Research On Fault Diagnosis Method Of Multilevel Inverter

With the increase of the speed of high-speed trains in China,the safety of highspeed trains has also attracted the attention of all walks of life.If the high-speed trains fail,the traffic will be chaotic.And the car will be destroyed and seriously harm the lives of the people.Therefore,how to accurately diagnose and isolate the power switch tube of the inverter in the high-speed rail traction system,so that the high-speed train traction system works normally,thus ensuring the safe and stable operation of the highspeed train becomes a top priority.This paper first analyzes and studies the common high-speed train traction inverters,and determines the most common diode-clamped three-level inverter in the high-speed train traction system as the research object,and analyzes the failure mechanism of the inverter.And the main performance characteristics,and detailed circuit analysis of the operating state of the three-level inverter.Secondly,this paper proposes a secondary fault diagnosis method for inverter based on nuclear fuzzy clustering and rule inference for the problem of similar fault waveforms in different fault types of high-speed train traction inverters.The variational mode decomposition algorithm(VMD)can eliminate the exponentially decaying DC offset problem and avoid the aliasing phenomenon.It is more effective in the signal processing of discrete nonlinear systems.Therefore,this paper uses VMD to extract the characteristics of the inverter fault signal.And build a fault signature vector table that accurately identifies all fault types.Then the neural network(BP)is used for learning training and the diagnosis results are output,and the weight and threshold of the neural network are optimized by the optimized genetic algorithm.Due to the shortcomings of genetic algorithm(GA),such as easy to fall into local optimal solution and slow convergence rate,the kernel fuzzy clustering(KFCM)algorithm is introduced to transform the classification problem into multi-objective optimization problem,and the genetic algorithm is easy to fall into the local optimal solution.And the shortcoming of slow convergence,the primary diagnosis method based on VMD-nuclear fuzzy clustering optimization genetic neural network is used to diagnose the primary fault of the inverter;The voltage signal of the bridge arm in the inverter is extracted,and a secondary diagnosis method combining VMD and rule reasoning is used to accurately identify the special fault of the inverter.Finally,through the experimental simulation,the diagnostic results of three optimization algorithms of KFCM-GA-BP,GA-BP and particle swarm optimization(PSO)optimized neural network are compared.The results show that the diagnostic accuracy of KFCM-GA-BP algorithm is Higher than the other two,the primary diagnostic method can quickly and effectively identify the primary fault of the inverter;The secondary diagnosis method combining VMD and rule reasoning is used to identify the special fault of the inverter.The data analysis shows that the method can accurately identify the special fault of the inverter.