Engineering Structures 99 (2015) 232–252

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

A case study on the application of passive control and seismic isolation

techniques to cable-stayed bridges: A comparative investigation through
non-linear dynamic analyses
M.D. Martínez-Rodrigo a,⇑, A. Filiatrault b,c
a
Dept. of Mechanical Engineering and Construction, Universitat Jaume I, 12071 Castellón, Spain
b
Dept. of Civil, Structural and Environmental Engineering, Univ. at Buffalo State Univ. of New York, Buffalo 14260, USA
c
University Institute of Advanced Studies (IUSS) of Pavia, 27100 Pavia, Italy

a r t i c l e i n f o a b s t r a c t

Article history: This paper investigates the seismic performance of an existing steel cable-stayed bridge located in a high
Received 30 September 2014 seismic zone. In 1988 the structure experienced the failure of one of its anchorage plates needing closure
Revised 16 March 2015 and structural repair as a consequence of seismic activity. In this study, the possibility of retrofitting the
Accepted 28 April 2015
original bridge with different passive supplemental damping and seismic isolation systems is proposed
Available online 18 May 2015
and assessed, in order to improve the response of the structure in its longitudinal direction. To this
end a Finite Element model of the bridge is developed and the structural response is evaluated through
Keywords:
non-linear dynamic response analyses under a set of historical ground motions of different intensities,
Cable-stayed bridges
Dampers
including a near field record. Strength degradation capabilities are introduced in the model, allowing
Seismic isolation the occurrence of brittle failure in all the members. Seismic performance indices referred to the dynamic
Passive control response of the unretrofitted structure and to the results of a pushover analysis are defined and used to
Strength degradation compare the proposed innovative solutions. The major improvement on the overall response of the
Non-linear dynamic analysis bridge is shown and conclusions regarding the most appropriate retrofit alternative for the particular
Seismic performance case study are determined.
Supplemental damping Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction in the convenience of permitting certain relative movement of the

In the last decades cable-stayed bridges have gained popularity
throughout the world for spans up to approximately 1 km due to
their appealing aesthetics, fast construction, efficient use of struc-
tural materials leading to small structural members and light
appearance, and increased stiffness when compared to suspension
bridges. This type of structure, characterized by a long fundamen-
tal period, large flexibility, light weight and low structural damp-
ing [1,2], is quite vulnerable to large amplitude oscillations when
excited by earthquake ground motions [3–6]. A number of studies
have been performed in the past related to the seismic behaviour
of cable-stayed bridges [5,7–9]. The seismic response of a
cable-stayed bridge depends to a great extent on how the bridge
deck is connected to the tower and the piers. Rigid connections
limit the deck horizontal displacements under earthquake action
but unavoidably increase the transmitted forces between the
superstructure and the substructure. There is a general agreement
⇑ Corresponding author. Tel.: +34 964728133; fax: +34 964728106.
E-mail address: mrodrigo@emc.uji.es (M.D. Martínez-Rodrigo).
deck at the pier and tower locations, to reduce the internal forces
at the base of these elements but, due to the low inherent damping,
important horizontal displacements are to be expected in that
location. This behaviour suggests seismic control techniques, i.e.
passive, active or semi-active control, as possible alternatives to
improve the performance of cable-stayed bridges under strong
earthquake ground motions [10].
This particular study is focusing on an existing metallic 183 m
long cable-stayed bridge located in a high seismic zone (Fig. 1). A
past inspection of the bridge revealed the complete failure of one
of the four anchorage plates connecting the deck steel girders to
one abutment. The bridge was closed and immediately repaired.
The investigation that followed this event [11,12], confirmed that
the failure had been the result of a ML = 6.0 magnitude earthquake,
during which peak horizontal ground accelerations of approxi-
mately 0.15 g were recorded in the epicentral region. The original
rigid welded connection of the bridge tower to the box girder at
the deck level, along with the high levels of longitudinal vibration
evidenced from previous studies, suggest that the introduction
of supplemental dampers, the application of base isolation

http://dx.doi.org/10.1016/j.engstruct.2015.04.048
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252
Fig. 1. Bridge under study.
techniques or a combination of both could effectively improve the
seismic performance of this particular bridge structure.
In a mitigation design (or redesign) context, two main objec-
tives are pursued: (i) to increase the structure fundamental period,
and therefore reduce the spectral acceleration demand levels and,
(ii) to enhance the energy dissipation capabilities of the bridge and,
thereby, increase damping. In this regard, three approaches may be
adopted: (i) introducing dampers that, taking advantage of the
structure motion during the earthquake, transform the input seis-
mic energy into heat i.e. fluid viscous dampers (FVDs), metallic or
friction dampers (MDs, FDs), shape memory alloys (SMAs); or into
kinetic energy i.e. tuned mass dampers (TMDs); (ii) isolating totally
or partially the bridge deck from the substructure, leading to min-
imum member forces but maximum deck horizontal displace-
ments; or (iii) a combination of (i) and (ii) which would take the
advantages of seismic isolation but while also controlling the level
of displacements through supplemental damping. In this particular
study, passive devices with and without isolation systems are con-
sidered, as opposed to active or semi-active techniques, as they do
not require external energy and therefore, if properly installed and
maintained, may constitute reliable, robust, simple and economic
alternatives.
The first studies dealing with the application of seismic isola-
tion techniques to improve the seismic performance of
cable-stayed bridges were conducted by Ali and Abdel-Ghaffar
[2,3]. The authors evaluated the seismic isolation of cable-stayed
bridges using elastomeric and lead-rubber bearings (LRBs), focus-
ing on the devices most appropriate constitutive models for analy-
sis and their optimal location and mechanical properties.
Constantinou et al. [13,14] proposed and evaluated experimentally
a seismic isolation system for bridges consisting on multidirec-
tional Teflon disc bearings and displacement control devices. The
former accommodated thermal movements and provided isolation
while the latter introduced a restoring force for re-centring the
bridge during the earthquake and for providing additional energy
dissipation capacity and rigidity for service loading. Iemura and
Padrono [15] investigated the seismic performance of a
cable-stayed bridge isolated by means of elastomeric and
hysteretic bearings, incorporating passive linear and semi-active
dampers. Casciati et al. [16] evaluated the dynamic performance
of the ASCE benchmark problem of a cable-stayed bridge isolated
with elastomeric LRBs by means of fragility curves. Weselowsky
and Wilson [17] examined the seismic response of cable-stayed
bridges isolated by means of LRBs under the action of near field
earthquake records. The authors concluded that there is an optimal
amount of isolation that will maximize the benefits of reducing
233
seismic forces and limiting the seismically-induced displacements.
Soneji and Jangid [18] compared the effectiveness of three systems
i.e. high damping rubber bearings (HDRBs), LRB and a friction
pendulum system (FPS) on the isolation of cable-stayed bridges,
finding the optimal amount of yield strength of LRBs and frictional
coefficient of FPS leading to the maximum reduction in the tower
base shear, while keeping the maximum displacement of the
superstructure to a minimum. They also proposed to combine iso-
lation techniques with passive dampers to enhance the control of
the maximum displacement in isolated structures.
Apart from the studies dedicated to seismic isolation, other
works have been conducted on the introduction of passive control
systems with and without seismic isolation. In particular, regard-
ing the application of metallic or friction devices, Nihihara et al.
[19] analysed the seismic performance of cable-stayed bridges
retrofitted with yielding MDs, considering different deck-tower
connections. The authors detected an important reduction of the
maximum displacements in the harp cable-system arrangement
case. Vader and McDaniel [20] compared the effectiveness of FDs
and FVDs with that of a new shear-link protection system for
cable-supported bridges. More recently, Camara [5] investigated
the possibility of introducing yielding MDs to control the trans-
verse response of the towers in cable-stayed bridges.
As for the effectiveness of improving the seismic performance of
cable-stayed bridges with passive FVDs, most of the limited studies
done in this regard considered hybrid control systems as well. That
is the case of Soneji and Jangid [21], who compared the seismic
performance of cable-stayed bridges with isolation systems with
and without passive hybrid systems, combining elastomeric or
sliding isolation systems in association with FVDs. He and
Agrawal [22] evaluated the effectiveness of passive FVDs and a
hybrid control system, consisting of passive FVDs installed in
parallel with semi-active dampers under a large number of near
field records. The semi-active dampers are only triggered in the
case of short duration pulses in order to protect the passive dam-
pers, resulting in a hybrid system quite capable of reducing peak
response quantities of the bridge. Valdebenito [23] analysed the
important role of the damping constant in the seismic performance
cable-stayed bridges retrofitted with FVDs.
Only a few authors have evaluated the possibility of introducing
shape memory alloy based devices in cable-stayed bridges.
Sharabash and Andrawes [10] pointed out the re-centring capabil-
ities, strain hardening for high strain levels and stress plateau lim-
iting the forces transmitted to the structure as clear advantages.
TMDs have also been considered as a possible measure to control
seismic induced vibrations in bridges. Koh et al. [24] investigated
the effect of installing several tuned liquid dampers in a simplified
model of a suspension bridge. The authors focussed on the spatial
distribution of the devices and its effect in controlling the response
of multiple modes.
In recent years, innovative passive and semi-active devices have
been investigated for the same application. Domaneschi and
Martinelli [25] compared the performance of a refined version of
the ASCE cable-stayed bridge benchmark model retrofitted with
electro-inductive devices, idealized by the Bouc–Wen hysteretic
law, with classical MDs. Ha et al. [26] proposed a complex damper
system combining a velocity dependent oil damper, for frequent
and small vibration amplitudes, and a displacement dependent
elasto-plastic damper, which activates for large amplitude
vibration pulses. The authors show that the proposed system could
significantly improve the seismic performance of long span
cable-stayed bridges.
Despite that the first studies on passive isolation of cable-stayed
bridges date back 20 years, there is a clear need for research on
passive protection systems applied to this type of structures [4],
along with the development of regulations and normative support
234 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252
[5]. In this study, the possibility of improving the seismic perfor-
mance of an existing bridge structure through passive control
and base isolation techniques is investigated. In particular the
hypothetic retrofit of the structure including hysteretic dampers,
fluid viscous dampers, tuned mass dampers and its seismic isola-
tion by means of friction pendulum bearings is evaluated, and
the different alternatives are compared under the action of several
earthquake ground motions, including a near field record. Strength
degradation capabilities are included in the model allowing the
occurrence of brittle failure and force redistribution in all the
members. The major improvement in the overall response of the
bridge is shown and conclusions regarding the most appropriate
retrofit alternative for the particular case studied are determined.
2. Original structure and numerical model
2.1. Description of the cable-stayed bridge structure
The bridge under study consists of a double leg steel tower,
double-plane fan-type cables and two steel box girders supporting
a composite concrete steel deck (see Fig. 2). The total length of the
bridge is 183 m divided into four identical spans between the cable
anchorages and the supports. There is a 4% upward slope from the
west to the east abutment along the deck.
The bridge deck is composed of a 165 mm thick concrete slab,
11 m wide, with two non-structural precast parapets (see Fig. 3).
Five longitudinal steel stringers are spaced at equal transverse
intervals of 2.4 m. Floor beams, transverse to the main girders at
equally spaced intervals of 7 m, transfer the stringer loads to the
two main box girders at the outer edges of the deck. The
1.5  3 m box girders are made of welded flanges, webs, stiffeners
and diaphragms. The cables are connected to the deck at the top
flange of the main box girders. The tower, consists of two
1.5  2.4 m rectangular box steel legs and a cross beam supporting
the deck. The top of the tower is 43 m tall. The thickness of the
flanges and webs of the box steel legs and the box girder equal
50 mm. Each leg of the tower is rigidly connected to the intersect-
ing box girder at the deck level.
The support system of the bridge is founded on rock. At each of
the abutments, roller supports resist the uplift forces generated by
the cables, which allow for sliding in the longitudinal direction of
the bridge. The bearings under each leg of the tower prevent hor-
izontal and vertical movements and permit rotation around the
transverse axis of the bridge only. The bridge incorporates four
cables per tower each composed of nine strands. Each strand has
a cross-sectional area of 65.1 mm2. The cables are constructed from
standard galvanized bridge strands with a Young’s modulus of
175 GPa, yield strength of 1500 MPa and ultimate strength of
1725 MPa. A full description of the bridge under study may be
found in Christopoulos and Filiatrault [27].
2.2. Numerical model of the original bridge structure
The dynamic response of the bridge in its longitudinal direction
under the action of a series of ground motions is investigated
numerically. To this end a planar Finite Element (FE) model of
the structure is implemented using the general purpose nonlinear
dynamic analysis computer program RUAUMOKO-2D [28]. Only
half of the bridge width is modelled, as the structure is symmetri-
cal and no seismic transverse input is considered. The model
includes one main box girder, representing the deck, discretized
in 19 2D-beam elements (see Fig. 4), one tower box girder (13
beam–column elements) and four sets of cables modelled as
tension-only truss elements. The nodal coordinates used in the
model are based on the assumed locations of field welded
connections of the box girder and the tower. One finite element
per cable is considered. The concrete deck, steel stringers and floor
beams are considered rigid and its dead load (25 kN/m for half of
the bridge) is applied to the girder. Soil structure interaction is dis-
regarded as the support system of the bridge is founded on rock
[11].
Cable-stayed bridges are characterized by their geometric non-
linear behaviour under static and dynamic loadings, especially in
the case of long spans. Nonetheless, previous authors have found
that there is only a slight difference between linear and geometri-
cally nonlinear seismic time-history analysis of cable-stayed
bridges under strong ground motions, provided that the analysis
is performed starting from the deformed equilibrium configuration
under dead loads [9,30,31]. The geometrical nonlinearity in this
type of structures originates from (i) the cables sag effect; (ii) the
action of compressive loads on the deck and the tower due to the
inclination of the cables; and (iii) the effect of large deflections
and displacements. From these three sources of nonlinearity, (ii)
and (iii) are implemented in the numerical model while only the
cable sag effect has been neglected since just one FE per cable
has been included, disregarding cable-structure interaction effects.
This assumption is a consequence of the high level of axial loads
experienced by the cables in the deformed equilibrium configura-
tion under the static loads in the particular bridge under study, the
relatively short length of the cables, and the moderate level of pre-
dicted inelastic response in these elements under the seismic
action [32–34]. The equivalent axial stiffness of each cable element
considered is 719, 1040, 1777 and 1843 kN/m, from the outer cable
of the East span to the West span shortest cable, respectively.
Regarding material nonlinearity, the inelastic behaviour of
beam and beam–column elements, used for the deck and the tower
respectively, is represented by the Giberson one-component frame
model [35], incorporating the possible formation of plastic hinges
at one or both ends of the elastic central length of an element.
These hinges are assigned a bi-linear hysteretic behaviour with a
curvature strain-hardening ratio of 0.02 (see Fig. 5a). The hinges
length is set equal to 90% of the member depth. The plastic resis-
tance of the hinges is based on an expected yield strength of
290 MPa and a Young’s modulus of 200 GPa. Beam elements are
also assigned a bilinear axial load-axial displacement hysteresis.
However there is no interaction between the axial and the
moment–curvature yield states. On the other hand, an axial
load-moment interaction as per [29] is considered for the tower
elements (see Fig. 5c). The inelastic tensile response of the cables
is modelled with a tension-only bi-linear hysteretic behaviour
(see Fig. 5b). The tensile strain-hardening ratio is set to 0.1. In all
the elements the possible occurrence of brittle failure has been
introduced by forcing the sudden degradation of the member
strength after the ultimate curvature ductility demand for frame
elements, l/, or elongation ductility demand for cable elements,
lD, has been exceeded, as shown in Fig. 5. Ductility demands are
defined as follows for the frame (deck and tower) and for the cable
elements:
l/ ¼ ð/p þ /y Þ/1
y lD ¼ ðDp þ Dy ÞD1
y ð1abÞ
where /y and /p are the yielding and plastic curvatures in the frame
elements, respectively, and Dy and Dp are the yielding and plastic
elongations in the truss elements modelling the cables.
The failure criterion assumed for all box girder and tower ele-
ments is based on a plastic end rotation limit of 0.016 rad, which
is lower than the plastic rotation capacity of a compact section
[29]. This corresponds to a moment increase due to strain harden-
ing of 0.1 Mp, as shown in Fig. 5a. Similarly, the failure criterion for
the cables is based on the tensile strength of the cables. These fail-
ure states are associated to maximum ductility demand levels of
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252 235

