The Effects of Heat Input and Wel d

Process on Hot Cracking in Stainless Steel

The Sigmajig test was used to quantify the hot-cracking
response of Type 316 stainless steel welded with the
GTA W, EBWand LBWprocesses
BY G. M. GOODWI N
ABSTRACT. The Sigmajig cracking re-
sponse of a single heat of 0.010-in.-thick
(0.25-mm) Type 316 stainless steel was
determined as a function of welding pro-
cess and parameter variations within a
process. It was found that changes in
welding parameters resulted in changes
in cracking response, even when gross
energy input was maintained constant.
Higher travel speeds and elongated wel d
pool shapes increased the cracking
response, as did increased heat input.
Cracking responses for all three welding
processes, gas tungsten arc (CTA), elec-
tron beam (EB), and pulsed solid-state
laser beam (LB), were found to follow a
common relationship wi th net heat input.
Both EB and LB welds showed higher
cracking resistance than GTA welds.
Introduction
The initial application of the Sigmajig
test (Ref. 1) used a standard set of wel d-
ing parameters wi th the CTA process to
evaluate the dependence of cracking
sensitivity on heat-to-heat compositional
KEY WORDS
Sigmajig Test
316 Stainless Steel
Hot Cracking Response
Heat Input
Electron Beam Welding
Pulsed Laser Welding
GTA Welding
Cracking Resistance
Threshold Stress
Argon Shielding
G M. GOODWIN is a member of the Metals
and Ceramics Division, Oak Ridge National
Laboratory, Oak Ridge, Tenn.
Paper presented at the 68th Annual AWS
Meeting, held March 22-27, 1987. in Chicago,
III.
effects in stainless steels. In this work, a
specific, relatively crack-sensitive heat of
Type 316 stainless steel was used to
study the effect of heat input variations
on cracking sensitivity using three wel d-
ing processes: gas tungsten arc, electron
beam, and pulsed solid-state laser.
There are numerous examples in the
literature documenting the effects of heat
input on hot cracking. Typically, an
increase in heat input increases cracking
sensitivity, but often, there are also com-
plicating factors to be considered, includ-
ing changes in solidification mode (Refs.
2-4) or compositional effects (Refs. 5-7).
Solidification growth morphology has
also been shown t o be an important
factor (Refs. 8, 9). Few studies have
attempted to isolate the effect of individ-
ual parameters (Ref. 10), and fewer yet
have addressed process-to-process varia-
tions (Ref. 11). In this study, we have
intentionally selected a single crack-sensi-
tive heat of material known to solidify as
primary austenite for the processes and
parameters investigated. The variations in
response are thus attributable as nearly as
possible to changes in stress at the trailing
edge of the wel d pool , which are deter-
mined by heat input, heat fl ow, thermal
gradients, and weld pool geometry.
Experimental Conditions
A single heat of 0.010-in.-thick (0.25-
mm) Type 316 stainless steel was used
throughout the experiment. Its composi-
tion is given in Table 1. Previous work
(Ref. 1) established a threshold cracking
stress
1
of 18 ksi (124 MPa), as compared
wi th other commercial heats ranging
from 15 to 53 ksi (103 to 365 MPa). It has
a reasonably high P 4- S content (0.042
wt-%), a Cr-Ni ratio of 1.47 (Ref. 12), a
predicted ferrite number (FN) of 0 (Ref.
13), and, as typified in Fig. 1, invariably
solidifies as primary austenite wi th a small
amount of residual eutectic ferrite.
Gas tungsten arc welds were pro-
duced under argon wi th a constant arc
length of 0.035 in. (0.88 mm). Using
DCEN and a 1/16-in. (1.6 mm) diameter
thoriated tungsten electrode wi th a 45-
deg included angle and a 0.010-in. trun-
cation, arc current and travel speed were
systematically varied above and below
the standard conditions of 20 A and 35
ipm (14.8 mm/s) (Ref. 1). Five specimens
were run at each combination of param-
1
Threshold cracking stress, a
m
in, is defined in
Ref. 1 as the applied stress above which
cracking first occurs.
I'.aa
y
:
-~~y^
Fig. 1 Typical
microstructure of gas
tungsten arc weld in
Type 316 stainless
steel. Heat 828013.
Solidification mode is
primary austenite with
a small amount of
eutectic ferrite.
Murakami's etch
88-S | APRIL 1988
Table 1Composition of 0.25-mm-Thick
Type 316 Stainless Steel (Heat 828013)
Element
C
Mn
P
S
Si
Ni
Cr
Mo
Cu
N
Cr / Ni ratio<
a
>
Predicted FN<
b
>
Primary
solidification
(c)
Composition
(wt-%)
0.018
1.70
0.032
0.010
0.34
12.16
17.04
1.98
0.05
0.047
1.47
0
Austenite
(a) Calculated f r om Ref. 12.
(b) Predicted f r om Ref. 13.
(c) Based on metallographic observat i on; see Fig. 1.
Fig. 2 Gas tungsten
arc weld macrostruc-
tures. A Standard
conditions: 20 A, 35
ipm (14.8 mm/s),
<r
min
= 18 ksi (124
MPa). B Low current:
16 A, 35 ipm,
a
min
= 31ksi(214
MPa). C-High travel
speed: 20 A, 49 ipm
(20.7 mm/s), rT
min
= 31
ksi. D Round weld
pool: 12 A, 21 ipm
(8.9 mm/s), a
mm
=
37.5 ksi (259 MPa).
EElongated weld
pool: 28 A, 49 ipm,
o-min = 7 ksi (48 MPa)
WELDING RESEARCH SUPPLEMENT I 89-s
100
80
g 60
<
g 40
20
HEAT 828013
207MPa (30ksi)
TRAVEL SPEED = 14.8 mm/s (35in./min)
AVERAGE OF FIVE TESTS 1 er
18 20 22
I (amps)
28
Fig. 3-
a?
o
<
o
- Cracking response to changes in weld current
TRAVEL SPEED (in./min)
21.0 24.5 28.0 31.5 35.0 38.5 42.0 45.5
80
60
4 0
20
0
8
A 1 1 1 1 1 1
\ HEAT 828013
\
T
207MPa(30ksi), I = 20amps
\

