Accepted Manuscript

Fatigue life of a dissimilar welded joint considering the weld residual stress:
Experimental and finite element simulation

Weiya Zhang, Wenchun Jiang, Xu Zhao, Shan-Tuang Tu

PII: S0142-1123(18)30002-1
DOI: https://doi.org/10.1016/j.ijfatigue.2018.01.002
Reference: JIJF 4531

To appear in: International Journal of Fatigue

Received Date: 24 October 2017

Revised Date: 20 December 2017
Accepted Date: 3 January 2018

Please cite this article as: Zhang, W., Jiang, W., Zhao, X., Tu, S-T., Fatigue life of a dissimilar welded joint
considering the weld residual stress: Experimental and finite element simulation, International Journal of
Fatigue (2018), doi: https://doi.org/10.1016/j.ijfatigue.2018.01.002

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Fatigue life of a dissimilar welded joint considering the weld residual stress:

Experimental and finite element simulation

Weiya Zhanga, Wenchun Jianga , Xu Zhaob, Shan-Tuang Tuc
a
State Key Laboratory of Heavy Oil Processing, College of Chemical Engineering, China University of Petroleum

(East China), Qingdao, 266580, PR China

b
Tianhua Research Institute of Chemical Machinery and Automation, Lanzhou 730060, PR China
c
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

Abstract

This paper investigated the fatigue life of a dissimilar welded joint between SAF2205

duplex stainless steel and 304 austenite stainless steel. A nonlinear fatigue damage

model based on continuum damage mechanics (CDM) is used to estimate the fatigue

life, and the effect of the residual stress is studied. The results show that large residual

stresses were generated in the dissimilar welded joint and agree well with the

indentation measurement results. The fatigue test specimens were cut from the welded

sample, and the residual stress relaxed significantly after longitudinal cutting,

especially for longitudinal residual stress. Considering the transverse residual stress,

the estimated fatigue life is more consistent with the experiment. The residual stress

mainly affects the mean stress, rather than the stress amplitude. As the longitudinal

residual stress relaxed significantly after cutting, the effect of welding residual stress

on fatigue life can be seriously overestimated. The width of fatigue test specimen

should be at least 30% of the as-weld sample for fatigue design in consideration of the

Corresponding author. Tel/Fax:+86 532 86980609;
E-mail address: jiangwenchun@126.com(Wenchun Jiang )
residual stress relaxation.

Keywords: welded joint; fatigue life; continuum damage mechanics; residual stress

1. Introduction

Dissimilar welded joints are widely used in aircrafts, shipbuilding, pressure vessels

and piping [1, 2]. However, the weld toe regions of joints are prone to fatigue crack

because they are often exposed to cycling load, which leads to the failure of the whole

structure [3]. In the meanwhile, the weld residual stresses are generated and have a

significant influence on the fatigue strength [4]. Therefore, it’s of great necessary to

take into account of the residual stress in assessment of fatigue life for dissimilar

welded joints.

In the welding process, large residual stresses are generated due to the concentrated

heat source, mismatch properties between the filler metal and base metal, constraints

and phase transformation [5-7]. Generally, the tensile residual stresses are detrimental

to the fatigue process, while the compressive residual stresses can prolong the fatigue

life. Extensive studies have been devoted to the fatigue behavior of welded structures

considering the weld residual stress [8-12]. As the residual stress is frequently related

to the plasticity induced crack closure effect, the approach of the effective stress

intensity factor is widely employed [13, 14]. The effect of the residual stresses on the

stress ratio was also considered [15], whereas the effect is noticeable only for stress

ratio R≤0.5 [16]. In the meanwhile, the presence of residual stresses changes the

mean stress [17]. In fatigue analysis, the residual stresses can be treated as mean
stresses in high cycle fatigue, while the effect of residual stress is tiny (106) in low

cycle fatigue regime as the initial cyclic plastic deformation relaxes the residual

stresses [18-20]. However, the present works are mainly focused on the similar weld

joints, how the residual stresses affect the fatigue behavior is still unclear for

dissimilar weld joints. Firstly, due to the different physical and mechanical properties,

the magnitude and distribution of residual stress for dissimilar weld joints are

different from the similar joints [21-23]. Moreover, the fatigue properties are different

for dissimilar weld metals [24, 25]. Therefore, a further study should be paid on the

fatigue behavior of dissimilar weld joints.

Linear elastic fracture mechanics (LEFM) [26, 27] and the continuum damage

mechanics (CDM) [28, 29] have been widely used to assess the fatigue life of weld

joint. The method basing on fracture mechanics puts more emphasis on estimating the

propagation life of fatigue crack, whereas the life of fatigue crack initiation is

neglected [30]. Serious underestimation of the total life maybe caused particularly in

high cycle fatigue regime, where the crack initiation occupies dominantly [31]. While

the CDM can estimate the fatigue life of weld structures including the fatigue crack

nucleation and growth [32]. Chang et al. [33] researched the high cycle fatigue life of

steel butt welds in the presence of residual stress by CDM, and the results show a high

accurate estimation taking into account the weld residual stress. Vázquez et al. [34]

proposed a multiaxial fatigue life model combined with Miner’s cumulative damage

rule, which considered the effect of the residual stress and its relaxation. Tensile
residual stress will accelerate the fatigue damage rate [35] while compressive residual

stress will delay the fatigue damage [36] and the relaxation is mainly occurs in first

few cycles [37]. In this paper, we also used CDM to evaluate the fatigue life of a

dissimilar welded joint considering the residual stress.

2. Experimental

2.1 Sample preparation

The base materials of the dissimilar welded joint are SAF2205 duplex stainless

steel and 304 austenitic stainless steel, and the welding metal E2209 is similar to

SAF2205. The 304 plate is as a stiffener welded to the SAF2205 main plate. Their

chemical compositions are listed in Table.1. The welding method is Gas Metal Arc

Welding (GMAW). The welding current, voltage and speed are130-140A, 26-28V and

18-20cm/min, respectively. The net line energy is 0.94 kJ/mm. The illustration of the

dissimilar welded joint is shown in Figs. 1.

