Uncertainty Reduction in Aeroelastic Systems with Time-Domain Reduced-Order Models

Prediction of instabilities in aeroelastic systems requires coupling aerodynamic and structural solvers, of which the former dominates the computational cost. System identification is employed to build reduced-order models for the aerodynamic forces from a full Reynolds-averaged Navier-Stokes solver, which are then coupled with the structural solver to obtain the full aeroelastic solution. The resulting approximation is extremely cheap. Two time-domain reduced-order models are considered: autoregressive with exogenous inputs, and a linear-parameter-varying-autoregressive-with-exogenous-input model. Standard aeroelastic test cases of a two-degree-of-freedom airfoil and Goland wing are studied, employing the reduced-order models. After evaluating the accuracy of the reduced-order models, they are used to quantify uncertainty in the stability characteristics of the system due to uncertainty in the structure. This is observed to be very large for moderate structural uncertainty. Finally, the uncertainty is reduced using simulated aeroelastic response data from flight tests to update structural parameters using the Bayes' theorem.