21st Congress of International Council of the Aeronautical Sciences, Melbourne, Australia, 13-18 September, 1998
Paper ICAS-98-1.10.5

IDENTIFICATION OF AIRCRAFT NON-LINEAR DYNAMICS USING VOLTERRA SERIES

Marques F. D., Belo E. M.
University of Sao Paulo, Brazil

Keywords: aircraft, non-linear dynamics, volterra series

An approach for a systematic identification of aircraft non-linear dynamics by means of a Volterra functional series is presented in this paper. Volterra functional series is an intuitively satisfying representation for continuous non-linear time-invariant dynamic systems. The rigorous mathematical formulation of Volterra functional series have motivated a variety of different functional representations in order to extend the range of systems modelling, as well as to overcome the difficulties in determining the Volterra kernels. Recently, with the advances in the theory of computational neural networks, it has been developed a particular network architecture that is shown to be equivalent to a discrete Volterra series. This methodology facilitates, in principle, the kernel calculation of any order. The neural network approach, to achieve a Volterra series, is applied for the case of an aircraft non-linear longitudinal dynamics. Results have shown that the approach performs well and provides suitable approximation of the non-linear behaviour of aircraft longitudinal dynamics. The easy implementation of this kernel identification methodology also contributes for further applications in aircraft non-linear analysis and control design.

view full paper