Probabilistic responses of

base-isolated structures to E1
Centro 1940 and Mexico
City 1985 earthquakes
Lin Su and Goodarz Ahmadi
Department of Mechanical and Aeronautical Engineering, Clarkson University, Potsdam,
NY 13676, USA
(Received February 1991; revised July 1991)

Stochastic earthquake responses of a base-isolated shear beam

structure with several base isolation systems, including the laminated
rubber bearing, the resilient-friction base isolator with and without
sliding upper plate, and the EDF system, are studied. The recently
developed stochastic models for the El Centro 1940 and the Mexico
City 1985 earthquakes which preserve the nonstationary evolutions
of amplitude and frequency content of ground accelerations are used
as seismic excitations. The method of time-dependent equivalent
linearization is utilized and the mean-square response statistics of the
base-isolated shear beam structure are evaluated. Statistically
estimated peak responses of the base-isolated structure are compared with the response spectra for the actual earthquake
accelerograms. Comparisons of the results with those for the fixedbase structure show that the base isolation systems can, in general,
be highly effective in reducing the mean square responses of the
superstructure. However, they are sensitive to ground excitations
with considerable energy at low frequencies.

Keywords: earthquake response, base-isolated structures

Base isolation design strategy for protecting structures in

seismic regions has attracted considerable attention in
the last two decades. Kelly t,2, has provided excellent
survey articles on historical developments and recent
progress on the subject. Therefore, works which are
directly related to the present study are only briefly
reviewed.
The laminated rubber bearing (LRB) is the most
extensively used base isolation system ~-3. A number of
buildings all over the world, including several in
California and Utah, have been constructed using
laminated rubber bearing base isolation foundation for
protection against earthquakes. The LRB consists of
alternating layers of rubber and steel with the rubber
being vulcanized to the steel plates. The mechanical
behaviour of this system is shown schematically in
Figure l(a). This schematic diagram also represents the
GERB isolation system which is composed of helical
springs and viscodampers 4.
Combining a rubber bearing and a friction element in
parallel, Mostaghel and co-workers 5-7 developed the
resilient-friction base isolator (R-FBI) system. This
isolator consists of concentric layers of Teflon coated

plates that are in friction contact with each other and

contains a central core of rubber or steel reinforced
laminated rubber. Figure l(b) shows a schematic diagram
of the resilient-friction base isolation system. The
Alexisismon base isolator developed by Ikonomou s has
similar mechanical behaviour.
A friction-type base isolation system was developed
under the auspices of Electricite de France (EDF) 9 for
nuclear power plants in regions of high seismicity. The
main isolator of the EDF system consists of a laminated
(steel-reinforced) neoprene pad topped by a lead-bronze
plate which is in frictional contact with a steel plate
anchored to the structure. The EDF isolator which
essentially uses an elastomeric bearing and friction plate
in series is schematically shown in Figure 1(c). For low
amplitude ground accelerations, the lateral flexibility of
the neoprene pad provides isolation. At high levels of
excitation, however, sliding will occur which provides
additional protection.
The desirable features of the EDF base isolator and
the R-FBI system were combined by Su et al. to who
proposed a new base isolation system with two different
frictional elements. The behaviour of this isolator which

0141-0296/92/040217-14
1992 Butterworth-Heinemann Ltd

Eng. StrucL 1992, Vol. 14, No 4 217

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

_=

ub

ub

41

=J

//

T}

//

I
I
I
I
t
I

_1

/
/

/
/
/
/

//

I
I
I
J
I

V'

/no

mb

--,

= ulg

//////.,'1
~

Structure

Base raft
J Base isolator

K/"////
ndotion

e
Figure 1 Schematic diagram of base-isolated systems. (a), LRB system; (b), R-FBI system; (c),

EDF system; (d), SR-F system, (e), structural model

was called the sliding resilient-friction (SR-F) base

isolation system is shown schematically in Figure 1 (d).
For low level seismic excitation, the system behaves as
a R-FBI unit. The sliding at the top friction plate occurs
only for a high ground acceleration which provides an
additional safety measure for unexpectedly severe earthquakes. An extensive numerical simulation study on performances of the SR-F base isolation system under a
variety of conditions has been carried out 11. Comparative studies of performances of different base isola-

218

Eng. Struct. 1992, Vol. 14, No 4

tion systems for various earthquake excitations have

been presented elsewhere t2-14
To account for the uncertainties involved, the probabilistic response analysis of base-isolated structures
has been considered. The response of a rigid block
sliding on a randomly moving foundation has been
analysed 15-18. The effects of the presence of the rubber
element in addition to the frictional element has been
studied 19,20. A probabilistic comparison of various base
isolation systems for rigid and flexible structures to

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

random white noise models of earthquake ground excitations has also been carried out 2t'22. These studies are of
particular interest for reliability analysis of various base
isolation systems.
The earlier probabilistic response analyses of baseisolated structures were limited to simplified structural
models and highly idealized random earthquake ground
excitations. In particular, probabilistic sensitivity
analysis of base-isolated structures to long-period
ground excitations has not been considered. Recently,
there have been several attempts to develop more
realistic models for simulating ground excitations which
allow for the appropriate nonstationary variations of the
amplitude and the frequency content of earthquakes 23-26. Such stochastic models for the El Centro
1940 and the Mexico City 1985 'earthquakes which are
quite convenient for random response analyses of structures have been suggested 27.
In this work, probabilistic comparative studies of performances of different base isolation systems, including
the laminated rubber bearing, the EDF system and the
resilient-friction base isolator with and without sliding
upper plate, are carried out. A nonuniform elastic shear
beam is used as the structural model. The recently
developed stochastic models 27 for the El Centro 1940
and the Mexico City 1985 earthquakes which preserve
the nonstationary evolutions of amplitude and frequency
content of ground accelerations are used as seismic
excitations. The methods of time-dependent equivalent
linearization and expansion by normal modes are
utilized for the response analysis. The equations governing the second-order response statistics are derived.
Numerical solutions of the resulting set of differential
equations are obtained by using a fourth-order
R u n g e - K u t t a - G i l l scheme. The root mean square
(RMS) responses are evaluated and the effectiveness of
different base isolation systems are studied.
Several Monte-Carlo digital simulations are also performed. Ensembles of 500 samples of earthquake excitations are generated and the response time histories of the
structure are evaluated. Ensemble averaging is used and
the response statistics of interest are determined. The
simulation results are compared with those obtained
from the statistical linearization for tests of accuracy.
Statistically estimated peak responses are compared
with those obtained for the actual accelerograms
and good agreements are observed. It is shown
that the base isolation systems can be quite effective in
reducing the structural response. The sensitivity of
various base isolation systems to long-period ground
excitation is also discussed.

