Aerodynamic Loading and Magnetic Bearing Controller Robustness Using a Gain-Scheduled Kalman Filter

Abstract

Modeling or predicting aerodynamic loading effects on rotating equipment has been a source of concern to those who wish to examine stability or response of critical components. The rotordynamic model of the system employed for such examination assumes greater importance for active bearings than for passive ones, if only because of the additional potential for instability introduced by the controller. For many systems, aerodynamic loading may vary widely over the range of operation of the bearings, and may depend on extended system variables. Thus, potential controllers for active magnetic bearings require sufficient robustness or adaptation to changes in critical aerodynamic loading parameters, as might be embodied in cross-coupled stiffness terms for compressor impellers. Furthermore, the presence of plant or measurement noise provides additional sources of complication. Here, the previous development of a nonlinear controller for a hypothetical single-stage centrifugal gas compressor is extended by comparing the compensator performance using a multivariable Luenberger observer against that of a stationary Kalman filter, both gain-scheduled for rotational speed. For the postulated system, it was found that the slower poles of the Kalman filter did not observably detract from controller convergence and stability, while predictably smoothing out the simulated sensor noise.