ZPW-2000 track circuit is one of the indispensable key equipment in China’s railway signal system.ZPW-2000 track circuit is a railway signal equipment that automatically completes the inspection of train occupancy status.Its work quality directly affects the efficiency of railway transportation,and it is also an important guarantee for the safe operation of trains.The track circuit fault mode is complex,and the maintenance personnel take a long time to deal with the fault,which seriously affects the order of railway transportation.It is the difficulty of field maintenance of railway signal.It has become a hot topic to explore various intelligent diagnosis methods to shorten fault processing time and improve the efficiency of track circuit fault maintenance.However,in the research of ZPW-2000 track circuit fault diagnosis,mostly based on the voltage value or current value obtained by the centralized monitoring system,the induction signal collected by the locomotive or the dynamic inspection vehicle,the steady-state value is analyzed and processed.This kind of method has high accuracy for the explicit fault diagnosis with clear fault mechanism or more historical fault information,but it may ignore the transient fluctuation state between normal and fault state under the steady-state condition,with little or no fault information.If only two states of normal and fault are distinguished,it is difficult to meet the diagnosis requirements of railway field for track circuit equipment.Therefore,in this thesis,aiming at the problem of insufficient timeliness under the steady-state condition,ZPW-2000 track transmission model is established by Simulink,and data is collected by the track circuit test base of Emei campus of Southwest Jiaotong University,and multi faults in the transient process are analyzed.The specific research contents are as follows:(1)Based on the Simulink model,the ZPW-2000 track circuit is modeled,and each component of track circuit is simulated and analyzed to provide data support for the subsequent track circuit transient fault diagnosis.(2)According to the system configuration of the track circuit test base in Emei campus,the software and hardware scheme of ZPW-2000 track circuit data acquisition is studied.According to the characteristics of track circuit data acquisition,select the appropriate acquisition unit,write Lab VIEW data acquisition operation interface,and complete the track circuit analog data acquisition.Based on the collected simulation data,the correctness of the Simulink track circuit model is verified.(3)Aiming at the problems that the transient fault characteristics of the rail circuit are not obvious and easily affected by noise,the Mel frequency cepstrum algorithm is introduced to realize the feature extraction of the transient voltage signal of the rail circuit fault.In order to further solve the problem of high feature extraction dimension of MFCC algorithm,principal component analysis(PCA)algorithm is used for dimensionality reduction.Taking rail circuit fault transient signal as input,after feature extraction and attribute reduction,Fisher’s criterion is used to find the optimal feature set.The results show that when eight-dimensional feature information is used,the feature extraction effect of the transient signal of the track circuit fault is better.(4)The fault transient signal diagnosis model of track circuit is studied.In this thesis,the characteristics of transient signals are analyzed,and the transient fault template library system is constructed.To solve the problem of redundancy caused by expanding template database,K-means clustering algorithm is used to realize the adaptive optimization of template database.By analyzing the principle of dynamic time rule(DTW)algorithm,DTW is used to match the test data and template database data dynamically,so as to realize the transient fault diagnosis of track circuit.The results show that the adaptive template library and DTW model can effectively diagnose the transient signal of track circuit.Finally,the research content of this thesis is summarized and the future research content is prospected. |