Journal of Constructional Steel Research 215 (2024) 108561

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

journal homepage: www.elsevier.com/locate/jcsr

Influence of corrosion on ultra-low cycle fatigue performance of steel

butt-welded joints with various welding methods
Yuelin Zhang a, b, Wei Wang a, b, Hongwei Huo c, Yang Wang c, Cheng Fang a, b, *
a
State Key Laboratory of Disaster Prevention in Civil Engineering, Tongji University, Shanghai 200092, China
b
College of Civil Engineering, Tongji University, Shanghai 200092, China
c
China 22MCC Group Corporation Ltd., Tangshan, Hebei 064000, China

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

Keywords: Preferential corrosion of welded joints in steel structures has been observed in a wide range of scenarios and
Welding method should receive enough attention. The current work focuses on the corrosion and ultra-low cycle fatigue (ULCF)
Accelerated corrosion behavior of low-alloy steel weld joints. Firstly, 1600 h neutral salt spray (NSS) tests were conducted on welded
Cyclic plasticity
joints made by different welding methods, and then topography scanning of corrosion pits was carried out. It was
Ultra-low cycle fatigue (ULCF)
Earthquake
found that the welded region is more susceptible to corrosion. Dynamic polarization curves of the welded joints
Butt-welded joints were measured, and the results show that the self-corrosion potential of the welded region is lower than that of
the base metal (BM) region, while the corrosion current density is higher than that of the BM region. Subse­
quently, cyclic tests were carried out on the corroded weld joints. The results show that corrosion can cause a
maximum reduction of 11.79% in yield load and 20.87% in ultimate load of the welded joints. Corrosion has
slight influence on the energy dissipation capacity of welded joints, but the maximum reductions in ULCF life
could reach 77.31% due to corrosion. Fractography analysis further shows that secondary cracks, map cracks,
and intergranular brittle fracture, apart from more common fatigue fracture, ductile fracture, and quasi-cleavage
fracture, appear on the fracture surfaces. Supplementary numerical simulation was conducted to enable a better
understanding of the failure processes. In the simulation, a geometric modeling method using topography
scanning data of corrosion surface was proposed. By using the proposed modeling approach, together with
Chaboche cyclic plasticity constitutive model and cyclic void growth model (CVGM) for capturing fracture
behavior, the predicted results show an acceptable agreement with the experimental ones.

1. Introduction fatigue issues of welded joints are also critical. For example, ultra-low
cycle fatigue (ULCF) factures of welded joints were widely observed
Steel structures (members) may not keep their original mechanical during the 1994 Northridge earthquake [7] and the 1995 Hyogo­
properties during the life-cycle stage owing to the ageing effect, which ken–Nanbu (Kobe, Japan) earthquake [8]. The SAC Steel Project was
can potentially affect the failure modes, especially for structures in then initiated focusing on developing guidelines for inspection, evalu­
coastal or marine environments [1]. The corrosion damage of steel ation and repair of existing and new steel moment resisting frame
structures increases the surface roughness and produces pitting pits with buildings [9], many parallel researches have been done in this field
different depths. The influence of corrosion on seismic performance of [10–14]. However, these studies have mainly focused on the behavior of
steel structures (members) has been studied by several researchers. For structural members, while the basic ULCF mechanism at material level
example, Apostolopoulos and Papadopoulos [2] reported that for rein­ [15–19] is insufficiently investigated. From a life-cycle performance
forcing steel bars S400, a mass loss of <2% causes a 22% reduction in the point of view, the problem of weldment corrosion is also non-negligible.
number of cycles to failure, while a 3% mass loss corresponds to a 47% Fig. 1(a) shows a case of preferential corrosion in heat affected zone
reduction in the number of cycles to failure. For Q345 mild steel plates, a (HAZ) of a weldment [20], and this phenomenon has been observed in a
5.73% mass loss could reduce the LCF life by 80% [3]. It should be noted wide range of coastal environments. For example, after 16 years' service
that steel members are usually connected by welding [4–6], and the of the Third Nanjing Yangtze River Bridge (the first long span bridge

* Corresponding author at: State Key Laboratory of Disaster Prevention in Civil Engineering, Tongji University, Shanghai 200092, China.
E-mail address: [email protected] (C. Fang).

https://doi.org/10.1016/j.jcsr.2024.108561
Received 2 December 2023; Received in revised form 22 January 2024; Accepted 18 February 2024
Available online 22 February 2024
0143-974X/© 2024 Elsevier Ltd. All rights reserved.
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

with steel towers in China), extensive corrosion damage was found on corrosion performance of Q345 steel welded joint, where pulsed tung­
the fillet weld in the crossbeam of the steel tower during the bridge sten inert gas (P-TIG) welding was used for priming, while submerged
inspection, as shown in Fig. 2(b) [21]. arc welding (SAW) was used for covering. After being placed in a hy­
There could be two main reasons leading to preferential corrosion of drofluoric acid environment for 100 h, the weld with only priming had
welded joints. Firstly, due to the unevenness of the weld joints' shape, no obvious corrosion, but many corrosion pits occurred on the SAW
the service condition of anti-corrosion coating at the joint region is covered joint since the microstructures of the weld changed after SAW
worse than that at the base metal (BM) region. Secondly, a weld joint covering. It is evident that preferential corrosion at welded areas are
often has stress concentration. Under service loads such as wind and influenced by several factors: the first factor is the inherent properties,
traffic induced vibrations, the coating for the welded material (WM) such as the microstructures of the materials [23,25] and the potentials of
tends to fall off, since the ductility of coatings are often worse than that different areas. The second one is welding technology parameters, such
of metals (Fig. 1(c)). Therefore, studying the corrosion of weldments has as welding current, voltage, joint configuration, pre-heating/post-
deep practical values. On the other hand, the material components of a heating temperature, ambient temperature, humidity, and even the
welded region are much more complex than that of the BM region, as wind velocity in-situ. The third one is the characteristic of the corrosive
BM, WM, and HAZ are all involved in a welded region. As a result, the environment, e.g., general atmospheric environment, industrial atmo­
corrosion characteristic of weld region is much more complex than that spheric environment, or marine atmospheric environment.
of the BM region. From the literature review, it can be found that previous works have
In 2015, Ma et al. [22] studied the corrosion characteristic of E690 been devoted to corrosion fatigue, weldment fatigue, and weldment
steel welded-joint under simulated marine atmospheric environment. corrosion. However, there are very limited studies on the (ultra) low
The results indicate that inter-critical heat affected zone (ICHAZ) is cycle performance of welded joints subjected to corrosion. The insuffi­
more susceptible to stress corrosion, due to lower strength, higher cient research data hinders a comprehensive understanding of the long-
negative potential and greater current density of this area. Song et al. service behavior of steel connections during an earthquake. The current
[23] investigated the corrosion characteristic of G390NH steel welded work focuses on the corrosion and ULCF behavior of low-alloy steel
joint under simulated industrial atmospheric environment. The results welded joints. Firstly, dog-bone flat-plate specimens were designed and
show that during the incipient stage of corrosion process, the main factor fabricated by using different welding methods, i.e., shielded metal arc
that influences corrosion is microstructure defect, while with the welding (SMAW), gas metal arc welding (GMAW), and submerged arc
development of corrosion process, the influence of potential difference welding (SAW), and then neutral salt spray tests were conducted on the
begins to exceed that of microstructure and becomes the most critical weld joints. After 1600 h of accelerated corrosion (corresponds to the
factor that governs the corrosion process. According to the values of corrosion of carbon steel for nearly 10 years in ISO C2 atmospheric
potential, the preferential sequence that corrosion happens is WM → BM environment, and for nearly 5 years in ISO C3 atmospheric environment
→ HAZ. Li et al. [24] performed corrosion fatigue tests on Q690qE steel [26]), topography scanning of corrosion pits was carried out. To reveal
weldment under simulated marine environment. Cross sectional the reason behind preferential corrosion, dynamic polarization curves of
topography observation results imply that groove corrosion occurs at the the weld joints were measured. Subsequently, cyclic tests were carried
coarse-grained heat affected zone (CGHAZ). The reason is that the out on the corroded joints, and the influence of accelerated corrosion on
content of Ni is low in this area, which leads to a Volta potential the yield load, peak load, energy dissipation capacity, as well as ULCF
discrepancy between the CGHAZ and WM. Gu et al. [25] studied the life of the weld joints is discussed. Fractography analysis was then

