Investigation To Estimation Of Wheel/Rail Contact Forces And Suspension Parameters For Railway Vehicles Based On Bayesian Processors

In the past decades, China experienced continuously increasing demands in railway transportation both in passenger numbers and freight-hauled tonnages. With increasing capabilities of the railway system, railway vehicles work in harsh conditions.This needs safer, stabler, more comfortable and more reliable railway vehicles. For increasing the safety, stability and reliablity of railway operation, estimations of the wheel-rail lateral contact forces and suspension parameters should be key technologies. They could help monitor the operation conditions of railway vehicles, and then avoid railway vehicles working in dangerous conditions, and also could provide indications for condition-based maintenance and active control of bogie.Recent researches are mainly for estimations of lateral suspension parameters rather than estimations of the vertical suspension parameters. This paper focuses on estimation of the vertical suspension parameters using several vertical dynamic modelsbased Bayesian Processors(BP), and estimation of wheel-rail lateral contact forces using discrete-time classical Bayesian processors. These BPs are validated by using the test-validated MBS model of the HXN5 locomotive. For this purpose, works of this paper are as follows:(1) According to a real HXN5 locomotive, a nonlinear MBS model is established in Simpack environment. For making the MBS model close to the real locomotive, a ring-line dynamic test of the HXN5 locomotive is carried out and then used to validate and improve the established MBS model. By validation, the simulation results of the MBS can be used as measurements and inputs of Bayesian processors with con?dence.(2) Four coupled lateral dynamic models of the HXN5 locomotive with different DOFs are derived for establishing discrete-time Classical Bayesian Processor(CBP).A multi-?lter strategy using single-wheelset CBP and two dual-?lter strategies using bogie-CBP(or half-vehicle-CBP) are presented for estimation of lateral contact forces and yaw moments of all six wheelsets. With measurements from the MBS simulation,results of these four ?lters are provided and discussed.(3) Two kinds of extended Classical Bayesian Processors(exCBP) and two kinds of Unscented Bayesian Processors(UBP) are provided for estimating vertical suspension parameters of railway vehicles, according to how the random track irregularities enter the railway vehicle system. The performances of these BPs on secondary suspension parameters estimation are discussed. In the case that the random track velocities are considered as process noises, the in?uences of the statistics properties of random track on parameter estimation are also studied.(4) Because the estimates by exCBP and UBP are sensitive to the intial conditions,a novel Mixture Bayesian Processor is introduced based on UBP, Particle Bayesian Processor and marginalization technique. The performances of the novel UBP-MBP on parameter estimation are studied by an example. The results of the example show that the UBP-MBP overcomes the disadvantages of the UBP and are more accurate than CBP-MBP. The novel MBP is then used to estimate parameters of the secondary vertical suspension for the four-wheelset vehicle. Finally, a HXN5 vertical dynamic model-based UBP-MBP is presented and validated using numerical results of the nonlinear HXN5 MBS model.