Intelligent visual detection method for the early surface damage of mine hoisting wire ropes

Abstract An intelligent visual detection method is proposed to identify early surface damage in operational mine hoist steel wire ropes (MHWRs), addressing challenges arising from complex surface morphology, subtle early-stage damage, and difficulties in identification. The method is based on an improved YOLOv5 (you only look once) network, a visual detection system has been developed, and on-site experiments and applied research are currently underway. Firstly, the operating conditions of the in-service MHWRs were analyzed. In response to their dynamic hoisting characteristics and complex surface morphology, a detection system framework based on high-speed visual perception and deep intelligent algorithms was proposed. Then, the Retinex and convolutional block attention module (CBAM) attention mechanism modules were introduced to solve the problems of uneven illumination and subtle early-stage damage, and a visual recognition network model for detecting surface early-stage subtle damage in MHWRs was constructed on the YOLOv5 base module, referred to as MineWR-net. Subsequently, based on the evolutionary process of surface damage in MHWRs, a dataset for early-stage damage was established, and performance comparative studies were conducted on various object detection algorithms. Finally, integrating practical operating conditions and application requirements, on-site experimental research was carried out for the MHWRs visual detection system. The results indicate that the designed MHWRs visual inspection system can achieve dynamic and clear acquisition of the surface image of serving steel wire rope under high-speed movement. Compared with other target detection algorithms, the average detection accuracy (Map) of MineWR-net is 82.3%, which has superior detection performance. This research can provide technical support for the industrial application deployment of the healthy operation and maintenance of MHWRs.