You are on page 1of 32

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2004; 33:543574 (DOI: 10.1002/eqe.360)

Analysis, testing, and implementation of seismic isolation


of buildings in Chile
Juan C. De la Llera; Carl Luders, Patricio Leigh and Henry Sady
Department of Structural and Geotechnical Engineering; Ponticia Universidad Catolica de Chile;
Casilla 306; Correo 22; Santiago; Chile

SUMMARY
This article summarizes the work done by the authors in seismic isolation over the past six years in
Chile. First, a general evaluation of the optimal values of the yield level of the isolation system is
performed, focusing on the idea of, but not restricted to, the use of leadrubber bearings. These optimal
values are obtained for two performance objectives: to minimize the base shear in the superstructure
and to control the isolator deformation. They were used in the design and construction of two important
isolated buildings that are described herein; a short description of the more relevant aspects of the design
and implementation of the isolation system in these two buildings is also presented. Furthermore, results
from a long testing program conducted on more than 260 full-size elastomeric isolators are summarized
and discussed. It is shown that these experimental results enable the elastomeric compounds to be
characterized quite accurately by testing reduced-scale specimens with elastomer thickness identical to
that used in the full-size isolators. Also, results from isolator constitutive modeling, scragging, and creep
in the short term are briey discussed. Inelastic analyses were performed in the structures in order to
evaluate realistic interstory drift and oor acceleration response reduction factors due to the isolation
design used. It is shown that in spite of the isolation system, minor inelastic excursions of the primary
structure are expected, leading to smaller drift and acceleration reduction factors than those obtained
from assuming an elastic response of the superstructure. In any case, seismic isolation is shown to be
a very competitive alternative, technically and economically, for building design in Chile. Although
it may be dicult to extrapolate this experience to other environments, the results presented herein
demonstrate that seismic isolation is a technique that can be eectively used to mitigate seismic hazards
in developing countries. Copyright ? 2004 John Wiley & Sons, Ltd.
KEY WORDS:

seismic isolation; elastomeric bearings; leadrubber bearings; optimal yield capacity;


Chilean code, isolator testing; building implementation

Correspondence

to: Juan C. De la Llera, Department of Structural and Geotechnical Engineering, Ponticia


Universidad Catolica de Chile, Casilla 306, Correo 22, Santiago, Chile.
jcllera@ing.puc.cl

E-mail:

Contract=grant sponsor: Chilean National Fund for Research and Technology, FONDECYT; contract=grant number:
1020774
Contract=grant sponsor: Fund for Foment and Technology, FONDEF; contract=grant number: D96I1008

Copyright ? 2004 John Wiley & Sons, Ltd.

Received 5 January 2003


Accepted 19 August 2003

544

J. C. DE LA LLERA ET AL.

INTRODUCTION
Seismic isolation has become a popular earthquake-resistant design technique for buildings and
bridges throughout the world. The technical advantages of this technique have been shown
by experimental work [13] and also by their successful response during earthquakes [4, 5].
In spite of this success, seismic isolation has not been incorporated into practice in some
earthquake-prone countries at a rate consistent with these advantages. Some arguments that
justify this are: (i) the natural conservatism of the profession, architects, structural engineers,
and codes, to incorporate new technologies in building and bridge design; (ii) the infrequent
occurrence of seismic events in some countries; (iii) cost-related issues regarding the devices
and their implementation; (iv) additional architectural and structural detailing; and (v) the lack
of knowledge of the users relative to the current earthquake design philosophy that admits
non-structural and structural damage during severe ground motions.
Because it is common to see seismic isolation stigmatized as an expensive technique, the
application of which should be primarily for critical facilities, one objective of this article is to
show the experience with base isolation in a country like Chile, within a framework of serious
budget limitations and other technological constraints. Although other applications of seismic
isolation are also increasing in this country, only its use in new building construction will
be emphasized herein, for which the implementation process is complex and where isolated
solutions must compete in direct costs with traditional earthquake-resistant schemes.
The rst base-isolated building in Chile was designed and built in 1992 [6] as a joint eort
between the professors and researchers at the Universidad de Chile, the Ministry of Housing, and Professor James Kelly of the University of California at Berkeley. The structure is a
four-story masonry building supported on eight medium damping rubber isolators (MDB). The
next step was a 3-year project funded by the government aimed at fostering the research and
implementation of seismic isolation and energy dissipation techniques in Chilean construction.
This project was developed at the Universidad Catolica de Chile and led to a new laboratory
for dynamic testing and vibration control. Several investigations with rubber and frictional isolation [7], and metallic and frictional dampers [8], have been developed since then. As a result
of this latter project, the rst seismic-isolated hospital was designed and constructed in Chile
in the year 2000 (H-UC). The 6-story R=C building is about 8000 m2 in plan and the isolation
system consists of 30 high damping isolators (HD) and 22 leadrubber bearings (LRB). Recently, a new seismically isolated institutional building for the Faculty of Engineering of the
Universidad Catolica (EF-UC) has been constructed (2002). The 5-story R=C structure is about
6000m2 and has 25 HD isolators, 17 LRB, and 11 steelteon frictional sliders. Most recently,
a new seismically isolated Military Hospital (MH) has been designed, with construction starting in January 2004. This 5-story R=C building has a total constructed area of about 50000 m2
with 164 isolators, 114 MDB and 50 LRB. Several aspects of the devices, earthquake response,
and design of the H-UC and EF-UC buildings are described in detail in this investigation.
During the design and construction processes of these buildings, analysis, design, and experimental aspects came up that the authors believe may be useful to the engineering profession
and scientic community. Because of space limitations only those aspects considered more relevant for structural engineering are reported in this article. Hence, construction aspects are not
reported; however, they are essential in a proper implementation of seismic isolation. In this article, the elastic and inelastic response of simplied models is rst used to estimate the optimal
yield capacity of LRB for impulsive-type and subduction-type ground motions. Then, selected
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

SEISMIC ISOLATION OF BUILDINGS IN CHILE

545

experimental results aimed at characterizing the horizontal cyclic behavior of MDB and LRB
from a sample of about 260 full-size seismic isolators are presented. Next, a more detailed description of the two buildings considered as examples is presented together with their expected
inelastic earthquake response. Finally, some practical observations relative to the analysis, design, and implementation of seismic isolation in these structures are presented; some of these
aspects are considered in the proposed building code for seismic isolation in Chile.
OPTIMAL YIELD CAPACITY
The rst aspect analyzed is the Optimal Yield Capacity (OYC) required for the isolation
system of a structure. In the case of LRB, the OYC is primarily related to the amount of
lead used in the design of these bearings; for MDB the OYC is dened by the mechanical
properties of the elastomer compound used. This yield capacity has an important eect on
the maximum base shear, torque, and acceleration of the superstructure, as well as the deformation demand in the isolation system. Along this line, Park et al. [9] looked for the OYC
of a bridge model subjected to dierent intensities of the NS component of the El Centro
record (1940). On the other hand, Ramallo et al. [10] have made a similar analysis for building models but without varying the ground motion intensity. This motivated the study of a
simplied 5-story building model, used in previous isolation investigations [10], subjected to
a suite of Chilean records and other well-known recorded ground motions for a wide range
of ground motion intensities. The objective of this parametric study was to dene the OYC
for the particular design of the H-UC and EF-UC buildings but results are extended to other
structures and ground motion characteristics.
Let us consider rst the earthquake response of the simplied 5-story building presented in
Figure 1. The parameters of this building are presented in Table I and were used previously
by Kelly et al. [11]. The building stiness k b is dened to generate an objective secant
design isolation period To = 2:5s. Because the isolation modes dominate the response of global
response quantities such as the base shear in the superstructure, a shear building model was
considered in the analysis with a single lateral degree-of-freedom at each oor level.
The equations of motion of the structure with an inelastic isolation system may be stated as
Mu + Cu + Ku + LT f = Mru g

(1)

where M, C and K are the mass, damping and stiness matrices of the superstructure, respectively; r is the excitation inuence vector; L is the kinematic transformation matrix between
the isolator deformation v and the degrees of freedom u of the structure, i.e., v = Lu; and
Table I. Structural model parameters.
Mass
(kg)
mb = 6800
M1 = 5897
M2 = 5897
M3 = 5897
M4 = 5897
M5 = 5897
Copyright ? 2004 John Wiley & Sons, Ltd.

