Axial Flux Studio / Field guide
Read results with context
Compare the same quantity, then inspect the evidence.
Printing keeps the current diagram selections and checklist state. Select the views you need before printing or saving as PDF.
On this page
Start with provenanceTorque needs angular resolutionMaps, envelopes and exportsInspect before you reportRead the retained exampleCheck convergence separately from agreementKeep a comparison record01Start with provenance
Read the reported tier, quality information and stale-state warnings alongside each number. A general uncertainty bound is not available for every route or operating point; do not infer one from a successful solve. Something that was not measured is left blank, never treated as zero.
Keep the design, tier, grid, operating point, temperature and study type with any number you quote. A loaded torque and a no-load cogging result answer different questions, and a peak voltage is not a phase RMS voltage.
Results lists stored runs, flux maps and sweeps with their quality verdicts. Compare warns when two runs are not comparable; check that before reading a difference as a design improvement. Agreement with a reference machine applies to that machine, its inputs and that quantity.
02Torque needs angular resolution
Mean torque, peak-to-peak ripple and cogging need different treatment. Small ripple can be comparable to grid or sampling noise. Inspect the waveform and convergence, and increase angular coverage or resolution when needed.
A mean d-q flux map cannot hold the angle-dependent harmonics behind ripple. Use an angle-resolved solve rather than inferring a ripple waveform from a map.
03Maps, envelopes and exports
Capability's envelopes and efficiency maps follow the tier of the flux map they read. Treat them as indicative above base speed. Losses are partly estimated, so read efficiency as indicative too.
The lookup-table export is consistent by construction at the grid points. It states its interpolation error, its reciprocity and how far the inductance tables are from the linkage tables' slope. Keep those figures with the tables.
04Inspect before you report
Visualize links waveform positions to field views. Use slices, local fields and radial distributions to investigate unexpected peaks, rather than relying on one attractive colour plot.
Report builds a standalone HTML record with plots and quality verdicts. Include the model assumptions and any unresolved warnings when you share it. A report communicates a calculation; it does not turn an unvalidated operating point into a certified rating.
05Read the retained example
The annotated capture above shows stored run 98, a native 3-D study of a one-sided 14-pole / 12-slot machine at 3,000 rpm, 40 A RMS and 49 rotor positions. The current working-design header belongs to a different machine: use the stored run identity when describing the result.
Start at the caveat banner, then read the quantity and units, then the angular sampling. These three checks help you decide what information is still needed. The full warning list, mesh refinement, temperature and software revision are not supplied by this capture record. An unreadable or absent condition is not permission to substitute one from another run.
06Check convergence separately from agreement
Convergence asks how the numerical result changes as the mesh, angular sampling or solver tolerances are refined. Reference agreement asks how a calculation compares with a specific published or measured quantity under matching conditions. Neither is a substitute for the other.
For a refinement comparison, retain the same machine, operating point, materials and conventions while changing the numerical setting under investigation. Compare the quantity you intend to report: a stable mean torque does not establish stable peak-to-peak ripple. Read solver residuals alongside physical outputs; a small algebraic residual is not by itself a bound on discretisation error.
The 49 positions in the retained capture describe its sampling; the capture does not provide a refinement series. No convergence percentage or uncertainty interval is assigned to it here.
07Keep a comparison record
| Keep with the result | Check before comparing |
|---|---|
| Stored run and machine identity | The working design header may have changed since the run |
| Solver route, version and numerical settings | Match the route and record mesh, tolerances and angular coverage |
| Speed, current, temperature and drive assumptions | Use the same physical operating point; state missing conditions |
| Quantity, units and convention | Phase/line, RMS/peak and mechanical/electrical angle must agree |
| Quality verdicts, warnings and source | Preserve unresolved caveats and distinguish simulation from measurement |
When the app flags runs as non-comparable, investigate that reason before presenting the difference as an improvement. For supported prototype comparisons, use the evidence register and its conditions rather than treating this application screenshot as validation data.
This guide uses a retained development capture and its screenshot record. No new solver campaign was run for the illustration.