Does anybody use a pure tetra mesh to solve the problem without applying any inflation layer at wall boundary?

I tried so many ways to improve the tet & tri mesh up to quality of at least 0.3 in ICEM V10. However, pyramid elements always showed up when prism mesh is extruded. Reasons giving by ICEM is that:

"58 node columns stopped during extrusion because prism couldn't retriangulate"

Any thoughts about this issue? Thanks for help.

Hi,

Yes, you can use tet only neshes. Just don't expect the boundary layer resolution to be any good. Of course for flows which don't have thin boundary layers (ie low Reynolds numkber flows) this is the case so it will work well. But if you use no prisms with a high Reynolds number flow you can guarantee your boundary layer accuracy will be terrible and your results rubbish.

Note that there is nothing wrong with pyramid elements. They run fine. If you are meshing a complicated geometry it is impossible to get all the tets to 0.3 quality and after prism meshing you often get elements with qualities much lower than that. Do your best to improve things, but often you are forced to run with a low quality mesh. Whether that is a problem or not depends on what you are modelling. CFX is a reasonably robust solver and can handle poor meshes in general, but some options (eg free surfaces) really do need a high quality mesh to run accurately.

Regards, Glenn Horrocks

Thanks, Glenn. I just found one of your old post regarding tetra mesh smoothing. I follow your 1st step and get a min overall quality of 0.43 for tri & tet elements.

For my case, inlet Reynolds Number ~ 2.5x10^5 and is considered a pretty high Re flow. My geometry is an air diffusing elbow connecting to circular pipe upstream and a rectangular channel downstream. Inlet pipe diameter is about 151mm.

I use 2 strategy for prism elements:

1. Create a single layer with total thickness and subdivide it later.

2. Specifying an initial height growth rate & number of layers.

Both methods end up with some pyramid elements. You mentioned that thin layer might give better quality. Do you have any suggestions about how to specify the prism layer in this case? The initial BL thickness at inlet pipe is about 7-10mm.

If pyramid elements are unavoidable, then what smoothing strategy would you normally use?

ICEM CFD Tutorial manual doesn't have any examples that end up with pyramid elements. That gives me an impression that these elements are not so good for CFD.

I hope you can help again with my questions. Really grateful for that.

Hi,

There is nothing wrong with pyramid elements. As long as their quality is OK then it will run fine. The problem generally is that pyramid elements are used to terminate prism stacks when it can't reach the full number of layers, and this occurs in regions of high mesh distortion so the pyramids often happen to be poor quality elements because of it.

Often the solver runs OK even with very poor quality elements. Generate the best quality mesh you can, but even if you still have some poor elements give the solver a go as it probably will converge anyway.

Glenn Horrocks

Which old post is that you are refering to? I cannot find it.

Perhaps this paper from Proc. of Building Simulation 2005 conference helps you a bit for the near wall mesh sizing. But also here prisms were used to resolve the wall boundary layer:

BS2005: Simulation of steady-state natural convection using CFD.

Good luck

TobiasZ

Just a thought Jane, but do you need to be using a tetrahedral mesh? The way you describe your geometry, it sounds like it would be fairly easy to mesh in Hexa. Perhaps I'm just not that great with Tetra, but I almost always seem to be able to get much better results with Hexa meshes.