Physics-Embedded Neural Networks: Graph Neural PDE Solvers with Mixed Boundary Conditions

TL;DR

This paper introduces physics-embedded neural networks based on E(n)-equivariant GNNs that effectively incorporate boundary conditions and predict long-term states in PDEs, outperforming classical and existing ML models in speed and accuracy.

Contribution

It presents a novel GNN-based PDE solver that explicitly considers boundary conditions and uses an implicit method for long-term predictions, enhancing reliability and generalization.

Findings

Outperforms classical solvers in speed and accuracy.

Successfully models flow phenomena in complex shapes.

Demonstrates high generalization on various geometries.

Abstract

Graph neural network (GNN) is a promising approach to learning and predicting physical phenomena described in boundary value problems, such as partial differential equations (PDEs) with boundary conditions. However, existing models inadequately treat boundary conditions essential for the reliable prediction of such problems. In addition, because of the locally connected nature of GNNs, it is difficult to accurately predict the state after a long time, where interaction between vertices tends to be global. We present our approach termed physics-embedded neural networks that considers boundary conditions and predicts the state after a long time using an implicit method. It is built based on an E(n)-equivariant GNN, resulting in high generalization performance on various shapes. We demonstrate that our model learns flow phenomena in complex shapes and outperforms a well-optimized classical solver and a state-of-the-art machine learning model in speed-accuracy trade-off. Therefore, our model can be a useful standard for realizing reliable, fast, and accurate GNN-based PDE solvers. The code is available at https://github.com/yellowshippo/penn-neurips2022.