Stiness
(kN=m)

Damping
(kN s=m)

kb = 232
k1 = 33732
k2 = 29093
k3 = 28621
k 4 = 24954
k 5 = 19059

c1 = 67
c2 = 58
c3 = 57
c4 = 50
c5 = 30
Earthquake Engng Struct. Dyn. 2004; 33:543574

546

J. C. DE LA LLERA ET AL.

m5

x5
k5

m5
k5

c5
m4

k4

m4
k4

c4

k3

c3
m2

k2

c2
m1

k1

c1

c1

mb

xb

LRB

g
ug
(a)

k2

c2

k1

c3
m2

m1

x1

c4
m3

m3
k3

c5

time

(b)

Figure 1. Simplied 5-story model used to compute the optimal yield capacity (OYC):
(a) base-isolated; and (b) xed-base.

f = f(v; v ; t) is the restoring force vector of the isolators. Thus the term LT f is the contribution
to the equilibrium, stated at the degrees of freedom of the structure, of the isolation restoring
forces. Equation (1) is frequently written in state-space format so as to integrate the linear
part of the structural response in exact form and the non-linear part by using, say, a predictor
corrector algorithm, or a RungeKutta method [12].
For this simplied parametric study, a simplied bilinear forcedeformation constitutive
relationship is assumed for the isolators. The inuence of two parameters is considered,
namely the yield force Qy , and the ratio between the post-yield and initial isolator stinesses, k Kf =Ki . To obtain a secant fundamental period To = 2:5 s, the post-yield stiness
is dened as Kf = k b = 232 kN=m; besides, the yield force Qy is related to the post yield force
Qf by Qf = (1 + k)Qy .
Let the optimal performance criterion for the OYC of the isolation system be dened
by either the minimum base shear of the superstructure, or by limiting the isolator deformations. For a given earthquake record, the range of normalized yield force considered is
Q y Qy =W = 0:005 0:4, where W is the seismic weight of the structure. Three values of k
are considered in the analysis, 6; 10 and 15, which are typical for LRB. The OYC of the
isolation system is obtained by minimizing the base shear of the structure as a function of Q y
for a range of peak ground acceleration (PGA) values. Because shear forces are of interest,
the records are scaled in order to obtain a functional relationship between the OYC and the
PGA values. A second criterion used was to dene the OYC required to limit the deformation
demand in the isolation system, which may be critical in some cases.
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

547

SEISMIC ISOLATION OF BUILDINGS IN CHILE


0.35

1
PGA = 0.17 g
PGA = 0.35 g
PGA = 0.53 g
PGA = 0.70 g
PGA = 0.88 g
PGA = 1.05 g
PGA = 1.23 g

0.9

Normalized base shear, V / W

0.3
0.8
0.25

0.7
0.6

0.2

0.5
0.15

0.4
0.3

0.1

0.2
0.05

Isolation deformation (cm)

0.1
0

40

80

35

70

30

60

25

50

20

40

15

30

10

20

10

0
0

(a)

0.05

0.1

0.15

Normalized yield force, QY

0.2

(b)

0.1

0.2

0.3

0.4

Normalized yield force, QY

Figure 2. Mean values of normalized base shear and isolation deformation as a function of the normalized yield force Q y of the isolation system: (a) Llo Lleo N10E; and (b) Sylmar N00E.

Figure 2 shows a typical relationship between the normalized base shear V V=W and the
normalized yield force Q y of the isolation system obtained for Sylmar (N10E component)
and Llo-Lleo (N00E component); the curves presented represent average values for the three
normalized stiness ratios k = 6; 10 and 15. As expected, the OYC in shear increases as the
PGA increases. For both records, such increase is almost linear in the space dened by V
and Q y [9]. The optimal yield values are, however, very dierent between the Sylmar and
Llo-Lleo records. For instance, OYC values for the Sylmar record vary between 4% and 25%
of the weight of the structure; the corresponding range for the Llo-Lleo record is between
0.5% and 3.5%, i.e., 7 to 12 times smaller. Besides, an increase of only 2% in the normalized
yield force Q y for the Llo-Lleo record leads to a large increase of 0.1 in the normalized based
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

548

J. C. DE LA LLERA ET AL.

shear V ; however, the same increase in Q y for the Sylmar record leads to an increase of only
0.04 in V . This implies, that the base shear in the structure is more sensitive to the yield
force in the former case. Furthermore, since the curves beyond the point of minimum shear
V show smaller slopes than the curves before that point, it is always convenient to select a
value of Q y slightly above the optimal value. Moreover, as should be expected, the isolation
deformations tend to decrease as the yield force increases for a given value of the PGA. The
trends indicated in Figure 2 by these two records turn out to be general for other records of
the same families, respectively [13].
Both performance criteria of the isolation system may be combined in a single plot as
shown in Figure 3. This gure presents the relationship between the normalized OYC, Q y ,
and PGA. Superimposed also is the normalized yield force required, as a function of the
PGA, to maintain simultaneously the isolator deformation demand at a specied value; these
restrictions are obtained from the points identied by the circular markers in the bottom
plots of Figure 2. Notice again that there is about a factor of 10 between the Q y values
for the Sylmar and Llo-Lleo records. For both records, the yield force required to control
the isolation deformation for values above the OYC increases rapidly with the PGA. These
deformation iso-performance curves may be used eectively in design in order to balance
base shear requirements and deformation requirements in the isolated structure. For instance,
it is apparent that forcing a small isolator deformation value for the true Sylmar record, say
D = 20 cm, would necessarily imply isolation yield forces larger than 30% of the weight of
the structure. On the other hand, for the Llo-Lleo record the required isolator yield forces are
less than 2% to keep a deformation demand of about 15 cm.
A summary of the normalized OYC values, Q y , and corresponding isolation deformations
for the nine dierent ground motions and three soil conditions considered are presented in
Figure 4. It is clear that dierent ground motions may lead to quite dierent required optimal
values and trends are hard to generalize. Results conrm, however, that the OYC in the
case of Chilean records and dierent soil conditions will rarely exceed 3%, associated with
isolation deformations of less than 40 cm. For soil conditions I and III (Table II), the OYC
is rather insensitive to variations of the PGA and, hence, the lead core is justied only for
controlling isolator deformations under serviceability conditions and for an eventual torsional
balance of the building plan. Instead, for soil type II there is a larger dependency of the OYC
with the PGA values. For instance, a value of 2.5% for the OYC would be adequate to keep
the normalized base shear in the superstructure less than 0.12 and isolator deformations less
than 20 cm for a rather large PGA of 0:8g.
For the other suite of ground motion records considered, Kobe and Newhall show the
smallest OYC requirements, say less than 10%. At the other end, the SCT, Taiwan, and
Sylmar ground motion records would require OYC values larger than 15% of the weight
of the structure. For instance, Q y = 0:25 would correspond to a PGA of about 0:25g in
the case of the SCT record; however, the same Q y value would correspond to a PGA
of 1:2g for the Sylmar record, i.e., about ve times larger. As should be expected, there
is a strong dependency of the required OYC values on the frequency distribution of energy of the ground motion. For all ground motions considered, linearity between the OYC
and PGA or isolator deformation is apparent and, hence, two design points would sufce to extrapolate or interpolate the behavior of the structure for other values of these
parameters.
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

549

SEISMIC ISOLATION OF BUILDINGS IN CHILE


0.08

Minimum base shear


Isolation deformation limit

0.07

D=2
0 cm

Normalized OYC

0.06

0.05

0.04

0.03

Llo Lleo N10E

D=

30 c

0.01

D=

25

cm

0.02

(a)

0.5
0.45

cm
cm
40
D=

20

0.25

35

30
D=

cm

D=

0.3

D=

25
cm

cm

0.35

D=

Normalized OYC

0.4

0.2
0.15
0.1

Sylmar N00E

0.05
0

0.2

0.4

0.6

0.8

1.2

1.4

Normalized PGA (g)

(b)

Figure 3. Optimal yield capacity for two performance criteria: (a) minimum base shear; and
(b) isolation deformation limit.