Fig. 2. General view of the bridge structure under study.

Fig. 3. Bridge deck cross section.

Fig. 4. Bridge model.

l/ = 6 and lD = 2.5 in the frame and cable elements, respectively.
Finally, strength degradation is included in all the girder and tower
elements as well as in the cable elements, once the maximum duc-
tility is exceeded, as indicated in Fig. 5. In the girder and tower ele-
ments the ductility levels at which degradation begins and stops
are selected as 6 and 1.05  6 = 6.3, with an associated residual
strength of 1% of the initial yield strength. As for the cables, the
degradation begins and stops when ductility values reach 2.5 and
1.05  2.5 = 2.625, respectively, for a residual strength of 1% of
the initial yield strength as well. Modelling this sudden failure of
structural elements may be of importance when the structure is
subjected to strong near field events and internal forces need to
redistribute if the structure locally fails.
The nonlinear dynamic response of the structure under the
action of several design ground motions applied in the longitudinal
direction of the bridge is obtained by direct integration, using the
Newmark-Beta constant acceleration method. Structural damping
was experimentally estimated in the bridge and values in the range
of 1–3.7% were obtained for the first five modes of vibration [12]. In
this case study, structural damping is incorporated as Rayleigh
damping, considering a 1% damping ratio assigned to the first
two modes of the structure to remain conservative. A static
236 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252
Fig. 5. (a) Bi-linear moment–curvature model for box-girder and tower elements; (b) bi-linear tensile force–elongation model for cables; (c) axial force-moment interaction
diagram for tower elements.
analysis of the bridge under the dead loads is performed prior to
any type of subsequent dynamic analysis, given the relevance of
starting from the deformed equilibrium position, as stated in pre-
vious studies [5,9,16,36].
2.3. Modal analysis of the original bridge structure
Fig. 6 shows the first four normal mode shapes, natural periods
Ti, associated damping ratios fi and effective modal masses Meff,i as
a percentage of the total mass Mtot computed with the previously
described FE model of the bridge, determined from the static equi-
librium solution under dead loads. When the bridge is analysed
under the dead loads the average percentage of the bending
moment with respect to the plastic moment computed along the
deck elements ends equals 46.2%.
There is a reasonable agreement between the first three lower
natural frequencies computed by the FE model and those obtained
experimentally in a previous investigation [12], as shown in
Table 1.
The sum of the first three effective modal masses is approxi-
mately 90% of the total mass of the bridge, therefore it is expected
that these three modes, and specially the second mode, contribute
the most to the total longitudinal seismic response of the bridge.
3. Pushover analysis of the original bridge structure
In this section the results of a pushover analysis performed on
the bridge model are presented. These results will allow the eval-
uation of the elastic and inelastic responses of the structure under
lateral loads, and will provide information regarding the structure
failure mechanism that should be expected under longitudinal
earthquake loading. In the pushover analysis, the bridge is sub-
jected to a progressively increasing lateral load profile, propor-
tional to the mass distribution of the deck and the tower, until
global failure of the structure occurs. The lateral load is applied,
starting from the equilibrium configuration under the dead loads,
slowly enough so that the inertia forces are insignificant. Fig. 7
shows the evolution of the base shear at the tower fixed support
versus the top of the tower horizontal displacement minus the sta-
tic solution (utop,st). These two quantities have been divided by the
total weight of the structure (half bridge, W = 12,042 kN) and by
the tower height (H = 43 m), respectively, in order to normalize
them. In this figure two curves are shown for the same pushover
analysis applied in the West–East direction (Fig. 7a), and in the
East–West direction (Fig. 7b), as different results are expected
due to the lack of symmetry of the cable arrangement.
Four different discrete damage events can be identified in
Fig. 7a: (1) the formation of the first plastic hinge in the tower
occurring next to the tower-deck connection when the base shear
reaches 9058 kN and for a horizontal displacement of the top of the
tower of 0.357 m; (2) the formation of the first plastic hinge in the
box girder occurring adjacent to the bridge-tower connection
almost simultaneously, when the base shear reaches 9079 kN for
a horizontal displacement of the top of the tower of 0.360 m; (3)
the brittle failure of the same first plastic hinge in the tower taking
place at the maximum base shear of 10,021 kN and a horizontal
displacement of the top of the tower of 0.539 m; and (4) the end
of the numerical analysis defined by the initiation of structural col-
lapse and appearance of large displacements.
Fig. 8 shows the location of the aforementioned damage events
(1)–(2) and (3). As it is shown, damage on the structure concen-
trates in the vicinity of the tower-deck connection. The structure
overall stiffness drops between events (1) and (2), but internal
forces redistribute still showing a certain level of lateral stiffness.
At the instant of event (3), the sudden failure of the tower hinge
leads to the global structure collapse. Yielding of cable elements
takes place after the pushover analysis is stopped at instant (4)
due to the occurrence of very large displacements.
When the load profile is applied at the same rate but in the
opposite direction (Fig. 7b), the bridge remains elastic until event
10 , when the box girder starts to yield for a base shear of
4089 kN, close to the tower-deck connection (see Fig. 8); as the lat-
eral load increases, for a top displacement of 0.283 m, the eastern
span cable yields (event 20 ); for event 30 the appearance of the first
plastic hinge at the tower occurs for a base shear of 8977 kN, very
similar to the base shear associated to the yielding of the tower
when the analysis is performed in the WE direction; almost simul-
taneously the eastern span cable fails (event 40 ). Finally, for event 50
the brittle failure of the plastic hinge in the tower that appeared for
30 leads to the collapse of the structure when the base shear
reaches its maximum value of 9931 kN. The analysis is stopped
at event 60 prior to the appearance of very large displacements.
Once again, damage on the structure mainly concentrates in the
vicinity of the tower-deck connection.
4. Input ground motions
The focus of this study is to assess the effectiveness of passive
damping and seismic isolation systems in improving the longitudi-
nal seismic response of the bridge under study, which was mostly
responsible for the damage experienced in a past earthquake. The
orientation of the control and isolation elements considered herein
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252 237

