Professional Documents
Culture Documents
Abstract: Fire is an accidental, severe hazard for bridges during their lifetime. Hangers in suspension bridges are among the most vul-
nerable components with respect to the hanger fatigue effect, and fatigue performance after fire exposure is vital to bridge safety. Therefore,
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.
a comprehensive assessment of the postfire hanger fatigue property is necessary. In this study, fatigue tests were conducted on steel wires
after various elevated temperatures, and a multiparameter Weibull model was adopted to describe the fatigue data. Based on the fatigue life
distribution of steel wires and the corresponding parallel systems, the hanger fatigue life was evaluated using the Monte Carlo simulation
and order statistics approach, and the S–N curves were obtained. The results demonstrated that the fatigue life of the hanger was signifi-
cantly lower than the mean life of the individual wires, and degraded as the exposure temperature increased. In addition, two small cables
consisting of 19 parallel steel wires were tested for verification, and the results were consistent with those of the analytical model. The
results of this study can be applied to quantify the extent of damage caused by fire and to assess the remaining hanger service life.
DOI: 10.1061/(ASCE)BE.1943-5592.0001938. © 2022 American Society of Civil Engineers.
Author keywords: Hanger; Steel wire; Fatigue; Parallel system; Postfire performance.
Introduction ductility) after heating (Lu et al. 2017; Zheng et al. 2006; Tao
2015), and they demonstrated inferior fire resistance owing to
Cable-supported bridges play a vital role in the traffic network grid, their low specific heat and high thermal conductivity. Conse-
and their safety has been a significant concern (Gong and Agrawal quently, hangers are adversely affected by fires. However, hangers
2016). In recent years, owing to the increase in traffic volume and are also prone to fatigue damage as they are subjected to significant
hazardous material transport, bridges have occasionally suffered cyclic fluctuations induced by vehicles and wind (Min Park et al.
from fires (Quiel et al. 2015), primarily caused by vehicular acci- 2013; Liu et al. 2017; Petrini and Bontempi 2011). In the research
dents (Garlock et al. 2012). For example, on March 4, 2017, a se- carried out by Zhang et al. (2021), it was found that the mechanical
vere vehicle collision occurred on the Runyang suspension bridge performance, especially the fatigue life, of the postfire steel wires is
(RSB) and caused a large fire that lasted for over an hour. The RSB significantly reduced as a result of the change in microstructure.
is one of the major connections between Zhenjiang and Yangzhou The most dramatic reduction of the fatigue life was found at expo-
crossing the Yangtze River in Jiangsu province, China. It was sure temperatures above 500°C. Since the postfire steel cables are
opened in 2005 and more than 40,000 vehicles passed through it likely to fail prematurely, it is necessary to investigate their fatigue
every day. Vehicles involved in the accident were thoroughly life quantitatively.
burned, although no casualties were reported. After the accident, In general, the fatigue properties of a hanger can be obtained via
an on-site inspection was immediately conducted. Four hangers fatigue tests, according to the specifications of the Ministry of
near the fire source were severely damaged, however, no broken Transport of China (MTC 2001). In this regard, it is stipulated
hanger wires were observed, as shown in Fig. 1. that the wire breakage rate in the hanger should not exceed 5%
Hangers are key load-carrying elements in suspension bridges after 2 × 106 fatigue cycles under a maximum load of 0.35 Pb (Pb
that connect the girder to the main cable (Jiang et al. 2009; Liu is the nominal breaking load) with a stress range of 150 MPa. How-
et al. 2019). Although the hangers mentioned previously did not ever, for cables consisting of a number of parallel steel wires,
fail after the fire, the observed damage poses a serious threat to owing to the capacity limitation of the test machine, evaluating
safety. Previous studies have revealed that high-strength steel ex- the fatigue behavior by testing the full cable cross section is imprac-
hibits deteriorated mechanical properties (such as strength and tical (Stallings and Frank 1991a).
