U-Net-based Chip Detection in CNC Machine

Removing chips from machine tools is critical to maintaining the quality and integrity of the machining process. However, this procedure also presents significant issues, including resource waste and processing time delays, particularly when the use of cutting oil for chip removal is constant or when frequent human inspection is required. Chip detection methods using traditional image processing are limited due to their vulnerability to environmental factors such as low lighting and dust. To address these limitations, we propose an approach using U-Net for segmenting the areas where chips accumulate within machine tools. Further, we suggest an optimal backbone for chip detection by modifying the existing backbone of the U-Net model. Despite complex environmental factors, our proposed method demonstrates robust segmentation performance showing its superiority over traditional image processing techniques.