Fan Et Al. (2022) - Fatigue Reliability-Based Replacement Strategy For Bridge Stay Cables A Case Study in China

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Structures 39 (2022) 1176–1188

Contents lists available at ScienceDirect

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

Fatigue reliability-based replacement strategy for bridge stay cables: A case


study in China
Ziyuan Fan a, Qiaowei Ye a, Xiang Xu a, b, Yuan Ren a, *, Qiao Huang a, Weizhen Li a
a
School of Transportation, Southeast University, Nanjing, China
b
School of Engineering, University of Edinburgh, Edinburgh, UK

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

Keywords: Due to repetitive external loads and aggressive environments, performance deterioration of stay cables is a major
Cable-stayed bridges concern for cable-stayed bridges. To balance failure risk and maintenance cost, this paper presents a case study of
Stay cables using a practical fatigue reliability-based method to form an optimal replacement schedule for stay cables. First,
Vehicle load model
based on long-term recorded vehicle information, the statistical analytic was performed to classify vehicles and
Fatigue reliability
Performance deterioration
establish probabilistic models of vehicle parameters (e.g., gross vehicle weight). Then, with the assistance of
Replacement strategy finite element model, structural responses in stay cables were generated under the simulated random traffic
loads. Meanwhile, the time-variant reliability index was introduced to quantitatively characterize the perfor­
mance deterioration, where the time-dependent corrosion inside the cables was considered. Based on the time-
variant reliability index, the cost analytic was conducted to obtain an optimal replacement schedule for stay
cables by minimizing the expected annual average cost. The developed method was applied to stay cables of the
Dashengguan Yangtze River Bridge, and the most critical cable No. A5 and the longest cable No. J21 were
emphasized. As a result, different stay cables have different optimal replacement timings, where the cable No. A5
has the shortest service life - replaced at the 17th service year with the minimum expected annual average cost of
¥ 44,551. Based on the optimal cable replacement timing for each stay cable, replacement schedule for stay
cables in practise could be determined accordingly.

1. Introduction various scenarios [7,8]. Ruan et al. [9] used weigh-in-motion (WIM)
data from a long-span bridge to generate random synthetic vehicle flow,
For long-span cable-stayed bridges, stay cables are crucial compo­ where the safety assessment of the bridge component was made for
nents to support main girders and transfer vehicle loads to towers [1,2]. different traffic conditions. Wang et al. [10] developed a traffic load
However, cables inevitably undergo performance deterioration over model to forecast the future vehicle loading based on recorded WIM
time due to cyclic vehicle loads and aggressive environments [3]. Once a data, where a framework of dynamic stress analysis for fatigue damage
sudden cable failure happens, it will damage the structure, cause enor­ prognosis under the future vehicle loading was presented. Xu et al. [11]
mous economic losses, or even lead to casualties. In this regard, to simulated the traffic load flow with major parameters of each vehicle to
ensure the integrity of cable-stayed bridges, the reliability assessment of estimate site-specific extreme traffic load effects, then compared with
stay cables is one of critical means [4]. the design values. Existing researches have shown that the vehicle model
Meanwhile, cables are vulnerable bridge components - suffering based method achieved great computational efficiency and accuracy,
from fatigues under repetitive vehicle loads [5]. To evaluate fatigue which is significant for the traffic load simulation and components
performance of stay cables, stress analytic of stay cables under traffic response analysis [12,13].
loads is necessary. Although the in-situ monitoring method is a Considering its comprehensive mechanical properties (e.g., high
commonly used means to obtain stress time histories [6], they suffer strength, good plasticity, and toughness), high-strength steel wire is the
from some limitations. For instance, sensors cannot cover all the loca­ most widely used material for bridge cables [14]. Although various
tions of interest due to economic issues. To overcome this disadvantage, methods were employed to prevent the environmental corrosion such as
a vehicle model based method is developed to simulate responses in using galvanizing, epoxy-coating, cement grout, and sheathing, a large

* Corresponding author.
E-mail address: magren@126.com (Y. Ren).

https://doi.org/10.1016/j.istruc.2022.03.093
Received 6 January 2022; Received in revised form 30 March 2022; Accepted 31 March 2022
Available online 7 April 2022
2352-0124/© 2022 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
Z. Fan et al. Structures 39 (2022) 1176–1188

Fig. 1. Practical fatigue reliability-based replacement strategy.

amount of wires inside the cable were still found to be corroded [15–17]. First, based on recorded vehicle information, the statistical tools were
This symptom can result in corrosion fatigue, leading to accelerated performed to divide vehicles into several types according to the number
performance deterioration of wires, premature failure of stay cables, or of axles and axle groups. For each type of vehicles, the probabilistic
even bridge collapse [18,19]. Previously, extensive investigations have models of vehicle parameters were established. Then, with the assis­
been conducted to reveal corrosion behaviours of wires and cables. Li tance of finite element model (FEM), a Monte Carlo based random
et al. [20] tested mechanical properties and residual fatigue life of vehicle load model was developed to simulate responses in stay cables.
corroded cables that collected from a cable-stayed bridge in China Meanwhile, the time-variant reliability index was introduced to quan­
serving for 18 years. Li et al. [21] conducted accelerated corrosion ex­ titatively characterize the performance deterioration according to the
periments to evaluate the variation in the corrosion depths of cable damage accumulation rule, where the time-dependent corrosion inside
wires over time. Zheng and Wang [22] studied corrosion fatigue damage the cables was considered. Based on the time-variant reliability index, a
evolution and predict the residual life of cable wires by the cellular cost analytic was conducted to obtain an optimal cable replacement
automaton based finite element method. The mutual promotion of schedule by minimizing the expected average annual cost.
corrosion and fatigue results in large performance degradation in cable
wires. Moreover, fracture mechanics and continuum damage mechanics 2. Methodology
were used to investigate the fatigue crack growth in corroded cable
wires, and the corrosion environment was found to accelerate the crack Since the annual average cost largely depends on the timing of
propagation [23,24]. Correspondingly, plenty of rehabilitation actions replacement, it is meaningful to develop a practical method to deter­
of the deteriorated cables were also needed to be carried out [25]. mine an optimal replacement schedule to lower potential costs [33].
Additionally, in order to assess the condition of stay cables, reli­ Based on the time-varying reliability degrading model and potential
ability based methods have been proposed [26]. In this respect, Faber costs analysis, a practical replacement strategy for each stay cable is
et al. [27] presented models for assessment of the strength and fatigue developed to minimize the expected annual average cost. The flowchart
life of cables, where inspection results were used to update the reli­ for the reliability based replacement strategy is shown in Fig. 1. First, the
ability. Liu et al. [28] assessed distortion-induced cracking in steel vehicle effects need to be evaluated to calculate the reliability index of
bridges with a fatigue reliability method integrating monitored data. stay cables. Then, deficiencies and damages are discussed based on field
Cheng [29] developed a reliability based procedure that can be used to inspection and maintenance records, and the costs are also determined.
assess the cable-stayed bridge towers under ship impact loading. Finally, the replacement strategy is determined by minimizing the ex­
Accounted for different scenarios in terms of future changes in opera­ pected annual average cost. By now, many cable replacements projects
tional situations, time-varying reliability results can also be generated in China were conducted as replacing all or a batch of cables at the same
[30]. As reliability based methods gained increasing acceptance, addi­ time. As a novel way, the proposed method combined random vehicle
tional research was developed to quantify the risk, cost, and benefit model, reliability and cost analysis, the results can be regarded as partial
associated with bridges [31]. The management planning that empha­ references in an application.
sizes analysis, prediction, optimization and decision-making under un­
certainties needs to be further studied [32]. 2.1. Performance deterioration and maintenance of stay cables
This paper presented a case study of using fatigue reliability-based
method to achieve an optimal replacement strategy for stay cables. For cable-stayed bridges, the design life is at least 100 years; whilst

