Annika Junker successfully completed her doctorate on the topic of “Data-driven modeling of nonlinear mechatronic systems in a form applicable for control engineering”. The first supervisor and DART follow-up group leader was Dr Julia Timmermann, the chair of the doctoral committee was Prof. Dr Ansgar Trächtler.
Abstract:
Model-based control design requires accurate insight into the dynamic behavior of the underlying physical system. Machine learning methods have the potential to reduce modeling efforts compared to the classic approach by effectively combining physical prior knowledge and models trained on measurement data. This dissertation develops methods to determine data-driven models for the control design of nonlinear mechatronic systems. For this purpose, the control applicability of Koopman operator-based methods, which approximate nonlinear dynamics by linear models, is analyzed. In addition, a novel method is developed for the data-driven determination of port-Hamiltonian models, which plausibly represent energy flows and can be directly utilized for passivity-based control design. Moreover, approaches for automatically updating the plant model used in the control loop are presented in case of model uncertainties or occurring changes in system dynamics during operation. Experimental and simulative studies demonstrate the outstanding prediction accuracy of the data-driven models and the high control performance. The findings of this dissertation make a significant contribution because the data-driven models exhibit a form that is highly usable for control engineering. They are physically interpretable and can be seamlessly integrated into existing analysis and design methods. This opens new perspectives for future applications and further developments.
