Analysis of perforated magnetic shields for electric power applications

Peter Sergeant, Ruth V. Sabariego, Guillaume Crevecoeur, Luc Dupré, C. GeuzaineIET 2009 · DOI

TL;DR

This paper tackles the challenge of designing perforated magnetic shields for electric power applications, which are crucial for reducing electromagnetic interference but are complex to model due to their 3D geometry introduced by perforations. The authors propose two methods to optimize these shields: a fast but less accurate modified 2D model and a more accurate but complex 3D thin-shell FEM combined with a space-mapping optimization algorithm.

What Problem Does This Paper Address?

Magnetic shields are essential for mitigating electromagnetic interference (EMI) in various electric power applications, especially at low frequencies. Traditionally, these shields are made of solid metallic sheets. However, in practice, perforations are often necessary for cooling or routing wires. These holes, even if small, break the symmetry of the shield, forcing the use of computationally expensive 3D numerical models for accurate analysis. This poses a significant challenge for design optimization, which typically requires numerous simulations. The paper aims to develop efficient methods for analyzing and optimizing the performance of these perforated shields, balancing accuracy with computational cost.

Methods

The paper presents two distinct approaches for analyzing and optimizing perforated magnetic shields:

  • Method 1: Modified 2D FEM (Fast and Easy-to-Implement)
    • This method starts with a classical 2D axisymmetric FEM model.
  • Perforations are approximated as "axisymmetric air gaps," effectively creating a segmented shield.
  • Crucially, the authors introduce modifications to this 2D model to better approximate the behavior of a 3D perforated shield. These modifications involve deriving equivalent air gap diameters and equivalent conductivities based on reluctance and eddy current considerations, respectively. These are derived for both low-frequency (flux shunting dominated) and high-frequency (eddy current dominated) regimes.
  • This approach is computationally inexpensive and suitable for quick, approximate designs.
  • Method 2: 3D Thin-Shell FEM with Space-Mapping Optimization (Accurate and Fast)
    • This method utilizes a 3D finite element model that employs a "thin-shell approximation." This approximation significantly reduces the computational burden of a full 3D model by reducing the shield to its mid-surface.
  • The optimization algorithm combines the fast, unmodified 2D FEM (as a "coarse model") with the accurate 3D thin-shell FEM (as a "fine model") using a space-mapping optimization technique.
  • Space mapping efficiently explores the design space by using the coarse model to guide the optimization of the fine model, significantly reducing the number of expensive 3D simulations required.
  • A specific variant, "aggressive space mapping" (ASM), is employed, which uses gradients of both the coarse and fine objective functions for parameter extraction, leading to better convergence.

Both methods are validated against experimental measurements for both unperforated and optimized perforated shields.

Datasets

Not discussed in the paper. The study relies on numerical simulations and experimental measurements performed by the authors.

Key Insights and Ideas

  • Perforations are a significant performance killer: Even small holes drastically reduce the shielding effectiveness of magnetic shields, primarily by reducing the effective material volume and creating pathways for magnetic flux.
  • Geometry matters, but material removal is key: The study suggests that the shielding deterioration is largely determined by the volume of material removed, meaning the number and size of holes are less critical than their total area, as long as the total removed volume is the same.
  • Frequency dictates the shielding mechanism: At low frequencies, flux shunting through the high-permeability material dominates. At higher frequencies, eddy currents induced in the shield become the primary shielding mechanism. This distinction is crucial for developing accurate models.
  • 2D models need clever adaptation: A standard 2D axisymmetric model is insufficient for perforated shields. However, by carefully deriving equivalent parameters (like effective air gap and conductivity), a modified 2D model can achieve reasonable accuracy, offering a good compromise between speed and precision.
  • Thin-shell FEM is a game-changer for 3D shields: The thin-shell approximation dramatically cuts down the computational cost of 3D FEM simulations for shields, making them more feasible for design and optimization.
  • Space mapping accelerates optimization: Combining a fast, approximate model (like the 2D FEM) with an accurate, but slower model (like the 3D FEM) within a space-mapping framework is a powerful strategy for efficient design optimization.

Key Results

  • Modified 2D model accuracy: The modified 2D FEM (2DM) showed deviations of less than 10% from the 3D FEM for a large hole diameter (76 mm) across a wide frequency range (0.1 Hz to 30 kHz). For smaller holes, the accuracy is even better.
  • 3D Thin-Shell FEM efficiency: The 3D FEM with thin-shell approximation was significantly faster than a conventional 3D FEM, with a CPU time of 26 seconds compared to more than double for a non-thin-shell approach.
  • Space-mapping optimization speedup: For a specific optimization problem (finding the optimal hole diameter for n=2), the space-mapping algorithm required only 4 evaluations of the fine (3D) model and 55 evaluations of the coarse (2D) model, compared to 12 evaluations of the fine model for a gradient algorithm. This highlights the significant computational savings.
  • Experimental validation: Both proposed methods were validated against experimental data, showing good agreement, particularly for the optimized perforated shield. The shielding factor for the optimized perforated shield was found to be around 0.6, compared to 0.3 for an unperforated one, demonstrating the effectiveness of the optimization.

Limitations and Future Work

  • Accuracy dependence on geometry: The accuracy of the modified 2D FEM is noted to be dependent on the specific geometry of the application. The derived equivalent parameters might need adjustment for different shield shapes or applications.
  • "Switching frequency" approximation: The modified 2D FEM uses distinct low- and high-frequency approximations for equivalent parameters. The transition between these approximations (the "switching frequency") can introduce a small "jump" in results, which is deemed acceptable but still a limitation.
  • Non-linear material effects: While the study used linear material properties (constant permeability), it acknowledges that for certain applications where shield regions might enter the Rayleigh or saturated zones, a non-linear model would be necessary.
  • Global optimum guarantee: The paper mentions that while the space-mapping algorithm found a global optimum in their example, it does not inherently guarantee a global optimum, especially with a limited number of fine model evaluations.
  • Future work: The authors suggest that for other applications, modifications to the equivalent parameters (deq, req) might be necessary. They also point out the need to check the influence of shield geometry on field intensity and the potential requirement for non-linear models in specific scenarios.

Summary updated

Abstract

The shielding performance of perforated magnetic shields for electric power applications is described. The shielding of an axisymmetric induction heating device is studied as a function of frequency, number of perforations and dimensions of the perforations. From the numerical point of view, the perforations cause the numerical model to be 3D. A numerical optimisation is carried out to find the optimal geometry with respect to the shielding factor and the volume of the shield. For the optimisation, two approaches are presented. The first approach is fast and easy-to-implement, but has limited accuracy. It uses a classical 2D axisymmetric model where the perforations are approximated by ‘axisymmetric air gaps’ resulting in a segmented shield. It is shown how to modify the 2D model to obtain results that are similar to the ones of a 3D model. The second approach is more accurate although quite fast, but more difficult to implement. It combines a 3D thin-shell finite element model with the unmodified 2D model in a space mapping optimisation algorithm. The validation of both models is based on experimental work for an unperforated shield and for the optimised perforated shield.
BibTeX
@inproceedings{sergeant2009analysis,
  author = {Peter Sergeant and Ruth V. Sabariego and Guillaume Crevecoeur and Luc Dupré and C. Geuzaine},
  title = {{Analysis of perforated magnetic shields for electric power applications}},
  booktitle = {IET},
  year = {2009},
  doi = {10.1049/iet-epa:20080203},
}

Loading related content...

Figure