Abstract:Complex spatial and temporal structures are inherent characteristics of turbulent fluid flows and comprehending them poses a major challenge. This comprehesion necessitates an understanding of the space of turbulent fluid flow configurations. We employ a diffusion-based generative model to learn the distribution of turbulent vorticity profiles and generate snapshots of turbulent solutions to the incompressible Navier-Stokes equations. We consider the inverse cascade in two spatial dimensions and generate diverse turbulent solutions that differ from those in the training dataset. We analyze the statistical scaling properties of the new turbulent profiles, calculate their structure functions, energy power spectrum, velocity probability distribution function and moments of local energy dissipation. All the learnt scaling exponents are consistent with the expected Kolmogorov scaling. This agreement with established turbulence characteristics provides strong evidence of the model's capability to capture essential features of real-world turbulence.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: How to use Diffusion Generative Models in deep learning to understand and generate the statistical properties of turbulence. Specifically, the author hopes to train a Denoising Diffusion Probabilistic Model (DDPM) to learn the turbulence vorticity distribution and generate turbulence snapshots that conform to the solutions of the two - dimensional incompressible Navier - Stokes equations. By analyzing the statistical scale properties of these generated samples, verify whether the model can capture the essential characteristics of real - world turbulence. ### Main research questions 1. **Generate new turbulence samples**: Can turbulence samples with real physical characteristics that are different from the training data set be generated? 2. **Consistency of statistical properties**: Can the generated turbulence samples maintain the same statistical properties as actual turbulence, such as energy spectra, structure functions, and local energy dissipation rates? 3. **Verification of Kolmogorov scaling law**: Do the generated turbulence samples conform to the scaling law proposed by Kolmogorov? ### Specific questions - **Inverse cascade phenomenon**: Can the phenomenon of energy transfer from high - frequency to low - frequency in two - dimensional turbulence be reproduced in the generated samples? - **Structure function**: Does the structure function of the generated samples match the theoretical expectation? - **Local energy dissipation**: Does the local energy dissipation of the generated samples conform to the theoretical prediction? ### Method overview - **Model selection**: Use the Denoising Diffusion Probabilistic Model (DDPM) as the generation model. - **Data set**: Train based on 5,000 vorticity images obtained from numerical simulations. - **Evaluation metrics**: Evaluate the model performance by comparing the statistical properties such as energy spectra, structure functions, and local energy dissipation between the generated samples and the real samples. ### Research significance If successful, this research will provide a new tool for learning and generating the statistical properties of turbulence, which can be used to increase the amount of data for turbulence analysis and improve the understanding of complex turbulence systems. In addition, compared with other generation models (such as GAN and VAE), the diffusion model shows better performance in generating realistic turbulence samples.

Turbulence Scaling from Deep Learning Diffusion Generative Models

Synthetic Lagrangian Turbulence by Generative Diffusion Models

Bayesian Conditional Diffusion Models for Versatile Spatiotemporal Turbulence Generation

Multi-scale Reconstruction of Turbulent Rotating Flows with Generative Diffusion Models

Comparison of Generative Learning Methods for Turbulence Modeling

Neural Differentiable Modeling with Diffusion-Based Super-resolution for Two-Dimensional Spatiotemporal Turbulence

Universality and scaling in compressible turbulence

Forecasting small scale dynamics of fluid turbulence using deep neural networks

Unfolding Time: Generative Modeling for Turbulent Flows in 4D

Recurrent flow patterns as a basis for turbulence: predicting statistics from structures

From Zero to Turbulence: Generative Modeling for 3D Flow Simulation

A new turbulence model based on scale decomposition

Numerical simulations of a stochastic dynamics leading to cascades and loss of regularity: applications to fluid turbulence and generation of fractional Gaussian fields

Benchmarking Autoregressive Conditional Diffusion Models for Turbulent Flow Simulation

Conditional Neural Field Latent Diffusion Model for Generating Spatiotemporal Turbulence

The turbulent cascade in five dimensions

A transformer-based synthetic-inflow generator for spatially developing turbulent boundary layers

A multiscale and multicriteria Generative Adversarial Network to synthesize 1-dimensional turbulent fields

An efficient training method to learn a model of turbulence

Scaling transition of active turbulence from two to three dimensions