Nonlinear Unsteady Aerodynamic Reduced-Order Modeling Using a Surrogate-Based Recurrent Framework

Abstract

The generation of comprehensive data sets during early design cycles of military aircraft is a major challenge due to highly complex flow phenomena which occur throughout the aircraft flight envelope. This calls for incorporating highly accurate methods early in the design process, while ensuring rapid turnaround times at the same time. Hence, to leverage information from an affordable and therefore limited number of high-fidelity simulations, numerical methods that are capable of predicting aircraft performance and stability and control characteristics at feasible cost play a significant role. In this paper, a nonlinear unsteady reduced-order model based on a surrogate-based recurrent framework for time-accurate forced motion predictions is applied. Two types of cost-efficient generic training maneuver simulations, a set of frequency sweeps and a modified Schroeder multi-sine signal, are used for model training data generation and assessed regarding their impact on model prediction accuracy and differences in computational cost. It is demonstrated as part of the NATO Science & Technology Organization research task group 351 that the surrogate-based recurrent framework is capable of providing accurate performance and stability estimates for a generic highly agile, multi-swept wing fighter aircraft at feasible cost.