Axial Flux Studio / Field guide
Follow the heat
From loss inputs through cooling paths to a temperature prediction—and its limits.
Printing keeps the current diagram selections and checklist state. Select the views you need before printing or saving as PDF.
On this page
Read the calculation as a chainSeparate inputs from predictionsCompare the same testNewer development evidenceWhat remains openWhat to keep with a temperature result01Read the calculation as a chain
A temperature prediction joins three things: heat generated, paths that carry it, and boundaries that remove it. Changing a contact, coolant inlet or loss distribution can change the hot spot even when total watts stay the same.
The diagram above is a conceptual heat path, not a solved temperature field. Select a heat source to see which assumptions to inspect. Real machines have several parallel heat paths.
02Separate inputs from predictions
- Material-temperature input: the temperature used for electromagnetic material properties. It is not a predicted operating temperature.
- Loss input: watts supplied to the thermal model, from an electromagnetic calculation, an engineering model or a prescribed heating test. Preserve that origin.
- Steady temperature: a prediction after heat generation and removal balance under the chosen boundaries.
- Transient temperature: a prediction over time, also dependent on initial temperatures and thermal masses.
- Coupled calculation: an outer iteration that feeds representative temperatures back into electromagnetic properties. Its convergence is separate from that of each field or thermal solve.
An internally balanced heat account does not prove that loss generation, contact conductances or cooling coefficients are physically correct.
03Compare the same test
The Ghent 4 kW reference's documented lumped-network winding prediction is 61.1 °C, against 62.7 °C in the 100 W DC, 1000 rpm test from Energies 11(11):2983, Table 3. This is a prescribed-heating comparison; it does not validate electromagnetic loss generation.
Another published campaign reports 56.0 °C for the same nominal test. The source notes retain that disagreement. For the 2983 magnet readings, the documented network predicts 30.2 / 28.4 °C against 30.5 / 28.1 °C at 1000 / 2000 rpm. The separate 2019 loaded-magnet campaign remains unexplained by the same thermal path. These are different tests, not interchangeable accuracy scores.
Read the Ghent dossier with its cooling and calibration qualifications. In any comparison, keep ambient, speed, heating, sensor location, test duration and temperature versus temperature-rise conventions together.
04Newer development evidence
Documented AF50 imported-machine thermal route check. At 28.3 A on q and 1000 rpm, the documented structured and tetrahedral electromagnetic routes give:
| Quantity | Structured route | Tet route |
|---|---|---|
| Torque | 0.5986 N·m | 0.6003 N·m |
| Winding mean | 207.62 °C | 207.62 °C |
| Winding maximum | 210.39 °C | 210.40 °C |
| Magnet mean | 41.77 °C | 41.50 °C |
| Total heat in / out | 81.05 W | 81.01 W |
Each route's heat input and output agree to the precision shown. The two columns give the balanced total for each route, not two measured heat flows.
These are internal route comparisons, not hardware measurements or a qualified operating rating. The winding temperature is approximately 208 °C at this chosen point; agreement between routes does not establish an acceptable continuous load.
Both routes use the same structured conduction model, shared copper loss and cooling films. The tet field supplies the loss inputs; its stator core loss is placed spatially onto the conduction grid. Other losses are distributed by volume. This is not a thermal solve directly on the tet mesh.
The retained report includes mesh warnings and the qualifications of the imported-model route. Where a structured conduction grid cannot be built, the tet route reports no temperatures and gives a reason. A missing temperature is not zero.
05What remains open
For the tet route, the coupled temperature-feedback loop and uncertainty band are not implemented in this record. Copper field-driven eddy heat is not yet placed from the local tet field. The slice route's temperatures remain open too.
The wider thermal programme also retains contact, cooling, calibration and measurement disagreements. A good fit can conceal an unidentified parameter or a missing heat path; fitting on a test is not an independent prediction of that same test.
06What to keep with a temperature result
Record the machine and route, loss sources and totals, heat-placement method, ambient or coolant inlet, cooling and contact assumptions, initial conditions if transient, and whether the value is a mean or a maximum. Preserve transfer warnings and energy-balance diagnostics.
Then ask separately: did it finish, did it converge, did it conserve the supplied heat, and has this quantity been compared with a relevant measurement? See results with context or prepare a solver-result report on Support.