Académique Documents
Professionnel Documents
Culture Documents
ABSTRACT. Fatigue tests have been residual stresses in the weld and to quantitative estimates to be made of
conducted on low carbon steel butt form the penetration defect into an the life spent in initiation. It is often
welds containing joint penetration active fatigue crack. conservatively assumed that the initi-
defects (less than complete joint ation period does not exist and that all
penetration). The growth of fatigue Introduction life is spent in crack propagation;
cracks from simulated penetration however, this philosophy is not justi-
defects was monitored by radi- The fatigue life of flawed materials fied on the basis of the observed data.
ography. These measurements allow- such as weldments containing inter- Previous fatigue studies by the
ed the total fatigue life to be sep- nal defects may be separated into two authors in weldments in the HY-series
arated into periods of crack initiation portions: that fraction of the life in steels7-9 have suggested that a con-
and crack propagation. which flaws sharpen to become ac- siderable portion of the fatigue life,
It was found that the rate of crack tive fatigue cracks (initiation), and that possibly as much as one-half, is spent
growth conformed to the expression: fraction in which the sharpened in crack initiation (or undetected
cracks steadily enlarge until fracture crack growth).
da = 0.27 X 10 13 (AK)5 occurs (propagation). In the current study, the authors
dN The propagation portion of the have investigated the fraction of life
fatigue life has been extensively spent in crack propagation in A36 low
The initiation period was found to oc- studied. Paris1 found that the rate of carbon steel weldments.10 The prin-
cupy approximately one half of the crack growth per cycle was propor- cipal experimental method involved
total fatigue life and consists of the tional to a power of the range in stress the radiographic monitoring of inter-
cycles necessary to shake down the intensity factor of the propagating nal crack growth from a simulated
crack. Recent studies have con- joint penetration defect. The results of
firmed the validity of this concept for a these experiments provided crack
wide range of materials.2 - 6 growth data which can be compared
F. V. LAWRENCE is Assistant Professor, The initiation period has proven to with the theoretically determined fa-
Metallurgical and Civil Engineering and be more difficult to describe or meas- tigue crack growth.
W. H. MUNSE is Professor of Civil Engi- ure. Furthermore, there are no suit-
neering, University of Illinois, Urbane able theories available which allow Weld Fabrication and Fatigue
Testing
The specimens were fabricated
from 3A in. (19mm) thickness ASTM
Table 1 - -Mechanical Properties and Chemical Composition A36 steel plate. The welding elec-
trode was a 1/16 in. (1.6 mm) diam
ASTM-A36 Mechanical Properties base wire equivalent to E70 grade
Yield Point 41,000 psi (283.5 N/mm 2 ) electrodes with a minimum ultimate
Tensile Strength 70,900 psi (488 N/mm 2 ) tensile strength of 70 ksi (482.6
Elongation in 8 in. 27% N/mm2). The mechanical properties
(315 mm) of the base metal and the chemical
Chemical Composition, wt % analysis of the base metal and elec-
Material C Mn P S Si
trode are listed in Table 1.
A36 base metal 0.23 0.9 0.015 0.03 The plate segments were joined
E70 1/16in. with a partial-penetration double-U
(1.6mm) wire 0.09 1.0 0.017 0.024 0.45 butt weld using gas metal-arc (GMA)
welding with A-5% 0 2 shielding gas.
W E L D I N G RESEARCH S U P P L E M E N T ! 221-s
After the specimens were machined
to shape and their weld reinforce-
ment removed, they were mounted in
a 50 kip (222.4 kN) Illinois lever-type
fatigue testing machine and tested to
failure under axial loadings at a rate of
300 cpm in ambient conditions.
Eleven tests on such penetration
defects were performed.
Fig. 1Partial penetration butt weld show- Three tests were periodically inter-
ing unwelded depth (2a) representing a rputed to allow radiographs to be
penetration defect in a fatigue loaded joint taken of the penetration void and thus
determine the amount of crack
growth from the defect. The radio-
graphs were taken parallel to the axis
of the weld so that the penetration
defect could be imaged in profile. The
width of the specimens was, there-
fore, necessarily limited to 1.5 in. to
permit the x-rays to penetrate the full
length of the weld. Crack advances of
0.01 in. (Aa = 0.005 in.) could be de-
tected and measured.
Sound Welds (10) ' ' ' ' '
indicates Specimen Previously Tested At Upon completion of the tests, the
A Lower Stress Level Without Failure
"1IOOO initial flaw dimensions were verified
I00- directly on the fracture surfaces.
