ARTICLE IN PRESS

Materials & Design

Materials and Design xxx (2006) xxxxxx www.elsevier.com/locate/matdes

Short communication

Application of Taguchi method in the optimization of cutting parameters for surface roughness in turning
M. Nalbant *, H. Gokkaya, G. Sur
Department of Mechanical Education, Faculty of Technical Education, Gazi University, 06503 Teknikokullar, Ankara, Turkey Received 21 July 2005; accepted 6 January 2006

Abstract In this study, the Taguchi method is used to nd the optimal cutting parameters for surface roughness in turning. The orthogonal array, the signal-to-noise ratio, and analysis of variance are employed to study the performance characteristics in turning operations of AISI 1030 steel bars using TiN coated tools. Three cutting parameters namely, insert radius, feed rate, and depth of cut, are optimized with considerations of surface roughness. Experimental results are provided to illustrate the eectiveness of this approach. 2006 Elsevier Ltd. All rights reserved.

1. Introduction In modern industry the goal is to manufacture low cost, high quality products in short time. Automated and exible manufacturing systems are employed for that purpose along with computerised numerical control (CNC) machines that are capable of achieving high accuracy and very low processing time. Turning is the rst most common method for cutting and especially for the nishing machined parts. In a turning operation, it is important task to select cutting parameters for achieving high cutting performance. Usually, the desired cutting parameters are determined based on experience or by use of a handbook. Cutting parameters are reected on surface roughness, surface texture and dimensional deviations of the product. Surface roughness, which is used to determine and to evaluate the quality of a product, is one of the major quality attributes of a turning product. Surface roughness is a measure of the technological quality of a product and a factor that greatly inuences manufacturing cost. It describes the geometry of the machined surfaces and combined with the surface texture. The mechanism behind the formation of surface roughCorresponding author. Tel.: +90 312 212 3952/6820; fax: +90 312 212 0059. E-mail address: nalbant@gazi.edu.tr (M. Nalbant). 0261-3069/$ - see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.matdes.2006.01.008
*

ness is very complicated and process dependent. To select the cutting parameters properly, several mathematical models [15] based on statistical regression or neural network techniques have been constructed to establish the relationship between the cutting performance and cutting parameters. Then, an objective function with constraints is formulated to solve the optimal cutting parameters using optimization techniques. Therefore, considerable knowledge and experience are required for this approach. In this study, an alternative approach based on the Taguchi method [68] is used to determine the desired cutting parameters more eciency. There were two purposes of this research. The rst was to demonstrate a systematic procedure of using Taguchi parameter design in process control of turning machines. The second was to demonstrate a use of the Taguchi parameter design in order to identify the optimum surface roughness performance with a particular combination of cutting parameters in a turning operation. The paper is organized in the following manner. An overview of the parameter design based on the Taguchi method is given rst. Then, the parameter design with the multiple performance characteristics is introduced. The experimental detail of using the parameter design to determine and analyze the optimal cutting parameters in turning operations is described next. Finally, the paper concludes with a summary of this study.

