Artificial Intelligence and Machine Learning-Based Modeling for Sustainable Wire Electrical Discharge Machining of Titanium Grade 4 Alloy

Abstract

This study tests the effect of Wire Electrical Discharge Machining (WEDM), on the hardness and surface finish characteristics of the Titanium Grade 4 alloy. Specifically, the study aims to optimize machining parameters such that material removal rate (MRR) will be maximized and energy efficiency maximized. To such ends, three soft computing methods, including Least Squares Support Vector Machine, Multi-gene Genetic Programming, and Fuzzy Logic, were employed to obtain the most accurate predictions of MRR and energy consumption (PC). The results showed that all three models could perform equally well on the training and testing datasets, as proven by their MRR and PC scores. Accuracy for the different models was evaluated using the coefficient of determination (R²), RMSE, and MAE. It was concluded that all three models had a high coefficient of determination (>95%) while showing lower error rates across the three responses. For model 1, the LS-SVM model was superior in all these performance metrics to the other two models, with RMSE of 0.080606, MAE of 0.065454, and MSE of 0.0064973. In contrast, the MGGP model recorded RMSE of 0.22056, MAE of 0.18994, and MSE of 0.048647 for model 2. Such error metrics, therefore, indicate that LS-SVM produces the minimum error, highlighting superiority in terms of accuracy and reliability of modeling.