1177
Z. Fan et al. Structures 39 (2022) 1176–1188

(5) remove rust, repair cracks or deformation in anchor and other


parts.
β0
In addition, as for limited corroded cables, it should sever partial
polyethylene protective cover to make wires exposed. Then, remove rust
Reliability index β

ΔtD on the surface of wires by steel brush, apply epoxy zinc rich primer, and
fill the gap between wires with grease. Finally, repair the polyethylene
a c protective cover.
For deteriorating bridges or components, maintenance actions can be
divided into protective and essential maintenance [35]. Protective
maintenance, such as crack repair and coating, is considered to prevent
or delay the component deterioration in a period of time (i.e., the
βTarget b d duration of maintenance effects). In contrast, essential maintenance is
intended to strengthen bridge components to a good state and improve
t1 t1+ΔtM t1+3ΔtM the level of resistance, usually with steel plate or fiber reinforced plastic
(FRP). To effectively quantify the performance deterioration, the reli­
Service time t ability index of stay cable is introduced and applied in the optimization
of replacement schedule [36]. Measures introduced above are all clas­
Fig. 2. Time-dependent reliability index with and without maintenance for
a cable.
sified as protective maintenance that cannot increase cable reliability.
Based on the maintenance actions of actual bridges, the schematic dia­
gram of time-dependent reliability of a deteriorating cable with and
for stay cables, the design life is only 20 to 30 years in mainland China
without maintenance is given in Fig. 2. The stay cables suffer from
[20]. Considering performance deterioration caused by various de­
repeated vehicle loads as well as environmental effects once the bridge is
ficiencies and damages during operational period, service lives of some
opened to traffic, which can be taken as the time of performance begins
stay cables are sometimes shorter than their design ones. Specifically,
to degrade.
bridge inspections show that corrosive environments inside the cables
Based on the specified deterioration model, the time-varying reli­
often result in corrosion [34]. Corrosion defects can significantly reduce
ability indexes can be derived under various material properties and
the cross sectional area and constitutive relation of the wires, initiate
time-dependent constraint conditions [37]. The practical model for stay
corrosion fatigue cracks under continuous traffic loading, eventually
cables is developed based on the following assumptions: (1) the cable
cause wire breakage [18]. Other common deficiencies of stay cables
performance begins to degrade when the bridge is opened to traffic, (2)
including protective system damages, wire mechanical wear, cable force
the duration of maintenance effects are equal, (3) the maintenance in­
deviation and water seepage in anchors.
tervals are equal or longer than the duration of maintenance effect, and
In view of the severe prognosis of deficiencies at cables, periodic
(4) the protective maintenance actions only prevent the component
inspection and maintenance actions should be taken to maintain cable
deterioration in duration of maintenance effect but cannot increases
performance. Some measures are introduced as follows:
reliability. The deterioration process with maintenance can be deduced
based on the original deterioration process without intervention and
(1) identify and seal the position of water seepage to keep the anchor
several related factors, including time to first maintenance (t1), duration
of stay cable clean and dry;
of maintenance effect (ΔtD), maintenance interval (ΔtM). Except the
(2) replace the protective oil in the anchor periodically, as well as
duration of maintenance effect, the deterioration process with mainte­
supply anti-corrosion paint to the screw rod, nut and other parts;
nance follows the original deterioration process, beginning from the
(3) replace the waterproof washer and damping washer at the joint of
position with an equal reliability index, for instance, curve ab parallel to
steel protective shell periodically;
curve cd in Fig. 2.
(4) repair damage in polyethylene protective cover;

Analysis of traffic data Traffic load simulatoin

The toll station database Video monitoring data Consider car-following theory

Generate vehicle parameters by


Monte Carlo sampling method
Vehicle type classification
Calculate response influence line
by FEM method
Axle number Wheelset number Wheelbase
Traffic load simulation using the
prepared data

Time-history results of vehicle


Statistical probability model
induced response
Build the probability models of vehicle parameters according to
traffic data, including vehicle type, gross weight, axle weight, Validate and save the results
wheelbase, arrival time, driving lane and speed.

Fig. 3. Flowchart of the developed vehicle load model.

1178
Z. Fan et al. Structures 39 (2022) 1176–1188

2.2. Vehicle load model


Z = g(X) = R − S (1)
A general approach to simulate vehicle load is using statistical where Z is defined as functional function of structure and the X is the
probability distributions obtained from bridge weighing measurements random variables series {x1, x2,…, xn}T.
as the basis for Monte Carlo simulation [38]. The procedure for vehicle The ability of bridge components to resist load effect decreases
load simulation is proposed as shown in Fig. 3. Vehicles need to be gradually with the increase of stress cyclic number. If taking damage
classified into different types according to their axle number, wheelset accumulation parameter D(n) as the control parameter, the critical state
number and wheelbase. Then, several important parameters that reveal function of internal damage can be described as:
traffic characteristics are extracted to build probabilistic models. The
g(Dc , C) = Dc − D(n) = 0 (2)
lane statistical distributions of each type of vehicle are also determined.
Hence, a field measurements based vehicle database can be established. where Dc is the Miner’s critical damage accumulation index, which
Meanwhile, the structural influence lines are calculated with the FEM can be regarded as random process that follows the normal distribution
method. Finally, to obtain the time history of vehicle-induced responses with a mean value of 1.0 and a standard deviation of 0.3; D(n) is the
corresponding to a specific period and traffic volume, the generated damage accumulation index, which is also random process that in­
dynamic vehicle flow that involves car-following theory and Monte creases monotonically with the cyclic number n.
Carlo sampling method is acted on structural influence lines. In view of that variable stress ranges inevitably exist in actual
After obtaining the load time history, the rain-flow counting method, structures, result from the randomness of loads, the damage accumula­
which have been widely used and recognized, can be used to count the tion index can be assumed to be the sum of the damage index caused by
stress range and corresponding cyclic number. each stress range Si. To descript the fatigue process by Miner’s law, D(n)
can be expressed as Eq. (3).

