Journal of Materials Processing Technology 153–154 (2004) 1033–1038

A study on the electrical discharge machining of conductive ceramics

I. Puertas∗ , C.J. Luis
Department of Mechanical Engineering, Manufacturing Engineering Section, Public University of Navarre (Spain),
Campus de Arrosadı́a s/n, Pamplona (Navarre), Navarre 31006, Spain

Abstract

In this work, a comparative study of the electrical discharge machining (EDM) of three different conductive ceramics has been carried out.
The ceramics used in the experiments were cobalt-bonded tungsten carbide (WC–Co), hot-pressed boron carbide (B4 C) and reaction-bonded
silicon carbide (SiSiC). This study has been done only for the finishing stages and the selection of the above-mentioned conductive ceramics
was motivated by their wide range of applications: machining and mining for WC–Co, thermal-neutron absorber and nozzles for B4 C
and wear parts and heat exchangers for SiSiC, among others. The comparative study has been carried out on the influence of the design
factors of intensity (I), pulse time (ti ) and duty cycle (η) over response variables such as surface roughness (evaluated through the Ra
parameter), volumetric electrode wear (EW) and material removal rate (MRR). This was accomplished through the technique of design of
experiments (DOE), along with multiple linear regression techniques, which allowed us to carry out the analysis with only a small number
of experiments.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Die-sinking EDM; Conductive ceramics; Surface roughness; DOE

1. Introduction terials. In line with current knowledge, the main inconve-

nience when applying the EDM technology to the field of
At present, EDM is a widespread technique used in ceramic materials, is the electrical resistivity of these mate-
industry for high-precision machining of all types of con- rials, where the limits are fixed between 100 and 300  cm
ductive materials such as metals, metallic alloys, graphite, [3].
or even some ceramic materials, of whatever intensity of In this work, a comparative study has been carried out cen-
hardness [1]. tred solely on the experience of the die-sinking EDM of three
The term, technical ceramic materials or advanced ce- different ceramic materials and for finishing stages. The
ramic materials, is a relatively new term, which is applied selected materials were hot-pressed boron carbide (B4 C),
to a range of various materials generally obtained from reaction-bonded silicon carbide (SiSiC) and cobalt-bonded
inorganic primary materials with a high grade of purity. tungsten carbide (WC–Co), whose fields of applications are
These primary materials are subjected to typical processes in constant growth. Accordingly, a comparative analysis has
in powder metallurgy technology and afterwards, to high been carried out on the influence of intensity, pulse time and
temperature sintering processes. With these materials, it is duty cycle on important technological variables such as sur-
possible to obtain high-density parts which have good tech- face roughness (Ra ), volumetric electrode wear (EW) and
nical properties related to hardness, mechanical resistance, material removal rate (MRR). This was accomplished us-
wear and corrosion at high temperatures [2]. ing the technique of design of experiments (DOE) and tech-
In spite of their exceptional mechanical and chemical niques of multiple linear regression analysis.
properties, ceramic materials have only achieved a partial
acceptance in the field of industrial applications due to the
difficulties in processing and the high costs associated with 2. Description of the experiments
their manufacture. Over the past few years, the advances
in the field of EDM have permitted the application of this 2.1. EDM machine used
technology to the manufacture of conductive ceramic ma-
The equipment used was a die-sinking EDM machine of
∗ Corresponding author. type ONA DATIC D-2030-S. The only modification which
E-mail address: [email protected] (I. Puertas). was made to the original equipment was the adaptation of a

