analiza aiqbench
This research establishes a rigorous validation pipeline using Lattice Boltzmann solvers to create high-fidelity datasets for training neural operators on complex turbulent flows. By benchmarking Fourier Neural Operators and U-Nets against physical metrics like drag coefficients, the work addresses critical gaps in scientific machine learning where data quality often limits model reliability. The study sets a new standard for evaluating AI surrogates in fluid dynamics through reproducible frameworks that prioritize physics-informed error correction over raw speed gains alone.
A Validated LBM Dataset and Pipeline for Surrogate Modeling of Turbulent 3D Obstructed Channel Flows
arXiv:2606.16765v1 Announce Type: new Abstract: Evaluating neural operators for 3D turbulent flow requires validated datasets with physical benchmarks. We present a reproducible pipeline generating training data for 3D channel flows around generated geometries at Re=1,000-10,000. Our lattice Boltzmann solver with cumulant collision operators is rigorously verified against experimental measurements (Strouhal number, drag coefficients, turbulent fluctuations) with comprehensive grid convergence studies at resolution 1024x512x512. Building upon an established framework, this validated pipeline enables standardized surrogate model comparison. We outline planned systematic evaluation of Fourier Neural Operator and U-Net variants on forecasting, super-resolution, and error correction tasks, using physics-informed metrics to assess turbulent energy cascade representation. Future work will compare computational efficiency between numerical solvers and neural surrogates, exploring practical application. We seek community feedback on our validation approach, planned benchmark methodology, and evaluation priorities for neural operators in turbulent flows.
Abstract:Evaluating neural operators for 3D turbulent flow requires validated datasets with physical benchmarks. We present a reproducible pipeline generating training data for 3D channel flows around generated geometries at Re=1,000-10,000. Our lattice Boltzmann solver with cumulant collision operators is rigorously verified against experimental measurements (Strouhal number, drag coefficients, turbulent fluctuations) with comprehensive grid convergence studies at resolution 1024x512x512. Building upon an established framework, this validated pipeline enables standardized surrogate model comparison. We outline planned systematic evaluation of Fourier Neural Operator and U-Net variants on forecasting, super-resolution, and error correction tasks, using physics-informed metrics to assess turbulent energy cascade representation. Future work will compare computational efficiency between numerical solvers and neural surrogates, exploring practical application. We seek community feedback on our validation approach, planned benchmark methodology, and evaluation priorities for neural operators in turbulent flows.
Machine Learning (cs.LG); Fluid Dynamics (physics.flu-dyn)
arXiv:2606.16765 [cs.LG]
(or arXiv:2606.16765v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2606.16765 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 155 cuvinte
Subiecte conexe
ChannelFlow-Tools: A Configuration-Driven Pipeline for Generating Machine-Learning-Ready Datasets of 3D Obstructed Channel Flows
arXiv · 19 aug. 2026
Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
Hugging Face · 27 iul. 2026
What We Learned Trying to Catch AI Liars: An Aletheia's Quest Retrospective
EleutherAI · 25 aug. 2026
Frontier AI Risk Report (February–March 2026)
METR · 19 mai 2026