Engineering Structures 29 (2007) 16531663

www.elsevier.com/locate/engstruct
Comprehensive evaluation of equivalent linear analysis method for
seismic-isolated structures represented by sdof systems
Murat Dicleli
a,
, Srikanth Buddaram
b
a
Department of Engineering Sciences, Middle East Technical University, 06531 Ankara, Turkey
b
Department of Civil Engineering and Construction, Bradley University, Peoria, IL 61625, USA
Received 27 December 2005; received in revised form 5 September 2006; accepted 12 September 2006
Available online 27 October 2006
Abstract
In this study, the equivalent linear (EL) analysis procedure used in the design of seismic-isolated structures is evaluated. The effect of several
parameters such as the intensity and frequency characteristics of the ground motion, isolator properties and the structure mass are considered in
the evaluation. First, the EL analysis procedure is evaluated using harmonic ground motions with various excitation periods and a suite of 15
seismic ground motions with various frequency characteristics. Then, the effect of the structure mass on the effective damping ratio and hence on
the accuracy of the EL analysis results is investigated. This is followed by regression analyses of the acquired data to assess the effect of various
parameters on the accuracy of the EL analysis results.
It is found that the accuracy of the EL analysis results is affected by the peak ground acceleration to peak ground velocity ratio of the ground
motion as well as the intensity of the ground motion relative to the characteristic strength of the isolator. It is also demonstrated that the effective
damping equation currently used in the design of seismic-isolated structures must incorporate the effective period of the structure and frequency
characteristics of the ground motion for a more accurate estimation of seismic response quantities. A new effective damping equation that includes
such parameters is formulated and found to improve the accuracy of the EL analysis results.
c 2006 Elsevier Ltd. All rights reserved.
Keywords: Seismic isolation; Nonlinear response; Equivalent linear analysis; Effective period; Effective damping
1. Introduction
For the design of seismic-isolated structures in the USA,
design specications such as AASHTO Guide specications
for seismic isolation design of bridges [1] and the international
building code [2] recommend various analysis methods. These
analysis methods are; equivalent lateral response procedure for
buildings, uniform load method for bridges, response spectrum
and nonlinear time history analysis methods. The rst three
methods are commonly used for the analysis of seismic-
isolated structures. These methods are based on theories of
linear elastic analysis. Since the behavior of seismic isolators
is non-linear in nature, equivalent linear elastic properties
need to be dened for the elastic analysis of seismic-isolated
structures. The equivalent linear elastic properties of the

Corresponding author. Tel.: +90 312 210 4451; fax: +90 312 210 4469.
E-mail address: mdicleli@metu.edu.tr (M. Dicleli).
isolators are expressed in terms of an effective stiffness and
an effective damping ratio [3] to account for the hysteretic
energy dissipation of the isolators. The formulations of these
equivalent linear elastic properties have been discussed by
Hwang and Sheng [4] and Hwang et al. [5] for isolators with
bilinear forcedeformation relationship. Using these equivalent
linear elastic properties, an equivalent linear (EL) analysis
procedure is followed to estimate the absolute maximum
seismically induced displacements and forces in the isolator.
The results obtained from EL analysis are approximations
of the actual maximum nonlinear response of the isolator.
Therefore, a comprehensive verication of the seismic response
quantities obtained from EL analysis is required to ensure
that the seismic-isolated structure performs as intended in the
design.
The accuracy of the results obtained from EL analysis
procedure for bridges has been studied by several researchers
[69]. These research studies were very useful in identifying
0141-0296/$ - see front matter c 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2006.09.013
1654 M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663
the relative accuracy level of the results obtained from EL
analysis procedures. However, most of them were concentrated
on specic design code procedures that involve the use of
smooth code design spectra and effective damping coefcients
to account for the hysteretic energy dissipation of the isolator.
Obviously, this type of an approach does not allow for an
absolute verication of the EL analysis results potentially
due to the approximations involved in the code procedure for
calculating the effective damping coefcient and simulated
ground motions used to match the smooth code design
spectrum. Furthermore, most of these studies have not explicitly
considered the effect of the isolator properties as well as the
frequency and energy content of the ground motion on the
accuracy of the EL analysis results. Therefore, a comprehensive
and code-independent evaluation of the EL analysis results for
seismic-isolated structures is required.
2. Research objective, scope and outline
The main objective of the presented research is to perform a
comprehensive and code-independent evaluation of the seismic
response quantities obtained from EL analysis procedure for
seismic-isolated structures and to propose improvements to
the procedure for more accurate prediction of such response
quantities.
The evaluation of the EL analysis results is performed
for seismic-isolated structures represented by isolators placed
on rigid supports (foundations) and supporting a rigid mass.
This allowed for a direct assessment of the uncertainties in
the EL analysis results, which are mainly associated with
the approximations related to the simulation of the actual
nonlinear behavior of the isolators. The evaluation of the
EL analysis results mainly involves the comparison of the
seismic response quantities obtained from EL analyses with
those obtained from nonlinear time history (NLTH) analyses.
The effect of several parameters such as isolator properties,
duration, intensity and frequency characteristics of the ground
motion are also considered in the evaluation of the EL analysis
results.
First, the EL analysis results are evaluated using a series
of harmonic ground motions with various excitation periods.
This allowed for the direct assessment of the effect of the
ground excitation period on the accuracy of the EL analysis
results. Then, the effect of seismic-isolated structure mass on
the effective damping ratio and hence on the accuracy of the
EL analysis results is investigated. Next, the EL analysis results
are evaluated using a suite of 15 seismic ground motions
with various frequency characteristics. This is followed by
regression analyses of the acquired data to assess the effect of
various parameters on the accuracy of the EL analysis results.
The regression analyses results are then used to incorporate
additional empirical relationships in the effective damping
equation to improve the accuracy of the EL analysis results.
At the end, the accuracy of the EL analysis results using the
improved damping equation is assessed and conclusions are
outlined.
Fig. 1. (a) Idealized hysteresis loop of a typical isolator. (b) Hysteresis loop of
a viscous damper.
3. Properties of typical seismic isolators
Seismic isolators may be classied into two groups as
rubber-based and sliding-based. Some examples of rubber-
based isolators are; low damping natural rubber bearings, high
damping natural rubber bearings and leadrubber bearings [10].
Some examples of friction-based isolators are Eradiquake [11]
and friction pendulum bearings [12].
The forcedisplacement hysteresis relationship of most
isolators is idealized as bilinear for design purposes as
illustrated in Fig. 1(a). In the gure, F
yi
is the characteristic
strength, k
i
is the elastic stiffness, k
p
is the post-elastic stiffness,
F
y
and U
y
are respectively the yield force and displacement and
F
d
and U
d
are respectively the design (or maximum) force and
displacement of the isolator. In the idealized hysteresis model,
the elastic stiffness, k
i
of the friction-based isolators is very
large; hence, U
y
is nearly equal to 0.
As mentioned earlier, methods based on linear elastic
theories are commonly used for the analysis of seismic-isolated
structures [1,2]. Such methods require the equivalent linear
elastic properties of the isolator, which are the effective stiffness
and effective damping ratio. The effective stiffness, k
e
, of the
isolator is graphically illustrated in Fig. 1(a). It is dened as
the slope of the dashed line connected from the origin to the
point of design (or maximum) displacement and force on the
hysteresis curve. Thus, k
e
is obtained by dividing the design
force, F
d
, by the corresponding design displacement, U
d
of the
isolator;
k
e
=
F
d
U
d
=
F
yi
U
d
+ k
p
. (1)
Neglecting the inertial effect of the structure mass supported
on the isolator, the effective damping ratio,
e
, which produces
the same amount of viscous energy dissipation as the hysteretic
energy dissipated at each cyclic motion of the isolator, is
M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663 1655
obtained by setting up the area under the hysteresis curve of
Fig. 1(a) to the area under the hysteresis loop of Fig. 1(b)
representing the energy dissipated due to viscous damping [10,
13]. Thus;

