Abstract :
[en] The present paper addresses the problem of modal identification of time-varying systems. The identification is based on a multivariate autoregressive moving-average model in which the time variability of the system is caught using a basis functions approach. In this approach, the time-varying regressive coefficients in the model are expended in the chosen basis functions and only the projection coefficients have to be identified. In that way, the initial time-varying problem then becomes a time-invariant one that can be solved. Because a multivariate model is used, in addition to the time-varying poles, the time-varying mode shapes may be identified too. The method is first presented and then applied on an experimental demonstration structure. The experimental structure consists of a supported beam on which a mass is travelling. The mass is chosen sufficiently large to have a significant influence on the dynamics of the primary system. This kind of problem is a classical example commonly used by many authors to test time-varying identification methods.
Scopus citations®
without self-citations
3