Nonlinear aerodynamic model identification using empirical mode decomposition

Abstract

A conceptual method based on the empirical mode decomposition algorithm is proposed to identify a high-fidelity full-flight envelope aerodynamic model, utilising flight data. The key idea is to recognise dominant phenomena of flight containing dissimilar amplitudes and frequencies by means of the empirical mode decomposition. Being distinguished and separated from each other, flight modes can be considered in the aerodynamic model, independently. To achieve the goal, an equation error method is utilised to identify six multi-input single-output systems for aerodynamic forces and moments. The inputs of the identification systems are intrinsic mode functions of flight parameters. The nonparametric identification problem uses the Hammerstein nonlinear ARMAX structure and estimates parameters by the least squares iterative algorithm. Flight tests of an unmanned aircraft in two complex manoeuvres are employed for the modelling and simulation. Several models with different nonlinear functions are introduced and trained by the first dataset. Then, to verify the identified model, the flight simulation is performed for the second dataset. Results indicate the appropriate performance of the identification method in nonlinear aerodynamics.