Effects of Porosity On Mechanical Properties of Aluminum Welds

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Effects of Porosity on Mechanical Properties of

Al umi num Welds


i n c r e a s i n g a r e a f r a c t i o n s o f p o r o s i f y in a w e l d c r o s s s e c t i o n
r e d u c e s s t r e n g t h , a n d f i n e p o r o s i t y w h e n p r e s e n t in s u f f i c i e n t
q u a n t i t y t o c o n t r i b u t e a t o t a l a r e a c o m p a r a b l e t o t h a t o f l a r g e
p o r e s e f f e c t s a lo s s in s t r e n g t h a s m u c h a s do l a r g e p o r e s
BY J. F. RUDY A ND E. J. RUP E RT
AB S T R AC T . T his experimental program
considered the effects of porosity in 2014
and 2219 aluminum alloy d-c gas tung-
sten-arc welds, the porosity was de-
veloped by moisture and/ or hydrogen
contamination of helium shielding gas.
Porosity with individual pores ranging
from 0. 005 to 0. 20 in. in diameter was
developed in a continuous pattern in 24
in. long welds. Welds were made in the
flat, horizontal and vertical positions on
x
/
4
and
3
/
4
in. thick material. Pore sizes,
shapes, and total amounts were measured
on fracture surfaces, and X-ray films;
these measurements were related to
mechanical properties in the transverse
and longitudinal weld axis, for bead-on
and bead-shaved conditions.
Increasing the area fractions of po-
rosity in a weld cross section reduces
strength; fine porosity ( < y
M
t h size),
when present in sufficient numbers to
contribute a total area comparable to
large pores, appears to effect the loss in
strength as much as do large pores.
Longitudinal axis tests indicate a less
significant effect of porosity when some
heat-affected metal is permitted in the
test cross section. Data are also given for
the loss in fatigue strength-life with in-
creasing porosity.
Mechanical property data from artifi-
cially defected (metal removal by drilling
and slotting) welds in 2014-T 6 are com-
pared with the porous weld data. T he
simpler defect geometry of drilled welds
allows derivation of simple area loss vs.
strength loss relationships, giving quan-
titative support to porous weld data gen-
eralizations. In the final analysis, strength
loss is shown to be directly proportional
J. F. RUDY, formerly with the Denver
Division of Martin Marietta C orp. . is with
General Electric Co.. Eveiidale, Ohio, and
E. J. R UPER T is Unit Head, Advanced
Manufacturing T echnology. D-^nv^r D'vl-
sion, Martin Marietta Corp. , Denver, Colo.
Based on paper presented at the AWS 48th
Annual Meeting held in Detroit, Mich.,
during April 24 to 28. 1967.
to loss in cross section area. '
C omment is included on methods of
obtaining intentional porosity, since these
techniques contribute to an understanding
of practical causes of porosity as ob-
tained in production welding.
In t r o d u c t i o n
T hi s study addresses the probl em of
defining sati sfactory criteria for ac-
cept ance or rejection of defects in
al umi num welds. T he whole challenge
of defining such cri teri a lies in the
recogni t i on t hat they cannot be con-
sidered i ndependent l y of such factors
as the cost or urgency of the items
bei ng inspected, t he design requi re-
ment of the weld, and the i mpact of
the repai r and scrappage alternati ves.
: ' * * '
T o clarify these consi derat i ons, it is
evi dent t hat t i me-consumi ng and ex-
pensive i nspecti on pr odcedur es are
not war r ant ed for low cost details
that can be t hrown away for a frac-
tion of the cost of a very detailed
inspection. C onversel y, an appreci abl e
effort can be economi cal l y di rected
t oward the mor e accurat e assessment
and i nspect i on of details havi ng hi gh
economi cal value, as are charact eri st i c
of l aunch vehicle pr ogr ams. A l so, it is
not enough to be able to state t hat a
given defect in an expensive fabrica-
tion is undesi rabl e. T he pert i nent
question is, "Is t hat defect as it exists
at t he t i me of inspection mor e or less
desirable t han a repai red wel d?" C om-
plicating the logic a little mor e, "A r e
Fi g. 1Effect of gr avi t y on pore d i s t r i bu t i on. A (t op)hor i zont al wel d macro cross
s e c t i on; B (top r i g h t )h or i zont a l wel d X-ray. C (cent er r i g h t )f l a t (down hand)
wel d macro cross s e c t i on; D (bot t om r i g h t )f l a t wel d X-ray
322-s I J UL Y 1970
WATER B UB B LEK
Fig. 2Schematic of c ont a mi na nt me t e r i ng system Fi g. 3Range of p or os i t y obt a i ne d f or e va l u a t i on, di s pl ay ed
on f r a c t u r e sur f aces
the properties of the defective weld,
as-is, more or less desirable than the
properties of the first repair, further
degraded by the probability that a
second repair will be required, or a
third . . . ?"
T he acceptance-repair working
document must operate in a very
complex technical and economic field.
As items of production become more
complex and expensive, it may be that
some industries will have to move
away from the concept of adopting a
blanket, across-the-board, accept-
reject ruling philosophy. As economics
justify, it may be well worthwhile to
prepare separate documents for welds
loaded in separable fashion. T he seem-
ingly drastic approach of perform-
ing a minor (i. e. , several man-days)
program; to provide laboratory sup-
porting data for a given defect in a
given detail is even justifiable, and has
been in fact pursued on occasion by
the several companies in the manufac-
ture of launch vehicle tankage.
T hus, the acceptance criteria of the
future may not take the same form as
those traditionally found in industry
today. T hey may instead take the
form of guidelines for analyzing the
specific "defect of the day. " In their
present form, the data of this paper
can best be used in that fashion. S pe-
cific wide-ranging acceptance criteria
are not derived and presented. How-
ever, existing criteria and critiques,
and guidelines, whereby more precise
and perhaps less convenient criteria
can be derived, are presented.
Ca u s e s a n d Pr o d u c t i o n o f Po r o s i t y
Porosity in any weld involves two
stages. T he gas must be introduced.
Ta b l e 1 Co n t a m i n a t i o n Me t h o d s Re q u i r e d t o Pr o d u c e Po r o s i t y
Le ve l s Te s t e d a t Va r i o u s We l d Po s i t i o n
M a t e r i a l
a nd l e ve l
2014-T6-
1
/, i n.
Le ve l One
Two
Th r e e
Fou r
> Fou r
2219-T87-V4 i n.
Le ve l One
Two
Th r e e
Fou r
>Fou r
2014-T6-
3
/
4
i n.
Le ve l One
Two
Th r e e
2219-T87-7* i n.
Le ve l One
Two
Th r e e
Fl a t
(80 c f h H.)
+20 F d e w p oi nt
5 c f h h y d r og e n
+20 F, 25 c f h H.
