Housner - The Behavior of Inverted Pendulum Structures During Earthquakes-1963 PDF
Housner - The Behavior of Inverted Pendulum Structures During Earthquakes-1963 PDF
February, 1963
ABSTRACT
During the Chilean earthquakes of May, 1960, a number of tali, slender structures survived
the ground shaking whereas more stable appearing structures were severely damaged. An
analysis is made of the rocking motion of structures of inverted pendulum type. It is shown
that there is a scale effect which makes tall slender structures more stable against overturning
than might have been expected, and, therefore, the survival of such structures during earth
quakes is not surprising.
lNTRODUCTION
The rigid block shown in fig. 1 will oscillate about the centers of rotation O and
O' when it is set to rocking. It is assumed that the coefficient of friction is suffi
ciently large so that there will be no sliding between the block and the base. The
significant properties of the block are its weight, W, its moment of inertia, 10 , about
1 Housner, G. W., "Limit Design of Structures to Resist Earthquakes," Proceedings of the
1956 World Conference on Earthquake Engineering, Earthquake Engineering Research Institute,
1956.
403
404 BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA
the point 0, and the location of the center of gravity a distance h above the base
and a distance b from the side of the block. As shown in fig. 1, the radial distance
from the center of rotation 0 to the center of gravity is R = V ~ + b2. When
the block is at rest, the line R makes an angle a with the vertical. The tilting of the
b!b
block from the vertical is measured by the angle 0.
0' 0
FI~. 1. A r o c k i n g block.
When the block is rotated through an angle 0, the weight of the block will exert
a restoring moment WR sin (a - 0). The equation of motion is then
d20
Io - WR sin (a - 0) ( 1)
dt 2
For tall, slender blocks having the angle a less than about 20 ° the sine of the angle
m a y be approximated by the angle and equation (1) may be written
[o 0 - WRO = - WRoe
This equation describes the free vibration of the block. Setting W R / I o = p~, the
equation becomes
- p20 = - p 2 o e
tion. The block will fall from 0 = 00 to 0 = 0 in a time t = T / 4 and at this instant
equation (2) becomes
T
0 = c ~ - (c~-Oo) coshp~-
or
eosh p T _ 1
4 1 - 00/a
4
/
/
Tw_R
5
/
2 /
J
J
J
I jJ
.f
/
0o 0.2 0.4 0.6 0.8 ID
eo
o¢.
FIG. 2. Period T of block rocking with amplitude 00.
T = 4 c) o s h - ~_( 1
p i ---04~ (3)
A graph of this equation is shown in fig. 2 where it is seen that the period is strongly
dependent upon the amplitude ratio 00/a. When 00/a is close to unity the period is
long, and when 00/a is close to zero the period is short.
During the rocking of a real block there would be, of course, a dissipation of energy
when the block impacted on the base. Thus, once during each half-cycle there would
be an increment decrease in the energy of vibration. Consequently, the period of
each half-cycle would be longer than that which preceded it. If the impact is as-
sumed to be inelastic (no bouncing), the rotation continues smoothly about the point
0' and the moment of momentum about 01 is conserved. The reduction in kinetic
energy during impact is
406 B U L L E T I N OF T H E S E I S M O L O G I C A L S O C I E T Y OF A M E R I C A
= 1 - ( 1 - ~0) c o s h p t
-- - ( 1 - ~o)p sinh pt
1 - ~-1
and
sinh 2 ptn = (i 1 ~2 1
where ~. is the amplitude following the n-th impact. The kinetic energy just after
impact is reduced to r times the energy just before impact. The velocity is therefore
reduced to x / r times the velocity before impact. This condition m a y be written
1 - (1 - = rr [1 -- (1 --
~ = 1 -- %/1 --r,[1 -- (1 - - ~ 0 ) ~]
A graph of this equation for r = 0.7 is presented in fig. 3 where the decrease in
amplitude for successive n is shown for several values of the initial displacement.
BEHAVIOR OF INVERTED PENDULUM STRUCTURES 407
It is seen that for large amplitudes the energy of vibration decreases rapidly but
for small oscillations the energy decreases slowly.
