j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378

journal homepage: www.elsevier.com/locate/jmatprotec

Material removal rate (MRR) study in the cylindrical wire

electrical discharge turning (CWEDT) process

M.J. Haddad a,∗ , A. Fadaei Tehrani b

a CAD/CAM Laboratory, Manufacturing Engineering Division, School of Engineering, Tarbiat Modares University, Tehran, Iran
b Manufacturing Engineering Division, Department of Mechanical Engineering, Isfahan University of Technology, Isfahan, Iran

a r t i c l e i n f o a b s t r a c t

Article history: As using wire EDM (WEDM) technology, complicated cuts can made through difficult to
Received 27 November 2006 machine electrically conductive components, the cylindrical wire electrical discharge turn-
Received in revised form ing (CWEDT) process was developed to generate precise cylindrical forms on complicate,
4 August 2007 hard and difficult to machine materials. A precise, flexible and corrosion resistance sub-
Accepted 8 August 2007 merged rotary spindle was designed and added to a conventional five axis CNC WEDM
machine to enable the generation of free-form cylindrical geometries. The hardness and
strength of the work material are no longer the dominating factors that affect the tool wear
Keywords: and hinder the machining process. The right selection of machining conditions is the most
Wire EDM important aspect to take into consideration in process related to the WEDM operations.
CWEDT This paper presents an investigation on the effects of machining parameters on material
MRR removal rate (MRR) in cylindrical wire electrical discharge turning (CWEDT) process. In this
DOE research, CWEDT of AISI D3 (DIN X210Cr12) tool steel is studied by using of statistical design
ANOVA of experiment (DOE) method. AISI D3 tool steel was used in the experiments because of its
RSM growing range of applications in the field of manufacturing tools, dies and molds as punch,
AISI D3 tool steel tapping, reaming and so on in cylindrical forms. The effects of EDM parameters such as
power, voltage, pulse off time and spindle rotational speed has been analyzed on MRR by
using analysis of variance (ANOVA). A model has been developed for MRR by using response
surface methodology (RSM). In order to study surface integrity, SEM and micro-hardness
tests were carried out in different machining parameters.
© 2007 Elsevier B.V. All rights reserved.

1. Introduction corner radii (Ho et al., 2004). During the WEDM process, the
material is eroded ahead of the wire and there is no direct
Wire electrical discharge machining (WEDM) is a widely contact between the workpiece and the wire, eliminating the
accepted non-traditional material removal process used to mechanical stresses during machining. The hardness and
manufacture components with intricate shapes and profiles. strength of the difficult to machine work material are no
It is considered as a unique adaptation of the conventional longer the dominating factors that affect the tool wear and
EDM process, which uses an electrode to initialize the spark- hinder the machining process. This makes the WEDM process
ing process. However, WEDM utilizes a continuously traveling particularly suitable for machining hard, difficult to machine
wire electrode made of thin copper, brass or tungsten of diam- materials. In addition, the cutting force in WEDM process is
eter 0.05–0.3 mm, which is capable of achieving very small small, which makes it ideal for fabricating miniature parts

∗
Corresponding author. Tel.: +98 311 391 5235; fax: +98 311 391 2625.
E-mail address: mjhaddad@modares.ac.ir (M.J. Haddad).
0924-0136/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2007.08.020
370 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378