+60 F, 25 c f h H,
30 c f h H
c

+20 F d e w p oi nt
5 c f h h y d r og e n
+60 F, 25 c f h H,,
30 c f h H
c
+60 F, 25 c f h Hi,
30 c f h H
e

+20 F d e w p oi nt
- 2 0 F, 5 c f h Hj
+25 F, 20 c f h H,
+20 F d e w p oi nt
- 2 0 F, 5 c f h H,
+25 F, 20 c f h H,
s
Hor i zont a l
(80 c f h H
e
)
+30 F d e w p oi nt
+20 F d e w p oi nt
25 t o 30 Fd e w p oi nt
30 F d e w p oi nt
30 t o 35 F d e w p oi nt
5 c f h Hs
+3 F d e w p oi nt
2 c f h Hs
+30 F d e w p oi nt

5c f h H
2
+20 Fd e w p oi nt
5 c f h H,
+30 F d e w p oi nt
6 c f h H,


45 F d e w p oi nt , 4
c f h Hs
+20 Fd e w p oi nt . 5
c f h R.
Ve r t i c a l
(50 c f h H For y
4
)
(80 c f h H For
:
' /
4
)
2, 3, & 5 c f h
Hy d r og e n*
4 c f h h y d r og e n
5 c f h h y d r og e n
* Sp or a d i c p or os i t y of
va r y i ng s i ze .
?. c f h h y d r og e n one
3, 4, 5 c f h h y d r og e n,
one

+20 F d e w p oi nt ,
3-5 c f h Hs

+20 F d e w p oi nt
+20 Fd e w p oi nt , 3
c f h Hs ;
5 F d e w p oi nt , 3
c f h H
;
;
5 c f h Hs
and the gas which is present in the
solidifying bead must be prevented
from escaping. If either of these con-
ditions are absent, a pore-free weld
should be produced. C onsidering the
escape possibility, gases find escape
more difficult as the solidification rate
increases, and as the solidification
geometry varies such that the upper-
most surface is not the last to solidify.
In every-day welding of launch vehicle
components, neither of these factors
can be varied at will in order to
reduce the incidence of porosity. Weld
travel speed ( < solidification rate) is
generally held to a moderate level of
approximately 5 to 15 ipm in order to
give the welding operator the greatest
chance to react to weld bead irregu-
larities. Faster welding speeds produce
an excessive amount of defective bead
before satisfactory corrections can be
introduced.
Solidification geometry is controlled
most effectively by selection of weld-
ing position. However, position is se-
lected by tooling concept, and cannot
be easily changed once the commit-
ment is made. Welds produced in the
horizontal position, which have as
their upper-most boundary the fusion
zone of the upper plate, present the
worst geometry for gas escape during
solidification. Flat welding and vertical
welding are both considerable im-
provements for allowing gas to es-
cape; welds so made do contain less
porosity. Horizontal welding also pro-
duces a nonrandom distribution of
porosity, since buoyancy tends to
cause the bubbles to collect at the
upper fusion zone rather than to be
distributed over the whole width of
the weld. T his distribution promotes a
plane of weakness; therefore, a given
volume of porosity can be expected to
exert a greater mechanical property
affect on horizontal welds compared
to flat or vertical. T his is illustrated by
the X-rays and micro cross section of
Fig. I.
C onsiderations of how the gas is
introduced to the weld bead are more
complex than escape mechanisms, and
the total understanding of the problem
is less complete. However, most of the
practical causes of gas introduction,
W E L D I N G RE SE A RCH S U P P L E M E N T | 323-S
I -
I I I ' ' I I I I L_
i :
$ -
s
\ <
1 E-
f 2
O -
U
o
a -
Q
O
U
t i
T;Ii Ii Ir ""I
Ti r i -
- T
ouv aamna
1
OHV aOHVT
-r
o o o o o o o o o o o o o o o
i ^j ^ . j i . > > '
O J C O t - L O l f i ^ C O N
o o o o ^
t ' J t ^J ,- ' I 1' * ' t ' J ^ . u .
o o o o o o o o o
3=
UJ
4
HOVi l OA OHV LNaHHiTD OHV
A i DaWOHd
324-s ] J U L Y 19 7 0
as determined by empirical problem-
cure experiences, appear to depend on
moisture or hydrocarbons as the fun-
damental cause. Other contaminants
such as silicone-containing lubricants
and even lead smears have been
shown to cause porosity. However,
these are not normally present in the
practical production problems en-
countered.
T he most frequent carrier of mois-
ture into the welding puddle appears
to be surface contamination, either of
the metals to be joined or of the filler
metal. C ontamination is usually in the
form of a hydrated oxide or (in the
case of chromate conversion coatings
which are introduced for corrosion
protection) a hydrated chromate. In
recognition of this problem, most
procedures for high quality aluminum
welding involve extensive surface
cleaning steps just prior to welding.
T he more reliable of these cleaning
procedures are those which involve a
non-smearing mechanical removal of
the surfaces to be melted as the last
preparation, and immediately prior to
welding. A n important characteristic
of mechanical removal techniques is
that no sharp corners or burrs can
remain. Such burrs, for reasons which
probably have to do with a high sur-
face-to-volume ratio, have been noted
to cause irregular inclusions, called
dross or oxide, in the weld bead. T he
porosity caused by inadequate re-
moval of hydrated surface contami-
nants is also irregular and drossy, as
opposed to spherical shapes suggesting
gas contamination.
T he objective of this experiment
was to measure the effects of porosity
on mechanical properties of welds. In
order to obtain sufficient repetition
for statistical validity, means were re-
quired for consistent and predictable
introduction of porosity into experi-
mental welds. T he means chosen for
artificial introduction of porosity were
centered on moisture and hydrogen
contaminants. Weld contamination
was first attempted via control of
surface of filler metal. A contamina-
tion gradient was obtained by each of
the following wire treatments applied
to a given as-received wire:
1. Alkaline solution cleaned.
2. Baked in air convection furnace
at 600 F for 4 hr.
3. Electrolytically charged with hy-
drogen.
4. Exposed in boiling water from 4
to 96 hr.
5. Humidity cabinet exposed at
160 Ff o r 4 h r .
6. A nodized with a range of
parameters from 5 to 25 amp/sq in.
current density, for exposure periods
from 1 to 5 min with varying bath
temperatures.
T hose treatments, which would be
expected to introduce moisture con-
tamination to the surface (humidity
exposure and anodizing) did provide
porous welds. Furthermore, the poros-
ity patterns obtained were of irregular
shape, similar to those found in practi-
cal production welding of aluminum.
However, these techniques had the
disadvantage of inadequate predicta-
bility of porosity level.
In order to obtain a more predicta-
ble sampling of a given porosity level,
other more reproducible methods
were sought. T he method finally se-
lected, by virtue of its predictability
and uniform distribution over the
whole length of a weld panel, was to
meter moisture or hydrogen additions
to the inert gas shielding. T his was
accomplished with the apparatus
shown in Fig. 2. T his plumbing ar-
rangement in the inert gas line al-
lowed the metering of a portion of the
helium through a water bubbler, and
back into the main stream. T he vol-
ume mixture of "bubbled" vs. "tank"
helium provided an effective control
of the dew point of the gas. Dew
point was measured directly by sam-
pling the gas at the port shown, imme-
diately prior to entry into the torch.
For some levels of porosity it was
found necessary to meter some tank
hydrogen in place of, or along with,
the moisture carrying helium. T hese
contamination means were successful
in providing a wide range of porosity.
T his range is best illustrated in Fig. 3,
showing a number of porous weld
metal fracture surfaces, arranged with
increasing porosity, from right to left.
Levels of contamination introduced
to obtain porosity for the various al-
loy, thickness, and welding position
combinations are shown in T able 1.
