EXPERIMENTAL STUDY OF ISOLATED BUILDING

GROUND EXCITATIONS

UNDER

TRIAXIAL

Downloaded from ascelibrary.org by Universidad Catolica De La Santisima Concepcion on 05/23/16. Copyright ASCE. For personal use only; all rights reserved.

By Jenn-Shin Hwang,1 Member ASCE, and Ting-Yu Hsu2

ABSTRACT: Unilateral, bilateral, and triaxial shaking table tests are conducted to study the seismic response
of a three-story base-isolated steel structure. The lateral force distribution formulas specified in the Uniform
Building Code are evaluated experimentally. A method to calculate the lateral force distribution is proposed and
experimentally verified. Seismic responses of the test model to the unilateral, bilateral, and triaxial ground
excitations are compared. The effects of vertical ground acceleration on the maximum acceleration of the superstructure and the maximum displacement of the isolation system are investigated. The maximum displacements of the isolation system under bilateral and triaxial ground motions are compared with those determined
from the 1.0 0.3 direction combination rule. In addition to the plan-symmetric condition, both one-way
and two-way eccentricities are introduced into the isolation system. The increase of the peak story acceleration
may be significant, compared with those obtained from unilateral and bilateral ground shakings. The experimental
results from the plan-eccentric condition also are compared with the conclusions of other analytical studies.

INTRODUCTION
In the past two decades, base-isolated structures have been
constructed to passively control seismic response. Extensive
studies have been conducted in laboratories to verify and enhance the design and analysis theories. Numerous shaking table studies for the seismic responses of base-isolated building
structures have been performed in the past. Kelly et al. (1980)
conducted shaking table tests of a base-isolated building structure with a fail-safe system on which the superstructure is carried when the bearings are displaced beyond their stability
limit. Kelly et al. (1986) tested a base-isolated rigid block by
lead-rubber bearings. The bearings were shaken up to the rollover displacement. Griffith et al. (1988) evaluated the effect
of column uplift on the seismic response of a base-isolated
building. Al-Hussaini et al. (1993) performed a series of tests
on a base-isolated building with the friction pendulum system.
Hwang and Ku (1997) correlated the experimental results of
a rigid block isolated by high damping rubber bearings with
those predicted by an analytical model. However, these shaking table studies are all limited to unilateral or unilateral plus
vertical ground excitations.
According to the processed displacement records at the base
floor of the USC Hospital during the 1994 Northridge earthquake (CSMIP 1994), the torsional response may be significant
and should be carefully considered for the design of isolation
bearings, utilities, and gaps. However, all of the base-isolated
models tested in the aforementioned shaking table studies are
plan-symmetric. Limited experimental evidence (Zayas et al.
1987; Nakamura et al. 1988) has been provided to substantiate
the effects of torsional coupling on base-isolated structures.
Nevertheless, extensive analytical studies have investigated the
effect of stiffness or strength eccentricity of base-isolated
structures. Lee (1980) investigated the effects of the eccentricities in the superstructure and in the isolation system of a
base-isolated single-story structure. Kelly and Pan (1983) determined the effects of the two-way eccentricity and damping
ratio of the isolation system. Eisenberger and Rutenburg

(1986) studied the effects of different ground motions on a

one-way eccentric base-isolated building. Nagarajaiah et al.
(1993) included a few parameters in their study such as the
superstructure flexibility, the uncoupled torsional to lateral frequency ratio, the eccentricities in the superstructure and isolation system, and the higher modes of vibration.
In this study, a three-story steel structure isolated by leadrubber bearings and natural rubber bearings was tested using
a three-axis shaking table. Unilateral, bilateral, and triaxial
ground shakings were used to investigate the seismic responses of the test structure. Plan-symmetric and plan-eccentric conditions were included in the test program. For the eccentric case, one-way and two-way eccentricities were
introduced into the isolation system to study the effects of
torsional coupling on the seismic responses of the test structure.
TEST MODEL AND TEST PROGRAM
A three-story base-isolated steel structure as shown in Fig.
1 is assumed to be a 0.4-scale model. The superstructure was
composed of three moment-resisting frames in the longitudinal
(X) direction and two exterior steel dual systems together with
a central moment-resisting frame were designed in the transverse (Y) direction. The smallest locally available rolled sections were selected for the construction of the test structure.

Prof., Dept. of Constr. Engrg., Nat. Taiwan Univ. of Sci. and Technol.,
P.O. Box 90-130, Taipei, Taiwan; also Div. Head, Nat. Ctr. for Res. on
Earthquake Engrg., Taiwan. E-mail: JSH@mail.ntust.edu.tw
2
Grad. Res. Asst., Dept. of Constr. Engrg., Nat. Taiwan Univ. of Sci.
and Technol., P.O. Box 90-130, Taipei, Taiwan.
Note. Associate Editor: Brad Cross. Discussion open until January 1,
2001. To extend the closing date one month, a written request must be
filed with the ASCE Manager of Journals. The manuscript for this paper
was submitted for review and possible publication on April 28, 1999.
This paper is part of the Journal of Structural Engineering, Vol. 126,
No. 8, August, 2000. ASCE, ISSN 0733-9445/00/0008-08790886/
$8.00 $.50 per page. Paper No. 20769.

