You are on page 1of 20

STRUCTURAL CONTROL AND HEALTH MONITORING

Struct. Control Health Monit. 2007; 14:895–914


Published online 26 October 2006 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/stc.188

Self-centring capacity of seismic isolation systems

Stefano Berton1,*,y, S. Infanti2, M. G. Castellano2 and H. Hikosaka3,z


1
ANTERIS S.r.l. Societá di Ingegneria, via F.lli Bandiera 7/7, Marcon, Venezia 30020, Italy
2
FIP INDUSTRIALE S.p.A., Via Scapacchió 41, Selvazzano Dentro, Padova I-35030, Italy
3
Emeritus Professor, Department of Civil Engineering, Kyushu University, Hakozaki 6-10-1,
Higashi-ku 812-8581, Fukuoka, Japan

SUMMARY

The ability of seismic isolators to develop restoring forces, also referred as self-centring capacity, is
commonly associated with the system ability to prevent excessively large unit deformations and possible
isolator failure. This concept is emphasized in some of the current standards or guide specifications for the
design of seismic isolated structures that usually impose minimum values of the isolator restoring force
capacity. In this paper a numerical study is performed to investigate the influence of the parameters that
characterize the model of a structure–isolator system on its self-centring capacity. The system is idealized as a
rigid mass mounted on a single isolator in which a bilinear force–displacement relation, which approximates
the nearly bilinear behaviour exhibited by most of the currently available devices, characterizes the isolator
response. Results from this study provide insights on the merit of current guide specifications and on the
overall isolation behaviour of the models. Copyright # 2006 John Wiley & Sons, Ltd.

KEY WORDS: base isolation; earthquake response; structural control; bilinear oscillator; self-centring

1. INTRODUCTION

During the last decades a significant number of buildings and bridges located in earthquake-
prone regions have been seismically protected through seismic isolation [1]. The basic idea of
this approach consists of interposing between the isolated structure and the foundation/support
system special devices called isolators that have the ability to reduce the level of the inertial
forces transmitted to the structure during a seismic event. This seismic force reduction comes at

*Correspondence to: Stefano Berton, ANTERIS S.r.l. Societá di Ingegneria, via F.lli Bandiera 7/7, Marcon, Venezia
30020, Italy.
y
E-mail: ss.berton@tiscali.it
z
Emeritus Professor.

Contract/grant sponsor: Japanese Society of the Promotion of Science

Received 15 September 2005


Revised 6 May 2006
Copyright # 2006 John Wiley & Sons, Ltd. Accepted 20 August 2006
896 S. BERTON ET AL.

the expense of the development of lateral deformations of the isolator bearings, which are, in
general, designed to possess sufficiently high lateral ductility.
The type of isolators that are used most for seismic protection of buildings are multilayered
laminated rubber bearings (RBs) [1, 2]. These bearings consist of pads of rubber compounds
that are reinforced with thin steel shims to increase the bearing axial stiffness. Usually, a high-
damping rubber compound is used to provide the isolation system with an equivalent viscous
damping ratio up to 15%. When a lead core is added, to further increase the isolator damping
ratio, these bearings are referred as lead rubber bearings (LRBs). In addition to isolate the
structure, RBs and LRBs posses a fairly good capacity of resetting themselves toward their
original position due to the intrinsic elastic characteristics of the rubber compounds. This
property is an important characteristic of any isolation system, usually referred as self-centring
or lateral restoring capacity. In addition to RBs and LRBs, other types of devices are employed
for seismic isolation of buildings and other civil structures such as bridges. These devices include
sliding isolators based on pure friction (PF) and hysteretic isolators. Sliding isolators dissipate
energy through the friction forces generated by the sliding components of the device. An
example of slider isolator is the friction pendulum system (FPS) that utilizes spherical sliding
surfaces [3]. Hysteretic isolation systems generally combine the use of low-friction sliding
bearings with special mild steel components that are designed to deform and yield, as the device
is activated [4, 5]. Due to their highly nonlinear behaviour and their energy dissipation capacity,
sliders and hysteretic isolators are able to limit efficiently the seismic force transmitted to the
isolated structures. However, in general, the higher the energy dissipation capacity the lower is
the self-centring ability of these systems.
Most of the current isolation devices exhibit a nearly bilinear behaviour [6]. Figures 1(a)–(d)
show a comparison between typical force–displacement responses of a bilinear model, a LRB,
an hysteretic isolator and a FPS, respectively. An ideal bilinear model is fully characterized by
three of the five parameters shown in Figure 1(a). The three parameters traditionally used for a
bilinear approximation of an isolation system are the strength force Q; the post-yield stiffness kp ;

Figure 1. Force–displacement responses: (a) idealized bilinear model; (b) lead-rubber isolator;
(c) hysteretic bearing; and (d) friction pendulum system.

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
SEISMIC ISOLATION SYSTEMS 897

and the yield displacement uy : The system strength Q is usually expressed as a portion of the
tributary structural weight W supported by the isolator. Typical Q values range between 3 and
12% of W; while typical values of the yielding displacement range from 0.2 to 0.5 mm for sliding
bearings (e.g. PF and FPS) [6, 7] to 10–50 mm, for LRB [2] and hysteretic bearings [4]. While the
initial stiffness has in general a small effect on the behaviour of the isolation system subjected to
moderate and strong ground motions, the post-yield stiffness is an important parameter that is
directly related to the self-centring capacity of the device.
Current codes for the design of base isolation structures, such as the ENV 1998-2:1994
Eurocode 8 [8], and the 1999 American Association of State Highway and Transportation Officials
(AASHTO) [9] impose minimum values of the restoring force capacity of the isolation systems.
The philosophy behind these requirements aims to limit cumulative build-ups of isolator
deformations that commonly associate with the system lack of restoring capacity. Large isolator
deformation demands associated with this cumulative effect have been observed in previous
studies on isolated structures subjected to near-fault-type ground motions [10]. In this paper a
parametric study is performed to evaluate the effect of the parameters that characterize the
model on the self-centring capacity of most currently used seismic isolation devices. The study
consists of performing a series of time history analyses using a set of representative input ground
motions in which the isolator–structure system is modelled as a single-degree-of-freedom
(SDOF) bilinear oscillator. Results from this study provide insights on the merit of current
guide specifications and on the overall isolation behaviour of the models. Although the
approach used is general, and therefore valid for any kind of isolated structures, particular
emphasis is given to seismic isolated bridges where isolators with highly nonlinear characteristics
have been used in several recent applications.

