Title: Atmospheric corrosion of gas metal arc welded dissimilar T-joints made of duplex stainless

steel and high-strength carbon steel

ed
Authors: Shuang Suna, 1, Chenying Liua, 1, Gonçalo Ferrazb, Burak Karabulutb, Brent Verhoevenb,
Xiongfeng Ruana, b, Raf Dewila, b, and Barbara Rossia, b
aDepartment of Engineering Science, University of Oxford, Oxford, United Kingdom

iew
bFaculty of Engineering Technology, KU Leuven, Sint-Katelijne-Waver, Belgium

1The first two authors contributed equally to this work.

Abstract
This paper investigates the corrosion behaviour of dissimilar T-joints made of duplex stainless

ev
steel and carbon steel under atmospheric environments. Stainless steel grade EN1.4062 was
welded to two different carbon steel grades S355J2 and P460NL2 via gas metal arc welding to
form T-joints. The fabricated specimens were subjected to 12-month exposure tests in four mildly
corrosive environments, distributed across 3 different European countries: Belgium, England, and

rr
Portugal. Samples were designed, fabricated, and exposed to continuously monitored
environments. Uniform, galvanic, and pitting corrosions were studied using microscopic scanning,
to develop guidance on the use of protective coating systems for dissimilar weldments. Corrosion
rates were quantitatively analysed. A corrosion database was subsequently formed to calibrate
ee
mathematical models for better prediction of the corrosion rate of carbon steel samples.

1 Introduction
Stainless steel (SS) and coated carbon steel (CS) are used in modern structures due to combined
p

high strength and durability, therefore leading to cost effectiveness. Their indefinitely recyclable
nature also offers a circular solution to the construction industry. Depending on their chemical
composition, the two metals greatly differ in cost, strength, ductility, and corrosion resistance. SS
ot

contains a high proportion of chromium (> 10.5% in mass), which reacts with oxygen to form a
self-repairing protective layer preventing corrosion. CS is susceptible to corrosion and therefore
requires a protective coating but has a generally higher supply availability at lower price.
tn

Recent studies have revealed that hybrid solutions, i.e., combining both types of metals, can bring
benefits such as tailored corrosion performance, improved sustainability, and reduced lifespan
cost [1], [2], [3], [4], [5]. A case study by Karabulut et al. has found that partially replacing carbon
rin

steel in a girder bridge with duplex stainless steel can lower both material use and maintenance
costs [2]. Hence, there has been a surge of interest to explore hybrid SS-CS structures in the built
environment to reduce overall cost and increase corrosion resistance.
ep

SS and CS alloys can be connected via dissimilar welding. The process causes microstructure
changes in the weld nugget and heat affected zone (HAZ) [6], [7], [8]. In particular, the corrosion
resistance of dissimilar weldments may worsen considerably. Recent scientific studies (see Table1,
[6], [7], [8], [9], [10], [11], [12]) have improved the understanding of the corrosion behaviour of
Pr

SS-CS dissimilar welds and provided guidance to effectively mitigate the risks with protective
coatings. Table 1 summarises prior work focusing on the corrosion performance of SS-CS
dissimilar welds.

1
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 In the literature, all families of SS have been investigated i.e., ferritic, austenitic, and duplex,
covering a wide range of yield strengths, whereas low carbon steel has been the primary focus of

ed
CS investigations, with yield strengths up to 350 MPa. Different welding processes have been
employed using common fillers like E2209. Standard electrochemical corrosion tests are usually
preferred. Important conclusions include: higher corrosion of the weldment due to galvanic
corrosion (see [12]), significantly lower corrosion performance owing to appropriate welding

iew
methods and fillers (see [6], [7], [8], [9], [10], [11]).

Despite the existing literature, little information exists on the corrosion behaviour of dissimilar
welds outside of laboratory conditions. Additionally, whilst a few duplex SS samples with yield
strength up to 560 MPa have been studied, there is usually a mismatch in the yield strength of

ev
the associated CS plates. All these factors have made it rather difficult to assess the true potential
of dissimilar SS-CS welds for construction use under real conditions.

To address the research gap, this study focuses on the atmospheric corrosion of SS-CS dissimilar

rr
welds in four European locations. Specifically, we have selected stainless duplex steel (i.e.,
EN1.4062) and low carbon steel (i.e., S355J2 and P460NL2) as base materials, to align with
industrial practice and select grades with close mechanical performance. Following the standard
gas metal arc welding process to form T-type joints between the two types of steel, a total of 84
ee
samples were designed and fabricated according to European standards ISO 8501-1 [13]. They
were subsequently placed in four European geographical locations for a year-long monitored
atmospheric corrosion test. The morphology and corrosion rate of each sample were analysed,
focusing on the heat affected zone (HAZ). Based on the findings, guidance on the necessary
p

protective coating on SS-CS welds is provided. Lastly, prediction models for CS corrosion are re-
evaluated, and a corrosion database has been formulated for carbon steel under mildly corrosive
environments to calibrate those models, thereby informing the standard set in ISO 9223:2012
ot

[14], i.e., corrosion of metals and alloys.

2 Experimental Procedure
tn

2.1 Base metals and specimen preparation

The lean duplex grade, i.e., EN1.4062, is chosen as the SS base material due to its high yield
strength compared to austenitic grades, cost effectiveness compared to traditional duplex grades,
and good corrosion resistance in medium corrosive environments. Two types of CS grades are
rin

considered in this study: the European Standard structural steel S355J2 and a high-quality
pressure vessel steel P460NL2. Both grades are low-carbon, readily weldable, and increasingly
employed in bridge structures. The base materials are joined by GMAW using a 309LSi electrode
of 1 mm diameter. The strength and chemical composition of each material is given in Table 2,
ep

according to EN 10025-2 [15] and EN 10025-3 [16].

According to the optimal selection given by Cools & Staepels [4], the welding parameters in Table
3 are adopted. Long samples were welded and then cut into small specimens for exposure.
Pr

Abrasive blast cleaning was conducted to remove weld notches and match the usual
requirements for steel plates prescribed in steel bridge specifications [17], i.e., Sa21⁄2 surface
quality.

