SoC-Based Early Failure Detection System Using Deep Learning for Tool Wear

This study aims to implement a system-on-chip (SoC) detection system for tool wear monitoring and alarms for high-precision machining processes. The proposed deep learning approach is trained by the collected sensors from real performed in a three-axial computer numerical control (CNC) machine center combined with different conditions of spindle speed and tightening torque. The corresponding vibrational and sound signals were collected by a three-axial accelerometer and micro-electro-mechanical-system (MEMS) microphone, and then the tool flank wear was measured by a camera. In this study, the flank wear was designed for early detection according to the ISO 8688-2:1989 standard. A deep learning model with frequency spectrum inputs of the collected signals was developed for tool wear prediction. In addition, to treat the machining variation for detection, two sensor fusion approaches are presented and implemented on an SoC-board (Pocket Beagle) for landing and cost reduction. The corresponding average detection accuracies were approximately 99.7% and 87.75% for the single and merged models, respectively. The results demonstrated the effectiveness and performance of the proposed approach.