ARTICLE IN PRESS
2 M. Nalbant et al. / Materials and Design xxx (2006) xxxxxx

2. Taguchi method Taguchi has developed a methodology for the application of designed experiments, including a practitioners handbook [1]. This methodology has taken the design of experiments from the exclusive world of the statistician and brought it more fully into the world of manufacturing. His contributions have also made the practitioner work simpler by advocating the use of fewer experimental designs, and providing a clearer understanding of the variation nature and the economic consequences of quality engineering in the world of manufacturing [1,2]. Taguchi introduces his approach, using experimental design for [2]: designing products/processes so as to be robust to environmental conditions; designing and developing products/processes so as to be robust to component variation; minimizing variation around a target value. The philosophy of Taguchi is broadly applicable. He proposed that engineering optimization of a process or product should be carried out in a three-step approach, i.e., system design, parameter design, and tolerance design. In system design, the engineer applies scientic and engineering knowledge to produce a basic functional prototype design, this design including the product design stage and the process design stage. In the product design stage, the selection of materials, components, tentative product parameter values, etc., are involved. As to the process design stage, the analysis of processing sequences, the selections of production equipment, tentative process parameter values, etc., are involved. Since system design is an initial functional design, it may be far from optimum in terms of quality and cost. The objective of the parameter design [9] is to optimize the settings of the process parameter values for improving performance characteristics and to identify the product parameter values under the optimal process parameter values. In addition, it is expected that the optimal process parameter values obtained from the parameter design are insensitive to the variation of environmental conditions and other noise factors. Therefore, the parameter design is the key step in the Taguchi method to achieving high quality without increasing cost. Basically, classical parameter design, developed by Fisher [10], is complex and not easy to use. Especially, a large number of experiments have to be carried out when the number of the process parameters increases. To solve this task, the Taguchi method uses a special design of orthogonal arrays to study the entire parameter space with a small number of experiments only. A loss function is then dened to calculate the deviation between the experimental value and the desired value. Taguchi recommends the use of the loss function to measure the performance characteristic deviating from the desired value. The value of the loss function is further transformed into a signal-to-noise (S/N)

ratio g. Usually, there are three categories of the performance characteristic in the analysis of the S/N ratio, that is, the lower-the-better, the higher-the-better, and the nominal-the-better. The S/N ratio for each level of process parameters is computed based on the S/N analysis. Regardless of the category of the performance characteristic, the larger S/N ratio corresponds to the better performance characteristic. Therefore, the optimal level of the process parameters is the level with the highest S/N ratio g. Furthermore, a statistical analysis of variance (ANOVA) is performed to see which process parameters are statistically signicant. With the S/N and ANOVA analyses, the optimal combination of the process parameters can be predicted. Finally, a conrmation experiment is conducted to verify the optimal process parameters obtained from the parameter design. In this paper, the cutting parameter design by the Taguchi method is adopted to obtain optimal machining performance in turning. !  y Nominal is the best: S=N T 10 log 2 1 sy ! n 1X 1 Larger-is-the better(maximize): S=N L 10 log n i1 y 2 i
n 1X 2 y n i1 i

2 !

Smaller-is-the better(minimize): S=N S 10 log

3 where , is the average of observed data, is the variance y of y, n is the number of observations and y is the observed data. Notice that these S/N ratios are expressed on a decibel scale. We would use S/NT if the objective is to reduce variability around a specic target, S/NL if the system is optimized when the response is as large as possible, and S/NS if the system is optimized when the response is as small as possible. Factor levels that maximize the appropriate S/N ratio are optimal. The goal of this research was to produce minimum surface roughness (Ra) in a turning operation. Smaller Ra values represent better or improved surface roughness. Therefore, a smaller-the-better quality characteristic was implemented and introduced in this study [9]. The use of the parameter design of the Taguchi method to optimize a process with multiple performance characteristics includes the following steps [11]:  Identify the performance characteristics and select process parameters to be evaluated.  Determine the number of levels for the process parameters and possible interactions between the process parameters.  Select the appropriate orthogonal array and assignment of process parameters to the orthogonal array.  Conduct the experiments based on the arrangement of the orthogonal array. s2 y

ARTICLE IN PRESS
M. Nalbant et al. / Materials and Design xxx (2006) xxxxxx 3

 Calculate the total loss function and the S/N ratio.  Analyze the experimental results using the S/N ratio and ANOVA.  Select the optimal levels of process parameters.  Verify the optimal process parameters through the conrmation experiment.

