Generalized Modal Decoupling Control for Active Magnetic Bearings

Control design for unstable, uncertain, and highly coupled MIMO (Multiple Input and Multiple Output) systems such as AMB (Active Magnetic Bearing) systems is extremely challenging and requires sophisticated synthesis strategies. While postmodern control methods such as µ-synthesis may be able to solve such problems, they require an accurate mathematical model of the system and specialized software tools for the controller synthesis. The resulting control algorithms are typically of excessively high order and therefore computationally demanding and challenging to implement. In contrast, the control culture of the magnetic bearing engineering community also offers classical, essentially single-input, single-output (SISO) methods to produce reliable and robust solutions to a wide variety of control problems. One well-known approach is translational-tilting modal-transformation control. However, this approach is limited in cases where the first bending mode is close to the desired closed-loop rigid body modes. This paper describes an approach for hand-synthesized, generalized modal decoupling control design for magnetically levitated rotors. The basic idea of this method is to use the singular value decomposition (SVD) to transform the decentralized inputs and outputs into generalized modal coordinates. This greatly simplifies the design of the controller, which in turn allows for much more sophisticated rotor dynamic designs by independently controlling the decoupled mode. This can be viewed as a generalization of the known decoupling transformations. The transformation works with both model- and measurement-based frequency response functions, in contrast to the known approaches in the literature, and is shown to excel where common classical control methods fail. Results and comparisons are demonstrated by an experimental evaluation of the frequency domain control performance of a supercritical gas turbine.