2
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
Table 1. Recent studies of dissimilar welds between SS and CS.
Refs. SS Strength1 [MPa] CS Strength [MPa] Thickness2 [mm] Welding Filler
e d
Corrosion Test
[12] N/A (F3)
[7]
[7] AISI 304 (A)
4
AISI 304 (A )
300 – 400
205
205
N/A
6
AISI 1020 (LC )
AISI 1020 (LC)
N/A
350
350
20
15
15
Laser
GTAW
SMAW9
8
N/A
N/A
N/A

ie w EC12
EC
EC

v
[9] AISI 304 (A) 205 Unspecified (LC) N/A 6 GTAW ER309L 3.5 wt% NaCl
[10] S30408 (A) 310 12Cr2Mo1R (LC) 310 40 GTAW10 N/A 3.5 wt% NaCl

e
5
[11] UNS 31803 (D ) 480 IS 2062 (LC) 250 5 SMAW E2209, E309 1M NaCl, 1M H2SO4
[8]
[6]
ER 2209 (D)
2205 (D)
560
448
ASTM A516-60 (LC) 220
16MnR (LA ) 7 345

r r 12
8
GMAW11
GTAW
ER2209
E(R)2209
EC
3.5 wt% NaCl
1 Strength: 0.2% proof strength (yield strength); 2 Thickness: base materials come with a shape of plate and are then welded together; 3 F: ferritic; 4 A: austenitic;

e
5 D: duplex; 6 LC: low carbon, also known as mild steel or plain C; 7 LA: low alloy; 8 GTAW: gas tungsten arc welding; 9 SMAW: shield metal arc welding; 10

combined with submerged arc welding; 11 GMAW: gas metal arc welding; 12 EC: electrochemical corrosion with unspecified information given on the solution.

p e
Table 2. Minimum yield strength and chemical composition of selected steels in this study.

t
Strength Chemical composition [wt.%]
Grade
[MPa] C Si Mn Cr Ni Mo Cu N

o
EN1.4062 530 0.02 - 2.00 22.00 2.00 0.40 - 0.20
S355J2 355 0.20 0.34 1.46 0.05 0.03 0.01 0.04 -
P460NL2

t
309LSi
460
410
n
0.20 0.32
0.02 0.80
1.10 0.02
1.80 23.50
-
13.50
0.01
-
0.04
-
-
-

ir n
Table 3. Welding parameters to fabricate the dissimilar T-joints.
Welding parameters
Horizontal Vertical Welding
Gas U I V Q
plate plate material

p
[V] [A] [cm/min] [kJ/mm]
EN1.4062 S355J2/P460NL2 309LSi Arcal 129 27 253 45.2 0.73

re
P
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 2.2 Rack of samples
As shown in Figure 1, seven types of configurations are investigated. The first three columns (a -

ed
c) are three base steels without welding, used as control groups. Their plate sizes and thickness
are annotated on the left side. The middle two columns (d - e) are T-joints of SS and CS with
weldments, whose geometric parameters are given on the right side. The last two columns (f - g)
depicts parts of the welded T-joints where the CS part has been removed to exclude their
influence during the atmospheric exposure programme. Two holes were drilled in each specimen

iew
to facilitate attachment to a test frame with rip-ties. Data collection equipment is also illustrated
(see paragraph 2.3.2). As shown in Figure 1, three rows of the same configurations have been
made for repeatability [18]. The test specimens were directly exposed away from any shelter. The
height of the lowest row of samples was 75 cm above ground level to prevent rain splashes. The
specimens were inclined at an angle of 45° facing south for maximum sunlight exposition.

ev
rr
p ee
ot
tn

Figure 1. Experiment rack of test samples with (a) EN1.4062 base plate; (b) S355J2 base plate; (c)
P460NL2 base plate; (d) Dissimilar T-joint of EN1.4062/S355J2 with 309LSi electrode; (e) Dissimilar
T-joint of EN1.4062/P460NL2 with 309LSi electrode; (f) Dissimilar weld of EN1.4062/309LSi
rin

(S355J2 removed); (g) Dissimilar weld of EN1.4062/309LSi (P460NL2 removed).

2.3 Atmospheric exposure preparation

2.3.1 Test sites
Four European cities were chosen to conduct the exposure experiments. The environmental
ep

corrosivity of each city is closely related to its distance to the sea and to the road with de-icing
salts. Specifically, the four cities, i.e., Sint-Katelijne-Waver (Belgium), Oxford (United Kingdom),
Casalinho da Foz (Portugal), and Knokke-Heist (Belgium), have the following distances to the sea:
40.9 km, 96.5 km, 18.8 km, and 0.8 km. All test sites were located near the meteorological stations
of each city, with distance to the road with de-icing salts between 0.01 – 0.1 km. According to
Pr

EN1993 Eurocode 3 [19], the corrosivity of each geographical location is low except for Knokke-
Heist which is medium.

2.3.2 Experimental measurements

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 The exposure programme started in September 2021 and specimens were withdrawn after 12
months in August 2022. The environmental parameters influencing the corrosion of steels, i.e.,

ed
chloride, temperature, and humidity, were monitored continuously at each test site. Specifically,
the chloride deposition was collected using a piece of dry gauze that was stretched and held in a
Plexiglas frame using the “dry plate” technique [20], as shown in Figure 1 (bottom left). The gauze
was washed each month, and the amount of deposited chloride was evaluated using ion
chromatography. The gauze was sheltered to provide maximum protection from rain.

iew
Temperature and relative humidity were recorded using a UX100-003 data logger, as shown in
Figure 1 (bottom right). The equipment was placed inside an insulation box where the air flow
was allowed from the sides and bottom. The influence of radiation reflection from the metal
surface below was minimised.

2.3 Post-corrosion assessment

ev
2.3.1 Experimental methodology
After one-year exposure programme, the samples were first subjected to microscopic
examination using a Hirox KH-8700 digital microscope. The qualitative analysis was conducted
using a scanning electron microscope (SEM) and an energy dispersive spectroscopy (EDS).

rr
Afterwards, the corrosion rate of each sample was evaluated based on their overall mass loss.
The loose corrosion products on each specimen were removed according to ASTM G1-03 [21], i.e.,
the corroded specimens were chemically cleaned in the 1000 mL solution with 100 mL of nitric
acid and 20 mL of hydrofluoric acid at 25°C for 10 minutes. The cleaning procedure was repeated
ee
six times until weight loss stagnation. After each cycle, the specimens were subsequently
measured on a digital scale with an accuracy of ±0.001g to calculate the difference from their
original weights.
p
2.3.2 Considered theoretical models for comparison
The collected corrosion data are compared with theoretical predictions using dose-response
functions based on previous literature and industrial standards, the details of which are
summarised in Table 4. The environmental parameters measured or collected from local stations
ot

are taken as the input of these functions, thus producing relevant predictions of corrosion rate.

