materials

Article
Prediction of Corrosive Fatigue Life of Submarine
Pipelines of API 5L X56 Steel Materials
Xudong Gao 1 , Yongbo Shao 1, *, Liyuan Xie 1 , Yamin Wang 2 and Dongping Yang 3
1 School of Mechatronic Engineering, Southwest Petroleum University, Chengdu 610500, China;
gxdgao193@163.com (X.G.); 18380156106@163.com (L.X.)
2 Centre for Infrastructure Engineering, Western Sydney University, Penrith, NSW 2751, Australia;
Y.Wang5@westernsydney.edu.au
3 Technology Inspection Center, China Petroleum & Chemical Corporation, Dongying 257062, China;
ydp17@126.com
* Correspondence: ybshao@swpu.edu.cn; Tel.: +86-28-83037601




Received: 28 January 2019; Accepted: 22 March 2019; Published: 28 March 2019


Abstract: Corrosive fatigue failure of submarine pipelines is very common because the pipeline is
immersed in a sea environment. In Bohai sea, many old pipelines are made of API 5L X56 steel
materials, and it is necessary to provide an accurate method for predicting the residual life of these
pipelines. As Paris law has been proven to be reliable in predicting the fatigue crack growth in metal
materials, the two constants in Paris law for API 5L X56 steel materials are obtained by using a new
proposed shape factor based on the analysis of experimental data measured from fatigue tests on
compact tension specimens immersed in the water of Bohai sea. The results of the newly proposed
shape factor show that, for a given stress intensity factor range (∆K), the fatigue crack growth rate
(da/dN) in seawater is 1.6 times of that that in air. With the increase of fatigue crack growth rate, the
influence of seawater on corrosive fatigue decreases gradually. Thereafter, a finite element model for
analyzing the stress intensity factor of fatigue crack in pipelines is built, and the corrosive fatigue life
of a submarine pipeline is then predicted according to the Paris law. To verify the presented method,
the fatigue crack growth (FCG) behavior of an API 5L X56 pipeline with an initial crack under cyclic
load is tested. Comparison between the prediction and the tested result indicates that the presented
method is effective in evaluating the corrosive fatigue life of API 5L X56 pipelines.

Keywords: API 5L X56 steel material; submarine pipelines; corrosive fatigue crack growth; Paris law
constants; stress intensity factor

1. Introduction
Submarine pipeline is regarded as the lifeline of offshore oil and gas transportation. As the
offshore oil industry boomed, the frequency of pipeline leaks increased. The submarine pipeline
is used in a harsh marine environment for a long time. It is vulnerable to failure due to complex
alternating stress and external factors [1–3]. The phenomenon of submarine pipeline hanging is
inevitable due to the scouring of sea water, and the hanging pipeline is subjected to alternating load
under the action of sea wave and it is, hence, sensitive to fatigue failure [4]. Therefore, it is of great
significance to study the corrosive fatigue of submarine pipelines in service to prevent the failure.
The crack growth rate under a given load depends on several factors, including crack length
and geometry, temperature, environment, test frequency, and the microstructure of the material.
Petit et al. [5,6] have investigated the detrimental effects of atmospheric conditions on fatigue crack
propagation in steel, Al and Ti-based alloys. The deleterious effects of water vapor were analyzed
based on experiments in a controlled partial pressure atmosphere, and modeling of intrinsic and

Materials 2019, 12, 1031; doi:10.3390/ma12071031 www.mdpi.com/journal/materials

Materials 2019, 12, 1031 2 of 19

environmentally assisted crack growth is proposed. At present, many researchers have studied the
influence of different environmental conditions on fatigue crack growth rate for different pipelines.
Javidi et al. [7] studied the stress corrosion cracking behavior of API 5L X52 steel under the near-neutral
and high pH conditions using slow strain rate tests and electrochemical methods. The results revealed
that the anodic dissolution at a crack tip was the dominant mechanism at near-neutral pH condition.
While at high pH medium, the hydrogen-based mechanism was dominant. Cheng et al. [8] established
a two-component model considering anode dissolution and hydrogen embrittlement based on the
fracture mechanics theory, and they applied the model for API 5L X65 pipeline steel. The results
showed that the model can control the shape of the simulated crack growth curve and the fatigue crack
growth (FCG) rate. As the global sustainable development strategy becomes the mainstream, offshore
natural gas, which is considered as a clean energy, has attracted the attention of various countries.
Therefore, many researchers have studied the FCG rate and the cracking behavior of pipeline steels in
different gas environments [9–14]. Wang et al. [9] studied the FCG rate of the base metal and the heat
affected zone of API 5L X70 pipeline steel under a hydrogen sulphide environment, and they found that
the existence of H2 S greatly accelerated the FCG rate. Zheng et al. [10] also obtained a similar law in the
study of API 5L X56 pipeline steel under an H2 S environment. In addition, it is found that the presence
of gaseous hydrogen also has a great impact on the FCG rate of steel, which exceeded an order of
magnitude larger than that of the FCG rate under laboratory air environment [11–14]. Amaro et al. [11]
studied API 5L X100 pipeline steel in high-pressure gaseous hydrogen environments, and they found
that the FCG response in hydrogen at da/dN < 3 × 10−4 mm/cycle is primarily affected by the hydrogen
concentration within the fatigue process zone, resulting in a hydrogen-dominated mechanism, and the
FCG response in hydrogen at da/dN < 3 × 10−4 mm/cycle results from fatigue-dominated mechanisms.
API 5L X56 pipeline steel is now used in the submarine pipelines in the Bohai sea of
China. Corresponding studies on X56 submarine pipeline steel have also been carried out in the
literature [10,15,16]. Yi et al. [15] simulated stress corrosion cracking and hydrogen permeation
behavior of X56 pipeline steel under atmospheric environment containing SO2 , and the results show
that hydrogen ion concentration decreases with time, indicating that the hydrogen concentration on the
surface decreases with the formation of corrosion rust. As the passivation film dissolves, the hydrogen
permeation rate increases and the stress corrosion cracking sensitivity of API 5L X56 increases with
the increase of SO2 concentration. Zhu et al. [16] studied the permeation action of API 5L X56 steel
in the sea mud with and without sulfate-reducing bacteria at corrosion and cathodic potential were
studied with an improved Devanathan–Stachurski electrolytic cell. The above studies on the API
5L X56 submarine pipeline are all influenced by a single factor, and the fatigue remaining life of the
pipeline with initial cracks in service is not analyzed.
Structural cracking is a major problem for large infrastructures including dams, bridges,
submarine pipelines and offshore wind turbine foundations. At the same time, reliability analysis
has become a concern of researchers. Based on the probability theory, reliability analysis of fatigue
fracture for engineering structures and components has been conducted, and proposed several analysis
models. Commonly used techniques include Monte Carlo simulation method and first-order reliability
method [17]. In Monte Carlo simulation, the random variable is formed by its distribution function.
For each Monte Carlo simulation sample, the sensitivity of each stress intensity factor (SIF) to the
random variable is calculated and compared with the obtained SIF to evaluate whether failure will
occur. Based on the above method, the fatigue life of the pipeline is analyzed [18–20]. An S-N curve
based on damage mechanics is another way to predict fatigue life. The S-N curve is used to predict the
fatigue life of crack in suspended pipeline. Wormsen et al. [21] carried out tests on low alloy machined
steel forgings and analyzed S-N curves in air and seawater environments with cathodic protection
measures. They provided specific plans for fatigue tests of submarine pipelines. Cunha et al. [22]
introduced an S-N curve from tensile test material properties, and they proposed a new algorithm to
evaluate the fatigue life of dented pipelines. Hong et al. [23] collected ASTM standard specimens from
the API 5L X42 gas pipeline, and FCG tests were subsequently performed. Additionally, S-N curve of
Materials 2019, 12, 1031 3 of 19

each specimen was estimated to verify the accuracy of the prediction model. Dong et al. [24] analyzed
the weld data recorded in a large number of literatures from 1947 to present by means of the new
mesh-insensitive structural stress method and the associated master S-N curve approach. According
to the characteristics of fracture mechanic materials, Lukács et al. [25,26] evaluated and calculated the
reliability of different pipeline structural unit models, during which different crack geometries were
assumed. Meanwhile, the stress intensity factors and safety factors were also determined for all cases.
Both methods are based on empirical formulas, model analysis and finite element simulation to obtain
the fatigue life. It is difficult to verify the accuracy of the predicted results in practical application.
The methods for predicting fatigue life of metal materials or structures can be divided into S-N
curves based on damage mechanics and Paris law based on fracture mechanics. At present, Paris
law-based on fracture mechanics is used to predict the corrosion fatigue life of cracked parts. The main
reason is that S-N curve based on damage mechanics is obtained by applying alternating loads on
smooth and non-cracked specimens. In fact, defects are inevitable when metal is processed, and so
the predicted life from S-N curve is dangerous. Paris law based on linear elastic fracture mechanics
analyzes the remaining life of cracked components, which is more instructive for practical engineering.
Moreover, Paris law can reflect the relationship between fatigue crack growth rate (da/dN) and stress
intensity factor range (∆K). It is only necessary to measure the Paris law constants (C and m) in a
certain environment through FCG tests. In this study, finite element analysis is carried out to analyze
the stress intensity factors under different crack depths in API 5L X56 submarine pipelines. The fatigue
crack growth life is then predicted from Paris law. To verify the accuracy of the prediction method, a
full-scale size API 5L X56 pipeline was tested in a fatigue loading condition. The effect of seawater on
the crack growth rate of API 5L X56 steel was also studied under a laboratory environment, and a new
method for predicting the fatigue life of API 5L X56 pipeline life was proposed.

2. Test on Fatigue Crack Propagation of API 5L X56 Steel Material

2.1. Mechanical Properties of API 5L X56 Steel Material

Fatigue crack growth test is based on Paris law to establish the relationship between the FCG rate
(da/dN) and the stress intensity factor range (∆K). For a linear elastic steel specimen, the SIF is linearly
related to the load and depends on the geometry and size of the steel specimen and the crack. Coupons
test and Charpy pendulum impact test are both carried out to obtain the mechanical properties of API
5L X56 steel materials including the elastic modulus (E), the yield strength (σy ), Poisson’s ratio (ν), and
the fracture toughness (KIC ). The impact energy is measured according to GB/T229-2007 standard for
“Metallic materials—Charpy pendulum impact test method” [27], and the experimental results are
listed in Table 1.

Table 1. Impact energy for API 5L X56 steel material.

Test ID Test Materials Test Temperature (◦ C) Impact Energy (J)

1 API 5L X56 20 158.27
2 API 5L X56 20 69.76
3 API 5L X56 20 124.66
4 API 5L X56 20 97.08
Mean Value - - 112.44

The fracture toughness of API 5L X56 steel material is estimated according to Equation (1)
provided in BS7910-2013 standard for “Guide to methods for assessing the acceptability of flaws in
metallic structures” [28] as follow

CV −20)(25/B)0.25 ] + 20,
p
KIC = [(12 (1)
Materials 2019, 12, x FOR PEER REVIEW 4 of 18

Table 2. Mechanical properties of API 5L X56 steel material.

