A study of Taguchi optimization method for identifying optimum surface roughness in CNC face milling of cobalt-based alloy (stellite 6)

Int J Adv Manuf Technol (2006) 29: 940–947 DOI 10.1007/s00170-005-2616-y ORIGINAL ARTICLE Eyup Bagci · Şeref Aykut A study of Taguchi optimization method for identifying optimum surface roughness in CNC face milling of cobalt-based alloy (stellite 6) Received: 4 December 2004 / Accepted: 15 February 2005 / Published online: 21 December 2005 © Springer-Verlag London Limited 2005 Abstract The aim of this work is to develop a study of Taguchi optimization method for low surface roughness value in terms of cutting parameters when face milling of the cobalt-based alloy (stellite 6) material. The milling parameters evaluated are feed rate, cutting speed and depth of cut, a series of milling experiments are performed to measure the surface roughness data. The settings of face milling parameters were determined by using Taguchi experimental design method. Orthogonal arrays of Taguchi, the signal-to-noise (S/N) ratio, the analysis of variance (ANOVA) are employed to find the optimal levels and to analyze the effect of the milling parameters on surface roughness. Confirmation tests with the optimal levels of cutting parameters are carried out in order to illustrate the effectiveness of Taguchi optimization method. It is thus shown that the Taguchi method is very suitable to solve the surface quality problem occurring the face milling of stellite 6 material. Keywords Analysis of variance (ANOVA) · Face milling · Stellite 6 · Surface roughness · Taguchi optimization method 1 Introduction Various kinds of cobalt-based alloys called ‘stellite’ have been used in fields requiring high heat and corrosion resistance and high wear strength, such as the nuclear, aerospace, and gasturbine industries [1, 2]. Because of their good quality, studies on the production of new kinds of cobalt-based alloys are still E. Bagci (u) TUBiTAK-UME, National Metrology Institute, P.K. 54, 41470, Gebze-Kocaeli, Turkey E-mail: eyup@gyte.edu.tr Tel.: +90 262 679 5000 Fax: +90 262 679 5001 S. Aykut Institute of Science and Technology, Marmara University, Istanbul, Turkey being carried out extensively. At the same time, some other products such as wires, plates, and welding electrodes made from these alloys have been used successfully in different fields. Cobase superalloys rely primarily on carbides formed in the Co matrix and at grain boundaries for their strength and the distribution, size, and shape of carbides depends on processing condition. Solid solution strengthening of Co-base alloy is normally provided by tantalum, tungsten, molybdenum, chromium, and columbium [1–3]. These alloys existing in a variety of more than 20 commercially available today, are being used extensively in high temperature applications requiring superior wear resistance, corrosion resistance, and heat resistance [3, 4]. The main usage area of cobalt based superalloys is the place where corrosion and temperature resistance are needed. Having more percentage of chrome in alloys gives better magnetic properties, corrosion resistance, and the working ability in higher temperatures. However, the most certain property is the resistance to temperature [5]. In recent years, cobalt has an important place especially in medicine applications and manufacturing of corrosion resistant materials. Most certain properties in the used area: • • • • High creep resistance, High structural stability, Resistance to thermal creep, Resistance to high thermal corrosion. The cutting force or the surface roughness models are widely used for predicting the cutting force [6–8] and the surface roughness [9–13], respectively. These models are needed to monitor the process to obtain machining accuracy and process efficiency. Until now, most of the developed surface roughness models have been used for the prediction of surface texture and the quantitative analysis of the machined surface. However, in general, the insert runout errors in a cutter body cannot be avoided in a face-milling operation and it quantitatively affects the analysis of the cutting force and the surface roughness. Consequently, it is difficult to determine an optimal feed rate based on the surface roughness model, since it will be affected by the insert run out errors [14]. 