Earthquake Resistant Design of Structures using the Variable Frequency Pendulum Isolator

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

Pranesh Murnal1 and Ravi Sinha, M.ASCE2

Abstract: The essential properties of sliding isolators used for earthquake resistant design are period shift, energy dissipation, and the
restoring mechanism. Isolation systems using a curved surface incorporate all of these in a single unit. The systems currently available
have limitations due to their period and restoring force characteristics. The writers have recently proposed a new isolator called the
variable frequency pendulum isolator VFPI, which overcomes these limitations while retaining the advantages. The oscillation frequency
of the VFPI continuously decreases with increase in sliding displacement, and the restoring force has an upper bound so that the force
transmitted to the structure is bounded. The mathematical formulation for the response of multi-degree-of-freedom MDOF structures
isolated using the VFPI has been discussed in this paper. Parametric studies have been carried out to examine the behavior of MDOF
structures and structure-equipment systems isolated with the VFPI, friction pendulum system, and pure friction isolator. The VFPI
performance is found to be effective and stable during low- and medium-intensity excitations, and fail-safe during high-intensity excitations. The VFPI is also found to exhibit robust performance for a wide range of isolator and ground motion characteristics, clearly
demonstrating its advantages for vibration control of MDOF structures.
DOI: 10.1061/ASCE0733-94452002128:7870
CE Database keywords: Isolation; Earthquake resistance structures; Seismic design.

Introduction
Use of base isolation systems has emerged as a very effective
technique for aseismic design of structures. In the base isolation
technique, a flexible layer or isolator is placed between the
structure and its foundation such that relative deformations are
permitted at this level. Due to the flexibility of the isolator layer,
the time period of motion of the isolator is relatively long; as a
result the isolator time period controls the fundamental period of
the isolated structure. For properly designed isolation system, the
isolator time period is much longer than those containing significant ground motion energy. As a result, the use of isolator shifts
the fundamental period of the structure away from the predominant periods of ground excitation. Extensive reviews of base isolation systems and its applicability are available in the literature
Kelly 1986; Buckle and Mayes 1990; Kelly 1993; Naeim and
Kelly 1999.
Practical isolation devices typically also include an energy dissipating mechanism so as to reduce deformations at the isolator
level. For example, the friction-type base isolator uses a sliding
surface for both isolation and energy dissipation, and has been
found to be very effective in reducing structural response Mostaghel et al. 1983. Due to the characteristics of excitation trans1

Former Graduate Student, Indian Institute of Technology Bombay,

Powai, Mumbai-400076, India.
2
Associate Professor, Indian Institute of Technology Bombay, Powai,
Mumbai-400076, India. E-mail: rsinha@civil.iitb.ac.in
Note. Associate Editor: Takeru Igusa. Discussion open until December
1, 2002. Separate discussions must be submitted for individual papers. To
extend the closing date by one month, a written request must be filed with
the ASCE Managing Editor. The manuscript for this paper was submitted
for review and possible publication on June 15, 2000; approved on May
11, 2001. This paper is part of the Journal of Structural Engineering,
Vol. 128, No. 7, July 1, 2002. ASCE, ISSN 0733-9445/2002/7870 880/$8.00$.50 per page.

mitted through the sliding surface, the performance of friction

isolators is relatively insensitive to variations in the frequency
content and amplitude of the input excitation, making performance of sliding isolators very robust Mostaghel and Tanbakuchi
1983. The simplest sliding isolator consists of a horizontal sliding surface pure friction, or PF system; however, it may experience large sliding and residual displacements, which are often
difficult to incorporate in structural design. Several systems have
been suggested in the past to accommodate the restoring mechanism along with sliding in order to reduce the sliding and residual
displacements to manageable levels Mostaghel and Khodaverdian 1987; Kelly 1988; Chalhoub and Kelly 1990; Mokha et al.
1990;
An effective mechanism to provide restoring force by gravity
has been utilized in the friction pendulum system FPS Zayas
et al. 1987; Zayas et al. 1990. In this system, the sliding surface
takes a concave spherical shape so that the sliding and recentering
mechanisms are integrated in one unit. The FPS isolator has several advantages over the pure friction isolator, and has demonstrated acceptable performance for many different structures and
excitation characteristics Mokha et al. 1991, 1996; Tsopelas
et al. 1996; Tsai 1997; Wang et al. 1998. The spherical sliding
surface of the FPS isolator provides a relatively constant time
period of oscillation, resulting in several practical disadvantages.
One main disadvantage is that FPS isolators can be effectively
designed for a specific level amplitude and frequency characteristics of ground excitation. Under more severe ground motions,
sliding displacements greater than the design displacements
occur. Since the FPS isolator restoring force is linearly proportional to sliding displacement, the additional sliding introduces
additional energy in the structure. As a result, the maximum intensity of excitation has a strong influence on FPS design. In
general, FPS isolators designed for a particular intensity of excitation may not give satisfactory performance during earthquakes
with much lower or higher intensity Sinha and Pranesh 1998.

870 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng., 2002, 128(7): 870-880

force can be expressed as the product of the equivalent spring

stiffness and the deformation, i.e.,
f R k x x

(2)

where k(x)instantaneous spring stiffness, and xsliding displacement of the mass. If the mass is modeled as a single-degreeof-freedom oscillator, the spring force restoring force can be
expressed as the product of the total mass of the system and
square of oscillation frequency

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

f R m 2b x x

Fig. 1. Free-body diagram of smooth sliding surface of isolator

showing forces at point of contact

The writers have recently developed a new isolator called the

Variable Frequency Pendulum Isolator VFPI that incorporates
the advantages of both the FPS and PF isolators Pranesh and
Sinha 1998, 2000a. The most important properties of this system
are: 1 its time period of oscillation depends on sliding displacement, and 2 its restoring force exhibits softening behavior. The
geometry of the sliding surface has been defined by a secondorder parametric equation. Using this parametric equation, the
isolator properties can be chosen to achieve progressive period
shift with variation in sliding displacements. Recent investigations on single-degree-of-freedom SDOF systems have shown
the VFPI to be very effective for a variety of excitation and structural characteristics.
In the present paper the performance of the VFPI for aseismic
design of multi-degree-of-freedom MDOF structures and
primarysecondary systems has been investigated. The structure
behavior is nonlinear and nonclassically damped and is not amenable to simple solution techniques. In order to simplify the solutions and improve understanding, mathematical formulation involving complex modal analysis has been developed for
analyzing MDOF structures isolated by the VFPI. The effectiveness of the VFPI for aseismic design has been demonstrated, for
example, in structure and primarysecondary systems.

