Students - David Adams, Zachary Benson, David Brewer, and Dillon Lisk
We designed an experiment to analyze the effect of Mach number and model resolution for the Lattice-Boltzmann method, a solver for the Boltzmann Transport Equation that approximates incompressible flows. Our goal was to determine the mesh size and Mach number required to determine accurate coefficient of drag values for incompressible flow (RE=20) over a cylinder. The Lattice-Boltzmann method requires the region of interest to be divided into a mesh of arbitrary dimensions. It then analyzes the behavior of the flow in each mesh cell. Smaller meshes are pretty quick to solve, but result in relatively imprecise results, while large meshes, which result in improved accuracy, become exponentially more costly from a computational perspective. To improve our computational efficiency, we made use of Matlab’s Parallel Computing Toolbox and CUDA GPU implementations as part of this project. Our results show how many degrees of freedom are generally required for an accurate coefficient of drag. One can see that the Mach # chosen has a significant effect on the results. A PDF version of a poster we presented at the Union University Scholarship Symposium can be found here.