EXPERIMENTAL PROGRAM
This section summarizes the results of dynamic testing of rubber isolators and reduced-scale
specimens conducted between 1997 and 2002 intended to select, design, and validate the cyclic
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

550

J. C. DE LA LLERA ET AL.

0.035

Zapallar
Llo Lleo
Via del Mar

Normalized OYC

0.03

0.025

0.02

0.015

0.01

0.005

0
0.35

El Centro
Sylmar
Newhall
Kobe JMA
Taiwan
Mexico SCT

Normalized OYC

0.3

0.25

0.2

0.15

0.1

0.05

0
0

0.2

0.4

0.6

0.8

Normalized PGA (g)

1.2

1.4

10

20

30

40

50

60

Deformation (cm)

Figure 4. Optimal yield capacity and isolation deformation corresponding to minimum base shear in the
isolation system for dierent ground motion records.

behavior of the isolation system for the H-UC, EF-UC, and MH buildings. Such a program
considered the testing of 100 by 100 mm square elastomer specimens up to rupture (over
200), quality control testing of more than 260 isolators under design conditions, and over 20
prototypes of isolators under maximum capable earthquake conditions, including simultaneous
shear=tension and shear=compression.
The set-up used initially for testing couples of isolators is presented in Figure 5(b) and
(c) and consists of a dynamic 1000 kN MTS actuator and a vertical loading system to apply
5000kN in compression and 400kN in tension. A new testing rig has been recently developed
to test individual isolators with a maximum axial load of 8000kN (Figure 5(d)). The stroke of
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

551

SEISMIC ISOLATION OF BUILDINGS IN CHILE

Table II. Seismic records considered in this investigation.


Group

Record

Direction

PGA
(g)

PGV
(cm=s)

PGD
(cm)

Soil I
Rock and sti soil
(Vs 900 m=s)

Zapallar a;b
Valparaso, UTFSMb
Quintayb

EW
N70E
T

0.305
0.176
0.260

13.470
14.600
19:351

1:898
3.143
3:503

Soil II
Sti soil
(400 m=sV 900 m=s)

San Felipeb
Llo-Lleoa;b
Melipillab

N80E
N10E
NS

0.434
0:713
0:687

17:774
40:295
34.289

3:705
10:786
13.300

Soil III
Soft soil
(Vs 400 m=s)

Valparaiso, Almendralb
Vina del Mar a;b
Llay Llayb

N50E
S20W
N80W

0.297
0.363
0:475

28:577
30.742
36.710

6:277
5:534
5:736

Soil II
Soil III
Soil II
Soil I
Other
Soil II
Other

Sylmar a;b
Newhalla;b
Corralitosb
El Centroa
tai068-butter a
Kobe, JMAa
Mexico, SCTa

N00E
N00E
N00E
N00E
N00E
N00E
N90E

0.843
0.591
0.63
0:348
0.568
0.822
0:171

128:880
94:73
55:2
33:45
116:27
81.30
60:610

30:670
28.81
12.03
12:36
56.20
17:69
21.160

a Used to determine optimal yield


b Considered for seismic analysis.

capacity.

the actuator is 1000mm and is connected to a 280gal=min pump through a manifold with three
25l: accumulators. The axial load during the test is maintained constant by the use of a passive
pressure control system connected to four axial load actuators of 2500kN each. The two beams
of the test rig (Figure 5(c)) are joined by a hinge (Figure 5(c)) that enables 3D rotation of the
actuator (primarily about a vertical axis); the beam on the left of the test rig is anchored to the
ground and the other is free to pivot laterally on top of a teon interface. The whole system is
controlled analogically for simple signals with an MTS 407 controller, or digitally for earthquake and real-time simulation using the software and hardware dSpace [14]. The test rig for
individual isolators (Figure 5(d)) works on a vertical plane; the isolator is xed in the upper
end to the vertical loading frame and on the other end to a displacement controlled sliding table
mounted on rollers. Thus, the 1000 kN shear may be essentially applied to a single isolator.
Elastomeric isolators of diameter  ranging from 600 mm to 900 mm were installed in the
H-UC, EF-UC, and MH buildings and considered in this research program. All these isolators
were tested up to a design shear deformation  = 150%; 20 additional isolators were tested
up to  = 200%, which takes the actuator to its maximum shear capacity. Rubber compounds
for these three buildings were provided by a local manufacturer and based on an exhaustive
destructive testing of reduced-scale 100 by 100 mm specimens and isolator prototypes. As
will be shown next, the correlation between the mechanical properties of these reduced-scale
specimens and the full-scale isolator properties enables one to predict the behavior of the
full-scale isolators from small to large shear deformations of the elastomer compound.
Figure 6 shows typical results for the secant shear modulus Ge(s) and eective damping
ratio (s)
of three reduced-scale specimens with a medium stiness compound subjected to
e
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

552

J. C. DE LA LLERA ET AL.

Lead-core

(a)

(b)

(c)

(d)

Figure 5. Experimental set-up for shear-compressiontension testing: (a) typical  = 600 mm isolator;
(b) shear-compression test; (c) test rig for isolator couples; and (d) test rig for individual isolators.

sinusoidal cyclic testing at frequency f = 0:1 Hz. The specimens have two elastomer layers
of the same thickness as that used in the full-scale isolators. The testing sequence included
shear deformation amplitudes  = 0:25, 0.5, 0.75, 1, 1.25, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5
and 6; this latter value was attained only in a few specimens. A typical shear failure of the
elastomer is shown in Figure 6(b); the propagation of rupture in the compound is caused by
the increase in shear deformation and starts for nominal shear strain 5, which was the
minimum performance established for these projects. It is interesting to mention that these
compounds all have elongations at rupture u 6. The smaller value of the shear deformation
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

553

SEISMIC ISOLATION OF BUILDINGS IN CHILE

0.12

14

0.1

12

0.08
(s)

16

10

0.06

0.04

(c)

(b)

Ge(s) (kgf/cm2)

(a)

Shear strain,

0.02

(d)

Shear strain,

Figure 6. Typical response of reduced-scale specimen, with a medium-stiness rubber compound: (a) scaled specimen; (b) rubber compound failure; (c) secant shear modulus, Ge ();
and (d) eective damping ratio, e ().

capacity u , as opposed to u , is attributed to: (i) the progressive damage that occurs in the
specimen as a result of the seven dynamic cycles applied at each deformation level and (ii)
the local shear strain concentrations in the corners of the square specimens. Figure 6(c) and
with shear strain for the medium stiness
(d) present a typical variation of Ge(s) and (s)
e
elastomeric compound. It is apparent that stiening of the specimen begins at  1:5 and
that the eective damping ratio always decreases with increasing . Two dierent elastomer
compounds were used in the isolation system of the H-UC and EF-UC buildings with nominal
secant shear values Ge = 5 kgf =cm2 (low stiness) and Ge = 7:5 kgf =cm2 (medium stiness).
Both compounds contain, in weight, over 90% natural rubber; damping is obtained by the use
of special mineral oils and proportions of reinforcing llers. Although not included here for
brevity, all experimental results lead to trends similar to those presented in this gure.
Similar testing results are presented for full-scale isolators of the H-UC building in
Figure 7. Both low- and medium-stiness elastomer compounds were used in the building.
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

554

J. C. DE LA LLERA ET AL.

14
13

Ge(i) (kgf/cm2)

G = 1.7844 2 5 .0684 + 8.4441

12

G = 1.7844 5.0684 + 8.4441

2
GG==2.5501
2.5501 7.4806
7.4806 ++13.108
13.108

11

R = 0.8275

R = 0.8558

10

9
5

8
7

6
3

5
0.14

= -0.9158e + 0.2337 1.20321/2 +


1.4014

0.22
0.2

= -0.8159e + 0.2491 1.12431/2 +


1.2603

0.12

e(i)

0.18
0.16

0.1

0.14
0.12

0.08

0.1
0.08

0.06

0.06
0

(a)

0.5

Shear strain,

1.5

0.5

1.5

Shear strain,

(b)

Figure 7. Shear modulus Ge(i) () and damping ratio e(i) () obtained from testing of actual-size isolators with low- and medium-stiness rubber compounds: (a) low stiness
compound; and (b) medium stiness compound.