Fig. 6. First four normal modes of vibration of the original bridge model.

Table 1
Experimental versus numerical natural periods for the lower flexural modes. Experimental damping ratios.

Mode no. Mode type Ti experimental (s) Ti numerical (s) ErrorTi (%) fi experimental
1 Longitudinal bending 1.85 2.11 14.0 0.0369
2 Longitudinal bending 0.85 0.87 2.4 0.0296
3 Longitudinal bending 0.57 0.56 1.8 0.0175

is such that they would almost not contribute to the reduction of are beyond the scope of this study. In order to analyze the
the seismic response of the bridge in the vertical direction. For dynamic performance of the bridge and investigate the possibility
this purpose other dedicated elements should be installed and of retrofitting the structure, one-dimensional horizontal ground
238 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252

1.00
Pushover WE (a) Pushover EW (b)
5'
3 4'
1 3'
0.75
|Rx| / W 2

0.50 2'

1'

0.25

4
6'
0.00
0.00 0.25 0.50 0.75 1.00 1.25 1.50 0.00 0.50 1.00 1.50 2.00 2.50
|utop - utop,st| /H (%) |utop - utop,st|/H (%)

Fig. 7. Base shear at tower support versus top of the tower horizontal displacement during pushover analysis. Load profile applied in the (a) West–East direction and (b) East–
West direction.

Fig. 8. Location of first plastic hinges and brittle failure in deck and tower during pushover analysis.

acceleration is assumed to be applied at the bridge supports. For
this purpose, 20 historical ground motions ranging from 6 to 7.3
in magnitude, which were scaled to match the 10% probability of
exceedance in 50 years uniform hazard spectrum for Los Angeles
[38], have been selected. These records were provided within the
SAC Joint Venture Steel Project [37] and are derived from the nor-
mal and perpendicular components of the 1940 Imperial Valley,
1992 Landers, 1989 Loma Prieta and 1996 North Palm Springs
earthquakes that have all occurred in Southern California. In addi-
tion, a particular near field ground motion, is selected as well as
input excitation. This input motion, derived from one horizontal
component of the ground motion recorded at the Rinaldi station
(distance 7.5 km) during the 1994 Northridge earthquake
(Moment Magnitude = 6.7), is characterized by a strong ground
acceleration pulse.
Table 2 summarizes the far field (LA01–LA20) and the near field
(NF13) ground motions main parameters. Columns 4–10 of Table 2
stand for the peak ground acceleration (pga), the root mean square
of acceleration (aRMS), the Arias intensity (IA), the Housner intensity
(HI), the significant duration referred to the 5% and 95% of the total
Arias intensity (Ds) and the predominant and the mean periods (Tp
and Tm).
In order to get some insight of the frequency distribution of the
5. Dynamic performance of original bridge structure
In this section the dynamic performance of the original bridge
structure is presented for the set of 21 earthquake records consid-
ered. The dynamic response of the bridge is numerically integrated
under the action of each ground motion. In order to compare the
performance of the bridge structure in its original configuration
and incorporating the retrofitted options (discussed later in this
paper), the following result quantities are obtained for each earth-
quake record:
– Maximum curvature ductility demand in the deck elements:
ldeck
/;max
– Maximum curvature ductility demand in the tower elements:
ltower
/;max
– Maximum elongation ductility demand in the cable elements:
lcables
D;max
– Total energy dissipation through plastic deformation of struc-
tural elements: Eps
– Seismic input energy at the end of each record: EI
– Maximum deck horizontal relative displacement measured at
the east abutment: urel
max;deck

seismic energy associated to the events under consideration, exact
elastic absolute acceleration spectra have been plotted for 1%
damping in Fig. 9, along with the first four natural periods of the
structure. The performance of the bridge is presented and analyzed
under the horizontal action of these 21 records in the following
sections.
– Maximum base shear at the tower fixed support: Rx,max
– Maximum bending moment of the tower at the deck level:
M tower
z;max
Fig. 10 represents the levels of maximum ductility demand
reached in the deck, tower and cables under each earthquake
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252 239

Table 2
Ground motion parameters [39].

Record Ground motion Station pga (g) aRMS (g) IA (m/s) HI (m) Ds (s) Tp (s) Tm (s)
LA01 Imperial Valley 1940 EC Valley Irr Dist 0.46 0.08 5.47 2.44 24.26 0.52 0.65
LA02 Imperial Valley 1940 EC Valley Irr Dist 0.68 0.09 7.00 2.55 24.52 0.26 0.56
LA03 Imperial Valley 1940 EC Array 5 James Road 0.39 0.05 1.78 2.38 8.52 0.16 1.12
LA04 Imperial Valley 1940 EC Array 5 James Road 0.49 0.05 1.65 1.55 7.09 0.34 0.75
LA05 Imperial Valley 1940 EC Array 6 Houston Road 0.30 0.05 1.28 1.99 10.80 0.06 1.44
LA06 Imperial Valley 1940 EC Array 6 Houston Road 0.24 0.04 1.00 1.36 11.23 0.3 0.79
LA07 Landers 1992 Barstow-Vineyard & H 0.42 0.04 2.19 1.78 22.26 0.72 0.95
LA08 Landers 1992 Barstow-Vineyard & H 0.43 0.04 2.26 1.74 22.24 0.32 0.78
LA09 Landers 1992 Yermo Fire Station 0.52 0.06 4.24 3.11 19.66 0.68 0.92
LA10 Landers 1992 Yermo Fire Station 0.36 0.05 3.40 2.05 20.72 0.22 0.67
LA11 Loma Prieta 1989 Gilroy Sewage plant 0.67 0.08 4.41 3.00 11.32 0.22 0.63
LA12 Loma Prieta 1989 Gilroy Sewage plant 0.97 0.10 6.58 1.67 6.40 0.2 0.37
LA13 Northridge 1994 County Fire Station 0.68 0.08 5.27 2.84 5.70 0.32 0.58
LA14 Northridge 1994 County Fire Station 0.66 0.08 5.31 3.27 5.52 0.3 0.66
LA15 Northridge 1994 Rinaldi Receiving Station 0.53 0.12 3.19 2.60 7.80 0.4 0.64
LA16 Northridge 1994 Rinaldi Receiving Station 0.58 0.13 3.95 3.43 7.03 0.3 0.79
LA17 Northridge 1994 Sylmar, Olive View 0.57 0.05 2.40 2.73 7.20 0.32 0.94
LA18 Northridge 1994 Sylmar, Olive View 0.82 0.07 5.02 3.60 5.30 0.36 0.70
LA19 North Palm Springs 1986 North Palm Springs 1986 1.02 0.09 8.24 1.83 8.54 0.16 0.35
LA20 North Palm Springs 1986 North Palm Springs 1986 0.99 0.11 10.97 3.56 6.78 0.22 0.40
NF13 Northridge 1994 Rinaldi Receiving Station 0.89 0.19 8.09 4.87 7.02 0.78 0.80

4.0 3.0
T4 T3 T2 T1 T4 T3 T2 T1
(a) (b)
3.0 LA01
2.0 LA07
LA02
LA03 LA09
Sa (g)

2.0
LA04
LA10
LA05
1.0
LA06
1.0

LA08
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0

5.0
T4 T3 T2 T1 T4 T3 T2 T1
(c) (d)
4.0 LA13
LA14
LA20
3.0 LA18
Sa (g)

2.0 LA15
LA16 NF13
LA11
1.0 LA12 LA19
LA17
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
T (s) T (s)

Fig. 9. Absolute acceleration response spectra (1% damping) for input records. — Bridge structure natural periods: T1 = 2.11 s, T2 = 0.87 s, T3 = 0.56 s, T4 = 0.34 s.

record. These values have been normalized dividing them by the under the LA16 record in the eastern span cable; the ultimate elon-
ultimate ductility capacity in each case. The structure remains gation ductility capacity is 2.5 for the cable elements.
elastic only under the action of the LA04 ground motion; it com- Fig. 12 represents the seismic input energy and the plastic
pletely collapses for the near field NF13 record, and experiences strain energy due to yielding of structural elements of the original
severe yielding under most of the remaining earthquakes. For all bridge under the 20 far field records. These two quantities have
earthquake records, yielding mainly concentrates in the connec- been expressed in terms of equivalent velocities as per:
tion between the tower and the deck, as indicated in Fig. 11. This sffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffi
is consistent with the results obtained earlier from the pushover 2EI 2Eps
VI ¼ V ps ¼ ð2abÞ
analysis. Apart from the NF13 record, the highest curvature ductil- M2 M2
ity demand reaches 4.49 in the deck and 1.95 in the tower under
the LA16 and LA02 records, respectively; note that the ultimate where M2 stands for the modal mass associated to the second mode,
curvature ductility capacity is 6.0 for the deck and tower elements. with a relevant contribution to the longitudinal seismic response of
For the cables, the maximum elongation ductility reaches 1.68 the bridge.The input energy is highest under the LA02 record. This
240 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252

deck tower cables

,max /6 ,max /6 ,max /2.5

Maximum ductility demand

LA01

LA02

LA03

LA04

LA05

LA06

LA07

LA08

LA09

LA10

LA12

LA13

LA14

LA15

LA16

LA17

LA18

LA19

LA20
LA11
Fig. 10. Maximum ductility demands in deck, tower and cable elements of the original bridge structure normalized by ultimate ductility capacities.