2. SELF-CENTRING REQUIREMENTS IN CURRENT GUIDE SPECIFICATIONS

In recent years, several countries have started to include in their seismic codes regulations to
provide guidance to the design of structures that incorporate seismic isolators or other seismic
protection systems. Examples of such codes are the Japanese 1996 Design Specifications for
Highway Bridges, Part V: Seismic Design code [11], the European ENV 1998-2:1994 Eurocode 8
[8], and in U.S.A., the 1999 American Association of State Highway and Transportation Officials:
Guide Specifications for Seismic Isolation Design [9] for the design of isolated bridges, and the
2003 International Building Code (IBC) [12] for building applications. Although in each country
the norms differ to account for local differences in seismic hazard and local construction
practices, in general they share the same main design philosophy. Furthermore, the
specifications for base isolated buildings included in the 1997 Uniform Building Code (UBC)
[13] (that includes essentially the same provisions for seismically isolated buildings of the more
recent 2003 IBC code) and the AASHTO for bridge applications are particularly important in
that those two codes have been used often as references for the development of analogous codes
outside U.S.A.
Most of the aforementioned norms include specific requirements on the isolation system that
are directly associated with the system ability to self-centre. These requirements usually impose
restrictions on the isolator restoring force capacity with the declared goal of preventing
cumulative build-ups of isolator deformations. For example, the 1999 AASHTO specifications
impose self-centring requirements to the isolation systems for bridge applications. In particular,

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
898 S. BERTON ET AL.

the 1999 AASHTO distinguishes between constant and non-constant restoring force isolators.
In the first case, which includes the majority of current devices, two criteria must be satisfied.
The first imposes that the restoring force developed at the design displacement, di ; is 0:0125W
1
greater (80 of the tributary weight W) than the corresponding value at 50% of di : The second
requirement imposes that the period corresponding to the tangent stiffness of the system at any
displacement value, must be smaller than 6 s. If DFi is the difference of the restoring force at di
and at 50% of di ; and the system response is idealized as bilinear, then the two AASHTO
criteria can be expressed by the following inequalities:
DFi 50:0125W ð1Þ

Tp 46 s ð2Þ
pffiffiffiffiffiffiffiffiffiffiffi
where Tp is the period corresponding to the post-yield stiffness kp (i.e. Tp ¼ 2p m=kp ).
The Eurocode 8 norms for the seismic design of bridges [8], also include a restoring force
criterion similar to the first criterion of the 1999 AASHTO (DFi 50:025W). Recently, the final
draft of the latest version of the European norms for the seismic design of bridges, Eurocode
8}Design of structures for earthquake resistance}Part 2: Bridges prEN 1998-2:2005 (EC8) [14]
has been released to the public. In this new document, when approved will supersede the current
ENV 1998-2:1994, new self-centring requirements have been included. Similar to the 1999
AASHTO, the new version of the EC8 norms requires two criteria to be simultaneously satisfied
by any isolation system; these new criteria are expressed by the following inequalities:
DFm 5dw Wdrm =dm ð3Þ

drm 4dm  dd da;max ð4Þ


where dm is the isolator maximum displacement capacity, DFm is the difference of the restoring
forces evaluated at dm and 50% of dm ; drm is the residual displacement of the isolator
corresponding to the isolator displacement capacity dm ; da;max is the maximum displacement
derived from the seismic analysis (eventually corrected by a coefficient of safety), and dw and dd
are two numerical coefficients. Note that the requirements expressed by Equations (1) and (3)
are similar, but the two second criteria for the AASHTO (i.e. Tp 46 s) and the EC8
specifications (Equation (4)) are quite different and uncorrelated. In Section 4 a comparison of
the self-centring requirements of the AASHTO and the new EC8 will be presented for several
different ground motions and several values of the parameters that characterize the isolation
system model.

3. NUMERICAL MODEL

3.1. Bilinear force–deformation model


Bilinear models are widely used to simulate the seismic responses of isolation systems [1, 2].
Figure 2 shows a schematic representation of a SDOF isolator model, together with its idealized
force–displacement loop for peak inelastic deformations of umax : A bilinear model response is
fully characterized by three independent parameters. Traditionally, these parameters are the
system strength Q; the yielding displacement uy ; and the post-yield stiffness kp : The yielding
force, Fy ; can also be used instead of Q; where the relation between these two parameters can be

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
SEISMIC ISOLATION SYSTEMS 899

(a) (b)

Figure 2. (a) SDOF model of an isolation system; and (b) force–displacement loop of a bilinear model.

written as
Q ¼ ð1  aÞFy ð5Þ
where a ¼ kp =ke ; and ke is the initial elastic stiffness. For most practical applications, the initial
stiffness ke is one or more orders of magnitude bigger than kp ; and therefore, Fy and Q are
usually almost equivalent.
Current design guidelines, such as the European Eurocode 8, allow under certain conditions
the use of a simplified linear procedure for the analysis of isolated structures. In that case, the
bilinear system is replaced by an equivalent linear model having an effective stiffness, keff ; and
effective damping ratio xeff : The effective stiffness is defined as the secant stiffness evaluated at
the maximum displacement umax (Figure 2(b)) and can be expressed as
Q
keff ¼ þ kp ð6Þ
umax
The effective damping ratio, xeff ; is calculated using the general definition of damping ratio, as a
portion of the ratio between the energy dissipated per cycle, EDC (Figure 2(b)), and the elastic
strain energy at the maximum displacement [15]:
1 EDC 2Qðumax  uy Þ
xeff ¼ ¼ ð7Þ
4p ES pkeff u2max

ffiffiffiffiffiffiffiffiffiffiffiffiffi period of the isolator Teff ; which is derived from


Another important parameter is thepeffective
the effective stiffness (i.e. Teff ¼ 2p m=keff ). The effective damping ratio and effective period
are two useful parameters that provide basic and practical information on the isolator dynamics
characteristics.

