Challenge
Output from SPH simulations is usually given by point clouds with millions of entities (billions in future), each of which contains local information on physical quantities such as temperature or mass density. While specialized tools produce visually appealing volume renderings (e.g. SPLOTCH), most state-of-the-art visualization packages fail to handle point clouds properly.
On the other hand, these packages offer a plethora of attractive possibilities for quantitative data analysis of gridded data, e.g., for producing contour plots on arbitrary planes through the simulation domain.
Solution
A code package was developed at RZG to create unstructured grids from SPH point data. The fast three dimensional Delaunay triangulation provided by
qhull is used. The resulting unstructured grid is written together with the point data in a legacy file format which can be read by applications such as Paraview or VisIt. A serial domain decomposition technique is implemented to keep the memory footprint of the program low. Hence, datasets of arbitrary size can be handled.
Cooperation
Klaus Reuter (RZG), Claudia Simion (TUM), Claudio Dalla Vecchia (MPE), Markus Rampp (RZG), Sadegh Khochfar (MPE)
Source code
The code package may be obtained upon request for use on RZG systems.
References
