Research on Cock States Detection of Cock Oven Inspection Robot Based on YOLO-v3

The state switching of the coke oven gas pipeline exchange cock plays a vital role in the safe production of coking. This paper designs and manufactures a rail-type mobile robot applied to coke oven basement inspection tasks to troubleshoot coke oven heat exchange equipment. Three detection methods are proposed for the opening and closing state of the exchange cock. Each detection method is analyzed according to the actual production process and requirements. The cock images are collected by the robot and made into a dataset. YOLO-v3 is used to train models of three detection methods on the same experimental platform and compare them. Finally, the optimal detection model is selected and deployed on the inspection robot. The detection accuracy rate reaches 99.1%, which effectively reduces the accident rate and maintains the stability of the coking system.