2.3. Reliability index based on damage accumulation ∑


k ∑
k
ni
D(n) = Di = (3)
i=1 i=1
Ni
In practical applications, the reliability index of stay cables can be
where Di is the damage owing the ith stress range; ni is the cyclic
calculated by first order second moment method based on the S-N curves
number experienced by the structural detail corresponding to the stress
and Miner’s law for damage accumulation. Considering the advantages
range Si and Ni is the critical cyclic number (i.e., fatigue failure) for a
of calculation simplicity and wide comprehension in applications,
structural detail corresponding to the stress range Si.
Miner’s law was used although some more refined approaches are now
The main factors that influence fatigue resistance of stay cables are
available.
the stress range S and the cyclic number N, which can be obtained from
The design of any structure or component requires that its resistance
statistical result of fatigue test or existing specifications. The relation­
R is greater than the load effect S (R > S). This requirement is expressed
ship between S and N can be expressed as:
as:

Fig. 4. Description of case bridge and stay cables: a) side view with sensor layout, b) vertical view with location of toll station and c) structure of stay cables.

1179
Z. Fan et al. Structures 39 (2022) 1176–1188

Fig. 5. The designed information of case bridge: a) force and length, b) area.

N × Sm = C⇒logC = logN + mlogS (4) where ADC is the average daily stress cycles.

where m and C are performance parameters of materials, whose


3. Case study
value can be obtained from existing specifications or test results.
Substituting Eq. (4) into Eq. (3), according to the S-N curves and
To evaluate the effectiveness of the proposed method, a case study of
equivalent rule, the fatigue damage accumulation index D(n) can be
stay cables on the Dashengguan Yangtze River Bridge (DYB) was per­
defined as:
formed for demonstration and providing the optimal replacement

k
ni ∑
k
Sm n ∑k
ni nSm schedule, which is significant to both existing and newly built cable
(5)
eq
D(n) = = ni × i
= × Sm
i = stayed bridges.
i=1
Ni i=1
C C i=1 n C

where n is the accumulated cyclic number for every stress range and 3.1. Description of the long-span bridge and stay cables
Seq is the equivalent stress range. The n and Seq can be calculated as Eq.
(6) and Eq. (7). The studied cable stayed bridge has a main span of 648 m, was
opened to traffic in 2005 and was named as the Third Nanjing Yangtze

k
n= ni (6) River Bridge before 2020. It has continuously served for almost 16 years.
i=1 It stretches across the Yangtze River and connects Nanjing City and its
[ ]1/m Pukou District. The overall layout of the bridge and its stay cables are

k
ni × Sm shown in Fig. 4. The main girder is designed as bidirectional 6 lanes. A
Seq = i
(7) sophisticated structural health monitoring (SHM) system was equipped
i=1
n
to continually track the operational situation of the bridge, which
The fatigue reliability index β is then defined as the mean value of Z included advanced anchor load cell (ALC) sensors to measure cable force
divided by its standard deviation. For engineering structures, reliability directly and high definition video cameras to monitor traffic. The high
index β represents a quantitative probability to measure ability of precision ALC sensors with an acquisition frequency of 10 Hz and a
achieving expected structural function within reference period. relative error of ± 1% were installed in the anchors of all stay cables.
Assuming all the variable are independent, the functional function of The DYB used a static weighing system (different from WIM system)
stay cables can be expressed as Eq. (8) according to the critical state which is located at bridge toll stations, a distance from the main bridge,
function in Eq. (2), hence the fatigue reliability index β can be deduced as shown in Fig. 4b. For each vehicle, its information such as arrival
and calculated by Eq. (9) [2,39]. moment, vehicle type, total weight, axle weight, axle distance, and
driving lane will be recorded by the system before entering into the main
Z = lnDc − lnD(n) (8)
bridge. The vehicle speed can be deduced and calculated from moni­
μZ μlnDC − lnn − mlnSeq + μlnC toring videos.
β= = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (9) The main girder is supported by 168 stay cables, numbered as in
σZ σ 2lnDC + σ 2lnC Fig. 4a. Each cable consists of 109 to 241 wires with a diameter of 7 mm
[40]. The structure of stay cable is shown in Fig. 4c. The designed area,
where μlnx and σlnx denotes the mean value and standard deviation of
length and pre-stressed force for all types of stay cables in DYB are
each random variable, whose value can be calculated as Eq. (10) and Eq.
shown in Fig. 5, which range from 4195 ~ 9275 mm2, 91.93 ~ 354.53 m
(11). The subscript × stands for variables Dc, C and N, respectively [37].
and 1650 ~ 4067 kN, respectively. The galvanized high-strength steel
( )
μx wires with good fatigue resistance were used in the cable design.
μlnx = ln √̅̅̅̅̅̅̅̅̅̅̅̅
̅ (10) The bridge was also equipped with professional management teams
1 + δ2x
to conduct periodic visual inspection and maintenance actions. The
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ general inspection procedure to stay cables of the DYB including:
( )̅
σ lnx = ln 1 + δ2x (11)
• The tightness of surface wears: damage or aging.
where considering the time-varying effect, the accumulated number • Check the integrity of each cable: cracking, bulging or deformation.
of cycles N(t) in the year t can be estimated as following equation: • Check the condition of each anchor: water seepage or corrosion.
• Check the appearance and condition of connection components at
N(t)= 365×ADC × t (12)
each socket: loose, corrosion, leak and aging of water-proof gasket.

1180
Z. Fan et al. Structures 39 (2022) 1176–1188

Fig. 6. Typical field visual inspection deficiencies in stay cables of the TNYB: a) surface damage, b) anti-corrosion paint of the steel protective shell falls off, c) oil
leakage at the joint, d) damage in polyethylene protective cover, e) cracking and corrosion in steel protective shell.

