Sie befinden Sich nicht im Netzwerk der Universität Paderborn. Der Zugriff auf elektronische Ressourcen ist gegebenenfalls nur via VPN oder Shibboleth (DFN-AAI) möglich. mehr Informationen...
Physics‐Informed Neural Network Method for Forward and Backward Advection‐Dispersion Equations
Ist Teil von
Water resources research, 2021-07, Vol.57 (7), p.n/a
Ort / Verlag
Washington: John Wiley & Sons, Inc
Erscheinungsjahr
2021
Quelle
Wiley Online Library
Beschreibungen/Notizen
We propose a discretization‐free approach based on the physics‐informed neural network (PINN) method for solving the coupled advection‐dispersion equation (ADE) and Darcy flow equation with space‐dependent hydraulic conductivity K(x). In this approach, K(x), hydraulic head, and concentration fields are approximated with deep neural networks (DNNs). We assume that K(x) is given by its values on a grid, and we use these values to train the K DNN. The head and concentration DNNs are trained by minimizing the residuals of the flow equation and ADE and using the initial and boundary conditions as additional constraints. The PINN method is applied to one‐ and two‐dimensional forward ADEs, where its performance for various Péclet numbers (Pe) is compared with the analytical and numerical solutions. We find that the PINN method is accurate with errors of less than 1% and outperforms some conventional discretization‐based methods for large Pe. Next, we demonstrate that the PINN method remains accurate for the backward ADEs, with the relative errors in most cases staying under 5% compared to the reference concentration field. Finally, we show that when available, the concentration measurements can be easily incorporated in the PINN method and significantly improve (by more than 50% in the considered cases) the accuracy of the PINN solution of the backward ADE.
Key Points
Physics‐informed neural network (PINN) method is proposed for forward and backward advection‐dispersion equations
The physics‐informed neural network (PINN) method has several advantages over some grid‐based discretization methods for high Péclet number problems
The physics‐informed neural network (PINN) method is accurate for the considered backward advection‐dispersion equations (ADEs) that otherwise must be treated as computationally expensive inverse problems