Materials 2019, 12, 1031 4 of 19
Material Elastic Modulus (GPa) Yield Strength (MPa) Poisson's Ratio 𝑲ΙC (𝐌𝐏𝐚 ⋅ √𝐦)
API 5L X56 200 340 0.3 188
where CV is the Charpy impact energy, B is the thickness of the specimens. Based on the data in Table 1,
2.2.
theDesign of C(T)
fracture Specimens
toughness of API 5L X56 steel material can be calculated. Combined with the tensile test
of metal materials,
The fatigue thegrowth
crack basic mechanical
test is the properties of API 5L
most commonly X56method
used steel material arethe
to study listed
FCGin rates
Tableof
2.
various homogeneous metal materials. The material used in the C(T) specimens is API 5L X56 steel
Table 2. Mechanical properties of API 5L X56 steel material.
material which is obtained from the submarine pipelines in Bohai sea. Standard compact tension,
√
C(T), specimens
Material are extracted
Elastic Modulusfrom submarine
(GPa) pipeline(MPa)
Yield Strength with a diameter of 219 mm
Poisson’s Ratio and ·a wall
KIC (MPa m)
thickness of 14
API 5L X56 mm. Figure 1a shows
200 that the plane direction
340 of the crack is perpendicular
0.3 to the
188 axial
direction and parallel to the axial direction respectively. Li et al. [29] carried out a load-controlled
three-point
2.2. Designbending test on API 5L X80 steel material. The results show that the FCG rate of the
of C(T) Specimens
specimens from different directions are basically the same. In this study, the sampling direction of
The fatigue crack growth test is the most commonly used method to study the FCG rates of
C(T) specimens is no longer distinguished.
various homogeneous metal materials. The material used in the C(T) specimens is API 5L X56 steel
C(T) specimens are designed according to GB/T 6398-2000 standard for “Metallic
material which is obtained from the submarine pipelines in Bohai sea. Standard compact tension, C(T),
materials―Fatigue testing―Fatigue crack growth method” [30] and ASTM E647-13 [31]. The wall
specimens are extracted from submarine pipeline with a diameter of 219 mm and a wall thickness of
thickness of API 5L X56 pipeline is 14mm. Under the premise of guaranteeing the machining
14 mm. Figure 1a shows that the plane direction of the crack is perpendicular to the axial direction
allowance, the thickness of C(T) specimens is selected as B=10mm and the width is selected as W=40
andFigure
mm. parallel
1btois the axial direction
a standard respectively.
C(T) specimen used inLifatigue
et al. [29] carried
crack outtest.
growth a load-controlled three-point
Overall, three specimens
bending
are testthe
tested on onSTM793
API 5L X80 steeltesting
fatigue material. The results
machine, and theshow that the
specimen FCG rate
marked withofX56-A-1
the specimens from
represents
different directions are basically the same. In this study, the sampling direction of
the one tested in an air environment while the other two specimens (marked with X56-S-1 and X56- C(T) specimens is no
longer
S-2) distinguished.
are tested in Bohai sea environment.

(a) (b)
Figure
Figure1.1.
Extraction location
Extraction and
location dimensional
and drawing
dimensional ofof
drawing C(T) specimens:
C(T) (a)(a)
specimens: Crack plane
Crack orientation;
plane orientation;
(b)
(b)C(T)
C(T)specimen.
specimen.

C(T)
The FCGspecimens
tests areare designed
carried out onaccording
the STM793 to GB/T 6398-2000
fatigue standardwhich
testing machine, for “Metallic materials—
is composed of a
Fatigue testing—Fatigue crack growth method” [30] and ASTM E647-13 [31]. The wall
loading device and control system. The loading device is used to provide cyclic load, while the control thickness of API
5L X56ispipeline
system used foris 14mm. Underloading
controlling the premise of guaranteeing
process and collecting the machining
testing data.allowance, the thickness
All specimens are
of C(T) specimens is selected as B = 10 mm and the width is selected as W
fabricated to produce a sharp crack with a certain length of approximately 3 mm from the machined = 40 mm. Figure 1b is a
standard C(T) specimen used in fatigue crack growth test. Overall, three specimens
V-notch using the K-decreasing technique to ensure that the K equation is not affected by the initial are tested on the
STM793
notch fatigue
shape. The testing machine,factor
stress-intensity and the specimen
range marked
is stepped withatX56-A-1
down represents
a constant the one
rate as the tested in
precracked
length increases, and the final ΔK equals to the minimum ΔK at the end of the precracking.in
an air environment while the other two specimens (marked with X56-S-1 and X56-S-2) are tested
Bohai sea environment.
Precracking in air at a frequency of 10 Hz at room temperature, and the crack length at the end is
recordedTheasFCGthe tests
initialare carried
crack out aon
length, the STM793 fatigue testing machine, which is composed of a
0 . Thereafter, the constant amplitude loading stage for fatigue
loading device and control system. The loading device
crack growth test is followed. The test is terminated is used
when the FCGto provide cyclic load,
rate exceeded 0.01 while the control
mm/cycle, and
system
the crackislength
used for controlling
at this loading
time is called theprocess and collecting
final crack length, af .testing data. All specimens are fabricated
to produce a sharp crack with a certain length of approximately 3 mm from the machined V-notch
using
2.3. the K-decreasing
Parameter technique
Selection in Fatigue Testto ensure that the K equation is not affected by the initial notch
shape. The stress-intensity factor range is stepped down at a constant rate as the precracked length
increases, and the final ∆K equals to the minimum ∆K at the end of the precracking. Precracking in
air at a frequency of 10 Hz at room temperature, and the crack length at the end is recorded as the
initial crack length, a0 . Thereafter, the constant amplitude loading stage for fatigue crack growth test is
Materials 2019, 12, 1031 5 of 19

followed. The test is terminated when the FCG rate exceeded 0.01 mm/cycle, and the crack length at
this time is called the final crack length, af .

2.3. Parameter Selection in Fatigue Test

The tests are carried out in both air and seawater conditions, and the seawater used in the test is
taken from Bohai sea of China where the submarine pipeline is located. C(T) specimens tested in a
seawater environment are soaked in seawater for more than 24 hours before precracking. All FCG
tests in air and in seawater environments are loaded in a sinusoidal wave form with a stress ratio of
0.1 and a constant amplitude. It is noted here that, even though the applied load ratio experienced
by the submarine pipeline may be negative, a previous study in the literatures [32–34] has found
that the FCG rates at R-ratio of −1.0 is similar or lower than that at R = 0.1 for steel in air and in
seawater environments. Therefore, the R-ratio of 0.1 used in this study to characterize the FCG rate is
relatively conservative. Previous study indicated that the effect of loading frequency, f , on the fatigue
crack growth tests varies from environment to environment [35]. To simulate the actual seawater
environment as much as possible and to consider the efficiency problem, the loading frequency in this
seawater test environment is 0.5 Hz. Previous studies found that the influence of loading frequency on
the FCG rate of steel in the air environment is negligible [34,36,37]. Therefore, FCG test in the air is
performed at a frequency of 10 Hz at room temperature. When selecting the maximum load, Pmax , the
load should not be too large to guarantee that the crack tip is in the small yielding range to meet the
requirements of Linear and Elastic Fracture Mechanics (LEFM). Simultaneously, the load should not be
too small to ensure that ∆K is greater than the stress intensity factor threshold, 4Kth . In addition, the
selection of the load is also within the effective control range of the fatigue testing machine. Testing
parameters such as loading frequency, f , maximum loading load, Pmax and load ratio, R, are shown in
Table 3.

Table 3. Parameters in fatigue crack propagation test.

Test ID Environment a0 (mm) af (mm) f (Hz) R Pmax (kN)

X56-A-1(CT) Air 12.98 25.03 10 0.1 11
X56-S-1(CT) Seawater 13.02 24.98 0.5 0.1 11
X56-S-2(CT) Seawater 12.95 22.51 0.5 0.1 9.5
Submarine Pipeline Specimen Seawater 5.5/52.1 13.9/60.2 0.5 0.1 235

2.4. Crack Growth Monitoring

In the FCG tests, a flexible method is used to measure the crack length development. Different
monitoring methods are used in different environments. In the air environment, the crack propagation
is monitored by using the Crack Opening Displacement (COD) gauge and the Back Face Strain (BFS)
method. Figure 2a shows the position of the COD gauge and strain gauge. Previous study in the
literature [35] verified the accuracy of the COD and BFS method for monitoring crack growth, and
the results showed that the BFS method can be used to replace the COD method. In this study, the
test data of API 5L X56 pipeline steel collected from two monitoring methods are compared to verify
the reliability of those two methods. Due to the difficulty in installing COD gauges in a seawater
environment, only the BFS method is used to monitor FCG in this condition. For the fatigue test in
seawater, the strain gauge surfaces are coated with a layer of AZ-710 anti-corrosive glue to protect
the strain gauge from the effect of corrosion. After the glue is solidified, a layer of waterproof tape is
applied to achieve double protection, as shown in Figure 2b. In addition, it should be noted that the
load frequency is up to 10 Hz. Therefore, the TST5912 dynamic signal test and analysis system is used
for data acquisition. During the cyclic loading process, the acquisition frequency of the dynamic strain
gauge in seawater is set to 100 Hz.
Materials 2019, 12, 1031 6 of 19

Materials 2019, 12, x FOR PEER REVIEW 6 of 18

(a) (b)
Figure
Figure 2. Fatigue
2. Fatigue crack
crack propagation
propagation monitoring
monitoring method:
method: (a) Monitoring
(a) Monitoring of crack
of crack growth
growth in air
in air
environment;
environment; (b) Monitoring
(b) Monitoring of crack
of crack growth
growth in seawater
in seawater environment.
environment.

