Int J Adv Manuf Technol

DOI 10.1007/s00170-012-4459-7

ORIGINAL ARTICLE

Improving the surface quality in wire electrical discharge

machined specimens by removing the recast layer
using magnetic abrasive finishing method
Sasan Khalaj Amineh & Alireza Fadaei Tehrani & Aminollah Mohammadi

Received: 14 March 2012 / Accepted: 8 August 2012

# Springer-Verlag London Limited 2012

Abstract This study explores the feasibility of removing Experimental results show the influence of MAF process on
the recast layer formed on aluminum alloy cylindrical speci- recast layer removal and surface roughness improvement.
mens machined by wire electrical discharge machining
(WEDM) by using magnetic abrasive finishing (MAF). Keywords MAF . WEDM . Recast layer . DOE . ANOVA
The WEDM is a thermal machining process capable of
accurately machining parts with high hardness or complex
shapes. The sparks produced during the WEDM process 1 Introduction
melt the metal’s surface. The molten material undergoes
ultra-rapid quenching and forms a layer on the surface Wire electrical discharge machining is a widely accepted
defined as recast layer. The recast layer may be full of nontraditional material removal process used to manufacture
craters and microcracks which reduce service life of materi- components with intricate shapes and profiles. During the
als tremendously, especially under fatigue loads in corrosive wire electrical discharge machining (WEDM) process, the
environments. This investigation demonstrates that MAF material is eroded ahead of the wire and there is no direct
process, can improve the quality of WEDM machined sur- contact between the workpiece and the wire, eliminating the
faces effectively by removing the recast layer. The present mechanical stresses during machining [1]. WEDM process
work studies the effect of some parameters, i.e., linear makes use of electrical energy generating a channel of
speed, working gap, abrasive particle size, and finishing plasma between the cathode and anode [2], and turns it into
time on surface roughness and recast layer thickness using thermal energy [3] at a temperature in the range of 8,000–
full factorial analysis. Three-level full factorial technique is 12,000 °C [4] or as high as 20,000 °C [5] initializing a
used as design of experiments for studying the selected substantial amount of heating and melting of material on
factors. In order to indicate the significant factors, the anal- the surface of each pole. A portion of this molten material is
ysis of variance has been used. In addition, an equation then ejected and flushed away. The remaining material
based on regression analysis is presented to indicate the resolidifies rapidly on the surface of the work piece. The
relationship between surface roughness and recast layer accumulated new structure on the surface is called the
thickness of cylindrical specimens and finishing parameters. “recast layer” due to its nonrevelation by etching [6, 7].
The recast layer formed by WEDM process increases surface
roughness, makes the surface hard and brittle, and decreases
the fatigue strength due to the presence of microcracks and
S. Khalaj Amineh (*) : A. Fadaei Tehrani : A. Mohammadi microvoids [8].
Department of Mechanical Engineering, Removing the damaged surface in a postmachining pro-
Isfahan University of Technology,
Isfahan 84156-83111, Iran cess greatly increases fabricating time and cost. Therefore,
e-mail: [email protected] this study presents magnetic abrasive finishing (MAF) pro-
A. Fadaei Tehrani cess to remove the recast structure for removing microcrack
e-mail: [email protected] and discharge craters. In MAF, the material is removed from
A. Mohammadi the surface of workpiece, by the circulation of the abrasive
e-mail: [email protected] particles in the magnetic field [9–11]. These particles form a
 Int J Adv Manuf Technol

