Reynolds Stress

Reynolds stress, representing the fluctuating momentum transport in turbulent flows, is a crucial yet challenging aspect of fluid dynamics modeling. Current research focuses on improving Reynolds-Averaged Navier-Stokes (RANS) simulations by developing data-driven closure models, often employing neural networks (including Graph Neural Networks) to predict the Reynolds stress tensor or its divergence, sometimes incorporating Bayesian methods to quantify model uncertainty. These advancements aim to enhance the accuracy and reliability of RANS simulations, particularly in predicting complex flows, with implications for various engineering applications such as turbomachinery design and virtual certification processes. Explainable AI techniques are also being used to better understand the underlying physics of turbulence and identify regions where model predictions are less accurate.

Papers

November 29, 2023

Enhancing Data-Assimilation in CFD using Graph Neural Networks
Michele Quattromini, Michele Alessandro Bucci, Stefania Cherubini, Onofrio Semeraro
Graph Neural Network Data Assimilation Navier Stokes CFD Simulation Reynolds Stress

July 5, 2023

A probabilistic, data-driven closure model for RANS simulations with aleatoric, model uncertainty
Atul Agrawal, Phaedon-Stelios Koutsourelakis
Model Uncertainty Aleatoric Uncertainty Closure Model Non Gaussian Aleatoric Uncertainty Reynolds Averaged Navier Stokes Reynolds Stress

February 2, 2023

Identifying regions of importance in wall-bounded turbulence through explainable deep learning
Andres Cremades, Sergio Hoyas, Rahul Deshpande, Pedro Quintero, Martin Lellep, Will Junghoon Lee, Jason Monty, Nicholas Hutchins, Moritz Linkmann, Ivan Marusic, Ricardo Vinuesa
Importance Aware Region Specific Velocity Field Explainable Deep Learning Coherent Structure Reynolds Stress

September 29, 2022

Evaluation of physics constrained data-driven methods for turbulence model uncertainty quantification
Marcel Matha, Karsten Kucharczyk, Christian Morsbach
Global Evaluation Data Driven Model Uncertainty Theoretical Physic Computational Fluid Dynamic Turbulence Model Turbulence Closure Reynolds Stress

March 31, 2022

A data-driven approach for the closure of RANS models by the divergence of the Reynolds Stress Tensor
Stefano Berrone, Davide Oberto
Data Driven Approach Inverse Divergence Turbulence Model Security Closure Reynolds Stress

February 11, 2022

Deep Learning to advance the Eigenspace Perturbation Method for Turbulence Model Uncertainty Quantification
Khashayar Nobarani, Seyed Esmaeil Razavi
Deep Learning Model Uncertainty Turbulence Model Reynolds Averaged Navier Stokes Reynolds Stress

Reynolds Stress

Papers

Enhancing Data-Assimilation in CFD using Graph Neural Networks

A probabilistic, data-driven closure model for RANS simulations with aleatoric, model uncertainty

Identifying regions of importance in wall-bounded turbulence through explainable deep learning

Evaluation of physics constrained data-driven methods for turbulence model uncertainty quantification

A data-driven approach for the closure of RANS models by the divergence of the Reynolds Stress Tensor

Deep Learning to advance the Eigenspace Perturbation Method for Turbulence Model Uncertainty Quantification