Nomenclature

Adj. MS adjusted mean squares

Adj. SS adjusted sum of squares
AD Anderson–Darling statistic
Coef. coefficient of factor and factor interaction in the
regression model
d.f. degree of freedom
DOE design of experiment
F F-test value
HAZ heat affected zone
HRC hardness Rockwell C
HV hardness Vickers
l cutting length (mm)
MRR material removal rate (mm3 /min)
MS mean squares
P probability value
r final radius of machined workpiece (mm)
RPM spindle rotational speed
R original radius of workpiece (mm) Fig. 1 – Illustration of the conventional and cylindrical wire
R2 R-squared statistic EDM. (a) Conventional wire EDM, (b) close-up view of the
Radj
2 adjusted R-squared statistic: correlation coeffi- gap and continuous electrical sparks and (c) cylindrical
cient wire EDT.
Ra surface roughness (␮m)
RSM response surface methodology
S root mean square parts using brass and carbide work material. Mohammadi
Seq. SS sequential sum of squares et al. (Mohammadi et al., 2005) investigated the turning by
S.E. Coef. standard errors for coefficients wire electrical discharge machining to evaluate the effects of
SS sum of squares machining parameters on MRR, surface roughness and round-
t machining time (min) ness by using Taguchi approach. Scott et al. (Scott et al., 1991)
T T-test value used a factorial design requiring a number of experiments
vf machine feed rate (machining cutting speed, to determine the most favorable combination of the WEDM
mm/min) parameters. They found that the discharge current, pulse
duration and pulse frequency are the significant control fac-
tors affecting the MRR and surface finish, while the wire speed,
(Schoth et al., 2005). One of the configurations of WEDM is wire tension and dielectric flow rate have the least effect. Liao
cylindrical wire electrical discharge turning (CWEDT). The et al. (Liao et al., 1997) proposed an approach of determining
concept of CWEDT is illustrated in Fig. 1. A rotary axis is the parameter settings based on the Taguchi quality design
added to a conventional five axis CNC wire EDM machine method and the analysis of variance. The results showed that
in order to produce cylindrical forms. The initial shape of the MRR and surface finish are easily influenced by the table
the part needs not to be a cylindrical form. The electrically feed rate and pulse on time, which can also be used to control
charged wire is controlled by the X and Y slides to remove the discharging frequency for the prevention of wire breakage.
the work material and generation of the desired cylindrical Huang and Liao (Huang and Liao, 2003) presented the use of
form (Qu et al., 2002a,b). Some turning wire EDM works have grey relational and S/N ratio analysis, which also display simi-
been reported for manufacturing small pins by Dr. Masuzawa’s lar results demonstrating the influence of table feed and pulse
research group at the University of Tokyo (Masuzawa et al., on time on the MRR. An experimental study to determine the
1985, 1994; Masuzawa and Tonshoff, 1997). The small diameter MRR and surface finish for varying machining parameters has
pins can be used as tools for 3D micro-EDM application (Qu et also been conducted (Rajurkar and Wang, 1993). The results
al., 2002b). Examples of the machined parts using the CWEDT have been used by presenting a thermal model to analyze the
method have shown in Refs. (Qu et al., 2002a,b; Masuzawa et wire breakage phenomena. In WEDM operations, MRR deter-
al., 1985, 1994; Masuzawa and Tonshoff, 1997; Mohammadi et mines the economics of machining and rate of production. In
al., 2005). Also, the application of a water-cooled submerged setting the machining parameters, the main goal is the maxi-
spindle extends the application of WEDM to cylindrical WEDM mum MRR. Literature lacks much to say about the use of wire
turning with rotation speed of up to 2800 rpm. This enables EDM for machining cylindrical forms (CWEDT) of AISI D3 tool
the production of gear wheels with integrated shaft for easy steel material, so the need has been felt towards the highlight-
gear assembly (Masuzawa et al., 2002). Qu et al. (2002a) derived ing of this process with the goal of achieving mathematical
a mathematical model for the material removal rate (MRR) models to enhance the process performance. The selection of
of the CWEDT process. The same authors (Qu et al., 2002b) this material was made taking into account its wide range of
investigated the surface integrity and roundness of CWEDT application in tools, dies and molds industries as punch, tap-
 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378 371

