A framework for fault detection method selection of oceanographic multi-layer winch fibre rope arrangement

•The collection of a substantial video dataset of the fibre rope in oceanographic winch winding process is built. The semantic segmentation algorithm was applied to fault detection in the fibre rope arrangement, and the performance of different models under various construction methods was validated for fault detection.•A model selection method was proposed. This method constructs evaluation indicators and assessment system to adapt models for varying engineering requirements.•Analysis scenarios were established for real-time monitoring during the maritime voyage of ships, for off-ship video analysis specifically targeting fault analysis, and for small detection devices used in embedded systems.•The training of the model underwent evaluation from various perspectives, affirming both the reliability and the practicality of the assessment approach. The quality of synthetic fibre rope arrangement is crucial for the orderly and efficient launch and recovery of marine exploration equipment in oceanographic winch systems. However, due to the unique properties of synthetic fibre ropes, faults are prone to occur during the rope winding process. This paper proposes a framework for fault detection method selection of oceanographic multi-layer winch fibre rope arrangement. An autonomous image dataset is used to train and evaluate the effectiveness and performance of various semantic segmentation models. Based on the performance metrics, a training model selection framework is proposed. The results of the fault detection method demonstrate that deep learning semantic segmentation algorithms perform well in the detection of rope winding arrangement faults. This evaluation framework provides guidance for the best model selection under various detection scenarios. It has made contribution to the fault detection and safe operation of oceanographic multi-layer winch fibre rope arrangement.