Boundary Layer mesh will not be created

[ghost]

Hello again,

I ran into problems when creating boundary layers on a geometry with several patches. After creating the surface mesh, I wanted to create the boundary layers on 14 of the 20 available patches with 25k triangles (these patches are next to each other, if this is important). After my 32 GB of RAM where not enough for this task, I went to another workstation, where the creation of the boundary layers is runnig for about 24 hours now on a 2,4 GHz core using about 140 GB of RAM and not dropping any message to the logfiles or the terminal it was startet from since the following:

creating boundary layer mesh) preparing prismatic layer

---

prism quads : 0 prism triangles : 0 stray quads : 0 stray triangles : 28538

---

WARNING: RestrictLocalH called, creating mesh-size tree

If you need it, I could post the engrid files, so you can work on this case.

Would be nice, if you could help me. Thank you.

Regards Andreas

[ogloth]

Hi,

it sounds a bit like NETGEN is going havoc, because it tries to create too many cells. 25k triangles should not require that much memory, unless you have a massive volume source somewhere. The boundary layer process itself does not consume much memory, but the initial tetra grid does. Is the mesh correctly oriented? In any case this should probably be caught somehow.

If you make the files available somewhere I could have a look.

Regards, Oliver

[ghost]

Hi Oliver, thank you for the quick reply.

there is noch volume source at all. Is there a tetra grid at all, when building the boundary layer mesh before creating the volume mesh? By the way, a general question: is it NETGEN or engrid, which creates the boundary mesh? I thought, NETGEN is only doing the volume mesh, but that might be wrong as well.

The files are available under http://139.30.100.166/kwijybo.tar.gz

Thank you, Andreas

[ogloth]

Correct, NETGEN only creates the tetras, but before creating the prismatic boundary layer, NETGEN is called once in order to create an initial tetra grid; This initial tetra the grid will be deleted again. If, for whatever reason, NETGEN thinks it needs more and more cells it will inflate until it crashes. This appears what is happening in your case.

[ogloth]

Andreas,

I have checked the case and the patches are oriented inside out. This does not fix the problem, however. The NETGEN library call enters a strange endless loop which floods the memory.

My suggestion for this case is to improve the surface mesh by assigning a different boundary code for every distinct geometric feature. This would avoid rounded edges and keep the geometric features intact. Maybe the case will mesh without problem then. Generally the geometry is fairly simple and should not pose any problem -- theoretically.

I will close this issue now, but I have created a new issue about better error handling for NETGEN calls.

(see https://github.com/enGits/engrid/issues/28).

Regards, Oliver

[ghost]

To complete this issue, I found the "solution": As I scaled the surface mesh after generation, the geometry was broken somehow. when I scaled the geometry in blender before improting it to engrid it worked. So the problem may be related to the scaling operation after surface and volumen mesh generation. Again thank you for your help, Oliver!