ping, reaming and so on in cylindrical forms. The present work

highlights the development of mathematical model for cor-
relating the inter-relationships of various CWEDT machining
parameters such as power, pulse off time, voltage and spin-
dle rotational speed on the material removal rate. A proper
design of experiments (DOE) was conducted to perform more
accurate, less costly and more efficient experiments. Full fac-
torial design technique is used widely in DOE (Montgomery,
2000) and was employed in this research to perform the exper-
imental design. Three different analyses were performed on
the data obtained from experiments where the interpretation
Fig. 3 – A sample of punches used for machining with
of their results is of great importance in any experimental
CWEDT process.
work. First, analysis of variance (ANOVA) was conducted to
help one to determine significant factors. Secondly, regres-
sion analysis was used to establish a relationship between
factors and responses by using response surface methodol- Table 1 – Factors, factor levels and factor designation
ogy (RSM). Thirdly, analysis of variance for regression was Factors Factor levels
applied to determine the accuracy of regression model (Myers
1 2 3
and Walpole, 1978). Surface integrity has been also studied
by scanning electron microscope (SEM) and measuring micro- Power: for each power level 10 11 –
hardness of layers to examine and investigate the sub-surface there is a corresponding
average current between
recast layers and heat affected zones (HAZ) of machined parts
wire and workpiece (level of
under specific process parameters.
current)
Pulse off time: time interval 6 8 10
between one discharge and
2. Experimental setup
the next (␮s)
Voltage: indicated potential 100 110 120
The experiments were run on a wire EDM machine type ONA R250 difference during ionization
CNC. The wire EDM machine was equipped with a rotary axis in order of the gap (V)
to produce cylindrical forms (Fig. 2). The experiments were aimed at Spindle rotational speed (rpm) 16 45 –
considering effects of several controllable factors on MRR. The straight
turning configuration was used during the tests. Theoretical equation
Hardness of work pieces (with 10 mm diameter) was 62 ± 2 HRC. Depth
(1) can be derived to describe the calculation for MRR:
and length of cut in all experiments were 2 and 10 mm, respectively.
According to initial tests done by Mohammadi et al. (Qu et al., 2002a;
MRR = (R2 − r2 )f (1) Mohammadi et al., 2005) using Taguchi approach in design of experi-
ments (DOE) and application of L18 (21 × 37 ) standard orthogonal array
where R is the original radius of the workpiece, r the final radius of the and primary tests accomplished on AISI D3 tool steel specimens by the
workpiece after machining and vf is the machining cutting speed or authors using ONA Aricut R250 technology manual and user’s guide,
feed rate. In this case for calculation of vf the equation, vf = l/t was used and L9 (34 ) Taguchi standard orthogonal array, factors and factor levels
where t is the machining time during the cutting length (l) on cylindri- were selected for determining reasonable process setting conditions,
cal surface of workpiece (Qu et al., 2002a). The specimens choosen in significant factors and their interactions (Table 1). Power, pulse off time,
this case were AISI D3 tool steel punches due to their growing range voltage and spindle rotational speeds are adopted as factors (inde-
of applications in the field of manufacturing dies and molds (Fig. 3). pendent variables) which vary during the experiments. None-variable
factors are presented in Table 2. These factors are set apart from the
experiment, and they are neither presumed to have important effect
on the process, nor can vary because of the equipment setup (Qu et
al., 2002a; Mohammadi et al., 2005). One of the most used techniques
for the design of experiments is the factorial design; which consists
of experimenting with all the possible combinations of variables and
levels. If the number of design factors’ levels is different, a kind of full
factorial design and mixed level design, is used. These mixed level fac-

Table 2 – Non-variable parameters in this research

Factors Factor levels

Maximum feed rate (mm/min) 1

Servo (V) 50
Wire tension (kg) 16
Wire speed (mm/s) 8
Dielectric 31
Inverse, finish Off
Fig. 2 – Spindle in five axis WEDM machine.
372 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378

Fig. 4 – Effects of factors on MRR. Fig. 5 – Interaction effects of factors on MRR.

torial designs are among the most widely used types of designs for little effect on MRR. Voltage, like power, has significant direct
process design and process improvement. In this research a mixed level effect on MRR, but in high values of voltage, MRR increases
design of 22 × 32 was designed for experimentation. Thus, 36 experi-
sharply. Also, Fig. 4 presents that increasing power results in
ments were conducted with parameter levels shown in Table 1. Each
significant MRR increase.
run is replicated two times so that the total number of runs was 72.
Although graphical assessment is the most intuitive means
The resolution of this factorial design allows us to estimate all the main
effects, factor interactions in this study. Note that the experiments were of effect consideration, the inferences made based on it is not
run in random order using randomization table, even the repetition of accurate thus cannot be reliable. Only in the case of com-
each run was not done, respectively. parison, the plot of factor effects is allowed to be used. The
interaction plot for the MRR can be observed in Fig. 5. As
has been stated, the strongest interactions are those between
3. Data analysis and discussion power and pulse off time, voltage and spindle rotational speed,
pulse off time and power.
Fig. 4 depicts the plot of factor effects on MRR. This plot can be ANOVA is used here to test the null hypothesis with regard
used to graphically assess the effects of factors on response. to the data gained through experiments. Through null hypoth-
Fig. 4 indicates that power, voltage, pulse off time and spin- esis it is assumed that there is no difference in treatment
dle rotational speeds have the most significant effect on MRR. means (H0 : 1 = 2 = · · · = a ). Table 3 is ANOVA table for MRR.
Furthermore, it is seen from Fig. 4 that pulse off time is recipro- Before any inferences can be made based on ANOVA table,
cally proportional to MRR. Also, Fig. 4 presents that increasing the assumptions used through ANOVA process have to be
power results in higher MRR. Spindle rotational speed presents checked.