0924-0136/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2004.04.197
1034 I. Puertas, C.J. Luis / Journal of Materials Processing Technology 153–154 (2004) 1033–1038

microfine control, which allowed us to obtain a mirror-type 2.4. Ceramic material of tungsten carbide (WC–Co)
surface finish for the case of steels. This additional equip-
ment consists of a series of stages formed by capacitors, The third material used in the experiments was 94WC–6Co
which provide variable capacity between the part and the tungsten carbide. This material was principally developed
electrode. in order to be used as cutting tool, due to its excellent
hardness properties (HRA 92). Furthermore, it possesses
a high compressive strength (5450 MPa), as well as good
2.2. Ceramic material of boron carbide (B4 C)
resistance to wear and oxidation at high temperatures. On
the other hand, the compound material made up of tung-
The first ceramic material used in the experiments was
sten carbide and cobalt has a high resistance to thermal
hot-pressed boron carbide (B4 C). It has excellent properties
shock and, consequently, it is capable of coping with rapid
of hardness (37 GPa), resistance to wear as well as fracture
changes of temperature. In the case of 94WC–6Co, it has a
toughness (3–4 MPa m1/2 ). Its density is low (2.50 g/cm3 )
thermal conductivity of 100 W/m K and a thermal expansion
but, on the other hand, it has a high compressive strength
coefficient of 4.3 × 10−6 K−1 , at a temperature of 200 ◦ C.
(2855 MPa) and a high module of elasticity (445 GPa), which
At present, although approximately 50% of the manufac-
makes its use as ballistic material one of its widest applica-
turing of this type of cemented carbides is used in machin-
tions. As 20% of natural boron is made up of the isotope of
ing applications, the number of alternative applications is
boron B10 , it presents a high neutron capture cross-section
quickly growing, where the most important areas are mining,
and so is usually used in neutron absorption applications
construction, transport and drilling in oil and gas installa-
such as lining in nuclear fission reactors.
tions, metal forming, structural components and as material
The samples of boron carbide used in the experi-
for tools in the field of forestry engineering.
ments were ground sheets of the following dimensions:
The samples of tungsten carbide were ground sheets of
50 mm × 50 mm × 4 mm. Moreover, the electrodes used
dimensions: 50 mm × 50 mm × 4 mm. As in the two pre-
were made of electrolytic copper as it is the most highly
vious cases, the electrodes used were made of electrolytic
recommended material for the EDM process of boron car-
copper because, according to the bibliographic sources con-
bide [4,5]. The selected copper electrodes (with positive
sulted [7,8], it is the most highly recommended material for
polarity for B4 C) used in this study for all the ceramic ma-
the EDM of tungsten carbide. Furthermore, it is also rec-
terials were of a prismatic shape and with a cross section of
ommended that negative polarity for the electrode should
8 mm × 12 mm.
be used in order to ensure a much more stable machining
process.
2.3. Ceramic material of silicon carbide (SiSiC)
2.5. Design factors
The second ceramic material used in this work was
reaction-bonded or siliconised silicon carbide (REFEL F® ). There are a large number of factors to consider within
The mechanical resistance (2000–3500 MPa to compres- the EDM process, but in this work we have only considered
sion and 310 MPa to tension) combined with a hardness the generator intensity (I), pulse time (ti ) and duty cycle
(25–35 GPa, Vickers hardness) which is higher than that (η). The reason why these three factors have been selected
of tungsten carbide explains its use as an element in high as design factors is that they are the most widespread and
temperature gas turbines as well as forming part of bear- frequently-used ones among EDM researchers.
ings and seals. Furthermore, it has a high thermal con-
ductivity (150–200 W m−1 K−1 , at 20 ◦ C) and a thermal 2.6. Response technological variables
expansion coefficient (4.3–4.6 × 10−6 K−1 , for an interval
of 20–1000 ◦ C) which gives it good resistance to thermal The response variables studied in this work were surface
shock. Reaction-bonded silicon carbide performs better un- roughness, material removal rate and electrode wear. The
der chemical corrosion than other ceramic materials such previous variables were chosen because of their importance
as tungsten carbide or alumina (Al2 O3 ). It is therefore within the EDM research field.
frequently used as lining for industrial furnaces. The surface roughness parameter which was selected as
The samples of silicon carbide were ground sheets of a response variable was the arithmetic average roughness of
dimensions: 50 mm × 50 mm × 5 mm. As in the case of the roughness profile (Ra ). This surface roughness parame-
the earlier ceramic material, the electrodes used were made ter is defined in accordance with UNE-EN-ISO 4287: 1999.
of electrolytic copper, as, according to the bibliographic When carrying out the surface roughness measurements over
sources consulted [6,7], it is the most highly recommended the ceramic materials, a phase corrected 2CR filter was se-
material for the EDM of silicon carbide. Conversely, the lected for the rugosimeter, along with a length of measure-
use of negative polarity for the electrode is recommended ment of 6.4 mm (8 mm × 0.8 mm). Moreover, a stylus tip
for this material, as the machining process can therefore be radius of 3 ␮m, a stylus speed of 1 mm/s and a stylus static
carried out in a more stable way. force of 0.12 mN were used.
 I. Puertas, C.J. Luis / Journal of Materials Processing Technology 153–154 (2004) 1033–1038 1035

