Advanced Systems Theory Applied to AMB Systems

In this thesis we look at the application of advanced systems theoretic algorithms for an experimental Active Magnetic Bearing System (AMB). These systems are inherently unstable and therefore require a feedback controller. Accurate modeling of these systems with input-output data is hindered due to the strong correlation between the inputs and noise. Recently a predictor based subspace identification algorithm has been developed specially catered to data from closed-loop experiments. We use this algorithm to build models for a flexible AMB spindle. In order to improve the accuracy of the model, we use frequency domain method which leads to a nonlinear optimization problem. We consider a variety of problem scenarios for designing controllers for rejecting disturbances that are caused due to mass imbalances. For ensuring performance against an uncertainty in the rotational speed we use a robust synthesis algorithm that is known to be convex. We synthesize H-infinity controllers and observe that for a model with flexible dynamics the resulting controllers are unstable. To realize the performance benefits of an unstable controller, we use a switching method based on the Youla parameters of the stabilizing controllers. We also design and implement LPV controllers for rejecting disturbances over a range of rotational speeds