2.2 Residual stress measurement

The methods for residual stress measurement can be divided into physical methods

and mechanical methods [38, 39]. The physical methods have the advantage that no

damage is produced to the components, including the X-ray, magnetic and ultrasonic

methods, etc. While for mechanical testing methods, the theory is very mature and has

the brevity and clarity. The shortcoming of mechanical methods is that destructive

damage is made on the components. The impact indentation method [40] for residual

stress measurement has the advantage that only small stress is superposed at the
testing position. Fig. 2 shows the schematic procedure of residual stress measurement

by impact indentation method. The method by impact indentation is to introduce an

additional stress field on the original residual stress field by applying an impact

indentation load. The residual stress is relaxed and a strain increment is generated

which could be detected by the strain gage. Therefore, the operation using the

indentation method is simple, leading a fairly good adaptation in complex testing

conditions.

The welding residual stress is measured by indentation method according to GB/T

24179-2009 [41]. The position of measurement is along the path on the surface as

shown in Fig. 1. The measurement process contains the following steps: Firstly, the

surface roughness is polished to Ra 6.3. Then the strain gages are pasted to the testing

points, and keeping the curing time 4h. Finally, the mechanical impact load is applied.

Before testing, the calibration was carried out on the material without residual stress.

The principal stresses are assumed to be parallel and normal to the welding direction,

which are calculated as follows:

E   e1  v e 2 
1  (1)
1  v2
E   e 2  v e1 
2  (2)
1  v2

where 1 and 2 are principle stresses parallel and normal to the welding direction,

respectively. e1 and e2 are the corresponding elastic strain tested by the instrument. E

and v are Elastic modulus and Possion’ ratio, respectively.

2.3 Fatigue test

 Usually, the surface treatment for welding structure should be made before

application. However, the structure is very complicated for our device used in

chemical industry. It’s difficult to smooth the weld toe before application. In this case,

we investigated the fatigue life of the original welding joint. After the measurement of

residual stress, 6 fatigue test specimens were directly cut from the welded sample

along the welding bead direction, and the width of the specimens is 2mm. Constant

stress amplitude tests with the loading ratio R=0 were performed using the MTS

machine in room temperature. Considering the results scattering, each group for

certain stress amplitude has two specimens. To avoid high temperature gradient in the

specimen caused by high frequency, and decrease the testing time, a favorable

frequency of 15 Hz is used.

3. Finite element analysis

3.1 Finite element model

A three-dimensional model was built and the meshing is shown in Fig. 3. To reduce

the computation time, one-quarter model was used and the corresponding symmetry

constraints were applied. As the welded joint was not fully penetrated, the mesh

between 304 and 2205 was not connected. The finite element meshing has a great

effect on the results and the mesh sensitivity on the results has been carried out.

Therefore, the mesh density at the welding region is refined. The minimum mesh size

at critical region is 0.5 mm and increased gradually far away from the weld region.

Finally, the number of elements and nodes for welded sample are 111980 and 120792,

respectively. The element type for welding temperature and stress analysis are DC3D8
and C3D8R, respectively. For the fatigue specimen, the mesh density along the

section is the same with the welded sample. The number of elements and nodes are

11198 and 17256, respectively.

3.2 Residual stress analysis

The welding process is modeled by a sequentially coupled method. Firstly, the

thermal analysis is made to obtain the temperature profiles by adopting the heat flux

gained from the welding arc. Then the temperature field is incorporated into the

mechanical analysis as the thermal load. The simulation of weld metal deposition is

achieved by function of element remove and add technique.

The input of heat flux, DFLUX is given by:

U  I 
DFLUX  (3)
V
where U and I are the welding voltage and current, respectively. ƞ is the arc efficiency

and V is the weld pass volume.

Meanwhile, the convection and radiation exist in the welding process [42],

obeying the Newton law (Eq. 4) and Stefan-Boltzman laws (Eq. 5), respectively.

qc  hc (Ts  T0 ) (4)

qr   0 0 [(Ts  273)4  (T0  273)4 ] (5)

where hc=10W/m2/K is the heat transfer coefficient. Ts and T0 are the surface

temperature and the ambient temperature, respectively. 0=0.9 is the emissivity and

0=5.6710-8 Wm-2K-4 is the Stefan-Boltzmann constant.

The total strain can be decomposed into three components as follows:

 total   e   p   th (6)
 e is elastic strain, p is plastic strain and th is thermal strain, respectively. Elastic

strain is modeled using the isotropic Hooke's law with temperature dependent Young's

modulus and Poisson's ratio. For the plastic strain, a plastic model is employed with

Von Mises yield criterion, temperature dependent mechanical properties and linear

isotropic hardening model. It is reported that the simulation results obtained by using

the perfect yield model is nearly identical to those obtained by using the model

considering the strain hardening [43]. Due to microscopic plasticity occurs in the high

cycle regime, the linear kinematic hardening model is used in next fatigue simulations.

The thermal strain is calculated using temperature dependent coefficient of thermal

expansion. Temperature dependent material properties of SAF2205 and 304 stainless

steels used in this model are shown in Table 2 [44,45]. Considering that welding

process has a very short heating process, creep was not included in the research. The

phase transformation will affect the residual stress to some extent due to the volume

change. However, for 2205 phase transformation occurs at the high temperature and

the yield strength is rather small. In this case, there is little effect phase transformation

on residual stress and not considered in this paper. Further research will be paid on the

quantitative effect of phase transformation on residual stress.

3.3 High cycle fatigue damage model

The continuum damage mechanics method has the advantages that depicting the

degeneration of the material strength. Kachanov and Rabotnov originally established

the continuum damage mechanics. The micro-cracks and micro-voids occurs at the

material are defined as a damage variable D. After that, the continuum damage

mechanics was widely used in the evaluation of fatigue, creep and creep-fatigue

interaction, etc [46,47]. The physical mechanism of fatigue damage is consistent with
the basic theory of damage mechanics. As the material degeneration is an energy

dissipation process, the dissipation potential   can be used to describe the damage

process. According to the orthogonal variable flow rule, the damage rate D is:
 
D (7)
Y
The energy dissipation potential   is:

S0 1
b  Y 
  (Y , Y , T , D,  )    Y (8)
S0  1  b 

where Y and T are damage strain energy release rate and temperature, respectively.

b and S0 are material constants.

