Identification of Magnetic Bearing System Using a Novel Subspace Identification Method

Active magnetic bearings (AMBs) are widely applied to support high-speed rotating machineries, especially in special environment. Identification is a useful method to obtain the mathematical model of an AMB system. Frequency domain subspace identification method (FDSIM) is an attractive identification technique. However, in practice one usually expects that the identified model possesses some properties, for example stability. FDSIM cannot ensure these expected properties. This paper firstly proposes a novel subspace identification method, FDSIM-lamda. This method can guarantee that the resultant model possesses predefined poles. Moreover, an optimization procedure is introduced to obtain the optimal poles based on FDSIM-lamda, the initial estimation of system poles and constraints on the system poles. The proposed method is then applied to identify an AMB system and validated by the experiments on a real AMB system.