Research On Method Of Fault Diagnosis For Suspension Controller Of Medium-Speed Maglev Train Based On SOM-BP Serial Neural Network

As a new type of maglev train with advanced technology,the electromagnetic(EMS)medium-speed maglev train combines the advantages of medium-low speed maglev train and high-speed maglev train.It not only has high traction efficiency,but also has a simple guiding structure.As the central part of suspension control system of the medium-speed maglev train,the suspension controller has many components,a complex structure and powerful functions.However,due to its simultaneous operation with the maglev train,internal parts and components of the controller are highly consumable.During the operation of the maglev train,the sudden failure of the suspension controller will directly reduce the stability of the suspension control system.In severe cases,the train may lose its suspension capacity directly.It is difficult to find out the failure source in a short time and carry out the maintenance work efficiently.Therefore,it is of great significance to study the fault diagnosis method of suspension controller of the medium-speed maglev train.In recent years,artificial neural network theory and its applied research are international frontier focused,providing a new research perspective for fault diagnosis.Based on the facts of engineering practices and the data characteristics of fault samples,this paper analyzes the operating principles and typical fault characteristics of the suspension controller of the medium-speed maglev train,and the fault diagnoses method based on the neural network model are proposed to diagnose the malfunction of suspension controller of medium-speed maglev train.The main contents and results of this research are displayed as follows:(1)The fault diagnosis model of suspension controller based on BP neural network and SOM neural network has been established and the fault diagnosis performance of the two models has been analyzed through simulation experiments.The experimental results show that the training efficiency and diagnostic accuracy of the two neural network models are not ideal and need to be further improved.(2)According to the limitations of BP neural network and SOM neural network,the competition layer results of SOM neural network is proposed as the input data of BP neural network.The fault diagnosis model of SOM-BP neural network is constructed and its fault diagnosis performance is analyzed through simulation experiments.The experimental results show that the training efficiency and diagnostic accuracy of the SOM-BP serial neural network fault diagnosis model are significantly better than the single neural network model.(3)Considering that the performance of SOM-BP tandem neural network can be easily affected by network connection weights and node thresholds,a particle swarm optimization(PSO)algorithm is proposed to assign network connection weights and node thresholds.The fault diagnosis model of SOM-BP serial neural network based on particle swarm optimization is constructed and its fault diagnosis performance is analyzed through simulation experiments.The experimental results show that the method of fault diagnosis of SOM-BP serial neural network based on particle swarm optimization is more effective in solving practical problems.