AVERAGE OF FIVE TESTS 1o-

I I I I I
9 10.4 11.9 13.3 14.8 16.3 17.8 19.3
TRAVEL SPEED (mm/ s)
49.0


2C .7
Fig. 4 Cracking response to changes in travel speed
I (amps)/TRAVEL SPEED (in./min)
12/21.0 14/24.5 16/28.0 18/31.5 2Q/35.0 22/38.5 24/42.0 26/45.5 28/49.0
80
g 60
<
40
20
1 1 1 1 i
HEAT 828013
207MPa (30ksi)
AVERAGE OF FIVE TESTS lo-

y

I I I I I
12/8.9 14/10.4 16/11.9 18/133 20/14.8 22/16.3 24/17.8 26/19.3 28/20.7
I ( amps) / TRAVEL SPEED (mm/s)
Fig. 5 Cracking response at constant current-to-travel speed ratio
eters to establish statistical variability.
Electron beam welds were made in a
7.5-kW Hamilton Standard hard vacuum
unit operating at 100 kV. Variations in
travel speed were accompanied by com-
pensating changes in beam current and
focus.
Pulsed solid-state laser beam welds
were made under argon shielding using a
400-W Raytheon Model SS-500, wi th
pulse rates, travel speeds, and power
levels varied such that the weld pattern
ranged from overlapping spots to a con-
tinuous pool.
Results and Discussion
Gas Tungsten Arc Welds
The standard conditions of 20 A and
35 ipm (14.8 mm/s) travel speed gave a
Sigmajig response of about 50% cracking
at 30 ksi (207 MPa). Figure 2 shows the
various GTA wel d macrostructures. As
plotted in Fig. 3, independently decreas-
ing the current caused a reduction in
cracking until at 16 A, which still results in
a full-penetration wel d, cracking ceased.
Increasing current increased cracking,
with 100% cracking (specimen separa-
tion) occurring at 26 A. The high standard
deviation for high cracking percentages
results from the fact that data between
about 60 and 100% cracking do not
occur, as noted in Ref. 1.
Variations in travel speed at constant
current caused comparable variations in
cracking response, as detailed in Fig. 4.
Decreasing travel speed increased crack-
ing, wi th 100% cracking at 24.5 ipm (10.4
mm/s), and increasing travel speed
decreased cracking, with zero cracking at
49 ipm (20.7 mm/s).
Both of the above effects are consis-
tent wi th results observed in actual prac-
tice, and both are explained by noting
that increasing heat input results in a
wider zone of plastic yielding and causes
a larger volume of metal t o expand and
subsequently contract, increasing shrink-
age stresses.
Maintaining essentially constant heat
input by varying both current and travel
speed proportionally resulted in the data
plotted in Fig. 5. At 12 A and 21 ipm (8.9
mm/s) and wi th a nearly circular wel d
pool , no cracking was observed. Crack-
ing increased as both current and travel
speed increased, and 100% cracking
occurred at 28 A and 49 ipm and wi th an
elongated mucronate wel d pool shape.
Although the change in wel d pool
shape itself clearly affects the observed
cracking response through geometric
stress concentration and alteration of
solidification growt h morphology at the
trailing edge of the pool , t wo other
possible contributions should be noted.
First, examination of Fig. 2 shows that the
weld wi dth increases roughly 30% going
90-s I APRIL 1988
from the circular to the elongated wel d
pool shape. This is undoubtedly due t o
the effect of the copper chill bar on
which the specimens are wel ded.
Although the arc heat input is maintained
constant, slower travel speeds allow a
higher fraction of this total heat input to
be extracted through the copper, result-
ing in a lower effective heat input to the
specimen (Refs. 14, 15). As noted above,
lower heat input reduces cracking.
Secondly, as pointed out by Chihoski
(Ref. 16), changes in wel d speed strongly
affect the stress distribution behind the
wel d pool. Wi thout an extensive stress