Another method involves deriving the hanger fatigue life based
1
Ph.D. Candidate, School of Civil Engineering, Southeast Univ., on the fatigue life of the constituent steel wires. The Post-
Nanjing 210096, P.R. China. Email: zhaolei.zhang@seu.edu.cn Tensioning Institute PTI (2007) pointed out that the relationship
2
Professor, Key Laboratory of Concrete and Prestressed Concrete between the fatigue strength of the cable and steel wire can be ex-
Structures of the Ministry of Education, Southeast Univ., Nanjing pressed as ΔScable = ΔSwire − c, where the value of c is 100 MPa.
210096, P.R. China (corresponding author). Email: guotong@seu.edu.cn Birkenmaier and Narayanan (1982) conducted fatigue tests on
3
Dept. of Bridge Engineering, School of Transportation, Southeast eight cables, each consisting of 19 steel wires, and based on the
Univ., Nanjing 210096, P.R. China. Email: zhongxiang@seu.edu.cn test results and the S–N curve of the steel wire, it was proposed
4
Postdoctoral Researcher, School of Civil Engineering, Southeast
to adopt the number of cycles at which 5% of the total wires
Univ., Nanjing 210096, P.R. China. Email: shiyuan_wang@seu.edu.cn
Note. This manuscript was submitted on February 15, 2022; approved
break as the fatigue life of the cable, and c = 140 MPa. Neverthe-
on June 6, 2022; published online on August 4, 2022. Discussion period less, these cursory methods are conservative and not sufficiently
open until January 4, 2023; separate discussions must be submitted for in- precise.
dividual papers. This paper is part of the Journal of Bridge Engineering, Although the occurrence possibility of bridge fire is low, it can
© ASCE, ISSN 1084-0702. lead to fatal consequence, or even the collapse of the bridges
In the Eq. (1), N is the fatigue life of the steel wire under stress
range ΔS.
γc = (cξA0)−1/α, where c is an unknown parameter; for a wire
with the length of L, ξ = L/L0 (L0 is the length of the sample seg-
ment with the same properties), represents the length effect and
is assumed to be 1 in this study; A0 is a constant that represents
the cross-sectional area of the wire (A0 = 3.14 × 52/4 = 19.625 in
this study).
K = K0 (1 − ms /mZ )γ , where ms and mZ are the mean stress of
the fatigue test and static strength, respectively. In this study, the
stress ratio was 0.5, therefore, ms = 1.5ΔS. It was assumed that
Fig. 4. Fatigue testing machine used in this study. mZ did not vary with the cyclic load, and its values were 1,514
and 1,144 MPa for specimens cooled from 500°C and 600°C, re-
spectively, based on tensile tests. Furthermore, γ is assumed to
Table 3. Fatigue life results obtained after 500°C and 600°C under be 0.5 (Rackwitz and Faber 1991).
