Fatigue tests and estimation of crack initiation andpropagation lives in AISI 304L butt-welds with
reinforcement intact
P. Johan Singha,*, B. Guhab, D.R.G. Acharb
aMaterials Technology Division, Indira Gandhi Center for Atomic Research, Kalpakkam, IndiabIndian Institute of Technology, Madras, Chennai, India
Received 22 October 2002; accepted 18 January 2003
Abstract
The crack process in a fatigue failure initiating at the toe of a butt weld with reinforcement intact normally comprisestwo major stages: (1) the crack initiation life (Ni): and (2) the crack propagation life (Np). The initiation period was
found to occupy approximately two-thirds of the total fatigue life and consists of the cycles necessary to shake downthe residual stresses in the weld. The crack propagation stage consists of two paths: (1) through thickness (Npx): and (2)along width direction (Npy). This model considers the influence of crack propagation direction in estimating the pro-
pagation life. The notch-root stress approach is used to estimate the crack initiation life, while the fatigue crack pro-pagation lives were estimated using fracture mechanics concepts. A test method is evaluated to estimate the fatiguecrack propagation life using fracture mechanics concepts. The experimental part consists of fatigue tests for gas metalarc welded (GMAW) butt joints of AISI 304L stainless steel with reinforcement intact. Constant amplitude fatigue
tests with stress ratio, R=0 were carried out using a 100 kN servo-hydraulic DARTEC universal testing machine witha frequency of 30 Hz. The predicted lives were compared with the experimental values. This study suggests that thecrack growth parameters are dependent on the crack propagation directions for the same material.
# 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Weld fatigue; Butt joints; Welded fabrications; Initiation life; Propagation life
1. Introduction
It has been recognized [1] that the fatigue life of welded elements without crack-like defect comprises twophases: fatigue crack initiation and fatigue crack propagation. Frequently, the Fatigue crack initiation lifeis predicted by the notch-root stress or notch-root strain approach [2,3]. The fatigue crack propagation lifeof welded elements is usually predicted by use of Paris’ equation [4,5]. The fatigue crack propagation
1350-6307/03/$ - see front matter # 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S1350-6307(03)00017-7
Engineering Failure Analysis 10 (2003) 383–393
www.elsevier.com/locate/engfailanal
* Corresponding author: Tel.: +91-4114-480232; fax: +91-4114-480081.
E-mail address: [email protected] (P.J. Singh).
portion of the fatigue life is a predictable quantity if da/dN as a function of �K has been empiricallyestablished [6,7].
Butt welds tested in fatigue with their reinforcement intact exhibit less fatigue resistance than plain plateor butt welds with their reinforcement removed due to the notch associated with the toe of the weld [7].Fatigue cracks invariably initiate at these locations (Fig. 1) and propagate through the heat-affected zone(HAZ) and base metal in the thickness direction in a direction perpendicular to the applied stress. Thecracks also propagate along the width direction. However, crack propagation is restricted to only the HAZregion along the width direction (that is, along the weld toe profile). The estimation procedure for crackpropagation life in both directions is different for the same material and geometry.
In this paper, constant amplitude fatigue test results of AISI 304L stainless steel butt welds with rein-forcement intact are reported and a new test method using fracture mechanics concepts is outlined andused to predict the fatigue crack propagation life Npy of the butt welds investigated.
2. Experimental
Butt joints with full penetration double-V groove welds were prepared using 6-mm thick AISI 304Laustenitic stainless steel cold rolled plate. The welds were made in the flat position with two passes per sideusing fully automatic gas metal-arc welding process, argon shielding gas and 308L electrodes. Great carewas taken to obtain the best possible joint alignment. Two types of testing specimens (unnotched andnotched) were prepared according to the configurations shown in Fig. 2. The unnotched specimens (Fig. 2a)were used for estimating crack initiation life and crack propagation life along the thickness direction. Thenotched specimens shown in Fig. 2b (SENT type) were used to estimate crack propagation life along widthdirection (Npy). The notches in Fig. 2b were placed in such a way that the roots were situated in the part ofthe heat-affected zone immediately adjacent to the weld (below the reinforcement), since this is the part at
Nomenclature
a, c crack depth, surface half cracklength
A currentai, af initial crack length, final crack
lengthb Basquin’s fatigue strength
exponent�f’ fatigue strength coefficientB specimen widthC, m Paris law coefficientsh bead heightKt, Kf stress concentration factor,
stress notch factorMs, Mt, Mk, �o correction factors in �K-
expressionN number of cyclesNi crack initiation cyclesNp crack propagation cycles
NT total fatigue lifeNpx number of propagation cycles
in thickness directionNpy number of propagation cycles
in width directionP applied loadr toe radiusR stress ratioRm UTSRp proof strengthTp specimen thicknessV voltage�� nominal stress range�K stress intensity factor range�m mean stress� flank angle� root angle�a notch-root stress amplitude�a crack advance step�Kth threshold �K
384 P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393
which the fatigue crack normally starts. The flank angle defining the weld geometry was measured for eachspecimen and the average value is found to be 68� with the standard deviation �4. Welds were tested by X-ray radiography for their soundness.
