On-machine tool wear monitoring system using strain-gauge signals and edge force coefficients with physics-informed machine learning

Official URL: https://dx.doi.org/10.1016/j.cirp.2026.04.030

This paper presents a hybrid framework for tool wear monitoring in milling integrating strain sensor-based force measurement signals (SSFM), mechanistic force-coefficient identification, and machine learning (ML). Strain sensors and dynamometer signals train ML models that estimate cutting forces across various cutting conditions. Reconstructed forces feed an inverse mechanistic model to estimate cutting and edge coefficients over tool life. These coefficients are then used by an XGBoost model to predict in-process flank wear. Experiments across multiple cutting conditions demonstrate real-time estimation, accurate force prediction, and reliable wear estimation, showing that physics-informed features improve robustness and production compatibility.