flexible magnetic abrasive brush which does not require

dressing. The MAF apparatus neither needs a very precise
worktable nor a very stiff structure since its cutting tool is a
unique flexible magnetic brush. Nevertheless, a mirror like
refined surface of high quality can be obtained easily. Fin-
ished surface neither showed a deteriorated layer nor micro-
cracks. MAF yielded better surfaces, especially of complex
shapes [12–14].
Until now, several researches have been done in order to
clarify the effective parameters in MAF process and its Fig. 2 Two-dimensional schematic of abrasive behavior during the
results such as surface roughness and amount of material process [20]
removal rate. Shinmura et al. have studied basic principle of
the MAF process and concluded that the stock removal and
surface finish value (Ra) increase as the magnetic abrasive revealed that the recast layer formed on Inconel 718 alloy
particle diameter “D” increases [15]. can be effectively softened by phosphoric acid coupled with
By using the ball-shaped magnetic abrasive pole with hydrochloric acid and the softened recast layer can be easily
special grooves, Lin indicated an adjunction in finishing effi- removed by a steel brush. Schumacher suspended special
ciency during the magnetic abrasive finishing of free-form powders in the dielectric fluid as a mean of improving surface
surfaces. The researcher found that working gap had the properties. The powder particles facilitate ignition process by
largest impact on the finishing quality [16]. It was confirmed creating a higher discharge probability and lowering break-
that automation of three-dimensional die and mold surface down strength of the insulating dielectric fluid [19].
polishing is possible using magnetic polishing tool by Kim et Most of the researchers have been using the MAF process
al. in 2007. They concluded that the polishing pressure can be to improve the surface roughness and material removal rate.
estimated by equivalent magnetic circuit and, thus, can be But hardly any information is available about removing the
changed to the desirable value by changing the working gap recast layer in WEDM machined surfaces. This study indi-
[17]. On the other hand, Wang and his colleagues [18] cates that the recast layer can be removed by MAF process
using steel grits as an abrasive and permanent magnets as a
magnetic flux supply. Figure 1a shows the workpiece with
Fig. 1 Cross-sectional of micro- its recast layer before finishing and Fig. 1b shows it after
samples (a) before the MAF finishing for a period of time. In addition, other approaches
process and (b) after the did not adequately pay attention to surveying the aluminum
MAF process
alloys workpieces while they have growing range applica-
tion in modern industries. Furthermore, in comparison with
the other conventional methods and with regard to the built-
up edge produced during polishing of ductile material, MAF
is suitable for inner surface polishing of aluminum alloy
tubes. The present work uses full factorial experimental
design to statistically investigate the effect of working gap,
rotational speed, finishing time, and abrasive particle size on

Fig. 3 External view of experimental setup

Int J Adv Manuf Technol

Table 1 Nonvariable parameters for the WEDM process Table 2 Nonvariable factors for the MAF process

Factors Factor levels Factors Factor level

Voltage (V) 110 Magnetic abrasive powders Carbon steel grits

Pulse off time (μs) 35 Magnetic poles Permanent poles (20×10×10 mm)
Power, for each level there is a 11 Lubricant SAE 40 oil (0.1 ml)
corresponding average current between
wire and workpiece (level of current)
Maximum feed rate (mm/min) 1
cylindrical specimens or for plates is microcutting and
Servo (V) 30
scratching [20, 21]. The abrasive behavior is primarily de-
Wire tension (Kg) 16
termined by the relative strengths of two opposing forces:
Wire speed (mm/s) 8
the magnetic force and the friction force acting on the
Dielectric 31
abrasive against the inner surface of the tube. When the
magnetic force is greater than the friction force, the abrasive
follows the magnetic field. This generates relative motion of
recast layer thickness and surface roughness. Since recast the abrasive against the inner surface, removing material
layer can be removed by MAF, similar procedures can be from the workpiece. The magnetic force on the abrasive is
transferred to recast layer removal in the WEDM machined the predominant component of finishing force, which is the
workpiece with different contour. result of the magnetic and centrifugal forces experienced by
the rotating magnetic abrasive [22].

2 Experimental setup 2.2 Magnetic abrasive finishing device

2.1 Abrasive behavior characteristic in the process According to the characteristic abrasive behavior in the
process, a finishing experimental device was designed and
Figure 2 shows a diagram of the abrasive behavior during made. Four rare earth permanent magnets were fixed in the
the internal magnetic abrasive finishing process using a pole inner surface of magnetic yoke. The abrasive particles are
rotation system for a nonferromagnetic workpiece. In the divided into two parts in the magnetic field. The angle
process, magnetic abrasives introduced into the workpiece between the two magnetic poles where magnetic abrasive
are attracted by the magnetic field and bear on the inner conglomerated is 90°. Figure 3 shows the external view of
surface of the workpiece. The rotation of the poles, which experimental setup. This setup with its embedded motor as a
consist of small permanent magnets, around the workpiece yoke driver can be a portable finishing mechanism. On the
causes the rotation of the magnetic field at the finishing area. other hand, it can be mounted on the cross slide of a lathe
The mechanism of finishing in MAF process, either for (AJ-725) using its spindle to rotate the workpiece.

