CFD Co-Processing Project

Co-processing is the term used for the processing of visualization information during the computational fluid dynamics (CFD) simulation run. With most of the top parallel computers having on the order of 10,000 processors (the highest ranked computer on the http://www.top500.org currently has 212,992 processors) it is unrealistic to expect to be able to efficiently analyze raw results from such machines on typical visualization machines.

Last year Kitware was awarded a Phase I SBIR, “Fluid Dynamics Co-Processing for Unsteady Visualization”, to solve the analysis issues created by the limited storage capacity and slow computational ability of typical visualization machines. The main goal of the project is to develop a “co-processing” ability using VTK for parallel CFD codes.

Currently, Kitware is able to extract CFD data directly from two different CFD codes (a Chimera/overset mesh and a mixed topology unstructured mesh) and co-process such things as mesh blocks, derived physical quantities, cut plane, and isosurfaces and output the results to files. Other future data extracts include streamlines, streaklines and pathlines.

Pressure contours and cut plane of mesh partitioning for an unstructured mesh CFD simulation.

An overset mesh CFD simulation of the velocity magnitude near the surface of a set of helicopter blades.