Because isolators used in the EF-UC building were all manufactured with the same medium
stiness compound, testing results are omitted herein for brevity. The trends for the secant
shear modulus Ge(i) and damping ratio (i)
e of the isolators follow similar trends as those of
the reduced-scale specimens. It is apparent that the low-stiness compound used was targeted
for high damping ratios, while the medium-stiness compound was for medium damping. At
least in the authors testing experience, there is almost inevitably a trade-o between larger
damping values, smaller shear deformation capacity u , and smaller stiness of the elastomer
compound. Hence, the designer should be aware that large damping is harder to obtain with
high-stiness compounds and that damping values larger than those presented in Figure 7 may
lead to elastomer compounds with reduced shear deformation capacity u as compared to that
of low-stiness compounds.
In light of the results of Figures 6 and 7, a relationship may be found between the properties
of reduced-scale and full-scale specimens. Such results are presented in Figure 8 where the
(s)
are plotted against each
secant shear modulus values Ge(i) Ge(s) and damping ratios (i)
e e
other. The results are interesting and show an almost perfectly linear relationship between the
two. For instance, the suggested equations for the medium stiness compound are
Ge(i) = 0:7684Ge(s) + 1:5504
Copyright ? 2004 John Wiley & Sons, Ltd.

(kgf =cm2 )

(2)

Earthquake Engng Struct. Dyn. 2004; 33:543574

555

SEISMIC ISOLATION OF BUILDINGS IN CHILE

0.14

13
0.12

11

Ge(s) = 2.5409 2 - 7.7910 + 13.4491

10
9
8
7

e(i) = -0.5038e + 0.1766


0.7531/2 + 0.8418

0.1
0.08
0.06

e(s) = -0.051e + 0.0485


0.13731/2 + 0.1581

0.04

Ge(i) = 2.06683 - 6.1863 + 11.9514

0.02

0.2

Isolator shear modulus, Ge(i)


(kgf/cm2)

Damping ratio, e

12

0.4

0.6

0.8

Shear strain,

1.2

1.4

0.2

1.6

0.12

11

10.5
10
9.5
9
8.5
8
7.5

0.4

0.6

0.8

1.2

1.4

1.6

Shear strain,

Isolator damping ratio, e(i)

Shear modulus, Ge (kgf/cm2)

14

Ge(i) = 0.7684Ge(s) + 1.5504

0.11

exp. fit
linear fit

0.1
0.09
0.08

e(i) = 1.3788e(s) 0.0051


0.07

7
6

10

11

12

13

0.06
0.02

(s)

0.04

0.06

0.08

Specimen damping ratio, e

Specimen shear modulus, Ge (kgf/cm^2)

0.1
(s)

Figure 8. Relationship between secant shear modulus and eective damping ratio determined from the
measured shear response of full-scale isolators and reduced-scale 100 100 mm specimens.

and for the damping ratio


(s)
(i)
e = 1:3788e 0:0051

(3)

Although these expressions may change among isolator manufacturers, a relationship of this
kind may still exist. This implies that it is possible to predict the values of the secant shear
modulus and eective damping ratios of the full-scale isolators by means of the 100 by
100 mm specimens for the range 1:5, which covers the usual design region. Furthermore,
these results may probably be extrapolated at least for 1:52:5 as well, and enable us to
foresee the performance of the isolator for a maximum capable earthquake condition. This is
why it seems advisable to request a 100 by 100 mm specimen for each manufactured isolator;
this would eventually make testing of full-scale isolators less necessary in the future.
The eect of axial loads on the cyclic behavior of  = 600 mm LRB isolators is evaluated
next. A lead core of diameter  = 100 mm was introduced in the central orice of a couple
of isolators with nominal shear modulus G = 7:5 kgf =cm2 . Results of testing these isolators at
dierent axial loads and deformation amplitudes are presented in Figure 9. Figure 9(a) shows
the behavior of the isolators for very small deformations ranging from 0:3 mm to 10 mm,
three axial loads N = 1000, 2000 and 3500 kN, and three sinusoidal excitation frequencies
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

556

J. C. DE LA LLERA ET AL.
Deformation, cm
1.28
150

0.96

0.64

0.32

0.32

0.64

0.96

1.28

N = var.

Shear force (kN)

100

50

50

100

Axial Force N = 1000 kN


Axial Force N = 2000 kN
Axial Force N = 3500 kN

150
0.08

0.06

0.04

0.02

0.02

0.04

0.06

0.08

16

24

32

Shear strain

(a)

Deformation, cm
32

800

24

16

N = 3500 kN

600

8
7

Shear force (kN)

400

4
3

200

2
1

0
200
1) = 0.125, E = 554 kNcm
2) = 0.25, E = 1503 kNcm
3) = 0.5, E = 3844 kNcm
4) = 0.75, E = 6118 kNcm
5) = 1,
E = 9989 kNcm
6) = 1.25, E = 11273 kNcm
7) = 1.5, E = 14222 kNcm
8) = 1.75, E = 17222 kNcm

400
600
800
2

1.5

(b)

0.5

0.5

1.5

Shear strain

Figure 9. Constitutive behavior of a couple of LRB ( = 600 mm) for small and large deformations and
dierent axial loads: (a) small deformations; and (b) large deformations.

f = 0:0887, 0.2 and 0:5 Hz. Figure 9(b) shows the cyclic behavior for larger deformations
ranging between 20 and 280 mm for the same axial loads, and a single excitation frequency
f = 0:1 Hz.
Properties of these forcedeformation cycles are summarized and compared for dierent
axial loads and excitation frequencies in Figure 10. By comparing the results of the small and
large deformation cycles, it is apparent that the value of axial load has an eect mainly on the
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

Copyright ? 2004 John Wiley & Sons, Ltd.

Normalized Stiffness
k / k = 1
0

10

15

20

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Shear strain,

(b)

10

15

20

25

30

25

0.5

1.5

2.5

30

10

15

f = 0.0887 Hz., N = 1000 kN


f = 0.2 Hz., N = 1000 kN
f = 0.5 Hz., N = 1000 kN
f = 0.0887 Hz., N = 2000 kN
f = 0.2 Hz., N = 2000 kN
f = 0.5 Hz., N = 2000 kN
f = 0.0887 Hz., N = 3500 kN
f = 0.2 Hz., N = 3500 kN
f = 0.5 Hz., N = 3500 kN

0.2

0.4

0.6

1.2

Shear strain,

0.8

1.4

1.6

1.8

Figure 10. Measured variation of the normalized stiness and eective damping ratio of LRB with axial load and
excitation frequency: (a) small deformations; and (b) large deformations.

(a)

Effective damping ratio,

20

SEISMIC ISOLATION OF BUILDINGS IN CHILE

557

Earthquake Engng Struct. Dyn. 2004; 33:543574

558

J. C. DE LA LLERA ET AL.

small deformation cycles (Figures 9 and 10). In the latter case, an increase in compression
force leads to better connement of the lead core, thus increasing the normalized isolator stiness K = k( = 1) =k = k sec =((2Ge A(i) +2A(l) Qy(l) )=Hr ) and eective damping ratio e . Moreover, as
should be expected, an increase in the excitation frequency leads to an increase in the isolator
stiness, but less than 10%; it also leads to a slight decrease in the damping ratio. The sensitivity of the response to dierent axial loads reduces with increasing shear deformation for the
level of deformations considered (Figure 10(b)). Indeed, for 0:4, the three curves are for all
design purposes essentially identical. Results may also be presented in terms of the normalized
hysteretic energy, e = Wd =Wd( = 1) , and normalized shear force, V = Vo =V( = 1) = Vo =(2Ge A(i) +
2A(l) Qy(l) ), but show identical trends [13]. Although these results may be slightly perturbed by
the scragging recovery present in the isolators, they clearly show that for compression stresses
in the isolators less than 130 kgf =cm2 , the eect of axial loads is negligible as seen from the
shear forcedeformation cycle of these isolators. In spite of this, axial forces are extraordinarily important in isolation design and evaluation of the maximum shear deformation of the
elastomer, stability properties, and long-term properties of the isolators.
Time variation of isolator properties is a permanent concern of users and designers. Some
results of scragging, creep, and time change of mechanical properties in the short term are
presented in Figure 11. Figure 11(a) shows a sequence of cycles performed to evaluate the
constitutive behavior in shear and scragging of the isolators. The elastomeric isolator selected ( = 900 mm) was subjected to the following tests: (a) one rst cycle at  = 0:25, 0.5,
0.75, 1, 1.25 and 1.5; (b) after each cycle in (a), seven cycles were performed dynamically
(f = 0:1Hz) with the same , the last of these seven cycles is presented by dashed lines in the
gure; (c) repeat (b) at increasing values of shear deformation  = 0:25, 0.5, 0.75, 1, 1.25 and
1.5only as an example, the maximum force attained in the last of these cycles for  = 1:25
is indicated by c in the gure; (d) repeat (c) 15 hours later. Notice that if the isolator deformation exceeds for the rst time the deformation corresponding to that of a previous cycle,
the curves catch the envelope of the unscragged sheardeformation constitutive relationship.
This implies that scragging is a deformation-dependent phenomenon and will restart every
time the isolator is taken to a new larger deformation. In our example, point a represents the
force for the unscragged envelope (upper bound) and point b is the maximum force where the
cyclic behavior becomes stable for a maximum  = 1:25. However, if a larger deformation,
say  = 1:5, is imposed on the isolator, the cyclic behavior for  = 1:25 now becomes stable
at point c, which is considerably lower than point b. Moreover, if the test is repeated several
hours later, the cycles become stable at point d, which is slightly below c. Thus, although the
isolator may have been scragged to a deformation, say o , scragging at larger deformations
o will decrease the isolator forces developed at o despite the previous scragging. This
eect should be considered in deciding the objective scragging deformation.
A second eect in design is the time recovery of the unscragged properties of the elastomer.
Compared in Figure 11(b) are the values of the secant shear modulus G obtained from several
tests of isolators performed about 96 hours apart; G1 and G2 represent the initial and last shear
modulus, respectively, for the same isolator under the same testing conditions. The apparent
trend is that for smaller deformations the isolator stiness increases (recovers) with time.
Because all the projects presented in this study are very recent there is no further information
that enables us to conrm this short-term behavior in the long term. Furthermore, typical
results from creep tests of isolators are presented in Figure 11(c). For these isolators, the nal
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