2.5. 2.5.
Fatigue Crack
Fatigue Propagation
Crack Propagation
ParisParis
law law
is commonly
is commonlyusedused
to describe the fatigue
to describe crack
the fatigue propagation
crack relationship,
propagation i.e., i.e.,
relationship, the the
relationship between
relationship the the
between FCGFCG
rate rate
(da/dN) and and
(da/dN) the stress intensity
the stress factor
intensity range
factor (∆K).(ΔK).
range The The
stressstress
intensity factor range can be defined as follows:
intensity factor range can be defined as follows:

∆K = K
∆K =K
max −max - K, min ,
Kmin (2) (2)
where K is the stress intensity factor, which is a parameter in linear and elastic fracture mechanics
where K is the stress intensity factor, which is a parameter in linear and elastic fracture mechanics and
and it is defined as follows:
it is defined as follows:
√
K = Fσ Fσ√πa,
K = πa, (3) (3)
where
where σ isnominal
σ is the the nominal
stress stress (the applied
(the applied stress), stress), a is the
a is the crack crackand
length, length,
F is aand F is a dimensionless
dimensionless shape
shapeItfunction.
function. It is
is indicated indicated
clearly in theclearly in the GB/T
GB/T 6398-2000 6398-2000
standard standard
that the thatofthe
SIF range C(T)SIF range ofisC(T)
specimens
specimens
calculated fromisthe
calculated
followingfrom the following equation:
equation:
ΔP a (4)
∆P a g( ),
∆K = ΔK =1/2 g (1/2 ), W (4)
BW BW W
where P is the applied load at both ends of the C(T) specimen, B is the specimen’s thickness, and W
where P is the applied load at both ends of the C(T) specimen, B is the specimen’s thickness, and W
is the specimen’s width. The solution of the shape function for a standard C(T) specimen is given in
is the specimen’s width. The solution of the shape function for a standard C(T) specimen is given in
References [30,31] as follows:
References [30,31] as follows:
a (2 + α)(0.886 + 4.64α -13.32α2 + 14.72α3 - 5.6α
4 ) (5)
g(
(2 W ) =
+ α)(0.886 + 4.64α − 13.32α 2 + 14.72α3 − 5.6α4 ,
a (1-α) 3/2
g( ) = , (5)
W (1 − α)3/2
where α=a/W. Equations (2) and (4) derive the equation of the SIF range as follows:
where α = a/W. Equations (2) and (4) derive the equation of the SIF 1 range a as follows: (6)
∆K = Kmax -Kmin = (Pmax -Pmin ) 1/2
g( ),
BW W
1 a
∆K = Kmax −Kmin = ( P max − Pmin ) g ( ), (6)
where da/dN is plotted against ΔK in log–log axes, and general BW 1/2threeW regions can be observed in FCG
trend. The three regions are named as the fatigue crack initiation (region Ⅰ), the stable crack growth
where da/dN
stage is plotted
(region Ⅱ), andagainst ∆K in log–log
the unstable fracture axes, and(region
stage generalⅢ).three
In regions
region Ⅰ, canthe
be observed in FCG
SIF is lower than the
trend. The intensity
stress three regions
factorarethreshold,
named as △ theKthfatigue
, and crack
there initiation
is no crack (region
growth.I), the stableΔK
When crack growth
exceeds stress
stageintensity
(region II), andthreshold,
factor the unstableit isfracture
in regionstage (region
II and III). Inbegins
the crack regiontoI, grow.
the SIFItisislower
worththan the stress
pointing out that
intensity factorgrowth
the crack threshold, ∆Ktha, and
follows power there is no
law, crack growth.
namely the so-called ∆K exceeds
WhenParis law, as stress
givenintensity factor
in Equation (7) in
threshold, it is in region II and the crack begins to grow. It is worth pointing out that the
region II. When ΔK increases to a certain value, it enters region III, where KImax is close to the KIc of crack growth
follows a power law,
the material, and namely
the FCGthe rateso-called
increasesParis law,until
sharply as given in Equation (7) in region II. When ∆K
the fracture.
increases to a certain value, it enters region III, where KImax is close to the KIc of the material, and the
da (7)
FCG rate increases sharply until the fracture. = C(ΔK)m ,
dN
where C and m are material constantsdarelated = C (to∆Kthe
)m , test conditions, and they can be obtained (7) by
regression analysis of FCG test data. dN
Materials 2019, 12, 1031 7 of 19

where C and m are material constants related to the test conditions, and they can be obtained by
regression analysis of FCG test data.
Two methods for calculating the FCG rate (da/dN) are recommended in GB/T 6398-2000. One is
the secant method and the other is the seven-point incremental polynomial method. Considering that
the secant method is relatively simple to operate, it is used to calculate the FCG rate in this study.

3. Analysis of Fatigue Crack Growth

According to the test conditions given in Table 3, fatigue tests are carried out in air and seawater
corrosive environments to obtain the crack propagation process. It is worth noting that the dynamic
strain gauge output is the relationship between time and strain when using the BFS measurement
method in the fatigue tests. Therefore, it is necessary to obtain the relationship between the strains
and the corresponding crack lengths of C(T) specimen. Such relationship is obtained through finite
element analysis by using ABAQUS software, and the calculation results are listed in Table 4. The
convergence analysis is carried out to analyze a C(T) specimen with a crack length of 13 mm, and it
is found that the result converges when the element size is 0.7 mm. Based on the numerical results
of the crack length and the back strain listed in Table 4, a fifth-order polynomial, which is presented
through a curve fitting technique by using DataFit software, is used to derive the relationship between
the crack length and the strain, and the detailed expression of the polynomial is listed as follows [34]:

a = C0 + C1 (µε ) + C2 (µε)2 + C3 (µε)3 + C4 (µε)4 + C5 (µε)5 , (8)

where C0 to C5 are the regression coefficients. The curve fitting results are shown in Table 5, and its
fitting accuracy can be assessed from the value of R2 (much close to 1.0).

Table 4. Relationship between crack length and back face strain.

Model ID Crack Length/mm Strain Value/µε

X56-13 13 668
X56-14 14 763
X56-15 15 870
X56-16 16 992
X56-17 17 1131
X56-18 18 1290
X56-19 19 1472
X56-20 20 1684
X56-21 21 1930
X56-22 22 2218
X56-23 23 2558
X56-24 24 2965
X56-25 25 3456
X56-26 26 4055
X56-27 27 4803

Table 5. Stress-strain relation coefficient.

Equation C0 C1 C2 C3 C4 C5 R2
Equation (8) 2.938 −2.060 × 10−2 −1.001 × 10−5 −2.935 × 10−9 −4.526 × 10−13 −2.802 × 10−17 0.99998
Equation (9) 0.973 −1685.761 −1.644 × 106 8.214 × 109 −7.381 × 1012 1 0.99999

In addition, Adedipe et al. [35] established a relation between crack length and strain considering
such factors as specimen’s material (elastic modulus E), specimen’s thickness B, specimen’s width W,
and the applied load P, which mathematical expression is represented by the polynomial as follows:

a/W = C0 + C1 U + C2 U2 + C3 U3 + C4 U4 + C5 U5 , (9)
predicted
predicted life
life is
is safe.
safe. Therefore,
Therefore, Equation
Equation (8)
(8) is
is used
used to
to predict
predict the
the crack
crack propagation
propagation in
in this
this study.
study.

Table
Table 5.
5. Stress-strain
Stress-strain relation
relation coefficient.
coefficient.
Equation
Equation CC00 C
C11 C
C22 C
C33 C
C44 C
C55 R2
R2
Materials 2019, 12, 1031 8 of 19
Equation
Equation (8)
(8) 2.938
2.938 -2.060E-02
-2.060E-02 -1.001E-05
-1.001E-05 -2.935E-09
-2.935E-09 -4.526E-13
-4.526E-13 -2.802E-17
-2.802E-17 0.99998
0.99998
Equation (9)
Equation (9) 0.973
0.973 -1685.761
-1685.761 -1.644E6
-1.644E6 8.214E9
8.214E9 -7.381E12
-7.381E12 11 0.99999
0.99999
where a/W is the aspect ratio of the C(T) specimen and C0 to C5 are the regression coefficients related
to theThe
The purpose
lengthof
purpose
crack oftoFCG
FCG tests
tests
width is
is to
ratio. to obtain
obtain the
the material
material constants
constants C C and
and m m ofof the
the API
API 5L 5L X56
X56 pipeline
pipeline
from
from Paris law and, thus, the relationship between the stress intensity factor range ΔK and the FCG
Paris law and, thus, the relationship between the stress intensity factor range ΔK and the FCG
rate
rate da/dN
da/dN can can be
be determined.
determined. The The GB/T U =6398-2000
GB/T [( | EBεW |[30]
6398-2000 /P)1/2
[30] and 1]−1 E647-2013
and+ASTM
ASTM E647-2013 [31] [31] standards
standards indicate (10)
indicate
that the shape function can be calculated using Equation
that the shape function can be calculated using Equation (5) for a standard C(T) specimen (5) for a standard C(T) specimen
where
with
with a/W
a/W| EBεW0.2.|/P
≥≥ 0.2. is a back
However,
However, surface studies
previous
previous strain parameter.
studies in
in the Compare
the literature
literature [38,39]
[38,39]thepointed
data calculated
pointed out
out that from
that the
the Equations
shape
shape function
function(8)
and (9);
solution the
of method
the standardwith higher
C(T) precision
specimen to
does determine
not follow the
the crack
equationlength
solution of the standard C(T) specimen does not follow the equation provided by standards [30,31] can
provided be selected.
by standards [30,31]
for Figurecracks
for shallow
shallow 3 shows
cracks with
withthea/W
fitting
a/W results
≤≤ 0.4.
0.4. To of thethe
To reduce
reduce crack
the length
error
error caused
causedandbythe
by thestrain
the shapevalue
shape of the
factor,
factor, theshape
the C(T) specimen
shape factor
factor of
of
the standard C(T) specimen is evaluated by numerical simulation before test data processing.from
using
the Equations
standard C(T) (8) and
specimen (9), respectively.
is evaluated The
by crack
numerical lengths calculated
simulation beforefromtestEquation
data (8) and
processing. The
The
Equation
SIFs
SIFs of
of cracks(9), respectively,
cracks with differentare
with different plotted
lengths
lengths inin
in the
thethe same coordinate
standard
standard C(T) systemare
C(T) specimens
specimens forcalculated
are comparison,
calculated as shownby
numerically
numerically byin
Figurethe
using
using the4. contour
It can beintegral
contour seen that
integral the crack
technique.
technique. The
Thelengths
numericalcalculated
numerical valuesby
values of the
of the two
the shape
shape formulas
functionare
function arealmost
are compared
compared the same.
with
with
From
the the
results data
from analysis,
Equation the
(5)results
and calculated
the new shape from Equation
factor (8)
proposed are
in relatively
[38]
the results from Equation (5) and the new shape factor proposed in [38] as shown in Figure 5. A shape as bigger,
shown in i.e.,
Figure the
5. predicted
A shape
life is safe.
function
function Therefore,
calculation
calculation Equation
equation
equation (8) is used
suitable
suitable for to predict
for this
this test
test is the crack propagation in this study.
is proposed.
proposed.