3.2. Equation of motion


A SDOF bilinear oscillator is used to model the dynamic behaviour of a seismic isolated
structure. The system is idealized as a rigid mass mounted on a single isolator (Figure 2(a)). The
equation of motion for the SDOF oscillator with mass m; and no damping, can be written as
FI ðtÞ
u. ðtÞ þ ¼ .ug ðtÞ ð8Þ
m

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
900 S. BERTON ET AL.

where uðtÞ is the relative displacement of the mass with respect to the ground, u. g ðtÞ is the input
ground acceleration (i.e. the double overdot represents the second derivative with respect to
time) and FI ðtÞ is the total force developed by the bilinear spring, hereafter referred as the
isolator force. The isolator force can be expressed as
FI ðtÞ ¼ kp uðtÞ þ QzðtÞ ð9Þ
where zðtÞ is a dimensionless quantity that varies from 1 to 1 and controls the portion of the
strength Q that contributes to the value of the isolator force. The quantity zðtÞ satisfies the
following nonlinear first-order differential equation [16]:
uy z’ þ b’ujzjn þ gj’ujzjzjn1  A’u ¼ 0 ð10Þ
where A; g; b and n are dimensionless parameters that control the shape of the hysteretic loop,
and therefore depend on the isolator characteristics. Due to the nonlinearity introduced by the
force–deformation behaviour of the bilinear model, the equation of motion expressed by
Equation (8) is integrated numerically in incremental form over a small time interval Dt using
the Newmark linear acceleration method.

4. PARAMETRIC ANALYSIS

4.1. Selection of parameters


The three independent parameters selected to characterize the bilinear model of an isolation
system are the post-yield period Tp ; the yield force Fy ; and the yield displacement uy : Tp is the
most important among the three parameters. It is traditionally considered directly associated
with the system restoring capacity and most of the current norms include criteria that are based
directly or indirectly on this parameter (Section 2). Several values within the two theoretical
limits of Tp ¼ Te (i.e. linear elastic system) and Tp ¼ 1 (perfect elastoplastic system) are used
for each analysis case.
Two values of yielding force are used for this parametric study through the two values of
r ¼ 5 and 10%; in which r is the ratio Fy =W: Similar to the system strength Q; the yielding force
is defined as a portion of the vertical load W: Note that, these two selected values of r; and
therefore of Fy ; are on the borders of typical range values for the majority of practical
applications.
Three different values of yielding displacements are used in this study: uy ¼ 0:5; 10 and
50 mm. The value of yielding displacement uy ¼ 0:5 mm was selected to represent pure slider
isolators or sliders with spherical surface such as FPSs (Figure 1(d)). Previous studies have
shown that values between 0.2 and 0.5 mm are appropriate to represent the small deformations
of the bearing components in the no-slip conditions [6, 7]. The other two values of the yielding
displacements, uy ¼ 10 and 50 mm; are selected because they represent typical yielding
deformations of other kind of isolators such as LRBs (Figure 1(b)) [17] and hysteretic bearings
(Figure 1(c)) [5, 6].

4.2. Input ground motions


A total of five records from five different recent earthquakes are used in this study. These
earthquakes are: the 1989 Loma Prieta earthquake, the 1994 Northridge earthquake, the 1995

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
SEISMIC ISOLATION SYSTEMS 901

Table I. Ground motion data.


Fault Soil PGA PGV PGD
Earthquake name distance typen amax vmax dmax
Record no. and station Component Mw (km) (g) (m/s) (m)
1 Kobe-KJMA 08 6.9 0.6 B 0.821 0.813 0.177
2 Northridge Rinaldi 2288 6.7 7.1 C 0.838 1.661 0.288
3 Loma Prieta Gilroy #2 908 6.9 11.2 C 0.322 0.391 0.121
4 Chi-Chi Taiwan TCU129 2708 7.6 1.18 C 1.010 0.60 0.502
5 Denali Alaska pump station #10 Fault normal 7.9 3.0 C 0.290 1.145 1.237
n
USGS soil classification.

Hyogo-Ken Nanbu earthquake (Kobe), the 1999 Chi-Chi Taiwan earthquake and the most
recent, the 2002 Denali Alaska earthquake. The parameters that characterize these ground
motions are listed in Table I. All records were recorded on soil type B or C based on the US
Geological Survey (USGS) soil classification, and were generated by seismic events having
moment magnitude MW ranging from 6.7 to 7.9. The peak ground accelerations (PGA) values
for these six ground motions varies from 0:322g for the East–West component of the Gilroy#2
Loma Prieta earthquake to 1:01g for the East–West component of the Chi-Chi Taiwan
earthquake recorded at the TCU129 receiving station. Great variability is also included in the
peak ground velocity (PGV) values that range from 0.391 m/s for the Loma Prieta record to the
1.661 m/s of the Northridge record. All the ground motions have been recorded at a relatively
short distance from the generating fault ranging from 0.6 to 11.2 km. However, only two of
these records present near-fault pulse-like characteristics, which include high PGV values. These
records are the Northridge-Rinaldi record and the normal components of the 2002 Denali
Alaska earthquake. Ground motion data of the unrotated Denali Alaska records were obtained
through a ftp link available in Reference [18] while all the other records were obtained from the
PEER Strong Ground Motion Database [19].

4.3. Parametric study approach

4.3.1. Model performances. For each input ground motion the two selected values of the
parameter r (5 and 10%), and the three of yielding displacement uy (0.5, 10 and 50 mm), lead to
six different analysis combinations. For each of these combinations, hereafter referred as an
analysis case, the post-yield period is varied from Tp ¼ Te to Tp ¼ 1: An average of 20 or more
values of Tp within the two theoretical limits are used so that an excess of 120 nonlinear time
history analyses are carried out for each input ground motion. Note that, because the yielding
force is expressed as a portion of the system mass weight, the results are completely independent
of the mass value which is set to m ¼ 1 kg throughout this study. The analysis outputs include
the model responses in terms of displacements uðtÞ; isolator force FI ðtÞ and isolator offset oðtÞ:
The isolator offset is defined as the inelastic portion of the isolator deformation, whose value at
the end of the simulation provides the model residual displacement. For each ground motion
and for each analysis case, several response peak values and other quantities of interest are
recorded. For example, the peak absolute value of the isolator displacement, uðtÞ; is calculated
and recorded into the variable ‘up’. Similarly, the variables ‘op’ and ‘frp’ represent the peak

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
902 S. BERTON ET AL.