559

SEISMIC ISOLATION OF BUILDINGS IN CHILE


800

700

a
b
600

Shear force (kN)

500

400

300

200

100

10

(a)
13

15

Deformation (cm)

G2 (kgf/cm2)

10

(b)

0.5

0.4

0.3

0.2

0.1

G560, N = 3900 kN
G7.560, N = 4500 kN
G5 60, N = 4500 kN

0.6

11

25

0.7

= 0.25
= 0.50
= 1.00
= 1.50

12

20

Deformation (cm)

10

11
2

G1 (kgf/cm )

12

13

(c)

10

20

30

40

50

60

70

Time (hours)

Figure 11. Scragging, recovery and creep in the short term for a typical isolator: (a) typical constitutive
behavior of  = 900 mm rubber isolator; (b) recovery of second shear modulus; and (c) isolator creep.

deformation set values are achieved in less than 30 hours. A large number of short-term creep
tests performed on these isolators demonstrate that for long-term compression stresses less than
130 kgf =cm2 , short-term creep deformations will rarely exceed 1 mm for these compounds and
isolator design. The uctuations observed in the nal deformation values may be attributed
to changes in ambient temperature.
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

560

J. C. DE LA LLERA ET AL.

The forcedeformation constitutive relationship of a LRB may be cast in terms of a recently


proposed model [15]. The model has fewer parameters than the original model proposed by
Pan and Yang [16] and is capable of representing peculiar phenomena of elastomeric bearings such as Mullins and scragging eects. In this model, the forcedeformation relationship
proposed is dened by a pseudo-linear relationship [15]:

f(v(t); v (t)) = k(v(t);


v (t))v(t) + c(v(t);

v (t))v(t)

(4)

where k(v(t);
v (t)) and c(v(t);

v (t)) are the deformation and deformation-rate dependent stiness and damping of the isolator, respectively; and v(t) and v (t) are the isolator deformation
and deformation velocity. The stiness may be computed through
t

a4 e(a9 0 f(v(t);v(t))v(t) dt )

k(v(t);
v (t)) = a1 + a2 v(t)2 + a3 v(t)4 +
cosh2 (a5 v (t))

(5)

and the damping coecient by


t
a6 + a7 v(t)2
(1 + e(a10 0 f(v(t);v(t))v(t) dt ) )
c(v(t);

v (t)) = 
a28 + v (t)2

(6)

where a1 to a10 are parameters that control the forcedeformation loop. For instance, the term
a9 times the integral of the work done by the isolator is intended to represent the loss of
stiness due to scragging of the isolator; the term a10 is related to the energy lost through
cycling of the isolator. Figure 12(a) shows the results obtained from this model for dierent
shear deformations and a constant compression load N = 3500 kN; the selected parameters a1
to a10 are dened in the enclosure in units of ton and cm. The matching observed between the
predicted and measured traces of the forcedeformation relationship is good, except for shear
deformations below  = 0:25. For other values of the axial load N in ton, just a8 needs to be
modied by considering the relationship a8 = 4:188(6:2+0:003N ). Also shown in Figure 12(b)
is the identied behavior of the lead core of the isolator, obtained by subtracting the measured
constitutive behavior of the isolator without the lead core from the total shear-deformation
behavior of the isolator including the lead core. It is observed that the identied yield capacity
in shear of the lead is about 10 MPa, which is the usual nominal value considered in design.

DESCRIPTION OF BUILDINGS
The rst building considered is a new hospital of the Universidad Catolica (H-UC) located on
the skirts of the Andes mountains in the east part of the city of Santiago (Figure 13(a)). The
building was designed in 1999 and its one-year construction was nished in early 2001. The
building has a total area of about 8000 m2 distributed in 6 stories (5 above ground level and 1
basement) and weighs about 10341ton above the isolation level. As indicated in Figure 14(a),
the building was built as a single unit with no construction joints due to the use of the isolation
system; its approximate plan dimensions are 78 m by 31 m. Story heights are equal to 3:4 m
in the upper 5 stories and 4 m at the basement. The structure is formed by ductile momentresisting frames with 80 cm deep beams with a typical 12 m span and columns with 75 cm
square sections in the superstructure and 90 cm at the basement. These latter columns are
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

561

SEISMIC ISOLATION OF BUILDINGS IN CHILE


Deformation (cm)
32

24

800

16

N = 3500 kN

16

24

32

Predicted
Measured

600

Shear force (kN)

400
200
0
200
400
600

a1 = 1964.8124

a2 = -1.0064

a3 = 0.0009043

a4 = 771.9865

a5 = 0.17741

a6 = 7625.2165

a7 = 0.5489

a8 = 1.1726

a9 = 1.1726

a10 = 1.1726

800
2

1.5

0.5

0.5

1.5

Shear deformation

(a)
15

Shear stress, MPa

10

10

15
0.8

0.6

(b)

0.4

0.2

0.2

0.4

0.6

0.8

Shear deformation

Figure 12. Comparison between the measured and predicted cyclic behavior of a LRB and identied
behavior of the lead-core: (a) LRB constitutive behavior; and (b) identied lead-core behavior.

laterally braced by ribs coming out of the perimeter retaining wall, and interior columns are
connected below the isolation level by 40 40cm axial R=C struts. Concrete with a cylindrical
strength fc = 300 kg=cm2 and steel with a nominal yield stress fy = 4200 kg=cm2 was used in
the construction of the structure. As shown in Figure 14, seismic isolation is placed at the
top of the basement, thus avoiding the construction of an extra oor slab for the building.
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

562

J. C. DE LA LLERA ET AL.

(a)

(b)

Figure 13. Isolated buildings described in this investigation: (a) Hospital building (H-UC); and
(b) Engineering faculty building (EF-UC).

Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

563

SEISMIC ISOLATION OF BUILDINGS IN CHILE

RB, H5-60 cm
RB, H8-60 cm
RB, H8-65 cm
RB, H8-70 cm
LRB, H5-60 cm
LRB, H8-60 cm

(a)

(b)

Figure 14. Hospital building considered in this study (H-UC): (a) building plan; and
(b) moment resisting plane 17.

Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

564

J. C. DE LA LLERA ET AL.

The isolation system consists of 52 elastomeric isolators of diameters 600, 650 and 700 mm,
22 of them with  = 100 mm lead cores, according to the planwise distribution presented
in Figure 14(a). The two elastomeric compounds presented earlier (Figure 7) were selected
for the structure, with design shear modulus G = 5 kg=cm2 and G = 7:5 kg=cm2 . The nominal
isolation periods of the structure are Tx = 2:80 s, Ty = 2:64 s and T = 2:27 s; the corresponding
periods of the xed-base structure are 0:41 s, 0:38 s and 0:23 s, respectively.
Since all isolators were tested prior to their installation in the building, their nal location
was selected in order to balance the structure in torsion. Indeed, the nal estimated nominal
eccentricities of the structure with measured stiness values are ex =b = 0:027 and ey =b = 0:011,
which are small for such an irregular structure. In order to increase the lateral-to-torsional
frequency ratio of the structure , the isolators with lead cores were placed closed to the
perimeter of the building (Figure 14(a)); the nal for the structure is estimated to be
about 1.2. The spectrum selected for the design of the isolation system is based on the
measured ground motion during the 1985 Chile earthquake, and denes for this structure
a maximum displacement for the isolator of 35 cm associated with a 10% probability of
exceedance in 100 years. Furthermore, based on the results presented earlier for the OYC, a
value Qy = 2:5% was selected for the design of the structure. Although not presented here
for the sake of brevity, architectonic details of this structure are rather complex and a large
number of interesting solutions were proposed for hanging a cable-stayed bridge used in the
main access of the structure, ramps, hanging staircases, hanging elevator shafts, pool, drains,
auditorium, piping, and so forth; the interested reader may check them elsewhere [17].
Figure 13(b) shows the new Engineering Faculty building of the Universidad Catolica
(EF-UC), which is located in the San Joaqun campus of this university in a SW district
of Santiago. The building was designed in the year 2000 and constructed during 2001. It
has a total surface area of approximately 6000 m2 distributed in 5 stories (4 above ground
level and 1 basement) and the superstructure weighs about 7456 tons. It has a rectangular
plan with plan dimensions 92 m by 14 m and, as in the previous case, it was built as a
single unit with no construction joints due to the lateral exibility of the isolation system
(Figure 15). The structure of the building is based on a combination of R=C shear walls
(e = 20 cm) and moment-resisting frames with reinforcement ductility details corresponding to
a zone of moderate seismicity. Such is the case because the shear walls constitute the lateral
load-resisting system and the columns are primarily a vertical load-carrying system. A typical
two span (9 m and 5 m) moment-resisting frame in the transverse direction is shown in Figure
15(b). Beams are either 20cm by 70cm, or 35cm by 75cm deep, rectangular exterior columns
are 40cm by 60cm, and interior columns are 45cm square sections. Nominal concrete strength
and steel reinforcement yield stress are the same as for the previous structure.
The structure is isolated at the foundation level with 42 identical elastomeric isolators
(Figure 15(a)) of diameter  = 600 mm, of which 17 have  = 100 mm lead cores, and 11 are
steelteonelastomer sliders. The structure has a retaining wall along its perimeter separated
from the main building structure by a 32cm gap, which is the maximum displacement expected
for an earthquake with a 10% probability of exceedance in 100 years. A single G = 7:5kg=cm2
elastomer compound was used for all isolators and sliders. The 300 mm slider moves on a
stainless steel sliding surface with  = 1200 mm diameter. The measured frictional coecient
varies approximately from 0.10 to 0.15 for the range of velocities and maximum axial load
of 68 ton assumed in the design. As in the previous case, the natural eccentricity of the
superstructure was corrected by the proper placement of the isolators in plan, leading to
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

SEISMIC ISOLATION OF BUILDINGS IN CHILE

565

G7.5-60
cm
RB, H860 cm
G7.5-60
60cm
cm
LRB, H8SL-30

(a)

(b)

Figure 15. Engineering Faculty building considered in this study (EF-UC): (a) building plan (basement);
and (b) moment resisting plane 7.

a nal design with maximum nominal eccentricities of ex =b = 0:003 and ey =b = 0:06. The
nominal isolation periods of the structure are Tx = 2:29 s, Ty = 2:29 s and T = 2:12 s; the
corresponding periods of the xed-base structure are 0:189 s, 0:32 s and 0:29 s, respectively.
Again, the isolators with lead cores were placed close to the perimeter of the structure leading
to = 1:08. Furthermore, an OYC Qy = 2:3% was selected in this case.
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

566

J. C. DE LA LLERA ET AL.

The nal costs of the structure, including all aspects of the isolation system, were US$
115=m2 and US$ 125=m2 for the H-UC and EF-UC buildings, respectively. In Chile these costs
are similar to those of conventional structures, positioning seismic isolation as a competitive
technique in the market. This leaving aside powerful arguments such as protection of contents
and earthquake-damage reduction in the isolated structure relative to a conventional structure.
Currently, three new modules are being constructed to the south of the EF-UC structure,
making this the longest building in the country (L 120 m) without construction joints.

ELASTIC AND INELASTIC BUILDING RESPONSE


In this section, the inelastic response of typical resisting planes of the H-UC and EF-UC
(Figures 14 and 15) is evaluated. A parametric study is performed for 12 recorded impulsive
and non-impulsive ground motions and variable yield capacity Qy of the isolation system
ranging between 2.5% and 15% of the weight of the structure. The inelastic analyses are performed assuming elasto-plastic 2D hinges at the column and beam ends of the superstructure.
Plastic behavior in the hinges is dened by an interaction surface between axial load and
bending moment computed from nominal section and reinforcement properties of the structural members. Each isolator is modeled by a bilinear constitutive relationship calibrated with
their experimentally measured constitutive behavior.
Four analyses are performed with dierent analytical models of the superstructure and
ground motion records, i.e.: (i) xed-base structure with linear behavior (L); (ii) xed-base
structure with inelastic behavior (NL); (iii) linear structure with inelastic behavior of the isolation system (LI); and (iv) inelastic structure with inelastic behavior of the isolation system
(NLI). Response reduction factors are dened as the ratio of the response of the xed-base
structure over the corresponding response of the isolated structure; i.e., R = r (o) =r (i) . Consequently, two reduction factors are dened: a linear reduction factor R(L) = r (L) =r (LI) , relating
the responses of the linear xed-base and isolated structure and the inelastic reduction factor R(NL) = r (NL) =r (NLI) , relating the responses of the inelastic xed-base and inelastic isolated
structure.
Figure 16 shows the forcedeformation responses of the isolation system of the H-UC building for three dierent yield force values, Qy = 2:5%, 5% and 15%, and three ground motions.
For the EW sti soil site ground motion Zapallar (Chile, 1985), the isolated structure has a
maximum deformation less than 1:5 cm; notice also that the minimum shear value occurs for
Qy = 2:5% and that for Qy = 15% the isolation system remains elastic. The NS component
of the Melipilla record (Chile, 1985) is more demanding and for Qy = 2:5%, which leads
to the smallest base shear, the maximum isolator deformation is about 10 cm. It is apparent
how the shear force in the isolation system increases as Qy increases, justifying the use of a
small OYC for the Chilean records. Furthermore, the N00E component of the Sylmar record
(Northridge, 1994) is quite demanding for the isolation system of this building, leading to
a deformation demand close to 50 cm for Qy = 2:5%. However, it is interesting to note that
for all Qy values, the maximum base shear demand in the structure is very similar, while the
maximum isolator deformations decrease signicantly for larger Qy values. Consequently, if
the H-UC building were subjected to the Sylmar record, the OYC should have been of the
order of 15% of the weight of the structure instead of the 2.5% used for its actual design.
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

567

SEISMIC ISOLATION OF BUILDINGS IN CHILE

Qy/W = 2.5 %
Qy/W = 5 %
Qy/W = 15 %

Zapallar EW

200
0

Isolator shear force (kN)

200
5
400
200

Melipilla NS

0
200
400
15
500

10

10

15

20

40

60

Sylmar N00E

500
60

40

20

Deformation (cm)

Figure 16. Typical bilinear forcedeformation response of an isolator of the H-UC building with inelastic
model of the isolation system and superstructure.