Fig. 1. Cases of preferential corrosion in welded zones.

Fig. 2. Geometry of the welded plate and weldment configuration (Unit: mm).

2
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

conducted to find out the fracture mechanism of the welded-joints under welding method, in which the melted welding rod or wire is used as
large strain amplitude cyclic loads. Finally, in-depth numerical simula­ filling material. The weld rod/wire and surface of the base metal are
tion was conducted to enable a better understanding of the failure melted through high temperature of electric arc and melted welding
processes. In the simulation, a geometric modeling method that con­ powder. SAW is applicable to weld members with large scale, and has
siders distribution randomness of corrosion pits by using topography high-quality welding characteristics. It is noted that the SAW welding
scanning data of corrosion pits was used. As for material property, method is often realized with no grooves and the BMs can be directly
Chaboche cyclic plasticity constitutive model and CVGM fracture model melted and reconnected by higher welding current and heat input,
were applied. whereas the other cases need grooves in which the weld materials are
filled. The joint configurations of the welded plates with different
2. Design and fabrication of specimens welding methods are shown in Fig. 2. The welding materials together
with welding parameters are listed in Table 1. As described by Fig. 2 and
Three different welding methods, namely, shielded metal arc weld­ Table 1, the butt joints in this study were produced by multi-layers
ing (SMAW), gas metal arc welding (GMAW), and submerged arc welds, and sufficient dwelling time (>1 min) was ensured in the study
welding (SAW), were used to make welded butt-joint plates. The three [28,29]. The chemical compositions of the base metal (BM) and weld
welding methods are widely used in engineering applications, and are materials (WMs) are summarized in Table 2. More details about the
recommended by Chinese code for welding of steel structures [27]. The welding technology parameters can be found in the authors' previous
three welding methods have different merits and application scenarios. work [6].
For example, the main advantage of SMAW is cost saving. GMAW is a After welding, the surface of the weld region was cleaned and grin­
common semi-automatic welding method that harnesses fusible welding ded, and then four dog-bone flat-plate specimens were cut off from each
wires as filling material. GMAW is suitable for various materials such as welded plate (Fig. 3 (a)). Due to the existence of welding deformation,
steel and aluminium. The characteristics of GMAW include high welding the specimens were flattened by a hydraulic machine (Fig. 3 (b)) to
speed and good welding quality. SAW is a widely-used automatic facilitate subsequent corrosion pits topography measuring and cyclic
tests. The dog-bone flat-plate coupons ready for test are shown in Fig. 3
Table 1
(c).
Welding process parameters.
Welding method SMAW GMAW SAW 3. Accelerated corrosion experiment
Pass Pass Pass 3, 4 Pass Pass Pass 1
1 2 and 5 1 2 and 2
3.1. Experimental setup

Model of welding rod

70S-6 422 J 70S-6 H08A Accelerated corrosion testing methods mainly include salt spray test,
or welding wire
Diameter of welding humidity-controlled test, corrosive gas test, electrolytic accelerated
1 3.2 1 3.2
rod or wire (mm) corrosion test and paste corrosion test, among which salt spray test is
Current (A) 232 135 140 360 regarded as a reliable and convenient method that well simulates the
Voltage (V) 70 25 27.4 30 effect of marine atmospheric environment [30]. A salt spray corrosion
Velocity (mm/s) 5 7.5 2 5 7.5 5.46
testing system mainly include a salt spray cabinet, a temperature control
instrument, a spraying device, and a salt spray collector [31]. The salt
Table 2 spray test in this study was carried out at Shanghai Shipbuilding Tech­
Main chemical compositions of the base metal (BM) and weld materials (WM) nology Research Institute, by using CZ-120 A type salt spray testing
(%). instrument. Testing conditions were set in accordance with pertinent
Welding Material C Si Mn P S regulations specified in Chinese code GB/T10125–2012 [31]. A total of
method eight specimens (two of them are auxiliary) were placed in the salt spray
BM Q345 0.18 0.32 1 0.013 0.008 instrument (Fig. 4). Depending on the considered welding method, the
SMAW 422 J 0.078 0.18 0.39 0.023 0.018 specimens were labeled as SMAW-C16–1, SMAW-C16–2, GMAW-C16–1,
GMAW WM
70S-6
0.105 0.975 1.34 0.015 0.013
GMAW-C16–2, SAW-C16-1, and SAW-C16-2. All the specimens under­
(ER50–6) went 1600 h of accelerated corrosion test. International Standard Or­
SAW H08A 0.1 0.03 0.55 0.014 0.011
ganization (ISO 12944-2: 1998) [26] classified the corrosivity of

Fig. 3. Fabrication procedures of the dog-bone coupons (a) Cutting (b) Flattening (c) coupons ready for tests.

3
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 4. Experimental setup (a) Salt spray test cabinet (b) Placed specimens.

