Ceramics International 42 (2016) 15671–15678

Contents lists available at ScienceDirect

Ceramics International
journal homepage: www.elsevier.com/locate/ceramint

Optimization of wire electrical discharge machining parameters for

cutting electrically conductive boron carbide
Ayan Pramanick a,n, Soumya Sarkar b, Partha Pratim Dey a, Probal Kr. Das b
a
Mechanical Engineering Department, Indian Institute of Engineering Science and Technology, Shibpur, Howrah 711103, India
b
Non-oxide Ceramics and Composites Division, CSIR-Central Glass and Ceramic Research Institute, Kolkata 700032, India

art ic l e i nf o a b s t r a c t

Article history: In this work, Pure boron carbide (B4C) was consolidated using spark plasma sintering (SPS) at 2050 °C
Received 26 February 2016 with a dwell of 10 min under 50 MPa uniaxial pressure in Argon atmosphere. The sintered specimen was
Received in revised form 499% dense and offered characteristic Vickers hardness and fracture toughness of 31.4 GPa and
23 May 2016
4.21 MPa-m0.5, respectively, at 4.9 N indentation load. The specimen showed satisfactory wire electrical
Accepted 3 July 2016
discharge machining (WEDM) performance because of its good electrical conductivity. The design of
Available online 5 July 2016
experiment (DOE) was arranged by L32 orthogonal array (OA) between the machining input parameters
Keywords: namely pulse on-time, pulse off-time, pulse peak current, dielectric fluid pressure and servo feed rate
Sintering and the output responses like machining speed and surface roughness (Ra). Regression models were
Boron carbide
employed to establish the numerical correlation between the machining parameters and output re-
WEDM
sponses. Experimental observations were utilized to formulate the first-order regression models to
Optimization
predict responses of WEDM. The optimized input parameters were 27 μs pulse on time, 48 μs pulse off
time, 180 A pulse peak current, 7 kg/cm2 water pressure and 2200 mm/min servo feed rate for the
WEDM performance to produce an optimum machining speed and Ra.
& 2016 Elsevier Ltd and Techna Group S.r.l. All rights reserved.

1. Introduction other nontraditional machining processes such as abrasive water

B4C has become quite popular for a wide range of engineering
applications like refractory, abrasive powder, neutron absorbent
cross section (
4000  10–28 m2), body and air-borne vehicular
armors against different ballistic threat levels due to its high
melting point ( 42400 °C), extreme chemical inertness (reacts
only with halogens at high temperature), low density (
2.51 g/cc),
very high hardness (
30 GPa) and flexural strength (4400 MPa).
B4C also possesses other favorable properties e.g. high thermal
conductivity (
40 W/m K at room temperature), low thermal
expansion coefficient (
5  10  6/°C) and low electrical resistivity
(
0.03 Ω m at room temperature) those make it suitable for
multifunctional applications and since, B4C is electrically con-
ductive, it has a bright prospect toward net-shape manufacturing
by WEDM technique [1]. The manufacturing condition is one of
the most imperative aspects to take into consideration in the
majority of modern manufacturing processes, particularly, in
techniques related to WEDM of electrically conductive ceramic
materials [2]. Use of conventional machining processes is limited
to machine such hard ceramics due to severe tool damage. Some
n
Corresponding author.
E-mail address: ayan_mech07@yahoo.co.in (A. Pramanick).
jet machining (AWJM) and laser beam machining (LBM) can be
used but the machining equipment as well as the machining
process both are quite expensive. Furthermore, for AWJM and
LBM, work-piece height is a constraint and the surface finish ob-
tained is also not good compared to that obtained in thermo-
electrical machining techniques like WEDM [3]. WEDM para-
meters were studied on hot pressed B4C sample and machining
responses like material removal rate (MRR) and Ra of the material
were evaluated [4]. Electrically conductive TiN/Si3N4 nano-
composite was hot pressed and processed by WEDM and its mi-
crostructure was investigated. The dependence of surface texture,
Ra and MRR on WEDM conditions such as power transistor num-
bers and pulse off-time was also analyzed. It was reported that
more power transistor numbers resulted in greater MRR and Ra.
More pulse off-time resulted in lower MRR [3]. Investigations were
also conducted on the effect of WEDM parameters on performance
characteristics i.e., MRR, kerf width and Ra. An optimum combi-
nation of process parameters was derived for large MRR and small
Ra using analysis of variance (ANOVA) [5]. Investigations were
carried out on the multiple response characteristics effects like
cutting speed, depth of cut and feed rate on Ra using the Taguchi
technique and Grey relational analysis by optimizing the EDM
process parameters [6,7]. The surface roughness prediction model
optimized the machining input variables to obtain high surface

http://dx.doi.org/10.1016/j.ceramint.2016.07.023
0272-8842/& 2016 Elsevier Ltd and Techna Group S.r.l. All rights reserved.
15672 A. Pramanick et al. / Ceramics International 42 (2016) 15671–15678

quality of Al2O3/SiC/TiC composite [8]. In this work, pure B4C was

fabricated using SPS technique and basic mechanical properties
e.g. bulk density (BD), apparent porosity (AP), Vickers hardness
(HV) and indentation fracture toughness (KIC) were evaluated
using standard procedures. Finally, optimization of WEDM para-
meters for machining of the fabricated specimens were carried out
using Taguchi method [9,10]. During machining, a detail experi-
mentation was performed to realize the effects of five WEDM in-
put parameters viz. pulse on-time (Ton), pulse off-time (Toff), pulse
peak current (IP), water pressure (WP) and servo feed rate (SF) on
output responses i.e., Ra and machining speed. Mathematical
models, comprising of the input and output parameters were de-
rived using DOE technique combined with regression analysis for
this purpose.