Table 4. Dose-response functions for CS corrosion published in the literature.

Ref.
tn

Function N1 Variable
(year)
Lien & San
𝜏 0.15 𝑃𝑑 0.22 RH3, T4, τ5, [22]
𝑅𝑐𝑜𝑟𝑟 = 0.0157 × 𝑅𝐻2.3 × ( ) × (1 + ) × 𝑒 −0.0019×(𝑇+20) 21
3800 50 Pd6 (2002)
rin

Knotkova et al.
𝑅𝑐𝑜𝑟𝑟 = 0.091 × 𝑃𝑑 0.56 × 𝜏 0.52 × 𝑒 𝑓(𝑇) + 0.158 × 𝑆𝑑 0.58 × 𝜏 0.25 × 𝑒 0.05×𝑇 [23]
𝑓(𝑇) = 0.103 × (𝑇 − 10) 𝑇 ≤ 10℃ 125 T, Pd, Sd7, τ
𝑓(𝑇) = −0.059 × (𝑇 − 10) 𝑇 > 10℃ (2003)
Klinesmith et al.
ep

𝜏 0.46 𝑃𝑑 0.62 𝑆𝑑 0.34 [24]

𝑅𝑐𝑜𝑟𝑟 = 13.4 × ( ) × (1 + ) × (1 + ) × 𝑒 0.02×(𝑇+20) 190 T, Pd, Sd, τ
3800 25 50 (2007)
Knotkova et al.
𝑅𝑐𝑜𝑟𝑟 = 0.09 × 𝑆𝑑 0.56 × 𝜏 0.53 × 𝑒 𝑓(𝑇) + 0.24 × 𝑃𝑑 0.47 × 𝜏 0.25 × 𝑒 0.05×𝑇 [25]
𝑓(𝑇) = 0.098 × (𝑇 − 10) 𝑇 ≤ 10℃ 119 T, Pd, Sd, τ
(2010)
Pr

𝑓(𝑇) = −0.087 × (𝑇 − 10) 𝑇 > 10℃

Standard ISO 9223
𝑅𝑐𝑜𝑟𝑟 = 1.77 × 𝑃𝑑 0.52 × 𝑒 0.02×𝑅𝐻+𝑓(𝑇) + 0.108 × 𝑆𝑑 0.62 × 𝑒 0.033×𝑅𝐻+0.04×𝑇 RH, T, [14]
𝑓(𝑇) = 0.150 × (𝑇 − 10) 𝑇 ≤ 10℃ 128
𝑓(𝑇) = −0.054 × (𝑇 − 10) 𝑇 > 10℃ Pd, Sd (2012)

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 1 N: Number of data points; 2 Rcorr: the first-year corrosion rate of carbon steel [μm/a]; 3 RH: the
annual average relative humidity [%]; 4 T: the annual average temperature [°C]; 5 τ: the annual

ed
average time of wetness [h/a], i.e., the duration during which the relative humidity exceeds 80%
and the temperature remains above 0°C; 6 Pd: the annual average sulphur dioxide deposition rate
[mg/(m2·d)]; 7 Sd: the annual average chloride deposition rate [mg/(m2·d)]; 8 f(T): a correction
factor specifically for steel.

iew
3 Results and Discussion
3.1 Environmental parameters
The annual averages of relative humidity, temperature, time of wetness, sulphur dioxide
deposition rate, and chloride deposition rate at all test sites for the period spanning September
2021 to August 2022 are summarised in Table 5. The data is obtained from experimental
measurements as well as the European Environment Agency [26] and meteorological stations.

ev
Table 5. Annual average of environmental parameters at four sites from Sep. 2021 to Aug. 2022.
Site RH[%] T [oC] Pd [mg/(m2·d)] Sd [mg/(m2·d)] Rainfall [mm] τ [h/a]
Source M1 M S2 M S C3
SKW 73.0 13.2 1.2

rr
8.0 292 3721
Oxford 79.2 12.2 2.5 7.6 302 4991
Casalinho da Foz 79.8 15.8 0.7 6.4 280 5030
Knokke-Heist 75.5 13.0 1.2 17.0 163 3349
ee
1 M: monitored, 2 S: station, 3 C: calculated

The more detailed monthly data are given in Figure 2. The temperature exhibits a roughly
unimodal distribution over the year. The deposition rates of sulphur dioxide at all test sites are
below 4 mg/(m2·d), indicating the lowest level of pollution according to ISO 9223 [14]. All the test
p

sites have medium chloride deposition rate whilst the coastal city, Knokke-Heist, has the highest
annual average. The humidity level is relatively consistent throughout the year and the time of
wetness observed ranges from 3349 h/a to 5030 h/a. This duration indicates long presence of
water films, which is likely to act as electrolytes and accelerate atmospheric corrosion of steels.
ot
tn
rin
ep

Figure 2. Monthly average values of humidity, time of wetness, rainfall, temperature, sulphur
dioxide deposition rate, and chloride deposition rate from Sep. 2021 to Aug. 2022.
Pr

3.2 Microstructure morphology of corroded samples

3.2.1 Uniform corrosion and morphology of rust layer

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 Pictures of the corroded T-joints are shown in Figure 3 after the year-long exposure programme.
The rust formation of each specimen mainly appeared in the CS plates, with little to no corrosion

ed
observed on the duplex SS surface as well as its nearby edges with the T-joints.

iew
ev
Figure 3. Corrosion of selected T-joints at four tests sets and the micromorphology of rust layers.

As for the CS plates, the exact rusted morphology varies by geographical sites and steel grades.

rr
As shown in Figure 3, specimens from the first three sites exhibit yellow rust (an indication of α-
FeOOH and Fe3O4) whilst the one at Knokke-Heist exhibit grey-green rust (an indication of γ-
FeOOH). The colour difference can be attributed to the effect of rain wash. For a CS plate, there
are normally two layers of rust on its surface. The inner one is fine and compact [27], and the
ee
outer layer has a looser and more porous structure [27], [28]. The outer one is less stable and can
easily flake off in the rain. During the exposure programme, the rainfall at Knokke-Heist was
significantly less than the others, and hence the outer surface was mostly maintained.