In addition to surface roughness, material removal rate

(MRR) and electrode wear (EW) were studied. Although
there are other ways of measuring material removal rate, in
this work we have obtained the MRR values through the
weight difference of the sample before and after undergoing
the EDM process. Furthermore, when calculating the values,
it was accepted that the densities of the ceramic materials
were 2.50 g/cm3 (B4 C), 3.10 g/cm3 (SiSiC) and 14.95 g/cm3
(WC–Co), in accordance with the information from the man-
ufacturer.
Although the concept of electrode wear can be defined in Fig. 1. Comparison of the I factor effect over Ra for the different ceramic
many different ways, in this current study we have selected materials.
the definition of electrode wear according to the differences
in the volumes of the electrode and the part, as this def- response variables, second order models were selected for
inition is the most commonly used among researchers. In all the cases, except for the Ra parameter of boron carbide
order to calculate the loss of material from the electrode, and the MRR variable of silicon carbide.
we have used the same technique mentioned earlier and we
have taken a density value of 8.96 g/cm3 for the case of elec-
trolytic copper, in line with the information from the man- 4. Comparative analysis of ceramic materials
ufacturer.
4.1. Comparative analysis of the Ra parameter

3. Design of the experiments Fig. 1 shows the intensity factor effect over the Ra param-
eter for the ceramic materials considered in this study.
The design which was finally chosen was a factorial de- As can be observed in Fig. 1, in both silicon carbide and
sign 23 with four central points, which provide protection tungsten carbide the Ra parameter tends to increase signif-
against curvature, and a total of 12 experiments were made icantly when intensity increases. Conversely, in the case of
[9,10]. Consequently, for the case of the response variables boron carbide Ra this tends to decrease with the increase in
which were not adequate for the previous first order model, intensity although, in this case, this factor does not turn out
this was widened by the addition of six star points, giving to be statistically significant for the considered confidence
then a central composite design made up of the star points level. In practice, this diminution tendency in the Ra param-
situated in the centres of the faces; that is to say, a total of eter with the intensity factor is exactly the opposite of what
18 experiments in the case of this second order model. would be expected a priori, as increases in pulse energy are
Table 1 presents the relationship between factors and their usually associated, by and large, with a worsening in the
selected variation levels, taking into consideration that the material surface roughness.
study wanted to focus on the case of finishing machining Fig. 2 shows the pulse time effect over the Ra parameter
stages. for the different ceramic materials under study.
Table 2 shows the design matrix resulting from the type of As can be seen in Fig. 2, the Ra parameter tends to in-
experiment selected in the case of the second order model, crease when pulse time is increased for all the considered
as well as the observations for the case of the three response materials. The most notable increase is produced for boron
variables which are considered in this study, where the in- carbide, the effect of this factor being rather less significant
tensity values 3, 4 and 5 are equivalent to 2A, 4A and 6A, in the case of silicon carbide. Furthermore, in this latter case
respectively. The design matrix for the first order model is the Ra parameter seems to tend to decrease somewhat for
obtained by merely eliminating the rows corresponding to
the star points of the model.
After having performed a series of curvature tests for each
of the ceramic materials and each of their corresponding

Table 1
Factors and levels selected for the experiments
Factors Levels

−1 +1

I 3 5
ti (␮s) 10 50
η 0.4 0.6 Fig. 2. Comparison of the ti factor effect over Ra for the different ceramic
materials.
1036 I. Puertas, C.J. Luis / Journal of Materials Processing Technology 153–154 (2004) 1033–1038

