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 />