Research On Fault Diagnosis Of Impedance Match Bond Based On Intelligent Algorithms

In electric railway, the unbalanced current caused by tractive power supply system may interfere with track circuit through conductive coupling. The impedance match bond, combining track circuit system and tractive power supply system, yields significant results of anti-jamming, so it plays an important role in reducing electrification interference in track circuit. However, it has a chance to break down because of the minute hidden danger or the rugged work environment. Actually, it ever caused "red zone" for sometimes. When an accident happens, this fault is concealed, so it takes a long time to look for and repair. Indeed, it reduces the operation efficiency of railway line. Therefore, it is necessary to study the fault diagnosis of impedance match bond. It can use intelligence algorithms to determine the fault types in terms of the real-time monitoring data which is collected. So, it can shorten hours of fault removal and ensure the safe and efficient railway transportation.In this context, this thesis analyzes the advantages and disadvantages of fault diagnosis which is usually used for common transformers and railway signal facilities, and then designs the fault diagnosis of impedance match bond by using rough set and fuzzy theory. It can determine the fault types accurately and ensure that signalmen can begin troubleshooting after knowing the fault type. So, it can speed up the progress of fault removal. The main contents of this thesis are as follows:This thesis focuses on how to design the accurate fault diagnosis of impedance match bond. The overall structure of the fault diagnosis system that can be used in railway actually is discussed, and the structure of each subunit and how to implement its function are analyzed in detail. Then, the content of fault diagnosis is introduced, it includes three parts:data collection, data processing, and fuzzy inference.In order to collect fault data, the integrated single feeding and single receiving 25 Hz phase detecting track circuit is built in the laboratory, and the locations of feature points needed to monitor are determined. In different scenarios which are simulated by several variables, the normal state and multiple failure states are emulated, and a mass of failure data is collected.In order to obtain the concise decision rules of fault diagnosis, rough set is used to analyze massive amounts of monitoring data. First, relative-attribute reduction algorithm is proposed. It can reduce the relative attribute of fault diagnosis decision table effectively; Then, the algorithm for decision rules acquisition is proposed. It can process the decision table obtained in the previous step, so the final decision rules of fault diagnosis are obtained and modeled rudimentarily.Finally, the fault model is refined based on fuzzy theory and fuzzy inference is selected to study first. Then, some representative fault samples are used to test the fault diagnosis so as to ensure that accuracy meets requirement.