Numerical Study of Gasoline Direct Injection Sprays across Different Operating Conditions using a Machine-Learning-Based Rate of Injection

Gasoline Direct Injection (GDI) is a key technology for increasing engine efficiency and reducing criteria pollutants, often in combination with boosted operation and engine downsizing. The development of advanced GDI injection systems is increasingly focused on refining the in-cylinder spray process, encompassing spray formation, mixing, and combustion, to achieve both homogeneous and stratified combustion. Given the broad spectrum of operating conditions, from cold-start to early and late injections at different engine loads, accurately modeling the spray formation and its evolution presents significant challenges. This study reports on a computational fluid dynamics investigation of spray morphology across different injection pressures under both cold-start and late-injection conditions. Due to a lack of information regarding the rate of injection (ROI) for the targeted injectors, a recently developed machine-learning-based model is used to extract the ROI from optical data of projected liquid volume (PLV). Comprehensive validation is conducted to assess the quality of the spray morphology predictions, including 2-D PLV maps and 3-D liquid volume fraction. Apart from the penetration length, a second-order momentum comparison is introduced to ensure adequate accuracy of the simulation results in terms of spray evolution. This research offers valuable insights for enhancing spray modeling and introduces valid alternative options for ROI estimation.