T he first level was generally obtained
by metering moisture to a dew point
of +20 F. Further contamination
was obtained by metering more mois-
ture, or mixing hydrogen in with the
moisture. A lthough not satisfactorily
indicated in T able 1, horizontal welds
generally gave more porosity for a
given level of contamination than did
flat welds. In the opposite sense, it was
difficult or impossible to obtain a
large amount of porosity at any prac-
tical level of contamination for the
vertical position. C ontamination with
this welding position caused excessive
arc instability before introducing ap-
preciable porosity, suggesting that the
gas escape from a vertical welding
puddle is quite efficient.
T his system of producing porosity
by shielding gas contamination was
apparently satisfactory. However, an
important undesirable aspect of the
porosity pattern became apparent
during the analysis of mechanical
property change. On attempting to
classify porosity patterns, it was noted
that shielding gas contamination tend-
ed to produce a very large number of
very small pores, so frequent and so
small that a precise characterization
system was difficult.
A nother complication of shielding
gas contamination is the influence of
gas composition on electrical charac-
teristics of the arc. In general, the
addition of moisture tends to make
the arc resistance increase, calling for
a higher arc voltage in order to
maintain an approximately constant
arc length. It was often necessary to
reduce current in order to compensate
for the more intense arc, holding an
approximately consistent puddle and
bead shape, for comparable contami-
nated and clean welds. Hydrogen con-
taminated arcs were somewhat less
unstable. C ombinations of contamina-
tion introduced complex changes on
the arc which defy simple generaliza-
Fi g. 5Surface appearance of c ont a mi na t e d 2014 hor i zont al p os i t i on wel ds. A (top
l e f t )mi nu s 22 F dew p oi nt (no marked change f r om zero c ont a mi na t i on wel d); B
(top r i g h t )h 22 F dew p oi nt ; C (bot t om l ef t )5 c f h H, mi xed t o 80 c t h He; D
(bot t om r i g h t ) 0" F dew p oi nt and 15 c f h H
2
(large humps cover large pores)
W E L D I N G RE SE A RCH S U P P L E M E N T I 325-s
Ta b l e 2 Ty p i c a l Po r e Co u n t s Fr o m Fr a c t u r e Su r f a c e
Nu mbe r of pores c ount ed on f r a c t u r e s u r f a c e .
P a ne l
s p e c i m e n
332- 13
332-2
332-20
v
i n.
d i a .
23
85
7
Va
i n.
d i a .
4
8
1
VM
i n.
d i a .
2
0
4
Vi e
i n.
d i a .
1
0
0
Va
i n.
d i a .
3
2
0
'A
i n.
d i a .
0
0
2
Tot al pore a r e a . Ul t i m a t e
Equi v. . % f r a c t u r e area . t e ns i l e
'/s i t h s Based on Gr i d i nt er- s t r e ng t h ,
c ou nt pore c ou nt cept me t h od ks i
181 15. 9 50 30. 1
189 16. 0 65 30. 4
175 15. 4 43 29. 4
ti on. In most cases, combi nat i ons of the
cont ami nant s caused less electrical in-
stability t han either cont ami nant by
itself. S ome compar at i ve ar c vol t age,
arc current , and arc length, (negat i ve
number s for proxi mi t y, i ndi cate bur-
ri ed-arc) are shown in Fi g. 4 as con-
t ami nat i on increases from 0 ( t ank
gas) up t hrough +2 5 dew point and
25 cfh of hydrogen. Er r at i c arc, and
the t endency for t he arc to bury itself
deeper unless voltage is i ncreased, is
an appar ent out growt h of increased
cont ami nat i on. For exampl e, the no-
cont ami nat i on arc is burri ed an opti-
mum 0. 020 in. with 12 arc volts,
while the +2 5 F dew poi nt , 25 cfh
H
2
(i nt o t ank He) was errat i c and
burri ed an excessive 0. 080 in. despite
an i ncrease in arc voltage to 16 v.
In addi t i on to arc instability, con-
t ami nat i on also had an influence on
weld bead surface charact eri st i cs.
S everal typical bead surfaces are
shown in Fi g. 5 as a function of
increasing cont ami nat i on. Low levels
of cont ami nat i on produce a rough
"gravel l y" surface, while great er levels
of cont ami nat i on produce gross dis-
tortions in bead cont our. It was neces-
sary to avoid the latter probl em as
much as possible since it would be
difficult, in mechani cal propert y analy-
sis, t o det ermi ne whet her vari ati ons
were a result of bead shape or porosi -
ty itself.
We ld in g Co n d itio n s
Welds were made in two materials,
2219-T 87 and 2014-T 6. Each materi-
al was represented in two thicknesses,
1
/
4
and
3
/
4
in. , and welds were made
in flat, horizontal, and vertical posi-
tions. Filler metals were 4043 for
2014-T 6 and 2319 for 2219-T 87.
Welding was d-c gas tungsten-arc
straight polarity with helium (plus
contaminant) as shielding gas. For
the V4 in. welds, a first pass was made
without filler metal, followed by a
second pass from the same side with
filler metal. T his procedure was
changed for vertical welding, where a
single pass with filler metal was more
satisfactory.
For the
3
/
4
in. thick welds, two
passesboth with filler wirewere
performed, one from each side. Joint
preparation was square butt in all
cases, with careful mechanical re-
moval cleaning procedures to avoid
introduction of contaminant by any
means other than the intentional
shielding gas source. It was desired to
use comparable welding parameters
for all welds of a given joint design.
However, differing shield gas composi-
tions necessitated changes in current,
voltage, and arc length in order to
minimize puddle geometry variation.
T he procedure for obtaining con-
taminated welds was to start with
welding parameters which had been
developed for the no-contamination
joint. As moisture and or hydrogen
was metered in, necessary adjustments
in welding parameters were made.
T rial welds were given a quick evalua-
tion by observation of porosity in a
weld fracture. On development of the
target contamination level for a given
welding condition, final verification
was obtained by X-ray, and the neces-
sary welds for the testing program
were produced. Mechanical properties
were measured primarily in tranverse
and longitudinal tension, with cursory
evaluation of the effects of porosity on
fatigue behavior.
Effe c t of Porosity on
Me c h a n i c a l Pro p e rtie s
De f e c t Cl a s s i f i c a t i o n Sy s te m
Classification of defects was per-
formed both before and after destruc-
tive testing. R ather non-specific and
arbitrary levels of badness were first
assigned by comparing the stepped
X-ray standards. Five levels, 0
through 4 from water clear to quite
bad, were adopted, as target porosity
levels for specimen production pur-
poses. A lso, the Martin Marietta C or-
porations' classification (Ref. EPS
55406) and the A B MA -PD-R -27A
systems were applied before fracture.
T he final nondestructive evaluation
performed was to measure total pore
area in the expected fracture plane,
by means of multiple angle X-ray
shots.
T he more intensive classification of
defects was performed by detailed ob-
servation of the exposed fracture sur-
faces. T his surface presents the most
convenient display of the defects
which were actually operable in frac-
ture propagation. Description of the
size-frequency distribution, the loca-
tion, and such other derived parame-
ters as total area, were obtained. Pre-
Fia. fiFracture sur f aces and X-rays g i vi ng pore c ou nt of
Table 2
S PEC I E '1Z-2Q
326-s J UL Y 1970
liminary analysis of these data indi-
cate that mechanical properties did
not depend primarily on such factors
as size distribution and location, but
appeared to depend on the'summation
of the defect area reduction to the
fracture surface area.