Table 3
Conversion of indoor accelerated corrosion time to typical atmospheric environment corrosion time.
Corrosion rate of ISO atmospheric environment K (μm/ C1 C2 C3 C4 C5
year) ≤1.3 1.3– 25 25– 50 50– 80 80– 200

Residential Xisha Yongxing Wanning City, Hainan

Typical environment Shanghai City Guangzhou City
interior Island Province
Corrosion rate K (μm/year) ≤1.3 17.8 36.4 146.9 551.4
1600 h corrosion induced thickness loss (μm) 174.42
Equivalent atmospheric corrosion time (year) 134 10 5 1 0.32

Table 4
Measured initial thicknesses and weights of the specimens.
Specimen No. Initial weight (g) Thickness (mm) Weight after corrosion (g) Weight loss Mean value

Point 1 Point 2 Point 3 Average

SMAW-C16–1 622.8 11.43 11.51 11.48 11.47 608.6 2.28%

1.98%
SMAW-C16–2 622.8 11.47 11.62 11.63 11.57 612.3 1.69%
GMAW-C16–1 634.4 11.42 11.47 11.51 11.47 623 1.80%
1.86%
GMAW-C16–2 630.8 11.49 11.47 11.48 11.48 618.7 1.92%
SAW-C16–1 638.2 10.89 10.8 10.99 10.89 623.9 2.24%
2.20%
SAW-C16–2 638.2 11.59 11.44 11.34 11.46 624.4 2.16%

atmospheric environments. The indoors environment of residential steel COMET 5 M laser scanner (Fig. 7). The minimum achievable accuracy
structures belongs to typical C1 environment regulated by ISO, and for scanning corrosion profiles is 1 × 10− 4 mm, determined by the
outdoor atmospheric environment generally belongs to ISO C2 or ISO C3 measuring accuracy of the optical scanning device. The model files ob­
environments. Table 3 provides the conversion of accelerated corrosion tained by scanning were imported to Geomagic Design X software
time to typical atmospheric environment corrosion time [26,32,33]. The (Fig. 8), and then “automatic partition” and “alignment wizard” com­
accelerated corrosion test of 1600 h can be equivalent to the corrosion of mands were applied to make the surface of a model parallel to the
carbon steel for nearly 10 years in the atmosphere of Shanghai City, and standard plan (Fig. 9). Finally, the clamping tools and other unnecessary
for nearly 5 years in the atmosphere of Guangzhou City. The weights and parts were removed, as shown in Fig. 10.
true thicknesses of the specimens, as well as the weights after corrosion The coordinates of two gauged surfaces (Side A and Side B depicted
are summarized in Table 4. Fig. 5 shows the appearances of the welded- in Fig. 11) on each specimen were output and summarized in Figs. 12-
joints after corrosion. Fig. 6 displays the side view of weldments 15. By statistical analysis, the corrosion depths of the BM regions and
corrosion. It can be seen from Fig. 5(b) and Fig. 6 that the corrosion in SAW welded joints approximately follow lognormal distribution, while
the welded zones is more severe than that of the BM area for the those of SMAW and GMAW welded joints approximately obey normal
considered specimens. distribution. The expectations and standard deviations of corrosion
depths of the BM and welded regions are listed in Table 5. It can be found
3.2. Topography scanning of corrosion pits that the average corrosion depths of the joint regions are much larger
than that of the BM region; specifically, the average corrosion depths of
Corrosion pits topography scanning was carried out at the Key Lab­ the joints with three different welding methods are 2.82, 2.12, and 3.06
oratory of Performance Evolution and Control for Engineering Struc­ times of their BM counterpart, respectively. (See Figs. 16–19.)
tures of Ministry of Education (PECES), Tongji University, by using

4
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 5. Corroded weldments (a) Before rust removal (b) After rust removal.

Fig. 6. Lateral views of weldment corrosion.

3.3. Electric potential analysis of Materials Science and Engineering (SMSE), by using Gamry Reference
600 electrochemical measuring system (Fig. 21). The system consists of
To explain why welded zone is more vulnerable to corrosion, dy­ an electrochemical workstation, a computer, and a three-electrode sys­
namic polarization curves of a typical region of interest (scanned area tem. In the three-electrode system, working electrode is the one to be
shown in Fig. 20) for the three weld joints were measured at the School

Fig. 7. Experimental setup of corrosion pits morphology scanning. Fig. 8. Primary model obtained by scanning.

5
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 11. Definition of the coordinate system.

the BM area.

4. Cyclic tests

4.1. Hysteretic curves

The label (code) of the considered specimens follows the sequences

of “Welding method-Corrosion time period-Loading protocol”, as
detailed in Table 7. Experimental setup was set in accordance with those
in the authors' previous works [3,34]. A pair of buckling-restraint
clamping plates were used to ensure that the specimens are under
axial tensile-compressive status. After testing, the load-displacement
Fig. 9. Model after coordinate transformation. curves of the specimens were recorded. Fig. 23 shows the typical load-
displacement curves of the corroded welded specimens during the first
measured, reference electrode is 3.5wt%NaCl solution, which is a few cycles. The yield and ultimate loads of the specimens are marked
reference electrode for measuring the electrode potential of electro­ with red fonts in the figure. There is no visible buckling during the tests,
chemical reaction, and platinum electrode plays the role of counter or due to the constraint effect of buckling restraint plates. The tension and
auxiliary electrode, which is only used to conduct current, to realize the compression asymmetry observed in Fig. 23, especially at large strain,
polarization of the electrode to be studied. may be caused by micro-crack propagation under tension and closure
The curves displayed in Fig. 22 are the dynamic polarization curves under compression, so the tensile peak load is slightly less than the
of the BM, WM, and HAZ for the three types of weldments. It can be compressive one. Minor tension and compression asymmetry is also the
found that the self-corrosion potentials of the BM are located above inherent property of steel itself.
those of the WM and HAZ, indicating that the corrosion tendency of the
BM is lower than those of the WM and HAZ. Under the same potential, 4.2. Load-carrying capacity
the current density of the BM is less than those of the WM and HAZ,
implying that the corrosion degree of the BM is lower than those of the The yield and ultimate loads of both uncorroded and corroded
WM and HAZ. In addition, the considered welded joints show apparent specimens are summarized in Table 8 and Table 9, and are plotted in
passivation behavior, i.e., passivation films are generated, which can Fig. 24. It can be found that accelerated corrosion has the most severe
refrain the corrosion process of the welded joints to a certain extent. impact on the yield load-carrying capacity of the SMAW welded spec­
By fitting the dynamic polarization curves illustrated in Fig. 22, the imen. After 1600 h of accelerated corrosion, the yield load of SMAW
self-corrosion potential Ecorr and corrosion current density icorr in the welded specimen is decreased by about 11.79%. For the SAW welded
different areas of the three considered weldments are obtained and are specimen, the yield load is decreased by about 8.41%. The corroded
listed in Table 6. For the three welding methods (i.e., SMAW, GMAW, GMAW welded specimen shows the least reduction in the yield load, i.e.,
and SAW), the self-corrosion potentials of the BM areas are about 260, by 7.93%. For ultimate load, when ±1% strain amplitude cyclic loading
60, and 380 mV higher than those of the WM area, and are about 440, was applied, the corroded SMAW welded specimen shows the least
220, and 500 mV higher than those of the HAZ. This confirms that the reduction (by 0.39%), and the corroded GMAW welded specimen ex­
WM and HAZ are more susceptible to corrosion compared to the BM, and hibits the greatest reduction in ultimate load (by 20.87%). With regards
shows that the corrosion resistance of SAW welded joint is the worst, to the case of ±3% strain amplitude cyclic loading, the ultimate load-
while the GMAW welded joint shows the best anti-corrosion capability. carrying capacities of the corroded welded specimen are decreased by
For icorr, the values of WM and HAZ are 1.31–2.5 times of the BM area. 6.24%–11.43% compared with those before corrosion.
Therefore, the WM and HAZ show more severe corrosion compared with

Fig. 10. Scanning results of corrosion pits morphology.

