Association-Rules-Based Research Of Analysis And Application For Relevant Parts Fault Diagnosis Of EMU

EMU(Electric Multiple units) fault diagnosis system is an application example of utilizing expert system in the field of EMU. Till now, there are still a few domestic researches which are consulted in this area. Relevant researches are lack of utilizing EMU’s information or analysis on such information. Based on the CSR(China Sounth Railway company) EMU ground information management system which is designed by Network Management Research Center, Beijing Jiaotong University, this paper designs and implements the EMU fault diagnosis system which is based on analyzing EMU operation and maintenance data. Works of this paper are as follows:Designing the EMU diagnostic knowledge database. By deeply analyzing EMU operation and maintenance data and application characters of fault diagnosis system, this paper designs the knowledge base in which association relationship can be seen as knowledge. In other words, this system can discover potential association rules from huge amount of EMU operation and maintenance data, save such relationships in the form of knowledge, and use such knowledge to diagnose EMU’s current status.Improving the association relationship mining algorithm. Due to that in every year, there will be millions of EMU operating data, which makes the issue that how to find out useful association rule knowledge an essential problem. By focusing on characteristics of EMU operation and maintenance data from existing EMU, this paper improves the FP-Growth algorithm. Through carefully analyzing and testing the performance of advanced algorithm, this paper proves that the improved algorithm is better than original one from the aspect of time consumption and space consumption.Designing and implementing the EMU fault diagnosis system. Based on advanced FP-Growth algorithm, this paper designs the knowledge acquisition application. Then, according to characters of knowledge in knowledge base, this paper designs and realizes inference engine module, explain application module, maintenance recommendation module and human-computer interaction module. All of these modules are carefully analyzed and designed, and tested through using instances. The test result shows that the system has high availability.