1
A i n . Tr a n s ve r s e Te n s i o n Re s u l t s
T he most promising system for
classifying defects, and one which
gave a reasonably valid basis for plot-
ting defect vs. mechanical properties,
was to count the actual frequency and
size of pores which appeared in the
fracture surface of the failed speci-
men. T his procedure may appear un-
reasonable since it is not, ol course, a
nondestructive test. However, it does
provide, by virtue of more precise
defect description, a basis for obtain-
ing an empirical relationship between
defect and behavior.
T hree examples of each welding
condition-porosity level panel weld
were taken representing the highest,
lowest, and near-average values of the
many tensiles given from the weld.
Each fracture surface was examined
exhaustively and data of the type illus-
trated in T able 2 were obtained. T he
particular samples used for T able 2
came from 2014-T 6,
l
/
4
in. thick,
horizontal position, tested with bead
on and contaminated to porosity
"level 4". T he three fracture surfaces
showed defects as tabulated, with fail-
ure surfaces and X-rays pictures in
Fig. 6. S pecimen number 332-13, for
example, contained 23 pores of
1
/
6 4
in. diameter, 4 pores of V32 in. , 2
pores of
3
/
64
in. , 1 pore of V
10
in. ,
and 3 pores of
3
/
3 2
in. T he term
"equivalent Vinths pores" converts
the cross sectional area of all the
pores listed to the number of
1
/
M
in.
pores which would, by themselves,
contribute the same total area. For
example, a single Vie '" P
o r e c o n
"
tributes the same area as sixteen
1
,
;4
in. pores, and thus the number of
1
/
18
in. pores is multiplied by 16 to obtain
its contribution to the " V M P
o r e a r e a
equivalents. "
Other systems for summing effects
of various pore sizes were tried, but
the one based on cross sectional area
met with greatest success as a tool for
predicting mechanical behavior. One
shortcoming of this system, as will be
evident in analyzing the plotted data
(particularly Figs. 7-11), is that pores
of size less than V
i ; 4
in. were not
considered. T his omission, not impor-
tant in normal welds, apparently
caused a sudden drop in mechanical
properties in the low range of meas-
ured areas.
T he data for the
1
/
i
in. tranverse
welds are plotted in Figs. 7 through
11. It is evident that this classification
system is quite successful in providing
a well defined locus of points which
relate ultimate tensile strength to de-
fect description for each plot. T he
important conclusion to be drawn
from this satisfactory correlation is
that the primary control of weld met-
al properties is the tola! area of poros-
ity in the fracture surface. T his im-
plies that the size of the individual
pores is not as important as the total
area of the pores of all sizes.
T he data of T able 2 were chosen
intentionally to illustrate this point.
T he three welds shown all have ap-
proximately the same total pore area,
175 to 189 equivalent
l
/4ths- T hey
also have approximately the same ulti-
mate tensile strength, i. e. , 29. 4 to 30. 4
ksi. However, the distribution of pore
size among the three samples is quite
different. 332-2 has a large number of
small pores, while 332-20 has most of
its total pore area represented in the
two V
s
in. pores. (T hese two
1
/
8
in.
pores contribute 128 of the total 175
sectional area for a given volume of
gas. T he following brief geometry ex-
ercise shows that, for a given volume
( V) , area (A ) decreases as number of
pores (N) increases:
4
V = 5 T r
3
TV
where r is pore radius. Solving for r,
3V '
equivalent
4
ths area. ) T hus.
despite different pore sizes, similar
total pore cross sectional areas result
in similar mechanical properties.
T his conclusion opposes the idea
that small pores are not damaging to
mechanical properties. T he statement
(tacitly accepted by some existing
weld acceptance criteria) that a given
volume of porosity is more innocous if
broken up into small pores, is also
refuted by this observation. In fact, if
pore cross: sectional area is the impor-
tant criteria, then a pore size is to be
desired which gives the least cross
4T T /V
A = 7r r
2
N
(area A, of great circles through N
spheres).
S ubstituting for r,
xy 2/3
Air N
Assuming V is constant, and grouping
other constants (K),
A = KN
1,S
, increasing as N increases.
T hus, a given volume of contami-
nant gas in a freezing puddle will
introduce larger cross section area if it
is present in the form of small pores.
However, a statistical distribution ar-
gument will also operate.
T his conclusion that strength loss is
a function of fracture plane loss
defines the essential requirements of a
nondestructive test. A testing and clas-
sification system will be more accurate
the more nearly it provides a readout
which reflects pore area fraction of
the (anticipated) fracture plane. T he
preciseness of area measure desired
must depend on the economic burden
or arriving at this measure, relative to
T ES T ED WITH BEAD OK
A R T IFIC IA L CONTAMINATION COUNTING
A LL PORES BY GRID INT ER C EPT
(5) (10)
' 1
- 1 -
100
(15)
ART1FICAL CONTAMINATION.
POR E COUNT
(%OF FRACTURE AREA)
(20) (25) (30) (35) (40) (45) (50)
200 300 4 00 500
T OT A L PORE AREA FOUND IN FR A C T UR E S UR FA C E.
EXPRES S ED IN EQUIVALENT NUMBER OF 1 64" POR ES
A*
BEAD SHAVED
OPEN
FILLED
S PLIT
SHIELD GAS
CONTAMINATION'
HYDROGEN
MOISTURE
MIXTURE
NONE
(A PPLIES T O FIGURES 7 THRU l l )
PORE COUNT
(% OF FR AC T UR E AREA)
(5) (10) (15) (20) (25) (30) (35) (40) (45) (50)
-r-
0 100 200 300 400 500
T OT A L PORE AREA FOUND IN FR AC T UR E S UR FA C E,
EXPRES S ED IN EQUIVALENT NUMBER OF 1/64" PORES
Fi g. 7St rengt h vs. pore area 2014-T6, f l at we l d , VA i n. transverse t e s t
W E L D I N G RE SE A RCH S U P P L E M E N T 327-s
BEAD ON
A HIGHEST STRENGTH FOR GIVEN PA NEL
D NEAR AVERAGE
y LOWEST
(A PPLIES TO FIGURES 7 THRU 13
50
g *
TOTAL PORE AREA FOUND IN FRACTURE S URFACE,
EXPRES S ED IN EQUIVALENT NUMBER OF 1/64" PORES
BEAD SHAVED
I (% OF FRACTURE AREA)
(5) (10) (15) (20) (25) (30) (35) (40) (45) (50)
TOTAL PORE AREA FOUND IN FRACTURE SURFACE
EXPRESSED IN EQUIVALENT NUMBER OF 1, 64" PORES
Fi g. 8St rengt h vs. pore area 2219-T87, f l a t we l d , % i n., t r ans -
verse t es t pore area by c ou nt
20
B EA D ON
(% OF FR A C T UR E AREA) T
(5) (10) (15) (20) (25) (30) (35) (40) (45) (50)
a
E -
a
40
3 0 -
w
T OT A L POR E AREA FOUND IN FR A C T UR E S UR FA C E,
EXPR ES S ED IN EQUIVA LENT NUMBER OF 1/64" POR ES
B EA D SHAVED
(% OF FR A C T UR E A R EA )
(5) (10) (15) (20) (2 5) " ( 30) (35) (45) (50)
0 100 200 300 400 500
T OT A L POR E AREA FOUND IN FR A C T UR E S UR FA C E,
EXPR ES S ED IN EQUIVA LENT NUMB ER OF 1/64" POR ES
Fi g. 9Strength vs. pore area 2014-T6, h or i zont a l we i d , % i n.,
transverse t est , pore area by c ount
2*
a
(% OF FR A C T UR E AREA)
(5) (10) (15) (20) (25)
- ~ 1
400
1
500
(50)
20
T OT A L POR E AREA FOUND IN FR A C T UR E S UR FA C E.