Fig. 11. Location of maximum ductility demands in deck, tower and cables of the original bridge structure and associated earthquake records.

record, followed by the LA16, leads to the maximum energy dissipa- displacement of the deck at the east abutment (instants 3 and 50 of
tion through plastic deformation of the structural members. In Fig. 7). In both cases results from the West–East and East–West
order to provide an adequate quantification of the structural dam- pushover analyses are averaged. Finally b is a non-negative
age accumulated under each earthquake record, taking into account non-dimensional constant that should be calibrated from experi-
both the maximum ductility demand and the cumulative hysteretic mental results for the system under consideration. A value of
energy associated to the number of cycles, Park and Ang damage 0.025 has been assumed for the steel structure on the basis of the
indices [40] have been calculated as per: recommendation of Park et al. [41]. Fig. 13 represents the value of
  the damage index DIPA for the 20 far field records. Apart from the
 rel;sw  near field event which in a few seconds provokes the collapse of
umax;deck  E
DIPA ¼   þ b  ps  ð3Þ the structure, the LA16, followed by the LA02 record, leads to the
PO   PO   PO 
uult;deck  Rx;yield   uult;deck  highest damage index. The main objective of the retrofit with pas-
sive supplemental damping elements is to dissipate the seismic
energy mechanically in order to reduce the energy dissipation asso-
In Eq. (3) urel;sw
max;deck is the maximum relative displacement of the East
ciated to plastic deformation of the structural members, thereby
abutment for each record measured from the static equilibrium
reducing the accumulated damage in the structure and the
configuration. On the other hand, RPO PO
x;yield and uult;deck are obtained maximum ductility demand. On the other hand, seismic isolation
from the pushover analysis. RPO
x;yield is the base shear at the first yield
techniques aim to directly reduce the seismic input energy trans-
(instants 1 and 10 of Fig. 7), and uPO mitted to the bridge structure.
ult;deck corresponds to the ultimate

Plastic strain energy Input energy

VI (m/s), Vps (m/s)

LA01

LA02

LA03

LA04

LA05

LA06

LA07

LA08

LA09

LA10

LA12

LA13

LA14

LA15

LA16

LA17

LA18

LA19

LA20
LA11

Fig. 12. Seismic input energy and plastic strain energy expressed in terms of equivalent velocities for the original bridge structure.
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252
Fig. 14 shows the computed maximum response parameters of
the original bridge structure under all the far-field ground motions
considered in terms of maximum deck relative displacement, max-
imum base shear and maximum bending moment of the tower at
the deck level. These variables have been normalized in order to
facilitate the comparison with analogous representations included
in the following sections:
 
 rel   
umax;deck  Rx;max 
rel
^ max;deck ð%Þ ¼
u   ^
 100 Rx;max ð%Þ ¼  100
uabs  jRx jpov er
deck pov er
 
 tower 
M z;max 
^ tower
M z;max ð%Þ ¼  tower   100
M  the longitudinal contraction and expansion of the steel core when
z pov er
where the quantities in the denominator stand for the total dis-
placement of the east abutment, the maximum base shear, and
the maximum bending moment of the tower at the deck level all
of them obtained during the pushover analysis for the time instant
when the maximum base shear is reached (events 3 and 50 of
Fig. 7a and b), and averaging the results between the West–East
and East–West pushover analyses.
In Fig. 14a the maximum horizontal relative displacement of
the deck, which takes place at the east abutment, ranges from
0.17 m under the LA04 record to 0.32 m under the LA16 record.
This trend is in clear correspondence with the observed maximum
yielding levels presented in Fig. 10. Fig. 14b shows the maximum
base shear computed at the tower support. The maximum value
reaches 9995 kN and occurs under the LA17 record. Finally in
Fig. 14c, maximum values of the bending moment of the tower
at the deck level are shown. The maximum bending moment
reaches 89,140 kNm and occurs under the LA16 ground motion.
The above response parameters of the original bridge structure
will be used in the following sections to evaluate the beneficial
effect of the different passive control retrofit strategies on the seis-
mic response of the cable-stayed bridge.
6. Retrofit of the bridge with metallic dampers
First, the possibility of improving the seismic response of the
bridge by incorporating Metallic Dampers (MDs) is evaluated.
MDs belong to the category of displacement-activated supplemen-
tal damping systems. These elements, when connected between
two parts of a structure experiencing a certain relative movement,
take advantage of the hysteretic behaviour of metals when
deformed into the post-elastic range to dissipate energy. MDs are
usually connected to a structure through stiff bracing members,
generally designed to remain elastic, and exhibit a hysteretic beha-
viour that can be idealized by an elastic-perfectly plastic load–dis-
placement relationship, as that shown in Fig. 15a, where Fa is the
DIPA
LA01
LA02
LA03
LA04
LA05
LA06
LA07
Fig. 13. Park and Ang damage index of the original bridge structure under the far field records.
241
load that activates the damper (or yield load), and the elastic stiff-
 is associated with the bracing elements axial stiffness
ness, k,
and/or the elastic stiffness of the yielding device itself. Particular
elements that fall into this category and that could be installed
in the bridge under study are buckling restrained braces (BRBs).
One of such BRB is the unbonded brace [42] shown in Fig. 15b. It
consists of a steel core member encased in a steel tube filled with
concrete. The steel core carries the axial load while the outer tube,
via the concrete, provides lateral support and prevents global
buckling. A thin layer of lubricating material along the steel core
at the concrete interface eliminates shear transfer and localized
strain concentrations during the elongation, while accommodating
ð4abcÞ
it yields both in tension and in compression.
For the bridge structure under study, it is assumed that MDs are
connected between two locations of the structure and dissipate
energy through tension–compression cycles. Bracing the bridge,
prior to the elements activation, will lead to a reduction in its nat-
ural periods. Earlier numerical studies [43] have indicated that the
best response of a structure retrofitted with hysteretic dampers is
obtained for low braced periods, that is, very stiff braces; therefore
high k values should be considered for the final retrofit solution.
For this reason, it is proposed to retrofit the bridge by connecting
the dampers through the braces between the tower and the bridge
deck instead of between the abutments and the deck, as longitudi-
nal deformations of the deck from service loads should not be pre-
vented. Two alternatives are considered in order to connect the
MDs to the structure as shown in Fig. 16: (i) connecting the MDs
underneath the deck between each tower support and two sections
of the girder on each side of the tower (marked in dashed trace),
and (ii) introducing the MDs above the deck level connecting a par-
ticular section of the tower to two girder sections (marked in
dotted-dashed trace).
As the bridge deck is at 4% slope and the dampers are supposed
to be incorporated into the structure after the latter has already
deformed under its own self-weight, special attention is given dur-
ing the modelling process in order to avoid that the devices carry
part of the gravity loads, which would not represent reality. To this
end, the dynamic analysis is performed in the absence of
self-weight. The deformed geometry of the bridge under its
self-weight is first obtained in the original configuration. The inter-
nal forces due to the self-weight are determined in all the mem-
bers, and the yield limits are modified so that an equivalent
hysteretic behaviour takes place when the self-weight loads are
eliminated. The hysteretic damper elements are then introduced
in the deformed configuration and the bridge response is obtained
under the earthquake records.
Christopoulos and Filiatrault [27] presented the
non-dimensional parameters governing the optimum activation
LA08
LA09
LA10
LA12
LA13
LA14
LA15
LA16
LA17
LA18
LA19
LA20
LA11
242 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252

(a)

max,deck
^u rel

(b)
Rx,max
^

(c)
^ tower
Mz,max

LA01
LA02
LA03
LA04
LA05
LA06
LA07
LA08
LA09
LA10

LA12
LA13
LA14
LA15
LA16
LA17
LA18
LA19
LA20
LA11

Fig. 14. Maximum response parameters of the original bridge structure: (a) maximum deck horizontal relative displacement, (b) maximum base shear at tower support, and
(c) maximum bending moment of the tower at the deck level.

Fig. 15. (a) Idealized load displacement relationship for MDs. (b) Components of an unbonded brace [27].