2.3 Procedure of experiments

In recent years, aluminum alloys have attracted attention of

engineers, researchers, and designers due to their growing
range of usage in the automotive industries and aerospace
applications [23, 24]. Particularly, 6xxx aluminum alloys
that have advantages such as corrosion resistance, medium

Table 3 Variable factors

Factors Factor levels

Working gap (air gap between 1 2 –

the poles and the surface
of work piece; mm)
Linear speed (m/s) 0.2 0.4 0.6
Finishing time (min) 5 12 19
Fig. 4 Scanning electron microscope photograph shows the recast Mesh size 60 120 –
layer structure of the WEDM machined surface
 Int J Adv Manuf Technol

Fig. 5 The surface

morphologies of specimens a
before MAF, b after 5 min
finishing with 0.4 m/s linear
speed, c after 12 min finishing
with 0.4 m/s linear speed, d
after 12 min finishing with
0.6 m/s linear speed

strength, and low cost, comparing to other aluminum alloys magnetic abrasive finishing process on recast layer thick-
[13, 23, 24]. The material used in this study was 6062 ness and surface roughness. The surface roughness is mea-
aluminum alloy. sured using a mobile roughness measurement (Mahr
In order to conduct the experiments, the specimens were Perthometer M2) with a 0.8 mm cutoff length (according
machined, using WEDM system (ONA R250) and cylindri- to DIN EN IS 0274:1998) and the scanning electron micro-
cal workpieces with 26.5 mm inner radiuses and 0.65 mm scope system is triggered to observe the thickness of the
thickness were prepared. The operating conditions of recast layer (Fig. 4). Multiple observations on surface
WEDM process were fixed for all of the specimens which roughness and thickness are taken and an average of these
are shown in Table 1. observations is taken as the final value of response. Non-
In this study, the designed experiments are aimed at variable factors are presented in Table 2. These factors
considering the effect of several controllable factors of denote the experiments conditions.

Fig. 6 a Normal probability plot and b residuals versus fitted values plot for ΔRa
Int J Adv Manuf Technol

Table 4 Analysis of variance

for ΔRa, using adjusted SS Source DF Seq SS Adj SS Adj MS F P
for tests
Working gap 1 0.45024 0.45024 0.045024 8.77 0.009
Linear speed 2 0.78055 0.78055 0.39028 7.6 0.005
Finishing time 2 2.58698 2.58698 1.29349 25.2 0.000
Mesh size 1 4.0656 4.0656 4.0656 79.2 0.000
Working gap×linear speed 2 0.2798 0.2798 0.1399 2.73 0.096
Working gap×finishing time 2 0.07308 0.07308 0.03654 0.71 0.506
Working gap×mesh size 1 0.41431 0.43431 0.41431 8.07 0.012
Linear speed×finishing time 4 0.30962 0.30962 0.07741 1.51 0.247
Linear speed×mesh size 2 0.36172 0.36172 0.18086 3.16 0.07
Finishing time×mesh size 2 0.36399 0.36399 0.182 3.55 0.053
Error 16 0.82134 0.82134 0.05133
Total 35 10.5095
R2 082.99 % Adj R2 092.22 %

