Lateral guidance of independently rotating wheel pairs using feedback linearization

Abstract

A main aspiration of the DLR project Next Generation train (NGT) is to include mechatronic systems in railway vehicles, because this opens a broad field to improve the running dynamics. Mechatronic systems enable the use of independently rotating wheels (IRW), which can crucially reduce wear due to the different wheel speeds, but afford an active guidance control to stabilize the lateral movement. However, the control design is challenging due to the complex non-linear system, mainly caused by the non-linear wheel-rail profile. Beyond that the guidance has to be robust to cover various scenarios regarding e.g. different velocities, track irregularities, curving and wear profiles. In this survey, a control design via feed-back linearization is presented, offering the possibility to explicitly address non-linearities in design. In this way, elaborate gain scheduling between linear designs in various operating points can be avoided. To enable the control synthesis an analytical model is established, analyzed and leads to a control structure, based on the feedback of different wheel speeds. Therefore, the advantage of IRW, allowing for low creepage and wear in every lateral position of the wheel pair is beneficially utilized. The control is implemented in Modelica, because it allows for automatically inverting models and using them conveniently in the control layout. In addition the gains of the established controller are adjusted in terms of stability and performance using multibody models of the DLR Railway Dynamics Library. The control is then transferred to a Simpack full scale multibody model via FMI and the performance is increased by control parameter optimization. Preliminary simulation results for a coach running on tracks with irregularities depict that a sufficient performance of the controller and a significant wear reduction compared to a conventional high speed train is attained. Beyond that, the usage of advanced inverse models in the control layout facilitates further refinement and the controller is to be tested on the 1:5 scaled roller rig.