Why the magnets sit inside the rotor
Nearly all traction machines use an interior permanent magnet topology — magnets buried in slots within the rotor lamination stack rather than bonded to its surface. The arrangement is usually a V, a double V, or a delta, and the choice is not stylistic.
For the magnet supplier this topology means simple geometry — rectangular blocks, usually, rather than arcs — and demanding everything else. The parts are unremarkable to make and difficult to specify, because the requirement is dominated by what happens to them thermally and magnetically once installed.
Lower-speed and lower-power traction machines, and many auxiliary drives, still use surface magnets with a retaining sleeve. The magnetic requirements are similar; the mechanical ones differ substantially, and the sleeve introduces the eddy loss problem discussed later. The general motor guide covers the topology comparison.
The thermal case is the design case
Everything difficult about a traction magnet follows from where it sits. The rotor is surrounded by an air gap, which is an excellent thermal insulator, and it is spinning, so conduction paths out are limited to the shaft and whatever cooling the design provides. Meanwhile the magnet is heated from three directions.
Typical rotor magnet temperatures in a traction machine run somewhere around 150 to 180 °C in sustained operation, with transient excursions above that. Compare that with the 80 °C maximum working temperature of a standard N-series grade and the entire grade problem is visible in one comparison.
What temperature does
Two things degrade with heat, and they matter differently. Remanence falls at roughly −0.11 to −0.12% per °C for NdFeB, which costs torque but is reversible — it comes back on cooling. Intrinsic coercivity falls much faster, around −0.5 to −0.6% per °C, and that is the dangerous one, because it raises the knee of the demagnetization curve toward the operating point. The derating calculator works the numbers for a specific grade and temperature.
The peak magnet temperature frequently occurs not under load but shortly after shutdown, when the machine stops, the coolant pump stops, and stored heat redistributes into a rotor that has no way to reject it. A thermal model that only covers driving cycles can miss the actual maximum by a wide margin.
Demagnetization at peak and at fault
Grade selection is not set by the normal operating point. It is set by the worst combination of temperature and opposing field the magnet will ever see, and in a traction machine that combination arrives under fault.
| Condition | Opposing field | Temperature | Design status |
|---|---|---|---|
| Continuous rated | Modest | High and sustained | Comfortable if the grade is right |
| Peak torque | High d-axis current | Rising | Must be checked explicitly |
| Deep flux weakening | Deliberately demagnetizing | High — occurs at high speed | Frequently the sizing case |
| Three-phase short circuit | Very high transient current | Whatever it was when the fault hit | The classic worst case |
| Inverter fault / uncontrolled generation | High | High | Must not destroy the machine |
The short-circuit case governs many designs. A fault produces a large transient current whose field opposes the magnets, at whatever temperature the rotor happened to be. If any part of any magnet crosses its knee, that part loses output permanently and the machine comes back with reduced torque, added imbalance and distorted back-EMF.
The corners go first
Demagnetization is not uniform across a magnet. The demagnetizing field concentrates at the corners and edges nearest the air gap and nearest the flux barriers, so the loss begins as a small region at one corner while the bulk of the magnet is unaffected. That is why a finite-element check of the worst-case field distribution matters more than a single averaged working point, and why localised remedies — chamfering a corner, adjusting a bridge, using a higher-coercivity grade only in the exposed segment — can be more efficient than upgrading the whole magnet.
A machine that has lost a corner of one magnet still runs. It shows slightly lower torque constant, a distorted back-EMF waveform, higher cogging and an unbalanced pole pattern. Those symptoms get attributed to the control system, the sensor or the build far more often than to the magnets. Measuring back-EMF harmonics before and after a fault event is the fastest way to identify it.
Coercivity without the dysprosium bill
Historically, high-temperature coercivity meant dysprosium or terbium substituted throughout the alloy. Heavy rare earths are the scarcest, most concentrated and most price-volatile part of the magnet, and loading several percent by weight into every magnet is expensive in both money and supply exposure.
