WELDING RESEARCH

SUPPLEMENT TO THE WELDING JOURNAL, AUGUST 1992 Sponsored by the American Welding Society and the Welding Research Council

Fatigue Life Prediction of Cruciform Joints Failing at the Weld Toe

A two-stage model appears to offer the most accuracy in predicting fatigue life
BY M . SKORUPA

ABSTRACT. The objective of this study is to select a theoretical model to predict the fatigue life of a weld. The experimental part consists of fatigue tests within the life range of from 105 to 2 X 106 cycles for cruciform welded joints in t w o structural steels under constant amplitude axial loading with a constant positive stress ratio: the fatigue crack initiating at the weld toe. The fatigue lives were then calculated according to the two-stage approach in which both the fatigue crack initiation phase and the crack propagation phase were considered, and according to the one-stage approach which assumes that the total fatigue life is spent in crack propagation. The local strain method was applied to calculate the fatigue crack initiation life, while the fatigue crack propagation life was estimated using fracture mechanics concepts. The important feature of these analyses was that the material properties needed were estimated via hardness measurements. The fatigue lives obtained by the two-stage model fall within the 0.95 confidence limits of the experimental data and are conservative in the sense that the average ratio of the observed-to-calculated life is 1.15. The one-stage model results in life estimates an order of magnitude lower than those derived from the two-stage model. The study suggests that the two-stage model rather than the one-stage model enables accurate fatigue life predictions even for welds containing undercuts.

and the two-stage models where N f = N, + N p

(2)

The approaches according to Equations 1a, 1b and 2 will be further termed as the P, I and l-P models, respectively. In Equations 1 and 2, Nf is the total fatigue of life, N p is the crack propagation life and N, is the crack initiation life. For small metal components, N, is defined as the number of cycles required to develop a dominant crack of a so-called initial size (ai), usually of the order of tenths of a mm. The meaning of N p in Equations 1a and 2 is not the same. N p in Equation 2 is the number of cycles spent in crack growth from a, to final fracture while N p in Equation 1a accounts also for the earliest phase of crack growth from a flaw or a microstructure defect. Usually, low-cycle-fatigue theory is adopted to estimate Nj, and fracture mechanics is employed to calculate N p . Disagreement prevails on which of the models defined by Equations 1 and 2 best reflects fatigue behavior of welds. A widespread opinion that defects, usually cracklike in shape, are unavoidable even in high-quality welds involves the predominant use of the P model (Refs. 1-3). However, successful estimates of weld fatigue life have also been made utilizing the two-stage approach (Ref. 4). Fatigue analysis of welds requires that strength effects of the welding process be

accounted for. The most important of these are: geometrical variability, change of material properties in the weld neighborhood and residual stresses. The objective of this paper was to select the theoretical model most adequate for predicting the total fatigue life of a weld. The study focused on a cruciform welded joint in t w o structural steels under constant amplitude axial loading. The dimensions of the weldment were chosen such that failure occurred at the weld toe. Fatigue lives calculated according to the two-stage and one-stage models for a defectless weld geometery and in the presence of undercut were compared with the observed results. Experimental Tests Specimens Cruciform welded specimens for fatigue testing were fabricated from structural steel plates. The steels were St3S (carbon mild steel) and 18C2A (low-alloy steel). The chemical compositions and mechanical properties of these steels are given in Table 1. The welding procedure was manual metal arc in the horizontal position. Based on Ref. 1, the ratios of the weld leg length I to the plate thickness t were chosen such to promote failure at the weld toe under axial fatigue loading. The St3S plates were welded in longer pieces and then saw-cut into narrow specimens. Each 18G2A specimen was welded separately. The specimen side surfaces were milled to discard weld start and stop areas, then ground. The specimen geometry is shown in Fig. 1 and the dimensions are given in Table 2. Variable geometry parameters (Fig. 2), knowledge of which is vital to analytical fatigue life prediction, namely, the weld contact angle fi, the notch root radius at the weld toe r, the depth of possible undercut at the weld toe d and misalignment (eccentricity e and angular distortion a),

KEY W O R D S Introduction Analytical procedures used for predicting fatigue behavior of notched structural members can be broadly classified as the one-stage models assuming that Nf = N p (la) or Nf = N (1b) M. SKORUPA is with the University of Mining and Metallurgy, Krakow, Poland. Modeling Fatigue Life Cruciform Joints Structural Steels Weld Toe Cracking Crack Initiation Crack Propagation Local Strain Fracture Mechanics Hardness Measure