Variable Frequency Pendulum Isolator Description

Mathematical Preliminaries
In order to develop the basic mathematical model of the isolator
sliding surface, consider the motion of a rigid block of mass m
sliding on a smooth curved surface of defined geometry, y
f (x). The restoring force offered by the curved sliding surface
can be defined as the lateral force required to cause a horizontal
displacement x. Assuming a point contact the various forces acting on the sliding surface when the block is displaced from its
original position at the origin of coordinate axes are shown in Fig.
1. At any instant the horizontal restoring force due to weight of
the structure is given by
f R mg

dy
dx

(1)

Assuming that the restoring force is mathematically represented

by an equivalent nonlinear massless horizontal spring, the spring

(3)

Here, b (x)instantaneous isolator frequency, and depends

solely on the geometry of sliding surface. In the friction pendulum system, which has a spherical sliding surface, this frequency
is almost constant and is approximately equal to g/R, where R
is the radius of curvature of the sliding surface Zayas et al.
1990.
Variable Frequency Pendulum Isolator Geometry
A sliding surface based on the expression of an ellipse has been
used as the basis for developing the sliding isolator with the desired properties Pranesh 2000. The equation of an ellipse with a
and b as its semimajor and semiminor axes, respectively, and with
coordinate axes as shown in Fig. 1, is given by
yb 1 1x 2 /a 2

(4)

Differentiating with respect to x, the slope at any point on the

curve is given by
dy
b
x
2
dx a 1x 2 /a 2

(5)

If the equation of the sliding surface is represented by Eq. 4, the

frequency of oscillation can be determined by substituting Eq. 5
in Eqs. 1 and 3, and the final expression is
2b x 2I / 1x 2 /a 2

(6)

where I2 gb/a 2 square of the initial frequency of the isolator

at zero sliding displacement. It can be seen that the frequency of
an elliptical curve is fairly constant for small displacements (x
a) and this value depends on the ratio b/a 2 . This implies that
the isolator frequency is inversely proportional to the square of
the semimajor axis and an increase in its value results in a sharp
decrease in the isolator frequency.
For the VFPI, the semimajor axis of the ellipse, a, has been
taken as a linear function of the sliding displacement x as below
Pranesh and Sinha 2000a.
axd

(7)

where dconstant. Substituting in Eq. 4, the expression for

ellipse can be simplified as

yb 1

d 2 2dx sgn x
dx sgn x

(8)

where sgn(x)signum function, which assumes a value of 1 for

positive sliding displacement and 1 for negative sliding displacement. It can be observed from Eq. 8 that the upper bound
of vertical displacement is equal to b, and it occurs only at infinitely large sliding displacement.
The slope at any point on the sliding surface is given as
dy
bd

x
2
dx dx sgn x d 2 2dx sgn x

(9)

JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 871

J. Struct. Eng., 2002, 128(7): 870-880

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Frequency and restoring force characteristics of variable frequency pendulum isolator VFPI and friction pendulum system FPS: a
frequency ratio and b normalized isolator force 0.02, VFPI: d0.3 m, b0.09 m; FPS: R1.0 m

To simplify the notations, a nondimensional parameter r

x sgn(x)/d is used. By substituting r and the initial frequency
I2 gb/d 2 in Eq. 9, and combining with Eqs. 1 and 3, the
isolator frequency at any sliding displacement can be expressed as
2b x

2I

(10)

1r 2 12r

In the above equations, parameters b and d completely define

the isolator characteristics. It can be observed that the ratio b/d 2
governs the initial frequency of the isolator. Similarly, the value
of 1/d determines the rate of variation of the isolator frequency,
and this factor has been defined as the frequency variation factor
FVF. It can also be seen from Eq. 10 that the rate of decrease
of the isolator frequency is directly proportional to the FVF for a
given initial frequency. The variation of oscillation frequency of a
typical VFPI with respect to the sliding displacement is shown in
Fig. 2a. For comparison purposes, the oscillation frequency of
the FPS with the same initial frequency has also been shown,
which is found to be almost constant. From this plot it is seen that
the oscillation frequency of the VFPI sharply decreases with increasing sliding displacement and asymptotically approaches
zero.
The forcedeformation hysteresis loops obtained from Eq. 3,
for example, the FPS and VFPI, are shown in Fig. 2b. In this
plot, the isolator force restoring forcefrictional force has been
normalized with respect to mg and the coefficient of friction is
taken as constant. It can be observed that the isolator force in the
VFPI first increases to reach its maximum value, and later slowly
decreases so as to asymptotically approach the frictional force at
large sliding displacement. This is an important property of the
VFPI, which limits the force transmitted to the structure. This
does not mean that the restoring capability is lost for larger displacements beyond the peak restoring force, as a smaller
amount of restoring force is always available. It is also observed
that for very small displacements variation of the restoring force
is approximately linear. In the FPS, on the other hand, the restoring force always varies linearly with the sliding displacements.
From this it can be concluded that VFPI behavior is similar to that
of the FPS for small displacements and its behavior is similar to
the PF system for very large displacement without significant loss
of restoring capability.
The profile of the sliding surface for the examples of the VFPI
and FPS has been shown in Fig. 3. It can be observed that the
VFPI is relatively flatter than the FPS, which results in smaller
vertical displacement for similar sliding displacements. This is an
additional advantage of the VFPI compared to the FPS since the

flatter sliding surface will induce smaller overturning forces in the

structure. A schematic diagram of the VFPI and its attachment
mechanisms is shown in Fig. 4. It has to be observed from this
Fig. 4 that, practically, the slider has to have a single point of
contact with the sliding surface so that the varying radius of curvature of the sliding surface does not interfere with smooth movement of the slider. Isolators with point contact at the sliding surface have been developed and used by many researchers and this
does not cause practical difficulties if proper choice of material is
made for example, see Shustov 1992; Kemeny 1997; Zhou et al.
1998.

