Indentify some aerodynamic parameters of a airplane using the spiking neural network

Abstract

The main objective of this study is to propose a method for identifying aerodynamic coefficient derivatives of aircraft attitude channel using spiking neural network (SNN) and Gauss-Newton algorithm based on data obtained from actual flights. Out of these, the SNN multi-layer network was trained by Normalized Spiking Error Back Propagation, in which, in the forward propagation period, the time of output spikes is calculating by solving quadratic equations instead of detection by traditional methods. The phase of propagation of errors backward uses the step-by-step calculation instead of the conventional gradient calculation method. SNN in combination with Gauss-Newton iterative calculation algorithm proposed in this study enables the identification of aerodynamic coefficient derivatives in a nonlinear model for aerodynamic parameters with higher accuracy and faster calculation time. The identification results are compared with the results when using the Radial Basis Function (RBF) network to prove the algorithm efficiency.