Statistical analysis on electrode wear in EDM of tool steel DIN 1.2714 used in forging dies

Journal of Materials Processing Technology 187–188 (2007) 711–714<br /> <br /> Statistical analysis on electrode wear in EDM of tool steel<br /> DIN 1.2714 used in forging dies<br /> H. Zarepour a,∗ , A. Fadaei Tehrani b , D. Karimi c , S. Amini a<br /> a<br /> <br /> Department of Manufacturing Engineering, School of Engineering (2),<br /> Islamic Azad University, Najafabad Branch, Isfahan, Iran<br /> b Department of Mechanical Engineering, Isfahan University of Techology, Isfahan, Iran<br /> c Department of Mechanical Engineering, Tarbiat Modarres University, Tehran, Iran<br /> <br /> Abstract<br /> Electrical discharge machining (EDM) is a widespread process which works very effectively in machining of difficult-to-cut materials and alloys<br /> in die and aerospace industries with high dimensional accuracies. However, this capability could be deteriorated due to electrode wear leading to<br /> decrease of process productivity. In this study, the effect of machining parameters of EDM process including on-time, current, voltage, the engaging<br /> time between workpiece and electrode, and pre-EDM roughing on electrode wear were experimentally investigated. Main effects of factors and<br /> interactions were considered in this paper and regression equation was derived. A L50 (21 × 511 ) Taguchi’s standard orthogonal array was employed<br /> as experimental design. Copper was used as electrode to machine the hot work tool steel 1.2714, which is widely used to make forging dies and<br /> mandrels.<br /> © 2007 Elsevier B.V. All rights reserved.<br /> Keywords: EDM; Electrode wear; Orthogonal array; Statistical analyses<br /> <br /> 1. Introduction<br /> Among the non-traditional methods of material removal processes, electrical discharge machining (EDM) has drawn a great<br /> a deal of researchers’ attention because of its broad industrial<br /> applications. This process is well suited for machining of casting<br /> and forging dies, powder metallurgy and injection molds, and<br /> aerospace parts.<br /> The process is a spark erosion method, eroding the workpiece by high frequency spark discharges [1]. EDM has a high<br /> capability of machining the accurate cavities of dies and molds.<br /> Nevertheless, electrode wear occurs during EDM process leading to a lack of machining accuracy in the geometry of workpiece<br /> [2]. Furthermore, electrode wear imposes high costs on manufacturers to substitute the eroded complicated electrodes by<br /> new ones for die making. In order to increase the machining<br /> efficiency, erosion of the workpiece must be maximized and<br /> that of the electrode minimized in EDM process [1]. Therefore,<br /> <br /> ∗<br /> <br /> Corresponding author. Tel.: +98 312 37 58930; fax: +98 312 37 56900.<br /> E-mail address: h-zare@iaun.ac.ir (H. Zarepour).<br /> <br /> 0924-0136/$ – see front matter © 2007 Elsevier B.V. All rights reserved.<br /> doi:10.1016/j.jmatprotec.2006.11.202<br /> <br /> studying the electrode wear and related significant factors would<br /> be effective to enhance the machining productivity and process<br /> reliability.<br /> A series of investigations have been conducted on electrode<br /> wear in EDM process. Soni and Chakraverti [3,4] studied the surface quality, material removal rate, wear ratio, and dimensional<br /> accuracy in EDM of alloy steels. Singh et al. investigated the<br /> effect of machining parameters on electrode wear in die-sinking<br /> EDM of En-31 tool steel with different electrode materials [5].<br /> Also, Luis et al. have carried out a study on electrode wear<br /> in EDM of silicon carbide using the technique of design of<br /> experiments [6].<br /> In this research, an experimental study is conducted to investigate on electrode wear in die-sinking EDM of DIN 1.2714 hot<br /> work tool steel. The selection of this material was made taking into account its wide range of applications in die and mold<br /> industries such as hammer and hydraulic forging dies. The aim<br /> of this study is to find out from which process parameters (factors) electrode wear is affected and to which factor interactions it<br /> is related. This is usually done by means of analysis of variance<br /> (ANOVA). Furthermore, regression analysis is used to establish the correlation between factors and response (tool wear).<br /> The appropriate degree of the polynomial regression equation is<br /> <br /> 712<br /> <br /> H. Zarepour et al. / Journal of Materials Processing Technology 187–188 (2007) 711–714<br /> <br /> found which is thought to be useful assessment of the predictive<br /> equation [7,8]. Finally, the optimal factor levels are obtained.