A comparative study of two-phase coupling models for a sparse-Lagrangian particle method

Simulations of spray combustion in statistically homogeneous turbulence with different droplet loadings are performed using a sparse-Lagrangian particle method called MMC. We compare different models for the distribution of the droplet source terms to the notional particles with corresponding CP-DNS data and solutions of a dense particle method that employs standard models for the two-phase coupling. Good agreement of unconditional mean and rms of the reactive scalars is found for both, the sparse particle method utilizing a one-to-one coupling technique between the droplets and the stochastic particles, and the dense particle method where the droplet mass is distributed equally to all particles within the computational cell. MMC predictions of the mixture fraction variance are somewhat superior to predictions by the dense particle method, but conditional fluctuations are underpredicted in both models. In contrast, MMC gives accurate predictions of the conditional mean temperature and its conditional variance if the droplet mass is preferentially distributed to particles closest to saturation conditions. The unconditional variance is, however, significantly overpredicted due to a pronounced peak at saturation conditions in the composition PDF. Attempts to incorporate the latter approach into the one-to-one coupling strategy are presented and allow for some control of gas-phase variance generation due to droplet evaporation. However, improvements have only been achieved for transitional periods but not for the entire duration of the spray combustion process, and - in the absence of suitable blending functions between the different approaches - the one-to-one coupling strategy currently seems the most appropriate choice for two-phase coupling in MMC.