Perpustakaan Digital ITB

Aircraft system identification is a crucial factor influencing the development of aircraft performance. The process of system identification involves estimation and modeling of various systems within an aircraft. Optimization methods play a pivotal role in enhancing the accuracy and efficiency of parameter estimation, addressing challenges such as non linearity, noise, and system complexity. The Output Error Method (OEM) is a widely used technique in aircraft system identification that can deal with measurement noise. The aim of the research is to explore and explain an alternative optimization methods in Output Error Method (OEM) framework for aircraft system identification and compare its performance with existing methods. The Gauss- Newton and Levenberg-Marquardt algorithms are widely used for estimating parameters in nonlinear models, with the former optimizing through successive linear approximations and the latter combining gradient descent and Gauss- Newton to enhance convergence stability. In this work, the Nelder-Mead and Broyden-Fletcher-Goldfarb-Shanno (BFGS) are compared with Gauss-Newton and Levenberg-Marquardt method as a baseline algorithm that works good within the OEM framework. The data utilized for evaluation is from book of R.V. Jategaonkar’s, specifically from a lateral-directional motion test case conducted on the ATTAS aircraft. Results indicate that Gauss-Newton and Levenberg-Marquardt method still produce best parameter estimates compared to the alternative methods.