Then, the damage rate can be expressed as:

   Y 
S0

D 
Y  b 
  Y  (9)

The energy dissipation potential density Y is:

 eq2 RV
Y  (10)
2 E 1  D 
2

Differentiating Y with respect to time, yields:

 eq RV
Y   eq (11)
E 1  D 
2

where RV is the stress triaxiality factor, which is expressed as:

RV  2 1  v  3  3 1  2v   H  eq 
2
(12)

where H is the hydrostatic stress and eq is Von Mises stress equivalent stress. E and

v are Young’s modules and Poisson’s ratio, respectively.

Submitting Eqs. (10) and (11) into Eq. (9), the fatigue damage rate can be obtained:
 S0
  eq2 RV  2 eq RV
D  
 2 Eb 1  D 2  2 E 1  D 2 eq
  (13)
2 eq2S0 1 RVS0 1
  eq
 2 E  0 b S0 1  D  0
S 1 2S 2

Assuming that the damage has a constant value in certain cycle, it holds that RV

equals to constant for proportional loading condition. Under the uniaxial state, the

damage during one cycle can be defined as follows:

D
N cycle
 Ddt

 max 2 RVS0 1
 2  2S 1d
0

 2E  1  D 
 min S0 1 S0 2 S0  2 (14)
b
 max B 2S0 1
 2 d
1  D 
 min 2 S0  2

 B (1  R 2 S0  2 )
B 

 
2S0  2
max

2S 0  2

 1  D 
min
2 S0  2
 S0  1  1 1  R  
2 S0  2
 a2S0  2 , R＜0
 
D     0  
2 S0  2
 1  D S  1 2 
 
N  B  max 
2S0  2
  min
2S 0  2
  B (1  R 2 S0  2 )
 a2S0  2 , 0  R＜1
 
 S0  1  1  D   S0  1  1 1  R  
2 S0  2 2 S0  2

 1  D 
2 S0  2

 2 
(15)

2 RVS0 1
where B  B  m ,  1 ,  u  , and the equations R   min  max ,
 2E 
S0 1
b S0

 a  1  R   max 2 are used here. Assuming that D=0, N=0; D=1, N=Nf. By

integrating Eq. (15), the fatigue life Nf is calculated:

 1 
2S0  2

 0   
S  1 1  R 
 2   a (2S0  2) , R <0
  2 S0  3 B 1  R

2S0  2
 
Nf   2S0  2
(16)
 1 
  0   2 
S  1 1  R 

  a (2S0  2) , 0  R <1
  2S0  3 B 1  R
2S0  2
 
 For completely reverse loading, the fatigue life can be expressed as:

Nf 
 S0  1  -(2S 2) 0
(17)
2  2S0  3 B1
1

The parameters of B-1 and S0 can be identified by the S-N curve under symmetric

loading. Here the case 0R<1 is discussed and the case R<0 is the same. From

Equations (16) and (17), we get:

2S0  2

 S0  1  1  R  
1
 S0  1  -(2S 2)  2   a (2S0  2) (18)
 
0

2  2S0  3 B1  2 S0  3 B 1  R
1 2S0  2

2S0  2
1 
2 B1  1  R    (2S0  2)
2   a  (19)
B  

1 R 2S0  2
   1 

Substituting Eq. (19) into Eq. (15), we have

 (2S0  2)
 
2 B1  a 
dD   1  (20)
2S0  2  a 
2S  2
  0
dN  S0  11  D 
Considering the effect of mean stress, the modified Goodman’s equation is used:

a 
b m 1 (21)
 a(-1) u

Then the fatigue damage at each cycle can be expressed as:

 2S0  2
dD 2 B1  a 
   (22)
dN  S0  1  1  b  m  u 1  D  

For welded structure, the stress amplitude and mean stress under multi-axial state

are written as:

1
1 3 min 
Aa    imax
, j   i , j  :  i , j   i , j  
2
min max (23)
2 2 

 mean 
2

1 max
,k 
 k ,k   kmin (24)

The multi-axial fatigue damage can be expressed as:

2S0  2
dD 2 B1  Aa 
   (25)
dN  S0  1  1  3b  mean  u 1  D  

Based on this damage model, a user subroutine was compiled by FORTRAN code

to calculate the fatigue damage evolution.

The damage parameters B-1, S0, and b could be determined by experimental data, as

listed in Table 3. The fatigue curves by experimental test for SAF2205 and 304 are

from Refs. [48-51], and the estimated fatigue curves by the present model are also

plotted in Fig. 4. It obviously shows that the estimated fatigue life by the present

fatigue damage model agrees well with the experimental results.

4. Results and discussion

4.1 As-weld residual stress distribution in the weld sample

Fig. 5 shows the contour of residual stress distribution in transverse, longitudinal

and normal direction. It obviously shows that there is stress concentration in the

welded T-joint region. For transverse stress acting normal to the direction of the weld

bead, the maximum is 400 MPa in the heat affected zone (HAZ) close to SAF2205,
whereas it is -93 MPa at the HAZ close to 304 stainless steel. For longitudinal

residual stress parallel to the direction of the weld bead, the welding region is

concentrated with a maximum tensile stress 586 MPa. In the HAZ close to SAF2205

and 304, they are 449 MPa and -97 MPa, respectively. The normal residual stress is

very small and it is not discussed in the following sections.

SAF2205 as the main plate of the component will suffer the cyclic stress in service.

Therefore, the HAZ close to SAF2205 is more sensitive to fatigue failure and should

be paid more attention. Therefore we measured the residual stress along path P (see

Fig. 1) in this zone. Fig. 6 shows the residual stress along path P by measurement and

FEM. In total, the measurement results have a good agreement with FEM, which

validates that the present FEM is right. For transverse residual stress, it is tensile

(~200MPa) at weld metal and then increases to the maximum 400MPa at HAZ, and

then finally decreases to zero gradually at base metal. For longitudinal residual stress,

it is also tensile state (~500MPa) at weld metal and HAZ and then decreases to zero at

base metal.

