Shot-to-shot spray variability is recognized as an important contributor to cycle-to-cycle variations in direct-injection engines. Large-eddy Simulations (LES) have been used to resolve more of the gas-phase turbulent fluctuations in Computational Fluid Dynamics (CFD) simulations, but the sources of shot-to-shot spray variability are not directly modeled in the Lagrangian parcel approach used most often for engine fuel spray simulations. Instead, the variability comes from variations in the spray boundary conditions. Currently, the most common way to introduce shot-to-shot variability in Lagrangian spray simulations is to vary the random seed used in the spray models for each spray realization. This approach was used to compare against recent near-nozzle (up to ≈9 mm downstream) projected mass density (PMD) measurements of a single-hole diesel injector. The injector has a nominal nozzle diameter of 110 μm, though analysis of experimental data indicates the actual diameter is approximately 118 μm. The measurements were acquired under non-vaporizing conditions using fast radiography with high-intensity X-rays available at the Advanced Photon Source at Argonne National Laboratory.
Spray simulations used the Dynamic Structure LES turbulence model along with a Lagrangian parcel approach with the blob injection model and modified KH-RT break-up model. 10 spray realizations were simulated each at three different minimum cell sizes, 62.5 μm, 93.75 μm and 125 μm. This is much finer mesh resolution than is typically performed for Lagrangian spray calculations and allows the simulations to capture more of the gas-phase turbulent fluctuations that develop from the different random seeds. The effect of the grid resolution on the predicted shot-to-shot variability was investigated. Larger minimum cell sizes increased the predicted variability in projected mass density slightly. This difference in predicted variability was less apparent in global spray quantities such as penetration length.
Initial turbulence kinetic energy (TKE) variations have also been proposed as a method to introduce shot-to-shot variations in spray simulations. 10 further simulations were run where the initial turbulence intensity was varied randomly between 0.1 and 1 m2/s2. The resulting spray variability was similar, but slightly lower than that introduced by varying the random seed.