Journal of Light Metals 2 (2002) 95–110

www.elsevier.com/locate/lightmetals

Study of damage initiation and fracture in aluminum tailor

welded blanks made via different welding techniques
H.R. Shakeri a, A. Buste a, M.J. Worswick a,*, J.A. Clarke b,1
, F. Feng b,
M. Jain b, M. Finn b
a
Department of Mechanical Engineering, University of Waterloo, 200 University Avenue West, Waterloo, Ont., Canada N2H 1G5
b
ALCAN Kingston Research and Development Centre, P.O. Box 8400, Kingston, Ont., Canada K7L 5L9

Abstract
The present paper addresses microstructural evaluation of failure in multi-gauge aluminum alloy tailor welded blanks (TWBs),
welded using non-vacuum electron beam and neodymium:yttrium–aluminum–garnet laser techniques. The limiting dome height test
is used to evaluate formability of the TWBs. Three gauge combinations utilizing AA5754 sheets are considered (2 to 1 mm, 1.6 to 1
mm and 2 to 1.6 mm). Different weld orientations have been considered: transverse and longitudinal. In general, TWBs show two
different types of fracture: weld failure and failure of the thinner aluminum sheet. Interaction of several factors determines the type
of failure occurring in a TWB specimen. These factors are weld orientation, morphology and distribution of weld defects as well as
the degree of constraint imposed by the thicker sheet on the thin sheet. The last factor usually depends on the difference in thickness
of the sheets and is represented here in terms of gauge ratio.
Ó 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Aluminum tailor welded blanks; Fracture; Nd:YAG laser; NVEB

1. Introduction num alloy sheet and the less familiar weldability re-
quirements for aluminum [2,3].
Tailor blanking is an established process used in- A combination of aluminumÕs high reflectivity, low
creasingly in automotive body components fabricated molten viscosity, and inherent oxide layer makes alu-
using steel sheet [1]. Tailor welded blanks (TWBs) con- minum a challenging material to be welded using
sist of multiple thickness, and sometimes multiple alloy, available welding systems [4]. Several difficulties, that
sheet metals, which are welded together into a single were reported to be common, are porosity and hot
blank of multiple gauge. In automotive applications, cracking in the fusion zone, reduction of strength and
these blanks are stamped to produce body panels. Nu- loss of alloying elements in the weld metal, and in the
merous advantages of this process arise from reducing case of laser welding, poor coupling during the welding
weight, decreasing part count, streamlining the assembly process due to high reflectivity of the material [3,5–10].
process, and cutting costs significantly by using singular Based on the relatively small number of published re-
TWBs instead of multiple blanks, which would have to ports, laser and non-vacuum electron beam (NVEB)
be individually stamped and assembled. Applying this welding techniques are among the processes that have
process to aluminum offers significant potential fur- shown good potential for high volume production of
ther weight reductions required for future automobiles. aluminum alloy TWBs required for auto body applica-
However, producing aluminum alloy TWBs is more tions [3,4].
challenging because of the lower formability of alumi- Two significant consequences of welding with high
intensity laser are rapid solidification and keyhole type
of penetration. The temperature gradient between the
*
Corresponding author. Tel.: +1-519-888-4567; fax: +1-519-888- laser and the workpiece increases as the energy density
6197. in the beam increases [10]. As a result of steep temper-
E-mail address: [email protected] (M.J. Wors-
wick).
ature gradient and rapid cooling, weld metals with finer
1
Currently with Alcan Global Automotive Products, 37676 structures and improved properties are attainable. Op-
Enterprise Court, Farmington Hills, MI 48331, USA. timization of the laser welding variables is very critical;
1471-5317/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved.
PII: S 1 4 7 1 - 5 3 1 7 ( 0 2 ) 0 0 0 2 8 - 7
96 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110

