[MOAB-dev] Puzzled

[Gerd Heber]

Iulian, thanks that's very helpful. The mesh comes from Abaqus, but I'm
getting it only "second hand", i.e., I don't have the original file, only
the output from a converter that was written for another code.
Obviously, this needs some rework. BTW, what's the most painless way
to get that VisIt (?) plugin to work?

Thank you very much.

Best, G.

-----Original Message-----
From: Grindeanu, Iulian R. [mailto:iulian at mcs.anl.gov] 
Sent: Monday, August 11, 2014 7:24 PM
To: Gerd Heber; MOAB dev
Subject: RE: Puzzled

yes, the issue is the model.
I had to turn on "Show internal zones" in visit, otherwise I could not see the inverted elements

In this picture, there are the 2 partitions, and some of the hexas have negative orientation, they do not show up with solid color, just as wireframe.

I still can't explain where are the 599 hexas, as I see only 500

the nodes on the interface between partitions are not resolved correctly if the elements are inverted (skinner does not identify them ) so 10 nodes are not shared (2x5) and this is why there are 10 more nodes in the "owned" category

(I highlighted 3 of those 10 nodes, A, B, C)

Again, how is this model created?

Thanks,
Iulian
________________________________________
From: moab-dev-bounces at mcs.anl.gov [moab-dev-bounces at mcs.anl.gov] on behalf of Grindeanu, Iulian R. [iulian at mcs.anl.gov]
Sent: Monday, August 11, 2014 6:12 PM
To: Gerd Heber; MOAB dev
Subject: Re: [MOAB-dev] Puzzled

hello,
Yes, it does not add up. Your code seems fine.

 There are other issues I can't explain with this mesh file.
How was this mesh generated?
It seems that elements have a "negative" orientation (negative volume).

moab uses the skinner to identify the boundary, a similar algorithm is used by paraview/visit to show the surface. if some elements are oriented differently, the skinner fails to identify the correct boundary, and it fails later in resolving the shared entities.

I can see only about 5x8x10~400  hexas, total, but mbsize reports 599;

It seems that there are elements filling a cube 10x10x10 about half are hexas, but 5x10x10 is 500, (again, mbsize reports 599)

So is half of the volume filled with prisms, and the other half with hexas?

Iulian
________________________________________
From: moab-dev-bounces at mcs.anl.gov [moab-dev-bounces at mcs.anl.gov] on behalf of Gerd Heber [gheber at hdfgroup.org]
Sent: Monday, August 11, 2014 3:15 PM
To: moab-dev at mcs.anl.gov
Subject: [MOAB-dev] Puzzled

I'm probably doing something stupid, but here it is:

The attached mesh contains 599 hexahedra and 1,140 prisms.
I has 1,408 nodes. It was partitioned into two partitions with mbpart.
When I run the attached program with 'mpirun -np 2 feawdy feawd.h5m' the output reads:

# hexes :       299     300
# prisms:       570     570
# vertexes:     781     781
# shared:       144     144
# owned:        781     637
# shared/owned: 144     0

The element counts add up nicely, but
I'm confused that the 'owned' counts don't add up to 1408.
What am I missing?

Thanks for your help.
G.