As efforts to improve efficiency and develop alternative fuels for internal combustion engines continue, the study of high-pressure injection sprays remains a critical research subject. The near-field region of these sprays remains difficult to capture experimentally and is prohibitively expensive to fully resolve numerically. The first part of this thesis is focused on assessment of a common, engineering-level spray model, the Lagrangian-Eulerian (LE) method, and illustrates its limitations in capturing the near-field accurately relative to higher-fidelity Volume of Fluid (VoF) simulations. In a modified LE implementation, droplet size is assigned from VoF data, but the resulting model still does not capture radial dispersion within the near-field satisfactorily. Specifically, in comparisons of mean projected mass, a varying spreading noted in the VoF is not observed. These findings motivate the development of an improved Lagrangian-based method for the near-field.
The second part of the thesis presents the development and assessment of a new Lagrangian method, the Near-field Lagrangian Dispersion Model (NFLDM), based on quantification of near-field dynamics from VoF simulations. Liquid transport in the NFLDM is governed by a mean flow treatment in the axial direction of the spray with a stochastic treatment for radial dispersion, utilizing VoF-based dynamic fields represented by a self-similarity model. The NFLDM demonstrates a significant improvement in predicting mean liquid mass distribution over conventional LE models in comparison with baseline VoF simulations spanning a range of relevant conditions, including multiple nozzles and fuels and a range of ambient pressures and injection velocities. Further assessments are conducted to validate the model’s performance at combustion conditions, including evaluations of single-hole combustion simulations and an implementation of a High Speed Direct Injection (HSDI) Diesel engine simulation. The results of these assessments demonstrate notable differences between the NFLDM and conventional LE models.