Explainable machine learning for enhancing predictive accuracy of cutting forces in hard turning processes

Predicting cutting forces in hard turning optimizes toolpaths, improving machining efficiency and precision, minimizing tool wear, and enhancing manufacturing processes. This study assesses the potential of ensemble machine learning models for predicting machining force components during the hard turning of AISI 52100 bearing steel. Firstly, it conducts a comprehensive evaluation of Random Forest, Gradient Boosting, XGBoost, and CatBoost machine learning models using experimental data collected during AISI 52100 bearing steel turning with a CBN cutting tool. Cubic Spline-based data augmentation is employed to enrich the data, further enhancing prediction quality. A comparative analysis of the considered models on both original and augmented datasets reveals significant performance improvements associated with the utilization of augmented data. Moreover, this study utilizes SHAP (SHapley Additive exPlanations) to elucidate model predictions, providing insights into the contribution of each feature. Results indicate that ensemble learning methods, particularly CatBoost and XGBoost, demonstrate satisfactory predictive results with an averaged R² of 0.96 and 0.945, respectively. Ensemble models like CatBoost and XGBoost, coupled with data augmentation, prove effective in predicting cutting forces during hard turning, emphasizing the potential for enhanced machining optimization.