Impact of liquid injector pressure on gas flow characteristics and evaporation rates: A combined Eulerian-Lagrangian approach and PSO-optimized XGBoost model

Abstract

In this study, the effect of increasing the pressure of liquid jet spray nozzles on the quantity of hot gas flow inside the cylinder and the evaporation rate of the sprayed liquid has been numerically investigated using the Eulerian-Lagrangian method. Additionally, the molar ratio of water vapor along the geometry has been predicted using an optimized machine learning model. Liquid water with a temperature of 300 K is injected into the cylinder containing a high-speed gas flow with a temperature of 2500 K by an arrangement of 5 injectors under different pressures. In this study, changes in static quantities and total flow such as pressure, temperature, mach number, evaporation rate, mole fraction and flow rate changes before and after liquid injection into the cylinder and also the effect of increasing injector pressure from 10 MPa to 50 MPa on this quantity have been investigated. Finally, the hyperparameters of the XGBoost model were optimized using the Particle Swarm Optimization (PSO) algorithm to predict the average molar ratio of water vapor under various conditions, resulting in the development of a combined PSO-XGBoost model. The results show that initially increasing the pressure of the liquid injectors causes significant changes in the gas flow quantities, but increasing the pressure from a certain range does not have a significant effect on the changes in the flow parameters. Increasing the liquid injector pressure up to about 20 MPa has a significant effect on the pressure and Mach number and up to about 30 MPa on the total temperature, increasing the evaporation rate and mass flow rate, but increasing the injector pressure more than these values does not have much effect on the flow characteristics. Also, the results obtained from predicting the molar ratio of water vapor using the PSO-XGBoost model demonstrated a strong correlation with the results from the numerical solutions, highlighting the model accuracy in predicting the molar ratio of water vapor under the complex flow conditions examined.