load of a hysteretic damper for a single storey structure under har-
monic ground motion. An important conclusion of this work is that
the optimum activation load depends on the frequency and ampli-
tude of the ground motion. For this reason, the retrofitted bridge is
first analysed under the LA16 earthquake record. This record lead
to the highest Park and Ang damage index in the original bridge
configuration under the far field records (Fig. 13). A parametric
study is conducted under the LA16 record for selecting the most
appropriate geometrical configuration of the MDs shown in
Fig. 16 and obtaining optimum values for the brace stiffness and
activation load. Each MD along with its connecting brace element
is introduced in the model as an axial tension–compression spring
with elastic-perfectly plastic behaviour.
From the results of this parametric study, the following conclu-
sions can be drawn: (i) the maximum ductility demand in the gir-
der, generally concentrated in element 15 (see Fig. 11), reduces
with an increase of the activation load of the MDs. For very high
activation loads, however, the level of forces transmitted by the
braces to the girder-deck junction become excessive and yielding
starts to occur at the tower close to the junction (generally in
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252
Fig. 16. Evaluated retrofit configurations with MDs. Thick trace: selected final design.
element 30 in Fig. 11). The activation load leading to the minimum
ductility demand in the girder while preventing yielding in the
tower is considered to be the optimum activation load of the
MDs; (ii) the ductility demand reduces with an increase of the
brace size of the MDs; (iii) bracing the structure from underneath
the deck leads to a better structural performance than introducing
the dampers above the deck; and (iv) the MD retrofit is more effec-
tive with an opening of the bracing elements, i.e. with an increase
of the angles a1 and a2 in Fig. 15. Nevertheless, as the length of the
braces increases, the buckling load decreases and larger sections
are required for these elements.
Based on the results of the parametric study under the LA16
record while keeping the brace sections to a minimum, the final
retrofit solution adopted is shown by the solid trace in Fig. 16.
Two MDs are introduced per tower underneath the deck connect-
ing the tower base with two sections of the girder forming angles
a1 = 57.5° and a2 = 60.2°, respectively. Steel hollow tube sections of
762 mm external diameter and 30 mm thickness are chosen for
both braces. An activation load of 2500 kN identical for all MDs
is selected. This load limits the maximum axial force in the bracing
members, which are designed such that buckling is prevented with
an adequate margin of safety. If a modal analysis is performed once
the bracing elements are included, the elastic natural periods of the
bridge structure reduce from the values shown in Fig. 9 to
T1 = 1.876 s, T2 = 0.703 s, T3 = 0.449 s and T4 = 0.299 s.
Retrofitting the bridge structure with the optimal MD configu-
ration improves significantly its longitudinal seismic response.
Collapse of the structure is now prevented under the near field
(NF13) record. The structure remains elastic or almost elastic
underground motions LA01 to LA11, LA16 and LA17, as shown in
Fig. 17. The maximum curvature ductility demand reaches only
2.44 in the girder during the NF13 event, while the maximum elon-
gation ductility demand in the cables reaches only 1.31 under the
same near field record. The tower remains elastic under all ground
motions. In general, the maximum demand decreases substantially
when compared with the original bridge configuration (see
Fig. 10).
Fig. 18 shows a comparison of the maximum response parame-
ters across all ground motions for the bridge structure retrofitted
with MDs (in black columns) with that of the bridge original con-
figuration (grey columns). The maximum displacement, base shear
and bending moment of the tower at the deck level have been nor-
malized as per Eq. (4) again.
The overall maximum deck relative displacement for all the far
field records takes place at the east abutment and reduces from
0.318 m to 0.111 m. The average reduction in maximum deck rel-
ative displacement across all 20 far field records reaches 79%,
showing the stiffening effect of the braces prior to yielding of the
243
MDs. In the case of the near field (NF13) record, which led to the
collapse of the original configuration of the bridge structure, the
maximum deck relative displacement when retrofitted with MDs
reaches 0.174 m. The maximum shear force in the tower base
reduces from 10,000 kN for the original bridge configuration (see
Fig. 13) to 5083 kN when the bridge structure is retrofitted with
MDs. The average shear force reduction across all 20 far field
records is 68%. The maximum base shear under the near field
record reaches 9650 kN. Finally, the maximum bending moment
in the deck (Fig. 18c) decreases from 89.1 MN m for the original
bridge configuration to 48.4 MN m for the bridge retrofitted with
MDs. The maximum bending moment in the bridge deck of the ret-
rofitted configuration takes place under the near field record,
reaching 90.0 MN m. The maximum bending moment shows an
average reduction of 67% across the 20 far field records after the
introduction of the MDs.
Note that in the proposed configuration the MDs axial forces are
transmitted to the pier foundation. Therefore, the horizontal force
that should be borne by the foundation exceeds the pier base shear
shown in Fig. 18b. In particular, under the near field record, which
is the worst scenario in this regard, the maximum horizontal force
transmitted to the foundation reaches 12,933 kN, exceeding by
approximately 31% the maximum horizontal force transmitted to
the foundation in the original configuration under the far field
records.
7. Retrofit of the bridge with fluid viscous dampers
In this section, retrofitting the cable-stayed bridge with Fluid
Viscous Dampers (FVD) is considered. In a FVD, energy is dissipated
by forcing the flow of silicone oil through small orifices in a piston
cylinder. The force produced by a FVD, FD, in the range of frequen-
cies of interest for earthquake response of structures, may be con-
sidered a function of the relative velocity between the device ends:
F D ¼ signðxÞC _a
_ D jxj ð5Þ
where CD is the damper constant, x_ the relative velocity between the
ends of the damper and a a velocity exponent, primarily controlled
by the design of the piston head orifices, and ranging between 0.5
and 1. When a < 1 a FVD behaves as a nonlinear viscous element.
The advantage of designing a FVD for a < 1 is the reduction of the
damper force at high velocities, in order to avoid overloading the
damper itself or the bracing system it connects to [44]. In this
bridge case study, linear FVDs (a = 1) are considered, as the
expected maximum forces are moderate. In a linear FVD subjected
to relative harmonic displacement, the maximum induced force is
linearly proportional to the excitation frequency, to the displace-
ment amplitude and to the damping constant. Fig. 19 illustrate
244 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252

deck tower cables

Maximum ductility demand

,max /6 ,max /6 ,max /2.5

NF13
LA01
LA02
LA03
LA04
LA05
LA06
LA07
LA08
LA09
LA10

LA12
LA13
LA14
LA15
LA16
LA17
LA18
LA19
LA20
LA11
Fig. 17. Maximum ductility demands in deck, tower and cables for the bridge retrofitted with MDs normalized by ultimate ductility capacities.

Original configuration Retrofitted with MDs (a)

max,deck
^u rel

(b)
Rx,max
^

(c)
^ tower
Mz,max

LA01
LA02
LA03
LA04
LA05
LA06
LA07
LA08
LA09
LA10

LA12
LA13
LA14
LA15
LA16
LA17
LA18
LA19
LA20
NF13
LA11

Fig. 18. Comparison of maximum response parameters of the bridge retrofitted with four MDs with Fa = 2500 kN with that of the original bridge structure: (a) maximum deck
horizontal relative displacement, (b) maximum base shear at tower support, and (c) maximum bending moment of the tower at the deck level.

the theoretical hysteresis response of a linear FVD under harmonic condition, FVDs do not exert any force, thereby not preventing the
excitation for different damping constants and forcing frequencies. structure dilatations and service loads related displacements.
The maximum developed force does not coincide with the maxi- In order to evaluate the effectiveness of retrofitting the bridge
mum relative displacement. This out-of-phase characteristic structure with FVDs, the response of the bridge is evaluated under
between peak force and displacement constitutes a clear advantage the 21 earthquake records considered. As only half of the bridge is
of FVDs when compared to other control devices, as lower forces are modelled, only two linear FVDs are introduced between the abut-
introduced in the structure as a consequence of the retrofitting pro- ment walls and rigid connectors fixed to the deck girders at a dis-
cess. In the particular bridge structure under study, connecting the tance of 2.5 m towards the centre of the bridge, in order to take
two box girders to the east and west abutments by means of hori- into account the dampers length and connecting elements. When
zontal FVDs is proposed, as shown in Fig. 19c. Under static loading the retrofitted bridge response is analysed, it is observed that both
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252
(a) CD =11
FD (kN) CD =3 MNs/m
=2
(b)
FD (kN)
CD =8.5
MNs/m
x (m)
Fig. 19. Hysteretic behaviour of a linear FVD under harmonic excitation for different (a) CD values and (b) excitation frequencies x; (c) connection of added FVDs at west
abutment.
the structural displacements and internal forces monotonically
decrease with increasing values of the damping constants of the
FVDs. Therefore optimal FVDs constants are selected as the mini-
mum values leading to an elastic or almost elastic response of
the bridge for the 20 far field motions considered, and minimum
yielding for the near field record (NF13) considered. After a few
iterations, CD = 8500 kN s/m was selected as the optimum damping
constant for each FVD. The installation of four FVDs with this con-
stant value corresponds to a 58.8% damping ratio in the second
mode of vibration, which contributes the most to the longitudinal
response These optimum damping constants allow the bridge
structure to remain elastic under all the ground motions except
for the LA18 and the NF13 records. For these two records, the max-
imum curvature ductility demands reach 1.58 and 1.36, respec-
tively, in the deck girders close to the deck-tower connection.
Neither the tower nor the cables yielded. The retrofit of the bridge
with FVDs not only avoids collapse of the structure under the near
field ground motion but results in an almost elastic response of the
structural members.
Fig. 20 shows a comparison of the response parameters for the
bridge structure retrofitted with the FVDs with that of the original
bridge configuration. The average maximum horizontal relative
displacement taken across the 20 far field ground motions, which
occurs at the east abutment, reduces on average by 38% with
respect to that of the original bridge structure. Note that for the
bridge structure retrofitted with FVDs, the maximum horizontal
displacement remains fairly constant across the various far field
motions with a maximum value of 0.172 m. For the maximum base
shear and maximum bending moment, the effect of the retrofit is
evident. For the far field records, the average maximum base shear
reduces by 65.4% compared to the original bridge structure and
reaches a maximum of only of 4133 kN, less than half of its value
before the retrofit. A similar reduction is observed for the maxi-
mum bending moment of the tower at the deck level, where the
maximum value reduces from 89.1 MN m for the original bridge
structure to 39.2 MN m for the bridge retrofitted with FVDs. As
for the bridge response under the near field record, the maximum
response is considerably lower than in the MD case, which is
related to the very good response of the FVDs under high velocity
pulses. The maximum relative displacement, base shear and bend-
ing moment at the tower reach 0.19 m, 5961 kN and 56.7 MN m,
respectively. A comparison between Figs. 20 and 18 show that
245
(c)
MNs/m
Box girder
=5
=4
=3
FVDs
West abutment
=
=2
the maximum response of the bridge is less affected by the input
ground motion when the retrofit is conducted with FVDs than with
MDs. This very desirable behaviour obtained with FVDs coincides
with a maximum force of 4143 kN and 4000 kN in the West and
East damper, respectively, which take place under the action of
the near field ground motion, and an average maximum damper
force at each of the two abutments of 2396 kN and 2267 kN under
the far field records. The maximum forces in the devices are sub-
stantially lower than the maximum capacity of commercial FVDs
nowadays [45]. Nevertheless the FVDs should be properly
anchored to the abutments, and these should be strengthen at
the connections if needed in order to absorb the predicted level
of transmitted forces.
8. Retrofit of the bridge with tuned mass dampers
In this section, retrofitting the cable stayed bridge using tuned
mass dampers is considered. TMDs or vibration absorbers are rel-
atively small mass-spring-dashpot units calibrated to be in reso-
nance with a particular mode of the structure, with the aim of
reducing its associated dynamic response. The devices must be
connected close to a location of maximum modal amplitude, and
are excited by the structural motion. In the particular case under
study, in order to install one or more unidirectional TMDs to mit-
igate the longitudinal sway of the bridge, two possibilities are
envisaged: (i) connecting the TMDs underneath the bridge deck
by means of a guiding system permitting the mass to slide longitu-
dinally with respect to the deck. In this case the TMDs could be
installed close to the abutments leaving the required clearance to
accommodate the maximum predicted displacements (see marks
1 and 2 in Fig. 21), and the sliding mass should be connected
between two fixed sections of the deck trough linear
spring-dashpot elements; or (ii) installing a TMD as a pendulum
damper hanging from the top of the bridge tower provided that a
transverse beam is added, from which the mass should be sus-
pended, as well as an auxiliary simple truss structure. The pendu-
lum length should be selected to achieve the desired TMD natural
frequency and a viscous damper should be connected between the
mass and the auxiliary structure (see mark 3 in Fig. 21).
Six TMD retrofit strategies were considered in this study, as
shown in Table 3. In three of them the installation of only one
246 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252

Original configuration Retrofitted with FVDs (a)

max,deck
^u rel

(b)
Rx,max
^

(c)
^ tower
Mz,max

LA01
LA02
LA03
LA04
LA05
LA06
LA07
LA08
LA09
LA10

LA12
LA13
LA14
LA15
LA16
LA17
LA18
LA19
LA20
LA11

NF13

Fig. 20. Maximum response of the bridge structure retrofitted with four FVDs with CD = 8500 kN s/m: (a) maximum deck horizontal displacement, (b) maximum base shear
at tower support, and (c) maximum bending moment of the tower at the deck level.

