Research On Fault Diagnosis For Track Circuits Based On Improved Decision Tree

The track circuit is essential basic signal equipment for the railway to ensure the train's safe operation,and its normal work or not will directly affect the railway transport efficiency.At present,the maintenance of track circuit is through regular maintenance or fault repair in railway site,the failure occurred with a large uncertainty and randomness,system diagnostic process exist blindness and complexity,lacking clear fault logic mechanism analysis.A major problem faced by maintenance personnel is how to diagnose system fault quickly and accurately.In view of the above problems,the main contents of this thesis are as follows;Firstly,introducing the equipment structure and working principle of the 25Hz phase-sensitive track system,and analyzing system common faults.The equivalent four-terminal network model of the phase-sensitive track circuit is established by using the uniform transmission line theory.Combined with the model specific parameters to derive each node monitoring value.Then,establishing the phase-sensitive track circuit's diagnosis decision table.For the fuzziness and uncertainty of system fault feature,this thesis proposing a fault diagnosis method based on PSO-FDT.The Min-Ambiguity algorithm is used to train the decision table to establish fuzzy decision tree.Meanwhile,PSO algorithm is used to optimize two parameters which influencing the performance of decision tree.Then extracting the rule from the established fuzzy decision tree and establish the diagnosis rule base as the basis for the fault diagnosis of the track circuit.The simulation results show that the PSO-FDT method reduces the blindness of the method based on the experience setting,and it is feasible to diagnose the phase-sensitive track circuit.Then,take the ZPW-2000A jointless frequency-shifted track circuit as the research object.For the complexity of the system composition and the redundancy of the fault feature,a combined network model based on C4.5 decision tree method is proposed to diagnose system fault.According to the structural characteristics of ZPW-2000A track circuit,a variety of failure modes are divided,and diagnostic model of the combined network is established to realize the hierarchical division of the track circuit fault from coarse to fine.Furthermore,the rough set attribute reduction method is adopted to extract the appropriate attributes for each tree node of the combined network.As a result,diagnostic decision tables of each tree node are obtained.Then,decision tables are trained to extract failure rules using C4.5 Algorithm.The simulation results show that the model can improve the performance of the diagnosis classifier in order to achieve better diagnostic results.Finally,according to the above research,the VC ++ 6.0 platform and MFC class library are used to visual programming,build the track circuit fault diagnosis system based on improved decision tree algorithm and realize the track circuit related fault diagnosis function.Inputting the relevant fault feature data into the diagnosis system,it can give the diagnosis result and the troubleshooting suggestions in time,and provide reference for system's fault handling and equipment maintenance.