MeshODENet: A Graph-Informed Neural Ordinary Differential Equation Neural Network for Simulating Mesh-Based Physical Systems
Journal of Applied Mechanics · 01.05.2026 · 93 (5)
DOI: 10.1115/1.4071488 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Applied Mechanics, online erschienen am 07.04.2026. Verfasst von Kangzheng Liu und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 28 Literaturangaben. Gefördert von: Arizona State University, National Academy of Sciences, National Institute of Food and Agriculture.
Abstract (Originalsprache)
The simulation of complex physical systems using a discretized mesh is a cornerstone of applied mechanics, but traditional numerical solvers are often computationally prohibitive for many-query tasks. While graph neural networks (GNNs) have emerged as powerful surrogate models for mesh-based data, their standard autoregressive application for long-term prediction is often plagued by error accumulation and instability. To address this, we introduce MeshODENet, a general framework that synergizes the spatial reasoning of GNNs with the continuous-time modeling of neural ordinary differential equations. We demonstrate the framework’s effectiveness and versatility on a series of challenging structural mechanics problems, including different elastic bodies undergoing large, nonlinear deformations. The results demonstrate that our approach significantly outperforms baseline models in long-term predictive accuracy and stability, while achieving substantial computational speed-ups over traditional solvers. This work presents a powerful and generalizable approach for developing data-driven surrogates to accelerate the analysis and modeling of complex structural systems.
Förderung
- Arizona State University
- National Academy of Sciences
- National Institute of Food and Agriculture
Zitieren
APA
Liu, K., & Ma, L. (2026). MeshODENet: A Graph-Informed Neural Ordinary Differential Equation Neural Network for Simulating Mesh-Based Physical Systems. Journal of Applied Mechanics, 93(5). https://doi.org/10.1115/1.4071488
Vancouver
Liu K, Ma L. MeshODENet: A Graph-Informed Neural Ordinary Differential Equation Neural Network for Simulating Mesh-Based Physical Systems. Journal of Applied Mechanics. 2026;93(5). doi:10.1115/1.4071488
BibTeX
@article{Liu2026,
author = {Liu, Kangzheng and Ma, Leixin},
title = {MeshODENet: A Graph-Informed Neural Ordinary Differential Equation Neural Network for Simulating Mesh-Based Physical Systems},
journal = {Journal of Applied Mechanics},
year = {2026},
volume = {93},
number = {5},
doi = {10.1115/1.4071488}
}
EUVIMED