Multi-Output Auto-Regressive Modeling of Transonic Unsteady Aerodynamics

This paper investigates a time-domain reduced-order modeling (ROM) approach for unsteady aerodynamics and its application to flutter prediction. A Multi-Output Auto-Regressive model with exogenous inputs (MOARX) is used to identify a linearized unsteady aerodynamic ROM from CFD data. The method's accuracy is first verified on a thin-airfoil benchmark. It is then applied to the Benchmark Supercritical Wing airfoil using URANS data at Mach 0.3 and 0.8 across various angles of attack. In subsonic attached flow, the MOARX model accurately predicts aerodynamic frequency-response functions and flutter onset, closely matching time-accurate aeroelastic CFD results. In transonic flow, despite increased sensitivity to angle of attack due to shock motion and separation-induced nonlinearities, the ROM captures key qualitative trends, including the transition from pitch-plunge coalescence flutter to pitch-dominated stall flutter and the associated dynamic pressure variation with angle of attack, consistent with experimental observations. The study establishes best practices for data-driven MOARX ROM construction, defines application limits in transonic separated-flow regimes, and demonstrates the potential of MOARX-based ROMs for efficient flutter behavior mapping. However, fully nonlinear CFD remains necessary at high angles of attack with significant flow separation.