[thinesunny]

Hello,

I have couple of questions related to ICEM:

1. I have generated a 3D structured grid for a turbomachinery ducted flow geometry (complexity:: medium, 3D). The grid metrics (determinants, angles) seem to be alright both in ICEM and OpenFOAM (checkMesh utility). But, when I run the solver (pisoFoam, running LES), the solution is diverging. A different mesh topology is working fine, though. So, it is something to do with grid itself. I am not sure how to identify what is the problem. Do you have any suggestions on this ?

2. I have the structured grid (3D) which shows good quality metrics before running any smoother on it. But, when I run any smoother, it becomes worse -- with negative determinants. I tried the suggestion in user manual to use high stabilization factors of around 8 to see if the mesh will retain its original quality while trying for orthogonality. But, it is also not working for me. Is there a way I can avoid this issue ?

I have a feeling that both the above problems are related. I am not sure if any one else has experienced this sort of strange issue.

Thanks
Sunil

[diamondx]

i stumbled upon this good post by simon:

Hexa solvers usually want a min angle of 9 to 18 degrees, hexa determinant should be above 0.1 or 0.2. Warpage and other factors should also be looked at.

Tetra mesh should have an "ICEM CFD quality" of greater than 0.2, but some cells below that are ok provided they are not in critical areas.

The ICEM CFD "Quality" metric is pretty harsh on Prism mesh (divides the prism into three virtual tetras and takes the worst normalized circumsphere ratio of the three), but usually Prisms of over 0.01 will run.

They just added a TGrid Skew quality metric to ICEM CFD 12.0 that matches TGrid exactly, if that is what you are looking for (remember TGrid is inverted so 0 is good and 1 is bad)

Of course your solver matters also... Some solvers are very tolerant of bad internal angles, but are sentitive to volumetric change. Others are the other way around. Some would prefer poor prisms to any pyramids, others would prefer a few decent pyramids rather than poor prisms.

I recomend checking the requirements with your solver's documentation. Keep in mind that there is usually a decent factor of safty there so they can blame the mesher if things don't converge.

[Far]

Would you like to post some pics ?

[thinesunny]

I am sorry, I cannot post any pictures (bound by a contract)