Sensorless control of active magnetic bearings

Active magnetic bearings (AMBs) levitate and support rotors using magnetic forces. The electromagnetic coils generate the force to keep the rotor centered, while position sensors such as eddy-current sensors are used to measure the rotor’s position for feedback control. However, these sensors often become the main source of reliability and cost issues, particularly in harsh operating environments.

Prof. Minkyun Noh’s research group at KAIST has developed a position estimator that eliminates the need for physical sensors. The proposed method combines a discrete-time voltage equation with Nyquist-frequency square-wave voltage injection to estimate rotor position using only phase voltage commands and measured phase currents. The approach was experimentally validated on an 8-pole radial magnetic bearing system. This system achieves stable levitation during take-off and dynamic tests. The experiments showed that the estimator achieved sensor-level accuracy, presenting position repeatability, linearity, and dynamic behavior closely matching conventional sensor-based control. This work was presented at the 2025 IEEE International Electric Machines and Drives Conference (IEMDC) on May 22, where it received the Best Oral Presentation Award [1].

The proposed algorithm estimates the rotor position using phase voltage commands and measured phase currents. It is derived from the discretized voltage equation under a quasi-static position assumption and enhanced by square-wave voltage injection at the Nyquist frequency. In contrast, conventional estimators typically rely on high-frequency injection combined with additional signal filters (e.g., low-pass filters), which can introduce phase delays and limit control bandwidth. By applying square-wave voltage at the Nyquist frequency, the proposed method eliminates the need for such post-processing filters. Accurate estimation is ensured only when both the discrete-time formulation and Nyquist-frequency injection are satisfied. In particular, the voltage injection is crucial for avoiding singularities in inductance estimation and maintaining estimator observability.

“Sensorless technology will be a cornerstone for next-generation rotating machinery, enabling more reliable and sustainable operation,” said Prof. Minkyun Noh, the corresponding author and a professor in the Department of Mechanical Engineering at KAIST.

The researchers expect that this technology will be advantageous in industrial applications such as heat pumps, turbo compressors, and power generators, where magnetic bearings are adopted to enable oil-free and reliable operation. Since the proposed estimator requires no additional mechanical modifications and relies only on existing phase voltage commands and current measurements, it can be seamlessly integrated into conventional systems. Beyond serving as a practical alternative to physical sensors, the estimator can also be applied for real-time condition monitoring, providing early indicators of potential issues such as unexpected rotor instability or vibration. With these features, the technology is expected to provide a practical solution for sensor replacement and predictive maintenance across various industrial applications.

This work was supported by the Korea Institute of Energy Technology Evaluation and Planning (KETEP) and the Ministry of Trade, Industry & Energy (MOTIE) of the Republic of Korea (No. RS-2023-00242282).

[1] M. Choi, D. Kim and M. Noh, “Sensorless Position Estimator for Active Magnetic Bearings Based on Non-linear Voltage Equation and Square-wave Voltage Injection,” 2025 IEEE International Electric Machines & Drives Conference (IEMDC), Houston, TX, USA, 2025, pp. 1238-1243.