TMD is proposed, and in the other three the installation of two
TMDs tuned to the same or different bridge natural frequencies
is proposed. In all the cases, except one, one TMD is tuned to the
second mode of the bridge, as it has the highest effective mass
and contributes the most to the structural dynamic response.
In order to evaluate the effectiveness of each configuration and
select the most appropriate one, the LA16 ground motion was
again considered. As in the previous cases, only one half of the
bridge was modelled and, as a first approximation, kinematic con-
straints were introduced so that the concentrated TMD masses
undergo the same vertical displacement but a relative horizontal
displacement with respect to the node to which these are con-
nected. Therefore, TMDs connected at locations (1) and (2) in
Fig. 21 were modelled as point masses, with horizontal movement
only and connected to a linear spring-damper element at nodes 2
and 19, respectively (see Fig. 4); the dead weight of each mass is
applied to these nodes. On the other hand, the pendulum TMD
was simulated as a first approximation as a concentrated mass
with horizontal movement only connected to node 21 through a
horizontal linear spring-dashpot element as well. Again, the dead
weight associated to this TMD is directly applied to the top of
the tower.
In each of the alternatives considered, either in the single
TMD or the multiple TMD case, the initial damper properties
were selected as follows. First the TMD mass, mTMD, was defined
as a small ratio, l, of the modal mass Mi associated to the
tuned mode (l = 0.02 was selected as a starting value). Then,
the optimum frequency and damping ratios between the TMD
and the bridge modal properties (g and f) were determined
by applying the optimum tuning conditions for minimizing the
root mean square value of the primary structure relative dis-
placement subjected to random load, as given in Eq. (6)
[46,27]. In Eq. (6) xTMD and cTMD are the TMD frequency in rad/s
and damping constant, respectively, and cc is the TMD critical
damping, cc = 2xTMD mTMD.
qffiffiffiffiffiffiffiffiffiffiffi
1  l2
mTMD ¼ l Mi g ¼ ! xTMD ¼ g 2p f i
1þl
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
lð1 þ 34lÞ
f¼ ! cTMD ¼ f cc ð6Þ
4ð1 þ lÞð1  l2Þ
where fi is the bridge mode frequency. Finally, the TMD stiffness
coefficient was obtained as per kTMD = xTMD2mTMD. In each configu-
ration considered, the TMD mass was progressively increased from
the initial l = 0.02 value and the bridge maximum ductility demand
was obtained under the LA16 record. Following this procedure, the
following conclusions were obtained: (i) when the TMD is installed
as a pendulum system (configurations 2 and 3), the maximum duc-
tility demand experienced by the structure under the LA16 record
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252
Fig. 21. TMD locations considered.
Table 3
Definition of possible retrofit alternatives using TMDs analyzed (optimum configu-
ration highlighted).
monotonically reduces with the mass ratio, l. In this case the struc-
ture performance is considerably better when the pendulum dam-
per is tuned to the second mode, rather than to the fundamental
mode, as it was expected. In the second configuration (tuning to
the 2nd mode), when l = 0.15 the maximum curvature ductility
demand reduces by 56.6% with respect to the original bridge config-
uration, while in the third one (tuning to the 1st mode) this reduc-
tion reaches only 32.3%. Moreover the required mass when the TMD
is tuned to the 2nd mode is only 60% of that when the TMD is tuned
to the first mode; (ii) the effectiveness of the retrofit when horizon-
tal TMDs are attached to the main girder by means of a guiding sys-
tem (cases 1, 4, 5 and 6) is considerably lower than that of the
pendulum system. The reduction of the maximum ductility experi-
enced by the superstructure reduces at a smaller rate with l and is
not monotonic. When the mass ratio exceeds a certain limit, the
maximum curvature ductility starts to increase again; (iii) the dif-
ference in terms of dynamic structural response between installing
one or two sliding horizontal TMDs tuned to the same frequency is
negligible. Therefore the results from configurations 1 and 6 are
very similar; and (iv) among configurations 4, 5 and 6, the best
performance is obtained for configuration 4, in which one TMD is
tuned to the fundamental frequency and the other one to the sec-
ond modal frequency. Nevertheless, if the best pendulum and the
best sliding TMD alternatives (configurations 2 and 4, respectively)
are compared, for the same mass ratio, l = 0.2, the maximum curva-
ture ductility reduces by 62.6% in the second configuration and in
only 24.28% in the fourth one, with respect to the original bridge
configuration. Moreover, the needed TMD mass in configuration
no. 2 is 37.2% of the corresponding mass needed in the fourth
configuration.
247
Based on the above observations, it was decided to consider as
most feasible TMD retrofit alternative the installation of one pen-
dulum TMD tuned to the second modal frequency of the bridge
(1.15 Hz), a mass ratio l = 0.20 with respect to the second modal
mass, and optimum tuning parameters those given by Eq. (6).
This mass ratio corresponds to a resulting mass of the TMD of
48 ton (2.7 m of side concrete block), which installation is ques-
tionable in practice. Once the final solution was defined, the
response of the bridge under the far field records and the near field
ground motion was determined. Results in terms of maximum
ductility demand are shown in Fig. 22. Collapse is prevented but
the structure remains completely elastic only under only the
LA04 and LA06 records. The tower remains elastic for all ground
motions. The maximum curvature ductility in the girder reaches
3.2 under the LA13 ground motion, and the maximum ductility
demand at the cables after the retrofit equals 1.32. If Fig. 22 is com-
pared with Fig. 17 and with the results of Section 7, it is evident
that structural damage is reduced to a much larger extent when
the bridge is retrofitted with metallic or viscous dampers rather
than with tuned mass dampers and, in this later case, an exces-
sively heavy mass needs to be introduced which installation in this
particular structure is unrealistic.
Finally, Fig. 23 shows a comparison of the maximum response
values reached in the original bridge configuration with that after
the TMD retrofit, in terms of the horizontal deck relative displace-
ment, base shear at the tower and bending moment of the tower at
the deck level for all ground motions. The maximum horizontal
displacement reduces in average by 24.4% under the far field
ground motions, but in a few cases the structure response is similar
or even worse after the TMD is introduced (e.g. LA05 record). The
maximum deck relative displacement after the retrofit reaches
0.25 m and takes place under the near field record. The maximum
base shear at the tower (Fig. 23b) and the maximum bending
moment of the tower at the deck level (Fig. 23c) reduce on average
by 20.22% and 18.33%, respectively under the LA01–LA20 records.
The overall maximum values of these two parameters occur when
the near field record is considered and reach 10,790 kN and
93.58 MN m, respectively. Again, when Fig. 23 is compared with
Figs. 18 and 20 it is evident that the beneficial effect of metallic
and fluid viscous dampers in the bridge seismic performance is
superior than in the TMD case. The main reason for this is the
change of effective period of the bridge structure, which yields
causing the TMD to be de-tuned.
248 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252

tower

Maximum ductility demand

deck cables
,max /6 ,max /6 ,max /2.5

NF13
LA01
LA02
LA03
LA04
LA05
LA06
LA07
LA08
LA09
LA10

LA12
LA13
LA14
LA15
LA16
LA17
LA18
LA19
LA20
LA11
Fig. 22. Maximum ductility demands in deck, tower and cables for bridge retrofitted with TMDs normalized by ultimate ductility capacities.

Original configuration Retrofitted with TMDs (a)

max,deck
^u rel

(b)
Rx,max
^

(c)
^ tower
Mz,max

LA01
LA02
LA03
LA04
LA05
LA06
LA07
LA08
LA09
LA10

LA12
LA13
LA14
LA15
LA16
LA17
LA18
LA19
LA20
LA11

NF13

Fig. 23. Maximum response of the bridge retrofitted with horizontal TMD: (a) maximum deck horizontal displacement, (b) maximum base shear at fixed support, and (c)
maximum bending moment of the tower at the deck level.

