Weld World (2013) 57:243–256

DOI 10.1007/s40194-012-0019-x

RESEARCH PAPER

An examination of electric servo-guns for the resistance spot

welding of complex stack-ups
Jerry Gould & Warren Peterson & James Cruz

Received: 7 January 2012 / Accepted: 20 August 2012 / Published online: 18 January 2013
# International Institute of Welding 2013

Abstract In this program, the capabilities of medium-fre- 1 Introduction

quency direct current (MFDC) electric servo resistance
welding guns for the welding of complex stack-ups were The automotive body in white has seen some of the more
investigated. Capabilities of interest from these guns include dramatic changes in the last two to three model generations
the ability to apply forge forces and currents, as well as the than in any other equivalent period in its history [1–3].
ability to sequence the current and forging profiles. The These changes are a result of twofold pressure being applied
complex stack-up under study included a 1-mm outside by the needs to improve fuel economy and body rigidity.
sheet attached to two 2-mm sheets. Work was done using Each of these factors also has multiple sub-influences. The
design of experiment (DOE) techniques. The experiment uncertain supply of crude oil, increased demand for gasoline
included a range of processing variations deliverable from both in developing and industrialized countries, and more
the MFDC servo-gun, as well as two electrode variations stringent government regulations to combat pollution are
(materials and sizes). Heat balance during resistance weld- some of the primary factors pushing automotive original
ing of this stack-up was found to be dominated by the equipment manufacturers (OEMs) to develop ever more
electrode variations. As electrode variations are not consid- efficient vehicles. The government is also having a marked
ered a solution for automotive complex stack-up applica- impact on the drive for body rigidity to accommodate more
tions, a best practice was defined from the DOE that challenging crash testing. The consumer demand for better
included similar-sized class 2 electrodes. This best practice handling and a more firm “feel” when driving also pushes
included short overall weld times, a moderate forge force, a OEM designers toward stiffer body structures.
significant (40 %) increase in current during forging, and a The need for increased fuel efficiency and improved
weld time relative to the forging portion of the weld sched- body rigidity are usually counterproductive to one another.
ule. This yielded nugget penetrations of roughly 50 % into If a body-in-white design had only to consider fuel efficien-
the thin attached sheet. It was noted that, at these penetra- cy, the gauges of the steels could simply be reduced to
tions, indentations on both sides of the joint reached 0.45 lighten the weight. Contrarily, if the needs were only to
mm and that process variations could be used to trade off address crash worthiness/rigidity issues, this could be ac-
penetrations with overall indentations. complished by increases in the steel gauges of the structure.
The balance between these issues was helped with the
Keywords (IIW Thesaurus) Resistance spot welding . introduction of high-strength steels (HSS) and advanced
Heat balance . Electric servo-gun . Design of experiments high-strength steels (AHSS) [2, 3]. A traditional B-pillar
frame member may have been composed of 2.5-mm-thick
270 MPa steel. With the use of HSS material, a designer
could both improve crash-worthiness and reduce weight by
Doc. IIW-2275, recommended for publication by Commission III "Re-
selecting 2.0-mm-thick DP 600 material.
sistance Welding, Solid State Welding and Allied Joining Processes" This improvement, naturally, does not come without pen-
alty. The costs, weldability, and formability of these higher-
J. Gould (*) : W. Peterson : J. Cruz
Edison Welding Institute, Columbus, OH, USA strength materials add new challenges to manufacturing.
e-mail: [email protected] One of these challenges is the welding of complex stack-
244 Weld World (2013) 57:243–256

ups. Complex stack-ups are typically defined as multiple An alternative approach is to look at applications of
sheets (more than two) with varying gauges. The most forge forces to improve attachment of the outside sheet
challenging of these for resistance welding are those in a complex stack-up. Forge forces (force increases at
attempting to attach a thin outer sheet to a stack-up of or near the end of the current pulse) have commonly
thicker materials. For automotive applications, this can in- been used to improve joint quality on both aluminum-
clude a thin sheet, nominally 0.6 to 0.8 mm, attached to a and nickel-based alloys [5]. However, systems capable
metal combination that can be up to 5 to 6 mm thick. of applying forge forces are not commonly applied for
Resistance welding of such stack-ups becomes a heat bal- automotive applications. This has changed recently with
ance problem. This heat balance problem becomes manifest the introduction of electric servo-guns [10–12]. Electric
through reduced nugget penetration in the thinnest outside servo-guns apply welding forces through the use of an
sheet in the stack-up. This problem has been addressed thor- electric motor and a mechanical actuation device (typi-
oughly in available reference literature for resistance welding cally a ball screw). Welding force, then, can be directly
[4, 5]. Standard solutions have included the use of different- controlled by the current supplied to the motor itself.
sized, as well as different-composition, electrodes to improve The systems originally offered advantages of extended
heat balance. These approaches essentially use electrodes with opening displacements and rapid force rises (shorter
different heat-conducting capabilities to promote nugget for- squeeze times). However, the close relationship between
mation in a way to allow attachment of the thinnest sheet. motor current and force allows forge cycles to be used
While such approaches have proven successful for many in automotive applications. The use of such forge forces
decades [4, 5], they are not suitable for current automotive has recently been demonstrated for improvements in
manufacturing. This is reflected in current recommended resistance weld quality for aluminum sheet [13].
starting resistance welding practices [6], suggesting a single- In this program, the ability of the servo-gun to accomplish
sized electrode for stack-ups up to several millimeters. For forge forces, as well as the coordination between forces and
these applications, there are concerns with the cost implica- currents, has been used to improve the weldability of complex
tions of maintaining several different-sized/material electrodes stack-ups. Weldability, in this case, was defined as the ability
for the various stack-ups that can occur on a vehicle. In to attach the thin outside sheet during the operation. To better
addition, the availability of electrodes with varying sizes/ characterize the relationships between weld gun parameters
materials increases the possibility that mistakes can be made, and final joint quality, design of experiment (DOE) techniques
potentially exacerbating the heat balance problems with these were used [14–16]. In this way, quantitative inferences could
candidate stack-ups. be drawn between various gun factors and a range of measures
Achieving heat balance for complex stack-ups using of final weld quality.
similar-sized electrodes has also been evaluated using numer-
ical modeling techniques [7]. This particular work addressed
heat balance when the electrode face size was relatively small 2 Experimental approach
compared to the total complex stack-up thickness. In that
work, it was found that relatively small electrodes could Complex stack-ups for resistance welding in an automo-
effectively draw the developing fusion zone close to the tive context can potentially cover a wide range of joint
electrodes, attaching the thin outside sheet. However, in this configurations. To reduce the vagueness implicit in the
case, excessive melting occurred at the outside sheet/middle term, discussions were held with a range of automotive
sheet contact surface, resulting in aggressive expulsion. This OEMs and tier 1 suppliers (Galiher 2007, Tower Auto-
study recommended not using small electrode sizes to attach motive, private communication; Bohr 2007, General
such thin outside sheets on these stack-ups. Motors Corporation, private communication; Pakalnins
A more pragmatic approach to achieving proper heat bal- 2007, Chrysler LLC, private communication; Wexler
ance in complex stack-ups has been studied by applying the 2007, Ford Motor Company, private communication;
“law of thermal similarities” [8, 9]. The law of thermal simi- Schnipke 2007, Honda of America, private communica-
larities is essentially a one-dimensional thermal analysis of tion). The answers received varied. Some manufacturers
spot weld stack-ups, predicting new processing schedules discussed developing a better understanding of welding
based on geometric ratios (based on thickness). The approach transformation-induced plasticity and twinning-induced
has been applied directly [8] and while incorporating both plasticity materials. Others talked about welding nonfer-
latent heats and dimensional changes (indentation) during rous materials, single-sided spot welding, and welding
welding [9]. While these approaches offer insights into weld- through adhesives. One constant discussed by each
ing practices for different total stack-up thicknesses, however, manufacturer, however, was the so-called thin/thick/thick
they offer only guidelines into attachments of a thin outer weld stack. These types of weld stacks are seen
sheet onto these stack-ups. throughout the body in white wherever two thick frames
Weld World (2013) 57:243–256 245

