Axial-Flux Rotor Magnet Demagnetization Temperature Selection: Temperature Is The First Subject of Selection

The most common mistake in magnet selection for axial-flux motors is to start from the grade table: first look at remanence, then at the maximum operating temperature, and pick a grade that is “good enough” to put into the BOM. But in an axial-flux structure, this order is exactly reversed. The rotor magnet temperature is not determined by the magnet itself; it is imposed on the magnet by the topology, the cooling, and the operating conditions. The real starting point of selection is to ask clearly: under the worst case in which short circuit, overload, and hot spots superimpose, how hot will the magnet become, and can it still withstand the reverse magnetic field at that temperature?

The Demagnetization Risk of Axial Flux Is “Written Into” the Structure

The selling point of axial-flux motors is clear: at the same power, weight and axial length can be halved compared with radial motors. But this compactness comes at a thermal cost.

In a double-stator single-rotor topology, the rotor is sandwiched between two stators, so the direct contact area between the rotor and the air is greatly reduced, and the heat-dissipation path is inherently narrow. At the same time, the magnets themselves generate eddy-current losses in the alternating magnetic field, and this heat is released directly inside the permanent magnets. The thermal radiation from the stators superimposed on the self-heating of the rotor makes it easy for the magnet temperature to approach or even cross the dangerous boundary under continuous high-power conditions. A recurring judgment in industry analyses is that the sandwich structure and high power density of axial-flux motors lead to extremely low thermal capacity, making the rotor magnets “highly prone to overheating.”

Therefore, magnet selection for axial flux cannot simply copy the experience of radial motors. The intuition in radial motors that “the rotor is cooled through the shaft and the air at the ends” basically fails in a disc-type topology.

What Actually Happens in Demagnetization: The Boundary Between Reversible and Irreversible

To understand selection, one only needs to grasp two temperature concepts.

Reversible demagnetization is the normal “thermal decay” of a magnet. As temperature rises, remanence decreases linearly, and it recovers after cooling, without permanent damage. NdFeB mainly operates in this range below 120°C. The problem appears above 120°C: the slope of reversible decay becomes more significant, and more critically, the boundary of irreversible demagnetization is pushed closer.

The physical criterion for irreversible demagnetization is the knee point. The demagnetization curve has an inflection point in the second quadrant. Operating points above the knee point are safe; once an external reverse magnetic field pushes the operating point below the knee point, the magnet cannot return after the external field is removed, and the remanence suffers permanent loss. The consequence of rising temperature is twofold: remanence decreases, and at the same time the knee point moves to the right (toward lower magnetic field), meaning that the same reverse magnetic field is more likely to cross the knee point at high temperature. For NdFeB, engineering practice generally treats around 150°C as the critical region for irreversible demagnetization; above this level, the risk of complete loss of magnetic performance is no longer a negligible small-probability event.

The letter suffix of a magnet grade essentially specifies the temperature at which the knee point can still remain far enough to the right. Common correspondences are: SH corresponds to about 150°C, UH to 180°C, EH to 200°C, and AH to 220°C. But there is a point that is very easily misread: the temperature indicated by the suffix is the temperature resistance of the material under specific permeance coefficient calibration conditions, not the “maximum operating temperature” that the magnet automatically obtains after being installed in the motor. The same UH magnet may run continuously at 130°C in a well-cooled prototype, but in a mass-production part with insufficient sandwich cooling, the hot spot may directly surge above 170°C.

Samarium Cobalt: The Fallback Option, but the Cost Must Be Calculated Clearly

When the temperature margin of NdFeB is truly insufficient, SmCo is an unavoidable alternative in engineering. SmCo has significantly higher temperature resistance; many grades can be used in applications at the 350°C level, and the Curie temperature is also in the 700–850°C range, far higher than the 310–350°C of NdFeB. In the high-temperature region, the slope of the irreversible demagnetization curve of SmCo with temperature is also much flatter than that of NdFeB.

