Research On Intelligent Fault Diagnosis For Track Circuits

Track circuit is one of the most important fundamental infrastructure in the railway signal system, with the rapid development of high-speed railway in our country, its "health" is directly related to the transport efficiency and the safety of the high-speed train. Presently, the maintenance mode of the track circuit in our country still remain in the traditional fault repair mode and planned repair mode, its faults having obvious uncertainty and ambiguity, and the blindness and complexity of the diagnostic process is strong. How to deal with its fault diagnosis quickly and effectively is still a major problem which the professional and technical personnel in railways have to face. So, the research on the high-speed railway track circuit intelligent fault diagnosis is of great significance.The main contributions in this thesis are as follows:Firstly, the study took the 25 Hz phase-sensitive track circuit in railway station as the research object, described the system composition and working principle, and analysis the system common faults. For the uncertainty and fuzziness of its faults, the study proposes a track circuit fault diagnosis improved method based on mamdani fuzzy BP neural network. By establishing diagnostic system model, this method used the adaptive-momentum BP learning method to train and optimize the model parameters, discussed the initial value of system parameters and given derivation process detailed. Simulation results show that under the same experimental conditions, the improved method reduce the training error and effectively improve the stability and convergence speed of the learning process. Introducing this method to 25Hz phase-sensitive track circuit faults diagnosis has high feasibility.Secondly, the study took the red light strap fault in the ZPW-2000A jointless frequency-shift track circuit as the research object, described the system composition and working principle. For the blindness and complexity of the fault diagnosis process, the study proposes a track circuit intelligent fault diagnosis method based on the FTA and multilevel fuzzy neural sub-networks. Through the establishment of macro tree and micro tree, this method divided the complex diagnosis process, used the FTA analysis logical to decide the fault source and extract the rules of fault diagnosis, builded several mamdani fuzzy neural subnets and connection them into the diagnosis model by the series-parallel method, used the momentum BP gradient optimization algorithm to optimize and adjust the model. Through the simulation analysis, show that the method is feasible and effective. This method can quickly locate the point of failure, greatly reduce the blindness and complexity of the fault diagnosis and improve the diagnosis speed and accuracy. It also can provide reference value for the intelligent diagnosis and maintenance of track circuit.Then, according to the fault diagnosis model which had been built in the study, designed the intelligent fault diagnosis system of track circuit. It used the VC++ 6.0 and the MFC class library to build the software development platform, and realized the function of the track circuit intelligent fault diagnosis by visual programming. Debugging results show that when the track circuit failure, after entering the relevant fault characteristic data into this system, it can diagnose fault timely, give a troubleshooting suggestion, and provide the basis for track circuit routine maintenance and troubleshooting.Finally, the content of this thesis is summarized, and the prospect of the further researching work is discussed.