analysis, which is beyond the scope of
this work, and clearly beyond the capa-
bilities of the author, it should be noted
that the local transverse stress at which
hot cracking initiates behind a wel d pool
must be exceedingly l ow, limited by the
composite strength of liquid plus solid
near its melting point. Thus, when the
normally predicted compressive stress in
that location is offset by the preapplied
test stress, the onset of cracking should
be expected to be sensitive to shifts in
the stress distribution, as detailed by Chi-
hoski.
The data shown in Figs. 3-5 are com-
bined in Fig. 6 to give an overall view of
the effect of current and/or travel speed
on cracking response. The constant heat
input (Fig. 5) data are a traverse from
lower right t o upper left of the figure.
The standard welding conditions of Ref. 1
are at the intersection of the t wo bold
lines in the center of the plot. The choice
of a test stress other than 30 ksi, or a
different heat of material, woul d, of
course, shift the results.
The effect of variations in current and/
or travel speed on cracking response can
also be reported by determination of the
threshold stress value, <T
m
j
n
. Table 2 tabu-
lates these values for the extreme GTA
conditions and compares them wi th val-
ues, to be discussed later, obtained with
20.7JA91
1001
[jn^.n.'.n0s
100
%
CRACKING
%
CRACKING
Fig. 6 Cracking response to changes in current and/or travel speed
the EB and pulsed LB processes. For the
GTA process, the standard conditions of
20 A and 35 ipm yield a threshold stress
of 18 ksi (124 MPa). The elongated wel d
pool formed at 28 A and 49 ipm drops
the threshold stress to 7 ksi (48 MPa),
whereas the lower heat input conditions,
achieved by either high travel speed (49
ipm) or low current (16 A), raise the
threshold stress to 31 ksi (214 MPa). The
highest cracking resistance (37.5 ksi/259
MPa) is obtained with the round wel d
pool (12 A and 21 ipm), at the same
nominal heat input as the standard and
elongated pool.
Electron Beam Welds
All three sets of welding parameters
used wi th the EB process gave higher
threshold stress values than any of the
GTA conditions. Figure 7 shows the EB
wel d macrostructures. Using the same
travel speed, 35 ipm, as the standard
GTA condition, with beam current and
focus adjusted t o give approximately the
same wel d wi dth, the measured thresh-
old stress value was 57.5 ksi (396 MPa).
Increases in travel speed to 70 and 105
ipm (29.6 and 44.5 mm/s), wi th compen-
sating changes in beam current and
focus, reduced the threshold stress val-
ues to 55 and 50 ksi (379 and 345 MPa),
respectively. As noted for the GTA
welds, this effect is attributed to a reduc-
tion in heat input (see Table 2), change in
weld pool shape, and change in stress
distribution at the trailing edge of the
wel d pool .
It is interesting to note that although
travel speed was varied by a factor of 3,
changes in the pool shape were such that
substructure size remained essentially
constant; this observation was confirmed
by both optical metallography and scan-
ning electron microscopy of crack sur-
faces.
Pulsed Laser Beam Welds
As with the EB welds, all sets of weld-
ing parameters used in the pulsed LB
process gave higher threshold stress val-
ues than any of the GTA conditions. The
laser beam wel d macrostructures are
shown in Fig. 8.
Table 2Sigmajig Threshold Stress (o
mi
) Values for Various Welding Conditions on O.OIO-in.-Thick (0.25-mm) Type 316 Stainless Steel, Heat
828013
Process
CTAW
EBW
LBW
Conditions
Elongated pool, 28 A, 49 ipm (20.7 mm/s)
Standard, 20 A, 35 ipm (14.8 mm/s)
Low current, 16 A, 35 ipm (14.8 mm/s)
High travel, 20 A, 49 ipm (20.7 mm/s)
