Robustness of Recursive Auto-tuning Strategy in Flexible Rotor and Magnetic Bearing Systems

Active magnetic bearings offer the capability to control rotor vibration and transmitted forces through the design of adaptive controllers. These controllers can take into account the changing operating conditions and variations in the system dynamics. A Recursive Open-Loop Adaptive Control (ROLAC) algorithm was developed, which minimizes the sum of squares of the measurements in frequency domain. ROLAC is used in addition to local PID controllers, which ensure the stability. Magnetic bearing dynamics depend on the eccentricity of the rotor with respect to magnetic poles. Non-concentric rotation of the rotor within the magnetic bearings causes changes in the bearing dynamic properties. Apart from system faults, non-concentric operation of the rotor can occur due to misalignment of the auxiliary bearings with respect to magnetic poles, changes in sensor calibration, and misalignment of bearings in systems where there are three or more bearings. This misalignment leads to changes in the magnetic bearing characteristics thereby affecting the overall system dynamics. This in turn may necessitate re-tuning of the ROLAC. The paper studies the effect of such misalignments on the controller performance and the robustness of the auto-tuning process. INTRODUCTION Magnetic bearings possess attractive characteristics for industrial and manufacturing applications due to their ability to support high-speed rotating loads with no friction, no lubrication and under extreme environmental conditions [1]. Since magnetic bearings can be actively controlled, they offer potential advantages compared with conventional bearings including vibration control, automatic balancing, condition monitoring and fault diagnostics [2]. However, due to their limited force capacity, active magnetic bearings have to incorporate retainer bearings to protect the laminations when rotor vibrations reach the clearance level. The future of magnetic bearing applications relies on addressing the safety and reliability issues in critical applications [3]. A range of control techniques have been developed to control active magnetic bearings under normal operation, to minimize rotor vibrations, and also to minimize transmitted forces [4–8]. An effective method of controlling synchronous vibration under varying operating conditions has been introduced by Burrows and Sahinkaya [4, 9, 10]. This open-loop adaptive control (OLAC) strategy, also referred as automatic balancing, can be extended to attenuate multi-frequency vibrations of the rotor [11]. However, because it depends on the steady state response and a Fourier transform of measured displacements, it may not be fast enough to prevent rotor contact with auxiliary bearings in response to a sudden change of unbalance. Therefore a recursive version of the algorithm has been developed, which updates the optimum force amplitude and phase at each sampling interval. This technique has improved the reaction speed of the controller to transient changes. It has also been shown that the ROLAC can prevent rotor contact with an