fields of solution exported to VTK shows up twice in Visit

[ragusa]

when exporting the solution fields to VTK and then opening the VTK file with Visit, the fields appear twice in the drop-down menus. For exmaple, one can see phi_g000 listed twice. There must be something silly going on.

[andrsd]

I literally ran into this about an hour ago. This is what I see in paraview:

Notice that one instance is a nodal version and the other is an elemental version of the same field.

[ragusa]

given that we use DG for discretization, values at a node are multiply defined: there is one value per cell that contains that node. This makes me think that the elemental version is what we want most of the time. However, some diffusion solver uses CFEM, for which nodal is correct.

[andrsd]

Do you know if there is a projection from the DG space into the CG space for visualization?

Also, the DG solution could be displayed in formats like VTK, ExodusII. You just need to duplicate the nodes, so you can store the value at a node multiple times. This would give users better idea of the actual solution. Simple graphics how this works (1D):

+------+ +------+
1      2 3      4

Then, values are stored into nodes 1, 2, 3, 4 the usual way with the fact that node 2 and 3 share the same physical location. Connectivity in the file is:

EDGE2: 1 2
EDGE2: 3 4

[ragusa]

I will the node duplication is what we do, through elemental, as nodes will be automatically duplicated then

[andrsd]

I played a little bit with Reed's balance problem over the weekend and discovered that the node duplication is already happening.

See the discontinuity at 0.5, 1.5, 2, 3, etc. Note that this is on a very coarse mesh to show the behavior.

[ragusa]

we should remove the elemental values for DFEM solution outputs.