System Identification for Active Magnetic Bearings at Nonzero Speeds

In this paper we consider the problem of obtaining a dynamic model of rotating machinery equipped with magnetic bearings for a given speed range. Theoretical models are typically not accurate enough and cannot be trusted to predict system behaviour for the whole operating range of the machine. In this work we first obtain a dynamic model of the system at several speed levels; second, an extrapolation/interpolation of these models is performed to produce an estimate of the behaviour for the whole speed range. This procedure overcomes two common issues in system identification for magnetic bearings: first, it is costly and time-consuming to perform system identification at many speed levels to cover the whole operating range. Second, it is typically not safe to reach certain speeds unless the stability and unbalance response at high speeds has been verified. This can be only done once a reliable model is available. From a technical perspective, our approach relies on local system identification techniques aimed at linear parameter varying (LPV) systems. Compared to existing work, we believe the main contributions of this work are: we do not rely on global approaches that might lead into nonconvex problems, our approach is amenable to automation, and it allows considering the presence of casing modes in the system.