I am currently trying to generate a mesh on a simple T junction using Netgen 1D-2D-3D and everything works until I try to add a viscous layer on every walls of my geometry. Indeed, it appears that when I specify a total thickness for the viscous layer which is below the value of refinement for the junction it does not succeed to generate the mesh. I have the following error : Viscous layers builder: Infinite loop at shrinking 2D mesh on face 13
I can force the creation of the boundary layers by specifying a total thickness value below the refinement region cell sizes (and also below the max cells sizes in the mesh) but it then create a viscous layer which is very thin…
Dear N469,
I have been able to fix the issue by switching from NETGEN 1D-2D-3D to NETGEN 3D. Unfortunately, it appears that the viscous layer that I obtain is not exactly what I am looking for, see just after the 90° angle which is not handled properly by the viscous layers:
I tried to reproduce your geometry and mesh but I don’t get any issue when I use Netgen 1D-2D-3D with viscous layers and a smaller than mesh element size thickness. I would also be curious to know the viscous layers settings you tried.
As for your second question, this looks like a visualization issue. Given the gray background I assume you’re looking at the mesh on Paravis, probably using a slice filter ?
If I do this with my mesh I observe the following:
Hello,
the second view of pieres message is a clip view of paraview to be exact, nevertheless it is not an visual effect of the crinkle set to off, it is because the cells of the BL are collapsed at that point.
I can not confirm as the geometry was not share (this helps to help the person) @PierreLCV if you can upload i can have a look. from what it looks like (really difficult to be sure just from the two images) there is an issue in the geometry itself, a super small edge or something like that in the geometry causing to give issues to the netgen algo. the geometry at least the section seeing in the image should not give any trouble (as you have seen louis) for a nice BL.
I am also trying to build viscous layers (VL) on a pipe junction. I got the same result as @PierreLCV — all the VL collapse into a single node at the 90º corner.
My geometry (attached) is a bit different: it is a divided cylinder (since after the junction I will use a hexa mesh) attached to a box. How can I get that nice VL corner just like @Louis.simvia?
P.S.: While studying this geometry, I found some issues when building viscous layers. Using quad-dominated elements either in the 3D mesh or on 2D surfaces to guide the VL always led to errors. Keeping the VL “Total thickness” smaller than the 3D mesh “Min. size” avoids errors. The Extrusion Method doesn’t change the final results, but Node Offset gives fewer errors and allows using a larger VL total thickness.
you are not going to be able to mesh in hexa the solid that is a block. hexahedron (i,j,k) requires that a valid quad mesh (“i.e. there should be eight nodes shared by three quadrangles and the rest nodes should be shared by four quadrangles” from SMESH/share/doc/salome/gui/SMESH/basic_meshing_algos.html) which is not your case.
when you will divide correctly the geometry, (dividing the hexagonal block into 9 different blocks) you will not be able to run correctly the viscous layer addition. this happends as you have a 3 block, one edge situation, as i commented in this post Bug SMESH hexahedron + layer generation issue with internal faces of blocks so at least with this kind of blocking you will not be able to add a BL to it.
