Evaluation of Sauter mean diameter prediction models in the jet breakup in cross flow by using ANN and CNN
摘要
When injected into a gaseous crossflow, a liquid jet undergoes complex atomization, breaking into small droplets. Understanding the droplet distribution during jet breakup is crucial; however, conducting CFD simulations of liquid jet breakup in a crossflow is computationally expensive. In this study, SMD prediction models based on the momentum flux ratio and Weber number were developed using ANN and CNN, and their prediction accuracies were compared to that of the RSM model. While both ANN and CNN models exhibit superior accuracy compared to the RSM model, challenges arise in predicting SMD values that change rapidly toward or from zero. To address these challenges and enhance prediction accuracy, an SMD correction method was introduced. This method significantly improves the models’ prediction accuracy, particularly in the upstream region of the domain, where rapid changes in SMD values occur near zero.