Round pool, 12 A, 21 ipm (8.9 mm/s)
35 ipm (14.8 mm/s), 1.0 mA
70 ipm (29.6 mm/s), 1.5 mA
105 ipm (44.5 mm/s), 2.0 mA
35 ipm (14.8 mm/s), 40 Hz, 64 W, 3 ms
35 ipm (14.8 mm/s), 100 Hz, 110 W, 3 ms
35 ipm (14.8 mm/s), 200 Hz, 107 W, 1 ms
52.5 ipm (22.2 mm/s), 200 Hz, 274 W, 1 ms
70 ipm (29.6 mm/s), 200 Hz, 314 W, 1 ms
Heat Input, J/in. ()/mm)
Cross
343 (13.5)
343 (13.5)
206 (8.1)
245 (9.6)
343 (13.5)
171 (6.7)
129(5.1)
114 (4.5)
110(4.3)
189 (7.4)
183 (7.2)
313(12.3)
269 (10.6)
Net
(a
>
120 (4.7)
120 (4.7)
72 (2.8)
86 (3.4)
120 (4.7)
86 (3.4)
64 (2.5)
57 (2.2)
22 (0.9)
38(1.5)
37 (1.4)
63 (2.5)
54 (2.1)
o-
mJn
, ksi (MPa)
7.0 (48)
18.0 (124)
31.0(214)
31.0(214)
37.5 (259)
57.5 (396)
55.0 (379)
50.0 (345)
80.0 (552)
70.0 (483)
65.0 (448)
60.0 (414)
55.0 (379)
(a) Assumed efficiencies: CTAW, 35% (Ref. 15); EBW, 50% (Ref. 9); LBW, 20% (Ref. 17).
WELDI NG RESEARCH SUPPLEMENT | 91-s
^^- ^^^I Lt ^^^J S^v ^
SH* ~ aa_ . I K - ^- - - ^ i > W - r^-a ' V JCa^^ffrta a ^ \ V . a ; . 4 ^ ,TalfcII*a*a-- . ^ a a ^ - ^ - ._- 4?S _
': ^Ay*$ " . A -<- """ .
Fig. 7 Electron beam weld
macrostructures. A35 ipm (14.8
mm/s), 1.0 mA, cr
min
= 57.5 ksi
(396 MPa); B-70 ipm (29.6
mm/s), 1.5 mA, cr
min
= 55 ksi (379
MPa); C- 105 ipm (44.5 mm/s),
2.0 mA, a
mjn
= 50 ksi (345 MPa).
400/um, 50X
92-s | APRIL 1988
A
7
Fig. 8 - Pulsed solid-state laser
beam weld macrostructures.
A-35 ipm (14.8 mm/s), 40
Hz, 64 W, 3 ms, <r
mn
= 80 ksi
(552 MPa); B-35 ipm, 100 Hz,
110 W, 3 ms, (r
m
in = 70 ksi
(483 MPa); C-35 ipm, 200 Hz,
107 W, 1 ms, <r
min
= 65 ksi
(448 MPa); D-52.5 ipm (22.2
mm/s), 200 Hz, 274 W, 1 ms,
<rmm = 60 ksi (414 MPa); E-70
ipm (29.6 mm/1.5), 200 Hz,
314 W, 1 ms, (7
min
= 55 ksi
(379 MPa)
At a travel speed of 35 ipm (14.8
mm/s), a pulse rate of 40 Hz wi th 3 ms
duration at a 64-W average power
resulted in a series of spots wi th about
75% overlap. Since this particular heat of
stainless steel did not crater crack under
these conditions, the result was that each
spot was essentially an independent
event, and the threshold stress value of
80 ksi (552 MPa) approached the ultimate
tensile strength of the base material (80.2
ksi/553 MPa in the annealed condition).
Increasing the pulse rate to 100 Hz (with 3
ms duration) and 200 Hz (with 1 ms
duration) required approximately twice
the power level (110 W versus 64 W) to
maintain the same wel d wi dth. Threshold
stress dropped to 70 and 65 ksi (483 and
448 MPa), respectively.
Increasing travel speed t o 52.5 and 70
ipm (22.2 mm/s and 29.6 mm/s) at 200
Hz, wi th concurrent increases in power
level, caused further reductions in thresh-
old stress to 60 and 55 ksi (414 and 379
MPa), respectively. The higher speed LB
welds resembled the EB welds, both in
physical appearance and in threshold
stress values.
Threshold Stress versus Heat Input
The threshold cracking stress values
for each of the welding process/parame-
ter combinations investigated versus net
heat input are plotted in Fig. 9. In deter-
mining net heat input, the following effi-
ciencies were assumed: GTA, 35% (Ref.
15); EB, 50% (Ref. 9); and LB, 20% (Ref.
17). It is debatable whether these effi-
ciency values are precisely correct for the
conditions of this experiment, but modest
changes in assumed efficiency do not
grossly affect the trend of the data. Wi th
the above noted exceptions where
changes in travel speed and wel d pool
shape affect threshold stress indepen-
dent of heat input, there is an overall
relationship for all three processes
between threshold stress and net heat
input. Although a straight line can be
fitted to the data with a correlation coef-
WELDI NG RESEARCH SUPPLEMENT 193-s
700
600
500
400
b 300
200
100



