Sematic segmentation for tool wear monitoring: a comparative study

In recent years, the development of intelligent manufacturing has gradually increased the requirements for tool condition monitoring (TCM) systems, and tool wear monitoring is a very important step in the TCM system. Since semantic segmentation can accurately segment the target area and the background, it has received widely attention in the field of tool wear monitoring. However, the semantic segmentation methods currently used in the field of TCM often rarely consider the real-time requirements of the monitoring system. In this research, we conduct a detailed comparison of recent semantic segmentation models. Considering the deployment environment and real-time requirements of TCM, we compare the four aspects of segmentation performance, inference speed, number of operations and model size on the tool wear dataset collected by ourselves, and summarize the various conditions that the model for TCM should have based on the comparison results.