(a)

(b) (c)

Figure 3. Example of analysis outputs: (a) plots of the normalized values of peak displacements, peak
offset, residual offset, peak isolator force and hysteretic energy; (b) model displacement time histories for
Tp ¼ Tpn ¼ 3:0 s; and (c) model force–displacement response for Tp ¼ Tpn ¼ 3:0 s:

absolute values of the isolator offset oðtÞ and of the isolator force FI ðtÞ; respectively. Other
quantities of interest recorded in each analysis include the model permanent deformation at the
end of the simulation, ‘or’, and the system total hysteretic energy Eh :
An example of analysis outputs is shown in Figure 3. The record no. 2 (i.e. Northridge
earthquake) is used for this example. The values of the model yielding force and yield
displacement are Fy ¼ 0:9804 N (i.e. r ¼ 10%) and uy ¼ 0:2 mm; respectively. More than 40
different post-yield periods are considered, ranging from Tp ¼ Te ¼ 0:09 s to Tp ¼ 1; mostly
distributed within Tp ¼ 10 s where the results are of interest for practical applications. The

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
SEISMIC ISOLATION SYSTEMS 903

model responses are shown in Figure 3(a), which depicts the plots of the normalized values of
the peak absolute displacements ‘up’, the peak absolute offsets ‘op’, the residual offset values
‘or’, the peak isolator force ‘frp’, and the total hysteretic energy Eh : All the variables
% which is the maximum
representing displacements (i.e. ‘up’, ‘op’ and ‘or’) are normalized by ‘up’
value of ‘up’, while, the peak isolator force and hysteretic energy are normalized in this example
by their maximum values, respectively. Of particular interest is the post-yield period Tpn ; which
correspond to the minimum of the ‘up’ curve within the practical limits of post-yield periods,
says within 1.5–10 s. Therefore, Tpn represents the optimal value of post-yield period
corresponding to the model that produces the smallest value of peak deformation. Hereafter,
the superscript ‘n ’ will be used to indicate the analysis result quantities associated with the
optimal post-yield period Tpn : For the example of analysis shown in Figure 3(a) the optimal
value of the post-yield period is 3.0 s.
Although the model performance in terms of deformations is of primary concern, it is also
important to verify the model ability to dissipate energy and its ability to effectively isolate the
system mass. The trend of the hysteretic energy as well as the trend of the peak isolator force
shown in Figure 3(a) are typical for all the cases analysed in this study. The hysteretic energy
always increases from its zero value at Tp ¼ Te to its final positive value at Tp ¼ 1; reaching its
maximum usually between Tp ¼ 225 s: The plots of the peak isolator force, after reaching its
maximum at elastic or close to elastic conditions, decreases monotonically and quite rapidly
toward its limit value at Tp ¼ 1 of frp ¼ Fy ¼ rmg: The variable ‘frp’ that represents the peak
isolator force values is important because it can be used to graphically verify the ability of the
model to reduce the force transmitted to the system mass. Figure 3(a) includes two horizontal
dashed lines representing the constant normalized values of mPGA, and the limit value Fy : The
points of the ‘frp’ curve below the mPGA line correspond to models which effectively isolate the
system mass by transmitting a force smaller than the tributary weight times PGA. On the
contrary, points above the line correspond to models that do not provide any isolation but
amplify the effect of the ground motion, and therefore are not of interest.
Figures 3(b) and (c) show the deformation time histories and the force–displacement response
for the optimal case Tpn ¼ 3:0 s: Both figures include the peak values of each quantity used for
n
the construction of the plots shown in Figure 3(a). Figure 3(b) includes the effective period Teff
n
and effective damping ratio xeff defined by Equations (6) and (7), respectively, determined using
as isolator design displacement the maximum value derived from the numerical simulation.

4.3.2. Comparison of code specification criteria. In the case of a bilinear idealization of the
isolator response, the self-centring criteria of the AASHTO and the EC8 norms can be written
as a set of inequalities that are direct or indirect functions of the design displacement di
(AASHTO, Equation (1)), of the isolator post-yield period Tp (AASHTO Equation (2)), and the
isolator displacement capacity dm (EC8, Equations (3) and (4)). In the AASHTO specifications
the design displacement di is defined as the lateral seismic displacement derived from the analysis
and, therefore, it corresponds to the peak displacement ‘up’. The displacement capacity dm ; used
in the EC8 specifications, is defined as the maximum design deformation that the isolator can
sustain, which includes the peak displacement derived from the seismic analysis, plus other
components such as the displacements due to the long-term effects and the thermal loadings.
For the purpose of comparing the self-centring criteria, the total design displacement dm is
calculated as the peak isolator displacement ‘up’ augmented by a constant lumped value that
includes temperature and long-term deformations, selected to be 100 mm. Because of this, di (i.e.

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
904 S. BERTON ET AL.

(a) (b)
Figure 4. Peak absolute displacement ‘up’ and self-centring norm specification functions for the models
with uy ¼ 0:2 mm and r ¼ 10% subjected to the North–West component of the Northridge earthquake
(Rinaldi): (a) AASHTO criteria; and (b) EC8 criteria.

‘up’) and dm (i.e. up þ 100 mm) for a given analysis can be associated with the corresponding
post-yield period Tp : Therefore, the self-centring criteria of these two norms can be plotted as
function of Tp and the corresponding limits of the post-yield periods graphically determined.
For example, in the case of the AASHTO norm, the first criterion expressed by the inequality
of Equation (1) can simply be written as F ¼ DFi  0:0125W50; and the function F plotted in
terms of the variable Tp : This can be done because the difference in restoring force DFi for a
given triplet of parameters that characterizes the model is only function of di ; for the
aforementioned reasons, and di is function of Tp : Then the intersection of the function F with
the zero constant value line provides the limit value of Tp below which this first AASHTO
criterion is satisfied. Since the second criterion is simply Tp 46 s; the smallest between 6 s and the
previous calculated value is the post-yield period limit imposed by the AASHTO norms. Figure
4(a) shows the graphical representation of the AASHTO criteria together with the model peak
displacement response ‘up’, for the example analysis case subjected to the record no. 2, whose
other output results are also shown in Figure 3. The first AASHTO criterion is represented by
the dashed–dotted line intersecting the zero constant value at Tp ¼ 7:66 s; which is bigger than
the constant value of 6 s corresponding to the second criterion. Therefore, Tp ¼ 6 s is the limit
imposed by the AASHTO norms for this case.
Figure 4(b) shows the corresponding representation of the EC8 criteria for the same example
case. As shown in the figure, the first EC8 criterion (dashed–dotted line) provides a limit of
Tp ¼ 7:88 s; where in the inequality of Equation (3), the recommended value of the numeric
coefficient dw ¼ 0:015 was used and the residual displacement drm was calculated from the total
design displacement dm (i.e. up þ 100 mm). The second criterion expressed by Equation (4)
(dotted line in Figure 4(b)) results in a smaller value of Tp ¼ 3:28 s; which is, therefore, the post-
yield period limit imposed by the EC8 norms for this particular case. Note that for the
application of the second EC8 criterion, the displacement da;max is set to the corresponding peak
displacement ‘up’ derived from the numerical simulation and the recommended value of
numerical coefficient dd ¼ 0:5 was used. The two derived post-yield periods show that the EC8
norms imposes a more restrictive limit for the example considered. The comparison of the
AASHTO and the EC8 norms for all the cases considered in this study, using the approach and
the assumptions discussed in this subsection, will be presented in the next part of this paper
together with the main results of this numerical study.