for instance, instability of the keyhole is found to be the formability, geometrical factors due to the differential
dominant cause of porosity in thin plates of some 5XXX thickness must be considered because the thickness
series aluminum alloys welded by laser [7]. change is a discontinuity in the blank. This thickness
Over the last few years, the development of a new change alters the stress and thus the strain that occurs
generation of laser, i.e. neodymium:yttrium–aluminum– during forming [2]. Based on several published results,
garnet (Nd:YAG) laser has emerged as a process with the formability of TWBs decreases as the strength–
wider variety of application in automotive welding. thickness ratio of the base material increases [18].
Specifically, the most significant strides have been made There are also reports suggesting that the formability
in terms of power output and beam delivery optics of the laser welds in steel TWBs is related to weld width
[11,12]. A significant feature of the Nd:YAG process is and weld hardness [17]. Weld width depends on heat
the possibility of using two beams (so-called dual beam input in the welding process and the hardness of weld/
technology). The double laser beam uses optical fibres to HAZ is mainly controlled by the composition of the
deliver beams from two individual Nd:YAG lasers. The base steel. Recent simulation and experimental study by
power of each beam can be manipulated to control Worswick and coworkers [19] has clearly shown that
the heat input within the joint. In welding of tailored aluminum alloy TWBs can be successfully deep drawn.
blanks, in particular, unbalancing the power to each TWB formability is also affected by the welding method,
beam means that the thick and thin sheets will be treated specifically under bending or stretch forming modes [20].
to a suitable amount of power. Once the weld para- The speed of the welding process has also been consid-
meters are established, production of welds of consistent ered as an influential factor on overall mechanical
quality is possible [11,13]. properties of the weld/TWB. For a given material and
Electron beam welding is also a high-energy-density welding process, there exists a welding speed that results
beam welding technique that has been commercialized in improved ductility in the weld [3]. However, differ-
since the late 1960s. This type of welding has been em- ences in welding conditions have produced only a mar-
ployed globally for producing a broad variety of auto- ginal effect on formability of laser-welded steel TWBs
motive components [14]. The main advantages of this [17].
method, compared to many of the fusion welding It has been suggested that the forming limit curve of
methods currently in use, are higher welding speed, the parent material cannot predict the onset of necking
lower heat inputs and greater depth-to-width aspect of steel TWB, even though the failure actually occurs
ratios [15]. Although the electron gun and electron op- in the parent material. To support this argument, it is
tical portions of any electron beam processing system proposed that in stretching area of the TWB specimen,
require a reasonably good vacuum, the workpiece itself the displacement of the stronger (or thicker) material is
may be located either in vacuum or in atmosphere. In restrained and the principal strains are then confined in
the latter case, however, as a result of interactions be- the weaker (or thinner) material and generate a defor-
tween electrons in the beam with the ambient air mole- mation instability. Consequently, formability of TWBs
cules, beam dispersion occurs. This dispersion effect can no longer be predicted with the conventional FLD
causes the beam to gradually increase in size with dis- [18]. Similar deformation instability has also been en-
tance traveled into the atmosphere. Consequently, in- countered in forming of AA5182-aluminum alloy TWBs
creasingly wider welds will be produced when longer with dissimilar gauges welded by gas tungsten arc
working distances are employed [15,16]. As a conse- welding process [21]. In that study, the thinner sheet
quence of beam-broadening, the NVEB welding process tends to undergo greater amount of deformation than
is not able to produce an impingement spot as small as the thicker sheet, resulting in premature failure, wrin-
that attainable with vacuum electron beam welding. kling, and shifting of the weld line.
Interestingly this same characteristic helps make the According to Gould et al. [22] in heat treatable
process more tolerant of seam alignment and work AA7075-T6 TWBs of similar thickness, welded using
height variations [16]. conductive heat resistance seam welding, there was a
Review of the literature suggests that although some considerable reduction in mechanical performance re-
studies have been done on formability of aluminum lated primarily to overaging of the HAZ during the
TWBs, most prior formability studies have considered welding process. In the same study, formability tests of
steel TWBs. In contrast to aluminum TWBs, the weld AA5754 TWBs of dissimilar thickness showed that good
and heat affected zone (HAZ) in steel TWBs are signif- ductility (as much as 30% strain) is achievable prior to
icantly stronger than the base material (in one study the failure.
welds and their HAZs were approximately twice as hard In another study by Naeem and Jessett [23], the re-
as the parent metal) [17]. Therefore, under common duction in performance in TWBs based on solid solution
forming operations, it is reasonable to expect that steel strengthened alloys, such as 5000 series alloys, is caused
TWBs will behave significantly differently from TWBs by a combination of factors including loss of alloying
made of aluminum alloys. In evaluation of TWB elements such as Mg. The addition of Mg using filler
 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110 97

metal, at least in some cases, improved the elongation to strain or near plane strain state depending on the weld
failure in 5000 series TWBs. They also reported a shift orientation. Aluminum spacers were used between the
from failure at the HAZ to weld failure during form- blanks and the tooling to account for the difference in
ability tests on 5000 series TWB specimens welded with thickness within the TWBs. Three gauge combinations
and without the filler material, respectively. Based on utilizing butt-welded 5754 sheets were considered (2 to 1
formability tests on 5182 aluminum alloy TWBs, Kridli mm, 1.6 to 1 mm and 2 to 1.6 mm). Two weld orien-
et al. [21] concluded that formability is significantly af- tations have been considered: transverse and longitudi-
fected by gauge combination and weld orientation. The nal to the main loading direction (long axis of the
formability of steel TWBs is determined by strength/ specimens) (Fig. 1). The rolling direction was perpen-
ductility ratio of the steels being joined and the direction dicular to the weld for all the specimens.
of the stresses imposed in the weld region [17]. Square grids of 2.8 mm were applied to the specimen
Due to the geometrical factors and other internal surface using a silk screening technique (Fig. 2). The
features including cast microstructure and various types local strains were obtained by measuring the size of the
of welding defects, the numerical description of form- grids before and after the tests by means of a video
ability in tailor blanking is very difficult. To reduce the probe and an in-house-developed image analysis system.
discrepancies between the simulated and experimental Specimens of different TWBs were tested to failure or
formability results, it is vital to obtain a sufficiently interrupted prior to failure, selected specimens sectioned
advanced understanding of damage initiation and frac- and examined by optical microscope. A JEOL JSM-840
ture in aluminum TWBs. Therefore, the present paper scanning electron microscope (SEM) combined with a
takes a microstructural approach to study fracture of Linke energy-dispersive X-ray microanalyzer (EDX)
aluminum alloy TWBs welded with two different tech-
niques: NVEB and Nd:YAG laser.

2. Experimental procedure

A typical composition of the parent aluminum alloy

AA5754 sheet used in this study is presented in Table 1.
The sheets were butt-welded using both a NVEB tech-
nique [24,25] and a Nd:YAG laser dual beam technique
[11,13]. NVEB welding was carried out with zero gap
between the sheets, a welding speed of 7.5 m/min, and
a 2.4 mm beam spot size at work. The other welding
parameters for NVEB welding, current and voltage
potential, were 17.8–20.2 MA and 175 kV, respectively.
Fig. 1. Weld orientations considered in the experiments, loading axis is
For the Nd:YAG welds, the beams were set at an angle
vertical.
of 60° to each other and 11° to the joint. Depending on
the gauge ratio, a 2300–2800 W power input allowed a
welding speed in the range of 7–11 m/min. To evaluate
the quality of the welds, specimens of as-welded TWBs
were prepared and examined using optical microscopy.
The limiting dome height (LDH) experiments were
performed using a 101.6 mm hemispherical punch ap-
paratus [26,27]. A clamping load of 500 kN and a punch
speed of 0.2 mm/s were used for the LDH experiments.
No lubricant was used to reduce the friction. Specimens
were either tested to failure or were unloaded at pre-
determined dome heights prior to necking. The geome-
try of the specimens (127  203 mm) provided a plane

