Model based observer synthesis for the longitudinal dynamics estimation of a wheelset

Abstract

The presented work illustrates a model based observer synthesis for the longitudinal dynamics estimation of a wheelset. This approach eases the trade-off between safety, comfort and wear that most research activities dealing with railway technology have to face. The developed observers are tested and validated at a test rig. The observer design process covers the three main steps: implementation of the observer dynamics, definition of the measurement signals, and specification of the observer cor-rection term via a multi-case optimization. The nonlinear analytical observer model is implemented using the Lagrange formalism and it takes the major nonlinearity of the wheel-rail interface into account. Furthermore, the rotational velocities of the wheelset and the roller are assumed to be outputs. The pneumatic pressure of the brake units and the torque of the roller drive are defined as inputs. Regarding the longitudinal railway dynamics the variation of the friction conditions in the wheel-rail interface and in the interface between brake pads and brake disc highly influence safety, comfort, and wear. Therefore, a parameter estimator and a disturbance observer are implemented, since these methods allow for a specific consideration of the varying friction conditions. The observer parameters are optimized using a multi-case optimization algorithm with measurement data sets. The received results proof that the observers accurately estimate the system states and provide reliable information regarding the longi-tudinal dynamics. Exploiting this information facilitates some promising applications to improve safety, comfort and wear all at once, e.g. advanced algorithms for the wheel slide and skid protection systems. Tasks to be tackled in the future are to adapt the observers to an entire train system and to validate them with data from track tests.