Mesh density for turbulent flow in pipes (how big is a problem?)

Dear friends, I appeal to your experience.

I would like to determine what mesh density (approximate) is suitable to estimate in a preliminary way the flow inside pipes (Re = 300000 approx and convective thermal conditions)
Obviously I know that it is necessary to use parameters of independence of mesh, in addition to yplus but this would help me to know “how big is a problem”

:slight_smile: regards

Hi Camilo,

I would have answered to use a mesh independence study but you could estimate the worst case scenario for example by saying that you want to resolve the smallest scales possible (Kolmogorov scales) which you could calculate for a DNS which we definitely do not want to do for such a high Reynolds number here. For the meshing you would need to use a finer mesh near the walls anyway as high gradients there are expected as in a Rayleigh Bénard convection scenario but I have no exact formula that would tell you how big the mesh would be except the one for DNS I have in mind right now - @Get_Barried & @1318980 any additional input from your side? Maybe you have a formula for this but I guess convergence study is the right approach here.

Best,

Jousef

Hi everyone,

Agreed with Jousef. A MIS is the best way to approach this. On top of the general MIS where you just increase refinement levels and keep cell counts within 1.5x, you can focus on specific areas.

For example you could be using a y+ of 200 and run that, keep the results, then move on to y+ 100 then y+ 50 and see how the results change without altering other parts of the mesh. This allows you to verify and probably identify the areas of the mesh you need to take care so that you can safely say your mesh is “good enough” .

Same technique can be applied to other areas of interest as well.

There is a formula for MIS but its case specific so you need to check the appropriate literature.

Cheers.

Regards,
Barry

2 Likes