e
=
4F
yi
(U
d
U
y
)
2k
e
U
2
d
. (2)
The derivation of the above equation makes the assumption
that the excitation frequency is identical to the natural
frequency of the EL system. This was rst recognized by
Jacobsen and Ayre [3] in the late fties. A response spectrum
obtained for the effective damping ratio,
e
, is then used in the
linear analyses to account for the hysteretic energy dissipated
by the isolator.
4. EL analysis procedure
The EL analysis procedure commonly followed for the
design of seismic-isolated structures where the isolators are
placed on rigid supports (or rigid foundations) is outlined
below.
1. Assume an isolator design displacement, U
d
.
2. Substitute the assumed design displacement, U
d
, and
properties of the isolator in Eq. (1) to calculate the effective
stiffness, k
e
of the isolator.
3. Substitute the assumed design displacement, U
d
, in Eq. (2)
to calculate the effective damping ratio,
e
.
4. Calculate the effective period, T
e
, of the seismic-isolated
structure. For a rigid mass, m, supported on an isolator, T
e
is
expressed as;
T
e
= 2
_
m
k
e
. (3)
5. Obtain a new design displacement from the displacement
response spectrum S
D
(
e
, T
e
) dened as a function of the
effective damping ratio and effective period.
6. Compare the new design displacement with the initially
assumed design displacement. If the difference between
the displacements is smaller than a preset tolerance level,
then stop the iteration. Otherwise, go to step 2 to continue
with the next round of iterations using the new design
displacement.
5. Ground motions used in the analyses
Ground motions are characterized by their peak ground
acceleration (A
p
) to peak ground velocity (V
p
) ratios, which
represent their dominant frequency and energy content. Ground
motions with intense long-duration acceleration pulses have
low A
p
/V
p
ratios whereas those with high frequency, short-
duration acceleration pulses have high A
p
/V
p
ratios. Earlier
research studies have revealed that the response of a structural
system to a ground motion may differ as a function of the
A
p
/V
p
ratio of the ground motion [14]. Consequently, the
verication of the EL analysis procedure needs to be performed
using ground motions with various A
p
/V
p
ratios to cover a
broad range of possibilities concerning the seismic response of
seismic-isolated structures.
Two sets of ground motions are used for the verication of
the linear analysis procedure. The rst set of ground motions is
a harmonic ground motion dened as follows;

