Journal of Materials Processing Technology 182 (2007) 107–116

Optimizing multiple quality characteristics via Taguchi

method-based Grey analysis
Lung Kwang Pan a,∗ , Che Chung Wang b , Shien Long Wei c , Hai Feng Sher d
aDepartment of Radiological Technology, Central Taiwan University of Science and Technology, Takun, Taichung 406, Taiwan
b Department of Industrial Design, Shih-Chien University, DaZhi, Taipei 104, Taiwan
c Department of Weapon System Engineering, Chung-Cheng Institute of Technology, National Defense University, Tashi, Taoyuan 335, Taiwan
d Department of Industrial of Management, HsiuPing Institute of Technology, TaLi, Taichung 412, Taiwan

Received 4 February 2005; received in revised form 12 January 2006; accepted 10 July 2006

Abstract
This paper presents and demonstrates the effectiveness of optimizing multiple quality characteristics of Nd:YAG laser welded titanium alloy
plates via Taguchi method-based Grey analysis. The modified algorithm adopted here was successfully used for both detraining the optimum
settings of machine parameters and for combining multiple quality characteristics into one integrated numerical value called Grey relational grade
or rank. The essential machining parameters were identified as (1) shielding gas, (2) laser energy, (3) traveling speed of the workpiece during
welding, (4) focusing position, (5) frequency of the impulse laser, and (6) pulse shape of the primary laser beam. Furthermore, the multiple quality
characteristics required herein included: (1) high ultimate tension stress, (2) small heat-affected zone, (3) large welding depth to width ratio, and
(4) low surface roughness following laser welding. Titanium alloy plates were butt welded under controlled machine parameter settings, machined,
and above quality characteristics were measured. The optimized machine parameter settings clearly improved the quality characteristics of welded
plates compared to quality levels achieved for conventional machine parameter settings. The tensile stress increased (922 ± 53 to 965.5 ± 46.5 MPa)
and the heat-affected zone decreased (0.27–0.22 mm) while the welding depth to width ratio remained the same at 0.76. The new approach increased
the surface roughness from Rmax of 14.2 to 34.4 ␮m.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Taguchi method; Grey analysis; Nd:YAG laser; Titanium alloy; Welding

1. Introduction ful. In contrast, Grey analysis was first proposed by Dr. Deng in
1982 to fulfill the crucial mathematical criteria for dealing with
Taguchi method has been extensively adopted in manufac- poor, incomplete, and uncertain system [8,9]. As implied by its
turing for nearly three decades to robustly design a product or name, which implies a shade between the absolutes of black and
process with a single quality characteristic (QC). The QC may white, Grey analysis can effectively recommend a method of
be to maximize the ultimate tensile stress of welded parts, min- optimizing the complicated inter-relationships among multiple
imize the surface roughness of parts after electrical discharge performance characteristics.
machining, or maximize the rate of debris removal in a milling A modified algorithm was thus established for combining
operation [1–7]. The unique calculation in Taguchi method algo- both Taguchi method and Grey analysis, and was successfully
rithm converts the quantitative assessment of specific QC into verified via butt welding of titanium alloy in this study. The
signal to noise, S/N, ratio. The S/N ratio quantifies the ability to multiple QCs of butt welding were demanded and assessed via
achieve the target value for a QC while minimize the variation consecutive measurements of 18 groups of machined samples.
from the target for a specific experimental condition or machine The multiple QCs demanded in this work included: (1) high ulti-
operating scenario. The applications of Taguchi method in either mate tension stress (UTS), (2) small heat-affected zone (HAZ),
machining or manufacturing field have been extremely success- (3) large welding depth to width ratio (D/W), and (4) low sur-
face roughness (SR) following laser welding. Furthermore, the
six essential machining parameters for Nd:YAG laser welding
∗ Corresponding author. in each group were optimized using the recommendation of
E-mail address: lkpan@ctc.edu.tw (L.K. Pan). Taguchi method. The six parameters as assigned herein included:

0924-0136/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2006.07.015
108 L.K. Pan et al. / Journal of Materials Processing Technology 182 (2007) 107–116

(1) shielding gas, (2) laser energy, (3) traveling speed of work- where xo implies the destination of the assigned value in the data
place during welding, (4) focusing position, (5) frequency of sequence.
the impulse laser, and (6) pulse shape of the primary laser beam. The weighted Grey relational coefficient is a weighting
Out of these six parameters, one has two possible levels and adjustment of the Grey relational coefficient and defines as
the others are three-level factors. In this study factor interac- n

