Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Structures 50 (2023) 1857–1867

Contents lists available at ScienceDirect

Structures
journal homepage: www.elsevier.com/locate/structures

Risk-based optimization of seismic mitigation devices constrained by


user-defined components importance
Jian Zhong a, Yuntao Zhu a, Yutao Pang b, *, Xinzhe Yuan c, You Dong d, Xinzhi Dang e, Kewen Xu f
a
Department of Civil Engineering, Hefei University of Technology, Hefei 230009, PR China
b
College of Engineering, China University of Geosciences (Wuhan), Wuhan 430073, PR China
c
Civil, Architectural and Environmental Engineering Department, Missouri University of Science and Technology, Rolla, USA
d
Department of Civil and Envriomental Engineering, The Hong Kong Polytechnic University, Hong Kong, PR China
e
Department of Bridge Engineering, Tongji University, Shanghai 200092, PR China
f
Anhui Comprehensive Transportation Research Institute Co., Ltd, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: The system optimization of complex structures is to achieve the overall best behaviour, however, specifications
Constrained optimization and engineers often pay more attention to the seismic performance of critical components. To comprehensively
User-defined threshold consider the requirement from theoretical research and practical engineering, a constrained optimization method
Seismic risk
of seismic mitigation devices for complex structures is proposed in this study, considering the user-defined
Cable-stayed bridges
importance of critical components. The seismic damage probability of individual components is defined as
Seismic mitigation devices
constrained functions (CFs). The overall probabilistic seismic performance of the bridge system is formulated as
the objective function (OF) to carry out optimization analysis within the admissible region obtained by CFs. The
surrogate functions between the variables and OF/CFs are formulated by the proposed Hazard-related Response
Surface Method (H-RSM). A three-dimensional long cable-stayed bridge with a mitigation device of the fluid
viscous damper (FVD) is constructed as the case-study model via FEM. Three constrained conditions are imposed
by restricting the CFs less than a user-defined threshold to satisfy different safety requirements of the pylon,
bearing and FVD. Different protection levels of the pylon are investigated to compare the constrained and un­
constrained optimization methods. The influence of the maximum displacement of FVD on the optimal variables
is also investigated.

1. Introduction technique to cope with the difficult problem. In this design method,
many objectives are constructed to evaluate the design and retrofit
Seismic isolation systems [1–3], energy dissipation and damper de­ effectiveness, such as the repair cost ratio of the column and bearing
vices [4–6], Self-centering systems [7–9] have been developed to reduce [18], sustainability and cost of retrofit actions [19], total recovery time
excessive seismic responses. Past studies of bridge seismic performance and skew of the recovery trajectory [20], cost of base isolation and the
during earthquake events show that shear force or bending moment displacement limits [21], safety and sustainability [22], reliability of the
would increase extremely if the girder-pier connection is firmly con­ structure and material costs [14,23], cost-benefit and sustainability
strained, while the weak connection would cause excessive seismic [24,25], initial cost and maximum inter-story drift [26], minimum
displacement [10]. Accordingly, the seismic mitigation strategy should system seismic risk [27]. Martins et al. [28] investigated the optimiza­
be designed to balance the seismic deformation (isolation devices) and tion of concrete cable stayed bridge with considering the multi-
seismic force (columns), which are usually hard to achieve. In past years, objectives of minimum cost, deflections, natural frequencies, service­
various kinds of seismic mitigation devices are proposed such as: fluid ability and ultimate limit states. In the research from Franchini et al.
viscous dampers, variable dampers and clutching inerter dampers [29], key design variables of tower and cable cross sections are opti­
[11,12]. mized, in which the seismic performance, repair cost and construction
Multi-objective optimized design [13–17] is regarded as an effective cost is comprehensively considered.

* Corresponding author.
E-mail address: pangyutao@cug.edu.cn (Y. Pang).

https://doi.org/10.1016/j.istruc.2023.02.124
Received 12 December 2022; Received in revised form 27 February 2023; Accepted 27 February 2023
Available online 10 March 2023
2352-0124/© 2023 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
J. Zhong et al. Structures 50 (2023) 1857–1867

