Abstract:This paper surveys machine-learning-based super-resolution reconstruction for vortical flows. Super resolution aims to find the high-resolution flow fields from low-resolution data and is generally an approach used in image reconstruction. In addition to surveying a variety of recent super-resolution applications, we provide case studies of super-resolution analysis for an example of two-dimensional decaying isotropic turbulence. We demonstrate that physics-inspired model designs enable successful reconstruction of vortical flows from spatially limited measurements. We also discuss the challenges and outlooks of machine-learning-based super-resolution analysis for fluid flow applications. The insights gained from this study can be leveraged for super-resolution analysis of numerical and experimental flow data.

What problem does this paper attempt to address?

The paper aims to address the problem of super-resolution reconstruction in fluid flow. Specifically, it uses machine learning methods to recover high-resolution vortex fields from low-resolution data. The main content of the paper is as follows: 1. **Background Introduction**: - Super-resolution technology is traditionally used in image processing, but in this paper, the authors apply it to the field of fluid mechanics, particularly vortex flow. - Compared to traditional techniques such as interpolation and example-based learning, machine learning can better reconstruct high-frequency information. 2. **Method Overview**: - The paper first introduces several commonly used machine learning models (such as fully connected networks, convolutional neural networks, and generative adversarial networks) and their applications in super-resolution. - It particularly emphasizes the advantages of convolutional neural networks (CNNs) in handling large vortex flow data and discusses how to avoid the curse of dimensionality. - It also explores the possibility of using generative adversarial networks (GANs) for super-resolution analysis without paired data. 3. **Physics-Inspired Loss Function**: - A loss function that combines physical laws is proposed to ensure that the reconstructed flow field meets physical conditions (such as conservation laws). - This approach helps achieve more accurate flow field reconstruction with a small amount of data. 4. **Case Study**: - The super-resolution reconstruction under supervised learning is demonstrated through the example of 2D decaying isotropic turbulence. - Experiments show that the physics-inspired model design can successfully reconstruct high-resolution vortex fields from low-resolution data. 5. **Challenges and Prospects**: - The current challenges of machine learning super-resolution analysis in fluid flow applications are discussed, including how to handle different input variable settings, super-resolution in large-scale turbulence simulations, and applications to real-world problems. - Finally, future research directions are proposed, including improving model performance and extending to more application scenarios. In summary, this paper aims to improve the accuracy and reliability of super-resolution reconstruction in fluid flow by integrating various machine learning methods and physics-inspired loss functions, thereby providing strong support for numerical simulation and experimental data analysis.

Super-Resolution Analysis via Machine Learning: A Survey for Fluid Flows

Super-resolution reconstruction of turbulent flows with machine learning

Single-snapshot machine learning for super-resolution of turbulence

Machine-learning fluid flow

Deep-learning-based super-resolution reconstruction of high-speed imaging in fluids

Super-resolution with dynamics in the loss

Machine learning-based vorticity evolution and superresolution of homogeneous isotropic turbulence using wavelet projection

Deep Learning Methods for Super-Resolution Reconstruction of Turbulent Flows

Unsupervised Denoising and Super-Resolution of Vascular Flow Data by Physics-informed Machine Learning

Machine learning-based vorticity evolution and super-resolution of homogeneous isotropic turbulence using wavelet projection

Super-resolution Reconstruction of Turbulent Flows with a Hybrid Framework of Attention

Comparison and analysis of various deep learning models for super-resolution reconstruction of turbulent flows

Machine Learning for Fluid Mechanics

A robust super-resolution reconstruction model of turbulent flow data based on deep learning

Vision-informed Flow Field Super-Resolution with Quaternion Spatial Modeling and Dynamic Fluid Convolution

A deep-learning super-resolution reconstruction model of turbulent reacting flow

Small-scale Reconstruction in Three-Dimensional Kolmogorov Flows Using Four-Dimensional Variational Data Assimilation

Enhancing High-Resolution Reconstruction of Flow Fields Using Physics-Informed Diffusion Model with Probability Flow Sampling

Super-resolution reconstruction of turbulent flows with a transformer-based deep learning framework

Optical Flow for Video Super-Resolution: A Survey

Deep Learning for Fluid Velocity Field Estimation: A Review