Table 3 – ANOVA for MRR

Source d.f. Seq. SS Adj. SS Adj. MS F P

Power 1 39.3667 39.3667 39.3667 114.52 0

Voltage 2 52.9639 52.9639 26.482 77.04 0
Pulse off time 2 9.7249 9.7249 4.8625 14.15 0
Spindle rotational speed 1 2.6531 2.6531 2.6531 7.72 0.009
Power × voltage 2 2.1556 2.1556 1.0778 3.14 0.056
Power × pulse off time 2 2.7977 2.7977 1.3988 4.07 0.026
Power × spindle rotational speed 1 0.1816 0.1816 0.1816 0.53 0.472
Voltage × pulse off time 4 0.3552 0.3552 0.0888 0.26 0.903
Voltage × spindle rotational speed 2 2.5551 2.5551 1.2776 3.72 0.034
Pulse off time × spindle rotational speed 2 2.6742 2.6742 1.3371 3.89 0.03
Power × voltage × pulse off time 4 0.4091 0.4091 0.1023 0.3 0.878
Power × voltage × spindle rotational speed 2 0.7159 0.7159 0.358 1.04 0.363
Power × pulse off time × spindle rotational speed 2 3.7201 3.7201 1.8601 5.41 0.009
Voltage × pulse off time × spindle rotational speed 4 1.0683 1.0683 0.2671 0.78 0.547
Power × voltage × pulse off time × spindle rotational speed 4 0.1083 0.1083 0.0271 0.08 0.988

Error 36 12.375 12.375 0.3437

Total 71 133.8248
 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378 373

Fig. 7 – Residuals vs. fitted values.

Fig. 6 – Normal plot of residuals.

Fig. 6 shows the normal plot of residuals. This plot is used to Table 4 – R2 and Radj
2 test for MRR regression model

test the normal distribution of errors. The distribution shown

Degree R2 (%) Radj
2 (%)
in Fig. 6 presents that the error normality assumption is valid.
The other two assumptions are shown to be valid by means of Linear 76.4 75
plot of residuals versus fitted values. This plot is illustrated in Linear + squares 78.2 76.2
Fig. 7. Linear + interaction 80.2 77
Full quadratic 82.1 78.4
The structureless distribution of dots above and below the
abscissa (fitted values) in Fig. 7 illustrates both the error inde-
pendency and variance constancy (Montgomery, 2000). rotational speed exhibits significant effect on MRR as well. The
Now since the assumptions are proved not to be violated effects of interaction between power and spindle rotational
through this experimentation ANOVA results which are listed speed, voltage and pulse off time, power and voltage and pulse
below are reliable. Confidence level is chosen to be 95% in this off time, power and voltage and spindle rotational speed, volt-
study. So the P-values which are less than 0.05 indicate that age and pulse off time and spindle rotational speed, power and
null hypothesis should be rejected and thus the effect of the voltage and pulse off time and spindle rotational speed have
respective factor is significant. It can be seen from Table 3 that no effect on MRR. The effects of interaction between power
power, voltage and pulse off time (P = 0.000) have the most sig- and voltage are on the verge of significance as it is indicated
nificant impact on MRR. Spindle rotational speed (P = 0.009) in Table 3.
has less effect than power, voltage and pulse off time. How- Regression analysis is performed to find out the relation-
ever, still seems to be significant. ship between factors and MRR. Table 4 indicates that full
The interaction effects between power and pulse off time, quadratic model is the best one in comparison with the oth-
voltage and spindle rotational speed, pulse off time and spin- ers that can be used with this factors and factor levels by Radj
2

dle rotational speed, power and pulse off time and spindle test. The R2 value indicates that the predictors explain 82.1% of

