PistonVale tutorial - issues about mesh quality
 

Hi guys,

I am a beginner with ICEM and of I am starting to built a little experience using the provided tutorials (with ICEM CFD 13.0).
I have a doubt about the PistonValve tutorial:
I am aware that the minimum quality target for a mesh with pyramids is about 0.05, and without pyramids is about 0.2.
The quality of the mesh that I obtain following the tutorial steps is reported in Figure 1 (in this case pyramids are present). (Total elements : 66943, PYRA_5 : 576)
I tried to play a little bit with the case creating a new mesh with more elements (Total elements : 93534, ); this time no pyramids are present and the quality of the mesh is reported in Figure 2.
As expected, the low quality elements are located in the region near the intersection of stem and port (where the thin cuts are defined): however I was not able to generate a mesh of decent quality without pyramids.
Is there a method to obtain a mesh with no pyramids and a good quality (>0.2) for this case?

Thanks

Andrea

Pyramids are used to join a hexa cell's quad face and a tetra cell's tri face. I find this happens if I build prism layers without floating them. When prisms are floated I don't get pyramids. So the key is to float the prisms to avoid pyramids.

You should look at this topic: prism generation. PSYMN gives a hyperlink to a very useful ANSYS presentation for ICEM prism generation. I use this for guidance alot for my Fluent and CFX simulations. This may help you.

I have been making some meshes recently whereby I smoothed the prims up to a quality of 0.3. Then froze them to further smooth the tetras. I found I needed to do this as it was more important to get higher tetra quality and that the highest quality prisms was okay. If I only smoothed the prisms up to 0.05 then nothing I did would improve the tetra quality which was very low. Just mentioning this since the presentation talks about prism smoothing.

Hi siw,

thank you for your reply (and for the useful link!)
Everything you said makes sense to me but I still have a doubt: what do you mean by floating the prism? Do you mean editing the prisms (using Edit Mesh tools) or using the "float" option in the Smooth Mesh Type window?

At the moment I am following this procedure for generating prisms:
- generate Octree tetras
- smoothing this mesh up to a decent quality (> 0.3)
- replace the Octree tetras with Delaunay TGlib
- smooth again the mesh if necessary
- generate the prisms with these settings:
Initial height: 0
Min prism quality: 0.00001
Orto weight: 0.35
Fillet ratio: 0.5
Max prism angle: 180
Number of surface smoothing steps: 0
Auto reduction enabled
- Smooth the mesh with PENTA_6 freezed
- Smooth again the mesh enabilng the smoothing for prisms and lowering Up to value to 0.02

Is this a correct approach in your opinion?

By floating the prisms I mean that in the prism mesh setup set the initial prism height and the total number of prisms both to zero. ICEM will then build you specified number of prism layers (so I enter a low number of layers and height ratio) so that there is a nice smooth volume transition between the adjacent prism and tetrahedral elements. After that you can then split those few layers into however many you really want and re-distribute them.

In terms of the smoothing I always check and smooth the prisms (PENTA_6) and tetra (TETRA_4) both set to float and smooth to a low quality value. I then set PENTA_6 to frozen and smooth the tetras to a higher quality value.

I would add to your list that at the octree stage that after making the octree volume that you delete the volume elements and smooth the triangular surface elements (don't forget to try the Laplace smoother and the uncheck it and re-smooth for the surface mesh only), this should speed things up. Then once your surface mesh is good then generate the volume mesh with Delaunay. Also another thing I do is to set the Prism Height Limit Factor to about 1.0 (you'll need to play with that) so that after the prism splitting and re-distribution I get a nice prism and tetra volume transition - you'll need to judge that yourself until you are happy because it's bad practice to have >20% volume transition.

Many thanks for the tips! I think I need to practise a lot to get used to ICEM, but at first sight it seems much more powerful than Gambit.

If it can be useful also to other people, I attach a picture of the mesh quality obtained following siw's tips and playing a little bit with the surface sizing when generating the Octree mesh...it seems pretty good ;)