∗
tions are not studied and therefore the L-18 orthogonal array xiωi
(k) = xi∗ (k)ωi , ωi = 1 (4)
was used to investigate these factors for significance using 18 i=1
trials. Grey analysis was then adopted to combine multiple QCs
into one numerical score, rank these scores, and determine the The weighting factor, wi is assigned equally as 0.25, 0.25,
optimal machine parameter settings. The revised laser welding 0.25, and 0.25 for ultimate tension stress, heat-affected zone,
setting then was confirmed and verified via ANOVA analysis for welding depth to width ratio, and surface roughness, respec-
each parameter adopted in this study. Further, some correlated tively in this work. The four requirements for laser butt welding
discussions on the basis of either the Taguchi method recom- discussed herein correspond to the two foregoing definitions.
mendation or Grey analysis were also explored to expand the Both the ultimate tension and the welding depth to width ratio
future applications of this revised methodology. in two thin plates following welding follow the first definition
and are “the-higher-the-better”; both the heated-affected zone
2. Grey analysis and the surface roughness along the welding path follow the
second definition and are “the-lower-the-better”.
The Grey analysis was first proposed many decades ago
but has been extensively applied only on the last decade. Grey 2.2. Grey relational grade
analysis has been broadly applied in evaluating or judging the
performance of a complex project with meager information. The Grey relational grade is defined as
However, data to be used in Grey analysis must be preprocessed Δmin + Δmax
γ(x0∗ , xiω
∗
) = Γi = (5)
into quantitative indices for normalizing raw data for another i
Δ∗0i + Δmax
analysis.
n
1
Δ∗0i = Δ0i (k) (6)
2.1. Grey relational coefficient n
k=1

The most crucial point for applying the Grey theory to prac- 0 < γ(x0∗ , xiω
∗
i
)≤1 (7)
tical case is how to quantify and grade the performances for
Δ0i (k) is the deviation sequence of the reference sequence x0∗ (k)
various ways of butt welding titanium parts. In doing so, the
and the comparability sequence xi∗ωi (k), i.e.:
Grey relational coefficient for each welding method is acquired
and ranked to determine optimal machine settings. Preprocess- Δ0i (k) = |x0∗ (k) − xiω
∗
i
(k)| (8)
ing raw data is a process of converting an original sequence into
a decimal sequence between 0.00 and 1.00 for comparison. Defi- Δmax = max max|x0∗ (k) − xjω
∗
j
(k)| (9)
nitions of data preprocessing depend on the characteristics of the
Δmin = min min|x0∗ (k) − xjω
∗
j
(k)| (10)
original data sequence [10–12]. If the expected data sequence is
of the form “the-higher-the-better”, then the original sequence The grade γ(x0∗ , xiω
∗ ), Γ implies the correlation between
i i
can be normalized as the reference sequence and the specific sequence xi to be
xio (k) − min xio (k) compared to. The evaluated grade fluctuates from 0.00 to
xi∗ (k) = (1) 1.00 and equals one if these two sequences are identically
max xio (k) − min xio (k)
where xio (k) is the original sequence, xi∗ (k) the sequence after
the data preprocessing, max xio (k) the largest value of xio (k), and
min xio (k) implies the smallest value of xio (k).
When the form “the-smaller-the-better” becomes the
expected value of the data sequence, the original sequence can
be normalized as
max xio (k) − xio (k)
xi∗ (k) = (2)
max xio (k) − min xio (k)
However, when the demanded value becomes the expected
value in the analysis, the original sequence can be normalized
as following to achieve the destination as close as possible:
|xio (k) − xo |
xi∗ (k) = 1 − (3) Fig. 1. The Nd:YAG laser welding system. A CNC machine system moves the
max{max xi (k) − xo ; xo − min xio (k)}
o
work along the welding path.
 L.K. Pan et al. / Journal of Materials Processing Technology 182 (2007) 107–116 109

coincident. Since the grade, Γ i , is evaluated through integrated 3. Experimental set-up

consideration of four requirements, the highest grade among
all 18 groups exactly stands for the optimized combination for The multiple QCs of the welded titanium plates are higher ultimate stress,
lower heat-affected zone, larger weld depth to width ratio, and lower surface
laser welding in this work. Furthermore, the grade also shows
roughness along the welded path. The next section summarizes all considerations
the contribution of the comparable sequence of demanded to regarding the experimental set-up.
the reference sequence. A higher grade also implies greater
dominance of the specific sequence over the other sequences.
3.1. Nd:YAG laser welding
Hence, Grey analysis involves a series of calculations that
include measurements of differences between original data A 500 W Nd:YAG laser welding device is used as a machining tool in this
sequences. study. Unlike the 10.6 ␮m wavelength of the CO2 laser beam, the wave-length