EXPR ES S ED IN EQUIVA LENT NUMBER OF 1/64" POR ES
BEAD SHAVED
2*
a
B EA D ON
! OF FR A C T UR E AREA) ^ ~ T
(5) (10) (15) (20) (25) (30)' (35) (40) (45) (50)
40
30
20
OF FR A C T UR E A R EA )
(5) (10) (15) (20) (25) (30) (35) (40) (45) (50)
T T
100 200 300 400 500
T OT A L POR E A R EA FOUND IN FR A C T UR E S UR FA C E,
EXPR ES S ED IN EQUIVA LENT NUMB ER OF 1/64" POR ES
B EA D SHAVED
(% OF FR A C T UR E AREA)
(5) (10) (15) (20) (25) (30) (35) (40) (45) (50)
0 103 200 300 400 500
T OT A L FOR E A R EA FOUND IN FR A C T UR E S UR FA C E,
EXPR ES S ED IN EQUIVA LENT NUMBER OF 1/64" POR ES
Fi g. 10Strength vs. pore area 2219-T87, hor i zont al we l d , Vi
i n. transverse t e s t , pore area by c ount
0 100 200 300 400 500
T OT A L POR E A R EA FOUND IN FR A C T UR E S UR FA C E,
EXPR ES S ED IN EQUIVA LENT NUMB ER OF 1/64" POR ES
Fi g. 11Strength vs. pore area 2014-T6, ve r t i c a l we l d , VA
transverse t e s t , pore area by c ount
328-s J UL Y 1970
Estimated
StarttMresl Repair
To Give Eqniv,,
. Properties
Fi g. 12Series of X-ray st andar ds f or VA i n. wel ds wi t h numbers i nd i c a t i ng f r a c t u r e pore c ount f or each speci men
the economic value of the part being
inspected. A shortcut attempt to come
up with convenient pore fraction area
measurements is to work backwards
using X-rays of fractures whose pore
area has already been measured, as
standards for rating other X-rays. For
example, an X-ray which looks similar
to one of the series shown in Fig. 12
may be assumed to have approximate-
ly the same pore fraction as that
which was actually counted for the
X-ray of Fig. 12. Ratings obtained in
this fashion were reasonably success-
ful as a basis for predicting mechani-
cal properties.
If economics justify, however, a
painstaking analysis of an individual
defect can be performed to obtain an
accurate measure of expected pore
area which will appear in the expected
fracture plane. For example, the two
X-rays of Fig. 13 are taken of the
same area. T his example is straight-
forward and does not require pains-
taking mapping of defects. T he
straight-on view (Fig. 13, left) illus-
trates that the greater part of the
porosity is contained in a plane which
is normal to the surface of the sheet.
On observing this X-ray it is obvious
where the expected fracture plane will
appear. However, the porosity is so
concentrated along this plane (or
"line" in the X-ray) that an accurate
count cannot be obtained because of
superposition. T he angle X-ray (Fig.
13, right) spreads the defect plane
out; a pore count can be obtained.
T he count from the angle shot divided
by the cross sectional area of the weld
K
g
W
w
u
H I *
H
H
50
40
30
P
20
gives an accurate measure of the total
pore fraction area. T his particular X-
ray was checked with an actual frac-
ture count, and results correlated
quite well.
How do existing aerospace industry
specifications rate as intruments foi
predicting mechanical properties? Fig-
ure 14 illustrates ultimate strength
INDIVIDUA L DA T A POINT S
T A KEN A T R A NDOM FR OM
C ONT A MINA T ED WELDS .
T 1
2 3
POR OS IT Y C LA S S
50 H
W
H
O
Z
Ed
w
w
w
EH
W
E->
4
P
40-
30
20
Fi g. 13Example of pore pl acement
wi t h i n we l d , possi bl e by m u l t i p l e X-rays
1 2 3 4 5
POROSITY CLASS
Fi g. 14Transverse t e ns i l e s t r e ng t h vs. ABMA s c at t er ed porosi t y c l a s s i f i -
c a t i on VA i n., bead on and bead shaved, mi xed t oget her
W E L D I N G RE SE A RCH S U P P L E M E N T | 329-s
ranges found within welds of given
classification levels according to
A B MA -PD-R -27A . T his specification
is rather ambiguous to apply, and has
the shortcoming that it appears to
place insufficient value on small pores;
that is the reason for the lack of
correlation between properties and
classification. T he very low values un-
der Class 1 (the highest quality) were
taken from samples which contained
large amounts of very fine porosity,
as illustrated in Fig. 15. T he same two
alloys are treated in similar fashion
for various classification levels in the
existing Martin Marietta specification,
EPS 55406, in Fig. 16. A gain, the
ultimate tensile strength scatter for a
given classification level is quite
broad. T he poor correlation between
properties and Martin Marietta's spec-
ification is also due to fine porosity.
% i n . Tr a n s ve r s e Te n s i o n Re s u l t s
T he classification system derived in
the foregoing discussion of
1
/
i
in.
welds was repeated for
3
/
4
in. welds.
T he major differences in the two thick-
nesses was that appreciably lower
levels of porosity, in terms of ratio of
pore area to fracture area, were ob-
tained in the
3
/
4
in. welds. T hus, the
range of porosity investigated was 0 to
12% fracture area instead of the 0 to
40% for the V
4
in. welds. However,
the summary derived in fashion simi-
lar to Figs. 7-11 for the V
4
in.
transverse welds,
1
illustrates the same
overall behavior.
Tr a n s ve r s e Te n s i l e , Ge n e r a l Di s c u s s i o n -
Im p r o ve d Are a Me a s u r e m e n t
Both the V
4
in- (Figs. 7-11) and
3
/
4
in. transverse tests indicate a de-
pendence of mechanical properties on
total pore area. T o a first approxima-
tion, we would expect this dependency
to be linearthat is, a given percent-
age area loss should impose a similar
UNCOUNTED TOO SMALL
Fi g. 15Very fi ne "u nc ou nt e d " pores as di spl ayed on f r a c t u r e surf ace and
X-ray of ve r t i c a l p os i t i on wel d wi t h H
:
c ont a m i na t i on, 2219, VA i n. X5
percentage of strength loss. T his as-
sumes no strengthening or stress riser
effects. However, the data show a
strength loss that appreciably exceeds
area loss. T his anomaly requires some
discussion.
