Design of a dynamic KALMAN-filter for a stray flux-based measurement system of magnetic bearings

Abstract:
The paper describes the design and application of a KALMAN-filter to improve the position signals of a stray flux-based measurement system for the rotor position of magnetic bearings. Due to the fact that the magnetic stray flux is weak and the HALL-sensors, which measure it are placed next to the bearing coils, the pulsed coil current leads to massive noise of the position signals. To ensure a proper operation and a stable levitation of the rotor it is highly necessary to improve the position signals. Conventional analogue filtering will affect an inacceptable loose of dynamic behaviour of the entire mechatronic system. Hence the application of a KALMAN-filter is a common approach for such a problem. The quality of the filter strongly depends on the accuracy of the model which describes the system. The model of all components of the system is formed in the state space where the possible system conditions are represented by the coefficients of the system matrix A, the input matrix B and the output matrix C. Their coefficients are constants in general. The stray flux-based measurement system shows a nonlinear characteristic which is cause by saturation effects of the laminated core. In addition to that the value of the measured stray flux not only depends on the rotor position it also is related to the coil current. To take this fact into account this paper proposes the design of a KALMAN-filter with varying coefficients of the process noise and observation noise covariance. This paper deals with the model generation where the filter is based on and its implementation on a real time computing system. Furthermore, other specifics of the stray flux-based measure-ment system are discussed (e.g. magnetic cross-coupling) which take negative effect on the position sensing and solution approaches are proposed. Concluding measurements are presented which show the improvement of the signal quality before and after the implementation of the dynamic KALMAN-filter.