30
30 30
30
Length(mm)

Length(mm)
CrackLength(mm)

CrackLength(mm)
25
25 25
25

20
20 20
20
Back
Back strian
strian from
from FE
Crack
Crack

FE Back
15
15 15
15 Back strain
strain from
from FE
FE
Equation（8）
Equation （9）
Equation（8） Equation （9）
10
10 -5000 -4000 -3000 -2000 -1000 10
00 10 -5000 -4000 -3000 -2000 -1000 00
-5000 -4000 -3000 -2000 -1000 -5000 -4000 -3000 -2000 -1000
Strain(με)
Strain(με) Strain(με)
Strain(με)
(a)
(a) (b)
(b)
Figure
Figure 3.
Figure 3. Relation
3. Relationbetween
Relation betweencrack
between cracklength
crack lengthand
length andstrain
and strainvalue
strain valuemeasured
value measuredby
measured bythe
by theBFS
the BFS (Back
BFS (BackFace
(Back Face Strain)
Face Strain)
Strain)
compliance
compliance method:
compliance method: (a)
method: (a)Fitting
(a) Fittingresults
Fitting results from
results from Equation
from Equation (8);
Equation (8); (b)
(8); (b) Fitting
(b) Fitting results
Fitting results from
results from Equation
from Equation (9).
(9).

30 The
30 16.5
16.5 The results
results from
from Eqation
Eqation (8)
(8)
(mm)

The
The results
results from
from Eqation
Eqation (8)
Length(mm)

(8) The results from Eqation (9)

(mm)

The results from Eqation (9)

Length(mm)

25 The results from Eqation (9)

The results from Eqation (9) 16.0
16.0
25
CrackLength

15.5
15.5
CrackLength

20
20 15.0
15.0
Crack

15 14.5
14.5
Crack

15
14.0
14.0
10
10
00 4000 5500
5500 6000
6000 6500
6500 7000
7000 7500
7500 8000
8000
4000 8000
8000 12000
12000 16000
16000 20000
20000 Cycle
Cycle (N) Cycle (N)
(N)
Cycle (N)
(a)
(a) (b)
(b)
Figure
Figure 4. Relation between crack length and strain value measured by the BFS compliance method:
Figure 4.
4. Relation
Relation between
between crack
cracklength
lengthand
andstrain
strainvalue
valuemeasured
measuredbyby the
the BFS
BFS compliance
compliancemethod:
method:
(a)
(a) Comparison
(a) Comparison
Comparison of
of the
of the
results
the results
from Equations (8) and (9); (b) Enlarged view of Figure
results from Equations (8) and (9); (b) Enlarged view of Figure 4a.
4a.

The purpose of FCG tests is to obtain the material constants C and m of the API 5L X56 pipeline
from Paris law and, thus, the relationship between the stress intensity factor range ∆K and the FCG rate
da/dN can be determined. The GB/T 6398-2000 [30] and ASTM E647-2013 [31] standards indicate that
the shape function can be calculated using Equation (5) for a standard C(T) specimen with a/W ≥ 0.2.
However, previous studies in the literature [38,39] pointed out that the shape function solution of
the standard C(T) specimen does not follow the equation provided by standards [30,31] for shallow
cracks with a/W ≤ 0.4. To reduce the error caused by the shape factor, the shape factor of the
 Materials 2019, 12, 1031 9 of 19

standard C(T) specimen is evaluated by numerical simulation before test data processing. The SIFs of
cracks with different lengths in the standard C(T) specimens are calculated numerically by using the
contour integral technique. The numerical values of the shape function are compared with the results
from Equation (5) and the new shape factor proposed in [38] as shown in Figure 5. A shape function
calculation equation suitable for this test is proposed.
Materials 2019, 12, x FOR PEER REVIEW 9 of 18

24 Shape factor from References [38] 24 Equation from References [38]

22 Equation (5) 22 Equation (5)

shape factor Y
Shape factor Y

20 The results from FE 20 Equation (11)

18 18 The results from FE
16 16
14 14
12 12
10 10
0.3 0.4 0.5 0.6 0.7 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70
Aspect ratio of crack (a/W) Aspect ratio of crack (a/W)
(a) (b)
Figure
Figure5.5.Shape
Shapefactor
factorerror
erroranalysis:
analysis:(a)
(a)Numerical
Numericalsimulation
simulationcompared
comparedwith
withrecommended
recommendedshape
shape
factor in the standard; (b) Comparison of new shape factor equation and old equation.
factor in the standard; (b) Comparison of new shape factor equation and old equation.

Figure 5a analyzes
Figure analyzesthetheshape
shapefactor of the
factor of standard C(T) specimen
the standard made ofmade
C(T) specimen API 5LofX56APIsubmarine
5L X56
pipeline material.
submarine pipelineThe resultsThe
material. show that show
results the ABAQUS simulationsimulation
that the ABAQUS is higher than the results
is higher than the calculated
results
from Equation
calculated (5) and from
from Equation the from
(5) and new equation proposed
the new equation in Reference
proposed [38]. It [38].
in Reference can be found
It can clearly
be found
that the shallow crack with a/W ≤ 0.4 has a larger error. Hence, the equation
clearly that the shallow crack with a/W≤ 0.4 has a larger error. Hence, the equation recommendedrecommended in the
in
standard
the cannot
standard be used
cannot in this
be used study.
in this The The
study. datadata
obtained fromfrom
obtained numerical simulation
numerical is fitted
simulation by using
is fitted by
a polynomial,
using and aand
a polynomial, newa shape factorfactor
new shape equation suitable
equation for calculating
suitable the stress
for calculating intensity
the stress factorfactor
intensity range,
∆K, ofΔK,
range, theofmaterial is presented
the material as expressed
is presented as expressedin Equation
in Equation(11).
(11).The
Thefitting
fittingdegree
degree ofof R
2
R isis0.9983.
2 0.9983.
Figure5b
Figure 5bshows
showsthethecomparison
comparisonof ofthe
theresults
resultscalculated
calculatedfrom
from different
different equations.
equations.
5 4 3 2 2
Y =−-1562.078α
Y = 1562.078α5 ++4225.567α
4225.567α4- −
4294.602α
4294.602α+32164.728α
+ 2164.728α- 538.762α+ 63.495,
− 538.762α + 63.495, (11)
(11)

4. FCG
3.1. FCG Behavior
Behavior ofofAPI
API5L5L X56Material
X56 Materialin in
AirAir
Thefatigue
The fatiguecrack
crackgrowth
growthtested
testedininthetheair
airisismonitored
monitoredby bythe
theCOD
CODgauge
gaugemethod
methodand andthe theBFS
BFS
method, both of which belong to the flexible method. To verify the accuracy
method, both of which belong to the flexible method. To verify the accuracy of the two measurement of the two measurement
methods,the
methods, thecollected
collecteddatadataofofspecimen
specimenX56-A-1X56-A-1are areanalyzed
analyzedand andcompared.
compared.The Thedata
datacollected
collectedfrom from
the COD gauge is processed by the STM793 control system, and the relationship
the COD gauge is processed by the STM793 control system, and the relationship between the FCG between the FCG rate,
da/dN, and the SIF range, ∆K, is finally displayed on the log–log axes. The
rate, da/dN, and the SIF range, ΔK, is finally displayed on the log–log axes. The data collected fromdata collected from BFS
method
BFS methodneeds to be
needs to transferred.
be transferred. TheThethreshold
thresholdstress intensity
stress factor
intensity range
factor value
range (∆K(𝛥𝐾
value th ) is) used for
𝑡ℎ is used
the assessment of the used initial stress intensity factor range values [40,41]. After
for the assessment of the used initial stress intensity factor range values [40,41]. After processing the processing the data
(contains
data the data
(contains of region
the data I), it I),
of region canit be
canseen in Figure
be seen 6 that
in Figure the FCG
6 that trends
the FCG observed
trends in X56-S-1
observed and
in X56-S-
1X56-S-2 specimens
and X56-S-2 suggestsuggest
specimens that thethat
√
threshold stress intensity
the threshold factor range
stress intensity for range
factor X56 steel
for in
X56seawater
steel inis
approximately ∆K th
seawater is approximately = 27 MPa · m. In addition, the threshold stress intensity factor range
𝛥𝐾𝑡ℎ =27 MPa ⋅ √m. In addition, the threshold stress intensity factor range of X56 steel
√
in X56
of the air environment
steel in the airwas directly inquired
environment from the inquired
was directly fatigue testing
from machine,
the fatigue∆Kthtesting
= 28 MPa · m.
machine,
𝛥𝐾𝑡ℎ =28 MPa ⋅ √m.

0.1 0.1
BFS test data of X56-S-1 BFS test data of X56-S-2
da/dN (mm/cycle)
da/dN (mm/cycle)

0.01 0.01

0.001 0.001

1E-4 Region II 1E-4 Region II

1E-5 1E-5
30 40 50 60 70 80 90100 30 40 50 60 70 80 90100
mm1/2)
△K(MPa· mm1/2)
△K(MPa·
for the assessment of the used initial stress intensity factor range values [40,41]. After processing the
data (contains the data of region I), it can be seen in Figure 6 that the FCG trends observed in X56-S-
1 and X56-S-2 specimens suggest that the threshold stress intensity factor range for X56 steel in
seawater is approximately 𝛥𝐾𝑡ℎ =27 MPa ⋅ √m. In addition, the threshold stress intensity factor range
ofMaterials
X56 steel in1031
2019, 12, the air environment was directly inquired from the fatigue testing machine, 10 of 19
𝛥𝐾𝑡ℎ =28 MPa ⋅ √m.

0.1 0.1
BFS test data of X56-S-1 BFS test data of X56-S-2

da/dN (mm/cycle)
da/dN (mm/cycle)

0.01 0.01

0.001 0.001

1E-4 Region II 1E-4 Region II

1E-5 1E-5
30 40 50 60 70 80 90100 30 40 50 60 70 80 90100
mm1/2)
△K(MPa· mm1/2)
△K(MPa·
(a) (b)
Figure
Figure6.6.Tested
Testeddata
datapoints in in
points log–log axes
log–log (contains
axes the the
(contains datadata
of region I): (a)
of region I):Specimen X56-S-1;
(a) Specimen (b)
X56-S-1;
Specimen
(b) X56-S-2
Specimen X56-S-2.
Materials 2019, 12, x FOR PEER REVIEW 10 of 18

When
Whenprocessing
processingBFS BFSdata,
data,thethedata
datapoints
pointsthat
thatmainly
mainlyexhibit
exhibitlinear
linearbehavior
behaviorareareselected,
selected,andand
they
theyare
areused
usedtotodraw
drawthe theaverage
averagecurvecurveand andthetheupper
upperandandthethelower
lowerlimits
limitson
onthethelog–log
log–logaxesaxes
(average curve ±±SD).
(averagecurve SD).All
Allthethe test
test points are
are re-selected
re-selectedby byusing
usingFORTRAN
FORTRANcompiler compiler software,
software, and and
the
the data
data outside
outside thethe upper/lower
upper/lower limit
limit areare eliminated.
eliminated. TheThe relationship
relationship between
between thethe
FCG FCG rate,
rate, da/dN,
da/dN, and
and
the the stress
stress intensity
intensity factorfactor
range, ∆K, isΔK,
range, is shown
shown in Figure
in Figure 7a. The7a. The relationship
relationship between between
the FCGtherateFCG
and
rate
the and theintensity
stress stress intensity factorobtained
factor range range obtained
by the two by the two methods
flexible flexible methods
is shownisinshown
Figurein7b.
Figure 7b.
It shows
Itthat
shows that the
the data data monitored
monitored by the COD by the CODand
method method
by theand
BFSby the BFS
method method
have a highhave a high
degree degree of
of coincidence,
coincidence,
indicating thatindicating that both
both methods canmethods
achieve thecanpurpose
achieveofthe purpose of
monitoring monitoring
crack crack
length. This length.
result This
provides
result
a basisprovides a basiscrack
for monitoring for monitoring
length usingcrack the BFSlength
method using
alonetheinBFS method
seawater alone in seawater
experimental tests. The
experimental tests. The
data are measured in data
two waysare measured in two
to calculate theways to calculate
material constantsthe Cmaterial
and m constants C and
of Paris law, andm the
of
Paris law, and
calculated the calculated
results are summarized resultsinare summarized
Table 6. in Table 6.

