Data-Driven Unsteady Aerodynamic Modeling for Fluid–Structure Interaction

Abstract

The piston theory class of simplified aerodynamic models that link body deformation to localized pressure fluctuation with respect to a mean steady state is attractive for modeling purposes as they constitute a computationally efficient model to calculate the unsteady pressure response on deforming bodies. However, as piston theories are valid for flow regimes where M>1.5, they are inaccurate for modeling the aerodynamic response associated with aeroelastic phenomena in subsonic or low-supersonic flow regimes. Unsteady computational fluid dynamics (CFD) simulations are used to learn the unsteady pressure response, and dynamic mode decomposition (DMD) is applied to the error between the CFD-computed pressure fluctuation and the piston-theory-predicted pressure fluctuation. This process is performed on two-dimensional and three-dimensional flow domains over various flow regimes. For both scenarios, the DMD analysis learns dominant spatial modes in the error that constitute low-order models, which may be used to improve the pressure response and are confirmed to be physically relevant. A method for computing aeroelastic stability with these leading modes is discussed and applied to fluid–structural beam and panel configurations. The results are compared against prior numerical and experimental investigations to quantify the improvement in prediction obtained by the DMD-augmented piston theory.