Generative AI for fast and accurate statistical computation of fluids
Research paper by Roberto Molinaro and collaborators. arXiv record dated 2025-02-03. Research area: Machine learning.
Resource information
- Authors
- Roberto Molinaro, Samuel Lanthaler, Bogdan Raonić, Tobias Rohner, Victor Armegioiu, Stephan Simonis, Dana Grund, Yannick Ramic, Zhong Yi Wan, Fei Sha, Siddhartha Mishra, Leonardo Zepeda-Núñez
- Research area
- Machine learning
- Application domain
- General AI research
- AI methods
- Diffusion models
- Research topics
- cs.LG, cs.NA, math.NA, physics.flu-dyn
- arXiv record date
- 2025-02-03
- Publication
- arXiv
- Upstream last modified
- Feb 4, 2025
- License
- CC BY 4.0
- Last record update
- Nov 4, 2026
Abstract
We present a generative AI algorithm for addressing the pressing task of fast, accurate, and robust statistical computation of three-dimensional turbulent fluid flows. Our algorithm, termed as GenCFD, is based on an end-to-end conditional score-based diffusion model. Through extensive numerical experimentation with a set of challenging fluid flows, we demonstrate that GenCFD provides an accurate approximation of relevant statistical quantities of interest while also efficiently generating high-quality realistic samples of turbulent fluid flows and ensuring excellent spectral resolution. In contrast, ensembles of deterministic ML algorithms, trained to minimize mean square errors, regress to the mean flow. We present rigorous theoretical results uncovering the surprising mechanisms through which diffusion models accurately generate fluid flows. These mechanisms are illustrated with solvable toy models that exhibit the mathematically relevant features of turbulent fluid flows while being amenable to explicit analytical formulae. Our codes are publicly available at https://github.com/camlab-ethz/GenCFD.