ARTICLE IN PRESS

Optics and Lasers in Engineering 46 (2008) 124132 www.elsevier.com/locate/optlaseng

Multi-objective optimization of Nd:YAG laser cutting of nickel-based superalloy sheet using orthogonal array with principal component analysis
Avanish Kumar Dubey, Vinod Yadava
Department of Mechanical Engineering, Motilal Nehru National Institute of Technology, Allahabad 211004, U.P., India Received 10 July 2007; received in revised form 17 August 2007; accepted 23 August 2007 Available online 24 October 2007

Abstract Laser cutting of nickel-based superalloy sheets, having wide applications in aircraft and rocket industries, is important from the quality of cut point of view. Keeping this in view, a hybrid approach of Taguchi method (TM) and principal component analysis (PCA) has been applied for multi-objective optimization (MOO) of pulsed Nd:YAG laser beam cutting (LBC) of nickel-based superalloy (SUPERNI 718) sheet to achieve better cut qualities within existing resources. The three-quality characteristics kerf width, kerf deviation (along the length of cut), and kerf taper have been considered for simultaneous optimization. The input parameters considered are assist gas pressure, pulse width, pulse frequency, and cutting speed. Initially, single-objective optimization has been performed using TM and then the signal-to-noise (S/N) ratios obtained from TM have been further used in PCA for multi-objective optimization. The results of MOO include the prediction of optimum input parameter level and their relative signicance on multiple quality characteristics (MQC). The responses at predicted optimum parameter level are in good agreement with the results of conrmation experiments conducted for verication tests. r 2007 Elsevier Ltd. All rights reserved.
Keywords: Laser cutting; Taguchi Method; Principal component analysis; Nd:YAG laser; Superalloy sheet; Multi-objective optimization

1. Introduction Advanced engineering materials have high demand in modern manufacturing industries due to their improved technological characteristics such as high mechanical strength and hardness, corrosion resistance, heat resistance and wear resistance. The most widely used advanced engineering materials are ceramics, composites and superalloys. Superalloys are generally made for elevated temperature applications where the resistance to deformation and stability are prime requirements. Nickel-based superalloys have wide applications in food processing equipments, caustic handling equipments and piping, chemical shipping drums, transducers, electrical and electronic parts, rocket motor cases, and aerospace and missile components [1]. These typical applications require
Corresponding author.

E-mail address: avanishdubey@yahoo.com (A.K. Dubey). 0143-8166/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.optlaseng.2007.08.011

stringent design requirements and close tolerances in manufactured products. But improved mechanical properties restrict the conventional manufacturing methods to process these materials. The non-conventional manufacturing processes are well suited for shaping advanced difcult-to-cut materials. Laser beam cutting (LBC) is one of the non-conventional cutting processes, most widely used for generating complex shapes and geometries in almost all engineering materials such as metals, non-metals, ceramics, composites, and superalloys [24]. LBC is a thermal cutting process and it is executed by moving a focused laser beam along the surface of the workpiece with constant distance; thereby generating a narrow cut kerf. The kerf fully penetrates the material along the desired contour. During the process, a part of the laser radiation is absorbed at the end of the kerf called the cutting front. The absorbed energy heats and transforms the prospective kerf volume into a molten, vapourized or chemically changed state which can be easily removed by a

ARTICLE IN PRESS
A.K. Dubey, V. Yadava / Optics and Lasers in Engineering 46 (2008) 124132 125

Excitation Source (Flash Lamp)

Bending Mirror

Chiller Unit

Lasing Medium (Nd:YAG Rod)

Rear mirror (Reflective)

Front Mirror (Partial reflective) Focusing Lens

Supply of gas jet to nozzle Workpiece CNC Controller Motor

Fig. 1. Schematic of Nd:YAG laser beam cutting system.