An alternative way to consider the multiple objectives is to construct ⎛ ⎞


a composite damage index as a comprehensive objective to assess the bi [ ( )]
⎜ln(ai (im) ) − ln(mC,i ) ⎟ ln im/mR,i
efficiency of seismic mitigation devices [30–32]. For instance, Zhang P[Di > Ci |IM = im] = Φ⎝ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ⎠=Φ
βR,i
(i
β2D,i + β2C,i
and Huo [30] developed a composite damage index by defining the
weight ratio of bridge piers and isolation bearings to be 0.75 and 0.25, = 1, 2, ..., nc )
respectively. Dezfuli and Alam [31] evaluated the carbon FRP-based (1)
elastomeric isolator by assigning different weights to the operational
specifications. Their result indicated that the weights significantly where mR,i and βR,i are the median value and std. deviation of normal
affected the optimal design parameters of FRP. cumulative distribution function (Φ[•]), respectively. The coefficients
Some major challenges remain to be addressed before achieving the (ai and bi) and the corresponding std. deviation (βD,i) are parameters of
optimal performance for long-span bridges: (1) For long-span bridges, the logarithmic linear function used to estimate the ith component
the importance of the multiple components is different. For example, the nonlinear seismic demand. mc,i and βc,i are the median value and std.
pylon is the most important for cable-stayed bridges [33]. Therefore, its deviation of the logarithmic normal distribution for the ith component
protection level should be higher than other components, such as capacity model. nc is the component amount of interests. The detailed
bearing, pier, etc. However, this important difference is not fully process of the generation for long bridge fragility analysis is thoroughly
addressed in previous research. (2) The seismic mitigation devices are presented by Zhong et al. [44]. Seismic risk of ith component (PA,i) is
optimally designed under uniform earthquake excitation, which is theoretically formulated as
reasonable for small structures (frames or short-span bridges) but is ∫∞ [ ( / ) ]⃒ ⃒
infeasible for long-span bridges owing to the variation of earthquake ln im mR,i ⃒⃒dλ(im)⃒⃒
PA,i = Φ ⃒ d(im)
arriving time, coherency and local sites [34,35]. The spatial variability im=0 βR,i d(im) ⃒ (2)
of ground motion may change the relative vulnerability of components (i = 1, 2, ..., nc )
[36,37], which would in turn impact the design of the mitigation devices
[38,39]. Besides, the input earthquake is usually assumed to be deter­ when the Hazard function (λ(im)) is estimated by a logarithmic linear
ministic, which does not account for the uncertainty of ground motions model (with two coefficients, k and k0): λ(im) = k0im− k, Eq. (2) is
[40]. degenerated as [41]:
To cope with the aforementioned challenges, a risk-based inequality- ( 2 )
constrained optimization method is proposed in this study to more k2 βR,i
PA,i = k0 m−R,ik exp (i = 1, 2, ..., nc ) (3)
efficiently optimize the design variables of the aseismic system. The 2
proposed final objective function is restricted by imposed constrained
Assuming the distribution of the occurrences of seismic events to be a
conditions. The seismic behavior of critical components (pylon, bearing,
Poisson process [45], the formulation of the lifetime seismic risk (PF,i)
pier, etc.), usually formulated as objective functions, is constructed as
during a specific period (T0) can be expressed as
constrained functions (CFs). The CFs are defined to be smaller than the
( 2 )
threshold of the user-defined target probabilities. The admissible k2 βR,i
parameter region can be obtained by solving the nonlinear inequality
− TPA,i − k
PF,i = 1 − e ≈ TPA,i = T0 k0 mR,i exp
2 (4)
CFs. The system seismic risk is constructed as the final objective function
(OF) to perform the optimization analysis within the obtained parameter (i = 1, 2, ..., nc )
domain, which could reflect the global performance of the whole bridge The system seismic risk (PF,Sys) of a long bridge is formulated by
[41,42]. The seismic risk functions of bridge components and the overall considering the joint probability density function of components.
system are developed by Zhong, Jeon, Ren 36. These functions are able Consequently, PF,Sys of a long-span bridge can be formulated as Eq. (5),
to account for the nonlinearity of structures and uncertainties of ground where the mR,Sys and βR,Sys are mean value and std. deviation, respec­
excitations. The hazard-related Response Surface Method (H-RSM) is tively. The details of the system seismic risk can be found in [41].
used to illustrate the mathematical model between the design variables ( 2 )
(DVs) and our proposed OF/CFs. k2 βR,Sys
A nonlinear cable-stayed bridge numerical model with a fluid viscous
k
PF,Sys = T0 k0 m−R,Sys exp (5)
2
damper (FVD) between the pylon and girder is built. Three components
(the pylon, bearing and FVD) and the system-level seismic risk functions
2.2. Hazard related RSM surrogate model
are formalized and characterized as CFs and OF, respectively. Optimi­
zation is performed on these CFs and OFs. The superiority of our con­
The RSM surrogate model is widely used to approximately construct
strained optimization method is demonstrated in comparison to the
a suitable analytical model and it is quite useful for solving optimization
unconstrained method. Moreover, the risk-informed inequality con­
problems. Concretely, the independent variables represent the param­
strained optimization method can flexibly characterize the importance
eters of the mitigation devices (X={x1,x2,…,xnv}), the normalized risk
of the components when assigned with different protection levels
(Ri(X)) of ith component is defined as the ratio of PF,i (X) compared to PF,i
(0)
(quantified by the acceptable seismic failure probability), which is quite
useful for engineers, experts and policymakers. Finally, the influence of expressed as Eq. (6-a), where PF,i (X) is the seismic risk of the bridge with
the allowable displacement of FVD on the optimal variables and mini­ aseismic devices represented by the parameters X, and PF,i is the seismic
(0)

mum risk is also investigated. risk of the bridge without the aseismic devices as a reference. Similarly,
the normalized risk (Ri(X)) of bridge system is defined as the ratio of
2. Methodology for risk-informed constrained optimization
PF,Sys (X) compared to PF,Sys expressed as Eq. (6-b), where PF,Sys (X) is the
(0)

2.1. Analytical solutions of seismic risk system seismic risk of the bridge with aseismic devices of parameters X,
and PF,i is the system seismic reliability of the reference bridge without
(0)

The likelihood of seismic demand (Di) exceeding capacity (Ci) the aseismic devices.
conditioned on a specific intensity measure (IM) is defined as the Normalized Reliability functions of ith component (Ri(X)) and the
fragility function [43]. Presupposing Di and Ci follow normal distribu­ bridge system (RSys(X)) are expressed as Eqs. (7-a) and (7-b), respec­
tions in logarithmic space, the fragility corresponding to ith component is tively, where the dispersion of the protected and reference bridge is
given as assumed to be identical.

1858
J. Zhong et al. Structures 50 (2023) 1857–1867

PF,i (X) ( / )− k ( ( )/ ) model for the ith component and bride system.
Ri (X) = = mR,i (X) m(0)
R,i exp k2 β2R,i − β2R,i0 2
P(0)
F,i (6a)
2.3. Construction of constrained optimization inequality functions
(i = 1, 2, ..., nc )

( / )− k ( ( )/ ) For the optimization of the seismic behavior of a long-span bridge,


PF,Sys (X)
RSys (X) = = mR,Sys (X) m(0)
R,Sys exp k2 β2R,Sys − β2R,Sys0 2 the primary requirement is to satisfy all the safety requirements of each
P(0)
F,Sys critical component, expressed as inequality functions. Deterministic-
(i = 1, 2, ..., nc ) based optimum design, where the objective function or behavioral
(6b) constraints are usually defined by force or displacements, in a deter­
ministic way, cannot account for the uncertainties associated with
( )− k structures or ground motions. In this study, probabilistic nonlinear
Ri (X) = mR,i (X)/m(0) (i = 1, 2, ..., nc ) (7a)
constraints are constructed and expressed as seismic-risk-related func­
R,i

( / )− k
tions (CFi(X)) of each component.
(0)
RSys (X) = mR,Sys (X) mR,Sys The admissible parameter region is defined by restricting CFi(X)
(7b)
̃i ) for ith component, which can be solved
smaller than the limit value (P
(i = 1, 2, ..., nc )
by Eq. (9-a), marked as Xi*. Fig. 1 (a) illustrates the variable range that
where mR,i
(0)
and β2R,i0 are the median value and std. deviation of the i th can fulfill the requirement of ith constraint. If the seismic risk is required
to be less than 0.04, the design variable must fall inside the contour
component’s fragility for the reference bridge, respectively. mSys,i and
(0)
marked as 0.04, highlighted in Fig. 1 (a). Fig. 1 (b) illustrates a three-
β2R,Sys0 are the median value and std. deviation of the fragility function objective case, where the admissible variable range (X*) is obtained
for the reference bridge system. by overlapping Xi*(i = 1,2,3) and solved by Eq. (9-b). Therefore, it is
The response surface method (RSM) is usually utilized to establish quite convenient to impose constrained conditions separately. The en­
the mapping relationship between the normalized risk and the input gineers, administrators, experts, or policymakers can identify different
variables (X) of the aseismic system. The ith constraint is defined as limit values (P̃i ) for different components according to their importance.
CFi(X), and the objective function is defined as OF(X). Hazard-related For example, since the pylon is the most vulnerable component of cable-
response surface method (H-RSM) surrogate model is proposed to ̃i can be defined as very low. On the other hand, since
stayed bridges, its P
construct CFi(X) and OF(X) associated with the design variables (X={x1, the bearings and other mitigation devices can be easily replaced, their
x2,…,xnv}), formulated by Eqs. (8-a) and (8-b). corresponding limit values can be defined relatively higher.
( )− k
(i)
∑nv ∑nv
(i) 2
nv − 1 ∑
∑ nv
X*i = solve{CFi (X) < P
̃i } (9a)
CFi (X) = a0 + a(i) x + a x + a(i)
x x
p=1
p p
p=1
pp p
p=1 q=p+1
pq p q
(8a)
X* = X*1 ∩ ...X *i ... ∩ X *nc (9b)
(i = 1, 2, ..., nc )
The objective function is defined as the overall system risk (PF,Sys) of
( )−