W E L D I N G RESEARCH SUPPLEMENT I 269-s

Table 1Chemical Composition and Mechanical Properties of the Steels Used for Welded Specimens Chemical Composition ("o)
Steel St3S 18C2A C 0.18 0.11

Mechanical Properties
Cr Cu 0.020 0.040 Su (MPa) 440 525 Sy (0.2%, MPa) 282 359 RA

Mn
0.47 1.78

Si
0.15 0.54

P 0.037 0.026

S 0.034 0.020

"0
58 52

0.020 0.020

Fig. 7 Test specimen geometry. were measured. The measurements of the fi and r values were made in side surfaces of the specimens using an optical microscope at a magnification of 10X. While only small variations in the < 3 values were observed due to very smooth weld profiles (Fig. 3A), the measured r values showed a large amount of scatter (Fig. 3B). Undercuts were macroscopically examined at magnification of about 70X in 19 longitudinal sections of the St3S specimens and in 35 sections of the 18G2A specimens; four weld toe regions being observed in each section. Preparation of that inspection involved wet polishing with successively finer grades of emery paper. Undercut depths were measured perpendicularly to the intended direction of loading from a horizontal projection of the base plate line Fig. 2. No undercuts except several very shallow (of depth be-

low 0.05 mm) depressions at the weld toe were detected in the St3S specimens. However, in the 18G2A specimen sections 82 undercuts were found for the total number of 140 weld toe regions inspected. The measured depths d did not exceed 0.1 mm (Fig. 3C), the notch root radius at the bottom of undercut r' being always larger than d. The way of fabrication of the 18C2A specimens achieved a very good joint alignment. Misalignments, namely eccentricity up to 20% of the plate thickness and angular distortions up to 3 deg (Fig. 2) were only found in part of the St3S specimens. As a rule, the values of e and a measured in both sides of a given specimen were not the same. The maximum values of e and a for the St3S specimens are listed in Table 3. Fatigue Tests The specimens were fatigue tested until failure by breaking into t w o parts in an electrohydraulic machine under constant stress amplitude at a frequency of = 1 5 Hz. The stress ratio R was 0.1 for the St3S specimens and 0.2 for the 18G2A speci-

Table 2 - -Dimensions of Welded Specimens Ratio l/t 1 1.5 1.5 Dimensions (mm) B L I 12 15 15 35 35 60 400 400 400

mens. All the 1/t = 1.5 specimens failed at the weld toe, while in part of the 1/t = 1 specimens weld failures were observed in accordance with other reported results (Ref. 1). Tables 3 and 4 give the fatigue test results for the specimens failing at the weld toe. A statistical analysis of the fatigue data for the Series 1/t = 1.5 St3S specimens and 18G2A specimens was carried out according to Ref. 5 assuming the following linear relationship between the fatigue life Nt- and the nominal stress range AS: logNf = A + B logAS (3) The small number of Series l/t = 1 St3S specimens failing at the weld toe prohibited performing a statistical analysis of the corresponding fatigue data. The fatigue test results together with the relevant regression lines and 0.95 confidence limits are shown in Figs. 4 and 5. The higher width of the scatter band for the St3S welds (Fig. 4) compared to that for the 18G2A welds (Fig. 5) may predominantly result from the occurrence of misalignments in the former specimens. The reason fatigue resistance of the St3S specimens is significantly higher than that of the 18G2A specimens may be attributed to

Material St3S 18G2A

t 12 10 10

Table 3Fatigue Test Results and Misalignments for St3S Specimens Misalignments Angular Eccentricity Distortion 1.5 2.0 1.5
0 1.5 1.0 1.0 0 1.5

Material

Ratio l/t

Nominal Stress Range AS (MPa) 298.5 248.3 195.0

Total Fatigue Life Nf (cycles) 78 160 160 000 245 500 907 800 1 224 500 1 349 000
146 000 254 000 260 000

St3S

288.5

309 000 201 246 282 630 476 671 844 988 500 000 000 000 000 000 000 000

0
1.0 2.0 2.0 1.0 1.0 2.0 1.0 0 2.0 1.0 1.5 0 0 1 2 0

264.3

1 . 5

236.4

205 Fig. 2 Variable geometry parameters and crack path for test specimens. 270-s I A U G U S T 1992

1 005 000 1 367 000 1 666 000 Unbroken

15i

iio
8 5

300 250 200

39 41 43 45 47 49 51 53 Weld Contact Angle /3, .15

Fatigue Data Points Regression Line Q95 Confidence Limits

Nf Observed Results

B
a _a o fi 5 0^ 0.5 15 2.5 3.5 4.5 5.5 6.5 75 85 Weld Toe Radius r. mm 801 g60 o
LO