are welded to an accompanying floor panel or skin side The relationships between these variables are presented
panel. Through these discussions, it was noted that the in Fig. 1. In the sequence shown, the secondary weld force
specifics of the stacks varied from manufacturer to always initiates 30 ms before the secondary welding current.
manufacturer. For this effort, a representative three To examine relationships between both the processing
high-weld stack-up for study has been defined including a variables described previously, as well as different electrode
2.0-mm DP 590 to a 2.0-mm DP 590 to a 0.7-mm 270-MPa configurations, DOE techniques were employed. The spe-
steel, with each sheet having a galvanneal coating. cific experiment selected was an 8-factor, 2-level, 32-run
Work in this program was done on a medium-frequency 2(8–3) fractional factorial design. This DOE is a resolution
direct current (MFDC) integral transformer electric servo C- IV design, in which 8 of the possible 28 2-factor interactions
frame welding gun. This unit is controlled by a proprietary are confounded. The variables in this experiment included:
ARO iBox control unit. This unit is fully integrated and
& Preliminary electrode force
allows easy program access and flexibility for program
& Ratio of the secondary to preliminary electrode force
modification. In addition, the ARO control unit also has
& Total weld time
predetermined outputs for data acquisition. Actual data ac-
& Fraction of the weld time used for primary current flow
quisition in this study was handled in two ways. First, all
& Number of pulses during secondary current flow
welds were monitored with a Miyachi 326B weld monitor.
& Ratio of secondary to primary current level
This was primarily done for a quick examination of root
& The material used for the electrode contacting the thin
mean square weld currents. More detailed monitoring was
sheet in the stack-up
done both through the ARO iBox interface (as noted previ-
& Ratio of the thick- to thin-side electrode diameters
ously) and with external instrumentation. Weld current and
force signals were taken directly from the ARO unit. The The first six of these factors are directly reduced from the
displacement was measured using a linear potentiometer. allowed process variables discussed previously. The last two
All data collection was done with a Yokogawa DL750 relate to variations in electrode configuration. Specifically,
digital storage oscilloscope. Resistance Welder Manufacturers Association (RWMA)
The ARO MFDC electric servo-gun has the ability to class 2 and RWMA class 3 copper electrodes were used
vary both force and current during a specific welding se- and machined to 6.4- and 8-mm face diameters as appropri-
quence. Control variables on this system include: ate. The stage 1 current was set proportionally to stage 1
weld time using the standard joule heating equation for
& Preliminary electrode force resistance spot welding (RSW), shown in Eq. 1:
& Duration of preliminary current flow
H ¼ I 2 Rt ð1Þ
& Preliminary weld current
& Secondary electrode force
& Secondary number of weld pulses where H is the heat generated in the weld, I is the applied
& Secondary weld pulse width welding current, R is the total resistance of the weld, and t is
& Secondary weld current the effective current on time. For stage 2 pulsation

Fig. 1 Sequencing of the welding current and force for the ARO MFDC electric servo-gun used in this study (diagram courtesy of ARO Inc.)
246 Weld World (2013) 57:243–256

schedules, the cool time was 30 ms. Actual levels for these The resulting experimental design associated with the
various factors were based on some preliminary experiments. confounding scheme of Table 1 is provided in Table 2. This
As mentioned previously, the actual experimental design table also provides the factor-level combinations for each
used in this study included eight two-factor interactions con- variable in the design. Of note, the total weld time shown in
founded with other two-factor interactions. Confounding Table 2 does not include the 30-ms cool times associated
refers to the inability to statistically differentiate the two with the secondary current flow pulsation weld schedules.
interactions in any subsequent analysis. To practically use this This experimental matrix was then converted into machine-
experiment, it was necessary to identify six of these potential programmable factors. This includes specific values of elec-
two-factor interactions that could be eliminated from consid- trode materials/sizes, as well as each process parameter of
eration. The selection of these six two-factor interactions was interest. This resulting revised matrix is provided in Table 3.
done using prior experience with RSW processes. This was Seven welds were produced for each run of the experiment.
done by first assessing the potential significance of each These welds were used for tensile testing (three samples),
candidate two-factor interaction based on this experience. chisel testing (three samples, results not included in this pa-
The six ranked lowest in potential significance were then per), and metallographic evaluations (one sample). A 31×
assumed negligible for the purposes of this experiment. 127-mm coupon size was used in the production of all test
The actual confounding arrangement (sets of confounded samples. Tensile testing was done in the tensile–shear config-
interactions) is characteristic for the specific DOE used. The uration. Specimen geometry and test methods matched the
actual confounded interactions then are based on assignment AWS/SAE D8.9M Recommended Practice [17].
of individual experimental input variables to specific col- For each sample tested, electrode indentation measure-
umns in the experimental design. For the experimental de- ments were first made at the top and bottom surfaces (the
sign used here, column assignments were made to assure latter not included in this paper). For tracking purposes, the
that the six lowest-ranked interactions were only confound- weld interface between the thin (top sheet) and thick middle
ed with those of higher assessed significance. This essen- sheet was designated as the 1–2 interface, while that be-
tially allowed the 20 top-ranked interactions to be included tween the two thick sheets (middle and lower sheet) was
in the subsequent analysis confounded (at worst) only with designated as the 2–3 interface. The peak tensile load was
the 6 lowest-ranked ones. The resulting confounding measured at both the 1–2 and 2–3 interfaces (the latter again
scheme (showing the pairs of two-factor interactions con- not included in this paper). Button sizes were measured at
founded in this experimental design) is presented in Table 1. each after all destructive tests.

Table 1 Matrix of confounded

interactions for the DOE used in Dominant interaction Secondary interaction(s)
this study [the determined dom-
inant interaction is listed in the Top electrode material×electrode diameter ratio Weld time×stage 1–2 WT timing, W2
first column, while the con- pulsation×force stage 1
founded secondary interaction(s) Top electrode material×weld time Electrode diameter ratio×stage 1–2 WT timing
are provided in the second
Top electrode material×W2 pulsation Electrode diameter ratio×force stage 1
column]
Top electrode material×force ratio 2–1 None
Electrode diameter ratio×weld time Top electrode material×stage 1–2 WT timing
Electrode diameter ratio×W2 pulsation Top electrode material×force stage 1
Top electrode material×I level 2–1 None
Electrode diameter ratio×force ratio 2–1 None
Electrode diameter ratio×I level 2–1 None
Stage 1–2 WT timing×force stage 1 Weld time×W2 pulsation
Weld time×force ratio 2–1 None
Weld time×force stage 1 W2 pulsation×stage 1–2 WT timing
Weld time×I level 2–1 None
W2 pulsation×force ratio 2–1 None
W2 pulsation×I level 2–1 None
Force ratio 2–1×stage 1–2 WT timing None
Force ratio 2–1×force stage 1 None
Force ratio 2–1×I level 2–1 None
Stage 1–2 WT timing×I level 2–1 None
Force stage 1×1 level 2–1 None
Weld World (2013) 57:243–256 247