nv ∑
nv nv − 1 ∑
∑ nv k
a long-span bridge (OF(X)). The next step is to determine the best
OF(X) = a(Sys)
0 + a(Sys)
p xp + a(Sys) 2
pp xp + a(Sys)
pq xp xq (8b) behavior for a long bridge and find out the corresponding design vari­
ables of the aseismic devices within the admissible range (X*) discussed
p=1 p=1 p=1 q=p+1

previously. When the objective function and definition domain are


where k is one of the seismic hazard parameters, xp is the pth independent
specified, it is then quite convenient to determine the minimum value
variable, nv is the amount of the variables, ap(i), app(i) and apq(i) are the
and corresponding parameters, formulated as Eq. (10). Generally
regression coefficients for the ith component, ap(Sys), app(Sys), and apq(Sys)
speaking, the optimal parameters (Xop) are capable to minimize the
are the regression coefficients for the bridge system.
objective function and fulfill the requirements of certain constrained
To provide the metadata of the surrogated model, a set of ns exper­
conditions. During the first step, the allowable variable ranges are ob­
iments are performed associated with parameters (Xj={xj1, xj2,…, xjnv, j
tained by fulfilling the seismic safety requirements of critical compo­
= 1,2,…,ns}). These ns sample points are designed using Latin Hyper­
nents. During the second step, the risk of the long bridge as a global
cube Sampling (LHS) technique to cover the very common range of the
system is formulated as OF(X) to perform optimization analysis within
parameters. Then ns pairs of experimental points {X(j), Ri(X(j))} and {X(j),
the admissible range (X*).
RSys(X(j))} are severed as the metadata to construct the H-RSM surrogate

X 1*
x2

X2*
X*
x2

X3*

x1 x1

Fig. 1. (a) The admissible design range to fulfill the requirement of the ith constraint, and (b) the overlapping of the design range associated with all the constraints.

1859
J. Zhong et al. Structures 50 (2023) 1857–1867

find Xop = [x1 , x2 , ...xnv ] where sgn(⋅) indicates the signum function, Cd and α are two variables
minimize OF(X)
̃1 (X = [Cd , α]) of FVD.
subject to CF1 (X) < P
̃2 (10)
CF2 (X) < P
... 4. Risk evaluation of the investigated bridge
CFnc (X) < P ̃ nc
Dynamic demands associated with the long-span bridge are obtained
3. Analytical model and simulation by performing numerous (ns = 100) nonlinear time history analyses
under a set of earthquake excitations from Shafieezadeh et al. [57] to
In past decades, cable-stayed bridges have been widely favored due consider the randomness of ground motions. The detailed information of
to its advantages of beautiful appearance, short operation time and small the ground motions is listed in Table A1, which inclueds 80 measured
size of components, etc. [46,47]. Therefore, a single tower cable-stayed ground motions and 20 synthetic ground motions. Moreover, the exci­
bridge with the configuration of 150 + 150 m is built as an illustration tations are simulated differently for each support point (pylons and
case, as shown in Fig. 2. Detailed information of the tower, pier, girder piers) to consider the spatial variability of ground motions (wave-pas­
and other critical components of the long-span bridge is described in sage effect, incoherence effect, site-response effect). Zhong et al. [41]
[48]. evaluated long-structure seismic risk under differential support motions,
OpenSEES platform [49] is utilized to perform seismic analysis of the which is used to construct the seismic risk in this study. Peak ground
bridge by building a nonlinear three-dimensional FEM (Fig. 3(a)). The velocity (PGV) is chosen to be the specific IM, due to its capability of
tower section is constructed using a fiber section to account for stiffness/ well estimating dynamic demand for long-span structures [41]. The
strength degradation, force-moment interaction, material plasticity hazard curve that a certain PGV will be exceeded for a certain time range
(Fig. 3(b)) and geometry nonlinearity, etc. Concrete-01 and Steel-01 are is expressed as λ(PGV) = k0PGV− k [58], where k0 and k are the two
adopted to simulate the concrete and longitudinal bar respectively, and parameters, which are determined to be 3.6 × 10− 4 and 1.926, respec­
the grades of the concrete and longitudinal bar are set as C50 and tively, following the work of Zhong et al [41].
HRB400. The girder is defined to be elastic according to field disaster Critical components including pylon, bearing and FVD are vulner­
investigation. The friction bearings, located on the pier to connect the able during earthquake, and their demands are recorded under time
girders, are fixed laterally, and modeled as bilinear elements (Fig. 3(c)) history analysis. The seismic demand parameters (EDPs) for pylon,
longitudinally. The friction coefficient, yield displacement and effective bearing and FVD are defined as the curvature ductility of pylon bottom
radius are set as 0.04, 2 mm and 5 m, respectively [50–53]. Ernst method section (μφ), the longitudinal displacement of the bearing (δb), and the
[54] is employed to reflect the large displacement and sage effect of the longitudinal deformation of FVD (δFVD) respectively, according to
cables (Fig. 3(d)). The stiffness of the pile foundation is simulated by a quantities of existed research [40,46,59,60]. The mean values and std.
foundation model consisting of six linear lumped springs [55]. deviations of the limit states of pylon curvature ductility, deformation of
To facilitate the seismic control of cable-stayed bridges, various FVD and bearings are obtained from the work of Ramanathan [61].
aseismic devices such as hysteretic elastomeric bearings, lead-rubber Table 1 summarizes the risk parameters of critical components and
bearings, friction pendulum system are proposed [56]. Among them, system of the long-span bridge.
fluid viscous dampers (FVDs) are especially wide-adopted due to its
effectiveness to reduce bridge dynamic response. Therefore, FVD is 5. Sensitivity analyses and optimization of FVD
chosen as the aseismic device and installed between the deck and the
pylon longitudinally in this study. The seismic force (fd) of FVD (Fig. 3 5.1. Formulation of the constrains and objective function
(e)) is determined by displacement (d) and velocity (ḋ), and is expressed
as Three constrains conditions are imposed to fulfill the requirement of
critical components, i.e., (1) to prevent the damage of the pylon (μφ), (2)
fd = Cd |ḋ|α sgn(ḋ) (11) to reduce the displacement of bearing (δb) to avoid deck unseating, and

Fig. 2. General layout of the single tower cable-stayed bridge (unit: cm).

1860
J. Zhong et al. Structures 50 (2023) 1857–1867

Kfc
fc

0.2Kfc
0.2fc

uu cu c

Fy
K
K

Fig. 3. (a) FEM of the single tower bridge (unit: m) (b) confined and unconfined concrete, (c) friction bearing, (d) stayed cable, (e) FVD.

