Fault Diagnosis Of The Stator Winding Of The Linear Synchronous Motor Of The Medium-speed Maglev Train

As a new type of transportation,the EMS medium-speed maglev train with a speed of200 km/ h has been extensively studied because of its advantages such as low noise,strong climbing ability,and low maintenance workload.The medium-speed maglev train is based on the original EMS-type low-speed maglev train keeping the suspension system basically unchanged,and the traction system is improved by adopting a long stator synchronous traction method.Compared with the ordinary wheel-rail track,the track has increased electromagnetic characteristics and requires electrical monitoring of parameters.The working environment of the traction system is harsh,and it is prone to failures that directly threaten the safe operation of trains,such as damage to the long stator winding cables,but the current field of fault research on medium-speed maglev traction motors is still at a blank stage.Therefore,this article takes the medium-speed maglev traction motor as the object to study the fault diagnosis method of the long stator winding,including the simulation analysis of the long stator winding fault,using CEEMDAN to extract the characteristics of the monitoring signal in the time and frequency domain,and Relief F to select the features of parameters,PSO-SOM-BP and PSO-SOM-SVM for fault diagnosis,etc.The main research contents and results are as follows:(1)Structural characteristics of medium-speed maglev traction motor and analysis of electromagnetic characteristics under normal operating conditions.Using Maxwell finite element simulation software to study the medium-speed maglev traction motor,the air gap flux density distribution of the motor,the influence of the power angle,stator current,and air gap length changes on the electromagnetic traction force and normal force are analyzed.The results also show that the traction system and the suspension system can realize mechanical decoupling under working conditions.(2)Simulation analysis of the stator winding fault of the medium-speed maglev traction motor.The fault types and causes of the long stator winding of the traction motor are summarized.The field-circuit coupling method is used to establish the simulation model of the single-phase short-circuit,two-phase short-circuit,three-phase short-circuit,and single-phase one-pole open circuit faults of the medium-speed maglev traction motor stator winding for the first time.The influence of each working condition on the related electrical quantities and electromagnetic force is analyzed.(3)Feature extraction of monitoring signal of traction motor condition monitoring signal.The wavelet soft threshold denoising is used to denoise the electromagnetic force signal,and the notch filtering method is used to defund the three-phase back EMF signal.The advantages of CEEMDAN method in signal feature extraction are studied.The time domain features of electromagnetic force,back EMF signal and Hilbert marginal spectrum feature after preprocessing are extracted by CEEMDAN to form the state feature vector of the stator winding of the traction motor.It is proposed to use the Relief FCorrelation method to extract the characteristic parameters of the traction motor state parameters that have sufficient correlation with the fault and remove the redundant features,and determine the set of characteristic parameter of traction motor fault.(4)Establish a fault diagnosis model for the traction motor.The methods of PSO-SOMBP and PSO-SOM-SVM for fault identification of the stator windings of medium-speed maglev traction motors are studied.The method of constructing characteristic parameters based on CEEMDAN and Relief F is used to construct characteristic vectors for the five types of working conditions of stator windings: normal operating conditions,single-phase short-circuit,two-phase short-circuit,three-phase short-circuit,and single-phase one-pole open circuit.The prediction accuracy rates of PSO-SOM-BP model and PSO-SOM-SVM model are 98.18% and 97.27% respectively.The experimental results show that the two diagnostic models proposed in this paper have achieved relatively ideal results in terms of prediction accuracy.In comparison,the prediction accuracy of the PSO-SOM-BP model is slightly higher,while the training efficiency and stability of the PSO-SOM-SVM model are higher.