Table 2 DOE based experimental plan used in this study

Trial Thin sheet side Thin/thick Total weld No. pulses Secondary/ Primary/ Primary weld Secondary/primary
no. electrode side electrode time (ms) for secondary primary weld total weld force (N) current level
material (class) diameter ratio current flow force ratio time (ms)

1 2 1 280 2 1.5 0.4 2,400 1.2

2 2 1 420 1 2.5 0.65 1,700 1.4
3 2 1 420 2 2.5 0.65 2,400 1.2
4 2 1 280 1 1.5 0.4 1,700 1.4
5 2 1 420 1 1.5 0.65 1,700 1.2
6 2 1 280 1 2.5 0.4 1,700 1.2
7 2 1 420 2 1.5 0.65 2,400 1.4
8 2 1 280 2 2.5 0.4 2,400 1.4
9 2 1.25 420 2 1.5 0.4 1,700 1.2
10 2 1.25 420 1 2.5 0.4 2,400 1.2
11 2 1.25 420 1 1.5 0.4 2,400 1.4
12 2 1.25 280 2 1.5 0.65 1,700 1.4
13 2 1.25 280 1 1.5 0.65 2,400 1.2
14 2 1.25 280 2 2.5 0.65 1,700 1.2
15 2 1.25 420 2 2.5 0.4 1,700 1.4
16 2 1.25 280 1 2.5 0.65 2,400 1.4
17 3 1 420 1 2.5 0.4 2,400 1.4
18 3 1 280 1 2.5 0.65 2,400 1.2
19 3 1 280 2 1.5 0.65 1,700 1.2
20 3 1 280 1 1.5 0.65 2,400 1.4
21 3 1 420 2 1.5 0.4 1,700 1.4
22 3 1 420 2 2.5 0.4 1,700 1.2
23 3 1 420 1 1.5 0.4 2,400 1.2
24 3 1 280 2 2.5 0.65 1,700 1.4
25 3 1.25 280 1 2.5 0.4 1,700 1.4
26 3 1.25 420 2 2.5 0.65 2,400 1.4
27 3 1.25 420 2 1.5 0.65 2,400 1.2
28 3 1.25 280 1 1.5 0.4 1,700 1.2
29 3 1.25 420 1 1.5 0.65 1,700 1.4
30 3 1.25 420 1 2.5 0.65 1,700 1.2
31 3 1.25 280 2 2.5 0.4 2,400 1.2
32 3 1.25 280 2 1.5 0.4 2,400 1.4

Metallographic inspections were done using standard & Button size at the 1–2 interface (taken from the tensile–
techniques. This included sectioning along the weld center- shear and chisel test samples)
line, grinding, polishing, and etching in a solution of satu- & Indentation at the outside thin sheet
rated picric acid in water. This etchant has been shown to & Nugget penetration into the outside thin sheet
successfully reveal the extents of weld nuggets in steel
Analysis of these results was done using standard EWI
resistance spot welds [18]. All welds were examined using
techniques [14–16]. These techniques include normalization
optical microscopy. Resulting micrographs were used to
of the response data, regression curve fitting, and presenta-
assess weld nugget penetration at the 1–2 interfaces.
tion in robustness plots.
Four measures of weld performance are then reported
Normalization is the process for converting non-normally
here as response variables. These measures were derived
distributed datasets to normal ones through the use of math-
from the mechanical testing and metallurgical evaluations
ematical transformations. This step is essential, as the most
described previously and include:
common statistical analysis approaches are based on the
& Tensile–shear strength at the 1–2 interface comparison of assumed normal subsets inference [19, 20].
248 Weld World (2013) 57:243–256

Table 3 DOE matrix converted into actual machine programmable factors

Trial Thin sheet side electrode Preliminary force and current Secondary force and current
no. configuration

Material Face diameter Force Weld time Weld current Force No. Pulse width Weld current
(class) (mm) (N) (ms) (kA) (N) pulses (ms) (kA)

1 2 6.4 2,400 110 10.0 1,600 2 85 12.0

2 2 6.4 1,700 270 8.0 1,890 1 150 11.2
3 2 6.4 2,400 270 8.0 2,670 2 75 9.6
4 2 6.4 1,700 110 10.0 1,130 1 170 14.0
5 2 6.4 1,700 270 8.0 1,130 1 150 9.6
6 2 6.4 1,700 110 10.0 1,890 1 170 12.0
7 2 6.4 2,400 270 8.0 1,600 2 75 11.2
8 2 6.4 2,400 110 10.0 2,670 2 85 14.0
9 2 8 1,700 170 9.3 1,130 2 125 11.2
10 2 8 2,400 170 9.3 2,670 1 250 11.2
11 2 8 2,400 170 9.3 1,600 1 250 13.1
12 2 8 1,700 180 9.1 1,130 2 50 12.7
13 2 8 2,400 180 9.1 1,600 1 100 10.9
14 2 8 1,700 180 9.1 1,890 2 50 10.9
15 2 8 1,700 170 9.3 1,890 2 125 13.1
16 2 8 2,400 180 9.1 2,670 1 100 12.7
17 3 6.4 2,400 170 9.3 2,670 1 250 13.1
18 3 6.4 2,400 180 9.1 2,670 1 100 10.9
19 3 6.4 1,700 180 9.1 1,130 2 50 10.9
20 3 6.4 2,400 180 9.1 1,600 1 100 12.7
21 3 6.4 1,700 170 9.3 1,130 2 125 13.1
22 3 6.4 1,700 170 9.3 1,890 2 125 11.2
23 3 6.4 2,400 170 9.3 1,600 1 250 11.2
24 3 6.4 1,700 180 9.1 1,890 2 50 12.7
25 3 8 1,700 110 10.0 1,890 1 170 14.0
26 3 8 2,400 270 8.0 2,670 2 75 11.2
27 3 8 2,400 270 8.0 1,600 2 73 9.6
28 3 8 1,700 110 10.0 1,130 1 170 12.0
29 3 8 1,700 270 8.0 1,130 1 150 11.2
30 3 8 1,700 270 8.0 1,890 1 150 9.6
31 3 8 2,400 110 10.0 2,670 2 85 12.0
32 3 8 2,400 110 10.0 1,600 2 85 14.0

Generally, such analyses are derived by comparing mean datasets to more accurately approximate normal conditions.
values and variances of sub-datasets, assuming that these are For the example above, arcsine functions, which include
normally distributed. Such analyses loose accuracy if the scaling terms, are considered effective [21, 22]. Other com-
datasets are not normally distributed [21, 22]. Transforma- monly used transformations include the square root, loga-
tions used to correct normality are typically 1:1 maps of the rithm, and power functions.
dataset into a normalized space. An example of a non- Normalized datasets are then curve fit using standard re-
normal dataset is the transverse failure strength of a welded gression techniques. Of note, once curve fitting is complete,
sample. Clearly, the probability of observing strengths be- the resulting equation must be back-transformed using the
low 0 or above the yield strength of the parent (surrounding) inverse function associated with any normalization treatment
material is 0. This is not consistent with a normally distrib- as described previously. Process robustness plots are graphical
uted dataset. Appropriate mathematical transformations to tools used to interpret the developed regression curves. These
improve normality generally extend the limits of such plots are essentially a series of one-dimensional sections
Weld World (2013) 57:243–256 249

