Design of Anti-Interference Control Scheme

Magnetic levitation systems are inherently open-loop unstable and highly nonlinear plants. Because electromagnetic attractive force is inversely proportional to the square of the air gap, even minute disturbances can lead to gap divergence and loss of control. Consequently, designing robust anti-interference control schemes is the pivotal step in transitioning magnetic levitation from merely achieving basic suspension to delivering high precision and unwavering stability.

Before formulating any disturbance-rejection mechanism, a thorough identification of disturbance sources and their respective characteristics is essential:

  • Payload Variations: Changes in passenger load on maglev trains or the placement of distinct weights on a levitation platform manifest as step or slowly varying forces along the vertical axis.
  • Track and Foundation Vibrations: Rail irregularities and ground tremors couple into the suspension air gap through the mechanical support structure, typically appearing as broadband displacement disturbances.
  • Electromagnetic Interferences: Switching noise from power amplifiers and ripple currents in the coils superimpose directly onto the control force.
  • Sensor Noise: High-frequency noise originating from displacement sensors is amplified through differential processing, inducing chattering in the control input.
  • Parametric Perturbations: Coil resistance fluctuations caused by thermal effects, alongside magnetic material saturation, result in drift within inductances and air-gap stiffness.

While the first three categories represent external disturbances, the latter two constitute internal uncertainties. A comprehensive anti-interference scheme must simultaneously address both fronts.
Engineering practices generally classify these strategies into three primary approaches:

  1. Feedback Suppression: Elevating controller gain or bandwidth using conventional methods like PID or LQR. While straightforward to implement, this approach is fundamentally constrained by actuator bandwidth and noise amplification, preventing infinite performance scaling.
  2. Feedforward Compensation: Designing a feedforward channel for measurable disturbances—such as previewing track irregularities ahead of a moving train—to proactively cancel out disturbance effects prior to output degradation. This yields rapid responses but relies heavily on disturbance measurability and high model accuracy.
  3. Disturbance Estimation and Compensation: Utilizing structures like Disturbance Observers (DOB) or Active Disturbance Rejection Control (ADRC) to estimate total lumped disturbances online and compensate for them in real time. Effective against unmeasurable disturbances as well, this has become the mainstream paradigm in modern magnetic levitation systems.

Furthermore, for operating environments plagued by severe parametric perturbations, $H_\infty$ robust control or sliding mode control can be deployed to guarantee worst-case stability.

Design Case Study: Magnetic Levitation Ball System with a Disturbance Observer

Consider the classic single-degree-of-freedom magnetic levitation ball system, governed by the dynamic equation:

$$m\ddot{x} = mg - \frac{k i^2}{x^2} + d$$

where $x$ represents the air gap, $i$ denotes the coil current, and $d$ signifies external disturbances. Linearization around the nominal operating point yields a second-order unstable plant. After establishing closed-loop stability via standard PID control, a disturbance observer is integrated through the following steps:

  1. Constructing the Nominal Model: Define the linearized nominal model as $G_n(s)$, whose inverse $G_n^{-1}(s)$ can be analytically derived.
  2. Designing the $Q$-Filter: The DOB structure is formulated as $\hat{d} = Q(s)\left[G_n^{-1}(s)u - y\right]$. A low-pass configuration $Q(s) = \dfrac{1}{(\tau s + 1)^2}$ is selected, with a bandwidth tuned between 100 Hz and 300 Hz to balance disturbance rejection speed against noise sensitivity.
  3. Control Input Augmentation: Convert the estimated disturbance into an auxiliary control signal $u_c = -\hat{d}/G_n^{-1}(s)$ and superimpose it directly onto the PID output.

During experimental validation, when a step force of approximately 0.2 N is abruptly applied to the stably levitated steel ball, a pure PID controller suffers an air-gap drop of about 1.5 mm with a recovery time exceeding 0.5 s. In contrast, integrating the DOB restricts the maximum drop to under 0.3 mm, compresses the recovery time to the order of 0.1 s, and avoids any noticeable amplification of high-frequency sensor noise.

Engineering Implementation Essentials

  • Sampling and Filtering: The cutoff frequency of the $Q$-filter must remain strictly below half the Nyquist frequency, and appropriate filtering blocks should be cascaded within $G_n^{-1}(s)$ to prevent estimation divergence caused by non-minimum phase zeros.
  • Bandwidth Hierarchy: The observer bandwidth must sit comfortably below the current loop bandwidth yet remain higher than the primary disturbance frequency band, establishing a well-structured multi-loop dynamic hierarchy.
  • Protection Logic: A safety mechanism should trigger a fallback to pure PID operation and issue an alarm if the estimated disturbance exceeds a predefined threshold, preventing system runaway should the observer fail.
  • Experimental Verification: Validate performance progressively following a structured sequence of "step loading, frequency sweep excitation, and long-duration endurance testing" to confirm disturbance rejection and overall robustness.

Summary

The core philosophy of anti-interference control design relies on a systematic workflow: "analyze first, compensate second, and verify thoroughly." By identifying the frequency bands and magnitudes of prospective disturbances, engineers can synthesize hybrid architectures combining feedback, feedforward, and disturbance observation. Tuning parameters such as the $Q$-filter bandwidth successfully arbitrates the trade-off between disturbance suppression capability and noise attenuation. For inherently unstable systems like magnetic levitation, a rigorously validated anti-interference framework remains the cornerstone of long-term reliable operation.