941 Fig. 1. Surface roughness profile 2 Surface roughness and measurement Surface roughness of a machined product could affect several of the product’s functional attributes, such as contact causing surface friction, wearing, light reflection, heat transmission, ability of distributing and holding a lubricant, coating, and resisting fatigue [15]. There are several ways to describe surface roughness. One of them is average roughness which is often quoted as Ra symbol. Ra is defined as the arithmetic value of the departure of the profile from the centerline along sampling length as shown in Fig. 1. It can be expressed by the following mathematical relationships [16]. 1 Ra = L
L |Y(x)| dx (1) 0 where Ra = the arithmetic average deviation from the mean line, and Y = the ordinate of the profile curve. There are many methods of measuring surface roughness, such as using specimen blocks by eye or fingertip, microscopes, stylus type instruments, profile tracing instruments, etc. A photo of the used tool while working is shown in Fig. 2. Perthometer M1 model surface roughness tool of Mahr firm was used in experimental work. The tools measuring surface roughness with Fig. 2. Surface roughness measurement probes, measure, and control in appropriate length and circumferences. The probe comes in and out holes while traveling on the surface. This movement is turned into electrical current by means of a coil or crystal. After increasing the current by using suitable units, its value is shown with a pointer or digitally. 3 Taguchi experiment: design and analysis 3.1 Taguchi methods Essentially, traditional experimental design procedures are too complicated and not easy to use. A large number of experimental works have to be carried out when the number of the process parameters increases. To solve this problem, the Taguchi method uses a special design of orthogonal arrays to study the entire parameter space with only a small number of experiments [17]. Taguchi is the developer of the Taguchi method [18]. Taguchi methods (orthogonal array) has been widely utilized in engineering analysis and consists of a plan of experiments with the objective of acquiring data in a controlled way, in order to obtain information about the behavior of a given process. The greatest advantage of this method is to save the effort in conducting experiments: to save the experimental time, to reduce the cost, and to find out significant factors fast. Taguchi robust design method is a powerful tool for the design of a high-quality system. He considered three steps in a process’s and product’s development: system design, parameter design, and tolerance design. In system design, the engineer uses scientific and engineering principles to determine the fundamental configuration. In the parameter design step, the specific values for system parameters are determined. Tolerance design is used to determine the best tolerances for the parameters [19]. In addition to the S/N ratio, a statistical analysis of variance (ANOVA) can be employed to indicate the impact of process parameters on surface roughness. In this way, the optimal levels of process parameters can be estimated. The analysis results of related subjects discussed above are given in the following sections. The steps applied for Taguchi optimization in this study are presented in Fig. 3. 942 Fig. 3. Steps applied in Taguchi optimization method 3.2 Plan of experiments Taguchi methods which combine the experiment design theory and the quality loss function concept have been used in developing robust designs of products and processes and in solving some confusing problems of manufacturing [20]. The orthogonal array selected was the L 27 (313 ) which has 27 rows corresponding to the number of tests with three columns at three levels, as shown in Table 2 the factors and the interactions are assigned to the columns. The outputs studied were surface roughness (Ra ). For the purpose of observing the effect influence degree of cutting conditions (feed rate, depth of cut, and cutting speed) in face milling, three factors, each at three levels, are taken into account, as shown Tables 1 and 2. 