different stress ranges (×104) The rest of α, m′ , c, and K0 are unknown parameters. This dis-
tribution was also supported by previous experiments (Castillo
ΔS after 500°C (MPa) ΔS after 600°C (MPa) et al. 1985; Matsukawa et al. 1988). By defining m = α/m′ and
√
Test no. 290 360 430 500 290 360 430 500 g(ΔS) = 1 − ms /mz , Eq. (1) can be rewritten as follows:
1 NF 115.95 38.98 17.07 175.00 71.42 27.77 9.22 α
ΔS α N m′
2 NF 109.19 40.90 22.07 151.35 60.49 21.52 12.40 FN (N , ΔS) = 1 − exp −
3 NF 93.57 42.42 16.02 182.22 64.43 17.93 9.84 γc K
4 NF 142.74 37.39 12.29 198.80 72.56 17.44 11.16 m
5 NF 121.12 36.44 13.25 168.58 43.90 26.52 12.16 c N
6 NF 122.89 32.74 18.55 178.95 52.28 20.36 8.62 = 1 − exp − m A0 (ΔS)α
K0 g(ΔS)
7 NF 84.48 38.95 19.79 188.09 65.56 19.69 8.08
8 NF 99.46 46.04 18.06 167.95 53.85 22.72 10.91
m
N
9 NF 97.87 36.09 15.25 158.08 61.14 23.17 8.71 = 1 − exp λA0 (ΔS)α (2)
10 NF 103.58 34.76 18.01 193.30 58.90 29.43 10.88
g(ΔS)
11 NF 106.57 31.03 15.51 160.94 77.44 24.31 8.55
12 NF 87.31 44.37 16.82 184.39 71.10 22.59 9.67 where λ = c/K0m ; the unknown parameters become α, m, and λ, and
13 NF 91.91 38.11 15.12 161.72 66.08 24.06 10.41 these parameters can be evaluated from the test results via maxi-
14 NF 112.57 36.04 15.98 171.61 55.62 24.84 7.04 mum likelihood estimation. The probability density function is
15 NF 101.21 39.21 16.22 177.24 68.40 22.77 9.25
dFN (N ; ΔS)
Note: NF indicates that no failure occurred until the cycle limit of 2 × 106. fN (N ; ΔS) =
dN
ΔS is the stress range. m−1 m
α N m α N
= λA0 (ΔS) exp −λA0 (ΔS)
g(ΔS) g(ΔS) g(ΔS)
were heated no higher than 400°C, all of them reached the test end (3)
point of 2 × 106 cycles without fracture under a mean stress range of
571.5 MPa with ±180 MPa amplitude, as required in the code GB/ The log-likelihood function is
⎡ ⎤
ln (λ) + ln (A0 ) + α ln (ΔSi ) + (m − 1) ln (Ni ) + ln (m)
n
m
LfN (N ; ΔS) = ⎣ Ni ⎦ (4)
i=1
−m ln (g(ΔSi )) − λA0 (ΔSi )α
g(ΔSi )
The previous equations can be solved using the BFGS algorithm (short for Broyden–Fletcher–Goldfarb–Shanno, a type of quasi-Newton
method). In addition, it is noteworthy that none of the specimens broke prior to the 2 million repeated cycle limit under a stress range
of 290 MPa after they were cooled from 500°C; therefore, the fatigue test data were in censored form. Several methods can be used to
manage censored data in engineering, such as the expectation–maximization (EM) algorithm (Dempster et al. 1977), which is an iterative
approach of maximum likelihood estimation that is widely used to estimate parameters in a probabilistic model when the data are incomplete
(Tian et al. 2011; Lan et al. 2017a). Each EM algorithm iteration includes two steps: The expectation step (E-step) and maximization step
(M-step). In this study, the endpoint of the tests was set at N0 = 2 × 106, and the conditional probability distribution function is expressed as
follows:
F(N ) − F(N0 )
F(N |N ≥ N0 ) =
1 − F(N0 )
m
α N α N0 m
1 − exp −λA0 (ΔS) − 1 + exp −λA0 (ΔS)
g(ΔS) g(ΔS)
= m
N
1 − 1 + exp −λA0 (ΔS)α
0
g(ΔS)
m
α N α N0 m
= 1 − exp −λA0 (ΔS) + λA0 (ΔS) (6)
g(ΔS) g(ΔS)
m−1
λmA0 (ΔS)α N
fN ≥N0 (N ) =
g(ΔS) g(ΔS)
m m
N N
× exp −λA0 (ΔS)α + λA0 (ΔS)α (7)
g(ΔS) g(ΔS)
Discussion
The test results are summarized, and the failure cycles are plotted as
a function of the cyclic stress range. As shown in Fig. 8, the number
(b)
of wire fatigue cycles increased and became scattered as the stress
range decreased. Comparing Figs. 7(a and b) at the same cyclic
stress range, the fatigue cycle decreased significantly as the expo-
sure temperature increased, indicating that high temperature can
significantly reduce the life of steel wires.
