CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   polyDualMesh results in non-planar faces and non-convex cells (http://www.cfd-online.com/Forums/openfoam-solving/106959-polydualmesh-results-non-planar-faces-non-convex-cells.html)

 tomislav_maric September 13, 2012 08:47

polyDualMesh results in non-planar faces and non-convex cells

Hi everyone,

I have an extremely simple domain: a box of dimensions [0,1]^3 which is meshed with tetrahedrons using the Netgen algorithm of Salome (version 6.4.0, 64 bit). When I use the polyDualMesh to generate a polyhedral mesh of the boxed domain, I get cells which are non-convex and have non-planar faces. Here is a picture of such a cell.

I've tried multiple feature angles, although the 90 degrees seems most sensible for a simple box domain. Any advice on this?

Here's the case with the .unv of the tetrahedral mesh.

Note: creating a more regular tetrahedral mesh using a decomposition of hex cells results in a combination of hex-poly cells, where the polyhedra are convex with planar faces, but the cell aspect ratio is kind of high (around 2).

Is there a way to get a good polyhedral mesh using purely tet mesh generation algorithms such as Netgen?

 deepsterblue September 13, 2012 10:18

Have you checked to ensure that your underlying tet-mesh is Delaunay? If not, you'll have to ensure that first.

Of course, you could easily pre-process the underlying tet-mesh (with smooth/remesh) before you convert it to polyhedra..

 tomislav_maric September 13, 2012 10:24

Quote:
 Originally Posted by deepsterblue (Post 381638) Have you checked to ensure that your underlying tet-mesh is Delaunay? If not, you'll have to ensure that first. Of course, you could easily pre-process the underlying tet-mesh (with smooth/remesh) before you convert it to polyhedra..
Thanks! Nope, didn't check this.... I'm working with Salome, so I'm not sure how it goes about generating tet meshes..

As for smoothing-remeshing: how can I be sure that the dual will result in convex cells? I mean, what parameters/workflow do I use for smooth/remesh in this case?

 deepsterblue September 13, 2012 10:29

I typically use the mesquiteMotionSolver in combination with moveDynamicMesh for a few time-steps. If you want to remesh as well, you can use dynamicTopoFvMesh (which can include a smoother as well).

Sorry for blatantly advertising my stuff, but couldn't help it...

 tomislav_maric September 13, 2012 10:32

Quote:
 Originally Posted by deepsterblue (Post 381640) I typically use the mesquiteMotionSolver in combination with moveDynamicMesh for a few time-steps. If you want to remesh as well, you can use dynamicTopoFvMesh (which can include a smoother as well). Sorry for blatantly advertising my stuff, but couldn't help it...
Hehehehe marketing is always good.. and your stuff is worth advertising, so no worries there! :D Thanks for the info, I will try it!

 All times are GMT -4. The time now is 16:02.