Research On Modeling Of Aerodynamic Coefficients For Flight Vehicles Based On Wavelet Network

Based on in-depth survey of mess references, this thesis firstly gives comprehensive overview of aerodynamic coefficient modeling and corresponding parameter estimation algorithm. Aerodynamic coefficient identification is a interdisciplinary subject between mechanics, informatics and control theory. This thesis also gives an overview of the development direction in this field, especially introducing the application of neural network in the aerodynamic coefficients identification, which gives the theoretical basis of using wavelet network in modeling of aerodynamic coefficients.Then, after introducing wavelet analysis and wavelet network basic theory the thesis begin to study the modeling criteria with optimizing algorithm, and design the structure of wavelet network. In case of high input dimension system model, taking norm of input vector as the input of wavelet network instead of using tensor product method to construct wavelet network ,which could solve the problems of high computation and curse of dimensionality. In the selecting of specific wavelet basis, this thesis first gets initial wavelet basis collection according to spectrum analysis, then gives the least squares regression algorithm to optimize wavelet basis collection based on the least estimation error criteria, which could also initialize the model parameters and increase the speed of convergence. Moreover, in order to capture the linear properties in aerodynamic coefficients, additional linear terms are added to the model, which make the wavelet network model has some definite physical significance in use.According to dynamics simulation for a missile model, flight data is obtained. The first half of the data is used for aerodynamic coefficients modeling, and the second half is used for verification of the resulting model based on pre-estimation of ballistic curve, the results indicate the wavelet network model of aerodynamic coefficients has generalization ability and applicability. In the end, some conclusions are drawn, which indicate that this method has practical significance of application.