Extended Kalman Filter Applied to an AMB System with Strong Magnetic Saturation

The main aim of the paper is to operate an AMB system inside the magnetically saturated region. We investigate how to stabilize a highly nonlinear AMB system with the actuator magnetic saturation by estimating the system states and parameters of the unconstrained bearing in real time. The stabilization is achieved by a combination of the controller and the extended Kalman filter to estimate the states of the plant. In the extended Kalman filter, the magnetization behavior has been modeled by using low-order piecewise polynomial approximation. The extended Kalman applied to the AMB system introduces an additional challenge because of the computational burden, numerical stability when computing the inverse of the innovation covariance matrix, and the non-continuous derivative in the computed Jacobeans. Feasibility of different control configurations is examined. As the case study we use the 1 DOF balancing beam model. The proposed solution for the system operation in the magnetic saturation provides such benefits as: the reduced size and costs of bearings, improved robust performance and stability, and an increased controllability region with respect to control variables.