Formulation

02U+~oOU
Ot

Ou(1, t)

u(0, t ) = 0 ,

- - = 0
ax

(2)

where u(x, t) is the deflection of the beam relative to its

base raft, x is the dimensionless height, t is the time, Ub
is the displacement of the base raft relative to the
ground, Co is the damping coefficient per unit mass,
and c is the speed of elastic shear wave. Here, A is the
cross-sectional area, which is assumed to vary as

A(x) = Ao e-Z~

(3)

where Ao is the cross-sectional area at the base of the

shear beam structure, a is a constant related to the
nonuniformity.
The general solution to equation (1) may be assumed
to be given as
N

u(x, t)= ~

q.(t)~.(x)

(4)

n=l

where N is the number of modes of vibration considered

and
q~,,(x) = fl.e ~x sin()~.x)

(5)

are the normal modes. The equations governing the

modal amplitudes q.(t) are given by
2 = - a.(fig + rio),
q. + 2~'.o~.q. + o~.q.
n=l,2

.....

(6)

Here, o~. and ~'. are the natural frequencies and the
damping coefficients of the shear beam structure which
are given by
wE = e20~.2 + a2),

~. _ Co

(7)

where o~t is the fundamental natural frequency. For

a = 0.1, the values of (.0./(~ 1 for the first ten modes of
vibration are listed in Table 1. In equations (6), the coefficients or. and/3, are defined as
.,, -

/3. ),.
)2 + az'

2
(2

-- sin(2X.)/~.)

in

(8)

where X. are the solutions to the eigenvalue equation

A schematic diagram of an elastic nonuniform shear

beam structural model with its base isolation system is
shown in Figure l(e). The equations governing the
motion of the base-isolated shear beam structure subject
to horizontal earthquake ground acceleration, au, are
given by

--Ot2

subject to boundary conditions

1 0 (Ou)
A Ox c2 -~x = -(~g + ab),
"O_<x_< 1 (1)

tan(h.) + ),.la = 0

(9)

For a = 0.1, the first ten values for 9~. are listed in

Table 1.
The equation governing the motion of the base raft is
given as
N

_r,=
mt

-a,

B.O.

(10)

n=l

Eng. Struct. 1992, Vol. 14, No 4

219

Probabflistic responses of base-isolated structures: Lm Su and G. Ahmadi

Table 1

Values of ;% and ~o./w~ for first ten modes of vibration for a = 1

10

,k,
w,,/c~1

1.63
1.00

4.73
2.90

7.87
4.81

11.01
6.73

14. 14
8.65

17.29
10.57

20.43
12.49

23.57
14.41

26.71
16.34

2 9 .8 5
18.26

where m,(m, = mo + too) is the total mass of the entire

structure, and mb and mo are the mass of the base raft
and the mass of the shear beam structure, respectively.
The coefficients B, in equation (10) and the mass
ratio R are defined as

B.

2aRot,
1 - e-2~,

R = --m
mt

(11)

( w ( t ) ) = O,

F; is the force exerted from the base isolator on mb. The

form of F, which depends on the nature of the base
isolator used will be discussed in the subsequent sections.
Ground excitations

The use of a stationary white noise process with a constant power spectrum as a stochastic model for earthquake ground acceleration has been suggested 2s'29. It is
now well known that the assumption of constant spectral
density may be a reasonable approximation only at the
bedrock level. The surface soil layer filters and
amplifies the ground acceleration. To account for these
effects, a filtered white noise model for earthquake
ground motion was suggested by Kanai 3 and Tajimi 31.
The ground acceleration power spectral density is given
by
S(co) = S,

1 + 4~'2(co/cog) 2
2 2
[ 1 - (co/w~)2] 2 + 4fsw
/cog2

where uy is the filter response, e(t) is a deterministic

amplitude envelope function and cos(t) is the timedependent predominant ground frequency. When
e(t) = 1 and %, is a constant, equations (13) and (14)
ensure that a s has the Kanai-Tajimi power spectrum
given by equation (12). In equation (14), w(t) is a Gaussian white noise process with the following statistics

(12)

e(t) = 9.44t 3 e x p ( - 1.1743t) + 3.723

(13)

ay + 2~gcos(t)af + co2g(t)uf = w(t)

(14)

with

Eng. Struct. 1992, Vol. 14, No 4

(16)

and
co~(t) = 7r{ 19.01 [exp(-0.0625t) - exp(-0.15t)]
+3.0}

(17)

For the Mexico City 1985 earthquake, ~'g = 0.12 and

e(t) =

if 0 _ t < 32.0 s

13.978 - 1.073(t - 32.0) if 32.0 _< t _< 42.9 s

2.2823

ag = - [2fgcos(t)u f + co~(t)uyle(t)

(15)

Here, 6( ) is the Dirac delta function, the angular

brackets, ( ) , denotes the expected value (ensemble
average), and So is the contant power spectrum of the
WN process.
For the El Centro 1940 earthquake, ~'g =0.42, as
suggested in the original Kanai-Tijimi model, is used
and the expressions for e(t) and cog(t) are 27

0.3915t + 1.45

where So is the constant power spectrum of the bedrock

white excitation, COg is the predominant ground frequency, and g's is the ground damping.
It is well known that the Kanai-Tajimi model cannot
properly represent nonstationary characters of seismic
ground excitations. A number of elaborated filtered
white noise and evolutionary models for earthquake
ground acceleration have, therefore, been proposed 23-26. Modified Kanai-Tajimi models for the El
Centro 1940 and the Mexico City 1985 27e~arthquakeshave
been developed by Fan and Ahmadi which can be
conveniently used for random response analysis of structures. Furthermore, they showed that the new model
preserves the nonstationary evolutions of amplitude and
frequency content of the original records. Accordingly,
the ground excitation is given by