2. Expeimental

2.1. Raw materials and specimen fabrication

B4C powder (H. C. Starck, Germany, HD-15A) was procured from

having B:C ratio of 3.7–3.8 and surface area of 15–21 m2/g was
used. The d50 of as-received powder was found to be
0.6 mm
when analyzed using a laser diffraction particle size analyzer (LS-
13–320-MW, Beckman Coulter, USA). The as-received powder was
directly poured into a graphite die-plunger assembly having
20 mm internal diameter and sintered at 2050 °C in a SPS furnace
(HP-D-25, FCT Systems GmbH, Germany) with a dwell of 10 min
under 50 MPa uniaxial pressure in Argon atmosphere to produce a

5 mm thick pellet (Fig. 1(a-b)). During the sintering cycle, fur-
nace was heated up to the peak temperature at a heating rate of
400 °C/min and after completion of the dwell time, cooling was
done at the rate of 200 °C/min.

2.2. Physicomechanical characterizations Fig. 1. (a) The SPS furnace holding the graphite die/plunger assembly within the
graphite top and bottom rams, (b) The sintered pure B4C after the SPS cycle.
Bulk density of the sintered B4C specimen was measured using
Archimedes water immersion principle in distilled water. Grinding-
Table 1
polishing of the sintered specimens was performed in a Spectrum Wire EDM parameters and their levels.
System 1000 instrument of LECO Corporation, USA. HV of the sin-
tered specimens was measured using a micro-Vickers hardness Level Ton Toff IP WP SF
tester (402 MVD, Wolpert-Wilson, Germany) at 4.9 N with 10 s (ls) (ls) (A) (kg/cm2) (mm/min)
dwell. KIC of the specimen was evaluated using direct crack mea-
1 27 48 180 7 2100
surement (DCM) technique and the well-known equations pro- 2 30 52 200 8 2200
posed by Niihara et al. [12,13] which were individually suitable for
median (c/a 4 2.5, where, ‘c’ is length of the crack emanating from
corners of Vickers impression and ‘a’ is half of the indentation di- off length of 6 mm. A typical outcome of roughness measurement
agonal) and Palmqvist crack (c/a o2.5) systems. HV and KIC data is shown in Fig. 2. Since, thickness of the sample was kept constant
were further analyzed using the 2-parameterWeibull statistics to (5 mm) during this work, the MRR (mm3/min) was equivalent to
obtain the much dependable characteristic hardness (HVCH) and the machining speed (mm/min). The machining speed (mm/min)
toughness (KIC, CH) values and corresponding Weibull modulus (m) was determined by noting the time taken in minutes for a parti-
[14–18]. cular length of cut (say, 4 mm) with the observation that variation
in kerf width is negligible for this set of experiments.
2.3. Machining of sintered B4C
2.4. Mathematical modeling
The machining operations were conducted in a CNC WEDM
(EUROCUT Mark-I, 734, India). Range of machining input data sets Mathematical models of machining responses were developed
at two levels are shown in Table 1. To study the effects of ma- by regression analysis from the experimental observations. Each of
chining parameters on performance characteristics and to trace the response function can be expressed [10] as:
out the optimal machining condition, a specifically designed ex-
n n n− 1 n
perimental procedure was required [19–22]. WEDM input data
was prepared by DOE. Process characterization was made using
Y = CO + ∑ Ci*X i + ∑ Cj*X j + ∑ ∑ Cij*X ij
i=1 i=1 i= 1 j= 2 (1)
ANOVA to identify the key input variables that affected Ra and
machining speed [22]. Ra values of the machined sample under where, Y is the response characteristic. Regression models were
different machining conditions were examined using an optical utilized to correlate the xi (1, 2,…., n) are coded levels of n quan-
surface profilometer (Contour GT-K, Bruker Corp., USA) with a cut- titative input process variables i.e. machining parameters, the
 A. Pramanick et al. / Ceramics International 42 (2016) 15671–15678 15673

Fig. 2. Typical Ra measurement output obtained from the non-contact profilometer.

terms C0, Ci, and Cij are the regression coefficients. By regression Surface roughness(Ra) = 157. 6 − 2. 649*Ton−1. 842*Toff
analysis, two mathematical models for two responses i.e. Ra and
−0. 289*IP−0. 23*WP−0. 0362*SF
Machining speed were derived as follows:
+0. 03156*Ton*Toff −0. 00094*Ton*IP
−0. 0363*Ton*WP +0. 00066*Ton*SF
+0. 00289*Toff *IP−0. 0766*Toff *WP
+0. 000366*Toff *SF +0. 02119*IP*WP
−0. 000002*IP*SF +0. 000437*WP*SF (2)

Table 2
WEDM input and response parameters.