suitable coaxial gas jet that accelerates the transformed material and ejects it out of the kerf (Fig. 1) [24]. The solid state Nd:YAG and gaseous CO2 lasers are most widely used industrial lasers due to their high powers [24]. Nd:YAG laser cutting becomes an excellent machining process because of high laser beam intensity, low mean beam power, good focusing characteristics due to very small pulse duration and narrow heat affected zone (HAZ). There has been growing interest in recent years in the use of pulsed Nd:YAG lasers for precision cutting of thin sheet metals and for applications that demand narrow kerf widths and intricate cut proles [5]. Due to its shorter wavelength (1.06 mm) in comparison to CO2 (10.6 mm), it is reected to a lesser extent by metallic surfaces and this high absorptivity of the Nd:YAG laser enables cutting of even highly reective materials with relatively less power. Materials having high thermal conductivity give poor results when cut by CO2 laser. Due to shorter thermal interaction time Nd:YAG laser gives better results. In pulsed mode, high incident peak power output facilitates thick material cutting. Furthermore, enhanced transmission through plasma, wider choice of optical materials and exibility in handling with the advent of bre optic beam delivery are also some of the interesting characteristics of the Nd:YAG laser [6]. The cutting quality of pulsed Nd:YAG laser depends on many controllable factors (operating parameters) such as pulse frequency, pulse energy, pulse width, cutting speed, assist gas type and its pressure. The proper control of these factors gives the desired cut quality [7]. Shanjin and Yang [8] have experimentally investigated and found the optimum range of parameters pulse frequency, cutting speed, type and pressure of assist gas for minimum HAZ during Nd:YAG laser cutting of titanium alloy sheet. Almeida et al. [9] applied the factorial designed experimental approach to determine the effects of pulse energy, overlapping rate and type of assist gas on the surface roughness and dross formation (edge irregularity) on the

Nd:YAG laser cut surface of pure titanium and titanium alloy sheet. Cosp et al. [10] have experimentally investigated and found the optimum parameters to avoid catastrophic break down during pulsed Nd:YAG laser cutting of 8 mm thick high vitried ceramic material (ne porcelain stoneware tile). Kuar et al. [11] have found the optimum value of lamp current, pulse frequency and cutting speed for minimum surface roughness during pulsed Nd:YAG laser cutting of silicon nitride ceramics. Lau and Lee [12] and Quintero et al. [13] have also experimentally studied the effect of parameters on MRR and HAZ, respectively, during cutting of ceramic composites. The convergingdiverging shape of laser beam prole, formation of recast layer, and dross adherence to bottom edge impose difculties in achieving the close geometrical tolerances during LBC. The three main quality characteristics of laser cut kerf that decides the kerf geometry are kerf taper (Kt), kerf deviation (Kd) along the length, and kerf width (Kw). The schematic of laser cut kerf is shown in Fig. 2. Researchers have experimentally studied the kerf qualities like Kt and/or Kw during LBC and concluded that these qualities can be optimized by proper control of process parameters [1417]. Some researchers have applied the Taguchi methodology (TM) to analyse and optimize the process parameters in order to achieve better cut qualities during LBC. Refs. [1820] show the application of TM to optimize the laser cutting parameters. Their ndings show a considerable improvement in cut qualities of different workpiece materials at optimum parameter levels. In TM-based experimental studies carried out so far, the researchers have optimized a single quality characteristic at a time. However, the performance of any manufacturing process depends on many quality characteristics and it is always desired to achieve an optimum parameter level that improves multiple quality characteristics (MQC) at the same time. For dealing with MQC, often pure engineering judgement is employed for process optimization which is

ARTICLE IN PRESS
126 A.K. Dubey, V. Yadava / Optics and Lasers in Engineering 46 (2008) 124132

Sheetmetal Top Kerf width

Thickness of sheet

performed by using L9 orthogonal array (OA) to obtain the results of single-objective optimization for Kw, Kd, and Kt, respectively. The signal-to-noise (S/N) ratios of each quality characteristics have been further used in PCA to optimize the MQC. Finally, the analysis of variance (ANOVA) was used to nd out the most inuential laser cutting parameter for multiple responses. 2. Methodology for multi-objective optimization For optimizing MQC simultaneously, the TM and PCA both are applied sequentially. This section gives a brief idea about these two approaches. 2.1. Taguchi method TM is a widely accepted method of design of experiments (DOE). It has proved to be an effective methodology for producing high-quality products at relatively low cost. This approach has been applied successfully in many US, Japan and European manufacturing rms, especially in automobile, electronics, food processing, and medical equipment industries [24,25]. The objective of Taguchi approach is to determine the optimum setting of process parameters or control factors, thereby making the process insensitive to the sources of variations due to uncontrollable or noise factors [24,25]. In this method, main process parameters or control factors which inuence on process results are taken as input parameters and the experiment is performed as per specically designed OA. The selection of appropriate OA is based on total degree of freedom (d.f.) which is computed as [24,25] d:f: number of levels 1 for each factor number of levels 1 number of levels 1 for each interaction 1. In general, S/N ratio (Z, dB) represents the quality characteristic for the observed data in the Taguchi DOE and mathematically it can be computed as [24,25]: Z 10 logMSD, (1)