Mathematical Formulation
The mathematical formulation has been developed for the response of a multistory building isolated by a sliding-type isolator.
However, the formulation is applicable to any MDOF structure.
We first consider an N-story shear structure isolated by a slidingtype isolator. The motion of the structure can be in either of two
phases: a nonsliding or sliding phase. In the nonsliding phase, the
structure behaves like a conventional fixed base structure since
there is no relative motion at the isolator level. When the frictional force at the sliding surface is overcome, there is relative
motion at the sliding surface, and the structure enters the sliding
phase. The total motion consists of a series of alternating nonsliding and sliding phases. The mathematical formulation of equations of motion in these two phases is briefly described below.

Fig. 3. Profile of sliding surfaces of friction pendulum system FPS

and variable frequency pendulum isolator VFPI: a FPS and b
VFPI

872 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng., 2002, 128(7): 870-880

gacceleration due to gravity. Substituting tan dy/dxb in Eq.

13 and simplifying, it reduces to

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

i1

m i x i x g m b x g m t g

dy
m t g
dx b

(14)

As seen earlier, the term m t gdy/dx b restoring force of the structure with mass m t for a given sliding displacement, x b Eq. 1.
Substituting Eqs. 1 and 3 in Eq. 14, this restoring force can
be expressed as a product of the total mass of the structure and
square of instantaneous isolator frequency, b (x). The condition
for the beginning of sliding finally reduces to

m i x i x g m b x g m t 2b x x b m t g

i1

(15)

Sliding Phase

Fig. 4. Schematic details of the variable frequency pendulum isolator VFPI and its attachment mechanism

Once the inequality Eq. 15 is satisfied the structure enters the

sliding phase and the degree of freedom corresponding to the base
mass also experiences motion. The equations of motion are now
given by
MxCxKxMrx g r f

where, M, C, and Kmodified mass, damping, and stiffness matrices of order N1, rmodified influence coefficient vector, and
f frictional force as given below:

Nonsliding Phase
In the nonsliding phase the structure behaves as a fixed-base
structure, since there is no relative motion between the ground
and base mass. The equations of motion in this phase are
M0 x0 C0 x0 K0 x0 M0 r0 x g

(11)

and
x b constant;

x b x b 0

(12)

where M0 , C0 , and K0 mass, damping, and stiffness matrices

of the fixed-base structure, respectively, x0 x 1 ,x 2 ,...,x N T
vector of displacements of the degrees of freedom DOFs of
the superstructure relative to the base mass excluding the DOF of
the base mass, x b displacement of the base mass (m b ) relative
to the ground, x g ground displacement, r0 influence coefficient
vector and the over dot indicates the derivative with respect to
time. Since the base mass does not move relative to the ground,
the velocity and acceleration of the base relative to the ground are
zero. However, the sliding displacement may be nonzero. The
structure is classically damped in this phase and, hence, Eq. 11
can be readily solved by usual modal analysis procedures Clough
and Penzien 1993 .

Initiation of Sliding Phase

When the structure is subjected to base excitation, it will remain
in the nonsliding phase unless the frictional resistance at the sliding surface is overcome. Assuming that the inertia force does not
affect the normal force at the sliding surface for small values of
theta see Fig. 1, the condition for the beginning of the sliding
phase can be expressed as

i1

(16)

m i x i x g m b x g cos m t g sin m t g cos

(13)

where the suffix iith DOF, m t total mass of the structure including the base mass m b , coefficient of sliding friction, and

K0

kb

M0
M0 r0

M 0 r0
T

mt

x0
x x ,
b

0
r 1

and

C0

(17)

f m t g sgn x b

Direction of Sliding
The direction of sliding depends on the signum function, which in
turn depends on the forces acting on the structure at the end of the
previous nonsliding phase. Once inequality Eq. 15 is satisfied,
the structure starts sliding in a direction opposite to the direction
of the sum of the total inertia force and restoring force at the
isolator level. So, we have

i1

sgn x b

m i x i x b x g m b x b x g m t 2b x b

i1

m i x i x b x g m b x b x g m t 2b x b
(18)

The signum function remains unchanged in a particular sliding

phase. The end of a sliding phase is governed by the condition
that the sliding velocity of the base mass is equal to zero, i.e.,
x b 0

(19)

Once the sliding velocity is zero, the structure may enter a nonsliding phase, reverse its direction of sliding, or have a momentary stop, and then continue in the same direction. To determine
the correct state, the solution process needs to continue using
equations of nonsliding phase wherein the sliding acceleration is
forced to zero and the validity of inequality Eq. 15 is checked. If
this inequality is satisfied at the same instant of time when the
sliding velocity is zero, it shows that there is a sudden stop at that
instant. The motion in the next instance of time is determined
from Eq. 18.
JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 873

J. Struct. Eng., 2002, 128(7): 870-880

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

Solution of Equations of Motion

The equations of motion can be directly solved by standard numerical integration techniques in order to obtain the time history
of responses. However, this approach is computationally expensive for large size problems and does not provide any insight into
the structure behavior. In this formulation, due to non-classical
damping, complex modal analysis has been used to uncouple the
equations of motion. The modal properties of the fixed-base structure are assumed to remain unchanged in the sliding phase to
reduce the size of the complex eigenvalue problem. The analysis
uses the nonlinear state vector approach Singh and Saurez 1992;
Sinha 1993; Pranesh 2000. The uncoupled equations are solved
by step-by-step integration procedures Pranesh and Sinha
2000a. The close-form final expressions for the evaluation of
modal response in both nonsliding and sliding phases are analogous to those for SDOF systems presented in Pranesh and Sinha
2000a.
Due to the highly nonlinear behavior of the system a very
small time step of the order of 2104 has been found suitable
for the step-by-step solution. The equations are linear in the nonsliding phase and may be nonlinear in the sliding phase. The
solution in any phase depends on the response at the end of previous phase. Further, a particular phase of response may last for a
very short duration. As a result, the transient component of response strongly influences the total response in any step. The
change in phase must be evaluated very precisely for accurate
determination of the structure response. This has been taken care
of by adaptively reducing the time interval to determine change in
phase very accurately conditions given by Eqs. 15 and 19.
The change in phase is considered to be accurately determined
when either the time interval is less than 1.01010 s or the
relative error is smaller than 1.01010, whichever is earlier.