FIG. 1.

Typical Plan and Elevation of Test Structure

JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 2000 / 879

J. Struct. Eng., 2000, 126(8): 879-886

Downloaded from ascelibrary.org by Universidad Catolica De La Santisima Concepcion on 05/23/16. Copyright ASCE. For personal use only; all rights reserved.

The superstructure was originally designed as a fixed-base

structure according to the seismic design specifications of
building structures in Taiwan (Department of Interior 1995;
Wen 1995), and the reduction of force demand due to the
incorporation of the base isolation system was not considered.
Lead bricks weighing 500 N each were attached to the floors
of the test model to simulate the seismic reactive mass. The
weights from the base floor to the roof were approximately
equal to 113, 92, 92, and 70 kN, respectively.
The isolation system was composed of eight circular leadrubber bearings and one natural rubber bearing for the phase
I tests, whereas six lead-rubber bearings and three natural
rubber bearings were used for the phase II tests. The diameter
of the circular lead-rubber bearing was 120 mm. The bearing
was composed of 16 rubber layers and 15 steel shim plates.
The thickness of each rubber layer was 3.5 mm, and the thickness of the steel plates was 1 mm. The diameter of the lead
plug was 25 mm. The dimensions of the natural rubber bearings were exactly the same as those of the lead-rubber bearings
without a lead-core. The isolation bearings were arranged to
yield a symmetric isolation system for phase I tests and eccentric isolation systems for phase II tests. The layouts of the
isolation system for the phase I and phase II tests are illustrated in Fig. 2, where ex and ey are the eccentricities in the
X- and Y-directions and are calculated based on the elastic
stiffness of the bearings, and Lx and Ly are the plan dimensions
of the test structure in the longitudinal (X) and transverse (Y)
directions. It should be noted that the isolation system layout
given in Fig. 2 for the phase II tests is not intended to duplicate
any possible practical layout of a base-isolated structure but
to introduce the stiffness and/or strength eccentricity into the
isolation system so that the effect of torsional coupling of the
isolation system can be investigated.
Based on the average axial load exerted on the isolation
system by the dead load of the superstructure, an axial load
of 40 kN is imposed to each bearing for the performance test
conducted in a cyclic manner. Very similar hysteresis characteristics have been observed for the entire set of lead-rubber
bearings. Typical test results are shown in Fig. 3. Additionally,
some other tests also are conducted with a varying axial load.
The test results have indicated that the characteristic strength

FIG. 2.
Tests

Layout of Isolation Bearings for Phase I and Phase II

FIG. 3. Typical Hysteresis Loops of Lead-Rubber Bearings

from Performance Test

TABLE 1.
Earthquake
names
(1)
1940 El Centro

Earthquakes Used for Tests

Station names
(2)
Irrigation District

Component
identifications
(3)

N-S(X ), E-W(Y ),
up(Z )
1989 Loma Prieta Capitola, Fire Station 360(X ), 90(Y ),
and up(Z )
1994 Kobe
JMA Station
N-S(X ), E-W(Y ),
up(Z )
1989 Loma Prieta Corralitos, Eureka
360(X ), 90(Y ),
Canyon Road
and up(Z )
1994 Northridge New Hall
360(X ), 90(Y ),
and up(Z )
1952 Taft
Lincoln School Tun- S69E(X ),
nel
N21E(Y ), and
up(Z )

Intensity
scale
(%)
(4)
100
100
50
120
60
200

Qd, effective stiffness Keff, and energy dissipation capacity

EDC (Rewrite 1996) of these lead-rubber bearings are dependent on the axial load. This result is consistent with what was
reported by Tyler and Robinson (1984).
The test program is composed of two phases. Six recorded
earthquakes summarized in Table 1 are used to investigate the
3D seismic responses of the test structure in the phase I tests.
These recorded ground motions are selected based on the reasons that some of the records have a broad frequency content
of strong shaking and some of the records have very high
vertical ground acceleration compared with the two horizontal
components. The superstructure and the isolation system are
so arranged to yield a symmetric-plan system. Four tests are
conducted for each recorded earthquake. The two horizontal
acceleration components of each earthquake are input to the
shaking table separately in the first two tests, followed by bilateral and triaxial excitations. In the phase II tests, six leadrubber and three natural rubber bearings are arranged to yield
one-way and two-way eccentricities in the isolation system.
Only two earthquakes, the 1940 El Centro and the 1989 Loma
Prieta recorded at Capitola, are used in the phase II tests. The
original time step of 0.02 of all ground motions is scaled to
0.0125 according to the scaling factor of the test model.
Displacement transducers and accelerometers are installed
at the corner and center columns of each story of frame A and
frame 3 to measure the seismic responses of the test structure.
Nine axial-shear load cells are mounted beneath the isolation
bearings to measure the shear forces and to monitor the axial
force variation of the bearings.

880 / JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 2000

J. Struct. Eng., 2000, 126(8): 879-886

TEST RESULTS
Phase I Tests (Symmetric Superstructure with
Symmetric Isolation System)

Downloaded from ascelibrary.org by Universidad Catolica De La Santisima Concepcion on 05/23/16. Copyright ASCE. For personal use only; all rights reserved.

