四川大学学报（自然科学版）2025，Vol.62Issue(5)：1254-1264,11.DOI:10.19907/j.0490-6756.250061
深度学习方法在计算流体力学中的应用
A review on the application of deep learning methods in computational fluid dynamics
摘要
Abstract
In computational fluid dynamics(CFD),constrained by the complexity of solving high dimen-sional nonlinear equations,traditional numerical calculation methods face some bottlenecks such as low com-putational efficiency,time-consuming grid generation,and difficulties in multi-physical field coupling model-ing.As an important branch of machine learning,deep learning paves a new way to this field through the col-laborative optimization of data-driven and physical constraints.In this paper,we briefly review the applica-tions of deep learning in fluid dynamics calculations.First,the basic models of deep learning are introduced,including the structures and characteristics of fully connected neural networks,convolutional neural net-works,recurrent neural networks,and their variants.Then,the applications of the deep learning methods in physical-information-based flow field modeling,operator-learning-based flow field reconstruction(based on DeepONet,FNO and other operator learning methods),and solving fluid problems with interfaces are intro-duced.Finally,the current challenges such as data,model performance and stability,physical interpretabil-ity,and computational resource requirements are summarized.Suggestions for future research directions are provided from aspects such as the fusion of data and physical models,model optimization and innovation,in-terpretability research,and exploration of efficient computational methods,offering a reference for promoting the research of this interdisciplinary field.关键词
流体计算/深度学习/物理信息/算子学习/界面问题Key words
Fluid computation/Deep learning/Physical information/Operator learning/Interface problem分类
数理科学引用本文复制引用
邓扬涛,张莘如,苏志杰,贺巧琳..深度学习方法在计算流体力学中的应用[J].四川大学学报（自然科学版）,2025,62(5):1254-1264,11.基金项目
国家自然科学基金(12371434) （12371434）