Table 5 – First regression model for MRR

Term Coefficient S.E. coef. T P

Constant 18.1541 0.16804 67.558 0

Power 2.2902 0.07515 9.839 0
Voltage −0.7931 0.09204 −11.167 0
Pulse off time 2.4353 0.09204 4.832 0
Spindle rotational speed 0.1788 0.07515 2.554 0.013
Voltage × voltage 0.0038 0.15942 2.354 0.022
Pulse off time × pulse off time −0.0301 0.15942 −0.755 0.453
Power × voltage 0.0105 0.09204 0.573 0.569
Power × pulse off time −0.22 0.09204 −2.39 0.02
Power × spindle rotational speed −0.0069 0.07515 −0.668 0.507
Voltage × pulse off time 0.0001 0.11273 0.026 0.979
Voltage × spindle rotational speed −0.0014 0.09204 −2.152 0.036
Pulse off time × spindle rotational speed 0.0039 0.09204 1.218 0.228

S = 0.6377; R2 = 82.1%; Radj

2
= 78.4%.
374 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378

Table 6 – Second regression model for MRR

Term Coefficient S.E. coef. T P

Constant 8.96796 20.5238 0.437 0.664

Power 3.23882 0.7387 4.385 0
Voltage −0.68118 0.3454 −1.972 0.053
Pulse off time 2.08759 0.951 2.195 0.032
Spindle rotational speed 0.13699 0.0688 1.991 0.051
Voltage × voltage 0.00375 0.0016 2.395 0.02
Power × pulse off time −0.21999 0.0905 −2.432 0.018
Voltage × spindle rotational speed −0.00137 0.0006 −2.189 0.032

S = 0.6268; R2 = 81.2%; Radj

2
= 79.2%.

Table 7 – ANOVA for second regression analysis for MRR

Source d.f. Seq. SS Adj. SS Adj. MS F P

Regression 7 108.682 108.682 15.526 39.52 0

Linear 4 102.223 51.744 12.936 32.93 0
Square 1 2.253 2.253 2.2534 5.74 0.02
Interaction 2 4.205 4.205 2.1027 5.35 0.007
Residual error 64 25.143 25.143 0.3929
Lack-of-fit 28 12.768 12.768 0.456 1.33 0.21
Pure error 36 12.375 12.375 0.3437

Total 71 133.825

the variance in MRR. The Radj 2 is 78.4%, which accounts for the (RSM), is shown in the following equation:
number of predictors in the model. Table 5 shows the coeffi-
cients of factors and factor effects in regression model. It can MRR (mm3 / min) = 8.96796 + 3.23882 × power − 0.68118
be seen from Table 5 that some interaction effects have no
significant effect in regression model, because their P-value is × voltage + 2.08759 × pulse off time
higher than 0.05, accepting that there is statistical evidence + 0.13699 × spindle rotational speed
of curvature in the first-order model, for a confidence level of
95%. Thus, since the proposed first-order model is suitable for + 0.00375 × voltage × voltage − 0.21999
a significance level ˛ of 0.05 is rejected and so, the second- × power × pulse off time − 0.00137
order model is selected. Table 6 shows the regression table for
the case of the proposed second-order model, where now, the × voltage × spindle rotational speed (2)
total number of degrees of freedom is equal to 71. As can be
seen in Table 5, all effects have a P-value less than 0.05, which
means that they are significant for a confidence level of 95%. Table 8 shows the verification of the test results. The
On the other hand, a value of 0.812 was obtained for the R2 - predicted machining parameters performance was compared
statistic, which signifies that the model explains 81.2% of the with the actual machining performance and a good agreement
variability of MRR, whereas the adjusted R2 -statistic (Radj2 ) is
was obtained between these performances. The above mathe-
0.792. The ANOVA table for the MRR using the fitted model matical model for MRR of CWEDT AISI D3 tool steel is of great
with linear, square and interaction terms is shown in Table 7. importance to the proper selection of machining parameters
This table indicates that all terms of the regression model during the machining of the cylindrical parts.
are significant at the confidence level of 95%. In this way, Fig. 8 shows the estimated response of MRR, varying the
the simplified model which presents the highest value for the factors of pulse off time and spindle rotational speed. As can
adjusted R2 -statistic by using response surface methodology be clearly seen in this figure, the MRR value tends to increase