Fig. 2. Close look of Nd:YAG laser welding system: (a) laser generating device; (b) control panel; (c) laser ejecting nozzle; (d) monitoring and adjusting stand; (e)
whole integrated system.
110 L.K. Pan et al. / Journal of Materials Processing Technology 182 (2007) 107–116

of the Nd:YAG laser beam is only 1.064 ␮m. The narrow impulse laser beam Table 1
can deposit energy of beam on a tiny zone, intensely affecting this specific area The chemical composition of each element in titanium alloy, Ti–6Al–4V
without relying on the gigantic power of the laser beam itself [13]. Furthermore,
Element wt.%
the low beam energy prevents the surface of the welding target from overheating
and ensures that the heat-affected zone is small. Al 6.0
A solid Nd:YAG laser beam is typically one of two types – continuous and V 4.0
impulse – defined according to the means of transferring energy. The impulse O 0.2
laser beam used herein, outperforms the continuous one for welding small com- H 0.015
ponents, because of its lower required initial power and precise positioning. Ni 0.05
Fig. 1 depicts the configuration of the Nd:YAG laser beam welding device. As Fe 0.4
demonstrated herein, the laser is generated, guided, and focused through an C 0.08
optical lens to concentrate within 0.60 ± 0.02 mm and then projected onto the
workpiece through the fiber optics with focal length 60 mm. A computer numer-
ical control (CNC) machine is used to hold the workpiece, position it precisely, Table 2
and move it along the welding path (Fig. 2). Some essential mechanical properties of titanium alloy, Ti–6Al–4V
Property Value
3.2. Titanium alloy
Tension stress, ultimate (MPa) 1060 ± 21
The attractive features of titanium alloys in terms of strength-to-weight ratio, Tension stress, yield (MPa) 1045 ± 45
stiffness, and anti-corrosion, anti-fatigue properties, as compared to other alloys, Fatigue strength (MPa) 140
make such alloy useful for fabricating precision components in heavy and light Shear strength (MPa) 550
industries. Furthermore, the excellent biologically compatiability of titanium Elongation (%) 10
alloy makes it the preferred alloy for fabricating artificial prosthetics and limbs. Hardness (HB) 326
Nevertheless, either the low thermal conductivity or the small elastic modulus Electrical resistivity ( cm) 0.000178
of titanium alloy has still limited the machining precision via conventional fac- Thermal conductivity (W/m K) 6.7
tricating process, since heat deposited onto the contact zone of the workpiece Melting point (◦ C) 1660
caused overheating in topical areas due to slow heat dissipation. The titanium
alloy, Ti–6Al–4V, used in this study, has been extensively applied in advanced
engineering because of its higher strength-to-weight ratio than other alloys with
designing experiments. The numbers indicate the various experimental layouts
similar composition. Tables 1 and 2 show the composition of Ti–6Al–4V and
or levels of the different factors. The six welding parameters as considered in this
some essential mechanical properties of Ti–6Al–4V.
work, shielding gas, laser energy, traveling speed of workpiece, welding focus
position, laser pulse frequency, and laser pulse shape were concerned with either
3.3. Ultimate tension stress, heat-affected zone, depth to width two or three different levels of each. Thus, a total of 486 (2 × 3 × 3 × 3 × 3 × 3)
ratio and surface roughness along the welding path different combinations are possible for setting up the welding process and
only 18 of these process designs needs to be tested using the Taguchi method
Table 3 lists the 18 welding parameter settings used to make the test pieces. approach.
In this study one two-level factor and five three-level factors are investigated. Each of the 18 trials or process designs is replicated three times and a total of
Only the main effects are of interest and factor interactions are not studied. 54 butt-welded test pieces are tested. Each butt-welded test pieces is made using
Therefore the degrees of freedom required for the study is eleven and Taguchi’s two 50 mm × 10 mm × 10 mm plate. Fig. 3 shows the size of the butt-welded
L-18 orthogonal array is used to define the 18 trial conditions [14,15]. Table 4 parts and the machined test specimen. The test specimen is used to determine
shows the arrangement of the samples into 18 groups as per Taguchi method of the ultimate tensile stress after the welding operation [1,2].

