Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

mesh-based simulation

A mesh-based simulation is a computational modeling technique used to predict the behavior of continuous physical systems by dividing a spatial domain into a structured or unstructured network of interconnected geometric elements, such as triangles, quadrilaterals, or polyhedra. Governing physical laws, typically expressed as partial differential equations for phenomena such as fluid dynamics, structural mechanics, or thermal transport, are approximated and numerically solved across the nodes, edges, and cells of this discretized structure using methods like finite element, finite volume, or finite difference analysis. This spatial discretization allows simulations to handle complex geometries and support adaptive resolution, enabling the density of the mesh elements to be dynamically adjusted to balance numerical accuracy with computational efficiency.

1 item

Learning Mesh-Based Simulation with Graph Networks

Learning Mesh-Based Simulation with Graph Networks

Tobias Pfaff, Meire Fortunato, Alvaro Sanchez-Gonzalez, Peter W. Battaglia

OrganizationsGoogle

Why you should read this

Introduces MeshGraphNets, a graph neural network framework that predicts complex physical dynamics on adaptive meshes across aerodynamics, structural mechanics, and cloth while running up to two orders of magnitude faster than traditional numerical solvers.

Mesh-based simulations are central to modeling complex physical systems in many disciplines across science and engineering. Mesh representations support powerful numerical integration methods and their resolution can be adapted to strike favorable trade-offs between accuracy and efficiency. However, high-dimensional scientific simulations are very expensive to run, and solvers and parameters must often be tuned individually to each system studied. Here we introduce MeshGraphNets, a framework for learning mesh-based simulations using graph neural networks. Our model can be trained to pass messages on a mesh graph and to adapt the mesh discretization during forward simulation. Our results show it can accurately predict the dynamics of a wide range of physical systems, including aerodynamics, structural mechanics, and cloth. The model's adaptivity supports learning resolution-independent dynamics and can scale to more complex state spaces at test time. Our method is also highly efficient, running 1-2 orders of magnitude faster than the simulation on which it is trained. Our approach broadens the range of problems on which neural network simulators can operate and promises to improve the efficiency of complex, scientific modeling tasks.

Added

2026-09-25