Selection Guide for Eddy Current Suppression Schemes for High-Speed Rotors with Segmented Magnets
You are here: Home » Blog » Blog » Industry Information » Selection Guide for Eddy Current Suppression Schemes for High-Speed Rotors with Segmented Magnets

Selection Guide for Eddy Current Suppression Schemes for High-Speed Rotors with Segmented Magnets

Views: 0     Author: Site Editor     Publish Time: 2026-11-20      Origin: Site

Inquire

facebook sharing button
twitter sharing button
line sharing button
wechat sharing button
linkedin sharing button
pinterest sharing button
whatsapp sharing button
kakao sharing button
snapchat sharing button
sharethis sharing button

Rotor eddy current loss in high-speed permanent magnet motors is an unavoidable engineering issue. As the electrical frequency rises and the harmonic content increases, eddy current loss in the permanent magnets climbs at a rate approximately proportional to the square of the frequency. The direct consequences are local temperature rise in the magnets, an increased risk of irreversible demagnetization, and continuous efficiency loss. Physically “cutting off” the eddy current paths—that is, dividing a solid magnet into several mutually insulated segments—is currently the most direct and mature suppression method in engineering practice. However, “segmentation” is far from a simple instruction: which direction to segment in, how many segments to use, how to insulate them, and how to balance loss reduction against manufacturing cost all constitute a selection problem that requires systematic judgment.

精密金属轴零件产品图.png

1. Start with the Physics: Does Your Magnet “Need to Be Cut”?

Before discussing a segmentation scheme, it is necessary to determine a preliminary question: whether the eddy current distribution inside the magnet is already significantly governed by the skin effect. The skin depth δ is inversely proportional to the square root of frequency. In the low- and medium-frequency range, the magnet dimensions are usually much smaller than the skin depth, and eddy currents are approximately uniformly distributed across the entire cross-section. In this case, segmentation in any direction can effectively cut off large loops, and the loss decreases monotonically as the number of segments increases. In high-frequency scenarios, however, the skin depth may shrink to a level comparable to or even smaller than the magnet dimensions. The current distribution inside the magnet is no longer uniform, and eddy currents tend to concentrate near the surface and end regions. At this point, the logic of segmentation undergoes a qualitative change.

A study on high-frequency, high-power-density permanent magnet motors provides a conclusion worth noting: at low frequencies, segmentation does reduce loss, but after entering the high-frequency range, increasing the number of segments may initially cause the loss to rise; only after the number of segments continues to increase to a certain level does the loss fall again. This non-monotonic “first rise, then fall” behavior originates from magnetic flux distortion and local eddy current redistribution near the segment interfaces at high frequencies. This means that high-frequency applications cannot simply copy the low- and medium-frequency intuition that “more segments are better”; they must be evaluated through electromagnetic field calculations based on the specific electrical frequency, magnet dimensions, and skin depth.

A practical preliminary criterion can be summarized as follows: first estimate the skin depth in the magnet material under the target operating conditions. If δ is of the same order as or smaller than the characteristic magnet dimension (thickness or pole arc width), the design has entered the “high-frequency segmentation-sensitive zone,” and the segmentation strategy must be re-examined using simulation or analytical models. If δ is significantly larger than the magnet dimensions, the conventional segmentation logic can be followed.

2. Direction Selection: Axial or Circumferential? The Answer Depends on the Intersection of Frequency and Process Constraints

Direction selection is the step most easily simplified to “choose whichever works better,” but in actual engineering there is no absolute advantage independent of operating conditions.

In low- and medium-frequency scenarios with a small number of segments, axial segmentation often shows higher “per-segment efficiency.” A three-dimensional finite element analysis of a large surface-mounted permanent magnet motor showed that when the number of segments does not exceed four, axial segmentation is more efficient than circumferential segmentation in reducing eddy current loss. A hub motor case also provides a quantifiable comparison: the eddy current loss of a solid magnet was 7 W; with two circumferential segments it dropped to 4 W; with two axial segments it dropped to about 2.5 W, so axial segmentation was better in terms of loss value. However, the same case immediately revealed a practical engineering constraint: the motor had a short axial dimension, and after axial segmentation the magnet pieces were difficult to clamp and position, making automated magnet bonding difficult to achieve. The final production scheme therefore still adopted circumferential segmentation. The optimal loss solution is not equal to the feasible manufacturing solution—this is a principle that must be embedded in the selection process.

As the number of segments increases or the frequency rises, the advantages of circumferential segmentation gradually become apparent. The reason is that circumferential segmentation directly reduces the pole arc width of each magnet. When this width approaches or becomes smaller than the skin depth, the effective “characteristic dimension” of the eddy current loop is substantially compressed, and the suppression effect truly takes hold. For high-frequency high-speed rotors, a proven and effective strategy is a “dual-direction coordinated” scheme based mainly on circumferential segmentation, supplemented by necessary axial segmentation: first cut the in-plane eddy current loops through circumferential subdivision, then arrange a small number of axial segments to suppress end effects, thereby achieving an acceptable compromise between loss suppression and manufacturing complexity.

In actual projects, circumferential segmentation is also naturally compatible with modular magnet assembly—each pole consists of several independent magnet blocks arranged along the arc surface, and the magnet bonding process can more easily connect with existing surface-mounted rotor production lines. Axial segmentation, by contrast, imposes higher requirements on magnet thickness consistency, end-face insulation treatment, and stack assembly positioning accuracy, and is more suitable for production lines with sufficient axial dimensions and automated stacking capability.