A comparison between the elastic and inelastic roof accelerations and 2nd-story drift of
the H-UC building for Melipilla and Sylmar is presented in Figure 17. The use of seismic
isolation in the structure always leads to a signicant reduction in interstory drift and accel(L)
(NL)
= 6:1 and R(NL)
eration, i.e., R(L)
A =4:9 for the Melipilla record; on the
 =5:7, RA =6:1, R
(L)
(L)
(NL)
other hand, R =2:9, RA =3:1, R =3:8 and R(NL)
A =1:48 for the Sylmar record. Therefore,
inelastic acceleration reduction factors are, as should be, smaller than the elastic ones. For
the inelastic case and Sylmar (Figure 17(b)), the 2nd-story drift is reduced from 2.3% to
about 0.6% and the maximum roof acceleration from 1g to 0:7g. By comparing parts (a) and
(b), the traces of interstory drift are strongly aected by the inelastic behavior of the superstructure. Indeed, for the Melipilla and Sylmar records the xed-base structure would end
up with permanent drifts of the order of 0.15% and 0.3%, respectively. Notice also that for
Melipilla the interstory drift demand for the isolated structure is below 0.2% and, hence, only
minor damage, if any, could be expected in non-structural components for this frame structure
subjected to this ground motion. These observations carry over to other ground motions and
to the EF-UC building as well [13].
Furthermore, Figure 18 shows the heightwise proles of displacement, drift, and acceleration
for the H-UC building subjected to the Melipilla and Sylmar ground motions. The inelastic
drift demand occurs in the second story and is about 1% and 2.3% for these two records,
respectively. Seismic isolation in the structure is able to reduce these drifts to about 0.16%
and 0.60% (Figure 17). Analogously, oor accelerations in the superstructure are also reduced
by seismic isolation from maximum values of about 1g for the inelastic case to 0:2g and 0:7g
for Melipilla and Sylmar, respectively. These accelerations increase with increasing values
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

568
Interstory drift (%) Acceleration (g)

J. C. DE LA LLERA ET AL.

0.22064

Interstory drift (%) Acceleration (g)

0.67657

0
0
1

Melipilla NS

1.3367
1.0176

Sylmar N00E

2nd story

0.5

0.17972

0.5

0.68269

10

20

30

40

50

60

2.0052

10

Time (sec)

(a)

(b)

fixed-base
isolated

2.1003

Roof

15

20

Time (sec)

Roof

Melipilla NS

0.2062

fixed-base
isolated

Sylmar N00E
1.0283
0.69426

0
0

1.0147

0.9936

1
2.2706

2nd story

0.5

0.16374

0.5

0.60373

1
2

1
0

10

20

30

40

Time (sec)

50

60

10

15

20

Time (sec)

Figure 17. Response history results for the xed-base and isolated H-UC building considering a linear
and inelastic model of the superstructure: (a) linear model of the superstructure; and (b) inelastic
model of the superstructure.

of Qy ; accelerations in the level immediately below the isolation interface also increase as
a result of this part of the structure being connected to the ground. This implies that the
story below the isolation level might result in larger member forces if the selection of the
OYC is inadequate. Since the yield force value Qy =W = 2:5% and the Melipilla record leads
in the superstructure to peak oor accelerations less than 0:20g, story drifts less than 0.2%,
and isolation displacements below 10 cm, this value was selected in the actual design of both
buildings considered in this investigation. This OYC could also work in principle for the
structure subjected to the Sylmar record; however, the isolation displacements would be over
50 cm, leading to a large gap and a costly isolation system. Because of this, a larger value of
shear capacity of the order of Qy =W = 15% would work better in such a case, reducing the
displacement demand of the isolation system to values below 30cm, peak oor accelerations in
the superstructure less than 0:7g, and story drifts below 0.6%. Consequently, these results show
the sensitivity of an isolation design to the dierent characteristics among ground motions.
Finally, notice that the earthquake responses estimated from linear and non-linear models of
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

569

SEISMIC ISOLATION OF BUILDINGS IN CHILE


5
Linear
Inelastic
LI
LIy 2.5
NLIFy 2.5
LIy 5
NLI
y5
LI
y 15
NLIy 15

Floor level

S
0

(a)

10

20

30

Displacement (cm)

0.5

1.5

Story drift (%)

0.5

1.5

Acceleration (g)

Floor level

S
0

(b)

50

Displacement (cm)

100

Story drift (%)

Acceleration (g)

Figure 18. Response of the H-UC building subjected to the Melipilla and Sylmar records for dierent
yield force and modeling of the superstructure: (a) Melipilla record; and (b) Sylmar record.

the superstructure are quite similar. Hence, these results warrant the use in design of an elastic
model for the superstructure together with an inelastic model of the isolation system.
In order to generalize these results for other ground motions, Figure 19 shows bands for
the ratio of the maxima of the interstory drift, R , and oor acceleration, RA , for the two
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

570

J. C. DE LA LLERA ET AL.

Figure 19. Band of reduction factors for acceleration and interstory drift in the superstructure of buildings
H-UC and EF-UC (rm soil): (a) Chilean ground motions; and (b) impulsive ground motions.

buildings as a function of Qy (2.5%15%). The results presented are for the actual rm
soil conditions of the two structures and the reduction bands obtained correspond to elastic
and inelastic models of the superstructure; results for other soil conditions lead to similar
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

SEISMIC ISOLATION OF BUILDINGS IN CHILE

571

conclusions and are summarized elsewhere [13]. It is apparent that seismic isolation always
leads to considerable reductions in interstory drifts that range between 2 and 7 times. These
reduction factors tend to decrease with increasing values of Qy . They are also more sensitive
to this parameter for Chilean records because the dissipation capacity of the lead cores is more
eective, as with most energy dissipators, only if the structure goes through multiple vibration
cycles. Furthermore, reduction factors for oor accelerations vary between 1.5 and 4 for both
families of ground motions and always decrease as Qy increases. Based on these results, an
OYC value less than 4% seems a good recommendation for Chilean records; however, larger
values will be needed for the design with impulsive ground motions in order to control the
deformation demand of the isolators. Finally, reduction factors based on the actual inelastic
response of the structure R(NL) are usually, but not always, smaller than reduction factors based
on the elastic response of the building R(L) . This should be expected since, for instance, oor
accelerations for the xed-base inelastic superstructure are smaller than those of the elastic
superstructure and, hence, the numerator of RA decreases for the inelastic case.

IMPLICATIONS FOR DESIGN AND PRACTICE


The results presented above have been used in calibrating a proposal for a new seismic isolation building code presented by a group of researchers and practitioners to the Chilean Institute
of Normalization [18]. This proposal is strongly based on the ICBO 2000 [19] and other previous design documents [18]indeed the initial document is a translation of the isolation
provisions of the UBC 97, but includes interesting modications (some in the commentary)
in aspects related to the local seismicity, design spectra denition, damping reduction factors,
isolation modeling, optimal values of shear capacity, experimental isolation results, and other
issues to make the text coherent with the local engineering practice.
An important step was to develop a NewmarkHall-type design spectra (Figure 20) consistent with the response spectra obtained for the 1985 Chile earthquake in all frequency regions.
Since the tendency has been to calibrate design spectra based on pseudo-acceleration response
spectra, reported spectral ordinates for long-period structures (T 2 s) are usually inadequate
since the exponents of the decaying branches of the spectrum do not satisfy the physically
expected trends. The spectra presented in Figure 20 is based on the work of the committee
[18] on the response spectra obtained for the 1985 earthquake [18]. Since for dynamic analysis
the whole spectrum is required, a simplied, say, constant pseudo-velocity spectra may lead
usually to an over-estimation of oor accelerations and forces in the superstructure.
Because of the organic nature of rubber and the inherent variability in the manufacturing process of isolators, testing some bearings and reduced-scale specimens of the elastomer
compound seems strictly necessary. In the authors opinion, based on testing so far over 400
dierent size and shape building and bridge isolators, quality control testing is at least as
important as prototype testing since it enables one to identify aws that are common to the
fabrication process, such as dierences in the properties of rubber compounds coming from
dierent batches of rubber, improper alignment of steel shims in plan and height (rarely seen
from the outside in the undeformed isolator due to the cover), local failures and cracks related to the mould and curing process, and more importantly, benchmarking for monitoring
the evolution of the isolation system in time. Although prototype testing does guarantee the
stability of the design based on testing a couple of isolators, it is fair to say that failing
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

572

J. C. DE LA LLERA ET AL.

Sa

1.4

Soil III

Normalized pseudo acceleration, Sa/g

1.2

2
.VV
T

AA
Soil III

4 2 . D
D
T2

Soil II

0.8

A
Soil II

T a Tb Tc

0.6

Td

Period (s)

Soil I
Soil I

0.4

0.2

0
0

0.5

1.5

2.5

3.5

Period, T (s)

Figure 20. Design spectrum proposed for the Chilean code of isolated structures (seismic
zone 2 and damping ratio  = 0:05).