Energy Formulation
Base isolators reduce the structural response by filtering the seismic excitations and by dissipating energy, thereby reducing the
energy that needs to be dissipated by the structure. Often it is very
difficult to decide on a proper trade off between the structural
deformations and isolator displacements for determination of isolator properties. The energy quantities are convenient to consider
since they involve all the response quantities and, hence, represent the overall response of the structure. As they are scalar, only
a single energy equation for the entire structure can be derived
irrespective of the number of degrees of freedom in the structure.
So, the energy quantities can represent the isolator performance in
a more unified manner and can be used to decide the overall
performance of the isolator.
The input energy due to ground motion is dissipated through
nonconservative energy dissipating mechanisms in the structure.
The remaining input energy is converted into a combination of
kinetic and potential energy. During the nonsliding phase there is
no relative motion at the isolator level and, hence, the entire input
energy is transmitted into the structure. Therefore, the formulation
developed by Uang and Bertero 1990 for fixed base structure is
applicable in this phase. The energy balance equation during the
sliding phase can also be similarly derived and has been presented
herein.
The input energy in the sliding phase can be defined as the
work done by the isolator force restoring forcefrictional force
due to ground displacements. The definition of input energy in
this phase is similar to the absolute input energy defined for a
fixed-base structure by Uang and Bertero 1990 as the isolator

force is identical to the inertia force. Due to sliding, some portion

of the input energy is dissipated through sliding friction and the
remaining energy is retained in different forms of conservative
energy.
The energy balance at any instant can be easily derived by
calculating the total work done by all conservative and nonconservative forces up to that instant. This can be achieved by calculating the differential work done by all the forces during a small
deformation of the structure dx0 , and then integrating to get the
total work done. The final expression for the energy balance is
found to be Pranesh 2000
2
1 T
0 M0x0t 21 m b x tb m t gy 21 xT0 K0x0
2x
t

m t g sgn x b dx b

xT0 C0x0 dt

xT0 M0r0 dx g

m b x b t dx g
(20)

Eq. 20 is similar to the absolute energy equation derived for the

conventional MDOF structure in the published literature, except
for the additional terms involving the potential energy due to
rising of the structure along the curved surface third term in Eq.
20 and the nonconservative energy term due to friction sixth
term in Eq. 20. There is no energy dissipation due to sliding
friction during the nonsliding phase. Eq. 20 can be written in
short as
E k E r E s E E E i

(21)

where E k sum of absolute kinetic energies of all the masses, and

E r and E s restorable potential energy due to rising of the structure along the sliding surface of the isolator and elastic energy
due to structural deformations, respectively. E and E energy
dissipated due to structural damping and sliding friction, respectively. As the sum of frictional force and the restoring force is
identical to the total inertia force, the term E i on rhsabsolute
input energy.
An isolator attempts to reduce the input energy (E i ), whereas
an energy dissipator attempts to dissipate a large portion of this
energy (E ). An ideal protective system will effectively incorporate both these features. If one assumes that the base isolated
structure remains elastic, the energy transmitted to the structure is
thus the conservative energy consisting of elastic strain energy,
kinetic energy, and potential energy due to the vertical rigid-body
displacement of the structure along the curved sliding surface,
neglecting the energy dissipated due to viscous damping. Hence,
the quantity of this conservative energy is a comprehensive indication of the effectiveness of the isolation system.

Response of Example Structure

The effectiveness of the VFPI to reduce the response of an example MDOF structure subjected to earthquake excitation has
been presented in this section. The NorthSouth NS component
of the El Centro 1940 earthquake has been chosen as the input
ground motion. In order to evaluate the effectiveness of the VFPI
under different levels of excitation, the recorded ground motion
has been scaled by factors called intensity factors 0.5, 1.0, and
2.0 representing the low, medium, and high intensities of excitation, respectively. These intensity factors correspond to peak
ground acceleration of 0.16, 0.32, and 0.64 g, respectively. Since
this earthquake has predominant periods at around 0.5 s, a fivestory shear structure is chosen for the study such that its funda-

874 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng., 2002, 128(7): 870-880

Table 1. Modal Properties of Fixed-Base and Isolated Structures

Mode

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

Fixed freq. Hz
Eff. modal mass %
Isolated freq. Hz
Eff. modal mass %

Isolator

0.49
99.93

1.96
87.95
3.64
0.07

5.72
8.72
6.92
0.00

9.02
2.42
9.76
0.00

11.59
0.75
11.93
0.00

13.22
0.16
13.31
0.00

mental period is close to 0.5 s. Using the formulation presented in

the paper, time-history analysis is carried out for the structure
isolated by the VFPI. To investigate the effectiveness of the VFPI,
the responses are compared with those of the structure isolated
with FPS and PF isolators. The FPS has been chosen with a radius
of 1.0 m so that its time period is around 2.0 s equal to the initial
period of the example VFPI. The example five-story shear building is represented as a lumped mass model with equal lumped
mass of 60,080 kg and equal story stiffness of 112600 kN/m for
each floor. The frequencies and modal properties for the fixedbase and isolated structures are given in Table 1. Since the natural
frequencies of a structure isolated by the VFPI change continuously with the isolator sliding displacement, the frequencies
shown in Table 1 thus indicate the upper bound on the frequencies
when the isolator displacement is zero. Previous investigations of
SDOF models have shown that FVF value between 2 and 10 per
m results in effective isolation by the VFPI Pranesh and Sinha
2000a. In this example, the VFPI chosen has an initial time period of 2.0 s and has FVF equal to 5 per m. The corresponding
values of isolator parameters b and d are 0.04 and 0.20 m, respectively. The analyses are carried out for three coefficients of friction, 0.02, 0.05, and 0.10. The structural damping is assumed
as 5% of the critical for all modes.

Modal Response Contributions

The contribution of different modes to the total response has also
been examined for the example structure subject to ground motions with intensity factor 1.0. The effective mass participation of
different modes for fixed-base and isolated structures are given in
Table 1. It is observed that the mass participation of higher modes
of the isolated structure is negligible when compared to the corresponding participation of fixed-base structures. In the isolated
structure, the first mode deformation is essentially rigid-body displacement of the structure at the isolator level, and the second
mode is due to the deformation in the structure. So, the response
of the structure in the sliding phase primarily depends on the first
two modes of response. However, it should be noted that the total
response consists of successive nonsliding and sliding phases and
the effective participation of higher modes contributing to the
overall response may not be negligible.