10 5 N|, N f , cycles

u ,0 6 10

6 2*10u

Fig. 4Observed and predicted with two-stage approach fatigue lives for St3S l/t = 7.5 welded specimens.

o fi20

"0 20 40 60 80 100 Depth ot Undercut d ^ m

F/g. J Measured variations of geometry parameters for test specimens: AWeld contact angle fi, BRoot radius at the weld toe r; C Depth of undercut d.

more favorable weld toe geometry (lack of undercuts) in the former. The present fatigue data fall within the reported scatter bands obtained with similar specimens (Ref. 1). Microscopic examination of the cracked specimens revealed that fatigue cracks initiated and propagated within the heataffected zone (HAZ). Fatigue crack paths were approximately straight lines sloped to the direction transverse to the main plate at an angle <p ~ 15 deg for l/t = 1 . 5 and rj> 8 deg for l/t = 1 (Fig. 2). An observation of the fracture surfaces indicated that the specimens failed when the crack depth reached on an average about 0.35L

Peterson's equation reads Kt-1 1 + Kf

1 +oc/t

(4)

where Kt is the elastic stress concentration factor, r is the notch root radius and a is a material constant given by 2.32 X 10 4 S U -1 ' 8 (mm), S being the ultimate strength of material in MPa. If the analytical Kt vs. r relationship is known, Equation 4 can be differentiated with respect to r to determine the critical notch root radius rc for which Kf obtains a maximum. For the defectless weld toe geometry, the following Kt vs. r relationship was obtained from finite element analysis (Ref. 6): K, = 1 + C f ( r / t p (5) where the constants C i and C2 depend on the ratio of 1/t. The Kf function given by Equation 4 combined with Equation 5 passes through a maximum for C 2 -E 1 rc = - ^ - - a (6) In the presence of undercut (Fig. 2) the theoretical stress concentration factor at the weld toe was approximated by (Ref. 7)

Fatigue Life Calculations

Geometrical Variability Based on the geometry measurements (Fig. 3), the variations in the angle fi are ignored and an average value of 8 = 45 deg is assumed. In order to cope with the variable nature of other geometry parameters at the weld toe, it is assumed, after Lawrence, ef al. (Ref. 4), that fatigue cracking starts at the location where the fatigue notch factor calculated through Peterson's equation adopts its maximum value.

Table 4Fatigue Test Results for 18G2A Specimens Nominal Stress Range AS (MPa) Total Fatigue Life N f l (cycles) 135 000 180 000 159 600 363 700 236 100 324 900 346 300 371 500 524 100 821 300 861 600 1 159 900 1 291 100 Unbroken

Ktu = K,(1+2V(d7FT]

(7)

w h e r e K t is given t h r o u g h Equation 5. Using Equations 4 and 7, the fatigue

Material

Ratio l/t

Table 5Empirical Relationships between Mechanical Properties and Hardness DPH of Steel Function of DPH 3.45 DPH 2.68 DPH - 138 3.3 DPH -I- 370 1

223
200 175 171 18C2A

Property Ultimate tensile strength, Su (MPa) Yield strength 0.2%, Sy (MPa) Fatigue strength coefficient, o-j'(MPa) Fatigue strength exponent, b

Source Ref. 8 Ref. 9 Ref. 9 Ref. 9

1.5

168

-glog(2.1 +
Cyclic yield strength, o-y'(MPa) Transition fatigue life, 2N tr (reversals) 266/DPH) 2.1 DPH 5.7 X 10s exp (-0.017 DPH) Ref. 8 Ref. 9

140

119

W E L D I N G RESEARCH SUPPLEMENT I 271-s

to occur in the St3S welds before cutting up the plates, they were relieved by the slicing operation (Ref. 1). Therefore, residual stresses were ignored in the present analysis. Residual stresses when present can be allowed for in the crack initiation analysis as a static prestress (Refs. 12 and 13). Crack Initiation Analysis
o Fatigue Data Points Regression Line 0.95 Confidence Limits Nf Observed Results 1Cfc Nf>Nj,Np, cycles 2*106

10-

Fig. 5 Observed and predicted with two-stage approach fatigue lives for I8G2A welded specimens.

notch factor for a weld with undercut was obtained as K t (1 4- v w ) ) - 1 From Equation 8, the Kfmax condition for a weld with undercut exists if f 1)2 4- 4Kt2 V(a 2 (K, ad)-a

(9) 2 K, a V ^ " J Based on the inspection of the weld toe regions described earlier, the St3S specimens were assumed to be defectless while for the 18G2A specimens the presence of undercuts was allowed for. According to the Kfmax concept, Equations 6 and 9 define the values of the notch root radius adopted in the crack initiation and crack propagation analyses for the St3S and 18G2A specimens, respectively. As the fatigue notch factor according to Equation 8 is a monotonic function of r and d, the minimum measured r value equal to 0.5 mm and the maximum measured d value equal to 0.1 mm were assumed for the 18G2A specimens.

