Objective The operation and maintenance of existing signaling equipment faces problems such as low maintenance efficiency, reliance on manual experience, complex fault modes across lines and equipment, and difficulties in deep integration of multi-source heterogeneous data. It is therefore necessary to study an intelligent diagnosis and fault repair model for signaling equipment to provide technical support for equipment operation and maintenance.

Method A log-based fault diagnosis method based on natural language processing and a fault information analysis method based on Word2Vec are introduced. A domain-adapted pre-training is performed using the BERT pre-trained model to capture deep semantic features of logs, and an intelligent diagnosis and fault repair model for signaling systems based on multi-source semantic information is established. On this basis, a multimodal maintenance knowledge base and intelligent analysis system based on system logs, fault reports, equipment manuals, and maintenance procedures are constructed, and a closed-loop mechanism of 'diagnosis-repair-feedback' for equipment operation and maintenance is established. The signaling system of a metro line in Beijing is taken for pilot application to verify the effectiveness of the proposed model and system.

Result & Conclusion  Compared with the traditional manual modes, the average fault diagnosis time of the intelligent diagnosis and fault repair model is reduced by 40%, and the comprehensive accuracy of fault diagnosis is improved to 92%, demonstrating its engineering applicability. The developed intelligent analysis system can achieve automatic fault localization, intelligent generation and dynamic optimization of maintenance strategies with a possession of continuous learning and optimization capabilities, effectively improving the level of intelligence and response efficiency of signaling equipment operation and maintenance in urban rail transit.