Generative diffusion models for synthetic trajectories of heavy and light particles in turbulence

Tianyi Li,Samuele Tommasi,Michele Buzzicotti,Fabio Bonaccorso,Luca Biferale

2024-06-07

Abstract:Heavy and light particles are commonly found in many natural phenomena and industrial processes, such as suspensions of bubbles, dust, and droplets in incompressible turbulent flows. Based on a recent machine learning approach using a diffusion model that successfully generated single tracer trajectories in three-dimensional turbulence and passed most statistical benchmarks across time scales, we extend this model to include heavy and light particles. Given the particle type - tracer, light, or heavy - the model can generate synthetic, realistic trajectories with correct fat-tail distributions for acceleration, anomalous power laws, and scale dependent local slope properties. This work paves the way for future exploration of the use of diffusion models to produce high-quality synthetic datasets for different flow configurations, potentially allowing interpolation between different setups and adaptation to new conditions.

Fluid Dynamics

What problem does this paper attempt to address?

This paper aims to solve the problem of how to generate the synthetic trajectories of heavy and light particles in turbulence. Specifically, the author extends the previous Diffusion Model (DM) used to generate the three - dimensional turbulent trajectories of a single tracer particle, enabling it to generate the synthetic trajectories of heavy and light particles with correct acceleration distribution, anomalous power law, and scale - dependent local slope characteristics. This is not only crucial for understanding the turbulent behavior in natural phenomena and industrial processes, but also paves the way for future use of the diffusion model to generate high - quality synthetic datasets under different flow configurations, thus potentially enabling interpolation between different settings and the ability to adapt to new conditions. The paper achieves this goal by introducing two different diffusion models (DM - 1c and DM - 3c), where DM - 1c is used to generate a single velocity component, while DM - 3c generates three velocity components simultaneously. These models can be trained to gradually denoise from pure Gaussian noise and finally generate trajectory samples that are highly consistent with the real data. The research results show that, whether for heavy particles, tracer particles or light particles, the diffusion model can accurately capture their complex topological - vortex structures in turbulence and the dynamic characteristics under different inertial conditions.

Generative diffusion models for synthetic trajectories of heavy and light particles in turbulence

Synthetic Lagrangian Turbulence by Generative Diffusion Models

Generation of cosmic ray trajectories by a Diffusion Model trained on test particles in 3D magnetohydrodynamic turbulence

Learning to Approximate Particle Smoothing Trajectories via Diffusion Generative Models

Bayesian Conditional Diffusion Models for Versatile Spatiotemporal Turbulence Generation

Heavy-Tailed Diffusion Models

Unfolding Time: Generative Modeling for Turbulent Flows in 4D

Diffusion Models for Generating Ballistic Spacecraft Trajectories

Geometric Trajectory Diffusion Models

Stochastic Reconstruction of Gappy Lagrangian Turbulent Signals by Conditional Diffusion Models

An end-to-end generative diffusion model for heavy-ion collisions

A Stochastic Model of Heavy Particle Dispersion in Turbulent Boundary Layer

Diffusion model approach to simulating electron-proton scattering events

Diffusion Policies for Generative Modeling of Spacecraft Trajectories

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

Benchmarking Autoregressive Conditional Diffusion Models for Turbulent Flow Simulation

Generative Modeling on Manifolds Through Mixture of Riemannian Diffusion Processes

Machine-Learning Solutions for the Analysis of Single-Particle Diffusion Trajectories

Generative Diffusion Models for Fast Simulations of Particle Collisions at CERN

Statistical models for the dynamics of heavy particles in turbulence