<br /> 2. Experimentation<br /> 2.1. Experimental design<br /> During EDM experiments, the input parameters (factors)<br /> were on-time (pulse-on duration), peak current, pulsed voltage,<br /> engaging time, and pre-EDM roughing accuracy. The pulse-off<br /> duration was kept constant which could effectively control the<br /> flushing of the debris from the gap, giving machining stability. Therefore, the effect of the pulse-off duration on machining<br /> process was not considered in the present work. Table 1 shows<br /> factors and factor levels assessed in this study.<br /> The design of experiments was performed through Taguchi<br /> approach of experimental design. Accordingly, a L50 standard<br /> orthogonal array was employed. All the main effects of factors<br /> and two-order interactions were desired to be considered in this<br /> study.<br /> 2.2. Experimental equipment and setup<br /> Fifty DIN 1.2714 steel specimens were machined for 50<br /> experimental runs in this work. The holes were created in the<br /> workpieces in five groups of dimensions as roughed cavities<br /> before EDM process. The allowances left for EDM die-sinking<br /> were considered equal to 0.25, 0.5, 0.75, 1, and 1.25 mm (preEDM rouging accuracy) for each group of dimensions. These<br /> dimensions are referred as 1, 2, 3, 4, and 5, respectively, in<br /> Table 1 and analyses. Hardness of the specimens was increased<br /> up to 48–50 HRC, nearly the same as the hardness of forging<br /> dies.<br /> Cylindrical copper rods with 99% purity and 8.98 g/cm3<br /> density were machined with good surface finish and exact<br /> dimensions as tool electrodes. Fig. 1 illustrates a sample of copper electrodes. Both electrodes and workpieces were deburred<br /> carefully providing stable conditions in EDM process. The specimens were machined on a “Pishtazan-Pulse Generator 120”<br /> die-sinking EDM machine which has an iso-pulse generator with<br /> a maximum 120 A current intensity.<br /> The experiments were carried out in petroleum-Behran oil 42<br /> dielectric (10:1, v/v) covering the workpiece by 10 mm. Final<br /> dimensions of machined holes in workpieces were exactly the<br /> same for all experimental runs. The electrode was negative polarity and the specimen was positive polarity during the EDM<br /> process. Weight of electrodes was measured after experiments<br /> <br /> Fig. 1. A sample of copper electrodes and related dimensions.<br /> <br /> using an accurate digital balance and the percentage of electrodes<br /> weight reduction was calculated as tool wear (wt.%).<br /> 3. Data analysis<br /> 3.1. Effect of factors and interactions on electrode wear<br /> Fig. 2 depicts the plot of main effects for electrode wear. Note<br /> that this plot illustrates data means versus factor levels. Based<br /> on this plot, the effect of each factor can be graphically assessed.<br /> Fig. 2a shows that pre-EDM roughing factor has a significant<br /> effect on electrode wear. It can also be seen from this figure that<br /> the effect of this factor is directly proportional to electrode wear.<br /> Also, it can be stated that by increasing the pre-EDM roughing<br /> factor electrode wear increases significantly. Fig. 2b portraits the<br /> effect of current on electrode wear. This figure presents that electrode wear is almost constant when current is increased from 5 to<br /> 10 A. By increasing the current form 10 to 20 A, electrode wear<br /> increases dramatically. Finally, one can interpret that current<br /> has a significant direct impact on electrode wear. Further experiments are needed to consider effect of current on electrode wear<br /> when current fluctuates between 5 and 15 A. Fig. 2c indicates<br /> the effect of on-time on electrode wear. This figure shows that<br /> on-time is reciprocally proportional to electrode wear. Fig. 2d<br /> suggests that voltage has a subtle effect on electrode wear. It<br /> is obvious from this figure that by changing voltage during this<br /> process electrode wear fluctuates insignificantly between 0.3 and<br /> 0.4%. So, it is concluded that the effect of voltage on electrode<br /> wear is almost negligible. Eventually, Fig. 2e shows the effect<br /> <br /> Table 1<br /> Factors and factor levels<br /> Factor<br /> <br /> Level 1<br /> <br /> Level 2<br /> <br /> Level 3<br /> <br /> Level 4<br /> <br /> Level 5<br /> <br /> On-time, A (␮s)<br /> Current, B (A)<br /> Voltage, C (V)<br /> Engaging time, D (s)<br /> Pre-EDM roughing, E<br /> <br /> 200<br /> 5<br /> 35<br /> 1<br /> 1<br /> <br /> 500<br /> 10<br /> 40<br /> 2<br /> 2<br /> <br /> –<br /> 15<br /> 50<br /> 3<br /> 3<br /> <br /> –<br /> 20<br /> 55<br /> 4<br /> 4<br /> <br /> –<br /> –<br /> –<br /> –<br /> 5<br /> Fig. 2. (a–e) Plot of main effects on electrode wear.