4 Experimental setup and cutting conditions 4.1 Cutting conditions and methodology Table 1. The process parameters and their levels Sample Cutting conditions Level 1 Level 2 Level 3 A B C Depth of cut (mm) Cutting speed (m/min) Feed rate(mm/min) 0.25 50 100 0.50 70 140 0.75 90 180 Experimental work is done on a CNC milling machine. Surface roughness is investigated by the effect of cutting rate, feed rate and cutting depth. Cutting speeds: 50, 70, 90 m/min, feed rates: 100, 140, 180 mm/min, and cutting depths are selected as 0.25, 0.50, 0.75 mm. Processibility parameter values are selected as recommended ISO plane milling standard values [21]. Ex- Table 2. An orthogonal array L 27 (313 ) of Taguchi 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Designation A Depth of cut (mm) B Cutting speed (m/min) C Feed rate (mm/min) A1 B1 C 1 A1 B1 C 2 A1 B1 C 3 A1 B2 C 1 A1 B2 C 2 A1 B2 C 3 A1 B3 C 1 A1 B3 C 2 A1 B3 C 3 A2 B1 C 1 A2 B1 C 2 A2 B1 C 3 A2 B2 C 1 A2 B2 C 2 A2 B2 C 3 A2 B3 C 1 A2 B3 C 2 A2 B3 C 3 A3 B1 C 1 A3 B1 C 2 A3 B1 C 3 A3 B2 C 1 A3 B2 C 2 A3 B2 C 3 A3 B3 C 1 A3 B3 C 2 A3 B3 C 3 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 1 1 1 2 2 2 3 3 3 1 1 1 2 2 2 3 3 3 1 1 1 2 2 2 3 3 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 Fig. 4. CNC milling machine 943 Table 3. Mechanical and technical specifications of the tool (SECO) Coating thickness (µm) Hardness (Rockwell C) Operating temperature (◦ C) Thermal expansion coefficient (10−3 ◦ K−1 ) Chamfer (mm) × (◦ C) 3-4 78 600 9.4 0.15 × 20 technique in coating is physical vapor deposition (PVD). PVDTiN coated tools are used. The coating thickness is 3–4 µm and hardness is 78 HRc. Mechanical and chemical properties of tools are listed in Tables 3 and 4. Geometry of the cutting tool is shown in Fig. 6. 4.3 Work-piece materials (stellite 6) Fig. 5. Symmetric milling perimental setup and milling methodology are shown in Figs. 4 and 5. 4.2 Cutting tools Because heat extraction is inevitable in the cutting process, cutting corner of contact edge causes thermal stresses and will become blind in a short time. This affect could be prevented by coating the surface of the cutting tool. The heat that comes from tool-chip contact surface will be reflected by the coated tool to chip. The most preferred Table 4. Tool holder and inserts standards Tool holder and insert standard R220.43–063–07W OFEN070405-ME15 Fig. 6. Tool holder and cutting tool geometries Various kinds of cobalt-based alloys called ‘stellite’ have been used in fields requiring high heat and corrosion resistance and high wear strength. The most generally used cobalt alloy, having excellent resistance to many forms of mechanical and chemical degradation over a wide temperature range. Particular attributes are its outstanding self-mated anti-galling properties, high temperature hardness, and high resistance to cavitation erosion, which results in its wide use as a valve seat material. This alloy is ideally suited to a variety of hard facing processes. Some of applications are in non-ferrous (super)alloys, magnets, high speed tool steels, ultrahigh strength alloy steels, abrasion-resistant cemented carbides for cutting tools, and stainless steels and they are preferred for stator vanes and diaphragms in gas turbines because of their excellent thermal shock- and corrosion resistance [5]. Other uses of cobalt include the cemented carbide cutting materials, where cobalt acts as a binder. Cobalt is I (±0, 025) s (±0, 025) D (mm) A (mm) H (mm) ap (mm) K Inserts 18,02 4,76 63 75 40 5 43 ◦ C 4 944 Table 5. Mechanical properties (a) and chemical analysis (b) of cobalt based superalloy (Stellite 6) a Density Tensile strength (N/mm2 ) Young’s modulus (N/mm2 ) Melting point Boiling point Hardness b Elements Silisyum Mangan Crom Nickel Molibden Wolfram Titanium Fe Tantal Carbon Cobalt 8, 9 g/cm3 183 Mpa 210 Mpa 1493 ◦ C 2993 ◦ C 43.6 HRc % wt 1,07 0,485 28,166 1,92 0,96 5,17 0,01 2,88 0,041 1,09 Balans The smaller is better quality characteristics can be explained as [23]:  n  1 2 S/N(η) = −10 × log (2) yi n i=1 where n = number of measurements in a trial/row, in this case, n = 3 and yi is the ith measured value in a run/row. S/N ratio values are calculated by taking into consideration Eq. 2. The surface roughness values measured from experiments and S/N ratio values are listed in Table 6. Ra response table for the feed parameter (A) at levels 1, 2, and 3 was created by using the Ra values between 1–3, 10–12 and 19–21 in Table 6. Ra response table for each level of the process parameters (cutting speed, feed rate, and depth of cut) was created in the integrated manner and Ra response results are given in Table 7. On the other hand, the same procedure for S/N response table including process parameters was applied and the S/N response results are listed in Table 8. The effects of process parameters resulting from the optimization process are plotted in Figs. 8 and 9. 