Performance evaluation of machine learning algorithms in predicting machining responses of superalloys

Abstract

This study explores the application of machine learning algorithms—gene expression programming (GEP), adaptive neuro-fuzzy inference system (ANFIS),andartificial neural networks (ANN)—topredictmachiningresponsesduringthemillingofInconel690,asuperalloyknown for its exceptional mechanical properties and oxidation resistance. Machining Inconel 690 presents significant challenges due to its toughness and work-hardening tendencies, which can lead to rapid tool wear and poor surface finish. Traditional optimization methods often rely on empirical models and trial-and-error approaches, which are time-consuming and costly. In contrast, machine learning techniques can effec tively model complex, nonlinear relationships between machining parameters and performance outcomes, such as surface roughness, cutting force, and cutting temperature. This study employs statistical metrics, including Root mean square error (RMSE), coefficient of determina tion (R2), and mean absolute percentage error (MAPE), to determine the predictive performance of the models. The results show that the GEP model achieved an R2 ranging from 0.944572 to 0.992999, with an RMSE between 0.015527% and 0.694523% and a MAPE ranging from 1.452397% to 4.947892%. ANFIS and ANN also demonstrated strong predictive capabilities, although GEP outperformed them. The importance of this study lies in its demonstration of advanced AI techniques as effective tools for optimizing machining processes, ultimately contributing to improved efficiency and quality in manufacturing superalloys.