Viscous layers not working with NETGEN_3D_Remote

Use of the platform
1

Hi everyone,

I’m currently working on generating a parallel mesh using Netgen 3D with multithreading (NETGEN_3D_Remote) in Salomé 9.14.0.

I’m trying to apply viscous layers using the ViscousLayers() method, but I haven’t been able to get it working in this parallel setup.

Interestingly, when I switch to the standard NETGEN_3D algorithm (non-parallel), the viscous layers work just fine.

Has anyone managed to apply viscous layers in combination with parallel meshing or specifically with NETGEN_3D_Remote?
Are there known limitations or any recommended workarounds for this case?

I’m using a relatively simple geometry (a box with a torus removed).

Thanks for any tips, advice, or examples anyone can provide.

2

when you mean netgen 3D in parallel, is the new implementations where the 1D and 2D are done in serial, and the 3D in parallel for each solid of a compound, no?
if thats the case i suspect that the issue is the geometry of the compound, the viscous layer addition does not like the nodes that are on an edge that needs to be moved (here is an example of what I mean Bug SMESH hexahedron + layer generation issue with internal faces of blocks - #5 by NabilG)
regards,

3

Yes, that’s exactly the method I’m using — as described here: Parallel Computing — Mesh 9.14.0 documentation.

I initially thought the issue might be due to internal surfaces between solids in a compound (as in your example), which could interfere with viscous layer generation.

To rule that out, I created a much simpler case: a single solid — a cube with a sphere removed. Even though this setup doesn’t benefit from parallelism, I ran it with NETGEN_3D_Remote, and I encountered the same problem: no viscous layers are generated.

For comparison, using NETGEN_3D (non-remote), the viscous layers are applied correctly.

I’ve attached the Python script used to generate the cube + sphere test case.
sphere_test_case.py (5.2 KB)