Received: 16 September 2019 Revised: 25 October 2019 Accepted: 28 October 2019

DOI: 10.1111/ffe.13156

ORIGINAL CONTRIBUTION

Fatigue strength degradation of corroded structural details:

A formula for S‐N curve

N.D. Adasooriya | D. Pavlou | T. Hemmingsen

Department of Mechanical and Structural

Abstract
Engineering and Materials Science,
University of Stavanger, Stavanger N‐ A formula is proposed to predict fatigue strength of corroded members and
4036, Norway joints of steel structures. The concept of the formula is first studied from
Correspondence
recently identified mechanism of corrosion fatigue. Hence, the corresponding
N. D. Adasooriya, Department of fatigue strength curve (i.e. S‐N curve) of corroded steel is presented. It is fur-
Mechanical and Structural Engineering ther improved to derive linear, bilinear or trilinear S‐N curve for corroded con-
and Materials Science, University of
Stavanger,Stavanger N‐4036, Norway. structional details of steel structures. The parameters of the corroded steel S‐N
Email: mudiyan.n.adasooriya@uis.no curve are determined based on the corrosion fatigue testing results of different
types of steel specimens in air, fresh water and seawater. Hence, the parame-
ters for the derived S‐N curve of corroded constructional details are predicted
based on the above parameters and tabulated for the detail categories given
in the Eurocode and DNVGL code. The proposed S‐N curve formula is com-
pared with full‐scale fatigue test results of several constructional details, and
the validity of the formula is confirmed.

KEYWORDS
corrosion fatigue, joints and connections, S‐N curve, steel structures

1 | INTRODUCTION Though many bridges are located onshore, de‐icing salt

may simulate marine environment to the bridges in
Corrosion is one of the main structural degradation pro- snowy regions.2 These failures emphasize the importance
cesses that affect integrity. Fractures due to cyclic stress of more accurate simulation of CF strength of structural
in corrosive media are designated as corrosion fatigue joints and connections (i.e. constructional details) in dif-
(CF) which is a type of environmental assisted cracking. ferent corrosive environments.3,5,6
The forty percent failures in oil and gas pipelines and Corrosion pit induced stress concentration models and
8% of failures in offshore steel structures are reported to fatigue notch factor are commonly used to determine
be caused degradation/deterioration due to corrosion. structures integrity.7-10 Fatigue crack initiation life is
Corrosion commonly takes place in the splash zone area determined when micro‐cracks propagate to certain criti-
of offshore structures as the coating in this area disap- cal length. Fatigue crack propagation life is determined
pears a few years after installation.1 Recent collapses of by fracture mechanics approach. Accurate determination
steel bridges due to corrosion are reported and some of modified corrosion‐fatigue stress intensity threshold
examples are the Silver Bridge in 1967, the Mianus River and the critical length has been reported as challenging
Bridge in 1990 and the Minnesota Bridge in 2007.2-4 and difficult.2-4 Therefore, limited amount of published

--------------------------------------------------------------------------------------------------------------------------------
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the
original work is properly cited.
© 2019 The Authors. Fatigue & Fracture of Engineering Materials & Structures published by John Wiley & Sons Ltd

