Parameter estimation of a railway vehicle running bogie using extended Kalman filter

This paper is concerned with fault detection problem in running bogie of high-speed railway vehicles. The running bogie of a railway vehicle is in general a highly-nonlinear, strongly coupled and time-varying dynamic system. Failures in some key components such as yaw damper can cause serious safety problems. In order to improve efficiency and accuracy of the in-service monitoring and fault detection, joint parameter and state estimation is desired. Even though the dynamic model could be represented by linear equations under certain conditions, the joint estimation yields a nonlinear filtering problem as the parameters are augmented to the state vector. In this preliminary study, EKF is adopted thanks its advantage in computational effort and the potential to develop an on-line algorithm. We investigated normal as well as failure modes of a vehicle under real track irregularity. Simulative results verify the feasibility of EKF even under multiple failures in the running bogie.