through the multidimensional process spaces defined by the surface of the thin attached sheet. It is of note in these
curves. Here, the curves themselves are first used to provide micrographs that the weld morphology appears to be nearly
an optimized set of conditions (based on the available data). two superimposed nuggets. This is believed to be an artifact
Then, through the curve-fitting equation, the factors of interest of the dual pulse schedule used for both of the trials shown
are varied one at a time over the range studied. The result is a in Figs. 2 and 3.
series of profiles showing the characteristic degradation of the
process associated with each factor as it is varied from its best- 3.2 Statistical analyses
practice condition. Such robustness plots can be used to both
identify critical processing factors and set limits on allowed As described previously, all response datasets were evaluat-
variations for any future manufacturing application. ed for normality prior to regression curve fitting. This was
done by generating a comparative normal dataset based on
the calculated means and standard deviations for the test
3 Results dataset (using commercially available statistical analysis
packages) and correlating this with the actual test data.
3.1 DOE experimental results Where normality could be improved, appropriate transfor-
mation functions were used. These transformation functions
The experimental results from the DOE trials are presented were developed in an iterative manner, maximizing the
in Table 4. This includes three replicates of the shear load to correlation of the response data with the comparative normal
failure at the 1–2 interface, seven replicates of the indenta- dataset. The resulting transformation equations and normal-
tion measurements for the thin attached sheet, six replicates ity confidence levels are provided in Table 5. All response
of the 1–2 interface nugget size, and one measurement of datasets with the exception of the shear load at the 1–2
weld nugget penetration into the thin sheet. Tensile–shear interface and thin sheet nugget penetration required normal-
data at the 1–2 interface (thin sheet attachment) showed only ity correction. The normality corrections used were, in gen-
a small variation over the experiment, from nominally 4 to eral, power functions, although the thin sheet side
6 kN. These results were reasonably consistent over the indentation data was corrected with a natural logarithm
three replicates taken. Here, loads to failure were seen to function. Generally, normality confidence levels of >98 %
vary by usually <10 % and never more than 20 %. were achieved. The exception was the shear load at the 1–2
Indentation measurements showed considerable variation interface, which showed a correlation of only about 80 %.
over the range of conditions studied in the DOE. The data Regression equations and statistical results for each of the
presented here show that indentations on the thin sheet side eight response variables are presented in Table 6. Each of
ranged from 0.2 up to 1.65 mm. There was also considerable these sets of results is based on first a stepwise regression to
scatter within each DOE run. The weld button size data for select factors for the curve fit model that could be correlated
the 1–2 interface presented here was taken directly from the to the results to at least an 80 % confidence level. Included
1–2 interface tensile–shear tests. in this table is the resulting regression model, as well as the
Nugget penetrations into the thin sheet were measured standard error and R2 value for the fit. Fits for the regression
from metallographic specimens. One weld was taken from analyses shown here range from moderate (transformed
each trial and used for this sectioning. For each micrograph, button size at the 1–2 interface, R2 056 %) to excellent
the minimum distance from the top sheet external surface (penetration, R2 094 %). Generally, the quality of the fit
and the apparent top sheet thickness were measured. Pene- degraded with two aspects of the respective datasets. These
trations were then calculated from the equation: included low overall variation in the data from trial to trial
and a high degree of variation within the trial. This is most
tS tR
% Penetration ¼  100 % ð2Þ clearly demonstrated in the case of the transformed button
tS
size at the 1–2 interface. Here, the average measured button
where tS and tR are the total and residual sheet thicknesses as size from trial to trial varied by <3 mm, and replicate results
measured from the micrographs, respectively. Penetrations within a trial were seen to achieve similar levels. The small
measured in this way are seen to vary from 0 up to 80 % variation in results from trial to trial reduces apparent de-
throughout this experiment. Micrographs showing extremes pendency between the response and input variables (reduc-
in penetration are shown in Figs. 2 and 3. Figure 2 is taken ing R2), while the variation within a trial increases the
from trial 7 and represents conditions of 0 % penetration. estimated error, again reducing the R2 values. These regres-
From this micrograph, it is clear that the weld nugget at best sion analyses, however, only include response variables
just touches the thin attached sheet, with no effective pene- which showed individual correlations of above an 80 %
tration. Alternately, the micrograph taken from trial 32 confidence level with from 13 to 18 of the input factors
(Fig. 3) shows a nugget that penetrates nearly to the exterior and two-factor interactions in each case.
250 Weld World (2013) 57:243–256

Table 4 Indentation measurements for each run of the DOE

Run Shear failure load at Thin sheet side indentation (mm) 1–2 interface button size (mm) Penn.
the 1–2 interface (kN) (%)

Rep. 1 Rep. 2 Rep. 3 Rep. 1 Rep. 2 Rep. 3 Rep. 4 Rep. 5 Rep. 6 Rep. 7 Rep. 1 Rep. 2 Rep. 3 Rep. 4 Rep. 5 Rep. 6

1 4.7 4.5 4.8 – – – – – – – 6.5 7.3 7.5 7.2 6.7 6.5 40

2 4.8 4.5 4.5 0.53 0.81 0.51 0.48 0.64 0.51 0.36 6.1 6.8 5.8 7.4 7.2 7.6 0
3 4.4 4.1 4.4 0.36 0.41 0.46 0.33 0.33 0.51 0.28 6.4 4.7 6.4 3.5 5.4 4.1 7
4 5 5 5.1 – – – – – – – 8 8 8 7.1 7.3 6.9 47
5 3.9 4.1 4 0.25 0.25 0.3 0.23 0.28 0.25 0.25 6.3 6.3 5.7 5.5 4.9 5 6
6 4.4 4.5 4.7 0.38 0.38 0.38 0.25 0.3 0.3 0.38 6.4 6.9 7.4 5.4 5.4 6.1 33
7 4.3 4.4 4.2 0.25 0.23 0.38 0.43 0.23 0.41 0.36 6.4 5.1 6.2 5.5 4.7 5.5 0
8 4.9 5.2 5.2 0.56 0.74 0.89 0.66 0.89 0.79 0.3 7.5 8 8.3 6.7 7.7 7.2 25
9 5.1 4.9 4.9 0.66 0.41 0.41 0.43 0.48 0.38 0.48 7.7 7.7 7.5 6.2 6.3 6.5 7
10 5 4.8 5.1 0.64 1.04 0.61 0.84 0.76 0.66 0.51 8.2 7.8 8.3 6.8 7.1 6.9 33
11 5.1 5.5 5.2 0.81 0.69 0.66 1.02 0.56 0.69 1.5 8.2 8.2 8.3 6.8 7.8 7.5 31
12 4.4 4.5 4.7 0.43 0.36 0.38 0.28 0.43 0.38 0.51 7 6.8 7.1 5.7 6.2 6.5 25
13 5.2 4.6 4.5 0.38 0.36 0.51 0.43 0.71 0.51 0.69 7.2 7.3 7 6.3 6.7 6.7 50
14 4.6 4.6 4.5 0.33 0.36 0.38 0.43 0.28 0.48 0.48 6.8 6.3 6.8 5.2 6.4 6.2 40
15 5 5.9 5.2 0.69 0.46 0.64 0.71 1.17 0.56 0.66 7.7 7.8 7.7 7.4 8.6 6.2 44
16 4.9 4.8 4.8 0.99 1.35 0.86 0.71 0.81 1.02 0.99 7.9 7.8 7.9 7.5 7.1 7.5 50
17 5 4.8 5.2 1.27 1.65 1.22 1.02 1.65 0.76 0.76 7.9 8.2 9.4 7.5 7.3 7.5 38
18 4.8 4.6 4.5 0.36 0.58 0.25 0.69 0.33 0.56 0.56 7.2 7 6.5 6.7 6.9 6.6 50
19 4.5 4.4 4.4 0.3 0.33 0.3 0.3 0.33 0.33 0.33 7.1 6.7 6.6 5.8 6 5.9 60
20 4.9 4.7 4.8 0.51 0.48 0.43 0.48 0.46 0.51 0.56 7.6 7.6 7.8 6.8 7 6.9 73
21 4.9 5 5.2 0.64 0.64 0.66 0.69 0.53 0.58 0.58 8 6.6 8 7.6 7.6 7.5 67
22 4.8 4.5 4.9 0.51 0.46 0.56 0.48 0.74 0.61 0.56 7.9 7.5 7.7 6.2 6.1 6.6 73
23 5.2 5.2 5.1 0.43 0.56 0.56 0.51 0.46 0.66 0.61 8 7.3 8.1 6.3 6.4 6.5 44
24 4.4 4.4 4.2 0.36 0.33 0.38 0.38 0.36 0.36 0.36 7 7 6.9 5.3 6.2 5.8 53
25 5.2 5.2 5.1 0.71 0.61 1.04 0.43 0.56 0.74 0.58 8.8 8.1 8.7 7.5 7.6 7.7 53
26 4.4 4.5 4.4 0.48 0.36 0.38 0.36 0.41 0.61 0.41 6.4 6.4 6.4 5.7 4.1 6.2 27
27 4.5 4.9 4.4 0.48 0.36 0.2 0.25 0.36 0.41 0.33 7 6.6 6.4 6.5 6.3 6.5 67
28 4 5.1 4.9 0.43 0.58 0.41 0.33 0.58 0.61 0.38 7.9 7.8 7.8 7.1 7 6.8 44
29 4.8 4.5 4.4 0.43 0.48 0.43 0.36 0.43 0.38 0.51 7.5 7.6 7.3 6.8 7.2 6.4 56
30 4.4 4.4 4.8 0.25 0.25 0.3 0.3 0.33 0.25 0.28 7 7.3 7.5 6.9 6.6 6.2 50
31 5 5.1 5.2 0.58 0.76 0.51 0.41 0.46 0.38 0.51 7.7 7.7 7 5.9 4.9 6.1 31
32 5.4 5.1 5.2 0.46 0.36 0.41 0.53 0.53 0.46 0.38 7.8 7.6 8 7.1 7.5 7.5 80