Table 8 – Results of confirmation tests for MRR model

Run Power Voltage Pulse off Spindle rotational Results of Results of Error of the
time speed regression model experiments model (%)

1 11 120 6 16 14.42 14,916 −3.34

2 10 100 10 45 9.62 9,475 1.5
3 11 110 8 45 11.76 11,948 −1.6
4 12 105 12 90 10.41 10,453 −0.38
 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378 375

Fig. 10 shows the estimated response surface for the MRR

parameter, according to the design parameters of power and
pulse off time, while the other factors remain constant in their
central values. As has been previously pointed out, this figure
shows the important influence that the design factor of power
possesses over the MRR parameter, so that when power is
increased, the MRR also tends to increase appreciably at least
up to a maximum value. Afterwards MRR tends to decrease,
for high values of the pulse off time factor and within the con-
sidered work interval. Furthermore, it can also be observed
that the MRR parameter tends to increase when the pulse off
time factor is decreased, especially for high values of power.
Fig. 8 – Estimated response surface of MRR vs. pulse off
The previous tendency of growth for this factor becomes less
time and spindle rotational speed.
intense as power decreases, with the MRR parameter actually
decreasing slightly, after reaching a peak, for values close to
the low level of power.

4. SEM results

SEM is used to examine the surface integrity and sub-surface

of CWEDT of AISI D3 tool steel. The cylindrical samples were
sliced in the radial direction. The surfaces of sliced cross-
sections were polished and etched to observe the sub-surface
damage. Although regular EDMed surfaces are isotropic and
Fig. 9 – Estimated response surface of MRR vs. voltage and have no specific texture or pattern (Qu et al., 2002b), CWEDT
spindle rotational speed. surfaces may have macro-ridges or circular arcs, in the cross-

with the lower pulse off time and spindle rotational speed val-
ues. In the other hand, by decreasing spindle rotational speed
values in the higher values of pulse off time, the MRR val-
ues increase. Also in the higher values of spindle rotational
speed by decreasing the pulse off time values, the MRR values
increases.
Fig. 9 shows the estimated response surface of MRR in func-
tion of the factors of voltage and spindle rotational speed,
while the pulse off time and power remains constant in their
central values. It is indicated that by decreasing the spindle
rotational speed values and increasing the voltage values, the
MRR maximizes. This figure shows the limited influence that
spindle rotational speed possesses over MRR for lower values
of voltage, but in the higher values of voltage by decreasing
the spindle rotational speed values, the MRR increases. In the
higher values of spindle rotational speed by increasing voltage,
the MRR increases, respectively.

Fig. 11 – SEM micrographs of macro-ridges and ideal arcs

on surfaces under high MRR = 14.916 (power = 11,
Fig. 10 – Estimated response surface of MRR vs. pulse off voltage = 120, pulse off time = 6 and spindle rotational
time and power. speed = 16).
376 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378

Fig. 13 – SEM micrographs of surfaces and cross-sections

under high MRR = 14.916 (power = 11, voltage = 120, pulse off
time = 6 and spindle rotational speed = 16).

thermal model to predict the thicknesses of damaged layers.

