A comparative analysis on metamodel-based predictive modeling of electrical discharge machining processes

To meet the ever-increasing requirements of higher dimensonal accuracy along with better surface quality in many of the modern-day manufacturing industries, electrical discharge machining (EDM) appears to be an efficient material removal process to generate intricate shape features on various hard-to-cut work materials. Due to complex material removal mechanism (thermal erosive action of high frequency electrical current), and involvement of varying input parameters and responses, development of an effective predictive model for an EDM process is really a challenging task. To resolve the issue, this paper thus proposes the applications of four metamodeling techniques, in the form of polynomial regression, kriging, radial basis function and gene expression programming (GEP) as effective prediction tools for two EDM processes. For the first EDM process, pulse current, pulse rate, duty cycle and voltage are treated as the input parameters, whereas, material removal rate (MRR) and average surface roughness are the responses. On the other hand, pulse-on time, peak current, gap voltage and flushing pressure are the input parameters, and MRR, tool wear rate, overcut and taper are the responses in the second EDM process. The prediction performance of all the considered metamodeling techniques is finally contrasted using various model accuracy metrics, like R-squared (R-2), adjusted R-squared (R-adj(2)), root mean square error (RMSE) and relative root mean square error (RRMSE). Based on the past experimental datasets, it is observed that GEP provides more accurate and stable overall prediction results for both the EDM processes under consideration, outperforming the other techniques with respect to maximum R-2 and R-adj(2), and minimum RMSE and RRMSE values. Thus, it can be applied as a potent prediction tool to any of the machining processes.