Research On Fault Diagnosis Of Traction Inverter Based On Compressed Sensing Theory

In the process of rapid urbanization,rail transit system occupies a important position,the safe and efficient operation of urban rail transit vehicle can provide strong guarantee for the city's social production,as an important part of locomotive traction inverter,it will requently switch between on and off states while the vehicle is running,the state is relatively complex,have higher failure rate,and the traction inverter fault,in addition to interfere with the normal social production will bring security hidden danger to the crew,most of the existing traction inverter fault diagnosis method based on the Nyquist sampling theorem,Massive signal acquisition is needed to ensure the accuracy of diagnosis,which leads to large amount of data to be processed and difficulties in data storage and transmission.Therefore,based on large data on traction inverter fault,rapid and accurate diagnostic method helps to improve the efficiency of the fault to solve urban social production and the security,based on the project of National Key R&D Program of China "without catenary power city rail vehicle key technology and equipment development" as the research background,a new fault diagnosis method for traction inverter based on compressed sensing theory is developed.Firstly,a simulation model based on MATLAB/Simulink was built based on the topology and working principle of the traction inverter.After knowing the causes and types of faults,all single and double tube faults were simulated based on the simulation model,and voltage signals under 22 working conditions were collected as the data basis for subsequent research.Secondly,the sparse representation performance of CS theory is studied,and a method to optimize the sparse representation performance based on double sparse dictionary model is proposed.Aiming at the fault voltage signal of traction inverter,this method can improve the compression measurement performance of CS theory,and the feasibility of the optimization method is verified by simulation experiments.Thirdly,a comparative analysis and research is performed on the measurement matrix of CS theory.The measurement matrix optimization method based on the Gram matrix is used to optimize the Gauss matrix.The correlation between the optimized measurement matrix and the sparse basis is lower,and the reconstruction error is more smaller than the traditional method,the feasibility of the optimization method is verified based on simulation experiments.Finally,SVM is introduced into the fault mode classification of the traction inverter,and a fault diagnosis method based on the CS-SVM model is proposed.The CS theory is used to process the fault voltage signal of the traction inverter.In the process of reconstructing the measurement values obtained by the compression measurement,the fault feature parametersare directly extracted,and they are used as the input of the support vector machine to finally complete the fault classification.Simulation results show that compared with the traditional method,this method has improved fault diagnosis accuracy and diagnosis speed,and the voltage signal data length that needs to be processed during the diagnosis process is greatly reduced,which can reduce the pressure on the data transmission storage device.