220

( w ( t ) w ( t + 7"))= 2rSo6(r)

if 42.9 < t _ 54.52 s

(18)

and
coe(t) = r{1.1 + 0 . 1 2 s i n [ 0 . 1 7 ( t - 2.0)] }

(19)

w e r e s u g g e s t e d 27. For both earthquakes the intensity of

white noise is So = 1.0 c m 2 s-3

Explicit equations governing the base raft displacement ub for various base isolation systems are described
below. For nonlinear base isolation systems, the
equivalent linearized forms are derived and the corresponding moment equations for evaluating secondorder statistics are also presented.
Moment equations

The derivation of the moment equations for the special

case of the R-FBI system is described in this section.
The mechanical behaviour of the R-FBI system is shown
schematically in Figure l(b). For the shear beam structure with the R-FBI system, the equation of motion for

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

with

the base raft is given by

al, + 2f,,o~,,ab + O~2oUb+ Ixt g sgn(ub)

ql

= - a~, -

'-0"
0

ql

(20)

E B"q"
tl= I

where ~0,, is the natural frequency of the base isolator, ~'0

is the damping coefficient, ~ is the friction coefficient
of the Teflon plates, g is the acceleration of gravity and
B, is a constant defined by equation (11). Equation (20)
describes the motion of the base raft during the sliding
phase, the nonsliding phases which are of negligible
duration for moderate to high intensity earthquakes are
neglected tS. A natural period To between 3 to 4.5 s and
0.03 </x~ _< 0.05 for the R-FBI systems have been
recommended -~-7. Here, To = 4 s, #~ = 0.04 and
~',, = 0.1 are used in the subsequent analysis. Note also
that when ~0o is small (very soft spring), equation (20)
reduces to that of a sliding joint (pure-friction)
isolator jS"tg. W h e n / ~ = 0, equation (20) becomes identical to the governing equation of a LRB base isolation
system.
Equation (20) may be replaced by its equivalent linear
system given a s 19

a5 + C,,ub + WoUb = --ag -- ~ .

ftb
u:,

B,,q.

(21)

n=|

where C,. is the equivalent damping coefficient.

Minimizing the mean-square error with respect to Ce,
using Gaussian statistics and assuming that (Ub) = O,
it follows that

W= w(t)

E21 E22 E23 E24

0
0
0
1
E =

(24)

at

0
0

0
0
(25)

E41 E42 E43 E44 E45 E46

0
0
0
0
0
1
0

E65 E66

where

E,i-

N
2

Ub

X=

E24

-*o]
E22
-2~',~o,
~o2cq
--,
. . . .
, E,3---1 - oqB
1 - t~jBl "
1 - oqBi
Ce ol I

1 -

at Bi

E41 = - B i E , _ l , E42 = - B I E 2 2 , E43 = - E 2 3 / O t l ,

E.,4 = - Ez4 /ot I
E45 = e(t)oG(t),"

E46 = 2 e ( t ) ~ e % ( t ) ,

E65 = -~%(t),2

E66 = - 2 ~ % ( t )
C,,(t) = 2~.,~o,, +/z~g

(26)

(22)
The equations governing the second-order moments
may be expressed in matrix form a s 33

Clearly the equivalent damping C,(0 is inversely proportional to (a~(t)) m. In spite of the fact that the
response of a nonlinear system to a Gaussian excitation
is, in general, a non-Gaussian process, the quasiGaussian linearization method is known to lead to results
with acceptable accuracy 32"33.
Equations (6), (13) and (21) must be solved
simultaneously to determine the modal amplitudes and
the base displacement. It is known that the first mode is
the dominant vibration mode for low- to medium-rise
fixed-base structures. References 34 and 35 show that
the first mode contains more than 85% to 90% of the
vibration energy of the structure. In the case of baseisolated structures, therefore, it may be sufficient to consider only the first mode of vibration of the superstructure.
Considering only the first mode of vibration (N = 1)
and substituting a s in (6) and (21) with the expression
given by (13), equations (6), (14) and (21) may now be
restated in the state variable form as
X = EX + W

(23)

Q = E Q + (EQ) r + G

(27)

where Q is the covariance matrix with Qo = ( x # / ) , and

G is a square matrix all of whose elements are zero
except G66 which is given by
G66 = 27rS,,

(28)

In the derivation of equations (27) and (28), use is made

of the fact that (q~) = ( q l ) = (uh) = ( a b ) = (U/)
= ( a t ) = 0. It should be pointed out that, for most
base isolation systems, residual slips or plastic drifts
exist. Nevertheless, since the excitation is a zero-mean
process, the expected values of those residual
displacements are equal to zero.
The explicit set of moment equations corresponding to
equation (27) have been described elsewhere 36. This
deterministic system of equations may be solved by a
variety of numerical techniques. The dependence of the
equivalent damping C,(t) on (u 2) = Q,~ as given by
equation (22) can be conveniently accommodated inth e
numerical solution scheme. The results will provide

Eng. Struct. 1992, Vol. !4, No 4

221

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

various second-order moment responses. The meansquare absolute acceleration transmitted to the roof of
the structure can also be easily evaluated. Using equations (4) and (6) and assuming ~ O j -- 1.0. it follows
that::
(//~) = ((// + tJh + / / e ) 2 )

= ~,(o:4(q~)

(29)
When equation (27) is solved, all the terms on the righthand side of (29) are known functions of time, and the
mean-square acceleration can be evaluated.
For other base isolation systems the second moment
equations can be formulated similarly to those of the RFBI system. Details of the derivations for the LRB, the
EDF and the SR-F base isolators are omitted here due
to lack of space, but may be found elsewhere 36. A
similar procedure is described in Reference 22.

