Assessment of subgrid dispersion models for large-eddy simulations of turbulent jet flows with dilute spray droplets

High-fidelity simulations for polydispersed sprays in the Eulerian–Lagrangian framework need to incorporate subgrid-scale effects in the par- ticle evolution equations. Although the quasi-linear evaporation rate formulation captures evaporating droplet statistics, further improve- ment is required when subgrid-scale velocity effects become essential. The subgrid dispersion model strongly affects droplets spatial distribution, and subsequently net evaporation rate, depending on how rapidly they are dispersed into the dry air region. The main original contribution of this study is to assess the performances of a number of commonly used dispersion models in a consistent manner, against a reference direct simulation results. The models considered are (i) discrete random walk, (ii) approximate deconvolution method, (iii) sto- chastic model based on the Langevin equation, and (iv) combined approximate deconvolution method with the Langevin equation. Mass and enthalpy transfer source terms together with droplet diameters and particle distributions were compared against corresponding direct numerical and large-eddy simulations without a model as reference cases. Numerical results at low Stokes and moderate Reynolds numbers indicate that the dispersion model choice does not affect Eulerian field averages or fluctuations. However, proper dispersion models are essential to capture droplet distributions in the far-field region after jet breakup for Stokes number smaller than unity. The unclosed Lagrangian momentum equation without any dispersion model most accurately reproduces direct numerical simulation in the near field.