3 Design of experiments 4 Experimental result and discussion

Factors and their levels are shown in Table 3. Rotational speed, 4.1 Data analysis
working gap, finishing time, and mesh size are adopted as
factors (independent variables) which vary during the experi- Figure 5 shows the surface morphologies of the specimens
ments. In order to perform more accurate and more efficient before and after finishing. The WEDM surface before fin-
experiment, a proper design of experiments is needed. ishing is full of droplets and has a porous morphology (see
Full factorial design is one of the design of experiments Fig. 5a). The recast layer would be relatively removed and
techniques which provide the opportunity of experimenting the surface would be relatively smoother after the finishing
with all the possible combination of variable and levels. The process (see Fig. 5b–d).
potential to build models and to gradually increase the
complexity of the models is one of the most important 4.1.1 Analysis of surface roughness
capabilities of factorial design of experiments. A kind of
full factorial design is mixed level design which is used One of the most powerful tools in statistical inference for the
when the number of the design factors’ level is different. comparison of products is analysis of variance. This method
In this research, a mixed full factorial design of 22 ×32 is quite robust to non-normality of the data. Although in
has been selected due to the number of factors and factor comparison with the other methods, analysis of variance
levels considered in the study. Subsequently, 36 experiments (ANOVA) tends to be less affected by violations of the
are conducted at parameter levels. The resolution of this model assumptions.
factorial design makes contribution to estimate all the main Before using ANOVA results, the assumptions used
effects and factors interactions in this study. through this technique should be checked. Figure 6a shows

Table 5 Modified analysis of

variance for ΔRa, using Source DF Seq SS Adj SS Adj MS F P
adjusted SS for tests
Working gap 1 0.4502 0.4502 0.4502 5.57 0.026
Linear speed 2 0.7806 0.7806 0.3903 4.82 0.016
Finishing time 2 2.587 2.587 1.2935 15.99 0.000
Mesh size 1 4.0656 4.0656 4.0656 50.26 0.000
Working gap×mesh size 1 0.4143 0.4143 0.4343 5.12 0.032
Error 28 2.2652 2.2652 0.0809
Total 35 10.5629
R2 078.56 % Adj R2 073.19 %
 Int J Adv Manuf Technol

insignificant parameters which their p value is greater than

0.05, are eliminated from the ANOVA table.
The modified ANOVA table is shown in Table 5, which
indicates that all of the effects have a p value less than 0.05.
It means that these effects are significant for confidence
level of 95 %. The R square (R2) value indicates that the
predictors explain 78.56 % of the variance in ΔRa. The
adjusted R2 is 73.19 % which accounts for the number of
predictors in the model. Both the values indicate that the
model fits the data well.
Figure 7 depicts the plots which can be used to graph-
ically assess the effects of factors on ΔRa. It is indicated
that, working gap, abrasive mesh size, rotational speed, and
Fig. 7 Main effects plot for ΔRa finishing time have significant effect on ΔRa. As can be
seen from Fig. 7, the working gap and the abrasive mesh
size are reciprocally proportional to ΔRa. Furthermore,
the normal plot of residuals. This is clear that the points increasing rotational speed and finishing time result in
generally form a straight line, though the residuals are higher ΔRa; however, in high values of these two factors,
normally distributed and the error normality assumption is ΔRa increases slowly. The interaction plot for the ΔRa can
valid. The pattern of residuals, the residuals versus fitted be observed in Fig. 8. It is clear from this plot that the
values plot, is shown in Fig. 6b. So the other two assump- strongest interaction is between the working gap and the
tions are shown to be valid by means of the structureless mesh size.
distribution of dots above and below the zero line which
illustrates both the error independently and variance constancy 4.1.2 Analysis of recast layer changes
[25].
Now, since the assumptions are proved not to be violated, Figure 9a shows the normal plot of residuals. The distribution
the ANOVA results are reliable. In the present study, the shown in Fig. 9a indicates that the error normality assumption
confidence level is assumed to be 95 %. It can be seen from is valid since the points generally form a straight line. The plot
Table 4 that the finishing time, linear speed, mesh size, and of residuals versus fitted values (Fig. 9b) shows the structure-
working gap (p<0.05) have the most significant effect on less distribution of dots above and below the zero line. So this
ΔRa. In order to have a more accurate analysis, other plot illustrated both the error independency and variance

Fig. 8 Interaction plot for ΔRa

Interaction Plot for Ra(µm)
Data Means
5 12 19 1 2 60 120
2.0
linear
speed(m/min)
1.5
0.2
linear speed(m/min)
1.0 0.4
0.6
2.0
finishing
linear
1.5
time(min)
speed(m/min)
50.2
finishing time(min)
1.0 120.4
190.6
2.0
working
finishing
linear
1.5
gap(mm)
time(min)
speed(m/min)
working gap(mm) 150.2
1.0 12
2 0.4
190.6

mesh size
Int J Adv Manuf Technol

Fig. 9 a Normal probability plot and b residuals versus fitted values plot for the recast layer changes

