Wire rope self-rotation measurement method based on sliding column block modeling and one-dimensional convolutional neural network

Abstract Wire ropes are widely used in lifting, transportation, mining, and marine engineering. Their complex braided structure brings excellent tensile strength and fatigue resistance. However, during use, additional stress is often generated due to rotation or torsion, which leads to increased local friction and accelerated wear, which in turn causes fatigue fracture, affecting service life and equipment safety. Therefore, monitoring the rotation angle and torsion state of wire ropes is of great significance to improving reliability and extending service life. This paper proposes a wire rope self-rotation measurement method based on sliding column block modeling and one-dimensional convolutional neural network (1D-CNN). The method first preprocesses the wire rope image, including region of interest extraction and axis fitting, and completes image distortion correction to improve geometric accuracy. Then, the sliding column block mechanism is introduced to perform local context modeling on the image pixel information, and the correlation map is constructed through normalized cross-correlation to filter out the interference area. Subsequently, the improved 1D-CNN was used to train and identify the column block classification results, so as to accurately extract the key image features reflecting the rotation speed. Experimental results show that the proposed method improves measurement accuracy by 113.5% after image information screening. Under the condition of a 1° rotation per frame, the average angular error is reduced from 0.098 91° to 0.046 35°, significantly outperforming traditional methods. Finally, tensioning device experiments verified the stability and accuracy of the proposed method under actual working conditions.