Other material properties of the HAZ were calculated through the following relationships that are known in the low-cycle-fatigue theory: cyclic strain hardening exponent n ' = b/c (10) cyclic strength coefficient K' =o- y '/(0.002) n ' (11) fatigue ductility coefficient as the mean of = (^'/K')1 and tf' = a,'(2N tr ) b -VE (13) Table 6 lists the material properties calculated from the relationships given in Table 5 and from Equations 10 to 13. The values of r(, r c ' and Ktmax in Table 6 result from Equations 4 to 9. Initial and Final Crack Size Where the l-P model (Equation 2) is imployed, the initial crack size a; must be defined in order to bridge between the initiation and propagation analyses. A common practice is to assume the ai value arbitrarily (usually 0.25 mm), although nonarbitrary definitions of aj were also proposed (Refs. 10, 11). For the considered geometry and material it was shown in Ref. 7 that the value of 0.25 mm lies between those calculated according to Refs. 10 and 11. Therefore, a: = 0.25 mm was adopted in this study. Based on the examination of fracture surfaces, the final crack size af = 3.5 mm was assumed. Residual Stresses The way of fabrication of both St3S and 18G2A specimens precluded occurrence of high residual stresses. With the narrow 18G2A specimens, making a weld produced a more or less even temperature across the plates and consequently could only create low residual stresses. Although relatively high residual stresses were likely (12)

The crack initiation lives (Nj) were estimated with the local strain approach (Ref. 14). The mathematical procedure to calculate the local stress and strain amplitudes (<r and e, respectively) and the local mean stress (<r0) is illustrated in Fig. 6 where R is the stress ratio, S and e denote the nominal stress and strain, respectively, and the subscripts " 1 " and " a " to stress or strain correspond to the value at the end of the first loading reversal and to the amplitude, respectively. Since the material properties involved in the monotonic stress-strain curve equation cannot be related to hardness, cyclic softening of HAZ material was ignored. The Ni values were calculated through the equation from Morrow (Ref. 15) a* = (fff' - <r0)(2Ni)b (14) in which only the material properties that can be directly related to hardness are utilized. To allow for the effect of cyclic relaxation of the local mean stress on Nj the constant o0 value in Equation 14 was replaced by (Ref. 16) <ro.2N = <T0 (2N-1)k (15) where ff0,2n is the current value of mean stress at any reversal 2N and k denotes the relaxation exponent. From Ref. 12, k was expressed through the following empirical relationship valid for metals: k = -31912.5 fp /(Ee tr ) (16) where ep is the plastic strain amplitude, tXr: the transition strain (half strain amplitude corresponding to the transition fatigue life 2N,r), and E: the Young modulus in MPa. The operational details and the programming flow diagrams of the procedure used to compute Nj can be found elsewhere (Ref. 7). Crack Propagation Analysis Since the material constants associated with fatigue crack propagation rate cannot be estimated from hardness, only the concepts where crack growth rate is described in terms of low-cycle-fatigue material properties were used in the crack propagation analysis, namely the LEFM model from Majumdar and Morrow (Ref. 17) and the EPFM approach of Usami (Ref. 18). Both above mentioned concepts were shown in Ref. 7 to provide for a welded joint in mild steel the N p values, which correlated well with those obtained using the Paris equation.

Estimating H A Z Material Properties through Hardness Measurements The properties of the HAZ material that are involved in the fatigue process of the test specimens may be different from those of the base metal. As HAZ material is difficult and expensive to test, all the required material parameters were either estimated from hardness measurements at the weld toe using empirical relationships given in Table 5 or assumed. The fatigue ductility exponent c = 0.6 was assumed. The relationships of Sy and <7y' with hardness depend on the type of steel. As discussed in Ref. 7, those assumed in Table 5 correlate best with reported test results for various HAZ materials in steels (Refs. 4, 9). 272-s I A U G U S T 1992

A c c o r d i n g t o the m o d e l of M a j u m d a r a n d M o r r o w (Ref. 17) the material ahead o f t h e crack tip w i t h i n the r e v e r s e d plastic z o n e is c o m p o s e d of the uniaxial " f a t i g u e e l e m e n t s " o f a w i d t h o f 2P* w h e r e p* is a " m i c r o s t r u c t u r e size." Assuming that f a tigue crack extension occurs d u e t o t h e successive fatigue failure of each e l e m e n t M a j u m d a r a n d M o r r o w d e d u c e the f o l l o w i n g expression f o r fatigue crack g r o w t h rate: da dN
iW c

