Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Engineering Structures 272 (2022) 114967

Contents lists available at ScienceDirect

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

Fatigue reliability analysis considering corrosion effects and integrating


SHM information
Xu Han a, Dan M. Frangopol b, *
a
Dept. of Civil and Environmental Engineering, ATLSS Engineering Research Center, Lehigh Univ. 117 ATLSS Dr, Bethlehem, PA 18015, USA
b
Professor and the Fazlur R. Khan Endowed Chair of Structural Engineering and Architecture, Dept. of Civil and Environmental Engineering, ATLSS Engineering Research
Center, Lehigh Univ, 117 ATLSS Dr, Bethlehem, PA 18015, USA

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

Keywords: Fatigue cracking is a common problem facing steel bridges. S-N curves have been adopted in design specifications
Fatigue to offer guidance on fatigue design of these bridges. Structural health monitoring (SHM) techniques can provide
Corrosion reliable information on stress states at fatigue-prone details using strain sensors, thereby circumventing complex
Reliability
theoretical analyses. When corrosion is present, simply relying on SHM information obtained through the
SHM
Finite element analysis
monitoring period may not be sufficient to reliably estimate the damage incurred by corrosion-enhanced fatigue
as corrosion is a long-term process and the time-variant effect of corrosion on fatigue needs to be determined. In
this paper, time-variant analysis is used to estimate the reliability of fatigue-prone details at the coped region of a
transverse diaphragm in a tied-arch bridge. Stress histograms obtained based on SHM information associated
with this case study are utilized to establish the distribution of the original effective stress. Finite element
analysis is conducted for (a) determining the spatial adjustment factor (SAF) to adjust the stress range obtained at
the location of strain sensors, and (b) assessing the effect of corrosion on the effective stress range. The reliability
analysis results show that for the fatigue details investigated in this case study, corrosion has a significant effect
on the fatigue reliability and, therefore, should be considered in the life-cycle management of fatigue-prone
details.

1. Introduction failure under the same loading condition have been recorded. In addi­
tion, the stress-cycles at fatigue-prone details in steel bridges induced by
Fatigue is a primary concern for steel bridges, as the connections vehicles have variable amplitudes [4]. The number of cycles within a
between different steel plates are prone to develop fatigue cracks under certain period of time is also random. Given these uncertainties, prob­
stress cycles induced by traffic. Without proper management, fracture abilistic approach should be adopted to determine the safety of fatigue-
may occur at fatigue-prone details, causing severe failure consequences. prone details of interest. A series of fatigue-test projects have been
In this regard, it is necessary to estimate the fatigue strength of fatigue- launched by the National Cooperative Highway Research Program
prone details and the associated remaining service life in order to (NCHRP) in the 1970s. Regression analyses were performed on the test
perform timely interventions. According to [1], the fatigue process can results and the relationship between the mean of log S and log N was
be divided into three phases, namely, the crack initiation phase, the obtained, where S and N are the stress range and number of cycles to
crack-propagation phase, and the final fracture stage. Since the begin­ failure, respectively. The mean log S – log N curve is shifted below and
ning of the 19th century, research has been conducted to investigate all parallel by two standard deviations [5]. This shifted curve is adopted in
three phases. A significant portion of research work was experiment- design specifications, such as [3], to account for the safety margin in
oriented, which finally led to the creation of S-N curves. The theory of design. So far, some studies have taken the uncertainties associated with
S-N curves and cumulative damage criteria [2] based on S-N curves have the S-N curve in AASHTO into consideration and performed reliability
been incorporated in design specifications all over the world [3]. Most of analysis on fatigue-prone details in steel bridges. Zhao et al. [6] con­
cyclic load tests are carried out on small-scale specimens with a constant ducted fatigue reliability analysis on a full-penetration butt weld using
stress range amplitude, and large dispersions of the number of cycles to AASHTO S-N curves. Tobias and Foutch [7] carried out fatigue

* Corresponding author.
E-mail addresses: xuh216@lehigh.edu (X. Han), dan.frangopol@lehigh.edu (D.M. Frangopol).

https://doi.org/10.1016/j.engstruct.2022.114967
Received 3 December 2021; Received in revised form 2 August 2022; Accepted 11 September 2022
Available online 29 September 2022
0141-0296/© 2022 Elsevier Ltd. All rights reserved.
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

constant amplitude are calculated based on the stress range histograms,


which are integrated into the S-N approach to determine the remaining
fatigue life. Information gathered from SHM facilitates the reliability
analysis of fatigue-prone details in steel bridges. Liu et al. [23] con­
ducted fatigue reliability assessment on the coped region of the trans­
verse diaphragm of a tied arch bridge in Pittsburg, PA. Kwon and
Frangopol [24] estimated fatigue reliability by first fitting the stress
range histogram with several probabilistic distributions, and calculating
equivalent stress range based on the fitted distributions. Soliman et al.
[25] performed reliability assessment of fatigue-prone details in I-39
bridge in Wisconsin using bilinear S-N curves. So far, fatigue analysis of
steel structures utilizing SHM information is abundant. Many studies
applied SHM on orthotropic steel deck for which fatigue is a critical
Fig. 1. Crack at the coped region of the transverse diagram (adapted problem [26,27]. Ni et al. [28] obtained stress range histogram using
from [16]). SHM and employed finite mixture distribution to represent the stress
range histogram. It is worth noting that SHM techniques have also been
reliability analysis on railway bridges using S-N approach, in which the used for reinforced concrete bridges to detect corrosion loss of steel
stress range is obtained through realistic traffic simulation. Imam and reinforcements [29]. SHM has been conducted on world-renowned
his co-workers conducted several studies on the fatigue reliability of bridges to facilitate fatigue reliability analysis, such as Great Sea Belt
riveted connections in railway bridges [8–10]. Bridge [30] and Manhattan Bridge [31]. In addition, advanced data-
Fatigue-prone details are classified into several categories in bridge driven models, such as machine learning model, have been adopted to
design codes based on their fatigue resistance. One type of fatigue-prone conduct fatigue reliability analysis integrating SHM information [32].
detail that is of special interest is the connection between longitudinal However, research efforts on multi-hazard analysis utilizing SHM in­
main girders and the transverse diaphragm of steel bridges. Distortion- formation are insufficient.
induced fatigue is prone to occur at the gap between the connection Corrosion is another major hazard facing steel bridges. The interac­
plate for the diaphragm and the top flange of the longitudinal girder. For tion between corrosion and fatigue may accelerate the fatigue crack
multi-girder bridges, the relative longitudinal displacement between growth, thereby leading to further deterioration of the fatigue-prone
two adjacent girders can lead to out-of-plane bending of the diaphragm, detail and a reduction of fatigue life. When corrosion exists, loss of
which results in the fatigue cracking at the end of the diaphragm [11]. material near the fatigue-prone detail will engender a larger stress under
For a tied-arch bridge, the relative displacement between the longitu­ the same loading condition. The resultant larger stress range will inflict a
dinal tie girder and top flange of the transverse diaphragm can cause more severe damage on the fatigue-prone detail. Tragic accidents such
out-of-plane displacement of the diaphragm, facilitating the develop­ as the collapse of the Silver Bridge in 1967 [33] are bracing reminders of
ment of cracks [12]. It is estimated that out-of-plane distortion or other the impact of corrosion-enhanced fatigue on the safety of structures.
unanticipated secondary stresses at fatigue-sensitive details accounts for Reliability analysis on coupled corrosion-fatigue has been conducted on
nearly 90 % of fatigue cracking cases [13]. Different retrofit techniques ship structures [34,35]. For steel bridges, Hosseini et al. [36] and Sah­
[13,14] have been proposed to address this issue, including drilling rapeyma et al. [37] conducted fatigue reliability assessment on a truss
holes at the end of the existing crack to arrest it, providing a positive bridge considering the amplification of stress range associated with the
attachment between longitudinal girder and transverse diaphragm, reduced section of the truss members due to corrosion. The truss can be
among others. Among all the retrofit techniques, softening the connec­ considered as a group of axially-loaded members, which makes the in­
tion by cutting off part of its upper end is a preferable option, as it is crease of stress ranges due to corrosion easy to be accounted for. How­
proved to be both cost-effective and efficient [15]. However, cutting off ever, in many cases where the structure of interest is in a complex stress
the upper part of the end of diaphragm (the retrofitted diaphragm is state, no closed-form solutions can be used to account for the increase of
referred to as coped diaphragm) creates new fatigue-prone details at the stress range under corrosion. Adasooriya and Siriwardane [38] consid­
coped region. When the coped diaphragm is subjected to large in-plane/ ered time-variant stress ranges in fatigue-prone details on a six-span
out-of-plane bending stress, fatigue cracks may develop at the corner of riveted railway bridge. The time-variant load stress histograms were
the coped region [16] shown in Fig. 1. Therefore, it is necessary to generated through the combination of traffic simulation and finite
evaluate the safety with respect to fatigue of the coped region. element (FE) modeling. In that study, corrosion and fatigue are
Numerous studies have been carried out on the fatigue behavior of the considered in a deterministic manner. It should be noted that perfor­
girder-to-diaphragm connections, such as Fisher and Keating [17], mance of structures under corrosion and fatigue should be assessed from
Stallings et al. [18,19], and Fraser et al. [20], among others. a probabilistic perspective as both hazards are subjected to un­
The stress state at fatigue-prone details at the connection between certainties. Under normal operation conditions, the stresses resulting
longitudinal girders and transverse beams can be hard to determine from traffic loads are within the elastic range. Therefore, the stress range
theoretically, as both in-plane and out-of-plane movements are involved associated with a fatigue detail under corrosion can be considered as the
in generating stress at such details. Other factors such as imperfect ge­ product of the stress range without corrosion effects and a corrosion-
ometry and residual stress can also influence the stress distribution at enhanced factor.
the fatigue-prone detail. Information obtained through structural health Currently, there has been a paucity of multi-hazard structural anal­
monitoring (SHM) can provide time-series strain data to make an ac­ ysis utilizing SHM information, especially for structures subjected to
curate diagnosis and prognosis of the safety level of the fatigue-prone corrosion and fatigue. Most of the analysis in this regard has been
detail. SHM has found its applications in fatigue evaluation projects focused upon long-span suspension bridges [39], while the analysis
associated with steel bridges, such as the I-64 Kanawha River Bridge could be on many other types of bridges. Li et al. [40] used acoustic
[21]. In these projects, strain gages are mounted on the location near the emission sensors to obtain information on fatigue behavior of corroded
fatigue-prone detail of interest by which the time history of stress bridge cables. That study focused on the censoring techniques and the
associated with the fatigue-prone detail can be recorded. Algorithms effect of corrosion on fatigue stress range was not quantified. Ye et al.
such as rainflow counting method [22] are used to transfer the recorded [32] conducted corrosion fatigue life assessment of a suspension bridge
stress history into stress range histograms. Equivalent stress ranges of using SHM information, yet the influence of corrosion on effective fa­
tigue stress range was quantified using the cross-sectional area loss,