Unilateral Tests
The fixed-base and base-isolated conditions of the model
are tested using random white noises to determine their natural
frequencies. To prevent yielding of the lead-rubber bearings,
a small peak ground acceleration of 0.03g is selected for the
white noise ground motion. For the base-isolated condition,
the first mode natural frequencies determined from the transfer
functions are equal to 2.16, 4.96, and 12.5 Hz, respectively,
in the longitudinal (X), transverse (Y), and vertical (Z) directions. For the fixed-base model, the first mode natural frequencies are equal to 2.97 and 4.96 Hz corresponding to the
longitudinal and transverse directions.
The results from unilateral earthquake tests indicate that the
hysteresis loops of the lead-rubber bearings do not produce
the same forces. A typical example obtained from the longitudinal test with the Capitola record is shown in Fig. 4, which
shows that the EDC, Keff, and Qd of the hysteresis loops are
quite different. This result is not the same as that observed
from the cyclic loading tests in which all lead-rubber bearings
reveal very similar hysteresis characteristics as shown in Fig.
3. The reason for this difference may be because the static
axial loads exerted on the bearings under the exterior center
columns (e.g., LRB2) are larger than those on the bearings
under the corner columns (e.g., LRB1). The average static axial loads exerted on the bearings under the center columns of
frames A and C, the center columns of frames 1 and 3, and
the corner columns are measured to be equal to 49, 50, and
25 kN, respectively. Because the shear force-displacement relationship of the bearings is dependent on the axial load (Tyler

and Robinson 1984), the energy dissipation of the bearings

under the center columns is larger than that under the corner
columns. However, the dependency of the hysteresis loops on
the axial load is often not considered in the practical design
in which the hysteresis loops of lead-rubber bearings having
the same lead core diameter and bearing dimension are usually
presumed to be the same. The Uniform Building Code (UBC)
[International Building Code Officials (IBCO) 1997] has required design engineers to consider the effect of axial load
when modeling the behavior of isolation bearings for a dynamic time history analysis. However, difficulties arise because lead-rubber bearings from different manufacturing processes or manufacturers may reveal different axial load
dependency. From this investigation, it is concluded that the
effect of the axial load on isolation bearings installed at different locations of a base-isolated structure should be considered for practical design.
In addition to the aforementioned, the hysteresis loop of
each bearing shown in Fig. 4 is not skew-symmetric as was
observed from the cyclic loading test result shown in Fig. 3,
because the overturning moment of the superstructure that undergoes a ground excitation has introduced an axial load variation into the bearings. For example, discrepancy is observed
between the traces a-b-c and c-d-e of the hysteresis loops of
the bearing LRB 2 as shown in Fig. 4(b). Corresponding to
these two loop traces, the axial load on the bearing varies from
traces a-b-c to c-d-e as shown in Fig. 4(c). Because the
loading variation trace c-d-e shows an increase on the compressive axial load and the load loading variation trace a-bc depicts a decrease on the compressive axial load, the area
encased in the loop trace c-d-e is larger than that covered by
the loop trace a-b-c. The effect of axial load induced by the
overturning of the superstructure should also be considered in
the practical analysis.
The maximum relative story displacements of the test model
subjected to unilateral (longitudinal) ground excitations are
briefly summarized in Fig. 5. In addition, the normalized peak
story acceleration with respect to the maximum shaking table
acceleration is shown in Fig. 6. In the following, these maximum story responses including those measured at the base
floor will be used to correlate the experimental results with
the lateral force distribution formulas given in the current design specifications such as UBC (IBCO 1997).
The lateral force distributed over the height of the structure
above the isolation interface is specified by the UBC as
Fx = Vs

wxhx

(1)

wihi

i=1

where Vs = minimum force that should be considered for the

design of the superstructure above the isolation interface,
which is determined from dividing the maximum base shear

FIG. 4. Hysteresis Loops of Lead-Rubber Bearings from Longitudinal Test of Capitola Record and Axial Load Variation of
LRB 2: (a) Corner Bearing LRB 1; (b) Center Bearing LRB 2; (c)
Axial Load Variation Corresponding to Loop Traces from a-b-cd-e of LRB 2

FIG. 5. Summary of Maximum Story Displacement of Unilateral (Longitudinal) Tests: (a) Capitola; (b) New Hall
JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 2000 / 881

J. Struct. Eng., 2000, 126(8): 879-886

Downloaded from ascelibrary.org by Universidad Catolica De La Santisima Concepcion on 05/23/16. Copyright ASCE. For personal use only; all rights reserved.

