Position estimation accuracy improvement based on accurate modeling of self-sensing active magnetic bearings

Abstract

The Self-sensing Active Magnetic Bearings (SSAMBs) which feature for the simplification of independent displacement sensors, contribute to both cost reduction and convenience in control and assembly. The key point of this technique is to accurately estimate rotor position signals directly from coil voltages or currents, thus the actuator itself also serves as a virtual displacement sensor. Although several feasible approaches have been proposed to realize the self-sensing process, there still lie limitations in estimation accuracy, robustness and dynamic performance of the system. In fact, due to the nonlinear property of the self-sensing process, few accurate models are available for quantitative analysis of the estimation accuracy, and this restricts further development of this technique. Focusing on the modulation type SSAMBs, an accurate analytical model of the self-sensing in frequency domain is firstly established in this paper. Eddy current effects and filter properties are also considered for better accuracy. Based on this model, we evaluate possible estimation error sources during the self-sensing process, and investigate if the estimation accuracy can be improved fundamentally by selecting main system parameters properly. In this way, the estimation accuracy gets further improved under existing compensation methods, and better system robustness and dynamic performance can be achieved. Finally, the analytical model and associated conclusions are validated through static experiment, 4-Degree of Freedom (DOF) self-sensing rotor levitation experiment and 0–5000rpm rotating operation experiment on a 4-DOF rigid rotor-radial SSAMBs platform.