Advanced techniques in milling process monitoring

Official URL: https://www.utis.tc/gorseller/files/UTIS%202024_Proceedings.pdf

This paper presents an advanced methodology for monitoring of milling process through a hybrid Physics-Based Machine Learning (PBML) approach. The proposed method integrates mechanistic force models with machine learning algorithms to predict cutting forces with high accuracy. By utilizing accurate simulated data to train machine learning models, the study addresses the challenges of high costs and extensive time requirements associated with real experiments. The PBML model demonstrates over 97% prediction accuracy across various materials, including unseen datasets. Additionally, in the second layer of machine learning, trained using enhanced simulations, process parameters are identified from cutting forces for use in monitoring and fault detection. the model's real-time fault detection capabilities are validated through experimental testing on complex geometries, confirming its effectiveness in identifying process deviations and enhancing manufacturing efficiency. This approach is poised to significantly enhance unmanned manufacturing environments by enabling precise process monitoring, fault detection, and parameter optimization, demonstrating strong potential for industrial deployment.