Even for a simple problem (3D) like a flow in a cylinder, I have trouble while using tet-mesh, while I get good results when I use `brick'. I am a new user and I assume I am missing out some detail. I'd be glad if someone could help. Thanks.

Tet meshes are difficult to converge, require a lot of cells, and even then the solution is often too diffusive. I only use tet mesh in small transition zones between hex blocks. (Sometimes I wonder if it's not better to just have a non-conformal mesh with only hex cells and eliminate tets altogether. I'd like to hear more opinions on this...)

If you really need tets, then try to have the boundary faces meshed with quads and use a good boundary layer. And always limit your cell skewness to <0.8 if you can.

I agree with Erwin (in particular for boundary layer), but have you used second order upwind? Have you check the volume of the tets at the center of your cylinder? It should not be too large.

Giacomo

hi Giacomo, you mean to say that, for tet-mesh we should go for second order upwind? tell me more about this. bye

Using tet-mesh the solution is often too diffusive, that is "takes some errors" because of the fluid flow direction.

In order to reduce this problem you have to use the second order upwind discretization. See solution controls panel help, or see manuals.

Giacomo

got good results thk's for good hints. bye

Does anyone have studied the diffussive effects of Tet-mesh under laminar as well as turbulent conditions? I generally use Tet-mesh because of ease of meshing (I use Fluent). And difficulty in converging with Tet-mesh is not an issue for me. But my flow geometries are always complex and I can't separate the diffusive efects due to the geometry itself or turbulence or meshing scheme. Any general comments to help decide the meshing under turbulent/laminar conditions? Thanks. PK