p o n s e Histories of B r o a d T a n k

T A B L E 2. M a x i m u m R e s p o n s e s of U n a n c h o r e d B r o a d T a n k Subjected to Different
Intensities of El Centro Record

(g) Ye (g) (ksi) (in,) (degrees) (see)


(1) (2) (3) (4) (5) (S) (7)
0.14 3.47 0.49 0.83 0.00 0.00 0.25
0.15 3.40 0.51 6,68 0.07 0.00 0,26
0.16 3.32 0.53 6,89 0.16 0.00 0,26
0,18 3.17 0.57 7.26 0.40 0.0(3 0,27
0.20 3.01 0.60 7.58 0.95 0.40 0,29
0,23 2.74 0.63 7.91 1.93 2,27 0.32
0.26 2.53 0.66 8.19 3.07 4.32 0.35
0.28 2.39 0.67 8.18 3,78 5.37 0,36
0,31 2.26 0.70 8.70 5.50 8,00 0.39
0.40 1.93 0.77 9.53 8.77 11,95 0.43
0.45 1.71 0.77 9,51 8.70 ll.97 0.45
0.50 1.51 0.76 9.38 8.08 11,22 0.44

3543
J. Struct. Eng. 1994.120:3525-3547.
o
t -vvvvvv
VVVVVV
, ~w'v'
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

"~---

-a
~ 3~ ~,=0.20g-~l~,

i. t v
vvvvvgv'vvvvv
:~g= 0.31g---~ ~ --

~7 A AIIAnRAAIIA~AAAA~
Z-2 VVVVVVUVVVVVVV
~g = 0.50g--~

I I I I
0 1 2 3 4 5 6
t, sec

FIG. 14. Pseudoacceleration Histories for Unanchored Broad Tank Subjected to


Different Intensities of El Centro Record

TABLE 3. Maximum Responses of Unanchored Broad Tank for Different Values


of Base Plate Thickness Subjected to El Centro Record

h A O'max Wmax A0max ~'max


(in.) .~g (ksi) (in.) (degrees) (sec)
(1) (2) (3) (4) (5) (6)
0.38 1.95 7.35 5.88 13.50 0.71
0.46 2.08 7.97 5.79 10.94 0.41
0.57 2.26 8.70 5.50 8.00 0.39
0.86 2.68 10.85 4.09 1.90 0.35
1.14 2.89 12.01 2.40 0.00 0.31

ground motion intensity is due to the "flat" nature of the M-t~ diagram,
which limits the maximum force that can be transmitted to the superstruc-
ture. For a fully anchored tank, the increase in A and Crmaxwould be pro-
portional to that in s

Effects of Base Plate and Wall Thicknesses


By virtue of their significant effect on the uplifting resistance of the base
plate (Malhotra and Veletsos 1994b), the thicknesses of the base plate and
3544
J. Struct. Eng. 1994.120:3525-3547.
TABLE 4. Maximum Responses of Unanchored Broad Tank for Different Values
of Shell Wall Thickness Subjected to El Centro Record

hs a heb f W A ~r,a~ Wma~ A0max 7"ma•


(in.) (in.) (Hz) (kips) ~g (ksi) (in.) (degrees) (see)
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

(1) (2) (3) (4) (5) (6) (7) (8) (9)


0.57 0.40 2.75 423 2.89 15.34 6.36 9.06 0.50
0.86 0.60 3.26 638 2.85 10.08 3.22 4.57 0.40
1.14 0.80 3.94 846 3.47 8.70 5.50 8.00 0.39
1.43 1.00 4.19 1058 3.14 7.03 3.12 4.19 0.33
1.71 1.20 4.62 1269 2.44 5.85 1.73 1.74 0.28
ah s Base course shell thickness used in computing uplift resistance of bas plate.
=

bhe = Effective uniform shell thickness used in computing f from (4) and Table 1.

the tank wall are expected to significantly influence the seismic response of
unanchored tanks.
Solutions for the critical responses of the broad tank are presented in
Table 3 for a range of base plate thicknesses. An increase in this thickness
is associated with a reduced effective natural period of the system T, and
with reduced uplifts and plastic rotations at the junction of the base plate
and tank wall. The effect on the values of A and O'max is controlled by the
characteristics of the system and the ground motion, and cannot be predicted
readily.
An increase in the thickness of the tank wall has a threefold effect: (1)
It increases the system's fixed-base natural frequency; (2) it increases the
weight of the tank wall and, hence, the resistance to uplifting; and (3) it
further increases the plate's resistance to uplifting by increasing the resis-
tance against inward movement at the plate boundary. The effects of these
changes on the critical responses of the broad tank are highlighted in Table
4. There are significant reductions in the magnitudes of the base uplift,
plastic rotation in the base plate, and compressive axial stress in the tank
wall.