Bottom Kerf width

Sheetmetal

Non-uniform Kerf (Kerf deviation)

Length of cut

Kerf width

Fig. 2. Schematic representation of laser cut kerf: (a) kerf taper and (b) kerf deviation.

very subjective and therefore always brings uncertainty to the decision making process. Antony [21] has suggested that normalized quality loss for each quality characteristic in TM can be used for optimizing the MQC simultaneously. But an uncertainty in the selection of weighting factor is always possible due to possible correlations amongst the MQC. This uncertainty can be overcome by using a hybrid approach of TM and principal component analysis (PCA). PCA has been proven to be useful for dealing with poor, incomplete and uncertain information. In PCA, a set of correlated variables (quality characteristics) are transformed into a set of uncorrelated principal components (PC). A weighting factor for a PC is determined based on its contribution percentage to total variance. Fung and Kang [22] have applied the TM and PCA sequentially to optimize the MQC at a time during injection-moulding process. The outcome of their study suggests that this hybrid approach is suitable for optimizing the MQC in manufacturing processes. Tzeng and Chen [23] have also tried the same hybrid approach in CNC turning process and found the considerable improvements in MQC. This hybrid approach has yet not been applied so far, in any of laser-based manufacturing processes for improving the MQC. In the present paper, three kerf qualities such as Kt, Kd, and Kw have been optimized simultaneously during pulsed Nd:YAG LBC of sheet with hybrid approach of TM and PCA. The control factors taken are: assist gas pressure (GP), pulse width (PW), pulse frequency (PF), and cutting speed (CS). The workpiece material is 0.7 mm thick nickelbased superalloy SUPERNI 718 (an equivalent grade to Inconel 718) which has wide applications in jet engines, pump bodies and parts. Initially, experiments have been

where MSD is the mean square deviation and commonly known as quality loss function. Depending on the experimental objective, the quality loss function can be of three types: lower-the-better (LB), higher-the-better (HB), and nominal-the-best (NB) type. In the case of kerf taper, kerf deviation, and kerf width, lower values of them are desirable. These quality loss values in the TM are called as LB-type and are computed as follows [24,25]: " # n 1X 2 MSD y , (2) n i1 i where yi is the observed data of quality characteristic at the ith trial and n is the number of repetitions at the same trial.

ARTICLE IN PRESS
A.K. Dubey, V. Yadava / Optics and Lasers in Engineering 46 (2008) 124132 127

The S/N ratio represents the desired part/undesired part and aim is always to maximize the S/N ratio whatever be the nature of quality characteristics. From the S/N ratio, the effective parameters having inuence on process results can be seen and the optimal sets of process parameters can be determined [24,25]. 2.2. Principal component analysis The PCA is a multivariate statistical method that selects a small number of components to account for the variance of original multi-response [26]. In PCA, the original dataset of MQC are converted into PC which is a linear combination of multi-responses obtained in a trial run. The procedure of PCA can be described as follows [22,23]: 1. The S/N ratios of each quality characteristics obtained from TM are normalized as x j i xi j xj , xj xj

where pi(k) is the kth PC corresponding to ith experimental run, vk(j) is jth element of kth eigenvector. 6. The total principal component index (TPCI) corresponding to ith experimental run (Pi) is computed as follows:
m X k1

Pi

pi k ek,

(7)

ek

eigk , m P eigk
k1

(8)

where eig(k) is the kth eigenvalue. 7. The TPCI for each experimental run is used to nd out the average factor effect at each level. The optimum parameter level that corresponds to the maximum TPCI is also predicted.