Table 2
Design of experiment matrix obtained for the case of the second order model
Number I ti (␮s) η B4 C SiSiC WC–Co

Ra (␮m) EW (%) MRR Ra (␮m) EW (%) MRR Ra (␮m) EW (%) MRR

(mm3 /min) (mm3 /min) (mm3 /min)
1 3 10 0.4 1.38 1.93 0.135 0.89 9.47 0.092 0.66 36.51 0.055
2 5 10 0.4 1.30 3.10 0.187 1.09 8.80 0.213 1.24 15.32 0.248
3 3 50 0.4 2.27 0.26 0.159 0.75 16.18 0.024 0.66 71.58 0.012
4 5 50 0.4 2.16 0.59 0.206 1.17 8.92 0.118 1.93 12.09 0.202
5 3 10 0.6 1.41 2.07 0.328 0.86 10.60 0.139 0.65 35.25 0.090
6 5 10 0.6 1.26 3.38 0.357 1.23 11.82 0.331 1.37 15.54 0.428
7 3 50 0.6 2.03 0.30 0.339 0.82 14.90 0.043 0.64 71.77 0.020
8 5 50 0.6 2.06 0.39 0.396 1.86 8.04 0.378 2.02 11.09 0.318
9 3 30 0.5 1.56 1.04 0.183 0.84 13.03 0.047 0.64 66.41 0.026
10 5 30 0.5 1.68 0.80 0.261 1.34 8.92 0.234 1.75 11.02 0.279
11 4 10 0.5 1.34 2.84 0.290 1.20 8.07 0.339 1.11 17.34 0.236
12 4 50 0.5 1.92 0.50 0.250 1.26 8.25 0.159 1.73 15.21 0.155
13 4 30 0.4 1.68 1.01 0.171 1.24 8.17 0.147 1.43 14.93 0.147
14 4 30 0.6 1.51 0.91 0.341 1.34 8.68 0.254 1.54 13.35 0.246
15 4 30 0.5 1.77 0.77 0.283 1.34 8.42 0.190 1.43 14.29 0.193
16 4 30 0.5 1.83 0.69 0.282 1.33 8.45 0.192 1.34 14.15 0.192
17 4 30 0.5 1.77 1.02 0.277 1.30 8.59 0.180 1.39 14.08 0.194
18 4 30 0.5 1.78 0.88 0.274 1.37 8.29 0.203 1.42 14.09 0.194

pulse time values close to its highest level of 50 ␮s. There- remains constant, surface finish tends to worsen because the
fore, the tendency of Ra with regard to this design factor stability of the process is affected.
is exactly what would be expected a priori, as decreasing
the frequency of the pulse usually leads to worse values of 4.2. Comparative analysis of the electrode wear EW
surface roughness.
Fig. 3 shows a comparative graph of the duty cycle effect Fig. 4 shows the main effect of the intensity factor over
over the Ra parameter for the three ceramic materials studied the response variable of EW, for the different materials con-
in this work. As can be observed in this figure, in the cases sidered in this study. In the case of boron carbide, the elec-
of silicon carbide and tungsten carbide, the Ra parameter trode wear tends to increase when intensity increases. On
increases when duty cycle is increased, especially for silicon the contrary, in the cases of silicon carbide and tungsten car-
carbide, while this effect is not statistically significant for bide, the wear value decreases significantly, with intensity
tungsten carbide, within the considered confidence level. down to a minimum value after which the electrode wear
In the case of boron carbide, the Ra parameter tends to de- increases with the increase in the previous factor. In prac-
crease lightly when duty cycle is increased, although, in this tice, the logical tendency to expect, a priori, would be of a
case, the effect of this factor is not statistically significant decrease in the electrode wear as we move towards stages
for the confidence level under consideration in this study. On with greater pulse energy, that is, growing intensity levels.
the other hand, this tendency of the Ra parameter to decrease Fig. 5 shows the effect of the pulse time factor over the
when duty cycle increases, within the considered work in- EW variable for the different ceramic materials. As can be
terval, is exactly the opposite of what would be expected in seen in Fig. 5, in the case of boron carbide, the electrode
practice, as, when pause time tends to decrease, which is wear value tends to decrease greatly with an increase in pulse
equivalent to increasing the duty cycle factor if pulse time time, which coincides with the behaviour which would be