U
g
= A
p
Sin
_
2
T
g
t
_
(4)
where

U
g
is the ground acceleration, T
g
is the excitation period
and t is time in seconds. For this harmonic ground motion, the
A
p
/V
p
ratio is expressed as;
A
p
V
p
=
2
T
g
. (5)
The main reason for using harmonic ground motions for
the verication of the EL analysis procedure is to have a clear
understanding of the effect of the frequency characteristics (or
the A
p
/V
p
ratio) of the ground motion on the accuracy of the
EL analysis results. Additionally, the availability of accurate
closed form solutions for the elastic response of single degree
of freedom systems subjected to such harmonic ground motions
facilitates the EL analysis procedure. Harmonic ground motions
with A
p
/V
p
ratios ranging between 5.23 s
1
and 20.0 s
1
are
used in the analyses. For the range of A
p
/V
p
ratios considered,
the excitation periods vary between 0.314 and 1.2 s.
The second set of ground motions involves a suite of 15
earthquakes with A
p
/V
p
ratios ranging between 5.50 s
1
and
21.5 s
1
. The details of the ground motions are presented in
Table 1. In the table, the strong motion duration is taken as
the bracketed duration [15] which is the time between the rst
and last exceedances of some threshold acceleration, taken as
0.05g. These ground motions are used for further verication
of the EL analysis procedure and for devising improvements
to the procedure for a more accurate prediction of the seismic
response quantities.
Ground motions with high A
p
/V
p
ratio generally contain
high frequency acceleration pulses, which load and unload
the structure in very short time intervals. Thus, even if the
isolator reaches its yield level, isolator displacements beyond
this yield point cannot be sustained for an extended period
of time as the seismic inertial force applied on the structure
remains above the isolators yield level only for a short duration.
Consequently, seismic and harmonic ground motions with very
high A
p
/V
p
ratios produce elastic or nearly elastic isolator
responses. Since the concept of equivalent linear analysis does
not apply to elastic response, the choice of the range of A
p
/V
p
ratio (5.23 s
1
and 21.5 s
1
) is found adequate for the purpose
of this study.
6. Parameters included in the analyses
A parametric study is conducted to test the accuracy of
the seismic response quantities obtained from EL analyses.
As mentioned earlier, the A
p
/V
p
ratio of the ground motion
is anticipated to affect the seismic response and hence the
accuracy of the seismic response quantities obtained from EL
1656 M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663
Table 1
Important features of earthquake records used in the analyses
Earthquake Station/component A
p
(g) V
p
(cm/s) A
p
/V
p
(1/s) Duration (s) Strong motion
duration (s)
San Fernando, 1971 8244 Orion Blvd./180