Exp.no. Machine Input Variables Machine Output Experimental Variables Overall Grey relational Grade (γi)

Ton (ls) Toff (ls) IP (A) WP (kg/cm2) SF (mm/min) Ra (lm) M/c Speed (mm/min)

1. 27 48 180 7 2100 3.77 0.453 0.885

2. 27 48 180 7 2200 3.83 0.437 0.977
3. 27 48 180 8 2100 3.38 0.460 0.729
4. 27 48 180 8 2200 3.73 0.432 0.951
5. 27 48 200 7 2100 2.93 0.481 0.592
6. 27 48 200 7 2200 3.08 0.454 0.688
7. 27 48 200 8 2100 3.48 0.548 0.598
8. 27 48 200 8 2200 3.51 0.451 0.793
9. 27 52 180 7 2100 2.64 0.435 0.704
10. 27 52 180 7 2200 3.02 0.463 0.651
11. 27 52 180 8 2100 2.20 0.496 0.489
12. 27 52 180 8 2200 2.69 0.468 0.594
13. 27 52 200 7 2100 2.34 0.499 0.494
14. 27 52 200 7 2200 2.71 0.491 0.544
15. 27 52 200 8 2100 2.15 0.529 0.435
16. 27 52 200 8 2200 2.64 0.508 0.506
17. 30 48 180 7 2100 3.05 0.495 0.581
18. 30 48 180 7 2200 3.51 0.454 0.783
19. 30 48 180 8 2100 2.88 0.516 0.522
20. 30 48 180 8 2200 3.43 0.463 0.733
21. 30 48 200 7 2100 2.54 0.549 0.443
22. 30 48 200 7 2200 2.96 0.508 0.545
23. 30 48 200 8 2100 2.50 0.639 0.371
24. 30 48 200 8 2200 3.01 0.511 0.548
25. 30 52 180 7 2100 2.65 0.531 0.475
26. 30 52 180 7 2200 3.08 0.468 0.646
27. 30 52 180 8 2100 2.08 0.544 0.412
28. 30 52 180 8 2200 2.49 0.482 0.541
29. 30 52 200 7 2100 1.99 0.595 0.361
30. 30 52 200 7 2200 2.66 0.539 0.466
31. 30 52 200 8 2100 2.09 0.594 0.368
32. 30 52 200 8 2200 2.55 0.530 0.466
15674 A. Pramanick et al. / Ceramics International 42 (2016) 15671–15678
Machining Speed= − 11. 45 + 0. 244*Ton−0. 0693*Toff
+0. 0144*IP +0. 768*WP +0. 00534*SF
−0. 00026*Ton*Toff +0. 000407*Ton*IP
−0. 00161*Ton*WP−0. 000130*Ton*SF
−0. 000052*Toff *IP−0. 00185*Toff *WP
+0. 000049*Toff *SF +0. 000433*IP*WP
−0. 000011*IP*SF −0. 000322*WP*SF
2.5. Taguchi method
Taguchi method utilizes two major tools which are (i) S/N
(signal/noise) ratio to measure the quality and (ii) orthogonal array
to encounter the factors affecting the machining performance [10].
According to Taguchi quality design concept, an L32 (25) i.e. 2 le-
vels 5 factor orthogonal array table with 32 rows was chosen for
the experiments (Table 2). The experimental observations were
further transformed into a (S/N) ratio using ANOVA [19].
2.6. Selection of (S/N) Ratio
The (S/N) ratio was obtained using Taguchi methodology. Since,
‘S′ and ‘N′ represented the desirable (mean) and undesirable
(standard deviation) values, respectively, (S/N) ratio mainly in-
dicates the extent of variation in performance characteristics. In
Taguchi method, a loss function is defined to measure deviation
between the experimental and desired value of a performance
characteristic [20,21]. This loss function further transformed into a
(S/N) ratio. In this particular DOE problem, there were two dif-
ferent types of performance characteristics namely, lower-the
better (LB) and higher-the-better (HB) [22].
2.6.1. Lower-the-better type problem
For this type of problem, the quality characteristic was taken as
continuous, non-negative and the most desirable value can be
zero. As there was no adjustment factor, the thrust was simply on
minimizing the quality loss without adjustment as in Eq. (4):
⎡1 n ⎤
LLB=−10log10⎢ ∑ yR2a ⎥
⎢⎣ n i=1
⎥⎦
2.6.2. Higher-the-better type problem
Here, the objective function was described to maximize the
machining speed without using any adjustment factor. Thus, the
higher-the-better (S/N) ratio can be expressed as:
⎡ n ⎤
1 1
LHB= − 10log10⎢ ∑ ⎥
⎢⎣ n 2
ymachine ⎥ nomenon through grain boundary regions and resulted in highly
i=1 speed ⎦
where, n is the number of observations. Hence, single response
optimization was achieved for optimizing the machining speed
and Ra separately.
2.7. Grey relational analysis theory
In this analysis, experimental data i.e. measured features of
quality characteristics were first normalized followed by Grey re-
lational coefficient evaluation to represent the correlation be-
tween the desired and experimental data. Then, overall Grey re-
lational grade was determined by averaging the Grey relational
coefficient corresponding to selected responses [10,23,24]. The
optimal factor setting for maximizing the overall Grey relational
grade can be performed by Taguchi method. During Grey relational
generation, the normalized data i.e. Ra, corresponding to LB cri-
terion was expressed as:
max. yi ( k ) − yi (k )
xi =
max. yi ( k ) − min. yi (k ) (6)
Similarly, data processing for machining speed related to HB
criterion was expressed as:
yi(k ) − minyi ( k )
(3) xi =
maxyi ( k)−minyi (k) (7)
where, xi(k), min. yi (k) and max. yi (k) represent the value after
Grey relational regression, the smallest value of yi (k) and the
largest value of yi (k) for kth response. An ideal sequence is x0 (k),
where, k ¼ 2 for the responses and i ¼ 1, 2, 3,…, 32. The Grey
relational coefficient ξi (k) was calculated using the following ex-
pression [10,25]:
∆min +ψ∆max
ξ i (k) =
∆0i( k)+ψ∆max (8)
where, Δ0i (k)¼ |x0(k) xi(k)| and ψ used as the distinguishing
coefficient that can vary between 0 and 1 and here, ψ was taken as
0.5. Δmax and Δmin are the global maximum and minimum values,
