A system for automated tool wear monitoring and classification using computer vision

This paper presents an approach for automated monitoring and classification of face milling tool wear using computer vision. A test setup with low-cost equipment for in-machine application is developed and used to generate an image dataset from worn and new tools. Different types of filters and segmentation techniques are applied and compared for image preprocessing. For wear detection, both classification and regression models using convolutional neural networks are evaluated. Best results were obtained with a combined model. This demonstrates that optical wear monitoring is feasible with low-cost equipment. However, potentials for improvement were identified in manual labeling and image quality.