Wire electrical discharge machining of AZ91 magnesium alloy; investigation of effect of process input parameters on performance characteristics

In this investigation, AZ91 magnesium alloy was used as experimental material [18]. This alloy is produced by die casting. Die casting process is one of the most effective methods for producing magnesium parts especially in the automotive industry. The chemical composition, physical and mechanical properties of AZ91 are presented in tables 1 and 2, respectively. Prior to conduct the WEDM experiments, the work piece dimension reached the desired dimension (140 * 60 * 14 mm) by the grinding process.

The machining process was performed on the 310 Robofil Charmilles Technologies WEDM machine. This machine uses isopulse type generator. In this experimental investigation, three parameters of the WEDM process including pulse on time (T_ON), pulse current (IP) and servo feed rate (SF) were considered as variable input parameters. The levels and values of these variable parameters are shown in table 3. The constant parameters of machining are presented in table 4. Selected parameters values were chosen from prior experiences and based on device capacity. Zinc coated brass wire with 0.25 mm diameter was used as electrode tool while de-ionized water used as a dielectric fluid. The EDX analysis of the wire electrode is shown in figure 2. During cutting of each sample, the cutting speed (V) was obtained from machine monitor and recorded after that machining conditions reached steady state. The cutting length was 40 mm in all experiments.

Zoom In Zoom Out Reset image size

In order to reduce the number of experiments needed to examine the effect of input parameters on output parameters and analysis of results, the design of experiments (DOE) technique and statistical analyses were used. To do that, the response surface methodology (RSM) with the Box-Behnken design was employed in this study. In the RSM, statistical and mathematical techniques are used for experimental modeling and expression of the correlation between an output variable (response variable) with one or more input variables (independent variables). Also, by using ANOVA, the contribution content of each input parameter on the output parameter can be found [19]. Design of experiments and analyzing of the results were done with the help of Design Expert 10 software. Overall 15 experiments with 3 factors in 3 levels were conducted. The DOE run order with experimental results of 3 response variables including MRR, KRFW, and SR is shown in table 5.

The kerf width was measured by an optical microscope with a resolution of 0.001 mm. Kerf was measured at three different places and their average was taken as a final response. The surface roughness was measured using 'Perthometer-M2, Mahr, Germany' device. Average roughness (Ra) which is common in manufacturing and production was used to the expression of surface roughness. Surface roughness was measured at 5 different positions and their average value was reported as Sr 3D images obtained by the Atomic Force Microscope (AFM) were used to examine the topography of machined surfaces. To do that 'Nanosurf Mobile S' device was used. Scanning electron microscope (SEM) images and EDX analysis were used to investigate the microstructure and metallurgical changes at the WED machined surface. To recast layer observation, the cross-section of WED machined surface was polished with 336 to 1500 SiC papers. SEM images were achieved by 'MIRA3- FEG, Tescan, Czech Republic'.

The MRR in the electrical discharge machining process indicates the process productivity. The main limitation of this process, which in some cases prevents competition with other machining methods, such as broaching, is the low rate of cutting. Equation (1) was used to obtain MRR.

$$\begin{eqnarray}&&{\rm{MRR}}\left({{\rm{mm}}}^{3}/\,{\rm{\min }}\right)={\rm{Kerf}}\,{\rm{Width}}* {\rm{Height}}\,{\rm{of}}\,{\rm{Cut}}* {\rm{Cutting}}\,{\rm{Speed}}\end{eqnarray} \tag{ 1 }$$

Figure 3 shows the response surface plots for the MRR. It is obvious in figure 3(a) that the MRR increases with the increase in pulse on time and pulse current. In this process, the amount of material removal per spark depends on the thermal energy of spark. As the energy of the sparks increases, the amount of applied heat to the work piece increases and more volume of the material melts and evaporates which resulting in high MRR [10]. The energy of the sparks is calculated from equation (2).

$$\begin{eqnarray}&&E=\displaystyle {\int }_{0}^{Ton}I.V.dt\end{eqnarray} \tag{ 2 }$$