4.2 Residual stress distribution in the fatigue test specimen

In the paper, the small specimen of fatigue test is cut from the as-weld big sample.

The sample is cut in the longitudinal direction. But it should note that the cutting will

relax the residual stress [52]. Here, the cutting process is also modeled by the method

in our previous work [53], which is not described here in order to decrease the paper

length. Fig. 7 shows the contour of transverse (a), longitudinal (b) and normal (c)

residual stress of the fatigue test specimen. It obviously shows that the residual

stresses are released after cutting. For transverse residual stress, the HAZ close to
SAF2205 are still concentrated and the magnitude is decreased to 344 MPa. For

longitudinal residual stress, the welding region almost relaxed completely to zero state.

It finds that the longitudinal cutting has mainly relaxed the residual stress in

longitudinal direction, which is similar to our previous finding [53], while the

transverse cutting will relax the residual stress in transverse direction, as proved by

[54]. Although the residual stress has relaxed significantly, there is still stress

concentration in transverse direction at the weld toe close to SAF2205 and its effect

on fatigue life is discussed in the following.

4.3 Fatigue life assessment

Not only the transverse residual stress is concentrated at the HAZ close to SAF

2205, but also the maximum stress amplitude and mean stress occur at the region

during cycles [55], as shown in Fig. 8. Due to the concentration of residual stress and

fluctuating stress, the fatigue failure initiates at the surface of HAZ close to SAF2205

and then propagates along the thickness, and the estimated damage location is the

same as the experimental as shown in Fig. 9, as similar found by Baumgartner et al.

[56]. Once the damage at the critical region reaches 0.99, the value at adjacent regions

are still rather smaller.

Base on CDM, the fatigue life is defined as the number cycles as the damage

reaches the critical value 0.99. Fig. 10 shows the fatigue damage evolution curve in

the whole fatigue life. It shows that the fatigue damage evolution has a nonlinear

trend in the process of material failure. The damage increases slowly at beginning,
and shows an exponential growth as the number of cycles increase. After the damage

reaching the critical value 0.99, the number of cycles at the critical damage value is

equal to the fatigue life. The corresponding S-N curves by FEM and experiment are

plotted in Fig.11. The effect of residual stress on fatigue life is also researched. It is

obvious that the residual stress affect the fatigue life remarkably. Considering the

residual stress, the estimated fatigue life is more consistent with the experimental

results. When the nominal stress amplitude is larger than 160 MPa, the fatigue life

without considering the residual stress are conservative. However, the deterioration of

residual stress on fatigue life becomes noticeable as the stress amplitude decreases,

when the nominal stress amplitude is below 160 MPa.

As described in section 3.3, the present fatigue damage model is a function of stress

amplitude and mean stress. The effect of residual stress on equivalent stress amplitude

and mean stress with different applied load is shown in Fig. 12. It obviously shows

that the residual stress has little effect on the stress amplitude, while it affects the

mean stress greatly. When the nominal stress amplitude is less than 160 MPa, the

mean stress with considering residual stress is larger, while they are smaller when the

nominal stress amplitude is larger than 160 MPa. When the nominal stress amplitude

is less than 160MPa, the residual stress mainly comes from the initial welding process.

However，the residual stress is combined with the cyclic load and initial welding

process when larger load is applied. This phenomenon can be explained by R6

approach [57]. For small-scale yielding condition, the stress intensity factor at the
weld toe can be superposed by:

K  Kp  Ks (26)

The indices “p” and “s” represent the “primary” and “secondary”. The residual

stress as the secondary stress intensity factor Kres is added to the maximum stress

intensity factor Kmax and minimum stress intensity factor Kmin during the cycle. In this

case, the stress intensity factor range and stress ratio R can be obtained:

K   Kmax  Kres    Kmin  Kres  (27)

K min  K res
R (28)
K max  K res

Therefore, the residual stress will not affect the stress intensity factor range, i.e.,

stress range but influence the stress ratio, i.e., mean stress.

Fig. 13 shows the residual stress relaxation and plastic strain at the HAZ close to

SAF2205 at different nominal stress amplitude. It is revealed that the initial residual

stresses (324 MPa) tend to relax during the cyclic loading, and the extent of relaxation

is mainly dependent on the stress amplitude applied. Furthermore, the corresponding

plastic strain due to load increase as the applied load level increase. Only the

stabilized residual stress and plastic strain are depicted because the changing process

is basically limited to the first few cycles which is also found by Lee et al. [32].

However, some researchers found that the residual stress relaxation occurs over the

whole fatigue life [35]. The relaxation of residual stress is dependent on the material

characterization of cyclic softening behavior [58,59]. The residual stress keeps

constant as the cyclic stress stabilized. Although the residual stress has relaxed, the
relaxed residual stress is enough to increase the mean stress. When the nominal stress

amplitude exceeds 160 MPa, the compressive residual stress induced by applied load

is larger than that without initial residual stress. This is due to the existence of initial

residual stress makes the material yielding more easily and larger plastic deformation.

Therefore, larger compressive residual stress is obtained after unloading, and finally

decrease the mean stress compared with that without residual stress. It is significant to

recognize that the released residual stress at HAZ is high enough to considerably

affect the fatigue life. This explains the discrepancy between the fatigue life of the

weld joint with and without the residual stresses.

4.3 Discussion

Compared Fig. 5 with Fig. 7, the longitudinal residual stresses were relaxed

significantly after cutting. As a result, the effect of welding residual stress on fatigue

life can be seriously overestimated if we access the fatigue life by using the cutting

small specimen. Fig. 14a shows the fatigue life of the big as-weld sample and the

cutting sample with different nominal stress amplitude. At the same fatigue life, the

nominal stress amplitude of the uncut weld sample is about one half of the cut sample.

