Data-Driven Modeling of Unsteady Aerodynamics for Fluid-Structure Interaction

Compliant structures are prominent in aerodynamic design and require quick and accurate unsteady aerodynamic models to predict their behavior. The state of the art consists of coupled fluid-structural simulation methods that are too numerically expensive for rapid structural stability assessment. Piston theory presents a simple, inexpensive aerodynamic model to predict the pressure fluctuations arising on a deforming surface, but is accurate only for high-speed flow regimes. This work investigates the use of dynamic mode decomposition (DMD) to develop unsteady aerodynamic models that remain numerically inexpensive yet offer accurate predictions in subsonic and low-supersonic flow regimes. DMD is used to develop a low-rank approximation to the difference between piston theory predicted pressure fluctuations and full-model predictions. Elements of aeroacoustic theory are used to inform the model. A canonical panel flutter scenario is considered as the motivating example. An improved unsteady aerodynamic model of the flow over a deforming panel is developed from a subset of the learned modes from DMD. The ability of the improved model to predict the stability of the panel configuration is presented, and the framework to develop models for alternate flow configurations is discussed.