Constant amplitude fatigue tests with stress ratio, R=0 were carried out using a 100 kN servo-hydraulicDARTEC universal testing machine with a frequency of 30 Hz. For each condition, 12–16 samples weretested. The specimens were tested to complete failure or to an endurance of 2 million cycles, if there was noevidence of fatigue cracking.
For the unnotched specimens, the crack initiation points were determined using resistance type straingauges (micro measurements# EA-06-060LZ-120) and the crack propagation data were determined usingfoil type crack propagation gauges (micro measurements# TK-09-CPB02-005/DP series). The crack pro-pagation gauges consist of a number of resistor strands connected in parallel. This is bonded to a specimenat the weld toe through a connector circuit and progression of a crack through the gauge pattern causessuccessive open circuiting of the strands, resulting in an increase in total resistance. The output is amplifiedand a step curve of strands broken versus time can be obtained in the computer.
The notched specimens are initially fatigue precracked to 0.5 mm and the crack propagation length alongthe width direction was measured using a travelling microscope (100�). The chemical composition andmechanical properties of base metal and filler metal are given in Tables 1 and 2. The welding processparameters used to fabricate the joints are presented in Table 3.
Fig. 1. Crack propagation paths (Npx and Npy) in butt-welded joints with reinforcement intact from the weld toe.
Table 1
Chemical composition (wt.%)
Material/electrode
C Si Mn P S Cr Ni Nb Cu Co N MoBase metal AISI 304L
0.022 0.35 1.79 0.026 0.001 18.32 8.23 0.002 0.40 0.08 0.055GMAW AWS A5.9 ER 308LSi
0.014 0.86 1.73 0.015 0.007 19.68 10.15 – 0.05 – – 0.05P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393 385
Table 2
Mechanical properties
Material/electrode
Plate thickness/electrodediameter (mm)
Rp (MPa)
UTS, Rm(MPa)
Elongation
(%)
Hardness
HB
0.2%
1%AISI 304L
6 297 328 622 54 170GMAWER 308Lsi
1.6/1.2 420 - 600 40 -Table 3
Welding process parameters
Process
Electrode diameter(mm)
V
A Welding speed(mm/c)
Wire feed rate
(m/min)
Gas flow rate
(argon)(lit/min)
GMAW (Fully automatic)
1.2 (root pass) 28–30 200–240 7.5–10 14 16DCEP
1.6 (II–pass) 30–34 235–280 7.5–12 15 16Fig. 2. Details of specimens: (a) without notch (b) with notch.
386 P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393
3. Results and discussion
3.1. Prediction of crack initiation life—using unnotched specimens
Ni was evaluated experimentally using the crack initiation criterion [8–10]. Here, in our investigations theinitiation criterion is the number of cycles required to grow 0.5-mm length of crack from the weld toeunder a particular stress range. To estimate Ni for a notched component, a local stress-life or strain-lifeapproach is commonly used [2,10]. The stress method relates Ni to ��, an equation from Basquin with theMorrow mean stress correction [11] often being used, namely
�a ¼ �0f � �m
� �2Nið Þ
b: ð1Þ
This is a linear relationship when plotted on log–log axes. The notch-root stress amplitude, the stress at theweld discontinuity can be related to the remote stress using the fatigue notch factor [12] so that Eq. (1) becomes:
Ni ¼1
2
D�2Kf
�0f � �m
0B@
1CA
1b
: ð2Þ
The prediction method also requires that the fatigue notch factor be calculated. The fatigue notch factorfor butt joints with reinforcement angle, �=68�, was calculated according to Yung [13] and is 1.774. Thefatigue crack initiation cycles were predicted using Eq. (2). The Basquin constants are given in Table 4.
3.2. Crack propagation along thickness direction—using unnotched specimens
The weld toe fatigue failure conditions of an as-welded joint are related to three special features:
A stress concentration due to the weld toe angle A tensile residual stress field resulting from non-uniform heating of the plate during welding The microstructure of heat affected zone (HAZ) at toe apex
These three features are particularly pronounced in a small region below the toe apex through whichcrack growth can take place.