And the stress amplitude and mean stress at the crack initiation region with different

nominal stress amplitude are depicted in Fig. 14b. It should note that the mean stress

of the big as-weld sample is rather larger than the fatigue test specimen at the same

nominal stress amplitude, whereas the stress amplitude is slightly smaller. Both the

stress amplitude and mean stress are affected by the specimen dimension, and finally
changes the fatigue life. This phenomenon can be attributed to “size effect” of cut

dimension [60]. Ma et al. [61] once studied the effect of cutting size of specimen on

residual stress profiles and fatigue crack of friction stir weld joints, and the results

show that larger compressive residual stresses were generated at the bigger the

specimen notch tips and decrease the fatigue crack growth rates. Boni et al. [62] also

carried out the comparative fatigue tests on friction stir welded joints with different

size. The used width of larger specimens was about thirteen times of the small

specimens, and found that the fatigue life of the larger specimen was about 40% of the

small specimen. Both their work demonstrated the significant difference between the

small and big specimen. Therefore, there is necessity to find a correlation between the

small and larger specimen. Fig. 15a shows the fatigue life with different specimen

width along the weld bead. The applied nominal stress amplitude is 100 MPa. The

fatigue life shows an exponential decrease trend and then keeps constant as the

sample width increases (>45mm, i.e., 30% width of the as-weld sample). Fig. 15b

shows the stress amplitude and mean stress with different specimen width. The stress

amplitude shows a slight linear decrease and the mean stress shows an exponential

increase. As the specimen width exceeds 30% width of the as-weld sample, the mean

stress increase slowly, i.e., the residual stress relaxed slowly, which is consistent with

our previous conclusion [53]. Therefore, the critical width for fatigue test specimen

should also be at least 30% of the as-weld sample in consideration of residual stress

relaxation induced by cutting.

5. Conclusion

This paper investigated the weld residual stress and assessed the fatigue life of

dissimilar welded joint between 2205 duplex stainless steel and 304 austenite stainless

steel. The weld residual stress is researched combined with indentation measurement

and finite element method. Fatigue estimation is made using a nonlinear fatigue

damage model base on CDM and the effect of residual stress on fatigue is researched.

Conclusion can be obtained as follows:

(1) Large residual stresses were generated in the dissimilar welded joint and the

calculated residual stress agrees well with that of the indentation measurement. The

residual stresses are mainly concentrated at the welding region and HAZ (~400 MPa)

that close to SAF2205. The longitudinal residual stress relaxed to zero state after

longitudinal cutting. However, the transverse residual stress at the HAZ close to

SAF2205 is still concentrated with the magnitude 344 MPa.

(2) The fatigue failure occurs in the surface of HAZ close to SAF2205, and the

estimated fatigue life agrees well with the experiment. Effect of welding residual

stress on fatigue estimation is remarkable. The residual stresses mainly affect the

mean stress, rather than the stress amplitude.

(3) As the longitudinal residual stress relaxed significantly after longitudinal cutting,

the effect of welding residual stress on fatigue life can be seriously overestimated.

The width of fatigue test specimen should be at least 30% of the weld sample for

fatigue design in consideration of the residual stress relaxation during cutting.

Acknowledge

The authors gratefully acknowledge the support provided by the National Natural

Science Foundation of China (51105380), Taishan scholar foundation of Shandong

Province (ts201511018) and the Natural Science Foundation for Distinguished Young

Scholars in Shandong Province (JQ201417) and Fundamental Research Funds for the

Central Universities (17CX05019 and 17CX06030).

Reference

[1] J. Enz, V. Khomenko, S. Riekehr, V. Ventzke, N. Huber, N. Kashaev, Single-sided

laser beam welding of a dissimilar AA2024–AA7050 T-joint, Mater Design, 76

(2015) 110-116.

[2] M. Froend, F. Fomin, S. Riekehr, P. Alvarez, F. Zubiri, S. Bauer, B. Klusemann, N.

Kashaev, Fiber laser welding of dissimilar titanium (Ti-6Al-4V/cp-Ti) T-joints

and their laser forming process for aircraft application, Opt Laser Technol, 96

(2017) 123-131.

[3] A. Carpinteri, J. Boaretto, G. Fortese, F. Giordani, I. Iturrioz, C. Ronchei, D.

Scorza, S. Vantadori, Fatigue life estimation of fillet-welded tubular T-joints

 subjected to multiaxial loading, Int J Fatigue, (2016).

[4] S.J. Maddox, Fatigue Strength of Welded Structures, 1991.

[5] T.L. Teng, C.P. Fung, P.H. Chang, W.C. Yang, Analysis of residual stresses and

distortions in T-joint fillet welds, Int J Pres Ves Pip, 78 (2001) 523-538.

[6] L. Zhu, S. Han, Q. Song, L. Ma, Y. Wei, S. Li, Experimental study of the axial

compressive strength of CHS T-joints reinforced with external stiffening rings,

Thin Wall Struct, 98 (2016) 245-251.

[7] M. Zain-Ul-Abdein, D. Nélias, J.F. Jullien, F. Boitout, L. Dischert, X. Noe, Finite

element analysis of metallurgical phase transformations in AA 6056-T4 and their

effects upon the residual stress and distortion states of a laser welded T-joint, Int

J Pres Ves Pip, 88 (2011) 45-56.

[8] Z. Barsoum, Residual stress analysis and fatigue of multi-pass welded tubular

structures, Eng Fail Anal, 15 (2008) 863-874.

[9] C. Zeng, W. Tian, W.H. Liao, The effect of residual stress due to interference fit on

the fatigue behavior of a fastener hole with edge cracks, Eng Fail Anal, 66 (2016)

72-87.

[10] D.T. Ngoula, H.T. Beier, M. Vormwald, Fatigue crack growth in cruciform

welded joints: Influence of residual stresses and of the weld toe geometry, Int J

Fatigue, 101 (2016) 253–262.

[11] K. Hemmesi, M. Farajian, A. Fatemi, Application of the critical plane approach

to the torsional fatigue assessment of welds considering the effect of residual

 stresses, Int J Fatigue, 101, Part 2 (2017) 271-281.

[12] J.E. Larue, S.R. Daniewicz, Estimateing the effect of residual stress on fatigue

crack growth, Int J Fatigue, 29 (2007) 508-515.