(3) to limit the deformation of FVD (δFVD) to guarantee its effectiveness


find Xop = [Cd , α]
during the earthquake. For Probabilistic seismic risk assessment ap­ ( )− 1.926
proaches, structural seismic safety is usually defined as a certain prob­ minimize OF(X) = 1 + 2.56Cd − 1.91Cd2 − 0.171Cd α

ability of damage during the lifetime. Therefore, the constrained ⎪

conditions are defined by restricting the risk-informed formulations ⎪



(
1 + 2.86Cd − 2.42Cd2 − 0.053Cd α
)− 1.926
< 0.42

⎪(
(CFi(X)) less than a specific limit value (P
̃i ), which is defined as Eq. (12): ⎪

⎪ 2
)− 1.926
(15)

⎨ 1 + 1.78Cd − 1.12Cd − 0.315Cd α < 0.43
̃i = Oi /P(0)
P (12) subject to ( 2
)− 1.926
F,i ⎪
⎪ 1 + 2.56Cd − 1.16Cd − 1.32Cd α < 0.43


⎪ 0 < Cd ⩽104
where Oi is the predefined acceptable failure probability (Oi) of ith




⎪ 0.2⩽α⩽0.5
component, P(0)F,i is the risk of the reference bridge. In this case, the ⎪


acceptable failure probabilities (Oi) of the pylon (μφ), bearing (δb) and
FVD (δFVD) are defined as 0.05, 0.1 and 0.1, which is documented in The process of the optimization imposed by nonlinear inequality
Table 2. It is noticed that the seismic risk (P(0)F,i) of the reference bridge conditions is discussed and compared with the unconstrained optimi­
shown in Table 1 is larger than Oi, indicating the necessity of installing zation in this study. By solving the nonlinear inequality conditions in the
aseismic devices. design space, the admissible variable range Xi* for each component can
The design variables (X(j) = [Cd(j), α(j)], j = 1,2,…,ns}) are generated be calculated, shown in the contour maps in Fig. 5. These ranges are then
using Latin hypercube sampling in the design space (α is designed in the overlapped (Fig. 6) to obtain X* for performing optimization analysis.
range of [0.2, 0.5], Cd (×104) is in the range of (0, 1.0]), where ns is the
total number of the samples. Then ns pairs of supporting points {X(j),
Ri(X(j))} and {X(j), RSys(X(j))} are severed as the meta data to construct 5.2. Comparison of constrained and unconstrained optimization
the hazard-related RSM surrogate model, formulated as Eq. (13).
The admissible domain (Xi*) of the variables (X = [Cd, α]) of FVD can
( )− 1.926
CF(Cd , α) = a0 + a1 Cd + a2 α + a11 Cd2 + a22 α2 + a12 Cd α (13) satisfy all the seismic performance requirements of each important
component (the pylon, bearing and FVD, etc.). Therefore, the next step
Note that the CF(Cd, α) is 1 when Cd is zero, i.e. 1 = a0 + a2α + a22α2, of the proposed optimization methodology aims to optimize the objec­
thus, the equation yields: tive function (OF(X)) to further improve the seismic structural perfor­
(
CF(Cd , α) = 1 + a1 Cd + a11 Cd2 + a12 Cd α
)− 1.926
(14) mance by minimizing the system seismic risk of the subjected bridge
with the obtained admissible design space.
The parameters (a1, a11, a12) are identified and listed in Table 3. The To investigate the importance of the FVD variables, i.e., Cd and α, on
correlation coefficient (R2) of the nonlinear regression (shown in the cable-stayed seismic risk as a whole bridge, the partial deviation of
Table 3) is 0.92, 0.96, 0.98 and 0.97 for the risk function of μφ, δb, δFVD OF(X) are formulated with respect to Cd and α in Eq. (16-a) and (16-b),
and the bridge system, demonstrating its accuracy and efficiency of the and illustrated in Fig. 7. The larger the absolute gradient is, the faster the
proposed H-RSM surrogate model. The model is drawn in Fig. 4, where OF(X) changes. It is clearly shown that the partial derivatives of OF(X)
filled circles are the metadata. with respect to Cd are from − 5 to 1, while the partial derivatives of OF
When the surrogate model is constructed, it is then transformed to a (X) respective to α change in the range of 0 to 0.1. Therefore, the seismic
typical mathematical problem of finding the minimum of the con­ risk is much more sensitive to damping coefficient Cd than the exponent
strained nonlinear multivariable function, formulated as Equation (15). α.
The bound constraints of Cd (×104) (0 < Cd ≤ 1) and α (0.2 ≤ α ≤ 0.5)
∂OF ( )−
are also represented in the equation as the design variables for practical (16a)
2.926
= − 1.926 1 + a1 Cd + a2 Cd2 + a3 Cd α (a1 + 2a2 Cd + a3 α)
∂Cd
engineering.

1861
J. Zhong et al. Structures 50 (2023) 1857–1867

Table A1 Table A1 (continued )


