Analysis of the influence of EDM parameters on surface quality, MRR and EW of WC–Co

Journal of Materials Processing Technology 153–154 (2004) 1026–1032<br /> <br /> Analysis of the influence of EDM parameters on surface<br /> quality, MRR and EW of WC–Co<br /> I. Puertas∗ , C.J. Luis, L. Álvarez<br /> Department of Mechanical Engineering (Manufacturing Engineering Section), Public University of Navarre (Spain),<br /> Campus de Arrosad´a s/n, Postal Code 31006 Pamplona, Navarre, Spain<br /> ı<br /> <br /> Abstract<br /> The adequate selection of manufacturing conditions is one of the most important aspects to take into consideration in the die-sinking<br /> electrical discharge machining (EDM) of conductive ceramics, as these conditions are the ones that are to determine such important<br /> characteristics as: surface roughness, electrode wear (EW) and material removal rate, among others. In this work, a study was carried out<br /> on the influence of the factors of intensity (I), pulse time (ti ) and duty cycle (η) over the listed technological characteristics. The ceramic<br /> used in this study was a cemented carbide or hard metal such as 94WC–6Co. Approximately 50% of all carbide production is used for<br /> machining applications but cemented carbides are also being increasingly used for non-machining applications, such as: mining, oil and gas<br /> drilling, metal forming and forestry tools. Accordingly, mathematical models will be obtained using the technique of design of experiments<br /> (DOE) to select the optimal machining conditions for finishing stages. This will be done only using a small number of experiments.<br /> © 2004 Elsevier B.V. All rights reserved.<br /> Keywords: EDM; Surface roughness; DOE<br /> <br /> 1. Introduction<br /> Electrical discharge machining (EDM) is a non-traditional<br /> manufacturing process based on removing material from a<br /> part by means of a series of repeated electrical discharges<br /> (created by electric pulse generators at short intervals) between a tool, called electrode, and the part being machined<br /> in the presence of a dielectric fluid [1]. At the present<br /> time, EDM is a widespread technique used in industry for<br /> high-precision machining of all types of conductive materials such as: metals, metallic alloys, graphite, or even some<br /> ceramic materials, of whatsoever hardness.<br /> In spite of their exceptional mechanical and chemical<br /> properties, ceramic materials have only achieved partial acceptance in the field of industrial applications, due to the<br /> difficulties of processing and the high cost associated with<br /> their manufacture. Over the past few years, the advances in<br /> the field of EDM made the application of this technology<br /> available for the manufacture of conductive ceramic materials. In line with current knowledge, the main inconvenience when applying the EDM technology to the treatment<br /> of ceramic materials is the electrical conductivity of these<br /> ∗ Corresponding author.<br /> E-mail address: inaki.puerta@unavarra.es (I. Puertas).<br /> <br /> 0924-0136/$ – see front matter © 2004 Elsevier B.V. All rights reserved.<br /> doi:10.1016/j.jmatprotec.2004.04.346<br /> <br /> materials. The threshold of electrical resistivity for these<br /> materials is considered to be between 100 and 300 cm<br /> [2,3].<br /> In this work, a study focused on the die-sinking EDM<br /> of a ceramic material such as cobalt-bonded tungsten carbide (WC–Co), whose field of applications is in constant<br /> growth, was carried out. Consequently, an analysis on the<br /> influence of intensity, pulse time and duty cycle over technological variables such as: surface roughness, electrode<br /> wear (EW) and material removal rate was performed. This<br /> was done using the technique of design of experiments<br /> (DOE) and techniques such as multiple linear regression<br /> analysis. The combined use of these techniques has allowed us to create both first and second-order models,<br /> which make it possible to explain the variability associated<br /> with each of the technological variables studied in this<br /> work.<br /> <br /> 2. Experimental<br /> In this section, there will be a brief description of the<br /> equipment used to carry out the EDM experiments, along<br /> with the ceramic material used and its dimensions. Also, the<br /> design factors used in this work will be outlined.