respectively, in the different data series [25–27]. After averaging
the Grey relational coefficients, the Grey relational grade γi was
computed as follows:
1 n
γi = ∑k = 1 ξ i(k) (9)
n
where, n is the number of process responses. The higher Grey
relational grade means that the corresponding parameter combi-
nation was closer to the optimal. The optimal levels of machining
response parameters were selected in the final step to predict and
verify the improvement of the performance characteristics.
3. Results and discussions
3.1. Physicomechanical properties of the SPS sintered specimen
Bulk Density of the sintered B4C was found to be
2.5 g/cc.
(4) Considering the theoretical density of pure B4C as 2.51 g/cc [1,28],
the present specimen was found to be
99.5% dense having
negligible porosity. This indicates that besides the effects of high
sintering temperature and pressure employed during the con-
solidation process, high electrical conductivity of B4C, facilitated
sufficient current conduction through the consolidating mass
during SPS cycle and eventually, enhanced the effective localized
heating of B4C particles that promoted the mass transport phe-
(5)
dense specimen without using any sintering aid [1]. Evidently,
present densification data fall on the higher side of available lit-
erature data [11,28–31]. Moshtaghioun et al. [28] obtained fully
dense pure B4C ceramic through SPS processing at Z1700 °C with
a dwell of 3–5 min. However, the authors employed relatively
higher pressure (i.e. 75 MPa) and slower heating rate (i.e. 100 °C/
min) compared to those used in the present study.
On the other hand, fully dense pure B4C was fabricated by
Hayun et al. [29] using SPS technique at 2050 °C with a dwell of
6 min under 32 MPa applied pressure. Evidently, although, the
sintering temperature was the same as used in the present study,
use of a relatively higher pressure (i.e. 50 MPa) and 4 min extra
dwell time ensured attainment of almost similar density of present
pure B4C at a heating rate that was 8 times higher than that used
by Hayun et al. [29]. Therefore, although, the present sintering
 A. Pramanick et al. / Ceramics International 42 (2016) 15671–15678
Fig. 3. 2-parameter Weibull plot of (a) Vickers hardness and (b) fracture toughness
data obtained at 4.9 N.
temperature was higher than that used by others, present sinter-
ing schedule effectively utilized the primary advantage of SPS
processing i.e., rapid densification at much higher heating rate
[11,28,31].
Since, dense B4C and its composites are mainly used in ad-
vanced structural applications, its mechanical performance is quite
important for designing purpose. As far as mechanical properties
of the present B4C sample was concerned, average HV and KIC at
4.9 N indentation load were found to be (30.43 7 2.19) GPa and
(3.96 7 0.59) MPa m0.5, respectively. Present HV and KIC values
were found to be quite comparable with available literature data
on B4C ceramics [28,29,31,32].
Considering the extreme brittleness, high flaw sensitivity and
catastrophic failure tendency of monolithic B4C, the 2-parameter
Weibull statistics was employed to evaluate the characteristic
property values which were primarily suitable for component
designing purpose [14–17]. The HVCH and KIC, CH values obtained
from the linear fit to ln{ln1/(1-p)} versus ln(HV0.5) or ln(KIC) (Fig. 3
(a-b)) were found to be 31.4 GPa and 4.21 MPa-m0.5, respectively,
having Weibull modulus (m) values of 17 and 8, respectively. Al-
though, the Weibull modulus (17) of HV data indicated significant
15675
Fig. 4. Comparison between predicted and experimental (a) Ra and (b) machining
speed values.
reliability of present B4C specimen against permanent deforma-
tion under applied stress, the lower modulus i.e. 8 for toughness
data suggested relatively higher sensitivity to the inherent flaws of
the studied specimen. Lower toughness data as well as limited
resistance to catastrophic failure is one of the limitations of pure
B4C ceramic and that can be minimized using proper reinforce-
ment [30,31,33]. However, to assess the precise machining re-
sponse behavior of pure B4C ceramic during WEDM process, the
present specimen was fabricated without using any sintering aid
or reinforcing phase.
3.2. Machining performance analysis and optimization
The predicted response values were evaluated using the
mathematical models presented in Eqs. (2),(3). It defined that how
much the model fits with the actual machining responses. The
experimental versus predicted responses showed good agreement
with each other (Fig. 4(a-b)). ANOVA was carried out to identify
the significant variables and to quantify their effects on response
characteristics (Tables 3a and 3b). The ANOVA depicted the degree
of freedom (DF) and sum of square (SS) values in sequential and
adjustable modes, error terms, P-value, R-square and adjusted
R-square [3].
The ANOVA table was generated at 95% confidence level and
the p-value represented the significance of machine input
15676 A. Pramanick et al. / Ceramics International 42 (2016) 15671–15678