2
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

which is a simplistic approach. Investigation on fatigue reliability of 2.1.2. Equivalent stress range calculation based on SHM information
structures subjected to corrosion-enhanced fatigue utilizing SHM in­ When data are obtained through SHM, a smoothing window is usu­
formation is worth carrying out. This SHM information is more reliable ally applied to the original data to remove random noise spikes. A
than that obtained through theoretical analysis such as FE modeling. In depopulation process may come next in order to reduce the data volume
this paper, the case study of a tied-arch bridge is presented and analyzed. while maintaining the characteristics of the original data such as peaks
Due to a lack of SHM information on corrosion, a hybrid investigation and valleys. Cyclic counting methods such as rainflow counting are
approach on the fatigue reliability analysis of fatigue-prone details applied to the processed data to obtain the stress range histograms. The
subjected to corrosion-enhanced fatigue is adopted. In this approach, random-amplitude load spectrum is converted into an equivalent
information on fatigue (such as stress range histogram) is obtained constant-amplitude load spectrum through the equation
through SHM, while FE analysis is conducted to account for the corro­ [ ]m1
sion effects on fatigue. Corrosion-enhanced factors associated with ∑K
ni m
Sre = ⋅S (3)
several load cases and different boundary conditions are calculated to i=1
Ntotal ri
obtain the stress range histogram under corrosion. The calculated fa­
tigue reliability profile takes into account the effect of corrosion on fa­ where Sre is the equivalent effective stress range associated with the
tigue reliability. The hybrid approach proposed herein to conduct life- variable-amplitude stress histogram; K is the number of stress range bin;
cycle probabilistic performance evaluation of transverse diaphragm ni (i = 1⋯K) is the number of stress cycles falling within i th stress range
under complex stress states is innovative. The information obtained bin; Ntotal is the total number of stress cycles associated with a stress cut-
from the case study presented is useful for decision-making for optimal off level Scut ; Sri (i = 1⋯K) is the mean value of the i th stress range bin;
life-cycle management of steel bridges in corrosive environments. m is the fatigue material constant in Eq. (1).
During the process of obtaining equivalent stress range, a cutoff
2. Fatigue and corrosion modeling stress level should be established so that any stress range less than the
cut-off will be ignored, regardless of the algorithm used. Different cut-off
2.1. Fatigue modeling levels have been proposed. According to [44], under variable-amplitude
cyclic loading, the cutoff threshold is about 25 % CAFL (Constant
2.1.1. S-N approach Amplitude Fatigue Limit) for welded details that are expected to have a
Two different approaches are widely used to conduct fatigue finite fatigue life, as stress ranges below one quarter of CAFL make very
assessment, namely S-N approach and fracture mechanics approach. The little contribution to the fatigue damage. For fatigue-prone details that
linear S-N approach [41,42] is adopted in this paper. are expected to have an infinite fatigue life, the cutoff threshold is
In the S-N approach, fatigue strength of a structural detail is char­ defined based on the relationship between effective stress range and the
acterized in terms of the relationship between stress range and number number of stress cycles associated with different cutoff levels. It has been
of cycles to failure. This relationship is established based on the scat­ demonstrated that as the cut-off level decreases, the relationship be­
tered data from cyclic load tests, as a larger stress range is observed to tween effective stress range and corresponding number of cycles ap­
correspond to a smaller number of cycles to failure. In the design proaches the S-N curve asymptotcally in log–log plot. As long as the
specifications, an S-N curve is obtained by shifting the mean S-N curve cutoff level is consistent with the slope of the S-N curve, considering
two standard deviations lower and is associated with a 2.3 % probability additional stress cycles at a lower cut-off level does not improve the
of failure assuming normal life distribution of a fatigue-prone detail damage assessment [12].
under a given stress range [5]. To assess the fatigue life of a structural
detail deterministically, the mean S-N curve should be used [43].
The S-N curves are illustrated as sloping straight lines in logarithmic 2.2. Corrosion modeling
scale. The associated equation is
For carbon steel plates, corrosion will start when the coating breaks
( )1/m
A down. In this paper, a uniform thickness loss over the plate surface is
S= (1)
N adopted. The time-variant thickness of a steel plate can be expressed as
[45].
where S is the nominal fatigue stress range; A is fatigue resistance co­
d(t) = d0 − dcorr (t) = d0 − rcorr (t − t0 ) (4)
efficient associated with a specific fatigue category; N is the number of
stress cycles; and m is a material constant indicating the slope of the S-N
where t is the time of bridge in service (year); d(t) is the thickness of the
curve.
web at time t; dcorr (t) is the thickness loss due to corrosion at time t
Multiple damage criteria are used regarding the cumulative damage
(mm); t0 is the coating life (year); d0 is the initial thickness (mm); rcorr is
incurred upon the fatigue-prone details by cyclic stress of variable
the annual corrosion rate (mm/year). It is worth mentioning that
amplitude. Miner’s rule, which is the simplest and most widely used
nonlinear corrosion models such as d(t) = AtB (A, B are coefficients
criterion, is adopted in this paper. According to this rule, the damage
associated with the power law equation) are also adopted in many
fraction of a certain stress range level increases linearly with the number
studies such as [46]. For carbon steel plates, the value of B is close to 1
of applied stress cycles. The total damage inflicted upon a fatigue-prone
[47]. In addition, linear corrosion model may reflect the long-term
detail
corrosion loss in a more accurate manner [48]. Therefore, linear

