Troubleshooting RANS Turbulence Solvers

This section provides guidance for fixing problems with Reynolds-Averaged Navier-Stokes (RANS) turbulence solvers.

Solution of the transported turbulence variables diverges

The residual for any of the transported turbulence variables increases significantly after the startup transient, indicating that the solution of the turbulence field is diverging.

Make sure the velocity field is not diverging and hence causing the turbulence model to diverge.
Slightly reducing the turbulence solver under-relaxation factor (say to between 0.3 and 0.5) might stabilize the solution.
In situations where mesh quality is impeding convergence, using the first-order convection scheme might help.

Solution of the transported turbulence variables converges slowly

The flow residuals are converging well, but the residuals for the transported turbulence variables are not decreasing as fast.

You can increase the turbulence solver under-relaxation factor if you are confident that the solution is otherwise progressing well, but a value of more than 0.9 is not recommended, since it might result in solution divergence.

Unrealistically high turbulent viscosity impedes flow solution

During the first few iterations, unrealistically high turbulent viscosity values are impeding the flow solution.

Reducing the turbulence solver under-relaxation factor will retard the update of turbulent viscosity.
It is possible to make this factor quite small (say 0.01) to cause only a minimal update, but this would typically be done for a few iterations only.

Warnings of limited turbulent viscosity occur

Repeated warnings occur that the turbulent viscosity is being limited in some cells.

This is both common and acceptable during the first few iterations due to startup transients.
If the warnings continue to occur as the flow solution nears convergence, it is possible to increase the Maximum Ratio to prevent the warnings.
Generally one would only increase the Maximum Ratio if the simulation has a very high Reynolds number and no errors seem to be present in the simulation setup. The turbulent viscosity ratio is proportional to a turbulent Reynolds number, where the velocity scale is typically given as $k^{1 / 2}$ and the length scale as $(k^{3 / 2}) / ε$ (for a model based on the eddy viscosity concept). If you expect very high turbulent Reynolds numbers (which occur when the mean flow Reynolds number is high, such as in a planetary boundary layer), you could try increasing the Maximum Ratio limit by one or two orders of magnitude. If you do not expect high turbulent Reynolds numbers, it is likely that another problem is present in your simulation, such as an incorrect boundary condition, or a mesh that hasn’t been scaled correctly.