subsampled or unstructured grid

First let me congratulate for this very promising project.

A friend of mine has asked me to evaluate OpenLB for use with his graphics design jobs. Currently he is using a rather old commercial tool which he has to replace, preferably with open source software.

His usual applications visualize all kinds of objects with water or beverages circulating around them. Because the objects can contain small scale structures, I assume that for an acceptable efficiency we need a subsampled LBM grid (by subsampled I mean that in the proximity of boundaries the coarsest grid is divided along 1,2 or 3 dimensions into 2,4 or 8 subcells, repeated recursively until the desired resolution is reached.)

Does OpenLB allow this subsampling, probably on an octree grid? I also read about LBM on unstructured grids. Are any of these implemented or planned?

Thank you so much

No, there is currently no support for locally refined grids in OpenLB. We do plan on adding such a feature to the project and are actively developing concepts and algorithms towards a code with grid refinement. My guess is that this will take quite a while, though. Grid-refinement abilities in OpenLB will certainly be based on structured grids, similar to the ones in the codes which use octrees to structure the data.

My suggestion to your friend is to simply try out OpenLB on a regular grid without sub-sampling. The LB code is actually much faster than what you get from many other CFD products, and you may obtain what you need by simply using a huge grid. This is especially true as you aim at producing nice graphics, as opposed to validating scientific data. Also, don’t hesitate to run the code on a parallel machine, if you have one available, as parallelization is automatic. Furthermore, a distinct advantage of using a regular grid is that it is more straightforward to extract graphics from the result of the simulation.

Actually, if you are interested in computer graphics based on LB simulations, you should also have a look at the web site of Nils Thuerey:

I think that Nils’ LB code, containing a model for free-surface flows, is available with an open-source license (he calls it El’Beem, or something alike).

I found this recent, 2008, papar with adaptive nested-grid and adaptive local-time stepping, here. It may be useful for future plans of grid refinement of OpenLB.

Best regard,

jlatt, you say there is no support for grid refinement in openlb.
Now I read at
“Grid refinement
The approach chosen in the OpenLB project for the representation of an inhomogeneous or anisotropic numerical space is to use local grid refinement techniques…”

Can somebody please clarify: Does OpenLB support grid refinement or doesn’t it?


No for the moment it does not support grid refinement, but it is currently under development.

Does it now?

The grid refinement will not be available in the next release. We don’t know exactly this one will be. You’ll have to be patient… :wink: