Skip to content

Recommended parameters for robustness

Yvan Fournier edited this page May 20, 2021 · 2 revisions

When setting up a computation, the following options are recommended for a good robustness, especially on meshes using tetrahaedra or distorted elements.

Default combinations may be more precise, and others even more robust (but more imprecise), but the following are reasonably good "all-around" settings.

Turbulence model

Depends on the computation and modeling requirements, but the most robust options are usually:

  • k-ω SST with "all y+" wall law.
  • k-ε linear production with 2 scale wall law.

Gradient reconstruction

  • Green-Gauss with iterative handling of non-orthogonalities is recommended.
    • If this fails, using a least-squares based gradient with extended neighborhood may be an alternative
    • Green-Gauss based on least-squares face values interpolation behaves similarly to that with iterative handling of non-orthogonalities, and can be made a bit smoother using an extended neighborhood, but have been observed to be less robust on a case with stratification (though it seemed more robust on others)

Other options

  • Handling of transposed gradient: off.
  • Pseudo-coupled velocity-pressure solver: off.
  • Improved pressure interpolation: on (by default starting with v7.0).
  • Convective scheme: SOLU (all variables, except for turbulence, which can be upwind).
  • Outlet condition: imposed pressure outlet.

For a steady case, use the "steady" (idtvar = 2) time scheme, with maximum CFL of 1.0.

  • prior to version 7.0, this was referred to as "local time step" (with a nother scheme named "steady" which has been deprecated).