Fatigue Fract Eng Mater Struct. 2019;1–13. wileyonlinelibrary.com/journal/ffe 1

2 ADASOORIYA ET AL.

works is found in the area of fatigue notch factor due to model the fatigue strength of material, full‐scale steel
pitting corrosion.9,10 Recent investigations of non‐passive members and joints (i.e. constructional details).
metals such as carbon steel reveals that CF cracks do not Detailed provisions and models/formulas are not avail-
initiate at pits even though those are available. This state- able in codes of practices to predict fatigue strength of cor-
ment concludes that CF crack initiation greatly occurs roded structural details of land‐based structures/onshore
without the formation of pits (i.e. absence of pitting cor- structures.2,3,15 Testing of different full‐scale details in
rosion) and it may occur in any corrosive media.11-14 simulated corrosive environment is complicated espe-
This mechanism is proved by means of a slip‐band prefer- cially in VHCF region (e.g. if the cyclic frequency is 1
ential dissolution model and hydrogen embrittlement Hz, for 108 cycles, it takes more than 3 years to finish a
theory.12-14 As these models are unable to represent CF single test). Therefore, the test results are limited and
crack initiation below yield stress and in the absence of the fatigue endurance of the majority of available results
hydrogen, Zhao et al11 investigated the CF crack initia- are reported in the range from 104 to 106 cycles. The
tion and propagation mechanism of a carbon steel in obtained results are also scattered as the test process is
details by stress controlled fatigue tests and microstruc- subjected to many variables and uncertainties. Those
tural analyses with scan electron microscopy, electron results are not sufficient to warrant variable amplitude
backscatter diffraction and transmission electron micros- fatigue limits of the details. A few past studies recommend
copy. The CF crack initiation and initial propagation the endurance limit to be neglected (i.e. S‐N curve should
mechanisms are highly governed by the peak stress level. be used without cut off limit) to take into account the
In the high cycle fatigue region of steel, a high possibility fatigue strength degradation due to the unexpected local-
of crack initiation in seawater is observed at the austenite ized corrosion (i.e. mild pitting or crevice corrosion) near
grain boundaries than at the pits or from initial cracks.11 the corroded details.15-18 The modified S‐N curves do not
A generalized model which precisely addresses the above match the available fatigue test results of corroded details.
mechanism is lacking in the literature for predicting life Health and Safety Executive (HSE) provisions and
of the defect free (i.e. without cracks or pits) steel struc- guidelines for fatigue strength have been presented for
tural details. Therefore, further research is required to the corroded structural details in offshore structures.19
develop models to predict fatigue strength of the corroded The fatigue strength curve for the welded tubular joints
steel members, joints and connections (i.e. constructional in seawater‐free corrosion was obtained by applying envi-
details) which are subjected to different types of ronmental reduction factor (ERF) to the relevant curve in
corrosion. air. The ERF is determined by the ratio of the number of
As a result, a strain‐life model was proposed based on cycles to fatigue failure in air to that in seawater/marine
the Smith–Watson–Topper model2,3 to simulate the CF environment for tubular joints at the same stress range.
behaviour of metals. The model takes into account the Average ERF value has been determined by very
state of corrosivity of environment, the stress level and scattered CF test results (results varies from 0.8 to 5.2).
the corrosive behaviour of the steel. Further verification The variation of defined test failure criteria (i.e. stopping
of the proposed function is recommended in the same criteria), damage histories of the joints and other environ-
articles as future studies. The procedure for determining mental factors may scatter the obtained fatigue lives. The
the model parameters requires a series of fatigue tests. results were obtained in the range from 6×104 to 2×106
The effect of pitting on the prediction of life is recom- cycles and it does not include either constant amplitude
mended for further studies.2,3 To use the model, transfor- fatigue limit (CAFL) or variable amplitude fatigue limit
mation of stress ranges into strain ranges is required to (VAFL) (i.e. 107 and 108 cycles respectively). A constant
use the model by using cyclic properties of steel and slope is proposed for the whole region of the corroded
fatigue notch factor. Due to mentioned complexities of S‐N curve. The same report has also mentioned the
the model and uncertainties of the model parameters, doubtfulness of the proposed ERF due to lack of data.19
few applications are reported. A stress‐life fatigue The modified corroded curve has the same slope with
strength curve for corroded material is recently pub- the air curve until 107 cycles. This violates CF behaviour
lished.4 The proposed curve depends on corrosion param- of steels, which has negligible difference between fatigue
eter, σ∞,cor, which should be determined by CF tests in lives in corrosive and non‐corrosive environments in the
very high cycle fatigue (VHCF) region in relevant corro- low‐cycle fatigue (LCF) region when compared to VHCF
sive media. The determination of σ∞,cor is complex as region.2-4,19-22
fatigue tests in the VHCF region is extremely time con- To overcome the previously mentioned problems, the
suming. The discussed methods/models are limited for main objective of this paper is to derive a generalized for-
determination of the σ∞,cor in the specimen scale. There- mula to determine the fatigue strength of structural
fore, it is required to have a simple formula/approach to joints/constructional details exposed to corrosive
ADASOORIYA ET AL. 3

media/environment. The concept of the proposed for- propagation is mainly governed by the peak stress level
mula has been studied by means of the fatigue test results and the stress range. Higher probability of crack initiation
of different types of corroded steel specimens in different has been observed in parent austenite grain boundaries in
corrosive environments. The parameters used in the for- HCF region. The crack propagation has been reported
mula are mainly dependent on corrosive environment along the parent austenite grain and ferrite lath bound-
(i.e. urban and marine) and the corresponding construc- aries.11 This phenomenon may be observed even below
tional detail (i.e. detail category), which can easily be the fatigue endurance limit and hence there is no safe
found in the fatigue design codes. The corrosive environ- stress level at which the fatigue life is infinite. The cracks
ment dependent parameters have already been conserva- are usually transgranular4,23 and, therefore, the specified
tively defined based on the corrosion fatigue testing number of cycles are stated as an endurance for the cor-
results of different types of steel in different environ- roded steel by assuming the material will endure. The
ments. The proposed formula can be easily used with degraded fatigue strength/endurance of the corroded steel
the trilinear or bilinear fatigue curves/S‐N curves of depends on the environmental, metallurgical and struc-
detailed categories provided in any fatigue design codes tural factors. These factors have direct effects on the cor-
of steels structures only in the high cycle fatigue (HCF) rosion rate which governs both strength degradation and
region. Initially the mechanism and the concept of the the overall stiffness of the structure.15,24 The fatigue
fatigue strength degradation of the material is discussed endurance/strength of corroded material reaches a
and hence the proposed relation is derived for steels threshold after a certain period of exposure to corrosive
exposed to corrosive media. The verification of the cor- environment. Fatigue test results of steel specimens show
roded parameters is done by comparing CF tests of differ- that the difference between fatigue lives in corrosive and
ent types of steel specimens tested in air, fresh water and non‐corrosive environments LCF region is negligible and
seawater. The relation for the fatigue strength curve is a significant larger difference is observed in VHCF
then developed for corroded constructional details. The region.2,4,20-22
verification of the proposed formula is done by compar-
ing experimental results of many full‐scale tests of cor-
3 | FATIGUE STRENGTH C URVE
roded structural details.
F O R CO R R O DE D ST E E L M A T E RIA L

