In order to solve the problems existing in automobile enterprises after-sale maintenance

such as over-reliance on the experience of maintenance technicians

inefficient access to maintenance manuals and ineffective use of maintenance historical data

a knowledge graph in the automobile faults based on idle after-sale maintenance data of an automobile enterprise is introduced. In view of the long text type of some fields in the data

a combination of rule preprocessing and deep learning model entity extraction method is proposed to mine and utilize the vehicle maintenance historical data to complete the construction of vehicle fault knowledge map. In order to effectively utilize the vehicle fault knowledge graph to assist maintenance technicians in fault diagnosis

a vehicle fault diagnosis process based on the knowledge graph is designed

which includes a fault diagnosis method combining the knowledge graph multi-entity and extreme gradient boosting. The effectiveness of the fault diagnosis method and the practical availability of the process are verified by using the experimental comparison and actual case test.