Grain boundary diffusion changed the economics. Heavy rare earth is applied at the surface of a sintered part and diffused along the grain boundaries, concentrating it where coercivity is actually determined rather than distributing it through the grain interiors where it mostly reduces remanence. The result is comparable coercivity for a fraction of the heavy rare earth content, with less remanence penalty.
| Route | Heavy RE content | Remanence impact | Constraint |
|---|---|---|---|
| Bulk Dy/Tb alloying | Highest | Notable reduction | Cost and supply exposure; the legacy approach |
| Grain boundary diffusion | Substantially lower | Small | Diffusion depth limits section thickness — thin parts only |
| Fine-grain processing | Reduced | Minimal | Process-intensive; producer-specific capability |
| Samarium cobalt | None | Lower Br throughout | Excellent hot behaviour; too little torque density for most traction |
| Ferrite / rare-earth-free | None | Far lower | Viable only with a machine redesigned around it |
The constraint on grain boundary diffusion is worth carrying into the design phase, because it is geometric. The heavy rare earth diffuses inward a finite distance, so the technique works well on parts that are thin in at least one direction and progressively less well as sections thicken. A magnet designed without that in mind may not be producible by the route you were counting on to hit the cost target. Ask early — it can influence how a magnet is segmented and how thick each piece is.
The rare-earth-free and reduced-dysprosium guide covers the wider set of options, including the machine-level changes that reduce magnet dependency rather than optimising it.
Segmentation and rotor loss
Sintered NdFeB conducts, so the magnets carry induced currents driven by every harmonic the rotor sees — slotting, MMF harmonics, and inverter switching. The resulting heat appears inside the magnet, which is the worst possible location given the thermal picture above.
Because eddy loss scales with the square of the dimension across the current path, dividing a magnet into segments reduces loss roughly with the square of the segment count. Two segments take loss to around a quarter in the idealised case; four to a few percent. Real machines realise less than that because of end effects and finite insulation, and the practical optimum for traction rotors usually lands between two and six segments axially.
Eddy currents and rotor losses works through the physics, the loss expression and the non-magnet levers — skew, slot and pole combination, magnetic wedges and switching frequency — several of which are cheaper than segmenting if the machine is not yet frozen.
Holding them in at speed
Traction rotors run to 15,000–20,000 rpm and beyond, and centrifugal load on a rotor magnet scales with the square of speed. In an interior topology the lamination bridges carry that load, which makes bridge design a coupled magnetic and structural problem: thick bridges are structurally comfortable and leak flux, thin ones are magnetically efficient and are the fatigue-critical feature of the rotor.
That last point connects directly to the boundary question of what is bought and what is done in house. Traction volumes are exactly the case where buying unmagnetized blanks and magnetizing the completed rotor is standard practice, because it makes automated placement tractable and improves pole consistency. The make versus buy guide covers the capital and fixture cost that decision carries, and retention design covers the mechanical calculation.
A magnet in a rotor pocket is not in a benign environment. Coolant leaks, condensation during thermal cycling and fretting against the lamination all reach it, and a corroding magnet in a rotor cannot be inspected or replaced without a teardown. Specify the coating for the real environment rather than for the assumption that the pocket is sealed — see coatings compared.
What sourcing has to carry
Traction magnets concentrate every difficult feature of the category into one part number: high volume, long qualification, heavy rare earth content, a concentrated supply base and an active regulatory environment. The engineering decisions above all have sourcing consequences.
The recycled-content point is worth flagging early to anyone writing a sustainability commitment. Recycled and blended feedstock lands most comfortably in standard coercivity ranges; the high-coercivity, heavy-rare-earth grades a traction rotor needs are the hardest to produce from secondary material. Committing to a recycled percentage across a programme without checking it against the traction part specifically is a commitment that may not be deliverable — rare earth magnet recycling covers what is realistically available.
For the commercial mechanics, price indexing covers writing the material component into a contract, supply risk monitoring covers the licensing and origin exposure, and second-source qualification covers what an alternative actually costs and how long it takes.