Research On Fault Diagnosis Of ZPW-2000A Track Circuit Based On Bayesian Network

ZPW-2000 A track circuit system is developed based on the UM71 track circuit by Chinese technology, which has been more widely applied in the railway system of our country, because it has strong anti-jamming and good transmission performance. The fail-safe problems of ZPW-2000 A track circuit system is becoming increasingly important with the development of high-speed railway. Because of the completeness of network of bidirectional reasoning mechanism and fault symptom analysis, Bayesian network fault analysis becomes one of the most common methods of intelligent analysis.ZPW-2000 A track circuit fault diagnosis system is established based on Bayesian network. The main contributions of this dissertation are as follows:(1) On the basis of researching structure and the work principle of ZPW-2000 A track circuit system, this dissertation establishs the track circuit failure mode of "failure cause-failure mode-failure effect" and effects analysis(FMEA) table. Combined with fault tree model, this dissertation has completely elaborated the relationship between each unit of the system and avoided the omission of symptom of the fault.(2) Based on the FMEA table and fault tree model, this dissertation has established the CME Bayesian network model with the model of "fault cause-fault mode-fault effect" with three layer structure. According to the fault polymorphism and the redundant structure of the ZPW-2000 A track circuit system, this dissertation has analyzed the model using CME Bayesian network and summed up the impact of the failure probability of the redundant structure, which improved the accuracy of the fault diagnosis. At the same time, the fuzzy theory was introduced into Bayesian network, which enhanced its ability to deal with incomplete information and solved the less prior information and difficult to acquire the problem of track system unit. Using the bidirectional reasoning function of the bayesian network, this dissertation has analyzed the conditional probability and causality between each system unit, which is proved effectiveness of bayesian method in the fault diagnosis field.(3) Based on Hugin8.1 and the Bayesian network toolbox of Matlab,the accuracy of Bayesian network can be verified through analying specific examples of the fault reasoning.By using the C# programming software, a ZPW-2000 A track circuit fault diagnosis system is set up. The fault diagnosis system has visibility and operability, which is provided maintenance recommendations to on-site technical personnel and improved efficiency of the proposed fault diagnosis method.