Decision Method of Operation and Maintenance of CBTC System Based on Knowledge Graph

At present, the failure disposal process for the CBTC system relies heavily on the personal experience of practitioners, and the level of intelligence is at a low level. At the same time, the valuable information in the failure log text is not effectively exploited. Aiming at these problems, this paper proposes a decision-making method for CBTC system operation and maintenance(O&M) based on knowledge graph. First, the knowledge graph construction process is built by combining top-down and bottom-up methods, and the schema layer of the whole failure O&M process is designed by combining expert knowledge. Then, the BERT pre-training model is applied to extract failure entities from the labeled dataset, and the experiment is used to prove the good performance of this model. Next, the extracted entities and their relationships are in the form of nodes and edges for failure knowledge graph is populated and visualized using Neo4j. Finally, the CBTC system O&M knowledge graph is applied to give the failure cause tracing and maintenance suggestion generation method, and the usability of the proposed method is verified with examples, which can effectively assist practitioners to make the O&M decision.