It follows from equations (3) and (6) that the successive half periods of vibra-
tion during the free rocking are given by
T-2~°Rtanh-1%/r~[l- (1-~o)2 ]
2
According to the foregoing analysis the free rocking of a block proceeds as follows.
At first there will be a few oscillations of large amplitude and slow rocking. The
amplitude decreases markedly with each impact and corresponding with this there
is an increase in frequency of oscillation to an infinite value. Actually, a freely
rocking block does not make a truly inelastic impact, but there is a slight bouncing
and sliding involved. This is another source of energy absorption and under favor-
I.O
r=OY
0.8
0.6
¢.
O.4
0.2
0 2 3 4 5 6 7 8
NUMBER OF IMPACTS
FIG. 3. Amplitude ~ subsequent to n - t h impact.
able conditions the block makes only some eight or ten impacts before being brought
to rest.
It is seen from this equation that a necessary condition for motion to be initiated is
that a/g > a, which specifies the fraction of g acceleration required to begin tilting
408 BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA
the block. The solution of equation (7), subject to the conditions that at t = 0
both the velocity and displacement are zero is
O- (g -1) ( cosh pt - 1)
(x
It is seen that if a/g > ~ and if the acceleration acts for a sufficient length of time,
the block will overturn. It is not necessary, however, for the base acceleration to
act continuously in order to overturn the block. For each value of a there is a time
tl during which the acceleration can generate a velocity sufficient to overturn the
block.
The condition for overturning is that the total work done by the inertial force
Wa/g is just equal to the difference in potential energy between positions 0 =
and 0 = 0. This condition may be written
cosh h = 1 + (9)
Figure 4 is a graph of this equation giving the duration, t,, of ground acceleration a
required to overturn the block.
The foregoing analysis is not realistic for earthquake ground motions since it
assumes constant ground accelerations of finite duration followed by a constant
velocity of the ground. This type of ground motion does not occur during earth-
quakes and hence it is not meaningful to discuss the overturning of blocks in terms
of percent g acceleration.
tW__R
'1 Io
2
\
01.0 1.2 1.4 1.6 1.8 2.0
_q.
go~
FIG. 4. C o n s t a n t a c c e l e r a t i o n a of d u r a t i o n tl r e q u i r e d for o v e r t u r n i n g .
a 1-1-
where
sin ~ -- ga/a; cos ¢ = %/1 -- (ga/a) 2
2.2
2.0
/;/
1.8 /
go~ 1.6
/
1.4
/
J
1.2 J
f
J
J
1.0
0 02. 0.4 0.6 0.8 i.O 1.2 1.4 1.6 1.8 2.0
P
FIG. 5. S i n u s o i d a l a c c e l e r a t i o n p u l s e a s i n cot r e q u i r e d for o v e r t u r n i n g .
Making use of equation (10) and rearranging terms, equation (11)a can be put in
the form
a _ 1 + (12)
ga
or
a _ ~//1 + I0 [27ry
where Tg is the period of the ground acceleration. This relation was first derived by
Kirkpatrick 2 and it expresses the conditions under which a single sine pulse will
overturn the block. As may be verified, this is a minimum condition as it insures
t h a t 0 = 0 w h e n 0 = a.
A graph of equation (12) is presented in fig. 5 for small values of ~o/p For large
values of w/p, say o~/p > 3, equation (12) can be represented by
a
- (12)b
ga p
2 Nirkpatrick, P., "Seismic Measurements by the Overthrow of Columns," Bull. of the
Seism. Soe. of Amer., Vol. 17, No. 2, 1927.
BEHAVIOR OF INVERTED PENDULUM STRUCTURES 411
or
a Jr
ga 2-T°~WRIo
0.)
- a Rg
or
aTg-a'2jr~3Rg2
where T~ is the period of the pulse. If we consider all blocks that have a given length
of diagonal, 2R, the equation may be written
aT~ _ Ka
2
where K is a constant. It is thus seen that the product of the amplitude of the
acceleration pulse by its duration must be proportional to the slenderness angle
a for a given R.
OVERTURNING BY EARTHQUAKE M O T I O N
The preceding paragraphs dealt with the overturning of the block by a square
acceleration pulse and by a half sine-wave pulse and equations (9) and (12) give
the minimum accelerations required to overturn a block by such single pulses.