(3) 3. Pulsed Nd:YAG laser beam cutting process In pulsed Nd:YAG LBC process the cutting is done due to succession of beam spots along the length of cut. The geometrical accuracy of laser cut kerf mainly depends on kerf taper along the thickness of sheetmetal, kerf deviation along the length of cut and kerf width. The appropriate selection of different input parameters and their levels that affect these quality characteristics can improve quality of cut. 3.1. Selection of cutting parameters (4) The experiment was performed on a 200 W pulsed Nd:YAG laser beam machining system with CNC work table. The oxygen was used as an assist gas. The different process parameters (or control factors) taken are: assist GP, PW or pulse duration, PF, and CS. Focal length of the lens used is 50.0 mm. Nozzle diameter (1.0 mm), nozzle tip distance (1.0 mm), and sheet material thickness (0.7 mm) were kept constant throughout the experiments. SUPERNI 718 (a Ni-based superalloy) sheetmetal, supplied by Mishra Dhatu Nigam Limited, Hyderabad (India) was used in the experiments as workpiece material. The chemical composition of SUPERNI 718 is given in Table 1. Three levels of each control factor have been selected without considering the interaction effect. The numerical values of control factors at different levels are shown in Table 2. An exhaustive pilot experimentation has been conducted to
Table 1 Percentage chemical composition of SUPERNI 718 Fe Ni 52.37 Cr 19.0 Cu 0.15 Al 0.5 Mo 3.05 Mn 0.18 Ti 0.9 Si 0.18 C 0.04 Cd+Ta 5.13

where x j is the normalized S/N ratio for jth quality i characteristic in ith experimental run, xi(j) is the S/N ratio for jth quality characteristic in ith experimental run, x(j) is the minimum and x(j)+ is the maximum of S/N ratios for jth quality characteristic in all experimental runs. 2. The normalized multi-response array for m quality characteristics and n experimental runs can be represented by matrix X* as 2 x 1 1 x 2 . . . 1 x 2 . . . 2 ... ... ... ... xn 2 . . . ... ... ... ... ... x m 1 3

6 x 1 6 2 6 X 6 . . . 6 6 ... 4 x 1 n

x m 7 7 2 7 . . . 7. 7 ... 7 5 x m n

3. The correlation coefcient array (Rjl) of matrix X* is evaluated as follows: covx j; x l i i ; sx j sx l i i

Rjl

j 1; 2; . . . ; m;

l 1; 2; . . . ; m, (5)

where covx j; x l is the covariance of sequences i i x j and x l; sx l is the standard deviation of i i i sequence x l. i 4. The eigenvalues and eigenvectors of matrix Rjl are calculated. 5. The PC are computed as follows:
m X j1

pi k

x j vk j, i

(6)

18.5

ARTICLE IN PRESS
128 A.K. Dubey, V. Yadava / Optics and Lasers in Engineering 46 (2008) 124132 Table 3 Experimental layout using L9 orthogonal array Level 3 4.0 1.4 28 60 Trial no. 1 2 3 4 5 6 7 8 9 Factor level A 1 1 1 2 2 2 3 3 3 B 1 2 3 1 2 3 1 2 3 C 1 2 3 2 3 1 3 1 2 D 1 2 3 3 1 2 2 3 1 0.2340 0.4060 0.4160 0.3280 0.4380 0.4380 0.3900 0.3800 0.4640 0.0300 0.0500 0.1200 0.0300 0.0300 0.1200 0.0400 0.0700 0.0200 0.4092 0.8185 1.2278 0.8185 0.6139 1.0231 1.2278 1.2278 0.4092 Kerf width (mm) Kerf deviation (mm) Kerf taper (1) Table 2 Control factors and their levels used in the experiment Symbol A B C D Factors Oxygen pressure Pulse width Pulse frequency Cutting speed Unit kg/cm2 ms Hz mm/min Level 1 2.0 0.6 18 20 Level 2 3.0 1.0 23 40

decide the parameter range for through cutting of SUPERNI 718 sheet of 0.7 mm thickness. The initial setting of parameters was: oxygen pressure 2 kg/cm2, pulse width 0.6 ms, pulse frequency 18 Hz, and cutting speed 20 mm/min. 3.2. Laser cutting performance evaluation The quality characteristics analysed are kerf taper, kerf deviation (along the length), and kerf width. Two cuts each of 10 mm length were obtained in each experimental run and for each quality characteristic an average quality value of two cuts has been taken. The top and bottom kerf widths were measured using the tool makers microscope (Model RTM-900, RADICAL Instruments, India) at 10 magnication. The kerf taper (Kt) angle has been calculated using following formula: K t deg top kerf width bottom kerf width 180 . 2p workpiece thickness

Table 4 S/N ratios for kerf width, kerf deviation, and kerf taper Experiment no. S/N ratios (dB) Kerf width (Kw) Kerf deviation (Kd) 1 2 3 4 5 6 7 8 9 Overall mean Z 12.6157 7.8295 7.6181 9.6825 7.1705 7.1705 8.1787 8.4043 6.6696 8.3711 30.4576 26.0206 18.4164 30.4576 30.4576 18.4164 27.9588 23.0980 33.9794 26.5847 Kerf taper (Kt) 7.7613 1.7396 1.7826 1.7396 4.2380 0.1984 1.7826 1.7826 7.7613 1.9660

