2016 Volume 68 Issue 1 Pages 43-49
We conducted an experiment and large-eddy simulations (LESs) of turbulent dispersion of a passive tracer in a cavity space. LESs were run on simple orthogonal numerical grids that had a grid spacing of H/200 and H/400 (where H is the cavity height). Although the grid spacing of H/200 in the cavity was sufficient to allow prediction of mean and variance of concentration, it underestimated concentration fluctuation at high frequency, as compared to the experiment. However, LES with a grid spacing of H/400 successfully provided probability density functions and power spectra of concentration that agreed very well with those in the experiment.