0.1 0.1
X56-A-1 BFS test data of X56-A-1
da/dN(mm/cycle)

COD test data of X56-A-1

da/dN (mm/cycle)

0.01 Mean curve 0.01

Mean curve±SD Mean curve
Mean curve±SD
0.001 0.001

1E-4 1E-4

1E-5 1E-5
30 40 50 60 70 80 90 30 40 50 60 70 80 90
1/2
△K(MPa·mm ) mm1/2)
△K(MPa·
(a) (b)
Figure
Figure7.7.Comparison
Comparisonofofdifferent
differentcrack
crackgrowth
growthmonitoring
monitoringmethods:
methods:(a)
(a)FCG
FCGtest
testdata
dataobtained
obtainedby
by
BFS method; (b) Comparison of test results of COD and BFS.
BFS method; (b) Comparison of test results of COD and BFS.

C(T)specimens
Table6.6.C(T)
Table specimensmaterial
materialconstants
constantsand
andremaining
remaininglife.
life.

Calculation from Equation Recommended by GB/T

Calculation from 6398-2000
Equation Recommended Calculation from Equation (11)
Sample ID Test Count Calculation from Equation (11)
Sample ID Test Count C bym GB/TR2 6398-2000
Cycles Error (%) C m R2 Cycles Error (%)
X56-A-1 29156 9.95C× 10−9 m2.854R2 0.95 Cycles
26283 Error (%)
−9.85 6.60C× 10−9 m2.950R2 0.95 Cycles
27851 Error (%)
−4.48
X56-S-1 15547 2.92 × 10−9 3.353 0.96 14159 −8.93 2.14 × 10−9 3.418 0.96 15273 −1.76
X56-A-1
X56-S-2
29156
16394
9.95E-09
3.76 × 10−8
2.854 0.950.95
2.636
26283
15358
-9.85
−6.32 6.60E-09
3.26 × 10−8
2.950 0.950.94
2.661
27851
16154
-4.48
−1.46
X56-S-1
Pipeline 15547
42079 3.76 × 10−83.353
2.92E-09 0.960.95
2.636 14159
38444/36651 -8.93
- 3.26 × 10−83.418
2.14E-09 0.960.94
2.661 15273
40935/39383 -1.76
-

X56-S-2 16394 3.76E-08 2.636 0.95 15358 -6.32 3.26E-08 2.661 0.94 16154 -1.46
Pipeline 42079 3.76E-08 2.636 0.95 38444/36651 - 3.26E-08 2.661 0.94 40935/39383 -

3.2. FCG Behavior of API 5L X56 Material in Seawater

Since the presence of seawater increases the difficulty of the fatigue test, the crack length is
monitored in the seawater environment by using only BFS method. Table 3 shows that fatigue tests
on two specimens are conducted in seawater to verify the accuracy of the tested results. The two sets
of data are preprocessed separately, and the processing method is the same as the BFS method data
Materials 2019, 12, 1031 11 of 19

5. FCG Behavior of API 5L X56 Material in Seawater

Since the presence of seawater increases the difficulty of the fatigue test, the crack length is
monitored in the seawater environment by using only BFS method. Table 3 shows that fatigue tests
on two specimens are conducted in seawater to verify the accuracy of the tested results. The two
sets of data are preprocessed separately, and the processing method is the same as the BFS method
data processing method in air. The relationship between the FCG rate, da/dN, and the stress intensity
factor range, ∆K, is shown in Figure 8. Figure 8a,b show the trend between the FCG rate and the
stress intensity factor range for specimens X56-S-1 and X56-S-2, respectively. As seen in Figure 8,
the relationship between da/dN and ∆K exhibits a linear change in log–log axes. According to this
relationship, the two sets of tested data are fitted linearly to obtain the Paris law constants (C and
m) of the API 5L X56 pipeline steel under free corrosion conditions, as shown in Table 6. The Paris
law constants obtained from the two sets of data are used for fatigue life prediction of the standard
C(T) specimen, and then the prediction results are compared with the tested results in Table 6 to judge
the accuracy of the prediction method. The results show that both sets of data can be used to predict
Materials
fatigue 2019, 12,ax standard
life of FOR PEER REVIEW 11 of 18
C(T) specimen. Since the predicted results of specimen X56-S-2 are closer to
the experimental value, the Paris law constants measured by X56-S-2 specimen are used to
the experimental value, the Paris law constants measured by X56-S-2 specimen are used to predict predict the
corrosion
the corrosionfatigue life life
fatigue of the submarine
of the pipeline.
submarine pipeline.

0.1 0.1
BFS test data of X56-S-1 BFS test data of X56-S-2
da/dN (mm/cycle)

da/dN (mm/cycle)

0.01
Mean curve 0.01 Mean curve
Mean curve±SD Mean curve± SD

0.001 0.001

1E-4 1E-4

1E-5 1E-5
30 40 50 60 70 80 90 30 40 50 60 70 80 90
1/2
△K(MPa·mm ) m1/2)
△K(MPa·
(a) (b)
FigureFigure
8. Tested data points
8. Tested in log–log
data points axes: (a)
in log–log Specimen
axes: X56-S-1;
(a) Specimen (b) Specimen
X56-S-1; X56-S-2
(b) Specimen X56-S-2.

6. Comparison
3.3. ofFCG
Comparison of FCGBehavior
Behaviorin in
AirAir
andand in Seawater
in Seawater
The FCG
The FCG datadatatested
testedininairair
and in seawater
and in seawaterare are
plotted together
plotted in theinsame
together log–log
the same axes asaxes
log–log shownas
in Figure
shown in 9.Figure
Figure9.9aFigure
shows9a a comparison of the tested
shows a comparison ofdata, and Figure
the tested data,9band shows
Figurea comparison
9b shows of a
fitting curves (with standard deviation). It indicates that the da/dN − ∆K curve
comparison of fitting curves (with standard deviation). It indicates that the da/dN-ΔK curve in air is in air is consistent
with the trend
consistent withofthe
steady
trendfatigue crackfatigue
of steady growth in region
crack growthII, i.e., the FCG
in region II, rates increases
i.e., the FCG rates linearly when
increases
the stress
linearly whenintensity factorintensity
the stress range increases. Afterincreases.
factor range calculation,After FCG rate, da/dN,
thecalculation, the FCG corresponding
rate, da/dN,
√
to the stress intensity
corresponding factor
to the stress intensity ∆K = range
range factor ΔK =·29.64
29.64 MPa m (the
MPainitial crack
⋅ √m (the length
initial crack of length
the X56-A-1
of the
specimenspecimen
is 12.98 mm) is 1.53mm) − 4
× 10 is mm/cycle. At the endAt of the
the test, the FCG rate corresponding
X56-A-1 is 12.98 1.53×10−4 mm/cycle.
√ end of the test, the FCG rate
to the stress intensity factor range ∆K = 75.06 MPa · m (crack termination
corresponding to the stress intensity factor range ΔK = 75.06 MPa ⋅ √m (crack termination length a f = 25.03
length aisf
mm)
− 3
× 10mm)mm/cycle.
=2.4
25.03 is 2.4 × 10 mm/cycle.
−3

For API 5L X56 pipeline material, there are similar laws in seawater as in air. The difference from
the air environment is that the FCC rate of the specimen in the seawater is increased faster, and the
slope of the da/dN-ΔK logarithmic curve in the seawater is small. This indicates that the influence of
seawater corrosion on the FCG rate is greater in the initial stage of fatigue crack growth. With the
increase of ΔK, the influence of seawater corrosion is reduced gradually, and the FCG rate in the
seawater is closer to the FCG rate in the air.
According to Table 6, da/dN = 2.57 × 10−4 mm/cycle corresponding to a0 = 12.98 mm in seawater
is obtained, which is about 1.68-times the FCG rate in air. The corresponding da/dN =3.63×10−3
mm/cycle at af = 25.03 mm is about 1.52-times the FCG rate in air. From the whole test process, the
FCG rate of API 5L X56 pipeline material in seawater is about 1.6-times that in air. However, the
influence of seawater on corrosion fatigue is reduced gradually with the increase of FCG rate.

0.1 0.1
seawater is closer to the FCG rate in the air.
According to Table 6, da/dN = 2.57 × 10−4 mm/cycle corresponding to a0 = 12.98 mm in seawater
is obtained, which is about 1.68-times the FCG rate in air. The corresponding da/dN =3.63×10−3
mm/cycle at af = 25.03 mm is about 1.52-times the FCG rate in air. From the whole test process, the
FCG rate
Materials of 12,
2019, API 5L X56 pipeline material in seawater is about 1.6-times that in air. However,
1031 the
12 of 19
influence of seawater on corrosion fatigue is reduced gradually with the increase of FCG rate.

0.1 0.1
Test data of X56-A-1 Mean curve of X56-A-1

da/dN(mm/cycle)
da/dN(mm/cycle)

0.01 Test data of X56-S-2 0.01 Mean curve ± SD of X56-A-1

0.001 0.001

1E-4 1E-4
Mean curve ± SD of X56-S-2
Mean curve of X56-S-2
1E-5
1E-5 30 40 50 60 70 80 90
30 40 50 60 70 80 90
1/2 mm1/2)
△K(MPa·
△K(MPa·mm )
(b)
(a)

Figure 9. Comparison of FCG rates of API 5L X56 steel in seawater and in air: (a) Comparison of tested
data on log–log axes; (b) Comparison of fitting curves on log–log axes.

