Learning Mesh-Based Simulation with Graph Networks
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:3IQKLXNQrecord.jsonopen to challenge →
read the original abstract
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.
This paper has not been read by Pith yet.
Forward citations
Cited by 25 Pith papers
-
Physics-Guided Dual-Stream Heterogeneous Graph Neural Network for Predicting Full-Field Structural Response of Stiffened Panels
DS-HGNN achieves lower RMSE for stress and displacement prediction on stiffened panels than six benchmark GNN models and matches top accuracy with 19-38% fewer training samples.
-
Domain-Validity-Gated Metamorphic Testing of Scientific ML Surrogates
A domain-validity rubric and MR-card format screen candidate metamorphic relations into auditable test assets for SciML surrogates, separating model violations from out-of-domain applications.
-
Bridging CAD and Data-Driven Design: Attributed Feature Graphs for Engineering Design
Attributed Feature Graphs (AFGs) represent CAD features as attributed nodes and relations as directed edges to enable GNN surrogate models that predict design performance with feature-level interpretability on the Car...
-
Discovering Physical Directions in Weight Space: Composing Neural PDE Experts
Fine-tuning neural PDE operators to regime endpoints reveals a physical direction in weight space that CCM uses to compose accurate merged models for new or extrapolated regimes from metadata or short prefixes.
-
Evaluating LLMs on Large-Scale Graph Property Estimation via Random Walks
EstGraph benchmark evaluates LLMs on estimating properties of very large graphs from random-walk samples that fit in context limits.
-
RynnWorld-4D: 4D Embodied World Models for Robotic Manipulation
A tri-branch diffusion model co-generates RGB, depth, and optical flow from a single RGB-D image, and an inverse dynamics head on its internal latents achieves state-of-the-art bimanual manipulation success rates.
-
NEXUS: Neural Energy Fields for Physically Consistent Contact-Rich 3D Object Dynamics
NEXUS introduces a graph-based neural energy-field model that derives forces from scalar energy and dissipation terms to achieve physically consistent contact-rich 3D dynamics.
-
Point Cloud Sequence Encoding for Material-conditioned Graph Network Simulators
PEACH uses a novel spatio-temporal point cloud sequence encoder plus auxiliary supervision to enable zero-shot adaptation of graph network simulators to unseen physical properties, outperforming mesh-based baselines i...
-
Mask-Morph Graph U-Net: A Generalisable Mesh-Based Surrogate for Crashworthiness Field Prediction under Large Geometric Variation
MMGUNet morphs coarse graph hierarchies with feature-aligned barycentric mapping and uses masked pretraining plus frozen edge layers to improve generalisability of mesh surrogates for crashworthiness prediction under ...
-
Mask-Morph Graph U-Net: A Generalisable Mesh-Based Surrogate for Crashworthiness Field Prediction under Large Geometric Variation
Mask-Morph Graph U-Net morphs coarse graph hierarchies with barycentric parameterization and applies masked supervised pretraining to improve generalizability of hierarchical GNN surrogates for crashworthiness predict...
-
U-HNO: A U-shaped Hybrid Neural Operator with Sparse-Point Adaptive Routing for Non-stationary PDE Dynamics
U-HNO uses adaptive per-point routing in a U-shaped hybrid architecture to achieve state-of-the-art accuracy on PDE benchmarks with sharp localized features.
-
Toward an Operational GNN-Based Multimesh Surrogate for Fast Flood Forecasting
A multimesh graph-neural-network surrogate conditioned on discharge and trained with pushforward produces accurate 6-hour flood forecasts in 0.4 s on GPU versus 180 min on 56 CPU cores for the reference Telemac2D model.
-
Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges
Geometric deep learning provides a unified mathematical framework based on grids, groups, graphs, geodesics, and gauges to explain and extend neural network architectures by incorporating physical regularities.
-
Towards Fast GNN Surrogates for CO2 Migration in Complex Geological Formations
A GNN surrogate with geometry-conditioned anisotropic message passing and autoregressive residual training produces competitive forecasts of gas saturation and liquid density for CO2 storage on the SPE11A benchmark wi...
-
Graph Mamba Operator: A Latent Simulator for Interacting Particle Systems
GraMO couples graph interactions and temporal state updates in one linear recurrence with input-dependent coefficients to simulate N-body, motion, and robotics systems with lower long-horizon error than prior GNN or S...
-
Instrumented data for causal scientific machine learning
Instrumented data augments observations with mechanistic models, uncertainty, and counterfactuals to enable causal interventions via Pearl's do-operator in scientific machine learning.
-
Physically Viable World Models: A Case for Query-Conditioned Embodied AI
Embodied AI requires query-conditioned world models that select the simplest physical abstraction sufficient to answer intervention queries.
-
LEIA: Learned Environment for Interactive Architected Materials
LEIA is a world model for autoregressive 3D simulation of architected materials under interactive loading, benchmarked on MicroPlate and applied to surrogate-guided de novo design search with finite-element validation.
-
Bridging Data and Physics: A Graph Neural Network-Based Hybrid Twin Framework
A GNN-based hybrid twin learns the ignorance component of physics simulations from sparse data and generalizes corrections across meshes, geometries, and loads in nonlinear heat transfer.
-
Towards a Certificate of Trust: Task-Aware OOD Detection for Scientific AI
A score-based diffusion model estimates joint likelihoods of inputs and regression predictions to detect out-of-distribution cases in scientific tasks, with the likelihood correlating to prediction error.
-
Linkify: Learning from Interface-Augmented Assembly Graphs
Linkify augments assembly graphs with corrected interface point clouds and trains GATv2 for masked part prediction, outperforming non-graph baselines on Fusion 360 data.
-
SuperCond-GNN: Scalable Graph Neural Network Surrogate for Superconducting Circuit Simulations
A GNN surrogate model for HTS magnet circuits achieves 4.3% mean MAPE on voltage prediction within the tested design space and supports zero-shot and few-shot generalization to new topologies.
-
Mesh Graph Neural Network Framework for Accelerating Finite Element Simulation for Arbitrary Geometries
Mesh graph network generalizes von Mises stress prediction to unseen 2D plate geometries and loads, outperforming conventional ML models on held-out cases.
-
Learning and Inferring Multiphase Flow Dynamics in Porous Media using Scientific Machine Learning: Application to the "FluidFlower" CO2 Injection Experiment
A CNN surrogate trained on high-fidelity simulations is embedded in MCMC to infer multiphase flow parameters from full-field FluidFlower experiment observations, yielding better simulation-experiment agreement than ma...
-
Performance and Explainability Requirements of Evolutionary Algorithms in Real-World Physics-Informed Optimization
Domain experts require fast convergence and some explainability from evolutionary algorithms in physics-informed optimization, with other needs varying by problem, revealing an application gap.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.