6
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 12. Coordinates of corrosion pits (BM region).

Fig. 13. Coordinates of corrosion pits (SMAW joints).

4.3. Energy dissipation capacity both uncorroded and corroded specimens, stabilized hysteretic cycles
are extracted and depicted in Fig. 25. The results are displayed in
Energy dissipation capacity is an important index to assess the hys­ Fig. 27. Overall, corrosion has a slight influence on the energy dissipa­
teretic performance of materials, and is usually reflected by equivalent tion capacity of the weld joints. The variation of EVD before and after
viscos damping (EVD) coefficient (marked as he in Fig. 26), whose corrosion is <6% in most cases.
calculation procedure is demonstrated in Fig. 26. To obtain the EVDs of

7
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 14. Coordinates of corrosion pits (GMAW joints).

Fig. 15. Coordinates of corrosion pits (SAW joints).

Table 5 4.4. Ultra-low cycle fatigue (ULCF) life

Variation of the distribution expectation and standard deviations with corrosion
time. The variations of the peak loads for the considered specimens are
Welding method BM SMAW GMAW SAW illustrated in Fig. 28 and Fig. 29. The number of cycles when the peak
Expectation (mm) 0.17 0.48 0.36 0.52
load is decreased to 85% of the value at cyclic stabilized loop is defined
Standard deviation 0.56 0.18 0.10 0.62 as the fatigue life [3,34,36]. It is noted that for specimens SMAW-C0–2
and GMAW-C16–1, macro crack was initiated outside the gauged region,
and the specimens experienced sudden fracture. As a result, there was no
apparent degradation of the ultimate loads during the whole loading

8
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 16. Corrosion distribution characteristic (BM, lognormal distribution).

Fig. 17. Corrosion distribution characteristic (SMAW, normal distribution).

Fig. 18. Corrosion distribution characteristic (GMAW, normal distribution).

9
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 19. Corrosion distribution characteristic (SAW, lognormal distribution).

Fig. 21. Gamry Reference 600 electrochemical measuring system.

Fig. 20. Schematic diagram of the dynamic polarization measuring area.

process, in which case the ULCF life is determined as the entire cyclic 5. Scanning electron microscope (SEM) analysis
number prior to failure. The comparison of the specimens' fatigue lives
before and after corrosion is demonstrated in Fig. 30. Under the two 5.1. Welding defects
different loading protocols (i.e., ±1% and ± 3% strain amplitudes), the
ULCF life of the SMAW welded specimens is reduced by 77.31% and As stated in Section 2, in the process of manufacturing butt-welded
8.57%, respectively, after experiencing corrosion. The ULCF life of the dog-bone specimens, the specimens were mechanically straightened to
GMAW welded specimens is reduced by 1.09% and 16.67%, respec­ flat by a mechanical testing machine to correct the welding deformation.
tively, after experiencing corrosion, and the ULCF life of the SAW wel­ By comparing Fig. 32 and Fig. 33, it can be found that small cleavage
ded specimens is reduced by 18.42% and 60%, respectively, after crack or generation of defects was not introduced in this process. There
experiencing corrosion. Overall, 1600 h of accelerated corrosion can is no obvious difference in initial defect between the as welded state and
cause a maximum reduction of about 77.31% in cyclic number at failure after-straightened states, implying that instantaneous cleavage fracture
for the welded specimen, while the minimum value of fatigue life surface or quasi-cleavage fracture were not generated in this process,
degradation is only 1.09%, indicating there is a great discreteness of the and the low-cycle fatigue life is minimally affected.
values. This is on one hand caused by distribution randomness of the
welding defects, and on the other hand caused by distribution
randomness of the corrosion pits.

10
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 22. Dynamic polarization curves for: (a) SMAW (b) GMAW and (c) SAW joints.

Table 6
Fitting results of the polarization curves.
Welding method SMAW GMAW SAW

Index BM WM HAZ BM WM HAZ BM WM HAZ

Self-corrosion potential Ecorr (mV) − 420 − 680 − 860 − 650 − 710 − 870 − 360 − 740 − 860
Self-corrosion potential difference (mV) 0 260 440 0 60 220 0 380 500
Corrosion current density icorr (μA/cm2) 0.12 0.22 0.22 0.13 0.17 0.17 0.1 0.19 0.25

Table 7 boundary of 422 J can be regarded as instantaneous fracture zone

Labels of all specimens. exhibiting ductile fracture surface with small ductile dimples. There are
No. Label Description also secondary cracks mixed in the fracture surface. In addition, quasi-
cleavage fracture was also observed on the fracture surface, and there
SMAW- SMAW welded specimen-0 h corrosion- ± 1% strain
1
C0–1 amplitude cyclic loading is a small amount of map cracks mixed in the middle. According to
SMAW- SMAW welded specimen-0 h corrosion- ± 3% strain Ref. [34], fracture patterns of uncorroded SAMW welded joints under
2
C0–2 amplitude cyclic loading large strain amplitude cyclic loads mainly include fatigue fracture,
SMAW- SMAW welded specimen-1600 h corrosion- ± 1% strain ductile fracture, and quasi-cleavage fracture. Therefore, the formation of
3
C16–1 amplitude cyclic loading
SMAW- SMAW welded specimen-1600 h corrosion- ± 3% cyclic strain
secondary cracks and map cracks is a unique fracture mode of corroded
4 SMAW joints.
C16–2 amplitude loading
GMAW- GMAW welded specimen-0 h corrosion- ± 1% strain For the GMAW welded joint (Fig. 34 (b)), due to the existence of
5
C0–1 amplitude cyclic loading welding-induced defect, corrosion medium contacts with internal metal,
GMAW- GMAW welded specimen-0 h corrosion- ± 3% strain
6 leading to the generation of a certain amount of corrosion products.
C0–2 amplitude cyclic loading
GMAW- GMAW welded specimen-1600 h corrosion- ± 1% strain Under cyclic loads, macro crack also initiated from the position where
7 obvious welding-induced defect is located at. With propagation of the
C16–1 amplitude cyclic loading