<br /> <br /> I. Puertas et al. / Journal of Materials Processing Technology 153–154 (2004) 1026–1032<br /> <br /> Fig. 1. Die-sinking EDM machine used.<br /> <br /> 2.1. Equipment used in the experiments<br /> The equipment used in order to carry out the EDM of<br /> tungsten carbide was a die-sinking EDM machine of type<br /> ONA DATIC D-2030-S. Fig. 1 shows a photograph of this<br /> equipment.<br /> The EDM machine was equipped with a magnetic grip<br /> in order to hold the parts in place, but, due to the kind<br /> of material in question, a mechanical grip was finally used<br /> instead. In Fig. 2, a photograph illustrating the whole fixture<br /> system employed in the experiments is shown. In addition,<br /> in the same figure, the type of flushing used for the EDM<br /> experiments in order to assure an adequate removal of the<br /> debris from the work gap is shown. In this case, due to its<br /> simplicity and the shallow machining carried out in these<br /> experiments, jet flushing was selected. The pressure used for<br /> the dielectric fluid was 20 kPa.<br /> 2.2. Material used in the experiments<br /> Accordingly, the ceramic material used in this case was a<br /> 94WC–6Co, which means that it has a proportion of cobalt<br /> of 6%. This material was principally developed in order to<br /> be used as a cutting tool, due to its excellent hardness properties (HRA 92 for 94WC–6Co). Furthermore, this material<br /> possesses a high compressive strength (5450 MPa), as well<br /> as good resistance to wear and oxidation at high temperatures. On the other hand, the composite material made up of<br /> tungsten carbide and cobalt has a high resistance to thermal<br /> shock and, consequently, it is capable of coping with rapid<br /> changes of temperature. In the case of 94WC–6Co, this has a<br /> <br /> 1027<br /> <br /> Fig. 2. Gripping and jet flushing system used.<br /> <br /> thermal conductivity of 100 W/m K and a thermal expansion<br /> coefficient of 4.3 × 10−6 K−1 , at a temperature of 200 ◦ C.<br /> At present, although approximately 50% of the manufacturing of this type of cemented carbides is used in machining applications, the number of alternative applications<br /> is quickly growing. Among these alternative applications,<br /> could be highlighted: mining of metallic and non-metallic<br /> materials, construction, transport and drilling in oil and gas<br /> installations, metallic materials forming, structural components and, finally, as material for tools in the field of forestry<br /> engineering.<br /> The samples of tungsten carbide were ground sheets of<br /> the following dimensions: 50 mm × 50 mm × 4 mm. Moreover, the electrodes used were made of electrolytic copper<br /> (with negative polarity) given that, according to the bibliographic sources consulted [4,5], it is the most highly recommended material for the EDM process of tungsten carbide.<br /> Furthermore, the copper electrodes were selected in a prismatic form with a transverse area of 8 mm × 12 mm.<br /> 2.3. Design factors and response technological variables<br /> analysed<br /> There are a large number of design factors to be considered within the EDM process, but in this work we have only<br /> considered the level of the generator intensity (I), pulse time<br /> (ti ) and duty cycle (η) [1,6].<br /> The surface roughness parameter selected as response<br /> variable, defined in accordance with UNE-EN-ISO 4287:<br /> 1999, was the arithmetic average roughness of the roughness<br /> profile, that is to say, the Ra parameter.<br /> <br /> 1028<br /> <br /> I. Puertas et al. / Journal of Materials Processing Technology 153–154 (2004) 1026–1032<br /> <br /> When carrying out the roughness measurements over the<br /> ceramic sheets, a phase corrected 2CR filter for the rugosimeter, along with a length of measurement or evaluation<br /> of 6.4 mm (8 mm × 0.8 mm), were selected. The values of<br /> the surface roughness parameter for each experiment were<br /> obtained from the arithmetic mean of the values of the measurements taken following three parallel directions and in an<br /> equidistant distribution over the total area subjected to the<br /> EDM process.