铀矿冶2025，Vol.44Issue(2)：29-37,9.DOI:10.13426/j.cnki.yky.2024.09.09
物理信息神经网络在地下水铀迁移对流-弥散方程中的应用
Application of Physics-Informed Neural Networks in Solving the Advection-Diffusion Equation of Uranium Migration in Groundwater
摘要
Abstract
In the in-situ leaching of uranium,uranium migration is influenced by both groundwater flow and solute diffusion,and this process can be effectively modeled using the advection-diffusion e-quation.Accurately modeling the variation of uranium concentration over time and space is crucial for predicting uranium migration in groundwater during in-situ leaching of uranium.Traditional numerical methods,such as the finite difference method,are computationally intensive and prone to errors while dealing with high-dimensional,complex problems.Therefore,this research aims to explore the appli-cability and accuracy of physics-informed neural networks(PINN)in solving the advection-diffusion e-quation.Through numerical simulations of the one-dimensional advection-diffusion equation,and by comparing the PINN solutions with numerical and analytical solutions.The results show that PINN provide higher accuracy and better alignment with the analytical solution over long-term simulations compared to numerical methods.Furthermore,PINN exhibit certain extrapolation capabilities.Addi-tionally,the introduction of dropout enhances the generalization ability and convergence speed of the PINN model,confirming the potential of PINN in solving complex physical problems.关键词
原地浸出采铀/铀迁移/对流-弥散方程/物理信息神经网络/有限差分数值模拟Key words
in-situ leaching of uranium/uranium migration/advection-diffusion equation/physics-in-formed neural network/finite difference numerical simulation分类
天文与地球科学引用本文复制引用
张喆安,彭智婷,白云龙,刘龙成..物理信息神经网络在地下水铀迁移对流-弥散方程中的应用[J].铀矿冶,2025,44(2):29-37,9.基金项目
中核集团基础研究项目(CNNC-JCYJ-202333,电驱动强化CO2+O2地浸机理研究). （CNNC-JCYJ-202333,电驱动强化CO2+O2地浸机理研究）