But the cost of SmCo must be put on the table at the same time. Its typical remanence is in the 0.95–1.1 T range, while NdFeB can reach 1.1–1.4 T; the maximum energy product is usually 25%–30% lower than that of NdFeB. This means that to achieve the same air-gap flux density, SmCo magnets need to be thicker and heavier, which in turn encroaches on the already tight axial space of axial-flux machines and may even increase rotor inertia and eddy-current losses. SmCo is also more expensive, and its supply chain maturity and delivery flexibility are not as good as those of NdFeB. Therefore, unless the thermal check clearly indicates that “even NdFeB upgraded to EH/AH cannot hold,” using SmCo to “buy” temperature margin is often a transaction paid three times over in weight, cost, and power density.

The Correct Order of Selection: Infer the Grade From the Magnet Temperature

An actionable temperature-based magnet selection process for axial flux should follow the order below, rather than starting from the grade table.

Step 1: Define the cooling first, then discuss the grade. The rotor cooling method of an axial-flux machine directly determines the “floor” of the magnet temperature. Air cooling is simple and has low parasitic losses, but the thermal conductivity and specific heat capacity of air are limited, and the cooling capability is strongly dependent on speed and environment. Oil cooling is highly expected in axial flux; experimental data show that a reasonable oil-cooling scheme can reduce the maximum temperature by 15–20°C further than pure water cooling, but oil cooling “carrying heat away” does not equal “eliminating heat.” The temperature difference between the edge and the center of a disc-type structure can still reach 30–50 K, and the local hot spot is the object of demagnetization verification. Before the cooling scheme is frozen, any grade selection lacks a reliable premise.

Step 2: Put the short-circuit condition into the temperature check. The temperature under normal operation is only the baseline. Demagnetization verification must cover the reverse magnetic field generated by a controller fault or winding short circuit, and this check must be performed at the highest temperature the magnet may reach under that condition. The design criterion in the literature is straightforward: the magnet thickness should be selected to avoid irreversible demagnetization after a three-phase short-circuit current at the lowest temperature. More generally, it is necessary to confirm that the operating point does not fall below the knee point at the superposition of the worst temperature and the worst demagnetizing field.

Step 3: Use “actual magnet temperature + short-circuit demagnetizing field” to check the grade suffix. If the magnet hot-spot temperature given by the cooling scheme and electromagnetic design is on the order of 140–150°C, and the short-circuit demagnetizing field is not severe, SH may enter the candidate list; once the hot spot approaches or exceeds 160°C, UH is a more reliable starting point; when the hot spot reaches above 180°C, EH or even higher grades must be seriously evaluated, or the cooling design itself must be reconsidered. The logic here is: the suffix is not a “performance level,” but a “capability level for preserving the knee point at a given temperature.” The cost of higher-suffix grades (UH/EH) is usually slightly lower remanence or a higher heavy-rare-earth addition; it is not free insurance.

Step 4: Use rotor segmentation and magnet geometry to “lighten the load” on temperature. Magnet selection is not an isolated material decision. The eddy-current loss of axial-flux rotor magnets is directly related to the eddy-current path length and surface area on the magnet surface, and permanent-magnet segmentation is an effective means of reducing eddy-current loss and suppressing temperature rise. When the grade has already been determined and the temperature margin is tight, reducing the magnet’s own heat generation through segmentation optimization is sometimes more economical than simply raising the grade suffix by one level, and it is also more conducive to preserving torque density.

A Practical Judgment Line

If this article is to be compressed into one sentence to bring into a design review, it is: the grade of an axial-flux rotor magnet should not be “selected”; it should be “calculated”—first use the cooling and short-circuit conditions to calculate what temperature the magnet will reach, then use that temperature to look up the knee point, and finally settle on the suffix.

For NdFeB, 120°C is the observation point at which reversible decay begins to become significant, and around 150°C is the region where irreversible demagnetization risk enters the level that must be seriously verified. These two numbers are not the “maximum operating temperature” in the grade table; rather, they remind designers that in the sandwich structure of axial flux, the temperature margin disappears faster than in radial machines. Treating temperature as the subject of selection rather than the object of the grade is what keeps demagnetization from turning from a material problem into a systemic reliability gap.