[13] C. Garcia, T. Lotz, M. Martinez, A. Artemev, R. Alderliesten, R. Benedictus,

Fatigue crack growth in residual stress fields, Int J Fatigue, 87 (2016) 326-338.

[14] P.A. Deschênes, J. Lanteigne, Y. Verreman, D. Paquet, J.B. Lévesque, M. Brochu,

A new experimental method to study the influence of welding residual stresses

on fatigue crack propagation, Int J Fatigue, (2017).

[15] C.D.M. Liljedahl, J. Brouard, O. Zanellato, J. Lin, M.L. Tan, S. Ganguly, P.E.

Irving, M.E. Fitzpatrick, X. Zhang, L. Edwards, Weld residual stress effects on

fatigue crack growth behaviour of aluminium alloy 2024-T351, Int J Fatigue, 31

(2009) 1081-1088.

[16] D. Wang, H. Zhang, B. Gong, C. Deng, Residual stress effects on fatigue

behaviour of welded T-joint: A finite fracture mechanics approach, Mater Design,

91 (2015) 211-217.

[17] M. Leitner, Influence of effective stress ratio on the fatigue strength of welded

and HFMI-treated high-strength steel joints, Int J Fatigue, (2017).

[18] C.M. Sonsino, Effect of residual stresses on the fatigue behaviour of welded

joints depending on loading conditions and weld geometry, Int J Fatigue, 31

(2009) 88-101.

[19] P. Rettenmeier, E. Roos, S. Weihe, Fatigue analysis of multiaxially loaded crane

 runway structures including welding residual stress effects, Int J Fatigue, 82

(2016) 179-187.

[20] X.K. Meng, J.Z. Zhou, C. Su, S. Huang, K.Y. Luo, J. Sheng, W. Tan, Residual

stress relaxation and its effects on the fatigue properties of Ti6Al4V alloy

strengthened by warm laser peening, Mat Sci Eng A-Struct, (2016).

[21] K.H. Chang, C.H. Lee, Finite element analysis of the residual stresses in T-joint

fillet welds made of similar and dissimilar steels, Int J Adv Manuf Tech, 41

(2009) 250.

[22] P.V. Ramana, G.M. Reddy, T. Mohandas, A.V.S.S.K.S. Gupta, Microstructure and

residual stress distribution of similar and dissimilar electron beam welds –

Maraging steel to medium alloy medium carbon steel, Mater Design, 31 (2010)

749-760.

[23] C.H. Lee, K.H. Chang, Comparative study on girth weld-induced residual

stresses between austenitic and duplex stainless steel pipe welds, Appl Therm

Eng, 63 (2014) 140-150.

[24] S. Mercan, S. Aydin, N. Özdemir, Effect of welding parameters on the fatigue

properties of dissimilar AISI 2205–AISI 1020 joined by friction welding, Int J

Fatigue, 81 (2015) 78-90.

[25] V. Infante, D.F.O. Braga, F. Duarte, P.M.G. Moreira, M.D. Freitas, P.M.S.T.D.

Castro, Study of the fatigue behaviour of dissimilar aluminium joints produced

by friction stir welding, Int J Fatigue, 82 (2015).

[26] Z. Barsoum, I. Barsoum, Residual stress effects on fatigue life of welded

structures using LEFM, Eng Fail Anal, 16 (2009) 449-467.

[27] C.H. Lee, K.H. Chang, Finite element computation of fatigue growth rates for

mode I cracks subjected to welding residual stresses, Eng Fract Mech, 78 (2011)

2505-2520.

[28] T.L. Teng, C.P. Fung, P.H. Chang, Effect of weld geometry and residual stresses

on fatigue in butt-welded joints, Int J Pres Vess Pip, 79 (2002) 467–482.

[29] S. Sarkani, G. Michaelov, D.P. Kihl, Stochastic fatigue damage accumulation in a

T-welded joint accounting for the residual stress fields, Int J Fatigue, 23 (2001)

71-78.

[30] Z. Wang, Q. Wang, Fatigue assessment of welds joining corrugated steel webs to

flange plates, Eng Struct, 73 (2014) 1-12.

[31] X. Yang, N. Li, Z. Jin, T. Wang, A continuous low cycle fatigue damage model

and its application in engineering materials, Int J Fatigue, 19 (1997) 687-692.

[32] C.H. Lee, K.H. Chang, V.N.V. Do, Modeling the high cycle fatigue behavior of

T-joint fillet welds considering weld-induced residual stresses based on

continuum damage mechanics, Eng Struct, 125 (2016) 205-216.

[33] V.N.V. Do, C.H. Lee, K.H. Chang, V.N.V. Do, C.H. Lee, K.H. Chang, High cycle

fatigue analysis in presence of residual stresses by using a continuum damage

mechanics model, Int J Fatigue, 70 (2015) 51-62.

[34] J. Vázquez, C. Navarro, J. Domínguez, A model to estimate fretting fatigue life

 including residual stresses, Theor Appl Fract Mec, 73 (2014) 144–151.

[35] F. Shen, B. Zhao, L. Li, C.K. Chua, K. Zhou, Fatigue damage evolution and

lifetime estimateion of welded joints with the consideration of residual stresses

and porosity, Int J Fatigue, 103 (2017) 272-279.

[36] Y. Shen, S.M. Moghadam, F. Sadeghi, K. Paulson, R.W. Trice, Effect of retained

austenite-Compressive residual stresses on rolling contact fatigue life of

carburized AISI 8620 steel, Int J Fatigue, 75 (2015) 135-144.

[37] Z. Zhan, W. Hu, F. Shen, Q. Meng, J. Pu, Z. Guan, Fatigue life calculation for a

specimen with an impact pit considering impact damage, residual stress

relaxation and elastic-plastic fatigue damage, Int J Fatigue, (2016).

[38] Wan Y, Jiang W, Li J, et al. Weld residual stresses in a thick plate considering

back chipping: neutron diffraction, contour method and finite element simulation

study. Mat Sci & Eng A, 2017.

[39] Zhou P, Jin X, Chen J, et al. Residual stress estimation in laminated ZrB2-SiC

ultra-high temperature ceramics with strong interfaces using X-ray diffraction

and indentation techniques. Ceram Int, 2017, 43(15): 12459-12465.