Table 3a Table 4a
Analysis of Variance for Ra (mm). Response table for (S/N) ratio of Ra (Lower is better).

Source DF Seq-SS Adj-SS Adj-MS P R2 (%) R2 adjusted (%) Level Ton (ls) Toff (ls) IP (A) WP (kg/cm2) SF (mm/min)

Ton 1 0.04344 0.04343 8.33 0.011 97.29 94.75 1  6.481  6.229  6.489  6.216  5.679
Toff 1 0.01094 0.01093 2.10 0.167* 2  5.615  5.867  5.607  5.881  6.417
IP 1 0.03457 0.03456 6.63 0.020 Delta 0.866 0.363 0.882 0.335 0.738
WP 1 0.00698 0.00698 1.34 0.264* Rank 2 4 1 5 3
SF 1 0.00225 0.00225 0.43 0.520*
Ton*Toff 1 0.28313 0.28312 54.30 0.000
Ton*IP 1 0.02703 0.02702 5.18 0.037 Table 4b
Ton*WP 1 0.00340 0.00340 0.65 0.431* Response table for (S/N) ratio of machining speed (Higher is better).
Ton*SF 1 0.00113 0.00112 0.22 0.648*
Toff*IP 1 0.02940 0.02940 5.64 0.030 Level Ton (ls) Toff (ls) IP (A) WP (kg/cm2) SF (mm/min)
Toff*WP 1 0.11400 0.11400 21.86 0.000
Toff*SF 1 0.01240 0.01240 2.38 0.143* 1  9.416  10.099  9.480  9.200  8.358
IP*WP 1 0.07900 0.07900 15.15 0.001 2  8.567  7.883  8.502  8.782  9.625
IP*SF 1 0.00813 0.00812 1.56 0.230* Delta 0.849 2.216 0.978 0.418 1.267
WP*SF 1 0.00053 0.00052 0.10 0.754* Rank 4 1 3 5 2
Error 16 0.08343 0.00521
Total 31 7.82887

largest delta was given rank 2 and so on.

The response Table 4a showed that the pulse peak current
Table 3b
possess larger impact over surface roughness followed by pulse on
Analysis of Variance for machining speed (mm/min).
time, servo feed rate, pulse off time and water pressure. The re-
Source DF Seq-SS Adj-SS Adj-MS P R2 (%) R2 adjusted (%) sponse Table 4b depicted that the pulse off time possessed the
better impact over machining speed followed by servo feed rate,
Ton 1 0.001399 0.001399 4.53 0.049 93.63 87.66
Toff 1 0.000220 0.000220 0.71 0.411*
pulse peak current, pulse on time and water pressure. The (S/N)
IP 1 0.000218 0.000218 0.71 0.413* ratio was utilized to identify the control factor settings that
WP 1 0.001474 0.001474 4.78 0.044 minimized the effect of noise on Ra and machining speed (Fig. 5(a-
SF 1 0.001451 0.001451 4.70 0.046
Ton*Toff 1 0.000020 0.000020 0.07 0.801*
Ton*IP 1 0.001193 0.001193 3.87 0.067*
Ton*WP 1 0.000047 0.000047 0.15 0.703*
Ton*SF 1 0.003030 0.003030 9.82 0.006
Toff*IP 1 0.000035 0.000035 0.11 0.742*
Toff*WP 1 0.000110 0.000110 0.36 0.559*
Toff*SF 1 0.000773 0.000773 2.51 0.133*
IP*WP 1 0.000150 0.000150 0.49 0.496*
IP*SF 1 0.001005 0.001005 3.26 0.090*
WP*SF 1 0.002069 0.002069 6.71 0.020
Error 16 0.004936 0.000308
Total 31 0.077479