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
SEISMIC ISOLATION SYSTEMS 905

4.4. Analysis results


All the analysis results are represented in five group figures, one for each of the five input ground
motions considered in this study (Figures 5–9). Each of these figures consists of a 3  2 subfigure
array representing the results of the six combinations of the parameters Fy and uy : Similar to the
example shown in Figure 3(a), the normalized values of the peak isolator displacement ‘up’, the
residual displacement ‘or’, and the peak isolator force ‘frp’ are plotted as functions of Tp : Only
the results within 2 and 9 s are shown in the figures since these values include the majority of
practical applications. The five group figures are organized in a way such that each column of
the 3  2 array represents the results of the three analysis cases having the same yielding force,
while each row represents the two cases with same yielding displacement. All the displacement
%
results (i.e. the ‘up’ and the ‘or’ curves) are normalized and represented as a percentage of ‘up’,
which is the maximum value of ‘up’ among all six analysis cases, while the ‘frp’ plots are
normalized by mPGA. The figures also include several response values at Tp ¼ Tpn such as the
peak displacements ‘upn ’, the peak isolator force ‘frpn ’ and the value of the optimal post-yield
period itself, Tpn : In addition, the effective periods and the effective damping ratios, evaluated for
the cases Tp ¼ Tpn ; are included in two small boxes located on the subfigure upper-left and
upper-right corner, respectively. The AASHTO and the EC8 limits, derived using the
assumptions and the procedure presented in Section 4.3.2, are represented by two vertical
lines positioned at the corresponding values of post-yield period.

Figure 5. Plots of the peak deformations and peak isolator force for the North–South component of
the 1995 Kobe earthquake (record no. 1).

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
906 S. BERTON ET AL.

Figure 6. Plots of the peak deformations and peak isolator force for the South–West component of
the 1994 Northridge earthquake (record no. 2).

Figure 5 shows the results of record no. 1 ground motion (Kobe). All the displacement values
are represented as a percentage of up % ¼ 549:6 mm; which is the maximum ‘up’ value recorded
for the case with r ¼ 5%; uy ¼ 50 mm and Tp ¼ 2 s (bottom-left subfigure). Note that up % (i.e.
100%) is represented in each subfigure by an horizontal dotted line. All the ‘frp’ curves are
normalized by mPGA (PGA ¼ 0:821g), which is also represented by the 100% dotted line when
used in conjunction with the ‘frp’ curves. The other horizontal line shown in each subfigure,
represents the limit of Fy ¼ rmg to which the ‘frp’ curves tend as Tp approaches infinity.
Although the plots of the results shown in Figure 5 follow similar trends and have similar shapes
for a given value of r; the optimal post-yield periods can vary considerably. In fact, for the three
r ¼ 5% cases, the optimal Tpn values are 5.5, 9.0 s and 1 for uy ¼ 0:5; 10 and 50 mm,
respectively. A different pattern is observed among the three r ¼ 10% cases in which the optimal
post-yield periods are similar, ranging from 2.5 to 3.25 s. The corresponding values of model
peak deformation ‘upn ’ range from 152:4 mm (r ¼ 10% and uy ¼ 0:5 mm case) to 286.1 mm
(r ¼ 10% and uy ¼ 50 mm case). The ability of the models to reduce the force transmitted to the
mass is quite good in all six optimal condition cases, as indicated by the relatively small values of
‘frpn ’ ranging from 0.05 (i.e. the theoretical limit for the r ¼ 5% cases) to 0:208mg: The other
plot shown in each subfigure represents the residual displacements ‘or’, which for this particular
ground motion and within the selected range of Tp are always a small portion of the peak
displacement. The effective periods and effective damping ratios at Tp ¼ Tpn included in the two
small boxes located in upper corners of each subfigure, reflect the variability of the Tpn values,

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
SEISMIC ISOLATION SYSTEMS 907

Figure 7. Plots of the peak deformations and peak isolator force for the East–West component of
the 1989 Loma Prieta earthquake (record no. 3).

ranging from 1.91 to 4.19 s for the effective periods, and from 22.4 to 50.4% for the effective
damping ratio. Regarding the restoring force requirements we can see that the application of the
AASHTO and the EC8 specifications impose different maximum values of post-yield periods.
The EC8 specifications, among the six cases shown in Figure 5, impose post-yield periods
ranging from 2.7 to 5.0 s, while the AASHTO imposes always a maximum value of 6 s except for
one case (r ¼ 5% and uy ¼ 0:5 mm case) in which the limit is 5.5 s. As a first general observation
we can see that the maximum values of post-yield period imposed by these norms are in some
cases smaller than the optimal Tpn (particularly for the EC8), prohibiting the use of systems that
produce not only the best performances in terms of deformations, but also significant lower
levels of seismic induced forces.
Figure 6 depicts the analysis responses of the model subjected to record #2 ground motion
(Northridge). This record, in addition to high values of PGA ð0:838gÞ; presents characteristics of
near-fault ground motion. Among the six analysis cases, the values of ‘upn ’ vary from 327.3 mm
(middle-right subfigure) to 533.8 mm (bottom-left subfigure). The displacement normalizing
% ¼ 872:1 mm which is the maximum value produced by the model for the case r ¼ 5%
value is up
and uy ¼ 50 mm: The general trends of the response plots for all the analysis cases are quite
similar together with the six values of Tpn that range between 2.5 and 3.25 s. The peak isolator
force values, ‘frpn ’, range from 0.221 to 0:362mg: It is interesting to notice that, within the
selected range of periods and for values bigger than the Tpn ; the ‘up’ curves are quite flat, and
therefore the model peak deformations present relatively small variations within this range of

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
908 S. BERTON ET AL.