Model for Fatigue Life Prediction Fig. 6. Convergence process of parameters in EM algorithm: (a) pa-
rameter λ; (b) parameter α; and (c) parameter m.
The hanger can be modeled as a bundle consisting of n steel wires
of the same diameter and length arranged in parallel, where the fric-
tion among the wires is disregarded, as shown in Fig. 9, ΔL denotes
Table 4. Estimated parameters of Weibull distribution for fatigue life of
the repeated load range applied to cable. Coleman (1958) and
steel wires
Phoenix (1978a) were the first to establish a theoretical fatigue
model for fiber bundles. The model was subsequently extended Temperature α m λ
by Rackwitz and Faber (1991) and applied to parallel-wire cables. 500°C 52.09 9.95 3.42 × 10−196
The fatigue load applied to the hanger was an axial non-negative 600°C 37.74 8.03 3.90 × 10−146
load with a constant range. Under cyclic loading, the wire with
the smallest failure time failed first in the system, followed by
that with the second-smallest failure time. One after the other fails, the force is redistributed among the remaining n-i compo-
breaks because of fatigue damage. Assume that any broken wire nents, and the stress range becomes,
loses its load-carrying capacity, and all surviving members equally
share the load at any time. This assumption is appropriate for hang- n
ΔSi = ΔS (9)
ers clamped at the two ends using an anchor. After the ith element n−i
(a) (b)
Fig. 7. Weibull probability plot of fatigue life of wires: (a) 500°C; and (b) 600°C.
(a) (b)
Fig. 8. Summary of steel wire fatigue test results: (a) 500°C; and (b) 600°C.
arranged in ascending order (i.e., N1 < N2 < … < Ni < … < N109).
The level of damage under cyclic loading is defined by the well-
known Miner’s fatigue rule: Mi = Ni/Nr, where Nr is the reference
fatigue life, for example, the mean life at a certain stress range.
The fatigue cycles of individual wires depend on the accumulation
of continuous fatigue damage derived from their load histories.
Under a constant axial load range, the first wire in the system
broke when the number of cycles reached N1, which is the
smallest value generated by the Monte Carlo method. Subse-
Fig. 9. Modeling of parallel wire bundles.
quently, the stress range among the remaining wires is n/(n−1)
ΔS. As the cyclic load continued, the second-smallest wire failed
where n = total number of wires in the hanger; i = number of bro- when the cumulative damage reached M2, and the number of cy-
ken wires; and ΔS = initial stress range. cles was
As shown in Eq. (9), the stress range among the surviving wires n −B
increases progressively, and the growing stress range accelerates N ′2 = Nr M1 + (M2 − M1 ) (10)
the cumulative damage in the remaining wires, resulting in faster n−1
wire breakage in the cable. Finally, all wires failed. In the general form, the number of cycles when the ith wire frac-
tures can be deduced by analogy, as shown in (11) (Stallings and
Frank 1991b). N ′n is the number of cycles when the last wire
Fatigue Life Prediction of Hanger
fails and is defined as the system fatigue life.
Monte Carlo Simulation
The Monte Carlo method was used to generate 109 random numbers −B
i
n
corresponding to the Weibull distribution, which simulates the fa- N ′i = Nr (Mj − M j−1 ) (11)
tigue life of each wire in the hanger. These random lives are j=1
n−j+1
1 −B
φ(x) = (12)
1−x
−B
j−1 n
Therefore, φ = (14)
n n−j+1
(a)
(b)
Fig. 12. Probability distribution of fatigue life of hanger under four load ranges at 5% breakage: (a) 500°C; and (b) 600°C.
where
F(u)[1 − F(v)] for 0≤u≤v
Γ(u, v) =
F(v)[1 − F(u)] for 0≤v≤u
As discussed previously, (18) and (19) represent the mean and
variance of the system, respectively, which can be solved via nu-
Downloaded from ascelibrary.org by Southeast University on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.