"

I
-if
I
I I
(cold worked)
UTS
(annealed)
A
A
A
D
I I
I
A
D
I
I
O
I
I I
O GTA- 35%
D EB - 5 0 %
A LASER- 20%

O
I I


o
o
o
100
75
50
25
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
NET HEAT INPUT (J/mm)
0
5.0
Fig. 9 Threshold cracking stress versus heat input. UTS = ultimate tensile strength
fi ci ent R = 0. 87, i nt ui t i on wo u l d suggest
t hat an exponent i al cur ve begi nni ng at
t he ul t i mat e tensi l e st r engt h at zer o heat
i nput and decayi ng t o zer o stress at ver y
hi gh val ues of heat i nput mi ght be mor e
appr opr i at e.
Concl usi ons
Based on t he above obser vat i ons, we
can concl ude f or this part i cul ar heat of
Type 316 stainless st eel , and i mpl y f or
ot her materi al s, t hat :
1) For each of t he t hr ee wel di ng pr o-
cesses, wel di ng paramet ers have a signif-
i cant ef f ect on cr acki ng response.
2) In general , i ncreasi ng heat i nput
i ncreases cracki ng response.
3) At least f or t he GTA process,
mani pul at i on of wel di ng par amet er s at
essentially const ant heat i nput alters t he
cracki ng response due t o changes in
wel d pool shape and stress di st ri but i on at
t he trai l i ng edge of t he wel d pool .
4) Compar ed wi t h t he GTA process,
bot h t he EB and pul sed LB processes
pr ovi de i mpr oved cracki ng resi stance.
These obser vat i ons are all in qual i t at i ve
agr eement wi t h results obt ai ned in actual
pract i ce. This not onl y conf i r ms t he abi l i ty
of t he Si gmaj i g test t o r espond appr opr i -
at el y t o changes in t he var i ous wel di ng
par amet er s, but f ur t her suggests that t he
quant i t at i ve nat ure of t he test wi l l pr ove
useful in sel ecti ng par amet er s (or pr o-
cesses) t o avoi d cracki ng, whi l e opt i mi z-
i ng t r avel speed, heat i nput or ot her
vari abl es of i nterest.
Acknowledgments
The aut hor woul d like t o t hank several
i ndi vi dual s f or t hei r cont r i but i ons t o this
wor k . ). D. Hudson was responsi bl e f or
t he GTA wel di ng, R. W. Reed pr oduced
t he laser beam wel ds, and B. G. Cross
per f or med t he EB wel di ng. The pr ogr am
was managed by t he Y-12 Plant Devel op-
ment Di vi si on, and R. A. Huber , T. M.
Must al eski , Jr., and M. H. Ri chey of t hat
or gani zat i on deser ve speci al t hanks f or
thei r pat i ence and counsel . The manu-
script was t yped by Kat hy Gar dner and
r evi ewed by J. F. Ki ng and T. M. Must al -
eski . The fi nal r epor t was edi t ed by
Geor ge Battle and pr epar ed by Al ma
McDonal d. This wor k was sponsor ed by
t he U.S. Depar t ment of Energy under
cont r act DE- AC05- 84OR21400 wi t h Mar -
t i n Mar i et t a Energy Systems, Inc.
References
1. Goodwi n, G. M. 1987. Development of