The kerf deviation (Kd) is the difference of maximum and minimum width of top kerf measured along the length of cut K d maximum top kerf width minimum top kerf width: The kerf width (Kw) is the width of top kerf. The kerf widths taken for top or bottom kerfs are the mathematical average of ve measurements of each cut taken at equal distances along the length of cut. 4. Determination of optimal cutting parameters In this section, the results of single-objective optimization using TM and multi-objective optimization (MOO) by hybrid approach of TM and PCA have been discussed. The verication results obtained on suggested optimum parameter levels have also been reported. 4.1. Orthogonal array experiments The total d.f. has been calculated without considering the interaction effect among different control factors. The total d.f. for four control factors each at three different levels is (31) 4+1 9. Therefore, an L9 OA has been selected for performing the experiments. The observed quality values for each quality characteristics kerf width

(mm), kerf deviation (mm), and kerf taper (deg) in different trials have been tabulated in Table 3. 4.2. Single-objective optimization using TM The S/N ratios (Z values) for Kw, Kd, and Kt have been calculated from experimental values of each quality characteristic (Table 3). The S/N ratio corresponding to each experimental run is given in Table 4. The factor effect of a parameter at any level is computed by taking the average of all S/N ratios at the same level. The effect of various factors at different levels for responses Kw, Kd, and Kt are shown in Table 5. Also, the graphical representations of factors effect at different levels are shown in Fig. 3. The optimum parameter level is the level corresponding to maximum average S/N ratio for a control factor. Therefore, the optimum parameter level for minimum value of Kw, Kd, and Kt are A1B1C1D1, A3B1C2D1, and A1B1C2D1, respectively. A better feel for the relative effect of the different parameters/factors can be obtained by the decomposition of the variance, which is commonly called ANOVA. It is a statistical technique to estimate quantitatively the relative

ARTICLE IN PRESS
A.K. Dubey, V. Yadava / Optics and Lasers in Engineering 46 (2008) 124132 Table 5 S/N response table for kerf width (Kw), kerf deviation (Kd), and kerf taper (Kt) Symbol Factor Mean S/N ratios (dB) Level 1 Kw A B C D Kd A B C D Kt A B C D
a

129

11
Gas pressure (kg/cm2) Pulse width (ms) Pulse frequency (Hz) Cutting speed (mm/min)

10

Level 2

Level 3

S/N Ratio (dB)

Oxygen pressure Pulse width Pulse frequency Cutting speed Oxygen pressure Pulse width Pulse frequency Cutting speed Oxygen pressure Pulse width Pulse frequency Cutting speed

9.3544a 10.1590a 9.3968a 8.8186a 24.9648 29.6246a 23.9907 31.6315a 2.5728 2.5728a 1.9268 6.5869a
a

8.0079 7.8014 8.0605 7.7262 26.4438 26.5254 30.1525a 24.1319 1.9264 1.3984 3.7468a 0.0804

7.7509 7.1528 7.6558 8.5683 28.3454a 23.6041 25.6109 23.9907 1.3987 1.9268 0.2243 0.6085

7 0 1 2 Factor Level 33
Gas pressure (kg/cm2) Pulse width (ms) Pulse frequency (Hz) Cutting speed (mm/min)

30
S/N Ratio (dB)

Optimum level.

27

S/N Ratio (dB)

contribution that each control factor or parameter makes on the overall measured response. The relative signicance of factors is often represented in terms of F-ratio or in percentage contribution. Greater the F-ratio more signicant will be the factor. The results of ANOVA for Kw, Kd, and Kt is given in Table 6. The results show that pulse width is the signicant factor for Kw, cutting speed is signicant for Kd, and pulse frequency and cutting speed both are signicant factors for Kt under more than 95% condence level. The contribution of factors in increasing order for Kw is CS, GP, PF, and PW, for Kd is GP, PW, PF, and CS, and for Kt is GP, PW, PF, and CS. 4.3. Multi-objective optimization using hybrid approach of TM and PCA The normalized S/N ratios for each quality characteristic Kw, Kd, and Kt against different experimental runs have been calculated using Eq. (3) and are shown in Table 7. The correlation coefcient array (Table 8) is obtained using Eq. (5). The eigenvalues and eigenvectors computed from correlation coefcient matrix are: 1.9280, 0.8298, 0.2423 and [0.3974, 0.6407, 0.6569], [0.9146, 0.3345, 0.2271], [0.0743, 0.6911, 0.7190], respectively. The three principal components PC1, PC2, PC3 and their integrated TPCI for each experimental run have been computed using Eqs. (6)(8) and tabulated in Table 9. The factor effect at each parameter levels have been computed by taking the average of all TPCI at that level and are given in Table 10. A graphical plot for factor effect at different levels is shown in Fig. 4. The optimum parameter level for MQC corresponds to maximum average TPCI for a control factor which is A1B1C2D1 i.e. GP at 2.0 kg/cm2, PW at 0.6 ms, PF at 23 Hz, and CS at 20 mm/min.