Fig. 3. Comparison of the η factor effect over Ra for the different ceramic Fig. 4. Comparison of the I factor effect over EW for the different ceramic
materials. materials.
 I. Puertas, C.J. Luis / Journal of Materials Processing Technology 153–154 (2004) 1033–1038 1037

Fig. 5. Comparison of the ti factor effect over EW for the different Fig. 7. Comparison of the I factor effect over MRR for the different
ceramic materials. ceramic materials.

As shown in Fig. 7, the MRR variable increases when the

logical to expect, as an increase in the pulse energy (in this
intensity factor increases, in all the cases under considera-
case through an increase in pulse time) is usually associated
tion. However, in the case of boron carbide, this tendency to
with a decrease in electrode wear. On the contrary, in the
increase reaches a peak after which it begins to decrease with
cases of both silicon carbide and tungsten carbide, for the
the increase in intensity. This tendency of the MRR variable
considered work interval, the wear value tends to increase
to increase with the design factor of intensity is what would
when the pulse time factor also increases.
be expected, a priori, as the quantity of removed material is
Fig. 6 shows a comparative graph of the effect of duty
proportional to the size or area of the electrical pulse used.
cycle over EW for the three ceramic materials which have
On the other hand, as can be seen in Fig. 7, this tendency
been considered in this work.
to increase with respect to the intensity factor is much more
In the cases of boron carbide and silicon carbide, the elec-
obvious for the cases of both silicon carbide and tungsten
trode wear value tends to increase slightly when the duty
carbide.
cycle factor increases. Meanwhile, in the case of tungsten
Fig. 8 shows a comparative graph of the effect of the
carbide, EW increases slightly up to a maximum value, sit-
pulse time factor over MRR for the different ceramic ma-
uated around the central value of duty cycle (η = 0.5), after
terials. In the cases of silicon carbide and tungsten car-
which it tends to decrease with the increase in the previous
bide, the value of MRR tends to decrease when there is
design factor. In practice, the tendency which would be ex-
an increase in the value of the pulse time factor, which
pected, based on the experience gained from other types of
does not coincide with what would be expected a priori.
materials, is that of a decrease in wear with the decrease in
In practice, as was commented in the earlier case, the in-
pause time, which is equivalent to increasing the duty cycle
crease in the quantity of pulse energy, in this case through
factor, if the pulse time factor remains constant. Neverthe-
an increase in pulse time, tends to increase the value of
less, as has been previously seen, the main effect of the duty
MRR, although it always reaches a point where it starts
cycle factor does not turn out to be statistically significant
to decrease even though the pulse time value continues to
for any of the ceramic materials studied.
increase.
On the other hand, in the case of boron carbide, the MRR
4.3. Comparative analysis of the material removal rate value tends to decrease with pulse time down to a mini-
MRR mum value, around its central value, after which MRR tends
to increase with this factor. However, in the case of this
Fig. 7 shows the main effect of the intensity factor over ceramic material, the main effect of pulse time does not
the MRR variable, for the different ceramic materials con- turn out to be significant, for the confidence level under
sidered. consideration.

Fig. 6. Comparison of the η factor effect over EW for the different ceramic Fig. 8. Comparison of the ti factor effect over MRR for the different
materials. ceramic materials.
1038 I. Puertas, C.J. Luis / Journal of Materials Processing Technology 153–154 (2004) 1033–1038

