基于DeepFeatureNet的侧向受限垂向浮力射流模拟研究OA

Buoyancy jets are widely present in environmental and industrial fields,playing a key role in applications such as wastewater discharge,ocean diffusion,and cooling water release.Due to the dynamics of buoyancy jets involving density differences,turbulent mixing,and interactions with the surrounding environment,predicting their behavior under different conditions becomes extremely complex.Therefore,a jet simulation method based on DeepFeatureNet is introduced and evaluated using a laterally constrained vertical buoyancy jet as an example.The method is also systematically compared with traditional machine learning methods,such as Support Vector Regression,Decision Trees,and Lasso Regression.Through comparing the predictive performance of different models,the applicability,strengths and weaknesses of each approach on complex datasets are assessed.The results show that DeepFeatureNet achieves excellent predictive accuracy on both the training and testing datasets,significantly outperforming traditional methods.Although Support Vector Regression and Decision Tree models also perform well,their predictive capabilities are slightly inferior to that of DeepFeatureNet.In contrast,Lasso Regression demonstrates relatively weaker performance in capturing data features.The stability of the models is further analyzed through confidence distribution plots,revealing that the confidence scores of DeepFeatureNet are more concentrated,indicating more stable predictive performance and a reduced likelihood of extreme biases.The findings highlight the significant potential of DeepFeatureNet in simulating fluid dynamics,pollutant transport and diffusion in water.The research findings provide valuable reference for future studies of water pollution simulations.