Maxwell 3D AC Magnetic: impedance boundary + anisotropic permeability + winding

[lalvarez]

Hello,

I am modeling stray losses in a power transformer under zero-sequence (homopolar)
excitation in Maxwell 3D, AC Magnetic (eddy current) solver, AEDT 2026 R1. The model
requires, at the same time:

1. Tank walls represented with the nonlinear impedance boundary condition (NLSIBC),
   since meshing the solid steel at 50 Hz skin depth is not practical.
2. Current-driven windings (winding setup with coil terminals), with per-segment
   excitation.
3. Laminated GO magnetic wall shunts, whose physically correct representation is an
   anisotropic nonlinear permeability (B-H curves in the lamination plane, collapsed
   permeability in the stacking direction).

The Maxwell help topic "Impedance Boundary" states that when a project includes an
impedance boundary condition, a material with anisotropic permeability and a winding
at the same time, it cannot be solved by the eddy current solver nor the 3D transient
solver. I have tested this systematically on 2026 R1 with a minimal synthetic model
(copper ring + coil terminal + stranded current winding; steel_1008 plate with
nonlinear impedance boundary, solve inside off; one test block whose material is the
only thing that changes):

- Block with **linear** anisotropic permeability (constant tensor 5000 / 5000 / 25):
  **solves normally**.
- Block with **nonlinear** anisotropic permeability (B-H curves per tensor component,
  different curve in the stacking direction): **solver aborts** with:

  > Maxwell 3D eddy current solver does not support the project when the following
  > conditions are applied at the same time in one design: 1. Impedance boundary
  > condition 2. Anisotropic permeability material 3. Winding setup

- Block with an anisotropic property type but **three identical B-H curves**
  (isotropic de facto): **also aborts** with the same message, so the check appears
  to be triggered by the declared property type (AnisoProperty + nonlinear), not by
  the actual values.

Questions:

1. Can Ansys confirm that the restriction applies only to **nonlinear** anisotropic
   permeability, and that linear anisotropic tensors are fully supported in this
   combination? The help topic does not make this distinction, and it matters for
   choosing a modeling strategy. If linear anisotropy is supported, is the solution
   accuracy guaranteed, or does the solver silently degrade anything?
2. Is there a roadmap to remove this limitation in an upcoming release?
3. What is the recommended workflow for laminated magnetic shunts when the tank must
   stay as NLSIBC and the excitation must be a winding? The options I see are:
   (a) linear anisotropic tensor for the shunt (no saturation), optionally iterating
   the in-plane permeability to a secant value at the operating point;
   (b) isotropic nonlinear B-H for the shunt plus a discrete air gap between shunt
   and wall sized as g = t_shunt · (1 − F) to restore the missing through-thickness
   reluctance of the stack;
   (c) replacing the winding with imposed current excitations, which does solve with
   anisotropy, but loses the winding/circuit coupling.
   Is any of these the intended path, or is there a better supported approach
   (e.g. lamination composition with stacking factor — does the lamination model
   trip the same check)?
4. Minor related issue: is the restriction expected to behave identically in 3D
   transient with the nonlinear impedance boundary introduced in 2021 R2?

Setup details: AEDT 2026 R1, Maxwell 3D, AC Magnetic, 50 Hz, adaptive setup. The
minimal reproduction project is a few MB and I can attach it if useful.

Thank you.