Table 6Mechanical Properties of St3S HAZ and 18G2A HAZ Material Property Hardness, DPH Ultimate strength, Su (MPa) Yield strength 0.2%, Sy (MPa) Young modulus, E (MPa) Cyclic yield strength, a y '(MPa) Fatigue strength coefficient, 07'(MPa) Fatigue strength exponent, b Fatigue ductility coefficient, tf' Fatigue ductility exponent, c Cyclic hardening exponent, n ' Cyclic strength coefficient, K'(MPa) Transition fatigue life, 2N tr (reversals) Peterson's material constant, a (mm) Critical notch root radius, rc [ r c ' ] (mm) Maximum fatigue notch factor, Kfmax St3S HAZ 18G2A HAZ 153 528 272 200000

170 587
318 200000 357

930 -0.094 0.878

-0.6

0.157 947
31678 0.242 0.167<a) 0.157<b> 1.995 w 2.93<b>

321 875 -0.097 0.937 -0.6 0.162 878 42294 0.292 [0.74]
3.12

- [ 4 ( 1 + n')e y ]

(17)
(a) l/t = 1.5 (b) l/t = 1

w h e r e AK is the stress intensity factor range, e / is the cyclic yield strain, and t h e meaning of t h e o t h e r symbols is explained in Table 6. T h e validity of Equation 17 is limited b y the c o n d i t i o n Rp 2p* (18) w h e r e R p is the r e v e r s e d plastic z o n e size. The physical i n t e r p r e t a t i o n of p* given b y M a j u m d a r and M o r r o w (Ref. 17) suggests that it represents the m e a n distance b e t w e e n the m a j o r m i c r o s t r u c t u r e d e f o r m a t i o n barriers. This distance is t h o u g h t t o b e closely related t o the length o f a n o n p r o p a g a t i n g surface crack at t h e fatigue limit (Ref. 19). A c c o r d i n g t o Usami (Ref. 18), the diameter (d p c ) o f such a crack in a metallic material is a material constant w h i c h can b e expressed as d p c = 1.633 X 1 0 - ( S y / E ) " ( m m ) (19) The a b o v e m e n t i o n e d findings f r o m Refs. 18 a n d 19 are the rationale f o r assuming in the present analysis that p* equals d p c . Equation 17 w a s only utilized t o calculate the crack p r o p a g a t i o n lives for the St3S specimens. T h e a p p r o a c h o f Usami (Ref. 18) w a s applied t o d e r i v e t h e N p estimates f o r t h e 18G2A specimens since Equation 18 w a s n o t satisfied f o r 18G2A H A Z material e v e n at t h e highest stress level c o n s i d e r e d . From e x p e r i m e n t a l tests o n various steels, Usami postulates a c o r relation b e t w e e n the n o r m a l i z e d crack g r o w t h rate ( d a / d n ) / a e , a e being the e f f e c t i v e crack length (half length of a central crack in an infinite b o d y ) a n d the e f f e c t i v e local strain range given b y Aeeff = Ae/(1-R), R < O Ateff = Ae , R > O (20) w h e r e Ae a n d R are the t o t a l local strain range and t h e local stress ratio, respectively, at a distance of a f r o m t h e n o t c h r o o t in the u n c r a c k e d b o d y . Usami's experimental data points can b e f i t t e d in w i t h t h e f o l l o w i n g relationships:
7 2

( d a / d N ) / a e = 12.3(Ae eff ) 2 Aeeff < 9 X 10~ 4 ( d a / d N ) / a e = 64800(Ae e f f ) 9 2 2 9 X l O " 4 < Aeeff < 5 X l O " 3 ( d a / d N ) ( / a e = 101(Ae e ff) 2 Ae ef f 2= 5 X 10~ 3

rule. The stress intensity f a c t o r solution w a s d e r i v e d b y finite e l e m e n t m e t h o d , as d e s c r i b e d t h o r o u g h l y in Ref. 6. Comparison of Predicted and (21) O b s e r v e d Results Two-Stage Approach In Figs. 4 a n d 5 the total fatigue lives Nf p r o v i d e d via the t w o - s t a g e l-P m o d e l (Equation 2) are c o m p a r e d w i t h the f a -

F r o m Ref. 18, Equation 21 p r o v e d valid f o r a n u m b e r of steels. In o r d e r t o calculate the N p values Equations 17 a n d 21 w e r e integrated n u merically in 24 steps e m p l o y i n g Simpson's

U) (I)
1_

ti
u (_
F 0 ,

/V
\ /

/ J)
0

' V ~ (

00

1 2

Number of reversals, 2N 6-| i,=Kf S,e.|

10

Fig. 6 Schematic of notch stress and strain calculation.