Time History of Response

The response quantities are evaluated by solution of the equations
of motion as discussed in the preceding sections. The main response quantities of interest are acceleration of the top story, sliding displacement of the isolator, and base shear. Time-history
plots for these response quantities of the example structure subjected to El Centro excitation with intensity factor 1.0 are shown
in Fig. 5. The maximum response values are also shown. It is seen
that there is substantial reduction in the maximum base shear for
the structure isolated by the VFPI, when compared to other isolation systems. These plots clearly demonstrate the effectiveness
of the VFPI in comparison with conventional FPS and PF systems. It can be observed from Fig. 5b that the base shear values
in the case of the VFPI tend to be limited to an upper bound. As
expected, the maximum accelerations and base shear in the structure with the VFPI are higher than in structure with the PF system. From Fig. 5c, it is observed that the sliding displacements
in the case of the VFPI increase and are even more than the PF
system. This is due to the fact that the isolator force in the VFPI
can act either as a restoring or driving force depending on the
direction of motion, whereas in the PF system the constant frictional force always opposes motion. However, the residual displacements in the VFPI are very small and are close to those of
the FPS, which clearly shows the effectiveness of the restoring
mechanism in spite of large sliding displacements. From these
response characteristics, it is therefore found that the VFPI retains
the main advantages of both FPS and PF isolators.

Fig. 5. Typical time-history response of the example shear building

subjected to El Centro 1940 NS ground motion: a top mass acceleration, b base shear, and c sliding displacement 0.02, FVF
5 per m, and 0.05
JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 875

J. Struct. Eng., 2002, 128(7): 870-880

Table 2. Modal Contribution of Maximum Response Quantities for Structures Isolated by Variable Frequency Pendulum Isolator VFPI and

Friction Pendulum System FPS

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

No. of superstructure modes

VFPI
Absolute acceleration of top g
Relative displacement of top m
Sliding displacement m
Base shear per unit weight

0.1594
0.0079
0.1272
0.0715

0.1649
0.0074
0.1268
0.0676

0.1881
0.0075
0.1269
0.0592

0.1934
0.0075
0.1269
0.0578

0.1918
0.0075
0.1269
0.0576

FPS
Absolute acceleration of top g
Relative displacement of top m
Sliding displacement m
Base shear per unit weight

0.1816
0.0102
0.1012
0.0923

0.2042
0.0095
0.1009
0.0957

0.2171
0.0096
0.1011
0.0986

0.2301
0.0096
0.1012
0.0994

0.2286
0.0096
0.1012
0.0996

The modal contributions of peak response quantities of the

example structure isolated using the VFPI and FPS are shown in
Table 2. It is observed from Table 2 that considering the first
mode of the superstructure gives an adequate estimate of the sliding displacement while this may lead to large errors in estimation
of other response quantities. This is because the sliding displacement is primarily controlled by the isolator mode, and hence the
sliding displacement of the base is not sensitive to contributions
from higher modes. It is also seen that the higher mode contributions are especially significant for the roof accelerations and base
shear. It is important to note that evaluation of base shear using
only the first mode contribution underestimates its value in the
case of the FPS, whereas it overestimates in the case of the VFPI.
Due to the highly nonlinear response of the structure, this phenomenon cannot be explained in terms of isolator characteristics
alone. However, this may be due to the highly nonlinear behavior
of the isolators accompanied by the stickslip motions in the case
of the FPS. Such motions are greatly reduced in the case of the
VFPI due to the force-softening characteristic of the restoring
force. It is further seen that for a reasonable estimate of the superstructure response quantities at least two modes of the superstructure are required. However, analysis with only one superstructure mode gives a reasonable approximation of the response
quantities and may be used for preliminary design of the isolator.
As the accelerations are likely to be underestimated by the first
mode, a larger number of modes must be considered when evaluating the structure response.

softening at smaller sliding displacement resulting in lower base

shear. As expected, there is a corresponding increase in the sliding
displacements when the FVF increases. However, this increase is
significant only for high-intensity excitation. It is also observed
that the maximum sliding displacements tend to reduce for very
high FVF. This is because the restoring force becomes very small
for FVF values greater than 10 per m and the response is prima-

Effect of Variable Frequency Pendulum Isolator

Geometry
To study the effect of geometry of the sliding surface, the variation of peak response quantities with the frequency variation factor has been considered see Fig. 6. The responses are plotted for
El Centro 1940 NS ground motions with intensity factors of 0.5,
1.0, and 2.0 representing low-, medium-, and high-intensity earthquakes, respectively. The coefficient of friction has been taken as
0.02. The reduction in acceleration with increase in the FVF is
found to be very significant for high-intensity earthquakes. It is
also observed that reduction in base shear with increase in values
of the FVF is substantial for both medium- and high-intensity
excitations. This can be explained in terms of the frequency variation characteristics of the VFPI. When the FVF is increased, there
is greater separation between dominant ground motion frequencies and instantaneous slider frequency due to the higher rate of
change of isolator frequency resulting in more effective isolation.
The force transmitted to the structure is also reduced due to force

Fig. 6. Effect of frequency variation factor on response of example

structures for scaled intensities of El Centro 1940 NS ground motion: a top floor acceleration, b base shear, c sliding displacement, and d residual displacement T I 2.0 s, 0.02, and
0.05

876 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng., 2002, 128(7): 870-880

Table 3. Ratio of Maximum Response of Structure Isolated by the Variable Frequency Pendulum Isolator to Maximum Response of Structure

Isolated by Frequency Pendulum System T I 2.0 s, T b 2.0 s.

Coefficient of friction

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

Intensity factor
Absolute acceleration of top
Relative displacement of top
Sliding displacement
Residual displacement
Base shear
Recoverable energy
Input energy

0.02

0.05

0.10

1.0

2.0

1.0

2.0

1.0

2.0

0.8391
0.7826
1.2540
3.6262
0.5784
0.4246
0.9189

0.5059
0.2898
2.1649
0.7693
0.2220
0.2147
0.6187

1.0064
0.9758
1.1183
1.0735
0.8079
0.3244
0.9691

0.8636
0.7587
1.2371
1.9000
0.5782
0.4477
0.8890

0.9913
1.0000
1.1970
0.9093
0.9805
1.0164
1.0000

0.9901
0.9542
1.0595
1.1365
0.7694
0.7223
0.9484

rily governed by the frictional force, and the behavior is similar to

the PF isolator. The residual displacements are quite small up to a
FVF value of around 10 per m for all excitation intensities. This
implies that the restoring mechanism of the isolator remains effective for a wide range of the FVF in spite of the increase in
maximum sliding displacement when the FVF value increases.

