Identification of Wheel-Rail Contact Condition Using Multi- Kalman Filtering Approach

Condition changes at the rail surface due to the fallen tree leaves and/or other contaminations can cause the low adhesion levels which present a serious challenge for the traction/braking control systems to avoid the problem of wheel slip/slide. This paper presents a multiple model based method for the identification of the adhesion limit to overcome the problem of the wheel slip/slide in poor contact conditions. The proposed scheme is an indirect method that exploits the dynamic properties of the conventional solid axle wheelset in response to changes in contact condition at the wheel-rail interface avoiding difficult and expensive measurement requirements. A nonlinear model of lateral and yaw dynamics of a conventional solid axle wheelset is used for the study. The non-linearity and changes in the interaction with the rail are modelled by using a set of non-linear creep/slip curves. The scheme consists of a bank of Kalman filters based on the linearized wheelset models. Each Kalman filter in the filter bank is optimally tuned to operate in a specific contact condition. Normalized root mean square values from the residual of each filter calculated using time moving windows are assessed to identify the operating condition of the wheelset