k
ni corrosion model is adopted herein.
D= (2) In addition to the corrosion rate, the spatial corrosion pattern can
Ni
influence the loss of sectional properties. Two types of spatial corrosion
i=1

is the sum of the damage associated with all the applied stress ranges, pattern have been proposed for longitudinal girders [49]. The first
where ni is the number of the i th stress range level (i = 1⋯k) and Ni is corrosion pattern (Pattern I) is for the scenario where corrosion extends
number of cycles to failure at the i th stress range level. Failure is deemed from the top surface of the bottom plate all the way upwards to the
to occur when D = 1.0. In practice, D may be less than 1.0 due to un­ intersection between the top flange and the web. In the second corrosion
certainties of various sources [5]. pattern (Pattern II), corrosion starts from the top surface of the bottom
plate and extends to the lower quarter point of web height. The first
corrosion pattern occurs at the end of the longitudinal beam, while the
second corrosion pattern applies to mid-span part of the longitudinal

3
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

measurement error for loading; Sre is the effective stress range calculated
according to Eq. (3); N(t) is the accumulated number of stress cycles
from 0 to t; t is time (year); NS(t) is the product of Sm
re and N(t), which
characterizes the damage index (NS) inflicted by the stress cycles during
SHM period.
If all the random variables are assumed lognormal, the time-variant
reliability index can be expressed as [50].
λΔ + λA − (λe + ln(λNS(t) ))
β(t) = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (6)
ξ2Δ + ξ2A + ξ2e + ξ2NS(t)

where λy and ξy denote the mean value and standard deviation of ln(y).
When the influence of corrosion is taken into account, cumulative NS
during SHM period will become time-variant if SHM is conducted at
different times during life-cycle. In that case, time-variant relationship
between cumulative NS value and t should be obtained. Based on [15],
the cumulative value of NS during Tg years can be calculated as
∫ Tg
( ) 365
NS Tg = × NSSHM (t)⋅(1 + α)t dt (7)
TSHM 0

where NSSHM (t) is the time-variant NS value during the SHM period
assuming that the SHM is conducted at a lead time of year t; TSHM is the
SHM duration; α is the annual increase rate of traffic volume on the
Fig. 2. Spatial corrosion patterns (a) Pattern I (b) Pattern II.
bridge.

4. Case study
beam. Illustrations of the two corrosion patterns are shown in Fig. 2.
In this paper, the influence of corrosion on fatigue will be repre­
4.1. Basic information on the Birmingham bridge
sented by an increase of effective stress range with the corrosion lead
time (i.e., time since corrosion begins). Effective stress ranges corre­
The investigation of the time-variant corrosion-enhanced fatigue
sponding to multiple representative thickness losses are calculated
reliability analysis is conducted on the coped region of the transverse
through FE analysis. Regression analysis is carried out to express the
diaphragm of the tied-arch Birmingham bridge in Pittsburg, PA. The
effective stress range as a function of thickness loss.
transverse diaphragm is connected with tie girders through bolted
connections; and is subjected to traffic loads transferred by stringers on
3. Reliability analysis
top of it. The bridge was originally built in 1976. In early 2000s, fatigue
cracks were found at almost all the transverse diaphragms at the con­
Based on [3] and Miner’s rule, the performance function associated
nections to the tie-girder. The thickness of the top and bottom flanges at
with fatigue reliability is expressed as
two end segments of transverse diaphragm (about 7436 mm from the
m N(t) NS(t) end) is 34.925 mm. The length of the transverse diaphragm is 32169
g(t) = Δ − e⋅Sre ⋅ = Δ − e⋅ (5)
A A mm. The top and bottom flanges are thickened to 57.15 mm in the
middle segments. The thickness of the web of diaphragm is 12.7 mm. A
where Δ is Miner’s critical damage accumulation index; e is retrofit action was conducted to alleviate the fatigue problem of

Fig. 3. Elevation of the Birmingham bridge (dimensions are in mm).

4
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

Fig. 4. Illustrations of the transverse diaphragm (a) typical cross section and (b) geometry of the cutout region (dimensions are in mm) (adapted from [12]).

Fig. 5. Directions of stress components and the principal stress at the cutout corner.

transverse diaphragms. After the retrofit in 2003, the Lehigh ATLSS 19). The elevation of the bridge is shown in Fig. 3. The typical section
Engineering Research Center was tasked with conducting SHM on the and the zoomed-in view of the diaphragm at the connection region are
retrofitted connection details [12]. The retrofit made was a cutout on the shown in Fig. 4. The fatigue behavior of the corner of cutout is inves­
diaphragm at connections to the tie-girder, thereby reducing the stiff­ tigated herein (the channel of the associated strain gage is named as Cl-
ness of connections. A total of 32 strain gages were attached on the II). This fatigue-prone detail is located in the area of greatest stress at the
diaphragm near the north end of the bridge (referred to as floorbeam connections [12]. At this area, both the horizontal (denoted as σzz ) and

5
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

Fig. 6. Schematics of loading conditions: (a) Load case I; (b) Load case II; (c) Load case III (Note: P = 66.7kN).

