Towards Practical Tool Wear Monitoring: A Lightweight Multi-Attention CNN Approach

Abnormal cutting tool conditions significantly impact machine downtime and lifespan, leading to increased waste and reduced productivity in modern industrial manufacturing. Effective monitoring of cutting tool conditions is therefore crucial for enhancing quality and minimizing production costs. This study presents a lightweight, multiattention Convolutional Neural Network (CNN) approach for tool wear monitoring (TWM) in milling operations, using acoustic signals. A comprehensive numerical analysis was conducted to evaluate the performance of the proposed model against established architectures, including MobileNetV2, ResNet50, EfficientNet, and DenseNet121. The findings indicate that the lightweight multi-attention CNN achieves competitive accuracy while maintaining high computational efficiency, providing a practical and efficient solution for real-time tool wear monitoring.