3. Number of Segments: There Is a “Saturation Point of Benefit,” and Even More So a “Cost Cliff”

More segments are not always better—this has become a consensus from simulation to production line. The rate at which eddy current loss decreases diminishes as the number of segments increases: the first few segments often capture the largest proportion of loss reduction, after which the marginal benefit rapidly narrows. Research on propulsion motors explicitly points out that when there are too many segments, the gain is extremely limited while process difficulty and processing cost rise rapidly. Three-segment designs for hundred-kilowatt-class motors and four-segment designs for megawatt-class motors are more pragmatic choices. Another evaluation of large wind power permanent magnets points to a similar conclusion: a sufficient number of axial segments can provide a good compromise between magnet stability and loss reduction, but the threshold of “sufficient” must be determined in combination with the specific pole pitch and frequency.

During selection, it is recommended to infer the number of segments from the “loss target” rather than pursuing the lowest loss with the “maximum number of segments.” A specific approach is as follows: set the upper limit of magnet eddy current loss allowed by the rotor temperature rise or efficiency, find the minimum segmentation combination that satisfies this limit through finite element parameter scanning, and then use this combination as the baseline scheme to evaluate whether its process cost is acceptable. If the manufacturing cost of the baseline scheme is too high, only then consider further optimization in direction combination or insulation method, rather than simply increasing the number of segments.

4. Insulation and Process: Whether Segmentation “Counts” Depends on the Interface

The eddy current suppression effect of segmented magnets is ultimately determined by the electrical insulation quality between segments. If a conductive path exists between adjacent magnet blocks, eddy currents will “bypass” the segment interface and continue to flow, making segmentation electromagnetically meaningless. The splicing of segmented magnets usually employs insulating adhesives or bonding layers containing insulating substrates, so that the joint surface simultaneously performs the dual functions of structural connection and electrical isolation. The national standard GB/T 45756—2025, Technical Specification for Splicing of Sintered Rare Earth Permanent Magnets, has systematically specified the materials, adhesive combinations, mechanical processing, and surface protection for the splicing of sintered rare earth permanent magnets intended to reduce eddy current loss. This provides a citable basis for acceptance of insulation interfaces in selection.

Two variables in process details are easily underestimated. The first is the effect of the segmentation method on magnetic flux: segmentation cutting itself causes a certain loss of magnet volume and a decrease in magnetic flux. Experimental data show that for 4 to 8 equal-part non-through-cut segmentation, the volume loss increases from about 0.96% to 2.24%, and the magnetic flux loss correspondingly increases from 1.12% to 2.18%. The second is the trade-off between through-cut segmentation and non-through-cut segmentation: non-through-cut segmentation is simpler in process and lower in cost, and also has a clear suppression effect on eddy currents, but its eddy current loss is still slightly higher than that of through-cut segmentation (in one comparison, 9.68 W versus 4.95 W), with an efficiency difference on the order of one percent. During selection, judgment should be made according to how tight the loss budget is: if the thermal margin is sufficient, non-through-cut segmentation may be a more economical solution for cost-sensitive products; if the rotor temperature rise is already close to the demagnetization safety boundary of the magnet, through-cut segmentation and its insulation reliability should be given priority.

5. An Operable Selection Framework

By bringing the above dimensions together, a sequential decision chain can be formed, rather than repeated trial and error among direction, number of segments, and process.

Step 1: Perform a frequency–skin depth screening. Estimate the skin depth in the magnet at the target maximum electrical frequency. If δ is significantly larger than the characteristic magnet dimension, conventional segmentation is sufficient, and the focus should shift to process feasibility. If δ is close to or smaller than the magnet dimensions, the design has entered the high-frequency sensitive zone, and subsequent decisions must be based on electromagnetic field simulation rather than empirical formulas.

Step 2: Conduct a preliminary screening at the direction level. For low- and medium-frequency applications with few segments, sufficient axial space, and stacking capability, the loss advantage of axial segmentation can be evaluated first. For high-frequency scenarios or cases with limited axial assembly, use circumferential segmentation as the baseline scheme. In high-frequency high-power-density applications, directly consider a “circumferential-primary, axial-supplementary” dual-direction coordinated structure, and avoid excessive subdivision in a single direction.

Step 3: Infer the number of segments from the loss target. Set an acceptable upper limit for magnet eddy current loss, and find the minimum segmentation combination that satisfies the condition through parameter scanning. At the same time, pay attention to the possible non-monotonic phenomenon in the high-frequency range, and do not replace “more accurate field calculation” with “more segments.”

Step 4: Make the final decision based on insulation reliability and process cost. Among the candidate schemes, compare the flux loss of through-cut and non-through-cut segmentation, the maturity of the bonding process, adaptability to automated production lines, and life-cycle cost. The final scheme should be the intersection of electromagnetic performance, thermal safety, and manufacturing economy, rather than the optimal value of a single indicator.

The essence of eddy current suppression in segmented magnets is to actively “design discontinuities” inside the magnet. The location, density, and insulation quality of the discontinuities determine the degree to which eddy currents are weakened; the assembly complexity, flux penalty, and insulation failure risk brought by the discontinuities determine whether this method can be sustained on the production line. For high-speed rotor designers, the core capability in selection is not to remember “whether axial or circumferential is better,” but to possess a framework awareness for judging “which discontinuity is the least costly and truly effective under the current frequency, current dimensions, and current production line conditions.”

Facebook
Twitter
LinkedIn
Instagram

WELCOME

SDM Magnetics is one of the most integrative magnet manufacturers in China. Main products : Permanent magnet,Neodymium magnets,Motor stator and rotor, Sensor resolvert and magnetic assemblies.
  • Add
    108 North Shixin Road, Hangzhou, Zhejiang 311200 P.R.China
  • E-mail
    inquiry@magnet-sdm.com​​​​​​​

  • Landline
    +86-571-82867702