For API 5L X56 pipeline material, there are similar laws in seawater as in air. The difference from
the air environment is that the FCC rate of the specimen in the seawater is increased faster, and the
slope of the da/dN − ∆K logarithmic curve in the seawater is small. This indicates that the influence
of seawater corrosion on the FCG rate is greater in the initial stage of fatigue crack growth. With
the increase of ∆K, the influence of seawater corrosion is reduced gradually, and the FCG rate in the
seawater is closer to the FCG rate in the air.
According to Table 6, da/dN = 2.57 × 10−4 mm/cycle corresponding to a0 = 12.98 mm
in seawater is obtained, which is about 1.68-times the FCG rate in air. The corresponding
da/dN = 3.63 × 10−3 mm/cycle at af = 25.03 mm is about 1.52-times the FCG rate in air. From the
whole test process, the FCG rate of API 5L X56 pipeline material in seawater is about 1.6-times that in
air. However, the influence of seawater on corrosion fatigue is reduced gradually with the increase of
FCG rate.

7. Prediction of Corrosion Fatigue Life of Submarine Pipeline with Initial Crack Defect
Based on Paris law and the finite element analysis for SIF, an approximate integration method is
presented for predicting the fatigue life of a submarine pipeline with an initial crack defect. To verify
the effectiveness of the presented method, a corresponding experimental test on a full-scale API 5L
X56 pipeline is carried out in laboratory.

7.1. Experimental Test on Corrosion Crack Growth of a Full-Scale Submarine Pipeline

The corrosion fatigue crack propagation test on a full-scale API 5L X56 submarine pipeline is
carried out to provide an actual basis for the prediction of corrosion fatigue life. The API 5L X56
submarine pipeline used in the test is fabricated in accordance with API SPEC 5L [42] issued by the
American National Standards Institute. The structure of the fatigue testing machine and the clamping
of the specimen are shown in Figure 10. In addition, the prefabricated surface crack shown in the figure
is located on the bottom surface of the specimen at the mid-span, and it is prefabricated by a circular
abrasive cutting wheel to produce a surface crack with a width of 1.5 mm, a length of 52.1 mm and a
depth of 5.5 mm. The pipeline is pinned at both ends on the supports, and a pressure in the sinusoidal
wave is applied at the mid-span. To ensure that the crack can expand during the fatigue loading
process, the amplitude of the cyclic loading must make the crack tip has a certain stress intensity factor
to drive the crack propagation and ensure that the pipeline does not produce a wide range of yield.
The amplitude of the cyclic load in the fatigue test is 235 kN. The loading frequency, waveform, stress
ratio, and other parameter settings are consistent with the FCG tests of the API 5L X56 C(T) specimens.
The specific experimental parameters are shown in Table 3.
the sinusoidal wave is applied at the mid-span. To ensure that the crack can expand during the fatigue
loading process, the amplitude of the cyclic loading must make the crack tip has a certain stress
intensity factor to drive the crack propagation and ensure that the pipeline does not produce a wide
range of yield. The amplitude of the cyclic load in the fatigue test is 235 kN. The loading frequency,
Materialsstress
waveform, 2019, 12, 1031 and other parameter settings are consistent with the FCG tests of the API 5L
ratio, 13 of 19

X56 C(T) specimens. The specific experimental parameters are shown in Table 3.

(a) (b)
Figure 10. Fatigue
Figure test on
10. Fatigue a full-scale
test submarine
on a full-scale pipeline:
submarine (a) Fatigue
pipeline: test equipment
(a) Fatigue for full-scale
test equipment for full-scale
pipeline; (b) Schematic
pipeline; view view
(b) Schematic and details of fatigue
and details crackcrack
of fatigue test test.

To simulate
To simulate the seawater
the seawater conditions,
conditions, natural natural
seawaterseawater is continuously
is continuously injected intoinjected into the
the crack-
crack-containing
containing part during part
theduring thethe
test. At test.same
At the same
time, time, non-corrosive
non-corrosive clean water
clean water is injected
is injected into the
into the
pipeline to monitor the penetration of the fatigue crack. A large amount of water leakage occurs whenwhen
pipeline to monitor the penetration of the fatigue crack. A large amount of water leakage occurs
the fatigue
the fatigue crackcrack is through
is through the tube
the tube wall.wall.
At theAtleft
theofleft of Figure
Figure 11,crack
11, the the crack has been
has been penetrated
penetrated and and
the water
the water startsstarts to drip.
to drip. After After
the end theofend
the of the fatigue
fatigue test, thetest, thesurface
crack crack surface is visible
is visible in the
in the cut pipecut
as pipe
as shown in Figure 11. It can be seen from the fatigue crack surface that
shown in Figure 11. It can be seen from the fatigue crack surface that the fatigue crack expands in the fatigue crack expands
both in bothand
length length anddirection.
depth depth direction.
However, However, the expansion
the expansion rate doesratenotdoes notuniformly
change change uniformly
during theduring
thecrack
fatigue fatigue crack propagation.
propagation. In the beginning,
In the beginning, the initial
the initial crack propagates
crack propagates faster faster
alongalong the depth
the depth
direction
direction of theofpipeline,
the pipeline,
whilewhile the crack
the crack propagates
propagates relatively
relatively slowlyslowly
alongalong the length.
the length. As theAscrack
the crack
depthdepth increases,
increases, the FCG
the FCG rate along
rate along the circumferential
the circumferential direction
direction increases.
increases.
Materials 2019, 12, x FOR PEER REVIEW 13 of 18

Figure 11.
Figure 11. Surface
Surface of
of fatigue
fatigue crack.
crack.

7.2.
4.2. Prediction on Fatigue Life by Using
Using Paris
Paris law
law
When
When thethe safety evaluation
evaluation of crack-containing
crack-containing components
components is is carried
carried out,
out, the material constants
of Paris law are only
only related
related to
to the
the test
test environment
environment and materials, and they are widely widely used for the
prediction
prediction of
of fatigue
fatigue life.
life. The life prediction can be performed
performed by integrating Paris law as shown shown inin
Equation
Equation (12).
(12).
Z a Z N
i +1
ai+1
da
da Ni+1
i +1
= dN, (12)
∫
ai ai CC(ΔK)(∆K )m Ni Ni ,
=
m ∫ dN (12)

where aai iand

where andai+a1i+1areare
thethe
crack depths
crack of any
depths twotwo
of any cracks in the
cracks stable
in the crack
stable growth
crack region,
growth Ni and
region, Ni Nand
i+1
are
Ni+1the
are number of cycles
the number corresponding
of cycles to the
corresponding tocrack length
the crack is ai is
length andai aand
i+1 . ai+1 .
Equation (12) shows that the fatigue life of the crack in region II can be predicted if the initial
crack length (a0 ) and the final crack length (a1 ) are known. The ΔK of the C(T) specimen can be
calculated according to Equation (5). In this paper, the new shape factor, which is calculated from the
presented equation based on finite element analysis, is used to calculate ΔK. The accuracy of the Paris
law constants obtained in the FCG test directly affects the prediction of the corrosion fatigue life of
the submarine pipeline. ΔK is calculated using two shape factor equations, and the corresponding
Materials 2019, 12, 1031 14 of 19

Equation (12) shows that the fatigue life of the crack in region II can be predicted if the initial crack
length (a0 ) and the final crack length (a1 ) are known. The ∆K of the C(T) specimen can be calculated
according to Equation (5). In this paper, the new shape factor, which is calculated from the presented
equation based on finite element analysis, is used to calculate ∆K. The accuracy of the Paris law
constants obtained in the FCG test directly affects the prediction of the corrosion fatigue life of the
submarine pipeline. ∆K is calculated using two shape factor equations, and the corresponding Paris
law constants are fitted based on this. Using Equation (12) to obtain the number of cycles of the C(T)
specimen and comparing it with the test results, it was found that the error was within 10% as listed in
Table 6. Using Equations (5) and (11), respectively, to calculate the Paris law constants, it is found that
the power exponent constants calculated from Equation (5) are usually smaller than the calculated
results from the new equation of shape factor. Table 6 shows that predicting the fatigue life of the C(T)
specimen using Equation (5) recommended in the standard produces more conservative predictions.
In contrast, the results calculated from Equation (11) can be used for expectations of the remaining life
of the specimens with a better result.

7.3. Finite Element Analysis for ∆K of Fatigue Crack in Submarine Pipeline

In recent years, finite element analysis using commercial software has been favored by researchers
because of its simple operation, low cost and strong applicability. ABAQUS is used here to carry out
numerical analysis for the stress intensity factor range (∆K) of the C(T) specimen, and the numerical
results are compared with the experimental values. It is found that the finite element analysis can
obtain better results. Therefore, ABAQUS is also used to carry out finite element analysis to analyze
the cracked pipelines. When using finite element software to simulate the crack propagation process of
submarine pipelines, the crack propagation process in the pipeline can be divided into several stages,
and the crack propagation life at the end of a certain stage can be used to predict the crack propagation
life at this stage. This method is called approximate integration method. In addition, there are growth
behavior in both the radial direction and the circumferential direction during the fatigue crack growth
process of the pipeline, and the FCG rates are different.
The established finite element model for cracked pipeline needs to calculate the SIF values along
the crack front. Such SIFs can be used in Pairs law to calculate the crack growth rate and rate ratio
in different directions. According to the ratio of the circumferential direction to the radial direction,
the circumferential fatigue crack direction of the pipeline can be matched. Depending on the ratio
of the radial and the circumferential FCG rates, the length of the cracks along the radial and the
circumferential directions can be matched. Because of the stress singularity at the crack tip, a 15-node
iso-parametric stress singular element is required for simulating the crack tip using the perimeter
integral technique. Except for the crack front region, the other regions are meshed with a 20-node
hexahedron. The reduced integration technique is introduced in the calculation process to improve
the calculation efficiency. The loading and the boundary conditions in the finite element analysis are
as consistent as possible with the testing conditions, and the finite element model of the pipeline are
shown in Figure 12.
 iso-parametric stress singular element is required for simulating the crack tip using the perimeter
integral technique. Except for the crack front region, the other regions are meshed with a 20-node
hexahedron. The reduced integration technique is introduced in the calculation process to improve
the calculation efficiency. The loading and the boundary conditions in the finite element analysis are
as consistent
Materials as possible with the testing conditions, and the finite element model of the pipeline
2019, 12, 1031 are
15 of 19
shown in Figure 12.

(a) (b)
Figure
Figure 12.12.Finite
Finiteelement
elementmodel
modelofofpipeline
pipelinewith
withinitial
initialcrack:
crack:(a)(a)Schematic
Schematicview
viewofofFEFEmodel;
model;(b)(b)
Details
Details ofofthethe
FEFE model
model.

During
Duringthe thepropagation
propagationprocess
processofofthe
thecorrosion
corrosionfatigue
fatiguecrack,
crack,the
thecrack
crackpropagates
propagatesalong
alongthe
the
radial and the circumferential directions of the pipeline and the expansion rate is different.
radial and the circumferential directions of the pipeline and the expansion rate is different. Taking Taking the
X56-S-2 specimen
the X56-S-2 as an example,
specimen the cracked
as an example, pipeline
the cracked needs needs
pipeline to establish a finiteaelement
to establish model model
finite element with
different crack lengths
with different crackinlengths
two directions
in twofordirections
analysis. The
for stress intensity
analysis. The factor
stress range corresponding
intensity to
factor ranges
different crack lengths
corresponding are shown
to different crackinlengths
Tablesare
7 and 8, respectively.
shown in Tables 7 and 8, respectively.

