High-fidelity reduced-order aerodynamic models : application to active control of engine inlets

An unsteady Euler flow solver is developed for subsonic and supersonic duct flows. A linearized solver is derived from the nonlinear set of equations. Results show the accuracy of both formulations. Arnoldi-based model order reduction is reviewed. Both single and multiple frequency point Arnoldi methods are presented. An orthogonal reduced-order basis is generated by iterating system matrices on the input vector at a set of selected points. The full-order system is then projected onto the reduced space. The resulting reduced-order model obtained matches moments of the original model transfer function at each frequency point selected. A specific implementation procedure is presented for a system with singular descriptor matrix, which arises from boundary conditions. A state space system is extracted from the algebraic system and boundary conditions are accounted for in the output via a feedthrough term. Results are presented for linearized analysis of a started supersonic diffuser. Both single point and multiple point Arnoldi methods are applied. The size of the reducedorder models is decreased by two orders of magnitude over the computational fluid dynamics model, while preserving accuracy of the relevant dynamics. Moreover, multiple point Arnoldi outperforms the single point method, and allows a further reduction in the number of states while offering a balance between model size and computational cost. The reduced-order models are used for active shock position control via bleed. The derived controller is shown to be effective for the full-order model, and the dynamics of interest are accurately captured. Thesis Supervisor: Karen E. Willcox Title: Assistant Professor