[wingcollide]

Hi Simon!
As I probably mentioned before I have to compare a structured mesh with an unstructured one. For the structured mesh I put the elements in the boundary layer at heights that will not vary with more than 10%. The problem is, with the unstructured mesh, the elements are arranged more like a zig zag puzzle and so when looking at the results the jump in the elements's height is quite big.The errors(with respect to Blasius) are not much higher than for the perfect structured mesh but my supervisor wants me to have a relavant comparison and so I wanted to know if is possible to get the uniform height of the tetras(or a variation of maximum10%).I already made a structured mesh with similar distribution of the elements as the unstructured mesh, and again the results are not very different than for a perfect structured mesh.
Let me know if you have an answer to my question and maybe a link where it is explained how cfx is solving the structured and unstructured grids and how the errors vary with the elements height differences(deltay).
Thank you
Camelia

[PSYMN]

Do you mean you are using an OCTREE Tetra mesher and are getting the sudden 2 to 1 size transitions? This is how OCTREE tetra is.

For the smoothest Tetra, you could start with an OCTREE mesh (or use the patch based surface mesher), but then delete the volume mesh and smooth the heck out of the rest of the mesh using the Laplace option. This Laplace option tries to smooth out the triangle sizes and equalize the angles. When you are satisfied with the surface mesh, use the Advancing front or Delaunay TGlib with AFT options to generate a new tetra mesh.

I will include some images to show the difference.

The TGlib AFT Delaunay is probably my new favorite bottom up tetra meshing option.

If you are talking about the jump between the prisms and the tetras, look up my threads with the key word "float". You want to float the prism height to achieve the best volume transition between prisms and tetras.

If you mean something else, please send a screen shot.

[wingcollide]

Hi Simon!
I am sorry I didn't reply for such a long time. Thank you once again for the reply. I already had my prism layer floating(if floating-if I understood correctly from the ICEM tips- is setting the initial height 0). And I tried to Laplace smooth my mesh but it alters my prism layer height even where I wouldn.t need it. My mesh looked fine to me( except at the leading where the prism layer height is much bigger than the tetra elements height) and it gave good results but at the processing part I had problems with how the data.I attached a caption from my boundary layer in one of my meshes. The red line-which would be ideal for a chart line will take points equally spaced, while the blue line will have more points not equally spaced.I asked my supervisor which one to consider when doing a structured mesh with similar distribution as the unstructured one and he told me to consider the distribution evenly spaced.
I didn't find the Delauney Tglib option but I will look in more depth after the holidays.
Let me know if you think my prism tetra transition is ok.
Thanks again and I hope you had a merry Chritsmas and you will have a happy New Year!
Camelia
3dunstructured1_0022ed.jpg

[PSYMN]

Sorry, I am not understanding what you are showing me in that image... I also don't get the blue and red lines... Perhaps it is too late.

However, I can tell you that we don't usually allow the prisms to move when we smooth... If you have access to the training material take a look at the section about smoothing for Prisms...

On the smooth DEZ, you will see that you can "smooth" certain element types. Usually, we "Freeze" the "Penta-6" elements so the smoother doesn't mess with them.

If you decide you want to smooth some of the worst prisms, you can smooth a subset (Under Edit Mesh => Repair Mesh => Smooth) or you can set the "Up to" value very low (0.01) and set just a few iterations so it will only work on the worst elements and not cause trouble elsewhere.

[hz283]

Originally Posted by PSYMN

Do you mean you are using an OCTREE Tetra mesher and are getting the sudden 2 to 1 size transitions? This is how OCTREE tetra is.

For the smoothest Tetra, you could start with an OCTREE mesh (or use the patch based surface mesher), but then delete the volume mesh and smooth the heck out of the rest of the mesh using the Laplace option. This Laplace option tries to smooth out the triangle sizes and equalize the angles. When you are satisfied with the surface mesh, use the Advancing front or Delaunay TGlib with AFT options to generate a new tetra mesh.

I will include some images to show the difference.

The TGlib AFT Delaunay is probably my new favorite bottom up tetra meshing option.

If you are talking about the jump between the prisms and the tetras, look up my threads with the key word "float". You want to float the prism height to achieve the best volume transition between prisms and tetras.

If you mean something else, please send a screen shot.

Hi Simon,

I am new here. Thank you very much for your helpful posts. In order to do the mesh smoothing, when I delete the volume mesh, all the meshes are deleted. There is nothing I can smooth. So Could you please explain more about the procedure? Thank you so much.

hz