Where, T_on = pulse duration (s),I = pulse current (A),V = machining voltage (V), andE = spark energy (J)

Zoom In Zoom Out Reset image size

According to equation (2), with the increase in pulse current and pulse on time the thermal energy of sparks and the intensity of heat will be increased and consequently more amount of workpiece material will be melted from the workpiece. Figures 2(b) and (c) show that the servo feed rate does not have an effect on the MRR since the applied thermal energy to the workpiece have no relation with the servo feed rate. Table 6 presents the analysis of variance (ANOVA) table for MRR. According to the presented data, the effect of factors A, B, AB, and A² on the MRR is significant because the P-value is less than 0.05 and the pulse on time has the greatest effect on the material removal rate due to 84.14% contribution in MRR. This phenomenon can be explained by the fact that because of high thermal conductivity of magnesium alloys heat is distributed rapidly into work material and the increase in sparking time leads to more distribution of heat which in turn causes to more volume of work affected by heat. According to results, during WEDM of magnesium, the MRR (thickness = 14 mm) reaches to 180 mm³ min⁻¹, while according to Soundararajan et al [8] maximum material removal rate in A413 aluminum alloy (thickness = 15 mm) was 36.241 mm³ min⁻¹ and almost 6 mm³ min⁻¹ in IN706 superalloy (thickness = 10 mm) based on Sharma et al [11]. The main reasons for this difference could be due to low melting point and a high thermal conductivity of magnesium alloys. According to these results, the use of wire-cut in mass production of magnesium parts can be promising due to the high cutting rate. Equation (3) expresses the predicted mathematical model for MRR.

$$\begin{eqnarray}&&\begin{array}{l}{\rm{MRR}}=+109.57313-113.06133* {{\rm{T}}}_{{\rm{On}}}-2.06553* {\rm{IP}}-8.62814* {\rm{SF}}\\ \,\,\,\,+7.53282* {{\rm{T}}}_{{\rm{On}}}* {\rm{IP}}+1.82065* {{\rm{T}}}_{{\rm{On}}}* {\rm{SF}}-0.098219* {\rm{IP}}* {\rm{SF}}\\ \,\,\,\,+101.06743* {\rm{T}}\_{{\rm{On}}}^{2}+0.010129* {{\rm{IP}}}^{2}+0.49123* {{\rm{SF}}}^{2}\end{array}\end{eqnarray} \tag{ 3 }$$

The width of the formed kerf in the WEDM process through the wire transition indicates the dimensional accuracy of the process [20]. In WEDM, in order to avoid short circuit and also dielectric fluid injection into the cutting area, there is a need for a gap between the wire and the work piece. Due to the specific distance between the wire and work piece sides, the width of the slit that will be created after the wire transition will be larger than the wire diameter. The kerf width is calculated from equation (3).

$$\begin{eqnarray}&&{\rm{Kerf}}\,{\rm{Width}}\left({\rm{mm}}\right)={\rm{Wire}}\,{\rm{Diameter}}+2* {\rm{Gap}}\,{\rm{Distance}}\end{eqnarray} \tag{ 4 }$$

Figure 4 shows the kerfs created in 6 samples of the experiments. It is clear from these images that the maximum kerf width is related to figure 4((b); ninth experiment) and the lowest kerf width is related to figure 4((f); thirteenth experiment). In these samples, the maximum and minimum thermal energy have been applied to the work piece, respectively. By comparing figure 4((a); tenth experiment) and figure 3((b); ninth experiment), it is obvious that by increasing the pulse on time, the kerf width increases. Also, the comparison of figure 4((c); third experiment) and figure 4((d); fifteenth experiment) shows that increase in pulse current leads to the kerf width increasing whereas figure 4((d); fifteenth experiment) and figure 4((e); seventh experiment) comparison reveals that although there is no significant difference in kerf widths, it could be resulted that increasing in servo feed rate caused kerf width to decrease. These results also can be obtained from figure 6 which indicates the RSM plots of kerf width. From table 7 it is found that pulse on time with 95% contribution in KERW response has the most effect on kerf width. According to ANOVA table for kerf width in addition to factor A factors B and A² also have a significant effect on kerf width (P-value less than 0.05). Increase in pulse on time and pulse current lead to increase in the depth of heat penetration in to work material and in turn it causes to more material removal from wire sides and large kerf. Similar to material removal rate, the kerf width in magnesium alloy is much higher than the other materials, which results from a low melting point and high thermal conductivity. These results are in perfect agreement with the results presented for WE43 Mg WEDM [7]. The developed mathematical model for kerf width is presented in equation (4).