A fatigue strength formula for corroded steel is presented

2 | M E C H A N I S M OF FA T I G U E based on previously published work.4 Values for corrosive
STRENGTH DEGRADATION OF parameters used in the formula are proposed in this
CO RRODED S T E E L paper to further simplify the previously published
formula. The validity of the proposed parameters is
Gliding can be seen in some of the grains due to alternat-
confirmed by comparing CF test results of different types
ing stresses. When dislocations reach a grain boundary,
of steel.
gliding is ceased.22,23 When the stress is reversed, the
movement of the grains retraces along the gliding plane.
Slight irregularities may prevent the smooth gliding and 3.1 | Formula for high cycle fatigue region
it roughens along original gliding plane. Roughness
restricts the movement and develops another parallel The formula presented in this section exists in a previous
gliding plane. Finally, these disorganized bands of mate- publication4 by the same authors. Based on Basquin's
rial cause separation between gliding planes while pro- law, the S‐N curve formula for corroded steel is pro-
ducing gaps and eventually developing cracks at high posed.4 The concept of the proposed formula is based on
stress ranges (i.e. above the endurance limit) or cease the degradation mechanism discussed in the section 2
the gliding at low stress ranges (i.e. bellow fatigue endur- and the derivation is presented in a previous paper.4
ance limit). Under corrosion, disorganized atoms are The final formula is shown in Eq. (1).
moving along a gliding plane with less activation energy


than without corrosion. The CF crack initiation can occur ðb−cÞ
σ a:cor ¼ σ ′f N cf;LCF N f where c
without the presence of pits and the corrosion‐fatigue    
σ ∞ N f ; FL
cracks are expanded by post reaction of corrosion for ¼ log (1)
σ ∞;cor log N f ;LCF
carbon steel.11 The slip‐bands preferentially dissolution
model and the hydrogen embrittlement theory11-14
describe this mechanism properly above the yield where σa,cor is the fatigue strength of corroded material,
strength. The mechanism of CF crack initiation and which corresponds to the number of cycles to fatigue
4 ADASOORIYA ET AL.

failure, N f . The term σ ′f is the fatigue strength coefficient TABLE 1A Corrosion fatigue limits of steel tested in natural
water4,23,25
and b is the Basquin's exponent. The endurance limit
(i.e. fatigue limit for high‐cycle fatigue) is σ∞ and σ∞,cor Fatigue Fatigue
is the endurance limit for the corroded material, which endurance endurance
corresponds to a specified number of cycles, Nf,FL. When limit for limit for
the stress amplitude is the yield strength, σy, the number uncorroded corroded
material, σ∞ material, σ∞, .
of cycles to fatigue failure of the uncorroded material is σ
Material (MPa) cor (MPa) ∞;cor ∞
Nf,LCF. σ
0.11% C steel, 172 110 0.64
annealed
3.2 | Proposed values of the parameters 0.16%C steel, 241 138 0.57
quenched and
The values of σ ′f , b, σ∞, Nf,FL and Nf,LCF are determined tempered
from the S‐N curve of uncorroded materials. The value 1.09%C steel, 289 158 0.55
of σ∞,cor has to be determined by fatigue testing in the annealed
VHCF region and it is a complicated process. Therefore, 3.5%Ni, 0.3%C 338 200 0.59
this section presents a reasonable accurate relation to steel, annealed
obtain σ∞,cor for structural steels in natural water 0.9%Cr, 0.1%V, 289 152 0.53
(i.e. similar to urban environment) and sea water 0.5%C steel,
(i.e. similar to marine environment). annealed
Revie and Uhlig20,23,25 performed fatigue testing for 13.8%Cr, 0.1%C 345 241 0.70
several grades of medium and low strength corroded steel,
and uncorroded steels. Corrosive process was carried quenched and
out in natural water. The values of σ∞ and σ∞,cor were tempered
determined corresponding to Nf,FL = 107 cycles. The 0.14%C,0.88%Cu 234 149 0.64
ratios σ∞,cor/σ∞ are listed in Table 1a which shows that steel, annealed
the ratio varies in the range of 0.53‐0.70. The mean and 12.9%Cr, 0.11%C 379 264 0.70
coefficient of variation (COV) values of the ratio steel,
are 0.61 and 0.1. Taking into account the 5% failure prob- hardened and
ability (i.e. 95% safe prediction), the design value of the tempered
σ∞,cor/σ∞ ratio can be proposed as 0.5 for conservative
prediction of fatigue strength of medium and low 1/3 of the life in air. This difference was greater when
strength structural steels which are subjected to corrosion the stress level was low.
in natural water media. It is recommended to use mean
and/or design value depending on the importance of the
case studied. 3.3 | Experimental verification of the
Boyer23 and Adasooriya et al.4 presented previous proposed parameters
fatigue test results in both air and saline/seawater. The
related S‐N curves were plotted for several grades of The corrosion fatigue test results of three different steel
medium and low strength steels. The σ∞ and σ∞,cor were types23 and one type of aluminium alloy28 have been
determined corresponding to Nf,FL =107 cycles. The ratios compared with the predicted fatigue lives by the proposed
σ∞,cor/σ∞ are tabulated in Table 1b which shows that the formulae and published by authors.4 The major objective
ratio varies in the range of 0.36‐0.59. The mean and the of this section is to verify the fatigue strength formula
coefficient of variation (COV) values of the ratio are related proposed parameters by comparing the experi-
0.46 and 0.21. Taking into account the 5% failure mental fatigue lives of the corroded specimens of differ-
probability (i.e. 95% safe prediction), the design value of ent materials in different corrosive environments.
σ∞,cor/σ∞ ratio can be proposed as 0.27 for conservative Figure 1 shows the comparison of predicted fatigue
prediction of fatigue strength of medium and low strength with test results of three types of carbon steel:
strength structural steels which are subjected to corrosion the mild steel (MS), cold twisted deformed steel (CTD)
in sea water media. The CF endurance limit of high and quenched and self‐tempered steel (QST). The pre‐
strength steel reduces to 90% when it is exposed to sea corroded specimens were subjected to fatigue testing in
water and hence the ratio is found to be 0.1.26 Yantao 50Hz and stress ratio, R, equal to ‐1.25 The surface of the
et al.27 concluded that the CF life was reduced to 1/2 to uncorroded specimens were polished to a roughness of
ADASOORIYA ET AL. 5