W E L D I N G RESEARCH SUPPLEMENT I 273-s

2*10c

1.0 ,N0 DEFECT

N.05y \ UNDERCUT Xf N f , cycles. Fig. 8 Predicted percentage of total life devoted to fatigue crack initiation as a function of total life for sound weld and for weld with undercut. tf 210o

i/t=io.
* 18G2A

a single data p o i n t , the predictions agree w i t h the e x p e r i m e n t a l results w i t h i n a fact o r of 2. T h e predictions are conservative in the sense that the average ratio of the o b s e r v e d - t o - c a l c u l a t e d life is 1.15. G e n erally l o w e r discrepancies b e t w e e n t h e actual and e s t i m a t e d results o b s e r v e d in Fig. 7 f o r t h e 18G2A specimens c o m p a r e d t o those f o r the St3S specimens are aparently d u e t o the higher scatter in t h e fatigue data in the latter case. In Fig. 8 the p r e d i c t e d p e r c e n t a g e o f total life spent in crack initiation is p l o t t e d as a f u n c t i o n of total life. For a s o u n d w e l d , the figure s h o w s that w i t h i n t h e c o n s i d e r e d life range crack initiation consumes a prevailing f r a c t i o n o f life w h i l e in the presence of u n d e r c u t , crack initiation life d o m inates only in the l o n g life regime (Nf > 5 X 10 5 cycles). One-Stage Approach As discussed previously, the d i f f e r e n c e b e t w e e n life estimates a c c o r d i n g t o the t w o - s t a g e a n d t h e one-stage (P m o d e l ) a p p r o a c h e s arises f r o m the different w a y s of p r e d i c t i n g the n u m b e r of cycles t o d e v e l o p a crack of size a;. In Fig. 9 t h e evaluations of that n u m b e r of cycles f o r t h e c o n s i d e r e d c r u c i f o r m w e l d b y using the l-P m o d e l a n d the P m o d e l are c o m p a r e d , Usami's c o n c e p t (Ref. 18, see Equations 2 0 a n d 21) being e m p l o y e d in the latter case. It can b e seen in Fig. 9 that t h e P m o d e l gives m o r e c o n servative predictions o n t h e life t o f o r m a crack of size aj than the l-P m o d e l . T h e discrepancies b e t w e e n that life estimate f r o m b o t h a p p r o a c h e s increase w i t h d e creasing stress level t o reach the value o f 9.1 f o r the s o u n d w e l d and 45 f o r the w e l d w i t h u n d e r c u t . Considering that the l-P m o d e l has b e e n previously s h o w n t o yield slightly c o n s e r v a t i v e Nj estimates (Figs. 4 , 5 a n d 7), f r o m Fig. 9, it is clear that f o r lives greater than 10 5 cycles t h e P m o d e l c a n n o t entirely a c c o u n t for t h e total fatigue life e v e n f o r a w e l d containing u n d e r c u t .

105 Observed Fatigue Life N f l cycles

Fig. 7Comparison between observed and predicted

106

2*106

fatigue lives for all test specimens.

tigue test data f o r the w e l d e d specimens. T h e discrepancies b e t w e e n the p r e d i c t e d AS vs. Nf diagrams and the regression lines are d e p i c t e d b y ratios o f Nf values given f o r the t w o stress levels that c o r r e s p o n d t o the highest a n d the l o w e s t stress range c o n s i d e r e d in t h e experimental tests. F r o m Figs. 4 a n d 5, it is seen that the p r e d i c t e d AS vs. Nf curves fall w i t h i n the scatter bands o f t h e fatigue data. Since the slopes

of the theoretical AS vs. Nf diagrams in Figs. 4 a n d 5 are l o w e r than these of the regression lines, t h e m e a n fatigue lives (represented b y the regression lines) are u n d e r e s t i m a t e d at higher stress levels and slightly o v e r e s t i m a t e d at l o w e r stress levels. T h e c o m p a r i s o n b e t w e e n t h e actual a n d calculated fatigue lives for all the fatigue specimens is given in Fig. 7. Except

oo <

50

\M0DEL WELD\. l-P T O E ^ \

NO DEFECT (St3S) UNDERCUT (18G2A)
~lI I I
ll
E

10Life to Form Crack of Size O j , cycles Fig. 9 Comparison of estimates on life to form a crack of size a, according to two-stage and onestage model for sound weld and for weld with undercut.