Figure 8. Plots of the peak deformations and peak isolator force for the East–West component of
the 1999 Chi-Chi Taiwan earthquake (record no. 4).

periods. For this reason, models corresponding to isolator periods bigger than Tpn ; for example,
around 6 s, might be preferable over the ones corresponding to the optimal values because
despite a small increase in the model deformation demands they provide significantly better
performances in terms of isolation capacity (i.e. smaller forces transmitted to the mass). The Tp
limits derived from the application of the self-centring requirements show again that the
AASHTO specifications produce in general higher values (always 6 s for this analysis series)
than the corresponding EC8 requirements. However, in one case (r ¼ 10% and uy ¼ 50 mm) the
EC8 results in a higher value (6.2 s). For this ground motion, the application of both norms
produces limits that are always bigger than the calculated optimal post-yield periods.
The results of the model subjected to record #3 are shown in Figure 7 (Loma Prieta). The
PGA, PGV and peak ground displacement (PGD) of this record are among the smallest within
the set of ground motions used in this study (Table I). Because of the small intensity of the input
ground motion, the ‘upn ’ values obtained from the analysis are smaller than the corresponding
values of the previous cases, varying from 57.5 mm (top-right subfigure) to 180.3 mm (bottom-
right subfigure). The displacement normalizing value for this analysis series is up % ¼ 288:6 mm:
Similar to the results of the Kobe record shown in Figure 5, the values of optimal post-yield
periods vary considerably ranging from Tp ¼ 2:75 s to Tp ¼ 1: It is interesting to notice that for
two of the six analysis cases shown in the figure, the best model performance in terms of isolator
deformation occurs for Tp ¼ 1: The peak forces ‘frpn ’ transmitted to the system mass are also
relatively small, although in the case corresponding to r ¼ 10% and Fy ¼ 10 mm the resulting

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
SEISMIC ISOLATION SYSTEMS 909

Figure 9. Plots of the peak deformations and peak isolator force for the N46.7E component of the
2002 Denali Alaska earthquake (record no. 5).

value of frpn ¼ 0:165mg represents a significant portion of mPGA (about 51%). As observed in
the previous analysis series, an increase of Tp within the selected limits corresponds in general to
a slow monotonic increase of the residual displacements, ‘or’. For this analysis series, the worst
case in terms of residual displacement is the one corresponding to r ¼ 10% and uy ¼ 10 mm:
For that case, the maximum value of or ¼ 163:6 mm occurs at Tp ¼ 1 (out of range of the
figure), which is about 90% of the isolator maximum displacement obtained from the same
analysis case (up ¼ 182:2 mm). The limits of post-yield periods determined from the application
of the EC8 norms are smaller than the corresponding values of the AASHTO norms. Similar to
the previous analysed cases, the EC8 limits are particularly small for the three r ¼ 10% cases in
which they vary from 2.4 to 3.2 s, and are always smaller than the Tpn values that produce
minimum model deformations. This is also true in four out of six cases for the AASHTO norms,
whose limit values are also relatively small for this ground motion, ranging from 3.4 to 5.7 s.
Therefore, despite the relatively small intensity of the input ground motion, both the EC8 and
the AASHTO specifications seem to be too restrictive, as in most instances they prohibit systems
that produce best deformation performances.
Figure 8 shows the analysis responses of the model subjected to record #4 (Chi-Chi Taiwan).
The PGA for this record is quite high ð1:01gÞ; the highest among the six ground motions used in
this study. The PGV and PGD are also moderately high (0.60 m/s and 0.502 m, respectively).
However, this record does not present near-fault pulse-like characteristics [20]. Despite the high
value of PGA, the model peak displacements obtained from this analysis series are relatively

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
910 S. BERTON ET AL.

small (‘upn ’ values varying from 97.7 to 269.2 mm), and significantly smaller than the
corresponding values of the Northridge analysis series (Figure 6). The normalizing displacement
% ¼ 489:4 mm that is the maximum deformation that occurs for the case r ¼ 5% and
value is up
uy ¼ 50 mm at Tp ¼ 1: Although the magnitude of the peak deformations is quite different
from that of the Northridge earthquake, the general trends of the responses of the two series
present some similarities. For example, the values of optimal post-yield periods are almost the
same among all six analysis cases and quite small, varying from 2.0 to 2.5 s. Despite the low
values of the optimal post-yield periods, the ability of the model to reduce the force transmitted
to the mass is quite good. In fact, the ‘frpn ’ values range from 0.166 to 0:212mg that are only
about 16 and 15 of mPGA, showing that base isolation, for this particular ground motion, is quite
effective for a wide range of the parameters characterizing the model response. For this analysis
series, the effective periods evaluated for the Tpn cases range from 1.42 to 2.38 s, while the
corresponding effective damping ratios vary from 4.8 to 31.9%. Similar to what observed in the
previous analysis series, the AASHTO specifications provide limit values that are always bigger
than the corresponding EC8 limits. For the AASHTO specifications the post-yield period limits
vary from 5.1 to 6.0 s, while the EC8 limits vary from 4.1 to 5.0 s for the r ¼ 5% cases, and from
2.6 to 3.0 s for the three r ¼ 10% cases. Despite their small values, both the AASHTO and the
EC8 limits are in this case always bigger than the corresponding optimal post-yield periods.
The responses of the model subjected to the input record #5 (Denali Alaska) are shown in
Figure 9. This record presents strong characteristics of near-fault pulse-like ground motions
[18]. Its PGA is relatively small (i.e. 0:290g; the smallest among the records used in study).
However, its PGV and PGD values are quite large (Table I). The plots of the maximum isolator
displacements, ‘up’, show similar trends among all six analysis cases with peak values recorded
around 2.0–3.0 s. However, the average of the ‘up’ values is significantly bigger for the three
r ¼ 5% cases, confirming the great influence of the yielding force Fy on the model deformation
response as observed in the previous analysis series. The maximum peak displacement used as
normalizing value is up % ¼ 1161:3 mm (the highest among the six analysis series) produced for
the case r ¼ 5% and uy ¼ 50 mm at the value of Tp ¼ 2:75 s: Although the trends of the ‘up’
plots are similar, the Tpn values are significantly different ranging from 3.5 s to 1: The peak
isolator force values, although small in absolute value, are always an important portion of
mPGA ranging from 0.10 to 0:173mg: The values of the effective periods and effective damping
ratios calculated at optimal conditions ranges from 2.09 to 3.97 s and from 16.5 to 53.4%,
respectively. Again, the application of the AASHTO specifications provide higher post-yield
period limits ranging from 5.1 to 6.0 s, while the EC8 result in limit values ranging from 2.8 to
5.8 s. Note that for five out of six cases, the EC8 limits are smaller than the calculated optimal
values.