TABLE 1B Corrosion fatigue limits of steel tested in sea water4,23

Fatigue Fatigue
endurance endurance
limit for limit for
uncorroded corroded
material, σ∞ material, σ∞, .
σ
Material (MPa) cor (MPa) ∞;cor ∞
σ
0.14%C,0.88%Cu 234 84 0.36
steel, annealed
0.16%C steel, 241 89 0.37
hardened and
tempered
27%Cr, 0.2%C 309 183 0.59
high
chromium
steel
12.9%Cr, 0.11%C 379 194 0.51
steel,
hardened and
tempered

Rz = 6 μm and enclosed in the natural water for 60 days.

Both the design and the mean values of the σ∞,cor/σ∞
ratio were used for these verifications as shown in Fig. 1.
Figure 2a shows the comparison of predicted S‐N
curves with the test results of E690 steel in air and simu-
lated sea water.11 The E690 steel was newly developed
for marine structures. The chemical composition,
mechanical properties and the test specification were
clearly discussed by Zhao et al..11 The round bar speci-
mens were subjected to stress‐controlled fatigue testing
in simulated sea water (i.e. 3.5wt% NaCl solution) at 1
Hz frequency. The CF test results of medium strength
shaft steel are compared with proposed S‐N formula in
Fig. 2b. The sample preparation, test specifications and
the mechanical properties have been clearly presented else
where.29 The fatigue tests were conducted at R=‐1 at the
frequency of 22 Hz. The corrosive media has been simu-
lated by 3.5 wt% NaCl and pH value 5 solution. The corro- FIGURE 1 Comparison of the proposed S‐N curve with
sion fatigue test results of butt‐welded joints of steel which experimental test carried out in natural water: (a) for mild steel
is commonly used for offshore platforms are compared (MS), (b) for cold twisted deformed steel (CTD), (c) for self
with proposed S‐N formula in Fig. 2c. The sample prepara- tempered steel (QST) [Colour figure can be viewed at
tion, test specifications and mechanical properties have wileyonlinelibrary.com]
been clearly presented else where.27 Four point bending
cyclic loading test was performed in the frequency of 0.5 4 | PROPOSED FATIGUE
Hz and the R = ‐1. The sea water circulation system was ST R E NGT H C UR V E FO R C O R R OD ED
used during the testing and 24.53 wt% NaCl solution was CONSTRUCTIONAL DETAILS
used. The pH value of the solution ranged from 7.5 to
8.5. Both the design and the mean values of σ∞,cor/σ∞ ratio The corroded material fatigue strength curve described
were used for these verifications as shown in Fig. 2. above was further improved to develop S‐N curves for
Figures 1 and 2 show a good match between the fatigue corroded constructional details/detail categories. The pro-
strength curve related proposed values of σ∞,cor/σ∞ and posed curve is obtained by modifying the design fatigue
corrosion fatigue test results of different types of steel. strength curves of categories/detail class of connections.
6 ADASOORIYA ET AL.

between the stress range, Δσ and the corresponding num-

ber of cycles to fatigue failure, NR, presented by Eq. (2) for
the uncorroded (i.e. fatigue design code given details)
curve.


1=m ð−1=mÞ
Δσ ¼ Δσ D N f ;CAFL N R (2)

where ΔσD is the stress range at the fatigue curve slope

changing point which corresponds to the N f ,CAFL cycles.
In some fatigue design codes, ΔσD is defined as the con-
stant amplitude fatigue limit.1,30 The slope of the fatigue
strength curve is ‐1/m. According to the Eurocode30 m
is equal to 3 when Δσ ≥ ΔσD, is equal to 5 when
ΔσD ≥ Δσ > ΔσL and it is infinite when Δσ ≤ ΔσL, where
ΔσL is the fatigue endurance limit of the detail which cor-
responds to N f ,VAFL. This is also defined as the fatigue
cut‐off limit or variable amplitude fatigue limit in some
fatigue design codes.1,30 Some of the fatigue curves of
the details do not have such a cut‐off limit.
According to the verified mechanism and the concept
of corroded fatigue endurance of steel mentioned in sec-
tions 2 and 3, the linear variation in the difference
between the fatigue strengths of the uncorroded and cor-
roded details has been observed as shown in Fig. 3. The
relative difference in log scale is linearly deducted from
the corrosion‐free fatigue strength to obtain the fatigue
strength of the corroded detail. Hence the fatigue strength
range of corroded constructional detail, Δσcor, corre-
sponding to NR, can be derived, if Δσcor ≥ ΔσD,cor,
   