5.2 Analysis of variance (ANOVA) This method was developed by Sir Ronald Fisher in the 1930s as a way to interpret the results from agricultural experiments. ANOVA is a statistically based, objective decision-making tool for detecting any differences in average performance of groups of items tested [23]. Fig. 7. View of material stellite microstructure when 100 and 200 times magnified also used in hardfacing alloys, where it is most effective above 650 ◦ C in retaining the hardness and wear resistance for which these alloys are known. Minor uses for cobalt-bearing alloys include metal-to-ceramic and metal-to-glass seals, watch springs, pen nibs, and medical and dental alloys. The chemical composition mechanical properties of workpiece material is given in Table 5. Figure 7 shows stellite 6 material’s microstructure as 100 and 200 times magnified. 5 Data analysis and results 5.1 Analysis of the S/N ratio In the Taguchi method, the term ‘signal’ represents the desirable value (mean) for the output characteristic and the term ‘noise’ represents the undesirable value (S.D.) for the output characteristic. Therefore, the S/N ratio is the ratio of the mean to the S.D. Taguchi uses the S/N ratio to measure the quality characteristic deviating from the desired value. There are several S/N ratios available depending on type of characteristic; lower is better (LB), nominal is best (NB), or higher is better (HB) [22]. Table 6. The Ra and S/N ratio values Experiment Surface numbers roughness value S/N ratio (dB) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 6,6509 6,2852 5,8827 13,2309 6,5951 9,9515 13,4733 10,7820 10,2572 6,3391 4,0132 2,6271 6,1079 5,7977 4,8825 8,3815 8,1343 6,2316 9,6561 5,4655 2,9260 7,9588 3,4915 3,2356 13,7649 11,6340 9,7090 0,465 0,485 0,508 0,218 0,468 0,318 0,212 0,289 0,307 0,382 0,630 0,739 0,495 0,513 0,570 0,381 0,392 0,488 0,329 0,533 0,714 0,400 0,669 0,689 0,205 0,262 0,327 945 Table 7. Ra response table for surface roughness Levels 1 2 3 ∆max − min Rank A (mm/tooth) B (m/min) C (mm) 0.363 0.51 0.458 0.147 3 0.531 0.482 0.318 0.213 1 0.343 0.471 0.517 0.174 2 Table 8. S/N response table for surface roughness Levels 1 2 3 ∆max − min Rank A (mm/tooth) B (m/min) C (mm) −9, 23 −5, 83 −7, 53 3,39 2 −5, 53 −6, 80 −10, 26 4,72 1 −9, 50 −6, 91 −6, 18 3,31 3 Fig. 8. Ra response table for surface roughness ANOVA helps in formally testing the significance of all main factors and their interactions by comparing the mean square against an estimate of the experimental errors at specific confidence levels. This is to be accomplished by separating the total variability of the S/N ratios, which is measured by the sum of the squared deviations from the total mean S/N ratio, into contributions by each of the design parameters and the error. First, the total sum of squared deviations SST from the total mean S/N ratio ηm can be calculated as [24]: n  (ηi − ηm )2 SST = (3) i=1 where n is the number of experiments in the orthogonal array and ηi is the mean S/N ratio for the ith experiment. The percentage contribution P can be calculated as below: P= SSd SST (4) where SSd is the sum of squared deviations. ANOVA results are illustrated in Table 7. Statistically, there is a tool called an F test named after Fisher [19] to see which design parameters have a significant effect on the quality characteristic. In the analysis, Table 9. ANOVA results for surface roughness for steelite 6 material Fig. 9. The effects of process parameters (S/N response table for surface roughness) F- ratio is a ratio of mean square error to residual, and is traditionally used to determine the significance of a factor. F ratio corresponding 95% confidence