International Journal of Fatigue 143 (2021) 105997

Contents lists available at ScienceDirect

International Journal of Fatigue

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

Fatigue life prediction of 316L stainless steel weld joint including the role of
residual stress and its evolution: Experimental and modelling
Wenchun Jiang a, *, Xuefang Xie a, Tianjiao Wang a, Xiancheng Zhang b, Shan-Tung Tu b,
Jinguang Wang c, Xu Zhao d
a
State Key Laboratory of Heavy Oil Processing, College of New Energy, China University of Petroleum (East China), Qingdao 266580, PR China
b
Key Laboratory of Pressure System and Safety (MOE), School of Mechanical and Power Engineering, East China University of Science and Technology, Shanghai
200237, PR China
c
Sinopec Engineering Incorporation, China Petrochemical Corporation, Beijing 100101, PR China
d
Tianhua Institute of Chemistry Machinery, Lanzhou 730000, PR China

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

Keywords: Weld residual stresses have a great effect on the fatigue strength of welding structures. As an internal stress, the
Weld residual stress residual stress is unfounded to be treated as the conventional mechanical load during the fatigue safety
Relaxation and redistribution assessment due to its relaxation and redistribution under cyclic loading. Thus, establishing an accurate model
Fracture morphology
considering the effects of residual stress is critical for fatigue strength design. Therefore, this paper develops a
Fatigue life prediction
sequential coupling residual stress analysis and a cyclic viscoplastic constitutive model to predict the residual
stress evolution and its role on the fatigue life, which has been validated by an experimental measurement of
residual stress and a SEM-based observation fatigue fracture morphology. The results show that the role of re­
sidual stress on fatigue life is mainly reflected by the increase of mean stress, and its increased magnitudes
depends on the stress redistribution. The redistribution of residual stresses is closely correlated to the fatigue
fracture behavior of weld joints. And a fatigue life prediction model of weld joints is proposed based on the
experimental data of base metal by further considering the residual stress and its redistribution. Finally, the
experimental results verify the proposed model on predicting both the fatigue failure location and lives of weld
joints, and the predicted lives are within the 1.5 error band.

1. Introduction inevitable even after the post-welding heat treatment [5–8]. Moreover,
as the internal stresses, the magnitude and distribution of residual
The weld joints play an important role in the manufacture of the stresses vary with the external loads [9–15], which makes it more
engineering components for its advantages of cost saving, good oper­ difficult to assess its role on the structural integrity. Therefore, it is
ating conditions, easy to automate and mechanize. But it usually be­ important to establish a calculation method to quantify their effects
comes the weakest point in the whole structures due to the occurrence of more accurately.
the geometrical discontinuity, welding defects and residual stresses, Understanding the residual stress evolution under cyclic loading is
especially under the cyclic loading [1–3]. And Harati et al. [4] have the basic for the accurate fatigue life prediction of weld joints. It is
evaluated the relative effects of weld toe geometry and residual stresses unpractical to comprehensively investigated the residual stress evolu­
by comparing the fatigue strength of cruciform welds produced by the tion by experimental methods due to the limitation of measuring areas
low transformation temperature filler material and the conventional and cost, as a result, the finite element (FE) simulation cooperating with
filler material, and proved that residual stress had a relatively larger the experimental verification has been a trend to reveal the evolution of
influence than the weld toe geometry on fatigue strength. In addition, residual stresses. Dattoma et al. [12] evaluated the stress state of a weld
with the improvement of welding technology and non-destructive in­ component subjected to a sinusoidal external load by the bilinear
spection methods, the geometrical morphology and welding defects isotropic hardening model, but the results cannot clarify the relationship
have been gradually under control, but the residual stresses are between the residual stress relaxation and the number of cycles. Lee

* Corresponding author.
E-mail address: [email protected] (W. Jiang).

https://doi.org/10.1016/j.ijfatigue.2020.105997
Received 23 June 2020; Received in revised form 13 September 2020; Accepted 12 October 2020
Available online 2 November 2020
0142-1123/© 2020 Elsevier Ltd. All rights reserved.
W. Jiang et al. International Journal of Fatigue 143 (2021) 105997