To compare the corrosion behaviours between different CS grades, the specimens at Knokke-
p

Heist are taken as an example. Enlarged microscopic images are obtained for both grades. As
shown in Figure 4, the rust layer on steel S355J2 was thinner, indicating better corrosion
resistance. The phenomenon is likely due to the difference in chemical compositions between the
two CS grades. According to Table 2, S355J2 comes with a higher proportion of nickel and
ot

chromium, which has been found to facilitate the formation of a protective layer [27], [28], [29],
[30], [31], [32]. Hence, only relatively loose rusts exist on the S355J2 grade while P460NL2 has a
wider spread and more uniformly distributed rust layer, which could significantly impact its
tn

structural performance.
rin
ep

Figure 4. Morphology of rust layers on carbon steel in Knokke-Heist after year-long exposure.

3.2.2 Galvanic corrosion

Pr

Whilst corrosion has primarily influenced CS plates, galvanic corrosion has also happened, i.e., if
two dissimilar metals are connected in a corrosive environment, the less noble metal takes the
role of the anode and corrodes at a faster rate than it should [33]. As shown in Figure 3, the

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 junctions between CS and SS plates are corroded, whilst the cases at Knokke-Heist are most
severe, attributed to high chloride deposition rate.

ed
To assess the electrochemical behaviour of the studied materials further, grades S355J2 and
EN1.4062 were placed in 1M sodium chloride solution for 12 hours. The potential of each material
was recorded continuously, with EN1.4062 between -0.05 V and -0.1 V and S355J2 between -0.55
V and -0.65 V. Hence, the potential difference measurement between the two grades ranges from

iew
+0.45 V to + 0.6V. Similar phenomenon was also obtained on grades P460NL2 and EN1.4062. This
observation demonstrates potential galvanic corrosion in the experiments and also informs the
galvanic series of metals in [33].

3.2.3 Pitting corrosion

Pitting corrosion was observed on the CS plates and nearby HAZ after the cleaning procedure. As

ev
shown in Figure 5, the T-joint specimen of grade P460NL2 at Knokke-Heist is taken as an example.

rr
ee
Figure 5. Pits on carbon steel and HAZ of the T-joint with grade P460NL2 at Knokke-Heist. Colour
p

maps showing the pit depths are also indicated where the lowest point is taken as zero.

A total of 20 pits were randomly selected across the CS plate and HAZ of each T-joint specimen.
The average pit depth is illustrated in Figure 6(a). In general, more and deeper pits exist on grade
ot

P460NL2 compared to those of grade S355J2. Additionally, more severe localised corrosion was
observed in the HAZ of each sample. The pits found in the HAZ near the CS plate are significantly
deeper in Knokke-Heist. This is due to the medium corrosivity in that site where the metal surface
tn

was almost completely deteriorated due to high chloride deposition rate. This observation
suggests that in such salt-rich environments, a coating system should be implemented as a
protection mechanism for dissimilar welds, particularly in the HAZ.
rin
ep
Pr

Figure 6. (a) Average pit depth on the surface of carbon steels and HAZ. (b) EDS analysis of
dissimilar T-joint between EN1.4062 (A) and 309LSi (B) at Knokke-Heist after exposure.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 Interestingly, no pits were detected in the HAZ between the austenitic 309LSi filler and EN1.4062
lean duplex SS plate at any test locations. As shown in Figure 6(b), there is a discernible fusion

ed
line between the base metal and filler, which is a result of the difference in the proportion of
nickel contents. Both materials also have similarly high proportion of chromium, which facilitates
forming protection films against corrosion.

3.3 Atmospheric corrosion rate

iew
3.3.1 Measured corrosion rate by weight loss
Figure 7(a) depicts the overall mass loss of a certain specimen as a function of the number of
cleaning cycles. Two approximately straight lines can be observed. Line AB represents the removal
of corrosion products and steel substrate, while line BC depicts the removal of the steel substrate
only. Point D represents the mass of the steel substrate at zero cleaning cycles and can be
obtained by extrapolating line BC to the ordinate axis. The entire mass loss due to corrosion is the

ev
distance between points A and D. According to ASTM G1-03 [21], the corrosion rate is
𝐾𝑊
𝑅𝑐𝑜𝑟𝑟 =
𝐴𝑡ρ
where K is the conversion factor 8.76 x 107, W is the mass loss [g], A is the exposure area [cm2],
t is the time of exposure [h], and ρ is the density of steel [g/cm3].

rr
p ee
ot

Figure 7. (a) Stabilised mass loss after chemical cleaning, as exemplified by the result of the T-
tn

joint with grade S355J2 in Knokke-Heist. (b) Sectioned T-joint for decomposed corrosion rate.

The averaged corrosion rate of each specimen is summarised in Table 6. According to corrosivity
category defined by ISO 9223 [14], the corrosion level of the SS plates falls into the lowest level
C1 and the one of all CS plates into C2. As expected, CS plates have corroded at a much faster rate
rin

due to the lack of a protective layer and S355J2 base plates corroded less than P460NL2, which
aligns with the observed uniform and pitting corrosion behaviour. The overall corrosion rates of
T-joints were lower than those of the CS-based plates, as a result of the added SS filler and base
plates. Compared to the samples where CS plates are entirely removed, i.e., N-S355J2 and N-
P460NL2, their corrosion rates are slightly higher than pure SS plates, indicating that dissimilar
ep

welds are more susceptible to corrosion.

Table 6. Averaged corrosion rate measured by weight loss after the year-long exposure.
Corrosion rate [µm/a]
Pr

Test site T- T- N- N-
EN1.4062 S355J2 P460NL2
S355J2 P460NL2 S355J2 P460NL2
SKW 0.42 9.23 12.50 4.63 4.96 1.50 1.59

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 Oxford 0.47 10.10 13.29 5.74 5.93 1.60 1.68
Casalinho da Foz 0.40 13.12 15.69 6.46 6.89 1.72 1.80

ed
Knokke Heist 0.41 16.02 18.28 8.42 8.87 2.22 2.30

3.3.2 Corrosion rate by decomposition

To evaluate how much dissimilar welding has influenced the corrosion behaviour of the T-joint
samples, each specimen is decomposed into four segments, i.e., two base materials and two

iew
“carbon-removed joints” as depicted in Figure 7(b). The “decomposed” corrosion rate is
calculated from the measured rates of individual base metals and carbon-removed joints in Table
6. Hence, the corrosion rate of a T-joint by decomposition goes as
∑4 𝑅𝑐𝑜𝑟𝑟−𝑖 𝐴𝑖
𝐷𝑅𝑐𝑜𝑟𝑟 = 𝑖=1 4
∑𝑖=1 𝐴𝑖
where Ai is the surface area of each section, and the effect of drilled holes has been taken into

ev
account for compensation. Rcorr-i is the corrosion rate of each segment, as obtained in Table 6.