• Testing the cable forces based on field frequency measurements. In view of the large number of probability distributions for each type
of vehicle as defined in Table 1, taking type 3 as an example for dis­
According to annual visual inspection records from 2011, the com­ cussion. The probability densities of gross weight, speed and time in­
mon deficiencies in stay cables of DYB include cracking and corrosion at terval were shown in Fig. 7, in which the corresponding parameters
steel protective shell, oil leakage at the anchor joint and damage at were illustrated by Table 2. The speed was modelled by t-distribution
polyethylene protective cover, as shown in Fig. 6. Bridge inspections while time interval was modelled by the exponential distribution. Time
revealed that the damage was concentrated at the lower ends of stay interval in Fig. 7c is defined as the moment of successive vehicles arrival,
cables, which was due to rainwater inflow and relative movements be­ and their behaviours are then simulated using car-following model. The
tween the cable and the girder [41]. The failure of the protection system gross weight was modelled by the mixture lognormal distribution with 6
for stay cables will accelerate corrosion in wires, while severe corrosion components. The vehicle distributions on each lane of the bridge can
and fatigue are the main failure mechanisms of stay cables. also be determined from traffic data, for instance, the proportions of
vehicle type 3 at fast lane, middle lane, and slow lane were 32.55%,
52.91% and 14.54%, respectively.
3.2. Traffic load simulation
3.2.2. The vehicle traffic volume
The traffic data between January 2006 and April 2009 were collected To estimate the vehicle-induced responses, a reasonably annual
from the bridge weighing system and video camera to establish the average daily traffic (AADT) should be determined as an input param­
vehicle model, which involve a total of 6,151,697 records. eter of the vehicle load model. With the DYB opening to traffic, road
network around the bridge has gradually formed and the traffic volume
3.2.1. Probability distributions of vehicle parameters showed a significant increase trend. The AADT increases from 12,399 to
There are many types of vehicles on highway bridges, which can be 35,758 during 2006 to 2019. The measured traffic volume data from
divided into two major categories according to load, namely cars and 2006 to 2019 are shown in Fig. 8.
trucks [8]. Through the preliminary processing of the data collected To avoid traffic jams at connection passageway of Nanjing City and
from the bridge weighing system, the axle number of vehicles passing its north district. Several other bridges and underwater tunnels located
DYB was mainly listed from two to six. According to the axle number, at Yangtze River near the case bridges have been constructed to reduce
wheelset number and wheelbase of vehicles crossing the bridge, a total traffic at rush hours. The average daily traffic volume of DYB maintained
of 12 types and their proportions were identified, which contained most at a stable level of slightly more than 30,000 in recent years. The traffic
vehicles (99.63%) passing the bridge, as shown in Table 1. volume increase of DYB was limited in the future. Considering the
Based on the bridge weighing system data and monitoring video, the operational situation, the AADT of 40,000 was adopted for the following
major parameters of vehicles including gross weight, axle weight, arrival discussion.
time, driving lane and speed were determined to develop the traffic load
model. The parameters of each vehicle type were treated as random 3.2.3. Evaluation of traffic load
variables. For these random variables that have one mode, they can be After determining the parameters of the developed vehicle load
assumed to satisfy a certain distribution. Then, the maximum likelihood model, the traffic load effects of DYB can be simulated with the assis­
estimation was used to calculate the parameters of the probability dis­ tance of FEM method. The details of validated FEM can be found in
tribution function. The Kolmogorov-Smirnov (KS) test was also applied former research [43]. The MATLAB programming platform was used to
to check the determined probabilistic model. However, for these random achieve Monte Carlo random sampling, then the generated vehicle
variables that have one mode, the expectation maximization (EM) al­ moved according to car-following theory. A step length of 1 s was set to
gorithm was used to calculate the parameters of probability models. In obtain real time locations of each vehicle on the bridge. The simulated
addition, the car-following theory was used to consider three situations, vehicle sequences were loaded on cable force influence lines that
namely acceleration, deceleration, and coast [42].

1181
Z. Fan et al. Structures 39 (2022) 1176–1188

Table 1 rain-flow counting method was used to determine the stress spectrum
Vehicle types and associated information. with a mean-range distribution from original data. According to the
Vehicle Graphic illustration Proportion Wheelbase (m) stress range spectrums of all cables, the calculated minimum equivalent
type (%) stress range was 6.64 MPa (No. A1) and the maximum one was 16.31
1 5.45 2.920 MPa (No. A5). It should be noted the cable No. A21 showed the
maximum cable force range, but it also had the maximum area that
about twice of cable No. A5. Fig. 11 shows the stress range spectrum of
the No. A1 and No. A5 cables with the AADT of 40,000. It indicates most
stress ranges in stay cables during the life period are at a low level.
2 53.87 4.862

