Particle clustering in turbulence: Bridging the gap between experiments and Simulations

Direct numerical simulations (DNS) of inertial clustering in turbulence are usually performed with monodisperse particles having uniform-random initial spatial distribution. In these simulations, clustering statistics (e.g. radial distribution function) are obtained using full 3D spatial distribution. However, due to realistic constraints, most experiments to date are done with significant polydispersity, non-uniform particle initial spatial distribution and lower dimensional sampling. Here we pick as an example a recent wind-tunnel study of spray injected droplets using phase Doppler interferometers (PDI). Comparisons of results from both sources are often ambiguous. In view of this, we ran two DNS with polydisperse particles (matching experimental conditions). One of these initiates with a uniform-random spatial distribution while the other initiates with particle concentrated in the center of the simulation volume to imitate spray injection. We study the effect of polydispersity and initial inhomogeneity on the 3D radial distribution function. In addition, we resample the particles in 1D to simulate measurement by a PDI and compare the results with the wind-tunnel data. We will present our findings, with emphasis on how they will ease analysis of experimental results.