10

Final Remarks A l t h o u g h based o n the material p r o p erties estimated in a r o u g h w a y , t h e fatigue lives p r e d i c t e d using the t w o - s t a g e

274-s I A U G U S T

1992

a p p r o a c h are in g o o d a g r e e m e n t w i t h t h e e x p e r i m e n t a l results. Utilizing in fatigue analyses o f w e l d s t h e H A Z material p r o p erties estimated via hardness m e a s u r e ments at t h e w e l d t o e w a s first p r o p o s e d b y L a w r e n c e , ef al. (Ref. 4). T h e subseq u e n t analyses, h o w e v e r , w e r e c o n f i n e d t o t h e predictions o f the l o n g life fatigue strength of w e l d s w i t h t h e use o f t h e o n e stage I m o d e l (Ref. 20). This study including t h e crack p r o p a g a t i o n p e r i o d estim a t e d according t o t h e c o n c e p t s in w h i c h the fatigue crack g r o w t h rate is expressed in terms o f t h e low-cycle-fatigue material properties, e n a b l e d life predictions w i t h i n the range o f f r o m 10 5 t o 2 X 1 0 5 cycles. Conclusions T h e t w o - s t a g e a p p r o a c h , including b o t h the fatigue crack initiation a n d p r o p a g a t i o n phases, enables o n e t o estimate total fatigue lives in t h e life regime o f 10 5 t o 2 X 10 6 cycles, w h i c h agree w i t h i n a fact o r o f 2 w i t h experimental data f o r crucif o r m w e l d e d specimens failing at t h e w e l d toe. T h e study supports t h e applicability o f h e a t - a f f e c t e d z o n e material properties estimated via hardness measurements t o w a r d s fatigue life predictions o f w e l d s . The one-stage a p p r o a c h , w h i c h n e glects t h e fatigue crack initiation phase, c a n n o t entirely a c c o u n t f o r t h e total f a tigue life, e v e n f o r w e l d s containing u n dercuts. References 1. Gurney, T. R. 1979. Fatigue of Welded Structures. Cambridge, U.K. 2. Yamada, K., and Hirt, M.A. 1982. Fatigue crack propagation from fillet weld toe. /. of the Structural Division. ASCE 108, No. ST7, pp. 1526-1539. 3. Smith, I.F.C., and Smith, R.A. 1983. Fatigue crack growth in a fillet welded joint. Eng. Fract. Mech. 18: 861-869. 4. Lawrence, F.V., Ho, N.)., and Mazumdar, P.K. 1981. Predicting the fatigue resistance of welds. Annual Review of Materials Science 8:402-425. 5. Statistical analysis of linear or linearized stress-life and strainTife fatigue data. 1980. ASTM Standard E739-80. 6. Skorupa, M Braam, H and Prij, J. 1987. Applicability of approximate K| solutions t o wards cracks at weld toes. Eng. Fract. Mech. 26: 669-681.

7. Skorupa, M. 1989. Predicting the fatigue life of welds. Scientific Bulletins of the Stanislaw Staszic Academy of Mining and Metallurgy. No. 1257. Mechanics, Bull. 18 (in Polish). 8. Morrow, J. 1965. Internal Friction, Damping and Cyclic Plasticity. ASTM STP 378, pp. 45-87. 9. McMahon, ).C, and Lawrence, F.V. 1984. Predicting fatigue properties through hardness measurements. FCP Report No. 105, University of Illinois at Urbana-Champaign. 10. Chen, W . 1980. A model for joining crack initiation and propagation analyses. Ph.D. Thesis, University of Illinois at Urbana-Champaign. 11. Cameron, A.D., and Smith, R.A. 1982. Fatigue life prediction for notched members. Int. J. of Pressure Vessels and Piping 10: 205217. 12. Burk, J.D. 1978. The effect of residual stresses on weld fatigue life. Ph.D. Thesis, University of Illinois at Urbana-Champaign. 13. Skorupa, M. 1990. Fatigue crack initiation life prediction for welded joints by low cycle fatigue approach. Fat. Fract. Eng. Mater. Struct. 13: 597-613. 14. Dowling, N.E., Brose, W.R., and Wilson, W.K. 1977. Fatigue Under Complex Loading Analysis and Experiments. Advances in Engineering. SAE 6: 55-84. 15. Morrow, J. 1968. SAE Fatigue Design Handbook, pp. 21-30. 16. Jhansale, H.R. and Topper, T.H. 1973. Cyclic Stress-Strain Behavior. Analysis, Experimentation, and Failure Prediction. ASTM STP 519, pp. 246-270. 17. Majumdar, S and Morrow, |. 1974. Fracture Toughness and Slow Stable Cracking. ASTM STP 559, pp. 159-182. 18. Usami, S. 1986. Small Fatigue Cracks. The Metallurgical Society Inc., pp. 559-585. 19. Miller, KJ. 1987. The behavior of short fatigue cracks and their initiation. Fat. Fract. Eng. Mater. Struct. 10: 75-91 and 93-113. 20. Yung, J.Y., and Lawrence, F.V. 1985. Analytical and graphical aids for the fatigue design of weldments. Fat. Fract. Eng. Mater. Struct 8:223-241. 21. Testin, R.A., Yung, j.-Y., Lawrence, F.V., and Rice, R.C. 1987. Predicting the fatigue resistance of steel weldments. Welding Journal 66: 93-s to 98-s.