Analysis technique
A fourth-order R u n g e - K u t t a - G i l l scheme is used for
numerical integration of various systems of governing
equations as well as the moment equations given by
(27). For structures with different natural periods, values

of ~',, =0.02 tor the modal damping coefficients,

R = 0.75 for the mass ratio and a =0.1 for the
nonuniformity coefficient are considered and the typical
values of parameters for different base isolation systems
as listed in Table 2 are used. As noted before, for the
SR-F base isolation system two sets of parameters are
used. SR-F1 designates a sliding resilient-friction isolation system with a natural period of 1.0 s which is
similar to that of the EDF system. SR-F2 identifies a SRF system with a natural period of 4.0 s which is essentially a R-FBI system with a sliding upper plate. The
recently developed stochastic models for the N00W
component of the E1 Centro 1940 and the N90W component of the Mexico City 1985 earthquakes which
preserve the time-dependent nature of the amplitude and
frequenc.y content of the ground accelerations are used
as seismic excitations in the analysis.
To assess the accuracy of the results, several MonteCarlo digital simulations are also performed. Segments
of Gaussian white noise processes are generated for ~(t)
and time histories of structural responses are calculated.
The first five modes of vibration of the shear beam structure are used in these simulation analyses. Statistics of
interest are estimated from ensembles of 500 sample
time histories. The simulated RMS responses are compared with those obtained from the linearization method.

E! Centro 1940 earthquake

Table 2

Expressions for the amplitude envelope function and the

time-dependent predominant ground frequency for the
El Centro 1940 earthquake are given by equations (16)
and (17). In this section, time-dependent mean-square
deflection, relative velocity and absolute acceleration
responses of the roof of the base-isolated shear beam
structure, as well as the mean-square relative base raft
displacements are evaluated. The peak responses of the
structure are statistically estimated and the results are
compared with those obtained for the actual El Centro
1940 earthquake record.

Values of parameters used for various base isolators

Isolator

Natural
period
To (s)

Damping
coeff,
~'o

Friction
coeff,
p.~

Friction
coeff.
/.t

R-FBI
LRB
EDF
S-RF1
S-RF2

4.0
2.0
1.0
1.0
4.0

O. 1 0
O. 1 0
0.10
0.10
O. 1 0

0.04
0.04
0.04

0.2
0.2
0.2

Cln

F-B

tO

4.-

SR - F2
S R - FI

EOF

<[

R-FBI

-0.5

I0
Time (s)

Figure 2

222

LRB

LRB ..~.~....

0.5
0.0

Sample time

15

R-FBI

0
v

-I

20

histories; and Fourier decompositions of absolute acceleration responses for

Eng. Struct. 1992, Vol. 14, No 4

I0

Frequency ( H z )
El C e n t r o 1 9 4 0

earthquake

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

3.0

40
,

F-B

.-.

~2.0-

i
/

t-

(~ 1.0-

./

.SR-F,

.-,0.6

I0

'

/ EpF.SR_FI

\.

.
i

15

20

~,

Time (s)

F-B

,~ 6

I0

15

ZO

Time (s)

0
4--

-=0.4

LRB

fJ

, ~, .
""

0
n

toO.2

0
C
Figure 3

I0

15

20

Time ( s )

I0

15

20

Time (s)

RMS responses for El Centro 1 9 4 0 earthquake

Sample responses
Responses of the shear beam structure with various base
isolators to digitally simulated nonstationary filtered
white noise model of the E1 Centro 1940 earthquake are
analysed. Sample absolute acceleration time histories at
the top of the structure with a natural period of 0.4 s are
shown in Figure 2(a). It is observed that base isolation
systems reduce the acceleration transmitted to the
superstructure to a significant extent. For the LRB
system, the acceleration response is a smoothly varying
time function. The acceleration responses for the frictional systems, however, contain a number of sharp
peaks.
Fourier decompositions of sample accelerations time
histories are shown in Figure 2(b). The frequency content of the acceleration responses can be clearly observed
in Figure 2. The acceleration transmitted to the fixedbase structure contains a sharp peak at a frequency of
about 2.5 Hz corresponding to the fundamental frequency of the structure. The acceleration Fourier spectra
for the R-FBI, the SR-F1 and the SR-F2 base isolation
systems are broad banded with no distinguishable major
peaks and contain certain high frequency components.
These high frequency components, which are generated
by the stick-slip shock loadings, contain most of their

energy in the frequency range of 0.5 to 5 Hz. The

Fourier spectra for the LRB and the EDF systems peak
at the natural frequencies of these isolators. Figure 3(b)
further shows that the Fourier spectrum for the LRB
system does not have significant high frequency components.

RMS Responses
For different base isolation systems, the nonstationary
RMS responses of the structure as obtained from equation (27) are evaluated. The RMS deflection, relative
velocity and absolute acceleration at the top of the shear
beam, as well as the RMS relative base raft displacement
responses are shown in Figure 3 by solid lines. A structure with a natural period of TI = 0.4 s was used in
these analyses. The Monte-Carlo digital simulation
results are shown by dashed lines in this figure for comparisons. As noted before, in the Monte-Carlo simulations, the first five modes of vibration of the shear beam
structure are used, while in the second-order moment
analysis, only the first mode of vibration is considered.
Nevertheless, the general agreement between the
predicted RMS responses and those obtained by the
digital simulation is reasonably good even for the
friction-type base isolation systems. The predicted RMS

Eng. Struct. 1992, Vol. 14, No 4

223

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

Response spectra

velocity and base raft displacement responses for the

SR-F1 system are somewhat lower than those obtained
by the Monte-Carlo simulations. It is conjectured that
these discrepancies are due to the loss of non-Gaussian
statistics of the response in the present Gaussian
linearization scheme.
Figure 3 shows that the use of base isolation systems
reduces the RMS acceleration, deflection and velocity
responses of the structure by a factor of 4 to 15 when
compared with those of the fixed-base structure. For
example, the peak RMS acceleration for the fixed-base
structure reaches to 0.75g, while those of base-isolated
structure are in the range of 0.06g to 0.2g. The reduction in the peak structural responses is at the expense of
the generation of relative base displacement. Figure 3
shows that the peak RMS displacement responses for the
R-FBI/SR-F2 and the LRB systems are about 8 cm,
while those for the EDF and the SR-FI systems are
about 2 to 3 cm. This figure also shows that the RMS
responses reach their peaks at about 3 to 5 s and then
decay rapidly to their stationary limits. This trend is in
agreement with the time variation of amplitude envelope
function e(t) as given by equation (16).