CONCLUSIONS

An approximate method has been presented for the evaluation of the


response to horizontal ground shaking of liquid-containing, unanchored
cylindrical tanks. The method is believed to account for all important pa-
rameters of the problem. Following are the principal conclusions of the
parametric studies reported here.
The ratio of liquid height to tank radius is the most important single
parameter governing the uplifting response of tanks. Base uplifting increases
the effective period of vibration of the system from that applicable to its
fully anchored condition; larger the height-to-radius ratio, the larger is the
resulting increase. A similar increase in the time period is obtained with an
increasing intensity of ground shaking.
Depending on the characteristics of the system and ground motion, an
increase in the effective period of vibration may increase or decrease the
system's maximum pseudoacceleration. For tank sizes encountered in prac-
tice, base uplifting normally reduces hydrodynamic pressures and the as-
sociated overturning base moment. This reduction is associated with a sig-
nificant amount of base uplift and plastic yielding at the junction of the base
3545
J. Struct. Eng. 1994.120:3525-3547.
plate and tank wall, and, because of the reduced contact of the wall with
the foundation, with significantly increased compressive stresses in the tank
wall.
Base uplifting reduces the component of damping associated with the
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

deformation of the tank, but this reduction is partly compensated by an


increase in damping due to yielding in the base plate. Because of the highly
"pinched" character of the moment-rotation diagram of the uplifted base
plate, the hysteretic component of damping is quite small.
Response quantities most sensitive to an increase in the intensity of ground
shaking are the maximum base uplift and plastic rotation at the boundary
of the base plate and tank wall. In contrast, the hydrodynamic pressures
and axial stresses in the tank wall are much less sensitive to such change.
An increase in either base plate thickness or tank wall thickness reduces
the magnitude of the maximum base uplift and plastic rotation in the base
plate. An increase in wall thickness also reduces axial stress in the tank wall.

ACKNOWLEDGMENTS
This paper is based on the doctoral dissertation the first writer prepared
at Rice University, under the supervision of the second writer. Financial
support for the research was provided by the Electric Power Research In-
stitute (EPRI), and is gratefully acknowledged, as are the many helpful
comments received from Dr. H. T. Tang of EPRI. The preparation of this
paper was supported in part by Project 558223, with Brookhaven National
Laboratory. The first writer also acknowledges the ASCE O. H. Ammann
Research Fellowship.

APPENDIX. REFERENCES
Auli, W., Fischer, F. D., and Rammerstorfer, F. G. (1985). "Uplifting of earthquake-
loaded liquid-filled tanks." Proc., 1985 Pressure Vessels and Piping (PVP) Conf.,
New Orleans, La., Vol. 98-7, 71-85.
Barton, D. C., and Parker, J. V. (1987). "Finite element analysis of the seismic
response of anchored and unanchored liquid storage tanks." J. Earthquake Engrg.
and Struct. Dynamics, Vol. 15, 299-322.
Cambra, F. J. (1982). "Earthquake response considerations of broad liquid storage
tanks." Rep. UCB/EERC-82/25, Earthquake Engrg. Res. Ctr., Univ. of Calif. at
Berkeley.
Clough, D. P. (1977). "Experimental evaluation of seismic design methods for broad
cylindrical tanks." Rep. UCB/EERC-77/lO, Earthquake Engrg. Res. Ctr., Univ.
of Calif. at Berkeley.
Clough, R. W., and Penzien, J. (1975). Dynamics of structures, McGraw-Hill, New
York, N.Y.
Electric Power Res. Inst. (EPRI). (1991). "A methodology for assessment of nuclear
power plant seismic margin." Rep. NP-6041-SL, Electric Power Res. Inst. (EPRI),
Palo Alto, Calif.
Ishida, K., and Kobayashi, N. (1988). "An effective method of analyzing rocking
motion for unanchored cylindrical tanks including uplift." J. Pressure Vessel Tech-
nol., Vol. 110, 76-87.
Malhotra, P. K., and Veletsos, A. S. (1994a). "Beam model for base uplifting analysis
of cylindrical tanks." J. Struct. Engrg., ASCE, 120(12), 3471-3488.
Malhotra, P. K., and Veletsos, A. S. (1994b). "Uplifting analysis of base plates in
cylindrical tanks." J. Struct. Engrg., ASCE, 120(12), 3489-3505.
Manos, G. C., and Clough, R. W. (1982). "Further study of the earthquake response
3546

J. Struct. Eng. 1994.120:3525-3547.


of a broad cylindrical liquid-storage tank model." Rep. UCB/EERC-82/07, Earth-
quake Engrg. Res. Ctr., Univ. of Calif. at Berkeley.
Natsiavas, S. (1987). "Response and failure of fluid-filled tanks under base excita-
tion," PhD thesis, Calif. Inst. of Technol., Pasadena, Calif.
Niwa, A. (1978). "Seismic behavior of tall liquid storage tanks." Rep. UCB/EERC-
Downloaded from ascelibrary.org by GEORGE MASON UNIVERSITY on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

78/04, Earthquake Engrg. Res. Ctr., Univ. of Calif. at Berkeley.


Shih, C. F. (1981). "Failure of liquid storage tanks due to earthquake excitation."
Rep. 81-04, Earthquake Engrg. Res. Lab., Calif. Inst. of Technol., Pasadena,
Calif.
Veletsos, A. S., and Nair, V. V. D. (1975). "Seismic interaction of structures on
hysteretic foundations." J. Struct. Div., ASCE, 101(1), 109-129.
Veletsos, A. S., and Tang, Y. (1990). "Soil-structure interaction effects for laterally
excited liquid-storage tanks," J. Earthquake Engrg. and Struct. Dynamics, Vol.
19, 473-496.
Veletsos, A. S., Tang, Y., and Tang, H. T. (1992). "Dynamic response of flexibly
supported liquid-storage tanks." J. Struct. Engrg., ASCE, 118(1), 264-283.
"Welded steel tanks for oil storage." (1988). AP1 Standard 650, 8th Ed., Am. Pet.
Inst. (API), Washington, D.C.

3547

J. Struct. Eng. 1994.120:3525-3547.

You might also like