The average decomposed corrosion rates of each T-joint are summarised in Table 7. Compared
to the actual rates of the T-joint samples given in Table 6, the decomposed rates are mostly lower

rr
except for the one made of P460NL2 at SKW. The results suggest that in general dissimilar welding
accelerate corrosion, due to galvanic corrosion between CS and SS plates and corrosion of HAZ
near the CS plate. The samples at Oxford and Knokke-Heist have been subjected to a higher
impact, as indicated by their relative RoCs.
ee
Table 7. Averaged corrosion rate of T-joints by decomposition after the year-long exposure.
Decomposed T-S355J2 Decomposed T-P460NL2
Test site 1
Corrosion rate [μm/a] RoC Corrosion rate [μm/a] RoC
p

SKW 4.55 -1.73% 5.21 5.04%

Oxford 4.97 -13.42% 5.54 -6.58%
Casalinho da Foz 6.26 -3.10% 6.42 -6.82%
Knokke-Heist 7.69 -8.67% 7.58 -14.5%
ot

1 RoC refers to ratio of change compared to the actual corrosion rate.

3.4 Corrosion prediction

tn

3.4.1 Prediction of carbon steel corrosion rate from literature

Using the dose-response functions from Table 4 and the environmental parameters in this study
as summarised in Table 5, the corrosion rate of CS plates can also be estimated at each site. The
results are compared in Figure 8. In general, the models by Knotkova et al. (2003) [23] and
Knotkova et al. (2010) [25], as well as from ISO 9223 standards [14] are relatively closer to the
rin

measured results. Meanwhile, the predicted values from the models by Lien & San (2002) [22]
and Klinesmith et al. (2007) [24] have exaggerated the potential corrosion behaviours at all sites,
which is likely because both functions were originally calibrated for more aggressive marine
environments. To improve the prediction accuracy, a corrosion database is formed to calibrate
existing dose-response functions in mildly corrosive environments.
ep

3.4.2 Corrosion database

Corrosion data points are collected from prior research where specimens were subjected to
similar atmospheric conditions to this study. Up to 11 pieces of literature have been identified.
Pr

Prior to statistical analysis, manual data filtering process has been carried out based on the
following criteria: (i) the corrosivity category must be the same as our study at C2; (ii) the time of
wetness must be maintained below 5500 h/a; (iii) the sulphur dioxide deposition rate and chloride
deposition rate must be kept below 4 mg/(m2·d) and 60 mg/(m2·d), respectively; (iv) in cases

10

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 where meteorological or pollution data are missing, the data point is excluded. After manual
filtering, a database consisting of 50 data points are formed in Table 8 from locations worldwide

ed
as marked in Figure 9.

iew
ev
rr
ee
Figure 8. Comparison between the estimated and measured corrosion rates of carbon.
p

Table 8. Database of corrosion tests used in this study compiled from literature sources.
Pd Sd Rcorr
No. Location Grade T [°C] τ [h/a] Refs.
[mg/(m2·d)] [mg/(m2·d)] [µm/a]
ot

1 Iguazu AISI 1006 22.9 5387 2.0 1.5 5.8 [25]

2 San Juan AISI 1006 19.2 911 2.0 1.5 4.6 [25]
3 Ahtari AISI 1006 4.0 3040 2.8 1.5 9.7 [25]
tn

4 Ahtari AISI 1006 4.0 3127 2.5 1.5 12.5 [25]

5 Ahtari AISI 1006 4.1 3381 1.5 1.5 11.3 [25]
6 Birkenes AISI 1006 5.2 3679 1.4 0.6 21.4 [25]
7 Birkenes AISI 1006 5.9 4608 1.0 0.6 18.6 [25]
rin

8 Birkenes AISI 1006 6.3 4187 1.0 0.6 21.8 [25]

9 Birkenes AISI 1006 7.5 4406 0.8 0.6 17.1 [25]
10 Birkenes AISI 1006 6.6 3968 0.8 0.6 20.7 [25]
11 Birkenes AISI 1006 6.2 3679 0.8 0.6 18.6 [25]
ep

12 EI Pardo AISI 1006 25.3 2427 3.1 1.5 16.3 [34]

13 EI Pardo AISI 1006 25.3 2716 3.8 1.5 17.0 [34]
14 EI Pardo AISI 1006 25.2 3522 3.3 1.5 15.6 [34]
15 Los Angeles CS 17.4 1956 0.5 33.0 17.3 [34]
Pr

16 San Juan CS 18.8 902 2.0 1.5 4.9 [34]

17 Caratinga CS 21.5 4222 0.8 8.9 8.6 [34]
18 Caratinga CS 20.9 4222 1.3 7.4 11.5 [34]

11

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 19 Caratinga CS 21.2 4222 1.7 1.6 13.1 [34]
20 Brasilia CS 20.4 3872 2.0 1.5 12.9 [34]

ed
21 Paulo Afonso CS 25.9 1507 2.0 1.5 17.3 [34]
22 Cotove CS 27.0 2891 0.3 1.5 19.6 [34]
23 / CS 14.4 3504 0.7 20.0 24.5 [35]
24 / CS 7.5 4406 0.3 2.7 6.2 [35]

iew
25 / CS 6.6 3968 0.6 29.1 23.5 [35]
26 / CS 15.3 4845 0.3 2.7 5.4 [35]
27 / CS 14.2 1500 0.8 3.0 7.4 [35]
28 Chennai CS 28.3 2791 0.9 1.7 19.0 [36]
29 Delhi CS 25.5 702 0.2 2.3 13.9 [36]

ev
30 Digha CS 26.6 2280 1.0 0.8 24.2 [36]
31 Easter Island SAE 1070 20.5 4690 1.2 50.0 30.0 [37]
32 Mawlamyine CS 23.8 5532 2.4 1.9 13.9 [38]
33 Yangon CS 28.9 4002 2.4 3.7 17.3 [38]

rr
34 Mandalay CS 28.6 1172 2.4 1.2 8.9 [38]
35 Rajiv Gandhi S235 28.0 1798 1.4 57.3 3.7 [39]
36 Fort George S235 28.0 1800 1.9 41.6 5.7 [39]
ee
37 Mutual Aid S235 28.3 1800 1.4 14.5 2.3 [39]
Medine Camp
38 S235 26.3 3272 1.6 51.4 6.3 [39]
de Masque
39 Xishuangbanna Q345 21.5 4320 0.1 0.1 14.9 [40]
p
40 Kralupy CS 11.0 3679 4.0 1.0 5.1 [41]
41 Phangnga CS 27.7 5090 2.6 28.6 18.2 [42]
42 Prague CS 11.0 3592 4.0 3.0 7.0 [41]
ot