<br /> <br /> H. Zarepour et al. / Journal of Materials Processing Technology 187–188 (2007) 711–714<br /> <br /> 713<br /> <br /> Table 2<br /> Values of estimated effects<br /> Term<br /> <br /> Effect<br /> <br /> Term<br /> <br /> Effect<br /> <br /> A<br /> B<br /> C<br /> D<br /> E<br /> A×B<br /> A×C<br /> A×D<br /> <br /> −0.1359<br /> 0.4351<br /> 0.0188<br /> −0.0909<br /> 0.2910<br /> −0.2059<br /> 0.0333<br /> −0.0610<br /> <br /> A×E<br /> B×C<br /> B×D<br /> B×E<br /> C×D<br /> C×E<br /> D×E<br /> <br /> −0.1710<br /> −0.0945<br /> 0.0440<br /> 0.2674<br /> 0.0582<br /> −0.0032<br /> 0.0018<br /> <br /> of engaging time between electrode and workpiece on electrode<br /> wear. This figure shows that the engaging time exhibits little<br /> effect on electrode wear. But unlike voltage, by increasing the<br /> engaging time electrode wear first increases and then stalls off<br /> insignificantly.<br /> Factor effects are calculated here to consider effect of factors<br /> on electrode wear more accurately than graphical assessment.<br /> Table 2 shows the magnitude of factor effects. Using this table<br /> one can sort factors in order of their significance. Also, this table<br /> illustrates the proportionality of factor effects to the response<br /> (electrode wear). Positive values of factor effects indicate direct<br /> proportionality of the respective factors, while negative values demonstrate that the corresponding factor is reciprocally<br /> proportional to the response.<br /> Note that two-way interactions are also included in Table 2.<br /> It can be seen from this table that current (B) has the most significant effect on electrode wear and is directly proportional to it.<br /> The interaction between engaging time and pre-EDM roughing<br /> (D × E) has the least effect on electrode wear and is directly proportional to it. One shortcoming to Table 2 is that the inferences<br /> made based on this table are only comparatively valid. To cope<br /> with the problem analysis of variance (ANOVA) is numerously<br /> used by experimenters. But ANOVA is not employed here since<br /> error normality and error independency assumptions are not satisfied. For more information on ANOVA assumptions consult<br /> reference [7].<br /> Normal probability plot of standardized effects is used here<br /> as an alternative to ANOVA approach. Although electrode wear<br /> does not follow the normal distribution, estimated effects can be<br /> assumed to exhibit normal distribution [9].<br /> Normal probability plot of standardized effects is shown in<br /> Fig. 3 to separate significant factors from insignificant factors.<br /> Factors and interactions which have departed considerably from<br /> the straight line have significant impact on electrode wear.<br /> 3.2. Regression analysis<br /> A quadratic polynomial regression equation, in terms of factors (Table 1), is obtained to fit experimental data. Least square<br /> method is used in regression analysis to find the coefficients of<br /> the equation. It is shown as Eq. (1).<br /> Wear (%) = 0.4895 + 0.0007A + 0.0068B − 0.0268C<br /> + 0.08049D − 0.1185E + 0.0016B2 + 0.0003C2<br /> <br /> Fig. 3. Normal probability plot of effects.<br /> <br /> + 0.0026D2 − 0.0037E2 − 0.0001AB<br /> − 0.0000004AC − 0.0002AD + 0.00007AE<br /> − 0.00013BC + 0.0072BD − 0.0022BE<br /> − 0.00072CD + 0.0030CE − 0.0056DE<br /> <br /> (1)<br /> <br /> Note that A2 has been removed from the equation since it is<br /> highly correlated with other variables.<br /> Regression statistics R2 and R2 are obtained equal to 98.7<br /> Adj<br /> and 97.8%, respectively. The R2 value indicates that the predictors explain 98.7% of variance in electrode wear. The R2 value<br /> Adj<br /> accounts for the number of predictors in the model. Both values indicate that the presented model fits the data very well.<br /> The analysis of variance for regression analysis is shown in<br /> Table 3. The p-value shows that the model is significant at α-level<br /> of 0.05.<br /> Model adequacy is checked by means of plot of residuals<br /> versus fits, plot of residuals versus order of the data, and normal<br /> plot of residuals. Fig. 4 shows the plot of residuals versus fits.