Method for Machining monitoring using accelerometry coupled with a dynamic digital twin brick for smart manufacturing

This document presents an innovative method of monitoring the machining process. This method uses acceleration measurement and a dynamic digital twin in conjunction with machine learning algorithms. The aim is to detect and reduce self-amplifying vibrations which frequently occur in Industry 4.0 contexts. Such vibrations compromise part quality, shorten tool life, and reduce overall productivity. Using the measurement of physical quantities, we propose using the stability lobe method to predict unstable machining areas. Machine learning techniques such as autoencoders, PCA, K-means, DBScan and OneClassSVM are integrated to complement this approach, reducing the dimensionality of the data and classifying the operating ranges as stable or unstable. We evaluate the performance of the models using various scores and compare them with the results obtained from the stability lobe diagrams. The latter enable the accurate identification of unstable areas, while the machine learning algorithms demonstrate their effectiveness in detecting this phenomenon. The proposed method improves the proactive detection of vibrations during machining, enabling real-time adjustment of cutting parameters to ensure stable, optimal production. This work fully supports the transition to smart manufacturing by providing an advanced artificial intelligence-based solution for monitoring and optimising machining processes.