The current paper presents the estimation of aerodynamic stability and control derivatives using a combinational approach of an extreme learning machine (ELM) network and Gauss-Newton based optimization method. The classical methods based on maximum likelihood estimation process have been widely used for estimating parameters on the principle of residual error minimization principle. But, an alternative search with the non-classical methods based on feedforward neural-network (FFN'N) has also been investigated in the estimation process whose performance is fully dependent on the generalization of the network. The generalization of the network primarily depends on the training algorithm which may produce a poor relationship due to trap of the error function in local minima. Such issues can be sorted out using ELM based network for non-linear mapping of the given input-output dataset. For the estimation of the parameters, the states of a predefined aerodynamic model are propagated through ELM network in the optimization process of Gauss-Newton method. In our research paper, three flight data have been used to estimate their respective aerodynamic parameters and their values are compared with the values of the standard estimation methods. Finally, the accuracy of the estimates is presented in terms of their corresponding standard deviations.
展开▼