Table 6 Analysis of variance

for the recast layer changes, Source DF Seq SS Adj SS Adj MS F P
using adjusted SS for tests
Working gap 1 72.049 72.049 72.049 13.57 0.002
Linear speed 2 216.949 216.949 108.474 20.43 0.000
Finishing time 2 172.734 172.734 86.367 16.27 0.000
Mesh size 1 189.847 189.847 189.847 35.76 0.000
Working gap×linear speed 2 7.496 7.496 30748 0.71 0.508
Working gap×finishing time 2 0.552 0.552 0.276 0.05 0.949
Working gap×mesh size 1 29.977 29.977 29.977 5.65 0.3
Linear speed×finishing time 4 3.653 3.653 0.913 0.17 0.95
Linear speed×mesh size 2 12.285 12.285 6.142 1.16 0.339
Finishing time×mesh size 2 0.293 0.293 0.146 0.03 0.973
Error 16 84.943 84.943 5.309
Total 35 790.778
R2 089.26 % Adj R2 076.50 %

Table 7 Modified analysis of

variance for the recast layer Source DF Seq SS Adj SS Adj MS F P
changes, using adjusted
SS for tests Working gap 1 72.049 72.049 72.049 18.47 0.000
Linear speed 2 216.949 216.949 108.474 27.81 0.000
Finishing time 2 172.734 172.734 86.367 22.14 0.000
Mesh size 1 189.847 189.847 189.847 48.67 0.000
Working gap×mesh size 1 29.977 29.977 29.977 7.69 0.01
Error 28 109.221 109.221 3.901
Total 35 790.788
R2 086.16 % Adj R2 082.74 %
 Int J Adv Manuf Technol

explain 86.16 % of the variance in the removed recast layer.

The adjusted R2 is 82.47 % which accounts for the number
of predictors in the model. Both the values indicate that the
model fits the data well.
Figure 10 shows the plot of main effects for the factors
which have been considered in this research. This graph
depicts the effects of factors on recast layer thickness. As
can be seen, working gap and abrasive particle size recipro-
cally affect the output. In addition, the finishing time and
rotational speed have direct proportion to the recast layer
thickness increase; that is, by increasing these factors, the
removed recast layer increase significantly. The interaction
plot for the changes in the recast layer can be observed
Fig. 10 Main effects plot for the recast layer changes in Fig. 11. It is clear from this plot that the interaction
between working gap and mesh size are the strongest
ones.
constancy [25]. Now, it is illustrated that the ANOVA results
are reliable since the assumptions are proved not to be violat- 4.2 Influence of parameters on surface improvement
ed. Table 6 shows that the interaction effects have no signif-
icant influence in ANOVA model. Because their p value is 4.2.1 Influence of linear speed
higher than 0.05, accepting that there is statistical evidence of
curvature in this model, for a confidence level of 95 %. It can The relation between the linear speed of the workpiece and
be seen from Table 6 that working gap, abrasive mesh size, the changes in the surface roughness and recast layer thick-
rotational speed, and finishing time (p<0.05) have significant ness is shown in Figs. 7 and 10. It shows that as the linear
effect on the recast layer thickness. Thus, other insignificant speed increases, the surface roughness and recast layer
parameters which their p value is greater than 0.05, are elim- thickness decrease. At higher linear speed, the magnetic
inated from the ANOVA table for a more accurate analysis. force cannot make the abrasive particles stay on the surface
The modified ANOVA table is shown in Table 7, which and this can be partially attributed to splashing of the few
indicates that all of the effects have a p value less than 0.05. abrasives. As a result, it is seen that the effect of linear speed
It means that these effects are significant for confidence on recast layer thickness and surface roughness in higher
level of 95 %. The R2 value indicates that the predictors values is less.