1
Table 2 Results of Fatigue Tests on Welded Specimens Containing 3/16-in. Joint Penetration Defect "
Crack
propagation
Life in life in Defect Specimen
Stress thousands thousands size, thickness,
Specimen rauge; ( a ) Stress of cycles, of cycles, in. (mm) in. (mm)
number ksi (N/mm 2 ) ratio (N T ) (Npm) (2a n ) (2b)
222-s | M A Y 1 9 7 3
0.30
0.25
LP 1 - 2 - 1 - 7
AK = A<r/wa~ sec
where: (C
- 6
AK = range in stress intensity fac-
tor
E
Aa = stress range to which the
0.20 member is subjected
5
1 a = half crack length
2b = specimen thickness
4
0.15 With AK values computed using
Eq. 1, a plot of AK versus da/dN was
- 3 constructed as shown in Fig. 6. A best
0.10 J fit curve was computed using the
method of least squares, and the two
- 2 standard deviation limits are indi-
cated.
da
= C (AK) (2)
0 20 40 60 80 100 120 140 l&S
dN
Cycles x 10 where:
Fig. 3 Half crack length versus cycles for specimen LP1-2-1. Points are the measured m, C = material constants
data. Curve is calculated
The values of m and C determined
for the data presented in Fig. 6 are:
m = 5.8 and C = 0.269 X 10-13 so
that for the data in Fig, 6:
030 - d - = 0.269 X 10"13 (AK)5'8* (3)
dN
Calculated Crack
Propagation Lives
From Eq. 2, the following can be
obtained.
fa' da ru<
dN (4)
1 clSo--/
Substituting AK from Eq. 1 and
writing Eq. 4 in terms of finite differ-
ences:
where:
a f = half crack length at failure
a o = initial half crack length
N pc = c r a c k p r o p a g a t i o n l i f e
Cycles x IO (calculated)
Fig. 4 Half crack length versus cycles for specimen LP1-2-4. Points are the mea- Aa = finite advance of crack
sured data. Curve is calculated
Using values of C and m from Fig.
6, the initial size of the penetration
void (a0) measured on the fracture
surface and the relationship of Eq. 5,
the number of cycles spent in crack
in Figs. 3, 4, and 5 respectively. The became very well defined due to propagation was computed for spec-
point at which crack propagation rapidly increasing crack opening dis- imens LP1-2-1, LP1-2-4, and LP1-2-5.
began was determined from periodic placements as evidenced by ever Table 3 gives a comparison of
radiographs taken during the tests. sharper x-ray images (see Fig. 4). measured and calculated crack prop-
During the early stages of a test, the Smooth curves (not shown) were agation lives, and the calculated crack
penetration void remained invisible, drawn through the data, and the slope propagation curves versus cycles are
even though the specimen was radio- of the curves (da/dN) at particular plotted in Figs. 3, 4, and 5 as solid
graphed while under peak (tensile) values of half crack length (af) were lines. The calculated crack length ver-
loads. Gradually, as the residual measured. The value of the range in sus cycle curves are in reasonable
stresses apparently shook down, the stress intensity factor (AK) was cal- agreement with the measured data,
penetration void became visible and culated for the geometry shown in Fig.
just prior to measurable crack growth 1 from the relation:11 *See Table 4
W E L D I N G RESEARCH S U P P L E M E N T l 223-s
and the calculated cycles of crack 0.30
propagation N pc - differ from the
measured values by n average of LP 1 - 2 - 5 -
about 30 percent.
0.25
There are several possible sources
for this error, which is not large con- -
sidering the uncertainties in the I
measurements. There are uncertain- 1 0.20
ties in defining the time at which
cracks begin to propagate (resolu-
tion of radiographic measurement at
f
4 2
0.005 in. [0.127 mm]). This error 0.15
mm / cycle
5000
io' io" 4 10
IOO
1 !
i
da _l3
5- 8 E
Y ~ = 0 , 2 6 9 x 10 ( A K ) - , E
dN / 2
&
1
da -13 ! 8
IOOO
dN x__^
%yyyyy~"~*
""~~i >
l d
L d j = 0,501 x 10 ( A o5'8
1
IO
IO 10
1 1 _
10
-5
10
500
224-s I MAY 1 9 7 3
ib*o.r m.
Table 4Comparison of Crack Growth I I I ,'
Parameters (For A K in ksiVinch and
da/dN in inch/cycle)
Material m C i
E70 weld metal
(as welded) 5.8 0.27X10-3
-HJ 1 1
i
i j
, 1 i); {
A36<a> 3.3 0.1X10'9 -I -i
r >- _M -4-i-l-
- y L
hf-|-
P r o p a g a t i o n Life
41
i 1
observed gradual development of a o.oi i
crack opening displacement indi- IO
(a) Data f r o m B a r s o m 3