et al. [13] and Cho et al. [14] developed a cyclic elastic-plastic consti­ extensively. Finally, a fatigue lifetime prediction model of weld joints
tutive model with mixed hardening rules to evaluate the evolution of was proposed base on the experimental data of base metal by further
weld residual stress under cyclic loading, and found that the relaxed considering the residual stress and its redistribution.
behavior is concentrated at first few cycles, particularly in the first cycle.
Xie et al. [15] further adopted a combined method of the FE simulation 2. Experimental
and X-ray measurement to reveal the relaxation rule of residual stress
under cyclic loading, and concluded that the residual stress relaxation is 2.1. Material and specimen preparation
the production of new plastic deformation caused by the superposition
between as-weld residual stress and applied loads. Thus, based on the The 316L SS plates with a thickness of 4 mm, which has been solution
existing research, it concludes that a cyclic constitutive model with a treated at 1070-1100℃ and then water quenched, were evaluated in this
great ability of describing the cyclic deformation behavior of weld joints study, and the chemical compositions are listed in Table 1. Then, the
is of great importance for the prediction of residual stress evolution standard plate fatigue specimens of base metal with the gauge di­
under cyclic loading. mensions of 20 × 25 mm were machined.
Apart from the prediction of residual stress evolution, investigating In addition, to avoid the influence of wire cutting process on the
the real role of residual stress on the fatigue strength assessment is also residual stress redistribution, the plate fatigue specimens of weld joints
crucial. The modified S-N model is commonly adopted by many design were machined by cutting a V-shaped groove at the base metal fatigue
codes for the fatigue assessment of weld structures, such as the ASME- specimens and then welded by the TIG weld with single pass. The sketch
VIII-2013 [16] in American, JB4732-2014 [17] in China, BS7608- of weld joints specimen is shown in Fig. 1. The welding parameters were
2015 [18] in England. The namely nominal stress [19,20], hot-spot strictly controlled with the voltage of 8–12 V, current of 50–60 A, argon
stress [20], notch stress [19] or structure stress [20–23] are used to gas flow of 8–10 L/min, and arc length was controlled at 2–4 mm. In
calculate the stress amplitude, and the fatigue life is subsequently pre­ addition, the auxiliary plates were used to fix both sides of the welding
dicted based on the different S-N curves of weld joints from the fatigue plate to avoid excessive bending deformation during the welding
tests. This is, the effect of residual stress on the fatigue life is involved in process.
the design of S-N curves, and there is no need to take the residual stress Furthermore, to better characterize the material used in this study,
into the calculation of stress amplitude additionally. However, the the microstructure was observed by the optical microscope. The
experimental results of weld joints are much more scattered than that of macroscopic morphology and typical microstructure distribution were
base metal due to its sensitivity with geometric structures and welding shown in Fig. 2. The microstructure of the based metal is single-phase
condition, which indicates that the accurate design of S-N curves of weld austenite with some twins; the side near the base metal in the heat
joints requires a large number of fatigue tests. Moreover, this method affected zone is the relatively coarse austenite grains, and the side near
can not reveal the real role of residual stress on the fatigue performance the weld is the austenite matrix with banded ferrite; the microstructure
of weld joints, which easily leads to an overconservative assessment. at the weld is equiaxed austenite and dendritic austenite with strip
Then, another method based on the fracture mechanics was attempted to ferrite.
analyze the effect of residual stresses. They incorporated the residual
stresses into the calculation of the fracture mechanics parameters, such 2.2. Fatigue tests and fracture observation
as the effective stress intensity factor [24–28] and cyclic-J integral [29],
and the various cracks propagation laws and numerical techniques, such The fatigue tests of both base metal and welding joints were per­
as the extended finite element method (XFEM) [29], node release formed by the MTS servo-hydraulic fatigue machine at ambient tem­
technique [25] or virtual crack closure technique (VCCT) [28], were perature. The stress imposed on the specimen was controlled as
subsequently adopted to simulate the fatigue cracks propagation triangular with a constant loading rate of 100 MPa/s. Due to the
behavior under the effect of residual stresses. But in those models, the occurrence of initial tensile residual stress, the real stress condition of
initial cracks with certain sizes and location must be predefined before weld joints is usually greater than the stress aroused by applied loads,
the cracks propagation, which is hardly realized in the engineering. As a which indicates that the residual stress has a significant role on the mean
results, the method based on the continuous damage mechanics is stress level of the weld joints. This is to say, a comprehensive under­
increasingly concerned due to its strong advancement in the prediction standing about the role of mean stress on the fatigue performance of the
of cracks nucleation [29–34]. The coupling of cyclic plasticity consti­ material is an important basic for the accurate consideration of residual
tutive model and fatigue damage variable characterized by the damage stress. Thus, in this study, the stress amplitude was controlled as con­
evolution laws combines the residual stresses evolution and fatigue stant of 200 MPa, but the mean stresses were different, i.e., 125 ± 200
damage accumulation, when the as-weld residual stresses were defined MPa, 150 ± 200 MPa, 175 ± 200 MPa and 200 ± 200 MPa. And the
as the initial stress-strain field. Cui et al. [33] established an elastic- deformation of gauge length during cyclic loading was monitored by a
plastic damage fatigue prediction model to calculate the fatigue life at clip gauge of 12 mm to better understand the cyclic deformation
three cases: without considering the residual stress, considering residual behavior. It is worth noting that for the weld joints fatigue specimens,
stress only without relaxation, considering both residual stress and the real stress and strain in the gauge length is heterogeneous due to the
relaxation. And they found considering residual stress and relaxation influence of welding residual stress and heterogeneous microstructure,
can provide more accurate result. However, its implementation process thus, in this study, the stress and strain in the weld joint fatigue tests
and cycle-by-cycle analysis are too complex and are very difficult to be refer to the nominal stress. Lastly, the fracture morphology was analyzed
widely used in engineering. A more easy, accurate and suitable model by SEM observation to reveal their fatigue failure mechanism.
for engineering utilization still needs to be developed.
Therefore, in this paper, the cyclic mechanical responses and fatigue
failure behaviors for 316L SS weld joints were firstly investigated by 2.3. Residual stress measurement
fatigue tests, and then the welding residual stresses and their evolution
under cyclic loading were subsequently analyzed by the combination of X-ray diffraction method was used to measure the residual stress
the X-ray measurement and numerical simulation. The underlying based on the Bragg’s law [35,36]:
mechanism of not only relaxation but also redistribution of residual λ = 2dsinθ (1)
stress is revealed by the quantified analysis of stress distribution under
different levels of applied loads, and the correlation between fatigue where λ is the wave length, d is the interatomic lattice spacing, and θ is
fracture behavior and the redistributed residual stress was discussed the diffraction angle. The test method is 2θ-sin2ψ [37], and the residual

2
W. Jiang et al. International Journal of Fatigue 143 (2021) 105997

Table 1
Chemical compositions of 316L stainless steel (%).
Element C Si Mn P S Ni Cr Mo N

Content 0.025 0.41 1.41 0.025 0.025 10.22 16.16 2.09 0.043

limited within 1 mm, as shown in Fig. 1. The measurement was per­

formed by the X-350A type stress gauge. The voltage and current are 25
kV and 7 mA, respectively. For the 316L stainless steel, (2 2 0) is used as
the measuring crystal face, and 2θ varies from 123◦ to 132◦ at a scanning
step of 0.1◦ .