Table 3 Cutting parameters and their levels (cutting speed is kept constant 300 m/min) Symbol A B C Cutting parameter Insert radius (mm) Depth of cut (mm) Feed rate (mm/rev) Level 1 0.4 0.5 0.15 Level 2 0.8 1.5 0.25 Level 3 1.2 2.5 0.35

3. Turning process experiments Turning is a widely used machining process in which a single-point cutting tool removes material from the surface of a rotating cylindrical workpiece. Three cutting parameters, i.e., feed rate, depth of cut, and insert radius must be determined in a turning operation. A common method of evaluating machining performance in a turning operation is based on the surface roughness. Basically, surface roughness is strongly correlated with cutting parameters such as insert radius, feed rate, and depth of cut [12]. Surface roughness is not strongly correlated with cutting speed therefore cutting speed is not used in this paper. Proper selection of the cutting parameters can obtain better surface roughness. Hence, optimization of the cutting parameters based on the parameter design of the Taguchi method is adopted in this paper to improve surface roughness in a turning operation. 3.1. Selection of cutting parameters and their levels The cutting experiments were carried out on an Johnford T35 CNC lathe using TiN coated tools with the grade of P-20 for the machining of AISI 1030 steel bars. In the tests used inserts were TNMG160404-MA, TNMG160408-MA and TNMG160412-MA. Mechanical properties and chemical composition of AISI 1030 carbon steel which was used in the experiments as shown in Tables 1 and 2, respectively. The initial cutting parameters were as follows: insert radius of 0.8 mm, a feed rate of 0.25 mm/ rev, and a depth of cut of 1.5 mm. The feasible range for the cutting parameters was recommended by a machining handbook, i.e., insert radius in the range 0.41.2 mm, feed rate in the range 0.150.35 mm/rev, and depth of cut in the range 0.52.5 mm. Therefore, three levels of the cutting parameters were selected as shown in Table 3. 3.2. Machining performance measure There are various simple surface roughness amplitude parameters used in industry, such as roughness average (Ra), root-mean-square (rms) roughness (Rq), and maximum peak-to-valley roughness (Ry or Rmax), etc. [13,14]. The parameter Ra is used in this study. The average roughness (Ra) is the area between the roughness prole and its mean line, or the integral of the absolute value of the roughness prole height over the evaluation length (Fig. 1). Therefore, the Ra is specied by the following equation: Z 1 L jY xj dx; 4 Ra L 0 where Ra is the arithmetic average deviation from the mean line, L is the sampling length, Y the ordinate of the prole curve. The average surface roughness Ra, is measured within the sampling length of 5.6 mm. 4. Determination of optimal cutting parameters In this section, the use of an orthogonal array to reduce the number of cutting experiments for determining the optimal cutting parameters is reported. Results of the cutting experiments are studied by using the S/N and ANOVA analyses. Based on the results of the S/N and

Fig. 1. Surface roughness prole.

Table 1 Mechanical properties of AISI 1030 carbon steel Elongation (%) 31.2 Hardness (HB) 126 Tensile strength (MPa) 463.7 Yield strength (MPa) 341.3 Thermal conductivity (W/m K) 51.9

Table 2 Chemical composition of AISI 1030 carbon steel, % weight C 0.276 Si 0.110 Mn 0.610 P 0.040 S 0.050

ARTICLE IN PRESS
4 M. Nalbant et al. / Materials and Design xxx (2006) xxxxxx Table 5 Experimental results for surface roughness and S/N ratio Experiment number Cutting parameter level A B C Measured Calculated Surface S/N ratio roughness for surface Designation roughness

