Nonlinear aircraft system identification using artificial neural networks enhanced by empirical mode decomposition

This paper aims to improve the performance of artificial neural networks used for the aircraft system identification by taking flight dynamic characteristics into consideration. In the proposed method, flight dynamic modes are recognized, isolated, and inputted individually to feed-forward neural networks. This method has several advantages such as being adaptive, involving all observable modes in the identification process, considering interactions between longitudinal and lateral-directional modes, and reducing noise effects. Simulated and real flight data of the HARV aircraft at high-angle of attack maneuvers are employed to train the neural networks and evaluate them. Results demonstrate improved accuracy and generality of the proposed method in comparison with the conventional ones.