$$\begin{eqnarray}&&\begin{array}{l}{\rm{KRFW}}=+0.29576+0.14987* {{\rm{T}}}_{{\rm{ON}}}-0.000385417* {\rm{IP}}+0.00304167* {\rm{SF}}\\ \,\,\,\,+0.000625* {{\rm{T}}}_{{\rm{ON}}}* {\rm{IP}}+0.000833333* {{\rm{T}}}_{{\rm{on}}}* {\rm{SF}}-0.000166667* {\rm{IP}}* {\rm{SF}}\\ \,\,\,\,-0.046333* {{{\rm{T}}}_{{\rm{ON}}}}^{2}+0.000197917* {{\rm{IP}}}^{2}-0.000148148* {{\rm{SF}}}^{2}\end{array}\end{eqnarray} \tag{ 5 }$$

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Some of the mechanical and physical properties of machined parts, such as corrosion resistance, friction, fatigue strength and loading capacity depend on the roughness of the machined surface. Therefore, this parameter plays a crucial role in the function of parts. Due to that, one of the most important parameters to determine the machinability of materials is surface roughness. The surfaces machined by the WEDM process are full of micro cavities, and craters. These craters are formed when the sparks hit the work piece. Figures 5(a) to (c) shows the response surface plots of surface roughness. It is found from response plots that SR increases with increase in pulse on time and pulse current whereas slightly decreases with increase in servo feed rate. These results can be explained by the fact that with an increase in pulse current and pulse on time, the intensity of heat and the duration of this amount of applied heat into work piece increases that in turn leads to larger and deeper craters formation on the surface and high surface roughness. ANOVA table for SR (table 8) indicates that pulse on time and pulse current have 86.44 and 3.25% contribution in SR respectively, whereas servo feed rate does not notably affect the surface roughness (0.19% contribution). It seems that increase in servo feed rate leads to decrease in the number of impacted sparks per unit area of the surface and in turn causes to decrease in Sr figures 6(a) and (b) show the surface topography images of WED machined surface obtained by AFM at highest and lowest spark energy, respectively. Generated craters are obvious in these images. It is clear from these images that the surface which machined by high thermal energy (figure 7(a)) has rough surface condition while the specimen which machined by low energy has a fine surface. Predicted mathematical model for SR is presented in equation (5).

$$\begin{eqnarray}&&\begin{array}{l}{\rm{SR}}=+1.34457+4.29277* T\_\mathrm{On}+0.069465* {\rm{IP}}+0.050775* {\rm{SF}}\\ \,\,\,-0.016125* T\_\mathrm{On}* {\rm{IP}}+0.013833* T\_\mathrm{On}* {\rm{SF}}+8.33333{\rm{E}}-005* {\rm{IP}}* {\rm{SF}}\\ \,\,\,-1.78733* {T\_\mathrm{On}}^{2}-9.73958{\rm{E}}-004* {{\rm{IP}}}^{2}-4.17593{\rm{E}}-003* {{\rm{SF}}}^{2}\end{array}\end{eqnarray} \tag{ 6 }$$