Three standard deviations (3a) has been a commonly

used estimate for a rough bound on the peak value of a
zero mean random process. The 3o-estimates are used
here to evaluate the statistical response spectra.
The statistical deflections, relative velocities, and
absolute accelerations, as well as base raft displacements of the base-isolated structures are evaluated
and the results are plotted against the fundamental natural period of the structure (T~ = 27r/w~) in
Figure 4, the response spectra for the fixed-base structure are also reproduced as a comparison. In Figure 4,
the solid lines correspond to the statistical estimates for
the peak responses and the dashed lines show the peak
responses obtained for the accelerogram record of the
N00W component of the El Centro 1940 earthquake. A
structural period larger than 1 s was not considered,
since the use of base isolation systems are only recommended for relatively compact structures (with less than
10 storeys). These low-rise structures are expected to
have natural frequencies greater than 1 Hz.
For various base isolators, Figure 4(a) shows the max-

i0 z

10 3
/

IO

~'~10z

~ I0

F-B

SR-FI

.~,"" - " " - " - . /

g
P

==

LRB

"0

~ ' " ' " "

i0-1

!B'""

R-FBI/SR-F2

i0 -z

I0-I

Natural period

IO

=E

~...-,""
!

i0 -I

(s)

\\~:". . . : L R B~ -.-"""
' . . . .....

R-FBI/SR- F2
i

i0 -I
i0 z

F-B

Noturol period (s)

EDF

LRB

c. om

" no

nj )
o
.,0

.-,,,-Ii :~.._ ---. ~

=E

.....

-.'- " -,_.' .-'-~-- .'-"

" ..........

s t'~.

......

I0"1

. . . .

.o

x
o

R FBI/SR F2

__

LRB

Natural period (s)

1
I

I
d

I0-'

Noturol period (s)

Figure 4 Variations of peak responses of structure with its natural period for El Centro 1940 earthquake

224

SR FI

- ...............

R- FBI/S,R-F2

i0 "1
C

................

....JI

Eng. Struct. 1992, Vol. 14, No 4

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

imum deflections at the roof of the shear beam structure

versus its fundamental natural period. As expected, the
peak deflection of the structure increases as its flexibility
increases. It is also observed that the base-isolated structure, generally, has much lower peak deflections than
those of the fixed-base structure. Minimum peak deflections are generated for the structure with the R-FBI and
the LRB base isolation systems. Figure 4(a) also shows
that the three standard deviation level appears to be a
reasonable estimate of peak deflections for various base
isolation systems. For the El Centro 1940 earthquake,
the 3o-estimates provide realistic upper bounds of the
peak deflections for all the systems discussed here.
The statistically estimated peak relative velocity
responses at the roof of the structure for various base
isolation systems are shown in Figure 4(b) by the solid
lines and compared with those for the E1 Centro 1940
earthquake accelerogram. It is observed that, although
certain deviations for the R-FBI and the SR-F1 systems
are observed, the statistically estimated response spectra
curves provide reasonable bounds on the peak velocity
responses of the structure. Figure 4 also shows that peak
velocities for the base-isolated structure are lower by
orders of magnitude when compared to those of the
fixed-base structure. The LRB system leads to the least
peak velocity responses among the base isolators considered.
The 3o-estimates for the peak absolute acceleration at
the roof of the structure versus TI for various base
isolation systems are shown in Figure 4(c) by the solid
lines. It shows that, although there are some deviations,
the statistical acceleration response spectra, (which are
also smooth functions of the natural period), are in
qualititative agreement with those obtained for the El
Centro 1940 earthquake accelerogram, It is also
observed that peak absolute accelerations for the baseisolated structure are lower by a factor of 4 to 10 when
compared with that of the fixed-base structure. Figure
4(c) also shows that the LRB system leads to the lowest
peak acceleration among the isolators considered. The
statistical results, however, suggest that the R-FBI/SRF2 system generates the smallest RMS acceleration. It
should also be emphasized that these differences are not
significant in practical applications.
One important design parameter for seismic isolation
systems is the maximum base displacement of the structure relative to its foundation. Figure 4(d) compares the
peak base displacements of the shear beam structure produced by different base isolators. The 3a-estimates for
the peak base raft displacements are shown in this figure
by the solid lines which are in reasonable agreement
with those obtained for the El Centro 1940 earthquake
accelerogram. It appears that the peak base displacements are almost constant for the entire range of the
natural period considered. Figure 4 also shows that all
base isolation systems lead to manageable peak base
displacements. The SR-F1 and the EDF systems
generate the lowest maximum base displacements, while
the LRB and the R-FBI systems produce the largest
displacements among the isolators considered.

Mexico City 1985 earthquake

It is believed that thc base isolation systems are
generally sensitive to long-period ground excitations.

However, the deterioration in the performance of

various base isolation systems for earthquake excitations
with considerable energy at low frequencies is not fully
understood. In particular, probabilistic response
analyses of base-isolated structures for long-period
ground excitations have not been attempted. Therefore,
it is of interest to perform such analyses for the random
model of the Mexico City 1985 earthquake. This earthquake contained prolonged sinusoidal behaviour at a frequency of about 0.5 Hz. The expressions for time
variations of the amplitude envelope function and
predominant ground frequency for the Mexico City
1985 earthquake are given by equations (18) and (19).
This stochastic model covers the duration of 28.54 to
83.06 s of the actual record. This is a rather long duration with its first 10 s containing relatively little energy.
In this study, the first 10 s are omitted and only the
strong motion part is considered. (Therefore, the time
t = 0 s here corresponds to the time t = 38.54 s in the
actual earthquake.) In addition to saving computer time,
another reason for omitting the initial low amplitude
part is the assumption of continuous sliding used in the
linearization models for the R-FBI and the SR-F base
isolation systems. As noted before, this assumption is
reasonable only when the intensity of ground excitation
is not too low.
In this section, the stochastic model for the Mexico
City 1985 earthquake is used as ground excitation. The
time-dependent mean-square deflection, relative
velocity and absolute acceleration responses at the roof
of the base-isolated shear beam structure, as well as the
mean-square relative base raft displacements are
evaluated and the results are plotted in several figures.
The peak responses of the structure are statistically
estimated and are compared with those obtained for the
actual Mexico City 1985 earthquake record.

