Sensor Based System Identification in Real Time for Noise Covariance Deficient Models

Defence science journal

article

Prashant Kumar¹  Sarvesh Sonkar¹  Ajoy Kanti Ghosh¹  Deepu Philip¹ 

[1] Indian Institute of Technology

关键词: Parameter estimation;    System identification;    Extended Forgetting Factor Recursive Least Square (EFRLS);    Frequency Transform Regression (FTR);    Aerodynamic derivatives;   

DOI  :  10.14429/dsj.72.17663

学科分类：社会科学、人文和艺术（综合）

来源: Defence Scientific Information & Documentation Centre

【 摘 要 】

System identification methods have extensive application in the aerospace industry’s experimental stability and control studies. Accurate aerodynamic modeling and system identification are necessary because they enable performance evaluation, flight simulation, control system design, fault detection, and model aircraft’s complex non-linear behavior. Various estimation methods yield different levels of accuracies with varying complexity and computational time requirements. The primary motivation of such studies is the accurate quantification of process noise. This research evaluates the performance of two recursive parameter estimation methods, viz.; First is the Fourier Transform Regression (FTR). The second approach describes the Extended version of Recursive Least Square (EFRLS), where E.F. refers to the Extended Forgetting factor. Also, the computational viability of these methods was analyzed for real-time application in aerodynamic parameter estimation for both linear and non-linear systems. While the first method utilizes the frequency domain to evaluate aerodynamic parameters, the second method works when noise covariances are unknown. The performance of both methods was assessed by benchmarking against parameter estimates from established methods like Extended Kalman Filter (EKF), Unscented Kalman Filter (UNKF), and Output Error Method (OEM).

【 授权许可】

All Rights reserved