Continuous time system identification of magnetic bearing systems using frequency response data

This paper focuses on identifying nonlinear unstable magnetic bearing systems using a linear model and frequency response data. Subspace system identification methods adopting a Laguerre network in the frequency domain were used to identify continuous time transfer functions for a test-stand magnetic bearing apparatus. The results show that a continuous time multivariable model was successfully identified based on the experimental data.

1 Introduction

Magnetic bearings (MBs) have been recognized by industry [1] to have the required attributes, which allow them to be used in applications where traditional bearings and their associated lubrication requirements are impractical. Additionally, as. they operate without contact between the rotating shaft and the stationary + bearing elements, frictional losses are minimized. But there are some rotational losses due to induced eddy currents. This greatly facilitates the high speed operation of rotating machines, and significantly reduces problems occurring when contact bearings are used.