Table 1
A typical composition, in wt.%, of aluminum alloy AA5754 parent
sheet material used in this study
Element Mg Mn Fe Si Cu Al
Fig. 2. A typical TWB specimen for LDH experiment with square
Percentage 3.18 0.38 0.17 0.08 0.01 Balance
grids.
98 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110

system was used for analysis of fracture surfaces. An As a result of relatively fast resolidification of the
accelerating voltage of 20 kV was used for all imaging weld metal, in the both NVEB and laser specimens,
and analysis work. there is a major difference in terms of the presence of the
second phase particles. While large, 1–15 lm, second
phase particles are present in all aluminum sheets re-
3. Experimental results gardless of the thickness; under the typical resolving
power of an optical microscope, no such large particles
3.1. Examination of undeformed as-welded TWBs are observable in the weld metal (Fig. 5). To investigate
this observation further, bulk samples were taken from
Both NVEB and Nd:YAG laser specimens of differ- the weld bead as well as the parent aluminum sheets for
ent gauge combinations were examined by optical mi- chemical analysis. Results have indicated that the bulk
croscopy. In both sets of specimens weld defects are chemistry shows only marginal differences between the
present; however, the presence of defects in the NVEB weld metal and the parent material in the both NVEB
specimens was much more pronounced than that in laser and Nd:YAG specimens. Analysis of weld metal sug-
specimens. Fig. 3 illustrates two typical weld defects in gests that there has been Mg losses of 6% and 4.4% (by
different specimens. Considering the morphology of the weight) in NVEB and Nd:YAG specimens, respectively,
defects, it can be concluded that most of the defects in however, the results are only accurate to within a margin
NVEB specimens were result of shrinkage during the of error of about 4%. Therefore, the changes in bulk
solidification of weld metal. The major defects in the chemistry should be considered as insignificant. Alter-
weld metal in laser specimens were gas-related. Fig. 4 natively, two possibilities were considered: alloying ele-
shows another class of defect in both groups, which is ments within those large particles (mainly Mg, Mn and
the roughness of the weld surface or other weld surface Fe) either (i) remained in solid solution with aluminum
defects such as undercuts (which appear as a surface after the welding process, and/or (ii) formed much
defect that eventually may act as a crack). Such defects smaller particles (that are not resolvable under optical
can be critical if the part is exposed to cyclic stresses. microscopy). Should the former be the case, this higher

Fig. 3. Weld defects with different morphology in (a) 2:1 Nd:YAG laser specimen, and (b) 2:1.6 NVEB specimen.

Fig. 4. Superficial defective features such as undercuts and porosity at the welds in (a) 2:1.6 Nd:YAG laser specimen; and (b) 2:1 NVEB specimen,
respectively.
 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110 99

Fig. 5. Absence of the large second phase particles in the weld metals in (a) Nd:YAG laser, and (b) NVEB specimens.

amount of alloying elements in solution would lead to a function of several different phenomena including solid
higher solid solution hardening effect than that in the solution hardening, work-hardening, hardening due to
parent aluminum sheets. Furthermore, EDX micro- the presence of a second phase and also due to the
analysis showed that up to 90% of Mg remains in solid presence of grain boundaries, etc. However, based on
solution and only about 10% is consumed by the second microhardness measurements of Stasik and Wagoner [3]
phase particles within the parent aluminum sheets. and Shakeri et al. [28], there is not an unfavourable and
Similar microanalysis on the weld, remembering that the or wide differential between the parent material and the
distribution of the elements is not uniform due to res- weld metal strength within the TWBs produced by either
olidification, showed that the amount of Mg that re- of the welding methods.
mains in solid solution within the weld region is less than
that in the parent materials for both types of welding 3.2. Comparison of limit strain and strain distribution in
techniques. Therefore, it can be concluded that the latter TWBs
possibility is the case and more refined particles formed
within the weld. These submicron particles may act as Fig. 6 is a plot of normalized limit strain values for
grain growth inhibitors or to improve hardness and TWBs versus loading directions for specimens of dif-
strength. In addition, they play a significant role in ferent gauge combinations as well as different welding
the void initiation stage of the ductile failure pro- techniques. While loaded longitudinally (0°), TWBs re-
cess. However, the resulting mechanical properties are a tain relatively a large portion of their limit strains

Fig. 6. Comparison of the retained limit strain as a function of weld orientation for Nd:YAG laser and NVEB TWBs of different gauge combi-
nations. At 0° (longitudinal loading) in all specimens failure begins in the weld. At 90° (transverse loading) all Nd:YAG specimens showed parent
failure, while NVEB specimens exhibited either weld failure (denoted with ‘‘W’’) or parent failure (denoted with ‘‘P’’).
100 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110

