Application of Deep Learning in Steel Cable Surface Damage Detection: A Case Study of the YOLOv8-SWRSD Model

Surface damage detection of steel cables is critical for industrial safety, yet traditional methods face inefficiency and subjectivity. This study focuses on surface damage detection of steel cables under fixed-camera scenarios across diverse scenes, and proposes YOLOv8-SWRSD, an enhanced deep learning model integrating RepNCSPELAN4, C2f_EMBC, and C2f_SWC modules to improve dual-scale defect recognition. Evaluated through generalization ability and robustness verification, compared to YOLOv8, the model achieves 1.5% and 1.2% improvements on abrasion, while broken wires show 8.2% and 12.4% improvements, respectively. It also maintains robustness under blurred and grayscale conditions. For broken wires detection, YOLOv8-SWRSD maintains a 1.8% mAP50 advantage over the typical model YOLOv8-Ghost under identical dataset conditions. This work delivers a reliable cable defect detection solution, advancing industrial inspection safety. Future work will develop novel architectures and attention mechanisms, enhancing detection of diverse cable damages.