0.13 23.9 5.5 59.60 17.92

Imperial Valley, 1940 El Centro/180

0.21 36.5 5.8 53.48 26.60

Loma Prieta, 1989 Oakland Outer Wharf/0

0.22 35.4 6.1 40.00 8.80

Loma Prieta, 1989 Oakland Outer Wharf/270

0.28 37.6 7.2 40.00 9.94

Northridge, 1994 Arleta and Nordhoff Fire Station/90

0.34 40.4 8.4 60.00 24.08

Kern County, 1952 Taft Lincoln Tunnel/69

0.16 15.7 9.7 54.38 19.60

Imperial Valley, 1940 El Centro/270

0.35 32.3 10.6 53.76 25.86

Santa Barbara, 1978 283 Santa Barbara Courthouse/222

0.20 16.3 12.2 12.60 4.15

Coalinga, 1983 36227 ParkeldCholame 5W/270

0.15 10.8 13.4 40.00 9.95

Northridge, 1994 Santa Monica City Hall Grounds/0

0.37 24.9 14.6 60.00 12.92

Whittier Narrows, 1987 24401 San Marino, SW Academy/360

0.20 12.8 15.6 40.00 4.54

Whittier Narrows, 1987 90079 Downey Birchdale/90

0.24 13.7 17.4 28.60 6.80

San Fernando, 1971 Pacoima Dam./196

1.08 57.5 18.4 41.70 33.6

Northridge, 1994 Santa Monica City Hall Grounds/90

0.88 41.8 20.7 60.00 13.58

Parkeld, 1966 Cholame, Shandon/40

0.24 10.8 21.5 26.20 7.90

Table 2
Properties used to represent various bearing types
k
i
(kN/m) k
p
(kN/m) Associated isolator type
10,000 2000 High damping rubber bearing
20,000 1000 Leadrubber bearing
20,000 2000 Leadrubber bearing
20,000 4000 High damping rubber bearing
200,000 2000 Friction-based bearings
analyses. Therefore, simulated harmonic and seismic ground
motions with various A
p
/V
p
ratios are used in the analyses.
Furthermore, the intensity of the ground motion relative to the
characteristic strength of the isolator is anticipated to affect the
response of seismic-isolated structures and hence the accuracy
of the seismic response quantities obtained from EL analyses.
For this purpose, a dimensionless ratio, A
p
W/F
yi
(W: weight
acting on isolator), which represents the ratio of the seismic
inertial force of a rigid structure to the characteristic strength of
the isolator, is used as a parameter in the analyses. Keeping the
W/F
yi
ratio constant, the ground motions are scaled between
0.1g and 0.5g to obtain A
p
W/F
yi
ratios ranging between 2
and 10. Additionally, the type and properties of the isolator
is anticipated to affect the seismic response and hence the
accuracy of the seismic response quantities obtained from EL
analyses. Accordingly, as demonstrated in Table 2, the elastic
stiffness, k
i
, and post-elastic stiffness, k
p
, of the isolator are
varied to represent various types of rubber-based and friction-
based isolators to cover a broad range of possibilities for
testing the accuracy of the seismic response quantities obtained
from EL analyses. For the harmonic ground motions, ground
excitation duration is also considered as a parameter in the
evaluation of the EL analysis results.
7. Analyses conducted
The NLTH analyses of the seismic isolated systems
considered in this study are conducted using the program
NONLIN [16]. A total of 243 NLTH analyses are conducted
to cover a broad range of parameters.
A total of 864 iterative EL analyses are conducted for
the evaluation of the EL analyses results and verication
of the results from the same EL analysis procedure with
proposed improvements. The EL analysis procedure utilizes
a displacement response spectrum obtained for the calculated
effective damping ratio at each iteration step. Thus, for the
harmonic and suite of 15 seismic ground motions considered in
the analyses, their displacement response spectra corresponding
to the calculated effective damping ratio is obtained at each
iteration step. The spectral displacement corresponding to
the calculated effective period is then obtained from the
displacement spectrum and used as part of the iterative EL
analysis procedure.
8. Evaluation of the EL analysis for harmonic excitations
Fig. 2(a) displays the ratios, U
E
/U
NL
, of the maximum
isolator displacements obtained from EL analyses to those
obtained from NLTH analyses as a function of the A
p
/V
p
ratio of the harmonic ground motion and for various A
p
W/F
yi
ratios ranging between 2 and 10. It is observed that the
U
E
/U
NL
ratios are generally smaller than 1.0. Thus, the EL
analysis procedure produces unconservative estimates of the
displacement response ranging between 67% and 95% of
the actual nonlinear response as a function of the A
p
/V
p
and A
p
W/F
yi
ratios. On the average, U
E
/U
NL
ratios vary
between 0.80 and 0.90. Generally, the U
E
/U
NL
data become
more scattered as the A
p
/V
p
ratio increases. This clearly
demonstrates that the difference between the EL analysis and
NLTH analysis results becomes larger for harmonic ground
motions with high frequency content.
Fig. 2(b) displays the ratios, F
E
/F
NL
, of the maximum
isolator forces obtained from EL analyses to those obtained
from NLTH analyses as a function of the A
p
/V
p
ratio of
the harmonic ground motion and for various A
p
W/F
yi
ratios.
The observations for the F
E
/F
NL
ratios are similar to those
for the U
E
/U
NL
ratios. The EL analysis procedure produces
unconservative estimates of the isolator force response ranging
M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663 1657
Fig. 2. Ratios of EL analyses results to NLTH analyses results (harmonic
ground motions). (a) Maximum isolator displacement ratio for k
i
=
20, 000 kN/m, k
p
= 2000 kN/m. (b) Maximum isolator force ratio for k
i
=
20, 000 kN/m, k
p
= 2000 kN/m. (c) Maximum isolator displacement ratio for
A
p
W/F
y
i = 6.
between 86% and 99% of the actual nonlinear response.
However, in the case of F
E
/F
NL
ratios, the data is less scattered.
This indicates that the EL analysis procedure predicts the
maximum isolator forces better than the maximum isolator
displacements. This can be explained by the following equation
representing the bilinear forcedisplacement relationship of
isolators:
F = F
yi
+ k
p
U. (6)
In the equation above the isolator force has two components:
the characteristic strength, F
yi
and the restoring force, k
p
U.
The inaccuracies in the isolator displacements resulting from
the EL analysis procedure affects only the restoring force
component of the total isolator force. Consequently, the overall
error in the estimation of the isolator force is less than that of
the isolator displacement.
Fig. 2(c) displays the U
E
/U
NL
ratios as a function of
the A
p
/V
p
ratio of the harmonic ground motion for various
isolation system properties. It is observed that the U
E
/U
NL
ratios range between 0.56 and 1.07 as a function of the A
p
/V
p
ratio and isolation system properties. On the average, U
E
/U
NL
ratios vary between 0.82 and 0.93. In general, for harmonic
ground excitations with low frequency content (A
p
/V
p
<
10), the U
E
/U
NL
ratios are similar for all isolation system
properties considered and fall within the range of 0.80 and
0.98. However, for harmonic ground excitations with higher
frequency content (A
p
/V
p
> 10), the U
E
/U
NL
ratios further
decrease as the elastic stiffness, k
i
, of the isolation system
increases. For instance, at A
p
/V
p
= 20, the U
E
/U
NL
ratios
are equal to 0.99, 0.87 and 0.56 for isolation systems with k
i
values of 10, 000 kN/m, 20, 000 kN/m and 200, 000 kN/m
respectively. This indicates that the EL analysis produces the
most unconservative estimates of the displacement response for
friction-based seismic isolation systems subjected to harmonic
ground motions with high A
p
/V
p
ratio.
The durations of the harmonic ground motions used in
the analyses are varied within the practical range of interest
(between 5 and 20 s, simulating the strong part of a typical
ground motion) to study the effect of the ground motion
duration on the prediction of the actual nonlinear envelope
responses by EL analyses. Analyses results have revealed that
the duration of the harmonic ground motions does not affect the
envelope of the nonlinear responses and hence the prediction
of such responses by EL analyses for the seismic isolation
systems considered in this study. This may not be true for
ground motions with very short duration.
9. Effect of structure mass on effective damping ratio
The seismic response of a seismic-isolated structure depends
on the amount of energy dissipated by the isolator during
each cyclic motion. Consequently, the quantication of the
energy dissipation capacity of an isolator is very important. To
quantify its energy dissipation capacity, the isolator is generally
subjected to a harmonic cyclic displacement with a period,
T
g
and amplitude U
d
and the resulting force is measured.
The energy dissipated per cycle is then calculated as the area
under the forcedisplacement hysteresis loop. In such tests,
the isolator possesses virtually no inertial mass. Thus, the
inertial forces are negligibly small compared to the forces
generated from elasticity and energy dissipation. Accordingly,
the effective damping ratio,
e
, of an equivalent elastic system
simulating the nonlinear behavior of the isolator is simply
calculated using Eq. (2).
However, for an isolator supporting a rigid mass, m, and sub-
jected to a harmonic cyclic displacement with a period, T
g
and
amplitude U
d
, the effective damping constant, C
e
, of an equiv-
alent elastic system simulating the nonlinear behavior of the
1658 M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663
Fig. 3. Effect of including T
e
and T
g
in damping equation (harmonic
ground motions). (a) Maximum isolator displacement ratio for k
i
=
20, 000 kN/m, k
p
= 2000 kN/m. (b) Maximum isolator displacement ratio
for A
p
W/F
y
i = 6.
isolator is related to the energy dissipated per cycle as [13,17];
C
e
= 2
e1
m
2
T
e
=
4F
yi
(U
d
U
y
)

