Model-based And Data-driven Traction Motor Fault Diagnosis Method For Urban Rail Trains

As China’s urbanization process continues to advance,urban rail trains,as an important part of the green transportation system,have injected new vitality into the travel of the majority of residents and local economic growth under their multi-level and high-quality transportation service network.As the "heart" of the urban rail train,the traction motor is crucial to ensure its stable and reliable performance.Nevertheless,due to the extremely harsh working environment and complex structure,the common faults of traction motors are more frequent,so the fault diagnosis of traction motors of urban rail trains is of special practical significance.In this thesis,a series of theoretical and practical application studies are conducted on the traction motor fault diagnosis of metro rail transit trains from Model-based and Data-driven perspectives.(1)Research on a dynamic modeling method of traction motor based on SVPWM vector control.To address the problem of traction motor vector control,the space vector pulse width modulation(SVPWM)control method is used to establish a dynamic model of traction motor,taking into account the existing "DC-AC" power supply system of urban rail trains.Firstly,the mathematical model of traction motor in a three-phase stationary coordinate system is simplified,and the reduced-order,decoupled mathematical model in the two-phase coordinate system is derived.Then,the inverter and traction motor are considered comprehensively,and the SVPWM vector control algorithm study is implemented based on the sector determination,base vector action time,and state switching time point.Finally,a dynamic simulation model of the traction motor is built in the MATLAB/Simulink environment according to the implementation logic of the control algorithm,and the simulation experiments show the reliability of the dynamic modeling method.(2)Research on a traction motor fault diagnosis method based on image representation.To deeply analyze the dynamic process of the traction motor,a traction motor fault diagnosis method based on image representation is proposed for the closed-loop characteristics and parameter uncertainty of the motor.Firstly,a traction motor statespace model is established by introducing the system uncertainty term.Secondly,an observer-based controller is proposed to ensure the stability of the motor system,and the image representation of the original and nominal systems is constructed on this basis.Finally,a dynamic simulation model of the traction motor is built in the MATLAB/Simulink environment according to the implementation logic of the control algorithm,and the simulation experiments show the reliability of the dynamic method.(3)Research on a traction motor fault diagnosis method based on key performance indicators.Considering that accurate models are difficult to build in actual systems,it is difficult to establish an accurate model in the actual system,while the massive and rich motor data can be directly obtained through sensors,and then from the data-driven perspective,a KPI-based Canonical Correlation Analysis(CCA)is proposed for the process monitoring problem related to the Key Performance Indicator(KPI)of traction motors.Canonical Correlation Analysis for traction motor fault diagnosis.Firstly,a residual-based traction motor fault diagnosis method is proposed by analyzing the residual signal and CCA characteristics and performing Singular Value Decomposition(SVD)on the correlation evaluation matrix.Secondly,the KPI-related information is considered in the projection subspace,and a KPI-CCA model is established based on the correlation relationship between variables.On this basis,the feature space is decomposed into KPI-related part and KPI-independent part,and the corresponding monitoring statistics and thresholds are designed to obtain more comprehensive fault diagnosis information.Finally,the feasibility of the method is verified by relying on the actual operation data of traction motors of urban rail trains.