Non-linear Lift Curve Modeling Using Neural Networks with Flight Test Data of a Subscale Fighter

Abstract

© 2022, American Institute of Aeronautics and Astronautics Inc. All rights reserved.System identification based on mathematical models is generally restricted to linear systems. To model nonlinear behavior, more complex mathematical models are needed and often not available. To model the nonlinear dynamics at high angle of attack of a fighter, a neural network method was applied. The system identification process used in this work used flight test data acquired from a remotely piloted Generic Future Fighter (GFF) subscale. After the application of the neural network, the non-linear effect present in the detachment of the wing boundary layer was possible to estimate. The neural network used was the Feedforward type and the optimization of the parameters was carried out with Backpropagation. Stall maneuvers were initially used to train the neural network (training cycle) and later a new stall maneuver was used to validate the identification (prediction cycle). The method demonstrated the ability to estimate the lift curve in a subscale fighter.