Study and prediction on macroscopic characteristics of free spray of typical alcohol fuels through experimentation and the artificial neural network

The use of alcohol fuels in automotive engines holds significant potential for reducing reliance on fossil fuels and mitigating greenhouse gas emissions. This study investigated the macroscopic spray characteristics of methanol, ethanol, isopropanol, n-butanol, and isobutanol using a constant volume chamber and the schlieren method. The effects of varying ambient temperatures, ambient pressures, injection pressures, and fuel temperatures on the spray characteristics of these alcohol fuels were systematically analyzed. An artificial neural network (ANN) was employed to predict spray characteristic parameters, with Bayesian optimization applied for hyperparameter tuning, leading to the development of an optimal ANN model utilizing fuel properties as input. The results revealed that although the overall trends in spray characteristics were consistent among different alcohol fuels under varying conditions, the magnitude of changes differed significantly due to variations in their physicochemical properties. The developed ANN model demonstrated high predictive accuracy (R2 > 0.98), closely matching experimental data. This study highlights the potential of ANN-based models to reduce experimental workload in the development of carbon-neutral fuels and advanced fuel injection systems, contributing to cleaner and more efficient combustion technologies.