Surface roughness and surface crack length prediction using supervised machine learning–based approach of electrical discharge machining of deep cryogenically treated NiTi, NiCu, and BeCu alloys

This study aims to investigate the impact of various input variables in electrical discharge machining (EDM) on specific responses, including surface crack length (SCL) and surface roughness (SR). The variables under scrutiny are the electrical conductivity of the workpiece tool, pulse-on time, gap voltage, pulse-off time, and gap current. The study focuses on generating a mesoscale square blind hole in both cryo-treated and untreated workpieces using electrolytic oxygen-free copper. Experimental design and statistical software were employed to facilitate the analysis, following Taguchi’s L18 (61 × 34) orthogonal array. Through heat map, it was determined that pulse on time, pulse off time, and gap voltage significantly influence surface roughness. On the other hand, workpiece electrical conductivity, gap current, gap voltage, and pulse on time were found to impact surface crack length. It can be seen from the study that the formation of surface cracks exhibited a decreasing trend at the initial level of conductivity of the workpiece, while SCL increased as the WEC was raised. Additionally, lower values of gap current were associated with greater crack length, whereas increasing the gap current reduced crack length. Furthermore, an increase in gap voltage corresponded to an increase in crack length, whereas crack length decreased with an increase in pulse on time. Machine learning regression methods employed in the study could predict surface roughness and surface crack length values with R -squared values more than 0.90.