ANOVA analyses, optimal cutting parameters for surface roughness are obtained and veried. 4.1. Orthogonal array experiment To select an appropriate orthogonal array for experiments, the total degrees of freedom need to be computed. The degrees of freedom are dened as the number of comparisons between process parameters that need to be made to determine which level is better and specically how much better it is. For example, a three-level process parameter counts for two degrees of freedom. The degrees of freedom associated with interaction between two process parameters are given by the product of the degrees of freedom for the two process parameters. In the present study, the interaction between the cutting parameters is neglected. Therefore, there are six degrees of freedom owing to the three cutting parameters in turning operations. Once the degrees of freedom required are known, the next step is to select an appropriate orthogonal array to t the specic task. Basically, the degrees of freedom for the orthogonal array should be greater than or at least equal to those for the process parameters. In this study, an L9 orthogonal array was used. This array has twenty six degrees of freedom and it can handle three-level process parameters. Each cutting parameter is assigned to a column and twenty seven cutting parameter combinations are available. Therefore, only twenty seven experiments are required to study the entire parameter space using the L9 orthogonal array. The experimental layout for the three cutting parameters using the L9 orthogonal array is shown in Table 4. Table 5 shows the experimental results and S/N ratio of surface roughness. 4.2. Analysis of the signal-to-noise (S/N) ratio As mentioned earlier, there are three categories of performance characteristics, i.e., the lower-the-better, the higher-the-better, and the nominal-the-better. To obtain optimal machining performance, the lower-the-better performance characteristic for surface roughness should be taken for obtaining optimal machining performance.
Table 4 Experimental layout using an L9 orthogonal array Experiment number Cutting parameter level A Insert rad. 1 2 3 4 5 6 7 8 9 1 1 1 2 2 2 3 3 3 B Dept of cut 1 2 3 1 2 3 1 2 3 C Feed rate 1 2 3 2 3 1 3 1 2 D Error

Insert Dept of Feed rad. cut rate 1 2 3 4 5 6 7 8 9 0.4 0.4 0.4 0.8 0.8 0.8 1.2 1.2 1.2 0.5 1.5 2.5 0.5 1.5 2.5 0.5 1.5 2.5 0.15 0.25 0.35 0.25 0.35 0.15 0.35 0.15 0.25 00 ++ 00 00+ 0+ ++ +0 ++0 2.170 6.385 13.630 2.435 5.215 1.800 2.535 0.875 1.515 6.734 16.104 22.695 7.730 14.347 5.1076 8.079 1.159 3.608

Table 5 shows the experimental results for surface roughness and the corresponding S/N ratio using Eq. (3). Since the experimental design is orthogonal, it is then possible to separate out the eect of each cutting parameter at dierent levels. For example, the mean S/N ratio for the insert radius at levels 1, 2 and 3 can be calculated by averaging the S/N ratios for the experiments 13, 46, and 79, respectively. The mean S/N ratio for each level of the other cutting parameters can be computed in the similar manner. The mean S/N ratio for each level of the cutting parameters is summarized and called the mean S/N response table for surface roughness (Table 6). In addition, the total mean S/N ratio for the nine experiments is also calculated and listed in Table 6. Fig. 1 shows the mean S/N ratio graph for surface roughness. The S/N ratio corresponds to the smaller variance of the output characteristics around the desired value. Fig. 2 shows the predicted versus actual values for S/N ratio of surface roughness. Here, it is seen that most of the points are close to this line. Fig. 3 shows the plot of predicted values versus residual values for S/N ratio of surface roughness. Deviations are very small each parameter and negligible. 4.3. Analysis of variance The purpose of the ANOVA is to investigate which of the process parameters signicantly aect the performance characteristics. This is accomplished by separating the total

Table 6 Response table mean S/N ratio for surface roughness factor and signicant interaction Symbol Cutting parameter Mean S/N ratio Level 1 A B C Insert radius Depth of cut Feed rate 15.178 7.515 3.561 Level 2 9.062 9.764 9.148 Level 3 3.509 10.470 15.041 Maxmin 11.669 2.955 11.480

Total mean S/N ratio = 9.249.

ARTICLE IN PRESS
M. Nalbant et al. / Materials and Design xxx (2006) xxxxxx 5

Fig. 2. The mean single-to-noise graph for surface roughness.

Fig. 3. Comparison of actual and predicted S/N ratio of surface roughness using regression analysis.

