Entity · paper

Label-Free Finite-Volume-Residual Training of Attention Graph Neural Networks for Coupled Thermo-Fluid Fields

paperactiveprovisionallabel-free-finite-volume-residual-training-of-attention-graph-neural-networks-for-coupled-thermo-fluid-fields-9b16ccca·1 events·first seen Jul 23, 2026

Aliases: Label-Free Finite-Volume-Residual Training of Attention Graph Neural Networks for Coupled Thermo-Fluid Fields

More like this (12)

Graph Convolutional Attention: A Spectral Perspective on Graph Denoising and Diffusion Graph Convolutional Attention: A Spectral Perspective on Graph Denoising and Diffusion Graph Convolutional Attention Graph Attention Network Conditional Generative Flow Networks A Physics-Informed Fourier-Wavelet Transformer for Multiscale Computational Fluid Dynamics Surrogate Modeling Beyond Fully Random Masking: Attention-Guided Denoising and Optimization for Diffusion Language Models NLL-Guided Full-Attention Layer Selection for Training-Free Sliding-Window Adaptation Cross-Layer Sparse Attention Graph Neural Networks MiniMax Sparse Attention Neuron-Aware Data Selection for Annotation-Free LLM Self-Distillation

Recent events (1)

4arXiv · cs.LG·Jul 23, 2026·source ↗

Label-free training of graph neural networks for 3D thermo-fluid simulation using finite-volume residuals

Researchers propose training attention graph neural networks (GNNs) as neural surrogates for 3D coupled thermo-fluid fields by minimizing finite-volume method (FVM) residuals directly on the mesh, eliminating the need for labeled CFD simulation data. The approach is evaluated across four scenarios including steady-state and parametric transient cases, achieving 2.3–2.8% normalized RMSE on steady benchmarks and outperforming a supervised baseline on transient cases. The method reduces model development cost by avoiding expensive data generation from conventional numerical solvers.

Label-Free Finite-Volume-Residual Training of Attention Graph Neural Networks for Coupled Thermo-Fluid Fields