The Study On Condition Monitoring Method Of Urban Rail Vehicle Based On Particle Filter

The article mainly discusses the condition monitoring method of urban rail vehicle based onparticle filter.Particle filter, Kalman filter, parameter estimation algorithm,vehicle system dynamicsare analyzed. The strategy of repeated uniform sampling is added to parameter estimation algorithm,which is applied to condition monitoring. The main contents of this article are introduced as follow:(1)The status and history of particle filter is introduced firstly, then is the application of particlefilter, emphasizing on the condition monitoring and fault diagnosis method base on it.The basicprinciple of particle filter is analyzed, including Monte Carlo method, Bayesian filter,standardparticle filter,sample importance resample particle filter. At last, the system condition monitoringbased on particle filter state estimation is analyzed. Simulation results indicate that the method canmonitor system fault, but can not fix where the fault is.(2)The parameter estimation based on particle filter and Kalman filter,as well as extendedKalman filter is analyzed. Simulation results indicate that parameter estimation based on extendedKalman filter depends on initial value.The deviation will be large if inappropriate value is chosen.Theparameter estimation based on particle filter and Kalman filter has nothing to do with initial value,canwork well. The system condition monitoring method based on particle filter parameter estimation cannot noly monitor system fault, but also parameters changes related with fault.(3)The dynamic model of urban rail vehicle is analuzed. The lateral dynamic equations andvertical dynamic equations of vehicle system are set up firstly. The lateral dynamic space model andvertical dynamic space model are gained by discreting the equations. The strategy of repeateduniform sampling is added to parameter estimation based on particle filter and Kalman filter, putforward a new condition monitioring method of vehicle system based on parameter estimation withthe improved strategy. Simulation results indicate that with the strategy of repeated uniformsampling,the disability of monitoring the change of parameters due to particle depletion can beavoided when the sudden failure of suspension system results in the change of parameters；and theparameter estimation can be done when the parameters have changed with the aging of vehiclesystem.