on the superstructure is obtained by directly reducing the

inelastic base shear force transmitted by the isolation system Vb by an RI factor. The inertia force on the base floor
is not excluded.
To illustrate these aforementioned discussions, the measured
base shear force in the isolation system is substituted for Vs in
(1) (i.e., RI is set to 1), and the calculated acceleration distributions are compared with those measured from the tests in
Fig. 6. In the figure, it is clear that the predicted distributions
of the peak story acceleration are quite different from the measured distributions in shape and in magnitude.
In this study, a procedure to calculate the lateral force distribution on a base-isolated building structure will be proposed
in the following. The procedure does not follow the UBC formula in which the design base shear force of the superstructure
is calculated by reducing the maximum inelastic shear force
of the isolation system by a reduction factor. Instead, the lateral force distribution on the whole base-isolated structure including the base floor is calculated first based on the maximum
inelastic shear force transmitted by the isolation system. Then,
the lateral force exerted on the superstructure may be reduced
by an RI factor, if desired.
Thus, it is intended first to ascertain whether the distribution
of the maximum lateral story displacements can be used to
calculate the lateral force distribution as is usually done for an
elastic fixed-base structure. The following formula is proposed
to calculate the lateral force distribution and to correlate with
the experimental results:
Fx = Vb

wxx

(2)

wii

i=0

FIG. 6. Summary of Measured and Predicted Distributions of

Normalized Peak Story Acceleration: (a) El Centro; (b) Capitola;
(c) New Hall; (d) Taft

force transmitted by the isolation system Vb by a reduction

factor RI; wi and wx = weight located at story levels i and x,
respectively; and hi and hx = height of story levels i and x,
respectively, above the base floor. According to (1), two anomalies arise:
1. Eq. (1) was adopted based on an assumption that the first
mode shape of an elastic multistory fixed-base building
structure can be approximately proportioned by the
height of the story level above the base (Chopra 1995).
This assumption is recognized to be acceptable for most
multistory fixed-base building structures, because the design base shear force Vs is obtained by reducing the
elastic base shear force by a factor such as Rw of UBC.
However, for a base-isolated structure, the total lateral
force Vs is calculated based on a reduction on the inelastic design base shear force Vb, which is determined
from an equivalent linear analysis procedure (Hwang et
al. 1996). The equivalent linear analysis is adopted because the maximum inelastic displacement response of a
structure can be approximated by the maximum elastic
displacement response obtained from an equivalent linear
system. Therefore, the first-mode shape of a base-isolated
structure may not be able to be represented by the proportion of the story height as is usually done for a fixedbase building structure.
2. As can be seen in (1), where the index i is starting from
i = 1, the inertia force exerted on the base floor should
be excluded from the total base shear force Vb when calculating the design shear force of the superstructure Vs.
However, according to UBC, the design base shear force

where Vb = total base shear force transmitted by the isolation

system; i and x = relative displacements of each story including the base floor to the ground. For i = 0, w0 = weight
of the base floor, and 0 = relative lateral displacement of the
isolation system to the ground, which is the design displacement of the isolation system denoted as DD in the UBC. Using
(2) and the measured maximum story displacements, the lateral acceleration distribution of the test structure subjected to
various ground motions can then be calculated. In Fig. 6, it is
seen that the calculated maximum story accelerations compare
well with the measured peak story accelerations. Therefore, it
can be concluded that, even though the seismic response of a
base-isolated structure is inelastic, the maximum story displacement can be used as the mode shape to calculate the
lateral force distribution as is usually done for an elastic fixedbase structure. One should be cautious that the weight and the
lateral displacement of the base floor are involved in the above
calculation using (2). However, in the practical design, these
story displacements, except for the relative displacement (or
design displacement) of the isolation system, are not known
prior to the calculation of the lateral force distribution. To
facilitate the procedure for practical analysis, it is proposed
that the story displacements i and x can be calculated prior
to the determination of the lateral force distribution using the
total base shear force Vb and an assumed lateral force distribution given by
fx = Vb

wx

(3)

wi

i=0

in which fx = assumed lateral force distribution to calculate the

lateral displacement i and x. Using (3), the story displacement can then be calculated using a simple elastic analysis.
This is because the total base shear force and the maximum

882 / JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 2000

J. Struct. Eng., 2000, 126(8): 879-886

Downloaded from ascelibrary.org by Universidad Catolica De La Santisima Concepcion on 05/23/16. Copyright ASCE. For personal use only; all rights reserved.

inelastic displacement of the isolation system predicted by (3)

are exactly equal to Vb and DD so that the stiffness of the
isolation system can be represented by the effective stiffness
Keff. Based on (2) and (3), the correlation between the calculated results and the experimental results is shown in Fig. 6,
which suggests that the proposed procedure be adopted for the
calculation of the lateral force distribution of a base-isolated
building. Once the lateral forces on the superstructure are determined, they can be reduced by RI, if desired, to obtain the
design force of the superstructure.