Table Radial
7. 7.
Table crack
Radial propagation
crack and
propagation loading
and cycles.
loading cycles.
√
Model ID ModelCrack Crack Depth ∆K
ID Depth ΔK (mm/cycle) da/dN
(mm/cycle) (MPa⋅ ·√𝐦)
da/dN(𝐌𝐏𝐚 m) Cycles Cycles
Model-1-R Model-1-R 5.5 5.5 20.848
20.848 1.0538E-04
1.0538 × 10 − 4 - -
Model-2-R Model-2-R 6 6 21.354
21.354 1.1220E-04
1.1220 × 10 − 4 4457 4457
Model-3-R Model-3-R 7 7 23.006
23.006 1.3695 × 10−4
1.3695E-04 7302 7302
Model-4-R Model-4-R 8 8 24.715
24.715 1.6636 × 10−4
1.6636E-04 6011 6011
Model-5-R Model-5-R 9 9 25.776
25.776 1.8532 × 10−4
1.8532E-04 5396 5396
Model-6-R 10 27.180 2.1341 × 10−4 4686
Model-6-R 10 27.180 2.1341E-04 4686
Model-7-R 11 27.714 2.2475 × 10−4 4449
Model-7-R 12 11 27.714 2.2475E-04 − 4 4449
Model-8-R 28.302 2.3767 × 10 4208
Model-9-R Model-8-R 13 12 28.302
29.561 2.3767E-04
2.6684 × 10−4 4208 3748
Model-10-R Model-9-R 13.2 13 29.561
30.693 2.6684E-04
2.9490 × 10 − 4 3748 678
Model-10-R 13.2 30.693 2.9490E-04 678
Table 8. Circumferential crack growth and loading cycles.
Table 8. Circumferential crack growth and loading cycles
√
Model ID Crack Length ∆K(mm/cycle) da/dN (MPa· m) Cycles
Model ID Crack Length ΔK(mm/cycle) da/dN (𝐌𝐏𝐚 ⋅ √−𝐦) 6
Cycles
Model-1-C 52.1 8.209 8.8265 × 10 -
Model-1-C 52.2 52.1 8.209 8.8265E-06 -
Model-2-C 9.522 1.3099 × 10−5 3817
Model-3-C Model-2-C 52.5 52.2 9.522
11.539 1.3099E-05
2.1839 × 10−5 3817 6868
Model-4-C Model-3-C 52.9 52.5 11.539
14.733 × 10−5
2.1839E-05
4.1841 6868 4780
Model-5-C Model-4-C 53.5 52.9 17.073
14.733 6.1934 × 10−5
4.1841E-05 4780 4844
Model-6-C 54.3 20.295 × −5
Model-5-C 53.5 17.073 9.8105 10
6.1934E-05 4844 4077
Model-7-C 55.3 22.699 × −4
Model-6-C 54.3 20.295 1.3214 10
9.8105E-05 4077 3784
Model-8-C 56.6 25.188 1.7429 × 10−4 3729
Model-7-C 55.3 22.699 1.3214E-04 3784
Model-9-C 58 27.199 2.1381 × 10−4 3274
Model-10-C Model-8-C 59.5 56.6 25.188
27.755 1.7429E-04
2.2564 × 10−4 3729 4210
Model-9-C 58 27.199 2.1381E-04 3274

Table 6 shows that the submarine pipeline experienced a total of 42079 loading cycles during the
corrosion fatigue test. The final crack parameters corresponding to the number of loading cycles are as
follows: the crack propagation in radial direction (crack depth) is 13.9 mm, and the crack propagation
in circumferential direction (crack length) is 60.2 mm. The final crack parameters simulated from finite
element analysis are as follows: the crack propagation in radial direction is 13.2 mm, and the crack
propagation in circumferential direction is 59.5 mm. The numerical simulation produces an accurate
 Table 6 shows that the submarine pipeline experienced a total of 42079 loading cycles during the
corrosion fatigue test. The final crack parameters corresponding to the number of loading cycles are
as follows: the crack propagation in radial direction (crack depth) is 13.9 mm, and the crack
propagation in circumferential direction (crack length) is 60.2 mm. The final crack parameters
Materials 2019, 12, 1031 16 of 19
simulated from finite element analysis are as follows: the crack propagation in radial direction is 13.2
mm, and the crack propagation in circumferential direction is 59.5 mm. The numerical simulation
produces an accurate
enough prediction on enough prediction
the fatigue on the fatigue
crack propagation, whichcrack propagation,
indicates that thewhich indicates
presented finite that the
element
presented finite element
model is suitable model the
for analyzing is suitable
SIFs of for
theanalyzing
crack in a the SIFs of the crack in a pipeline.
pipeline.

7.4. Analysis
4.4. Analysis of
of Corrosion
Corrosion Fatigue
Fatigue Crack
Crack Propagation
Propagation
When fatigue
When fatigue crack
crack propagates
propagates in the pipeline, it is is found
found that
that the
the crack
crack growth
growth ratesrates in
in the
the radial
radial
direction and in the circumferential direction are not the same. To reflect
direction and in the circumferential direction are not the same. To reflect the expansion rates of the the expansion rates of the
crack in
crack in the
the depth
depth direction
direction and and inin the
the length
length direction
direction more
more clearly
clearly andand intuitively,
intuitively, thethe data
data inin two
two
different directions are compared in Figure 13. The figure shows that the
different directions are compared in Figure 13. The figure shows that the initial crack grows faster initial crack grows faster
along the
along the radial
radial direction
direction of of the
the pipeline
pipeline at at the
the beginning
beginning of of the
the FCG
FCG testtest while
while the the crack
crack propagates
propagates
relatively slowly
relatively slowly along
along the
the circumferential
circumferential direction.
direction. AsAs the
the crack
crackdepth
depthincreases
increasesto toaacertain
certainextent,
extent,
the growth of the crack along the circumferential has a certain hindrance to
the growth of the crack along the circumferential has a certain hindrance to the radial expansion, and the radial expansion, and
the rate
the rate of
of crack
crack propagation
propagation along along the
the circumferential
circumferential increases,
increases, i.e.,
i.e., the
the slope
slope of of the
the curve
curve becomes
becomes
larger when
larger when thethe number
number of of cycles
cycles ininthe
thefigure
figureisisapproximately
approximately15,000.
15,000.Subsequently,
Subsequently, when when the the crack
crack
grows to a certain extent in the circumferential direction, the FCG rate in that
grows to a certain extent in the circumferential direction, the FCG rate in that direction will decrease. direction will decrease.
The number of cycles per
The per stage
stageisisaccumulated
accumulatedtotocalculate
calculatethethefatigue
fatiguelife lifeofofthethepipeline
pipeline extending
extending in
different directions. Numerical analysis of radial crack propagation revealed
in different directions. Numerical analysis of radial crack propagation revealed that the crack that the crack experiences
40,935 loading
experiences cycles
40,935 when cycles
loading it propagates
when itfrom 5.5 mmfrom
propagates to 13mm
5.5 mmin depth,
to 13mm which resultswhich
in depth, in an error
resultsof
−2.7%
in compared
an error of −2.7% to the experimental
compared to the measurement. The crack experiences
experimental measurement. The crack a total of 39,383
experiences a loading
total of
cycles in
39,383 circumferential
loading cycles indirection from 52 mm
circumferential to 58 mm
direction from based
52 onmmnumerical
to 58 mm simulation,
based on which has an
numerical
error of −6.4%
simulation, whichcompared
has an toerrorthe experiment measurement.
of −6.4% compared to theIf experiment
the life prediction is carriedIfout
measurement. thefrom
life
Equation (5),
prediction the above
is carried outtwo
from errors are −(5),
Equation 8.6% theand −12.9%,
above respectively.
two errors are −8.6% and −12.9%, respectively.

0.001
da/dN(mm/cycle)

1E-4

1E-5
Radial direction of crack
Circumferential direction of crack
1E-6
0 10000 20000 30000 40000
Cycles(N)

FCG rates
Figure 13. FCG
Figure rates in
in the
the radial
radial and
and circumferential direction of the pipeline.
pipeline

The predictions
The predictions showshow thatthat thethe predicted
predicted values
values from
from two
two equations
equations are
are smaller
smaller thanthan the
the
experimental values, indicating that the predicted results are within the safe
experimental values, indicating that the predicted results are within the safe life of the cracked life of the cracked
pipeline, and
pipeline, and the
the predictions
predictions have have guiding
guiding significance
significance for
for the
the corrosion
corrosion fatigue
fatigue life
life evaluation
evaluation forfor
pipelines. In addition, it is found that the results predicted by the new equation with
pipelines. In addition, it is found that the results predicted by the new equation with the shape factor the shape factor
presented in
presented in this
this study
study areare closer
closer toto the
the experimental
experimental values
values while
while the
the predictions
predictions based
based on on the
the shape
shape
factor recommended
factor recommendedin inthe
thestandard
standardare aretoo
tooconservative.
conservative.In In summary,
summary, through
through thethe combination
combination of
of FCG test and the finite element simulation, fatigue crack growth in a pipeline
FCG test and the finite element simulation, fatigue crack growth in a pipeline can be predicted can be predicted
accurately and
accurately and effectively.
effectively. ItIt is
is also
also worth
worth noting
noting that
that the
the fatigue
fatigue life
life prediction
prediction needs
needs to to verify
verify the
the
shape factor equation recommended in the standard to obtain more
shape factor equation recommended in the standard to obtain more accurate prediction. accurate prediction.

8. Conclusions
Tests on corrosion fatigue propagation of the API 5L X56 pipeline in Bohai sea are carried out.
To obtain the material constants in Paris law, the standard C(T) specimens made of API 5L X56 steel
material in air and seawater are tested, respectively. The stress intensity factors in both C(T) specimens
and in pipeline specimen are calculated from finite element analysis. Using Paris law, the fatigue life of
Materials 2019, 12, 1031 17 of 19

submarine pipelines in Bohai sea is predicted. Based on the above studies, the following conclusions
are drawn:

1. Based on finite element analysis to calculate the stress intensity factor of C(T) specimen, it is
found that the recommended equation of shape factor in GB/T 6398-2000 produces relatively
lower values. Especially for the shallow crack (a/W < 0.4), such shape factor provides prediction
with a large error.
2. By comparing the Paris law constants obtained from equations in the standard and from the
presented equation in this study, it is found that the Paris law constants calculated from the
equation recommended in the standard are usually smaller than the results calculated from the
new presented equation of the shape factor. When the fatigue life is predicted by using the
equation recommended in the standard, the prediction is too conservative.
3. The datasets obtained from the FCG test show that for a given ∆K, the FCG rate (da/dN) in
seawater is 1.6-times that in air and the effect of seawater on corrosion fatigue decreases with the
increase of the FCG rate.
4. When using the finite element analysis to simulate the expansion of fatigue crack in the pipeline,
the fatigue crack propagation starts to expand faster along the pipeline radial direction while the
crack propagates relatively slowly along the circumferential direction in the early stage. When
the crack depth increases to a certain extent, the expansion rate in the circumferential direction
also increases. When the crack expands to a certain extent in the circumferential direction, the
fatigue crack growth rate will decrease, which is in agreement with the experimental results.
5. Based on the combination of the FCG test and finite element simulation, Paris law can be used to
predict fatigue crack growth of submarine pipeline accurately and effectively.