3.3 Process robustness plots anchor conditions that are consistent with current industry
practice [17]. Specifically, this meant using class 2 electro-
The regression analyses were then used to prepare process des and a matching (6.3-mm) face diameter on the thick
robustness plots for each of the response variables studied. sheet side of the joint. Anchor conditions for these plots then
The resulting plots for the eight responses evaluated in this included:
study are provided in Figs. 4, 5, 6, and 7. All plots have been
anchored at the same baseline set of conditions for the input & Class 2 upper electrode
factors. These conditions have been defined empirically & A 6.4-mm electrode face diameter on the thick sheet side
through comparison of the plots for the different response of the joint
characteristics. Performance was optimized in these plots by & A total weld time of 284 ms
maximizing weld nugget penetrations into the thin sheet, & Forty percent of the overall weld time in stage 1
button sizes, and joint strengths, while minimizing indenta- & A single weld pulse for stage 2
tions. In addition, allowance was made to provide a set of & A 2,400-N stage 1 weld force
Weld World (2013) 57:243–256 251

button sizes below 6.8 mm or above 8.4 mm. The process

robustness plot for the degree of electrode indentation at the
thin sheet surface is provided in Fig. 6. Here, the anchor
conditions used resulted in nominally 0.45 mm of indenta-
tion into the thin attached sheet. This corresponds to roughly
45 % indentation of the sheet on that side of the joint.
Further, the robustness plot suggests that indentation is
sensitive to a range of the DOE factors. Specifically, greater
indentations could be driven by larger thick sheet side
electrode diameters (up to 0.85 mm), higher forge/weld
force ratios (up to 0.75 mm), as well as a class 3 electrode
Fig. 2 Cross-section of the weld in trial 7, showing 0 % penetration
on the thin sheet side (up to 0.65 mm). Conversely, inden-
into the attached thin sheet (this weld was made under conditions of a tations could be reduced by use of relatively longer stage 1
class 2 electrode on the thin sheet side, a 6.3-mm electrode on the thick currents (down to 0.33 mm), pulsation during the stage 2
sheet side, a 420-ms weld time, two pulses in the main current, a current (0.36 mm), a lower relative stage 2 current
secondary to primary force ratio of 1.5, a primary to total weld time
ratio of 0.65, a 2,400-N weld force, and a secondary to primary current
(0.35 mm), and a lower welding force (0.38 mm).
ratio of 1.4) Finally, the process robustness plot for the degree of
penetration into the outside thin sheet is presented in
& A 50 % increase in weld force from stage 1 into stage 2 Fig. 7. At anchor conditions, this penetration was estimated
& A 40 % increase in weld current from stage 1 into stage 2 at roughly 0.53 mm. This penetration, however, is affected
by a number of the DOE factors. Particularly, variations in
the electrode configurations themselves appear most affec-
The resulting plot for the weld shear strength at the 1–2 tive at increasing penetrations. For example, increasing the
interface is shown in Fig. 4. Predicted strengths at this thin sheet side electrode to class 3 copper increases penetra-
interface using the defined anchor conditions are roughly tions to over 0.8 mm, while the use of an 8-mm electrode on
5.2 kN. It can be seen that none of the DOE factors varied the thick sheet side electrode results in estimated penetra-
individually, resulted in strengths <4.5 or >5.5 kN. A com- tions of nearly 0.7 mm. Penetration is most adversely affect-
parable process robustness plot for the measured button size ed by lengthening the weld time. Here, increasing the weld
at the 1–2 interface is provided in Fig. 5. At anchor con- time to 417 ms reduces the estimated penetrations to
ditions, the plot suggests that a button size of almost 8 mm <0.1 mm. Higher relative stage 2 weld currents, lower weld
can be achieved. This is in excess of the generally accepted forces, shorter relative stage 1 welding times, and use of
6√t criteria incorporated in many recommended practices pulsation were all detrimental to nugget penetrations. The
and specifications [4–6, 17]. Of note, none of the plotted effect in each case was <0.15 mm.
variations from these anchor conditions resulted in predicted

4 Discussion

4.1 Influence of electrode variations on heat balance

in complex stack-ups

It is of note in this study that overall best penetrations into

the thin attached sheet of this stack-up were achieved using
variations in electrodes. This included using differentially
sized electrodes (larger electrode on the thick sheet side of
the joint), as well as different thermal conductivity electro-
des. The results here generally confirm the recommenda-
tions made in both the Resistance Welding [4] and Welding
Fig. 3 Cross-section of the weld in trial 32, showing 80 % penetration [5] Handbooks. That is, changing the heat extraction capa-
into the attached thin sheet (this weld was made under conditions of a bilities of the electrodes on either side of the joint can
class 3 electrode on the thin sheet side, an 8-mm electrode on the thick effectively move the heat center of the stack-up. The effects
sheet side, a 280-ms weld time, two pulses in the main current, a
secondary to primary force ratio of 1.5, a primary to total weld time
of different electrode materials and electrode sizes can be
ratio of 0.4, a 2,400-N weld force, and a secondary to primary current understood by employing a simple heat flow analysis. In
ratio of 1.4) this analysis, heat balance is assessed by comparing the
252 Weld World (2013) 57:243–256