[40] Xue H, Yan L, Ge R, et al. Residual stress research on Heavy Rail using X-ray

diffraction and impact indentation method. International Conference on

Intelligent Systems Research and Mechatronics Engineering. 2015.

[41] Chinese National Standard: Metallic materials-Residual stress determination-The

indentation strain gage method, GB/T 24179-2009.

[42] Wu S C, Peng X, Zhang W H, et al. The virtual node polygonal element method

for nonlinear thermal analysis with application to hybrid laser welding. Int J Heat

& Mass Transfer, 2013, 67(3):1247-1254.

[43] Yan DJ, Liu XS, Li J, Yang JG, Fang HY. Effect of strain hardening and strain

softening on welding distortion and residual stress of A7N01-T4 aluminum alloy

by simulation analysis. J Cent South Univ Technol 2010;17:666–73.

[44] E.F. Rybicki, R.B. Stonesifer, Computation of Residual Stresses due to Multipass

Welds in Piping Systems, J Press Vess-T Asme, 101 (1979) 149.

[45] W. Jiang, X.P. Xu, J.M. Gong, S.T. Tu, Influence of repair length on residual

stress in the repair weld of a clad plate, Nucl Eng Des, 246 (2012) 211-219.

[46] Kang G, Liu Y, Ding J, et al. Uniaxial ratcheting and fatigue failure of tempered

42CrMo steel: Damage evolution and damage-coupled visco-plastic constitutive

model. Int J Plasticity, 2009, 25(5):838-860.

[47] J. Lemaître, R. Desmorat, Engineering damage mechanics : ductile, creep,

fatigue and brittle failures, (2005).

[48] G. Chai, Fatigue behaviour of duplex stainless steels in the very high cycle

regime, Int J Fatigue, 28 (2006) 1611-1617.

[49] M. Nakajima, Y. Uematsu, T. Kakiuchi, M. Akita, K. Tokaji, Effect of quantity of

martensitic transformation on fatigue behavior in type 304 stainless steel,

Procedia Eng, 10 (2011) 299-304.

[50] Pessoa D F, Herwig P, Wetzig A, et al. Influence of surface condition due to laser
 beam cutting on the fatigue behavior of metastable austenitic stainless steel AISI

304. Eng Fract Mech, 2017.

[51] Polak J. Cyclic plastic response and fatigue life of duplex and superduplex

stainless steels. Kovove Materialy, 2005, 43(4): 280.

[52] Dattoma V, Giorgi M D, Nobile R. On the evolution of welding residual stress

after milling and cutting machining. Comput & Struct, 2006,

84(29–30):1965-1976.

[53] Jiang W, Woo W, An G B, et al. Neutron diffraction and finite element modeling

to study the weld residual stress relaxation induced by cutting. Mater & Design,

2013, 51(5):415-420.

[54] Qian Z Y, Chumbley S, Johnson E. The Effect of Specimen Dimension on

Residual Stress Relaxation of the Weldments. Adv Mater Res, 2014,

996(996):820-826.

[55] D.T. Ngoula, H.T. Beier, M. Vormwald, Fatigue crack growth in cruciform

welded joints: Influence of residual stresses and of the weld toe geometry, Int J

Fatigue, 101 (2017) 253-262.

[56] J. Baumgartner, T. Bruder, Influence of weld geometry and residual stresses on

the fatigue strength of longitudinal stiffeners, Weld World, 57 (2013) 841-855.

[57] Budden PJ, Sharples JK. Treatment of secondary stresses. In: Ainsworth RA,

Schwalbe K-H, editors. Comprehensive structural integrity; volume 7: practical

failure assessment methods. Elsevier; 2003. p. 245–87.

[58] Kang G, Kan Q, Yu C, et al. Whole-life transformation ratchetting and fatigue of

super-elastic NiTi Alloy under uniaxial stress-controlled cyclic loading[J].

Materials Science & Engineering A, 2012, 535(7):228-234.

[59] Hensel J, Nitschke-Pagel T, Ngoula D T, et al. Welding residual stresses as

needed for the prediction of fatigue crack propagation and fatigue strength. Eng

Fract Mech, 2017.

[60] Carpinteri A, Spagnoli A. A fractal analysis of size effect on fatigue crack growth.

Int J Fatigue, 2004, 26(2):125-133.

[61] Ma Y E, Staron P, Fischer T, et al. Size effects on residual stress and fatigue

crack growth in friction stir welded 2195-T8 aluminium – Part I: Experiments.

Intl J Fatigue, 2011, 33(11):1417-1425.

[62] Boni L, Lanciotti A, Polese C. “Size effect” in the fatigue behavior of Friction

Stir Welded plates. Int J Fatigue, 2015, 80:238-245.

List of figure captions

Figure
Caption
number
Fig. 1 Illustration of the welded T-joint
Schematic procedure of residual stress measurement by impact indentation
Fig. 2
method
Fig. 3 Mesh of the welded sample and fatigue specimen
Fatigue life of SAF2205 duplex stainless steel and 304 austenite stainless
Fig. 4
steel
Contour of transverse (a), longitudinal (b) and normal (c) residual stress
Fig.5
distribution
Residual stress distribution along the path by indentation and finite
Fig. 6
element method
Contour of transverse (a), longitudinal (b) and normal (c) residual stress
Fig.7
distribution after cutting
Contour of stress amplitude (a), mean stress (b) of the fatigue test
Fig. 8
specimen during cycle combined with residual stress
Fig. 9 Fatigue crack initiation position of the weld T-joint
Fatigue damage curve as number of cycles increase with different applied
Fig. 10
load
Fig. 11 Fatigue life comparisons between experimental and FEM
Comparisons of stress amplitude and mean stress with and without
Fig. 12
considering the residual stress
Transverse residual stress relaxation and plastic strain at different nominal
Fig. 13
stress amplitude
Fatigue life estimation (a), and stress amplitude and mean stress (b) of
Fig. 14
weld sample and fatigue test specimen considering the residual stress
Fatigue life (a), and stress amplitude and mean stress (b) with different
Fig. 15
specimen width