List of selected ground motions and relevant supporting information. Number PGA PGV(m/ Sa02 Sa10 SaT1(g) M R
Number PGA PGV(m/ Sa02 Sa10 SaT1(g) M R (g) s) (g) (g) (km)
(g) s) (g) (g) (km)
74 0.063 0.037 0.156 0.046 0.002 6.0 39.6
1 0.159 0.212 0.423 0.161 0.024 6.9 28.2 75 0.117 0.092 0.189 0.127 0.003 6.0 39.6
2 0.443 0.317 0.962 0.356 0.009 6.9 14.5 76 0.299 0.225 0.556 0.208 0.003 6.0 56.8
3 0.367 0.396 1.679 0.316 0.016 6.9 14.4 77 0.063 0.043 0.187 0.053 0.001 6.0 31.3
4 0.212 0.378 0.587 0.311 0.011 6.9 16.1 78 0.071 0.055 0.113 0.082 0.001 6.0 34.2
5 0.323 0.160 0.669 0.097 0.005 6.9 24.2 79 0.041 0.035 0.097 0.067 0.001 6.0 35.7
6 0.215 0.400 0.311 0.584 0.036 6.9 28.2 80 0.118 0.063 0.390 0.062 0.001 6.0 39.8
7 0.279 0.390 0.492 0.434 0.028 6.9 25.8 81 0.159 0.611 1.264 0.758 0.027
8 0.209 0.359 0.593 0.273 0.042 6.9 28.8 82 0.443 0.843 0.972 0.563 0.068
9 0.42 0.439 0.836 0.379 0.016 6.7 15.8 83 0.367 0.659 0.566 0.401 0.060
10 0.242 0.215 0.699 0.236 0.005 6.7 23.9 84 0.212 0.649 0.613 0.653 0.109
11 0.24 0.172 0.433 0.170 0.005 6.7 29.5 85 0.323 0.751 0.947 0.892 0.139
12 0.206 0.091 0.835 0.062 0.004 6.7 25.4 86 0.215 0.682 3.306 0.580 0.027
13 0.358 0.226 0.747 0.323 0.005 6.7 25.5 87 0.279 0.933 1.482 1.023 0.040
14 0.159 0.120 0.458 0.108 0.002 6.7 22.3 88 0.209 0.996 1.104 1.136 0.030
15 0.482 0.448 1.078 0.485 0.013 6.7 13.0 89 0.42 0.976 0.882 0.666 0.035
16 0.443 0.292 0.853 0.471 0.014 6.7 12.3 90 0.242 0.841 2.836 0.852 0.023
17 0.174 0.167 0.490 0.193 0.018 6.6 21.2 91 0.24 1.382 1.345 2.453 0.045
18 0.258 0.449 0.642 0.274 0.021 6.7 13.9 92 0.206 0.970 0.612 0.965 0.056
19 0.207 0.321 0.399 0.299 0.032 6.7 24.4 93 0.358 1.621 1.797 1.848 0.049
20 0.211 0.269 0.481 0.341 0.04 6.7 13.3 94 0.159 1.589 1.437 1.083 0.057
21 0.057 0.185 0.166 0.154 0.012 6.8 46.0 95 0.482 0.938 2.903 0.795 0.129
22 0.139 0.125 0.359 0.211 0.009 6.9 57.4 96 0.443 1.539 2.283 1.640 0.091
23 0.141 0.129 0.410 0.132 0.011 6.9 43.4 97 0.174 1.376 1.597 1.179 0.093
24 0.103 0.142 0.202 0.138 0.012 6.9 31.6 98 0.258 2.206 1.999 1.184 0.146
25 0.112 0.129 0.266 0.093 0.018 6.9 32.6 99 0.207 2.097 1.060 1.417 0.129
26 0.278 0.338 0.433 0.523 0.015 6.9 36.3 100 0.211 1.217 0.853 0.845 0.061
27 0.079 0.063 0.262 0.083 0.001 6.7 56.1
28 0.136 0.071 0.288 0.088 0.004 6.7 49.6
29 0.322 0.213 0.764 0.233 0.005 6.7 30.9
30 0.123 0.105 0.400 0.123 0.003 6.7 59.3 Table 1
31 0.23 0.124 0.574 0.146 0.002 6.7 47.6 Probability damage parameters of critical components of the single tower
32 0.068 0.075 0.189 0.129 0.005 6.7 46.6 bridge.
33 0.077 0.094 0.170 0.198 0.006 6.7 36.3
Components EDPs Reference bridge
34 0.152 0.085 0.204 0.099 0.004 6.7 42.4
mR,i PF,i
(0) (0) (0)
35 0.063 0.074 0.168 0.079 0.03 6.7 37.7 βR,i
36 0.067 0.077 0.180 0.155 0.002 6.7 43.6 Pylon μφ 0.64 0.45 0.12
37 0.186 0.132 0.242 0.127 0.003 6.7 32.7
Bearing δb 0.45 0.40 0.23
38 0.067 0.062 0.161 0.096 0.003 6.7 54.1
System – 0.42 0.39 0.32
39 0.194 0.131 0.326 0.119 0.004 6.7 60.0
40 0.153 0.096 0.485 0.124 0.003 6.7 39.3
41 0.078 0.148 0.272 0.095 0.014 6.5 23.8
42 0.254 0.277 0.450 0.348 0.011 6.5 28.7 Table 2
43 0.134 0.125 0.247 0.076 0.013 6.5 15.5
Objective safety of engineering demand parameters.
44 0.116 0.197 0.332 0.172 0.025 6.5 18.2
45 0.139 0.133 0.403 0.099 0.011 6.5 21.9 Components μφ δb δFVD
46 0.110 0.228 0.157 0.094 0.018 6.5 15.1
Oi 0.05 0.1 0.1
47 0.04 0.036 0.143 0.085 0.002 5.8 21.7
P
̃i = Oi /P(0) 0.42 0.43 0.43
48 0.076 0.106 0.189 0.151 0.004 5.8 17.6 F,i

49 0.032 0.052 0.089 0.092 0.003 6.2 29.4


50 0.212 0.079 0.387 0.08 0.002 6.2 29.4
51 0.2 0.119 0.357 0.165 0.005 6.2 14.6
52 0.113 0.064 0.327 0.042 0.004 6.2 14.0
53 0.083 0.082 0.171 0.113 0.001 5.8 25.0 Table 3
54 0.171 0.083 0.331 0.081 0.001 5.8 22.0 The parameters for the fitted two-dimensional quadratic function.
55 0.176 0.059 0.357 0.044 0.001 5.8 19.4
EDPs a1 a11 a12 R2
56 0.333 0.195 0.521 0.246 0.003 6.0 16.9
57 0.059 0.029 0.097 0.036 0.001 6.0 28.1 μφ 2.86 − 2.42 − 0.0530 0.92
58 0.141 0.195 0.399 0.260 0.003 6.0 18.3 δb 1.78 − 1.12 − 0.315 0.96
59 0.151 0.129 0.590 0.195 0.002 6.0 16.3 δFVD 2.56 − 1.16 − 1.32 0.98
60 0.133 0.099 0.240 0.146 0.002 6.0 24.5 System 2.56 − 1.61 − 0.171 0.97
61 0.044 0.039 0.132 0.049 0.001 6.5 49.0
62 0.131 0.099 0.240 0.101 0.002 6.4 47.3
63 0.1 0.085 0.124 0.076 0.002 6.4 50.7
64 0.128 0.140 0.201 0.104 0.004 6.5 49.3 ∂OF ( )−
(16b)
2.926
= − 1.926 1 + a1 Cd + a2 Cd2 + a3 Cd α a3 Cd
65 0.147 0.110 0.337 0.071 0.004 6.5 32.6 ∂α
66 0.351 0.281 0.513 0.352 0.018 6.5 43.6
67 0.069 0.092 0.250 0.101 0.012 6.5 35.9
As discussed, the exponent α of FVD has very little influence on the
68 0.057 0.041 0.185 0.035 0.003 6.5 31.5 system seismic risk, and is selected to be 0.3 practically and conve­
69 0.167 0.074 0.427 0.064 0.002 6.5 54.1 niently. The partial deviation with respect to Cd is plotted in Fig. 8 (a). It
70 0.073 0.077 0.136 0.093 0.004 5.8 37.3 is shown that the multiple objectives compete. More specifically, the
71 0.142 0.064 0.402 0.077 0.001 6.2 38.1
optimal coefficient Cd (×104) are 0.59, 0.88 and 0.90 for the pylon,
72 0.071 0.081 0.156 0.149 0.005 6.2 32.5
73 0.036 0.043 0.098 0.077 0.003 6.2 30.3 bearing and mitigation devices, respectively. If the Cd keeps on
increasing after the pylon ductility (μφ) achieves its minimum seismic
risk at Cd = 0.59, the risk of δb and δFVD would decrease, but at the cost of

1862
J. Zhong et al. Structures 50 (2023) 1857–1867

Cd

Cd

Cd
Cd Cd Cd

Fig. 4. The H-RSM surrogate model of CFs for (a) μφ, (b) δb, (c) δFVD associated with damping coefficient (Cd) and exponent (α).