compared to the parent sheet. In addition, in this line perpendicular to the weld. The measurements from
loading orientation regardless of gauge combination or the thinner sheet are on the left in the figure. In the re-
welding technique, failure in all specimens begins in gion adjacent to failure, major strain values of up to
the weld and then propagates into the parent sheets. 20% are measured. Fig. 7b illustrates the measured mi-
TWB specimens welded using the NVEB technique, nor strains for the 2:1 gauge combinations. It is appar-
performed slightly better at 2:1 and 1.6:1 gauge combi- ent that the major strains are much higher in the thinner
nations, while Nd:YAG specimens exhibited a higher sheet and high minor strains in thick sheet to accom-
retained limit strain at the 2:1.6 gauge combination. modate punch curvature.
Under transverse loading condition, all Nd:YAG spec- Fig. 8a shows the measured major strains for the
imens showed parent failure. Unlike the Nd:YAG NVEB and laser 1.6:1 transverse specimens. Fig. 8b
specimens, the NVEB specimens exhibited different presents the measured major strains for the laser and
failure modes. The highest gauge ratio (2:1) specimens NVEB 2:1.6 transverse specimens.
failed in the parent metal, whereas the low gauge ratio Considering the observed failure modes highlighted
(2:1.6) specimens exhibited weld failure. Both parent in Figs. 7 and 8, it can be concluded that while loaded
and weld failures occurred for 1.6:1 NVEB specimens. transversely, Nd:YAG laser-welded TWBs have only
As Fig. 6 illustrates, the NVEB specimens that failed shown failure of the thin parent sheet regardless of the
under weld failure mode exhibited a drastic reduction in gauge combination. In contrast, NVEB-welded speci-
the retained limit strain. The figure also suggests that, mens of different gauge combinations showed different
overall TWBs perform better when loaded longitudi- failure modes. At the highest gauge ratio, 2:1 gauge
nally. combination, NVEB-welded specimens also showed
Fig. 7a compares the major strain distribution in parent sheet failure mode. Specimens with the lowest
transverse loaded 2:1 NVEB and laser TWBs along a gauge ratio, 2:1.6 gauge combination, only showed weld

Fig. 7. Comparison of major and minor strain distributions for transverse loaded, laser and NVEB, 2:1 TWB specimens at the measured punch depth
at failure: (a) 2:1 TWB major strains, (b) 2:1 TWB minor strains.

Fig. 8. Comparison of major strain distributions for transverse welded TWB specimen at the measured punch depth at failure: (a) 1.6:1 TWB major
strains, (b) 2:1.6 TWB major strains.
 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110 101

failure. Both failure modes were observed while testing For instance, Fig. 10 is the same crack in Fig. 9. The
NVEB specimens with 1.6:1 gauge combination. crack initiation and propagation mechanisms can be
observed in Fig. 10a. Voids are evident ahead of the
3.3. Fracture of NVEB welds crack tip and crack advances occur through void co-
alescence within the crack tip. This image shows the role
Examination of both failed/necked specimens clearly of welding defects in the both nucleation and propaga-
shows strong role of weld defects in both initiation and tion of damage and final failure in NVEB welds. Fig.
propagation of failure. These weld defects are strongly 10b clearly shows that the pre-existing defects start to
anisotropic in terms of microstructural morphology. open-up while approached by a growing crack. It should
Therefore, their presence would affect the properties be noted that, in this case since we are dealing with a
differently in different directions. Fig. 9 is a typical case defective weld; there is no real nucleation stage.
of centerline weld failure in a NVEB specimen loaded in As mentioned earlier, the cast structure of the weld
transverse direction. In this specimen, failure begins at metal is clearly distinguishable from that of wrought
the center of the LDH specimen and propagates along parent sheets (Fig. 9). Dendritic growth is one of the
the weld. As the figure shows, the crack propagates characteristics of this cast structure that arises from the
through the equiaxed grains at the center of the weld. non-equilibrium solidification of the weld metal. Fig. 11
Measurement of the strain, along a line perpendicular illustrates dendritic growth in the same specimen as in
to the weld in a 2:1.6 gauge combination, showed that Figs. 9 and 10. This observation also suggests that the
the major strain at a distance about 10 mm away from present weld defects are mostly the result of micro-
the weld in the thin sheet is only 8.5%. By moving to- shrinkage during the solidification of the weld metal.
wards the weld, the major strain decreases and reaches a To further examine the weld defects, the fracture
value of roughly 2% in the vicinity of the weld in the thin surfaces were also examined by SEM. Under low mag-
sheet. Similarly, on the thick sheet side, the maximum nification dimple-like features were observed (Fig. 12a),
value of the major strain is about 5% at a location about however, examination of the surface at higher magnifi-
12.5 mm away from the weld. This value also decreases cation showed that these dimple-like features were ac-
to about 4% in the vicinity of the weld. These results tually patches of dendrites (Fig. 12b). This observation
indicate that the weld failed at a very low strain level. is in agreement with earlier observations and leads to
this conclusion that the low strains to failure are due to
existence of dendritic structures and microshrinkage
in the weld. Very shallow and small dimples as well
as globular particles can be observed in some areas be-
tween the dendritic patches (Fig. 12c and d). Since this
welding technique does not use vacuum shielding, exis-
tence of these globular particles may imply absorption
of oxygen during the welding operation.
In NVEB specimens loaded longitudinally (loading
axis was along the weld line), failure begins inside the
weld and then propagates into both sheets perpendicular
to the weld. Strain measurement suggests that in a 2:1.6
gauge combination, strain localization begins at about
21–22% strain. Fig. 13 shows the necked region of the
weld in one of these specimens, where the strain is esti-
mated above 22%. For this type of failure, there is an
Fig. 9. Centerline weld failure in a NVEB specimen loaded in the opportunity to compare the fracture surface of the weld
transverse direction. with that of the parent aluminum sheet. Fig. 14a shows

Fig. 10. (a) Presence of welding defects along a growing crack and in front of it; and, (b) higher magnification of the region in the vicinity of crack tip
that shows how the existing weld defects are opening up while being approached by a growing crack.
102 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110

Fig. 11. (a) Dendritic structure of the weld and microshrinkage defects along the failed region and ahead of the advancing crack; and, (b) higher
magnification of the crack tip region in (a).