Sample responses
Using the nonstationary Kanai-Tajimi model of the
Mexico City 1985 earthquake, a number of accelerograms are digitally simulated. Responses of the shear
beam structure with various base isolators to these
simulated ground excitations are analysed. Sample
absolute acceleration time histories at the roof of the
structure with tl = 0.4 s are shown in Figure 5(a). It is
observed that, with the exception of the R-FBI and the
SR-F2 systems, the isolators amplify the transmitted
acceleration. The amplification for the LRB system is
about 100% when compared to that of the fixed-base
structure. Figure 5 also shows that the acceleration
responses are smoothly varying time functions. This is
due to the almost sinusoidal behaviour of the simulated
Mexico City 1985 earthquake.
Fourier decompositions of the acceleration time
histories are shown in Figure 5(b). It is observed that the
acceleration transmitted to the structure generally contains a sharp peak at a frequency of about 0.5 Hz. The
peak has a particularly high amplitude for the LRB
system. The Fourier spectra for the R-FBI and the SRF2 base isolation systems contain energy at high
frequencies but with no dominant peaks.
The results presented in Figure 5 show that the base
isolation systems considered do not function properly
when subjected to the simulated Mexico City 1985
earthquake. The R-FBI and the SR-F2 systems are the

Eng. Struct. 1992, Vol. 14, No 4

225

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

LRB
m

-~

Ilfh

SR-F2

iL__

FB

SR - F2

EDF

0.5

hO

0.0
-0.5~

I0

20

30

R- FBI

40

Time (s)

I0

Frequency(Hz)

Sample time histories and Fourier decompositions of absolute acceleration responses for Mexico City 1985 earthquake

Figure 5

only systems which appear to provide small amounts of

protection.
For two simulated accelerograms, sample responses of
the relative base raft displacements for the EDF and the
SR-F1 are shown in Figure 6. Solid lines in this figure
are the total base displacement, Ub, and dashed lines
represent the deflection of the elastometric part of the
isolator, y. It is observed that at certain times large slips
occur at the upper friction plates for both base isolation
systems. The structure then appears to vibrate around
the new equilibrium locations due to the presence of the
rubber element. The upper samples show that the
residual slip displacements remaining at the end of the
earthquake could become quite large. It should be
emphasized that all frictional base isolation systems will
experience residual relative base displacements after a
seismic event. Therefore, a provision to recentre the
structure may be required. The laminated rubber bearing
(LRB) is perhaps the only base isolation system that will
recentre itself at the end of the ground motion.

RMS responses
The nonstationary RMS responses of various baseisolated structures are evaluated. It is assumed that the
fixed-base shear beam structure has a natural period of
T~ = 0.4 s. The corresponding set of second-order
moment equations are solved. The resulting RMS
deflection, relative velocity and absolute acceleration at
the top of the shear beam, as well as the RMS relative
base raft displacement responses are shown in Figure 7
by solid lines. The RMS responses as obtained by the
Monte-Carlo digital simulations are plotted in this figure
by dashed lines as a comparison. Although some deviations are observed, the general agreement is reasonable.
Certain discrepancies are noticed for the SR-F1 and the
EDF systems. The predicted RMS responses are
somewhat higher than the simulation ones for the deflection and absolute acceleration but are lower for the
velocity. As noted before, it is conjectured that these
discrepancies are due to the assumption of quasi-

"E

,,

,.::v';,
,,

.' i,,,..;i~,j,~,'.
j
v'/'...
",.':".,
'."

~ ..... :

~-~

q)

='-"~"'"'"

................

E
@

~0

glO

_8 no
t/)
.B

r~

0
-I0

-JO

8
Figure 6

226

I0

20
30
Time (s)

40

I0

Sample time histories of base displacement responses for Mexico City 1985 earthquake

Eng. Struct. 1992, Vol. 14, No 4

20

Time (s)

30

40

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

1.0

~ ' , . .~~ ~ ~ ,

~ 3t

~EDF

"~

"0

F-B

0.5

0.0

i0

.g

20
30
Time (s)

40

I0

20
30
TimeJs)

LRB . ~ ~ ; I ~ ,

/ ==
=

40

R-FBI/SR-F2

I io"

~ 0.1

a:

.
0

0
I0

~
20
30
Time (s)

0
40

I0

20
30
Time (s)

40

Figure 7 RMS responses for Mexico City 1985 earthquake

Gaussian statistics of the response used in the linearization scheme.

From Figure 7 it is observed that the LRB system
significantly amplifies the RMS response of the structure for this long period earthquake due to resonance.
The EDF and the SR-F1 base isolation systems also do
not reduce the RMS deflection, velocity and acceleration
responses of the structure. The R-FBI and the SR-F2
base isolators are the only ones which appear to reduce
the deflection, velocity and acceleration RMS response
to a certain extent.
Figure 7 shows that the peak RMS displacement
response for the LRB system reaches about 30 cm due
to the resonance between the earthquake excitation and
the base isolator. The corresponding RMS base
displacement for the R-FB1/SR-F2 systems is about
22 cm. The linearization methods predicts RMS base
displacements of about 3 to 4 cm for the EDF and the
SR-FI base isolation systems. The Monte-Carlo simulations, however, lead to a relatively large residual base
displacement of about 10 cm. The discrepancy is due to
the approximations involved in the Gaussian linearization method and in estimating the mean stick-slip
parameter 2-'36. It is observed from Figure 6, that large
base displacement occurs at t = 24 (about the time of
peak excitation) which essentially remains as a residual

base displacement. This peculiar behaviour is not properly accounted for in the present linearization
method. From Figure 7, it is also observed that the RMS
responses reach their peak values at about 20 to 22 s and
then decay rapidly and approach their stationary limits.