8
GMAW- GMAW welded specimen-1600 h corrosion- ± 3% cyclic fatigue crack, quasi-cleavage fracture surface and ductile fracture sur­
C16–2 strain amplitude loading face generated around.
SAW welded specimen-0 h corrosion- ± 1% strain amplitude
9 SAW-C0–1
cyclic loading
For the SAW welded joint, since automatic welding was used, no
SAW welded specimen-0 h corrosion- ± 3% strain amplitude apparent welding defect was formed. However, the heat input of the
10 SAW-C0–2
cyclic loading SAW welding method is much higher than that of the other two welding
11 SAW-C16–1
SAW welded specimen-1600 h corrosion- ± 1% strain methods. The microstructures of the BM at middle region experienced an
amplitude cyclic loading
obvious change when the welding process starts, and the grain size is
SAW welded specimen-1600 h corrosion- ± 3% cyclic strain
12 SAW-C16–2
amplitude loading rather coarse [6], which also led to an apparent change of the material
property. As a result, under fatigue loads, the fracture surface is rather
flat and smooth, and exhibits a characteristic of brittle fracture domi­
5.2. Fractography analysis nated by quasi-cleavage fracture and intergranular fracture, where no
void coalescence was observed among the crystals. During subsequent
To better understand the ultra-low cycle fatigue (ULCF) fracture welding processes, the previous heat input plays a role of pre-heating, so
mechanism of weld joints under cyclic loads, microscopic fractography the fatigue performance of the SAW welded joint gets better. Thus, the
was conducted at SMSE, Tongji University. Sigma 300 VP serial block- surrounding fractures exhibit fatigue fracture characteristics. Finally,
face scanning electron microscopy (SEM), as shown in Fig. 31, was ductile dimples appear at the instantaneous fracture zone of the joint.
used to measure the fracture surfaces. The scanned results are shown in Overall, the fracture mechanism of corroded welded-joints under
Fig. 34. For the SMAW welded joint (Fig. 34(a)), macro cracks initiated large strain amplitude loads is much more complex than that of uncor­
from various zones, where crack initiation zone 1 (middle zone) was roded ones. Except for fatigue fracture, ductile fracture, and quasi-
generated under the effect of welding-induced imperfection, while zone cleavage fracture, secondary cracks, map cracks, and intergranular
2 (edge zone) was generated due to surface corrosion. The macro cracks brittle fracture also appear on the fracture surfaces of the welded
propagate outward after initiation, followed by the formation of radial specimens. In addition, under the coupling effect of welding-induced
steps and stripes. Since two different weld materials, 70S-6 and 422 J, defects and corrosion pits, macro cracks of corroded welded-joints can
were involved in the SMAW welded joint, a clear boundary was formed initiate from various positions.
between the two different weld materials. Both sides outside the

11
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 23. Load-displacement curves of the corroded specimens under cyclic loads.

Table 8 of the corrosion pits. The initial model contains information including
Yield loads of the specimens with different welding methods. the coordinate point information of a specimen's external surface and
Welding ±1% ±3% Average the triangular meshes of the specimen's external surface. The irregular
method triangular meshes of the specimen's external surface are shell meshes
SMAW 50.68 63.70 57.19 rather than solid meshes, so this kind of mesh cannot be used to conduct
Uncorroded Corrosion induced
GMAW 60.37 52.76 56.57 FE modeling directly. A series of processing must be conducted,
[34] degradation
SAW 51.76 56.01 53.89 including importing the initial model to Rhino 7 software, and then
SMAW 51.56 49.33 50.44 11.79% using “QRMesh” command to convert the irregular meshes to regular
Corroded GMAW 49.65 54.51 52.08 7.93%
SAW 48.34 50.36 49.35 8.41%
quadrangular meshes, where the amount of the quadrangular meshes
can be adjusted through a QRSettings command. After the conversion of
mesh type, these quadrangular meshes can be rebuilt to a closed surface,
Table 9 and the closed surface can be filled to a solid model (Step 3 in Fig. 35).
Peak loads of the specimens with different welding methods. Finally, the solid geometric model can be generated and imported to FE
Welding ±1% ±3% software for pre-processing (Step 4 in Fig. 35).
method

SMAW 75.22 Corrosion 86.28 Corrosion 6.2. Constitutive model and its parameters
Uncorroded
GMAW 80.60 induced 90.45 induced
[34]
SAW 70.85 degradation 78.12 degradation Chaboche combined hardening constitutive model [40–43] based on
SMAW 74.93 0.39% 80.90 6.24%
Corroded GMAW 63.78 20.87% 84.01 7.12%
the von Mises flow law was applied. Isotropic hardening and kinematic
SAW 65.36 7.75% 69.19 11.43% hardening are contained in this model, as expressed as Eqs. (1) and (2),
respectively:
( )
6. Numerical simulation (1)
pl
σ 0 = σ y + Q 1 − e− bε

6.1. Geometric model C( )

α= 1 − e− γεpl
+ α1 e − γεpl
(2)
γ
Due to the high complexity of the corroded specimens' surface
topography, geometrical shapes and distribution characteristics of the where σ 0 is the yield surface of each constant amplitude loading cycle, σy
corrosion pits cannot be accurately modeled by using simplified corro­ is yield stress, Q is the maximum variation value of the yield surface, b is
sion pits such as cylindrical pits [37], conical pits [38], and hemi­ the ratio of yield surface variation with the increase of plastic strain, εpl
spherical pits [39]. To solve this problem, a method that uses is the equivalent plastic strain (PEEQ) at the end of each cycle, α is back
topography scanning results of the corrosion pits for creating finite stress, the ratio of C/γ is the maximum variation of back stress, while γ
element (FE) models is proposed in this paper, with the detailed defines the variation ratio of back stress that changes with the variation
modeling procedure shown in Fig. 35. The first step is to prepare the of plastic strain. As can be seen in Eqs. (1) and (2), several parameters
corroded specimen's initial model obtained by topography measurement such as Young's modulus E, Poisson's ratio μ, yield stress σ y, isotropic

12
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 24. Load-carrying capacity of the welded specimens with different welding methods: (a) yield load-carrying capacity (b) ultimate load-carrying capacity.

Fig. 25. Stabilized hysteresis curves of the coupons (a) ±1% strain amplitude loading (b) ±3% strain amplitude loading.

hardening parameters Q and b, kinematic hardening parameters C and γ, ∑ ∫ ε2 ∑ ∫ ε2

are involved in this model. According to Ref. [3], corrosion has no in­ VGI cyclic = exp(|1.5T|)dεpl − exp(|1.5T|)dεpl
fluence on the material property of steels, but only causes geometrical tensile cycles ε1 compressive cycles ε1

loss of the specimen surface. Therefore, for each considered zone, the (3)
combined hardening constitutive model parameters of the corroded
specimens are set in consistent with those of the uncorroded specimens. where ε1 and ε2 represent the plastic strains at the beginning and end of a
The details of the considered material parameters are given in Table 10. loading excursion, T is stress triaxiality:
/
T = σ m σeq (4)
6.3. Fracture prediction model
where σm is mean or hydrostatic stress:

While it is difficult to consider various fracture patterns in a single 1

σ m = (σ 1 + σ 2 + σ 3 ) (5)
fracture prediction model, the fatigue involved in this paper belongs to 3
the scope of ultra-low cycle fatigue (ULCF). For this kind of fatigue, the
fracture position and cycle number at failure can be predicted by using where σ1, σ 2, and σ 3 are three principle stresses.
( )
cyclic void growth model (CVGM) [44,45]. In this model, a void growth
VGI critical critical accumulated
cyclic = VGI monotonic exp − λεp (6)
index marked as VGIcyclic was defined, with the evolution equation of:

13
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 29. Variation of peak load with cyclic number (±3%).

