A Validated LBM Dataset and Pipeline for Surrogate Modeling of Turbulent 3D Obstructed Channel Flows

2026-06-15 · Source: Machine Learning · Field: Science & Research — Engineering & Applied Sciences, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

A new, validated pipeline and dataset are presented for surrogate modeling of turbulent 3D obstructed channel flows. This reproducible pipeline generates training data for 3D channel flows around generated geometries at Reynolds numbers ranging from 1,000 to 10,000. It utilizes a lattice Boltzmann solver with cumulant collision operators, rigorously verified against experimental measurements including Strouhal number, drag coefficients, and turbulent fluctuations. Comprehensive grid convergence studies were performed at a resolution of 1024x512x512. This established framework aims to enable standardized comparison of neural operators, such as Fourier Neural Operator and U-Net variants, across forecasting, super-resolution, and error correction tasks, using physics-informed metrics to assess turbulent energy cascade representation.

Key takeaway

For research scientists developing or evaluating neural operators for turbulent fluid dynamics, this validated LBM dataset and pipeline offer a standardized benchmark. You can utilize this resource to rigorously compare Fourier Neural Operator and U-Net variants across forecasting, super-resolution, and error correction tasks. Consider adopting the proposed physics-informed metrics to accurately assess turbulent energy cascade representation in your models, ensuring robust and comparable results.

Key insights

A validated LBM dataset and pipeline enable standardized evaluation of neural operators for 3D turbulent flows.

Principles

• Validated datasets are crucial for neural operator evaluation.
• Grid convergence studies ensure solver accuracy.
• Physics-informed metrics assess turbulent flow representation.

Method

The pipeline generates training data using a lattice Boltzmann solver with cumulant collision operators, verified against experimental measurements and grid convergence studies at 1024x512x512 resolution.

In practice

• Use the pipeline for FNO/U-Net variant evaluation.
• Apply physics-informed metrics for turbulent energy cascade.
• Compare computational efficiency of solvers vs. surrogates.

Topics

• Lattice Boltzmann Method
• Turbulent Flow Modeling
• Neural Operators
• Surrogate Models
• Dataset Validation
• Computational Fluid Dynamics

Best for: AI Scientist, Research Scientist

Related on AIssential

• FLAME 2023: Diving into the Future of Fluid Dynamics and Machine Learning — AI techniques are transforming simulation by integrating physics-based modeling with data-driven machine learning.
• Physics-informed convolutional neural networks for fluid flow through porous media — Physics-informed CNNs accurately predict fluid flow in porous media, significantly accelerating traditional LBM simulations.
• Operator learning for the 2D incompressible Navier-Stokes equations: a conformal prediction approach in the data-scarce regime — The perturbation-based conformal prediction framework quantifies uncertainty in neural operators efficiently, especially in data-scarce regimes, by comparing noisy and original label predictions.
• Neural Operator-Based Surrogate Model for CFD:Helical Coil Steam Generator in Small Modular Reactor — Neural operator-based surrogates, combined with ROMs and multi-scale techniques, enable real-time CFD for SMR digital twins.
• Reformulating Neural Operators in $d+1$ Dimensions for Embedding Evolution — Neural operators redefined in d+1 dimensions using Schrödingerisation better capture system evolution and achieve superior performance.
• Droplet-LNO: Physics-Informed Laplace Neural Operators for Accurate Prediction of Droplet Spreading Dynamics on Complex Surfaces — PI-LNO uses Laplace transforms and physics-informed neural networks to model droplet spreading dynamics efficiently.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.