Cd Cd Cd

Fig. 5. Admissible domain of damping coefficient (Cd) and exponent (α) for each CFs (a) μφ, (b) δb, (c) δFVD.

influence of constrained conditions on the optimal parameters of FVD,


four acceptable failure probabilities of the pylon (Oi, i = 1,2,3,4) are
identified as 0.05, 0.045, 0.04, 0.035, respectively. Following the pro­
b
cedure and steps of Section 5.1 and 5.2, the corresponding minimum
FVD seismic risk of the whole long-span bridge is obtained and documented
in Fig. 9.
It is shown that the protection level of the pylon (Oi) has a significant
influence on the optimal parameters and the seismic risk. If the peak
point (0.82, 8.7%) of the system risk curve falls in the admissible vari­
able range, the optimal Cd will always be 0.87. For instance, when the
acceptable failure probability of the pylon (Oi) is identified as 0.05,
0.045, the constrained optimal parameters of the FVD are identical with
the unconstrained solution, shown in Fig. 9. However, when the
acceptable failure probability of the pylon is set to be less than 0.04, the
optimal parameters are different from the unconstrained solution. The
explanation of this solution can be attributed to the narrowing of the
admissible domain, which can be clearly observed in Fig. 9.
More comprehensive illustration of the constrained optimization in
terms of two variables (i.e. Cd and α) are shown in Fig. 10. The white
Cd
hollow circles represent the global optimal Cd conditioned on different
Fig. 6. Admissible domain of damping coefficient Cd and the exponent α values of α. Four admissible domains associated with Oi are plotted. It is
considering CFs. clearly shown that the optimal parameters are different from the un­
constrained solution for some constrained cases (e.g., when Oi of pylon is
the lower performance of the pylon ductility. Therefore, we need to set to be less than 0.04).
consider the bridge as a global system. The risk curves of pylon, bearing, The overall system optimization of the bridge must be premised on
FVD and bridge system corresponding to Cd are illustrated in Fig. 8 (b), the security of individual components. The proposed optimization
where the minimum risk of individual components and system are method allows to define the seismic protection priority for different
labeled as well as the admissible domain. The system risk curve within components, which can satisfy all the predefined component objectives,
the domain is plotted by a thick black line. It is seen that the optimal Cd and achieve the best seismic performance and least seismic risk.
to obtain the best system performance within the domain is 0.82.
The importance of individual components as well as the whole bridge
seismic performance is fully considered in the proposed risk-informed 5.3. Sensitivity analysis and optimization under different constraints.
optimization method. The pylon is usually deemed as the most impor­
tant component of the long-span bridge. To further examine the The sensitivity analyses of the FVD parameters show that the mini­
mum risk of the FVD deformation is larger than that of the pylon

1863
J. Zhong et al. Structures 50 (2023) 1857–1867

OFSys OFSys
Cd

Cd CdCd

Fig. 7. Sensitivity analysis of system seismic risk with respect to (a) damping coefficient Cd and (b) exponent α.
Cd

Cd Cd

Fig. 8. (a) Partial deviation of seismic risk with respective to Cd and (b) risk variation corresponding to Cd.

ductility. The best bridge system performance is achieved by decreasing


the risk of the FVD at the cost of increased risk of pylon ductility.
Pylon Oi Therefore, FVD displacement capacity should be enlarged. The previous
Pylon Oi Pylon Oi Pylon Oi
research did not focus sufficiently on quantifying FVD design deforma­
tion capacity (mC-FVD). The sensitivity analyses of mC-FVD are carried out
to determine the influence of mC-FVD on the minimum seismic risk, as
well as the corresponding optimum parameters of Cd and α. Following
the procedure described above, the optimal Cd and the corresponding
seismic risk of the whole bridge are determined and drawn in Fig. 11.
The white hollow circles represent the global optimal Cd given a specific
value of α, and the red lines display the boundary of the admissible
design space constrained by the pylon, bearing and FVD. It can be found
that the contour map changes with different mC-FVD, and the value of OF
tends to be smaller with a larger capacity of FVD (mC-FVD). The con­
strained solution for optimization remains consistent with the uncon­
Cd Cd Cd Cd
strained (X(u)op) when mC-FVD is greater than 0.3. For smaller mC-FVD, the
Fig. 9. The influence of acceptable failure probability of the pylon (Oi), (a) X(u)op falls outside of the admissible range, and constrained optimal
0.05, (b) 0.045, (c) 0.04 and (d) 0.035, on the optimal parameters of FVD. parameters are then located on the right boundary, indicating that the

1864
J. Zhong et al. Structures 50 (2023) 1857–1867

Cd
Fig. 10. The contour map of system risk constrained by different conditions.

6. Conclusions
mC-FVD = mC-FVD =
Compared to short-span bridges, the seismic assessment and opti­
mization of long bridges should focus on: (1) nonlinearity of material
and structures, (2) uncertainty and spatial variability of ground motions,
(3) multiple components and their different importance, and (4) the
overall behavior of the whole bridge, etc. To deal with the challenges,
this paper proposes a reliability-oriented constrained optimization
method to optimize the design parameters of the aseismic devices sub­
Cd Cd jected to constrained conditions. During the optimization, the seismic
risk of each component is limited according to their importance so that
some important components will not face a higher risk when improving
mC-FVD = mC-FVD =
the seismic performance of the whole bridge. The probabilistic perfor­
mance of each component is constructed as the inequality constrained
functions (CFs) and must be fulfilled to obtain the admissible design
parameter ranges of mitigation devices. The system-level seismic reli­
ability is formulated as the objective function (OF) to perform the
optimization analysis within the obtained range. The surrogate model
between the variables and the seismic risk is formulated by Hazard-
Cd Cd
related Response Surface Method (H-RSM).
A three-dimensional single-tower cable-stayed bridge FEM is con­
Fig. 11. The contour map of bridge system risk for different deformation ca­ structed as an illustrative case with fluid viscous damper (FVD) as the
pacity (mC-FVD) of bearing and FVD (a) 0.2 m, (b) 0.3 m, (c) 0.4 m and (d) 0.5 m. aseismic device. Three risk-informed constrained functions (deforma­
tion of bearing, curvature ductility of the pylon, displacement of the
constrained conditions significantly influence the optimal solution. FVD) and the objective function are formalized. The effectiveness of our
Conclusions can be clearly obtained in Fig. 12, where optimal Cd are proposed constrained optimization method investigated in comparison
plotted for given values of α to be 0.3. The optimal Cd for the bridge with the unconstrained method. The risk-informed constrained optimi­
system is decreased when the mC-FVD changed from 0.2 to 0.6, but the zation method is also demonstrated to characterize the importance of
trend becomes flat gradually, and finally converges to a specific value the component with different protection levels (quantified by the
(Cd = 0.59). It can be explained as the failure of FVD or bearing does not acceptable seismic failure probability), which is quite useful for engi­
contribute to the bridge system with a sufficiently large value of mC-FVD. neering practicality. The influence of the allowable displacement of FVD
Therefore, it can be concluded that the deformation capacity of the FVD on the optimal variables and minimum risk is also investigated.
should be enlarged to be 0.5. Beyond 0.5, however, it does not signifi­ The study enables the identification of different reliability thresholds
cantly change the optimal parameters of the damper or reduce the sys­ for each critical component by separately considering their importance,
tem risk of the cable-stayed bridge anymore. which also benefits the engineering practicality of long-span bridge
design.