Fig. 26. Calculation of EVD [35].

Fig. 30. Comparison of the welded specimens' fatigue lives before and
after corrosion.
Fig. 27. Calculated results of EVD.

Fig. 31. Experimental setup of SEM.

Fig. 28. Variation of peak load with cyclic number (±1%).
According to the discrepancy of material property and approximate
geometric dimensions of the WM, the model is divided into two parts,
where VGIcritical
monotonic is the critical value of VGI on the occasion of mono­ one is assigned for BM material property, and the other part is assigned
tonic loading, λ is treated as a material-dependent damage ability co­ for WM material property, as shown in Fig. 36. HAZ is assumed to share
efficient. εaccumulated
p is defined as the PEEQ that has accumulated up to the same material property with WM, and the purpose is to greatly
the beginning of each “tensile” excursion of loading. For the BM and WM improve modeling and calculation efficiency at the expense of very small
considering different welding methods, VGIcritical
monotonic was determined in
calculation accuracy. There are additional reasons for making such an
the range from 2.2 to 3.2, and λ ranges from 0.15 to 0.25, in accordance assumption: Firstly, the size of HAZ is difficult to determine, noting that
with Ref. [34]. An USDFLD subroutine was coded according to CVGM the method of measuring Vikers hardness [46] is not accurate. Secondly,
fracture model. More details about the iteration procedure of the sub­ the typical width of HAZ is about 3 mm, so it's not possible to fabricate
routine can be found in Ref. [34]. such small specimens for material property test [47]. Thirdly, the
boundary line of HAZ is very complicated and difficult to fully capture.

14
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 32. Welding imperfections (as welded state).

Fig. 33. Welding imperfections (after-straightened state).

15
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 34. Microscopic fractography of: (a) SMAW (b) GMAW (c) SAW welded joints.

16
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 34. (continued).

Fig. 35. Flow chart of the proposed modeling method.

Fourthly, an HAZ actually includes several different areas such as of buckling-restraint clamping plates was used to prevent the specimen
overheating area, normalizing area, incomplete normalizing area, and from buckling during the testing process. To simulate the effect of
tempering area, and their material properties can only be estimated buckling-restraint plates, the freedom of X direction on lateral surfaces
approximately. According to the method proposed in Subsection 6.1, the of the specimens was confined. The loading process is conducted by
part of the specimen between the clamping tools of the testing machine using displacements at both ends of the model, as shown in Fig. 38.
was modeled, as shown in Fig. 37. As mentioned in Subsection 4.1, a pair

17
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Table 10
Cyclic plasticity constitutive model parameters [34].
Parameter Unit BM WM for SMAW WM for GMAW WM for SAW

±1% ±3% ±1% ±3% ±1% ±3% ±1% ±3%

E MPa 1.92E+05 1.93E+05 1.91E+05 1.94E+05 1.91E+05 1.91E+05 1.94E+05 1.94E+05

μ – 0.26 0.26 0.26 0.26 0.26 0.30 0.26 0.26
σy MPa 277.61 243.82 319.20 298.10 319.20 307.49 258.91 244.39
Q MPa 96.01 89.99 50.00 65.19 93.98 110.85 97.77 99.64
b – 2.08 30.00 23.80 30.00 30.00 13.67 19.94 20.00
C1 MPa 6887.52 7338.38 6989.78 7048.57 7586.98 6693.32 7129.63 7337.65
γ1 – 98.86 81.40 40.00 83.45 50.00 84.15 50.00 84.78

6.4. Predicted results

In true experimental scenarios, even if a crack appears, compressive

capability of the specimen still exists. To consider this, the strength of
the damaged element was reduced to a certain extent, rather than
directly deleting the element. Under the cyclic load of ±3% strain
amplitude loading, the variations of load with cycle numbers predicted
by the Chaboche cyclic plasticity constitutive model and CVGM fracture
model are demonstrated in Fig. 39. For the SAW welded joint, the ULCF
life prediction results of both uncorroded and corroded specimens are
highly consistent with the test results. For the other two kinds of welded
specimens, the predicted error ranges from 6.67% to 22%, and with the
increase of cycle number at failure, the error of calculation accumulates
gradually.
Taking the case of GMAW as an example, the variations of VGIcyclic
and the corresponding critical values of the uncorroded and corroded
specimens under tensile-compressive hysteretic loads with cycle number
are illustrated in Table 11 and Table 12, respectively. It can be seen that
with the increase in cycle number, the value of VGIcyclic increases
Fig. 36. Material assignment.
gradually, while the critical value decreases gradually. For the uncor­
roded specimen, when loaded to the 8th cycle, no VGIcyclic value of any
element exceeds its corresponding critical value, but fracture damage
has already occurred at part of the elements of the corroded specimen.
When loaded to the 9th cycle, the VGIcyclic values of some elements at the
middle region of the uncorroded specimen begin to exceed their critical
values. Besides the difference in the cycle when failure initiates, another
apparent distinction between the failure modes of the uncorroded and
corroded specimens is that the macro crack of the uncorroded specimen
only initiates at the middle region, while for the corroded counterpart,
micro crack also initiates from the position where a critical corrosion pit
is located. This is consistent with the multi-crack origin phenomenon
observed by SEM. Under tensile-compressive cyclic loads, due to the
existence of corrosion pits, the stress status of the edge of pit changes
from uniaxial to triaxial, with significant stress concentration at edge of
the corrosion pit. As a result, when VGIcyclic begins to exceed VGIcritical
cyclic ,

Fig. 37. Load and boundary conditions. macro crack initiates at the edge of the critical corrosion pit. Overall, the
ULCF damage process of the corroded weld joints under cyclic loads can
be approximately simulated by using CVGM.

Fig. 38. Loading protocols.

18
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Fig. 39. Variation of load with cycle number.

Table 11
Evolution of VGI and its critical value (uncorroded specimen).

Cycle
8 9 10 11
number

critical
VGI cyclic

VGIcyclic

7. Conclusions of the joints under large strain amplitude cyclic loads. Finally, in-depth
numerical simulation was conducted to further interpret the test results.
In this paper, dog-bone flat-plate coupons were designed and fabri­ According to the results obtained by the experiments and numerical
cated using different welding methods, namely, SMAW, GMAW, and simulations, the following conclusions can be drawn:
SAW, and then neutral salt spray tests were conducted on the welded-
joints. After 1600 h of accelerated corrosion, topography scanning of 1) After 1600 h of accelerated corrosion, the mass wastages of the three
corrosion pits were carried out. Dynamic polarization curves of the weld welded specimens are about 1.98%, 1.86%, and 2.20%, respectively.
joints were also measured. Cyclic tests were then carried out, followed The average thicknesses of the welded regions are decreased by
by fractography analysis conducted to find out the fracture mechanism about 0.48 mm, 0.36 mm, and 0.52 mm, respectively, indicating that

19
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

Table 12
Evolution of VGI and its critical value (corroded).