d
dpc e E

k
Kl Kf (Kfmax) Ki (Ktu)

k' I n' N Nf Ni NP Ntr

r(rc) r' (rc')

R RP S Su Sy a

P % f / p*

a
Co
Cf' ffy

<P Subscripts

Fatigue ductility exponent Depth of undercut Diameter of nonpropagating surface crack Nominal strain Modulus of elasticity Relexation exponent Stress intensity factor Fatigue notch factor (maximum value) Stress concentration factor (at the bottom of undercut) Cyclic strength coefficient Weld leg length Cyclic hardening exponent Number of cycles Total fatigue life Crack initiation life Crack propagation life Transition fatigue life Weld toe radius (critical value) Root radius at the bottom of undercut (critical value) Stress ratio Reversed plastic zone size Nominal stress Ultimate tensile strength Yield strength 0.2% Angular distortion Weld contact angle Notch strain Cyclic plastic notch strain Transition strain Fatigue ductility coefficient Cyclic yield strain Microstructure size from Majumdar and M o r r o w equation Notch stress Notch mean stress Fatigue strength coefficient Cyclic yield strength Slope of crack path

Appendix
a ae 3f ai a Crack size Equivalent crack size Final crack size Initial crack size Peterson's material c o n stant Fatigue strength e x p o n e n t

Amplitude 1 The first reversal A Range Abbreviations Elastic-plastic fracture EPFM mechanics HAZ Heat-affected zone Linear-elastic fracture LEFM mechanics

WELDING RESEARCH SUPPLEMENT I 275-s

Nitrogen in Arc Welding A Review

WRC Bulletin 369 December 1991

By IIW Commission II In 1983, Commission II of the International Institute of Welding (IIW) initiated an effort to review and examine the role of nitrogen in steel weld metals. The objective was to compile in one source, for future reference, the available information on how nitrogen enters weld metals produced by various arc welding processes, what forms it takes in these welds, and how it affects weld metal properties. This bulletin contains 13 reports and several hundred references related to Nitrogen in Weld Metals that has been prepared as a review to show the importance nitrogen has in determining weld metal properties. Publication of this report was sponsored by the Welding Research Council, Inc. The price of WRC Bulletin 369 is $85.00 per copy, plus $5.00 for U.S. and $10.00 for overseas, postage and handling. Orders should be sent with payment to the Welding Research Council, Room 1301, 345 E. 47th St., New York, NY 10017.

Research on Modern High-Strength Low-Alloy Steel Welding

WRC Bulletin 373 June 1992

(1) Influences of Steel Composition and Welding Procedure on the HAZ Toughness of Thick-Section Structural Steels
By P.L. Harrison and P. H. M. Hart

(2) Heat-Affected Zone Properties of Thick-Section Microalloyed Steels A Perspective

By F. Heisterkamp, K. Hulka and A. D. Batte

(3) Experience in Fabricating New Types of Offshore Plate and Linepipe

By P. Tuvnes and I. Harneshaug

(4) Influence of Local Brittle Zone on HAZ Toughness of TMCP Steels

By S. Aihara and K. Okamoto The four papers contained in this Bulletin were presented at the Conference on "Metallurgy, Welding and Qualification of Microalloyed (HSLA) Steel Weldments," held at Houston, Tex., November 6-8, 1990. The American Welding Society holds the copyrights and is the source of these papers. Publication of this document was sponsored by the Welding Research Council, Inc. The price of WRC Bulletin 373 is $40.00 per copy, plus $5.00 for U.S. and $10.00 for overseas, postage and handling. Orders should be sent with payment to the Welding Research Council, Room 1301, 345 E. 47th St., New York, NY 10017.

276-s I AUGUST 1992