Edge/face ordering of high order geometries seem inconsistent with corresponding 1D ordering

[Vahid-Galavi]

Introduction
If we consider a high order 2D geometry, for example quadrilateral2D8 or quadirlateral2D9, the nodal ordering is as follows:

Quadrilateral2D8/       Quadrilateral2D9/
Quadrilateral3D8:       Quadrilateral3D9:

3-----6-----2           3-----6-----2 
|           |           |           | 
|           |           |           | 
7           5           7     8     5 
|           |           |           | 
|           |           |           | 
0-----4-----1           0-----4-----1 

If we consider a high order 1D geometry, i.e. Line2D3 or Line3D3, the ordering is as follows:

Line2D3/
Line3D3:

0-----2----1

So, (in all geometries) we get first the corner nodes and then the middle nodes which are consistent.

Problem
Now if we look at GenerateEdges() functions which return edge or face node numbers, I found that the node numbers that I get in these elements: Triangle2D6/Triangle3D6, Quadrilateral2D8/Quadrilateral3D8 and Quadrilateral2D9/Quadrilateral3D9 are inconsistent with node numbering of Line2D3/Line3D3. These functions return the first, the middle and the last nodes instead of the first, the last and the middle. This implies that in order to check if a line belongs to a face, we need to do a redundant sorting.

This also the case in GenerateEdges of Hexahedron3D20 Hexahedron3D27, Prism3D15.

Questions

• Why do we use a different node numbering for edges?
• Can we modify the node numbering of edges to make them consistent with 1D geometries?

[philbucher]

This is an important observation, I agree that this is a problem!
However I don't think it is possible to change the ordering, as this is sort of imposed by the meshers we use
Would changing the GenerateEdges function work? I think this might be possible as it is used less

[Vahid-Galavi]

In my opinion, we only need to change GenerateEdges.

[philbucher]

that makes sense to me
did you also check if the same problem exists with GenerateFaces?

[Vahid-Galavi]

that makes sense to me did you also check if the same problem exists with GenerateFaces?

@philbucher I am checking now.

[Vahid-Galavi]

@philbucher What is the ordering number for surface loads in 3D? Is there a documentation?

[philbucher]

hm not sure
@loumalouomega @AlejandroCornejo do you know?

[loumalouomega]

hm not sure
@loumalouomega @AlejandroCornejo do you know?

We consider gid criteria...

[Vahid-Galavi]

Then the ordering that I get from GenerateFaces of 3D geometries are not consistent. They are defined in reverse order. So, they have to be changed, I think.

[philbucher]

Then the ordering that I get from GenerateFaces of 3D geometries are not consistent. They are defined in reverse order. So, they have to be changed, I think.

makes sense to me. No more comments from my side

@KratosMultiphysics/technical-committee ?

[RiccardoRossi]

@KratosMultiphysics/technical-committee also agrees ... just we are a little afraid of things that can get broken.

In any case if you have a PR please go ahead and see if it passes the tests. Is so...green light on our side

[Vahid-Galavi]

Thanks @RiccardoRossi! At the moment, I do not have a PR for this issue but I will make it as soon as I find some time.

[ddiezrod]

Issue shoud be solved after #10668