Topic: Correct node numbering in VTK?

Hi,

I am using VTK to understand better how OOFem works. I found out that the node numbering in VTK output is not the same as in the input file (I guess this is probably due to renumbering for more efficient solving). Is there any way to get the input and VTK output files to correspond with each other?

At least having an output file with the renumbering "function" (input node number vs. output node number), a simple script could be written.

Besides, it would be nice to be able to use a kind of include directive in input files, to separate different sections (e.g., mesh from problem data).

Thanks,

Santiago

2

Re: Correct node numbering in VTK?

Hi Santiago,

the numbering is different, that is right. This is not due to renumbering for efficient solving, because in this case the equation numbering (which is related to degrees of freedom) is changed, and this has nothing to do with node numbering (at least in oofem).
The node renumbering in vtk export comes from the fact, that export is done on a region basis. Regions (represented by cross section id at the moment) allow to have smoothing (which is always done in order to bring internal values from integration points to nodes) only on a certain part of a domain and not through the internal boundary, for example. They come also handy when domain is made from two materials and one needs to visualize internal variable available in only one material. Then one region (without internal variable) is skipped and recovery is done only on second region.
This region based recovery requires node renumbering, since nodes on internal region boundary have to be doubled. Unfortunately renumbering takes place also when there is only a single region.

But there is a simple way how to deactivate region based recovery (and in effect to keep nodal numbering). In sm/src/vtkexportmodule.h just comment out line "#define RBR_SUPPORT" and recompile. This can be used for homogeneous problems containing only one type of material model.

If you need values of any variable for specific node/element you can always get them from oofem output file (possibly using extractor tool for more comfortable extracting).

Any suggestion here is welcome.

>Besides, it would be nice to be able to use a kind of include directive in input files, to separate different sections (e.g., >mesh from problem data).
I agree. Unfortunately could not do it right now, but it is on my todo list.

Borek

Re: Correct node numbering in VTK?

A side-effect question:

Is it implemented the node (or equation) renumbering for optimized solving?

Thanks,

sas

4

Re: Correct node numbering in VTK?

Hi Santiago,

yes, there is a built-in equation renumbering, based on Sloan algorithm. It can be turned on using "-rn" on command-line, or setting "renumber" parameter to 1 in analysis record.

Borek

Re: Correct node numbering in VTK?

Excellent! Thanks

PS: The first option is described in the OOFEM Input Data Format Specification, but the second is not.