parameters. It may be visualized from Table 3a that among the

linear effects, the Pulse off time, flushing dielectric water pressure
and servo feed were insignificant factors (* marked) as the P-va-
lues (i.e. 40.05) shown 0.167, 0.264 and 0.520 respectively. Si-
milarly, from Table 3b it was visualized that among the linear ef-
fects of P-values, pulse off time (0.411) and pulse peak current
(0.413) values shown insignificant effects (i.e. 4 0.05) on ma-
chining speed. The other linear parameter effects were found to be
significant (i.e. P- value o0.05). R-square terms described the
extent of variation in the observed responses that explained by the
model. Adjusted R-square represented the adjusted number of
terms in the model. As shown in Tables 3a and 3b, the R-square
value was found to be 93.63 for machining speed and 97.29 for Ra.
This suggested the adequacy to fit the data in the model very well.
R-square adjusted terms also represented good agreement with
the R-square value as 94.75% (Table 3a) and 87.66% (Table 3b) for
Ra and machining speed, respectively.
Based on Taguchi's optimization, (S/N) ratio of machining re-
sponses are shown in Tables 4a and 4b representing the measure
of robustness. The delta value measured the size of the effect by
taking the difference between the highest and the lowest char-
acteristic average of a factor like Ton, Toff etc. The ranks in a re-
sponse table were used to identify the input factor which had the
greatest effect on the WEDM response value. The factor with the
largest delta value was given as rank 1, the factor with the second Fig. 5. (S/N) ratio plots for (a) Ra and (b) machining speed.
 A. Pramanick et al. / Ceramics International 42 (2016) 15671–15678 15677

Table 5
Table for mean of the overall grey relational grade.

Level Ton (ls) Toff (ls) IP (A) WP (kg/cm2) SF (mm/min)

1 0.6645 0.6713 0.6672 0.6147 0.5288

2 0.5163 0.5095 0.5137 0.5661 0.6521
Delta 0.1481 0.1618 0.1535 0.1535 0.1233
Rank 3 1 2 5 4

Fig. 7. (S/N) ratio plot for Grey relation grade.

technique [6,20,21]. It was noticed that among the five input

parameters, Toff had the most significant effect on Grey relational
grade followed by IP, Ton, SF and WP represented the least sig-
nificant effect on the Grey relation grade (Table 5). The graphical
illustration of parameters is shown in Fig. 7. Analysis of factors
contributed to fix the optimal process parameters for maximum
Grey relational grade as Ton ¼27 ms, Toff ¼ 48 ms, IP¼180 A,
Fig. 6. Main effect plots (a) Ra and (b) machining speed values.
WP¼ 7 kg/cm2 and SF¼2200 mm/min.

b)). The curve showing larger extent of inclination was treated as
the most significant curve, while, the curve being horizontal to the
mean line expressed the less significant effect over the responses.
From Fig. 5(a) it may be understood that pulse off- time, servo feed
rate and pulse peak current graphs showed highly significant ef-
fect on surface roughness followed by Pulse on-time and WP. Si-
milarly, it may also understood from the Fig. 5(b) that pulse on-
time, pulse peak current and servo feed rate showed high sig-
nificance effect on machining speed than pulse off-time and water
pressure. The optimal surface roughness as well as machining
speed value was obtained at Ton ¼30 m s, Toff ¼52 m s, IP¼200 A,
WP ¼8 kg/cm2 and SF ¼2100 mm/min.
The main effect plots of surface roughness and machining
speed is shown in Fig. 6(a-b), respectively. The Ra value decreased
linearly with increase in Ton, Toff, IP and WP levels. But in case of SF,
it showed linear and increasing slope with increase of level. The
decreased WP value depicted steep slope than other parameters
(Fig. 6a). It may be understood from the Fig. 6(a) that the all
3.3. Confirmatory tests
The confirmatory test [10] is an essential step to validate the
optimization results which was derived by Grey Taguchi technique
by using a set of experimental results. A new set of three experi-
ments was conducted to verify the actual performance of the final
optimized set of machining parameters and the average values of
the output responses, i.e., machining speed and surface roughness
(Ra) were noted as shown in Table 6. The variation between the
predicted and experimental results of the output responses i.e.,
machining speed and Ra were found to be 3.26% and 1.56% re-
spectively. These observations clearly shows that multiple perfor-
mance characteristics of the WEDM for machining of almost
Table 6
The confirmatory test.
Optimal set of Parameters