Cycle
8 9 10 11
number

critical
cyclic

cyclic

the SAW welded specimen is more susceptible to corrosion under CRediT authorship contribution statement
neutral salt spray environment, while the GMAW welded specimen
exhibits the lowest corrosion sensitivity. Yuelin Zhang: Writing – original draft, Visualization, Validation,
2) The corrosion depths of the BM regions and SAW welded joints Software, Methodology, Investigation, Formal analysis, Data curation.
approximately follow a lognormal distribution, while those of the Wei Wang: Writing – review & editing, Supervision, Resources, Project
SMAW and GMAW welded joints approximately obey normal dis­ administration, Funding acquisition, Conceptualization. Hongwei Huo:
tribution. The average corrosion depths of the welded joints are Data curation, Formal analysis, Investigation, Methodology, Software,
much larger than that of the BM regions; specifically, the average Visualization, Writing – original draft. Yang Wang: Data curation,
corrosion degrees of the joints with the three different welding Investigation, Methodology, Software, Visualization, Writing – original
methods are 2.82, 2.12, and 3.06 times of their BM counterparts, draft. Cheng Fang: Writing – review & editing, Supervision, Resources,
respectively. Project administration, Methodology, Funding acquisition,
3) The reason why welded region shows preferential corrosion is that Conceptualization.
self-corrosion potential of the welded region is lower than that of the
BM region, while the corrosion current density (icorr) is higher than
Declaration of competing interest
that of the BM region. Under the three different welding methods, the
self-corrosion potentials of the BM areas are about 260, 60, and 380
The authors declare that they have no known competing financial
mV higher than those of the WM area, and are about 440, 220, and
interests or personal relationships that could have appeared to influence
500 mV higher than those of the HAZ. For icorr, the values of WM and
the work reported in this paper.
HAZ are 1.31–2.5 times of the BM area.
4) Accelerated corrosion lasting for 1600 h can cause a maximum
Data availability
reduction of 11.79% in yield load and 20.87% reduction in ultimate
load of Q345 weld joints. Corrosion has a slight influence on the
Data will be made available on request.
energy dissipation capacity of the welded joints. Under the two
different loading strain amplitudes of ±1% and ± 3%, the maximum
Acknowledgement
reductions in ULCF lives of the welded-joints are 77.31% and 60%,
respectively. Due to the distribution randomness of both welding-
The financial supports from the National Natural Science Foundation
induced defects and corrosion pits, the influence of corrosion on
of China (NSFC) with Grant Nos. 52078359 and 51820105013 are
each index of the specimens has a great discreteness.
gratefully acknowledged. Support for this study was also provided by
5) The fracture mechanism of corroded welded-joints under large strain
“Shuguang Program” which is supported by Shanghai Education
amplitude loads is much more complex than that of the uncorroded
Development Foundation and Shanghai Municipal Education Commis­
ones. Except for fatigue fracture, ductile fracture, and quasi-cleavage
sion (22SG18).
fracture, secondary cracks, map cracks, and intergranular brittle
fracture also appear on the fracture surfaces of the welded speci­
References
mens. In addition, under the coupling effect of welding-induced
defects and corrosion pits, macro cracks of corroded weld joints [1] Q. Zhang, J. Wen, Q. Han, H. Zhuge, Y. Zhou, Life-cycle assessment and prediction
initiate from various positions. on ultimate capacity of corroded Q690 steel columns with H-section under bi-
6) The ULCF performance of the corroded welded joints under large directional cyclic loading, Thin-Walled Struct. 195 (2024) 111391.
[2] C.A. Apostolopoulos, M.P. Papadopoulos, Tensile and low cycle fatigue behavior of
strain amplitude cyclic loads can be approximately simulated by the corroded reinforcing steel bars S400, Constr. Build. Mater. 21 (2007) 855–864.
combination of Chaboche cyclic plasticity constitutive model and [3] Y. Zhang, C. Fang, W. Wang, Experimental and numerical study on cyclic behavior
CVGM fracture model, further studies should be focused on the of corroded Q345 steel, J. Constr. Steel Res. 196 (2022) 107369.
[4] Y. Zhang, Y. Wang, The influence of welding mechanical boundary condition on
development of more efficient and comprehensive fracture models.
the residual stress and distortion of a stiffened-panel, Mar. Struct. 65 (2019)
259–270.
[5] Y. Zhang, L. Tian, The effect of joint configuration on residual stress and distortion
of the 304 stainless steel multi-pass welded joints, Mater. Today Commun. 30
(2022) 103070.