Fig. 4. Comparison of residual and predicted S/N ratio of surface roughness using regression analysis.

variability of the S/N ratios, which is measured by the sum of the squared deviations from the total mean of the S/N ratio, into contributions by each of the process parameters and the error. First, the total sum of the squared deviations SST from the total mean of the S/N ratio g can be calculated as [15]. m m m m X X X X 2 2 gi  g g2 2gi  SS T g g i
i1 i1 i1 m X i1 i1

" #2 t m X sg 2 1 X i g SS P m i1 i t j1

where p represent one of the experiment parameters, j the level number of this parameter p, t the repetition of each level of the parameter p, sgj the sum of the S/N ratio involving this parameter p and level j. The sum of squares from error parameters SSe is SS e SS T SS A SS B SS C 7 The total degrees of freedom is DT = m 1, where the degrees of freedom of the tested parameter Dp = t 1. The variance of the parameter tested is VP = SSP/DP. Then, the F-value for each design parameter is simply the ratio of the mean of squares deviations to the mean of the squared error (FP = VP/Ve). The corrected sum of squares SP can be calculated as: b P SS P DP V e S The percentage contribution q can be calculated as: q bP S SS T 9 8

m X i1 m X i1

g2 2m2 m2 g g i g2 i 1 m "

m X i1

g2 m2 g i 5

#2 gi

where m is the number of experiments in the orthogonal array, e.g., m = 9 and gi is the mean S/N ratio for the ith experiment (Fig. 4). The total sum of the squared deviations SST is decomposed into two sources: the sum of the squared deviations SSP due to each process parameter and the sum of the squared error SSe. SSP can be calculated as:

ARTICLE IN PRESS
6 M. Nalbant et al. / Materials and Design xxx (2006) xxxxxx Table 8 Results of the conrmation experiment for surface roughness Initial cutting parameters Level Surface roughness (lm) S/N ratio (dB) Improvement of S/N ratio A2B2C2 2.68 8.567 Optimal cutting parameters Prediction A3B1C1 3.913 10.502 Experiment A3B1C1 0.8 1.935

Statistically, there is a tool called the F-test named after Fisher [10] to see which process parameters have a signicant eect on the performance characteristic. In performing the F-test, the mean of the squared deviations SSm due to each process parameter needs to be calculated. The mean of the squared deviations SSm is equal to the sum of the squared deviations SSd divided by the number of degrees of freedom associated with the process parameter. Then, the F-value for each process parameter is simply a ratio of the mean of the squared deviations SSm to the mean of the squared error SSe. Usually the larger the F-value, the greater the eect on the performance characteristic due to the change of the process parameter. Table 7 shows the results of ANOVA for surface roughness. It can be found that the insert radius and feed rate are the signicant cutting parameters for aecting the surface roughness. The change of the depth of cut in the range given by Table 3 has an insignicant eect on surface roughness. Therefore, based on the S/N and ANOVA analyses, the optimal cutting parameters for surface roughness are the insert radius at level 3, the feed rate at level 1, and the depth of cut at level 1. 4.4. Conrmation tests Once the optimal level of the process parameters is selected, the nal step is to predict and verify the improvement of the performance characteristic using the optimal level of the process parameters. The estimated S/N ratio ^ using the optimal level of the process parameters can g be calculated as [5] ^ gm g
q X i1

the initial cutting parameters to the optimal cutting parameters is 10.502 dB. The improvement of the S/N ratio for the individual performance characteristic is shown in Table 8. Based on the result of the conrmation test, the surface roughness is decreased 3.35 times, in the foregoing discussion; the experimental results conrm the prior parameter design for the optimal cutting parameters with the multiple performance characteristics in turning operations. 5. Conclusions This paper has presented an application of the parameter design of the Taguchi method in the optimization of turning operations. The following conclusions can be drawn based on the experimental results of this study:  Taguchis robust orthogonal array design method is suitable to analyze the surface roughness (metal cutting) problem as described in this paper.  It is found that the parameter design of the Taguchi method provides a simple, systematic, and ecient methodology for the optimization of the cutting parameters.  The experimental results demonstrate that the insert radius and feed rate are the main parameters among the three controllable factors (insert radius, feed rate and depth of cut) that inuence the surface roughness in turning AISI 1030 carbon steel.  Surface roughness can be improved simultaneously through this approach instead of using engineering judgement. The conrmation experiments were conducted to verify the optimal cutting parameters. The percentage contributions of insert radius, feed rate and depth of cut are 48.54, 46.95 and 3.39, respectively.  In turning, use of greater insert radius (1.2 mm), low feed rate (0.15 mm/rev) and low depth of cut (0.5 mm) are recommended to obtain better surface roughness for the specic test range.  The improvement of surface roughness form initial cutting parameters to the optimal cutting parameters is about 335%.  Deviations between actual and predicted S/N ratio of surface roughness are small each parameter. This research demonstrates how to use Taguchi parameter design for optimizing machining performance with minimum cost and time to industrial readers.