Fig. 11 Interaction plot for the

recast layer changes
Interaction Plot for removed recast layer(µm)
Data Means
5 12 19 1 2 60 120
15
linear
10 speed(m/min)
0.2
linear speed(m/min)
5 0.4
0.6
15
finishing
linear
10 time(min)
speed(m/min)
finishing time(min) 50.2
5 120.4
190.6
15
working
finishing
linear
10 gap(mm)
time(min)
speed(m/min)
working gap(mm) 150.2
5 2 0.4
12
190.6

mesh size
Int J Adv Manuf Technol

Table 8 Regression model for ΔRa Table 10 ANOVA for the regression model of ΔRa

Predictor Coefficient SE Coefficient T P Source DF SS MS F

Constant 1.9708 0.2781 7.25 0.000 Regression 4 639.26 159.82 32.71 0

Linear speed 0.8973 0.3069 2.92 0.006 Residual error 31 151.49 4.89 –
Finishing time 0.045836 0.008769 5.23 0.000 Total 35 790.78 –
Working gap −0.2237 0.1003 −2.23 0.033
Mesh size −0.0112 −6.71 −6.71 0.003
R2 073.5 % Adj R2 070.0 %
size. So in this case, larger magnetic particles act with greater
force on the surface. However, the effective cutting edges of
the larger particles are fewer than it is for small ones, so it
4.2.2 Influence of working gap decrease the removal rate of the recast layer. Although, ex-
tremely enlarging the abrasive size, increase the size of micro-
According to Figs. 7 and 10, by increasing the working gap, chips which are eroded from the surface and it deteriorates the
the changes in recast layer thickness and ΔRa decrease. surface roughness. Consequently, increasing the abrasive size
This is due to the fact that the magnetic flux between the have its own special limits.
permanent poles is constant, so the larger gap decreases the
flux density and finishing pressure. As a result, the removed
material decreases and this causes the adjunction of the 5 Regression analysis
remained recast layer and reduction of surface roughness
improvement. As a result of the regression analysis, Eq. 1 presents the
linear relationship between factors and ΔRa (response).
4.2.3 Influence of finishing time ΔRa ¼ 1:97 þ 0:897 linear speed

Figures 7 and 10 show the effect of finishing time on recast þ 0:0458 finishing time  0:224 working gap
layer thickness and surface roughness. By increasing the
 0:0112 mesh size ð1Þ
finishing time, the number of abrasive particle cycles in
the cylindrical specimens increases. Thus, the surface And also Eq. 2 shows the linear relationship between factors
microchipping increases and it improves the surface rough- and the removed recast layer (response).
ness by removing the recast layer. However, increscent
chips which are removed from the surface, mix with abra- Removed recast layer ¼ 10:2 þ 14:7 linear speed
sive particles, and reduce their efficiency. So passing time
would not have its previous effect on surface improvement. þ 0:380 finishing time
 2:83 working gap
4.2.4 Influence of abrasive particle size
 0:0765 mesh size ð2Þ
The recast layer thickness decreases and the surface roughness
improves when the mesh size decreases (abrasive particle size In these equations, the unit of ΔRa and the removed recast
increases). This is because of the fact that the magnetic force layer will be micron if the linear speed is in meter per
which acts on abrasive particle is proportional to the particle second, finishing time in minute and working gap in
millimeters.
The above mathematical models for surface roughness
Table 9 Regression model for the recast layer changes and recast layer thickness are of great importance to the

Predictor Coefficient SE Coefficient T P

Constant 10.227 1.998 5.12 0.000 Table 11 ANOVA for the regression model of the recast layer changes
Linear speed 14.719 2.256 6.52 0.000
Source DF SS MS F P
Finishing time 0.37955 0.06446 5.89 0.000
Working gap −2.8294 0.7369 −3.84 0.001 Regression 4 7.7598 1.94 21.45 0
Mesh size −0.07655 0.01228 −6.23 0.000 Residual error 31 2.803 0.0904 –
R2 080.8 % Adj R2 078.4 % Total 35 10.8629 –
 Int J Adv Manuf Technol

Table 12 Results of confirma-

tion tests for ΔRa Run Working Linear speed Finishing Mesh Results Results of Regression
gap (μm) (m/min) time (min) size of model experiments error (%)

