Real-time capable nonlinear pantograph models using local model networks in state-space configuration

Modern pantograph current collectors for high-speed trains are mechatronic systems that are increasingly equipped with active control to maximize their dynamic performance. To realize a high-quality contact, decrease wear, and increase speed, it is necessary to use high-fidelity co-simulation and hardware-in-the-loop (HIL) testing tools, as well as modern model-based control concepts. In all these areas, efficient, real-time-capable, and accurate models of the pantograph dynamics are required. This paper proposes such efficient model structures and identification procedures to calibrate these models from measurement data of a full-scale pantograph. The pantograph dynamics are thereby modeled by a network of local linear models, which provides a globally nonlinear system model, local interpretability, and high computational efficiency. The proposed identification methods are designed to especially exploit expert knowledge to maximize model quality under the present system and measurement limitations (loss of contact, limited excitation bandwidth). The methods are successfully validated with respect to the white-box model as well as with measurement data of a real, full-scale pantograph, and the obtained models provide high accuracy while fulfilling real-time requirements.