Deep Learning-Based Intelligent Defect Detection of Cutting Wheels with Industrial Images in Manufacturing

Yang, Shaojie; Li, Xiang; Jia, Xiaodong; Wang, Yinglu; Zhao, Haodong; Lee, Jay

Abstract The cutting wheel is an important tool in the television liquid crystal display (LCD) panel manufacturing process. The degradation of the cutting wheel significantly affects the LCD panel quality. Currently, there is few effective approaches that can detect the degradation of the cutting wheel at the working station for health monitoring purpose, due to the small size of the component and the complex manufacturing operation. That leads to high economic costs in the production lines in the real industries. In order to address this issue, this paper presents a deep convolutional neural network-based method for defect detection of the cutting wheels using the industrial images. An end-to-end health monitoring system is built based on machine vision, which directly takes the raw images as inputs, and outputs the detection results. That facilitates the industrial applications since little prior knowledge on image processing and fault detection is required. The experiments on a real-world cutting wheel degradation dataset are carried out for validation. High fault diagnosis testing accuracies are obtained, that indicates the proposed method offers an effective and promising approach for the cutting wheel health monitoring problem.