Smaller accelerations than are specified by these equations may set the block to
rocking but will not overturn it. It is, however, possible to overturn the block with
smaller accelerations if a number of pulses act successively. This would be the
case of strong-motion earthquake ground acceleration, which may be thought of
as a train of pulses which, in general, have periods that are short compared to the
rocking periods of structures. The resulting behavior of the block is indicated by
the following approximate analysis.
For purposes of illustration, consider ground acceleration composed of a sequence
of N discrete step changes ±Av in ground velocity, that are randomly distributed
over time 0 < t < tl, and have equal probability of being positive or negative in
sign. This represents an idealized earthquake ground motion whose average velocity
response spectrum (undamped) is a constant. The effect of such ground motion on
structures is the same as if the ground were at rest and impulsive inertial forces
F~At were acting through the center of mass, where
F~At = ±MAy (n = 1, 2 . . . N)
412 BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA
The response of a linearly elastic structure, such as is shown in fig. 6, to the random
succession of forces F~ is to be excited into oscillations. The effect of the n-th im-
pulse, F,At, is to produce a change in velocity Av, and hence to produce a change in
kinetic energy of vibration
A K E . = ½ M(v~ + Av)2 -- ½ Mv~ 2
where v~ is the velocity of the structure at the time that F~At acts. If a large num-
ber of such idealized earthquakes were to act upon the structure, the A K E ~ would,
~X
in general, be different for each of the earthquakes. However, the average incre-
ment change in energy, A K E , , averaged over all the earthquakes would be
A K E ~ = M v~(Av) + ½M(Av)~
and since v-~, the average velocity is zero and the average Av is also zero, it follows
that
AKE. = ½M(Av)~
E,~ = ½M(Av)2n (n = 1, 2 . . . N )
This represents an average building up of energy with the number of pulses and
hence the average rate of energy input, P, is
1 M(~v)2 N
P = 2 t~
This constant power input P will cause the average amplitude of vibration to in-
crease proportional to %/t-.
If the structure has viscous damping, the power input, P, will build up the
amplitude of vibration until the average rate of energy loss iust balances the input.
BEHAVIOR OF INVERTED PENDULUM STRUCTURES 413
The energy loss per cycle for a linearly damped oscillator is 4~rkE, where ~ is the
fraction of critical damping, and E is the energy of vibration. The average time rate
of energy loss is then 4~r~ P,/T, where T is the period of vibration. An energy balance
for the system then is3
0E_ p _ 4 ~ (13)
Ot T
The solution of this equation is
= ~ 1 -- exp
(v))-- t
AO,~ - RF~ At _ R M A v
Io Io
and hence
A K E ~ -- MR(Av)O~ + 1 ( M R ) ~ (Av)~
2 Io
and the ensemble average is
AKE,~ - 1 ( M R ) 2 (Av)2
2 Io
If there were no energy loss during the rocking, the buildup of energy would give
on the average
o~ "V [o Rg
When the idealized earthquake motion satisfies equation (15) there is a 5 0 per cent
probability that the ground motion will overturn the block. If a is greater than the
quantity on the right side of equation (15) the probability of overturning is less
than 50 per cent, and if a is smaller than this quantity the probability of over-
turning is greater.
If there is an energy loss during the rocking of the block, with a fraction r E of
the energy of vibration being lost at each impact on the base as described by equa-
tion (5), an energy balance would require
0;g -- p --
2rE
Ot T
T = 4 eosh_l 1
P 1 0o
og
PE = WRa 2 --
and for the forced vibration of the block the average instantaneous amplitude of
rocking, O0, is defined by
= WRoe 2 --
or
Ot 2 cosh-1
2oe2 _=
1 -- ~ / ~ (16)
BEHAVIOR 0F INVERTED PENDULUM STRUCTURES 415
This may be compared with equation (13) for the elastic structure. For very small
E, the denominator of the right side approaches a value of 2 and equation (6) has
essentially the same form as equation (13) :
aE _ P _ prE
-- -
at 2
On the other hand, when the amplitude of vibration approaches the unstable point
(E = WR/2a:2) the denominator becomes extremely large and the last term in
equation (16) may be neglected and the equation written
In this case the block is effectively undamped. It is thus seen that, although the
equivalent fraction of critical damping remains constant, as the amplitude of
rocking increases, the rate of energy dissipation decreases strongly and this in tum
promotes an increased amplitude of vibration. It may, therefore, be expected that
the behavior of the rocking block could be quite variable in that relatively small
ground motion may fortuitously build up the amplitude at the beginning of the
ground motion and lead to overturning the block.