Effect of Coefficient of Friction

There exists a wide choice of materials for the isolator sliding
surface including PTFE (0.02) and Teflon (0.05 0.15).
The choice of sliding surface strongly affects the performance of
the isolation system. The performance of isolators for different
coefficients of friction has been investigated in this section. The
coefficients of friction equal to 0.02, 0.05, and 0.10 have been
chosen to span the practical range of sliding friction coefficients
of the VFPI. The peak responses and energy quantities of the
example system for different coefficients of friction are tabulated
in Table 3. The results are presented for medium and high intensities of base excitation. These results are normalized with respect
to the corresponding peak response quantities of the structure
isolated with a conventional FPS having a time period equal to
the initial time period of the VFPI, and hence the results directly
indicate the effectiveness of the VFPI in comparison to that of the
FPS.
The response results show that the VFPI is much more effective in reducing the peak acceleration and relative displacement at
the top of the structure in comparison with the FPS for a low
coefficient of friction and high intensity of excitation. This does
not mean that the VFPI is not effective for the higher coefficient
of friction and lower levels of excitations but that its advantage
with respect to the FPS in decreasing peak responses is reduced.
On the other hand, the FPS is marginally more effective in dissipating the energy for higher coefficients of friction. This implies
that the isolation characteristic remains effective in the case of the
VFPI for all coefficients of friction. The reduction in base shear
when using the VFPI is significant for all coefficients of friction
and intensities of excitation. There is also an expected increase in
the peak sliding and residual displacements for all coefficients of
friction.

The effect of the coefficient of friction on the performance of

the VFPI and FPS can be better understood by considering the
energy quantities Table 3. The use of the VFPI with the example
structure does not lead to a considerable reduction in input energy
except for the low coefficient of friction and high-intensity excitation, but exhibits a substantial reduction in the recoverable energy. For lower coefficients of friction, the FPS dissipates lesser
frictional energy thereby transmitting more energy into the structure. In the case of the VFPI, the structure response depends on
the effective isolation at all levels and lesser energy dissipation
does not affect its performance. For a lower coefficient of friction
the energy transmitted through the VFPI is bounded by the peak
isolator force and the frequency separation results in substantially
superior performance when compared to the FPS.
The effect of excitation intensity has also been investigated
from energy considerations Table 4. Table 4 gives the ratio of
peak energy quantities for high-intensity excitation intensity
factor2.0 to peak energy quantities for medium-intensity excitation intensity factor1.0 for sliding surfaces with different
coefficients of friction. It is found that input energy increases at a
higher rate when using the FPS than when using the VFPI. This is
particularly significant for the low coefficient of friction typically
chosen for VFPI. The low-energy ratio in the VFPI compared to
the FPS clearly demonstrates the continued effectiveness of isolation even for very high intensity of excitation. The recoverable
energy quantities show a greater reduction when using the VFPI
when compared to the FPS, showing that the energy dissipation
characteristics of the VFPI are also more robust than that of the
FPS. It is also interesting to note that in the case of the VFPI, the
increase in the energy transmitted to the structure with increase in
intensity of excitation is lower than the corresponding increase in
the input energy. This implies that the response of the structure
isolated by the VFPI is not significantly affected by the intensity
of excitation.

Response of Isolated PrimarySecondary Systems

Many structures support substructures or secondary systems
whose safety and functional integrity during earthquakes is essential. Examples include sensitive equipment and piping in indus-

Table 4. Ratio of Maximum Energy for High-Intensity Excitation to Maximum Energy for Medium Intensity Excitation
Coefficient of friction

0.02

0.05

0.10

Isolator Type

VFPI

FPS

VFPI

FPS

VFPI

FPS

Input energy
Recoverable energy

2.57
1.27

4.25
6.27

2.65
2.58

3.00
5.57

3.34
2.47

3.61
3.38

JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 877

J. Struct. Eng., 2002, 128(7): 870-880

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

trial facilities, structures containing delicate instruments, and any

structure controlled by a vibration absorber or tuned mass damper
see, for example, Pranesh and Sinha 2000b. The secondary systems are typically characterized by the following properties: 1
very small mass compared to the structure mass, 2 tuning of
fundamental frequency to that of the structure, and 3 nonclassical damping due to different damping ratios of the structure and
secondary system Igusa and Der Kiureghian 1985. A low-mass
secondary structure responds to the acceleration of its supporting
floor similar to the response of the primary structure to ground
excitations. However, the floor accelerations differ from typical
ground motions in severity and characteristics. As a result, the
response characteristics of secondary systems are strongly influenced by their dynamic properties relative to those of the primary
structure, such as the frequency ratio ratio of frequency of the
secondary system to the fundamental frequency of the primary
system, mass ratio ratio of mass of the secondary system to the
mass of the primary system, and damping in the secondary system Igusa and Der Kiureghian 1985. The secondary systems
exhibit highly amplified responses when the frequency ratio is
nearly equal to unity. The isolation of the primary system not only
changes the dominant frequency of excitation at the base of the
secondary system, but also reduces the amplitude of the excitation. The formulations derived earlier are also applicable for
analysis of primarysecondary systems and can be directly used.
In the present investigations the response of light equipment
mounted on top of the isolated five-story shear structure considered earlier has been investigated. The secondary system has a
mass (m eq) equal to 1% of the total mass of the structure including the base mass and is attached to the top story. The structural
damping and the damping in the equipment are both assumed to
be equal to 5% of the critical damping to eliminate the influence
of nonclassical damping from this evaluation. While the effect of
nonclassical damping is very important in determining the behavior of the secondary system, its influence is relatively similar for
both isolated and fixed-base structures. Therefore, noninclusion of
the nonclassical damping parameter does not limit the scope of
this investigation. The equipment has been modeled by a mass
with a linear spring and a damper attached to the top floor of the
structure.

