Non-linear normal modes and non-parametric system identification of non-linear oscillators

Citation
X. Ma et al., Non-linear normal modes and non-parametric system identification of non-linear oscillators, MECH SYST S, 14(1), 2000, pp. 37-48
Citations number
24
Categorie Soggetti
Mechanical Engineering
Journal title
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
ISSN journal
08883270 → ACNP
Volume
14
Issue
1
Year of publication
2000
Pages
37 - 48
Database
ISI
SICI code
0888-3270(200001)14:1<37:NNMANS>2.0.ZU;2-D
Abstract
The Karhunen-Loeve (K-L) decomposition procedure is applied to a system of coupled cantilever beams with non-linear grounding stiffnesses and a system of non-linearly coupled rods. The former system possesses localized non-li near normal modes (NNMs) for certain values of the coupling parameters and has been studied in the literature using various asymptotic techniques. In this work, the K-L method is used to locate the regions of such localized m otions. The method yields orthogonal modes that best approximate the spatia l behaviour of the beams. In order to apply this method simultaneous time s eries of the displacements at several points of the system are required. Th ese measurements are obtained by a direct numerical integration of the gove rning partial differential equations, using the assumed modes method. A two -point correlation matrix is constructed using the measured time-series dat a, and its eigenvectors represent the dominant K-L modes of the system; the corresponding eigenvalues give an estimate of the participations (energies ) of these modes in the dynamics. These participations are used to estimate the dimensionality of the system and to identify regions of localized moti on in the coupling parameter space. The same approach is applied to a syste m of non-linearly coupled rods. Through the comparison of system response r econstructions of the responses using a simple K-L mode and a number of phy sical modes, it is shown that the K-L modes can be used to create lower-ord er models that can accurately capture the dynamics of the original system. (C) 2000 Academic Press.