Machine learning predicting the transport mechanisms and entrainment characteristics of negative buoyant jets

Fountains injected into homogeneous fluids, characterized by combined temperature and concentration effects, are common in both natural and environmental settings. In this study, the capacities of several machine learning models, including support vector regression, multi-layer perceptron, random forests, XGBoost, CatBoost, AdaBoost, and LightGBM, were investigated to clarify the transient flow behavior of fountains. The results indicated that the multi-layer perceptron was superior to the other models as it provided improved coefficient of determination, root mean squared error, and mean absolute error. This study confirmed that the machine learning techniques have great potential to study the transient flow behavior of fountains.