The Potential of Machine Learning Methods for Separated Turbulent Flow Simulations: Classical Versus Dynamic Methods

Feasible and reliable predictions of separated turbulent flows are a requirement to successfully address the majority of aerospace and wind energy problems. Existing computational approaches such as large eddy simulation (LES) or Reynolds-averaged Navier–Stokes (RANS) methods have suffered for decad...

Full description

Saved in:
Bibliographic Details
Main Author: Stefan Heinz
Format: Article
Language:English
Published: MDPI AG 2024-11-01
Series:Fluids
Subjects:
Online Access:https://www.mdpi.com/2311-5521/9/12/278
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850049609156526080
author Stefan Heinz
author_facet Stefan Heinz
author_sort Stefan Heinz
collection DOAJ
description Feasible and reliable predictions of separated turbulent flows are a requirement to successfully address the majority of aerospace and wind energy problems. Existing computational approaches such as large eddy simulation (LES) or Reynolds-averaged Navier–Stokes (RANS) methods have suffered for decades from well-known computational cost and reliability issues in this regard. One very popular approach to dealing with these questions is the use of machine learning (ML) methods to enable improved RANS predictions. An alternative is the use of minimal error simulation methods (continuous eddy simulation (CES), which may be seen as a dynamic ML method) in the framework of partially or fully resolving simulation methods. Characteristic features of the two approaches are presented here by considering a variety of complex separated flow simulations. The conclusion is that minimal error CES methods perform clearly better than ML-RANS methods. Most importantly and in contrast to ML-RANS methods, CES is demonstrated to be well applicable to cases not involved in the model development. The reason for such superior CES performance is identified here: it is the ability of CES to properly account for causal relationships induced by the structure of separated turbulent flows.
format Article
id doaj-art-382707db6ee74483bf6e48192cbf6d27
institution DOAJ
issn 2311-5521
language English
publishDate 2024-11-01
publisher MDPI AG
record_format Article
series Fluids
spelling doaj-art-382707db6ee74483bf6e48192cbf6d272025-08-20T02:53:41ZengMDPI AGFluids2311-55212024-11-0191227810.3390/fluids9120278The Potential of Machine Learning Methods for Separated Turbulent Flow Simulations: Classical Versus Dynamic MethodsStefan Heinz0Department of Mathematics and Statistics, University of Wyoming, 1000 E. University Avenue, Laramie, WY 82071, USAFeasible and reliable predictions of separated turbulent flows are a requirement to successfully address the majority of aerospace and wind energy problems. Existing computational approaches such as large eddy simulation (LES) or Reynolds-averaged Navier–Stokes (RANS) methods have suffered for decades from well-known computational cost and reliability issues in this regard. One very popular approach to dealing with these questions is the use of machine learning (ML) methods to enable improved RANS predictions. An alternative is the use of minimal error simulation methods (continuous eddy simulation (CES), which may be seen as a dynamic ML method) in the framework of partially or fully resolving simulation methods. Characteristic features of the two approaches are presented here by considering a variety of complex separated flow simulations. The conclusion is that minimal error CES methods perform clearly better than ML-RANS methods. Most importantly and in contrast to ML-RANS methods, CES is demonstrated to be well applicable to cases not involved in the model development. The reason for such superior CES performance is identified here: it is the ability of CES to properly account for causal relationships induced by the structure of separated turbulent flows.https://www.mdpi.com/2311-5521/9/12/278computational fluid dynamicsmachine learninglarge eddy simulation (LES)Reynolds-averaged Navier-Stokes (RANS) methodshybrid RANS-LES methods
spellingShingle Stefan Heinz
The Potential of Machine Learning Methods for Separated Turbulent Flow Simulations: Classical Versus Dynamic Methods
Fluids
computational fluid dynamics
machine learning
large eddy simulation (LES)
Reynolds-averaged Navier-Stokes (RANS) methods
hybrid RANS-LES methods
title The Potential of Machine Learning Methods for Separated Turbulent Flow Simulations: Classical Versus Dynamic Methods
title_full The Potential of Machine Learning Methods for Separated Turbulent Flow Simulations: Classical Versus Dynamic Methods
title_fullStr The Potential of Machine Learning Methods for Separated Turbulent Flow Simulations: Classical Versus Dynamic Methods
title_full_unstemmed The Potential of Machine Learning Methods for Separated Turbulent Flow Simulations: Classical Versus Dynamic Methods
title_short The Potential of Machine Learning Methods for Separated Turbulent Flow Simulations: Classical Versus Dynamic Methods
title_sort potential of machine learning methods for separated turbulent flow simulations classical versus dynamic methods
topic computational fluid dynamics
machine learning
large eddy simulation (LES)
Reynolds-averaged Navier-Stokes (RANS) methods
hybrid RANS-LES methods
url https://www.mdpi.com/2311-5521/9/12/278
work_keys_str_mv AT stefanheinz thepotentialofmachinelearningmethodsforseparatedturbulentflowsimulationsclassicalversusdynamicmethods
AT stefanheinz potentialofmachinelearningmethodsforseparatedturbulentflowsimulationsclassicalversusdynamicmethods