Δσ D N f ;CAFL
logðΔσ cor Þ ¼ logðΔσ Þ − log
Δσ D;cor log N f ;LCF
 
logN R − logN f ;LCF (3)

where ΔσD,cor is the stress range corresponding to N f ,CAFL

cycles at the intersection of the two corroded fatigue
FIGURE 2 Comparison of the proposed S‐N curve with curves. Here, Nf,LCF is the number of cycles to fatigue fail-
experimental test carried out in simulated sea water: (a) for E690 ure of uncorroded details at the intersection point of their
steel, (b) for Shaft steel, (c) for steel butt welded joints [Colour HCF and LCF regions. This has been generally consid-
figure can be viewed at wileyonlinelibrary.com] ered as 104 cycles, which is the lowest value of NR given
in fatigue design codes.1,30 The proposed formula for the
The reduction factors, which represent environmentally fatigue strength of corroded detail categories can be sim-
assisted corrosion fatigue, are utilized for this modifica- plified as,
tion. The linear, bilinear or trilinear design fatigue curves
of details used in different fatigue codes can be modified Δσ cor N R −c
¼ where c
to obtain the proposed fatigue curve for corroded details. Δσ N f ;LCF   
Δσ D N f ;CAFL
¼ log
(4)
4.1 | Derivation of fatigue strength curve Δσ D;cor log N f ;LCF
for corroded details
By substituting Δσ from Eq. (2), the proposed formula
The trilinear fatigue strength curve of uncorroded con- for the fatigue strength for the corroded detail can be
structional detail is presented in Fig. 3.30 The relation obtained as
ADASOORIYA ET AL. 7

FIGURE 3 Schematic representation of fatigue strength curve of uncorroded and corroreded detail categories [Colour figure can be viewed
at wileyonlinelibrary.com]

 1  1
 and the environment dependent parameters of the
−c−
Δσ cor ¼ Δσ D N cf;LCF N f ;CAFL
m NR m (5) detail. These parameters are commonly determined by
full‐scale fatigue tests in the VHCF region and it is
If Δσcor ≤ ΔσD,cor, the fatigue strength of corroded con- costly in terms of both resources and time as the loading
structional detail can be obtained, frequency is very low for full scale testing. The test
results (i.e. experimental fatigue lives) are scattered as
  logΔσ D;cor − logΔσ L;cor the test process is subjected to many variables and
log Δσ D;cor − logðΔσ cor Þ ¼ uncertainties. These ΔσD,cor and ΔσL,cor can be
  logN f ;CAFL − logN f ;VAFL
 logN f ;CAFL − logN R determined by fracture mechanics theories when there
are corrosion pits. Recent investigations of carbon steel
(6)
reveals that CF cracks initiate in any corrosive media
or, due to a different mechanism and the presence of
the pits is not necessary.11-14 Therefore, this section pro-
Δσ cor NR
c poses a reasonably accurate relation to obtain ΔσD,cor
¼ where c and ΔσL,cor for structural steels in fresh water (i.e. simi-
Δσ D;cor N
 f ;CAFL    lar to urban environment) and sea water (i.e. similar to
log Δσ D;cor N f ;CAFL
¼ (7) marine environment) based on the parameters used in
Δσ L;cor log N f ;VAFL
the proposed formula for corroded material as described
When Δσcor ≤ ΔσD,cor, the proposed formula for in section 3.2.
fatigue strength for corroded details can be obtained as, The values of ΔσD,ΔσL,m,N f ,LCF,N f ,CAFL and N f ,
30
VAFL are presented in Table 2 to determine Eurocode
h i
c given fatigue curves. The following relations are obtained
Δσ cor ¼ Δσ D;cor N − c
f ;CAFL N R (8)
to determine the values of ΔσD,cor and ΔσL,cor by interpo-
lations and extrapolations of σ∞,cor/σ∞ ratios correspond-
The parameters c and c depend on the CF endurance
ing to N f ,CAFL and N f ,VAFL.
of the construction details and their determination are
discussed in the following sub section.
0:9 1:33
Δσ D;cor σ ∞;cor Δσ L;cor σ ∞;cor
¼ and ¼ (9)
Δσ D σ∞ Δσ L σ∞
4.2 | Parameters used in the proposed
curve
The corresponding mean and conservative values
The values of ΔσD,ΔσL,m,N f ,LCF,N f ,CAFL and N f ,VAFL are (i.e. design value= mean‐2×standard deviation) are listed
directly obtained from the code providing fatigue in Table 2. Similarly, for DNV code, the following rela-
strength/S‐N curves of constructional details tested in tions can be derived. The corresponding values are listed
air.1,30 The ΔσD,cor and ΔσL,cor are the corrosive state in Table 2.
8 ADASOORIYA ET AL.

TABLE 2 Parameters used in the proposed fatigue strength curve of corroded details

Parameter Constructional details in Eurocode Constructional details in DNV code

N f ,LCF 104 104

N f ,CAFL 5×106 107
N f ,VAFL 108 108
Δσ L 0.549 0.631
Δσ D
Corrosion Marine environment Urban environment Marine environment Urban environment
parameters Mean Conservative Mean Conservative Mean Conservative Mean Conservative
value value value value value value value value
Δσ D;cor 0.497 0.308 0.641 0.536 0.46 0.27 0.61 0.50
Δσ D
Δσ L;cor 0.356 0.175 0.518 0.40 0.356 0.175 0.518 0.40
Δσ L

Δσ D;cor σ ∞;cor Δσ L;cor σ ∞;cor 1:33 parameters for obtaining corroded S‐N curves in any
¼ and ¼ (10) detail category given in any fatigue codes.
Δσ D σ∞ Δσ L σ∞