9. Seismic isolation of the bridge structure main girders rest on friction pendulum bearings (FPBs) at the cen-
tral pier; and (iii) the central pier is fixed at the base (see Fig. 24).
In this last section, the dynamic performance of an identical The FPB is a friction-type sliding bearing that uses gravity as the
cable-stayed bridge seismically isolated is analysed under the 21 restoring force. The system consists of an articulated friction slider
earthquake records. Even though this solution would imply a that travels on a spherical concave lining surface (see Fig. 25a).
major modification of the structural system in a real retrofit situa- Each deck girder rests on the bearing top plate, which accommo-
tion, it is included for comparison purposes with the previously dates the articulated slider. This connection avoids the transmis-
described control solutions. The seismic response of the cable sion of bending moments to the substructure. The sliding
stayed bridge is evaluated considering that (i) the upper and lower horizontal forces, developed as a consequence of the seismic loads,
parts of the tower are disconnected at the deck level; (ii) the deck are transmitted through the bottom plate of the bearings fixed to
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252
the top of the central pier. FPB performance exhibits two phases:
(i) a static phase controlled by friction in the sliding interface, in
which the FPB supported bridge responds as if it were convention-
ally supported, and (ii) a dynamic phase in which the slider travels
on the concave surface leading to high relative displacements
between the deck and the pier top.
An average coefficient of friction in the interface l = 0.05 was
assumed herein, which is a common value used in practice for
PTFE-stainless steel interfaces. The FPB lateral force–displacement
relation is modelled as that of a horizontal spring with bilinear
hysteretic behaviour (see Fig. 25b). A very high value of the initial
horizontal stiffness, k0 = 106 kN/m, is provided in order to repro-
duce the response of a rigid friction spring. The yielding force
and the post-yielding stiffness are computed in terms of the weight
that each FPB supports, W = 9.98 MN, and the radius of the lining
surface R, is set to 1 m. This radius is selected from the acceleration
response spectra (Fig. 9), as it corresponds to an isolated friction-
less structural period Ti in the vicinity of 2 s, as per R = g(Ti/2p)2,
beyond the periods associated with the largest spectral accelera-
tions of all the earthquake records considered.
As for service loads, wind locks were supposed to be installed
on each bearing. These wind locks should be designed to prevent
any displacement of the friction pendulum bearings under static
and non-seismic loading, but would break off under very small
seismic excitation. It was assumed that break away seismic forces
are small enough to prevent the need to introduce them in the
numerical model.
In addition to reducing the ductility demand in the main ele-
ments of the superstructure and substructure, the maximum seis-
mic displacement of the friction pendulum isolator is limited to
±300 mm, in order to avoid pounding between the deck and the
abutments. To increase damping and reduce this displacement,
two linear fluid viscous dampers per tower pier are inserted con-
necting each pier with the main girder. The damping constants
(CD = 2000 kN s/m for each damper) were determined for the par-
ticular FPB system to reduce the maximum displacement below
the limitation under all the ground motions.
The seismic response of the bridge was then obtained under
the 21 records. The isolated structure remains elastic under all
the ground motions, including the near field event. Fig. 26 com-
pares a summary of the results obtained for the isolated bridge
with those of the original bridge configuration in terms of (a)
deck horizontal relative displacement at the east abutment, and
(b) maximum base shear at the pier foundation. In Fig. 26c the
maximum bending moment at the pier base has been included
as well. The maximum horizontal relative displacement at the
east abutment reduces on average by 57.7% (maximum value
Fig. 24. Seismically isolated bridge model.
249
now of 0.18 m) with respect to that of the original bridge config-
uration for the LA01–LA20 records. The base shear greatly reduces
when the isolation system is incorporated. The average reduction
reaches 60% with a maximum value of 5263 kN. As for the
response under the near field record, the maximum relative dis-
placement at the east abutment and the maximum base shear
reach 0.22 m and 5981 kN, respectively. Finally, the maximum
bending moment at the foundation takes place under the near
field (NF13) event reaching 76.4 MN, transmitted to the pier foun-
dation. This desirable behaviour is attained with an average max-
imum force of 1022 kN in the FVDs under the far field records,
and a maximum force of 2652 kN under the action of the near
field ground motion. These values are considerably below current
commercial devices maximum capacity. The maximum horizontal
displacement at the friction pendulum bearings takes places
under the far field record and reaches 0.21 m, below the assumed
limit of 300 mm. For this value of maximum displacement, the
effective natural period of the friction pendulum isolated bridge
attains 2.1 s, as it was expected. In this first mode of vibration
the superstructure nearly behaves as a rigid body and the dis-
placement concentrates across the isolation system, leading to a
remarkable reduction in the lateral forces resisted by the
structure.
10. Performance indices definition and comparative results
In this section the bridge response in its original configuration is
compared to that obtained for the different retrofit and isolation
alternatives. First, the total input energy expressed in terms of
equivalent velocity, as per Eq. (2ab), is represented in Fig. 27 for
all the records and final configurations. Under the far field records,
the input energy is reduced to a greater extent in average when the
bridge is retrofitted with FVDs, followed by the MDs alternative
and the isolation solution. In fact, the equivalent velocity reduces
30.3%, 24.3% and 15.3% in the three cases, respectively. This magni-
tude only reduces 5.5% in average when the bridge is retrofitted
with the TMDs, exceeding the input energy when the TMD is
installed its initial value under eight out of the twenty far field
records. These results are consistent with the overall performance
observed for the different solutions and presented in what follows
in the form of performance indices.
For the purpose of comparing the effectiveness of the retrofit
alternatives analysed, five performance indices relating the maxi-
mum response of the retrofitted structure with that of the original
bridge configuration are defined and calculated for the far field
records LA01–LA20 (Eqs. (7a)–(11a)) and the near field record
(Eqs. (7b)–(11b)) as follows:
250 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252

Fig. 25. (a) Single curvature friction pendulum bearing scheme. (b) Theoretical hysteresis loop of the bilinear spring that approximates the FPB dynamic behaviour.

Original configuration Base isolation with FPS (a)

max,deck
^u rel

(b)
Rx,max
^

(c)
^ base
Mz,max

LA01
LA02
LA03
LA04
LA05
LA06
LA07
LA08
LA09
LA10

LA12
LA13
LA14
LA15
LA16
LA17
LA18
LA19
LA20
LA11

NF13

Fig. 26. Maximum response of the bridge seismically isolated: (a) maximum deck horizontal displacement, (b) maximum base shear, and (c) maximum bending moment at
fixed support.

Original MDs FVDs TMDs FPBs

VI (m/s)

LA01

LA02

LA03

LA04

LA05

LA06

LA07

LA08

LA09

LA10

LA12

LA13

LA14

LA15

LA16

LA17

LA18

LA19

LA20
LA11

NF13

Fig. 27. Seismic input energy expressed in terms of equivalent velocity for the original configuration and the retrofit and isolation alternatives.
 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252 251

   ret
 rel   rel 
umax;deck ret u max;deck 
LA01-LA20

TMDs
(a)

TMDs
TMDs
LA01—LA20
Iffu ð%Þ ¼    100 Inf
u ð%Þ ¼ 
uabs 
 NF13
 100 ð7abÞ
 rel original

FVDs
umax;deck LA01—LA20 deck pov er

TMDs

TMDs
FPBs

FPBs
   

FVDs
FVDs

MDs
MDs
MDs
Rx;max ret Rx;max ret
IffR ð%Þ ¼ LA01—LA20 nf
 100 IR ð%Þ ¼ NF13
 100 ð8abÞ

MDs


MDs
Rx;max original jRx jpov er
LA01—LA20

FVDs

FVDs

FPBs
FPBs

FPBs
   
 tower ret  tower ret Iu
ff ff ff I
ff
I
ff
M z;max LA01—LA20 Mz;max  IR IM
IffM ð%Þ ¼    100 Inf   NF13  100
M ð%Þ ¼  tower  ð9abÞ
 tower original M NF13

TMDs
M z;max LA01—LA20 z pov er (b)

