Enhancing Generalization Capability of Tool Wear Prediction in Drilling Process Using Nonlinear Regression

The rapidly increasing demand for automation and efficiency in human resources within flexible manufacturing systems has made automated and systematic tool condition monitoring (TCM) and high-precision real-time anomaly detection essential for the cutting process. Drilling is crucial among cutting operations because inaccuracies can result in errors in the subsequent machining operations. The manufacture of even a single product requires drills of varying tool diameters. However, studies aimed at enhancing the applicability of TCM for tool diameter have not been conducted. This study proposes a new nonlinear regression model that improves the prediction accuracy for tool wear during the drilling process. The model is developed through a series of cutting experiments across tool diameters to ensure its reliability. We utilize the cutting force in the thrust direction to develop the nonlinear regression model based on both thrust force and tool diameter. Computational simulations demonstrate that the proposed method achieves higher prediction accuracy than conventional linear regression model and deep learning approaches specialized in time series data.