Response spectra
The 3a-estimates are used in this section for evaluating
peak structural responses. The statistically evalued
response spectra for various base-isolated structures are
shown in Figure 8 by the solid lines. The response spectra for the fixed-base structure are also reproduced in
this figure for comparison. The dashed lines in Figure
8 correspond to the peak responses obtained for the
accelerogram of the N90W component of Mexico City
1985 earthquake.
Figure 8 shows that as TI increases the peak structural deflections and velocities increase, while the peak
accelerations remain almost constant. The three standard
deviation estimates provide realistic upper bounds on the
peak responses for the base-isolated structure under the
Mexico City 1985 earthquake excitation. It is also
observed from ths figure that peak responses for the
base-isolated structure with the LRB, the EDF and the
SR-F1 systems are higher than those of the fixed-base

Eng. Struct~ 1992, Vol. 14, No 4 227

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

103

I0 z

,.-., I0

F-B

~"~ I0 z

E
U
t-

SR-FI

O
,D
,0-,,

}
"0
X

'"'"

R- FBI/SR-F2

'~o I0
x

"" R-FBI/SR-F2
10-2

Natural

i0 -I

IO'l

period (s)

0-1
Natural period (s)
i0 z

.9

EDF

SR-FI

LRB

R- Ffll/SR-F2

_o

I0,

LRB

'"

~" , ~ ' w ' . z ; o ~ t

0
)<

0
=E

R- FBI/SR-F2
I0 -I

O-I

Figure 8

Natural period (s)

o-i

Natural period (s)

Variations of peak responses of structure with its natural period for Mexico City 1985 earthquake

structure by a factor of 1.2 to 2.5. Thus, the base isolation systems appear not to function properly for this
long-period earthquake.
Figure 8(d) compares the peak base displacements of
various base-isolated structures. The statistically
estimated peak base raft displacements appear to be in
reasonable agreement with those obtained for the actual
Mexico City 1985 earthquake accelerogram. This figure
also shows that the peak displacements are almost constant for the entire range of the natural period considered.
It is also observed that the SR-F1 and the EDF systems
generate the lowest maximum base displacements
among the isolators considered while the LRB system
produces the largest one.
The presented results show that the base isolated
structures are quite sensitive to long period ground
excitations. Therefore, the use of base isolation systems
with the typical values of parameters as listed in Table
2 in regions which have the potential of generating
earthquakes with considerable energy at low frequencies
should be avoided. It is, of course, conceivable that new
base isolation systems may be developed, or the existing
ones may be redesigned in order to reduce the sensitivity
to long period ground excitations.

228

Eng. Struct. 1992, Vol. 14, No 4

Conclusions
Earthquake responses of a base-isolated shear beam
structure with different base isolation systems have been
probabilistically analysed. In contrast to the earlier
studies which used simple white or filtered white noise
ground motion models, here, the recently developed
models with evolving amplitude and frequency for the El
Centro 1940 and the Mexico City 1985 earthquakes are
used as seismic excitations. The method of timedependent equivalent linearization is utilized and the
mean-square response statistics of the base-isolated
shear beam structure are evaluated. Statistically
estimated peak responses of the base-isolated structure
are compared with the response spectra for actual earthquake accelerograms.
Based on the results presented, the following conclusions may be drawn
The nonstationary (quasi-Gaussian) equivalent
linearization technique combined with the secondorder moment equations provides a systematic and
computationally efficient tool for response analysis
of base-isolated structures to random earthquake

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

excitations. Comparisons with the performed digital

simulations show that, although certain deviations
for the frictional base isolation systems are
observed, the method is reasonably accurate.
The statistically generated response spectra for the
recent stochastic models of the El Centro 1940 and
the Mexico City 1985 earthquakes are in good agreement with the actual records.
For response analyses of base-isolated shear beam
structures subject to random earthquake excitations,
the assumption that the first vibration mode of the
structure is dominant is reasonable.
Base isolation systems are generally highly effective
in reducing the RMS responses of the superstructures for common earthquakes such as the El Centro
1940 earthquake.
Base isolation systems are sensitive to long period
ground excitations. For the Mexico City 1985 earthquake, which contains considerable energy in the
low frequency range, the base isolators with typical
design parameters as listed in Table 2 do not function properly.
The acceleration response time histories of a structure with a friction type base isolator are broad band
processes and contain high frequency components.
The acceleration responses for the LRB and the EDF
systems are narrow band processes and their Fourier
spectra contain sharp peaks at the natural frequencies of the isolators.
For the El Centro 1940 earthquake, the EDF system
provides the highest RMS acceleration responses
and the R-FBI and SR-F2 and the LRB systems lead
to the lowest ones among the base isolation systems
considered. Furthermore, the SR-F1 and the LRB
systems generate the lowest and the highest RMS
base displacement responses, respectively.

The results show that the base isolation systems are

highly effective in protecting the structure from earthquake damage. However, these devices are quite sensitive to long-period seismic excitations and tlieir use in
regions which have the potential to generate earthquakes
with considerable energy at low frequencies should be
avoided.
It should also be emphasized that the presented results
were for specific values of parameters for various base
isolation systems. While the conclusions are not sensitive to small variations in material properties, should
the parameters of the isolators be different from the
assumed values, the performance results could alter
significantly. In particular, large variations in the value
of the friction coefficient, strongly affect the performance of frictional base isolation systems.

Acknowledgements
The authors wish to thank Mr. Fa-Gung Fan for many
helpful discussions. This work has been partially supported by the National Center for Earthquake Engineering R e s e a r c h , S t a t e U n i v e r s i t y o f N e w Y o r k at B u f f a l o
under grant NCEER 882012A.

References
Kelly. J. M. 'Aseismic base isolation', Shock Vib. Dig., 1982, 14,
17-25

