ANALYSIS ON OFFLINE AND ONLINE IDENTIFICATION METHODS FOR AIRCRAFT STABILITY AND CONTROL DERIVATIVES

摘要

Stability and control characteristics analysis has long been the important research area of aircraft flight dynamics, which is the critical factor of control system design, performance evaluation and integrated design of aircraft. The linear perturbation equations describing aircraft longitudinal and lateral motion, which are derived from nonlinear dynamic equations based on the small perturbation theory, are characterized with matrices called stability and control derivatives in flight control theory. There are two main methods to obtain these derivatives, theoretic deduction and parameter identification, where the latter is a valuable complement for the former one. Offline and online parameter identification are utilized in engineering application with different emphasis. Offline methods are commonly used to obtain linear dynamic model of aircraft under specific operation conditions, with complicated aerodynamic shape or dynamic characteristics, where the model could be used to investigate the stability and control characteristics. Online methods are commonly used in fault detection or flight adaptive control, where the derivatives are estimated with Kalman filters. Aircraft longitudinal and lateral stability and control characteristics are discussed here with online and offline identification methods. Firstly, the small perturbation dynamic equations under rudder perturbation are deduced, and the expressions of all stability and control derivatives are given. Secondly, the Unscented Kalman filter (UKF) method and maximum likelihood estimation (MLE) method are verified with aerodynamic data of a small unmanned aerial vehicle ANCE, where UKF proves to be an adequate online estimation method by the consistent results and its asymptotic approximation to the theoretic values. We also compare the effects of random noises on the estimation accuracy and modes response eigenvalues for these two methods. The results show that UKF has better noise-resistance than MLE, and that UKF prevails in longitudinal derivatives estimation and modes response analysis while maintaining equal performance in lateral direction.