24

21 0 1 2 Factor Level 3 4

c
8 6 4 2 0 -2 0 1 2 Factor Level 3 4
Gas pressure (kg/cm2) Pulse width (ms) Pulse frequency (Hz) Cutting speed (mm/min)

Fig. 3. Effect of factor levels on S/N ratio for: (a) kerf width, (b) kerf deviation, and (c) kerf taper.

The ANOVA given in Table 11 shows the contribution of different factors in increasing order as: gas pressure, pulse frequency, pulse width, and cutting speed. The graphical representation of contribution of different control factors on TPCI of MQC is shown in Fig. 5. The improvement in predicted TPCI at the optimum level is found to be 0.10 as compared with initial parameter setting. The value of kerf width (mm), kerf deviation (mm), and kerf taper (deg) at this optimum level are found to be

ARTICLE IN PRESS
130 A.K. Dubey, V. Yadava / Optics and Lasers in Engineering 46 (2008) 124132 Table 6 Results of ANOVA for Kw, Kd, and Kt Symbol Kw A B C D Error Pooled error Total Kd A B C D Error Pooled error Total Kt A B C D Error Pooled error Total
a

Factors

Degrees of freedom

Sum of squares

Mean squares

Contribution (%)

Oxygen pressure Pulse width Pulse frequency Cutting speed

2 2 2 2 2 8

4.4506 15.0160 4.9808 1.9651a 1.9651 26.4125 17.2316a 54.3873 61.2200 114.6465 17.2316 247.4853 2.0747a 2.0758a 18.6193 96.5051 4.1505 119.2748

2.2253 7.5080 2.4904 0.9825 0.9825 8.6158 27.1936 30.6100 57.3232 8.6158 1.0373 1.0379 9.3096 48.2525 1.0376

2.2649 7.6417 2.5347

13.0758 56.8519 18.8577 7.4399

100 3.1562 3.5527 6.6532 6.9627 21.9760 24.7368 46.3246

Oxygen pressure Pulse width Pulse frequency Cutting speed

2 2 2 2 2 8

100 8.9746 46.5161 1.7394 1.7403 15.6104 80.9099

Oxygen pressure Pulse width Pulse frequency Cutting speed

2 2 2 2 4 8

100

Pooled factor. Table 9 Principal component scores and their integrated TPCI Experiment no. 1 2 3 4 5 6 7 8 9 Average P 0:4623. PC1 1.5500 0.6330 0.0634 0.9395 0.9435 0.1425 0.4937 0.3086 1.2976 PC2 0.4287 0.0688 0.1459 0.1208 0.3250 0.0393 0.0270 0.1662 0.5616 PC3 0.1100 0.0869 0.0119 0.3070 0.0874 0.1131 0.4426 0.2296 0.0279 TPCI 1.1058 0.3948 0.0820 0.6620 0.5235 0.0933 0.3605 0.2628 0.6763

Table 7 Normalized S/N ratios for kerf width, kerf deviation, and kerf taper Experiment no. Kerf width (Kw) 1 2 3 4 5 6 7 8 9 1.0000 0.1951 0.1595 0.5067 0.0842 0.0842 0.2538 0.2917 0.0000 Kerf deviation (Kd) Kerf taper (Kt) 0.7737 0.4886 0.0000 0.7737 0.7737 0.0000 0.6131 0.3008 1.0000 1.0000 0.3690 0.0000 0.3690 0.6308 0.1660 0.0000 0.0000 1.0000

Table 8 Correlation coefcients among the targeted quality characteristics Correlation coefcient Kw Kd Kt Kerf width (Kw) 1.0000 0.2495 0.3181 Kerf deviation (Kd) 0.2495 1.0000 0.7541 Kerf taper (Kt) 0.3181 0.7541 1.0000

Table 10 Response table for TPCI Symbol Factors A B C D

a

Level 1 Level 2 Level 3 MaxMin Rank 0.53a 0.71a 0.49 0.77a 0.43 0.39 0.58a 0.28 0.33 0.28 0.32 0.34 0.20 0.43 0.26 0.49 4 2 3 1

Oxygen pressure Pulse width Pulse frequency Cutting speed Optimum level.