parameters affect the Ra value other than water pressure due to

Input Parameters Ton (ls) Toff (ls) IP (A) WP SF
the slope is quite horizontal with the mean line. The main effect (kg/cm2) (mm/min)
plot for machining speed in Fig. 6(b) shows that the value in-
creased with increase in levels of Ton, IP and WP but decreased Levels 27 48 180 7 2200
with the increase in level of SF only. In Fig. 6(b), the Toff and WP Parameters Predicted Average result of Experimental responses
value showed fewer effects on machining speed due to the linear Machining Speed 0.429 0.443
variation with very narrow slope with the mean line. The Grey (mm/min)
Surface roughness (mm) 3.83 3.89
relational grade was utilized as the performance measure for the Grey relation grade 0.981
case of multi response optimization and evaluated using Taguchi
15678 A. Pramanick et al. / Ceramics International 42 (2016) 15671–15678
theoretically dense pure B4C having adequate mechanical prop-
erties of acceptable reliability were significantly improved through
this study.
4. Conclusions
The major outcomes of this work may be summarized as
follows:
i. Phase pure B4C ceramic having
99.5% theoretical density was
obtained by SPS processing at 2050 °C under 50 MPa with
10 min dwell. 2-paramter Weibull statistical analysis of ex-
perimental data points showed unique characteristic hardness
(31.4 GPa) and toughness (4.21 MPa-m0.5) values of the sintered
specimen with adequate structural reliability (Weibull moduli
ranged from 8 to 17).
ii. The mathematical models of Ra and machining speed were
found to be accurate because the predicted response para-
meters were quite close to the actual response parameters.
iii. It is observed from the ANOVA results of the individual input
factors, that the Pulse on-time and pulse peak current have
strong effect on the response Ra and pulse on-time, water
pressure and servo feed rate have the strong effect on ma-
chining speed respectively.
iv. While, the most significant machining parameter for Ra was
found to be pulse peak current followed by pulse on-time,
servo feed rate, pulse off-time and water pressure from the (S/
N) ratio analysis, for machining speed, the pulse off-time was
found to be the most important followed by servo feed rate,
pulse peak current, pulse on-time and water pressure.
v. The Grey Taguchi analytical technique was utilized for WEDM
parameter optimization of pure B4C ceramic to get the best
machining responses. Based on Grey relational analysis, the
optimized set of input parameters for machining of dense pure
B4C was 27 ms Ton, 48 ms Toff, 180 A IP, 7 kg/cm2 WP and SF of
2200 mm/min. To validate the optimization results, a new set of
3 experiments was conducted and the average output re-
sponses i.e., machining speed and surface roughness were
found to be close to the predicted values obtained from the
optimization model.
Acknowledgements
The authors like to thank the Director, CSIR-Central Glass and
Ceramic Research Institute, Kolkata, India for his kind permission
to carry out the research work. The co-operation of the staff
members of NOCC Division, of CSIR-CGCRI, Kolkata, India and IIEST,
Shibpur, India for performing the experiments are thankfully ac-
knowledged. Financial assistance of CSIR, India under ESC-0104 to
carry out the research is also acknowledged.
References
[1] B.M. Moshtaghioun, D.G. García, D.A. Rodríguez, High-temperature creep de-
formation of coarse-grained boron carbide ceramics, J. Eur. Ceram. Soc. 35
(2015) 1423–1429.
[2] V. Domnich, S. Reynaud, A.H. Richard, M. Chhowalla, Boron Carbide: Structure,
Properties, and Stability under Stress, J. AM. Ceram. Soc. 94 (2011) 3605–3628.
[3] C. Zhang, Effect of wire electrical discharge machining (WEDM) parameters on
surface integrity of nanocomposite ceramics, Ceram. Int. 40 (2014) 9657–9662.
[4] R. Bobbili, V. Madhu, A.K. Gogia, An experimental investigation of wire elec-
trical discharge machining of hot-pressed boron carbide, Def. Technol. 11
(2015) 344–349.
[5] N. Tosun, C. Cogun, G. Tosun, A study on kerf and material removal rate in wire
electrical discharge machining based on Taguchi method, J. Mater. Proc.
Technol. 152 (2004) 316–322.
[6] C. Tzeng, Y. Lin, Y. Yang, M.C. Jeng, Optimization of Turning Operations with
Multiple Performance Characteristics Using the Taguchi Method and Grey
Relational Analysis, J. Mater. Proc. Technol. 209 (2009) 2753–2759.
[7] J.L. Lin, C.L. Lin, The use of the orthogonal array with grey relational analysis to
optimize the electrical discharge machining process with multiple perfor-
mance characteristics, Int. J. Mach. Tools Manufac. 42 (2002) 237–244.
[8] K.M. Patel, P.M. Pandey, P.V. Rao, Determination of an optimum parametric
combination using a surface roughness prediction model for EDM of Al2O3/
