Eddy3D stops at ~110 iterations even when maxIter is increased (non-converged residuals)

[MK_Archi]

Hi everyone,
I’m running CFD simulations in Eddy3D for several façade panel geometries.
Most of my models run normally and reach the specified maximum number of iterations (e.g. 500 or 1000).
However, a few specific geometries always stop at around iteration 110, regardless of how high I set maxIter.

Here is the issue:

• maxIter is set to 500 or 1000, but the solver consistently stops around iteration 110
• The residual plot shows clear divergence and no convergence at that point
• The probe velocity becomes extremely unrealistic (e.g. 2.8e+9 m/s)
• The panel curvature and overall geometry are very similar to other models that run successfully
• Boundary conditions, domain size, and mesh settings are identical

Because other nearly identical geometries run completely normally, I am unsure what is causing only these specific models to terminate early.

My questions:

1. Under what conditions does Eddy3D/OpenFOAM stop iterations before reaching maxIter, even when the residuals have not converged?
2. If the geometry is almost the same as successful cases, what else could cause this kind of non-convergent behaviour?
3. Are there recommended checks to identify where the divergence occurs in these cases?
4. How can I stabilize the solver so that these geometries also run to the full number of iterations?

I have attached the residual plot, which shows the divergence around iteration ~100.

Any advice or ideas would be greatly appreciated.
Thank you!

[minggangyin]

The gunplot indicates that the OpenFOAM simulation has diverged. You need to adjust the mesh, fvScheme, or fvSolution settings to achieve convergence.

[QiH]

Check your mesh quality using checkMesh, also the fvSchemes/fvSolution as suggested above, e.g. a smaller relaxation factor may helps (not always)