vertical stress (denoted as σ yy ) are on the same order of magnitude as the in-plane and out-of-plane loading. A more accurate SAF distribution can
maximum principal stress. The direction of maximum principal stress is be obtained based on the time-series of SAF. However, time-series SAF is
about 45◦ , which is the direction of the Cl-II strain gauge. It has been only of value when the real traffic load data that generates the measured
verified through FE analysis that the value of the maximum principal time-variant strain data is available. Unfortunately, this is not the case
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
stress can be approximated as σ 2yy + σ2zz [12]. As the radius of the for the Birmingham bridge. It is also worth noting that conducting traffic
simulation integrating FE models can be computationally expensive. In
cutout has been saw cut and ground smooth, the fatigue-prone detail can this paper, SAF is obtained based on research reported in [23]. Three
be categorized into category A in [3]. This category has the highest fa­ loading conditions are considered to obtain SAF. Load case I (LC I)
tigue resistance coefficient. An illustration of the directions of stress consists in applying 2.54 mm (0.1 in.) out-of-plane displacement to the
components and maximum principal stress of the strain gauge at the top flange of the diaphragm at the transverse location of web stiffeners.
cutout corner is shown in Fig. 5. The amplitude of the out-of-plane displacement is based on the
SHM was conducted for 40 days on the Birmingham bridge. Through maximum recorded value during the SHM period. In Load cases II and
data collected at CL-II and other locations, it was determined that the III, a 66.7kN (15kips) wheel load is applied through the deck and lon­
response of channels of strain gages at the connections are a combina­ gitudinal stringers to the diaphragm on the exterior (LC II) and interior
tion of both in-plane and out-of-plane response. In-plane response makes (LC III) traffic lane of the bridge.
a major contribution to the stress-range cycles, while out-of-plane Schematics of the three loading conditions are shown in Fig. 6. Two
response either increases or decreases the total surface stress [12]. boundary conditions are considered for the transverse diaphragm. The
Through rainflow counting, the time history of stress cycles collected at first boundary condition (BC1) is a fixed boundary condition where both
Cl-II was converted to the stress histogram. The stress cycles of which the in-plane and out-of-plane displacement are constrained. In the second
range is below 3.45 MPa (0.5ksi) are ignored as no fatigue damage is boundary condition (BC2), in-plane rotation is allowed while the out-of-
made by those cycles [12]. Detailed information regarding the SHM plane displacement is constrained. The real boundary condition of the
process and the SHM data is provided in [12]. Liu et al. [23] pointed out transverse diaphragm is believed to be somewhere between these two
that the stress at the location of strain gage may not be the most critical boundary conditions. It should be noted that relaxing the in-plane
stress near the cutout region. Therefore, FE models should be built to rotational degree of freedom in BC2 has no effect on the structural
find the maximum stress locations near the strain gages. response to Load case I as out-of-plane displacement at the top flange
The ratio of the stress at the maximum stress location over the stress only induces out-of-plane rotational movement at the end of transverse
at the strain gage location is referred to as the spatial adjustment factor diaphragm. It is also worth noting that the boundary condition may be
(SAF). SAF is sensitive to loading conditions, which means that it is time- better characterized by building a global finite element model including
variant as the loading condition is changing constantly under moving the main girders.
traffic. A rigorous approach to calculate SAF is to conduct traffic simu­
lation on the FE model, thereby obtaining time series of SAF under both

6
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

Table 1
SAF factors.
Boundary condition Loading condition CL-II Maximum SAF
Principal stress (absolute value) in the corner of cutout
σyy (MPa) σzz (MPa) Maximum principal stress (absolute)

I Load case I 20.59 − 27.97 21.21 30.90 1.46


Load case II 6.67 17.26 19.95 41.47 2.08
Load case III 28.75 53.35 65.40 129.57 1.98
II Load case II − 9.94 − 10.83 − 15.93 − 30.49 1.91
Load case III − 10.49 − 11.31 − 16.69 − 31.94 1.91

Fig. 8. Determination of cutoff level.


Fig. 7. PDF of original and SAF-modified stress ranges.

4.2. SAF modification Table 2


Corrosion rate data of carbon steel plate in moderate marine environment.
FE models are built for the transverse diaphragm in ABAQUS 2017 Exposure condition Corrosion rates
using C3D20R elements [51]. A global seed size of 50 mm is assigned for (μm/year)
the geometry model. The seed size of arc near the cutout is decreased to Marine 36a
2 mm to ensure fine mesh. The FE model has more than 410,000 ele­ Marine (Roof of Washington Hotel by the shore of Limon Bay) 12b
ments. A mesh sensitivity analysis has been conducted to make sure that Kure Beach (250 m lot), North Carolina 21.7c
further mesh refinement will not improve the accuracy of the FE analysis Point Reyes, California 11.9c
Barcelona 20.4d
results significantly. Full connection between diaphragm and the angles
Cadiz 12.5d
of bolted connections is assumed herein (i.e., no slip or detachment Cabo Negro 45.3d
between the angle and the diaphragm). In the ABAQUS modelling pro­ Alicante (30 m lot) 48.3d
cess, the surfaces of the web of diaphragm and the angles in contact are Alicante (100 m lot) 8.4d
tied together. The algorithm associated with tie constraint ensures that Block Island (Rhode Island) 242.7e
Kure Beach (250 m lot), North Carolina 118.0e
no relative displacement between the two contacting surfaces exists 171.0f
even if the two surfaces are meshed using different mesh densities 142.1 g
(which is the case in this FE model). As the fatigue-prone detail of in­ 81.8 h
terest is some distance away from the angles, it is not necessary to build 71.9i
Point Reyes, California 76.8j
FE models of the bolted connections in a very detailed manner. SAF
Kure Beach (200 m lot), North Carolina 20.7 k
values associated with these representative load cases are related to the
maximum principal stress (absolute value), which plays a critical role in Note: (a) Based on [54]; (b) Based on [55]; (c) Based on [56]; (d) Based on [57];
magnifying the stress range recorded by strain gauges. SAF values (e) Based on [58]; (f) Based on Reference 9 in [59]; (g) Based on Reference 11 in
[59]; (h) Based on Reference 12 in [59]; (i) Based on [60]; (j) Based on [61]; (k)
associated with three load cases and two boundary conditions are given
Based on [62];
in Table 1. In probabilistic analysis, the SAF factor is represented by a
uniform distribution, with its lower bound and upper bound equal to
those of the SAF data in Table 1. A modification process given SAF value cutoff stress level is selected as 17.23 MPa (2.5ksi), which represents
is applied to the original stress histogram. Random samples are gener­ 10.4 % of CAFL of the category A (165 MPa). When the modification is
ated based on the histogram of the original stress ranges (assuming made on the stress range histogram, the cutoff stress level should also be
uniform distribution within each bin). Then random samples of SAF are updated. After comparing the curve of cutoff stress versus effective
generated based on the triangular distribution. The multiplication of number of cycles and the slope of S-N curves of category A (as shown in
two samples yields the modified stress range samples considering SAF. Fig. 8), the new cutoff stress level associated with SAF modification is
Histograms of the original samples and the SAF-modified samples are selected as 34.45 MPa (5.0ksi), which is about 20.8 % of CAFL of cate­
shown in Fig. 7. gory A.
The effective stress Sre is calculated using Eq. (3). Based on [12], the

7
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

Fig. 9. Stress contours near coped region (a) no corrosion (b) 50-year corrosion under spatial pattern I and Load case II (Note: In the figure legend, “S” stands for the
stress; 75% refers to the averaging threshold for nodal stress calculation).

4.3. Influence of corrosion on the effective stress range two spatial corrosion patterns are both considered in building FE models
involving corrosion thickness loss. At lead time t, the thickness loss is
When corrosion exists, the most critical stress at the fatigue-prone calculated using Eq. (4). The thickness of corroded steel plates will be
detail can be calculated as reduced in FE models. Load cases II and III are applied on FE models of
corroded diaphragm to obtain αcorr associated with in-loading condi­
Scr (t) = Scr,ori ⋅SAF⋅αcorr (t) (8)
tions. For out-of-plane loading, as the horizontal displacement should be
larger when corrosion occurs, Load case I cannot be used to obtain αcorr .
where Scr,ori is the original stress at the location of Cl-II obtained through
Therefore, a fourth load case (i.e., LC IV) is adopted to obtain αcorr in the
SHM data; SAF is the spatial adjustment factor; and αcorr (t) is the time-
case of out-of-plane loading (the out-of-plane movement at the bound­
variant corrosion-enhanced factor. For the fatigue detail investigated
ary is restrained in this case), in which the top flange at the cross section
herein, αcorr (t) can be represented by the ratio of the maximum principal
of web stiffener is subjected to a transverse force of 0.1kN.
stress (absolute value is used if the cutout region is in compression under
von Mises stress contours with an averaging threshold of 75 % for
external loads) at Cl-II in the FE model associated with corrosion lead
nodal stress under LC II and BC II are shown in Fig. 9 for the coped region
time t to the combined stress at Cl-II in the original FE model with no
of (a) the original FE model and (b) the FE model associated with 50
corrosion.
years of corrosion (Pattern I). It can be seen that due to corrosion
Corrosion rate of carbon steel plate is highly dependent upon the
thickness loss (the web thickness is reduced to 9.45 mm from 12.7 mm),
ambient atmosphere. In this paper, corrosion rates of carbon steel plates
stress value at the corner of cutout increases with the lead time of
in moderate marine environment are considered. Detailed information
corrosion, which indicates an acceleration of fatigue damage rate at the
on corrosion rate data is shown in Table 2. The mean corrosion rate of
fatigue-prone detail.
the data in Table 2 (0.067 mm/year) is used herein.
The stresses at Cl-II associated with corrosion lead time of 10, 20, 30,
Given the fact that the bridge has been in service for almost 30 years
40, and 50 years are calculated. The information on αcorr at those years
when the SHM was conducted, the remaining coating life of the dia­
associated with different load cases and spatial corrosion patterns is
phragm is not considered (i.e., t0 = 0 in Eq. (4)). The aforementioned