2
T
g
U
2
d
(7)
where T
e
is the effective (modal) period and
e1
is the effective
modal damping ratio of the equivalent elastic system. Knowing
that:
m = k
e
_
T
e
2
_
2
(8)
and solving for
e1
from Eq. (7):

e1
=
T
g
T
e
4F
yi
(U
d
U
y
)
2k
e
U
2
d
. (9)
Comparing Eqs. (2) and (9),
e1
is expressed as:

e1
=
T
g
T
e

e
. (10)
The above equation relates the effective modal damping
ratio of a seismic-isolated structure with nite mass to the
effective damping ratio of the same structure with no mass. This
clearly demonstrates that the effective damping ratio used in
the design of seismic-isolated structures must somehow involve
the effective period of the structure and dominant period (or
frequency content) of the ground motion.
The EL analyses of the same seismic-isolated systems
subjected to harmonic ground motions are conducted using
Eq. (9) for the effective damping ratio instead of Eq. (2).
This second set of analyses is conducted to demonstrate the
effect of including T
g
and T
e
in the effective damping equation
on the accuracy of the EL analyses results. The period, T
g
,
of the harmonic ground motion and the effective period, T
e
,
of the seismic-isolated system are substituted in Eq. (9) at
each iteration step of the EL analysis procedure to obtain the
effective modal damping ratio,
e1
. The analyses results are
depicted in Fig. 3.
Fig. 3(a) displays the ratios, U
E
/U
NL
, of the maximum
isolator displacements obtained from EL analyses to those
obtained from NLTH analyses as a function of the A
p
/V
p
ratio of the harmonic ground motion and for various A
p
W/F
yi
ratios. Comparison of this gure with Fig. 2(a) reveals that
including T
g
and T
e
in the effective damping equation resulted
in considerable improvement to the accuracy of the EL analysis
results. Overall, the scatter of the data is reduced. Now,
the EL analysis produces more reasonable estimates of the
displacement response ranging between 92% and 112% of the
actual nonlinear response as a function of the A
p
/V
p
and
A
p
W/F
yi
ratios. On the average, U
E
/U
NL
ratios are more
conservative and vary between 1.03 and 1.09.
Fig. 3(b) displays the U
E
/U
NL
ratios as a function of the
A
p
/V
p
ratio of the harmonic ground excitation for various
isolator properties. Comparison of this gure with Fig. 2(c)
reveals that including T
g
and T
e
in the effective damping
equation resulted in considerable improvement in the accuracy
of the EL analysis results for various range of isolator properties
considered in this study. Now, the EL analysis produces more
reasonable estimates of the displacement response ranging
between 82% and 116% of the actual nonlinear response for
various isolator properties. On the average, U
E
/U
NL
ratios are
more conservative and vary between 0.98 and 1.10.
10. Evaluation of the EL analysis for seismic excitations
Fig. 4(a) displays the U
E
/U
NL
ratios as a function of
the A
p
/V
p
ratio of the seismic ground motion and for
various A
p
W/F
yi
ratios ranging between 2 and 10. It is
observed that the U
E
/U
NL
ratios are generally smaller than
1.0 with only a few exceptions. On the average, the EL
analysis produces mostly unconservative estimates of the actual
nonlinear displacement response with U
E
/U
NL
ratios ranging
between 0.64 and 1.11. It is observed that generally, the scatter
of the data increases for larger A
p
W/F
yi
ratios. Furthermore,
the thick solid curve representing the average of the results
from all A
p
W/F
yi
cases indicates that the difference between
the EL and NLTH analyses results becomes larger for ground
motions with high frequency content (large A
p
/V
p
ratio). This
is in agreement with the ndings of the analyses cases involving
harmonic ground motion.
Fig. 4(b) displays the U
E
/U
NL
ratios as a function of the
A
p
/V
p
ratio of the seismic ground motion for various isolator
properties. It is observed that, on the average, U
E
/U
NL
ratios
are mostly unconservative and vary between 0.70 and 1.22.
In general, the difference between the EL and NLTH analyses
M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663 1659
Fig. 4. Ratios of EL analyses results to NLTH analyses results (seismic
ground motions). (a) Maximum isolator displacement ratio for k
i
=
20, 000 kN/m, k
p
= 2000 kN/m. (b) Maximum isolator displacement ratio for
A
p
W/F
y
i = 6. (c) Maximum isolator displacement ratio for various A
p
/V
p
ratios.
results becomes larger for ground motions with large A
p
/V
p
ratios regardless of the properties of the isolator.
Fig. 4(c) displays the variation of U
E
/U
NL
ratios as a
function of the A
p
W/F
yi
ratios. It is observed that generally
the scatter of the data increases for larger A
p
W/F
yi
ratios.
However, the thick solid curve representing the average of the
results from all A
p
/V
p
cases indicates that, on the average,
the EL analysis produces more reasonable estimates of the
displacement response for larger A
p
W/F
yi
ratios.
11. Variation of effective damping ratio as a function of
period shift
Earlier, it has been demonstrated that the effective damping
ratio,
e
, currently used in the design of seismic-isolated
structures (Eq. (2)) must somehow involve the effective period
of the structure and dominant period (or frequency content) of
the ground motion (Eq. (9)) for a more accurate estimation of
the nonlinear response quantities obtained from EL analyses.
To picture the variation of the actual effective damping ratio
as a function of the effective period and to assess the accuracy
of the existing effective damping equation, rst, the effective
damping ratios,
er
, required to produce EL analysis results
equal to those of the NLTH analyses are obtained for the cases
considered in this study. This required an iterative analysis
procedure using the bisection search method [18]. The effective
damping ratios,
e
, currently used in the design of seismic-
isolated structures are also calculated using Eq. (2). Next, the

