Performance limitations and self-sensing magnetic bearings

Magnetic bearings levitate a rotating object (typically, a rotor) with a magnetic field, and are unstable in open-loop operation. Position feedback control is required to maintain the rotor in a centered position. Typically, a separate position sensor is used to directly measure the rotor's position for use in this feedback loop. Self-sensing magnetic bearings use the bearing's electrical signals to estimate rotor position.; There are two primary classes of self-sensing methods: state estimation and ripple-based. State estimation approaches model the bearing as a linear, time-invariant system and treat the rotor position as a state to be estimated as part of linear, time-invariant feed-back control. Ripple-based approaches rely on effects of the driving switching amplifier to estimate position. The self-sensing method presented here is a ripple-based approach, in that it generates estimates of position from switching voltage and current signals.; The primary impediment to many schemes is achieving acceptable robustness to plant parameter variations. System theory for linear, time invariant plants provides explicit measures of achievable robustness, but there are no such analyses for ripple-based approaches. While many researchers discuss such problems, none has provided quantified measures of such robustness. Many researchers of ripple-based approaches claim that their techniques achieve higher stability robustness than the state estimation approaches, but none has provided theoretical or experimental measures that attempt to quantify such robustness improvement. Furthermore, for ripple-based approaches, there is no design and analysis method to replace the current ad-hoc methods of ripple-based design and implementation.; Simulation and experimental data demonstrate that the self-sensing method presented here can make a viable active magnetic bearing system. Gain margin comparisons demonstrate that our ripple-based approach appears to have higher robustness than what would be expected from a corresponding ripple-less self-sensing scheme. Furthermore, robustness measures calculated for a linear, periodic model of the bearing suggest that the ripple is the mechanism which provides added robustness to self-sensing systems.