summarized in Table 2 in which (Ubx)max = maximum relative

displacement of the isolation system measured from the unilateral tests in the longitudinal (X ) direction; (Uby)max = maximum relative displacement of the isolation system measured
from the unilateral tests in the transverse (Y ) direction; and
(Ub)max = maximum relative displacement of the isolation system in any horizontal direction measured from the bilateral or
triaxial tests. It is interesting to note in Table 2 that the maximum displacement responses of the isolation system measured from the bilateral and triaxial tests are not necessarily
smaller than those determined from the 1.0 0.3 direction
combination rule. For example, the maximum displacements
measured from the triaxial tests of the Capitola and Taft
records are about 15% larger than those obtained from the
1.0 0.3 combination rule. Therefore, the design displacement of the isolation system determined using the direction
combination rule may not be conservative. In addition, it also
is observed in Table 2 that the maximum displacements of the
isolation system measured from the triaxial tests are all slightly
larger than those from the bilateral tests. In particular, the triaxial test with the Capitola record has generated a maximum
displacement at the isolation bearings 10% larger than that of
the corresponding bilateral test. Fig. 8 shows the comparison
between the displacement traces of the isolation system measured from the bilateral and triaxial tests of the Capitola record. This comparison shows that the effect of vertical ground
acceleration may have to be carefully considered when determining the maximum displacements of the isolation system.
To further investigate the effect of vertical ground excitation, the measured hysteresis loops of the bearings LRB 1 and
LRB 2 from the bilateral and triaxial tests of the Capitola
record are compared in Fig. 9. As can be seen in the figure,
the hysteresis loops of the bearings LRB 1 and LRB 2 in the
longitudinal direction under the triaxial excitation show a very
significant effect of the vertical ground acceleration, compared
with those under the bilateral excitation. The irregularity of
the hysteresis loops may be caused by the axial load variation
on the bearings induced by the vertical ground excitation. An
example shown in Fig. 10 indicates the variation of axial loads
on the bearings LRB 1 (under corner columns) and LRB 2
(under center column) measured from the triaxial test is significantly larger than those measured from the bilateral test.
Higher frequency responses have resulted in a large increase
on the axial load variation of the bearings. The maximum negative (or tensile) axial load variations attributed to the combined effects of the overturning moment and the vertical
ground excitation are about 43 and 88 kN, respectively,
on bearings LRB 1 and LRB 2. These maximum vertical load
variations are comparably larger than the average static axial
loads of 25 and 49 kN on the bearings. Therefore, uplift may
have occurred in these bearings. This again demonstrates the
need to consider the effect of vertical load variation for the
analysis and design of the bearings.

Bilateral and Triaxial Tests

Among the bilateral and triaxial test results, the peak story
acceleration and the maximum displacement of the isolation
system are of particular interest, because the peak story acceleration is related to the inertia force exerted on the superstructure and the maximum displacement of the isolation system is important for the design of the isolation bearings,
utilities, and gaps. The normalized peak story accelerations
measured in the longitudinal direction are summarized in Fig.
7, in which the difference among the unilateral, bilateral, and
triaxial test results is insignificant because, under bilateral and
triaxial excitations, the isolation bearings may yield in any
lateral direction once the total shear force reaches the yield
surface of the bearings. When the lead-rubber bearings yield,
the isolation effect commences and the transmission of the
ground acceleration into the superstructure is limited. Thus,
the maximum longitudinal or transverse acceleration component is not necessarily larger than that of the unilateral tests.
On the other hand, the maximum displacements of the isolation system determined from the unilateral, bilateral, and triaxial tests and the 1.0 0.3 direction combination rule are

FIG. 7. Longitudinal Normalized Peak Story Accelerations: (a)

Capitola Tests; (b) New Hall Tests
TABLE 2.

Maximum Dislacement Responses to Unilateral, Bilateral, and Triaxial Excitations of Phase I Tests
Bilateral

Unilateral
Earthquake (Ubx)max
(mm)
records
(2)
(1)
Capitola
Corralitos
El Centro
Kobe
New Hall
Taft

30.77
38.74
17.80
29.64
49.07
26.98

Triaxial

Occurring
time
(s)
(3)

(Uby)max
(mm)
(4)

Occurring
time
(s)
(5)

(Ub)max
(mm)
(6)

Occurring
time
(s)
(7)

(Ub)max
(mm)
(8)

5.95
4.96
1.30
5.54
3.46
2.54

19.57
45.88
13.47
24.19
27.02
22.07

5.89
4.96
7.73
5.23
4.93
2.84

38.34
54.28
21.79
33.32
56.30
37.44

5.94
4.66
1.29
5.26
3.45
2.53

42.07
55.81
21.85
33.54
59.70
38.29

1.0 0.3 Rule

Occurring
time
(Ubx)max 0.3(Uby)max 0.3(Ubx)max (Uby)max
(s)
(mm)
(mm)
(9)
(10)
(11)
5.96
4.68
1.29
5.54
3.46
2.54

36.64
52.50
21.84
36.90
57.18
33.60

28.80
57.50
18.81
33.08
41.74
30.16

JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 2000 / 883

J. Struct. Eng., 2000, 126(8): 879-886

Downloaded from ascelibrary.org by Universidad Catolica De La Santisima Concepcion on 05/23/16. Copyright ASCE. For personal use only; all rights reserved.