Table 3
Eighteen groups of samples were evaluated in this work
Group no. Shielding gas Laser energy (W) Traveling speed (mm/min) Laser focus (mm) Pulse frequency (Hz) Pulse shape

1 Ar 200 650 0.0 80 I

2 Ar 200 700 −0.3 120 II
3 Ar 200 750 −0.6 160 III
4 Ar 260 650 0.0 120 II
5 Ar 260 700 −0.3 160 III
6 Ar 260 750 −0.6 80 I
7 Ar 320 650 −0.3 80 III
8 Ar 320 700 −0.6 120 I
9 Ar 320 750 0.0 160 II
10 He 200 650 −0.6 160 II
11 He 200 700 0.0 80 III
12 He 200 750 −0.3 120 I
13 He 260 650 −0.3 160 I
14 He 260 700 −0.6 80 II
15 He 260 750 0.0 120 III
16 He 320 650 −0.6 120 III
17 He 320 700 0.0 160 I
18 He 320 750 −0.3 80 II

The notations in the last column; I, II and III imply the various shape of laser pulse as depicted in Fig. 4. Units for the factors are shown beneath the respective factor
in parenthesis.
 L.K. Pan et al. / Journal of Materials Processing Technology 182 (2007) 107–116 111

Table 4
Standard orthogonal array L-18 of 18 different groups following Taguchi method suggestion
Sampling no. Factor A Factor B Factor C Factor D Factor E Factor F

1 1 1 1 1 1 1
2 1 1 2 2 2 2
3 1 1 3 3 3 3
4 1 2 1 1 2 2
5 1 2 2 2 3 3
6 1 2 3 3 1 1
7 1 3 1 2 1 3
8 1 3 2 3 2 1
9 1 3 3 1 3 2
10 2 1 1 3 3 2
11 2 1 2 1 1 3
12 2 1 3 2 2 1
13 2 2 1 2 3 1
14 2 2 2 3 1 2
15 2 2 3 1 2 3
16 2 3 1 3 2 3
17 2 3 2 1 3 1
18 2 3 3 2 1 2

The numbers in each column indicate the experimental levels considered for the specific factors A–F.

The six welding parameters as depicted in Table 3 are defined as 4. Results and interpretation of data

(A) Shielding gas. Implying the covering gas during welding process [16]. Table 5 lists the mean ultimate tension stress, heat-affected
(B) Energy. The specific energy of the applied laser welding device [13]. zone, depth to width ratio, and surface roughness along the weld-
(C) Traveling speed of workpiece. The speed at which the workpiece is moved
along the welding path.
ing path for 18 groups of measured samples and Table 6 shows
(D) Laser focus position. The applied laser beam can be focused exactly onto the Grey relational grade evaluated for each group.
the surface or beneath the surface of the workpiece.
(E) Pulse frequency. The frequency of the laser beam; in addition; the pulse
time and delay time of assigned laser frequency in this study are categorized
into three different combinations: (1) 2.5 and 12.5 ms for 80 Hz; (2) 2.2 and
8.33 ms for 120 Hz; (3) 1.7 and 6.25 ms for 160 Hz.
(F) Pulse shape. The pulse shape of the laser beam applied here; as depicted
in Fig. 4. The three differently shaped laser pulses can be manipulated
precisely by control the panel of this specific device.

Fig. 3. The top drawing shows the two plates that are butt-welded together to
make the part shown in the middle. The bottom drawing shows the machined
test specimen used in the tensile stress test. All dimensions are in mm. Fig. 4. Three differently shaped pulses applied in this work.
112 L.K. Pan et al. / Journal of Materials Processing Technology 182 (2007) 107–116

Table 5
The average ultimate tension stress, heat-affected zone, depth to width ratio, and surface roughness along the welding path for 18 groups, respectively as evaluated
in this work
Group no. Ultimate tension stress (MPa) Heat-affected zone (mm) Depth to width ratio (mm) Surface roughness, Rmax (␮m)

