Analysis of Tool Wear Evolution in CNC Machining Using Multi-Sensor Data and Deep Learning
Please login to view abstract download link
Tool wear monitoring remains a key challenge in CNC machining due to its strong impact on process stability, component quality, and tool life management [1]. This work investigates the evolution of wear in coated carbide drilling tools on X45NiCrM04 steel workpiece by combining multi-sensor signal acquisition with deep learning–based tool condition classification. Electrical current drawn by the machine tool, cutting force signals, and acoustic emissions captured through a microphone are monitored during machining in order to capture the progressive changes associated with tool degradation. A set of controlled machining experiments is performed using new cutting tools with progressive monitored wear to generate labelled datasets (with distinct wear states). Convolutional Neural Networks (CNNs), with and without Long Short-Term Memory (LSTM) are employed to automatically learn discriminative representations from raw sensor signals, enabling the classification of tool condition without relying on biased visual inspection [2]. The performance of the proposed approach is assessed for individual sensors and for multi-sensor fusion strategies. The results indicate that combining electrical, mechanical, and acoustic sensing improves the robustness and accuracy of tool wear classification compared to single-sensor approaches. The proposed methodology demonstrates the potential of deep learning–based multi-sensor monitoring as a scalable and nonintrusive solution for tool condition monitoring in CNC machining environments.