4.5. General considerations from analysis results


Within the limitations imposed by the model assumptions and by the limited number of cases
analysed, the results of this study provide the basis for a discussion on the influence of the three
independent parameters uy ; Fy and Tp ; and on the merits of the self-centring criteria imposed by
AASTHO and EC8 norms. As a general comment, it is noted that some combinations of the
parameters used in this study correspond to devices that might not be available with the current
technology and/or materials. This fact partially explains the high values of the peak
deformations often obtained for certain parameter combinations. However, it is important to

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
SEISMIC ISOLATION SYSTEMS 911

point out that this is not a limitation since this parametric study focuses on the relative influence
of the parameters on the model response. The values of the effective periods and effective
damping ratios show often high variability within the cases analysed. This is not a surprise since
these practical parameters are meaningful only when associated with the maximum deformation
of the device, whose values in this study are always highly variable even within the same analysis
case.
In terms of the influence of the parameter uy ; the analysis results show that the average values
of the model peak displacements always increase as this parameter is varied from 50 to 0.5 mm.
However, the variation of this parameter does not change significantly the trends of the plots of
the peak displacement responses, which in general present very similar shapes within the three
cases having the same value of r: Table II shows the average values for each input ground
motion of the peak deformations, upaverage ; calculated considering all the results of the
analyses with same uy : The last column of the table shows the variations in terms of percentage,
Dupaverage ; of the values of the average peak deformations corresponding to uy ¼ 50 and 0:5 mm:
Considering all the ground motions, Dupaverage decreases 39.05% when uy is varied from 50
to 0.5 mm, with the maximum variation ð49:83%Þ obtained for record #3 and the smallest
ð24:18%Þ for record #2. The difference in magnitude of the up responses due to the variation of
uy ; can be partially attributed to the better energy dissipation capacity associated with models
having same characteristics but smaller uy ; which in general produce larger values of effective
damping ratio. Another contribution to this difference in magnitude of the deformation
responses can be associated with the larger elastic (and therefore recoverable) deformations of
the models with larger yielding displacements and smaller pre-yield stiffnesses.
Results from this study indicate that the parameter Fy has always a strong influence on the
deformation of the model and, therefore, on its self-centring capacity. The yielding force Fy not
only affects significantly the magnitudes of the model peak displacements, but also in some

Table II. Average values of ‘up’ (same uy analyses).


upaverage (mm)
Earthquake record no. uy ¼ 50 uy ¼ 10 uy ¼ 0:5 Dupaverage (%)
1 324.37 242.11 211.06 34.93
2 564.09 430.46 427.70 24.18
3 198.00 171.40 99.33 49.83
4 275.09 198.55 169.86 38.25
5 564.74 365.91 293.35 48.05

Table III. Average values of ‘up’ (same r analyses).


upaverage (mm)
Earthquake record no. r ¼ 0:05 r ¼ 0:1 Dupaverage (%)
1 276.77 241.60 12.71
2 543.62 404.55 25.58
3 171.58 140.91 17.85
4 260.63 168.37 35.40
5 535.29 280.71 47.56

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
912 S. BERTON ET AL.

instances, the overall trend of the plots of the displacement responses. Similar to Table II,
Table III shows the average values of the peak deformations, upaverage ; this time calculated
considering all the analyses with same r and therefore same Fy : Considering all the input ground
motions, Dupaverage decreases 27.82% as r changes from 5 to 10%. The maximum decrement of
47:56% is obtained for record #5 and the smallest for record #1, in which the variation is
12.71% as shown in the last column of Table III. When r is increased from 5 to 10%, the ability
of the system to reduce the force transmitted to the isolated mass decreases due to the higher
theoretical limit (i.e. rmg) to which the ‘frp’ curves tend as Tp tends to 1: Nevertheless, it is to
be noted that the isolation ability of the model calculated at Tpn or higher values of post-yield
period having r ¼ 10% is sufficiently good for all the cases considered.
The results of this study confirm that the parameter Tp influences both the model deformation
response as well as its ability to self-centre. However, within the selected practical limits, the
influence of this parameter has been found to be smaller than expected. Although the influence
of Tp depend greatly on the nature of the input ground motion, the peak displacement curves
are frequently flat. In some instances, an increase of post-yield period results in smaller isolator
deformations as shown in some of the analysis cases of Figures 5, 7 and 9, in which the smallest
values of ‘up’ are obtained with models corresponding to Tp ¼ 1: These results contradict the
assumption that larger post-yield periods are always associated with larger cumulative
deformations and larger permanent displacements. In addition, an increase of the post-yield
period within the selected practical limits, usually does not influence considerably the peak
isolator deformation ‘up’. On the other hand, small increments of Tp are often associated with
more significant improvements of the isolator ability to transmit lower forces to the system
mass.
The final considerations of this section are for the comparison of the AASHTO and EC8 self-
centring specifications. Within the assumptions stated in Section 4.3.2, the two criteria of the
new version of the EC8 result in maximum post-yield periods that are in general more restrictive
than the corresponding AASHTO limits. In addition, the application of the new EC8 often
produces limit values that are smaller than the optimal post-yield periods Tpn : Therefore, the
EC8 norm (in particular, the newly introduced second criterion) seems to be too restrictive, as in
most instances it prohibits the use of the system with best displacement performance and better
isolation capacity. Conversely, the application of the AASHTO specifications, often results in
limit values of Tp ¼ 6 s defined by its second criterion, which are usually larger than the
corresponding Tpn values. In the cases in which Tpn exceed the AASHTO limit, the deformations
of the model corresponding to the optimal periods do not improve significantly with respect to
the corresponding performances at the AASHTO limits. Therefore, for the cases analysed in this
study, the combination of the two AASHTO criteria provides reasonable limits, although the
results also indicate that the 6 s of the AASHTO second criterion could be increased, for
example up to 7 or 8 s, without resulting in excessive isolator deformation demands and without
changing significantly the required level of system restoring capacity.