er
/
e
ratios are calculated and displayed in Fig. 5 as a function
of the period shift, T
e
/T
i
where T
i
represents the initial elastic
period of the seismic-isolated structure. For the seismic-isolated
systems considered in this study, T
i
is calculated using the
elastic stiffness, k
i
, of the isolator and the rigid mass supported
on the isolator. The data presented in Fig. 5 reveal that Eq.
(2), which is currently used in the design of seismic-isolated
structures, generally yields effective damping ratios larger than
those required to accurately estimate the nonlinear response by
EL analysis. This obviously produces unconservative estimates
of the actual nonlinear responses as demonstrated earlier.
Although there is a considerable scatter in the data, it is
observed that the
er
/
e
, ratio approaches unity with increasing
period shift (or increasing effective period). This indicates that
the EL analyses may yield more reasonable estimates of the
actual nonlinear response of seismic-isolated structures with
larger effective periods. This is conrmed by the analysis
results presented earlier. It was observed that ground motions
with lower A
p
/V
p
ratio generally produce larger displacements
and hence larger effective periods. Accordingly, EL analysis
procedure yields more reasonable estimates of the actual
nonlinear response of seismic-isolated structures subjected to
ground motions with lower A
p
/V
p
ratio and hence larger
effective period as conrmed by Fig. 4(a) and (b). Furthermore,
it was observed that larger effective periods are associated with
larger A
p
W/F
yi
ratios that represent the intensity of the ground
motion relative to the characteristic strength of the isolator.
Accordingly, EL analysis yields more reasonable estimates
of the actual nonlinear response of seismic-isolated structures
subjected to ground motions with larger A
p
W/F
yi
ratio and
hence larger effective period as conrmed by the average of the
results (thick solid line) presented in Fig. 4(c).
Based on the above discussions, it may be concluded that
the effective period implicitly includes the effect of the A
p
/V
p
ratio or frequency characteristics (dominant period) of the
ground motion. Thus, Fig. 5 represents the variation of
er
/
e
ratios as a function of the effective period (or period shift) as
well as, implicitly, the intensity and frequency characteristics
(dominant period) of the ground motion.
1660 M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663
Fig. 5. Ratio of actual effective damping to effective damping calculated using
Eq. (2) as a function of the effective period shift.
12. Improvement of the effective damping ratio
The data presented in Fig. 5 is used to incorporate the
effective period in the existing damping equation to improve
the prediction of nonlinear responses by EL analysis. For this
purpose, a procedure similar to that followed by Iwan [19]
to estimate the inelastic spectra from elastic response spectra
is employed to obtain an analytical equation representing the
variation of the
er
/
e
ratio as a function of the effective period,
T
e
. Iwan [19] used two parameters; the period shift, T
e
/T
i
and effective damping,
e
, to dene an optimal effective linear
system for any particular inelastic system. Accordingly, the
logarithms of the
er
/
e
, ratios presented in Fig. 5 are rst
plotted as a function of the logarithms of the relative period
shifts (T
e
/T
i
1) of the seismic-isolated systems considered in
this study. Then a minimumleast square t of the loglog data is
performed to obtain the following equation for the
er
/
e
ratio
as;

er

e
=
_
0.41
_
T
e
T
i
1
_
. (11)
The above equation is plotted in Fig. 5. It is the analytical
representation of the variation of the actual damping ratio
relative to Eq. (2) currently used in the design of seismic-
isolated structures. Substituting Eq. (2) in place of
e
in Eq.
(11) and solving for
er
, the improved effective damping ratio
is expressed as:

er
=
4F
yi
(U
d
U
y
)
2k
e
U
2
d
_
0.41
_
T
e
T
i
1
_
. (12)
The new effective damping ratio is a function of the
properties of the forcedisplacement hysteresis loop of the
isolator, design displacement as well as the dynamic properties
of the isolated structure, which are also functions of the
intensity and frequency characteristic of the ground motion.
Equations similar to that derived above were also proposed by
Iwan [19] (Eq. (13)) and Hwang et al. [20] (Eq. (14))

e
= 13.52
_
T
e
T
i
1
_
0.3952
(13)