The proposed values in Table 2 are compared with 4.3 | Experimental verification of the
environmental reduction factors (ERF), which are proposed curve
obtained from experimental results of welded plate and
tubular joints in marine environment. The ERF is the The corrosion fatigue test results of structural details
Δσ D Δσ L exposed to corrosive media are compared with predicted
inverse of the ratios and . The value varies
Δσ D;cor Δσ L;cor fatigue lives by the proposed formula to confirm its valid-
from 1 to 5.2. An average ERF value of 3 has been recom- ity in this section. Full‐scale fatigue testing results of
mended for both 16 mm welded plate and tubular joints rolled plates fabricated from SMA weathering steel2,31
at 106 cycles of endurance in freely corroding environ- are compared with the prediction of Eqs 5 and 8 as shown
ment.19 The proposed values have a good match with in Fig. 4. This steel is commonly used in Japan for brid-
the ERF values. Even though this paper only presents ges, agricultural vehicles, railway wagons, water pipes
the parameters for the Eurocode and DNV codes, the and etc. The tests were performed in tension compression
same procedure can be applied to determine the constant amplitude loading. Corroded and uncorroded

FIGURE 4 Comparison of proposed S‐N curve with fatigue tests of 2 years weathered rolled SMA plates [Colour figure can be viewed at
wileyonlinelibrary.com]
ADASOORIYA ET AL. 9

test results correspond respectively to plates taken after subjected to weathering in air and tests were performed
two years of weathering and unweathered plates. In in a sheltered environment, urban environment parame-
lower stress levels, corroded specimens fatigue endurance ters conservatively were selected from Table 2 for
is closer to uncorroded plates fatigue endurance as there predicting the S‐N curves. The predicted S‐N curve shows
were no severe rust pits formed after two years of a good match with the test results as shown in the Fig. 4.
weathering. The weathered steel is allowed to rust in Fatigue test results of corroded bridge girders extracted
order to form protection coating and the severity of rust from Trolley Bridge32 are compared with the prediction of
pits governs the fatigue endurance.3 Therefore, lower the proposed formula as shown in Fig. 5. The beams were
stress level is not sufficient to increase the stress concen- standard rolled, 4.72 m long and made of carbon steel
tration of rust pits and this is the reason why the same after 85 years of corrosion. The bridge was located urban
fatigue life for weathered and unweathered plates can environment and subjected to moderate corrosivity. The
be observed. The Eurocode detail category 160 represents 22 beams which had not been subjected fatigue in service
fatigue strength of rolled steel plates.30 As the plate was (i.e. operating stress were well below fatigue limit) were

FIGURE 5 Comparison of proposed S‐N curve with fatigue test of corroded carbon steel beams [Colour figure can be viewed at
wileyonlinelibrary.com]

FIGURE 6 Comparison of proposed S‐N curve with full‐scale fatigue test of girth welded pipes under variable amplitude loading [Colour
figure can be viewed at wileyonlinelibrary.com]
10 ADASOORIYA ET AL.

FIGURE 7 Comparison of proposed S‐

N curve with fatigue test of tubular joints
in seawater‐free corrosion, thickness 16
mm: (a). parameters obtained from mean
S‐N curve in air, (b). parameters obtained
from design S‐N curve in air [Colour
figure can be viewed at wileyonlinelibrary.
com]

FIGURE 8 Comparison of proposed S‐N curve with Aghoury's strain‐life model [2,3] predicted fatigue endurances for different corrosion
factors [Colour figure can be viewed at wileyonlinelibrary.com]
ADASOORIYA ET AL. 11

subjected to bending fatigue test. The Eurocode detail cat- amount of corrosion and rusting were observed in girth
egory 160 represents the fatigue strength of rolled carbon welded region only for few pipes and specimens. The
steel.30 The proposed S‐N curve for both marine and S‐N curve for uncorroded girth welded details were taken
urban environments was plotted in Fig. 5 by using Eqs from the same reference and the predicted air curves are
5 and 8. The model parameters were selected from shown in Fig. 6. The model parameters for the proposed
Table 2 to correspond detail category 160. Figure 5 shows S‐N curve of the corroded girth welded pipes were
a good agreement with the proposed formulae for 70% of obtained from the predicted air curve and Table 2. Simi-
the test results. larly, Eqs 5 and 8 are used to predict S‐N curve of the cor-
Figure 6 shows a similar verification with the fatigue roded items as shown in Fig. 6 compared with the
testing results of full‐scale girth welded pipes tested under corresponding experimental results.
variable amplitude loading.33 The resonance fatigue test- HSE offshore technology report19 has provided free
ing technique was used for those full‐scale pipes. Rotating corrosion fatigue strength tubular joints used in offshore
redial force was applied to one end to excite the pipe close structures based on surveying some fatigue test results.
to its first mode of vibration. A bending moment was, The determination of ERF was also mentioned in the
then, generated to induce a variable amplitude stress report. Fatigue tests of tubular joints with both the toe
state. Alternatively, girth welded pipe strips were also ground weld and as welded were considered for the test.
tested by ordinarily hydraulic test machine. Considerable Proposed S‐N curves for the marine environment were

FIGURE 9 Comparison of Aghoury's

strain‐life model [2,3] predicted fatigue
endurances and experimental fatigue lives
of sheltered steel beams tested in moist
saltwater environment with: (a). proposed
S‐N curve plots, (b). proposed S‐N curve
formula predicted fatigue lives [Colour
figure can be viewed at wileyonlinelibrary.
com]
12 ADASOORIYA ET AL.