Fig. 12. (a) A general view of fracture surface in a case of centerline weld failure, in a transverse-loaded NVEB specimen, (b) higher magnification of
the dimple-like feature in (a) that shows dendritic structure, (c) tiny shallow dimples and globular particles in some areas; and, (d) higher magni-
fication of the globular particles.

fracture surfaces of both weld metal and the parent closer to the parent sheet much smaller dimples are
sheet. The weld area is easily recognizable due to ab- present (Fig. 14b). By moving towards the central re-
sence of large dimples that are present in the parent gions of the weld, root or upper surface in the weld those
sheet area. In general the weld fracture surface is much dimples become shallower and are replaced by a much
smoother than that of the parent material. More de- smoother surface, characteristics of sheared fracture
tailed examination reveals that in the areas of the weld (Fig. 14c and d). The finer dimple structure is a conse-
 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110 103

3.4. Fracture of Nd:YAG laser welds

In Nd:YAG laser TWBs, weld failure was only ob-

served in longitudinally loaded specimens. Failure of the
thin parent sheet was the usual failure mode in speci-
mens that were loaded in transverse direction. Since this
section of the paper is dedicated to failure in weld,
failure of the parent sheet material will be discussed in
following section.
Failure in longitudinally loaded welds initiates in the
weld and then propagates into the parent sheets normal
to the weld line. Measurement of the major strain on a
Fig. 13. The necked region of a 2:1.6 NVEB specimen loaded longi- 2:1.6 gauge combination around a necked region sug-
tudinally.
gests that strain localization occurs at approximately
22–25% strain. Fig. 15a is from a necked region of the
weld (specimen sectioned parallel to the sheet face). The
quence of alteration of the second phase in the weld. major strain is estimated above 25%. This figure shows a
In the parent metal, large particles act as nucleation significant increase in weld defects in the necked region.
sites for dormant voids, which result in large dimples. A Fig. 15b shows the resulting final failure under similar
larger number of smaller void-nuclei in particle-free conditions. Examination of the fracture surfaces by
weld producing a fine dimple structure. In addition, SEM also yields observations similar to those from the
many welding defects were observed close to the upper NVEB specimens.
surface (Fig. 14c) and also in the central areas (Fig. 14d Fig. 16a shows a general view of the fracture surface
and e). that includes both the parent sheet and the weld regions.

Fig. 14. (a) Fracture surface of a NVEB specimen loaded longitudinally that contains both parent sheet (large dimples) and weld metal regions, (b)
very small dimples in the weld near the sheet metal, (c) welding defect near the upper surface of the weld, (d) welding defects at the central region of
the weld; and, (e) one of the defects in (d) at higher magnification.
104 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110

Fig. 15. (a) The necked region of a longitudinally loaded Nd:YAG weld that shows growth in the weld defects; and, (b) resulting final failure under
similar condition as in (a).

Fig. 16. Fracture surface of a 2:1.6, Nd:YAG laser TWB loaded longitudinally, (a) general view of fracture surface of both the parent sheet and the
weld, (b) dimples in the parent sheet, (c) dimples in the weld near the parent sheet, (d) sheared region in the central region of the weld, (e) shrinkage
defect in the weld; and, (f) higher magnification of the dendritic structure inside the defect in (e).

There is a clear difference in dimple size between the 3.5. Analytical evaluation of the role of defects
parent sheet (Fig. 16b) and those in the weld closer to
the sheet (Fig. 16c). Moving from the transitional region In order to produce a more quantitative evalua-
of the weld towards more central regions, the tiny tion of the role of defects in the weld failure process,
dimples eventually disappear and are replaced by a rel- analysis of the unetched images of weld specimens
atively smooth surface (Fig. 16d). The fracture surface was carried out. Fig. 17 shows the result of image ana-
also exhibits many welding defects of gas- or shrinkage- lysis of the weld in Nd:YAG laser and NVEB speci-
origin (Fig. 16e and f). mens loaded longitudinally. Fig. 17a is the actual optical
 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110 105

Fig. 17. (a) The necked region of Nd:YAG laser weld, (b) digital image of (a) produced by image analysis software, (c) digital image of a an un-
deformed portion of Nd:YAG weld, (d) and (e) actual and digital images of the necked region of NVEB weld, respectively, (f) an undeformed section
of NVEB weld, (g) and (h) chart of defect area of the weld versus major strain for the Nd:YAG laser and NVEB welds, respectively.

micrograph of Nd:YAG laser weld near the necked This analysis is very useful in terms of providing a
region. Fig. 17b shows the digital image produced tangible evaluation on the role of existing defects in both
by image analysis software from Fig. 17a. In Fig. types of welds, however, its shortcomings must not be
17c, a region of undeformed region of the weld is overlooked. First of all the whole analysis is 2D, while
shown. the real events are 3D. Secondly, defects are randomly
Fig. 17d–f are actual images of near the necked re- distributed in the original blanks and even close to onset
gion, digitally processed image of ‘‘d’’, and actual image of strain localization, damage localization occurs in
of an undeformed region of the weld in NVEB speci- certain intervals. In addition, as discussed earlier, there
men, respectively. Fig. 17g is a summary of the image are two different types of welding defects in the speci-
analysis for the Nd:YAG laser specimen that shows mens, i.e., gas-defects and shrinkage defects. Gas defects
clear increase of the weld defect area by major strain. are globular in shape and easily deform without growing
Fig. 17h shows similar result for the NVEB specimen. larger, while shrinkage defects usually contain sharp
Clearly in both cases significant increase in the areas corners and easily expand under stress. The current
covered by weld-defects is observable. It is interesting to image analysis cannot distinguish these types of defects
notice that even at relatively high strain level the per- from each other. At last it should also be noted that in
centage of the weld defect area approaches only one this analysis, strain was measured for each individual
percent of the surface. It may imply that the material grid on the surface, while in reality––specially in the
approaches instability at relatively small percentage of vicinity of the necked region, strain distribution is not
defects. uniform even inside each grid.
106 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110