Experimental Verification
Test Procedure
To verify the accuracy of the proposed model, two small cables
consisting of 19 parallel wires were fabricated. The shape and di-
mensions of the test cable are illustrated in Fig. 16. First, the sam-
ples were exposed to temperatures of 500°C and 600°C and then
cooled to ambient air temperature. The entire heating process
was identical to that of the wire.
Finally, the fatigue tests were conducted. A hydraulic servocon-
Fig. 17. Fatigue test of cable.
trol testing machine (Fig. 17) with a dynamic capacity of 1,000 kN
was used to impose constant load ranges of 134.2 and 160.3 kN at
10 Hz under force control, which imply that the initial stress range
Table 5. Load ranges of cable fatigue tests (kN) of a single wire in the cable was 360 and 430 MPa, respectively
Exposure Load Maximum Minimum Mean (Table 5). The number of cycles for each wire break was manually
No. temperature range load load load recorded.
Cable 500°C 160.3 320.7 160.3 240.5
no. 1
Result and Comparison
Cable 600°C 134.2 268.5 134.2 201.3
no. 2 The number of cycles at which the wire broke was compared be-
tween the simulation and experimental results, as shown in
Fig. 18. The fatigue behavior of the cables obtained through the
where Nh = fatigue life of the hanger; and ΔL = load range. C de- simulation was consistent with the experimental results, demon-
notes the probability level, and D is a constant. The least-squares strating the accuracy of the proposed method. Alternatively, it
method was used to fit the mean hanger life under the four load was observed that after approximately half of the wires in the
ranges; the results are shown in Fig. 13. cable broke, the remaining wires simultaneously collapsed, which
Similarly, the life at any failure probability can be obtained. For may be because the maximum fatigue load exceeded the residual
example, the P–S–N curves at a 5% probability of failure based on a capacity of the cable. The fracture surface morphology obtained
5% wire breakage life are also shown in Fig. 13. by scanning electron microscopes also supports this conclusion.
Fig. 19 shows that approximately half of the photographs display
Effect of Number of Wires on Fatigue Life of Hanger the typical fatigue fracture characteristic without a ductile phenom-
Fig. 14 plots the mean life and the life at a 5% probability of failure enon at the broken parts; however, the other half show necking phe-
varies with the number of wires (also taking a hanger subject to a nomenon with some radiating fibers.
(a) (b)
Fig. 18. Comparison between simulation and test results: (a) Cable No. 1; and (b) Cable No. 2.
used to manage the censored data. Based on the fatigue life distri-
bution of the steel wires and the corresponding parallel system,
the fatigue cycles of the hanger subjected to a constant load
range were derived using Monte Carlo simulation and the order
statistics approach, and the S–N curves were obtained. Finally,
the proposed model was experimentally verified. The following
conclusions were drawn:
1. The fatigue properties of steel wires degrade when they are
subjected to elevated temperatures. When heated below 500°C,
all specimens could resist 2 million cyclic loads without frac-
ture, as required in the standard; however, when they were ex-
posed to temperatures exceeding 500°C, their fatigue lives
decreased significantly. Similar to the steel wire, the hanger
life decreased as the exposure temperature increased, and its re-
(a) sidual performance depended on the maximum temperature
reached. This indicates that a fire can reduce the service life
of a hanger.
2. The wire life was well modeled by the Weibull distribution, and
the hanger life was proven to obey a normal distribution based
on the parallel system.
3. The hanger life is far lower than the mean life of its constituent
wires owing to the effect of the growing stress range as individ-
ual wires break, which is consistent with the experimental ob-
servations. The hanger life is insensitive to the number of
wires in the hanger; however, its variability decreases as the
number of wires increases.
4. It is reasonable to adopt the number of cycles at a 5% probability
of wire failure as the hanger fatigue life at which 5% of the wires
break.
The present study can be used as a reference for the safety eval-
(b)
uation of cable-supported bridges after a fire.