e
=
_
_
_
2
_
1
k
p
k
i
_ _
1
U
y
U
d
_
e

_
1 +
k
p
k
i
_
U
d
U
y
1
__
_
_
_
_
U
d
U
y
_
0.58
6 10
k
p
k
i
. (14)
The proposed equations Eqs. (2) and (12)(14) are plotted
in Fig. 6 together with the data representing the actual effective
damping ratios,
er
, required to produce EL analysis results
equal to those of the NLTH analyses. In the gure, the data
points (
er
) correspond to various seismic isolation systems
(with various properties; k
i
, k
p
, F
yi
and m) subjected to a
variety of ground motions scaled to a range of intensities. In
plotting the gure, rst, the actual effective stiffness, k
e
=
F
d
/U
d
, of the seismic isolation system is calculated for each
data point using the isolator design force, F
d
, and displacement,
U
d
, obtained from the NLTH analyses. Next, the calculated
effective stiffness and the mass, m, of the seismic isolation
system are substituted in Eq. (3) to obtain the effective period,
T
e
, for each data point. Then, the initial elastic periods, T
i
,
and hence the T
e
/T
i
ratios are calculated using the isolation
system properties corresponding to each data point. Finally,
Eqs. (2) and (12)(14) are calculated using the isolation system
properties corresponding to the relevant T
e
/T
i
ratio for each
data point and the results are plotted as a function of the
T
e
/T
i
ratios. The gure demonstrates that the damping equation
(2) currently used in the design of seismic isolated structures
generally overestimates the actual damping ratio. However, the
improved damping equation proposed in this study generally
yields reasonable estimates of the actual effective damping
ratio for all ranges of period shifts. Similarly, the effective
damping equation proposed by Hwang et al. [20] generally
yields reasonable estimates of the actual effective damping ratio
for T
e
/T
i
< 2.5. However, the effective damping equation
proposed by Iwan [19] does not follow the general pattern of
the actual damping ratio data. It underestimates the effective
damping ratio for T
e
/T
i
< 2.5. For larger period shifts,
damping equations proposed by both Hwang et al. [20] and
Iwan [19] overestimate the actual damping ratio.
13. Comparison of EL analysis results using the proposed
and other effective damping equations
EL analyses of the seismic-isolated systems considered
in this study are re-conducted using the damping equation
proposed in this study (Eq. (12)) and those proposed by
Iwan [19] and Hwang [20]. The U
E
/U
NL
ratios for the EL
analyses results with the existing damping equation, the one
proposed in this study and those proposed by Iwan [19] and
Hwang [20] are illustrated in Fig. 7.
Fig. 7(a) compares the U
E
/U
NL
ratios obtained using the
proposed damping equation (Eq. (12)) with those obtained
using the existing damping equation (2) and those proposed
by Iwan [19] and Hwang [20] as a function of the A
p
/V
p
ratio of the seismic ground motions for the average of the
M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663 1661
Fig. 6. Comparison of results from various damping equations with actual
effective damping data.
Fig. 7. Comparison of U
E
/U
NL
ratios obtained using the improved and
existing damping equations and those proposed by others. (a) U
E
/U
NL
vs
A
p
/V
p
ratios for the average of cases involving all A
p
W/F
y
i ratios for seismic
ground motions. (b) U
E
/U
NL
vs A
p
W/F
y
i ratios for the average of all A
p
/V
p
cases for seismic ground motions. (c) U
E
/U
NL
vs A
p
/V
p
ratios for the average
of cases involving all A
p
W/F
y
i ratios for harmonic ground motions.
results for all the A
p
W/F
yi
ratios considered in this study. It is
observed that the proposed damping equation generally yields
more reasonable estimates of the actual nonlinear response
of the seismic-isolated systems considered in this study in
comparison with other damping equations. The U
E
/U
NL
ratios
for the proposed damping equation generally uctuate around
1.0 whereas those for the existing damping equation (2) are
generally smaller than 1.0 (unconservative). Furthermore the
proposed damping equation yields more reasonable estimates
of the actual nonlinear responses for A
p
/V
p
> 9.5 in
comparison with Iwans [19] damping equation and for
A
p
/V
p
< 9.5 and A
p
/V
p
> 17.5 in comparison with
Hwangs [20] damping equation.
Fig. 7(b) displays the U
E
/U
NL
ratios as a function of
the A
p
W/F
yi
ratio for the average of the results for all
the seismic ground motions with various A
p
/V
p
ratios. The
gure demonstrates that the proposed damping equation yields
more reasonable and consistently conservative estimates of
the actual nonlinear responses of the seismic-isolated systems
considered in this study compared to other damping equations.
While Eq. (2) yields unconservative estimates of the actual
nonlinear responses for the whole range of A
p
W/F
yi
ratio
considered in this study, the damping equation proposed by
Hwang [20] yields unconservative estimates of the actual
nonlinear responses for A
p
W/F
yi
> 6.
Fig. 7(c) is similar to Fig. 7(a), but the results are presented
for the harmonic ground motions. In the gure, a notable
improvement in the prediction of the actual nonlinear response
is observed when the proposed damping equation (Eq. (12)) is
used in the EL analyses compared to the results obtained using
the existing damping equation (Eq. (2)) and those proposed
by Iwan [19] and Hwang [20]. The gure also proves that the
proposed damping equation yields more reasonable estimates
of the actual nonlinear response regardless of the type of ground
motion used in the analyses.
The dispersion, d, of the U
E
/U
NL
ratios with respect to
1.0 for the existing and proposed damping equations and those
proposed by Iwan [19] and Hwang [20] are calculated for each
set of A
p
W/F
yi
ratios using the following equation [21];
d =
n

