Standard

Turbulence model development using machine learning methods for a channel flow. / Garmaev, Sergei; Yakovenko, Sergey.

в: AIP Conference Proceedings, Том 2504, № 1, 030015, 16.02.2023.

Результаты исследований: Научные публикации в периодических изданиях › статья по материалам конференции › Рецензирование

Harvard

Garmaev, S & Yakovenko, S 2023, 'Turbulence model development using machine learning methods for a channel flow', AIP Conference Proceedings, Том. 2504, № 1, 030015. https://doi.org/10.1063/5.0133600

APA

Garmaev, S., & Yakovenko, S. (2023). Turbulence model development using machine learning methods for a channel flow. AIP Conference Proceedings, 2504(1), [030015]. https://doi.org/10.1063/5.0133600

Vancouver

Garmaev S, Yakovenko S. Turbulence model development using machine learning methods for a channel flow. AIP Conference Proceedings. 2023 февр. 16;2504(1):030015. doi: 10.1063/5.0133600

Author

Garmaev, Sergei ; Yakovenko, Sergey. / Turbulence model development using machine learning methods for a channel flow. в: AIP Conference Proceedings. 2023 ; Том 2504, № 1.

BibTeX

@article{0583fdd2435d4b9eb8dc985bbff2afb0,
title = "Turbulence model development using machine learning methods for a channel flow",
abstract = "Reynolds-averaged Navier-Stokes (RANS) equations have a wide range of industrial applications to modelling of turbulent flows. However, RANS models suffer from lack of accuracy in comparison with Direct Numerical Simulation (DNS) due to uncertainties in modelling of Reynolds stresses. Recently it has been shown that methods of statistical learning can be used to model the Reynolds stress tensor using high-fidelity DNS data as a reference. In this study, we compare methods of machine learning for the Reynolds stress anisotropy prediction in a plane channel flow using publicly available DNS data for high Reynolds number flows. Machine learning methods compared in this work include previously proposed tensor-basis neural networks, tensor-basis random forest and gradient boosting methods on a similar tensor-basis.",
author = "Sergei Garmaev and Sergey Yakovenko",
note = "The research was carried out within the state assignment of Ministry of Science and Higher Education of the Russian Federation (project No. 121030500149-8).; 2021 Actual Problems of Continuum Mechanics: Experiment, Theory, and Applications : XXVIII Всероссийская конференция с международным участием «Высокоэнергетические процессы в механике сплошной среды», посвященная 100-летию со дня рождения Н.Н. Яненко ; Conference date: 20-09-2021 Through 24-09-2021",
year = "2023",
month = feb,
day = "16",
doi = "10.1063/5.0133600",
language = "English",
volume = "2504",
journal = "AIP Conference Proceedings",
issn = "0094-243X",
publisher = "American Institute of Physics",
number = "1",

}

RIS

TY - JOUR

T1 - Turbulence model development using machine learning methods for a channel flow

AU - Garmaev, Sergei

AU - Yakovenko, Sergey

N1 - The research was carried out within the state assignment of Ministry of Science and Higher Education of the Russian Federation (project No. 121030500149-8).

PY - 2023/2/16

Y1 - 2023/2/16

N2 - Reynolds-averaged Navier-Stokes (RANS) equations have a wide range of industrial applications to modelling of turbulent flows. However, RANS models suffer from lack of accuracy in comparison with Direct Numerical Simulation (DNS) due to uncertainties in modelling of Reynolds stresses. Recently it has been shown that methods of statistical learning can be used to model the Reynolds stress tensor using high-fidelity DNS data as a reference. In this study, we compare methods of machine learning for the Reynolds stress anisotropy prediction in a plane channel flow using publicly available DNS data for high Reynolds number flows. Machine learning methods compared in this work include previously proposed tensor-basis neural networks, tensor-basis random forest and gradient boosting methods on a similar tensor-basis.

AB - Reynolds-averaged Navier-Stokes (RANS) equations have a wide range of industrial applications to modelling of turbulent flows. However, RANS models suffer from lack of accuracy in comparison with Direct Numerical Simulation (DNS) due to uncertainties in modelling of Reynolds stresses. Recently it has been shown that methods of statistical learning can be used to model the Reynolds stress tensor using high-fidelity DNS data as a reference. In this study, we compare methods of machine learning for the Reynolds stress anisotropy prediction in a plane channel flow using publicly available DNS data for high Reynolds number flows. Machine learning methods compared in this work include previously proposed tensor-basis neural networks, tensor-basis random forest and gradient boosting methods on a similar tensor-basis.

UR - https://www.scopus.com/record/display.uri?eid=2-s2.0-85149622390&origin=inward&txGid=25c539850285d4b7d293c0f02c3150fd

UR - https://www.mendeley.com/catalogue/652010d7-ea68-3fe4-975d-389ac17d5624/

U2 - 10.1063/5.0133600

DO - 10.1063/5.0133600

M3 - Conference article

VL - 2504

JO - AIP Conference Proceedings

JF - AIP Conference Proceedings

SN - 0094-243X

IS - 1

M1 - 030015

T2 - 2021 Actual Problems of Continuum Mechanics: Experiment, Theory, and Applications

Y2 - 20 September 2021 through 24 September 2021

ER -

ID: 59659996