Radial Basis Function (RBF) for Non–Linear Dynamic System Identification

Abstract

Masalah utama dalam pengenalpastian sistem ialah memilih struktur model yang sesuai. Dalam artikel ini, rangkaian fungsi asas jejarian menggunakan pelbagai fungsi asas dilatih untuk mewakili sistem dinamik tak linear masa diskrit dan keputusannya dibandingkan. Algoritma kuasa dua terkecil ortogon digunakan untuk memilih model rangkaian asas jejarian termudah. Untuk menerangkan tatacara pengenalpastian, dua contoh pemodelan sistem tak linear dibincangkan. Kata kunci: fungsi asas jejarian; pengenalpastian sistem; pemodelan sistem tak linear; algoritma kuasa dua terkecil ortogon One of the key problem in system identification is finding a suitable model structure. In this paper, radial basis function (RBF) network using various basis functions are trained to represent discrete-time nonlinear dynamic systems and the results are compared. The orthogonal least square algorithm is employed to select parsimonious RBF models. To demonstrate the identification procedure, two examples of modelling nonlinear system were included. Key words: radial basis function; system identification; non-linear system modelling; orthogonal least square algorithm