Wire-breakage prediction during WEDM of Ni-based superalloy using machine learning-based classifier approaches

Abstract

Wire-breakage (WB) during Wire electric discharge machine (WEDM) is a critical issue since it halts the machining process and leads to increased energy consumption, lower productivity, and diminished product quality. Current study applies machine learning-based classification approaches: logistic regression, random forest, decision tree, and Gaussian Naïve Bayes for prediction of wire-breakage instances during WEDM of nimonic 263 executed under varying cutting conditions of peak current (IP), spark-gap voltage (SV), pulse-on time (Ton) and pulse-off time (Toff). The decision tree and Gaussian Naïve Bayes' accuracy have equal accuracy (96%), whereas the logistic regression has a comparatively lesser accuracy (92%). Random forest has the highest accuracy of a hundred percent for the testing data. This research demonstrates that binomial classification algorithms may be employed offline to anticipate input parameters contributing to wire breaking. Avoiding the parametric combination leading to wire breakage during machining may enhance productivity and improve product quality. Graphical abstract: (Figure presented.) © The Author(s), under exclusive licence to Springer-Verlag France SAS, part of Springer Nature 2023.