2.4. Experimental results

2.4.1. Cyclic mechanical response

The cyclic strain response reflects the evolution of cyclic deformation
behavior for the material under stress-controlled loading. Fig. 3(a) and
(b) show the maximum strain evolution of the base metal and weld joints
at each cycle, respectively. It is clearly found that comparing to the
cyclic deformation behavior of base metal, the accumulation of defor­
mation for weld joints exhibited a three-stages evolution as well, i.e., the
decelerating increase phase, the saturated phase and the accelerating
increase phase, but their magnitudes were slightly smaller at the same
Fig. 1. The sketching specimen and the distribution of the measuring points. stress level, particularly under the higher stress levels. The asymmetric
loading condition and intrinsic cyclic hardening property of 316L SS is
stress is calculated by: responsible for the primary decelerating ratcheting response at the first
few cycles until entering a stable cyclic response stage. Then, the final
σ x = K⋅M (2) significant increase of deformation at the last stage indicates the
occurrence of a macro-crack and its propagation which reduces the
E π
K= − g ctgθ0 (3) stiffness of specimens. It is worth noting that the period of final macro-
2(1 + ν) 180 cracks propagation only maintained hundreds of cycles, while the
saturated cyclic response was dominant during the whole life, which
∂(2θψ x )
M= (4) indicates that the fatigue life for the weld joints in the experiments can
∂(sin2 ψ )
be approximately equal to the initiation period of macro-crack, and it is
improper for the methods based on the fracture mechanics to ignore the
where θ0 is the diffraction angle under a stress-free state, Ψ is the angle
cracks initiation when fatigue assessment of weld joints. Furthermore,
between the normal crystal surface and the material surface, and K is the
comparing the experimental results from two repeated tests, a great
stress constant. A linear relationship exists between 2θ and sin2ψ , and M
consistency was observed for the base mental, but for the weld joints, its
is the slope between the diffraction angle 2θ and sin2ψ . M is calculated if
difference was significant. This further leads to a more discrete distri­
more than three points (2θ, sin2ψ ) are determined.
bution of fatigue life of weld joints, which will be discussed at following.
The as-weld residual stress and its saturated redistribution (after 50
cycles) were measured. The distribution of measuring points is along the
weld toe with a interval of 2.5 mm, and the distance with weld toe is

Fig. 2. The macroscopic and microscopic morphologies of the 316L SS weld joints.

3
W. Jiang et al. International Journal of Fatigue 143 (2021) 105997

(a) 0.16 (b) 0.12

Base metal Weld joints
Stress amplitude:200MPa Stress amplitude: 200MPa
0.14
Mean stress: 0.10 Mean stress:
125MPa-1 125MPa-2 125MPa-1 125MPa-2
0.12 150MPa-1 150MPa-2 150MPa-1 150MPa-2
175MPa-1 175MPa-2 175MPa-1 175MPa-2
0.08 200MPa-1 200MPa-2
Maxmium strain

200MPa-1 200MPa-2

Maxmium strain
0.10

0.08 0.06

0.06
0.04

0.04
0.02
0.02

0.00 0.00
0 50000 100000 150000 200000 250000 300000 0 25000 50000 75000 100000 125000 150000 175000

Number of cycles Number of cycles

Fig. 3. The evolution of maximum strain at each cycle for the base metal (a) and weld joints (b).

2.4.2. Fatigue failure behavior welded toe, and no welding defect was observed at the crack initiation
The fatigue lives of base metal and weld joints at different mean zone. On the other hand, the direction of cracks propagation was
stresses are discussed in Fig. 4. It is clearly shown that the level of mean changed from along the width direction to thickness direction. The
stress has a strong effect on the fatigue life of the materials, i.e., the different fracture characteristics between weld joints and base metal are
fatigue life decreased with the increase of the mean stress. In addition, generally attributed to the role of geometric structure, but its correlation
the distribution of fatigue life for weld joints was more scattered than with the residual stress is still not clear.
that of the base metal, but they were all much shorter than that of base
metal regardless of the mean stress. It is worth noting that the scattered 3. FE modelling
distribution of fatigue life requires a large number of fatigue tests to
determine the accurate S-N curve of weld joints for its fatigue assess­ In this paper, FE analysis was performed to calculate the as-weld
ment, and the period and cost of fatigue tests increase the difficulty for residual stress and its redistribution under cyclic loading. The whole
the engineering application. Thus, it is very meaningful if an accurate analysis can be divided into two stages, i.e., the identification of as-weld
weld joints fatigue life prediction model can be established based on the residual stress based on the sequential coupling thermo-mechanical
experimental data of base metal by further considering the role of re­ analysis and a subsequent fatigue calculation based on a cyclic visco­
sidual stress. plastic constitutive model.
Furthermore, the fatigue fracture characteristics of the base metal The FE geometric model was established according to the dimension
and weld joints were concerned. Figs. 5 and 6 illustrate the fracture of weld joints fatigue specimen, and the mesh near the weld zone was
morphologies of base metal and weld joints under the cyclic loading of refined, as shown in Fig. 7. To realize the balance of the computational
175 ± 200 MPa, respectively. It can clearly see that the base metal cost and accuracy, the mesh sensitivity of FE results was performed, and
exhibited typical three-stages properties, i.e., the cracks initiation zone finally, 22,224 elements and 25,875 nodes were meshed.
with some precipitations at one side of the parallel section, subsequent
propagation zone along the width direction and final fractured zone
with lots of the dimples. However, the fracture morphology of weld 3.1. As-weld residual stress analysis
joints is different. On the one hand, the rupture occurred at near of the
The sequential coupling thermo-mechanical method was adopted to
calculate the as-weld residual stress, i.e., the welding thermal histories
was firstly calculated by a transient thermal analysis, and then were
300000
used as an input for the mechanical analysis. In the thermal analysis, the
275000 Base metal weld heating process was simulated by a heat source with the double
Weld joints
ellipsoid proposed by the Goldak [38], and its movements along the
250000
weld direction was modeled by a ABAQUS DFLUX user subroutine:
Fatigue lifetime(cycle)

