Estimation Of The Wheel–Rail Contact Parameters For A Conical Solid–Axle Railway Wheelset Using Linear Integral Filter Method

This paper addresses the specific problem of estimating the ‘conicity’ and ‘creep coefficients’ values of a conical railway wheelset, which vary significantly as the vehicle runs along straight tracks with lateral irregularities. The performances of the continuous–time (C–T) least–squares error (LSE), least–absolute error (LAE) and least–absolute error with variable forgetting factor (LAE+VFF) estimators that employ the linear integral filter (LIF) method are compared. Each estimator was designed based on the fifth order model of a single solid–axle wheelset suspended to ground via lateral springs and dampers connected in parallel. For each simulation, the LAE + VFF estimator performed the best because the algorithm combines the least–absolute error identification that has an instrumental variable element to overcome estimation bias problem and the variable forgetting factor for fast tracking and smooth steady–state estimation. The estimator was then directly applied to a 14th–order two–axle railway vehicle system where two separate fifth order wheelset model estimators were used to estimate the front and rear wheelsets parameters independently. Since the vehicle body of the two–axle vehicle was effectively decoupled from its wheelsets, the LAE + VFF estimator produced similar estimated front and rear wheelset parameters values, and hence simplifying the estimation process of the more complex 2–axle railway vehicle model.