The general solution for the stress intensity factor range of a semi-elliptical fatigue crack (neglecting theplasticity correction term), as would occur at the toe of weld, is given by [4,14]:
DK ¼MsMtMk
�oD�
ffiffiffiffiffiffi�a
p: ð3Þ
Table 4
Measured constants
Crack growth direction
Paris constants Basquin constants (from initiation analysis)m
C b (�f0��m)Along thickness (Npx)
3.8 3.20e-12 �0.131 1140Along width (Npy)
3.6 1.54e-12P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393 387
s allows for the effect of a free surface at the mouth of the crack. Mt is a correction term for finite
Mthickness and depends on a/tp, and a/2c. �o is the complete elliptic integral and depends a/2c. In our ana-lysis, the effect of crack ellipticity is not considered since the initial crack length (ai=0.5 mm) is consideredsufficiently large to avoid both the low �K region and the effects of ellipticity. Its value may be obtainedfrom standard tables and �o=1 when a/2c=0 [4]. Mk gives a correction according to the local stress con-centration of the weld and depends on �. For a fracture mechanics treatment Maddox [4] introduced theconcept of Mk as a magnification of the stress concentration factor, which would be present for a crack ofthe same geometry but without the presence of the weld. Since the factor Mk is influenced by the stressconcentration factor at the toe, its effect decreases as the crack length increases. Therefore, by including thefactor Mk, it is assumed that, in the initial stage (when a=0), the crack tip is a stress field completelydefined by the magnified stress field of the stress concentration. The product MsMtMk/�o, for butt weldedjoints with reinforcement intact can be expressed as [15]:MsMtMk=F0 ¼ 1:1A1 þ 0:6635A�2
a
tp
�þ 0:5255A�
3
a
tp
�2
þ0:4566A�4
a
tp
�3
þ0:4153A�5
a
tp
�4
; ð4Þ
�
where A1=1.280, A2=�7.40, A3=51.77, A4=�130.40 and A5=112.68 when �=60 .3.3. Crack propagation along width direction—using notched specimens
For automatically welded joints, the weld profile can be very even, and uniform through cracks (a/2c=0)are more likely to occur [16,17]; hence the Gross [18] solution applies:
Fig. 3. Fatigue crack growth curves of unnotched and notched specimens.
388 P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393
MsMt=F0 ¼ 1:122 � 0:231�a
B
� þ 10:55�
a
B
� 2
�21:7�a
B
� 3
þ33:19�a
B
� 4
: ð5Þ
When the crack propagates along the width direction in the welded joints with the reinforcement intact,the variation in Mk along the crack path depends on the weld profile evenness and geometry. The Mk
variation with crack length can be neglected in this case. In automatic welded joints, the weld profile isconsidered to be even [16,17]. The authors assumed Mk equal to the stress concentration factor at theweld toe, which is determined by the weld geometry and the nature of the remote stresses. Thus Eq. (5)becomes:
MsMtMk=F0 ¼ Kt 1:122 � 0:231�a
B
� þ 10:55� a
B
� 2
�21:7�a
B
� 3
þ33:19�a
B
� 4� �
; ð6Þ
where Kt=1.415 when the weld flank angle � 68 degrees.
3.4. Prediction of crack propagation life
The fracture mechanics analysis for most structural steels is based on Paris power law [6],
Fig. 4. Crack propagation lives of unnotched and notched specimens.
P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393 389
da=dN ¼ C DKð Þm: ð7Þ
The stress intensity factor range is a function of the crack length and stress level. Therefore, for testsconducted at a constant stress range, �K becomes a direct function of instantaneous crack length, and thenumber of cycles of crack propagation may be expressed as:
Np ¼
ðafai
1
C DKð Þm da: ð8Þ
The number of cycles of repeated stress necessary to advance the crack from an initial flaw size to a finalcrack size can be calculated using Eq. (8) if an analytical function for the range of stress intensity factor canbe found and integrated. The above integral may be evaluated by finite difference method (Simpson’s rulefor numerical integration) with the aid of a computer.
Np ¼Xafai
DaC DKð Þ
m : ð9Þ
Fig. 5. Comparison between predicted and experimental crack initiation life of unnotched specimens.
390 P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393
Using either Eq. (8) or (9) and letting af equal half the thickness of the specimen, it is possible toestimate the number of cycles spent in crack propagation Npx during the fatigue life of a weld if ai isknown, the material constants C and m are known, and if an analytical function for the SIF range can befound. Similarly, the propagation life Npy can be found out by considering af as equal the width of thespecimen.
The crack growth rate for both the crack propagation directions was calculated. The relationshipbetween the SIF range and the corresponding crack growth rate, da/dn for both specimen types on alog–log scale in terms of best fit line is shown in Fig. 3. The data points mostly correspond to thesecond stage of the sigmoidal relationship of Paris’ equation. �Kth for notched and unnotched speci-mens is approximately 12.5 MPa m1/2 and 8 MPa m1/2 respectively. m and C are given in Table 4. Thedifference in Paris constants may be attributed to the crack propagation direction. The Mk-factorinfluences the crack growth properties (m and C) in the x and y directions. Also, In the case of Npx, thefatigue crack initiates in the coarse-grained heat affected zone (HAZ) then propagates successivelythrough the coarse-grained HAZ, the fine grained HAZ, and finally the base metal. However, in thecase of Npy, the crack growth is confined only in the HAZ regions. It is also seen that the propagationlife (Npy) along the width direction in the notched specimens is short (Fig. 4) when compared to Npx ofunnotched specimens.