1 623.0 27.3 0.69 27.3

2 548.5 26.9 0.66 26.9
3 251.5 24.6 0.46 24.6
4 597.0 27.5 0.75 27.5
5 956.0 31.1 0.72 31.1
6 739.5 88.0 0.75 88.0
7 906.5 73.9 0.78 73.9
8 868.5 54.4 0.76 54.4
9 768.0 40.7 0.62 40.7
10 197.5 21.2 0.45 21.2
11 949.5 34.4 0.76 34.4
12 207.5 15.5 0.45 15.5
13 528.5 30.2 0.73 30.2
14 973.0 126.6 0.77 126.6
15 942.0 27.5 0.72 27.5
16 554.0 48.5 0.58 48.5
17 954.5 28.5 0.68 28.5
18 922.0 275.4 0.66 275.4

As shown, group 11 listed in Table 6 has the highest grade of ally, the ideal ultimate tension stress, heat-affected zone, depth
0.993. Groups 15 and 17 have the second and third highest grade, to width ratio and surface roughness were obtained in groups
0.986 and 0.981, respectively. The machining parameters for 14, 12, 7 and 12 with values 949.5 MPa, 0.12 mm, 0.78 ␮m
laser welding in group 11 are: (1) shielding gas, helium; (2) laser Rmax and 15.5 ␮m Rmax , respectively. An extreme performance
energy, 200 W; (3) traveling speed of workpiece, 700 mm/min; was also implied in group 12, with two out of four QCs, heat-
(4) welding focus position, 0 mm, exactly onto the surface of affected zone and surface roughness having almost the highest
the workpiece; (5) laser pulse frequency, 80 Hz; (6) laser pulse grade while the other two, tension stress and depth to width
shape, shape III (cf. Section 3.3, Fig. 4). To judge further the ratio, were barely graded. Clearly, it is easier to meet a single
precise performance of group 11, the ultimate tension stress was QC requirement in laser welding than to meet multiple require-
949.5 MPa, the fourth highest value; the heat-affected zone was ments. Furthermore, Taguchi method can rapidly and reliably
0.14 mm, the fourth lowest value; the depth to width ratio was suggest a layout for laser welding with numerical verification on
0.76, the third largest value of all 18 groups. However, the surface machined performance which overcomes the conflict between
roughness of welding surface along the path reached 34.4 ␮m different requirements and still achieves the user specified
Rmax , the 10th smallest value among all 18 groups. Addition- goals.

Table 6
The evaluated Grey relational coefficients and grades for 18 groups
Group no. Grey relational coefficient Grey relational coefficient after weighted Grey relational grade

UTS HAZ D/W SR UTS (wi = 0.25) HAZ (wi = 0.25) D/W (wi = 0.25) SR (wi = 0.25) Γi Rank

1 0.55 0.96 0.71 0.95 0.14 0.24 0.18 0.24 0.972 6

2 0.45 0.85 0.63 0.96 0.11 0.21 0.16 0.24 0.962 12
3 0.07 0.91 0.04 0.96 0.02 0.23 0.01 0.24 0.934 16
4 0.52 0.81 0.89 0.95 0.13 0.20 0.22 0.24 0.972 6
5 0.98 0.71 0.80 0.94 0.24 0.18 0.20 0.23 0.980 4
6 0.70 0.84 0.90 0.72 0.17 0.21 0.22 0.18 0.971 8
7 0.91 0.00 1.00 0.78 0.23 0.00 0.25 0.19 0.956 13
8 0.87 0.30 0.92 0.85 0.21 0.08 0.23 0.21 0.964 10
9 0.74 0.78 0.51 0.90 0.18 0.19 0.13 0.22 0.964 10
10 0.00 0.97 0.01 0.98 0.00 0.24 0.00 0.24 0.933 18
11 0.97 0.96 0.93 0.93 0.24 0.24 0.23 0.23 0.993 1
12 0.01 1.00 0.00 1.00 0.00 0.25 0.00 0.25 0.934 16
13 0.43 0.87 0.84 0.94 0.11 0.22 0.21 0.23 0.969 9
14 1.00 0.88 0.95 0.57 0.25 0.22 0.23 0.14 0.979 5
15 0.96 0.89 0.81 0.95 0.24 0.22 0.20 0.24 0.986 2
16 0.46 0.70 0.38 0.87 0.11 0.17 0.09 0.22 0.947 15
17 0.98 0.82 0.70 0.95 0.24 0.20 0.17 0.23 0.981 3
18 0.93 0.99 0.63 0.00 0.23 0.24 0.15 0.00 0.951 14
 L.K. Pan et al. / Journal of Materials Processing Technology 182 (2007) 107–116 113