Table 5 Normalization equa-

tions for each response factor Response variable Normalization equation Normality confidence level (%)
in this study (also included in
this table are the normality Shear load at the 1–2 interface None 99.5
confidence levels for the data Shear load at the 2–3 interface Xshear2 3 ¼ ðshear2 3 Þ1:7 99.2
transformed with these Xchisel2 3 ¼ ðchisel2 3 Þ1=6
Chisel load at the 2–3 interface 98.1
equations)
Button size at the 1–2 interface Xbutton1 2 ¼ ðbutton1 2 Þ2:35 99.6
Button size at the 2–3 interface Xbutton2 3 ¼ ðbutton2 3 Þ2:6 98.5
Thin sheet side indentation Xindentthin ¼ lnðindentthin Þ 99.1
Thick sheet side indentation Xindentthick ¼ ðindentthick Þ1=4 99.3
Thin sheet penetration None 99.7

heat-extracting ability of the electrodes on the two sides of Since ft +fT 01, this equation can be expressed as heat bal-
the stack-up. For steady-state conditions, the heat extraction ance from the thin side as:
balance can be approximately expressed as:
 1
Qt kt At ft ¼   ð5Þ
 ¼ ð3Þ kt A t
1þ
QT k T AT kT A T

 
where Qt and QT are the heat extractions, kt and kT are the
For the stack-up considered here (1+2+2 mm), at-
thermal conductivities, and At and AT are the effective areas
tachment of all three sheets is best accomplished by
for the electrodes on the thin and thick sides of the joints,
locating the heat center in the middle of the center
respectively. If two further simplifications are made, that
sheet. This corresponds to ft 040 %. Through application
thermal gradients in the steel on the thin and thick sides
of Eq. 5, the associated ratio of electrode characteristics
are assumed to be linear from the peak temperature and the  
kt A t
electrode face temperatures on both sides are the same, then kT AT ¼ 1:5. For the study conducted here, the maxi-
biasing of the heat center can be expressed as: mum ratios of electrode areas and thermal conductivities
ft kT A T were 1.6 and 2 (room temperature thermal conductivity
¼ ð4Þ values), respectively. Not surprisingly, through the
fT kt A t
course of this study, it was possible to change the
where ft and fT are the fraction (of the stack-up) penetration degree of nugget penetration in the thin sheet from 0
of the heat center from the thin and thick sides, respectively. to 100 %, largely related to electrode variations.

Table 6 Regression equations and statistical support data for each response used in this study (transformed factors are indicated by an “X” prefix)

Response Preliminary regression equation Standard R2

error (%)

Shear load at the Ten1–204,762−235 Stg1–2Timing+92.2 ElectDia+85.7 ILevel1–2−79.3 UpElect×ElectDia+64.4 197 74.6
1–2 interface ForceStg1−62.6 Stg1–2Timing×ILevel1–2−50.5 UpElect×ILevel1–2−50.5 WeldTime×ForceStg1−
41.2 UpElect×ForceRto−41.2 ForceStg1×ILevel1–2−39.4 ForceRto×ForceStg1−33.8 W2Pulse×
ILevel1–2−28.3 W2Pulse
Button size at the XAvgBS1–2096.4−13.7 Stg1–2Timing+8.66 ILevel1–2−7.92 W2Pulse−5.44 UpElect×ElectDia+ 19.6 56.1
1–2 interface 4.93 ElectDia−3.59 Stg1–2Timing×ILevel1–2−3.57 WeldTime×ForceStg1−3.37 UpElect−3.08
W2Pulse×ILevel1–2−2.75 ElectDia×W2Pulse−2.59 W2Pulse×ForceRto−2.14 WeldTime−2.09
UpElect×ForceRto+2.02 ForceRto×ILevel1–2
Thin sheet side XIndE-10−3.99−0.156 Stg 1–2Timing+0.136 ILevel1–2−0.148 UpElect×ElectDia+0.103 0.221 72.5
indentation ForceStg1−0.0837 W2Pulse−0.0637 W2Pulse×ILevel1–2+0.100 ForceRto+0.0443 ForceRto×
ForceStg1−0.0615 ElectDia×WeldTime+0.0527 ForceRto×ILevel1–2+0.0518 ElectDia−0.0399
UpElect×ForceRto−0.0357 ElectDia×W2Pulse+0.0319 WeldTime−0.0273 ElectDia×ForceRto−
0.0295 WeldTime×ForceStg1+0.0261 WeldTime×ILevel1–2
Thin sheet Penetration00.408+0.134 UpElect−0.0638 WeldTime−0.0538 UpElect×ElectDia+0.0500 UpElect× 0.0825 93.7
penetration WeldTime−0.0494 ForceRto×ForceStg1−0.0444 UpElect×ForceRto−0.0394 Stg1–2Timing×
ILevel1–2+0.0350 UpElect×W2Pulse−0.0312 WeldTime×ForceStg1−0.0281 ForceRto+0.0275
ElectDia×WeldTime−0.0275 ForceRto×ILevel1–2−0.0256 WeldTime×ILevel1–2−0.0250
ElectDia×W2Pulse+0.0250 Stg1–2Timing×ForceStg1+0.0244 WeldTime×ForceRto−0.0238
Stg1–2Timing+0.0225 ElectDia
Weld World (2013) 57:243–256 253

6 9

Average Button Size 1-2 (mm)

5.5
7
Tensile Peak Load 1-2 (N)