225000 For the first half ellipsoid heat source:

200000
√̅̅̅ ( ) ( ) ( )
6 3f1 Qf x2 y2 z2
q(x, y, z) = √̅̅̅exp − 3 2 exp − 3 2 exp − 3 2 (1)
175000 πaf bc π b af c

150000 For the second half ellipsoid heat source:

√̅̅̅ ( ) ( ) ( )
125000 6 3f2 Qr x2 y2 z2
q(x, y, z) = √̅̅̅exp − 3 2 exp − 3 2 exp − 3 2 (2)
100000
πar bc π b ar c

75000 where q is heat flux density. Q is the energy of heat source. a, b and c are
half-axis lengths of ellipsoids and independent of each other. f1 and f2
50000
100 120 140 160 180 200 220 represent the energy distribution coefficients of the first and the second
Mean stress(MPa) half ellipsoids, respectively. To account for heat losses, the convection
and radiation are both taken into consideration. And the thermal pa­
Fig. 4. The fatigue lives of base metal and weld joints at different mean stress. rameters for 316L SS are shown in Table 2 [40].

4
W. Jiang et al. International Journal of Fatigue 143 (2021) 105997

Fig. 5. The fracture morphologies for the base metal under 175 ± 200 MPa.

Fig. 6. The fracture morphologies for the weld joints under 175 ± 200 MPa.

Then, for the subsequent mechanical analysis, due to the fact that
microstructure transformation has an insignificant effect on weld re­
sidual stress and deformation [39], the metallurgical phase trans­
formation is not considered. In this study, the classical thermal-elastic-
plastic small deformation theory was employed, and the total strain of

Table 2
Thermal parameters of the 316L SS.
T/℃ Thermal Density/ Specific heat/ Thermal expansion/
conductivity/ Kg × m− 3 J × Kg− 1 × ◦ C− 1
10− 5 × K− 1
W × m − 1 × ◦ C− 1

20 14.12 7966 502 1.539

200 16.69 7966 514 1.621
300 18.11 7966 526 1.686
400 19.54 7966 638 1.373
500 20.96 7966 550 1.778
Fig. 7. Finite element mesh and local refinement of weld joints.
700 23.81 7966 575 1.872
1000 28.08 7966 611 1.927

5
W. Jiang et al. International Journal of Fatigue 143 (2021) 105997

materials is decomposed into three components:

∑
N

dε = dεe + dεp + dεt (5) α= αk (11)

k=1

where εe, εp and εt are the elastic strain, plastic strain and thermal strain [ ]
2 p
tensor, respectively. And εt was calculated by the temperature- α̇k = ζk γ ε̇ − μk αk ṗ − H(fk )αk 〈ε̇p : Kk − μk ṗ〉 (12)
3 k
dependent thermal expansion coefficient, while the εe was modeled by
the isotropic Hooke’s law with temperature dependent Young’s modulus
fk = ‖αk ‖2 − γ 2k = 0 (13)
E and Poisson’s ratio μ. And our another paper [40] has proved that the
mixture hardening model has a better ability of predicting the residual αk
stress. Thus, the Chaboche mixed hardening model were employed, as Kk = (14)
‖αk ‖
follows.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ where N is the number of back stresses, and ζk and γk are material pa­
Fy = 1.5(s − α) : (s − α) − R (6)
rameters, and μk is the ratcheting coefficient.
The fatigue experimental data of base mental was used to determine
∑
N
α= αk (7) the material parameters involved in the cyclic constitutive model, and
k=1 confirmed values are listed in Table 4. To further verify the accuracy of
the cyclic viscoplastic constitutive model, an user subroutine UMAT was
dαk = Ck dεp − γk αdp (8) written to simulate the mechanical behavior of 316L SS based on the
fully implicit integration method. And Fig. 8(a) and (b) gives the com­
R = R0 + Q∞ (1 − e− bp
) (9) parison of experimental and simulated results of the 316L SS under a
representative cyclic loading 175 ± 200 MPa. It can be seen that the
where Fy is the classical Von-Mises yield function. s and α are deviatoric
cyclic constitutive model used in this paper can accurately describe the
and back stress tensor, respectively. R is a scalar, which represents the
cyclic mechanical behavior in the first few cycles, and can meet the
effective stress. N is the number of back stress, and in this study, is
requirements of calculating residual stress redistribution.
considered as 2. Ck and γk are the material parameters. R0 and Q∞ are the
initial and saturated effective stress, respectively. And b determines the
4. Discussion
hardening rate. p is the equivalent accumulation plastic strain, deter­
mined by the consistency condition in the rate-independent framework.
4.1. As-weld residual stress and its redistribution
And the involved mechanical material parameters at different temper­
atures of the 316L SS are listed in Table 3 [40].
Due to the relatively stronger constraints along the transverse di­
rection, the transverse residual stress is dominating. And further
3.2. Fatigue analysis
considering that the cyclic loading was controlled along the transverse
direction, thus, only the transverse residual stress is concerned at the
The fatigue analysis was used to determine the redistribution of re­
following discussion.
sidual stress under cyclic load, which distributes the as-weld residual
The as-weld and redistributed residual stresses along P1 under the
stress as the initial stress field. It has been introduced that the cyclic
cyclic loading of 150 ± 200 MPa are illustrated in Fig. 9(a), and the P1
constitutive model is the key for the accurate prediction of the residual
was defined perpendicular to the welding direction on the surface of the
stress evolution, and the only little difference was observed for the cyclic
weld joint. It can be observed that the as-weld residual stress was much
mechanical response of base mental and weld joints. Thus, in this study,
higher at the HAZ (heated affect zone) and weld zones, and reached to
only the base metal cyclic viscoplastic constitutive model was consid­
near 200 MPa. But after the cyclic loading, the residual stress is relaxed
ered. Apart from the Von-Mises yield function and isotropic Hooke’s
significantly, particularly at the first cycles. In addition, the weld toe at
law, a time-dependent flow rule and nonlinear kinematic hardening law
the interface between HAZ and weld zones had a higher saturated re­
considering the ratcheting behavior were included, as following:
sidual stress. Thus, to better understand the evolution of residual stress,
√̅̅̅
3 Fy n s − α the as-weld and redistributed stresses along P2 are analyzed in Fig. 9(b),
ε̇vp = 〈 〉 (10)
2 K ||s − α| |
Table 4
where the symbol < > is the Macauley operator, when x ≤ 0, <x> = 0; Material parameters of cyclic viscoplastic constitutive model.
when x > 0, <x> = x. K and n are material constants that reflect the Isotropic parameters:
material ratio. The evolution of α is determined by the Ohno-Karim E = 197000 MPa, ν = 0.3, R = 105 MPa
model [41] which has been proved has a good ability in predicting the Viscosity parameters:
ratcheting behavior, as follow: K = 290 MPa, n = 4.5
Hardening parameters:
ζ1 = 6948, γ1 = 61.2 MPa, ζ2 = 5, γ2 = 67.6 MPa, μ = 0.1

