Online Aerodynamic Parameter Estimation And Adaptive Disturbance Rejection Control For Hypersonic Vehicle

Hypersonic vehicle has great strategic significance and is widely concerned and studied by many countries at present.Its flight speed is high,the flight speed domain and the flight airspace are wide,the flight ability is strong.The complex reentry flight environment makes the hypersonic vehicle system show strong nonlinear,strong coupling,strong uncertainty and other characteristics,among which the uncertainty of aerodynamic characteristics is particularly significant,which puts forward higher requirements for the stabili ty of the control system.Aiming at the unpowered hypersonic vehicle with significant aerodynamic characteristics uncertainty,this paper studies to reduce its aerodynamic characteristics uncertainty and achieve strong robustness of reentry attitude contro l.Specific research contents are as follows.Firstly,the kinematics and dynamics of hypersonic vehicle are modeled,and the common coordinate systems and their transformation relations during reentry process are introduced.The center of mass and attitude kinematics and dynamics equations of hypersonic vehicle are derived and established,and the attitude dynamics model is linearized by feedback to obtain the linear decoupling model,which provides the model basis for subsequent aerodynamic parameter identification and attitude control.Secondly,the intelligent online aerodynamic parameters estimation methods of hypersonic vehicle based on S-SVM and LS-SVM are proposed respectively.Real-time flight data are preprocessed to obtain learning samples.In S-SVM estimation method,S-SVM with nonlinear modeling capabilities is used for online aerodynamic modeling,where flight states are as input and aerodynamic force/moment coefficients are as the output.And based on noise level of samples,hyper-parameters of S-SVM and number of training samples are determined.Combined with the idea of numerical differentiation,the aerodynamic derivative of the hypersonic vehicle is obtained.In LS-SVM method,the state variables of different orders are selected as the sample input,and the aerodynamic force/moment coefficients are taken as the output.The online aerodynamic modeling is completed based on the LS-SVM,and the mathematical derivatives of the analytical model is obtained.Through the estiamtion of aerodynamic parameters,the aerodynamic uncertainty of the control model is reduced,which lays the foundation for subsequent attitude control.Then,for hypersonic vehicle with the characteristics of strong nonlinearity,strong coupling and strong uncertainty,a reentry attitude control method based on active disturbance rejection control(ADRC)is studied.Based on the feedback linear decoupling control model,a nominal ADRC attitude control system is designed.The tracking differentiator,the ESO and the nonlinear feedback control law of the second-order system are designed,and the internal and external disturbances of the system are estimated by ESO.Combined with the estimations of aerodynamic parameters,the aerodynamic uncertainties in the nonlinear control model a re classified into known parts of the model,so that a new affine linearized control model without aerodynamic uncertainties is established.Then,the ADRC based on the estiamtions of aerodynamic parameters is designed to reduce the dependence on the observation ability of the ESO.Lastly,aiming at the characteristics of strong nonlinearity,strong coupling and strong uncertainty,a hypersonic vehicle attitude control method based on model reference adaptive is studied.A NN-MRAC controller is designed for the linear decoupling system with second-order feedback.The internal and external disturbances are reconstructed by neural network,and the model parameters of neural network are updated by adaptive laws.The stability of the system is proved by Lyapunov theory.Furthermore,based on the affine linearization control model without aerodynamic uncertainties,a NN-MRAC based on parameter identification is designed to reduce the dependence on the observation ability of the neural network observer and reduce the conservatism of the system.