Reduction of sensing cycle duration in self-sensing active magnetic bearings using voltage asymptote prediction

This work aims at reducing the required time to obtain a self-sensed position signal in active magnetic bearings. Specifically, the self-sensing concept of Direct Digital Inductance Estimation should be enhanced by a prediction algorithm. Ordinarily, the current slope as a result of a coil voltage step is measured via a transformer, whereby a single measurement of the transformer's secondary coil voltage is sampled after the voltage has converged towards its asymptote value. This paper proposes to predict the asymptote by fitting an exponential curve to multiple measurements sampled before the transformer voltage has converged. While a lower accuracy of the self-sensed position signal is to be expected with this approach, the shorter duration of the sensing phase allows for more of the amplifier voltage to be used to actuate the magnetic bearings. This in turn can be useful in situations where the magnetic bearings reach their dynamic limits in supplying the requested coil currents. A single-axis magnetic bearing test bench driven by a switching amplifier in differential current control is used to evaluate this approach at different flotor positions. Statistical evaluations of the position signals obtained by the conventional and by the prediction method are compared. Finally, an outlook is given on the feasibility of this approach and possible use cases.