3.6. Parent sheet failure and formability. During the deformation process, these
particles, which usually are less deformable than the
In fractured specimens that did not exhibit weld surrounding matrix, suppress the deformation in the
failure, parent metal failure occurred through necking of matrix [29]. Eventually as a result of stress build-up,
the thinner sheet away from the weld. Fig. 18a is from a the particles either fracture or debond from the matrix
transverse loaded 2:1.6 Nd:YAG specimen that frac- and produce voids inside the material. Fig. 18c shows
tured through the thinner sheet, about 5 mm away from formation of a void as a result of debonding of a par-
the weld. Measurement of major strain in the vicinity of ticle, ahead of the advancing crack in Fig. 18a. After this
the failed region suggests that the thinner sheet failed at point, failure process would proceed by growth of these
about 22% strain. Fig. 18b shows the Nd:YAG weld voids towards each other.
region adjacent (parallel) to the failed region in Fig. 18a. To obtain a better understanding on the chain of
It is interesting to notice, although, the weld is full of events during the failure of the parent aluminum thin
defects, fracture still took place in the thinner parent sheet, specimens were also sectioned perpendicular to
sheet. Thus a competitive process exists between weld plane of the sheet and direction of fracture (Fig. 19). At
and parent metal failure. In general the extent of dam- a relatively advanced stage of strain localization, a sig-
age in the weld region will increase as the gauge ratio nificant amount of necking has occurred macroscopi-
decreases. cally, corresponding to about 50% reduction in area,
As mentioned before, an influential factor in prop- (Fig. 19a); however, a relatively small percentage of void
erties and failure of metallic material is the presence of damage is observable in the necked region (Fig. 19b and
second phase or inclusions in the material. In most of c). As the figure shows, early void damage is the result of
the cases the second phase particles increase the strength fracture or debonding of larger second phase particles.
of the material and at the same time reduce its ductility Fig. 19d is closer to the final stage of the failure process

Fig. 18. (a) Advancing fracture in the thin sheet of a TWB with transverse laser weld, (b) weld regiion of the same specimen as in (a); and, (c) void
initiation at the second phase particle ahead of the failed region in (a).
 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110 107

Fig. 19. Failure process in the parent aluminum sheet in a transverse loaded Nd:YAG specimen: (a) early stage of strain localization, (b) and (c) void
damage at the second phase particles in the necked region in ‘‘(a)’’, (d) extensive damage in the necked region near final stage of the process, (e) void
damage in (d) outside the shear bands; and, (f) realignment of the voids inside the bands.

(68% reduction in area) and shows extensive damage in previous results on the mutual contribution of void
the necked region. Closer examination shows that most damage and shear band formation in the failure process.
of the void damage is still associated to the second phase Fig. 20a shows a general view of the fracture surface
particles (Fig. 19e). However, the final stage of fracture that contains both dimpled and sheared regions. The
process proceeds via formation of shear bands. A strong fraction of sheared area to areas covered with dimples
evidence for shear band formation is realignment of varies between different locations within the specimen.
void damage and the presence of second phase particles In the dimpled regions, a dual size distribution of dim-
oriented at 45° to the loading axis (Fig. 19f). No evi- ples is observable, Fig. 20b. The larger dimples are evi-
dence of such alignment is observed outside the shear dence of void nucleation at the second phase particles,
band regions (Fig. 19e). whereas the smaller dimples are result of ductile rup-
Observation from fracture surfaces of the specimens ture of inter-void ligaments. The smaller dimples likely
with parent sheet failure by SEM are consistent with the formed within the shear banding final separation. Larger
108 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110

Fig. 20. (a) General view of fracture surface that contains dimpled and sheared region, (b) dual size distribution of dimples; and, (c) large dimples in
the sheared region.

dimples are even present in the sheared areas (Fig. 20c), rection. These defects produce strain gradients within
implying that at least some extent of void nucleation at the weld metal upon loading the TWB. However, in
second phase particles took place prior to failure via terms of morphology and distribution, these defects are
shear banding. anisotropic. Therefore, their effects on the properties
vary in different orientations of the weld relative to the
loading direction. Based on the result of the current
4. Discussion investigation, welding defects dominate when the weld
line and loading direction are parallel. When the loading
The difference in weld bead formability between the axis is perpendicular to weld line, other parameters in-
NVEB and Nd:YAG laser welds are not directly related cluding the gauge ratio, amount of reinforcement at the
to the respective welding energy sources. The two series top and/or bottom of the weld and presence or absence
of welds were made on two different systems and the of superficial defects such as undercuts also play an
laser welds were made after the study on NVEB welds. important role on overall properties.
The primary system differences employed for the NVEB Weld defects are not the only consequences of the
and laser welds were, (1) the NVEB system did not have cast structure of the weld. In addition, it is reasonable to
seam tracking capability which the laser system had, and expect lower ductility in the weld due to its inferior cast
(2) the application of the inert gas shielding was poor microstructure compared to wrought microstructure of
in the case of the NVEB. Beam welding without seam the parent sheet. In terms of changes in chemical com-
tracking may result in variable heat input and, conse- position, although large second phase particles usually
quently, variable weld bead volume along the joint. This present in the parent sheet are not observable in the
variable heat input will cause defects varying from lack weld, there are contradicting reports on this issue. Ac-
of fusion defects to contiguous microshrinkage regions cording to Davies et al. [4], the weld metal typically
at the weld root. With respect to inert gas shielding, shows very little change when compared to the parent
both a NVEB and laser weld bead reach very high sheet metal. However, Moon and Metzbower [10], re-
temperatures and the metal will oxidize to form a thick ported a significant change in chemical composition of
oxide film on the surface. Results of the present study the weld metal compared to that in the parent sheet. In
along with those of Pastor et al. [7] clearly shows that fact, EDX analysis of the weld metal along with the bulk
keyholes are present during the welding process of these analysis and microstructural evidences in this study
TWBs and keyhole stability plays an important role in suggests that, although, marginal reduction of some
defect formation. For instance, the turbulence of the of the elements occurs, more importantly, the alloying
keyhole weld can break off the oxide film on the surface elements are distributed inhomogeneously. This is in-
and drop it into the weld bead as an inclusion. Good deed a structural imperfection that may produce harm-
inert gas shielding prevents this oxidation. The inclu- ful consequences and may also explain the inconsistency
sion and the contiguous microshrinkage will lower the of the one of the observations in this study.
formability of a weld. Also, with the reduction of sheet Microstructural inspection of the undeformed Nd:
ratios it has to be remembered that the weld bead YAG laser TWB specimens suggested that the vast
thickness (i.e., height) decreased and consequently the majority of the welding defects in these specimens are
weld bead imperfections will likely have a greater in- gas-related. However, similar inspection of the de-
fluence since the ratio of their size to the weld thickness formed specimens clearly showed that a significant
will increase. amount of welding defects, other than gas-generated
As pointed out throughout the current paper, the role defects, are present in the specimens. This implies that as
of welding defects in initiation of damage and final a result of inhomogeneous distribution of alloying ele-
failure is very significant depending on the loading di- ments, which in turn is result of non-equilibrium solid-
 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110 109