Table 3
Thermal-mechanical properties of 316L SS.
T/℃ E/MPa σ0 /MPa C1 /MPa γ1 C2 /MPa γ2 Q∞ /MPa b

20 191,200 125.60 156435.00 1410.85 6134 47.19 153.4 6.9

275 179,600 97.60 100631.00 1410.85 5568 47.19 154.7 6.9
600 137,750 90.90 64341.00 1410.85 5227 47.19 150.6 6.9
750 129,750 71.40 56232.00 1410.85 4108 47.19 57.9 6.9
900 116,320 66.20 0.05 1410.85 292 47.19 0.0 6.9
1000 100,000 31.82 0.00 1410.85 0 47.19 0.0 6.9
1100 80,000 19.73 0.00 1410.85 0 47.19 0.0 6.9
1400 20,000 2.10 0.00 1410.85 0 47.19 0.0 6.9

6
W. Jiang et al. International Journal of Fatigue 143 (2021) 105997

(a) 0.050 (b) 450

Loading condition:175±200MPa
0.048 Loading condition:175±200MPa 400 Experiment
0.046 FE
FE Model
Exp-1 350
0.044
Exp-2
0.042 300
Maxmium strain

Stress(MPa)
0.040 250
0.038
200
0.036
150
0.034
0.032 100

0.030 50
0.028
0
0.026
0 1 2 3 4 5 6 7 8 9 10 11 0.00 0.01 0.02 0.03 0.04
Number of cycles Strain

Fig. 8. Comparison of simulated and experimental results: (a) Maximum strain at first 10 cycles and (b) Strain-stress curves at first 5 cycles.

toe was translated to be tensile, particularly at two sides where the

(a) 300 30
tensile residual stress accumulated with the increase of cycles. This
Weld
redistribution and relaxation phenomenon was also verified by the
250 25
Redistributed residual stress (MPa)

experimental measurement.
Path1
As-weld residual stress (MPa)

200 20 The evolution of residual stress under cyclic loading with different
mean stress was similar to that at the cyclic loading of 150 ± 200 MPa
150
N=1 N=5
15 discussed above, but the saturated magnitudes were different. Fig. 10
shows the saturated residual stress along P2 under different mean stress.
100 10 It is easily found that the redistribution and relaxation of residual stress
N=50 increased with the increase of mean stress, and the smaller stress at
Tensile N=10
50 5 middle and greater stress at two sides of weld toe were easily identified
for the great mean stress.
0 0

-50 -5 4.2. Cyclic process of weld joints under the effect of residual stress
Compressive

-100 -10 To better understand the evolution mechanism of residual stress, the
-15 -10 -5 0 5 10 15
real-time distribution of residual stress along P2 under different external
Distance along P1 (mm) loading levels at the first cycle is illustrated in Fig. 11(a) and (b). It is
(b) 400 40 interesting that when the external loads are below 150 MPa during the
N=50 loading process, a nearly homogeneous increase of residual stress is
300 N=10 30 observed, and Fig. 11(b) further implies that the decrease of stress is
uniform during the unloading process regardless of the levels of external
Redistributed residual stress (MPa)

200 20 loading as the result of elastic deformation. This indicates that for the
As-weld residual stress (MPa)

N=5 high cycles fatige of weld joints when the external loads are smaller, the
100 Tensile 10
N=3
0 0 50
Cyclic amplitude: 200MPa
N=1
Mean stress:
-100 -10 45 125MPa
Compressive 2.5mm
Path2
150MPa
Path2
-200 -20 40 175MPa
200MPa
35
Residual stress(MPa)

-300 -30

-400 -40 30
0 5 10 15 20

Distance along P2 (mm) 25

Fig. 9. The as-weld and redistributed residual stress along (a) P1 and (b) P2 20
under the cyclic loading of 150 ± 200 MPa.
15

and P2 is defined as along weld toe. The as-weld residual stress exhibited 10
a compressive-tensile-compressive distribution due to the fact that the
cooling rate of weld mental at the middle zone was slower than that at 5
two sides. However, a significant relaxation and redistribution was 0 5 10 15 20
observed after cyclic loading due to the new equilibrium between Distance along path2 (mm)
external and residual loading, i.e., the residual stress along whole weld
Fig. 10. The saturated residual stress along P2 under different mean stress.

