Adaptive Linear and Extended KALMAN Filter applied to AMB with Collocated Position Measuring

In the majority of cases, magnetically levitated rotors are equipped with externally mounted position sensors, i.e. a dislocation between bearing and sensor plane occurs. The disparity between actual and measured rotor position can cause system instability due to natural oscillations of the shaft. This paper proposes a method for avoiding this problem by integrating capacitive position sensors straight into the stator laminations of a radial magnetic bearing. The stronger sensor noise originating from the proximity between bearing coils and sensors can be reduced effectively by means of a stochastical state estimator. Different types of KALMAN filters used for these purposes, both linear and extended ones, are presented in this paper. It is shown how unknown bearing parameters can be estimated by means of the filter. Concluding, a LQG controller (optimal controller combined with optimal state estimator) is presented and compared to an optimally parametrised PID controller. All methods proposed in this paper are approved by measurements.