The Development of Neural Network Techniques for the System Identification of Aircraft Dynamics

INTRODUCTION A great deal of time and effort is spent by aircraft manufacturers in the identification and estimation of the parameters that properly describe the aerodynamics of a particular aircraft. The derivatives of the aircraft forces, with respect to the input state have been the basis of many design and development techniques and tools, such as dynamic stability analysis, performance, simulation and control. The system identification of the aerodynamic model using neural networks involves the selection of the necessary input and output states to the system, representation of the model, and parameter estimation of the characteristic quantities, and in this case the aerodynamic force and moment derivatives. The application of neural networks as both a model descriptor and parameter estimation technique also requires a close selection of model inputs, to most efficiently utilise the method. System identification of the entire dynamic system as opposed to the parameter est