Aerodynamic modelling of quasi steady stall using neural-network based Gauss Newton method

The maximum likelihood estimators are commonly used in the estimation of aerodynamic model coefficients but, their performance depends on the initial guess values and the noise level of the data. Some of its variants are iterative in nature hence, computational cost is higher. An alternative investigation is performed with the artificial neural network (ANN) where an accurate non-linear mapping is produced with the input-output dataset. And further, the trained network is utilized in the Gauss-Newton based optimization process to estimate the optimal paramters. Our objective of the research is to produce an optimal neural network for non-linear mapping of the quasi steady stall data of aircraft and further, used it to estimate the aerodynamic coefficients. Finally, the validation of the estimates is presented with their Cramer-Rao bounds.