Table 1 chemical composition of SAF2205 and 304 stainless steel

Materials C Si Mn Cr Ni Mo N Austenite (%)

2205 0.03 0.8 1.2 25 7 4.8 0.3 55
304 0.06 1.0 2.0 17.0-19.0 8.0-11.0 - - -

Table 2 Temperature dependent properties for 2205 and 304

Temperature(C) 20 200 400 600 800

Material properties for 2205 duplex stainless steel

Young’s Modulus(GPa) 210 175 152 140 112
Yield strength(MPa) 510 322.2 278.5 218.5 170.4
Poisson’s ratio 0.27 0.29 0.33 0.29 0.26
3
Density(kg/m ) 7886 7815 7781 7762 7738
-6
Thermal expansion(1/C10 ) 13.64 14.37 17.05 17.71 18.07
Thermal conductivity(W/mC) 19.2 21.3 24.2 24.1 23.0
Specific Heat(J/kgC) 400 503 633 800 895

Material properties for 304 stainless steel

Young’s Modulus(GPa) 199 180 166 150 125
Yield strength(MPa) 206 153 108 82 69
Poisson’s ratio 0.28 0.28 0.28 0.28 0.28
3
Density(kg/m ) 8010 7931 7840 7755 7667
-6
Thermal expansion(1/C10 ) 16.0 17.2 18.2 18.6 19.5
Thermal conductivity(W/mC) 15.26 17.6 20.2 22.8 25.4
Specific Heat(J/kgC) 500 544.3 582 634 686
Table 3 Fatigue damage parameters for 2205 and 304
Materials B-1 S0 b u
-37
2205 1.9510 4.97 0.57 750
-34
304 2.1810 4.44 0.97 505
 Fig.1 Illustration of the welded T-joint

20

Unit: mm

12
100

304

Pat
hP
25

2205

100 150
Fig. 2 Schematic procedure of residual stress measurement by impact indentation

method

Steel punch rod:

certrain load transformation between
resistance and strain
Test point
Strain gage:
orthometric distribution
¦Å
2

¦ 1Å
Fig. 3 Mesh of the welded T-joint
Fig. 4 Fatigue life of SAF2205 duplex stainless steel and 304 austenite stainless

steel

600

500
Stress amplitude (MPa)

400

300

200 Experimental Predicted

304 (R= -1) 304 (R= -1)
304 (R=0.1) 304 (R=0.1)
100 2205 (R= -1) 2205(R= -1)
2205 (R= 0) 2205(R= 0)
0 3
10 104 105 106 107
Fatigue life, N f
Fig. 5 Contour of transverse (a), longitudinal (b) and normal (c) residual stress

distribution
Fig. 6 Residual stress distribution along the path by indentation and finite element

method

800
600
400
Residual stress (MPa)

200
0
-200
Transverse stress by indentation
-400 Longitudinal stress by indentation
-600 Transverse stress by FEM
Longtitudinal stress by FEM
-800
0 5 10 15 20 25
Distance along the path (mm)
Fig. 7 Contour of transverse (a), longitudinal (b) and normal (c) residual stress

distribution after cutting

Fig. 8 Contour of stress amplitude (a), mean stress (b) of the fatigue test specimen

during cycle combined with residual stress

Fig. 9 Fatigue crack initiation position of the weld T-joint
Fig. 10 Fatigue damage curves as number of cycles increase with different applied

load

1.0 Critical value 0.99

0.8
Nominal stress amplitude
Fatigue damage

0.6 200 MPa

160 MPa
120 MPa
0.4

0.2

0.0

101 102 103 104 105 106

Number of cycles, N
Fig. 11 Fatigue life comparisons between experimental and FEM

220
Predicted by FEM with RS
Predicted by FEM without RS
Nominal stress amplitude (MPa)

200
Experimental results

180

160

140

120

104 105 106 107

Number of cycles to failure, N f
Fig. 12 Comparisons of stress amplitude and mean stress with and without

considering the residual stress

800
Stress amplitude without residual stress
700 Stress amplitude with residual stress
Mean stress without residual stress
600 Mean stress with residual stress
Stress (MPa)

500

400

300

200

100
60 80 100 120 140 160 180 200 220
Nominal stress amplitude (MPa)
Fig. 13 Transverse residual stress relaxation and plastic strain at different nominal

stress amplitude

400 0.020
Combined RS
300
Transverse residual stress (MPa)

RS with load
Plastic strain with initial RS 0.016
200
Plastic strain without initial RS
100 0.012

Plastic strain
0
0.008
-100
-200 0.004
-300
0.000
-400
120 140 160 180 200
Nominal stress amplitude (MPa)
Fig. 14 Fatigue life estimation (a), stress amplitude and mean stress (b) of weld

sample and fatigue test specimen considering the residual stress

(a) 300
Initial weld sample
Nominal stress amplitude (MPa)

250 Fatigue test specimen

200

150

100

50

0 3
10 104 105 106 107
Number of cycles to failure, N f
 (b) 600
Stress amplitude of initial weld sample
500 Stress amplitude of fatigue test specimen
Mean stress of initial weld sample
400 Mean stress of fatigue test specimen

Stress (MPa) 300

200

100

0
8 10 12 14 16 18 20
Nominal stress amplitude (MPa)

Fig. 15 Fatigue life (a), stress amplitude and mean stress (b) with different specimen

width
7
(a) 10

Fatigue life of weld T-joint

Number of cycles to failure, N f

106

105 30% width of

the as-weld sample

104
0 20 40 60 80 100 120 140 160
Specimen width (mm)
(b) 300
280
260
240

Stress (MPa)
Stress amplitude
220 Mean stress

200
180
160
140
0 20 40 60 80 100 120 140 160
Specimen width (mm)
 Research highlights

·Fatigue life of a dissimilar welding joint considering residual stresses are studied.

·The fatigue life is estimated by a nonlinear fatigue damage model.

·The effect of the residual stress on fatigue life is studied.

·The cutting on residual stress relaxation leads to fatigue life overestimated.