Table 7
The calculated η(S/N) and corresponding Grey relational grade for 18 groups
Group no. η(S/N) Grey relational grade

1 −0.247 0.972
2 −0.336 0.962
3 −0.593 0.934
4 −0.247 0.972
5 −0.175 0.980
6 −0.256 0.971
7 −0.391 0.956
8 −0.318 0.964
9 −0.318 0.964
10 −0.602 0.933
11 −0.061 0.993
12 −0.593 0.934
13 −0.274 0.969
14 −0.184 0.979
15 −0.122 0.986
16 −0.473 0.947
17 −0.167 0.981
Fig. 5. η vs. various welding parameters. The highest η values for various param-
18 −0.436 0.951
eters implies the optimized setting of that specific parameter in this work.

5. Verification and discussion (W) 200, 260, and 320 in this work (cf. Table 3). Fig. 5 plots the
rearranged averages η versus various welding parameters.
5.1. Incorporating with Taguchi method Fig. 5 shows the optimal value of each welding parameter as
(A) shielding gas, argon; (B) laser energy, 260 W; (C) traveling
Taguchi method is used to broaden the application of Grey speed of workpiece, 700 mm/min; (D) welding focus position,
relational analysis herein (cf. Table 3) [17,18]. The calcu- 0 mm, exactly onto the surface of workpiece; (E) laser pulse
lated Grey relational grade was taken as the inspected value frequency, 80 Hz; (F) laser pulse shape, shape III. Since “the-
in Taguchi method (cf. Table 6). The purpose of using Taguchi higher-the-better” η is exactly the required value according to
method calculation is then to obtain the highest Grey relational Taguchi method calculation (cf. Eq. (11)).
grade.
Thus, the Grey relational grade calculated for each assigned 5.2. Verifying integrated performances of laser welding
group was dealt as an inspected value in Taguchi method defi-
nition and reorganized as follows [19]: The optimization of an Nd:YAG laser in butt welding a Ti
 n  alloy thin plate can, therefore, be examined according to Taguchi
1  −2 method recommendation. Table 8 shows the original and opti-
η = −10 log yi (11)
n mized setting of the laser beam for welding the same Ti alloy.
i=1
In addition, the original setting in this work was determined
where η represents the inspection index, defined as the signal to from similar facility in other factory. As shown, the ultimate
noise ratio (S/N; unit, dB). A larger η was considered preferable, tension stress changes from 922 ± 53 to 965.5 ± 46.5 MPa; the
since the Grey relational grade of Ti alloy thin plate laser welding heat-affected zone decreases from 0.27 to 0.22 mm, the welding
was expected to be of the form “the-higher-the-better”. A higher depth to width ratio along the path holds in 0.76, whereas the sur-
η(S/N) value also implies a superior integrated performance for
laser butt welding, since the major signal dominates the noise. Table 8
yi was the Grey relational grade in each group, i, and n was the Comparison between original setting and optimal recommendation of laser beam
number of repeated trial in each group. The number of repeated onto Ti alloy thin plate butt welding
trial was one, since only one relational grade was acquired in Welding parameter Original setting Optimal
each group for this particular calculation of S/N. recommendations
Table 7 lists the η(S/N) with the corresponding Grey rela-
A. Shielding gas He Ar
tional grade for every group. As clearly depicted, the highest B. Laser energy (W) 260 260
grade for group 11, 0.993, was converted to the largest η value, C. Traveling speed (mm/min) 700 700
−0.061, as well. The η-value for each group can be further rear- D. Laser focus position (mm) −0.3 0.0
ranged and averaged on the basis of various factors (cf. Table 4). E. Pulse frequency (Hz) 120 80
F. Pulse shape Type II Type III
For instance, averages from groups 1 to 8 and from groups 9 to 18
stand for the contributions from various shielding gases: Ar and Ultimate tension stress (MPa) 922 ± 53 965.5 ± 46.5
He; further, averages for groups 1–3, 10, 11 and 12, groups 4–6, Heat-affected zone (mm) 0.27 0.22
Depth to width ratio 0.76 0.76
13, 14 and 15, and groups 7–9, 16, 17 and 18 correspond to the Surface roughness, Rmax (␮m) 14.15 36.9
contributions from various factors B1, B2, and B3; laser energy
114 L.K. Pan et al. / Journal of Materials Processing Technology 182 (2007) 107–116