i gm g

10

g where gm is the total mean of the S/N ratio, i , is the mean S/N ratio at the optimal level, and q is the number of the process parameters that signicantly aect the performance characteristic. The estimated S/N ratio using the optimal cutting parameters for surface roughness can then be obtained and the corresponding surface roughness can also be calculated by using Eq. (3). Table 8 shows the results of the conrmation experiment using the optimal cutting parameters of surface roughness. Good agreement between the predicted machining performance and actual machining performance is shown. The increase of the S/N ratio from
Table 7 Results of the analysis of variance for surface roughness Source of variation Insert radius Depth of cut Feed rate Error Total Degree of freedom 2 2 2 2 8 Sum of squares 204.377 14.295 197.735 4.680 421.087 Mean square 102.189 7.148 98.867 2.340 F ratio 43.668 3.054 42.248 Contribution (%) 48.54 3.39 46.95 1.11 100

ARTICLE IN PRESS
M. Nalbant et al. / Materials and Design xxx (2006) xxxxxx 7

Further study could consider more factors (e.g., cutting speed, materials, lubricant, etc.) in the research to see how the factors would aect surface roughness. Also, further study could consider the outcomes of Taguchi parameter design when it is implemented as a part of management decision-making processes. References
[1] Taguchi G. Introduction to quality engineering. Tokyo: Asian Productivity Organization; 1990. [2] Ross PJ. Taguchi techniques for quality engineering. New York: McGraw-Hill; 1988. [3] Phadke MS. Quality engineering using robust design. Englewood Clis (NJ): Prentice-Hall; 1989. [4] Disney BJ, Pridmore WA. Taguchi methods: applications in world industry. UK: IFS Publications; 1989. [5] Yang H, Tarng YS. Design optimization of cutting parameters for turning operations based on the Taguchi method. J Mater Process Technol 1998;84:1229. [6] Abuelnaga AM, El-Dardiry MA. Optimization methods for metal cutting. Int J Mach Tool Design Res 1984;24(1):118.

[7] Chryssolouris G, Guillot M. A comparison of statistical and AI approaches to the selection of process parameters in intelligent machining. ASME J Eng Ind 1990;112:12231. [8] Chua MS, Rahman M, Wong YS, Loh HT. Determination of optimal cutting conditions using design of experiments and optimization techniques. Int J Mach Tools Manuf 1993;33(2): 297305. [9] Montgomery DC. Design and analysis of experiments. 4th ed. New York: Wiley; 1997. [10] Fisher RA. Statistical methods for research worker. London: Oliver & Boyd; 1925. [11] Nian CY, Yang WH, Tarng YS. Optimization of turning operations with multiple performance characteristics. J Mater Process Technol 1999;95:906. [12] Shaw MC. Metal cutting principle. New York: Oxford University Press; 1984. [13] Webmaster PDI. (2004, August 20). Surface Metrology Guide. [www document]. Available from: http://www.predev.com/smg/ parameters.htm. [14] Yang JL, Chen JC. A systematic approach for identifying optimum surface roughness performance in end-milling operations. J Indus Technol 2001;17(2 ):18. [15] Lin TR. Experimental design and performance analysis of TiNcoated carbide tool in face milling stainless steel. J Mater Process Technol 2002;127:17.