Numerical Investigation of Mesh/Turbulence/Spray Interaction for Diesel applications 2005-01-2115

It has been established by many authors that numerical simulations of Lagrangian Diesel sprays are particularly sensitive to the computational grid and that the most sensitive sub-model is the collision model by O'Rourke [1]. Although recent models by Schmidt & Rutland [2] and Nordin [3] remove this grid dependency, the intention with this paper is not to study drop/drop interaction. The paper focus on the drop/gas interaction, and the collision model has therefore been turned off.

While most papers on Diesel spray simulations focus on modeling the atomization/breakup, few investigate the effect of turbulence/spray interaction. This paper will show that the turbulence model plays a significant role in how the spray behave on different grids, and we present an easy way to reduce the grid dependency by limiting the turbulent length scale in the liquid core region. This idea is based on the suggestion by Stiesch [4], where it was proposed that the turbulent length scale, in the jet region, should be limited by the jet diameter. This constraint was implemented, so that only where liquid is present, the turbulent length scale is limited by an input parameter L_sgs. It is shown that this constraint has a positive effect on the spray behavior and especially the vapor penetration. The difference in spray penetration on the different meshes is reduced such that the penetration on the coarsest mesh becomes more similar to that of the finer mesh, while the penetration on the finer mesh is relatively unchanged.

Since most industrial applications are based on RANS k-ε type models, we investigated three version, the standard k-ε and its RNG and Launder-Sharma modifications. The goal being to evaluate their effect on the spray behavior and sensitivity to the computational mesh variation.

Finally, a comparison with spray penetration experiments in a high-pressure, high temperature constant volume vessel, is performed, where we demonstrate a good agreement between numerical simulations and experiments.