A detailed stress map does not become a core material model simply by taking its mean. Magnetic response can depend nonlinearly on stress, orientation and operating point. Evaluating a law at the average stress can therefore differ from averaging the responses calculated from the local states.
Start with the scale of the observation and the scale of the model element. A diffraction gauge volume, surface strain map and microstructural calculation sample different physical regions. Their values cannot be merged without defining what each represents.
Write the transfer rule explicitly
For an illustrative scalar response f, the average of f(sigma) need not equal f of the average sigma. This statement describes nonlinear averaging; it does not prescribe the correct homogenization for a core. The actual response may require a stress tensor, material direction, magnetic history and local field rather than one scalar stress.
Record the weighting used to transfer observations. Volume weighting, diffraction signal weighting and magnetic-energy weighting are different choices. Use a rule connected to the observable and model question, not whichever produces the easiest dataset.
| Transfer stage | Information preserved | Check before using it downstream |
|---|---|---|
| Local observation | Sampled region, directions and uncertainty | Is it representative of the element? |
| Material response evaluation | Local stress and magnetic state | Is the law valid for those states? |
| Averaging or homogenization | Weighting and boundary assumptions | Is nonlinear behavior retained adequately? |
| Core-scale input | Effective law and validity region | Does the element interpretation match the solver? |
| Independent load case | Observations excluded from fitting | Does the transfer predict a new state? |
This record makes uncertainty visible at each transition instead of hiding all of it in one fitted parameter.
Compare two transfer hypotheses
Evaluate a mean-state approximation and a distribution-aware alternative using the same supported input data. Compare their predictions under a load state not used to choose the transfer parameters. If both reproduce that state within uncertainty, the simpler representation may be sufficient for that question.
If the predictions differ, identify which regions and response quantities are responsible. Correlation between local field and stress can matter: a heavily stressed region carrying little relevant flux may contribute differently from the same region at a higher local magnetic demand.
Keep coordinates and histories connected
Transform local stress into the same material frame used by the magnetic law. Preserve the relation between rolling direction and the model’s geometry. A mean tensor in global axes cannot replace orientation-resolved states without testing that simplification.
Separate calibration from validation. Agreement with the observations used to fit the effective law establishes a fit, not predictive transfer. Retain an independent mechanical state, specimen region or relevant excitation comparison wherever the model will support a decision.
The resulting model package should include the transfer rule, its demonstrated range and the unresolved scale assumptions. Its value is a verified relationship between local evidence and the core calculation. A dense micro-stress map without that relationship can add apparent detail while leaving the model’s actual decision accuracy unchanged.
Technical references
- Materials / MDPI. Magnetostrictive Properties of the Grain-Oriented Silicon Steel Sheet under DC-Biased and Multisinusoidal Magnetizations.
- NIST. NCAL: Diffraction Stress Measurement Under Applied Load.