7
W. Jiang et al. International Journal of Fatigue 143 (2021) 105997

(a) 440
External loading 400
Without residual stress
420
375MPa
400 300
380

Stress response (MPa)

325MPa
300 200
150±200
Residual stress (MPa)

250MPa 175±200
200 100
100
0
0
150MPa -100 Point A
-100
With residual stress
-200 -200

-300 50MPa
-300
-400 0MPa
-400
0 5 10 15 20 25 0 10 20 30 40 50 60
Distance along P2 (mm) Time (s)

(b) 425 Fig. 12. The loading process of weld joints under the effect of residual stress.
External loading
400
375MPa 4.3. Fatigue life prediction model
375
The accurate prediction of fatigue failure location of weld joints is
Residual stress (MPa)

350 important for its life prediction, and the geometric characters and
325MPa
welding defects are generally considered by the traditional researches,
325 as its correlation with residual stress distribution still requires intensive
investigation. Fig. 13 compares the distribution of both as-weld and
100
75MPa saturated residual stresses with experimental fracture morphology of the
75 specimens under a typical cyclic loading of 175 ± 200 MPa. It is an
50 meaningful finding that at the one side of weld toe where fatigue crack
preferentially initiated during the experimental process, the as-weld
25
0MPa residual stresses is significantly compressive and the redistributed
0 stresses changed to be tensile. It is well known that the compressive
0 5 10 15 20 25
stress can restrain the initiation of cracks. This indicates that the redis­
Distance along P2 (mm) tributed residual stresses is more consistent with the experimental re­
sults, and the prediction of fatigue failure location must take the residual
Fig. 11. The evolution of residual stress along P2 at first cycle during (a)
stress redistribution into consideration.
loading and (b) unloading process.
Furthermore, it has been proved before that the experimental fatigue
life of weld joints is significantly scattered, and it could be improved if a
redistribution and relaxation of residual stress can be ignored. But for
weld joints fatigue life prediction model can be established based on the
the greater external loads, the distribution of residual stress is changed
experimental data of base metal. Also, the role of residual stress on the
significantly, i.e., the increase at the initial tensile stress regions is lower
cyclic process of weld joints is mainly reflected by the increase of mean
than that at the initial compressive stress regions. And the real stress at
stress. In other words, a base mental fatigue life prediction model which
two sides are even slightly greater than that at the middle when the
includes the role of mean stress is essential for the fatigue assessment of
external loads are up to 375 MPa. This redistribution of residual stress
weld joints. Kang et al. [42] investigated the cyclic mechanical behavior
along P2 must be closely related to the heterogeneous hardening
and fatigue failure mechanism of 304 SS at various load levels, and
behavior induced by the heterogeneous distribution of as-weld residual
proposed a uniaxial fatigue life prediction model by considering both
stress. The as-weld tensile residual stress indicates a more easiler yield
ratchetting damage and fatigue damage. A fatigue parameter FP was
and more significant hardening, resulting in the stresses are translated
defined by a linear superposition of the two kinds of damage, and the
towards the region where the as-weld residual stress is compressive. As a
relationship of mean stress, stress amplitude and fatigue life were
result, the redistribution of residual stress is significant after the elastic
established, as following:
unloading. That is, not only the relaxation but also the redistribution are
( ) ( )
required to be considered for the low cycle fatigue or ratchetting life σ a − σf− 1 + k σmax − σ f− 1
prediction of weld joints. FP = (15)
E
Fig. 12 shows the variation of the stress response of point A at one
side of weld toe with and without residual stress under the external N = a × [exp(b × FP) ] (16)
cyclic loading of 150 ± 200 MPa. As expected, when the residual stress is
ignored, the cyclic stress response of this point is completely consistent where σ a is the stress amplitude, σ max is the maximum stress, E is the
with the external loading. But it is very interesting that the level of mean elastic modulus, and σf-1 is the fatigue limit. k is the weighting factor of
stress was increased to 180 MPa as the stress amplitude was still 200 ratcheting damage depended on the cyclic hardening/softening char­
MPa after 5 cycles due to the redistribution, although the as-weld re­ acteristics of material to describe the sensitivity of the fatigue life to
sidual stress were compressive and reaches to near 350 MPa. This in­ ratcheting deformation. As the cyclic hardening material, the k for 316L
dicates that the role of residual stress on the cyclic stress response of the SS is considered as 0.1. Then, based on the fatigue lives for 316L SS base
material is mainly in the form of increasing mean stress, but the iden­ metal in Fig. 4, the material parameters a and b were identified as 3.01
tification of its magnitude is required to take the relaxation and redis­ × 106 and − 1.45 × 104 by the least square method, respectively.
tribution under cyclic loading into consideration. After the identification of base mental fatigue life prediction model,

8
W. Jiang et al. International Journal of Fatigue 143 (2021) 105997

Fig. 13. The comparison of as-weld and redistributed residual stress and experimental fracture morphology.