<br /> In addition to surface roughness, other very important<br /> response variables which are of interest when studying EDM<br /> processes, are material removal rate (MRR) and electrode<br /> wear, as shown in Eqs. (1) and (2).<br /> MRR =<br /> EW =<br /> <br /> volume of material removed from part<br /> time of machining<br /> <br /> volume of material removed from electrode<br /> volumen of material removed from part<br /> <br /> (1)<br /> (2)<br /> <br /> Although other ways of measuring MRR and EW do exist,<br /> in this work the material removal rate and electrode wear<br /> values have been calculated by the weight difference of the<br /> sample and electrode before and after undergoing the EDM<br /> process.<br /> <br /> 3. Design of the experiments<br /> The design which was finally chosen was a factorial design 23 with four central points [7], which provide protection against curvature, consequently carrying out a total of<br /> 12 experiments. For the case of the response variables which<br /> were not adequate for the previous model, this was widened<br /> by the addition of six star points, giving a central composite design made up of the star points situated in the centres<br /> of the faces. So, the case of the second-order model turned<br /> out to be made up of a total of 18 experiments, the previous<br /> 12 from the first-order model plus the six star points. The<br /> graphs presented here were done using STATGRAPHICS®<br /> plus, Version 5.0.<br /> Table 1 presents the relationship between the design factors and their corresponding selected variation levels, taking<br /> into account that the study wanted to focus on the finishing machining stages, owing to the influence which a good<br /> surface quality, in the case of ceramics, has over properties<br /> such as fatigue strength and wear. Consequently, the intensity levels chosen for the case of the intensity factor were 3<br /> <br /> I<br /> <br /> ti (␮s)<br /> <br /> η<br /> <br /> Ra (␮m)<br /> <br /> EW (%)<br /> <br /> MRR<br /> (mm3 /min)<br /> <br /> 1<br /> 2<br /> 3<br /> 4<br /> 5<br /> 6<br /> 7<br /> 8<br /> 9<br /> 10<br /> 11<br /> 12<br /> 13<br /> 14<br /> 15<br /> 16<br /> 17<br /> 18<br /> <br /> 3<br /> 5<br /> 3<br /> 5<br /> 3<br /> 5<br /> 3<br /> 5<br /> 3<br /> 5<br /> 4<br /> 4<br /> 4<br /> 4<br /> 4<br /> 4<br /> 4<br /> 4<br /> <br /> 10.0<br /> 10.0<br /> 50.0<br /> 50.0<br /> 10.0<br /> 10.0<br /> 50.0<br /> 50.0<br /> 30.0<br /> 30.0<br /> 10.0<br /> 50.0<br /> 30.0<br /> 30.0<br /> 30.0<br /> 30.0<br /> 30.0<br /> 30.0<br /> <br /> 0.4<br /> 0.4<br /> 0.4<br /> 0.4<br /> 0.6<br /> 0.6<br /> 0.6<br /> 0.6<br /> 0.5<br /> 0.5<br /> 0.5<br /> 0.5<br /> 0.4<br /> 0.6<br /> 0.5<br /> 0.5<br /> 0.5<br /> 0.5<br /> <br /> 0.66<br /> 1.24<br /> 0.66<br /> 1.93<br /> 0.65<br /> 1.37<br /> 0.64<br /> 2.02<br /> 0.64<br /> 1.75<br /> 1.11<br /> 1.73<br /> 1.43<br /> 1.54<br /> 1.43<br /> 1.34<br /> 1.39<br /> 1.42<br /> <br /> 36.51<br /> 15.32<br /> 71.58<br /> 12.09<br /> 35.25<br /> 15.54<br /> 71.77<br /> 11.09<br /> 66.41<br /> 11.02<br /> 17.34<br /> 15.21<br /> 14.93<br /> 13.35<br /> 14.29<br /> 14.15<br /> 14.08<br /> 14.09<br /> <br /> 0.055<br /> 0.248<br /> 0.012<br /> 0.202<br /> 0.090<br /> 0.428<br /> 0.020<br /> 0.318<br /> 0.026<br /> 0.279<br /> 0.236<br /> 0.155<br /> 0.147<br /> 0.246<br /> 0.193<br /> 0.192<br /> 0.194<br /> 0.194<br /> <br /> (2 A) and 5 (6 A). On the other hand, levels of 10 and 50 ␮s<br /> as well as levels of 0.4 and 0.6 were selected for pulse time<br /> and duty cycle, respectively. The reason for using levels of<br /> intensity instead of intensities is due to the requirements of<br /> the programming of the EDM equipment.<br /> Table 2 shows the design matrix resulting from the type of<br /> experiment selected in the case of the second-order model,<br /> as well as the observations for the case of the three response<br /> variables which are considered, where the intensity values<br /> 3, 4 and 5 are equivalent to 2 A, 4 A and 6 A, respectively.<br /> The design matrix for the first-order model is obtained by<br /> merely eliminating the rows corresponding to the star points<br /> of the model, which means rows 9–14, inclusive.<br /> <br /> 4. Analysis of surface roughness<br /> A first-order model was proposed for the response variable of the Ra parameter, but this was rejected as a result of the variance analysis for the proposed curvature test.