Proceedings of the
World Congress on Micro and Nano Manufacturing (WCMNM 2022 )
19–22 September 2022, Lueven, Belgium
doi:10.3850/978-981-18-5180-3_RP19-0022
The Automatic Tool Wear Monitoring System for Micro-Milling Application with Image-Based Wear Detection
1Mechanical Engineering Department, Universitas Indonesia
2Mechanical Engineering Department, Universitas Katolik Indonesia Atma Jaya
ABSTRACT
Tool wear monitoring (TWM) is the one of crucial aspects in achieving a high-quality micro-milling process. The unmonitored wear progression may cause an early catastrophic tool-breakage during the micro-milling process, leading to the process shutdown. Furthermore, the micro-milling with the unmonitored tool wear may also fail to produce the specified surface roughness. This paper presents the development of an automatic tool wear monitoring system for the micro-milling application to solve the aforementioned problems. The presentation focuses on the direct approach of the micro-tool wear monitoring by using Convolutional Neural Networks (CNN) and U-Net segmentation technique. The wear occurrence on the micro-tool is inferred from the micro-tool images captured by a camera. From the experiment, the approach achieved 87.6% success in distinguishing the micro-tool type by using CNN and up to 78.1% success in segmenting the wear region by using U-Net segmentation technique. The wear detection algorithm was developed as the component of the bigger system, i.e., the automatic TWM system.
Keywords: Micro-milling, tool wear monitoring, convolutional neural network, U-Net.
1Mechanical Engineering Department, Universitas Indonesia
2Mechanical Engineering Department, Universitas Katolik Indonesia Atma Jaya
ABSTRACT
Tool wear monitoring (TWM) is the one of crucial aspects in achieving a high-quality micro-milling process. The unmonitored wear progression may cause an early catastrophic tool-breakage during the micro-milling process, leading to the process shutdown. Furthermore, the micro-milling with the unmonitored tool wear may also fail to produce the specified surface roughness. This paper presents the development of an automatic tool wear monitoring system for the micro-milling application to solve the aforementioned problems. The presentation focuses on the direct approach of the micro-tool wear monitoring by using Convolutional Neural Networks (CNN) and U-Net segmentation technique. The wear occurrence on the micro-tool is inferred from the micro-tool images captured by a camera. From the experiment, the approach achieved 87.6% success in distinguishing the micro-tool type by using CNN and up to 78.1% success in segmenting the wear region by using U-Net segmentation technique. The wear detection algorithm was developed as the component of the bigger system, i.e., the automatic TWM system.
Keywords: Micro-milling, tool wear monitoring, convolutional neural network, U-Net.
