Applied Nonlinear Controls Lab

Active Magnetic Bearings

Active magnetic bearings (AMBs) are receiving increasing amounts of attention due to their unique characteristics. AMBs use actively controlled electromagnetic forces to control the motion of a rotor or other ferromagnetic body in air. Contact-free suspension leads to important advantages relative to conventional rolling element or hydrodynamic bearings: reliability, low-maintenance and losses, and higher speeds in extreme environments without requiring complex lubrication systems. Examples of applications include:

  1. high-performance vacuum pumps due to absence of wear, lubrication systems, and lubricant contamination of the process
  2. Machining spindles used for high speed cutting which benefit from the high rotation speeds and accurate positioning that are enabled by contact free operation and active controls.

Research in the Applied Nonlinear Controls Lab is focused on the design and implementation of various nonlinear control schemes for a five degree-of-freedom system from SKF Magnetic Bearings. The aim of the research is to design a controller which can guide a high-speed rotating shaft along a desired trajectory over the bearing air gap. This tracking control objective has application to precision high-speed boring of elliptical holes and positioning stages. However, operating the shaft over a large fraction of the bearing air gap demands consideration of the nonlinear force-current-position relationship. As a result, control can benefit from modeling and direct compensation of this nonlinear behaviour.

A. Comparison of bias-based and zero-bias nonlinear controllers

Initial research efforts focused on the fundamental issue of bias currents and how they affect the performance of nonlinear tracking controllers. Specifically, a desired bearing force can be generated in a non-unique manner from the input coil currents. Conventional AMB operation utilizes bias currents to facilitate linear control designs and to produce a higher force response. However, most of the proposed nonlinear control techniques have focused on avoiding these energy-wasting bias currents by complementary, or switching, control schemes. Our work, which is reported in [1], compares bias-based and zero-bias nonlinear controllers. Both designs are demonstrated experimentally to achieve tracking of a non-rotating shaft; however, when subjected to disturbances through high-speed rotation the bias-based nonlinear design is shown to be more robust for a given voltage saturation level. Hence zero-bias operation, although minimizing control effort, puts higher demands on the control hardware to achieve the same dynamic specifications as bias-based control.

B. Disturbance estimation and suppression

Achieving precision tracking of a shaft when it is subject to rotation is a challenging problem due to the induced synchronous vibration resulting from mass unbalance. The use of nonlinear state estimators is an approach which integrates naturally which the aforementioned tracking control schemes. A constant and harmonic disturbance observer allows us to actively estimate the forces resulting from vibration, shaft loading, and steady state model error. Feeding these estimated forces back into the control loop allows us to balance the shaft and compensate in real time for loading effects and changes in the speed setpoint. The figure below shows engagement of constant and harmonic disturbance compensation to the shaft rotating at 10,000 rpm.

After compensation, the shaft is driven to the origin with a vibration amplitude of 1 micron. At this point, the shaft may be guided along a desired path in the air gap. This is shown in the figure below, where the shaft is rotating at 10,000 rpm and moved along an 2 Hz elliptical orbit. This work is reported in [2].

C. Inertial autocentering

Present research efforts are focused on a related vibration control problem known as inertial autocentering. Previously, we applied a synchronously rotating magnetic force to cancel out the unbalance forces and keep the shaft rotating about its geometric center. An alternative approach is to allow the shaft to rotate about its inertial center, eliminating the vibration forces which usually get transmitted to the bearing housing and platform. Inertial autocentering can be achieved by rendering the controller unresponsive to the synchronous vibration - and this can be solved by reformulating the disturbance estimation and compensation problem. Inertial autocentering is a unique capability of AMBs, and is an advanced control algorithm which is beneficial to applications with low noise and vibration requirements.

This research effort is supported at the University of Alberta by an NSERC equipment grant. The co-applicant on this equipment grant is Professor H.J. Marquez. The construction of the test bench has benefited from the support of Bazooka Electronics (Edmonton, AB) and Carsten Collon from TU-Dresden Institut für Regelungs- und Steuerungstheorie.


[1] T.R. Grochmal & A.F. Lynch, 'An Experimental Comparison of Nonlinear Tracking Controllers for Active Magnetic Bearings', Control Engineering Practice, Vol.15, No.1, pp.95-107, Jan. 2007.

[2] T.R. Grochmal & A.F. Lynch, 'Precision Tracking of a Rotating Shaft with Magnetic Bearings by Nonlinear Decoupled Disturbance Observers', IEEE Transactions on Control Systems Technology, Vol.15, No.6, pp.1112-1121, Nov. 2007.