Revealing underlying mechanisms of ambient air mass entrainment around subcooled and flash boiling fuel sprays by a hybrid data-driven framework

Fuel–air interaction inside a gasoline direct injection engine is a key factor affecting the fuel mixture formation and combustion efficiency. In this study, fluorescent particle image velocimetry and Mie scattering are applied to capture the air velocity field and spray morphology simultaneously. This work aims to reveal the dominant factors and underlying mechanisms affecting air mass entrainment under various superheated conditions ranging from subcooled to flash boiling. First, the spray-induced air velocity and mass entrainment rate are examined at the spray boundary. Next, a hybrid data-driven framework consisting of eXtreme Gradient Boosting (XGBoost) and data fitting models is proposed. The XGBoost model identifies two dimensionless parameters, superheated degree (Pa/Ps) and ambient-to-100 kPa density ratio (ρa/ρ100), as the major factors affecting the entrained air mass. Afterward, the data fitting results demonstrated that the air mass entrainment rate can be expressed by a generalized equation using the above two dimensionless parameters, Pa/Ps and ρa/ρ100. The influence of Pa/Ps is more significant, with the air mass entrainment rate decaying exponentially with respect to Pa/Ps. At last, a critical value of 1.7 for Pa/Ps is experimentally determined to distinguish the influence of superheated degree on the air mass entrainment. When Pa/Ps is less than 1.7, Pa/Ps strongly affects the air mass entrainment. Otherwise, the dependence of air entrainment on Pa/Ps is much reduced. The quantitative correlation between air mass entrainment rate, superheated degree, and air density ratio provides good insights into the impact of flash boiling on ambient air mass entrainment.