Awesome Machine Learning for Fluid Mechanics

A curated list of machine learning papers, codes, libraries, and databases applied to fluid mechanics.

Frameworks

TensorFlow is a well-known machine learning library developed by Google.
PyTorch is another framework for machine learning developed at Facebook.
Scikit-learn is all-purpose machine learning library. It also provides the implementation of several other data analysis algorithm.
easyesn is a very good implementation of echo state network (reservoir computing). ESN often finds its application in dynamical systems.
EchoTorch is another good implementation for ESN based upon PyTorch.
PySINDy is a package with several implementations for the Sparse Identification of Nonlinear Dynamical systems (SINDy). It is also well suited for a dynamical system.
PYPARSVD is an implementation for singular value decomposition (SVD) which is distributed and parallelized which makes it efficient for large data.

Turbulence modeling in the age of data, 2019. (Paper)
A perspective on machine learning in turbulent flows, 2019. (Paper)
Machine learning for fluid mechanics, 2020. (Paper)
A Perspective on machine learning methods in turbulence modelling, 2020. (Paper)
Machine learning accelerated computational fluid dynamics, 2021. (Paper)

Reynolds averaged turbulence modeling using deep neural networks with embedded invariance, 2016. (Paper)
Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, 2019. (Paper)
From deep to physics-informed learning of turbulence: Diagnostics, 2018. (Paper)
Data-driven fractional subgrid-scale modeling for scalar turbulence: A nonlocal LES approach, 2020. (Paper)
Stable a posteriori LES of 2D turbulence using convolutional neural networks: Backscattering analysis and generalization to higher Re via transfer learning, 2021. (Paper)
Subgrid modelling for two-dimensional turbulence using neural networks, 2018. (Paper | Code)
Neural network models for the anisotropic Reynolds stress tensor in turbulent channel flow, 2019. (Paper)
A machine learning framework for LES closure terms, 2020. (Paper)
A neural network based shock detection and localization approach for discontinuous Galerkin methods, 2020. (Paper)

Reduced-order modeling of advection-dominated systems with recurrent neural networks and convolutional autoencoders, 2020. (Paper | Code)
Time-series learning of latent-space dynamics for reduced-order model closure, 2020. (Paper | Code)
Reservoir computing model of two-dimensional turbulent convection, 2020. (Paper)
Predictions of turbulent shear flows using deep neural networks, 2019. (Paper | Code)
A deep learning enabler for nonintrusive reduced order modeling of fluid flows, 2019. (Paper)
Echo State Network for two-dimensional turbulent moist Rayleigh-Bénard convection, 2020. (Paper)

Deep learning in turbulent convection networks, 2019. (Paper)
Time-resolved turbulent velocity field reconstruction using a long short-term memory (LSTM)-based artificial intelligence framework, 2019. (Paper)
Unsupervised deep learning for super-resolution reconstruction of turbulence, 2020. (Paper)
Nonlinear mode decomposition with convolutional neural networks for fluid dynamics, 2020. (Paper)

Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations, 2020. (Paper)
Forecasting of spatiotemporal chaotic dynamics with recurrent neural networks: a comparative study of reservoir computing and backpropagation algorithms, 2019. (Paper)
Data-assisted reduced-order modeling of extreme events in complex dynamical systems, 2018. (Paper)

KTH FLOW: A rich dataset of different turbulent flow generated by DNS, LES and experiments. (Simulation data | Experimental data | Paper-1)
Vreman Research: Turbulent channel flow dataset generated from simulation, could be useful in closure modeling. (Data | Paper-1 | Paper-2)
Johns Hopkins Turbulence Databases: High quality datasets for different flow problems. (Database | Paper)
CTR Stanford: Dataset for turbulent pipe flow and boundary layer generated with DNS. (Database | Paper)
sCO2: Spatial data along the tube for heated and cooled pipe under supercritical pressure. It includes around 50 cases, which is a good start for regression based model to replace correlations. (Data | Paper-1 | Paper-2)

A first course on machine learning from Nando di Freitas: Little old, recorded in 2013 but very concise and clear. (YouTube | Slides)
Steve Brunton has a wonderful channel for a variety of topics ranging from data analysis to machine learning applied to fluid mechanics. (YouTube)
Nathan Kutz has a super nice channel devoted to applied mathematics for fluid mechanics. (YouTube)
For beginners, a good resource to learn OpenFOAM from József Nagy. OpenFOAM can be adapted for applying ML model coupled with N-S equations (e.g. RANS/LES closure). (YouTube)