1865
J. Zhong et al. Structures 50 (2023) 1857–1867

Risk
Cd

mC-FVD mC-FVD

Fig. 12. The optimal Cd (a) and corresponding minimum bridge system risk (b), dependent on deformation capacity of FVD.

Declaration of Competing Interest [17] Gidaris I, Taflanidis AA, Lopez-Garcia D, Mavroeidis GP. Multi-objective risk-
informed design of floor isolation systems [J]. Earthq Eng Struct Dyn 2016;45(8):
1293–313.
The authors declare the following financial interests/personal re­ [18] Xie Y, Zhang J. Optimal Design of Seismic Protective Devices for Highway Bridges
lationships which may be considered as potential competing interests: Using Performance-Based Methodology and Multiobjective Genetic Optimization
Jian Zhong reports financial support was provided by the National [J]. J Bridg Eng 2016;22(3):04016129.
[19] Dong Y, Frangopol DM, Saydam D. Pre-earthquake multi-objective probabilistic
Natural Science Foundation of China. Jian Zhong reports financial retrofit optimization of bridge networks based on sustainability [J]. J Bridg Eng
support was provided by the National Key Research and Development 2014;19(6):04014018.
Program of China. [20] Zhang W, Wang N, Nicholson C. Resilience-based post-disaster recovery strategies
for road-bridge networks [J]. Struct Infrastruct Eng 2017;13(11):1404–13.
[21] Wang Q, Fang H, Zou X-K. Application of Micro-GA for optimal cost base isolation
Acknowledgement design of bridges subject to transient earthquake loads [J]. Struct Multidiscip
Optim 2010;41(5):765–77.
[22] Tapia C, Padgett JE. Multi-objective optimisation of bridge retrofit and post-event
This work is funded by the National Key Research and Development repair selection to enhance sustainability [J]. Struct Infrastruct Eng 2016;12(1):
Program of China (2019YFE0112300), National Natural Science Foun­ 93–107.
dation of China (52178135). The supports are gratefully acknowledged. [23] Papavasileiou GS, Charmpis DC. Seismic design optimization of multi-storey steel-
concrete composite buildings [J]. Comput Struct 2016;170:49–61.
[24] Sabatino S, Frangopol DM, Dong Y. Sustainability-informed maintenance
References optimization of highway bridges considering multi-attribute utility and risk
attitude [J]. Eng Struct 2015;102:310–21.
[1] Jangid R. Optimum friction pendulum system for near-fault motions [J]. Eng Struct [25] Dong Y, Frangopol DM, Sabatino S. Optimizing bridge network retrofit planning
2005;27(3):349–59. based on cost-benefit evaluation and multi-attribute utility associated with
[2] Fenz DM, Constantinou MC. Spherical sliding isolation bearings with adaptive sustainability [J]. Earthq Spectra 2015;31(4):2255–80.
behavior: Experimental verification [J]. Earthq Eng Struct Dyn 2010;37(2): [26] Lagaros ND, Papadrakakis M. Seismic design of RC structures: A critical assessment
185–205. in the framework of multi-objective optimization [J]. Earthq Eng Struct Dyn 2010;
[3] Morgan TA, Mahin SA. Achieving reliable seismic performance enhancement using 36(12):1623–39.
multi-stage friction pendulum isolators [J]. Earthq Eng Struct Dyn 2010;39(13): [27] Zhong J, Wan HP, Yuan W, et al. Risk-informed sensitivity analysis and
1443–61. optimization of seismic mitigation strategy using Gaussian process surrogate model
[4] Jangid R. Optimum lead–rubber isolation bearings for near-fault motions [J]. Eng [J]. Soil Dyn Earthq Eng 2020;138:106284.
Struct 2007;29(10):2503–13. [28] Martins AMB, Simões LMC, Negrão JHJO. Optimization of concrete cable-stayed
[5] Chen X, Li C. Seismic assessment of earthquake-resilient tall pier bridges using bridges under seismic action[J]. Comput Struct 2019;222:36–47.
rocking foundation retrofitted with various energy dissipation devices [J]. Struct [29] Franchini A, Sebastian W, D’Ayala D. Surrogate-based fragility analysis and
Control Health Monit 2020;12. probabilistic optimisation of cable-stayed bridges subject to seismic loads[J]. Eng
[6] Dezfuli FH, Alam MS. Effect of different steel-reinforced elastomeric isolators on Struct 2022;256:113949.
the seismic fragility of a highway bridge [J]. Struct Control Health Monit 2017;24 [30] Zhang J, Huo YL. Evaluating effectiveness and optimum design of isolation devices
(2):e1866.1861-e1866.1815.. for highway bridges using the fragility function method [J]. Eng Struct 2009;31(8):
[7] Shinozuka M, Chaudhuri SR, Mishra SK. Shape-Memory-Alloy supplemented Lead 1648–60.
Rubber Bearing (SMA-LRB) for seismic isolation [J]. Probab Eng Mech 2015;41: [31] Dezfuli FH, Alam MS. Multi-criteria optimization and seismic performance
34–45. assessment of carbon FRP-based elastomeric isolator [J]. Eng Struct 2013;49:
[8] Billah AM, Alam MS. Seismic fragility assessment of concrete bridge piers 525–40.
reinforced with superelastic shape memory alloy [J]. Earthq Spectra 2014;31(3): [32] Zhong J, Hu Z, Yuan W, Chen L. System-based probabilistic optimization of fluid
101–3. viscous dampers equipped in cable-stayed bridges [J]. Adv Struct Eng 2018;21(12):
[9] DesRoches R, Delemont M. Seismic retrofit of simply supported bridges using shape 1815–25.
memory alloys [J]. Eng Struct 2002;24(3):325–32. [33] Wen JN, Han Q, Xie YZ, Du XL, Zhang J. Performance-based seismic design and
[10] Guan Z, Li J, Xu Y. Performance Test of Energy Dissipation Bearing and Its optimization of damper devices for cable-stayed bridge [J]. Eng Struct 2021;237.
Application in Seismic Control of a Long-Span Bridge [J]. J Bridg Eng 2009;15(6): [34] Katerina Konakli ADK. Stochastic dynamic analysis of bridges subjected to spatially
622–30. varying ground motions [R]. Technical Report 2011-105, Pacific Earthquake
[11] Jangid RS. Seismic performance assessment of clutching inerter damper for Engineering Research Center, University of California, Berkeley: Berkeley, CA,
isolated bridges[J]. Pract Period Struct Des Constr 2022;27(2):04021078. 2011.
[12] Madhekar SN, Jangid RS. Variable dampers for earthquake protection of [35] Bi KM, Hao H, Chouw N. Influence of ground motion spatial variation, site
benchmark highway bridges[J]. Smart Mater Struct 2009;18(11):115011. condition and SSI on the required separation distances of bridge structures to avoid
[13] Taflanidis AA. Optimal probabilistic design of seismic dampers for the protection of seismic pounding [J]. Earthq Eng Struct Dyn 2011;40(9):1027–43.
isolated bridges against near-fault seismic excitations [J]. Eng Struct 2011;33(12): [36] Zhong J, Jeon JS, Yuan W, Desroches R. Impact of Spatial Variability Parameters
3496–508. on Seismic Fragilities of a Cable-Stayed Bridge Subjected to Differential Support
[14] Sajedi S, Huang Q, Gandomi AH, Kiani B. Reliability-based multiobjective design Motions [J]. J Bridg Eng 2017;22(6):04017013.
optimization of reinforced concrete bridges considering corrosion effect [J]. ASCE- [37] Soyluk K, Dumanoglu A. Spatial variability effects of ground motions on cable-
ASME J Risk Uncertainty Eng Syst Part A: Civil Eng 2016;3(3):04016015. stayed bridges [J]. Soil Dyn Earthq Eng 2004;24(3):241–50.
[15] Bocchini P, Frangopol DM. Restoration of bridge networks after an earthquake: [38] Xie W, Sun LM, Lou ML. Shaking table test verification of traveling wave resonance
Multicriteria intervention optimization [J]. Earthq Spectra 2012;28(2):426–55. in seismic response of pile-soil-cable-stayed bridge under non-uniform sine wave
[16] Konak A, Coit DW, Smith AE. Multi-objective optimization using genetic excitation [J]. Soil Dyn Earthq Eng 2020;134.
algorithms: A tutorial [J]. Reliab Eng Syst Saf 2006;91(9):992–1007. [39] Xie W, Sun L. Experimental and numerical verification on effects of inelastic tower
links on transverse seismic response of tower of bridge full model [J]. Eng Struct
2019;182:344–62.