8
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

Table 3
Corrosion-enhanced factor.
Lead Spatial Boundary Loading Moderate marine
time corrosion condition condition
Maximum αcorr
(year) pattern (BC) (LC)
principal stress
(absolute value)
at Cl-II (MPa)

0 – BC I LC II 19.95 –
LC III 65.40 –
LC IV 12.03 –
BC II LC II − 15.93 –
LC III − 16.69 –
10 Pattern I BC I LC II 20.75 1.04
LC III 68.46 1.05
LC IV 13.70 1.14
BC II LC II − 17.02 1.07
LC III − 17.81 1.07
Patter II BC I LC II 20.33 1.02
LC III 66.54 1.02
LC IV 12.52 1.04
BC II LC II − 16.16 1.01
LC III − 16.89 1.01 Fig. 10. Time-variant profiles of corrosion-enhanced factor.
20 Pattern I BC I LC II 21.41 1.07
LC III 71.07 1.09
LC IV 15.31 1.27 SAF-modified stress range samples. As αcorr value associated with out-of-
BC II LC II − 17.02 1.07 plane loading (LC IV) is much larger than that associated with in-plane
LC III − 17.81 1.07 loading (LC II and LC III), the influence of corrosion on the damage index
Patter II BC I LC II 20.46 1.03
LC III 66.86 1.02
(NS) in these two loading scenarios are investigated separately. In the
LC IV 12.62 1.05 case of in-plane loading, the mean value of four αcorr values associated
BC II LC II − 16.25 1.02 with in-plane loading cases is adopted as the multiplication factor.
LC III − 16.94 1.02 Following the same methodology as that associated with Fig. 8, the
30 Pattern I BC I LC II 22.10 1.11
effective stress range (Seff ), daily number of cycles (Ndaily ), as well as the
LC III 73.88 1.13
LC IV 17.19 1.43 daily damage index (NSdaily ) associated with different cutoff levels
BC II LC II − 19.33 1.21 (assuming m = 3 in Eq. (1) [23]) are determined as shown in Table 4. It
LC III − 20.20 1.21 can be seen that under the same lead time and loading condition,
Patter II BC I LC II 20.61 1.03
corrosion Pattern I results in a larger effective stress Sre than corrosion
LC III 67.21 1.03
LC IV 12.72 1.06 Pattern II. For corrosion Pattern II, the time-variation of NSdaily is rela­
BC II LC II − 16.35 1.03 tively small as the associated corrosion-enhanced factor αcorr is very
LC III − 16.99 1.02 close to 1. It should be noted that traffic simulation can render a more
40 Pattern I BC I LC II 22.82 1.14
accurate distribution of αcorr . However, due to the unavailability of real
LC III 76.91 1.18
LC IV 19.40 1.61 traffic load data, the traffic simulation approach is not followed in this
BC II LC II − 20.71 1.30 case study.
LC III − 21.64 1.30 Uncertainty in NS plays an important role in calculating reliability
Patter II BC I LC II 20.76 1.04 index. For the curves showing the relationship between Seff and N, such
LC III 67.58 1.03
LC IV 12.83 1.07 as the two curves with markers in Fig. 8, when the cutoff stress range
BC II LC II − 16.45 1.03 threshold is lower than a certain value, the curves are approaching the S-
LC III − 17.04 1.02 N curve asymptotcally in log–log plot. The NS values associated with
50 Pattern I BC I LC II 23.57 1.18 these cutoff levels are used to determine the mean and COV of the NS. It
LC III 80.16 1.23
has been observed from the NS data in this study that the COV of NS is
LC IV 22.04 1.83
BC II LC II − 22.29 1.40 close to 0.25, regardless of the level of corrosion. Therefore, a COV of
LC III − 23.27 1.39 0.25 is chosen for NS.
Patter II BC I LC II 20.93 1.05 The relationship between mean damage index NS and t is quantified
LC III 67.99 1.04
through regression analysis. First, as both the effective stress range and
LC IV 12.95 1.08
BC II LC II − 16.56 1.04 number of effective stress cycles are obtained through the stress range
LC III − 17.097 1.02 histogram, the number of effective stress cycles during SHM can thereby
be obtained. NS can be considered as a function of corrosion thickness
loss dcorr , which is a function of t. Polynomial fitting analysis is con­
shown in Table 3. Time-variant profiles of αcorr are plotted in Fig. 10. As ducted to determine the relationship between these two variables. The
indicated under corrosion Pattern I, corrosion leads to a significant in­ relationship between NSdaily (daily NS value) and dcorr is fitted separately
crease of stress under the same loading condition. In the case of corro­ for each combination of the spatial corrosion pattern and loading sce­
sion Pattern II, the increase of stress under the same loading condition in nario (regarding whether it is in-plane or out-of-plane) as the ranges of
the presence of corrosion is not substantial. It can also be seen that the NSSHM in the four cases are different. The fitting results are shown in
influence of corrosion in the case of out-of-plane loading is more sig­ Fig. 11. Second-order polynomial fitting is adopted for the cases asso­
nificant than that associated with in-plane loading, as the corrosion- ciated with corrosion Pattern I while third-order polynomial regression
enhanced factor associated with the former is significantly larger than is conducted for the cases associated with corrosion Pattern II to ensure
that associated with the latter. Under the in-plane loading, corrosion the quality of fitting (i.e., achieving a low value of the root means square
exerts a larger influence under the fixed boundary condition (BC-I) error (RMSE)). For corrosion Pattern I, the regression model of NSdaily is
compared with the pinned boundary (BC-II).
expressed as
The obtained corrosion-enhanced factors are used to multiply the

9
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

Table 4
Effective stress and damage inflicted by stress cycles (SAF considered for corrosion cases).
Lead time Spatial corrosion Loading Cutoff stress Effective stress range Daily effective number of Damage inflicted by cyclic load per
(year) pattern condition (MPa) (MPa) cycles day (MPa3)

0 (SAF only) – – 34.45 49.13 137 1.62 × 107


10 Pattern I In-plane 31.01 47.14 208 2.18 × 107
Out-of-plane 37.90 54.89 149 2.47 × 107
Pattern II In-plane 34.45 49.50 142 1.73 × 107
Out-of-plane 34.45 50.00 152 1.89 × 107
20 Pattern I In-plane 34.45 50.64 165 2.14 × 107
Out-of-plane 41.34 60.42 159 3.52 × 107
Pattern II In-plane 34.45 49.60 144 1.76 × 107
Out-of-plane 34.45 50.18 155 1.96 × 107
30 Pattern I In-plane 41.34 58.21 127 2.50 × 107
Out-of-plane 48.23 69.40 144 4.81 × 107
Pattern II In-plane 34.45 49.71 146 1.80 × 107
Out-of-plane 34.45 50.32 158 2.02 × 107
40 Pattern I In-plane 41.34 59.58 145 3.07 × 107
Out-of-plane 48.23 72.71 199 7.66 × 107
Pattern II In-plane 34.45 49.84 148 1.84 × 107
Out-of-plane 34.45 50.47 162 2.08 × 107
50 Pattern I In-plane 41.34 60.94 169 3.82 × 107
Out-of-plane 55.12 82.88 196 1.12 × 108
Pattern II In-plane 34.45 49.96 151 1.88 × 107
Out-of-plane 34.45 50.70 166 2.16 × 107

