Simulation and Optimization of Venturi Injector by Machine Learning Algorithms

Abstract

This paper discusses the problem of low injection rates from Venturi injectors. The optimal combination of key structural parameters for Venturi injectors was investigated using a simulation software platform based on machine learning algorithms. The research considered different nozzle diameters under an inlet pressure of 0.3 MPa and outlet pressure of 0.1 MPa. For the various nozzle diameters, the optimal ranges of the contraction angle (20°–30°), diffusion angle (8°–10°), throat length (40–50 mm), and ratio of throat diameter to nozzle diameter (1.5–1.66) were found, and the parameter combinations that maximized the injection rate were determined. A regression model was used to predict the maximum injection rate with different nozzle diameters. For a nozzle diameter of 4 mm, the maximum injection rate increased by about 200% compared with the original model. In addition, a regression model for the prediction of the injection rate based on injector structural parameters was construction using data from physical injector models and verified by a three-dimensional printer. The model may be used to quickly and effectively design or predict the injection rate for different structural parameters of the Venturi injector.