| description abstract | Recent improvements in computing power and numerical techniques have enabled us resolve minute details of spray plume behavior and its wall boundary interactions, and made detailed spray simulations using large-eddy simulation (LES) turbulence models available to many more engineers. However, guidelines for parcel-based spray simulation boundary and initial conditions are still based on results from lower-resolution and Reynolds averaged Navier–Stokes (RANS) simulations. Hence, it is necessary to critically examine those assumptions and compare their impact with results using a RANS turbulence model. Three different parameters, including whether a simulation includes a detailed injector tip geometry or a flat surface, whether parcels are initialized at the counterbore exit, which is more common, or at the nozzle exit, and whether to use an experimentally derived rate of injection or one-way coupling with a separate internal nozzle volume of fluid simulation, were examined with an LES turbulence model. Both local data close to the injector and global penetration results were used to compare simulations. Local data, such as the local liquid volume fraction, showed greater variation between the conditions, which may have an impact on mixing and combustion predictions in engine applications. Spray penetration and other global measures demonstrated limited sensitivity to the boundary conditions/initialization procedure. Results were also compared with prior results that used a RANS turbulence model. RANS simulations had overall smoother responses to the changes, as would be expected, but LES simulations showed similar trends in the effects for the measured variables. | |