Time-History Response
The typical time-history plot of absolute acceleration and displacement of the equipment relative to the floor are given in Figs.
7a and b. The equipment is chosen with a frequency of 3.85 Hz,
which is tuned to the second natural frequency of the isolated
structure and represents the most severe case of tuning. It is observed from the time-history plots that there is considerable reduction in the peak response of the equipment in comparison with
both the equipment on the fixed-base structure and the structure
isolated by the FPS. It is to be noted that at around t5.5 s, the
response of the equipment on the structure isolated by the PF
system is greater than that on the structure isolated by the VFPI
and FPS isolators. The typical time history for recoverable energy
of equipment kinetic energystrain energy is shown in Fig.
7c. From Fig. 7c it is observed that the maximum recoverable
energy in the equipment is greatly reduced in the structure isolated by the VFPI in comparison to that isolated by the FPS.
However, as expected, it is higher than that of a PF system. As the
absolute kinetic energy has been considered, the effect of rigidbody motions is implicitly included. This is manifested through

Fig. 7. Typical response and energy time histories of light equipment

on the isolated structure subjected to El Centro 1940 NS ground
motion: a equipment acceleration, b equipment displacement,
and c recoverable energy 0.02, m eq0.01m t , eq0.05, frequency variation factor10 per m

the various peaks in the energy time history for the FPS isolated
structure. These peaks are drastically reduced in the case of VFPI
and PF systems.

Floor Response Spectra of Equipment

The maximum response of equipment can be conveniently studied in terms of its floor response spectra. The floor response spectra enable one to evaluate the effectiveness of different isolation
systems for secondary systems with various properties.
The floor response spectra for displacement and acceleration
are shown in Fig. 8. The displacement spectra are normalized
with respect to the peak displacement of equipment mounted on
the fixed-base structure equal to 0.28 m. The effects of the
structureequipment interaction, including the effect of equipment on structure, are fully considered in the analysis. In the case
of a fixed-base structure the equipment acceleration response will
be maximum when the equipment frequency tunes with the fundamental frequency of the structure approximately 1.85 Hz.
However, for base-isolated structures, the first natural frequency
is the isolator frequency, which is much lower than that of both
the fixed-base structure and the equipment. So, the possibility of
equipment frequency tuning with the isolator frequency is very
unlikely and has not been included. The equipment response
shows a peak when its frequency is close to the second frequency
of the isolated structure. This can be clearly observed from the
acceleration spectra Fig. 8b, wherein the peak in the FPS occurs at 3.85 Hz, which is close to the second isolated frequency
see Table 1. It is to be noted that amplification of the response

878 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng., 2002, 128(7): 870-880

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Floor response spectra for light equipment on example structure subjected to El Centro 1940 NS ground motion: a equipment
displacement spectra and b equipment acceleration spectra 0.02, m eq0.01 m t , eq0.05, FVF10 per m

due to tuning is almost nonexistent in the case of the VFPI. This

is due to the variation in oscillation frequency of the VFPI isolated structure. It is further observed that the response of the
equipment mounted on the structure with the VFPI performs better for the entire range of equipment frequencies. The VFPI is
effective even for flexible equipment, whereas a conventional
FPS shows a higher response. This shows that the performance of
secondary system or equipment is relatively independent of the
frequency content and amplitude of excitation when the structure
is isolated by the VFPI. The use of the VFPI for structure isolation, therefore, also acts as an effective device for passive vibration control of secondary systems.
The maximum equipment acceleration when its natural frequency is tuned to the natural frequencies of the isolated structure
is shown in Table 5. It is found that for all equipment frequencies,
the equipment experiences a greater response in the structure isolated by the FPS than in the structure isolated by the VFPI. This
is due to the constant fundamental period of the structure isolated
by the FPS wherein the structure behaves like a narrow banded
filter. On the other hand, the fundamental period of a structure
isolated by the VFPI continuously changes, causing it to behave
as a wideband filter.

Discussions and Conclusions

The effectiveness of a recently developed isolation system, the
variable frequency pendulum isolator, for vibration control of
MDOF systems has been investigated in this paper. Mathematical
formulations involving complex modal analysis have been proposed for analysis of the structure isolated by the VFPI subjected
to ground motions. Closed-form expressions have also been developed using the mode-synthesis approach. A five-story shear
structure has been analyzed for different intensities of El Centro
1940 NS ground motion. The response of the secondary system
mounted on the example structure has also been considered. The
results are compared with that of the fixed-base structure and the

Table 5. Maximum Acceleration of Equipment with Frequency

Tuned to Natural Frequencies of Isolated Structure

Equipment frequency Hz

3.73

7.14

10.0

12.5

Structure isolated by FPS g

Structure isolated by VFPI g

0.635
0.440

0.510
0.353

0.368
0.328

0.306
0.262

structure isolated by FPS and PF systems in order to evaluate the

relative effectiveness of these isolation systems.
From these investigations it is found that the VFPI is very
effective in reducing the response of structures and secondary
systems when compared to FPS and PF isolators. The VFPI reduces the response substantially without losing the restoring capability, thereby combining the advantages of both FPS and PF
isolators. It is also found that the response of a MDOF structure
isolated by the VFPI is dominated by the isolator and fundamental
modes of the structure.
Based on the investigations of the response of MDOF systems
isolated by the VFPI, the following conclusions can be drawn:
1. The VFPI uses the sliding surface derived from the elliptical
curve and provides a displacement-dependent time period as
well as force-softening restoring force characteristics.
2. Mathematical formulation for analysis of MDOF structure
isolated by the VFPI involving complex modal analysis has
been presented. The mode synthesis approach has been used
to simplify the analysis and the final expressions are similar
to that of a fixed-base structure.
3. The VFPI is very effective in reducing the response of both
structures and primarysecondary systems under a variety of
excitations. The performance of the VFPI is found to be
robust and superior to that of FPS and PF isolators.
4. The VFPI acts as an isolator combined with the effective
energy dissipation and restoring mechanism for all of the
coefficients of friction. The VFPI with a low coefficient of
friction (0.02) is very effective in reducing the response
of the structure.
5. The response of a structure at the isolator level can be estimated with reasonable accuracy by considering the first two
modes of the isolated structure. A greater number of modes
is required for estimating structural response quantities.
6. The behavior of the structure and primarysecondary system
isolated by the VFPI is relatively independent of the frequency content and amplitude of excitation.
7. The effect of the higher mode may be significant for estimating structural acceleration response quantities, and hence
contribution of higher modes must be considered for the
analysis of primarysecondary systems.