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The machined surfaces by WEDM are filled with micro cracks, carters, cavities, voids, splashes, melted debris. By applying the enough voltage between the two electrodes (wire and work piece), after dielectric fluid resistance breakage, due to the ionization of the medium molecules, plasma channel is created and the temperature of this channel reaches to 8000 °C–12000 °C, which is enough for melting and evaporation of every material [13]. After the end of pulse on time period and plasma channel collapse, due to the high exploding pressure, a few molten materials are thrown out of the melt hole and flushed away by a dielectric fluid from the machining zone and the remaining melted materials that have not been removed by dielectric are rapidly cooled due to water coolant effect and result in debris, splashes and micro voids formation [21]. Figure 8 shows the SEM images of machined surfaces. It is clear from this images that ninth experiment (figure 8(b)) which has been machined with the highest level of pulse on time and pulse current has the roughest surface condition and the number of micro cavities and amount of debris and splashes is more than the other specimens. Surfaces which machined by the lowest level of pulse on time (figures 8(a) and (f)) have the better and uniform surface condition. By reducing the pulse on time, the duration of applying heat into workpiece reduces and a small amount of material is melted and dielectric fluid flushes the molten material away from machining zone better, so it is expected that the surfaces become more uniform and fine. Comparison of figures 8(c) and (d) results in that increase in pulse current leads to a non-uniform surface with more micro cavities and holes. Increase in pulse current results in an increase in heat intensity which in turn causes to more material fusion however because of constant dielectric liquid pressure and injection rate during all machining condition, dielectric liquid efficiency decreases with increase in more material fusion. Decreasing in dielectric liquid efficiency leads to more debris and splashes generation in the surface. Based on figures 7(c) and (e), there are no significant differences in microstructures which indicate that micro structure condition does not change notably by servo feed rate.

Zoom In Zoom Out Reset image size

EDX analysis (figure 9) shows the elemental analysis of WED machined surface. As can be seen, in addition to the base metal components, oxygen and carbon are observed and content of Zn and Cu have been increased whereas the content of magnesium and aluminum has been decreased. The presence of oxygen is due to the decomposition of water molecules and the increase in the percentage of copper and zinc elements are due to the transformation of wire materials to the machined surface. In this process, both the cathode and the anode are melted, and when the machining zone is washed, some of the zinc coated brass wire materials are solidified together with the work piece materials at the sample surface and resulted in Cu and Zn reaching. In WED machined surfaces thermal induced micro cracks are produced because of high-temperature sparks. It is worth to say that as seen in microstructure images the number of these micro cracks even in high thermal sparks energy is low. The main reason for this phenomena could be high thermal conductivity of magnesium alloy that prevents thermal energy concentration on the surface.

Zoom In Zoom Out Reset image size

The main feature of WED machined parts is the presence of a thin white layer on the surface of the parts. After the electrical discharge, it is not possible to remove all molten material from the melt hole, and the remaining materials are re-solidified by a dielectric fluid and lead to the formation of a recast layer on the top surface of the work piece [22]. The thickness of this layer is non-uniform throughout the surface. Figure 10 shows the depth of recast layer in 2 samples (ninth and thirteenth experiments) which machined with the highest and lowest input energy, respectively. As can be seen in figures 9(a) and (b), by increasing pulse on time and pulse current the recast layer thickness increases. When the pulse duration and the current are at the highest level, the amount of molten material increase and by the fact that pulse interval time and dielectric liquid pressure are constant in all machining conditions, in turn, more molten material results in the more re-solidified material. Heat affected zone (HAZ) is at the bottom of recast layer. Due to the heating and rapid cooling, the characteristics of the recast layer and the HAZ are different from the base metal, and these areas are dangerous in terms of mechanical function because the micro structure has been changed, thermal residual stresses have been generated and impurity has been penetrated [23]. In sensitive applications such as the aerospace industry, this layer is harmful and should be removed as much as possible. To remove this layer, it is needed to trim-cut with low spark energy after rough-cut in WEDM or utilizing other non-tradition machining techniques such as chemical machining (ChM), electrochemical machining (ECM) or abrasive flow machining (AFM).

Zoom In Zoom Out Reset image size