SiC/TiC ceramic composite, Mater. Manuf. Process 24 (2009) 675–682.
[9] M.S. Phadke, Quality Engineering using robust design, AT & T Bell Laboratories,
PTR, Prentice-Hall Inc, 1989, pp. 231–249.
[10] R. Bobbili, V. Madhu, A.K. Gogia, Multi response optimization of wire-EDM
process parameters of ballistic grade aluminium alloy, Eng. Sci. Technol. Int. J.
18 (2015) 720–726.
[11] L. Xiaoguang, et al., Densification behavior and related phenomena of spark
plasma sintered boron carbide, Ceram. Int. 40 (2014) 4359–4366.
[12] K. Niihara, R. Morena, D.P.H. Hasselman, Evaluation of KICof Brittle Solids by
the Indentation Method with Low-Crack-to-Indent Ratios, J. Mater. Sci. Lett. 1
(1982) 13–16.
[13] K. Niihara, A fracture-mechanics analysis of indentation induced Palmqvist
crack in ceramics, J. Mater. Sci. Lett. 2 (5) (1983) 221–223.
[14] G.D. Girolamo, F. Marra, C. Blasi, E. Serra, T. Valente, Microstructure, me-
chanical properties and thermal shock resistance of plasma sprayed nanos-
tructured zirconia coatings, Ceram. Int. 37 (2011) 2711–2717.
[15] S. Sarkar, P.K. Das, Statistical analysis of mechanical properties of pressure less
sintered multiwalled carbon nanotube /alumina nanocomposites, Mater.
Chem. Phys. 137 (2012) 511–518.
[16] K.P. Sanosh, A. Balakrishnan, L. Francis, T.N. Kim, Vickers and Knoop Micro-
hardness Behavior of Coarse and Ultrafine-grained Titanium, J. Mater. Sci.
Technol. 26 (2010) 904–907.
[17] S. Sarkar, A. Dey, P.K. Das, A.K. Mukhopadhyay, A. Kumar, Evaluation of Mi-
cromechanical Properties of Carbon/Carbon and Carbon/Carbon–Silicon Car-
bide Composites at Ultralow Load, Int. J. Appl. Ceram. Technol. 8 (2011)
282–297.
[18] G.D. Quinn, J.B. Quinn, A practical and systematic review of Weibull statistics
for reporting strengths of dental materials, Dent. Mater. 26 (2010) 135–147.
[19] J.B. Saedon, N. Jaafar, Md. A. Yahaya, N.H. Saad, MD. S. Kasim, Multi-Objective
Optimization of Titanium Alloy through Orthogonal Array and Grey Relational
Analysis in WEDM, Proc. Technol. 15 (2014) 832–840.
[20] I. Puertas, C.J. Luis, A study of optimization of machining parameters for
electrical discharge machining of boron carbide, Mater. Manuf. Process. 19
(2004) 1041–1070.
[21] T.P. Bagchi, Taguchi Methods Explained: Practical Steps To Robust Design
ISBN-0-87692-80804, Prentice Hall Of India Private Ltd., 1993, pp. 11–19.
[22] R. Hicks, V.K. Turner, Fundamental Concepts in the Design of Experiments
(Fifth Edition), Oxford University Press, 1999, pp. 134–139.
[23] Y.F. Hsiao, Y.S. Tarng, W.J. Huang, Optimization of plasma arc welding para-
meters by using the Taguchi method with the Grey relational analysis, Mater.
Manuf. Process. 23 (2008) 51–58.
[24] A.R. Vikas, K. Kumar, Kumar, , Effect and Optimization of various Machine
Process Parameters on the Surface Roughness in EDM for an EN41 Material
using Grey-Taguchi, Proc. Mater. Sci. 6 (2014) 383–390.
[25] S. Datta, A. Bandyopadhyay, P.K. Pal, Grey based Taguchi method for optimi-
zation of bead geometry in submerged arc bead-on-plate welding, Int. J. Adv.
Manuf. Technol. 39 (2008) 1136–1143.
[26] S. Dewangan, S. Gangopadhyay, C.K. Biswas, Multi-response optimization of
surface integrity characteristics of EDM process using grey-fuzzy logic-based
hybrid approach, Eng. Sci. Technol, Int. J. 18 (2015) 361–368.
[27] L.M. Maiyar, R. Ramanujam, K. Venkatesan, J. Jerald, Optimization of ma-
chining parameters for end milling of Inconel 718 super alloy using Taguchi
based grey relational, Anal., Proc. ENG 6 (4) (2013) 1276–1282.
[28] B.M. Moshtaghioun, F. Luis, C. Hernández, D.G. García, S.D.B. Martín, D.
A. Rodríguez, A. Monshi, M.H. Abbasi, Effect of spark plasma sintering para-
meters on microstructure and room-temperature hardness and toughness of
fine-grained boron carbide (B4C), J. Eur. Ceram. Soc. 33 (2013) 361–369.
[29] S. Hayun, S. Kalabukhov, V. Ezersky, M.P. Darie, N. Frage, Microstructural
characterization of spark plasma sintered boron carbide ceramics, Ceram. Int.
36 (2010) 451–457.
[30] J. Yin, Z. Huang, X. Liu, Z. Zhang, D. Jiang, Microstructure, mechanical and
thermal properties of in situ toughened boron carbide-based ceramic com-
posites co-doped with tungsten carbide and pyrolytic carbon, J. Eur. Ceram.
Soc. 33 (2013) 1647–1654.
[31] C. Sun, Y. Li, Y. Wang, L. Zhu, Q. Jiang, Y. Miao, X. Chen, Effect of alumina ad-
dition on the densification of boron carbide ceramics prepared by spark
plasma sintering technique, Ceram. Int. 40 (2014) 12723–12728.
[32] M. Mashhadi, E.T. Nassaj, V.M. Sglavo, Pressureless sintering of boron carbide,
Ceram. Int. 36 (2010) 151–159.
[33] L. Farbaniec, J.D. Hoganand, K.T. Ramesh, Micromechanisms associated with
the dynamic compressive failure of hot-pressed boron carbide, Scr. Mater. 106
(2015) 52–56.