( )
2
NSdaily = 11.24dcorr − 1.99dcorr + 27.47 × 106 (9a) coefficient (A) are 82.0 × 1012, 39.3 × 1012, and 14.4 × 1012MPa3 for
categories A, B, and C, respectively [6]. The COV of A is 0.45 [6]. Two
in the case of in-plane loading and different annual traffic increase rates, 0 % and 5 %, are considered [15].
( ) Time-variant fatigue reliability profiles associated with the corrosion
2
NSdaily = 1.82dcorr + 3.00dcorr + 25.89 × 106 (9b) Patterns I and II assuming an annual corrosion rate of 0.067 mm are
shown in Figs. 12 and 13. It can be seen that if corrosion Pattern I occurs,
in the case of out-of-plane loading.
reliability profiles associated with corrosion (i.e., Cases IV and V) are
For corrosion Pattern II, the regression model of NSdaily is expressed
much lower than those associated with non-corrosion (i.e., Cases I). The
as decrease of reliability profile is more severe when the increase of
( 3
NSdaily = 4.74dcorr 2
− 29.0dcorr
)
+ 65.3dcorr + 250 × 105 (9c) effective stress range is associated with out-of-plane loading actions.
From Fig. 12(a), it can be seen that the fatigue reliability profile asso­
in the case of in-plane loading and ciated with Case V is somewhere between those associated with Cases II
( ) and III when corrosion lead time is larger than 35 years. This indicates
3
NSdaily = 1.49dcorr 2
− 89.96dcorr + 23.66dcorr + 249.3 × 105 (9d) that the fatigue-prone detail should be downgraded to Category C when
Case V and corrosion Pattern I occur. If only in-plane loading is
in the case of out-of-plane loading. In Eq. 9, the NSdaily is in MPa3 and considered in conjunction with corrosion Pattern I (i.e., Case IV), the
dcorr is in mm. reliability index at year 50 is higher to that associated with the fatigue
resistance coefficient of category B (i.e., Case II). Therefore, under these
4.4. Reliability analysis results conditions, the fatigue-prone detail should be downgraded from cate­
gory A to B. From Fig. 12(b), it can be seen that when corrosion Pattern II
Reliability analysis is conducted based on Eqs. (6), 7, and 9. A total of occurs, the associated reliability profiles (Cases IV and V) are only
5 cases are considered in this paper. In the first three cases, NS value slightly lower than that of Case I, indicating that corrosion Pattern II has
associated with SAF-modified stress histogram is used to characterize only a minor effect in a typical moderate marine environment. Com­
the time-variant fatigue damage. In Case I, the fatigue resistance of fa­ parison between Fig. 12 and Fig. 13 indicates that the annual rate of
tigue category A in [3] is used to represent the corrosion resistance. The traffic volume increase exerts a significant impact on the fatigue reli­
adverse impact of corrosion on fatigue may be regarded as downgrading ability. Due to the high fatigue resistance of fatigue Category A, the
the fatigue category of the detail. In this sense, fatigue resistance coef­ fatigue detail is safe even when the most critical corrosion pattern and
ficient of fatigue categories B and C in [3] is adopted in Cases II and III, load condition occurs. However, in severe marine environment where
respectively. In Cases IV and V, time-variant NS value due to corrosion is the corrosion rate can be several times larger than that in moderate
used to conduct reliability analysis. αcorr associated with in-plane marine environment, maintenance may need to be performed due to a
loading is adopted in Case IV while αcorr associated with out-of-plane lower fatigue index profile. Maintenance may also need to be carried out
loading is adopted in Case V. Fatigue resistance of category A is adop­ if a high annual increase rate of traffic volume occurs (e.g., 10 %).
ted in Cases IV and V.
The daily NS value calculated using Eq. (7) multiplied by 365 is 5. Conclusions and future work
considered as the mean NS at year t. Lognormal distribution is assumed
for all the other random variables in Eq. (6) [6,15,52,53]. Mean and 5.1. Conclusions
COV of the critical damage accumulation index (Δ) are 1 and 0.3,
respectively [52]. The mean and COV of the error factor (e) are set as 1 This paper investigates the influence of corrosion on the fatigue-
and 0.03, respectively [53]. The mean values of the fatigue resistance prone details of steel bridges utilizing SHM information, with

10
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

Fig. 11. Relation between damage index and effective stress range.

emphasis on fatigue details at the coped region of the diaphragm-to- loading, as well as the real spatial corrosion pattern. Annual in­
girder connection of steel bridges. Both the spatial adjustment and crease of traffic volume may also make a big difference in fatigue
corrosion-enhanced factors are calculated using FE models. The reliability index for steel structures.
approach adopted in this paper can be used to investigate fatigue-prone 3. The influence of the connection rigidity on the SAF is small. When
details under complex stress states when the influence of corrosion corrosion Pattern II occurs, the influence of connection rigidity on
needs to be accounted for. The following conclusions are drawn: the corrosion-enhanced factor is insignificant. When corrosion
Pattern I occurs, a lower rigidity tends to increase the corrosion-
1. For the spatial adjustment factor (SAF), the range calculated using enhanced factor when the thickness loss is significant.
representative load cases is not large. Therefore, using the SAF range
associated with representative load cases will not generate a signif­ 5.2. Future work
icant error compared with fitting an SAF distribution considering the
entire load history consisting of large amount of real traffic data. It should be noted that distortion-induced fatigue cracking at the
2. When thickness loss occurs due to corrosion, for a transverse dia­ connection between the transverse diaphragm and main girder is an
phragm subjected to both in-plane and out-of-plane loading, out-of- extensive research field. The stress range at the fatigue-prone details
plane loading exerts a more severe impact on fatigue reliability than near the connection between the transverse diaphragm and main girder
in-plane loading. For the same loading condition, corrosion Pattern I is influenced by the geometry of the cutout, such as the height, width as
is much more critical than corrosion pattern II. It is beneficial to well as the fillet radius. The influence of these geometrical parameters
determine relative weights of out-of-plane loading and in-plane on the fatigue life of details in the connection region of transverse

11
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

Fig. 12. Fatigue reliability profiles associated with 0% annual traffic vol­
Fig. 13. Fatigue reliability profiles associated with 5% annual traffic vol­
ume increase.
ume increase.

diaphragms merits further investigation. In addition, to fully capture the


6. Data Availability Statement
influence of SAF and corrosion effects on the stress range obtained
through SHM, it is encouraged to install weight-in-motion (WIM) sys­
All the data, models, and codes that support the findings of this study
tems to measure the real traffic load that generates the time-series of
are available from the corresponding author upon reasonable request.
strain data. It is also worth pointing out that the corrosion model
adopted herein is a theoretical one. A more detailed characterization of
CRediT authorship contribution statement
the corrosion thickness loss at the diaphragm can be performed if SHM
data on the thickness loss are available. Finally, a steel bridge, or even a
Xu Han: Conceptualization, Methodology, Software, Validation,
steel member in a bridge is associated with multiple fatigue-prone de­
Data curation, Writing – original draft, Writing – review & editing,
tails. The proposed approach can be further enhanced by including the
Visualization. Dan M. Frangopol: Writing – review & editing, Super­
ability of analyzing steel bridges with multiple fatigue-prone details
vision, Project administration, Funding acquisition.
under corrosive environments using a time-variant system reliability-
based approach. For the system reliability analysis associated with
Declaration of Competing Interest
multiple fatigue-prone details, correlations of load effects and the re­
sistances among different fatigue-prone details are crucial. SHM activ­
The authors declare that they have no known competing financial
ities can help shed light on these correlations. Also, the proposed
interests or personal relationships that could have appeared to influence
approach can be extended to include the effects of other hazards such as
the work reported in this paper.
earthquakes [63,64] on the life-cycle safety of corroded bridges with
fatigue-prone details.

