Algorithms for self-sensing magnetic bearings using current signals and least square identification

This paper discusses self-sensing algorithms based on least square identification for active magnetic bearings under strong eddy current influence. The considered algorithms use current data which are acquired at high frequency to estimate the air gap size by means of the actuator inductance. Eddy currents lead to a nonlinear current increase. This can result in poor position estimation accuracy when linear approximations are used. To face this problem, two approaches are considered. In the first an exponential trial function is introduced to compensate the eddy currents. Potential problems arise, because the exponent cannot be fitted with least square identification and therefore has to be taken as constant. The other utilizes the current data of two counteracting electromagnets, which are driven asymmetrically. To compensate eddy currents as well as movement induction the signals are added up, unlike other publications where they are subtracted. The estimation algorithms are investigated with measured current data. The first reduces the error to 40 % of a reference algorithm. The second shows further improvement to 20 % of the reference. However, the error absolute values are still too high for industrial applications. The paper closes with a proposal for further improvement.