Research On Fault Detection And Separation Of Urban Rail Vehicle Suspension System Based On Track-side Vibration Signals

The traditional fault diagnosis method for the suspension system of urban rail vehicles is based on the signals measured from car body and bogies.Vibration acceleration sensors need to be arranged on the four top corners of the vehicle body and the four top corners of the front and rear bogies to collect the corresponding vehicle body under different working conditions of the suspension system.In this condition,each vehicle needs to be equipped with a certain number of sensors in order to collect the signals of different fault conditions.This method will also increase the cost and complexity of the fault diagnosis.Therefore,in order to reduce the cost and complexity of fault detection,this paper proposes a method based on the track-side detection for detecting faults in the suspension system of urban rail vehicles.By arranging sensors on the track-side,the vibration response signals of the rails under different conditions of the suspension system are collected.In order to make the tack-side detection of urban rail vehicle suspension system fault diagnosis implemented,the research work in this paper can be divided into the main parameters of suspension system monitoring,fault condition detection and fault separation.The main research contents are as follows:(1)The vehicle and rail rigid-flexible coupling modeling and track-side vibration signal collection method is studied.According the failure mode of the vehicle suspension system,a rigid-flexible coupling model of the vehicle track is established based on SIMPACK-ABAQUS,and the track-side sensor layout rule is proposed.Based on the rigid-flexible coupling model,a fault simulation experiment platform is established,and the acceleration signal of the track-side vibration is collected by the sensor.This is the basis for fault diagnosis of vehicle suspension systems.(2)The monitoring method of main parameters of urban rail vehicle suspension system is studied.Based on the established wheel-rail rigid-flexible coupling simulation model,the dynamic response of the suspension system and the related mechanical parameters caused by wheel-rail excitation on the wheel-set,bogie,and vehicle body were monitored to verify the regularity of vibration signal transmission.For the monitoring of the secondary suspension parameters,combined with the vehicle dynamics model,the calculation results of the main parameters are obtained by using the frequency domain algorithm and the least squares algorithm.For the monitoring of the primary suspension parameters,the parameter detection is performed using acceleration and angular acceleration of the bogie.The cross-correlation relationship between the front and rear suspensions on the bogie and the consistency of wave crests are compared,then the rationality of the indicators is verified by a rigid-flexible coupling fault simulation model.(3)The fault detection method of suspension system for urban rail vehicles was studied.For the track-side vibration acceleration signals collected under different fault conditions of the suspension system,the characteristics are extracted using time-domain,frequency-domain and spectral refinement analysis methods,and the characteristics of the track-side signal characteristics between different fault conditions are analyzed.Based on the methods of Fourier transform and chirp Z-transform,fault detection between different fault conditions is realized.(4)The method of fault separation based on signal feature extraction is studied.Based on the signal decomposition methods of local mean decomposition,empirical modal decomposition,and variational modal decomposition,the signal components corresponding to the vibration acceleration from the track-side under different fault conditions of the suspension system are obtained,and the optimal components are calculated by using correlation coefficients.Then using the fast spectral kurtosis method extract the frequency band interval the sensitive components located.Based on the feature extraction method,the optimal signal component features are extracted as a sample feature library for fault separation.Finally,the least-squares support vector machine and fuzzy C-means clustering are used for fault separation,and the accuracy of fault separation under different fault feature extraction algorithms is obtained.