a new hot-cracking test - the Sigmajig. Weld-
ing Journal 66(1):33-s t o 38-s.
2. Cieslak, M. J., Ritter, A. M., and Savage,
W. F. 1982. Solidification cracking and analyti-
cal electron microscopy of austenitic stainless
steel wel d metals. Welding Journal 61(1):1-s t o
8-s.
3. Lippold, J. C. 1985. Centerline cracking in
deep penetration electron beam welds in
Type 304L stainless steel. Welding Journal
64(5):127-s t o 136-s.
4. Kujanpaa, V. P., Suutala, IM., Takalo, T.,
and Moisio, T. 1979. Correlation between
solidification cracking and microstructure in
austenitic and austenitic-ferritic stainless steel
welds. Weld. Res. Int. 9(2):55-76.
5. Cieslak, M. J., and Savage, W. F. 1985.
Hot-cracking studies of Alloy CN-7M. Welding
lournal 64(5): 119-s t o 126-s.
6. Patterson, R. A., and Milewski, J. O.
1985. GTA wel d cracki ng-Al l oy 625 t o 304L.
Welding Journal 64(8):227-s t o 231-s.
7. Golderer, W. , Schwab, R., and Stytzle, R.
1985. Avoidance of hot cracks when welding
a material combination consisting of tool steel
and a magnetically soft material wi th a solid
state laser. Welding and Cutting 37(12):E211-
E213.
8. Davi d, S. A., and Liu, C. T. 1982. High-
power laser and arc welding of thori um-
doped iridium alloys. Welding Journal
61(5):157-s t o 163-s.
9. Komizo, V., Pungshon, C. S., Gooch, T.
G., and Blakeley, P. J. 1982. Effects of process
parameters on centerline solidification in EB
welds. Met. Constr. Br. Weld. J. 18(2):104R-
111R.
10. Koren, A., Roman, M., Weisshaus, I.,
and Kaufman, A. 1982. Improving the wel d-
ability of Ni-base superalloy 713C. Welding
M/ma/61(11):348-s t o 351-s.
11. Grigor 'Yants, A. G., Fishkis, M. M., and
Andriyakhin, V. M. 1980. A comparison of the
capacities of laser, electron-beam and arc-
wel ded joints for resisting hot cracking.
Avtom. Svarka, No. 10, pp. 11-14.
12. Suutala, N. 1982. Effect of manganese
and nitrogen on the solidification mode in
austenitic stainless steel welds. Met. Trans.
13A(12):2121-2130.
13. Delong, W. T., Ostrom, G., and Szuma-
chowski, E. 1956. Measurement and calcula-
tion of ferrite in stainless steel wel d metal.
Welding Journal 35(11):526-s to 533-s.
14. Niles, R. W., and Jackson, C. E. 1975.
Wel d thermal efficiency of the GTAW pro-
cess. Welding Journal 54(1):25-s to 32-s.
15. Key, J. F., Smartt, H. B., Chan, J. W. , and
Mcllwain, M. E. 1982. Process parameter
effects on arc physics and heat f l ow in GTAW.
Proceedings of international Conference on
Welding Technology for Energy Applications.
Gatlinburg, Tenn., May 16-19.
16. Chihoski, R. A. 1979. Expansion and
stress around aluminum wel d puddles. Weld-
ing Journal 58(9):263-s to 276-s.
17. Personal communication, Prof. T.
DebRoy, Pennsylvania State University, State
College, Pa.
94-s | APRIL 1988