level in calculation of process parameters accurately is F0.05,2,26 = 3.37. P value reports the significance level (suitable and unsuitable) in Table 9. Percent (%) is defined as the significance rate of process parameters on drill bit temperature. The percent numbers depicts that depth of cut, feed, and cutting speed have significant effects on surface roughness. It can observed from Table 9 that depth of cut (A), cutting speed (B), feed rate (C), affect surface roughness by 17.88%, 38.27%, 20.14%, the stellite 6 material surfaces, consecutively. The depth of cut, feed rate, and spindle speed factors present statistical and physical significance on drill bit temperature, because Test F > Fα = 5% as shown in Table 9. Source of variation Degree of freedom (DOF) Sum of squares (S) MS F ratio (F) P value (P) Contribution P (%) A B C Error Total 2 2 2 20 26 0,113645 0,243224 0,128008 0,150596 0,635472 0,056822 0,121612 0,064004 0,007530 7,55 16,15 8,50 0,004 0,000 0,002 17,88% 38,27% 20,14% 23,69% 100% 946 Table 10. L27 (313 ) standard orthogonal array table with factors A, B, and C arranged in columns 2, 5 and 6, respectively. The interactions among factors are indicated as in columns 1, 7, 8, 9, 11 and 12 Experimental run 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 1 2 3 4 5 6 7 8 9 10 11 12 13 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 B×C 0 0 0 1 1 1 2 2 2 0 0 0 1 1 1 2 2 2 0 0 0 1 1 1 2 2 2 A 0 0 0 1 1 1 2 2 2 1 1 1 2 2 2 0 0 0 2 2 2 0 0 0 1 1 1 0 0 0 1 1 1 2 2 2 2 2 2 0 0 0 1 1 1 0 0 0 1 1 1 2 2 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 B 0 1 2 0 1 2 0 1 2 1 2 0 1 2 0 1 2 0 2 0 1 2 0 1 2 0 1 C 0 1 2 0 1 2 0 1 2 2 0 1 2 0 1 2 0 1 1 2 0 1 2 0 1 2 0 B×C 0 1 2 1 2 0 2 0 1 0 1 2 1 2 0 2 0 1 0 1 2 1 2 0 2 0 1 A× B 0 1 2 1 2 0 2 0 1 1 2 0 2 0 1 0 1 2 2 0 1 0 1 2 1 2 0 A×C 0 1 2 1 2 0 2 0 1 2 1 0 0 1 2 1 2 0 1 2 0 2 0 1 0 1 2 0 1 2 2 0 1 1 2 0 0 1 2 2 0 1 1 2 0 0 1 2 2 0 1 1 2 0 A× B 0 1 2 2 0 1 1 2 0 1 2 0 0 1 2 2 0 1 2 0 1 1 2 0 0 1 2 A×C 0 1 2 2 0 1 1 2 0 2 0 1 1 2 0 0 1 2 1 2 0 0 1 2 2 0 1 6 Determination of the minimum surface roughness Using the before mentioned data, one can predict the optimum surface roughness performance using the cutting parameters as: Predicted Mean (Minimum roughness) = A1 + B3 + C1 − 2x(Y) = 0.363 + 0.318 + 0.343 − 2(0.443) = 0.138 µm. (5) Similarly, the maximum S/N ratio is calculated to determine whether or not the minimum surface roughness is acceptable. Also, the maximum S/N ratio varies from the min = −11 dB to max = +∞ dB. The S/N ratio could be predicted as: Predicted S/N Ratio (Maximum) = η A1 + η B3 + ηC1 − x(η) = −9.23 − 10.26 − 9.50 + 2(7.53) = − 13.93 dB A confirmation of the experimental design was necessary in order to verify the optimum cutting conditions. 7 Confirmation tests The confirmation experiment is very important in parameter design, particularly when screening or small fractional factorial experiments are utilized. In this study, a confirmation experiment was conducted by utilizing the level of optimal process parameters (A1 B3 C1 ) in the part surfaces. The purpose of the confirmation experiment in this study was to validate the optimum cutting conditions (A1 B3 C1 ) that were suggested by the experiment that corresponded with the predicted value. In this research, the confirmation runs with the optimum cutting condition A1 B3 C1 resulted in response values of 0.147, 0.143 and 0.140 µm. Each Ra measurement was repeated at least three times. Therefore, the optimum surface roughness (Ra = 0.143 µm) can be obtained under the above-mentioned cutting condition in the Deckel Maho CNC vertical milling machine. (6) where η is the average value of surface roughness or S/N ratio. With this prediction, one could conclude that the machine creates the best surface roughness (Ra = 0.138 µm) within the range of specified cutting conditions (Table 1). The Ra value of 0.138 µm is the smallest value involving in experimental measurements. 8 Conclusions This study has discussed an application of the Taguchi method for investigating the effects of cutting parameters on the surface roughness value in the face milling of stellite 6 material. 