3.3. Reliability based assessment for stay cables

3 3.28 3.987 + 1.377 To select the most critical stay cable as a sample to study, Fig. 12
shows the fatigue damage D(n) of cables in 1 day (24 h), which were
calculated according to Miner’s law for damage accumulation as Eq. (3).
4 3.69 1.811 + 5.331 Cable No. A5 was regarded as the most critical stay cable. Meanwhile,
considering that cable No. J21 is the longest cable with almost the
maximum force that has drawn more attention by the bridge manage­
ment department, it was also selected.
5 2.86 1.791 + 4.216
The corrosion of steel wires is a crucial factor resulting in the failure
+ 1.361
of cables. The property experiments of accelerated corroded high-
strength steel wires in the laboratory and naturally corroded bridge
6 1.32 3.525 + 5.810
wires showed that the performance parameters m and C of the steel wires
degraded with service time [45]. As a sophisticated process, researches
on corrosion deterioration and damage mechanism was still insufficient.
7 11.39 3.525 + 5.590
+ 1.312 On the other hand, there was no interior inspection for stay cables of the
case bridge. A simplified corrosion model was used for illustration. In
8 12.50 3.525 + 6.811 former accelerated corrosion experiments for galvanized high-strength
+ 1.313 + 1.313 steel wires [21], the corrosion model for uniform depth of wires can
be expressed as:
9 0.56 3.233 + 1.376
+ 5.867 + 1.312
{
ψ 1 tγe1 te ⩽tc
du (te ) = (13)
10 3.31 3.233 + 1.376 ψ 1 tγc1 + ψ 2 (te − tc )γ2 te > tc
+ 6.811 +
1.313 + 1.313 where du (te ) is the uniform corrosion depth; te is the exposure period;
11 0.04 1.890 + 2.416
tc is the moment that the zinc coating be completely consumed; ψ 1 and γ1
+ 5.897 + 1.312 is the corrosion parameters of zinc coating; ψ 2 and γ2 is the corrosion
parameters of steel. In that study [21], the parameters wereψ 1 = 4.381,
12 1.36 1.890 + 2.416 γ 1 = 1.024, ψ 2 = 0.672, γ 2 = 0.915, respectively. The DYB is located near
+ 6.811 +
an industrial area. According to a material corrosion investigation of
1.313 + 1.313
China, the 24 μm galvanized zinc coating was completely consumed in
3.9 years when exposure to environment of the DYB location [46]. The
calculated from the FEM (see Fig. 9) to compute the time history of transformation of cable wires corrosion rate between experimental and
vehicle-induced cable force. In view of the symmetry for case bridge, local actual environmental condition can be deduced according to the
only 42 stay cables, accounted for one quarter of total number were equivalent uniform corrosion depth:
investigated. te
(14)
γ
d = ψ actual,1 tactual
1
= ψ 1 tγe1 ⇔ = 1.35 days/year
The simulated cable force time history within one hour of the 5 tactual
typical stay cables, including short cable No. J1, No. A1 and long cable
where ψ actual,1 and tactual are the parameter and time of zinc coating
No. J21, No. A21, were extracted as examples to show the results in
corrosion in actual environmental condition. That indicated corrosion in
Fig. 10. An impact factor was used to consider the roughness of the road
actual environmental condition for 1 year can be simulated by 1.35 days
surface and the dynamic effect of moving vehicles on the bridge. As the
in the used experimental condition of the research. For cable wires used
cable forces can also be continually measured by sensor of the bridge
by DYB, the initial fatigue parameters can be regarded as m0 = 3.645 and
SHM system, the simulated responses have been compared with the
C0 = 1015.075. By time scale transformation and regression fitting from
monitoring data, whose detail was demonstrated in former research
the exited research [21,47], the time-varying performance parameters m
[44]. According to the simulation results with the AADT of 40,000, the
(t) and C(t) can be identified as:
significant difference in force existed for various cables, such as ampli­
tude (-49.82 ~ 236.06 kN for No. J21, − 34.48 ~ 56.98 kN for No. J1, m(t)= − 0.329ln(t) + 3.645 (15)
− 31.25 ~ 58.58 kN for No. A1, − 20.59 ~ 238.19 kN for No. A5 and
− 89.39 ~ 290.90 kN for No. A21) and mean value (51.15 kN for No. C(t)= 10− 0.973ln(t)+15.075
(16)
J21, 4.26 kN for No. J1, 4.83 kN for No. A1, 40.78 kN for No. A5 and
64.08 kN for No. A21). where t is the years of service period. Therefore, substituting the
In order to calculate the equivalent stress range for 42 stay cables already known performance parameters into Eq. (9), the time-varying
(No. A1 ~ A21 and No. J1 ~ J21), cable forces were obtained with the reliabilities for cables No. A5 and No. J21 with the AADT of 40,000
simulation time of 24 h. The force data must be converted into corre­ were illustrated in Fig. 13. The time to corrosion initiation was assumed
sponding stress time history by dividing cable force by area. Then, the as that bridge started service. The length effect of cable was not
considered in the calculation, which requires further study.

1182
Z. Fan et al. Structures 39 (2022) 1176–1188

0.06 0.035
Measured Measured
Mixture lognormal 0.03 t-distribution
0.05
distribution
0.025

Probability density
Probability density

0.04
0.02
0.03
0.015
0.02
0.01
0.01 0.005

0 0
0 10 20 30 40 50 60 70 80 0 20 40 60 80 100 120 140 160 180 200
(a) Vehicle weight of type 3 (ton) (b) Vehicle speed of type 3 (km/h)

0.08
Measured
0.07 Exponential distribution
0.06
Probability density

0.05

0.04

0.03

0.02

0.01

0
0 20 40 60 80 100 120 140 160
(c) Time interval between successive vehicles (s)

Fig. 7. The examples of probability density distributions of vehicle parameters: a) gross vehicle weight for type 3, b) vehicle speed for type 3, c) time interval
between successive vehicles.

regarded as a target reliability index according to related study [37]. In


Table 2
fact, the stress range caused by cyclic vehicle loads is negligible for an
Parameters for probability distribution of vehicle type 3.
intact stay cable, however, corrosion damage will lead to serious local
Vehicle Distribution type Distribution parameters stress concentration and influence its service life. Considering the great
variable
influence of corrosion on the performance, it must be taken into account
Gross weight Mixture lognormal w1 = 0.129; μ1 = 3.38; σ1 = in the design, maintenance, and replacement for stay cable.
distribution 0.060
w2 = 0.011; μ2 = 2.57; σ2 =
0.596
3.4. Optimization of replacement schedule
w3 = 0.197; μ3 = 3.24; σ3 =
0.099
w4 = 0.304; μ4 = 2.54; σ4 = In view of the contradiction between increasing bridge management
0.169 demand and limited financial, as well as smaller social impacts, the
w5 = 0.205; μ5 = 3.49; σ5 = optimization of stay cable replacement schedule should be carried out.
0.136
To avoid cable sudden failure in service, the corroded cables of many
w6 = 0.154; μ6 = 3.02; σ6 =
0.175 bridges were replaced according to inspection and assessment results in
Speed t-distribution μ = 83.99; σ = 13.82; ν = 6.03 China. For instance, the cables of Chongqing Shimen Bridge (served for
Time interval Exponential distribution λ = 1/13.6015 17 ~ 20 years), Nanchang Bayi Bridge (served for 12 years) and
Guangzhou Haiyin Bridge (served for 7 years) were all replaced during
the design service period [48]. However, the criteria to decide whether a
The initial reliability indexes of the selected cable were 6.20 for No.
cable should be replaced have not been established, and most of cable
A5 and 7.21 for No. J21, as Fig. 13 shows. It can be seen that the reli­
replacement projects were usually conducted based on inadequate
ability index of stay cables gradually decreases with service time. The
judgment. Minimization of the expected annual average cost is one
degradation of reliability index is less obvious for original cables
approach for determining the optimal replacement schedule, which is
compared with corroded cables. Under the condition of considering
useful to minimize cost while preserving relative safety margin of bridge
corrosion, the reliability index of stay cables significantly drops down,
components.
and the decline range increased with time. It is about 20 years when
The total cost for bridge components during service period includes
reliability index of all corroded cables fell below 3.0, which can be
inspection cost, maintenance cost, replacement cost and failure cost. The

1183
Z. Fan et al. Structures 39 (2022) 1176–1188

Monthly average daily traffic volume (vehicles) 50000

40000

30000

20000

10000

0
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
Time (year)
Fig. 8. The measured average daily traffic volume from 2006 to 2019.