plotted using the formulae given in Eqs 5 and 8 for thick- member and joint scale. This is the main reason of having
ness 12mm tubular joints as shown in Fig. 7. The proposed a good agreement with the full‐scale fatigue test results.
S‐N curve in Fig. 7a was based on the parameters of mean The values of the curve parameters have been proposed
S‐N curve of tested tubular joints in air while Fig. 7b was only for detail categories given by Eurocode and DNV
based on the design curve. The marine environment code. The same procedure can be utilized with the values
model parameters were taken from Table 2 and hence proposed for the corroded steel to determine the model
both the design and the mean S‐N curves are plotted for parameters for any detail category given in any code.
corroded tubular joints. Measured hot spot stresses were The main advantage of the proposed formula is that it
used for this comparison. The S‐N curves has a very good requires only the fatigue strength curve of the construc-
agreement with the corresponding experimental fatigue tional detail in air, which is given in relevant codes, to
lives. The previously proposed free corrosion (FC) fatigue predict the corresponding S‐N curve in a corrosive media.
curve19 seems to be over conservative when the endurance
is lower than 106 cycles. However, it is doubtful in VHCF
region. The previous FC curve intersects the new curves NOMENCLATURE
between 105 and 106, where the test results were used for
ERF determination. This indirectly indicates that the b Basquin's exponent
determined model parameters shown in Table 2 have a m negative inverse slope of the S‐N curve
good agreement with the ERF values.19 Nf number of cycles to fatigue failure
As discussed in the introduction, Aghoury proposed a Nf,FL endurance number of cycles
strain‐life model based on the Smith–Watson–Topper Nf,LCF number of cycles to fatigue failure of the
model2,3 to simulate the CF behaviour of metals. The uncorroded materials at the yield strength
S‐N curves predicted by that model for three different cor- N f ,CAFL number of cycles at constant amplitude fatigue
rosion rates are compared in Fig. 8 with the correspond- limit
ing S‐N curves predicted by Eqs 5 and 8 for marine and N f ,VAFL number of cycles at variable amplitude fatigue
urban environments. The S‐N curve for air was predicted limit
based on corrosion factor zero. The figure shows good NR number of cycles to fatigue failure
agreement between proposed models. Δσ stress range
Aghoury2,3 compared his model estimated fatigue lives Δσcor fatigue strength range of corroded construc-
with the experimental fatigue behaviour of the corroded tional detail
rolled beams of A588 steel. The beams were boldly ΔσD stress range at constant amplitude fatigue limit
exposed for 67 months under metal deck in moist salt ΔσD,cor stress range at intersecting points of two slopes
water (i.e. sheltered in marine environment). The fatigue of corroded fatigue at number of cycles at
lives predicted by both models are compared with exper- constant amplitide fatigue limit
imental lives in Fig. 9. The figures confirm the validity of ΔσL stress range at variable amplitude fatigue limit
Eqs 5 and 8 and its parameters listed in Table 2. ΔσL,cor stress range at number of cycles ar variable
amplitude fatigue limit
σa,cor fatigue strength of corroded material
5 | CONCLUSIONS σy yield strength
σ ′f fatigue strength coefficient
The comparisons of the predicted fatigue curves with σ∞ endurance limit (i.e. fatigue limit for high‐
fatigue test results of corroded steel in different corrosive cycle fatigue)
media conclude the validity of a proposed formula and σ∞,cor endurance limit for corroded material.
the validity of the proposed mean and conservative values
for parameters used in the formula. The proposed curve
does not require any material parameter or corrosive
media specific parameter except the fatigue strength ORCID
curve (i.e. S‐N curve) obtained in air. N.D. Adasooriya https://orcid.org/0000-0001-8412-8009
The S‐N curves predicted by the derived formula are
in a very good agreement with experimentally obtained
fatigue lives of corroded constructional details considered RE FER EN CES
in the present study. The physics behind the CF has been 1. DNV GL Recommended practices. Fatigue design of offshore steel
studied from the literature in a micro structural level, structures DNVGL‐RP‐0005. Norway: Det Norske Veritas; April;
specimen scale and finally incorporated into structural 2016.
ADASOORIYA ET AL. 13