T hat the percent strength reduction
should be in direct proportion to the
percent area reduction caused by the
presence of the flaw, has been verified
by artificial flaws in Martin Marietta
C orporation (Denver) report C R -
64-198, "Weld R epair and Inspection
Procedures. "-' T he technique of this
study was to remove metal (introduce
artificial flaws) from thcweld bead by
drilling, as shown in Fig. 17. For most
of the work, the hole diameter was
1
/
i
the tensile specimen width, thus
allowing a simple conversion from
drill penetration percent to percent
area loss; 50% drill penetration equals
25% area loss, etc.
Data from these experiments are
presented in Fig. 18. T he heavy,
straight line plots the A strength pro-
portional to A area expectation. It
can be seen that the actual data fall
close to this plot, but real drilled weld
strengths are slightly less than pre-
dicted. However, the difference be-
tween linear plot and actual data are
much less for Fig. 18 than Figs. 7-11;
further, the difference between actual
and theoretical can be rationalized by
considering the influence of bending,
illustrated in Fig. 19. T his plot com-
pares drilled hole strength vs. linear
area loss strength for four different
conditions. T he first condition illu-
strates the relationship when area has
been removed by the relatively shal-
low drill hole, only 20% of depth.
Here, actual strengths fall 10 to 15%
below predicted strength. T he next
plot pair for 40% drill depth shows a
closer relationship between predicted
and acutal, while the 60% and
through-drill data show superposition
of actual and prediction data. T he
mechanical difference between shal-
low holes and the through holes lies
g

r a
, - t H >
EH
H
<
EH
J
P
50
4 0 - -
3 0 -
20
POR OS IT Y NUMBER
a
g
EH
CO
_ . M ^
O H I O
H
H
W
EH
<

5
5 0 - -
40'
30
20
T r
2 3 4 5
POR OS IT Y NUMBER
Fi g. 16 Transverse t e ns i l e s t r e ng t h vs. Ma r t i n Ma r i e t t a 's st andards, VA i n., bead on and bead shaved, mi xed t oget her
330-s J ULY 1 9 7 0
Fig. 17Typical 2014-T6 d r i l l e d s t a t i c t e ns i l e s p e c i me n, VA i n. fl at bot t om hol es: A (t op l e f t )s p e c i me n before t e s t i ng ; B (top
r i g h t )t y p i c a l f a i l u r e p a t t e r n; C (bot t om l e f t )f a i l u r e cross sect i on on 20% d e p t h ; D (bot t om r i g h t )f a i l u r e cross s ec t i on on
60% dept h
in the elimination of bending stresses
for the through hole.
T hese data, along with similar data
from other plots, are replotted on Fig.
20 for nine different hole diameter to
specimen width ratios. For the plot of
DI(W-D) = 1 , a through hole
(100% hole depth) constitutes a 50%
loss in strength. T his prediction is
borne out; the 0 depth and 100%
depth data are as expected. However,
the real data do not connect these
points with a straight line, but deviate
in the nonconservativeweaker than
predicteddirection. T he difference
between the straight line and the real
data is maximized for the shallow
holes, as would be expected of a
bending effect.
T he plot of Fig. 21 illustrates an-
other aspect of strength loss caused by
porosity defects. T he real data show
50- ,
no effect out to approximately 15%
loss in area. From 15 to 30%, the
strength loss appears to be coincident
with a linear plot which passes
through the zero defect ordinate at
approximately 38 ksi. However, the
as-welded strength of 6061-T 6 is only
32 ksi. T he typical fracture location at
the 32 ksi strength level is, however,
not in the weld metal but removed
from the weld metal, out into the
overaged heat affected zone. T hus, the
data indicate that the intrinsic
strength of the weld metal is 38 ksi,
and that this zone loses strength in
direct proportion to its loss in area,
finally becoming the weaker link when
its strength under-matches the over-
aged zone; this shifting of the weaker
link occurs when the weld metal
strength (as degraded) becomes less
than 32 ksi.
T he conclusion from these hole
drilling data is that, ignoring bending
effects, strength loss is directly propor-
tional to the loss of load carrying
area.
Why do the drilled hole samples
show a strength reduction proportion-
al to the pore area, to a first approx-
imation, while the porosity data of
Figs. 7-11 indicate a much greater
strength loss? T wo answers seem pos-
sible:
1. T he contamination used to ob-
tain porosity causes strength reduclion
by other mechanisms.
2. A rea loss due to porosity was
M/Y&r'r~
>-3
W
H
O
H
m 20
co
1"
LINEAR AREA LOSS PR EDIC T ION
* 1/4" HOLES
3/ 8" HOLES
A 1/2" HOLES
10
1 1 1 1 r -
20 30 40 50 60
% DEPT H OF WELD REMOVED BY DRILLING
1 1 P -
10 20 30
% WELD CROSS S ECTION REMOVED BY DRILLING
Fig. 18Strength of welds with metal removed by drilling
- 1
70
W E L D I N G RE SE A RCH S U P P L E M E N T 331-s
incorrectly estimated by restricting
the fracture surface count to include
only those pores over V^t h in diam-
eter.
An improved method of pore area
estimating, by grid intercept, supports
the latter. T he grid intercept method
is illustrated in Fig. 22, showing X10
photographs of fracture surfaces for
data X and Y of Fig. 7, top. S uperim-
posed on these fracture surfaces is a
rough grid, with 36 interceptions be-
tween lines A to F and 1 to 6 for Fig.
22A , and 42 (A to F, 1 to 7) for Fig.
22B . A method described by C heever
et al
3
states that the number of inter-
cepts which fall over a pore (obvious
example, D-4 in Fig. 22A ), divided by
the number of total possible intercepts,
is the porosity area fraction. Careful
counting gave 20 pore intercepts for
Fig. 22A (point X, Fig. 7) and 18 for
Fig. 22B (point Y) . C onverting to
porosity, point X has
2 0
/
8 6
= 57%,
and point Y has
1
/4
2
= 4 3 %.
S imilar recounts for all data of Fig. 7
gave the plot of strength percent vs.
pore area as shown as "all pores" in
Fig. 7. T his is a very important plot,
since the raw data now show good
correlation with the expected, propor-
tional area-loss relationship.
T he third (middle) plot of Fig. 7 is
for porosity which was obtained under
production welding conditions without
intentional contamination; it was not
characterized by large numbers of
small pores. For example, the frac-
tures of Fig. 23 show porosity that
was obtained in normal production
welding. T he panels were pre-
production checkouts to verify weld-
ing procedures; these did contain
porosity which was found, by later
investigation, to have been a result of
a combination of inadequate cleaning
procedures and hydrated chromate
surface. T hese fractures show the
readily discernible flaws, surrounded
by solid metal which does not contain
a background of additional indiscerni-
ble flaws.
T he "real weld" plot ( + ) of Fig. 7
obeyed the area loss principle better
than the artificial contamination plot.
T his shows that a more accurate meas-
ure of flaw area was obtained by
cA S - /~
LINEAR AREA LOSS PR EDIC T ION
DR ILLED 20% THROUGH
0.5
1. 0
! -
5
2. 0 2. 5 3 0
INCREASING HOLE DIA MET ER
W^TJ
Fi g. 19Plot of s t r e ng t h loss as hole becomes larger, f or each of f our hole d e p t h s . Shal l ow holes show
par t ur e f r om area loss p r e d i c t i on. Be nd i ng ef f ect s woul d be maxi mi zed wi t h shal l ow holes
3. 5
greater de-
332-s J UL Y 1970
100
40 50 60 70
% HOLE DEPTH
Fig. 20Summary pl ot of st r engt h vs. hole d e p t h , i l l u s t r a t i ng be nd i ng
ef f ect s
pore counting of fractures which did
not contain unusually fine porosity.