5. CONCLUSIONS

This paper presents the results of a parametric study performed to evaluate the self-centring
capacity of seismic isolation systems. The parameters used to characterize the bilinear oscillator
model are: the yielding displacement uy ; the yielding force Fy ; and the post-yield period Tp :

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
SEISMIC ISOLATION SYSTEMS 913

Criteria that impose minimum requirements on the restoring capacity of isolation systems are
included in current norms such as the prEN 1998-2:2005 Eurocode 8 (EC8) and the 1999
American Association of State Highway and Transportation Officials (AASHTO). The
motivations for these requirements include that an insufficient restoring capacity could lead
to the development of significant permanent deformations of the isolator during an earthquake.
The model performances have been evaluated in terms of peak displacement, permanent
deformation and in terms of its capacity of reducing the seismic force. From the results of this
study the following conclusions can be drawn:
1. Within the selected practical limits, the influence of the post-yield period Tp on the model
peak deformation response and self-centring capacity has been found to be smaller than
expected. The peak deformation curves, for values of Tp between 4 and 9 s, are frequently
flat showing a relative small influence of this parameter in that range of periods. The
yielding force Fy has always a significant influence on the magnitude and general trend of
the model deformation responses and on the ability of the system to self-centre. However,
most of the current guide specifications are based on criteria that consider only the
difference of the restoring force at 100% and at 50% of the isolator peak displacement,
neglecting the importance of Fy (or equivalently of the system strength Q) on the restoring
capacity of the system.
2. The evaluation of the requirements imposed by current norms in terms of self-centring
capacity shows that the combination of the two new criteria introduced in the final draft of
the EC8 is often too restrictive. In many instances, the maximum values of post-yield
period imposed by these norms were found to be smaller than the optimal periods that
correspond to models with best performances in terms of deformation demand and with
better isolation capacity. The requirements of the current version of the AASHTO
specifications, which includes a criterion based on a maximum value of post-yield period
(i.e. Tp 46 s), imposes limits of Tp that, in general, include systems with optimal
performances. However, the results of this study also indicate that the 6 s criteria of the
AASHTO could be increased, for example, to 7 or 8 s, without changing significantly the
required restoring capacity of the system.
3. In this study, the analyses carried out using ground motion records having near-fault
pulse-like characteristics were found to produce significant higher model deformations,
confirming the results obtained in previous studies on isolated structures subjected to such
ground motions.

ACKNOWLEDGEMENTS

The support for this study from the Japanese Society of the Promotion of Science (JSPS) through a
research grant to the first author is gratefully acknowledged.

REFERENCES
1. Skinner RI, Robinson WH, McVerry GH. An Introduction to Seismic Isolation. Wiley: New York, 1993.
2. Kelly JM. Earthquake-Resistant Design with Rubber (2nd edn). Springer: London, 1997.
3. Zayas VA, Low SS, Mahin SA. A simple pendulum technique for achieving seismic isolation. Earthquake Spectra
1990; 6(2):317–333.

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc
914 S. BERTON ET AL.

4. Highway Innovative Technology Evaluation Center (HITEC). Evaluation finding for FIP-energy absorbing systems,
L.C.C. sliding bearings. CERF Report: HITEC 98-05, Highway Innovative Technology Evaluation Center (HITEC),
1998.
5. Castellano MG, Infanti S, Saretta G. Validation of steel hysteretic dampers through shaking table tests. Proceedings
of the 5th World Congress on Joints, Bearings and Seismic Systems for Concrete Structures. Rome, Italy, 2001.
6. Makris N, Black CJ. Dimensional analysis of bilinear oscillators under pulse-type excitations. Journal of Engineering
Mechanics (ASCE) 2004; 130(9):1019–1031.
7. Mosqueda G, Whittaker AS, Fenves GL. Characterization and modeling of friction pendulum bearings subjected to
multiple components of excitation. Journal of Structural Engineering (ASCE) 2004; 130(3):433–442.
8. European Committee for Standardization (CEN). ENV 1998-2:1994 Eurocode 8: Design of Structures for Earthquake
Resistance Part 2: Bridges. Bruxelles, Belgium, December 1994.
9. American Association of State Highways and Transportation Officials (AASHTO). Guide Specifications for Seismic
Isolation Design. American Association of State Highways and Transportation Officials: Washington, DC, 1999.
10. Jangid RS, Kelly JM. Base isolation for near-fault motions. Earthquake Engineering and Structural Dynamics 2001;
30(5):691–707.
11. Kawashima K. Seismic isolation design code for highway bridges. Proceedings of the 7th International Seminar on
Seismic Isolation, Passive Energy Dissipation and Active Control of Vibrations of Structures. Assisi, Italy, 2001.
12. International Code Council (ICC). International Building Code. Falls Church, VA, 2003.
13. International Conference of Building Officials. Uniform Building Code (UBC). Whitter, CA, 1997.
14. European Committee for Standardization (CEN). prEN1998-2:2005 Eurocode 8: Design of Structures for Earthquake
Resistance}Part 2: Bridges Final Draft. Bruxelles, Belgium, March 2005.
15. Chopra AK. Dynamics of Structures: Theory and Applications to Earthquake Engineering (2nd edn). Prentice-Hall:
Upper Saddle River, NJ, 2001.
16. Wen YK. Method of random vibration of hysteretic systems. Journal of Engineering Mechanics (ASCE) 1976;
102:249–263.
17. Ryan KL, Chopra AK. Estimation of seismic demands on isolators based on nonlinear analysis. Journal of
Structural Engineering (ASCE) 2004; 130(3):392–402.
18. Ellsworth WL, Celebi M, Evans JR, Jensen EG, Kayen R, Metz MC, Nyman DJ. Near-field ground motion of the
2002 Denali fault, Alaska, earthquake recorded at pump station 10. Earthquake Spectra 2004; 20(3):597–615.
19. Pacific Earthquake Engineering Research (PEER). PEER Strong Motion Database 2000. http://peer.berkeley.edu/
smcat
20. Lee WHK, Shin TC. Strong motion instrumentation and data. Earthquake Spectra 2001; 17(S1):5–18.

Copyright # 2006 John Wiley & Sons, Ltd. Struct. Control Health Monit. 2007; 14:895–914
DOI: 10.1002/stc

You might also like