Phase II Tests (Symmetric Superstructure with

Eccentric Isolation System)

FIG. 8. Displacement Traces of the Isolation System of Capitola Tests: (a) Bilateral Test: (b) Triaxial Test

FIG. 9. Longitudinal Hysteresis Loops of Isolation Bearings of

Capitola Tests: (a) LRB 1Bilateral Test; (b) LRB 1Triaxial
Test; (c) LRB 2Bilateral Test; (b) LRB 2Triaxial Test

FIG. 10. Axial Load Variation of Isolation Bearings of Capitola

Tests: (a) LRB 1Bilateral Test; (b) LRB 1Triaxial Test; (c)
LRB 2Bilateral Test; (b) LRB 2Triaxial Test

It has been reported by Lee (1980) that, if the eccentricity

of the isolation system is zero or nearly so, the rotational displacement of the base floor remains small even if the eccentricity of the superstructure is large. Therefore, in the phase II
tests, only the eccentricity of the isolation system will be designed, and the superstructure will remain symmetric. The arrangement of the isolation system for the phase II tests is summarized in Fig. 2 and is divided into four cases: (1) ex /Lx = 0,
ey /Ly = 0; (2) ex /Lx = 0, ey /Ly = 0.08; (3) ex /Lx = 0, ey /Ly =
0.16; and (4) ex /Lx = 0.16, ey /Ly = 0.16. Corresponding to these
eccentricities, the four cases are respectively referred to as the
symmetric case, moderately eccentric case, largely eccentric
case, and two-way eccentric case.
Because the story acceleration measured at the center of
mass is related to the inertia force exerted on the test structure,
the effect of eccentricity of the isolation system on the peak
story acceleration will be investigated first. The normalized
peak story accelerations measured at the center of mass in the
longitudinal (X ) direction are summarized in Fig. 11 in which
it can be seen that all bilateral tests have produced almost the
same peak story accelerations as the unilateral (longitudinal)
tests. This result is the same as that observed in the phase I
tests. However, from the results of the triaxial tests shown in
Fig. 11, it also is observed that the vertical ground excitation
of the Capitola record has induced a dramatic increase in the
peak story acceleration. This is different from that seen in the
symmetric system of the phase I tests. The peak story accelerations on the superstructure have been increased as much as
100%. The measured acceleration time histories at the first
story obtained from the bilateral and triaxial tests for case 3
are compared in Fig. 12 in which it can be seen that the higher
mode response becomes much more significant in the triaxial
test than in the bilateral test. However, Fig. 13 shows that the
acceleration time histories measured at the base floor are not
affected by the vertical ground acceleration as much as those
of the superstructure, because the response of the base floor is
primarily attributed to the lower mode (isolated mode) vibration rather than the higher mode (nonisolated mode) vibration.
Based on the aforementioned, it can be concluded that the
vertical ground acceleration may induce a significant increase
on the maximum story acceleration for the cases with eccentricity in the isolation system, whereas the acceleration at the
isolation interface is less affected.

FIG. 11. Longitudinal Normalized Peak Story Accelerations

Measured at Center of Mass: (a) Case 3Capitola Tests; (b)
Case 4Capitola Tests

884 / JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 2000

J. Struct. Eng., 2000, 126(8): 879-886

Downloaded from ascelibrary.org by Universidad Catolica De La Santisima Concepcion on 05/23/16. Copyright ASCE. For personal use only; all rights reserved.

The maximum longitudinal (X ) displacements including

those induced by the translational and rotational responses of
the isolation system are summarized in Table 3, in which
(Ubx)max = maximum relative displacement of the isolation system measured at the center of mass; (Ucx)max = maximum rotational displacement of the isolation system at the corner of
the base floor, which is calculated by multiplying the rotation
angle by the Y-directional moment arm measured from the
center of rigidity (elastic center) to the corner; (Ucx)max = maximum relative displacement recorded at the corner of the base
floor; and Ucx = rotational displacement component contributing to (Ucx)max at the time (Ucx)max occurs and is calculated
by multiplying the rotation angle by the Y-directional moment arm. From examining Table 3, a few discussions can be
drawn as follows:
Comparing cases 13 of the unilateral (X) tests, it is concluded in most other studies that an increasing eccentricity
ey /Ly induces an increasing rotational displacement response.
Correlating cases 1 and 2 of the unilateral (X ) tests, it is
observed that torsional coupling reduces the maximum
displacement responses (Ubx)max at the center of mass for
a moderate eccentricity of ey /Ly = 0.08. This confirms the
conclusion of Pan and Kelly (1983). However, for a larger
eccentricity case such as case 3 (ey /Ly = 0.16), the displacement response (Ubx)max at the center of mass is increased compared with case 1.
Contradicting the conclusion of Pan and Kelly (1983) that
for two-way torsional coupling the eccentricity in the direction of ground excitation reduces the peak rotational
response, it is found in this study that under a unilateral
excitation the maximum rotational displacement (Ucx)max
of the two-way eccentricity case (case 4) is much larger
than that of the one-way eccentricity case (case 3). The
maximum rotational displacements (Ucx)max of case 4 are
61 and 86% larger than those of case 3 corresponding to
the longitudinal tests of the Capitola and El Centro
records. In addition, comparing the results of unilateral
tests of cases 3 and 4, the maximum displacement responses measured at the center of mass (Ubx)max of the
two-way eccentricity case (case 4) are about 10% larger

FIG. 12. Longitudinal Acceleration Time Histories Measured

at Center of Mass of First Story: (a) Bilateral Capitola Test; (b)
Triaxial Capitola Test