5 6

5
4.5
4
Upper Electrode Class
Lower Electrode Dia
3 Total Weld Time
4 Stg2 Pulsation
Force Ratio Stg2/ Stg1
Upper Electrode Class Welding Timing Stg1/Stg2
Lower Electrode Dia 2
Force Stg1
Total Weld Time I Level Stg2/Stg1
3.5 Stg2 Pulsation 1
Force Ratio Stg2/ Stg1
Welding Timing Stg1/Stg2
Force Stg1 0
I Level Stg2/Stg1
3 Class 2 Upper Electrode Class 3
Class 2 Upper Electrode Class 3 6.4-mm Lower Electrode Diameter 8.0-mm
6.4-mm Lower Electrode Diameter 8.0-mm 284 Total Weld Time 417
1 Pulse Stage 2 Pulsation 2 Pulses
284 Total Weld Time 417
1.5 Force Ratio (Stg1/Stg2) 2.5
1 Pulse Stage 2 Pulsation 2 Pulses
0.40 Weld Timing (Stg1/Stg2) 0.65
1.5 Force Ratio (Stg1/Stg2) 2.5
1700-N Force Stg 1 2400-N
0.40 Weld Timing (Stg1/Stg2) 0.65
1.2 I Level (Stg1/Stg2) 1.4
1700-N Force Stg 1 2400-N
1.2 I Level (Stg1/Stg2) 1.4
Fig. 5 Process robustness plot showing the effects of the DOE varia-
Fig. 4 Process robustness plot showing the effects of the DOE varia- bles on the average button size at the thin/thick (1–2) interface (the plot
bles on the peak tensile load at the thin/thick (1–2) interface (the plot is is anchored at base conditions of a class 2 electrode on the thin sheet
anchored at base conditions of a class 2 electrode on the thin sheet side, side, a 6.4-mm electrode face diameter on the thick sheet side of the
a 6.4-mm electrode face diameter on the thick sheet side of the joint, a joint, a 284-ms total weld time, with 40 % of the overall weld time in
284-ms total weld time, with 40 % of the overall weld time in stage 1, a stage 1, a single weld current pulse for stage 2, a 2,400-N stage 1 weld
single weld current pulse for stage 2, a 2,400-N stage 1 weld force, a force, a 50 % increase in weld force from stage 1 into stage 2, and a
50 % increase in weld force from stage 1 into stage 2, and a 40 % 40 % increase in weld current from stage 1 into stage 2)
increase in weld current from stage 1 into stage 2)
electrodes during replacements where large numbers of
4.2 Manufacturing considerations of the selection caps may be replaced simultaneously. As might be
of baseline conditions in this study anticipated, reversal of caps on a complex stack-up
would tend to move the heat center away from the
The discussion above underscores the dominance of desired location, with catastrophic results in a manufac-
electrode variations in achieving attachment of thin turing environment.
sheet outside elements in complex stack-ups. However, For these reasons, use of dissimilar electrodes cannot
as mentioned previously in this report, the application be considered a solution to attaching thin outer sheets in
of different electrodes to the two sides of such stack- a complex stack-up. As a result, the use of similar-
ups in actual automotive manufacture can be problem- sized, similar-material electrodes were selected in the
atic at best. The use of different electrodes has direct process robustness studies conducted here as part of
cost implications. Specifically, both electrode types the baseline conditions. Definition of a best welding
must be inventoried, adding cost to the process. In practice for servo-guns must then be focused on opti-
addition, there are concerns that, in modern automotive mizing weld times, as well as weld/forge force and
manufacturing, a single gun may be used to make spot current ratios.
welds on a range of stack-ups as part of a single
operation. Mandating dissimilar electrode combinations 4.3 Best practices for attachment of thin sheets in complex
for specific complex stack-up joints carries the sugges- stack-ups
tion of dedicated welding guns for specific joints. The
use of such dedicated guns inherently increases capital Optimization of the remaining DOE variables in this study
requirements and reduces flexibility. Both directly affect to achieve best penetrations on the thin sheet side of the joint
manufacturing costs in a detrimental way. The final suggested the use of short overall weld times, moderate
concern associated with dissimilar welding electrodes forge force to weld force ratios, and high forge current to
is related to maintenance. In high-volume manufactur- weld current ratios. Of these, weld time effects were dom-
ing environments, there is concern about confusing inant. The use of shorter weld times to achieve improved
254 Weld World (2013) 57:243–256

1 90
Upper Electrode Class
0.9 80
Lower Electrode Dia

Total Weld Time 0.8 70

Indentation E-1 (mm)

Stg2 Pulsation

Penetration 1-2 (%T)

0.7
Force Ratio Stg2/ Stg1 60
0.6
50
0.5
40
0.4
30
Upper Electrode Class
0.3 Lower Electrode Dia
Total Weld Time
Stg2 Pulsation
20
0.2
Force Ratio Stg2/ Stg1
Welding Timing Stg1/Stg2 10
0.1 Force Stg1
I Level Stg2/Stg1
0 0
Class 2 Upper Electrode Class 3
Class 2 Upper Electrode Class 3
6.4-mm Lower Electrode Diameter 8.0-mm
6.4-mm Lower Electrode Diameter 8.0-mm
284 Total Weld Time 417
284 Total Weld Time 417
1 Pulse Stage 2 Pulsation 2 Pulses
1 Pulse Stage 2 Pulsation 2 Pulses
1.5 Force Ratio (Stg1/Stg2) 2.5
1.5 Force Ratio (Stg1/Stg2) 2.5
0.40 Weld Timing (Stg1/Stg2) 0.65
0.40 Weld Timing (Stg1/Stg2) 0.65
1700-N Force Stg 1 2400-N 1700-N Force Stg 1 2400-N
1.2 I Level (Stg1/Stg2) 1.4 1.2 I Level (Stg1/Stg2) 1.4

Fig. 6 Process robustness plot showing the effects of the DOE varia- Fig. 7 Process robustness plot showing the effects of the DOE varia-
bles on the measured indentation at the thin sheet (E-1) surface (the bles on the nugget penetration into the thin attached sheet (1–2 inter-
plot is anchored at base conditions of a class 2 electrode on the thin face) (the plot is anchored at base conditions of a class 2 electrode on
sheet side, a 6.4-mm electrode face diameter on the thick sheet side of the thin sheet side, a 6.4-mm electrode face diameter on the thick sheet
the joint, a 284-ms total weld time, with 40 % of the overall weld time side of the joint, a 284-ms total weld time, with 40 % of the overall
in stage 1, a single weld current pulse for stage 2, a 2,400-N stage 1 weld time in stage 1, a single weld current pulse for stage 2, a 2,400-N
weld force, a 50 % increase in weld force from stage 1 into stage 2, and stage 1 weld force, a 50 % increase in weld force from stage 1 into
a 40 % increase in weld current from stage 1 into stage 2) stage 2, and a 40 % increase in weld current from stage 1 into stage 2)

heat balances is not unknown. Shorter weld times (and penetration. This suggests that there may be some sec-
higher currents) are known to react more aggressively with ondary growth of the weld nugget (during forging), im-
available contact resistances, preferentially heating the var- proving the resulting penetrations.
ious interfaces in the stack-up. As a result, there is tendency
to more readily heat the thin/thick interface, increasing both 4.4 Tradeoffs between penetrations, indentations, and joint
the degree of subsequent nugget penetration and likelihood strengths
of attachment. Of interest, higher forces and moderate forge/
weld force ratios both improved penetrations. In the former As described above, under baseline practice conditions as
case, this is believed to be related to the ability to grow defined in this DOE, penetrations of over 50 % (0.5 mm)
larger weld nuggets before expulsion (thin/thick sheet but- could be obtained into the thin sheet. It is of note that,
ton sizes also increased). once these baseline practices are defined, joint strengths
The increases in penetration related to the reduced rela- are relatively insensitive to any of the process variations
tive forge forces is believed to be associated with indenta- studied (not including electrode variations). The major
tion effects. It is observed that, at higher relative forge tradeoff appears to be between penetrations and indenta-
forces, indentations also increased. This increase in inden- tions. Under the baseline conditions, electrode indentions
tation has the effect of suddenly changing the geometry of of roughly 0.45 mm (45 % of the thin sheet thickness)
the weld relative to the temperature field. As this occurs, are observed. This may or may not be considered exces-
there is a redistribution of heat in accordance with the new sive based on some automotive specifications. As sug-
geometry. This redistribution appears to reduce the fraction gested above, the DOE results do suggest some potential
of the remaining thin sheet contained within the nugget, tradeoffs between nugget penetration and indentation.
effectively reducing effective penetrations on this side of Specifically, increased relative weld time as well as re-
the joint. duced forces each individually offer reductions in elec-
Finally, as noted above, higher forge/weld current trode indentation of roughly 0.1 mm, while maintaining
ratios appeared to have a beneficial influence on nugget nugget penetrations >0.4 mm (40 %). Most beneficial
Weld World (2013) 57:243–256 255