MDs

MDs
TMDs
TMDs
 ret

FPBs
 

FVDs
 

FVDs
 

FPBs
FVDs
l/;max 

TMDs
l/;max ret

MDs

MDs
LA01—LA20

TMDs
ff nf NF13
Il/ ð%Þ ¼    100 Il/ ð%Þ ¼  100 ð10abÞ

MDs
 original 6:0
l/;max LA01—LA20

FVDs

FVDs
FPBs

FPBs

 
   cables ret
lcables ret lD;max 
D;max LA01—LA20
IfflD ð%Þ ¼    100 Inf
lD ð%Þ ¼
NF13
 100
lcables original 2:5
D;max LA01—LA20
For the far field records, the indices are computed as the ratio
between the mean value across the 20 earthquake records of each
maximum structural response in absolute value (Iffu : relative dis-
placement at the east abutment, IffR : base shear, IffM : bending
moment of the tower at the deck level, Iffl/ : curvature ductility
demand in deck and tower elements, and IfflD : elongation ductility
demand at the cables) to the same quantity reached in the original
bridge configuration. As per the performance of the structure under
the near field event, due to the fact that the cable-stayed bridge col-
lapses in the original bridge configuration, the indices are computed
dividing the maximum response quantities after the retrofit in
absolute value by the total displacement of the east abutment
nf
(Inf
u ), the maximum base shear (I R ), and the maximum bending
moment of the tower at the deck level (Inf
M ) all of them obtained dur-
ing the pushover analysis for the time instant when the maximum
base shear is reached (events 3 and 50 of Fig. 7a and b), and averag-
ing the results between the West–East and East–West pushover
analyses. As for indices Inf nf
l/ and IlD , the maximum ductility demands
reached in the retrofitted cases under the near field record are
divided by 6.0 and 2.5, respectively, which correspond to the ductil-
ity levels associated to the starting of strength degradation (see
Fig. 5).
Fig. 28 shows the values reached by the previously defined per-
formance indices for (a) the far field records and (b) the near field
record corresponding to all the retrofit alternatives. The following
can be observed: (i) the retrofit of the cable-stayed bridge with
MDs leads to the maximum control of the bridge deck relative dis-
placements, both under the far field and under the near field
records. Nevertheless, the difference among the four options anal-
ysed in this regard reduces for the near field record characterized
by a strong ground velocity pulse; (ii) the maximum internal forces
experienced in the structure are similarly reduced with the intro-
duction of MDs, FVDs or FPBs, but again, the performance of MDs
when the structure is subjected to the near field record is worse
in this regard; (iii) the cable-stayed bridge only remains com-
pletely elastic when the structure is isolated, both for the far field
and near field records. Among the remaining three alternatives, the
retrofit with FVDs leads to the highest reduction in curvature duc-
tility demands and elongation levels; and (iv) the beneficial effect
of retrofitting the structure with TMDs is very limited, especially in
regards to the reduction of the seismic internal forces. Even with
the introduction of considerably large masses the improvement
FPBs
nf nf nf nf
Iunf IR IM I I
ð11abÞ
Fig. 28. Performance indices for the retrofit alternatives: (a) far field records and (b)
near field record.
of the bridge performance is substantially poor, when compared
to the other three alternatives.
11. Conclusions
This paper addressed the dynamic behaviour of an existing steel
cable-stayed bridge, which experienced severe damage in a previ-
ous earthquake. The main objective was to assess the possibility of
applying passive supplemental damping and seismic isolation
techniques in order to improve the longitudinal seismic perfor-
mance of the bridge structure. For this purpose, a Finite Element
model of the bridge was developed and the structural response
evaluated through non-linear dynamic response analyses under a
set of 21 seismic records of different intensities, including a near
field ground motion. Strength degradation capabilities were intro-
duced in the model, allowing the occurrence of brittle failure and
force redistribution in all the members.
Four retrofit alternatives were considered: retrofitting the
structure with metallic dampers, fluid viscous dampers, tuned
mass dampers and isolating the super-structure with the introduc-
tion of two friction pendulum bearings between the deck and the
central pier. The original (unretrofitted) model of the structure
experienced severe yielding under the far field records and col-
lapsed under the near field ground motion. For each retrofit alter-
native, the most appropriate configuration of dampers was
selected in order to optimize the beneficial effects of the devices.
Then, the optimal element properties and brace sizes, if needed,
were determined. Finally, an optimum response for each situation
was provided in terms of displacements, internal forces and max-
imum ductility levels reached. Finally, a set of ten performance
indices relating the retrofitted or isolated bridge maximum
response with the original one (or with the pushover results) were
defined and computed. From the analysis of these indices, conclu-
sions were drawn regarding the effectiveness of the different alter-
natives for the particular case under study.
From the analyses performed it was concluded that (i) the
initial collapse of the bridge experienced under the near field
record was prevented with the four proposed alternatives; (ii)
the cable-stayed bridge only remained completely elastic when
the structure was isolated, both for the far field records and
for the near field one; (iii) fluid viscous dampers were the next best
252 M.D. Martínez-Rodrigo, A. Filiatrault / Engineering Structures 99 (2015) 232–252
option in this regard and do not require a major modification of the
initial structure; (iv) the introduction of tuned mass dampers in
this particular case is not effective and may even aggravate the
bridge seismic performance when compared to the original situa-
tion. This is due to the change of effective period of the bridge
structure, which yields causing the tuned mass damper to be
de-tuned; (v) the introduction of metallic dampers is very effective
for the far field records and leads to a remarkable control of the
bridge displacements for all the records. Nevertheless, the perfor-
mance of the retrofitted structure in this case is considerably worse
under the near field event, when compared with the fluid viscous
damper solution or the isolation retrofit, that seem less affected
by input ground motion.
Acknowledgements
The authors express their gratitude to the Spanish Ministry of
Economics and Competitiveness (Ministerio de Economía y
Competitividad, MINECO), for the financial support received in the
framework of the program ‘‘José Castillejo’’ with reference number
JC2011-0304 for mobility of researchers.
References
[1] Public Works Research Institute (PWRI). Seismic design procedure of cable-
stayed bridges. Part I: dynamic characteristics of cable-stayed bridges based on
field vibration test results. Technical report, Tsukuba, Japan; 1986 [in
Japanese].
[2] Ali HM, Abdel-Ghaffar AM. Modeling the nonlinear seismic behavior of cable-
stayed bridges with passive control bearings. Comput Struct 1995;54(3):
461–92.
[3] Ali HEM, Abdel-Ghaffar AM. Seismic energy dissipation for cable-stayed
bridges using passive devices. Earthq Eng Struct Dynam 1994;23:877–93.
[4] AASHTO. Guide specifications for seismic isolation design. 3rd ed. Washington
DC (USA): American Association State Highway and Transportation Officials;
2010.
[5] Camara A. Seismic behaviour of cable-stayed bridges: design, analysis and
seismic devices. Ph.D. thesis. Spain: ETSI Caminos Canales y Puertos,
Universidad Politécnica de Madrid; 2011. <http://oa.upm.es/10036/>.
[6] Asgari B, Osman SA. Application of isolation systems in the seismic control of
cable-stayed bridges: a state-of-the-art review. In: EMESEG’10: 3rd WSEAS
international conference on engineering mechanics, structures, engineering
geology, Corfu, Greece; 2010.
[7] Abdel-Ghaffar AM. Cable-stayed bridges under seismic action. In. Cable-stayed
bridges: recent developments and their future: proceedings of the seminar,
Yokohama, Japan; 1991. p. 101–11.
[8] Dyke SJ, Caicedo JM, Turan G, Bergman LA. Phase I benchmark control problem
for seismic response of cable-stayed bridges. In: Semi-active control. J Struct
Eng-ASCE; 2001 [special issue].
[9] Ren WX, Obata M. Elastic-plastic seismic behaviour of long-span cable-stayed
bridges. J Bridge Eng-ASCE 1999;4:194–203.
[10] Sharabash AM, Andrawes BO. Application of shape memory alloy dampers in
the seismic control of cable-stayed bridges. Eng Struct 2009;31:607–16.
[11] Filiatrault A, Tinawi R, Massicotte B. Damage to cable-stayed bridge during
1988 Saguenay earthquake I: pseudostatic analysis. J Struct Eng-ASCE 1993;
119(5):1432–49.
[12] Filiatrault A, Tinawi R, Massicotte B. Damage to cable-stayed bridge during
1988 Saguenay earthquake II: dynamic analysis. J Struct Eng-ASCE
1993;119(5):1450–63.
[13] Constantinou MC, Reinhorn AM, Mokha A, Watson R. Displacement control
device for base-isolated bridges. Earthq Spectra 1991;7(2):179–200.
[14] Constantinou MC, Kartoum A, Reinhorn AM, Bradford P. Sliding isolation
system for bridges: experimental study. Earthq Spectra 1992;8(3):321–44.
[15] Iemura H, Pradono MH. Passive and semi-active seismic response control of a
cable-stayed bridge. J Struct Contr 2002;9:189–204.
[16] Casciati F, Cimerallo GP, Domaneschi M. Seismic reliability of a cable-stayed
bridge retrofitted with hysteretic devices. Comput Struct 2008;86:1769–81.
[17] Wesolowsky MJ, Wilson JC. Seismic isolation of cable-stayed bridges for near
field ground motions. Earthq Eng Struct Dynam 2003;32:2107–26.
[18] Soneji BB, Jangid RS. Effectiveness of seismic isolation for cable-stayed bridges.
Int J Struct Stab Dynam 2006;6:1–20.
[19] Niihara Y, Takeda T, Ichinomiya T, Suzuki R. Improvements in earthquake
resistance of PC cable-stayed bridges by hysteresis dampers. In: Proc of the 3rd
US–Japan workshop on earthquake protective systems for bridges, Berkeley,
USA; 1994.
[20] Vader T, McDaniel CC. Influence of energy dissipation systems on the seismic
response of cable-supported bridges. In: Proc 11th world conference on
earthquake engineering, Vancouver; 2004.
[21] Soneji BB, Jangid RS. Passive hybrid systems for earthquake protection of
cable-stayed bridge. Eng Struct 2007;29:57–70.
[22] He WL, Agrawal AK. Passive and hybrid control systems for seismic protection
of a benchmark cable-stayed bridge. Struct Control Health 2007;14:1–26.
[23] Valdebenito GE. Passive seismic protection of cable-stayed bridges applying
fluid viscous dampers under strong motion. Ph.D. thesis. Spain: Universitat
Politècnica de Catalunya; 2009. <http://hdl.handle.net/10803/6264>.
[24] Koh CG, Mahatma S, Wang CM. Reduction of structural vibrations by multiple-
mode liquid dampers. Eng Struct 1994;17(2):122–8.
[25] Domaneschi M, Martinelli L. Performance comparison of passive control
schemes for the numerically improved ASCE cable-stayed bridge model.
Earthq Struct 2012;3(2):181–201.
[26] Ha DH, Park JH, Park KS, Park W, Choo JF. Optimization and complex dampers
for the improvement of seismic performance of long-span bridges. KSCE J Civ
Eng 2010;14(1):33–40.
[27] Christopoulos C, Filiatrault A. Principles of passive supplemental damping and
seismic isolation. Pavia (Italy): IUSS Press; 2006.
[28] Carr AJ. RUAUMOKO-inelastic dynamic analysis program. Christchurch (New
Zealand): University of Canterbury; 2002.
[29] AISC. Specification for structural steel buildings. Chicago (IL, USA): American
Institute of Steel Construction Inc, ANSI/AISC 360-05; 2005.
[30] Fleming JF, Egeseli EA. Dynamic behaviour of a cable stayed bridge. Earthq Eng
Struct Dynam 1980;8:1–16.
[31] Nazmy AS, Abdel-Ghaffar AM. Non-linear earthquake response analysis of
long-span cable-stayed bridges: theory. Earthq Eng Struct Dynam 1990;19:
45–62.
[32] Ni YQ, Wang JY, Ko JM. Modal interaction in cable-stayed Ting Kau bridge. In:
Proceedings of the international conference on advances in structural
dynamics, Hong Kong; 2000.
[33] Wilson JC, Gravelle W. Modelling of a cable-stayed bridge for dynamic
analysis. Earthq Eng Struct Dynam 1991;20:707–21.
[34] Freire AMS, Negrao JHO, Lopes AV. Geometrical nonlinearities on the static
analysis of highly flexible steel cable-stayed bridges. Comput Struct 2006;84:
2128–40.
[35] Sharpe, RD. The seismic response of inelastic structures. Ph.D. thesis. New
Zealand: University of Canterbury; 1974.
[36] Ren WX, Peng XL. Baseline finite element modelling of a large span cable-
stayed bridge through field ambient vibration tests. Comput Struct
2005;83:536–50.
[37] SAC joint venture. Develop suites of time histories. Project task 5.4.1. Draft
Report; 1997.
[38] http://nisee.berkeley.edu/data/strong_motion/sacsteel/motions/la10in50yr.
html, January 2015.
[39] Kramer SL. Geotechnical earthquake engineering. Prentice-Hall; 1996.
[40] Park YJ, Ang AH. Mechanistic seismic damage model for reinforced concrete. J
Struct Eng-ASCE 1985;111(4):722–39.
[41] Park YJ, Ang AH, Wen YK. Damage-limiting aseismic design of buildings.
Earthq Spectra 1987;3(1):1–26.
[42] Wada A, Saeki E, Takeuchi T, Watanabe A. Development of unbonded brace.
Column technical publication no. 115:91-96, Nippon Steel, Japan; 1989.
[43] Filiatrault A, Cherry S. Seismic design of friction damped braced steel plane
frames by energy methods. Earthquake engineering research laboratory report
UBC-EERL-88-01. Vancouver (Canada): University of British Columbia; 1988.
[44] Symans MD, Constantinou MC. Passive fluid viscous damping systems for
seismic energy dissipation. ISET J Earthq Technol 1998;35(4):185–206.
[45] http://www.taylordevices.com/, August 2014.
[46] Constantinou MC, Soong TT, Dargush GF. Passive energy dissipation systems
for structural design and retrofit. MCEER monograph no. 1. Multidisciplinary
Center for Earthquake Engineering Research, University at Buffalo; 1998.