43 Berlin CS 10.0 4030 1.6 12.0 4.8 [41]

44 Barcelona CS 15.0 4468 3.0 3.0 6.6 [41]
45 Birkenes CS 6.0 3942 1.0 7.0 7.3 [41]
tn

46 UEMA CS 27.2 4920 2.8 8.3 21.8 [43]

47 Santa Rita CS 27.8 3420 3.0 16.6 24.6 [43]
48 Miranda CS 28.2 3390 2.0 13.0 24.3 [43]
49 São Mateus CS 28.2 3897 1.9 9.5 19.8 [43]
rin

50 Peritoró CS 28.5 3560 1.3 9.5 20.2 [43]

ep
Pr

Figure 9. Maps of atmospheric corrosion test sites gathered from the literature.

3.4.3 Calibration of dose-response functions

12

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 According to Table 4, when only T, τ, Pd, and Sd are considered, the dose-response functions in
the literature [24], [25] exhibit two distinct forms. The corrosion rate of each model can be

ed
rewritten as
𝑎 𝑎
𝑅𝑐𝑜𝑟𝑟−𝑎 = 𝑎1 𝑃𝑑 2 𝜏 𝑎3 𝑒 𝑎4 𝑇+𝑎5 + 𝑎6 𝑆𝑑 7 𝜏 𝑎8 𝑒 𝑎9 𝑇
𝜏 𝑏3 𝑃𝑑 𝑏5 𝑆𝑑 𝑏7
𝑅𝑐𝑜𝑟𝑟−𝑏 = 𝑏1 ( ) (1 + ) (1 + ) 𝑒 𝑏8 (𝑇+𝑏9 )
𝑏2 𝑏4 𝑏6
where b2, b4, b6, and b9 are the average of annual time of wetness, sulphur dioxide deposition

iew
rate, chloride deposition rate, and temperature, respectively.

The corrosion database is used to calibrate the unknown coefficients, i.e., ai and bi, of the two
models, which are determined by the least square method using descent algorithm in Python.
The final coefficients for the two calibrated dose-response functions, i.e., Model a and Model b,
are given in Table 9.

ev
Table 9. Values of all coefficients present in the calibrated dose-response functions.
Eq. a1 a2 a3 a4 a5 a6 a7 a8 a9
Rcorr-a 7.93 -0.15 0.1 -0.04 0 2.3e-4 -0.12 0.73 0.19

rr
Eq. b1 b2 b3 b4 b5 b6 b7 b8 b9
Rcorr-b 16.4 3384 0.3 1.7 0.01 9.4 0.3 0.002 18.7

The 50 corrosion points from the database are plotted in Figure 10 using models [24], [25] and
ee
the calibrated ones to compare measured and predicted corrosion rates.
p
ot
tn
rin
ep

Figure 10. Predicated corrosion rates from calibrated dose functions versus measured results.

The existing three models either overestimate the corrosion rates or have a relatively wide
variance. Model b has managed to reduce the variance, yet the overestimation still exists. In
Pr

contrast, Model a exhibits good alignment between actual and predicated results with a small
variance.

13

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 The recalibrated dose functions are finally used to predict the corrosion rates of CS plates in our
study. The results are compared with the measured rates as shown in Figure 11. Model a has a

ed
higher predication accuracy, which is the same as its similar formats as shown in Figure 9. Model
b goes with a bigger discrepancy, but it should be noted that the difference has already been
significantly reduced compared to its equivalence in Figure 9.

iew
ev
Figure 11. Predicated corrosion rates from calibrated dose functions versus measured results.

rr
To better compare the efficiency of the two calibrated functions as well as those from literature,
the root mean square error (RMSE) of each model at the four tested sites for two CS grades is
calculated as follows with results in see Table 10.
ee
1 4
RMSE = √ ∑ (𝑓𝑖 − 𝑦𝑖 )2
4 𝑖=1
where 𝑓𝑖 is the predicted value and 𝑦𝑖 is the corresponding real-measured value. Among all of
them, the new calibrated dose function Rcorr-a has the highest prediction accuracy for grade S355J2
p
whilst the model in [23] outperforms others for grade P460NL2.

Table 10. RMSE of the models from literature and by calibration of two CS grades. Unit: g/(m2·a)].
Model S355J2 P460NL2
ot

Ref. [22] 30.64 27.83

Ref. [23] 5.14 3.26
Ref. [24] 15.69 12.86
tn

Ref. [25] 5.98 3.73

Ref. [14] 5.89 3.73
Rcorr-a 2.74 4.22
Rcorr-b 11.22 8.37
rin

5 Conclusions
This work focusses on the atmospheric corrosion of dissimilar welded T-joints between duplex
stainless steel (EN1.4062) and carbon steel (S355J2 and P460NL2) in mildly corrosive
environments in European cities. Based on the experimental results and formulated corrosion
database, we have drawn the following conclusions.
ep

Uniform corrosion occurred on all CS plates on the dissimilar T-joints due to the absence of a
protective layer. Little corrosion was observed on SS plates and carbon-removed dissimilar welds,
which suggests that welding did not significantly influence uniform corrosion of the SS-CS junction.
Galvanic corrosion was noticed and validated by electrochemical tests and decomposed corrosion
Pr

rates. Pitting corrosion also happened across all specimens. In particular, the HAZ on the side of
CS exhibited more severe pitting corrosion. This is attributed to decarburisation and grain growth
as a result of element diffusion during the welding process. The pits found in the HAZ of the CS