tradeoffs between these two characteristics will undoubt- electrode size and conductivity. The results of that anal-
edly be defined by the individual application. ysis were consistent with the DOE results.
4. Electrode variations cannot be considered a solution
to heat balance problems for complex stack-ups in
5 Conclusions an automotive environment: The automotive manu-
facturing environment makes use of differential elec-
In this program, the resistance spot weldability of com- trodes for RSW prohibitive. Concerns include
plex stack-ups has been investigated with respect to the inventorying the different electrodes, proper mainte-
capabilities of new-generation MFDC electric servo-guns. nance with such electrodes, and the fact that indi-
These systems include capabilities for varying both cur- vidual guns are often used to make differing stack-
rent and force profiles during the welding cycle. This up joints on a single assembly.
investigation was conducted on a specific complex 5. Nugget penetrations into the thin sheet on the order of
stack-up, incorporating a 1- to 2- to 2-mm arrangement. 50 % could be achieved using the developed baseline
Weldability was assessed using DOE techniques. Factors conditions: The developed baseline conditions included
studied in this experiment included the use of differen- similar-diameter class 2 electrodes, short overall weld
tial electrode materials and sizes, primary welding times, a smaller fraction of the time during welding
forces, total welding times, time fractions of primary (before forging), a single pulse during forging, higher
current flows, numbers of pulses in the secondary cur- (2,400-N) weld forces, a moderate (50 %) increase in
rent flow, ratios of forge to welding currents, and ratios force during forging, and a higher (40 %) increase in
of forge to welding forces. Response variables in this current during the forge stage.
DOE included tensile–shear strengths at both the thin/ 6. Indentations of <0.5 mm were observed at baseline
thick and thick/thick interfaces, the chisel strengths at conditions in this study: These indentations represent
the thick/thick interface, button sizes at both the thin/ up to 50 % the thin attached sheet thickness. This level
thick and thick/thick interface, indentations at the thin of indentation may be problematic for some automotive
sheet and thick sheet sides of the joint, and nugget applications.
penetration into the thin sheet. The DOE was then used 7. Tradeoffs are observed between nugget penetrations and
to produce robustness plots for each response variable. indentations when RSW complex stack-ups: It was not-
Comparison of these robustness plots allowed definition ed that several factors could be used to reduce indenta-
of some baseline processing conditions for creating tion at the expense of penetration. These include
these joints. Individual plots then allowed interpretation reduced forces and longer fractions of the weld time
of performance variations associated with each variable prior to forging. The exact compromise of indentations
studied in the DOE. Specific conclusions from this work and penetrations will depend on the specifics of the
include: application.
8. Joint strengths were not strongly affected by process
1. Baseline practice for RSW complex stack-ups is primar- (or electrode) variations from baseline conditions:
ily based on nugget penetrations into the thin sheet: Once baseline conditions were established, varia-
Nugget penetration into the thin sheet was used as the tions in individual factors did not change the
primary measure of weld quality in the DOE study strengths by more than 10–20 %. These results
conducted here. This measure was found to be the most indicate an inherent robustness of RSW thin/thick/
sensitive to the process/electrode variations studied and thick stack-us.
is widely accepted in industry. 9. The process features enabled by MFDC electric
2. Weld nugget penetration results were dominated by servo-guns appear to offer advantages for RSW
electrode considerations in this study: Weld nugget pen- complex (thin/thick/thick) stack-ups: The results pre-
etrations were found to be greatly enhanced by using a sented in this report suggest that the ability to apply
class 3 electrode on the thin sheet side, as well as by a relatively high forge forces, as well as to increase
larger-diameter electrode on the thick sheet side. These currents during forging, and coordinate current var-
effects were stronger than for any of the process-related iations with the forging profile all improve penetra-
variables included in this study. tions during resistance welding thin/thick/thick
3. Influence of electrode variations on heat balance is stack-ups.
consistent with previous work: The observations made
regarding the effects of electrode variations are consis-
tent with previous work. This was supported with some Acknowledgments The EWI acknowledges the contribution of the
limited analysis quantifying the effects of differential State of Ohio, Department of Development and Thomas Edison
256 Weld World (2013) 57:243–256

Program, which provided funding in support of Edison Technology 11. Slavik SA (1999) Using servo-guns for automated resistance weld-
and Industry Center Services. ing. Welding Journal 78(7)29–33
12. Barthelemy P (2004) Servo weld gun—present and future. Sheet
Metal Welding Conference XI, Detroit AWS Section, Detroit,
Paper 3-7
References 13. Grimes P (2006) Advantages of using servo force control when
resistance welding aluminum sheet metal. Sheet Metal Welding
Conference XII, Detroit AWS Section, Detroit, Paper 6-2
1. Dinda S, Diaz R (1995) The partnership for a new generation of 14. Lehman LR, Gould JE (1994) A study of resistance spot welding
vehicles (PNGV) and its impact on body engineering. Proceedings manufacturability using design-of-experiments. International Body
of the IBEC 95, Advanced Technologies and Processes, IBEC Engineers Council (IBEC) 94 Proceedings, Advanced
Ltd., pp 5–8 Technologies and Processes, IBEC Ltd., Warren, pp 154–163
2. Crooks MJ, Miner RE (1996) The ultralight steel auto body pro- 15. Lehman LR, Gould JE (1995) A design-of-experiments evaluation
gram completes phase I. Journal of Metals 48(7):13–15 of resistance spot welding manufacturability—part 2: multiple
3. Bleck W (1996) Cold rolled, high-strength sheet steels for auto factor effects. IBEC 95 Proceedings, Advanced Technologies and
applications. Journal of Metals 48(7):26–30 Processes, IBEC Ltd., Warren, pp 88–99
4. Resistance Welder Manufacturers Association (2003) Resistance 16. Lehman LR, Gould JE (1996) A design-of-experiments evaluation
welding manual, 4th edn. Resistance Welder Manufacturers of resistance spot welding manufacturability—part 3: optimization
Association, Miami and process robustness studies. IBEC 96 Proceedings, Advance
5. Welding Handbook, 9th Ed (2007) Welding processes, part 2, Vol. Technologies and Processes, IBEC Ltd., Warren
3rd edn. American Welding Society, Miami, pp 1–48 17. American Welding Society (2002) Recommended practices for test
6. Auto-Steel Partnership (2008) A/SP starting resistance spot weld methods for evaluating the resistance spot welding behavior of
schedules for AHSS. Auto-Steel Partnership, Detroit automotive sheet steel materials. AWS/SAE D8.9M:2002
7. Lu F, Karagoulis MJ, Dong P (2000) The influence of different 18. Gould JE (1994) Modeling primary dendrite arm spacings in
sheet combinations on nugget development during resistance spot resistance spot welds, part 2—experimental studies. Welding
welding of thick stack-ups. Sheet Metal Welding Conference IX, Journal Research Supplement 73(5):91s–100s
Detroit AWS Section, Detroit, Paper 1-4 19. Diamond WJ (2001) Practical experiment designs for scientists
8. Fong M, Tsang A, Ananthanarayanan A (2000) Development of and engineers, 3rd edn. Wiley, New York
the law of thermal similarity (LOTS) for low-indentation cosmetic 20. DeGroot MH (1986) Probability and statistics, 2nd edn. Addison
resistance welds. Sheet Metal Welding Conference IX, Detroit Wesley, Reading
AWS Section, Detroit, Paper 5-6 21. Ahrens WH, Cox DJ, Budhwar G (1990) Use of the arcsine and
9. Agashe S, Zhang H (2002) Selection of schedules based on heat square root transformations for subjectively determined percentage
balance in resistance spot welding. Sheet Metal Welding data. Weed Science 38(4–5):452–458
Conference X, Detroit AWS Section, Detroit, Paper 1-2 22. Osborne J (2002) Notes on the use of data transformations.
10. Anderson C, Wiermaa C, Morel MK (2000) Developments in Practical Assessment, Research & Evaluation 8(6). Retrieved from
resistance spot welding. Practical Welding Today 4(6):38–40 http://PAREonline.net/getvn.asp?v08&n06