Machine learning-based predictive model for wire electric discharge machining of Incoloy 800H

Abstract

In an era of continuous technological advancement, the wire electrical discharge machining (WEDM) process has evolved to address the ever-increasing demand for precision and efficiency in manufacturing. Among difficult-to-cut materials, Incoloy 800H is widely used in aerospace and nuclear applications. The current work incorporates machine learning (ML) algorithms, that is, decision tree, support vector machines, logistic regression, random forest, artificial neural network, linear regression, XGBRegressor, and K-nearest neighbors, to model and optimize the process during WEDM of Incoloy 800H. The novelty lies in the comprehensive investigation of the effects of varying process parameters, namely, pulse off time (Toff), pulse on time (Ton), servo voltage (Sv), and wire tension (Wt) on output parameters, including wire breakage and surface roughness (SR). The predicted outcomes by the XGBRegressor model exhibit strong agreement with the experimental results, demonstrating a high correlation coefficient (R) of 0.945. The XGBRegressor proved to be the most effective model for predicting SR, achieving the lowest mean absolute error and mean squared error among all ML algorithms evaluated. In conclusion, the findings underscore the potential of integrating ML with statistical techniques to enhance predictive capabilities and process optimization in advanced manufacturing domains.