Explainable AI for tool wear prediction in turning

Temperature is the key indicator: The research identified tool temperature as the most significant factor in predicting tool failure, meaning operators can focus monitoring efforts on this single measurement rather than trying to interpret complex machine learning outputs.

XAI makes AI decisions understandable to workers: By using the Shapley method to explain predictions, the framework allows machining operators—not just data scientists—to understand why the AI predicted a tool would fail, supporting human decision-making rather than replacing it.

Multiple sensor data improves predictions: The system combined four different types of measurements (acceleration, acoustics, temperature, and spindle speed) to make tool wear predictions, demonstrating that responsible AI implementation benefits from diverse data sources rather than relying on a single indicator.

Practical application for manufacturing: This framework addresses a real operational need—predicting tool failure before it happens—which can reduce production downtime and waste while helping workers maintain better control over their equipment.

Transparency enables trust and accountability: By making the machine learning classifier's reasoning visible through explainability techniques, operators and managers can verify the system is working correctly and identify potential problems, rather than treating the AI as a "black box."