An experimental investigation and machine learning predictions to enhance microhardness in powder mixed electrical discharge machining

This study examines the powder mixed electrical discharge machining method, using electrolyte copper as a dielectric medium, to enhance the microhardness of the machined surfaces. The present dielectric medium in the EDM process, electrode made up of electrolytic copper (99.9%) was used to machine the workpiece. As an input parameter, pulse on time, gap current and pulse off time were used to improve the microhardness while reducing surface roughness. Using the Taguchi L₂₇ orthogonal array, the design of experiment and analysis of the experimental findings are performed. The metallic powders were chosen so that they would improve the properties of the die steels. Machine learning algorithms were used to train and test the experimental data obtained after machining of D2, H13 and OHNS die-steels. It also includes the study of feature importance and heatmaps which depict the significance of the chosen input parameters on the microhardness of these die-steels. R-squared value greater than 0.80 for machine learning algorithms were used to obtain scatter plots. In this study, multivariate linear regression was also used to obtain the equation including input parameters to predict the microhardness values and calculate percentage errors for predicted and actual microhardness values.