T his argument concludes that pore
area measurementwhen adequately
accomplishedis the valid predictor
of tensile properties.
T he data of Fig. 20 indicate that
bending stresses might require consid-
eration if the pore is near one surface.
However, this effect could be offset by
allowing bead build-up to remain.
Lo n g i t u d i n a l Te n s i l e Te s t Re s u l t s
S ummary plots for ultimate tensile
strength, with the load parallel to the
direction of welding, are given in Fig.
24 for the V
4
in. thickness. Insuffi-
cient porosity was obtained for the
3
/
4
in. specimen, on a percentage of cross
sectional area basis, to provide a
meaningful plotting range. However,
the "clusters" which were obtained all
show a strength which exceeds 40 ksi,
averaged over the whole width of the
tensile specimen.
A similar statement can be made
about elongation in this specimen,
measured parallel to the direction of
welding in the presence of appreciable
Haws. Measured elongation in two in-
ches exceeds 5% regardless of porosi-
ty up to the highest levels produced,
compared with a no-defect elongation
of 16%.
Fa t i g u e Re s u l t s
R eturning to the drilling removal
data
2
reproduced in Fig. 25, a pro-
portional strength fall off is indicated
for 10,000 cycle life, initiating at 75%
of the zero defect static load for
2014-T 6. T his indicates that stress
raisers associated with non-crack flaws
do not exert an important influence on
fatigue life; load bearing area still
appears to be the controlling cri-
terion.
R eplotting the weld porosity data of
Fig. 26 as the dotted line on Fig. 25
shows a similar slope, but the strength
of the actual porous welds are consid-
erably lower than the strength of the
drilled welds. Once again, the best
explanation appears to lie in the diffi-
culty of measuring real pore area. Solv-
ing for a correction factor of approx-
imately X5 from Fig. 7, pore count
vs. grid intercept moves the weld
porosity plot of Fig. 25 up close to the
75% drilled hole data.
In f l u e n c e of Oth e r De f e c t s
During the course of the program,
non-gaseous porosity defects were en-
countered in limited number. Several
comments can be offered on each.
Dross and Oxide Inclusions. T hese,
as opposed to a gas inclusion, were
e nc ount e r e d in s ever al no- c on-
tamination specimens. However, it
was never found to any great extent.
When found in sizes of approximately
1
i;i to
3
,,_, in. in a
1
4
in. trans-
verse tensile specimen, its influence on
transverse tensile strength was unde-
tectable. In fact, in a given panel
represented by 6 to 16 tensile speci-
mens, the specimen containing such
dross was never the lowest strength of
the group, compared to its "water
LINEAR AREA LOSS PREDICTION BASED ON
C A LC ULA T FD WELD MET AL STRENGTH
20 2*
%WELD AREA REMOVED BY DRILLING
Fig. 21Effect of i ncreased d r i l l i ng of 6061-T6 wel ds, showi ng del ayed i nf l u e nc e of pseudo por osi t y wh i l e f r a c t u r e
c ont i nu e s t o favor non-wel d met al (heat -af f ect ed zone) path
W E L D I N G RE SE A RCH S U P P L E M E N T 333-s
7
6
5
k
3
J
1
- i i ^ - l M
^
1
D i
Fi g. 22Fracture sur f aces of porous we l d s ; g r i d is super i mposed to a l l ow measure of pore area per cent by: (Dore i nt er-
c e p t s )/(t ot a l i nt e r c e p t s ) x 100 = pore area per cent . P (l e f t )p oi nt X (Fi g. 7), 57%; B (r i g h t )-
clear" parallels. Dross, when not ex-
posed to a surface of the specimen,
exhibited satisfactory fatigue life.
However, the dross did act as a fa-
tigue starting point.
Lack of Penetration. S everal speci-
mens showed lack of penetration in the
3
/
4
in. thickness due to difficulties in
obtaining adequate penetration in the
vertical welding position. Several
comments can be made which are of
interest concerning this defect. Most
importantly, it exerts a marked influ-
ence on the ability of the weld to
withstand a transverse load. C oupled
-poi nt Y (Fi g.
C ODE
7), 43%
60
W
H
E
CO
w
w
:|
50.
40-
^
30
B EA D ON B EA D S HA VED
FLA T
HOR IZ .
V:-
(% OF FR A C T UR E A R EA )
(12. 5) (25) (37. 5)
"I
L
-|
(50)
d^tWWXftr _ _ t Q l I ! _ .
^ ^ H j i ' 'spR j
I 1 | "}*:
&L .-
:
H> *; ''
:4
s
' 9 -
- 1i r 1 i r
0 25 50 75 100 125 150
T OT A L POR E A R EA FOUND DM FR A C T UR E S UR FA C E
EXPR ES S ED IN EQUIVA LENT NUMB ER OF 1/ 64" POR ES
W
EH
g
w
K
EH
CO
H
w
H

EH
J
P
6 0 _
A
50
40.
30
(%OF FR A C T UR E A R EA )
(12. 5) (25) (37. 5) (50)
Fi g. 23Nat ur al l y oc c u r r i ng porosi t y in
p r od u c t i on we l d . Note size and d i s t r i bu -
t i on of pores
1 1'-i 1" 1 h r
0 25 50 75 100 125 150
T OT A L POR E A R EA FOUND IN FR A C T UR E S UR FA C E
EXPR ES S ED IN EQUIVA LENT NUMB ER OF 1/64" POR ES
Fig. 24Summat i on of l ong i t u d i na l u l t i m a t e t e r s i l e s t r e ng t h p l ot s vs.
f r a c t u r e pore area f or VA i n. l ong i t u d i na l t e ns i l e t e s t i ng
334-s J UL Y 1970
30
20
STATIC STRENGTH,
AREA LOSS PREDICTION
(R EF: FIGURE 18)
X= l / 4 " DIA. HOLES
0 = 3 / 8 " DIA. HOLES
A = 1/2" DIA. HOLES,
MEAN VALUE FOR NO FAILURE.
MEAN VALUE FOR NO FAILURE.
MEAN VALUE FOR NO FAILURE.
PORE COUNT
MOVES UP TO 75% /
LINE IF ADJUSTED FOR /
ACCURATE PORE COUNT Z FATIGUE %REDUCTION, LINEAR TO NO
' DEFEC T VALUE OF 75% STATIC STRENGTH
r~
10 20 30 40 50
% DEPTH REMOVED BY DRILLING
60 70
I 1 1 1 1
-
10 15 20 25 30
% WELD AREA REMOVED BY DRILLING OR BY POROSITY
Fig. 25Loss of s t a t i c s t r e ng t h (Fi g. 18), d r i l l e d wel d 10,000 c y c l e f a t i g u e s t r e ng t h , and porous wel d 10,000 cycl e
f a t i g u e st r engt h (Fi g. 24), as f u nc t i on of area loss. The t i r e e r e l a t i ons h i p s are l i near wi t h area, bu t porcus wel d
dat a are lower t h a n expect ed
with this is the observation that in no
case was this defect detected radio-
graphically; it was only detected on
cross sectioning of the weld panel for
tensile testing, or on the fracture sur-
face. T his defect also has great influ-
ence on fatigue behavior.
