MOOSE seems unable to recognize 1D boundary on 3D mesh

[richmondodufisan]

Hi all,

I have a problem which I have encountered in the past that is coming up again. I use gmsh for meshing.

When I create a 2D mesh, and define 1D physical groups (lines), MOOSE is able to recognize those lines as boundaries and do things like apply Neumann conditions, use VectorPostProcessors like SideValueSampler, etc.

However, when I create a 3D mesh and define 1D physical groups, I get the error:

(VectorPostprocessors/upper_right_z_disp/boundary):
    the following side sets (ids) do not exist on the mesh: sample_location (19)

    MOOSE distinguishes between "node sets" and "side sets" depending on whether
    you are using "Nodal" or "Integrated" BCs respectively. Node sets corresponding
    to your side sets are constructed for you by default.

    Try setting "Mesh/construct_side_list_from_node_list=true" if you see this error.
    Note: If you are running with adaptivity you should prefer using side sets.

This only shows up for 1D physical groups. 2D physical groups/boundaries work fine. Also, "construct_side_list_from_node_list=true" does not change anything.

Based off of this behavior, If I understand correctly, the way it works is that "sides" can only be 1 dimension less than the dimension of the mesh. But then what do I do if I want to apply a line load on a 3D mesh? Or in my case (reason for making this post), sample a value along a curved line? LineValueSampler can't work for me there because that can only define straight lines.

Interestingly enough, a 0D physical group had no issues applying a Dirichlet BC.

I saw this answer #24174 from @jwpeterson , where it was suggested to add "lower_dimensional_block" to my mesh file, so I went ahead and did:

$MeshFormat
4.1 0 8
$EndMeshFormat
$PhysicalNames
2
0 18 "loading_point"
1 19 "sample_location" lower_dimensional_block
$EndPhysicalNames
$Entities
7 12 8 1
1 0 0 0 0 
2 0 0.002 0 0 
3 0 -0.002 0 0 
4 0 0 0.002 1 18 
5 0 0 -0.002 0 
6 0.002 0 0 0 
7 -0.002 0 0 0 
1 -6.389406329081476e-05 -6.389406329081475e-05 -6.389406329081475e-05 6.389406329081476e-05 0.002063894063290815 0.002063894063290815 0 2 4 -2 
2 -2.334153974494834e-05 -2.334153974494803e-05 -0.002023341539744949 2.334153974494834e-05 0.002023341539744949 2.334153974494846e-05 0 2 2 -5 

but nothing changed. How can I fix this so I can sample points along a curved line in 3D? I attached here my gmsh file, a simple sphere, and I want to sample the value of the displacement along a curved line on the surface from the top to the right side.

// Gmsh project created on Wed Oct 16 19:47:43 2024
SetFactory("OpenCASCADE");
//+
Point(1) = {0, 0, 0, 0.00005};
//+
Point(2) = {0, 0.002, 0, 0.00005};
//+
Point(3) = {0, -0.002, 0, 0.00005};
//+
Point(4) = {0, 0, 0.002, 0.00005};
//+
Point(5) = {0, 0, -0.002, 0.00005};
//+
Point(6) = {0.002, 0, 0, 0.00005};
//+
Point(7) = {-0.002, 0, 0, 0.00005};
//+
Circle(1) = {4, 1, 2};
//+
Circle(2) = {2, 1, 5};
//+
Circle(3) = {6, 1, 2};
//+
Circle(4) = {2, 1, 7};
//+
Circle(5) = {4, 1, 3};
//+
Circle(6) = {5, 1, 3};
//+
Circle(7) = {6, 1, 3};
//+
Circle(8) = {7, 1, 3};
//+
Circle(9) = {4, 1, 7};
//+
Circle(10) = {4, 1, 6};
//+
Circle(11) = {6, 1, 5};
//+
Circle(12) = {5, 1, 7};
//+
Curve Loop(2) = {1, -3, -10};
//+
Surface(1) = {2};
//+
Curve Loop(4) = {9, -4, -1};
//+
Surface(2) = {4};
//+
Curve Loop(6) = {2, -11, 3};
//+
Surface(3) = {6};
//+
Curve Loop(8) = {2, 12, -4};
//+
Surface(4) = {8};
//+
Curve Loop(10) = {5, -7, -10};
//+
Surface(5) = {10};
//+
Curve Loop(12) = {7, -6, -11};
//+
Surface(6) = {12};
//+
Curve Loop(14) = {6, -8, -12};
//+
Surface(7) = {14};
//+
Curve Loop(16) = {5, -8, -9};
//+
Surface(8) = {16};
//+
Surface Loop(1) = {7, 6, 5, 1, 2, 4, 3, 8};
//+
Volume(1) = {1};
//+
Surface Loop(2) = {5, 1, 2, 4, 3, 6, 7, 8};
//+
Physical Point("loading_point", 18) = {4};
//+
Physical Curve("sample_location", 19) = {10};

[GiudGiud]

Hello

I think from previous posts we ve seen that the gmsh reader does not support reading 1D sets inside a 3D mesh. You should try to make sure of this on a simple case

If your 1D nodeset is simple enough it should be possible at add it after loading the mesh in moose.

Guillaume

[richmondodufisan]

Sure, here's the mesh (had to zip to upload here). Thank you!

eyeball_3D_coarse.zip

Also here in case you can't use zip files (finer mesh): https://github.com/richmondodufisan/purple/blob/main/Cornea_Linear_Elastic/eyeball_3D.msh

[GiudGiud]

Thanks! I ll have time later this week

[GiudGiud]

does this work?

[Mesh]
    [fmg]
        type = FileMeshGenerator
        file = eyeball_3D_coarse.msh 
    []
    [nodes]
        type = ParsedGenerateNodeset
        input = 'fmg'
        expression = '(abs(y) < 1e-8) & (abs(x * x + z * z - 0.002 *0.002) < 1e-8 ) & (z > 0)'
        new_nodeset_name = 'circle'
    []
[]

I get a half circle of nodes on the outside

[richmondodufisan]

Yes that does! Thank you so much!! That is weird though, why would it work with "abs" and not with doing them individually?

[GiudGiud]

abs lets us specify tolerances easier