<br /> On the other hand, a value of 0.9843 is obtained for the<br /> R squared statistic (R2 ) and the adjusted R squared statis¯<br /> tic (R2 ) is 96.67%, for the case of the new second-order<br /> model. Moreover, the equation for this model is shown in<br /> Eq. (3):<br /> Ra = −3.04141 + 2.19567I − 0.0168304ti − 3.61815η<br /> −0.0000934524ti2 − 0.003125ti η + 2.7619η2<br /> <br /> (3)<br /> <br /> Levels<br /> −1<br /> <br /> I<br /> ti (␮s)<br /> η<br /> <br /> Number of<br /> experiment<br /> <br /> −0.262381I 2 + 0.0084375Iti + 0.3125Iη<br /> <br /> Table 1<br /> Factors and levels selected for the experiments<br /> Factors<br /> <br /> Table 2<br /> Design of experiment matrix for the second-order model<br /> <br /> +1<br /> <br /> 3<br /> 10<br /> 0.4<br /> <br /> 5<br /> 50<br /> 0.6<br /> <br /> where the values of the considered variables have been specified according to their original units.<br /> Fig. 3 shows the Pareto chart for the effects corresponding to the Ra parameter. As can be clearly seen, all the bars<br /> of the diagram which go beyond the vertical line correspond<br /> <br /> I. Puertas et al. / Journal of Materials Processing Technology 153–154 (2004) 1026–1032<br /> <br /> Fig. 3. Pareto chart for the effects of Ra .<br /> <br /> to the effects which are statistically significant, for a confidence level of 95%. Therefore, there are four significant<br /> effects which, in descending order of contribution, are: the<br /> factor of intensity, the factor of pulse time, the interaction<br /> of these two factors and, finally, the pure quadratic effect of<br /> intensity.<br /> As can be seen in Fig. 4, the values of the Ra parameter increases significantly with the factor of intensity, which<br /> also happens to be the most significant effect of all. On the<br /> other hand, the Ra parameter also tends to increase with<br /> the factors of pulse time and duty cycle, this last factor,<br /> as was seen earlier, not being significant for the 95% confidence level. The performance of the Ra parameter faced<br /> with the three design factors is what would be expected, a<br /> priori, as generally, the increase of energy in each pulse,<br /> be it through an increase in the value of intensity or pulse<br /> time, or the decrease in pause time, which is equivalent in<br /> this case, to an increase in the value of duty cycle, usually leads to a poorer surface roughness of the manufactured<br /> parts.<br /> Fig. 5 shows the estimated response surface for the Ra<br /> parameter, according to the design parameters of intensity<br /> and pulse time, whilst the duty cycle factor remains constant<br /> in its central value of 0.5.<br /> As has been previously pointed out, this figure shows us<br /> the important influence that the design factor of intensity<br /> possesses over the Ra parameter, so that when intensity is<br /> increased, the Ra parameter also tends to increase appreciably at least up to a maximum value, after which it tends to<br /> decrease, for low values of the pulse time factor and within<br /> <br /> 1029<br /> <br /> Fig. 5. Estimated response surface of Ra vs. I and ti .<br /> <br /> Fig. 6. Estimated response surface of Ra vs. I and η.<br /> <br /> the considered work interval. Furthermore, it can also be<br /> observed that the Ra parameter tends to increase when the<br /> pulse time factor is increased, especially for high values of<br /> intensity. The previous tendency of growth for this factor<br /> becomes less intense as we move towards lower values of<br /> intensity, with the Ra parameter actually decreasing slightly,<br /> after reaching a peak, for values close to the low level of<br /> intensity (I = 3).<br /> Fig. 6 shows the estimated response surface of the Ra parameter according to the factors of intensity and duty cycle,<br /> whilst the pulse time factor remains constant in its central<br /> level value, which is 30 ␮s.<br /> As has been previously stated, the graph in Fig. 6 shows<br /> the great influence that the intensity factor has over the Ra<br /> parameter, in such a way that it tends to increase significantly<br /> with the intensity factor, for any value of duty cycle and<br /> within the considered work interval. Furthermore, the graph<br /> also shows the lack of influence that, in this case, duty cycle<br /> has over the Ra parameter.