2 Kelly, J. M. 'Aseismic base isolation: review and bibliography', Soil

Dyn. Earthquake Engng., 1986, 5, 202-216
3 Kelly, J. M. and Hodder, S. B. 'Experimental study of lead and
elastometric dampers for base isolation systems in laminated
neopreme bearings', Bull. New Zealand Nat. Soc. for Earthquake
Enging., 1982, 15, 5 3 - 6 7
4 Huffmann, G. R. 'Full base isolation for earthquake protection by
helical springs and viscodampers', Nuclear Engng. Design, 1985, 84,
331-338
5 Mostaghel, N. 'Resilent-friction base isolator', Pep. UTEC 84-097,
The University of Utah, 1984
6 Mostaghel, N.. Hejazi, M. and Khodaverdian, M. 'Response of
structures supported on resilient-friction base isolator', Proc. Third
U.S. Nat. Conf. on Earthquake Engineering, Charlestown, South
Carolina. August 1986, pp. 1993-2003
7 Mostaghel. N. and Khodaverdian, M. 'Dynamics of resilient-friction
base isolator (R-FBI)'. Earthquake Engng. Struct. Dyn., 1987, 15,
379 - 390
8 lkonomou. A. S. 'Alexisismon seismic isolation levels for translational and rotational seismic input', Proc. 8WCEE, San Francisco,
2 1 - 2 8 July 1984, V, 975-982
9 Gueraud. R., NoeI-Leroux, J.-P., Livolant, M. and Michalopoulos,
A. P. "Seismic isolation using sliding-elastometer bearing pads',
Nuclear Engng. Design. 1985, 84, 363-377
10 Su. L.. Ahmadi. G. and Tadjbakhsh. I. G. 'Comparative study of
base isolation systems', J. Engng. Mech. Div. ASCE, 1989, 115,
1976-1992
I I Su. L.. Ahmadi. G. and Tadjbakhsh, I. G. 'Performance of a sliding
resilient-friction base isolation system for a shear beam structure',
J. Struct. Engng, ASCE, 1991. 117, 165-181
12 Su. L.. Ahmadi. G. and Tadjbakhsh. I. G. 'A comparative study of
performances of various base isolation systems - part 1: shear beam
structures'. Earthquake Engng. Struct. Dyn. 1989, 18, 11-32
13 Su, L.. Ahmadi, G. and Tadjbakhsh. I. G. 'A comparative study of
performances of various base isolation systems - part II: sensitivity
analysis'. Earthquake Engng. Struct. Dvn. 1990, 19, 2 1 - 3 3
14 Fan. F.-G.. Ahmadi, G. Mostaghel, N. and Tadjbakhsh, I. G. 'Performance analysis of aseismic base isolation systems for a multistorey building'. Soil Dyn. Earthquake Engng. 1991, 10, 152-171
15 Crandall, S. H., Lee, S. S. and Williams, J. H. "Accumulated slip of
a friction-controlled mass excited by earthquake motions', J. Appl.
Mech. Trans. ASME 1974, 41, 1094-1098
16 Ahmadi. G. 'Stochastic earthquake response of structures on sliding
foundation', Int. J. Engng. Sci. 1983. 121, 93-102
17 Constantinou, M. C. and Tadjbakhsh. I. G. 'Response of a sliding
structures to filtered random excitation', J. Struct Mech. 1984,
12, 401-418
18 Su, L., Orabi, I. I. and Ahmadi. G. 'Nonstationary earthquake
response of a sliding rigid structure'. Int. J. Engng. Sci. 1988, 26.
(9), 1013-1026
19 Sa. L., and Ahmadi, G. "Response of frictional base isolation systems
to horizontal-vertical random earthquake excitations', Probabilistic
Engng. Mech. 1988, 3, (1), 12-21
20 Lin, B. C., Tadjbakhsh, I. G. Papageorgiou. A. and Ahmadi, G.
'Response of base-isolated buildings to random excitations with a
Clough-Penzien model', Earthquake Engng. Struct. Dvn. 1989, 18,
49 - 62
21 Su, L., Ahmadi. G. and Tadjbakhsh, 1. G. 'A prohahilistic comparative study of various base isolation systems', Mech. Struct.
Machines, 1990. 18, 107-133
22 Su. L., Ahmadi, G. and Tadjbakhsh, I. G. 'Responses of baseisolated shear beam structures to random excitations', Probabilisdc
Engng. Mech. 1990. 5. 3 5 - 4 6
23 Amin, M. and Ang. A. H. S. "Nonstationary stochasitc model of earthquake motions', J. Engng. Mech. Div. ASCE, 1968, 94, (EM2),
559 -583
24 Cakmak. A. S., SheriL R. I. and Ellis, G. W. 'Modelling earthquake
ground motions in California using parametric time series methods.
Soil D3'n. Earthquake Engng. 1985, 4, (3), 124-131
25 Ellis, G. W. and Cakmak, A. S. 'Modelling earthquake ground
motions in seismically active regions using parametric time series
methods, technical report NCEER-87-0014, National Center for
Earthquake Engineering Research, SUNY at Buffalo, August 1987
26 Lin, Y. K. and Yong, Y. 'Evolutionary Kanai-Tajimi earthquake
models', J. Engng. Mech. Div. ASCE, 1987, 113, (8), 1119-1137
27 Fan, F.-G. and Ahmadi, G. 'Nonstationary Kanai-Tajimi models for
El Centro 1940 and Mexico City 1985 earthquakes'. Probabilistie
Engng. Mech. 1990, 15, 171-181
28 Housner, G. W. 'Properties of strong motion earthquakes', Bull.
Seismol. Soc. Amer. 1955. 45, 197-218

Eng. Struct. 1992, Vol. 14, No 4

229

Probabilistic responses of base-isolated structures: Lin Su and G. Ahmadi

29

Bycrofi, G. N. "White noise representation of earthquakes', J.

Engng. Mech. Div. ASCE, 1960, 86, (EM2), 1 - 1 6

30 K-anal, K. "Semi-emplricaJ formula for the seismic characteristics of
the ground', Univ. Tokyo Bull. Earthquake Res. Int. 1957, 35,
309-325
31 Tajimi, H. "A statistical methd of determining the maximum response
of a building structure during an earthquake', Proc. 2nd Worm Conf.
Earthquake Engng. Tokyo and Kyoto, 1960, Vol II, pp 781-798
32 Wen, W. K. 'Equivalent linearization for hysteretic systems under
random excitation', J. Appl. Mech. 1980, 47, 150-154

230

Eng. Struct. 1992, Vol. 14, No 4

33 Lin, Y. K. Probabilistic theory of structural dynamics, McGraw Hill,

1967
34 Wirching, P. H. and Yao, J. T. P. 'Modal response of structures',
J. Struct. Div., ASCE 1970, 96, 879-883
35 Su, L. and Ahmadi, G. 'Earthquake response of linear continuous
structures by the method of evolutionary spectra', Engng. Struct.
1988, 10, (1), 4 7 - 5 6
36 Su, L. and Ahmadi, G. 'Probabilistic responses of base-isolated
structures to El Centro 1940 and Mexico City 1985 earthquakes',
Report No. M1E-191, Clarkson University, August 1989