Fig. 21. W- or V-shape defects in the deformed region of a Nd:YAG laser weld.

ification and dendritic growth, there is certain amount starts at the larger clustered particles and preexisting
of structural imperfection––‘‘invisible defects’’––that are voids are also present after rolling. At a certain point
not detectable in undeformed condition. Upon loading close to the end of failure process, shear bands form in
the specimen, specifically along the weld line, these the necked region. The voids located inside the bands
‘‘invisible defects’’ open up and turn into ‘‘V’’- or ‘‘W’’- first rotate about 45° relative to the loading direction
shape defects (Fig. 21). and then the remaining material fails under shear. Void
As chemical analysis and microstructural examina- sheeting within the shear band is the final mode of shear
tion of undeformed specimens has shown, the large fracturing.
second phase particles present in the parent materials
are substituted with more refined submicron particles in
the weld. This issue is important because of the role of 5. Conclusions
these particles in void nucleation and growth (VNG)
mechanism in ductile fracture of metallic materials. 1. Two types of failure modes, parent metal and weld
Based on VNG mechanism, voids nucleate at particles metal failure, were observed in TWBs with varying
in the material. Consequently, when the void elongates gauge combinations produced by NVEB or Nd:YAG
to the initial intervoid spacing, void coalescence occurs laser welding techniques.
[30]. Recent work of Worswick et al. [31,32] on damage 2. For TWBs in general, the tests have shown that the
percolation model demonstrated how damage initiates location of the fracture or neck is dependent on sev-
and percolates within particle clusters as well as large eral factors including the gauge ratio (tthick =tthin ), weld
particles. Therefore, in absence of large particles, it is orientation, the amount of reinforcement at the upper
reasonable to expect relatively better ductility within the and lower surface of weld and also existence or ab-
weld metal as well as the overall TWB (since the parti- sence of superficial defects.
cle-induced voids would be initiated at a higher amount 3. If the other parameters maintain the same, as gauge
of stress/strain). However, due to its cast microstructure ratio (tthick =tthin ) increases, parent metal formability
and the presence of defects, the ductility of the weld is seen to decrease.
remains lower than the parent metal. 4. For transverse loaded TWBs, both failure modes
As mentioned in the result section, in cases where were observed in specimens with NVEB welds. The
weld failure did not occur, fracture of the thin parent Nd:YAG specimens exhibit better performance in
sheet away from the weld was the failure mode. In these the absence of weld failure.
cases as the gauge ratio increases, the possibility of 5. In TWB specimens loaded longitudinally, regard-
failure in the thin parent sheet increases. It should be less of the welding technique, failure initiates in the
noted that in these conditions, the constraint imposed weld and then propagates into the parent sheets. Al-
on the thinner sheet by both the thicker sheet and the though, the NVEB specimens were produced under
weld is very important factor. The higher the gauge poorly controlled welding conditions, the longitu-
ratio, the larger the constraint and the higher the pos- dinally loaded NVEB specimens exhibited a slightly
sibility of failure in the thinner parent sheet. better behavior than the Nd:YAG laser specimens.
The micromechanism of parent sheet failure is seen to 6. Although the cast structure of the weld makes it infe-
be a combination of void nucleation and damage and rior to the parent sheet material, depending on the
shear band formation. Early damage occurs when the loading direction, the weld/TWB properties are also
second phase particles fracture or debond from the strongly affected by existing welding defects, their
aluminum matrix as a result of stress build up imposed type and distribution. Therefore, it is reasonable to
by the surrounding deforming material. Nucleation expect lower ductility/formability in TWBs compared
110 H.R. Shakeri et al. / Journal of Light Metals 2 (2002) 95–110

