Fault Diagnosis Of Impedance Match Bond In High Speed Railway Based On Optimized Support Vector Machine

Railway transportation plays an important role in the transportation system,and its safety and reliability are highly crucial.The track circuit is a significant part of the railway signal system.As a combination of strong and weak electricity in the track circuit,the Impedance bond can not only provide the return channel of the traction current,but also restrain the unbalanced current interference caused by the traction power supply system.However,on the one hand,impedance match bond in the high-speed railway has complex structure,harsh working environment and hidden fault location.On the other hand,the high-speed railway line is isolated with external space and only has a short maintenance duration or "repair window",once the fault occurs,it may cost a long time to find and maintain,because it is difficult to locate the fault point quickly,so operation efficiency and safety will be influenced.Clearly,the research on the fault diagnosis method of impedance match bond has engineering application value.This thesis focuses on the fault diagnosis of BES(K)impedance match bond applied in ZPW-2000 A track circuit in high-speed rail.The main work of this thesis is as follows:Firstly,after structure,working principle and application scenarios of BES(K)impedance match bond are introduced,the similarities and differences between high speed and existing railway impedance match bond are compared,and the main methods of transformer modeling are discussed while the circuit simulation model of BES(K)impedance match bond is constructed by using Multisim.Then,the fault types and their mechanism are analyzed,and the circuit simulation model is used to simulate the consequence of each fault on the circuit characteristics.All faults are classified by the "frequency-consequence" risk matrix.Secondly,referring to the combination of fault categories classified by risk level,the thesis focuses on the use of intelligent algorithm for fault diagnosis.According to the overall structure of ZPW-2000 A track circuit,the fault simulation and test circuit of impedance match bond is carried on,and the fault data under a variety of operation scenarios are collected,including train operation scene and maintenance scene;after the collected data are classified and stored,the expert discrete method is selected,and the packaging method is used for feature selection to get fault sample data.After comparing the principle of three main algorithms and the results of model diagnosis,SVM(support vector machine)shows the best classification performance among the three algorithms to be suitable for small sample data processing.Further,differential evolution algorithm is used to optimize SVM.Based on the fault maintenance characteristics of different operation scenarios,the fault data of the same risk level are trained to get different optimal SVM models.Then the models with the same risk level but different operation scenarios are aggregated to form a all faults faults diagnosis combination model and a high risk faults diagnosis combination model.The typical measured data are used to verify the two models,and the fault diagnosis results show that the accuracy of high-risk fault diagnosis reaches 100%,which meets the requirements of fault diagnosis design indicators.Finally,the fault diagnosis software is developed with C# programming language,and the fault diagnosis function,fault data management function and fault diagnosis system update function are realized,and the fault diagnosis system architecture and communication mode are designed.