0.246, 0.01, and 0.2728, respectively, after conducting the conrmation run to verify the results (Table 12). The scanned image of laser cut kerfs at initial parameter setting and that of optimum parameter setting suggested by hybrid

approach are shown in Figs. 6 and 7, respectively. The results of conrmation test show that two quality characteristics Kd and Kt have been improved considerably

ARTICLE IN PRESS
A.K. Dubey, V. Yadava / Optics and Lasers in Engineering 46 (2008) 124132 131

while third quality characteristic Kw deteriorates slightly. Therefore, overall improvement in MQC has been registered. 5. Conclusions The concluding remarks of the paper can be summarized as follows:

(4) The contribution of different control factors on MQC is gas pressure8%, pulse width33%, pulse frequency11%, and cutting speed48%. The cutting speed is found to be the most signicant parameter in this operating range.

48%
Percentage Contribution
A Gas pressure B Pulse width C Pulse frequency D Cutting speed

(1) In single-objective optimization using TM, pulse width, and cutting speed is the signicant factor for kerf width and kerf deviation, respectively, while both pulse frequency and cutting speed are signicant for kerf taper. (2) The optimum parameter levels predicted in singleobjective optimization for minimum value of kerf width, kerf deviation, and kerf taper are A1B1C1D1, A3B1C2D1, and A1B1C2D1, respectively. (3) As compared to initial parameter setting the MQC has been improved by using hybrid approach of TM and PCA. The optimum value of control factors for overall improvement in MQC is gas pressure 2 kg/cm2, pulse width 0.6 ms, pulse frequency 23 Hz, and cutting speed 20 mm/min. The value of kerf width, kerf deviation, and kerf taper at optimum parameter level is 0.246 mm, 0.01 mm, and 0.27281, respectively.
0.8
Gas pressure (kg/cm2)

33%

8%

11%

C D Control Factors

Fig. 5. Contribution of different control factors on TPCI of multiple quality characteristics. Table 12 Results of conrmation experiment for multi-objective optimization Initial Factor level Kerf width (mm) Kerf deviation (mm) Kerf taper (mm) A1B1C1D1 0.2340 0.0300 0.4092 Optimized A1B1C2D1 0.2460 0.0100 0.2728

Total PC Score Index (TPCI)

0.6

Pulse width (ms) Pulse frequency (Hz) Cutting speed (mm/min)

0.4

0.2

0.0 0 1 2 Factor level 3 4

Fig. 6. Scanned image of laser cut kerf at initial parameter level: (a) top view and (b) bottom view.

Fig. 4. Effect of factor levels on TPCI.

Table 11 Results of ANOVA for TPCI Symbol A B C D Error Pooled error Total
a

Factors Oxygen pressure Pulse width Pulse frequency Cutting speed

Degrees of freedom 2 2 2 2 4 8

Sum of squares 0.07a 0.30 0.10a 0.43 0.17 0.90

Mean squares 0.035 0.150 0.050 0.215 0.042

F 3.57 1.19 5.12

Contribution (%) 7.78 33.33 11.11 47.78

100

Pooled factor.

