Visual Sensing Inspection for the Surface Damage of Steel Wire Ropes With Object Detection Method

Abstract

Nondestructive detection of the steel wire rope (SWR) is an essential scientific means to ensure the safe operation of rope systems, in which the visual sensing inspection method can directly grasp the surface damage condition of SWRs. However, due to the influence of algorithm performance, timely and accurate detection of SWR surface damage has always been a challenging task. At present, manual visual inspection is still the primary method in SWR maintenance. This article proposes an intelligent detection method of SWR surface damage based on an improved you only look once (YOLO) by introducing the object detection algorithm of deep learning (DL) for the first time, called WR-YOLO. The classic version-YOLOv3 was studied, and the backbone network was changed to improve its speed. To improve the algorithm adaptability under different working conditions, the dataset of SWR surface images, including three main states of health (HE), broken wire (BR), and wear (WE), was established. The proposed WR-YOLO was tested on the established dataset and compared with other YOLO algorithms. The results show that the proposed WR-YOLO can achieve a mean average precision (mAP) of 93.92% with a detection speed of 116 FPS, which is better than other YOLO algorithms in performance. Finally, the dynamic detection test was carried out, and the results show that it can achieve an accurate dynamic detection effect. Therefore, this method provides the most advanced visual inspection solution for rope surface damages, which can provide the basis for guiding engineering practice of HE maintenance.