1 2 0.4 12 60 1.768 1.76 −0.45

2 1 0.6 5 60 1.841 1.997 7.81
3 2 0.2 19 120 1.228 1.277 3.84

proper selection of finishing parameters during the magnetic the WEDM recast layer removal process and the conclu-
abrasive finishing of the cylindrical tubes. sions are as follows:
Tables 8 and 9 show the coefficient of factors and the
effect of factors in regression model for the outputs, ΔRa, 1. The MAF process can remove the recast layer which is
and removed recast layer, respectively. The ANOVA tables produced by WEDM on the inner surface of aluminum
for regression model with linear term are shown in Tables 10 alloy specimen, thus a finished surface without any
and 11 for the mentioned responses. As can be seen, the microcracks can be obtained by using this technique.
term of regression model are significant at the confidence 2. It is illustrated that the linear speed and mesh size are
level of 95 %. the highly effective parameters on surface roughness
The verification of the test results are shown in Tables 12 and recast layer thickness. It is observed that by increas-
and 13 for the surface roughness improvements and the ing the linear speed and abrasive particle size, the recast
recast layer thickness decreases, respectively. The predicted layer removal and surface roughness improvement
finishing parameters performance is compared with the ac- increase.
tual finishing performance and a good agreement was 3. The more time spend on finishing the specimens, the
obtained between these performances. more improvement is achieved in the surface roughness
since the removed recast layer has increased. However,
by passing time, the abrasive particle mixed with the
6 Microhardness analysis eroded material which weakens the abrasive influence.
The other effective parameter on the recast layer thick-
Figure 12 shows the changes that occurs in the surface ness and surface roughness is working gap value. The
hardness of specimens after the WEDM and although after removed recast layer and ΔRa are apparent to have an
the MAF process. It is indicated that the recast layer which increasing trend with decrease in working gap. Because
is the result of resolidification of molten material, has great- at distances, closer to the magnets, the significant mag-
er hardness than the matrix. This is because of the thermal netic flux density and so the magnetic force acting on
characteristic of the WEDM process. After removing the the abrasive particles increase.
recast layer completely, through MAF process, the surface 4. The estimated models for the surface roughness and
hardness decreases. However, it is still greater than the base recast layer thickness is indicated to be appropriate,
material, which indicates the presence of heat-affected zone using ANOVA, regression analysis and confirmation
although with a smaller thickness. tests. The mathematical models which show the relation
between the inputs and the responses can be used to
achieve the predetermined results.
7 Conclusions 5. The optimum parameter levels that give the best surface
finish improvement were determined. The ideal param-
This study employed the MAF mechanism and the full eter settings to achieve optimum output could be linear
factorial experimental design to explore the feasibility of speed, 0.6 m/s; finishing time, 19 min; mesh size, 60;

Table 13 Results of confirma-

tion tests for the recast layer Run Working Linear speed Finishing Mesh Results Results of Regression
changes gap (μm) (m/min) time (min) size of model experiments error (%)

1 2 0.4 12 60 10.39 11.09 6.31

2 1 0.6 5 60 13.5 14.31 5.66
3 2 0.2 19 120 5.52 5.37 −2.79
Int J Adv Manuf Technol

