[MOAB-dev] canonical face connectivity from 10-nodded tet mesh

[Jason Kraftcheck]

Julien Vignollet wrote:
> Dear Moab developers,
> 
> I am a new (and very modest!) moab user experiencing difficulties when
> trying to retrieve the canonical face connectivities in a mesh of 10
> nodded tetrahedrons (order based on Tautges system from the paper
> "Canonical numbering systems for finite-element codes") . I am trying to
> use moab in a 3D finite element code to model multi-phasic hyperelastic
> elements. I am using Gmsh and Paraview for pre- and post-processing.
> Attached is the basic mesh I am using to get started.
> 
> Apart from a small problem when importing the gmsh file (it seems nodes
> 7 and 8 are swapped), 

Are you certain that it is nodes 7 and 8 that are swapped?  I reviewed the
MOAB code for loading Gmsh files and there is a bug.  It appears that the
last two nodes for a 10-node tetrahedron will be in the swapped.  That is,
either nodes 8 and 9 are swapped or nodes 9 and 10 are swapped, depending on
whether you're counting from 0 or 1.

> my main difficulty arises when I try to access the
> faces connectivity (see extract of code at the end of this message).
> This technique worked successfully to obtain the element connectivities
> (i.e. the 10 nodes of the tets in a canonical order) when I loop over
> the elements, but it fails when I loop over the faces. The handle conn
> gets populated with the correct 6 face nodes but not in a canonical
> order. The vertex nodes are in the right order, but the mid-nodes are
> not and I don't see any consistent pattern.
> 

This is most likely a side effect of the above bug.  Because the tetrahedra
have incorrect node order, the triangles generated for the faces of the
tetrahedra are also incorrect.

- jason