the fatigue life of weld joints can be predicted by considering the

300000
maximum tensile residual stress after the redistribution as the increase
With residual stress
of mean stress. To further reveal the role of residual stress on the fatigue
Without residual stress
life, the calculated results without considering the residual stresses were
250000
also performed in Fig. 14. It is clearly shown that the results considering
residual stress have a better agreement with the experimental results,
Prediction life(cycle)

and are within the 1.5 error band. But the predicted fatigue lives were
200000
still greater than the experimental results, which is due to the fact that 1.5 Error band
only the effect of residual stress was considered in this study, and stress
concentration due to the weld toe was ignored although it was enough 150000
small. In addition, in this study, to better reveal the role of residual stress
on the fatigue performance of weld joints, a simple uniaxial specimens
were analyze. If it is generalized to the multiaxial stress state in the 100000
engineering, the finite element analysis method used in this paper is still
applicable to calculate the saturated residual stress after the redistri­
bution and relaxation. But a multiaxial base mental fatigue life predic­ 50000
tion model is required to replace the Kang model [42] used in this paper, 50000 100000 150000 200000 250000 300000
such as another Kang model in the Ref. [43]. Experimental life(cycle)

5. Conclusion Fig. 14. The comparison of predicted model and experimental data.

In this paper, the role of residual stress on the fatigue performance of (3) A fatigue life prediction model of weld joints has been proposed
weld joints was extensively researched by the combination of FE simu­ by considering the role of residual stress after its relaxation and
lation and experimental measurement. The correlations of the residual redistribution as the mean stress, and was proved to have a great
stress redistribution, the fatigue failure location and life of weld joints agreement with experimental results.
were revealed. And a fatigue life prediction model of weld joints was
proposed on the basic of the experimental data of base metal by further Declaration of Competing Interest
considering the residual stress. The main conclusions are listed, as
follows: The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
(1) Under the cyclic loading, the as-weld residual stresses are not the work reported in this paper.
only released but also redistributed significantly due to the
compatibility of plastic deformation, and its evolution mainly
Acknowledgments
happens at the few cycles, particularly the first cycles.
(2) The role of residual stress on the cyclic process of weld joints is
The authors gratefully acknowledge the support provided by the
mainly presented by the increase of mean stress, and the
National Key R&D Program of China (2018YFC0808800), Chang Jiang
increased magnitudes depends on its redistribution and
Scholars Program, Taishan Scholar Construction Funding
relaxation.
(ts201511018), Fundamental Research Funds for the Central

