State Estimation and Run-Out Reduction for Magnetically Levitated Motor Systems

Magnetically levitated motor systems create opportunities for new applications of electromechanical power conversion where high-speed and contact-free operation are required. These systems require position sensors and control algorithms for stable levitation which often require filters to reduce feed-back noise. Noisy position feedback creates challenges for high-bandwidth estimation of velocity and disturbance force; however, these unmeasured states can be helpful for high-performance control algorithms. This common problem in drives for electric motors has been solved using the Luenberger-style motion state observer. This paper investigates the applicability of the observer to digitally-controlled magnetically levitated systems. First, the observer's embedded plant model is derived for discrete-time implementation. Then, it is shown that the most advantageous use-case of the observer is different for magnetically levitated systems versus rotary motor systems. Unlike motor systems, the zero-lag filtering property of observers is minimally useful, while the ability to estimate and reject disturbances yields significant performance improvements for applications that care about run-out. In a prototype bearingless motor, rotor run-out is reduced by 5x by using control techniques based on the proposed observer.