Fig. 9. The cable force influence lines of 42 stay cables.

Fig. 10. The examples of simulated cable force time history.

1184
Z. Fan et al. Structures 39 (2022) 1176–1188

0.27
No.A1 No.A5
0.15
0.20

Probability
Probability

0.10
0.13

0.05 0.07

0 0
0 2 4 6 8 10 12 14 16 0 5 10 15 20 25 30 35 40
(a) Stress range (MPa) (b) Stress range (MPa)

Fig. 11. Stress range spectrum: a) No. A1, b) No. A5.

expected annual average cost can be estimated as follows:


[ ]/
CA = CI + CM + CR + CF × Pf (TR ) TR (17)

where CA is the cumulative annual average cost; CI is the cumulative


inspection cost; CM is the cumulative maintenance cost; CR is the
replacement cost; CF is the failure cost for each stay cable, which is
estimated to be ¥ 50 million; Pf(TR) is the failure probability in service
time TR, which can be converted from reliability index and TR is the
service time until replacement. The calculated CA is not discounted to
present value in this paper, considering cable replacement periods are
shorter compared with bridge reference period. However, it still needs
further study when considering multiple-replacement and bridge life-
cycle cost.

Table 3
Cost of different intervention actions for each cable in DYB.
Intervention Cost (CNY Time interval Duration of effect (year)
Fig. 12. Vehicle induced daily fatigue damage of stay cables. ¥) (year)

Inspection 1333 2 –
Maintenance 8333 5 2
Replacement 597583 – Improve performance to the
original condition

9
A5
8 A5-Corroded
7 J21
J21-Corroded
Reliability index β

6
5
4
3
2
1
0
0 10 20 30 40 50 60 70 80 90 100
Time (year)
Fig. 13. Time-varying reliability indexes of original and corroded stay cables.

1185
Z. Fan et al. Structures 39 (2022) 1176–1188

×104
8 70

Expected annual average cost (¥)


No. A5 No maintenance actions No. A5
7 1

With maintenance actions 60


Reliability index β

6 50
5
40
4
30
3
2 20
(17, 44551)
1 10
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
(a) Time (year) (b) Time (year)
Fig. 14. Optimization for cable No. A5 with maintenance actions: a) reliability index, b) expected annual average cost estimation.

4
8 70 ×10
Expected annual average cost (¥)
No. J21 No maintenance actions No. J21
7 1

With maintenance actions 60


Reliability index β

6 50
5
40
4
30
3
2 20
1 10 (32, 25520)
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
(a) Time (year) (b) Time (year)
Fig. 15. Optimization for cable No. J21 with maintenance actions: a) reliability index, b) expected annual average cost estimation.

In this paper, the costs within the analysis are determined according replacement for each cable in DYB are listed in Table 3, together with
to the experience and extensive investigation in China. The cost of corresponding time interval and duration of effect. The periodic main­
different intervention actions, including inspection, maintenance and tenance actions are carrying out moisture-proof measures, replacing