Fig. 6. Comparison between the predicted curve for fatigue crack propagation (FCP) life and experimental data for unnotched spe-
cimens.
P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393 391
3.5. Scatter diagram
The fatigue crack initiation lives were evaluated for unnotched specimens using the notch-root stressapproach and are represented in Fig. 5 for comparison with experimental values. A good agreement is shown.
The fatigue crack propagation lives so calculated for unnotched and notched specimens are comparedwith those of test results in Figs. 6 and 7 respectively. It is shown that the experimental values of fatiguecrack propagation lives (Npx and Npy) agreed well with those of the predicted lives.
4. Conclusions
The fatigue crack initiation and propagation lives of AISI 304L stainless steel butt welds with reinfor-cement intact under constant amplitude load can be well predicted separately.
A new test procedure using fracture mechanics concepts is outlined and used to predict the fatigue crackpropagation life in the y-direction (Npy) of the butt welds investigated.
Because the microstructure along the crack path and the stress concentration factor (Kt) are remarkablydifferent for unnotched and notched specimens, the fatigue crack propagation properties vary in both thecases for the same material.
Fig. 7. Comparison between the predicted curve for fatigue crack propagation (FCP) life and experimental data for notched speci-
mens.
392 P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393
Acknowledgements
This work was done at I.I.T., Madras as part of the first author’s PhD programme and financed byAvesta Polarit Research Foundation, Sweden. The authors are indebted to Avesta AB and in particularProf. Hans Nordberg for provision of the stainless steels and the welding materials.
References
[1] Zheng Xiulin, Lu Baotong, Cui Tianxie, Lu Xiaoyan, Lin Chao. Fatigue tests and life prediction of 16 Mn steel butt welds
without crack-like defect. International Journal of Fracture 1994;68:275–85.
[2] Lawrence FV, Mattos RJ, Higashida Y, Burk JD. Estimating the fatigue crack initiation life at welds [ASTM STP 648]. Phila-
delphia: ASTM; 1978 134–158.
[3] Bhuyan G, Vosikovsky O. Prediction of fatigue crack initiation lives for welded plate T-joints based on the local stress-strain
approach. Int J Fatigue 1989;11(3):153–9.
[4] Maddox SJ. An analysis of fatigue cracks in fillet welded joints. International Journal of Fracture 1975;11:221–43.
[5] Itoh Y. Fatigue life prediction for GMA welded butt joints in a C-Mn-Si steel. Welding Journal 1987;2:50s–6s.
[6] Paris PC, Erdogan F. A critical analysis of crack propagation laws. J Basic Engng 1963;85:528–38.
[7] Lawrence FV. Estimation of fatigue-crack propagation life in butt-welds. Welding Journal 1973;52:212s–20s.
[8] Jack R, Price AT. The use of crack initiation and growth data in calculation of fatigue lives of specimen containing defects. Metal
Const 1971;3:416–9.
[9] Testin RA, Young JY, Lawrence FV, Rice RC. Predicting the fatigue resistance of steel weld metals. Welding Journal 1987;4:93
s-98s.
[10] Zheng Xiulin. A further study on fatigue crack initiation life-mechanical model for fatigue crack initiation. Int J Fatigue 1986;
8(1):17–21.
[11] Morrow J. In: Graham F, editor. SAE fatigue design handbook. 1968. p. 21–30.
[12] Peterson RE. Stress concentration factors. New York: John Wiley and Sons; 1987.
[13] Yung JY, Lawrence FV. Analytical and graphical aids for the fatigue design of weldments. Fatigue Fract Engng Mater Struct
1985;8(3):223–41.
[14] Maddox SJ. Assessing the significance of flaws in welds subjects to fatigue. Welding Journal 1974;53(9):401s–409s.
[15] Burk JD, Lawrence FV. Influence if bending stresses on fatigue crack propagation life in butt joint welds. Welding Journal 1977;
2:61s–66s.
[16] Sahili A, Albrecht P. Fatigue life of welded stiffeners with known initial cracks [ASTM STP 833]. ASTM; 1984, 193–217.
[17] Gurney TR. Finite element analysis of some joints with the welds transverse to the direction of stress. Weld Res Int 1976;6(4):40–
72.
[18] Gross B et al. Stress intensity factors for a single-edge-notch tension specimen by boundary collocation of a stress function.
NASA Tech. Note D-2395, August 1964.
P.J. Singh et al. / Engineering Failure Analysis 10 (2003) 383–393 393