14

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 grades were notably deeper in Knokke-Heist due to high chloride deposition rate. Contrarily, the
HAZ on the side of the SS exhibited high pitting resistance, due to the formation of protective

ed
films on stainless steel grades. These findings have provided guidance on implementing suitable
protection at appropriate areas, i.e. the CS plates, nearby HAZ, and the junction between CS and
SS plates should be protected in salt-rich environments however, the stainless steel plates could
be left unprotected.

iew
Additionally, the evaluation of previous literature on corrosion prediction have revealed that
some models could lead to too conservative results as they were formulated in marine
environments of high corrosivity. A newly formulated corrosion database established specifically
for mildly corrosive environments, was used to re-calibrate two prediction functions from prior
research. The comparative analysis has informed the best model for CS corrosion under similar
conditions.

ev
References
[1] Y. Bouassida et al., Bridge design to eurocodes Worked examples, vol. 1, no. 1. 2012.
[2] B. Karabulut, G. Ferraz, and B. Rossi, ‘Lifecycle cost assessment of high strength carbon
and stainless steel girder bridges’, J Environ Manage, vol. 277, 2021, doi:

rr
10.1016/j.jenvman.2020.111460.
[3] T. Yabuki, Y. Arizumi, T. Shimozato, S. Guezouli, H. Matsusita, and M. Tai, ‘Hybrid Stainless
Steel Girder for Bridge Construction’, International Journal of Structural and Construction
Engineering, vol. 11, no. 5, 2017.
ee
[4] Y. ; S. A. Cools, ‘Microstructural and mechanical investigations on dissimilar welding’, KU
Leuven, 2021.
[5] K. ; B. N. V. Sterckx, ‘Painted, stainless or hybrid steel box girder bridges: Comparative life
cycle cost assessment’, KU Leuven, 2020.
p
[6] S. Wang, Q. Ma, and Y. Li, ‘Characterization of microstructure, mechanical properties and
corrosion resistance of dissimilar welded joint between 2205 duplex stainless steel and
16MnR’, Mater Des, vol. 32, no. 2, 2011, doi: 10.1016/j.matdes.2010.07.012.
[7] W. Chuaiphan, S. Chandra-Ambhorn, S. Niltawach, and B. Sornil, ‘Dissimilar welding
ot

between AISI 304 stainless steel and AISI 1020 carbon steel plates’, in Applied Mechanics
and Materials, 2013. doi: 10.4028/www.scientific.net/AMM.268-270.283.
[8] P. Luchtenberg, P. T. de Campos, P. Soares, C. A. H. Laurindo, O. Maranho, and R. D. Torres,
‘Effect of welding energy on the corrosion and tribological properties of duplex stainless
tn

steel weld overlay deposited by GMAW/CMT process’, Surf Coat Technol, vol. 375, 2019,
doi: 10.1016/j.surfcoat.2019.07.072.
[9] T. E. Abioye, O. E. Ariwoola, T. I. Ogedengbe, P. K. Farayibi, and O. O. Gbadeyan, ‘Effects of
welding speed on the microstructure and corrosion behavior of dissimilar gas metal arc
rin

weld joints of AISI 304 stainless steel and low carbon steel’, in Materials Today:
Proceedings, 2019. doi: 10.1016/j.matpr.2019.06.383.
[10] Y. Huang et al., ‘Microstructure and corrosion characterization of weld metal in stainless
steel and low carbon steel joint under different heat input’, Mater Today Commun, vol. 29,
2021, doi: 10.1016/j.mtcomm.2021.102948.
ep

[11] P. B. Srinivasan, V. Muthupandi, W. Dietzel, and V. Sivan, ‘An assessment of impact

strength and corrosion behaviour of shielded metal arc welded dissimilar weldments
between UNS 31803 and IS 2062 steels’, Mater Des, vol. 27, no. 3, 2006, doi:
10.1016/j.matdes.2004.10.019.
[12] W. Wu, S. Hu, and J. Shen, ‘Microstructure, mechanical properties and corrosion behavior
Pr

of laser welded dissimilar joints between ferritic stainless steel and carbon steel’, Mater
Des, vol. 65, 2015, doi: 10.1016/j.matdes.2014.09.064.

15

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 [13] ISO 8501-1:2007 Preparation of steel substrates before application of paints and related
products — Visual assessment of surface cleanliness Part 1: Rust grades and preparation

ed
grades of uncoated steel substrates and of steel substrates after overall removal of
previous coatings. 2007.
[14] ISO, ‘Corrosion of metals and alloys – Corrosivity of atmospheres – Classification,
determination and estimation’, ISO FDIS, vol. 13B, 2012.
[15] (CEN) European Committee for Standardisation, ‘BS EN 10025-2:2004 Hot rolled products

iew
of structural steels. Part 2: Technical delivery conditions for non-alloy structural steels’,
British Standard, vol. 2, 2004.
[16] British Standards, ‘Hot rolled products of structural steels— Part 3: Technical delivery
conditions for normalized/normalized rolled weldable fine grain structural steels’, BS EN
10025-3:2004, vol. 3, 2004.
[17] Y. Fujino, ‘Corrosion protection of steel bridges’, Corrosion Engineering, vol. 38, no. 10,

ev
1989, doi: 10.3323/jcorr1974.38.10_546.
[18] ISO 8565, ‘Metals and Alloys. Atmospheric Corrosion Testing. General Requirements’,
2011.
[19] J. C. Taylor, ‘ EN1993 Eurocode 3: Design of steel structures ’, Proceedings of the Institution
of Civil Engineers - Civil Engineering, vol. 144, no. 6, 2001, doi: 10.1680/cien.2001.144.6.29.

rr
[20] International Organization for Standardization, ‘Corrosion of metals and alloys-Corrosivity
of atmospheres-Measurement of environmental parameters affecting corrosivity of
atmospheres Corrosion’, Reference number ISO, vol. 9225, 2012.
[21] A. G1-03, ‘Standard Practice for Preparing, Cleaning, and Evaluating Corrosion Test
ee
Specimens.’, ASTM Special Technical Publication, no. Reapproved, 2017.
[22] L. T. H. Lien and P. T. San, ‘The effect of environmental factors in carbon steel atmospheric
corrosion; The prediction of corrosion’, in ASTM Special Technical Publication, 2002. doi:
10.1520/stp10886s.
p
[23] V. Kucera et al., ‘UN/ECE ICP materials dose-response functions for the multi-pollutant
situation’, Water, Air, and Soil Pollution: Focus, vol. 7, no. 1–3, 2007, doi: 10.1007/s11267-
006-9080-z.
[24] D. E. Klinesmith, R. H. McCuen, and P. Albrecht, ‘Effect of Environmental Conditions on
ot