50
Service time (year)

40

30

20

10
A21 A15 A8 A1 J14 J14 J21
Serial number
Fig. 16. Replacement schedule for stay cables of DYB.

1186
Z. Fan et al. Structures 39 (2022) 1176–1188

wrapping tape of polyethylene tube, repairing the damage of cable and 5. Data availability statement
its components, etc. The replacement includes the whole process of
producing new-cable, old-cable removal, installation. With the appli­ All data of reliability analysis are available from the corresponding
cation of the quick-replacing technique, it is possible to achieve author by request.
replacement for partial cables under normal vehicle flows, hence elim­
inating the cost of disruption to traffic.
The optimization results of replacement strategies are shown in Declaration of Competing Interest
Fig. 14 and Fig. 15. Note that corrosion has been considered as described
above. It is indicated that the maintenance measures can delay the The authors declare that they have no known competing financial
deterioration of stay cables within a period of time. For cable No. A5, interests or personal relationships that could have appeared to influence
when it is replaced in the 17th year of service, the expected annual the work reported in this paper.
average cost reaches its lowest point of 44,551. For cable No. J21, when
it is replaced in the 32th year of service, the expected annual average Acknowledgements
cost reaches its lowest point of 25,520. Since failure probability is low in
earlier stages, the expected annual average cost decreases initially with The study was supported by the Postgraduate Research & Practice
longer expected service life. However, it then increases with increasing Innovation Program of Jiangsu Province (No. KYCX19_0099), the
failure probability and more maintenance effort. There is an optimal Academician Special Science Research Project of CCCC (No. YSZX-03-
level of replacement schedule that will minimize the expected annual 2021-01-B and YSZX-03-2021-02-B), the Jiangsu Transportation Sci­
average cost. Once the replacement time of stay cables excesses opti­ ence and Technology Project (No.2020Y19-(1)), and the Scientific
mum opportunity, the potential cost of failure enhances to enlarge the Research Project of CICO (No. ZC20210604XBFW000200/01).
expected annual average cost. Therefore, integrated consideration for
the performance and the economy, the optimal replacement schedule
References
can be determined as 17th year and 32th year for cable No. A5 and cable
No. J21, respectively. The conclusion drawn from the developed [1] Takena K, Miki C, Shimokawa H, Sakamoto K. Fatigue resistance of large-diameter
methodology is also similar to that from engineering experience. cable for cable-stayed bridges. J Struct Eng 1992;118(3):701–15. https://doi.org/
10.1061/(ASCE)0733-9445(1992)118:3(701).
Similarly, the optimal replacement schedule for all stay cables was
[2] Hou N, Sun L, Chen L. Cable reliability assessments for cable-stayed bridges using
determined, as shown in Fig. 16. Different stay cables have different identified tension forces and monitored loads. J Bridge Eng 2020;25(7):05020003.
optimal replacement timings. In practice, other factors including target https://doi.org/10.1061/(ASCE)BE.1943-5592.0001573.
reliability index and cable designed life should be considered together to [3] Xu X, Huang Q, Ren Y, Zhao DY, Zhang DY, Sun HB. Condition evaluation of
suspension bridges for maintenance, repair and rehabilitation: a comprehensive
an ensure final decision. Many cable replacement projects in past years framework. Struct Infrastruct Eng 2019;15(4):555–67. https://doi.org/10.1080/
were conducted as replacing all cables at the same time. The presented 15732479.2018.1562479.
study provides a theoretical foundation and guidance for management [4] Yang DY, Frangopol DM. Risk-based inspection planning of deteriorating
structures. Struct Infrastruct Eng 2022;18(1):109–28. https://doi.org/10.1080/
and replacement, which could be extended to other case studies. How­ 15732479.2021.1907600.
ever, this study focused more on single cable local optimization and [5] Fu Z, Ji B, Yang M, Sun H, Maeno H. Cable replacement method for cable-stayed
admitted some stochasticity. To reduce the cost, multiple cables with the bridges based on sensitivity analysis. J Perform Constr Facil 2015;29(3):04014085.
https://doi.org/10.1061/(ASCE)CF.1943-5509.0000460.
similar condition can be replaced at a mobilization. Further research [6] Li ZX, Chan T, Zheng R. Statistical analysis of online strain response and its
about global optimization will be studied in the future. application in fatigue assessment of a long-span steel bridge. Eng Struct 2003;25
(14):1731–41. https://doi.org/10.1016/S0141-0296(03)00174-3.
[7] Chen Z, Xu Y, Xia Y, Li Q, Wong K. Fatigue analysis of long-span suspension bridges
4. Conclusions
under multiple loading: case study. Eng Struct 2011;33(12):3246–56. https://doi.
org/10.1016/j.engstruct.2011.08.027.
This paper presented a case study of using fatigue reliability-based [8] Ma R, Xu S, Wang D, Chen A. Vehicle models for fatigue loading on steel box-girder
bridges based on weigh-in-motion data. Struct Infrastruct Eng 2018;14(6):701–13.
method to achieve an optimal replacement strategy for stay cables.
https://doi.org/10.1080/15732479.2017.1359308.
The following concluding remarks can be drawn from this research: [9] Ruan X, Zhou J, Caprani CC. Safety assessment of the antisliding between the main
cable and middle saddle of a three-pylon suspension bridge considering traffic load
(1) To evaluate the performance deterioration of stay cables with and modeling. J Bridge Eng 2016;21(10):04016069. https://doi.org/10.1061/(ASCE)
BE.1943-5592.0000927.
without maintenance and obtain an optimal replacement [10] Wang FY, Xu YL, Sun B, Zhu Q. Dynamic stress analysis for fatigue damage
schedule, the fatigue reliability index can be used based on the prognosis of long-span bridges. Struct Infrastruct Eng 2019;15(5):582–99. https://
damage accumulation and vehicle load simulation. doi.org/10.1080/15732479.2019.1569072.
[11] Xu X, Xu YL, Ren Y, Huang Q. Site-specific extreme load estimation of a long-span
(2) By using the collected traffic data, the probability models of cable-stayed bridge. J Bridge Eng 2021;26(4):05021001. https://doi.org/10.1061/
various vehicle parameters including vehicle type, vehicle (ASCE)BE.1943-5592.0001700.
weight, speed, and time interval, can be established for the traffic [12] Chen ZW, Xu YL, Wang XM. SHMS-based fatigue reliability analysis of
multiloading suspension bridges. J Struct Eng 2012;138(3):299–307. https://doi.
load simulation together with FEM, car-following theory and org/10.1061/(ASCE)ST.1943-541X.0000460.
Monte Carlo method. [13] Zhou J, Ruan X, Shi X, Caprani CC. An efficient approach for traffic load modelling
(3) The reliability index of stay cables gradually decreases with ser­ of long span bridges. Struct Infrastruct Eng 2019;15(5):569–81. https://doi.org/
10.1080/15732479.2018.1555264.
vice time. Under the condition of corrosion, the decline rate of the
[14] Nakamura S, Suzumura K. Experimental study on fatigue strength of corroded
reliability for stay cables is accelerated, which indicates the bridge wires. J Bridge Eng 2013;18(3):200–9. https://doi.org/10.1061/(ASCE)
importance of anti-corrosion for cables. BE.1943-5592.0000366.
[15] Xu J, Chen WZ. Behavior of wires in parallel wire stayed cable under general
(4) The optimal replacement schedule for stay cables can be deter­
corrosion effects. J Constr Steel Res 2013;85:40–7. https://doi.org/10.1016/j.
mined through cost analysis to minimize the expected annual jcsr.2013.02.010.
average cost. For the DYB investigated in this study, different stay [16] Jiang C, Wu C, Cai CS, Jiang X, Xiong W. Corrosion fatigue analysis of stay cables
cables have different optimal replacement timings, where the under combined loads of random traffic and wind. Eng Struct 2020;206:110153.
https://doi.org/10.1016/j.engstruct.2019.110153.
cable No. A5 has the shortest service life - replaced at the 17th [17] Liu ZX, Guo T, Han DG, Li AQ. Experimental study on corrosion-fretting fatigue
service year with the minimum expected annual average cost of ¥ behavior of bridge cable wires. J Bridge Eng 2020;25(12):04020104. https://doi.
44,551. org/10.1061/(ASCE)BE.1943-5592.0001642.
[18] Barton SC, Vermaas GW, Duby PF, West AC, Betti R. Accelerated corrosion and
embrittlement of high-strength bridge wire. J Mater Civ Eng 2000;12(1):33–8.
https://doi.org/10.1061/(ASCE)0899-1561(2000)12:1(33).

