Continuous time state-space model identification with application to magnetic bearing systems

This thesis presents the identification of continuous time linear multi-variable systems using state-space models. A data-driven approach in realization by the subspace methods is carried out in developing the models. In this thesis, the approach by subspace methods is considered for both open-loop and closed-loop continuous time system identification. The Laguerre filter network, the instrumental variables and the frequency sampling filters are adopted in the framework of subspace model identification. More specifically, the Laguerre filters play a role in avoiding problems with differentiation in the Laplace operator, which leads to a simple algebraic relation. It also has the ability to cope with noise at high frequency region due to its orthogonality functions. The instrumental variables help to eliminate the process and measurement noise that may occur in the systems. The frequency sampling filters are used to compress the raw data, eliminate measurement noise so to obtain a set of clean and unbiased step response data. The combination of these techniques allows for the estimation of high quality models, in which, it leads to successful performance of the continuous time system identification overall. The application based on a magnetic bearing system apparatus is used to demonstrate the efficacy of the proposed techniques.