Applying of Bellman´s dynamic programming to the control of magnetic bearings

Magnetic bearings with premagnetization can be described as a system either with the current as input which can take continuous values or with the voltage as input which can take only discrete values because of the used inverter. The controller for the latter has to be nonlinear. One method to design optimal controllers with boundaries is the dynamic programming of Bellman. This paper describes how this method can be used to create a switching controller for a magnetic bearing. It is explained how the cost-function affects the stiffness. Experimental and simulation results are shown