Research On Classification Method Of Fault Of ZPW-2000A Jointless Track Circuit

The track circuit is an indispensable part of the railway signal system,which can provide operation information to the train,check the integrity of the track section and whether the track section is occupied by the train.Its work directly affects the safety and operation efficiency of railway transportation.However,the complex structure of track circuit and the fact that some equipment is located outdoors lead to high failure rate and diversity of faults,resulting in high work intensity and low accuracy of fault judgment.At present,the fault identification of the track circuit is mainly carried out by the staff to analyze the microcomputer monitoring data.This method has the problems of strong subjectivity and long fault identification time,which cannot meet the requirements of timely and accurate identification of fault types at this stage.Therefore,it is necessary to find a classification method that can accurately and objectively identify the fault types of track circuits to ensure the reliable operation of track circuits.In this thesis,according to the characteristics of the ZPW-2000 A jointless track circuit and based on the existing fault diagnosis research,the following research work is carried out:Firstly,the transmission model of the jointless track circuit under adjustment state is established.It is used to solve the problem that it is difficult to obtain the field fault data of the track circuit.The small amount of fault data restricts the learning ability of the classifier,which leads to the problem of low accuracy.The establishment of the transmission model is mainly based on the four-terminal network,the uniform transmission theory and the analysis of the working principle and electrical characteristics of the ZPW-2000 A jointless track circuit.By setting different fault parameters to simulate the frequent faults of the track circuit and calculating the values of each monitoring point,the track circuit fault data set is obtained.Secondly,this thesis uses the ReliefF algorithm and Sequential Forward Feature Selection(SFFS)method for feature parameter selections.The relief F algorithm is used to assign weights to all feature parameters in the original feature parameter set according to the classification ability,and sort them according to the weight size.Then the SFFS method is used to select the sorted feature parameters.This method takes into account the changes in classifier accuracy when selecting feature parameters.It is used to solve the problem that the selecting feature parameters and fault discrimination are carried out independently in the existing research on track circuit fault classification,so that the beneficial feature parameters for fault classification accuracy may be eliminated,resulting in low accuracy.Thirdly,this thesis proposes a track circuit fault classification method based on model fusion to solve the problem that the single algorithm considers the angle one-sided,resulting in low fault classification accuracy.C4.5 decision tree classifier,Adaboost.M2 classifier and Extreme Learning Machine(ELM)classifier are used for the preliminary classification of track circuit faults.The optimal weight of the combined model is calculated according to the error rate of the three classifiers.The three classifiers are then weighted.The classification results of the three classifiers are fused to obtain the final fault classification results.Finally,this thesis verifies the above track circuit fault classification system.After obtaining the fault data set through the transmission model,the Relief F algorithm and SFFS method are used to select the features.Therefore,the optimal feature subsets of the three classifiers of C4.5 decision tree,Adaboost.M2 and ELM are obtained.And the three optimal feature subsets are divided into a training set and a testing set.The training and testing of the three classifiers were completed.Then the proportion of the three classifiers in the output of the combined model is calculated according to the misjudgment rate.To verify the validity and feasibility of the proposed track circuit fault classification model,four comparative test schemes are designed in this thesis.The accuracy rate,accuracy rate,recall rate and F1-Score were taken as evaluation indexes to evaluate the proposed track circuit fault classification model,which were respectively 97.38%,97.36%,97.45% and 97.41%.The simulation results show that the fault classification system proposed in this thesis has a good effect on track circuit fault classification,and also has certain advantages compared with other methods.