in shear a reasonably fabricated natural rubber isolator under maximum capable earthquake
conditions is dicult, even after a large number of cycles. Therefore, since the results of
reduced-scale 100 by 100 mm specimens have shown an excellent agreement with the true
behavior of full-scale specimens, and they are cheaper, they should be required for describing
a compound throughout the whole shear deformation range. In any case, the designer must
bear in mind that variabilities in stiness and damping ratio within 15% should be expected
for the same isolator manufacturer and design.
An existent or induced crack in the elastomer will usually increase progressively as deformation increases. Therefore, the initiation of any crack in the elastomer should be delayed by
a good geometric design of the isolator mould, including primarily a well-designed foot and
rubber vents. Sharp corners and other singularities should be avoided since they generate large
local shear strains that initiate unbonding or cracking of the rubber laminates. From the tests
performed in scaled specimens, nal unbonding failures should be precluded since they tend
to occur suddenly as opposed to rupture in the elastomer that leads to a failure that requires
several cycles to propagate throughout the laminate (Figure 6(b)).
Specic damping reduction factors for non-impulsive ground motions were derived for this
code proposal. The reduction factors due to damping B for non-impulsive ground motions
may be computed for the three dierent soil conditions by (Table III):
B(; T ) =

1
Bo (Bo 1) exp(a( 0:05)T )

(7)

where Bo = 2(1 + )=(1 + 14:68 0:865 ); a is a coecient specied in Table III; and  is the
damping ratio. These reduction factors were computed by using ground motions compatible
with the design spectrum associated with the three dierent soil conditions. Because of the
larger number of strong motion cycles present in non-impulsive ground motions relative to
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

SEISMIC ISOLATION OF BUILDINGS IN CHILE

573

Table III. Coecient a (Equation (7)).




Soil I

Soil II

Soil III

0.10
0.15
0.20
0.25
0.30
0.50

396.9
180.7
117.9
94.0
68.5
36.9

293.1
124.6
76.1
54.3
42.0
22.2

224.5
98.0
57.1
39.6
30.4
16.1

impulsive motions, these factors are, as should be, larger than those specied in the ICBO
2000 code [19].
Although not described herein for reasons of brevity, architectonic aspects in seismically
isolated structures are of great importance and should not be overlooked during the structural
design phase. It is precisely this point that makes isolation projects special in the sense that
they require a stronger interaction between the architect and structural engineer; a poor design
of these architectonic details may counter balance the merits of seismic isolation in terms of
structural response. Although every structure has specic architectonic details, there are certain
typologies of solutions that tend to repeat in order to accommodate the lateral displacement
between the xed-to-ground structures and components and the superstructure. The interested
reader may nd a collection of low-cost joint solutions and details for the surrounding gap,
facades, staircases, ceilings, handrails, elevators, piping, and other situations elsewhere [17].
As stated before, the cost of the bare structure including isolation was US$ 115 and US$
125 per square meter for the H-UC and EF-UC buildings, respectively. This value is within
the usual range for conventional structures and thus makes isolated structures very competitive.
The reader may wonder about the direct cost of these structures without isolation. Although
estimates for both buildings were generated in terms of preliminary designs to compare between the isolated and conventional solutions, the numbers are always questionable since a
nal project would be required for estimating real construction costs in each case. However,
it is the authors opinion that direct costs between the two solutions ended up being quite
similar mainly because no extra slab was required in the H-UC building, all shear walls were
eliminated, and the building has no construction joint, which is required for structures of
such dimensions in plan. For the EF-UC building, extra costs due to isolation were counterbalanced by: (i) the elimination of the construction joint in the structure (ii) a more economic
design of the foundation system; and (iii) a lower density of shear walls and reinforcement
in columns than for conventional structures of this kind in the country.

CONCLUSIONS
This article summarizes the work done by the authors over the past six years in Chile
with seismically isolated structures. The work has included an extensive testing program,
design, and implementation of elastomeric isolation in two reinforced concrete structures of
quite dierent characteristics in structure and foundation soil. Currently, results of this applied investigation will be used in the new Military Hospital, a structure with 164 elastomeric isolators, the largest isolated structure in South America. It has been shown by this
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

574

J. C. DE LA LLERA ET AL.

structural implementation that seismic isolation is a competitive earthquake-resistant construction technique in this country and it is expected to expand rapidly in the years to come.
Furthermore, a new building code for seismically isolated structures is currently approved and
will probably be available for the profession in November 2003. This technology is being
transferred to the profession through seismic isolation courses taught in the Civil Engineering curricula of the best universities in the country as well as through the interaction in
projects between the dierent universities and the engineering profession, such as with the
Military Hospital. It is expected that the experience developed in this country may be of use to
other countries in the region which share similar seismic characteristics and social=economic
environments.
ACKNOWLEDGEMENTS

This investigation has been supported by the Chilean National Fund for Research and Technology,
FONDECYT under Grant # 1020774. Part of the research was also supported by the Fund for Foment
and Technology, FONDEF under Grant #D96I1008. The authors are grateful for this support.
REFERENCES
1. ATC 17-1. Seminar on Seismic Isolation, Passive Energy Dissipation, and Active Control. Applied Technology
Council. Vol. 1, San Francisco, 1993;1435.
2. Anderson TL. Earthquake Engineering Research Institute. Seismic isolation: from idea to reality. Earthquake
Spectra 1990; 6(2):
3. Raufaste N. Survey Report on Framing of the Guidelines for Technological Development of Base-isolation
Systems for Buildings. Building and Fire Research Laboratory, National Institute of Standards and Technology,
April 1992.
4. Holmes T, Somers P. Northridge earthquake of 17 January 1994. Reconnaissance report. Earthquake Spectra
1996; 11(Supplement C):243251.
5. Fujita T. Demonstration of eectiveness of seismic isolation in the HanshinAwaji earthquake and progress of
applications of base-isolated buildings. Report on the 17 January 1995 Kobe Earthquake, Japan. Institute of
Industrial Science, University of Tokyo, Japan; 197216.
6. Moroni MO, Sarrazn M, Borocheck R. Experiments on a base-isolated building in Santiago, Chile. Engineering
Structures 1998; 20:720 725.
7. Almazan JL. Accidental and natural torsion in isolated structures with the FPS systems. Ph.D. Thesis, Faculty
of Engineering. Universidad Catolica de Chile, 2001. In Spanish.
8. De la Llera JC, Esguerra C, Almazan JL. Earthquake behavior of structures with copper energy dissipators.
Earthquake Engineering and Structural Dynamics 2004; 33:329358.
9. Park JG, Otsuka H. Optimal yield level of bilinear seismic isolation devices. Earthquake Engineering and
Structural Dynamics 1999; 28:941955.
10. Ramallo JC, Johnson EA, Spencer BF. Smart base isolation systems. Journal of Engineering Mechanics
(ASCE) 2002; 128(10):10881099.
11. Kelly JM, Leitmann G, Soldatos AG. Robust control of base-isolated structures under earthquake excitation.
Journal of Optimization Theory and Applications 1987; 53:159180.
12. Inaudi JA, De la Llera JC. SAT-Lab: Structural Analysis Toolbox. www.sat-lab.com, 2002.
13. Leigh P. Aislacion ssmica de estructuras con sistemas hbridos de goma y aleaciones: analisis, experimentacion,
y diseno. Master Thesis, Faculty of Engineering. Universidad Catolica de Chile, 2002.
14. dSPACE. Solutions for Control: Catalog and Training Program 1999. info@dspaceinc.com, Germany, 1999.
15. Hwang JS, Wu JD, Pan TC, Yang G. A mathematical hysteretic model for elastomeric isolation bearings.
Earthquake Engineering and Structural Dynamics 2002; 31:771789.
16. Pan TC, Yang G. Nonlinear analysis of base-isolated MDOF structures. Proceedings of the 11th World
Conference on Earthquake Engineering, Paper No 1534, Mexico, 1996.
17. SIRVE. Innovative Systems for Vibration Reduction. www.sirve.cl, 2002.
18. Asociacion Chilena de Sismologa e Ingeniera Ssmica. Proposal for a Chilean code of seismically isolated
structures. Grupo de Proteccion Ssmica, ACHISINA, March 2001. In Spanish.
19. ICBO International Building Code. Seismically isolated structures: Section 1623. International Code Council,
Inc. 2000.
Copyright ? 2004 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2004; 33:543574

You might also like