A Method of Vehicle Fault Diagnosis Supporting Multi-value-chain Collaboration

摘要

To improve the collaborative efficiency of multi-value-chain in vehicle maintenance, we designed a case-based vehicle fault diagnosis method by analyzing the traditional diagnosis methods of vehicle fault and collecting user's actual requirements. In this paper, we summarized corpus of vehicle fault descriptions, analyzed user's language features and extracted proper nouns in vehicle field as an extended dictionary to improve the accuracy of text segmentation. We also built a database of vehicle maintenance cases and a vehicle structure tree to support this diagnosis method. Then using vectorization method to process fault description corpus and trained the semantic vectorization model with both statistics and topics. After that, we used vector distance algorithm to compute semantic similarity and return the optimal case in database. Finally, we located the exact position of current fault in vehicle structure tree. The experimental result shows that the accuracy of vehicle fault diagnosis achieved 86.7% by this method. The accuracy of vehicle fault locating is also achieved 81.8%. This case-based vehicle fault diagnosis system can be used for online fault consultation. It can also expand the collaborative business mode of vehicle service value chain and improve the quality and efficiency of vehicle maintenance service.