Fig. 6. The workpiece was welded under different laser energy, 160 and 260 W,
respectively. The other machining parameters for this specific case arc traveling
speed 700 mm/min, pulse shape III, pulse frequency 120 Hz, laser focus position
Fig. 7. The workpiece was welded under different laser pulse frequency, 80 and
0.0 mm and under shielding gas Ar.
120 Hz, respectively. The other machining parameters for this specific case are
laser energy 260 W, traveling speed 700 mm/min, pulse shape III, laser focus
face roughness on the workpiece degrades from Rmax of 14.15 to position −0.3 mm and under shielding gas Ar.
36.9 ␮m. Thus, the optimization specified herein is successful.
The apparent changes in three out of four machining require- arrangement of various levels in each factor preserves the same
ments are easily quantified although the change in machining frequency among 18 groups. Consequently, the evaluated data
setting is comparatively negligible. In addition, Figs. 6–8 depict can also be rearranged to indicate cross interaction between fac-
the various combinations of machining parameters for laser tors (cf. Table 4), since the assessed Grey relational grade can
welding applied in this work. The photos were shot in the pre- be easily rearranged to imply the cross interactions between any
liminary state of this study, thus, some parameters were not well two factors. Fig. 9 illustrates three varieties of cross interaction
optimized yet. However, as clearly shown, the inappropriate set- between factors; (A), (B), and (C) show laser energy versus laser
ting of the parameters can still be reflected onto the welded focus, laser frequency versus laser focus, and laser frequency
performance and degraded the tension stress. versus pulse shape, respectively. As clearly shown in part (A) of
The inappropriate setting of the parameters can be exactly Fig. 9, three lines which were barely intersected were interpreted
reflected onto the welded performance. as indicating the existence of no cross interaction between these
two factors, while laser frequency and laser focus had medium
5.3. Cross interactions between machining parameters cross interaction with each other (cf. Fig. 9B). As illustrated here,
the Grey grade decreased with the laser frequency when the laser
The Taguchi method can not only provide the solid sugges- focus was held at 0.0 mm moreover, the grade trend decreased
tion in recommending the dominant parameter for single QC, rapidly and intersected with another grade trend when the laser
but also offer an effective algorithm for clarifying the specific focus was held at −0.6 mm. The cross interaction between the
cross interaction between parameters. The unique orthogonal two parameters thus might mislead the regular Taguchi method
 L.K. Pan et al. / Journal of Materials Processing Technology 182 (2007) 107–116 115

Fig. 9. The cross interaction between factors for various cases: (A) laser energy
versus laser focus; (B) laser frequency versus laser focus; (C) laser frequency
vs. pulse shape.

inspected on the basis of various laser focus. Since the laser focus
at 0.0 mm achieved the highest Grey grade, the optimum laser
focus was, therefore, determined at 0.0 mm. The follow Tip veri-
fication of optimization was also confirmed this suggestion. Part
(C) had the most complicate cross interaction among the three
cases. The Grey grade fluctuated significantly with changes of
laser frequency under different pulse shapes. Thus, the primary
suggestion of optimized setting for pulse shape was adjusted to
the III type, since pulse shape III maintained the highest grade
among all the evaluated results. Additionally, maintaining the
laser pulse shape at III type could obtain a quasi-linear corre-
Fig. 8. The workpiece was welded under different laser focus: −0.3, 0.0, lation of Grey grade to laser frequency. This linear correlation
+0.3 mm, respectively. The other machining parameters for this specific case
between two machining parameters could minimize the conflict
are laser energy 260 W, traveling speed 700 mm/min, pulse frequency 80 Hz,
laser pulse shape III and under shielding gas He. in optimizing other factors.

