An Explainable AI Framework for Predictive Tool Condition Monitoring in High-Speed Machining of Ti-6Al-4V: Interpretable Intelligence

This paper focuses on developing an effective tool condition monitoring (TCM) system by integrating machine learning (ML) and explainable AI (XAI) techniques for high-speed machining of Ti-6Al-4V alloy, where rapid tool wear can compromise both productivity and product quality. Features extracted from raw acoustic emission (AE) and vibration signals, collected under six different machining conditions, were used to train various supervised ML algorithms. As a key objective, the performance of Linear Discriminant Analysis (LDA) and Quadratic Discriminant Analysis (QDA) was evaluated and compared with other classifiers, including Decision Tree (J48), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Random Forest (RF). Tool condition classification accuracy improved when features from both signals were fused at the feature level. After training, XAI techniques, SHapley Additive Explanations (SHAP) and Partial Dependence Plots (PDPs), were employed to provide post-hoc explanations of the RF model’s predictions. Significant features influencing tool condition classification were identified, and explanations for specific classification instances were analyzed using SHAP. This study highlights the effectiveness of combining accurate ML models with XAI tools to enhance transparency and interpretability in tool wear prediction.