Author Contributions: X.G, L.X. and Y.S. designed the experiments; Y.W. conducted the experimental tests on
full-scale pipeline; X.G. analyzed the data; D.Y. contributed materials for tests.
Funding: This study is supported by Scientific Innovation Group for Youths of Sichuan Province (Fund No.
2019JDTD0017) and the National Key R&D Program of China (2016YFC0802305).
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Ossai, C.I. Advances in asset management techniques: An overview of corrosion mechanisms and mitigation
strategies for oil and gas pipelines. Isrn Corros. 2012, 2012, 1–10. [CrossRef]
2. Fekete, G.; Varga, L. The effect of the width to length ratios of corrosion defects on the burst pressures of
transmission pipelines. Eng. Fail. Anal. 2012, 21, 21–30. [CrossRef]
3. Pilkey, A.K.; Lambert, S.B.; Plumtree, A.K. Stress Corrosion Cracking of X-60 Line Pipe Steel in a
Carbonate-Bicarbonate Solution. Corrosion 1995, 51, 91–96. [CrossRef]
4. Aljaroudi, A.; Khan, F.; Akinturk, A.; Mahmoud, H.; Premkumar, T. Risk assessment of offshore crude oil
pipeline failure. J. Loss Prev. Process Ind. 2015, 37, 101–109. [CrossRef]
5. Petit, J.; Sarrazin-Baudoux, C.; Hénaff, G. An overview on environmentally assisted fatigue crack propagation.
In Proceedings of the 7th International Conference on Fatigue, Beijing, China, 8–12 June 1999; pp. 2221–2228.
6. Petit, J.; Sarrazin-Baudoux, C. Some critical aspects of low rate fatigue crack propagation in metallic materials.
Int. J. Fatigue 2010, 32, 962–970. [CrossRef]
7. Javidi, M.; Horeh, S.B. Investigating the mechanism of stress corrosion cracking in near-neutral and high pH
environments for API 5L X52 steel. Corros. Sci. 2014, 80, 213–220. [CrossRef]
8. Cheng, A.; Chen, N.Z. Corrosion fatigue crack growth modelling for subsea pipeline steels. Ocean Eng. 2017,
142, 10–19. [CrossRef]
9. Wang, J.; Li, X.Y.; Zhang, Y.L.; Miao, W.; Li, S.R. Corrosion fatigue crack growth rates of X70 base metal and
heat affected zone in H2 S corrosive environment. Br. Corros. J. 2004, 50, 18–25. [CrossRef]
Materials 2019, 12, 1031 18 of 19

10. Zheng, C.B.; Huang, Y.L.; Yu, Q.; Huo, C.Y. Effect of H2 S on stress corrosion cracking and hydrogen
permeation behaviour of X56 grade steel in atmospheric environment. Br. Corros. J. 2013, 44, 96–100.
[CrossRef]
11. Amaro, R.L.; Rustagi, N.; Findley, K.O.; Drexler, E.S.; Slifka, A.J. Modeling the fatigue crack growth of X100
pipeline steel in gaseous hydrogen. Int. J. Fatigue 2014, 59, 262–271. [CrossRef]
12. Ronevich, J.A.; D’Elia, C.R.; Hill, M.R. Fatigue crack growth rates of X100 steel welds in high pressure
hydrogen gas considering residual stress effects. Eng. Fract. Mech. 2018, 194, 42–51. [CrossRef]
13. Ronevich, J.A.; Somerday, B.P.; Feng, Z. Hydrogen accelerated fatigue crack growth of friction stir welded
X52 steel pipe. Int. J. Hydrog. Energy 2017, 42, 4259–4268. [CrossRef]
14. Cheng, A.; Chen, N.Z. Fatigue crack growth modelling for pipeline carbon steels under gaseous hydrogen
conditions. Int. J. Fatigue 2017, 96, 152–161. [CrossRef]
15. Yi, G.; Zheng, C. Hydrogen permeation through x56 pipeline steel in atmospheric environment and its
implication on SCC. Int. J. Electrochem. Sci. 2012, 7, 10633–10643.
16. Zhu, Y.; Huang, Y.; Zheng, C.; Yu, Q. The hydrogen permeation investigation of API X56 steel in sea mud.
Mater. Corros. 2015, 58, 447–451. [CrossRef]
17. Madsen, H.O.; Krenk, S.; Lind, N.C. Methods of Structural Safety; Prentice-Hall: Englewood Cliffs, NJ,
USA, 1986.
18. Leira, B.J.; Næss, A.; Næss, O.E.B. Reliability analysis of corroding pipelines by enhanced Monte Carlo
simulation. Int. J. Press. Vessel. Pip. 2016, 144, 11–17. [CrossRef]
19. Ossai, C.I.; Boswell, B.; Davies, I.J. Application of markov modelling and monte carlo simulation technique
in failure probability estimation—A consideration of corrosion defects of internally corroded pipelines. Eng.
Fail. Anal. 2016, 68, 159–171. [CrossRef]
20. Hasan, S.; Sweet, L.; Hults, J.; Valbuena, G.; Singh, B. Corrosion risk-based subsea pipeline design. Int. J.
Press. Vessel. Pip. 2018, 159, 1–14. [CrossRef]
21. Wormsen, A.; Fjeldstad, A.; Kirkemo, F.; Muff, A.D.; Reinås, L.; Macdonald, K.A. Fatigue analysis of low
alloy steel forgings used in the subsea industry. Int. J. Fatigue 2016, 96, 43–66. [CrossRef]
22. Cunha, S.B.; Pasqualino, I.P.; Pinheiro, B.C. Stress-life fatigue assessment of pipelines with plain dents.
Fatigue Fract. Eng. Mater. Struct. 2010, 32, 961–974. [CrossRef]
23. Hong, S.W.; Koo, J.M.; Chang, S.S.; Kim, J.W.; Kim, J.H.; Hong, S.K. Fatigue life prediction for an API 5L X42
natural gas pipeline. Eng. Fail. Anal. 2015, 56, 396–402. [CrossRef]
24. Dong, P.; Hong, J.K. The master S-N curve approach to fatigue of piping and vessel welds. Weld. World 2004,
48, 28–36. [CrossRef]
25. Lukács, J. Application of Conditional Fracture Toughness Values for Reliability Assessment Calculations. In
Design, Fabrication and Economy of Metal Structures: International Conference Proceedings; Jármai, K., Farkas, J.,
Eds.; Springer: Berlin/Heidelberg, Germany, 2013; pp. 633–638.
26. Lukács, J. Two Methods for Determination of Fatigue Crack Propagation Limit Curves and their Application
for Different Materials. Key Eng. Mater. 2007, 345–346, 395–398.
27. GB/T229-2007. Metallic Materials—Charpy Pendulum Impact Test Method; Standardization Administration of
China: Beijing, China, 2007.
28. BS7910-2013. Guide on Methods for Assessing the Acceptability of Flaws in Fusion Welded Structures; British
Standards Institution (BSI): London, UK, 2013.
29. Li, L.; Yang, Y.H.; Xu, Z.; Chen, G.; Chen, X. Fatigue crack growth law of API X80 pipeline steel under
various stress ratios based on J-integral. Fatigue Fract. Eng. Mater. Struct. 2015, 37, 1124–1135. [CrossRef]
30. GB/T 6398-2000. Metallic Materials—Fatigue Testing—Fatigue Crack Growth Method; Standardization
Administration of China: Beijing, China, 2000.
31. ASTM: E647-13. Standard Test Method for Measurement of Fatigue Crack Growth Rates. In Annual Book of
ASTM Standards; ASTM: West Conshohocken, PA, USA, 2008.
32. Thorpe, T.W.; Scott, P.M.; Rance, A.; Silvester, D. Corrosion fatigue of BS 4360:50D structural steel in seawater.
Int. J. Fatigue 1983, 5, 123–133. [CrossRef]
33. Adedipe, O.; Brennan, F.; Kolios, A. A relative crack opening time correlation for corrosion fatigue crack
growth in offshore structures. Fatigue Fract. Eng. Mater. Struct. 2016, 39, 395–411. [CrossRef]
34. Adedipe, O.; Brennan, F.; Kolios, A. Review of corrosion fatigue in offshore structures: Present status and
challenges in the offshore wind sector. Renew. Sustain. Energy Rev. 2016, 61, 141–154. [CrossRef]
Materials 2019, 12, 1031 19 of 19

35. Adedipe, O.; Brennan, F.; Kolios, A. Corrosion fatigue load frequency sensitivity analysis. Mar. Struct. 2015,
42, 115–136. [CrossRef]
36. Dhinakaran, S.; Prakash, R.V. Effect of low cyclic frequency on fatigue crack growth behavior of a Mn–Ni–Cr
steel in air and 3.5% NaCl solution. Mater. Sci. Eng. A 2014, 609, 204–208. [CrossRef]
37. Shih, Y.S.; Chen, J.J. The frequency effect on the fatigue crack growth rate of 304 stainless steel. Nucl. Eng.
Des. 1999, 191, 225–230. [CrossRef]
38. Mehmanparast, A.; Brennan, F.; Tavares, I. Fatigue crack growth rates for offshore wind monopile weldments
in air and seawater: SLIC inter-laboratory test results. Mater. Des. 2017, 114, 494–504. [CrossRef]
39. Mehmanparast, A.; Davies, C.M.; Dean, D.W.; Nikbin, K.M. The influence of pre-compression on the creep
deformation and failure behaviour of Type 316H stainless steel. Eng. Fract. Mech. 2013, 110, 52–67. [CrossRef]
40. Taylor, D. Compendium of Fatigue Thresholds and Crack Growth Rates, 1st ed.; EMAS: Warley, UK, 1985.
41. Taylor, D.; Liu, J.C. Sourcebook on Fatigue Crack Propagation: Threshold and Crack Closure, 1st ed.; EMAS: Warley,
UK, 1993.
42. API SPEC 5L: Specification for Line Pipe; American Petroleum Institute: Washington, DC, USA, 2012.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).