Research On Prediction Method Of Shunt Malfunction Of ZPW-2000A Jointless Track Circuit

As an essential part of train operation control system in China,track circuit can not only check the state of sections and the integrity of lines,but also transmit the running information to the train.However,since track circuit is located outdoors,its working state is easily affected by weather,environment,line pollution,etc.For example,the resistance of the ballast bed decreases when rainy and snowy weather aggravate the leakage,causing the red-light strap fault.In addition,the lines that have not been running for a long time or in wet areas,oxides are easy to form on the rail surface,so that the train wheel sets cannot short the rail well,resulting in shunt malfunction.Thus,it is urgent to find out and repair the faults in track circuit system timely,with the purpose of improving the operation efficiency and avoiding accidents.This thesis mainly focuses on shunt malfunction and puts forward a prediction method based on researching the difference of the shunt current generated by the shunt state of track circuit in the normal and fault condition.Due to the large amount of failure data required to prediction models,which are confidential that cannot be obtained in large quantities.Therefore,this thesis by analyzing the structure and working principle of ZPW-2000 A jointless track circuit,and depend on transmission line theory and Kirchhoff’s law,four-terminal network models are established for the components with better insulation protection,and six-terminal network models are built for the components such as main track circuit and syntonic sections with serious leakage to ground.The equivalent model of track circuit is cascaded according to the structure of track circuit,which is used to simulate and analyze the three types of shunt malfunction signals and compare with the normal signal.The validity of the model is verified according to the existing field data.The simulation results show that the maximum relative error of the track circuit equivalent model is 6.41% in regulated state,and the simulated shunt current amplitude envelope curve in shunted state is consistent with the field locomotive induced voltage amplitude envelope curve.Thus,the accuracy of the model is higher than that of the traditional four-terminal network.The traditional fault detection methods of shunt malfunction of track circuit depend on the professional ability of electrical service personnel,which is prone to misjudgment,omission and other problems.In addition,the traditional methods are suitable for faults happened,but cannot play the role of protection and early warning.Thereby,the combined analysis method is proposed in this thesis,which is used to analyze and predict shunt malfunction of track circuit.Firstly,the shunt current signal is decomposed by variational mode decomposition(VMD)to extract the energy entropy feature,and then weighted fused with the deep expression feature extracted by cascade block convolution neural network(CB-CNN)to obtain the sensitive feature for fault prediction.Since the decomposition effect of VMD will be affected by the quadratic penalty factor and the number of mode decompositions,the improved whale optimal algorithm(IWOA)is used to optimize the two parameters.The simulation results show that VMD can effectively decompose the shunt current signal,and the shunt current signal in normal condition and shunt malfunction can be distinguished by the decomposition diagram of each mode component.What’s more,VMD optimized by IWOA can effectively search for the optimal parameter combination solution of quadratic penalty factor and mode decomposition number.The key to the prediction of shunt malfunction of track circuit is to establish the prediction model.The equivalent model of track circuit is used to simulate the shunt current signal,so as to train and verify the prediction model.Besides,five experimental schemes are designed by VMD,convolution neural network(CNN)and CB-CNN to verify the effectiveness of the combined analysis method adopted in this thesis.Comparison with other methods,the combined analysis method has achieved higher values in the four evaluation indexes of accuracy,precision,recall and F1-socre.The results show that the combined analysis method has better prediction of shunt malfunction of track circuit.