FIG. 13. Longitudinal Acceleration Time Histories Measured

at Center of Mass of Base Floor: (a) Bilateral Capitola Test; (b)
Triaxial Capitola Test
TABLE 3.
Isolation
system
layout
(1)

Earthquake
inputs
(2)

Maximum Longitudinal Displacement Responses of Isolation System of Phase II Tests

(Ubx)max
(mm)
(3)

Occurring
time
(s)
(4)

(Uc x)max
(mm)
(5)

Occurring
time
(s)
(6)

(Ucx)max
(mm)
(7)

Occurring
time
(s)
(8)

(Uc x)
(mm)
(9)

5.34
6.48
6.61
6.61
6.51
6.54
6.60
6.09
6.53
6.13

43.83
38.39
40.10
42.54
50.99
50.18
53.08
61.66
61.17
62.11

6.06
6.50
6.08
6.06
6.53
6.09
6.11
6.09
6.10
6.13

1.41
2.77
1.54
1.36
9.89
8.84
10.72
16.30
18.37
19.71

3.80
3.48
8.15
2.41
3.81
2.93
2.93
3.84
3.86
3.86

29.42
26.15
27.00
27.03
32.72
32.05
31.96
39.00
41.44
41.86

3.78
3.75
3.78
3.78
3.79
3.81
3.81
3.81
3.83
3.83

0.85
0.28
0.42
0.35
4.67
3.66
3.45
8.92
12.66
12.46

(a) Capitola Earthquake

Case 1
Case 2
Case 3
Case 4

X
X
X
X
X
X
X
X
X
X

Y
YZ
Y
YZ
Y
YZ

42.94
36.64
38.77
41.97
43.56
43.48
46.00
49.31
47.48
48.22

6.51
6.05
6.08
6.09
6.53
6.09
6.10
6.09
5.66
5.68

28.57
25.91
26.64
26.73
29.18
29.98
30.05
32.24
32.38
33.04

3.78
3.75
3.78
3.78
3.79
3.80
3.80
3.81
3.81
3.81

1.51
2.91
3.74
4.22
10.12
10.67
12.08
16.30
18.65
19.71
(b) El Centro Earthquake

Case 1
Case 2
Case 3
Case 4

X
X
X
X
X
X
X
X
X
X

Y
YZ
Y
YZ
Y
YZ

1.34
0.86
2.56
2.75
4.95
4.58
4.70
9.21
12.80
12.79

JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 2000 / 885

J. Struct. Eng., 2000, 126(8): 879-886

Downloaded from ascelibrary.org by Universidad Catolica De La Santisima Concepcion on 05/23/16. Copyright ASCE. For personal use only; all rights reserved.

than those of the one-way eccentricity case (case 3). This

displacement increase may be attributed to the cross-axis
effect reported by Lee and Liang (1998).
As shown in the test results of the two-way eccentricity
case (case 4), the maximum displacements measured at
the corner (Ucx)max are much larger than those measured
at the center of mass (Ubx)max. In addition, the contribution
of rotational component Ucx to the maximum corner displacement (Ucx)max is as much as 30%. Comparing Ucx
with (Ucx)max, it is realized that a very high percentage of
the maximum rotational displacement (Ucx)max has contributed to (Ucx)max. In addition, the time lags between
(Ucx)max and (Ucx)max are all very small for the two-way
eccentricity case. This finding is in contrast to that reported by Pan and Kelly (1983) and Kung and Pecknold
(1982) that, for a base-isolated structure, the torsional
coupling in the isolation system may not significantly affect the transient displacement response due to the time
lag between the maximum lateral and torsional responses
and the influence of damping in the isolation system.
However, it should be noted that the conclusion made by
those authors is for a small eccentricity case.
A comparison between the results from bilateral and triaxial tests for both one-way and two-way large eccentricity cases shows that the maximum displacements are not
significantly affected by the vertical ground acceleration.
CONCLUSIONS
Based on the 3D shaking table tests of a 0.4-scale threestory base-isolated steel structure, some conclusions of significance can be drawn as follows:
1. Plan-symmetric superstructure and isolation system:
Under a unilateral excitation, the effects of the axial
force variation of the lead-rubber bearings have to be
carefully considered for design. The axial force variation may be induced by the dead load distribution on
the bearings at different locations in the building, the
overturning moment of the superstructure, and the excitation of the vertical ground acceleration.
A method different from that specified in the current
design practices has been proposed to calculate the lateral force distribution on the superstructure. The appropriateness of the proposed method is verified by the
experiment.
The results from the bilateral and triaxial tests have
indicated that the maximum displacement determined
from the 1.0 0.3 direction combination rule may
not be conservative.
2. Plan-symmetric superstructure and eccentric isolation
system:
The vertical ground acceleration may dramatically increase the maximum story accelerations of the superstructure, whereas the maximum acceleration measured
at the base floor is not significantly affected. In addition, the effect of vertical ground acceleration on the
maximum displacement of the isolation system is less
significant than that of the plan-symmetric case.
Under a unilateral excitation, the maximum displacement
at the center of mass of the base floor may be reduced
for a small eccentricity case and increased for a large
eccentricity case, compared with the symmetric case.
Under a unilateral excitation, the maximum rotation
displacement at the corner of the isolation system with
a two-way eccentricity is much larger than the oneway eccentricity case. This may be classified as one of
the cross-axis effects.
Under a bilateral excitation of a two-way eccentricity

