"5G+machine vision" helps Angang steel strip inspection
Wang Xianfeng, a reporter from China Economic Herald, China Development Network, reported that recently, the "5G based machine vision strip surface inspection platform" project was applied in the cold rolling plant of Angang Steel, which not only realized distributed, long-term, stable, online identification and storage of strip surface defect information, but also realized subsystem data sharing. Each system can bring more than 1.8 million yuan of economic benefits to users every year.
The project is undertaken by Ansteel Information Industry Company in conjunction with Ansteel Group Beijing Research Institute. Angang Steel cooperated with China Mobile to carry out the project construction of "R&D and application of 5G based machine vision strip surface inspection platform" in Liaoning Province, which realized the application of machine vision quality inspection and production site monitoring scenarios, and was selected into the practice of key industries in the steel industry in Ten Typical Application Scenarios and Five Key Industries Practice of "5G+Industrial Internet".
It is understood that in terms of machine vision quality inspection, industrial cameras are deployed to take high-definition pictures and collect quality inspection data, and 5G network is used to transfer the collected high-definition image data such as 4K/8K at the cold rolling site back to the operating room platform. The platform's visual AI analysis capability is used to process and analyze the images and complete the real-time detection of strip steel surface defects; At the same time, the rolling speed of strip steel is very high. The flatness of strip steel is fed back through the reflective zebra stripes on the surface of strip steel, which is used for real-time detection of strip steel production quality and provides basis for adjustment of tension rolls. After the plan is deployed, the detection rate of conventional defects in the strip steel is more than 95%, and the online comprehensive defect classification rate is more than 90%, which improves the yield rate and reduces the downtime for strip breakage and roll change caused by strip defects. Through the analysis of the pictures collected on site, the detection accuracy is more than 99.99%; At the same time, the camera shall be analyzed separately to determine whether there is a large amount of dust covering the camera, and the lens shall be cleaned in time, which can save about 1 million yuan of belt maintenance cost every year.