Corrosion Rates’, Journal of Materials in Civil Engineering, vol. 19, no. 2, 2007, doi:
10.1061/(asce)0899-1561(2007)19:2(121).
[25] D. Knotkova, K. Kreislova, and S. W. Dean, ISOCORRAG International Atmospheric Exposure
Program: Summary of Results. 2010. doi: 10.1520/ds71-eb.
tn

[26] ‘European Environment Agency’, 2024. Accessed: Jul. 24, 2024. [Online]. Available:
https://www.eea.europa.eu/en
[27] X. L. Gao, Y. Han, G. Q. Fu, M. Y. Zhu, and X. Z. Zhang, ‘Evolution of the rust layers formed
on carbon and weathering steels in environment containing chloride ions’, Acta
rin

Metallurgica Sinica (English Letters), vol. 29, no. 11, 2016, doi: 10.1007/s40195-016-0472-
4.
[28] G. Su and X. Gao, ‘Comparison of medium manganese steel and Q345 steel on corrosion
behavior in a 3.5 wt % NaCl solution’, Materials, vol. 10, no. 8, 2017, doi:
10.3390/ma10080938.
ep

[29] A. HIGGINSON, R. C. NEWMAN, and R. P. M. PROCTER, ‘ChemInform Abstract: The

Passivation of Fe‐Cr‐Ru Alloys in Acidic Solutions’, ChemInform, vol. 21, no. 7, 1990, doi:
10.1002/chin.199007028.
[30] J. Guo, M. Seo, Y. Sato, G. Hultquist, C. Leygraf, and N. Sato, ‘ELECTROCHEMICAL BEHAVIOR
AND SURFACE COMPOSITION OF COPPER CONTAINING FERRITIC STAINLESS STEEL IN
Pr

SULFURIC ACID SOLUTION.’, Boshoku gijutsu, vol. 35, no. 5, 1986, doi:
10.3323/jcorr1974.35.5_283.

16

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912
 [31] M. Reffass, R. Sabot, C. Savall, M. Jeannin, J. Creus, and P. Refait, ‘Localised corrosion of
carbon steel in NaHCO3/NaCl electrolytes: Role of Fe(II)-containing compounds’, Corros

ed
Sci, vol. 48, no. 3, 2006, doi: 10.1016/j.corsci.2005.02.016.
[32] R. E. Melchers, ‘Effect of small compositional changes on marine immersion corrosion of
low alloy steels’, Corros Sci, vol. 46, no. 7, 2004, doi: 10.1016/j.corsci.2003.10.004.
[33] ‘Specialty Steel Industry of North Ammerican: galvanic corrosion’,
https://www.ssina.com/education/corrosion/galvanic-corrosion/ . Accessed: Jul. 24, 2024.

iew
[Online]. Available: https://www.ssina.com/education/corrosion/galvanic-corrosion/
[34] B. Chico, D. De la Fuente, I. Díaz, J. Simancas, and M. Morcillo, ‘Annual Atmospheric
Corrosion of Carbon Steel Worldwide. An Integration of ISOCORRAG, ICP/UNECE and
MICAT Databases’, Materials, vol. 10, no. 6, 2017, doi: 10.3390/ma10060601.
[35] S. K. Coburn, C. P. Larrabee, H. H. Lawson, and O. B. Ellis, ‘Corrosiveness of Various
Atmospheric Test Sites as Measured by Specimens of Steel and Zinc’, in ASTM Special

ev
Technical Publication, 1968. doi: 10.1520/STP34098S.
[36] D. D. N. Singh, S. Yadav, and J. K. Saha, ‘Role of climatic conditions on corrosion
characteristics of structural steels’, Corros Sci, vol. 50, no. 1, 2008, doi:
10.1016/j.corsci.2007.06.026.
[37] J. F. Marco, M. Gracia, J. R. Gancedo, M. A. Martín-Luengo, and G. Joseph,

rr
‘Characterization of the corrosion products formed on carbon steel after exposure to the
open atmosphere in the Antarctic and Easter Island’, Corros Sci, vol. 42, no. 4, 2000, doi:
10.1016/S0010-938X(99)00090-6.
[38] W. Thandar, Y. Y. K. Win, T. Khaing, Y. Suzuki, K. Sugiura, and I. Nishizaki, ‘Investigation of
ee
Initial Atmospheric Corrosion of Carbon and Weathering Steels Exposed to Urban
Atmospheres in Myanmar’, International Journal of Corrosion, vol. 2022, 2022, doi:
10.1155/2022/4301767.
[39] Y. Seechurn, B. Y. R. Surnam, and J. A. Wharton, ‘Marine atmospheric corrosion of carbon
p
steel in the tropical microclimate of Port Louis’, Materials and Corrosion, vol. 73, no. 9,
2022, doi: 10.1002/maco.202112871.
[40] J. Huang et al., ‘Atmospheric corrosion of carbon steels in tropical and subtropical climates
in Southern China’, Materials and Corrosion, vol. 71, no. 8, 2020, doi:
ot

10.1002/maco.201911487.
[41] T. T. J. Prošek, ‘Exposure site catalogue: catalogue of atmospheric corrosion field exposure
sites in europe.’, 2021.
[42] N. S. Palsson, K. Wongpinkaew, P. Khamsuk, S. Sorachot, and W. Pongsaksawad, ‘Outdoor
tn

atmospheric corrosion of carbon steel and weathering steel exposed to the tropical-
coastal climate of Thailand’, Materials and Corrosion, vol. 71, no. 6, 2020, doi:
10.1002/maco.201911340.
[43] Y. C. Sica, E. D. Kenny, K. F. Portella, and D. F. Campos Filho, ‘Atmospheric corrosion
rin

performance of carbon steel, galvanized steel, aluminum and copper in the North Brazilian
coast’, J Braz Chem Soc, vol. 18, no. 1, 2007, doi: 10.1590/S0103-50532007000100017.
ep
Pr

17

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4987912