Longitudinal load behavior, such as
found in a girth weld in a pressure
tank, is not influenced to a measurable
extent by lack of penetration (LOP) .
T his is not a surprising observation,
since the LOP lamination is parallel to
the direction of loading and would not
be expected to restrict the load carry-
ing ability. However, this argument is
not recognized in the assessment, in
existent aerospace criterion docu-
ments, of LOP for a longitudinal
stress orientation. One instance of lack
of fusion, a defect similar to LOP,
behaved in similar fashion in longitu-
dinal tensile testing. T hus, it appears
that a worthwhile gain can be realized
if criteria are written to recognize the
direction of major stress application
Discu ssio n
In considering the data which relate
expected mechanical properties to the
porosity in aluminim welds, it is neces-
sary to recognize that the alternative
to accepting the porous weld is either
repair or scrap. T he scrappage alterna-
tive becomes less attractive, if not
statistically close to impossible, as
launch vehicle tankage increases in
size.
Figure 27 is a plot of reduction in
tensile strength with increasing num-
ber of repairs in 2014-T 6 welds. It is
necessary to consider the probability
of multiple repairs because repair
welding is not always successful the
first time. R ecent production experi-
ence indicates that V
4
ol first repairs
are unacceptable and must be re-
paired again;
1
/
2
of these second
repairs are unacceptable, and approx-
imately
2
/
3
of the third repairs are
unacceptable. T hus, there is consider-
10'
, 6-
100, 000
J .
u <
o
10, 000
able risk that the repaired welds will
ultimately be lower in mechanical
properties than the defective weld
which was initially rejected.
By comparing Fig. 27 with Fig. 7 it
is possible to equate properties of
porous and repaired welds. It is inter-
esting to note that a third repair (Fig.
0 2 5 , 0 0 0 PS I S T R ES S LEVEL
D 20, 000 PS I S T R ES S LEVEL
A 15, 000 PS I S T R ES S LEVEL
V 10, 000 PS I S T R ES S LEVEL
NOT E: C IR C LED SYMBOLS R EPR ES ENT
DEFEC T S IN A DDIT ION T O
POR OS IT Y.
S OLID S YMB OLS A R E 2219.
1?% 18%
%AREA LOS S
24% :o %
+-
56% 42%
_1_
. 1 1 i
0 10 20 :30 40 50 50
T OT A L POR E A R EA FOUND IN FR A C T UR E S UR FA C E,
EXPR ES S ED IN EQUIVA LENT NUMB ER OF 1/64" POR ES
Fi g. 26Fatigue l i f e vs. f r a c t u r e pore count f or VA i n. ma t e r i a l . Al l oys and
p os i t i ons are mi xed
W E L D I N G RE SE A RCH S U P P L E M E N T 335-s
IDEAL MEAN CURVE
2. 33 - SIGMA LIMIT
3. 00 - SIGMA LIMIT
4 5 6 7
NUMBER OF R EPA IR S
Fi g. 27Weld s t r e ng t h vs. number or repai rs on 2014-T6 a l u m i nu m , VA i n.
t h i c k st ock
27typical strength of 36 ksi) has
the mechanical properties equivalent
to a weld which has sufficient porosity
to be well outside of standards of
acceptability, that is, to a weld con-
taining 25 to 35 area % porosity.
T hus more damage may be done by
repair than is repaired by it.
A nother aspect of practical applica-
tion of the general rule of percent
strength loss equaling percent area
loss, is consideration of the denomo-
nator in the area loss fraction. In a
uniaxial tensile specimen the denomo-
nator is- simply the area of the tensile
specimen. In a real pressure vessel the
incremental or base line area is not so
straightforwardly assigned. T hus di-
rect use of strength loss predictions
should be verified on real part, or
simulated part destructive tests prior
to use.
Co n c l u s i o n s
T he experience in fabricating the
porous 2014-T 6 and 2219-T 87 alumi-
num specimens and the mechanical
property evaluations of defective
welds, suggest the following conclu-
sions:
1. T he loss in transverse tensile
strength in the weld metal in alumi-
num welds is proportional to the loss
of sound metal area in the plane of
the expected fracture. T his statement
was derived from data with essentially
spherical (no sharp corners) pores.
2. Longitudinal tensile testing
shows less detrimental effect from
porosity than does transverse tensile
data. For many purposes defects can
be largely overlooked when the major
load is parallel to the welding direc-
tion.
3. Fatigue life, transverse tension-
tension loading, also shows a linear
degradation in stress to fail at given
cycle life, this reduction being propor-
tional to area loss in the fracture
plane.
4. Porosity obtained in horizontal
position welding is more damaging to
mechanical properties than other posi-
tions because the pores are not ran-
domly distributed throughout all pos-
sible fracture planes, but are instead
favorably situated in a given plane
which is normal to a transverse load.
This becomes a plane of weakness.
5. Pore location, except as in 4
above, has little effect on strength-
area loss relationship. Excessive con-
centration near one surface of the
bead (most likely the top) can impose
a bending component which amplifies
the stress somewhat. An artificial de-
fect specimen, in which the entire
defect was contained in a surface
layer 20% of thickness in depth, re-
duced the strength by less than 10%
of the linear prediction value.
6. Measuring area loss, as the in-
put to the strength loss prediction, is
difficult to do nondestructively, but
can be accomplished with careful mul-
tiple X-ray procedures provided that
(a) the pore density is not too high,
and (b) the individual pores are not
too small ( < y
6 4
in. ).
7. Size distribution of pores does
not effect the strength loss-area loss
relationship. Small pores, if present in
sufficient numbers to influence the
area loss, will reduce strength ac-
cordingly.
8. Moisture and/or hydrogen me-
tered to the shied gas provide an ef-
fective means of producing a predict-
able amount of porosity. However,
side effects in arc stability, are tem-
perature and jet action, and in bead
surface, are encountered.
Acknowledgements
A cknowledgement is given to Mar-
shall S pace Flight C enter, NA S A , for
sponsorship of this program as part of
an overall effort to improve aluminum
welding, and for permission to publish
these results. Particular appreciation is
expressed to Messrs. Earl Hasemeyer
and R. Hoppes: MS FC Project
Officers whose helpful comments kept
the program properly oriented.
References
1. R upert , K. J. . and R udy. J. F. .
"A nal yt i cal and S t at i st i cal S t udy on the
Effects of Porosi t y Level on Wel d Joi nt
Per f or mance". MS FC C ont r act NA S 8 11335.
Fi nal R epor t . March 1966.
2. Mar t i n Mari et t a C orporat i on (Denver)
R epor t C R -64-198. "Wel d R epai r and In-
specti on Pr ocedur es " . S ept. 1964.
3. C heever, D. L. . Kammer . P. A . . Mon-
roe, R . E. . and Mar t i n. D. C . "Wel di ng
B ase Metal I nvest i gat i on", proceedi ngs
MS FC A l umi num Weld Devel opment C om-
plex Meet i ng. J anuar y 19. 1966.
-Jutli
See page 340-s regarding the presentation of technical papers at the AWS
or& . . . 52nd A nnual Meeting in San Francisco during April 26-30, 1971.
336-s J U L Y 1 9 7 0

You might also like