Control Boost of an Active Magnetic Bearing System with DOBs

Yupeng Zheng, Hyeong-Joon AhnAsian 2022 · DOI

TL;DR

This paper demonstrates how Disturbance Observers (DOBs) can significantly "boost" the performance of Active Magnetic Bearing (AMB) systems compared to industry-standard PID control. By implementing DOBs in both voltage-controlled and current-controlled architectures, the researchers achieved up to a 60.7% improvement in command following and a 28% reduction in overshoot, making these systems far more robust against vibrations and load changes.

What Problem Does This Paper Address?

Active Magnetic Bearings (AMBs) are a marvel of modern engineering, allowing spindles to rotate without physical contact, eliminating friction and the need for oil lubrication. However, they have a "vibration problem." Because they rely on magnetic fields to levitate a rotor, they are highly sensitive to disturbances like rotor mass-imbalance and centrifugal forces, which manifest as synchronous vibrations.

While the industry heavily relies on conventional PID (Proportional-Integral-Derivative) control for its simplicity, PID often struggles to maintain stability and precision when faced with dynamic loads or high-frequency disturbances. The authors aim to bridge this gap by applying Disturbance Observers—a robust control tool—to a Magnetic Levitation System (MLS) to see if they can provide a measurable "boost" in stability and command-following accuracy.

Methods

The researchers utilized a Magnetic Levitation System (MLS) as their experimental platform, which consists of an electromagnet levitating a steel ball. They approached the control problem in two stages:

  1. Voltage-Controlled MLS: A baseline setup where a PID controller is augmented with a DOB. The DOB's job is to estimate the difference between the nominal plant model and the actual system behavior (the disturbance) and feed a corrective signal back into the system.
  2. Current-Controlled MLS: A more complex, dual-loop architecture. This includes:
  3. An inner PI Current Control loop with its own DOB to compensate for changes in coil inductance as the ball moves.
  4. An outer PID Position Control loop with a second DOB to handle external disturbances and improve tracking.

The system was modeled using nonlinear equations for electromagnetic force and coil inductance, which were then linearized around an equilibrium point (10mm) for controller design.

Datasets

As this is an experimental control systems paper, it does not use traditional "datasets" in the machine learning sense. Instead, the authors evaluated their system using:

  • Physical Parameters: A 32g steel ball (with variations at 19g and 39g to test robustness).
  • Disturbance Profiles: Sinusoidal disturbances at various frequencies (0.5 Hz to 2.0 Hz) and amplitudes (0.01V).
  • Reference Commands: Sinusoidal input commands with a 0.001m amplitude to test tracking performance.

Key Insights and Ideas

  • The "Mass-Robustness" Paradox: Interestingly, the researchers found that as the mass of the steel ball increased (from 19g to 39g), the disturbance rejection ratio actually improved (from 79.9% to 90.2%). This suggests the DOB-augmented system handles higher inertia more effectively in this specific configuration.
  • Dual-Loop Synergy: In the current-controlled model, the DOBs serve two distinct purposes. The current-loop DOB compensates for the physical nonlinearities of the electromagnet (inductance changes), while the position-loop DOB handles the "external" world.
  • Simplicity vs. Performance: The paper reinforces that DOBs are "plug-and-play" enhancements. They don't require replacing the existing PID infrastructure but rather wrap around it, making them a highly practical upgrade for industrial AMB systems.

Key Results

The addition of DOBs provided substantial quantitative improvements across the board:

  • Command Following: In the voltage-controlled setup, command-following performance improved by an average of 60.7%.
  • Disturbance Rejection: Low-frequency disturbances (0.5 Hz) were rejected at a rate of 83.6%.
  • Transient Response: In the current-controlled setup, the settling time was reduced by 0.06s, and maximum overshoot was slashed by 28% during initial levitation.
  • Overall Robustness: For the current-controlled position loop, disturbances in the 0–3 Hz range were removed by an average of 69.1%.

Limitations and Future Work

The primary limitation noted is the frequency dependence of the DOB. While it is exceptionally effective at low frequencies (0.5–1.0 Hz), its performance deteriorates as the disturbance frequency increases (dropping to 34.5% rejection at 2.0 Hz).

Future work would likely involve optimizing the "Q-filter" within the DOB to handle higher-frequency synchronous vibrations, which are common in high-speed AMB spindle applications. Additionally, testing the system on a full rotating AMB spindle (rather than a single-axis ball levitator) would be the logical next step for industrial validation.

Summary updated

BibTeX
@inproceedings{zheng2022control,
  author = {Yupeng Zheng and Hyeong-Joon Ahn},
  title = {{Control Boost of an Active Magnetic Bearing System with DOBs}},
  booktitle = {Asian},
  year = {2022},
  doi = {10.3850/978-981-18-6021-8\_or-02-0002.html},
}
Figure