9
W. Jiang et al. International Journal of Fatigue 143 (2021) 105997

Universities (17CX05019) and Project for Public Service Capability [22] Kyuba H, Dong P. Equilibrium-equivalent structural stress approach to fatigue
analysis of a rectangular hollow section joint. Int J Fatigue 2005;27:85–94.
Improvement of Industrial Technology Foundation and Common Tech­
[23] Dong P, Hong JK, Osage DA. The master S-N curve method: an implementation for
nology Promotion of Industry Quality in 2018 (CMIF-ZHL-20190102- fatigue evaluation of welded components in the ASME B&PV Code, Section VIII,
S001). Div 2 and API 579–1/ASME FFS-1. New York: WRC Bulletin; 2010.
[24] Cui C, Zhang Q, Bao Y, Bu Y, Luo Y. Fatigue life evaluation of welded joints in steel
bridge considering residual stress. J Constr Steel Res 2019;153:509–18.
References [25] Taheri S, Fatemi A. Fatigue crack behavior in power plant residual heat removal
system piping including weld residual stress effects. Int J Fatigue 2017;101:
[1] Fricke W. Fatigue analysis of welded joints: state of development. Mar Struct 2003; 244–52.
16(3):185–200. [26] Barsoum Z, Barsoum I. Residual stress effects on fatigue life of welded structures
[2] Ferro P, Berto F, James MN. Asymptotic residual stresses in butt-welded joints using LEFM. Eng Failure Anal 2008;23(2):191.
under fatigue loading. Theor Appl Fract Mech 2016;83:114–24. [27] Servetti G, Zhang X. Predicting fatigue crack growth rate in a welded butt joint: the
[3] Lopez-Jauregi A, Esnaola JA, Ulacia I, Urrutibeascoa I, Madariaga A. Fatigue role of effective R ratio in accounting for residual stress effect. Eng Fract Mech
analysis of multipass welded joints considering residual stresses. Int J Fatigue 2009;76(11):1589–602.
2015;79:75–85. [28] Jacob A, Mehmanparast A, D’Urzo R, Kelleher J. Experimental and numerical
[4] Harati E, Karlsson L, Svensson LE, Dalaei K. The relative effects of residual stresses investigation of residual stress effects on fatigue crack growth behaviour of S355
and weld toe geometry on fatigue life of weldments. Int J Fatigue 2015;77:160–5. steel weldments. Int J Fatigue 2019;128:105196.
[5] Yaghi Anas H, Hyde Thomas H, Becker Adib A. Comparison of measured and [29] Ngoula DT, Beier HT, Vormwald M. Fatigue crack growth in cruciform welded
modelled residual stresses in a welded P91 steel pipe undergoing post weld heat joints: influence of residual stresses and of the weld toe geometry. Int J Fatigue
treatment. Int J Press Vessels Pip 2020;181. 2017;101:253–62.
[6] Uzun F, Korsunsky AM. On the analysis of post weld heat treatment residual stress [30] Do VNV, Lee CH, Chang KH. High cycle fatigue analysis in presence of residual
relaxation in Inconel alloy 740H by combining the principles of artificial stresses by using a continuum damage mechanics model. Int J Fatigue 2015;70:
intelligence with the eigenstrain theory. Mater Sci Eng 2019;752(3):180–91. 51–62.
[7] Ren S, Li S, Wang Y, Deng D, Ma N. Finite element analysis of residual stress in [31] Zhan Z, Hu W, Shen F, Meng Q, Pu J, Guan Z. Fatigue life calculation for a
2.25Cr-1Mo steel pipe during welding and heat treatment process. J Manuf specimen with an impact pit considering impact damage, residual stress relaxation
Processes 2019;47:110–8. and elastic-plastic fatigue damage. Int J Fatigue 2017;96:208–23.
[8] Behnam S, Hassan S, Mahdi R, Morteza T. Effects of post weld heat treatment on [32] Lee CH, Chang KH, Do VNV. Modeling the high cycle fatigue behavior of T-joint
residual stress and mechanical properties of GTAW: the case of joining A537CL1 fillet welds considering weld-induced residual stresses based on continuum damage
pressure vessel steel and A321 austenitic stainless steel. Eng Fail Anal 2018;94: mechanics. Eng Struct 2016;125:205–16.
396–406. [33] Cui C, Zhang Q, Bao Y, Bu Y, Ye Z. Fatigue damage evaluation of orthotropic steel
[9] Mikkola E, Remes H, Marquis G. A finite element study on residual stress stability deck considering weld residual stress relaxation based on continuum damage
and fatigue damage in high-frequency mechanical impact (HFMI)-treated welded mechanics. J Bridge Eng 2018;23(10). 04018073.1-04018073.10.
joint. Int J Fatigue 2017;94(1):16–29. [34] Shen F, Zhao B, Li L, Kai C, Zhou K. Fatigue damage evolution and lifetime
[10] Estefen SF, Gurova T, Werneck D, Leontiev A. Welding stress relaxation effect in prediction of welded joints with the consideration of residual stresses and porosity.
butt-jointed steel plates. Mar Struct 2012;29(1). Int J Fatigue 2017;103:272–9.
[11] Hensel J, Nitschke-Pagel T, Tchoffo Ngoula D, Beier HT, Tchuindjang D, Zerbst U. [35] Bouffioux C, Pesci Raphaël, Boman R, Caillet N, Ponthot JP, Habraken AM.
Welding residual stresses as needed for the prediction of fatigue crack propagation Comparison of residual stresses on long rolled profiles measured by X-ray
and fatigue strength. Eng Fract Mech 2017. S0013794417303326. diffraction, ring core and the sectioning methods and simulated by FE method.
[12] Dattoma V, De Giorgi M, Nobile R. Numerical evaluation of residual stress Thin Walled Struct 2016;104:126–34.
relaxation by cyclic load. J Strain Anal Eng Des 2004;39(6):663–72. [36] Ahmed II, Adebisi JA, Abdulkareem S. Investigation of surface residual stress
[13] Lee CH, Chang KH, Do VNV. Finite element modeling of residual stress relaxation profile on martensitic stainless steel weldment with X-ray diffraction. J King Saud
in steel butt welds under cyclic loading. Eng Struct 2015;103:63–71. Univ - Eng Sci 2018;4:183–7.
[14] Cho J, Lee CH. FE analysis of residual stress relaxation in a girth-welded duplex [37] Gou R, Zhang Y, Xu X, Sun L, Yang Y. Residual stress measurement of new and in-
stainless steel pipe under cyclic loading. Int J Fatigue 2016;82:462–73. service X70 pipelines by X-ray diffraction method. NDT and E Int 2011;44(5):
[15] Xie XF, Jiang W, Luo Y, Xu S, Gong JM, Tu ST. A model to predict the relaxation of 387–93.
weld residual stress by cyclic load: experimental and finite element modeling. Int J [38] Goldak J, Chakravarti A, Bibby M. A new finite element model for welding heat
Fatigue 2017;95:293–301. sources. Metall Mater Trans B 1984;15(2):299–305.
[16] ASME-VIII-2013, Rules for construction of pressure vessels. American Standard [39] Deng D, Liang W, Murakawa H. Determination of welding deformation in fillet-
Press; 2013. welded joint by means of numerical simulation and comparison with experimental
[17] JB4732-2005, Steel pressure vessels-Design by analysis. Beijing: China Standard measurements. J Mater Process Technol 2007;183(2–3):219–25.
Press; 2005. [40] Jiang W, Luo Y, Wang BY, Woo W, Tu ST. Neutron diffraction measurement and
[18] Schijve J. Fatigue of Joints. Fatigue of Structures and Materials. Netherlands: numerical simulation to study the effect of repair depth on residual stress in 316L
Springer; 2004. stainless steel repair weld. J Pressure Vessel Technol 2015;137(4):041406.
[19] Al Zamzami I, Davison B, Susmel L. Nominal and local stress quantities to design [41] Abdel-Karim M, Ohno N. Kinematic hardening model suitable for ratchetting with
aluminium-to-steel thin welded joints against fatigue. Int J Fatigue 2019;123: steady-state. Int J Plast 2000;16(3–4):225–40.
279–95. [42] Liu Y, Kang G, Gao Q. Stress-based fatigue failure models for uniaxial
[20] Al Ibrahim, Zamzami Luca Susmel. On the accuracy of nominal, structural, and ratchetting–fatigue interaction. Int J Fatigue 2008;30(6):1065–73.
local stress based approaches in designing aluminium welded joints against fatigue. [43] Liu Y, Kang G, Gao Q. A multiaxial stress-based fatigue failure model considering
Int J Fatigue 2017;101:137–58. ratchetting-fatigue interaction. Int J Fatigue 2010;32(4):678–84.
[21] Dong P. A structural stress definition and numerical implementation for fatigue
analysis of welded joints. Int J Fatigue 2001;23:865–76.

10