navier stokes no-slip element normals

[mark smith]

In the flow solver does it matter which direction the element surface normals point for the no-slip BC ? if it does should they point into or out of the fluid?
Like wise for conjugate heat transfer at the interface between solid & fluid?

Thanks in advance
Mark

[raback]

Hi Mark,

No. No-slip conditions just set all components to zero.

If you have slip conditions its more complex because only normal component should be zero, and the tangential components are free. Then in generic case you need to work with Normal-Tangential coordinate system.

-Peter

[mark smith]

Hi Peter,
I am evaluating pointwise as a mesher for elmer and have a mesh produced by one of their resellers for my tests.
When a surface BC is defined for the no-slip surfaces I get a surface mesh whose normals point into the fluid, you also get a second BC group which is the wetted surface of the solid which the fluid touches it's identical to the no-slip surface but this surface mesh has the normals pointing in the opposite direction.
If in my simulation I use the mesh that points into the fluid the simulation runs and converges where as if I use the mesh where the normals point out of the fluid the simulation will not converge. This seems to imply that the direction of the normals is important?
Best Regards
Mark

[raback]

Hi Mark

The direction of the normals is important only if the BC is not zero. Perhaps the BCs are not pointing to the same nodes i.e. you have duplicate sets of nodes at the interface?

-Peter

[mark smith]

Hi Peter,
Yes it may well be that there are duplicate sets of nodes at the interface and probably duplicate surface elements as well, are there any Elmer tools which can identify this sort of problem and remove them from the mesh. This appears to be a problem with pointwise meshing software when outputting a gmsh mesh, it looks OK in gmsh but gives problems with conjugate heat transfer simulations after converting to Elmer format using Elmergrid. Are there any meshers that you can recommend? even commercial ones?
Best regards.
Mark

[mzenker]

Hi Mark,

the Salomé meshing module can make meshes without duplicated surfaces. See the wiki page: http://www.elmerfem.org/elmerwiki/index ... e_to_Elmer.
Instead of drawing it, you can also import your geometry. The essential is to make partitions before meshing.

HTH,

Matthias

[mark smith]

Hi Matthias,
Thanks for the link. Is Salome able to generate meshes for CFD/ conjugate heat transfer with prism boundary layers? What mesh engine does it use?
I use Gmsh for simple geometries but I haven't found a way to generate CFD meshes with tet cores & prism boundary layers & pyramid transition elements. For complex geometries I tend to use Comsol to make the mesh & export it for conversion with Elmergrid which works well without any problems (so far) it's just Comsol is an expensive tool just for meshing.

Best regards
Mark

[mzenker]

Hi Mark,

I am not sure about prisms and pyramids. For 3D (tetra) meshes with Salomé, I have only used netgen as meshing engine so far. There seem to be plans to include gmsh, but I don't know if that can be expected to be done anytime soon. There also seem to be more meshing engines available, including commercial ones, but I have never tried them out. You can find information on Salomé here: http://www.salome-platform.org. There is also a forum where you can post questions which are usually answered within 1 day. You may find some hints already by searching the forum.

There is also the GiD mesher (http://www.gidhome.com) which is commercial, but not expensive. I don't know if GiD can do what you need, when Salomé can't.

HTH,

Matthias

[mark smith]

Hi Matthias,
I have GiD & have used it for a number of years but it cannot produce boundary layer meshes other than having finer tet elements near the wall and bigger ones towards the core of the fluid.
I'll do some more research on Salome and try it out.
Regards
Mark

[MrChips]

I have also recently started evaluating a Pointwise-Elmer pairing as Elmer seems to have trouble using Salome meshes when there's a viscous layer applied. These runs typically end with a "Bubbles for element: 706 are not implemented" error. I just started producing some trial meshes in Pointwise with their T-rex solver to produce viscous layers. I like it much better than Salome.

I have not progressed far enough to say if I have the same problem, but I do have problems getting convergence with the NS solver (still troubleshooting, it is likely other things). I have been converting my meshes like this: Pointwise ---> .nas --(with gmsh)-> gmsh --(elmergrid/gui)-> .mesh.