IDENTIFICATION OF MULTISCALE STATE-SPACE MODELS FROM INPUT-OUTPUT DATA

摘要

This paper is concerned with the problem of identifying state-space models at different time scales on the basis of input-output data. The simplest approach consists of obtaining a model at the original time domain and then propagating the resulting model matrices to larger scales. The alternative approach proposed in the present work consists of directly identifying a separate model for each time scale. For this purpose, the model equations are conveniently rewritten in terms of transformed input variables. A simulated case study involving the flexible dynamics of an aircraft is presented for illustration. System identification was carried out by the N4SID subspace method. The identified models were evaluated in terms of the eigenvalue locations as well as the magnitude of the prediction errors. The results reveal that the proposed approach is less sensitive to measurement noise compared to the identification at the original time domain.