case, the maximum displacement at the corner of the

isolation system is larger than that at the center of
mass. In addition, a very high percentage of the maximum rotational displacement has contributed to the
maximum displacement at the corner of the isolation
system, at the time the maximum corner displacement
occurs. There is nearly no time lag between the maximum rotational displacement and the maximum corner displacement.
ACKNOWLEDGMENTS
This study was supported by the National Science Council of Taiwan
under Grant No. NSC-88-2625-Z011-004. The support of the National
Center for Research on Earthquake Engineering, for providing excellent
experimental facilities, is appreciated.

APPENDIX.

REFERENCES

AASHTO. (1999). Guide specifications for seismic isolation design,

Washington, D.C.
Al-Hussaini, T. M., Zayas, V. A., and Constantinou, M. C. (1994). Seismic isolation of multi-story frame structures using spherical sliding
isolation system. Rep. No. NCEER-94-0007, National Center for
Earthquake Engineering Research, Buffalo.
Chopra, A. K. (1995). Dynamics of structurestheory and applications
to earthquake engineering, Prentice-Hall, Englewood Cliffs, N.J.
CSMIP strong-motion record from the Northridge, California earthquake
of January 17, 1994. (1994). California Dept. of Conservation, Division of Mines and Geology, Sacramento, Calif.
Department of Interior. (1996). Seismic design specifications for buildings, Taiwan.
Eisenberger, M., and Rutenberg, A. (1986). Seismic base isolation of
asymmetric shear buildings. Engrg. Struct., 8(1), 28.
Griffith, M. C., Kelly, J. M., Coveney, V. A., and Koh, C. G. (1988).
Experimental evaluation of seismic isolation medium-rise structures
subject to uplift. Rep. No. 88/02, Earthquake Engrg. Res. Ctr., University of California, Berkeley, Berkeley, Calif.
Hwang, J. S., Chiou, J. M., Sheng, L. H., and Gates, J. H. (1996). A
refined model for base-isolated bridges with bi-linear hysteresis characteristics. Earthquake Spectra, 12(2), 245274.
Hwang, J. S., and Ku, S. W. (1997). Analytical modeling of high damping rubber bearings. J. Struct. Engrg., ASCE, 123(8), 10291036.
International Conference of Building Officials (ICBO). (1997). Uniform
building code, Whittier, Calif.
Kelly, J. M., Beucke, K. E., and Skinner, K. E. (1980). Experimental
testings of a friction damped aseismic base isolation system with failsafe characteristics. Rep. No. 80/18, Earthquake Engrg. Res. Ctr., University of California, Berkeley, Berkeley, Calif.
Kelly, J. M., Buckle, I. G., and Tsai, H. C. (1985). Earthquake simulator
testing of a base-isolated bridge deck. Rep. No. 85/09, Earthquake
Engrg. Res. Ctr., University of California, Berkeley, Berkeley, Calif.
Kung, S. Y., and Pechnold, D. A. (1982). Effect of ground motion characteristics on the seismic response of torsionally coupled elastic systems. Rep. No. 82-2009, Dept. of Civ. Engrg., University of Illinois,
Urbana-Champaign, Ill.
Lee, D. M. (1980). Base isolation for torsion reduction in asymmetric
structures under earthquake loading. Earthquake Engrg. and Struct.
Dyn., 8(3), 349359.
Lee, G. C., and Liang, Z. (1998). On cross effects of seismic responses
of structures. Engrg. Struct., 20(46), 503509.
Nagarajaiah, S., Reinhorn, A. M., and Constantinou, M. C. (1993). Torsion in base-isolated structures with elastomeric isolation systems. J.
Struct. Engrg., ASCE, 119(10), 29322951.
Nakamura, T., Suzuki, T., Okada, H., and Takeda, T. (1988). Study of
base isolation for torsional response reduction in asymmetric structures
under earthquake motion. Proc., 9th World Conf. on Earthquake Engrg.,
Vol. 5, International Association for Earthquake Engineering, 675680.
Pan, T. C., and Kelly, J. M. (1983). Seismic response of torsionally
coupled base isolation structures. Earthquake Engrg. and Struct. Dyn.,
11(6), 749770.
Tyler, R. G., and Robinson, W. H. (1984). High-strain tests on leadrubber bearings for earthquake loading. Bull. New Zealand Nat. Soc.
for Earthquake Engrg., Waikanae, New Zealand, 17(2), 90105.
Wen, Y. K. (1995). Building reliability and code calibration. Earthquake Spectra, 11(2), 269296.
Zayas, V. A., Low, S. S., and Mahin, S. A. (1987). The FPS earthquake
resisting system: Experimental report. Rep. No. 87/01, Earthquake
Engrg. Res. Ctr., University of California, Berkeley, Berkeley, Calif.

886 / JOURNAL OF STRUCTURAL ENGINEERING / AUGUST 2000

J. Struct. Eng., 2000, 126(8): 879-886