6. Lim LC, Lee LC, Wong YS, Lu HH (1991) Solidification micro-

structure of electro-discharge machined surfaces of tool steels.
Mater Sci Technol 3(7):239–248
7. Lee LC, Lim LC, Narayanan V, Venkatesh VC (1988) Quantifica-
tion of surface damage of tool steels after EDM. Int J Mach Tool
Manuf 28(4):359–372
8. Liao YS, Huang JT, Chen YH (2004) A study to achieve a fine
surface finish in wire-EDM. J Mater Process Technol 149:165–171
9. Jain VK (2004) Advanced machining processes. Allied, New
Delhi
10. Singh DK, Jain VK, Raghuram V (2006) Experimental investiga-
tions into forces acting during a magnetic abrasive finishing pro-
cess. Int J Adv Manuf Technol 30:652–662
11. Girma B, Joshi SS, Raghuram MVGS, Balasubramaniam R (2006)
Fig. 12 Vickers hardness of the surface at different steps of machining An experimental analysis of magnetic abrasives finishing of plane
and finishing surfaces. Mach Sci Technol 10(3):323–340
12. Yan BH, Chang GW, Cheng TJ, Hsu RT (2003) Electrolytic
magnetic abrasive finishing. Int J Mach Tool Manuf 43:1355–1366
working gap, 1 mm for the material; and process con- 13. El-Taweel TA (2008) Modelling and analysis of hybrid electro-
ditions applied in this study. The experimentation with chemical turning magnetic abrasive finishing of 6061 Al/Al2O3
composite. Int J Adv Manuf Technol 37:705–714
these parameter settings results in 18.04 μm as removed
14. Kim J, Choi M (1997) Development of the magnetoelectrolytic
recast layer and 2.39 μm as ΔRa (reduction in surface abrasive polishing system (MEAPS) and finishing characteristics
roughness from its initial value). Considering the above of a Cr-coated roller. Int J Mach Tool Manuf 37:997–1006
mentioned values as the input of Eqs. 1 and 2, the 15. Shinmura T, Takazawa K, Hatano E (1990) Study on magnetic
abrasive finishing. Ann ICRP 39:325–328
optimum ΔRa can be calculated as 2.482 μm and the
16. Lin CT, Yang LD, Chow HM (2007) Study of magnetic abrasive
optimum removed recast layer can be calculated as finishing in free-form surface operations using the Taguchi meth-
18.82 μm that shows around −3.8 and −4.3 % error, od. Int J Adv Manuf Technol 34:122–130
respectively, in predicted output in regression analysis. 17. Kim J-D, Noh I-H (2007) Magnetic polishing of three dimensional
die and mold surfaces. Int J Adv Manuf Technol 33(1–2):18–23
6. The microhardness tests show the differences between the
18. Wang CC, Chow HM, Yang LD, Lu CD (2009) Recast layer
surface hardness before and after the machining and the removal after electrical discharge machining via Taguchi analysis:
finishing. It is indicated that the MAF process can de- a feasibility study. J Mater Process Technol 209:4134–4140
crease the surface hardness by removing the recast layer. 19. Schumacher BM (1990) About the role of debris in the gap during
electrical discharge machining. Ann ICRP 39(1):197–199
20. Yamaguchi H, Shinmura T, Kaneko T (1996) Development of a
new internal finishing process applying magnetic abrasive finish-
References ing by use of pole rotation system. Int J Japan Soc Prec Eng 30
(4):317–322
21. Givi M, Fadaei Tehrani A, Mohammadi A (2011) Polishing of the
1. Ho KH, Newman ST, Rahimifard S, Allen RD (2004) State of the aluminum sheets with magnetic abrasive finishing method. Int J
art in wire electrical discharge machining (WEDM). Int J Mach Adv Manuf Technol. doi:10.1007/s00170-011-3753-0
Tools Manuf 44:1247–1259 22. Yamaguchi H, Shinmura T (2000) Study of an internal magnetic
2. Shobert EI (1983) What happens in EDM. In: Jameson EC (ed) abrasive finishing using a pole rotation system. Prec Eng 24:237–
Electrical discharge machining: tooling, methods and applications. 244
SME, Dearbern, pp 3–4 23. Altenpohl DG (1998) Aluminum, technology, applications, and
3. Tsai HC, Yan BH, Huang FY (2003) EDM performance of Cr/ Cu- environment. A profile of a modern metal, 6th edn. Warrendale,
based composite electrodes. Int J Mach Tools Manuf 43(3):245–252 TMS, pp 360–364
4. Boothroyd G, Winston AK (1989) Non-conventional machining 24. Troeger LP, Starke EA (2000) Microstructural and mechanical
processes. Fundamentals of machining. Marcel Dekker, New York, characterization of a superplastic 6xxx aluminum alloy. Mater
p 491 Sci Eng 277(1–2):102–113
5. McGeough JA (1988) Electrodischarge machining. Advanced 25. Montgomery DC (2000) Design and analysis of experiments.
methods of machining. Chapman & Hall, London, p 130 Wiley, New York