20
Y. Zhang et al. Journal of Constructional Steel Research 215 (2024) 108561

[6] Y. Zhang, X. Liu, Y. Zhou, Y. Shi, Influence of welding method on residual stress [26] ISO 12944-2: 1998, Paints and Varnishes-Corrosion Protection of Steel Structures
and metallography of a mild steel welded butt-joint plate, J. Constr. Steel Res. 199 by Protective Paint Systems [S]. 473, British Standards Institution, London, 1998.
(2022) 107640. [27] GB 50661–-2011, Code for Welding of Steel Structures [S], China Construction
[7] Robert Tremblay, Peter Timler, Michel Bruneau, Andre Filiatrault, Performance of Industry Press, Beijing, 2011.
steel structures during the 1994 Northridge earthquake, Can. J. Civ. Eng. 22 (1995) [28] Tianyi Zhang, Man-Tai Chen, Ao Cai, Huan Cao, Xiang Yun, Hui Zhao, Chen Shen,
338–360. Yuelong Zhang, Behaviour of high strength steel butt joints exposed to arctic low
[8] Michel Bruneau, Performance of steel bridges during the 1995 Hyogoken–Nanbu temperatures, Thin-Walled Struct. 192 (2023) 111157.
(Kobe, Japan) earthquake—a North American perspective, Eng. Struct. 20 (12) [29] Man-Tai Chen, Ao Cai, Madhup Pandey, Chen Shen, Yuelong Zhang, Hu. Lili,
(1998) 1063–1078. Mechanical properties of high strength steels and weld metals at arctic low
[9] James O. Malley, SAC steel project: summary of phase 1 testing investigation temperatures, Thin-Walled Struct. 185 (2023) 110543.
results, Eng. Struct. 20 (4–6) (1998) 300–309. [30] Wang Fengping, Li Jielan, Ding Yanwei, Metal Corrosion and Protection
[10] Wei Wang, Fangfang Liao, Yiyi Chen, Ductile fracture prediction and post-fracture Experiment [M], Chemical Industry Press, Beijing, 2014, p. 60P.
path tracing of steel connections based on micromechanics-based fracture criteria [31] GB/T10125–2012, Corrosion Tests in Artificial Atmospheres-Salt Spray Tests,
[J], Eng. Mechan. 31 (3) (2014) 101–108, 115. Standards Press of China, Beijing, 2013.
[11] Fangfang Liao, Wei Wang, Yiyi Chen, Extremely low cycle fatigue fracture [32] Liu Xin, Hu. Shi, Anti Corrosion and Fire Resistant Coating of Steel Structures [M],
prediction of steel connections under cyclic loading [J], J. Tongji Univ. (Nat. Sci.) Chemical Industry Press, Beijing, 2005, pp. 20–79.
42 (4) (2014) 539–546, 617. [33] Wu Zhang Quancheng, Chen Jiaguang Jiansheng, et al., Study on corrosion
[12] C. Fang, M.C.H. Yam, A.C.C. Lam, L.K. Xie, Cyclic performance of extended end- resistance of weathering steel and carbon steel in offshore atmosphere, Mater. Sci.
plate connections equipped with shape memory alloy bolts, J. Constr. Steel Res. 94 Eng. 19 (2) (2001) 12–15.
(2014) 122–136. [34] Yuelin Zhang, Cheng Fang, Wei Wang, Haowen Hou, Cyclic plasticity and ULCF
[13] C. Fang, W. Wang, C. He, Y.Y. Chen, Self-centring behaviour of steel and steel- behavior of steel butt-joints considering different welding methods, Int. J. Fatigue
concrete composite connections equipped with NiTi SMA bolts, Eng. Struct. 150 173 (2023) 107684.
(2017) 390–408. [35] X. Zhang, S. Zheng, X. Zhao, Experimental and numerical study on seismic
[14] C. Fang, W. Wang, C. Qiu, S. Hu, G.A. MacRae, M.R. Eatherton, Seismic resilient performance of corroded steel frames in chloride environment, J. Constr. Steel Res.
steel structures: a review of research, practice, challenges and opportunities, 171 (2020) 106164.
J. Constr. Steel Res. 191 (2022) 107172. [36] Yvbin Guo, Cheng Fang, Yue Zheng, Post-fire hysteretic and low-cycle fatigue
[15] Gang Shi, Meng Wang, Yuanqing Wang, Fei Wang, Cyclic behavior of 460 MPa behaviors of Q345 carbon steel, J. Constr. Steel Res. 187 (2021) 106991.
high strength structural steel and welded connection under earthquake loading [J], [37] J.K. Paik, J.M. Lee, M.J. Ko, Ultimate compressive strength of plate elements with
Adv. Struct. Eng. 16 (3) (2013) 451–466. pit corrosion wastage [J], Proc. Inst. Mechan. Eng. Part M J. Eng. Mar. Environ.
[16] J. Veerababu, R. Sunil Goyal, K. Laha Sandhya, Low cycle fatigue behaviour of 217 (4) (2003) 185–200.
grade 92 steel weld joints, Int. J. Fatigue 105 (2017) 60–70. [38] Tatsuro Nakai, Hisao Matsushita, Norio Yamamoto, Hironori Arai, Effect of pitting
[17] Su-Juan Guo, Run-Zi Wang, Haofeng Chen, Fu-Zhen Xuan, A comparative study on corrosion on local strength of hold frames of bulkcarriers (1st report), Mar. Struct.
the cyclic plasticity and fatigue failure behavior of different subzones in CrNiMoV 17 (2004) 403–432.
steel welded joint, Int. J. Mech. Sci. 150 (2019) 66–78. [39] Yi Huang, Yan Zhang, Gang Liu, Qi Zhang, Ultimate strength assessment of hull
[18] Huiyong Ban, Juncheng Zhu, Gang Shi, Cyclic loading tests on welded connections structural plate with pitting corrosion damnification under biaxial compression,
of stainless-clad bimetallic steel and modelling, J. Constr. Steel Res. 171 (2020) Ocean Eng. 37 (2010) 1503–1512.
106140. [40] J.L. Chaboche, G. Rousselier, On the plastic and viscoplastic constitutive
[19] Wenyuan Zhang, Lijing Zeng, Experimental investigation and low-cycle fatigue life equations—part I: rules developed with internal variable concept, J. Press. Vessel.
prediction of welded Q355B steel, J. Constr. Steel Res. 178 (2021) 106497. Technol. 105 (2) (1983) 153–158.
[20] T.G. Gooch, P.H.M. Hart, Review of welding practices for carbon steel deaerator [41] J.L. Chaboche, Time-independent constitutive theories for cyclic plasticity, Int. J.
vessels, in: Paper 303, presented at Corrosion/86, National Association of Plast. 2 (2) (1986) 149–188.
Corrosion Engineers, Houston, TX, March 1986. [42] J.L. Chaboche, Constitutive equations for cyclic plasticity and cyclic
[21] Chao Jiang, C.S. Wen Xiong, Linyun Zhou Cai, Yichen Zhu, Fatigue assessment of viscoplasticity, Int. J. Plast. 5 (1989) 247–302.
fillet weld in steel bridge towers considering corrosion effects, Eng. Fail. Anal. 143 [43] J.L. Chaboche, Unified cyclic Viscoplastic constitutive Equations-1: development,
(2023) 106901. capabilities, and thermodynamic framework, Unified Const. Laws Plastic Deform.
[22] H.C. Ma, Z.Y. Liu, C.W. Du, H.R. Wang, X.G. Li, D.W. Zhang, Z.Y. Cui, Stress (1996) 1–68.
corrosion cracking of E690 steel as a welded joint in a simulated marine [44] A.M. Kanvinde, G.G. Deierlein, Cyclic void growth model to assess ductile fracture
atmosphere containing Sulphur dioxide, Corros. Sci. 100 (2015) 627–641. initiation in structural steels due to ultra low cycle fatigue, J. Struct. Eng. 133 (6)
[23] Jialiang Song, Zhaoliang Li, Kui Xiao, Hong Luo, Junhang Chen, Yu Wei, (2007) 701–712.
Xin Zhang, Chaofang Dong, Study on the local corrosion behaviour and mechanism [45] Christopher Smith, Amit Kanvinde, Gregory Deierlein, Calibration of continuum
of bogie steel welded joints, Corros. Sci. 208 (2022) 110709. cyclic constitutive models for structural steel using particle swarm optimization,
[24] Qi Li, Qiong Yao, Liyang Sun, Hongchi Ma, Chuang Zhang, Ning Wang, Effect of J. Eng. Mech. 143 (5) (2017) 04017012.
micro-galvanic corrosion on corrosion fatigue cracking of the weld joint of high [46] Wen-Yu Cai, Yan-Bo Wang, Guo-Qiang Li, Richard Stroetmann, Comparative study
strength bridge steel, Int. J. Fatigue 170 (2023) 107568. on strength of TMCP and QT high-strength steel butt-welded joints, J. Constr. Steel
[25] Gu Yufen, Lu Na, Xu Youwei, Yu Shi, Gang Zhang, Qingling Sun, Microstructure Res. 197 (2022) 107447.
characteristics of Q345R-steel welded joints and their corrosion behavior in a [47] Fangfang Liao, Wei Wang, Yiyi Chen, Parameter calibrations and application of
hydrofluoric acid environment, J. Nucl. Mater. 574 (2023) 154214. micromechanical fracture models of structural steels, Struct. Eng. Mech. 42 (2)
(2012) 153–174.

21