[alpha23]

Hi all,

I am dealing with a simple turbulent case: wind tunnel experiment to be reproduced in OF, a simple cube laying on ground. Reynolds is quite low (37000), assigned velocity profile of variables U, k and epsilon at inlet, no slip walls for ground and cube, symmetry on lateral and upper walls, inletoulet with assigned pressure at outlet.

I have a TET mesh with a good quality (see attached checkMesh log). My yplus is in the range (20-200), I am using for k and epsilon standard wall functions on no slip walls:
k => type kqRWallFunction; value uniform 0.3;
espilon => type epsilonWallFunction; value uniform 0.006;
nut => type nutkWallFunction; value 0;

I am uning simpleFoam solver.

My problem is the following: for both kEpsilon and RNGkEpsilon first order Scheme is converging, when I shift to a second order one (typically Gauss linearUpwind for divSchemes), restarting from the converged solution, it spikes, than starts to drop but reaches an intermediate value for residuals (around 1e-4) and starts to oscillate around that value never diminuishing.

You can find fvSchemes and fvSolution attached.

Additional info. I have experimental data (U W velocity components profile ) to compare with solution. kEpsilon converged solution (first order obviously) is decent but the wall zone is not catching the correct behaviour, in particular in the recirculation zone after the CUBE. RNGkEpsilon behaves better (always first order) catching the correct wall value. But I want to compare second order results, and I am not getting why I have this oscillatory behaviour.

Any advice? Thanks a lot in advance

BlueSkies,

Alex

[karlli]

Hi Alex,
In my limited experience, I would suggest that you try one of the strictly TVD schemes (eg. limitedLinear) for the turbulent quantities. I have found this more stable than the linearUpwind scheme (even with cell limiters).

I did not quite understand the different combinations in your fvSchemes-file, but I would also suggest that you only use limiters on the variables where it is necessary (i.e. not as default).

Best regards,
Karl