A general framework for automated physics-based reduced-order modeling of electromechanical systems

Abstract

Physics-based models of electromechanical systems, such as finite element-based models and/or high-fidelity magnetic equivalent circuits, accurately represent underlying magnetic devices. However, these models usually introduce hundreds to thousands of state variables and are computationally intensive. Moreover, including relative motion in the physics-based dynamic modeling of electromechanical systems is not a trivial task. In this paper, relative motion is incorporated in highly accurate full-order models that are based on geometrical and material data. Automated linear and nonlinear order-reduction techniques are introduced to mathematically extract the essential system dynamics in the desired bandwidth, thus preserving both accuracy and computational efficiency. The resulting reduced-order systems are verified using finite element-based models and magnetic equivalent circuits in both time and frequency domains.