sten carbide, it has been observed that the wear value de-
creases with the increase in intensity until reaching a min-
imum value. Furthermore, the electrode wear tends to de-
crease with the increase in pulse time in the case of boron
carbide. On the other hand, in the cases of silicon carbide
and tungsten carbide, just the opposite behaviour was ob-
served. Finally, it has been observed that EW tends to in-
crease when duty cycle is increased in the cases of boron
carbide and silicon carbide, whilst in the case of tungsten
carbide, it increases up to a maximum value after which it
Fig. 9. Comparison of the η factor effect over MRR for the different
ceramic materials.
decreases.
With regard to material removal rate, it has been con-
firmed that in all cases this increases when intensity in-
Fig. 9 shows a comparative graph of the effect of duty
creases, although, in the case of boron carbide, this tendency
cycle over MRR for the three ceramic materials studied.
to grow reaches a maximum value after which it starts to
As shown in Fig. 9, the value of material removal rate
decrease. Furthermore, the fact of increasing pulse time
tends, in all the ceramic materials which were studied, to
produces a decrease in material removal rate, in the cases
increase when duty cycle has also been increased. This co-
of silicon carbide and tungsten carbide. In the case of boron
incides with what would be expected in practice, based on
carbide, it decreases with pulse time down to a minimum
the experience acquired in the field of EDM for other ma-
value, after which it tends to increase. On the other hand,
terials, as, when pause time is reduced, which is equivalent
it has been observed that material removal rate tends to
to increasing duty cycle, the MRR value tends to increase.
increase when duty cycle is increased for all the cases.
Furthermore, as can be seen in the previous Fig. 9, this ten-
dency for material removal rate to increase with duty cycle
is much more pronounced in the case of boron carbide than References
in the case of the other two remaining ceramic materials.
[1] EDM Technical Manual, Poco Graphite Inc., Texas, 1994.
[2] J.R. Davis, ASM Materials Engineering Dictionary, ASM Interna-
5. Conclusions tional, USA, 1992.
[3] I. Puertas, C.J. Luis, A revision of the applications of the electri-
The interest in the study of die-sinking EDM when ma- cal discharge machining process to the manufacture of conductive
ceramics, Rev. Met. Madrid 38 (5) (2002) 358–372.
chining conductive ceramics is a consequence of the diverse [4] I. Cabanes, J.A. Sánchez, L.N. López de Lacalle, A. Lamikiz, Mecan-
problems brought about by the machining of these materials izado por electroerosión de componentes cerámicos de precisión,
by other manufacturing processes such as laser machining, IMHE, May 2001, pp. 63–72.
abrasive jet machining, ultrasonic machining and, even, di- [5] J.A. Sánchez, I. Cabanes, L.N. López de Lacalle, A. Lamikiz, De-
amond wheel grinding, among others. velopment of optimum electrodischarge machining technology for
advanced ceramics, Int. J. Adv. Manuf. Technol. 18 (2001) 897–
It has been found that surface roughness increases when 905.
intensity increases, for the cases of silicon carbide and [6] C.F. Noble, A.J. Ajmal, L.F. Green, Electro-discharge machining of
tungsten carbide. Conversely, in the case of boron carbide, silicon carbide, in: Proceedings of the International Symposium of
roughness tends to decrease with an increase in intensity. Electro Machining—ISEM 7, Birmingham, 1983, pp. 305–312.
Accordingly, it has been shown that an increase in pulse [7] X. De Maidagan, J.A. Sánchez, J.I. Llorente, R. Lenzen, T. Nöthe,
Electroerosión: Llave para un crecimiento real de la aplicación de
time produces an increase in surface roughness. Further- las cerámicas?, IMHE, June/July 2000, pp. 43–48.
more, the fact of increasing duty cycle produces a worsening [8] S.H. Lee, X.P. Li, Study of the effect of machining parameters on
in surface roughness, especially in silicon carbide. On the the machining characteristics in electrical discharge machining of
other hand, in the case of boron carbide, just the opposite tungsten carbide, J. Mater. Process. Technol. 115 (2001) 344–358.
behaviour has been observed. [9] D.C. Montgomery, Design and Analysis of Experiments, Wiley, New
York, 2001.
With respect to electrode wear, in the case of boron car- [10] I. Puertas, C.J. Luis, A study on the machining parameters optimi-
bide, it has been observed that it increases when intensity sation of electrical discharge machining, J. Mater. Process. Technol.
also increases. On the contrary, in silicon carbide and tung- 143–144 (2003) 521–526.