947 In the face milling processes, cutting conditions have different cutting speed, depth of cut, and feed rate values. As shown in this study, the Taguchi method provides a systematic and efficient methodology for the design optimization of the cutting parameters with far less effect than would be required for most optimization techniques. From the analysis of result in face milling using the conceptual S/N ratio approach and ANOVA, the following can be concluded from the present study: 1. Statistically designed experiments based on Taguchi methods were performed using L-27 orthogonal array to analyze surface roughness as response variable. 2. Conceptual S/N ratio and ANOVA approaches for data analysis draw similar conclusion. 3. Statistical results indicate that surface roughness is significantly influenced (at 95% confidence level) by cutting speed, feed rate, and depth of cut. 4. In this study, the analysis of confirmation experiments has shown that Taguchi parameter design can successfully verify the optimum cutting parameters (A1B3C1), which are depth of cut = 0.25 mm (A1), cutting speed = 90 m/min (B3), and feed rate = 100 mm/min (C1). 5. The optimum surface roughness (Ra = 0.143 µm) can be obtained under the above-mentioned cutting condition in the Deckel Maho CNC vertical milling machine. Further study could consider more factors ( different insert geometry, materials, lubricant, cooling strategy etc.) in the research to see how the factors would affect surface roughness. References 1. Sullivan CP, Donachie MJ, Moral FR (1970) Cobalt base superalloys, Co Monograph Series, Centre d’Information du Cobalt, Brussels 2. Antony KC (1983) Wear resistant cobalt-base alloys. J Metal 35(2): 52–60 3. Crook P (1993) Properties and selection: nonferrous alloys and special-purpose materials, vol 2. Metals Handbook, ASM International, pp 446 View publication stats 4. Agarwal SC, Ocken H (1990) The microstructure and galling wear of a laser-melted cobalt-base hardfacing alloy. Wear 140(2):223–233 5. Bristow DJ et al. http://www.shieldalloy.com/cobaltpage.html 6. Fu HJ, Devor RE, Kapoor SG (1984) A mechanistic model for prediction of the force system face milling operation. ASME J Eng Ind 106:81–88 7. Kline WA, Devor RE (1983) The effect of runout on cutting geometry and for in milling. J MTDR 23(2–3):123–140 8. Ruzhong Z, Wang KK (1983) Model of cutting force pulsation on face milling. Ann CIRP 321:21–26 9. Kline WW, Devor RE, Shareef IA (1986) The prediction of surface accuracy in end milling. ASME J Eng Ind 104:272–278 10. Sutherland JW, Devor RE (1986) An improved method for cutting force and surface error prediction in flexible end milling systems. ASME J Eng Ind 108:269–279 11. You SJ, Ehmann KF (1989) Scallop removal in die milling by tertiary cutter motion. ASME J Eng Ind 111:213–219 12. Elbestawi MA, Ismail F, Yuen KM (1993) Surface topography characterization in finish milling. Int J Mach Tools Manuf 34(2): 245–255 13. Ismail F, Elbestawi MA, Du R, Urbasik K (1993) Generation of milled surface including tool dynamics and wear. ASME J Eng Ind 115: 245–252 14. Ehmann KF, Hong MS (1994) A generalized model of the surface generation process in metal cutting. Ann CIRP 43(1):483–486 15. Lou MS, Chen JC, Li C (1998) Surface roughness prediction technique for CNC end- milling. J Ind Technol 15(1):1–6 16. Yang JL, Chen JC (2001) A systematic approach for identifying optimum surface roughness performance in end-milling operations. J Ind Technol 17:2 17. Yang WH, Tang YS (1998) Design optimization of cutting parameters for turning operations based on the Taguchi method. J Mater Proc Technol 84(1–3):122–129 18. Taguchi G (1990) Introduction to quality engineering. Asian Productivity Organization, Tokyo 19. Ross PJ (1996) Taguchi techniques for quality engineering. McGraw– Hill Int Edn, Singapore 20. Tsao CC, Hocheng H (2004) Taguchi analysis of delamination associated with various drill bits in drilling of composite material. Int J Mach Tools Manuf 44:1085–1090 21. ISO 8688-1 (1989) Tool life testing in milling, Part I. Face milling, 1st edn 22. Phadke MS (1989) Quality engineering using robust design. PrenticeHill, Englewood Cliffs, NJ 23. Minitab User Manual Release 13.2 (2001) Making data analysis easier. MINITAB Inc, State College, PA, USA 24. Harold R (1992) Analysis of variance in experimental design. Springer, Berlin Heidelberg New York