bad convergence of k-omega SST model
 

Hi foamers,

I am trying to do an internal flow computation using k-omega SST model in OpenFOAM. The solver used is simpleFoam.

After reading several posts and theoretical, I am a bit lost yet...I cannot reach a convergent solution of my model.

The mesh is created with cfMesh. About 5 layers of prism cells are placed close to the wall of the bifurcated pipe. The first layer height is about 5e-05.

I have tried several BCs for k, omega and nut. I attach the main 3 options I have used.

Simulation residuals never fall below 1e-03 (for pressure it uses to be 1e-02).

So, here are my questions:

1. The yPLus value I have obtained using yPlusRAS utility (some people say that it is ystar, I have not read the code yet) is about 2-3. I want to resolve boundary layer ,so, could this be the problem? Should I go to a yplus value of 1 or below? Is this really the theretical yPlus? Can I apply wall functions in this case (I think wall function approach is better for a not well resolved mesh at the boundary, let's say yPlus between 30 and 300 or so)?

2. CheckMesh present some warnings, but I guess these come from the cells close to the wall, which are expected to be high aspect ratio cells. What about non orthogonal correctors?

3. As I am not finding a convergent solution, I started to think in solvers in fvOptions and discretization schemes in fvSchemes. I attach it too. I would like to go to second order schemes.

4. I know that my initial values could be better computed, but at the beginning the simulation residuals fall fast. The poblem is that at some point, they stay constant in a quite high value, although both flow rate and averaged outlet velocity remain constant. Do you think initial values could have influence in the convergence behaviour after several iterations?

I am using OF 2.2.2, but I will update to 2.3 soon.

Thanks in advance. I cannot solve this!

NOTE that in attached files sometimes one of the inlets is treated as a wall. This is not a mistake, I have decided to close it for some reasons, but the residuals behaviour is almost the same...

Sorry; I forgot to attach fvSolution and fvSchemes files.

I include also Residual plot, checkMesh output and meshDict (one of them) used.

Just one more thing, why I cannot compute yplus when I use something different than nutUSpalding WF or similar at the wall?

Hi Liam,

a few thoughts on some of your questions:

What is your expansion ratio. I usually start with 1.5.

Interesting topic! Yes I think it calculates ystar. There is topic about this on this forum which I can't find right now. Some member also wrote a tool which calculates yplus instead of ystar. If you want to resolve the boundary layer you shouldn't use wall functions. I would start with an yplus value of 1 and see where it takes you. Yes, for an yplus value of 30 or greater you absolutely should use wall functions.
You can play around with nonOrthogonalCorrectors in fvSolution. But in general it is often better to aim for a better mesh first.