! WRa:2 = ! W MR
2
2
Sv
2 2 g Io
or
(17)
It is seen that this has the same form as equation (15). For slender structures
MR2/Io has a value close to unity so that equation (17) can be approximated by
Sv
a: = -== (17 )a
-VgR
This equation may be interpreted as stating that for a given spectrum value Sv ,
a block having an angle a: given by equation (17)a will have approximately a 50 %
probability of being overturned. The dimensions of the structure enter only through
the length R and as equation (17)a shows, the larger the value of R the more stable
against overturning is the block. That is, comparing two blocks of the same propor
tions, the larger block is more stable than the smaller. There is thus an unexpected
scale effect which is explained by the fact that the ground motion is not scaled
416 BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA
with the block. If equations (17)a, (9), and (12)a are compared, setting MR2/I0 =
1 anda = b/h, there is obtained for smalla
Earthquake motion: b = S. �
Square pulse:
Sine pulse:
I
h
It is .seeu-from these equations that the semiwidth b required for stability does not
increase linearly with h as physical intuition indicates, but is affected less strongly,
or in other words, tall slender structures are more stable than might be supposed.
GoLF-BALL WATER TANK
Let h be the height of the center of gravity and D the diameter at the base as
shown in fig. 7, and suppose that the structure is rigid and is free to rock (no anchor
bolts). For a structure of such proportions h may be taken equal to R, and /0 may
be taken equal to MR2 • Equation (17) then becomes
D S.
a =
2h = Vgh
or
BEHAVIOR OF INVERTED PENDULUM STRUCTURES 417
This value of Sv gives 50 per cent probability of overturning, and it is seen that the
result is independent of the mass; that is, the stability is not affected by whether
the tank is completely full or completely empty if h may be taken to be a constant.
Experience with earthquakes in the United States indicates that a value of Sv =
2 ft/sec represents strong ground motion. If the tank has h = 100 ft and D = 10 ft,
the required Sv to give a 50 per cent probability of overturning is
It may be concluded, therefore, that for an Sv = 2 ft/sec, it is more likely that the
tank would remain standing than that it would overturn. If the tank has anchor
bolts that stretch and absorb energy during the rocking, there will be even less
probability that the tank will overturn. It is, therefore, not so surprising that some
golf-ball tanks have remained standing during earthquakes. There is, of course, a
certain probability that a tank of these proportions will be overturned by a ground
motion having Sv = 2.0 or even less than 2.0.
For comparison, suppose that the tank is tightly anchored to the foundation so
that it cannot rock but can only bend and that the period of vibration is 3 seconds.
In this case the maximum shear force generated would be
and
!_ 271" Sv 271"
= = � = 0_13
W T g 3 32.2
Thus the tank should be designed for 13 per cent g. It might appear from this that
it would be advantageous to design tall slender structures to act as rocking blocks;
however, a 50 per cent probability of survival for a rocking structure is not a satis
factory design. At present it is not known how to design a rocking structure to have
specified small probability of failure. It may also be noted that, if the tank is partly
full, the sloshing water introduces another mode of vibration which will affect the
behavior of the structure.
SuMMARY
The free oscillations of a rocking block are analyzed and the rocking period and
energy loss are set forth. The overturning of a block by a constant, horizontal
acceleration, by a single sine pulse, and by an earthquake type excitation are ex
amined. It is shown that there is an unexpected scale effect which makes the larger
of two geometrically similar blocks more stable than the smaller block. It is also
shown that the stability of a tall slender block subjected to earthquake motion is
much greater than would be inferred from its stability against a constant horizontal
force. In the light of these facts, the occasiona! survival of a slender structure that
is apparently highly unstable is not surprising.
CALIFORNIA !NSTITUTE OF TECHNOLOGY
PASADENA, CALIFORNIA