2. El Aghoury I. Numerical tool for fatigue life prediction of cor- measurements in riveted, bolted and welded structures: Sustain-
roded steel riveted connections using various damage models, able Bridges‐Background document SB3.4. Sweden: Digital
PhD thesis. Canada, Montreal: Concordia University; 2012. Vetenskapliga Arkivet; 2007.
3. El Aghoury I, Galal K. Corrosion‐fatigue strain‐life model for 19. HSE OTH 92 390. Background to new fatigue guidance for steel
steel bridge girders under various weathering conditions. J joints and connections in offshore structures, offshore technology
Struct Eng. 2014;140(6):Article number 04014026. report. UK: Health Safety Executives; 1999.
4. Adasooriya ND, Hemmingsen T, Pavlou D. Fatigue strength deg- 20. Revie RW, Uhlig HH. Corrosion and Corrosion Control, An
radation of metals in corrosive environments. Proceedings of first Introduction to Corrosion Science and Engineering, 4th edn. US:
conference of computational methods in offshore technology‐ Wiley and Sons; 2008.
COTech 2017, IOP Conf Ser: Mater Sci Eng: IOP Publishing; 21. Glaser W, Wright LG. Mechanically assisted degradation. ASM
Stavanger, Norway: 2017: Article number 276 012039. Handbooks ASM Internationals. 1992;1992:137‐144.
5. Roberge PR. Handbook of Corrosion Engineering. 1st ed. New 22. Gangloff RP. Environmental cracking ‐ corrosion fatigue. In:
York: McGraw Hill; 2000. Baboian R, Dean SW Jr, Hack HP, Hibner EL, Scully JR, eds.
6. Mikkelsen O, Rege K, Hemmingsen T, Pavlou D. Numerical Corrosion Tests and Standards Manual. US: ASTM Interna-
estimation of the stop holes‐induced fatigue crack growth retar- tionals; 2005.
dation in offshore structures taking into account the corrosion 23. Boyer HE. Atlas of Fatigue Curves. USA: ASM International;
effect. Proceeding of International Society of Offshore and Polar 1986.
Engineers conference series. San Francisco: American Society of 24. Yasser Y, Jeom KM. Ultimate strength reliability analysis of cor-
Mechanical Engineers, 2007: ISOPE‐I‐17‐552: 451‐458. roded steel‐box girder bridges. Thin‐Walled Struct. 2011;49:
7. Hoeppner DW, Chandrasekaran V, Taylor AMH. Review of 157‐166.
pitting corrosion fatigue models. Proceedings of ICAF Interna- 25. Bandara CS. Fatigue damage assessment of steel structures and
tional Conference. Melbourne, Australia: 1995: 1‐25. components, Improvements in stress‐life approach. Germany:
8. Acuña N, González‐Sánchez J, Ku‐Basulto G, Domınguez L. LAP Lambert Academic Publishing; 2015.
Analysis of the stress intensity factor around corrosion pits 26. Luiz OAA. Corrosion, Chapter 7, Machinery Failure Analysis
developed on structures subjected to mixed loading. Scr Mater. Handbook. Texas: Gulf Publishing Company; 2006:83‐99.
2006;55:363‐366.
27. Li Y, Hou B. Corrosion fatigue of welded joints of steel for
9. Ian T. The effect of notches and pits on corrosion fatigue marine platform. Indian J Eng Mat Sci. 2003;13:467‐470.
strength, PhD thesis. UK: Sheffield Hallam University; 2004.
28. Matthew W, Paul DE, Huseyn O, Murat T. The effect of surface
10. Sharifi Y, Rahgozar R. Fatigue notch factor in steel bridges due corrosion damage on the fatigue life of 6061‐T6 aluminum alloy
to corrosion. Arch Civ Mech Eng. 2009;9(4):75‐83. extrusions. Mater Sci Eng A. 2017;690:427‐432.
11. Zhao T, Liu Z, Du C, Dai C, Li X, Bowei Z. Corrosion fatigue 29. Li S‐X, Akid R. Corrosion fatigue life prediction of a steel shaft
crack initiation and initial propagation mechanism of E690 steel material in sea water. Eng Fail Anal. 2013;34:324‐334.
in simulated sea water. Mat Sci Eng A. 2017;708:181‐192.
30. NS EN 1993‐1‐9. Eurocode 3: Design of steel structures ‐ Part 1‐9:
12. Vucko F, Bosch C, Delafosse D. Effects of cyclic plastic strain on Fatigue. Belgium: European Committee for Standardization;
hydrogen environment assisted cracking in high‐strength steel. 2005.
International hydrogen conference (IHC 2012): hydrogen‐
31. Kunihiro T, Inove K, Fukuda T. Atmospheric exposure study of
materials interactions. WY, United States: American Society of
weathering steel. Research Lab. Report Br. 71‐08. Tokyo, Japan:
Mechanical Engineering; 2012.
Ministry of Construction; 1972.
13. Vucko F, Bosch C, Aoufi A, Delafosse D. Palladium coating on
32. Albrecht P, Shabshab CF, Wulin L. Remaining fatigue strength of
quenched‐tempered martensitic steel for hydrogen electrochemical
corroded steel Beams, IABSE reports. International Association
permeation tests. Technical Reports‐ ENSMSE‐SMS‐2014‐01.
of Bridge and Structural Engineering publications; 1990: 71‐84.
France: EMSE‐00951142; 2014.
33. Zhang YH, Maddox SJ. Fatigue testing of full‐scale girth welded
14. Wang R. Corrosion fatigue of metal materials. Xi'an: Press of
pipes under variable amplitude loading. J Offshore Mech Arctic
Northwestern Polytechnical University; 2001.
Eng. 2014;136(2):Article number 02140‐1.
15. Adasooriya ND, Siriwardane SC. Remaining fatigue life estima-
tion of corroded bridge members. Fatigue Fract Eng Mater
Struct. 2014;37(6):603‐622.
16. Wahab MA, Sakano M. Corrosion and biaxial fatigue of welded
structures. J Mat Proc Tech 3. 2003;143–144:410‐415. How to cite this article: Adasooriya ND, Pavlou
17. Nguyen KT, Garbatov Y, Soares CG. Spectral fatigue damage
D, Hemmingsen T. Fatigue strength degradation of
assessment of tanker deck structural detail subjected to time‐ corroded structural details: A formula for S‐N
dependent corrosion. Int J Fatigue. 2013;48:147‐155. curve. Fatigue Fract Eng Mater Struct. 2019;1–13.
18. Gerhard S, Christian K, Bertram K, Hensen W. Condition assess- https://doi.org/10.1111/ffe.13156
ment and inspection of steel railway bridges, including stress