1187
Z. Fan et al. Structures 39 (2022) 1176–1188

[19] Liu ZX, Guo T, Hebdon HM, Zhang ZL. Corrosion fatigue analysis and reliability [34] Suzumura K, Nakamura S. Environmental factors affecting corrosion of galvanized
assessment of short suspenders in suspension and arch bridges. J Perform Constr steel wires. J Mater Civ Eng 2004;16(1):1–7. https://doi.org/10.1061/(ASCE)
Facil 2018;32(5):04018060. https://doi.org/10.1061/(ASCE)CF.1943- 0899-1561(2004)16:1(1).
5509.0001203. [35] Ahn S, Kim W. On determination of the preventive maintenance interval
[20] Li H, Lan CM, Ju Y, Li DS. Experimental and numerical study of the fatigue guaranteeing system availability under a periodic maintenance policy. Struct
properties of corroded parallel wire cables. J Bridge Eng 2012;17(2):211–20. Infrastruct Eng 2011;7(4):307–14. https://doi.org/10.1080/15732470802587129.
https://doi.org/10.1061/(ASCE)BE.1943-5592.0000235. [36] Barone G, Frangopol DM. Reliability, risk and lifetime distributions as performance
[21] Li SL, Xu Y, Li H, Guan XC. Uniform and pitting corrosion modeling for high- indicators for life-cycle maintenance of deteriorating structures. Reliab Eng Syst
strength bridge wires. J Bridge Eng 2014;19(7):04014025. https://doi.org/ Saf 2014;123:21–37. https://doi.org/10.1016/j.ress.2013.09.013.
10.1061/(ASCE)BE.1943-5592.0000598. [37] Liu ZX, Guo T, Correia J, Wang LB. Reliability-based maintenance strategy for
[22] Zheng Y, Wang Y. Damage evolution simulation and life prediction of high-strength gusset plate connections in steel bridges based on life-cost optimization. J Perform
steel wire under the coupling of corrosion and fatigue. Corros Sci 2020;164: Constr Facil 2020;34(5):04020088. https://doi.org/10.1061/(ASCE)CF.1943-
108368. https://doi.org/10.1016/j.corsci.2019.108368. 5509.0001493.
[23] Mahmoud KM. Fracture strength for a high strength steel bridge cable wire with a [38] Enright B, O’Brien EJ. Monte Carlo simulation of extreme traffic loading on short
surface crack. Theor Appl Fract Mech 2007;48(2):152–60. https://doi.org/ and medium span bridges. Struct Infrastruct Eng 2013;9(12):1267–82. https://doi.
10.1016/j.tafmec.2007.05.006. org/10.1080/15732479.2012.688753.
[24] Cui CJ, Chen AR, Ma RJ. An improved continuum damage mechanics model for [39] Guo T, Chen YW. Fatigue reliability analysis of steel bridge details based on field-
evaluating corrosion-fatigue life of high-strength steel wires in the real service monitored data and linear elastic fracture mechanics. Struct Infrastruct Eng 2013;9
environment. Int J Fatigue 2020;135:10554. https://doi.org/10.1016/j. (5):496–505. https://doi.org/10.1080/15732479.2011.568508.
ijfatigue.2020.105540. [40] Fan ZY, Huang Q, Ren Y, Zhu ZY, Xu X. A cointegration approach for cable
[25] Mehrabi AB, Ligozio CA, Ciolko AT, Wyatt ST. Evaluation, rehabilitation planning, anomaly warning based on structural health monitoring data: an application to
and stay-cable replacement design for the Hale Boggs Bridge in Luling. Louisiana cable-stayed bridges. Adv Struct Eng 2020;23(13):2789–802. https://doi.org/
Journal of Bridge Engineering 2010;15(4):364–72. https://doi.org/10.1061/ 10.1177/1369433220924793.
(ASCE)BE.1943-5592.0000061. [41] Sun HH, Xu J, Chen WZ, Yang JX. Time-dependent effect of corrosion on the
[26] Catbas FN, Susoy M, Frangopol DM. Structural health monitoring and reliability mechanical characteristics of stay cable. J Bridge Eng 2018;23(5):04018019.
estimation: long span truss bridge application with environmental monitoring data. https://doi.org/10.1061/(ASCE)BE.1943-5592.0001229.
Eng Struct 2008;30(9):2347–59. https://doi.org/10.1016/j. [42] Wang FY, Xu YL. Traffic load simulation for long-span suspension bridges. J Bridge
engstruct.2008.01.013. Eng 2019;24(5):05019005. https://doi.org/10.1061/(ASCE)BE.1943-
[27] Faber MH, Engelund S, Rackwitz R. Aspects of parallel wire cable reliability. Struct 5592.0001381.
Saf 2003;25(2):201–25. https://doi.org/10.1016/S0167-4730(02)00057-7. [43] Xu X, Huang Q, Ren Y, Zhao DY, Yang J, Zhang DY. Modeling and separation of
[28] Liu M, Frangopol DM, Kwon K. Fatigue reliability assessment of retrofitted steel thermal effects from cable-stayed bridge response. J Bridge Eng 2019;24(5):
bridges integrating monitored data. Struct Saf 2010;32(1):77–89. https://doi.org/ 04019028. https://doi.org/10.1061/(ASCE)BE.1943-5592.0001387.
10.1016/j.strusafe.2009.08.003. [44] Ren Y, Zhu Z, Fan Z, Huang Q, Zhang T. Estimation of extreme cable forces of
[29] Cheng J. Reliability analysis of the Sutong Bridge Tower under ship impact cable-stayed bridges based on monitoring data and random vehicle models. Adv
loading. Struct Infrastruct Eng 2014;10(10):1320–9. https://doi.org/10.1080/ Civ Eng 2021;2021:1–15.
15732479.2012.757792. [45] Zhuang M, Miao C. Fatigue reliability assessment for hangers of a special-shaped
[30] Farreras-Alcover I, Chryssanthopoulos MK, Andersen JE. Data-based models for CFST arch bridge. Structures 2020;28:235–50. https://doi.org/10.1016/j.
fatigue reliability of orthotropic steel bridge decks based on temperature, traffic istruc.2020.08.067.
and strain monitoring. Int J Fatigue 2017;95:104–19. https://doi.org/10.1016/j. [46] Cao C. Material corrosion in natural environment of China. Beijing: Chemical
ijfatigue.2016.09.019. Industry Press; 2005.
[31] Frangopol DM, Kong JS, Gharaibeh ES. Reliability-based life-cycle management of [47] Li Y, Yang TT, Xia LZ. Analysis of corrosion effect on fatigue reliability of stay
highway bridges. J Comput Civil Eng 2001;15(1):27–34. https://doi.org/10.1061/ cables under traffic and wind. in Chinese J Hunan Univ (Nat Sci) 2020;47(5):
(ASCE)0887-3801(2001)15:1(27). 107–15. https://doi.org/10.16339/j.cnki.hdxbzkb.2020.05.012.
[32] Frangopol DM, Dong Y, Sabatino S. Bridge life-cycle performance and cost: [48] Zheng X, Xie X, Li X. Experimental study and residual performance evaluation of
analysis, prediction, optimisation and decision-making. Struct Infrastruct Eng corroded high-tensile steel wires. J Bridge Eng 2017;22(11):04017091. https://doi.
2017;13(10):1239–57. https://doi.org/10.1080/15732479.2016.1267772. org/10.1061/(ASCE)BE.1943-5592.0001114.
[33] Kong JS, Frangopol DM. Life-cycle reliability-based maintenance cost optimization
of deteriorating structures with emphasis on bridges. J Struct Eng 2003;129(6):
818–28. https://doi.org/10.1061/(ASCE)0733-9445(2003)129:6(818).

1188

You might also like