1866
J. Zhong et al. Structures 50 (2023) 1857–1867

[40] Wei B, Hu ZL, He XH, Jiang LZ. Evaluation of optimal ground motion intensity [52] Wei B, Fu Y, Li S, et al. Parameter optimization analysis of plane friction coupling
measures and seismic fragility analysis of a multi-pylon cable-stayed bridge with effect[J]. Mech Based Des Struct Mach 2021:1–24.
super-high piers in Mountainous Areas [J]. Soil Dyn Earthq Eng 2020;129. [53] Zhong J, Zhu Y, Han Q. Impact of vertical ground motion on the statistical analysis
[41] Zhong J, Jeon JS, Ren WX. Risk assessment for a long-span cable-stayed bridge of seismic demand for frictional isolated bridge in near-fault regions[J]. Eng Struct
subjected to multiple support excitations [J]. Eng Struct 2018;176:220–30. 2023;278:115512.
[42] Pang Y, He W, Zhong J. Risk-based design and optimization of shape memory alloy [54] Zhong J, Pang Y, Jeon JS, Desroches R, Yuan W. Seismic fragility assessment of
restrained sliding bearings for highway bridges under near-fault ground motions long-span cable-stayed bridges in China [J]. Adv Struct Eng 2016;19(11):
[J]. Eng Struct 2021;241:112421. 1797–812.
[43] Zhong J, Mao Y, Yuan X. Lifetime seismic risk assessment of bridges with [55] Ingham TJ, Rodriguez S, Donikian R, et al. Seismic analysis of bridges with pile
construction and aging considerations. Structures 2023;47:2259–72. foundations[J]. Comput Struct 1999;72(1–3):49–62.
[44] Zhong J, Jeon JS, Yuan WC, DesRoches R. Impact of Spatial Variability Parameters [56] Yi J, Zhou J, Ye X. Seismic control of cable-stayed bridge using negative stiffness
on Seismic Fragilities of a Cable-Stayed Bridge Subjected to Differential Support device and fluid viscous damper under near-field ground motions. J Earthq Eng
Motions [J]. J Bridg Eng 2017;22(6). 2022;26(5):2642–59.
[45] Nuti C, Vanzi I. To retrofit or not to retrofit? [J]. Eng Struct 2003;25(6):701–11. [57] Shafieezadeh A, Ramanathan K, Padgett JE, DesRoches R. Fractional order
[46] Zhong J, Xu W, Dang X, et al. Effect of near-fault vertical ground motions on failure intensity measures for probabilistic seismic demand modeling applied to highway
mode of long-span sea-crossing cable-stayed bridges[J]. Ocean Eng 2022;266: bridges [J]. Earthq Eng Struct Dyn 2012;41(3):391–409.
113005. [58] Cornell CA. Calculating building seismic performance reliability: a basis for multi-
[47] Pang Y, Wu X, Shen G, et al. Seismic fragility analysis of cable-stayed bridges level design norms [C], Proceedings of the 11th world conference on earthquake
considering different sources of uncertainties[J]. J Bridg Eng 2014;19(4): engineering; Acapulco, Mexico, 1996; 5707-5712.
04013015. [59] Yang T, Wei Y, Zhong J. Potential bias of conventional structural seismic fragility
[48] Zhong J, Jeon JS, Shao YH, Chen L. Optimal Intensity Measures in Probabilistic for bridge structures under pulse-like ground motions: Bias evaluation and strategy
Seismic Demand Models of Cable-Stayed Bridges Subjected to Pulse-Like Ground improvement[J]. Soil Dyn Earthq Eng 2023;166:107787.
Motions [J]. J Bridg Eng 2019;24(2):04018118. [60] Zhong J, Shi L, Yang T, et al. Probabilistic seismic demand model of UBPRC
[49] Mckenna F, Scott MH, Fenves GL. Nonlinear Finite-Element Analysis Software columns conditioned on Pulse-Structure parameters[J]. Eng Struct 2022;270:
Architecture Using Object Composition [J]. J Comput Civ Eng 2010;24(1):95–107. 114829.
[50] Li S, Wei B, Tan H, et al. Equivalence of friction and viscous damping in a spring- [61] Ramanathan KN. Next generation seismic fragility curves for California bridges
friction system with concave friction distribution[J]. J Test Eval 2020;49(1): incorporating the evolution in seismic design philosophy. Atlanta, Georgia Institute
372–95. of Technology; 2012 [D]. dissertation,.
[51] Wei B, Xiao B, Fu Y, et al. Effect of simulation accuracy of shear keys shear state on
seismic response of friction pendulum bearing. Structures 2022;39:1189–203.

1867

You might also like