to the parent metal. The loss in formability is signif- [16] D.E. Powers, G.R. Laflamme, Welding Journal 71 (6) (1992) 47–
icantly influenced by the quality of the weld especially 52.
[17] W. Waddell, S. Jackson, E.R. Wallach, The influence of the weld
for the low thickness ratios. To prevent the adverse structure on the formability of laser welded tailored blanks, in:
effects on TWB performance, the welding defects Proceedings of IBECÕ98, SAE P-334, SAE paper no. 982396, SAE
should be kept under a yet to be determined accept- Publication, Warrendale, PA, 1998, pp. 257–268.
able level. [18] F. Cayssials, An industrial application of specific forming limit
curves for tailor welded blanks, in sheet metal forming for the
new millennium, in: Proceedings of the IDDRG2000-21st Bien-
nial Congress-Ann Arbor, MI, USA, June, 2000, pp. 17–22.
Acknowledgements [19] Y. Lee, M.J. Worswick, M. Finn, W. Christy, M. Jain, Simulated
and experimental deep drawing of aluminum alloy tailor welded
Financial support provided by Materials and Manu- blanks, in sheet metal forming for the new millennium, in:
facturing Ontario (MMO) and Alcan International Ltd. Proceedings of the IDDRG2000-21st Biennial Congress-Ann
Arbor, MI, USA, June, 2000, pp. 1–7.
is greatly appreciated. The authors thank Mr. William [20] E. Sagstr€om, H. Larsson, B. Nilsson, B. Wadman, The formabil-
Christy of Alcan for his assistance in fabricating the ity of tailor-welded blanks depending on welding methods and
welded samples. material combinations, in sheet metal forming for the new
millennium, in: Proceedings of the IDDRG2000-21st Biennial
Congress-Ann Arbor, MI, USA, June, 2000, pp. 9–16.
[21] G.T. Kridli, P.A. Friedman, A.M. Sherman, Formability of
References aluminum tailor-welded blanks, SAE technical paper no. 2000-01-
0772, SAE Publication, Warrendale, PA, 2000.
[1] M.F. Shi, K.M. Pickett, K.K. Bhatt, Formability issues in the [22] J.E. Gould, D. Workman, L.R. Lehman, Application of conduc-
application of tailor welded blank sheets, in: Proceedings Sheet tive heat resistance seam welding for joining a 7075-T6 alloy and
Metal and Stamping, SAE SP-944, SAE paper 930278, SAE a 5754 dissimilar thickness combination, SAE technical paper no.
Publication, Warrendale, PA, 1993, pp. 27–35. 1999-01-3162, SAE Publication, Warrendale, PA, 1999.
[2] E.R. Pickering, M.A. Glagola, R.M. Ramage, G.A. Taylor, [23] M. Naeem, R. Jessett, Aluminum tailored blank welding with and
Production and performance of high steel gta welded aluminum without wire feed, using high power continuous wave Nd:YAG
tailor welded blanks, in: Aluminum Application for Automotive laser, in: Proceedings of IBECÕ98, SAE P-334, SAE paper no.
Design, SAE SP-1097, SAE Paper 950722, SAE Publication, 982395, SAE Publication, Warrendale, PA, 1998, pp. 247–255.
Warrendale, PA, 1995, pp. 67–73. [24] G. Laflamme, G. Schubert, Welding of tailored blanks in
[3] M.C. Stasik, R.H. Wagoner, Forming of tailor-welded aluminum atmosphere with the electron beam, in: Proceedings of the
blanks, in: J.D. Bryant (Ed.), Aluminum and Magnesium for International Body and Engineering Conference, IBEC Publica-
Automotive Applications, The Minerals, Metals and Materials tion, 1996.
Society (TMS), Warrendale, PA, 1996, pp. 69–83. [25] G. Schubert, Non-vacuum electron beam (NVEB) welding of
[4] R.W. Davies, H.E. Oliver, M.T. Smith, G.J. Grant, JOM 51 (11) aluminum, in: Proceedings of the AlumitechÕ97 Conference, The
(1999) 46–50. Aluminum Association, Washington DC, 1997, pp. 1200–1210.
[5] S. Venkat, C.E. Albright, S. Ramasamy, J.P. Hurley, Welding [26] S.S. Hecker, Metals Engineering Quarterly 14 (4) (1974) 30–36.
Journal 76 (7) (1997) 275s–282s. [27] A.K. Ghosh, Metals Engineering Quarterly 15 (3) (1975) 53–64.
[6] M.J. Cieslak, P.W. Fuerschbach, Metallurgical Transactions B–– [28] H.R. Shakeri, Y. Lee, M.J. Worswick, F. Feng, W. Christy, J.A.
Process Metallurgy 19B (4) (1988) 319–329. Clarke, Weld failure in formability testing of aluminum tailor
[7] M. Pastor, H. Zhao, R.P. Martukanitz, T. Debroy, Welding welded blanks, SAE paper no. 2001-01-0090, SAE Publication,
Journal 78 (6) (1999) 207s–216s. Warrendale, PA, 2001.
[8] J. Mazumder, Journal of Metals 34 (7) (1982) 16–24. [29] D.L. Davidson, in: R.K. Everett, R.J. Arsenault (Eds.), Metal
[9] P. Martin et al., in: M. Sahoo, C. Fradet (Eds.), Proceedings of Matrix Composites: Mechanisms and Properties, Academic Press,
the International Symposium on Light Metals, CIM, Montreal, New York, 1991, pp. 217–234.
1998, pp. 409–423. [30] L.M. Brown, J.D. Embury, in: Proceedings of the 3rd Interna-
[10] D.W. Moon, E.A. Metzbower, Welding Journal 62 (2) (1983) tional Conference on the Strength of Metals and Alloys,
53s–58s. Cambridge, 1973.
[11] R. Mueller, H. Gu, N. Ferguson, Nd:YAG laser welding for [31] M.J. Worswick, A.K. Pilkey, D. Lloyd, S. Court, Damage
automotive manufacturing applications, SAE technical paper no. characterization and damage percolation modeling in aluminum
1999-01-3148, SAE Publication, Warrendale, PA, 1999. alloy sheet, SAE paper no. 2000-01-0773, SAE Publication,
[12] J.A. Vaccari, American Machinist (1992) 37–48. Warrendale, PA, 2000.
[13] R. Hsu, A. Engler, S. Heinemann, The gap bridging capability in [32] M.J. Worswick, A.K. Pilkey, C.I.A. Thomson, D.J. Lloyd, G.
laser tailored blank welding, in: Proceedings of the ICALEO Burger, Percolation damage predictions based on measured
1998, Section F, pp. 224–231. second phase particle distributions, in: D. Northwood, M.
[14] D. Powers, G. Schubert, Welding Journal 79 (2) (2000) 35–38. Shehata (Eds.), Proceedings of IMS, Ottawa, Ontario, Canada,
[15] D.E. Powers, Welding Journal 76 (11) (1997) 59–62. July 26–29, 1998.