Abstract:This paper presents a new physics‐driven GraphSAGE method to simulate physical processes governed by PDEs. Graph representations are utilized to reduce the demand for evaluated points. Discrete weak forms of PDEs are used to supervis the training. A distance‐related edge feature and a feature mapping strategy are proposed to promote the training for irregular PDEs. As a result, well‐trained surrogate models are several times faster than the finite element method. Physics‐informed neural networks (PINNs) have successfully addressed various computational physics problems based on partial differential equations (PDEs). However, while tackling issues related to irregularities like singularities and oscillations, trained solutions usually suffer low accuracy. In addition, most current works only offer the trained solution for predetermined input parameters. If any change occurs in input parameters, transfer learning or retraining is required, and traditional numerical techniques also need recomputation. In this work, a physics‐driven GraphSAGE approach (PD‐GraphSAGE) based on the Galerkin method and piecewise polynomial nodal basis functions is presented to solve computational problems governed by irregular PDEs and to develop parametric PDE surrogate models. This approach employs graph representations of physical domains, thereby reducing the demands for evaluated points due to local refinement. A distance‐related edge feature and a feature mapping strategy are devised to help training and convergence for singularity and oscillation situations, respectively. The merits of the proposed method are demonstrated through a couple of cases. Moreover, the robust PDE surrogate model for heat conduction problems parameterized by the Gaussian random field source is successfully established, which not only provides the solution accurately but is several times faster than the finite element method in the experiments.

PhyMPGN: Physics-encoded Message Passing Graph Network for spatiotemporal PDE systems

PhyGNNet: Solving spatiotemporal PDEs with Physics-informed Graph Neural Network

Combining physics-informed graph neural network and finite difference for solving forward and inverse spatiotemporal PDEs

A fully differentiable GNN-based PDE Solver: With Applications to Poisson and Navier-Stokes Equations

Physics-informed MeshGraphNets (PI-MGNs): Neural finite element solvers for non-stationary and nonlinear simulations on arbitrary meshes

Learning time-dependent PDE solver using Message Passing Graph Neural Networks

PhyCRNet: Physics-informed convolutional-recurrent network for solving spatiotemporal PDEs

Physics-Informed Graph-Mesh Networks for PDEs: A hybrid approach for complex problems

TGN: A Temporal Graph Network for Physics Prediction

RBF-MGN:Solving spatiotemporal PDEs with Physics-informed Graph Neural Network

A Spatial-Temporally Adaptive PINN Framework for 3D Bi-Ventricular Electrophysiological Simulations and Parameter Inference

Physics-informed graph neural Galerkin networks: A unified framework for solving PDE-governed forward and inverse problems

Mesh-based GNN surrogates for time-independent PDEs

TG-PhyNN: An Enhanced Physically-Aware Graph Neural Network framework for forecasting Spatio-Temporal Data

Multi-resolution partial differential equations preserved learning framework for spatiotemporal dynamics

A Physics‐Driven GraphSAGE Method for Physical Field Simulations Described by Partial Differential Equations

P^2C^2Net: PDE-Preserved Coarse Correction Network for Efficient Prediction of Spatiotemporal Dynamics

Physics-embedded Fourier Neural Network for Partial Differential Equations

A Physics-driven GraphSAGE Method for Physical Process Simulations Described by Partial Differential Equations