<br /> <br /> 5. Analysis of electrode wear<br /> <br /> Fig. 4. Graph of the main effects of Ra .<br /> <br /> A first-order model was proposed for the response variable of electrode wear such as the one proposed for the<br /> case of the Ra parameter, but this was rejected after observing the results for the curvature test. With the second-order<br /> model, the values obtained for the R squared statistic and<br /> the adjusted R squared statistic were 0.9692 and 0.9345, respectively. The equation for the adjusted model is shown in<br /> <br /> 1030<br /> <br /> I. Puertas et al. / Journal of Materials Processing Technology 153–154 (2004) 1026–1032<br /> <br /> Fig. 9. Estimated response surface of EW vs. I and ti .<br /> <br /> Fig. 7. Pareto chart for the effects of EW.<br /> <br /> Eq. (4):<br /> EW = 315.152 − 181.578I + 2.37548ti + 269.637η<br /> +21.8262I 2 − 0.495464Iti + 0.370479Iη<br /> −0.00153199ti2 + 0.0144416ti η − 274.99η2<br /> <br /> (4)<br /> <br /> where the values of the considered variables have been specified according to their original units and, as in the previous<br /> case, no kind of simplification has been done.<br /> Fig. 7 represents the Pareto chart for the effects corresponding to the EW response variable. As can be clearly<br /> observed in this diagram, there are four significant effects<br /> which, in descending order of significance, are: the main effect of intensity, the pure quadratic effect of intensity, the<br /> interaction effect of intensity and pulse time and, finally, the<br /> main effect of pulse time, whilst the rest of the factors are<br /> not significant for a confidence level of 95%.<br /> Fig. 8 shows the graph of the main effects for each of the<br /> factors which have been considered in this study. As can be<br /> seen in Fig. 8, the most influential factor over EW is intensity, in such a way that the value of the wear decreases<br /> greatly when intensity is increased, at least down to a minimum value after which the value of EW begins to grow.<br /> This tendency is what would be expected, a priori, as higher<br /> values for the intensity factor usually lead to lower values<br /> of electrode wear.<br /> With regard to the pulse time factor, as it is shown in the<br /> above figure, the value of EW tends to increase when this<br /> factor is increased, within the studied work interval. This<br /> tendency is exactly the opposite of what was expected, as an<br /> <br /> Fig. 10. Estimated response surface of EW vs. I and η.<br /> <br /> increase in pulse time is usually associated with a decrease<br /> in electrode wear.<br /> Furthermore, the value of EW tends to increase with duty<br /> cycle up to a maximum value, situated approximately at its<br /> central value, after which it starts to decrease, although, on<br /> the other hand, the effect of this factor is not statistically<br /> significant for the considered confidence level. In general<br /> practice with other types of materials, an increase in duty<br /> cycle reducing the pause time is usually associated with a<br /> reduction in electrode wear as well.<br /> Fig. 9 shows the estimated response surface of electrode<br /> wear, varying the factors of intensity and pulse time. As can<br /> be clearly seen in this figure, the wear value tends to decrease<br /> with the intensity factor down to a minimum value after<br /> which it tends to increase. Moreover, this minimum value<br /> of wear moves gradually towards higher intensity values as<br /> the pulse time value increases, within the variation interval<br /> which was considered in this work.<br /> Fig. 10 shows the estimated response surface of EW in<br /> function of the factors of intensity and duty cycle, whilst<br /> the pulse time factor remains constant in its central value of<br /> 30 ␮s. This figure shows the limited influence that duty cycle<br /> possesses over electrode wear for any value of intensity, if<br /> the value of pulse time is fixed in its central level.<br /> <br /> 6. Analysis of material removal rate<br /> <br /> Fig. 8. Graph of the main effects of EW.<br /> <br /> For the response variable of material removal rate (MRR),<br /> initially a first-order model was proposed, but this was rejected after considering the results obtained for the model<br /> <br />