12
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

Data availability Eng Mech 2011;26(3):471–80. https://doi.org/10.1016/j.


probengmech.2011.02.001.
[25] Soliman M, Frangopol DM, Kown K. Fatigue assessment and service life prediction
Data will be made available on request. of existing steel bridges by integrating SHM into a probabilistic bilinear S-N
approach. J Struct Eng 2013;139(10):1728–40. https://doi.org/10.1061/(ASCE)
Acknowledgments ST.1943-541X.0000584.
[26] Liu Y, Zhang H, Liu Y, Deng Y, Jiang N, Lu N. Fatigue reliability assessment for
orthotropic steel deck details under traffic flow and temperature loading. Eng Fail
The authors are grateful for the financial support received from the Anal 2017;71:179–94. https://doi.org/10.1016/j.engfailanal.2016.11.007.
US National Science Foundation Grant CMMI 1537926. The authors [27] Yu S, Ou J. Fatigue life prediction for orthotropic steel deck details with a
nonlinear accumulative damage model under pavement temperature and traffic
would like to thank Dr. David Y. Yang, Portland State University, for his loading. Eng Fail Anal 2021;126:105366. https://doi.org/10.1016/j.
help in the conceptualization of research ideas in this study. The authors engfailanal.2021.105366.
would also like to thank Ian C. Hodgson, Lehigh University, for [28] Ni YQ, Ye XW, Ko JM. Monitoring-based fatigue reliability Assessment of steel
bridges: Analytical model and application. J Struct Eng 2010;136(12):1563–73.
providing information of the geometry of the Birmingham bridge. The https://doi.org/10.1061/(asce)st.1943-541x.0000250.
opinions and conclusions presented this paper are those of the authors [29] Heitner B, OBrien EJ, Yalamas T, Schoefs F, Leahy C, Décatoire R. Updating
and do not necessarily reflect the views of the sponsoring organization. probabilities of bridge reinforcement corrosion using health monitoring data. Eng
Struct 2019;190:41–51. https://doi.org/10.1016/j.engstruct.2019.03.103.
[30] Farreras-Alcover I, Chryssanthopoulos MK, Andersen JE. Data-based models for
References fatigue reliability of orthotropic steel bridge decks based on temperature, traffic
and strain monitoring. Int J Fatigue 2017;95:104–19. https://doi.org/10.1016/j.
[1] Committee ASCE. Fatigue reliability: Introduction. J Struct Div 1982;108:3–23. ijfatigue.2016.09.019.
[2] Miner MA. Cumulative fatigue damage. J Appl Mech 1945;12:A159–64. [31] Tochaei EN, Fang Z, Taylor T, Babanajad S, Ansari F. Structural monitoring and
[3] AASHTO. AASHTO LRFD Bridge Design Specifications. Washington, D.C.: remaining fatigue life estimation of typical welded crack details in the Manhattan
American Association of State Highway and Transportation Officials (AASHTO); Bridge. Eng Struct 2021;231:111760. https://doi.org/10.1016/j.
2017. engstruct.2020.111760.
[4] Fisher JW, Mertz DR, Zhong A. Steel Bridge members under Variable Amplitude [32] Ye XW, Liu T, Ni YQ. Probabilistic corrosion fatigue life assessment of a suspension
Long Life Fatigue Loading - NCHRP Report 267. Washington DC: Transportation bridge instrumented with long-term structural health monitoring system. Adv
Research Board; 1983. Struct Eng 2017;20(5):674–81. https://doi.org/10.1177/1369433217698345.
[5] Fisher JW, Kulak GL, Smith IFC. A Fatigue Primer for Structural Engineers. [33] Lichtenstein AG. The Silver Bridge collapse recounted. J Perform Constr Facil
Chicago, Illinois: National Steel Bridge Alliance, American Institute of Steel 1993;7(4):249–61. https://doi.org/10.1061/(ASCE)0887-3828(1993)7:4(249).
Construction; 1998. [34] Moan T, Ayala-Uraga E. Reliability-based assessment of deteriorating ship
[6] Zhao Z, Haldar A, Breen FL. Fatigue-reliability evaluation of steel bridges. J Struct structures operating in multiple sea loading climates. Reliab Eng Syst Saf 2008;93
Eng 1994;120(5):1608–23. (3):433–46. https://doi.org/10.1016/j.ress.2006.12.008.
[7] Tobias DH, Foutch DA. Reliability-based method for fatigue evaluation of railway [35] Han X, Yang DY, Frangopol DM. Probabilistic life-cycle management framework
bridges. J Bridg Eng 1997;2(2):53–60. https://doi.org/10.1061/(ASCE)1084-0702 for ship structures subjected to coupled corrosion–fatigue deterioration processes.
(1997)2:2(53). J Struct Eng 2019;145:04019116. https://doi.org/10.1061/(ASCE)ST.1943-
[8] Imam BM, Righiniotis TD, Chryssanthopoulos MK. Probabilistic fatigue load 541X.0002406.
spectra for riveted railway bridges. Proc 10th Int Conf Appl Stat Probab Civ Eng, [36] Hosseini A, Sahrapeyma A, Marefat M-S. A reliability-based methodology for
New York: Taylor & Francis; 2007. considering corrosion effects on fatigue deterioration in steel bridges - Part I:
[9] Righiniotis TD, Imam BM, Chryssanthopoulos MK. Fatigue analysis of riveted Methodology. Int J Steel Struct 2013;13(4):645–56. https://doi.org/10.1007/
railway bridge connections using the theory of critical distances. Eng Struct 2008; s13296-013-4006-x.
30(10):2707–15. https://doi.org/10.1016/j.engstruct.2008.03.005. [37] Sahrapeyma A, Marefat M-S, Hosseini A. A reliability-based methodology for
[10] Imam BM, Righiniotis TD. Fatigue evaluation of riveted railway bridges through considering corrosion effects on fatigue deterioration in steel bridges - Part II: Case
global and local analysis. J Constr Steel Res 2010;66(11):1411–21. https://doi. study of Neka Bridge. Int J Steel Struct 2013;13(4):657–70. https://doi.org/
org/10.1016/j.jcsr.2010.04.015. 10.1007/s13296-013-4007-9.
[11] Cousins TE, Stallings JM, Lower DA, Stafford TE. Field evaluation of fatigue [38] Adasooriya ND, Siriwardane SC. Remaining fatigue life estimation of corroded
cracking in diaphragm-girder connections. J Perform Constr Facil 1998;12(1): bridge members. Fatigue Fract Eng Mater Struct 2014;37(6):603–22. https://doi.
25–32. org/10.1111/ffe.12144.
[12] Connor RJ, Fisher JW, Hodgson IC, Bowman CA. ATLSS Report No. 04-04: Results [39] Mashayekhi M, Santini-Bell E, Eftekhar Azam S. Fatigue crack detection in welded
of Field Monitoring Prototype Floorbeam Connection Retrofit Details on the structural components of steel bridges using artificial neural network. J Civ Struct
Birmingham Bridge. Bethlehem, PA: Lehigh University; 2004. Heal Monit 2021;11(4):931–47. https://doi.org/10.1007/s13349-021-00488-7.
[13] Connor RJ, Fisher JW. Identifying effective and ineffective retrofits for distortion [40] Li D, Ou J, Lan C, Li H. Monitoring and failure analysis of corroded bridge cables
fatigue cracking in steel bridges using field instrumention. J Bridg Eng 2006;11: under fatigue loading using acoustic emission sensors. Sensors 2012;12:3901–15.
745–52. https://doi.org/10.1061/(ASCE)1084-0702(2006)11. https://doi.org/10.3390/s120403901.
[14] Dexter RJ, Ocel JM. Manual for Repair and Retrofit of Fatigue Cracks in Steel [41] Wöhler A. Über die Festigkeitsversuche mit Eisen und Stahl. Berlin: Ernst und Korn;
Bridges. McLean, VA: Federal Highway Administration (FHWA); 2013. 1870.
[15] Liu M, Frangopol DM, Kwon K. Optimization of retrofitting distortion-induced [42] Basquin OH. The exponential law of endurance tests. Proc-Am Soc Test Mater
fatigue cracking of steel bridges using monitored data under uncertainty. Eng 1910;10:625–30.
Struct 2010;32(11):3467–77. https://doi.org/10.1016/j.engstruct.2010.07.016. [43] Kwon K. Reliability Assessment, Performance Prediction and Life-Cycle
[16] Roeder CW, MacRae GA, Kalogiros AY, Leland A. Fatigue cracking of riveted, Management of Fatigue Sensitive Structures Based on Field Test Data. Bethlehem,
coped, stringer-to-floorbeam connections. Olympia, Washington: Washington PA: Lehigh University; 2011.
Department of Transportation; 2001. https://wsdot.wa.gov/research/reports/fu [44] Fisher JW, Nussbaumer A, Keating PB. ATLSS Report No. 92-04: Resistance of
llreports/494.1.pdf. Welded Details under Variable Amplitude Long-life Fatigue Loading. Bethlehem,
[17] Fisher JW, Keating PB. Distortion-induced fatigue cracking of bridge details with PA: Lehigh University; 1993.
web gaps. J Constr Steel Res 1989;12(3-4):215–28. [45] Southwell CR, Bultman JD, Hummer JC. In: Estimating of service life of steel in
[18] Stallings M, Cousins TE, Christopher BP. Evaluation of Fatigue Cracking. I-65 seawater. Seawater Corros Handb. New Jersey: Noyes Data Corporation; 1979.
Mobile Delta Crossing Bridges-Volume II. Montgomery, Alabama: The State of p. 87–374.
Alabama. Highway Department; 1993. [46] Kayser JR, Nowak AS. Reliability of corroded steel girder bridges. Struct Saf 1989;
[19] Stallings M, Cousins TE, Rotto RK, Reid CB. Evaluation of Fatigue Cracking in I-65 6:53–63.
Mobile Delta Crossing Bridges-Volume I. Montgomery, Alabama: The State of [47] Díaz I, Cano H, Chico B, De La Fuente D, Morcillo M. Some clarifications regarding
Alabama Highway Department; 1993. literature on atmospheric corrosion of weathering steels. Int J Corros 2012.
[20] Fraser REK, Grondin GY, Kulak GL. Behaviour of Distortion-induced Fatigue Cracks https://doi.org/10.1155/2012/812192.
in Bridge Girders. Alberta, Canada; 2000. [48] Melchers RE. Probabilistic models for corrosion in structural reliability assessment
[21] Connor RJ, Fisher JW. ATLSS Report No. 01-14: Report on field measurements and - Part 1: Empirical models. J Offshore Mech Arct Eng 2003;125:264. https://doi.
assessment of the I-64 Kanawha River bridge at Dunbar, West Virginia. Bethlehem, org/10.1115/1.1600467.
PA: Lehigh University; 2001. [49] Kayser JR. The Effects of Corrosion on the Reliability of Steel Girder Bridges. Ann
[22] Downing S, Socie D. Simple rainflow counting algorithms. Int J Fatigue 1982;4(1): Arbor, Michigan: University of Michigan; 1988.
31–40. https://doi.org/10.1016/0142-1123(82)90018-4. [50] Kwon K, Frangopol DM. Bridge fatigue reliability assessment using probability
[23] Liu M, Frangopol DM, Kwon K. Fatigue reliability assessment of retrofitted steel density functions of equivalent stress range based on field monitoring data. Int J
bridges integrating monitored data. Struct Saf 2010;32(1):77–89. https://doi.org/ Fatigue 2010;32:1221–32. https://doi.org/10.1016/j.ijfatigue.2010.01.002.
10.1016/j.strusafe.2009.08.003. [51] Dassault Systems. ABAQUS 2017 User’s Manual. Providence, RI; 2016.
[24] Kwon K, Frangopol DM. Bridge fatigue assessment and management using [52] Wirsching PH. Fatigue reliability for offshore structures. J Struct Eng 1984;110:
reliability-based crack growth and probability of detection models. Probabilistic 2340–56. https://doi.org/10.1061/(ASCE)0733-9445(1984)110:10(2340).