ARTICLE IN PRESS
132 A.K. Dubey, V. Yadava / Optics and Lasers in Engineering 46 (2008) 124132 [9] Almeida IA, Rossi W, Lima MSF, Berretta JR, Nogueira GEC, Wetter NU, et al. Optimization of titanium cutting by factorial analysis of the pulsed Nd:YAG laser parameters. J Mater Process Technol 2006;179:10510. [10] Cosp JP, Valle AJR, Fortea JG, Soto PJS. Laser cutting of highvitried ceramic materials: development of a method using a Nd:YAG laser to avoid catastrophic breakdown. Mater Lett 2002;55:27480. [11] Kuar AS, Doloi B, Bhattacharyya B. Experimental investigations on Nd:YAG laser cutting of silicon nitride. Int J Manuf Manage 2005;24:18191. [12] Lau WS, Lee WB. Pulsed Nd:YAG laser cutting of carbon bre composite materials. Ann CIRP 1992;39(1):17982. [13] Quintero F, Pou J, Lusquinos F, Boutinguiza M, Soto R, PerezAmor M. Quantitative evaluation of the cuts performed on mullitealumina by Nd:YAG laser. Opt Lasers Eng 2004;42:32740. [14] Ghany KA, Newishy M. Cutting of 1.2 mm thick austenitic stainless steel sheet using pulsed and CW Nd:YAG laser. J Mater Process Technol 2005;168:43847. [15] Prasad GVS, Siores E, Wong WCK. Laser cutting of metallic coated sheet steels. J Mater Process Technol 1998;74:23442. [16] Thawari G, Sarin Sundar JK, Sundararajan G, Joshi SV. Inuence of process parameters during pulsed Nd:YAG laser cutting of nickelbase superalloys. J Mater Process Technol 2005;170:22939. [17] Mathew J, Goswami GL, Ramakrishnan N, Naik NK. Parametric studies on pulsed Nd:YAG laser cutting of carbon bre reinforced plastic composites. J Mater Process Technol 1999;8990:198203. [18] Li C-H, Tsai M-J, Yang C-D. Study of optimal laser parameters for cutting QFN packages by Taguchis matrix method. Opt Laser Technol 2007;39(4):78695. [19] Tam SC, Lim LEN, Quek KY. Application of Taguchi method in the optimization of the laser-cutting process. J Mater Process Technol 1992;29:6374. [20] Lim S-H, Lee C-M, Chung WJ. A study on optimal cutting condition of a high speed feeding type laser cutting machine by using Taguchi method. Int J Precision Eng Manuf 2006;7(1):1823. [21] Antony J. Simultaneous optimization of multiple quality characteristics in manufacturing processes using Taguchis quality loss function. Int J Adv Manuf Technol 2001;17:1348. [22] Fung C-P, Kang P-C. Multi-response optimization in friction properties of PBT composites using Taguchi method and principal component analysis. J Mater Process Technol 2005;170:60210. [23] Tzeng Y-F, Chen F-C. Multiobjective process optimization for turning of tool steels. Int J Machining Machinability Mater 2006; 1(1):7693. [24] Phadke MS. Quality engineering using robust design. Englewood Cliffs, NJ: Prentice-Hall; 1989. [25] Ross PJ. Taguchi techniques for quality engineering. New York: McGraw-Hill; 1988. [26] Antony J. Multi-response optimization in industrial experiments using Taguchis quality loss function and principal component analysis. Qual Reliab Eng Int 2000;16:38.

Fig. 7. Scanned image of laser cut kerf at optimum parameter level: (a) top view and (b) bottom view.

(5) In multi-objective optimization, the loss in some quality characteristics is always possible as compared to singleobjective optimization but overall quality is improved. In present case, the kerf width deteriorates slightly but other two quality characteristics kerf deviation and kerf taper have been improved considerably.

Acknowledgement Financial assistance for this work was granted by the Department of Science and Technology (DST), Government of India, through project no. SR/S3/MERC-0076/ 2006 entitled Experimental and Numerical Study of Nd:YAG Laser Beam Cutting of Advanced Engineering Materials. Authors would like to acknowledge Mishra Dhatu Nigam Limited, Hyderabad (India), for valuable help in supplying the workpiece material in time. References
[1] Kalpakjian S. Manufacturing engineering and technology. AddisonWesley; 1995. [2] Luxon JT, Parker DE. Industrial lasers and their applications. Prentice-Hall; 1985. [3] Chryssolouris G. Laser machiningtheory and practice (mechanical engineering series). New York: Springer; 1991. [4] Steen WM. Laser material processing. New York: Springer; 1991. [5] Norikazu T, Shigenori Y, Masao H. Present and future of lasers for ne cutting of metal plate. J Mater Process Technol 1996;62:30914. [6] Meijer J. Laser beam machining (LBM), state of the art and new opportunities. J Mater Process Technol 2004;149:217. [7] Dubey AK, Yadava V. Experimental study of Nd:YAG laser beam machiningan overview. J Mater Process Technol, in press. [8] Shanjin L, Yang W. An investigation of pulsed laser cutting of titanium alloy sheet. Opt Lasers Eng 2006;44:106777.