<br /> It is clearly observed from this plot that residuals have constant<br /> variance. Plot of residuals versus order of the data is illustrated<br /> in Fig. 5. This figure indicates that residuals are independent of<br /> one another.<br /> Also, normal plot of residuals is shown in Fig. 6.<br /> Anderson–Darling statistic (AD) and the p-value calculated<br /> based on it, are shown in this figure. It is easily found out<br /> that residuals are normally distributed. The presented discussion<br /> implies that the predictive model is adequate.<br /> <br /> Table 3<br /> ANOVA table for regression analysis<br /> Source<br /> <br /> DF<br /> <br /> SS<br /> <br /> MS<br /> <br /> F<br /> <br /> p<br /> <br /> Regression<br /> Residual error<br /> <br /> 19<br /> 28<br /> <br /> 4.21979<br /> 0.05659<br /> <br /> 0.22209<br /> 0.00202<br /> <br /> 109.90<br /> <br /> 0.000<br /> <br /> Total<br /> <br /> 47<br /> <br /> 4.27638<br /> <br /> DF: degrees of freedom, SS: sum of Squares, MS: mean squares, F: F-value, p:<br /> p-value.<br /> <br /> 714<br /> <br /> H. Zarepour et al. / Journal of Materials Processing Technology 187–188 (2007) 711–714<br /> <br /> of all compared with effects of noises (uncontrollable factors).<br /> S/N ratio statistic (η) can be obtained by Eq. (2):<br /> η = −10 log10<br /> <br /> 1<br /> n<br /> <br /> n<br /> i=1<br /> <br /> 2<br /> yi<br /> <br /> (2)<br /> <br /> where yi is the ith observation of a treatment combination and n<br /> is the number of replications.<br /> The factor level which produces the largest η is detected as<br /> the factor level which pertains to the optimal condition. Accordingly, the optimal levels of A, B, C, D, and E factors would be<br /> equal to 2, 1, 1, 4, and 1, respectively. The corresponding value<br /> of each factor level can be found out referring to Table 1.<br /> 4. Conclusions<br /> Fig. 4. Plot of residuals vs. fits.<br /> <br /> In this research, an experimental investigation was performed<br /> to consider the electrode wear in EDM process of DIN 1.2714<br /> tool steel and the following results were concluded:<br /> 1. On-time, current, and pre-EDM roughing as factors along<br /> with on-time/current, on-time/pre-EDM roughing, and<br /> current/pre-EDM roughing as interactions, were found to<br /> have significant effect on electrode wear.<br /> 2. The predictive model was derived out of regression analysis.<br /> 3. The adequacy of predictive model was approved.<br /> 4. Optimal level of factors was presented using S/N ratio analysis.<br /> Acknowledgements<br /> <br /> Fig. 5. Plot of residuals vs. the order of the data.<br /> <br /> The authors appreciate Abzaran Co. and Industrial Cooperation Office of Isfahan University of Technology for supporting<br /> this research. Also, a special thanks to M. Pourehsan and M.<br /> Neshatpour for their close cooperation.<br /> References<br /> <br /> Fig. 6. Normal plot of residuals.<br /> <br /> 3.3. Determination of the optimal condition<br /> Optimal condition is detected by means of S/N ratio method.<br /> The rationale behind this method is to find a condition under<br /> which the effect of signals (controllable factors) is the greatest<br /> <br /> [1] C.C. Liu, Microstructure and tool electrode erosion in EDM of TiN/Si3 N4<br /> composites, Mater. Sci. Eng. J. 363 (2003) 221–227.<br /> [2] Y.Y. Tsai, T. Masuzawa, An index to evaluate the wear resistance of the electrode in micro-EDM, J. Mater. Process. Technol. 149 (1–3) (2004) 304–309.<br /> [3] J.S. Soni, G. Chakraverti, Investigative study on metal removal rate and wear<br /> ratio in EDM of high carbon high chromium die steel, J. Ind. Eng. 71 (1991).<br /> [4] J.S. Soni, G. Chakraverti, Effect of electrode material properties on surface<br /> roughness and dimensional accuracy in electro-discharge machining of high<br /> carbon high chromium die steel, J. Ind. Eng. 76 (1995) 46–51.<br /> [5] S. Singh, S. Maheshwari, P.C. Pandey, Some investigations into the electric discharge machining of hardened tool steel using different electrode<br /> materials, J. Mater. Process. Technol. 149 (2004) 272–277.<br /> [6] C.J. Luis, I. Puertas, G. Villa, Material removal rate and electrode wear study<br /> on the EDM of silicon carbide, J. Mater. Process. Technol. 164–165 (2005)<br /> 889–896.<br /> [7] D.C. Montgomery, Design and Analysis of Experiments, John Wiley & Sons,<br /> 2000.<br /> [8] H.M. Raymond, E.W. Ronald, Probability and Statistics for Engineers and<br /> Scientists, Macmillan Publishing Co., Inc., New York, 1978.<br /> [9] R.L. Mason, R.F. Gunt, J.L. Hess, Statistical Design and Analysis of Experiments, John Wiley & Sons, 2003.<br /> <br />