recommendation and make it necessary to reconsider from a 6. Conclusion

fresh perspective of optimizing in machining parameter set-
ting. To further optimize the laser focus, the laser frequency Optimization of multiple QCs of Nd:YAG laser welded tita-
as adopted in the revised setting herein was, first, determined as nium alloy parts using Taguchi method-based Grey analysis
80 Hz, after which the performance of Grey grades was carefully was studied in this paper. The calculation of the Grey rela-
116 L.K. Pan et al. / Journal of Materials Processing Technology 182 (2007) 107–116

tional grade helped to quantify the integrated performance of [3] H.C. Su, Y.S. Liao, J.T. Huang, A study on the machining-parameters opti-
multiple QCs required in laser welding. The complex interac- mization of wire electrical discharge machining, J. Mater. Process. Technol.
tions in laser welding involved the shielding gas, laser energy, 71 (1997) 487–493.
[4] Y.L. Su, et al., Design and performance analysis of ticn-coated cemented
traveling speed of the workpiece, laser focus, pulse frequency carbide milling cutters, J. Mater. Process. Technol. 87 (1999) 82–89.
and pulse shape. The optimization of laser welding, by calcu- [5] Y.L. Su, et al., Design of a titanium nitride-coated we milling cutter, J.
lating the Grey relational grade and using the recommenda- Mater. Process. Technol. 86 (233) (1999) 236.
tion of Taguchi experimental design for determining welding [6] D.C. Ko, et al., Methodology of perform design considering workability
parameters was successful. Accordingly, the optimal combi- in metal forming by the artificial neural network and Taguchi method, J.
Mater. Process. Technol. 80 (1998).
nation of welding parameters for laser welding included. Ar [7] Y.C. Hsiao, K.C. Ho, L.K. Pan, C.C. Wang, Optimization of Nd:YAG laser
as the shielding gas, with a 260 W laser, workpiece speed of welding onto magnesium alloy via Taguchi analysis, Opt. Laser Technol.
700 mm/min, laser focus distance of 0.0 mm, pulse frequency 37 (1) (2005) 33–42.
of 80 Hz, and type III pulse shape. The ultimate tension stress [8] J. Deng, Control problems of grey systems, Syst. Control Lett. 5 (1982)
changes from 922 ± 53 to 965.5 ± 46.5 MPa; the heat-affected 288–294.
[9] J. Deng, Introduction to grey theory, J. Grey Syst. 1 (1) (1989) 1–24.
zone decreases from 0.27 to 0.22 mm, the welding depth to [10] J.H. Wu, Z. Wang, L. Zhu, Grey relational analysis of correlation of errors
width ratio along the path maintains a level of 0.76, and the sur- in measurement, J. Grey Syst. 8 (1) (1996) 73–78.
face roughness on the workpiece reduces from Rmax of 14.15 to [11] L. Ma, J. Du, F. Zhu, M. Yi, The grey relational analysis of the dielectric
36.9 ␮m. constant and others, J. Grey Syst. 8 (3) (1996) 287–290.
[12] J. Deng, X. Tan, Y. Yang, Grey relational analysis factors in hypertensive
with cardiac insufficiency, J. Grey Syst. 10 (1) (1998) 75–80.
Acknowledgment [13] M.S. William, Laser Material Process, New York, 1991, pp. 113–116.
[14] G. Taguchi, Introduction to Quality Engineering, Asian Productivity Orga-
The authors would like to thank the National Science Council nization, Tokyo, 1990.
of the Republic of China for financially supporting this research [15] D.S. Chou, L.K. Pan, B.D. Chang, Optimization for solidification of low-
under contract no. NSC 89-2212-E-014-022. level-radioactive resin using taguchi analysis, Waste Manage. 21 (2001)
762–772.
[16] Annual Book of ASTM Standards. Technical Report V02-02, ASTM, 1992.
References [17] C.P. Fung, Manufacturing process optimization for wear property of fiber-
reinforced polybutylene terephthalate composites with grey relational anal-
[1] W.H. Yang, Y.S. Tarng, Optimization of the weld bead geometry in gas ysis, Wear 306 (2003) 254–298.
tungsten arc welding by the Taguchi method, Int. J. Adv. Manuf. Technol. [18] C.L. Lin, J.L. Lin, The use of orthogonal array with grey relational analysis
14 (1998) 549–554. to optimize the electrical charge machining process with multiple perfor-
[2] Y.S. Tarng, W.H. Yang, Design optimization of cutting parameters for turn- mance characteristics, Int. J. Mach. Tools Manuf. 42 (2002) 237–244.
ing operations based on the Taguchi method, J. Mater. Process. Technol. [19] M.S. Phadke, Quality Engineering using Robust Design, Prentice Hall,
84 (1998) 122–129. Englewood Cliffs, New Jersey, 1989.