Fault Diagnosis Method For On-board Equipment Of Ctcs Based On Rough Set Theory

The safety of high-speed train’s operation is very important.Train control system with critical safety characteristics is one of the key systems for the safe operation of high-speed train.The improvement of science and technology has provided a guarantee for the reliability of train operation control system.However,during the whole operation process of high-speed train,the unpredictability of environment and complexity of system itself indicate that the occurrence of fault is inevitable.Therefore,effective fault diagnosis method is the most important measure to ensure the safe operation of train control system.At present,the research of train control system’s fault diagnosis mainly focuses on the system-level fault analysis methods,such as Dynamic Bayesian Network.And there are few papers talking about the fault diagnosis which based on the data generated by device.In order to solve the problem above,this thesis proposes a fault diagnosis method which is based on Rough Set Theory from the view of data mining.The main contents of this dissertation are as follows:(1)This thesis elaborates the development of fault diagnosis technology detailedly.And the research object of this thesis,namely,the ATPCU-LOG of 300T on-board equipment is introduced.Taking the specific characteristics of ATPCU-LOG into consideration,the method of fault diagnosis based on Rough Set Theory is proposed from the view of data mining.And the basic principle of this method is analyzed.(2)Besides the basic algorithm of attribute reduction of Rough Set Theory,a new attribute importance measure algorithm is introduced and the inclusion relation is also taken into consideration.These improvements make the attribute reduction results more reliable.Finally,take the data of BTM unit and TIU unit as examples to accomplish the simulations with Matlab software based on the Neural Network algorithm,the combination of Rough Set Theory and Neural Network,and the Support Vector Machine algorithm respectively.The feasibility and effectiveness of the Rough Set Theory are verified by comparing the simulation results.(3)Based on the fault diagnosis rules which are obtained from the rough set attribute reduction and take the actual work flow of Data Analysis Center as the background,this thesis designs a software to make the process of data analysis and report generation intelligent.And the performance of this software is verified by comparing the results with daily records.In this thesis,the fault diagnosis method based on Rough Set Theory is put forward from the view of data mining.The feasibility of Rough Set Theory in the implement of fault diagnosis of train control system’s on-board equipment is verified by comparing the simulation results of the Neural Network algorithm,the combination of Rough Set Theory and Neural Network,and the Support Vector Machine algorithm.Finally,based on the rules above,the data analysis software is designed.It is the preliminary study about the application of Rough Set Theory in fault diagnosis of train control system.