Unsteady aerodynamic modeling based on POD-observer method

Abstract

A new hybrid approach to constructing reduced-order models (ROM) of unsteady aerodynamics applicable to aeroelastic analysis is presented by using proper orthogonal decomposition (POD) in combination with observer techniques. Fluid modes are generated through POD by sampling observations of solutions derived from the full-order model. The response in the POD training is projected onto the fluid modes to determine the time history of the modal amplitudes. The resulting data are used to extract the Markov parameters of the low-dimensional model for modal amplitudes using a related deadbeat observer. The state-space realization is synthesized from the system’s Markov parameters that are processed with the eigensystem realization algorithm. The POD-observer method is applied to a two-dimensional airfoil system in subsonic flow field. The results predicted by the ROM are in general agreement with those from the full-order system. The ROM obtained by the hybrid approach captures the essence of a fluid system and produces vast reduction in both degrees of freedom and computational time relative to the full-order model.