The fault knowledge graph of electromechanical equipment can not only store the knowledge of equipment faults, but also combine the reasoning technology on the graph to mine new fault modes and conduct causal traceability of faults, which is conducive to achieving the health management of electromechanical equipment. A design method for the knowledge graph of electromechanical equipment based on large models is proposed: The ontology framework is designed around the system structure principle and fault case data. Combined with the offline deployment of large models to identify named entities and relationships, knowledge extraction based on large models is achieved. On the basis of knowledge fusion, the storage and visualization of the fault knowledge graph are realized with the Neo4j database. Fault reasoning was carried out based on the framework of Cypher language in combination with the expert thinking mode. The fault knowledge graph construction and reasoning were carried out taking a certain type of unmanned equipment as the object, verifying the feasibility and effectiveness of the proposed method.