[Paraview] Streamlines with an octree/hierarchical grid

[Andy Bauer]

Hi Ethan,

The issue here is that unstructured grids (which is what you have here)
doesn't support non-conforming grids (e.g. octree grids with "hanging"
nodes). So in this case the stream tracer filter is trying to traverse the
grid by using cell adjacency information and because there isn't proper
connectivity information at the interfaces between different octree levels
the grid traversal thinks it has hit a boundary and stops.

The proper way to model this octree grid in VTK is through a
vtkNonOverlappingAMR, vtkOverlappingAMR or vtkHierarchicalBoxDataSet. See
http://www.vtk.org/doc/nightly/html/classvtkUniformGridAMR.html for
information on that and either the VTK or ParaView users guides.

Cheers,
Andy

On Tue, Jul 12, 2016 at 4:27 PM, Ethan Hereth <e.a.hereth at gmail.com> wrote:

> Andy,
>
> Thanks again.
>
> I don't think that there are any gaps, slicing the dataset surely doesn't
> seem to indicate that to me. I don't follow you with respect to using the
> clip filter to 'convert the AMR grid to an unstructured grid with hex
> cells' What does that mean? I use the clip filter a lot, but for isolating
> parts of the mesh that I'm particularly interested in. Am I missing
> something here?
>
> I'll attach a state file demonstrating the issue I'm seeing on a much
> simpler example case as well as the data. Maybe you'll have some time to
> glance at it and point out my blunder? This case has three dimensional flow
> but you'll notice how the streamlines simply terminate in the middle of the
> flow field for no apparent reason, usually where elements in the grid
> change level.
>
> I do sincerely appreciate your help here. It would help me out a lot if
> you could help me figure out how to make nice streamlines with this kind of
> data.
>
> Cheers,
>
> Ethan Alan
>
> On Tue, Jul 12, 2016 at 4:08 PM Andy Bauer <andy.bauer at kitware.com> wrote:
>
>> Hi Ethan,
>>
>> I'm not sure updating to a newer version of ParaView will help here but
>> it should be an easy thing to test out.
>>
>> I think the issue is that the grids aren't quite fitting tightly together
>> like you think they are. It's tough to say why that is without having
>> access to the data set but I'd suggest doing some slice planes through
>> where the streamlines terminate early to see if there are any gaps there.
>> The slice plane will automatically remove the ghost/blanked cells so it
>> should be much easier to see any gaps between the octree leaves.
>>
>> I don't think the cell types should matter here but if you want to test
>> out my assumption you can use the clip filter to convert the AMR grids to
>> unstructured grids with hex cells. This will blow up the memory usage so
>> it's not a good long term solution but ok for testing.
>>
>> Best,
>> Andy
>>
>> On Tue, Jul 12, 2016 at 3:47 PM, Ethan Hereth <e.a.hereth at gmail.com>
>> wrote:
>>
>>> Andy,
>>>
>>> Thanks for your rapid response. The streamlines do indeed seem to end
>>> near interfaces between leaves in the grid, but specifically when they are
>>> at different levels in the octree; i.e. when one element has four neighbors
>>> across any of its faces. I should be able to give an example of this
>>> behavior. But I thought I'd give you this update first. At least most of
>>> the times, in the cases that I've been experimenting with this afternoon,
>>> the reason for termination is 1/'out of domain'
>>>
>>> I'm using a newish version of paraview (4.4.0) but I'll try on 5.0 soon.
>>> Also, I'm not sure that this is important, but he elements are VTK Voxels,
>>> not hexahedra.
>>>
>>> Thanks for your help,
>>>
>>> On Tue, Jul 12, 2016 at 2:30 PM Andy Bauer <andy.bauer at kitware.com>
>>> wrote:
>>>
>>>> Hi Ethan,
>>>>
>>>> I've been able to use streamlines with a general multiblock data set
>>>> that uses blanking. This may be slightly different than your data set but I
>>>> don't see why it shouldn't work for you.
>>>>
>>>> Are the streamlines terminating near the interfaces between different
>>>> leaves of the octree grid? If you look at the ReasonForTermination cell
>>>> data output that should give you an idea. Note that values for that are:
>>>>
>>>>    1. out of domain
>>>>    2. not initialized
>>>>    3. unexpected value
>>>>    4. streamline max length reached
>>>>    5. maximum number of integration time steps reached
>>>>    6. stagnation (velocity too close to zero)
>>>>
>>>>
>>>> There are termination criteria for the filter (e.g. the options under
>>>> Integration Parameters and Streamline Parameters in the Properties panel,
>>>> some of them are advanced options so click on the gear to get access to
>>>> those) so you may want to play with those.
>>>>
>>>> If you still have problems with this, could you share your data set for
>>>> us to test with?
>>>>
>>>> Cheers,
>>>> Andy
>>>>
>>>> On Tue, Jul 12, 2016 at 2:14 PM, Ethan Hereth <e.a.hereth at gmail.com>
>>>> wrote:
>>>>
>>>>> Good afternoon,
>>>>>
>>>>> I am trying to generate streamlines on a solution generated on an
>>>>> octree style grid with hanging nodes/faces. I have cell centered data. The
>>>>> problem is that the streamlines always end too early. I assume this has to
>>>>> do with the incomplete connectivity inherent within this type of grid.
>>>>>
>>>>> My questions is, has anybody dealt with this before? And if so, what
>>>>> might I try to get this to work?
>>>>>
>>>>> I've tried creating a 3D Delaunay tessellation from the cell centers
>>>>> (obtained from the Cell Centers filter) but, at least for this case,
>>>>> Paraview crashes. Also, converting the cell data to point data does not
>>>>> take care of the problem.
>>>>>
>>>>> I've thought about attempting to interpolate my data onto a uniform
>>>>> structured mesh and use that new dataset as the source for the streamline
>>>>> filter, but I'm not sure how to go about doing that.
>>>>>
>>>>> Does anybody done this before or have any ideas about how I might
>>>>> solve this problem?
>>>>>
>>>>> Thanks for your time,
>>>>>
>>>>> Ethan Alan
>>>>>
>>>>> _______________________________________________
>>>>> Powered by www.kitware.com
>>>>>
>>>>> Visit other Kitware open-source projects at
>>>>> http://www.kitware.com/opensource/opensource.html
>>>>>
>>>>> Please keep messages on-topic and check the ParaView Wiki at:
>>>>> http://paraview.org/Wiki/ParaView
>>>>>
>>>>> Search the list archives at: http://markmail.org/search/?q=ParaView
>>>>>
>>>>> Follow this link to subscribe/unsubscribe:
>>>>> http://public.kitware.com/mailman/listinfo/paraview
>>>>>
>>>>>
>>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview/attachments/20160713/654b8560/attachment.html>