1=1

U
E
U
NL
1

n
(15)
where n is the number of data points. The results are plotted as
functions of the A
p
/V
p
ratios of the seismic ground motions
in Fig. 8(a) and harmonic ground motions in Fig. 8(b). As
observed fromFig. 8(a) and (b), the proposed damping equation
produces less dispersed data in most cases. This indicates
that the proposed damping equation does not only improve
the accuracy of the average EL results, it also improves the
accuracy of the individual EL analysis results for various
A
p
/V
p
and A
p
W/F
yi
ratios.
14. Conclusions
In this study the accuracy of the EL analysis in predicting
the actual nonlinear response of seismic isolated structures
1662 M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663
Fig. 8. Dispersion of U
E
/U
NL
data for improved and existing damping
equations and those proposed by others. (a) Seismic ground motions. (b)
Harmonic ground motions.
is evaluated. The existing damping equation is modied to
improve the accuracy of the EL analysis. Following are the
conclusions drawn from this research study:
It is observed that the EL analysis generally produces un-
conservative estimates of the actual maximum displacement
response of seismic isolators using the existing damping
equation.
However, the EL analysis predicts the maximum isolator
forces better than the maximum isolator displacements as
the inaccuracies in the prediction of isolator displacements
affect only the restoring force component of the total isolator
force.
Analysis results have revealed that the difference between
the EL analysis results and the actual nonlinear responses
becomes larger for ground motions with high frequency
content (high A
p
/V
p
ratio) regardless of the properties of
the isolator.
It is also observed that, on the average, the EL analysis
produces more reasonable estimates of the actual nonlinear
responses for ground motions with large intensity relative to
the characteristic strength of the isolator (larger A
p
W/F
yi
ratios).
Accordingly, as ground motions with larger intensity and
lower A
p
/V
p
ratio generally produce larger displacements
and hence larger effective periods, it may be concluded that
the EL analysis yields more reasonable estimates of the
actual nonlinear response of seismic-isolated systems with
larger effective periods.
Furthermore, it is demonstrated that the effective damping
ratio used in the design of seismic-isolated structures must
somehow involve the effective period of the structure and
dominant period (or frequency content) of the ground
motion for a more accurate estimate of the actual nonlinear
response of the seismic-isolated structure.
Accordingly, an improved effective damping equation that
incorporates the effective period is proposed. It is found
that the proposed damping equation yields more reasonable
estimates of the actual effective damping compared to
several other damping equations found in the literature.
The EL analyses of the seismic-isolated systems considered
in this study are re-conducted using the proposed damping
equation and other damping equations found in the literature.
It is found that the proposed damping equation yields
more reasonable estimates of the actual nonlinear responses
regardless of the type of ground motion used in the analyses.
Furthermore, it is observed that the proposed damping
equation reduces the dispersion of the U
E
/U
NL
data. This
indicates that the proposed damping equation does not only
improve the accuracy of the average EL results, it also
improves the accuracy of the individual EL analysis results
for various A
p
/V
p
and A
p
W/F
yi
ratios.
References
[1] AASHTO. Guide specications for seismic isolation design. Washington
(DC); 1999.
[2] International Code Council. International building code. Falls Church,
(Virginia); 2000.
[3] Jacobsen LS, Ayre RS. Engineering vibrations. New York: McGraw-Hill;
1958.
[4] Hwang JS, Sheng LH. Equivalent elastic seismic analysis of base-isolated
bridges with leadrubber bearings. Engineering Structures 1994;16(3):
2019.
[5] Hwang JS, Sheng LH, Gates JH. Practical analysis of base-isolated
bridges with bilinear hysteresis characteristics. Earthquake Spectra 1994;
10(4):70527.
[6] Hwang JS, Sheng LH. Effective stiffness and equivalent damping of base-
isolated bridges. Journal of Structural Engineering ASCE 1993;119(10):
3094101.
[7] Hwang JS. Evaluation of equivalent linear analysis methods of bridge
isolation. Journal of Structural Engineering ASCE 1996;122(8):9726.
[8] Hwang JS, Chang KC, Tsai MH. Composite damping ratio of seismically
isolated regular bridges. Engineering Structures 1997;19(1):5262.
[9] Franchin P, Monti G, Pinto PE. On the accuracy of simplied methods for
the analysis of isolated bridges. Earthquake Engineering and Structural
Dynamics 2001;30(3):36382.
[10] Naeim F, Kelly JM. Design of seismic isolated structures; From theory to
practice. Chichester (UK): Wiley; 1999.
[11] HITEC (Highway Innovation Technology Evaluation Center). Evaluation
ndings for R.J. Watson Inc. sliding isolation bearings. Technical
evaluation report. Reston (VA): ASCE. 1998.
[12] Zayas VA, Low SS. Seismic isolation of bridges using friction pendulum
bearings. In: Proceedings of the 1999 structures congress structural
engineering in the 21st century. Reston (VA): LA ASCE; 1999. p. 99102.
[13] Chopra AK. Dynamics of structures: Theory and applications to
earthquake engineering. Upper Saddle River, NJ: Prentice Hall; 2001.
[14] Dicleli M, Bruneau M. An energy approach to sliding of single-span
simply supported slab-on-girder steel highway bridges with damaged
bearings. Earthquake Engineering and Structural Dynamics 1995;24(3):
395409.
[15] Kramer SL. Geotechnical earthquake engineering. NJ: Prentice Hall;
1996.
[16] NONLIN. Federal emergency management agency. Washington (DC);
2000.
M. Dicleli, S. Buddaram / Engineering Structures 29 (2007) 16531663 1663
[17] Makris N, Chang S. Effect of damping mechanisms on the response
of seismically isolated structures, PEER Report 1998/06. (Berkeley):
Pacic Earthquake Engineering Research Center, Collage of Engineering,
University of California; 1998.
[18] Chapra SC, Raymond C. Numerical methods for engineers: With software
and programming applications. New York (NY): McGraw Hill; 2002.
[19] Iwan WD. Estimating inelastic response spectra from elastic spectra.
Earthquake Engineering and Structural Dynamics 1980;8(4):37588.
[20] Hwang JS, Chiou JM, Sheng LH. Establishment of an equivalent linear
model of seismically isolated bridges using a system identication
method. In: Proc., 4th US conf. on lifeline earthquake engrg. New York
(NY): ASCE; 1995. p. 41623.
[21] Kottegoda NT, Rosso R. Statistics, probability and reliability for civil and
environmental engineers. New York (NY): McGraw Hill; 1997.