References
Buckle, I. G., and Mayes, R. L. 1990. Seismic isolation: History, application, and performanceA world view. Earthquake Spectra,
EERI, 62, 161201.
JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 879

J. Struct. Eng., 2002, 128(7): 870-880

Downloaded from ascelibrary.org by MIGUEL BENDEZU ROMERO on 11/15/16. Copyright ASCE. For personal use only; all rights reserved.

Chalhoub, M. S., and Kelly, J. M. 1990. Sliders and tension controlled

reinforced bearings combined for earthquake isolation system. J.
Earthquake Eng. Struct. Dynam., 193, 333344.
Clough, R. W., and Penziene, J. 1993. Dynamics of structures, 2nd Ed.,
McGraw-Hill, New York.
Igusa, T., and Der Kiureghian, A. 1985. Dynamic characterization of
two-degree-of-freedom equipment-structure systems. J. Eng. Mech.,
1111, 119.
Kelly, J. M. 1986. Aseismic base isolation: Review and bibliography.
J. Soil Dynam. Earthquake Eng., 54, 202216.
Kelly, J. M. 1988. Base isolation in Japan, 1988. Rep. No. UCB/
EERC-38/20, Earthquake Engineering Research Centre, Univ. of
California, Berkeley, Calif.
Kelly, J. M. 1993. State of the art and state of the practice in base
isolation. Seminar on Seismic Isolation, Passive Energy Dissipation
and Active Control (ATC-17-1), Applied Technology Council, Redwood City, Calif.
Kemeny, Z. A. 1997. Ball-in-cone seismic isolation bearing. U.S.
Patent No. 5599106, U.S. Patent and Trademark Office, Washington,
D.C.
Mokha, A., Constantinou, M. C., and Reinhorn, A. M. 1990. Teflon
bearings in base isolation I: Testing. J. Struct. Eng., 1162, 438
454.
Mokha, A., Constantinou, M. C., Reinhorn, A. M., and Zayas, V. A.
1991. Experimental study of friction-pendulum isolation system.
J. Struct. Eng., 1174, 12011217.
Mokha, A., Amin, N., Constantinou, M. C., and Zayas, V. A. 1996.
Seismic isolation retrofit of large historic building. J. Struct. Eng.,
1223, 298 308.
Mostaghel, N., Hejazi, M., and Tanbakuchi, J. 1983. Response of sliding structures to harmonic support motion. J. Earthquake Eng.
Struct. Dynam., 113, 355366.
Mostaghel, N., and Khodavedian, M. 1987. Dynamics of resilient friction base isolator R-FBI. J. Earthquake Eng. Struct. Dynam., 153,
379390.
Mostaghel, N., and Tanbakuchi, J. 1983. Response of sliding structures to earthquake support motion. J. Earthquake Eng. Struct.
Dynam., 116, 729748.
Naeim, F., and Kelly, J. M. 1999. Design of seismic isolated structures:
From theory to practice, Wiley, New York.
Pranesh, M. 2000, VFPI: An innovative device for aseismic design.
PhD thesis, Indian Institute of Technology, Bombay, India.

Pranesh, M., and Sinha, R. 1998. Vibration control of primary

secondary systems using variable frequency pendulum isolator.
Proc., 11th Symposium on Earthquake Engineering, Univ. of Roorkee,
Roorkee, India, 697704.
Pranesh, M., and Sinha, R. 2000a. VFPI: An isolation device for aseismic design. J. of Earthquake Eng. Struct. Dynam., 295, 603 627.
Pranesh, M. and Sinha, R. 2000b. Aseismic design of tall structures
using variable frequency pendulum isolator. Proc., 12th World Conf.
Earthquake Engineering, Paper No. 284, Auckland, New Zealand.
Shustov, V. 1992. Base isolation: Fresh insight. Proc., 10th World
Conf. on Earthquake Engineering, Balkema, Rotterdam, 4, 1983
1986.
Singh, M. P., and Suarez, L. E. 1992. Dynamic condensation with
synthesis of substructure eigenproperties. J. Sound Vib., 1591,
139155.
Sinha, R. 1993. Nonclassically damped structures. Lecture Notes of
One Week Interaction ProgrammeState of the Art on Seismic Qualification of Structures and Systems, Bhabha Atomic Research Centre,
Mumbai, India, 71 80.
Sinha, R., and Pranesh, M. 1998. FPS isolator for structural vibration
control. Proc., Int. Conf. on Theoretical, Applied, Computational and
Experimental Mechanics (CD Volume), IIT, Kharagpur, India.
Tsai, C. S. 1997. Finite-element formulations for friction pendulum
seismic isolation bearings. Int. J. Numer. Methods Eng., 401, 29
49.
Tsopelas, P., Constantinou, M. C., Kim, Y. S., and Okamoto, S. 1996.
Experimental study of FPS system in bridge seismic isolation. J.
Earthquake Eng. Struct. Dynam., 251, 6578.
Uang, C. M., and Bertero, V. V. 1990. Evaluation of seismic energy in
structures. J. Earthquake Eng. Struct. Dynami., 191, 7790.
Wang, Y. P., Chung, L. L., and Liao, W. H. 1998. Seismic response
analysis of bridges isolated with friction pendulum bearings. J.
Earthquake Eng. Struct. Dynam., 2710, 10691093.
Zayas, V. A., Low, S. S., and Mahin, S. A. 1987. The FPS earthquake
resisting system: Experimental report. Rep. No. UCB/EERC-87/01,
Earthquake Engineering Research Centre, Univ. of California,
Berkeley, Calif.
Zayas, V. A., Low, S. S., and Mahin, S. A. 1990. A simple pendulum
technique for achieving seismic isolation. Earthquake Spectra,
EERI, 62, 317333.
Zhou, Q., Lu, X., Wang, Q., Feng, D., and Yao, Q. 1998. Dynamic
analysis on structures base isolated by a ball system with restoring
property. J. Earthquake Eng. Struct. Dynam., 278, 773791.

880 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng., 2002, 128(7): 870-880