New Direct Closed-loop Identification Method for Unstable Systems and Its Applications to Magnetic Suspension System

Abstract

Closed-loop identification is essentially needed in control of an unstable plant stabilized by a feedback controller. A new direct closed-loop identification approach is proposed based on an output inter-sampling scheme, in which by taking faster sampling of the output than the input of the system the restrictive identifiability condition is removed. In a case of unstable system, since the noise model becomes non-minimum phase system, the prediction error method is not available. In order to solve these problems we clarify that the output inter-sampling yields the SIMO structure of the plant model and give a new identification method in which the denominator polynomial can be identified in an open-loop manner and the numerator polynomial can also be obtained by a modified instrumental variable method. The proposed algorithm is validated in an experimental study using a magnetic suspension system.