Research On Intelligent Identification Algorithms Based On Fixed-wing Aircraft

In recent years,the performance of aircrafts has been continuously improved,and it has been widely used in military reconnaissance,surveillance,communication and civil fields.The research work has received extensive attention from researchers at home and abroad.In actual tasks,people put forward higher requirements on the control performance of the aircrafts,so it is of great significance to study the effective aircraft modeling method.In this paper,considering the characteristics of nonlinear,strong coupling and time-varying of the F-16 fixed-wing aircraft system under unmanned combat demand,we focused on four aspects of system identification research,including excitation signal design and neural network identification.Hammerstein-Wiener model identification and block-oriented identification.(1)According to the dynamic frequency response characteristics of the F-16 and the characteristics of five commonly used excitation signals,two types of aircraft identification excitation signals are selected and designed.The F-16 Simulink model based on wind tunnel test data was compiled,and the model was reconstructed in the new version of MATLAB and verified by simulation.This model can accurately simulate the flight state of the aircraft and provide reliable experimental data for subsequent identification experiments.(2)Considering the dynamic nonlinear characteristics of F-16 system,an identification method based on recurrent neural network is proposed.Using the data generated by the simulation model,neural network identification experiments based on two different recurrent neural networks were carried out.After determining the network type,the network structure parameters for the F-16 system identification were selected through trial and error and comparison.Experiments show that the global recursive NARX neural network identification works better than the local recursive Elman network identification,and it can accurately predict the state change of the aircraft,which is an effective neural networks identification method.(3)Aiming at the shortcomings of neural network identification for the lack of internal mechanism of aircraft system and specific meaning of model parameters,a method based on Hammerstein-Wiener model for aircraft identification is established.This method fits the F-16 system with a "Nonlinear-Linear-Nonlinear" model.Compared with the neural network identification method,the order and zero-pole of the system are analyzed.However,this method has a limitation of the unrecognizable intermediate quantity,and it cannot obtain a unified model structure for all output variables,which has certain limitations.(4)In order to further refine the description of the internal system,a block identification method of “Nonlinear-Linear-Nonlinear” structure is designed.Combined with the information flow of the system,the information of aircraft quality and rotational inertia are introduced.The structural constraints are added in the identification process.The F-16 aircraft system is divided into several linear and nonlinear modules,which are respectively identified and block-learned by appropriate models.Experiments show that the proposed block-oriented identification method deepens the understanding of the internal system,reduces the number of parameters that need to be estimated in the identification,and can effectively improve the accuracy of the F-16 aircraft system identification.