Fatigue failure Stress rupture
Fig. 19. Fracture surface of cables: (a) Cable No. 1; and (b) Cable Data Availability Statement
No. 2.
All test data are available from the corresponding author by request.
Conclusion
Acknowledgments
The fatigue properties of hangers after fire exposure were investi-
gated in this study. Several fatigue tests were conducted on steel Support from the Natural Science Foundation of Jiangsu Province
wires after cooling from various elevated temperatures. A Weibull under Grant No. BK20210255 is gratefully acknowledged. Support
distribution was adopted to describe the fatigue life of the steel from the Innovation Program for Bridge Engineering Research
wire under specified stress ranges, and the EM algorithm was Center of Southeast University is gratefully acknowledged.
Dempster, A. P., N. M. Laird, and D. B. Rubin. 1977. “Maximum likeli- tem of parallel members.” SIAM J. Appl. Math. 34 (2): 227–246. https://
hood from incomplete data via the EM algorithm.” J. R. Stat. Soc. doi.org/10.1137/0134021.
39: 1–22. PTI (Post-Tension Institute) Cable-Stayed Bridge Committee. 2007.
Faber, M. H., S. Engelund, and R. Rackwitz. 2003. “Aspects of parallel Recommendations for stay cable design, testing and installation. 5th
wire cable reliability.” Struct. Saf. 25 (2): 201–225. https://doi.org/10 ed. PTI DC45.1-12. Farmington Hills, MI: PTI.
.1016/S0167-4730(02)00057-7. Quiel, S., T. Yokoyama, K. Mueller, L. Bregman, and S. Marjanishvili.
Freudenthal, A. M., and E. J. Gumbel. 1953. “On the statistical interpreta- 2015. “Mitigating the effects of a tanker truck fire on a cable-stayed
tion of fatigue tests.” Proc. R. Soc. London, Ser. A 216 (1126): bridge.” In Proc., 2nd Int. Conf., on Performance-Based and
309–332. Life-Cycle Structural Engineering, edited by D. Fernando, J.-G. Teng
Garlock, M., I. Paya-Zaforteza, V. Kodur, and L. Gu. 2012. “Fire hazard in and J. L. Torero, 1002–1012. St Lucia, Queensland, Australia: Univ.
bridges: Review, assessment and repair strategies.” Eng. Struct. 35: of Queensland.
89–98. https://doi.org/10.1016/j.engstruct.2011.11.002. Rackwitz, R., and M. H. Faber. 1991. “Reliability of parallel wire cable
Gong, X., and A. K. Agrawal. 2016. “Safety of cable-supported bridges under fatigue.” In Proc., Int. Conf., on Applications of Statistics &
during fire hazards.” J. Bridge Eng. 21 (4): 04015082. https://doi.org Probability in Civil Engineering, 166–175. San Francisco, CA:
/10.1061/(ASCE)BE.1943-5592.0000870. International Civil Engineering Risk and Reliability Association.
Jiang, J. H., A. B. Ma, W. F. Weng, G. H. Fu, Y. F. Zhang, G. G. Liu, and SAC (Standardization Administration of China). 2017. Hot rolled steel
F. M. Lu. 2009. “Corrosion fatigue performance of pre-split steel wires wire rod for prestressed steel wire and strand. GB/T 24238-2017.
for high strength bridge cables.” Fatigue Fract. Eng. Mater. Struct. Beijing: SAC.
32 (9): 769–779. https://doi.org/10.1111/j.1460-2695.2009.01384.x. SAC (Standardization Administration of China). 2018. Hot-extruded PE
Kodur, V., L. Gu, and M. E. M. Garlock. 2010. “Review and assessment of
protection paralleled high strength wire cable for cable-stayed bridge.
fire hazard in bridges.” Transp. Res. Rec. 2172: 23–29. https://doi.org
GB/T 18365-2018. Beijing: SAC.
/10.3141/2172-03.