For the die-sinking EDM process, the depth of the damaged
layer was reported to be from 30 to 100 ␮m for an AISI 4130 steel
workpiece machined with pulse on-time of 100–300 ␮s. Qu et
al. (2002b) has studied the sub-surface heat affected zones
Fig. 12 – SEM micrographs of macro-ridges and ideal arcs (HAZ) and recast layers of carbide and brass in cylindrical wire
on surfaces under low MRR = 10.453 (power = 10, EDM. Under high MRR at 14 ␮s pulse on time, the recast layer,
voltage = 100, pulse off time = 10 and spindle rotational about 3 ␮m thick, can be clearly recognized on the surface of
speed = 45). carbide parts. Thinner recast layers, less than 2 ␮m, exist on
samples machined using shorter pulse on time. Bubbles can
be identified in the recast layers of carbide samples. The depth
section. Figs. 11 and 12 show the surfaces of two CWEDT of the heat affected zone (HAZ) is estimated to be about 4, 3
parts. and 2 ␮m on the carbide samples with 14, 5 and 2 ␮s pulse
As it is indicated in Fig. 11, in high MRR conditions, the dis- on time, respectively. Also very thin recast layers, about 1 ␮m,
charge energy released by increasing in the power and voltage can be observed on the cross-section of three brass samples.
values and lower pulse off time. Therefore, the molten mate- No heat affected zone (HAZ) can be recognized on brass sam-
rial of workpiece surface has high fluency. In this case, because ples, possibly due to the good thermal conductivity of brass.
of high cooling rate the big spherical grains can grow up in
order to subtraction of the surface energy. Fig. 12 shows that
in low voltage and power values and higher pulse off time,
the feasibility of big spherical grains invention does not exist.
This causes the smaller roughness and higher adherence in
the surface.
The recast layer is defined as the material melted by elec-
trical sparks and resolidified on the surface without being
ejected nor removed by flushing. Below the recast layer is
the heat affected zone (HAZ). For the tool steel material, the
chromium carbide melts and resolidifies in the heat affected
zone (HAZ). The molten chromium carbide dissolved in the
martensite base, therefore the chromium carbide grains have
no time to grow up and then this phase will be small. This
is observed in the SEM micrographs of the cross-section and
is used to identify the depth of heat affected zone (HAZ). In
the recast layer, the solution of chromium carbide is complete
and the phase is martenzite, respectively. Rajurkar and Pandit Fig. 14 – SEM micrographs of surfaces and cross-sections
(Rajurkar and Pandit, 1984) have studied the recast layer and under low MRR = 10.453 (power = 10, voltage = 100, pulse off
heat affected zones (HAZ) of EDM surfaces and developed a time = 10 and spindle rotational speed = 45).
 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378 377

Table 9 – Results of micro-hardness measurement

Factors Results of experiments Micro-hardness results (HV)

Run Power Voltage Pulse off time Spindle rotational speed MRR Ra HAZ Bulk material

1 11 120 6 16 14.92 7.48 696.66 946

2 10 100 10 45 9.475 5.2 818.75 946
3 12 105 12 90 10.45 6.78 716.5 946

energy. The MRR is, in general, higher for the higher discharge
energy.

5. Discussion

In this work, a study on the influence of the most relevant

CWEDT factors over material removal rate (MRR) has been
carried out. The study has been made for AISI D3 tool steel
used in manufacturing tools, die and mold components, such
as punches, reaming, tapping and so on in cylindrical forms.
In order to achieve this, DOE and response surface methodol-
ogy (RSM) have been employed to model the above mentioned
response variables by means of equations in the form of poly-
nomials. The design finally chosen to accomplish the present
study was a mixed full factorial one of type 22 × 32 . The design
Fig. 15 – SEM micrographs of surfaces and cross-sections factors selected in this case were power, voltage, pulse off time
(power = 12, voltage = 105, pulse off time = 12 and spindle and spindle rotational speed, where all of them, except for the
rotational speed = 90). last one, are parameters widely used by the machinists to con-
trol the WEDM machine generator. Full quadratic model was
proposed for MRR from the results obtained in its correspond-
ing R2 and Radj
2 tests. In the case of MRR, the only influential
The heat affected zone (HAZ) may exist but cannot be identi- design factors and interaction effects, for a confidence level
fied in brass samples. SEM micrographs of the cross-section of of 95%, were power, voltage, pulse off time, spindle rotational
tool steel parts machined under both high and low MRR exper- speed and interactions between power and pulse off time, volt-
iments are shown in Figs. 13 and 14. The recast layer, craters in age and spindle rotational speed, pulse off time and spindle
the recast layer and heat affected zone (HAZ) of two-tool steel rotational speed, power and pulse off time and spindle rota-
samples are identified in Figs. 13 and 14. Under high MRR the tional speed. The variation tendencies of the latter were those
recast layer, about 40–50 ␮m thick can be clearly recognized on expected a priori and according to these, in order to obtain a
the surface. Thinner recast layers, less than 15–20 ␮m, exist high value of MRR within the work interval of this study, both
on samples machined by lower MRR conditions. As shown design factors, power and voltage, should be fixed as high as
in Fig. 14, the depth of the heat affected zone (HAZ) is esti- possible and the pulse off time and spindle rotational speed
mated to be about 20–25 and 8–13 ␮m on the two samples with factors, should be fixed as low as possible. The macro-ridges,
high and low MRR, respectively. It can be seen from Fig. 15 surface craters, recast layers and heat affected zones (HAZ)
that the discharge energy is lower than experiment shown in were observed, and their sizes were estimated using the SEM.
Fig. 14.
The micro-hardness of sample 1 is as highest MRR, sam-
references
ple 2 as best surface roughness and sample 3 with medium
MRR, were measured by an optical microscope. The hardness
of heat affected zone (HAZ) and bulk material was measured.
The load during measuring the micro-hardness was 981 mN. Ho, K.H., Newman, S.T., Rahimifard, S., Allen, R.D., 2004. State of
the art in wire electrical discharge machining (WEDM). Int. J.
Table 9 shows the results of micro-hardness measurement. It
Machine Tools Manuf. 44, 1247–1259.
can be seen that the hardness values have decreased in the Huang, J.T., Liao, Y.S., 2003. Optimization of machining
heat affected zones (HAZ). In the HAZ the carbide grains melt parameters of wire EDM based on grey relational and
and resolidify by high cooling rate. In this case, the carbide statistical analyses. Int. J. Prod. Res. 41 (8), 1707–1720.
grains have no time to grow up and sediment in the marten- Liao, Y.S., Huang, J.T., Su, H.C., 1997. A study on the machining
zite basis, with small grains. It can be seen from Table 5 that parameters optimization of wire electrical discharge
machining. J. Mater. Process. Technol. 71 (3), 487–493.
by increasing the discharge energy the grains grow up slowly
Masuzawa, T., Tonshoff, H.K., 1997. Three-dimensional
and therefore the hardness of the HAZ decreases. Discharge
micromachining by machine tools. CIRP Ann. 46, 621–628.
energy is an important EDM process parameter that affects Masuzawa, T., Fujino, M., Kobayashi, K., Suzuki, T., 1985. Study on
the thickness of the recast layer and HAZ. Higher power and micro-hole drilling by EDM. Bull. Jpn. Soc. Prec. Eng. 20 (2),
voltage with lower pulse off time generate higher discharge 117–120.
378 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 1 9 9 ( 2 0 0 8 ) 369–378