13
X. Han and D.M. Frangopol Engineering Structures 272 (2022) 114967

[53] Frangopol DM, Strauss A, Kim S. Bridge reliability assessment based on monitoring. [60] ASTM Committee B-3. Report of subgroup of subcommittee VII on corrosiveness of
J Bridg Eng 2008;13:258–70. https://doi.org/10.1061/(ASCE)1084-0702(2008) various atmospheric test sites as measured by specimens of steel and zinc. Proc.
13. ASTM 59, West Conshohocken, PA: American Society for Testing and Materials
[54] Kucera V, Mattsson EM. Atmospheric corrosion of Bimetallic Structures. Atmos. (ASTM); 1959, p. 183–201.
Corros., Hoboken, NJ: Wiley; 1982. [61] Melchers RE. Transition from marine immersion to coastal atmospheric corrosion
[55] Southwell CR, Bultman JD. Corrosion of Metals in Tropical Environments-Final for structural steels. Corrosion 2007;63:500–14. https://doi.org/10.5006/
Report of 16-year Exposure. Washington D.C.: Naval Research laboratory; 1975. 1.3278401.
[56] Knotkova D, Kreislova K, Dean, Jr. SW. International Atmospheric Exposure [62] FHWA. Improved Corrosion-Resistant Steel for Highway Bridge Construction
Program: Summary of Results. West Conshohocken, PA: American Society for (FHWA-HRT-11-061). Washington, D.C.: Federal Highway Administration; 2011.
Testing and Materials (ASTM); 2012. [63] Decò A, Frangopol DM. Life-cycle risk assessment of spatially distributed aging
[57] Morcillo M, Simancas J, Feliu S. Long-term atmospheric corrosion in Spain: Results bridges under seismic and traffic hazards. Earthq Spectra 2013;29(1):127–53.
after 13-16 years of exposure and comparision with worldwide data. Atmos. https://doi.org/10.1193/1.4000094.
Corros. ASTM STP 1239, Philadelphia, PA: American Society for Testing and [64] Akiyama M, Frangopol DM, Ishibashi H. Toward life-cycle reliability-, risk-, and
Materials (ASTM); 1995, p. 195–214. resilience-based design and assessment of bridges and bridge networks under
[58] Copson HR. Long-time atmospheric coorosion tests on low-alloy steels. ASTM independent and interacting hazards: Emphasis on earthquake, tsunami and
Proc., West Conshohocken, PA: American Society for Testing and Materials corrosion. Struct Infrastruct Eng 2020;16(1):26–50. https://doi.org/10.1080/
(ASTM); 1960, p. 650–66. 15732479.2019.1604770.
[59] Albrecht P, Hall Jr TT. Atmospheric corrosion resistance of structural steels.
J Mater Civ Eng 2003;15:2–24. https://doi.org/10.1061/(ASCE)0899-1561(2003)
15.

14

You might also like