Introduction

Now a day, one of the major problems faced by the manufacturing industries is optimum selection of machining parameters to enhance the performance parameters like material removal rate and surface roughness. In conventional method, process parameters are selected based on trial and error method and by experienced operators. Wire EDM is the one of the efficient method to machine hard steels and ceramic components with precise components with complex geometry [1,2,3,4,5]. The authors explained about the design of experiments introduced by Taguchi, an orthogonal array of process parameters designed by covering all the levels of input parameters to conduct the experiment efficiently and analyses influence the of performance characteristics (MRR & SR) in terms of voltage, peak current, Interval time and pulse duration [6,7,8,9,10].

Panta Madhu, et al., worked on the aluminium metal matrix nanocomposites in wire EDM machining and optimized the process using VIKOR analysis [3]. The authors discussed the multi-objective optimization in wire EDM machining process and discussed about the influence of process parameters, while achieving more than one objective simultaneously [11,12,13,14,15]. In this research the variables such as spark variable gap size and non-uniformity of spark distribution near the wire are considered for analysis to increase the performance of machining in the wire EDM [16,17,18,19,20]. The hybrid method of combining ANN with genetic algorithm is used for the optimization in WEDM process [21,22,23,24,25].

The study on the WEDM process parameters for Aluminium 5454 alloy is carried out by changing the angles of work pieces 30°, 45° and 60° and experiments are performed to provide slots. The taper angle, pulse-off time and peak discharge current are considered as the controllable parameters to obtain the optimum process parameters. Suryapavan Cheruku, et al., developed an algorithm for wire EDM machining of Inconel 718 to reduce the surface roughness and used different techniques like ANN, SVM and GA to predict the surface roughness of Inconel 718 components [8]. Boopathi Sampath, et al., used helium assisted near dry wire cut electric discharge machining (NDWEDM) method molybdenum wire was used to cut M2-HSS material. The dielectric pressurized non-reacting helium gas mixed with minuscule water to form a fluid is used to overcome cooling and flush-out debris. Here, Taguchi method was used to have new experimental set up for WEDM to obtain the optimum parameters [26,27,28,29,30].

Akash Singh, et al., discussed about the manufacturing of aluminium based MMCs in stir casting process and looked over for EDM machining which is used to obtain deep holes and complex contours from the obtained composites. To examine, the obtained parameters grey regression analysis [GRA] was used and compare those with the RSM values. Finally, it was observed that GRA values are better than RSM values [10]. Ramakrishan. R, et al., elaborates the process of optimization in wire EDM process by design of experiments (Taguchi L9 orthogonal array and utilization of different techniques like ANN and MRSN. A good improvement was obtained [11]. For predicting mechanical properties, KNN and ANN algorithms are employed and contrasted. It is quite helpful to save a lot of time and efforts when using machine learning predictions [12].

Objectives

The main objectives of this work is to.

  • Minimize surface roughness, Minimize tool wear rate and maximize material removal rate by selecting the proper combination of pulse on time, input peak current and gap voltage.

The overall objectives of this work is to.

  • To improve the productivity by proper selection of process parameters by the optimization of process parameters.

  • To reduce experimentation time and cost.

  • To increase the performance of machining by finding influential parameters on performance characteristics.

Methodology

Figure 1 presents the flow chart on the method. The input variables were input peak current, gap voltage, pulse on time and flushing pressure and the ranges for input parameters were fixed based on limitations of machine manufactures and the input parameter levels are tabulated which is shown in Table 1. The experiments are conducted and the variation of input parameters on tool wear rate, MRR and average surface roughness were investigated. Minitab is used for the analysis and generation of surface plots are carried out to find the influencing parameter on output characteristics.

Fig. 1
figure 1

Flow chart on the method

Full size image
Table 1 Input variables and its levels
Full size table

After the machining process, the output parameters are obtained and shown in Table 2. The obtained valued are utilized to train the ANN model in the python in excel format. Then the accuracy of ANN model is measured and explained using the plot. Then the Minitab is used for Anova analysis and to generate the surface plots for the RSM method.

Table 2 Experimental values of inputs and outputs
Full size table

Results and Discussions

The surface plot graphs of the output variables for MRR, Tool wear rate, and average surface roughness versus various input parameters such as input peak current, gap voltage pulse on time and flushing pressure are plotted and shown in Figs. 2, 3, 4. By observing the surface plots above, it is observed that the pulse on time and peak current are directly proportional to material removal rate, as the pulse on time and peak current increases, there will be increase in MRR which leads to reduction in wire vibration and good accuracy surface finish. Graph refers, increase in removal rate is obtained due to increase in Ip and Ton [31,32,33,34,35]. It can be seen that at some point of gap voltage the MRR is increased then it is reduced, so there should be limit in maintaining gap voltage. In case of surface roughness (Ra), flushing pressure has more influence compared to other input parameters [36,37,38,39,40,41,42]. By referring graphs, it is also observed that surface roughness improved by variations in pressure. Over all, it is observed that the surface plots with various parameters of input and output determines the relation between them, from which it is able to predict the accurate input values to reduce wire vibration and the proper output [43,44,45,46,47,48,49,50].

Fig. 2
figure 2

MRR versus combination of input parameters

Full size image
Fig. 3
figure 3

Tool wear rate versus combination of input parameters

Full size image
Fig. 4
figure 4

Average surface roughness versus combination of input parameters

Full size image

The value from Table 2 is given to the artificial neural network modelling and testing using python 3.2. The predicted material removal rate is plotted against actual values MRR which is shown in Fig. 5 [51,52,53,54,55,56,57,58,59,60]. The R-Square value of above model is 84.7. From the graph, it is inferred that most of the predictions are falling under the feasible region of prediction which shows good accuracy of the algorithm [61,62,63,64,65,66,67,68,69,70].

Fig. 5
figure 5

MRR prediction using ANN versus actual value

Full size image

Conclusions

In this paper, process optimization on wire EDM machining of A365/WC metal matrix composites is carried out using RSM method and ANN. It is observed that Input current and Ton have more influence on Material Removal rate compared to other parameters. ANN model is developed to predict the MRR value of EDM machining process and accuracy of model is calculated to confirm the proper prediction. The predicted and actual values are compared to find suitable model for engineers for optimization in manufacturing. This work can be extended by applying the similar method to other machining process and also other materials. It also can be investigated using other data analysis techniques to increase the accuracy of predictions made by the techniques.