Masuzawa, T., Kuo, C.-L., Fujino, M., 1994. A combined electrical 1: concept, design, and material removal rate. J. Manuf. Sci.
machining process micronozzel fabrication. CIRP Ann. 43, Eng. 124 (3), 702–707.
189–192. Qu, J., Shih, A.J., Scattergood, R.O., 2002b. Development of the
Masuzawa, T., Okajima, K., Taguchi, T., Fujino, M., 2002. cylindrical wire electrical discharge machining process, part
EDM-lathe for micro-machining. CIRP Ann. 51, 2: surface integrity and roundness. J. Manuf. Sci. Eng. 124 (3),
355–358. 708–714.
Mohammadi, A., Fadaie Tehrani, A., Emanian, E., Karimi, D., Rajurkar, K.P., Pandit, S.M., 1984. Quantitative expressions for
Jafarpisheh, H., 2005. A new approach in turning by wire some aspects of surface integrity of electro-discharge
electrical discharge machining to evaluate the effects of machined components. ASME J. Eng. Ind. 106 (2), 171–177.
machining parameters on MRR. In: Proceedings of the Tehran Rajurkar, K.P., Wang, W.M., 1993. Thermal modeling and on-line
International Congress on Manufacturing Engineering, monitoring of wire EDM. J. Mater. Process. Technol. 38 (1–2),
Tehran, Iran, December 12–15. 417–430.
Montgomery, D.C., 2000. Design and Analysis of Experiments. Schoth, A., Forster, R., Menz, W., 2005. Micro-wire EDM high
John Willy & Sons Inc. aspect ratio 3D micro-structuring of ceramics and metals.
Myers, R.H., Walpole, R.E., 1978. Probability and Statistics for Micro Syst. Technol. 11, 250–253.
Engineers and Scientists. Macmillan Publishing Co. Inc., New Scott, D., Boyina, S., Rajurkar, K.P., 1991. Analysis and
York. optimization of parameter combination in wire electrical
Qu, J., Shih, A.J., Scattergood, R.O., 2002a. Development of the discharge machining. Int. J. Prod. Res. 29 (11), 2189–2207.
cylindrical wire electrical discharge machining process, part