SAC (Standardization Administration of China). 2019. Hot-dip zinc or
Lan, C. 2009. “Fatigue properties assessment theory of parallel wire cable.”
zinc-aluminum coated steel wire for bridge cables. GB/T
J. Shenyang Jianzhu Univ. 25 (1): 56–60.
17101-2019. Beijing: SAC.
Lan, C., D. Ren, Y. Xu, N. Li, and Z. Liu. 2017a. “Fatigue property assess-
Stallings, J. M., and K. H. Frank. 1991a. “Cyclic fatigue life of cables.”
ment of parallel wire stay cable II: Fatigue life model for stay cable.”
Eng. Fract. Mech. 38: 341–347. https://doi.org/10.1016/0013
Chin. Civ. Eng. J. 50 (7): 69–77.
Lan, C., Y. Xu, D. Ren, N. Li, and Z. Liu. 2017b. “Fatigue property assess- -7944(91)90013-Q.
ment of parallel wire stay cable I: Fatigue life model for wire.” Chin. Stallings, J. M., and K. H. Frank. 1991b. “Stay-cable fatigue behavior.”
Civ. Eng. J. 50 (6): 62–70. J. Struct. Eng. 117 (3): 936–950. https://doi.org/10.1061/(ASCE)0733
Li, H., C. M. Lan, Y. Ju, and D. S. Li. 2012. “Experimental and numerical -9445(1991)117:3(936).
study of the fatigue properties of corroded parallel wire cables.” Stigler, S. M. 1974. “Linear functions of order statistics with smooth weight
J. Bridge Eng. 17 (2): 211–220. https://doi.org/10.1061/(ASCE)BE functions.” Ann. Stat. 2 (4): 676–693. https://doi.org/10.1214/aos
.1943-5592.0000235. /1176342756.
Liu, Z., T. Guo, M. H. Hebdon, and Z. Zhang. 2019. “Measurement and Tao, Z. 2015. “Mechanical properties of prestressing steel after fire expo-
comparative study on movements of suspenders in long-span suspen- sure.” Mater. Struct. 48 (9): 3037–3047. https://doi.org/10.1617
sion bridges.” J. Bridge Eng. 24 (5): 04019026. https://doi.org/10 /s11527-014-0377-5.
.1061/(ASCE)BE.1943-5592.0001386. Tian, Z., J. Yang, and X. Zhong. 2011. “A method for reliability analysis of
Liu, Z., T. Guo, L. Huang, and Z. Pan. 2017. “Fatigue life evaluation on incomplete fatigue life based on an expectation maximization algo-
short suspenders of long-span suspension bridge with central clamps.” rithm.” J. Beijing Univ. Chem. Technol. 38 (3): 104–107.
J. Bridge Eng. 22 (10): 04017074. https://doi.org/10.1061/(ASCE)BE Weibull, W. 1951. “A statistical distribution function of wide applicabil-
.1943-5592.0001097. ity.” J. Appl. Mech. 18: 293–297. https://doi.org/10.1115/1.4010337.
Lu, J., H. Liu, and Z. Chen. 2017. “Post-fire mechanical properties of low- Zhang, Z., T. Guo, S. Wang, J. Liu, and L. Wang. 2021. “Experimental
relaxation hot-dip galvanized prestressed steel wires.” J. Constr. Steel study on post-fire properties of steel wires of bridge suspender.”
Res. 136: 110–127. https://doi.org/10.1016/j.jcsr.2017.05.012. Structures 33: 1252–1262. https://doi.org/10.1016/j.istruc.2021.04.099.
Matsukawa, A., M. Kamei, T. Mizoguchi, and Y. Sasaki. 1988. “Fatigue Zheng, W., Q. Hu, and H. Zhang. 2006. “Experimental research on the me-
resistance analysis of parallel wire strand cables based on statistical the- chanical property of prestressing steel wire during and after heating.”
ory of extremes.” Int. J. Fatigue 57 (7): 205–210. J. Build. Struct. 27 (2): 120–128.