Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network

Shirui Luo; Madhu Vellakal; Seid Koric; Volodymyr Kindratenko; Jiahuan Cui

doi:10.1007/978-3-030-59851-8_9

Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network

Shirui Luo, Madhu Vellakal, Seid Koric, Volodymyr Kindratenko, Jiahuan Cui

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Identifying the appropriate parameters of a turbulence model for a class of flow usually requires extensive experimentation and numerical simulations. Therefore even a modest improvement of the turbulence model can significantly reduce the overall cost of a three-dimensional, time-dependent simulation. In this paper we demonstrate a novel method to find the optimal parameters in the Reynolds-averaged Navier–Stokes (RANS) turbulence model using high-fidelity direct numerical simulation (DNS) data. A physics informed neural network (PINN) that is embedded with the turbulent transport equations is studied, physical loss functions are proposed to explicitly impose information of the transport equations to neural networks. This approach solves an inverse problem by treating the five parameters in turbulence model as random variables, with the turbulent kinetic energy and dissipation rate as known quantities from DNS simulation. The objective is to optimize the five parameters in turbulence closures using the PINN leveraging limited data available from costly high-fidelity DNS data. We validated this method on two test cases of flow over bump. The recommended values were found to be Cϵ1 = 1.302, Cϵ2 = 1.862, Cμ = 0.09, σK = 0.75, σϵ = 0.273; the mean absolute error of the velocity profile between RANS and DNS decreased by 22% when using these neural network inferred parameters.

Original language	English (US)
Title of host publication	High Performance Computing - ISC High Performance 2020 International Workshops, Revised Selected Papers
Subtitle of host publication	ISC High Performance 2020 International Workshops, Frankfurt, Germany, June 21–25, 2020, Revised Selected Papers
Editors	Heike Jagode, Hartwig Anzt, Guido Juckeland, Hatem Ltaief
Publisher	Springer
Chapter	9
Pages	137-149
Number of pages	13
ISBN (Electronic)	978-3-030-59851-8
ISBN (Print)	978-3-030-59850-1
DOIs	https://doi.org/10.1007/978-3-030-59851-8_9
State	Published - Oct 20 2020

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	12321 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Keywords

Neural network
Physics embedded machine learning
Turbulence modeling

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Online availability

10.1007/978-3-030-59851-8_9

Library availability

Discover UIUC Full Text

Cite this

Luo, S., Vellakal, M., Koric, S., Kindratenko, V., & Cui, J. (2020). Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network. In H. Jagode, H. Anzt, G. Juckeland, & H. Ltaief (Eds.), High Performance Computing - ISC High Performance 2020 International Workshops, Revised Selected Papers: ISC High Performance 2020 International Workshops, Frankfurt, Germany, June 21–25, 2020, Revised Selected Papers (pp. 137-149). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12321 LNCS). Springer. https://doi.org/10.1007/978-3-030-59851-8_9

Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network. / Luo, Shirui; Vellakal, Madhu; Koric, Seid et al.
High Performance Computing - ISC High Performance 2020 International Workshops, Revised Selected Papers: ISC High Performance 2020 International Workshops, Frankfurt, Germany, June 21–25, 2020, Revised Selected Papers. ed. / Heike Jagode; Hartwig Anzt; Guido Juckeland; Hatem Ltaief. Springer, 2020. p. 137-149 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12321 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Luo, S, Vellakal, M, Koric, S , Kindratenko, V & Cui, J 2020, Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network. in H Jagode, H Anzt, G Juckeland & H Ltaief (eds), High Performance Computing - ISC High Performance 2020 International Workshops, Revised Selected Papers: ISC High Performance 2020 International Workshops, Frankfurt, Germany, June 21–25, 2020, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12321 LNCS, Springer, pp. 137-149. https://doi.org/10.1007/978-3-030-59851-8_9

Luo S, Vellakal M, Koric S , Kindratenko V, Cui J. Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network. In Jagode H, Anzt H, Juckeland G, Ltaief H, editors, High Performance Computing - ISC High Performance 2020 International Workshops, Revised Selected Papers: ISC High Performance 2020 International Workshops, Frankfurt, Germany, June 21–25, 2020, Revised Selected Papers. Springer. 2020. p. 137-149. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-59851-8_9

Luo, Shirui ; Vellakal, Madhu ; Koric, Seid et al. / Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network. High Performance Computing - ISC High Performance 2020 International Workshops, Revised Selected Papers: ISC High Performance 2020 International Workshops, Frankfurt, Germany, June 21–25, 2020, Revised Selected Papers. editor / Heike Jagode ; Hartwig Anzt ; Guido Juckeland ; Hatem Ltaief. Springer, 2020. pp. 137-149 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{7f6de0073d9a458daa610b599b9ee74f,

title = "Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network",

abstract = "Identifying the appropriate parameters of a turbulence model for a class of flow usually requires extensive experimentation and numerical simulations. Therefore even a modest improvement of the turbulence model can significantly reduce the overall cost of a three-dimensional, time-dependent simulation. In this paper we demonstrate a novel method to find the optimal parameters in the Reynolds-averaged Navier–Stokes (RANS) turbulence model using high-fidelity direct numerical simulation (DNS) data. A physics informed neural network (PINN) that is embedded with the turbulent transport equations is studied, physical loss functions are proposed to explicitly impose information of the transport equations to neural networks. This approach solves an inverse problem by treating the five parameters in turbulence model as random variables, with the turbulent kinetic energy and dissipation rate as known quantities from DNS simulation. The objective is to optimize the five parameters in turbulence closures using the PINN leveraging limited data available from costly high-fidelity DNS data. We validated this method on two test cases of flow over bump. The recommended values were found to be Cϵ1 = 1.302, Cϵ2 = 1.862, Cμ = 0.09, σK = 0.75, σϵ = 0.273; the mean absolute error of the velocity profile between RANS and DNS decreased by 22% when using these neural network inferred parameters.",

keywords = "Neural network, Physics embedded machine learning, Turbulence modeling",

author = "Shirui Luo and Madhu Vellakal and Seid Koric and Volodymyr Kindratenko and Jiahuan Cui",

note = "Funding Information: Foundation{\textquoteright}s Major Research Instrumentation program, grant #1725729, as well as the University of Illinois at Urbana-Champaign. Funding Information: Acknowledgments. This work utilizes resources supported by the National Science Publisher Copyright: {\textcopyright} 2020, Springer Nature Switzerland AG.",

year = "2020",

month = oct,

day = "20",

doi = "10.1007/978-3-030-59851-8_9",

language = "English (US)",

isbn = "978-3-030-59850-1",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "137--149",

editor = "Heike Jagode and Hartwig Anzt and Guido Juckeland and Hatem Ltaief",

booktitle = "High Performance Computing - ISC High Performance 2020 International Workshops, Revised Selected Papers",

address = "Germany",

}

TY - GEN

T1 - Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network

AU - Luo, Shirui

AU - Vellakal, Madhu

AU - Koric, Seid

AU - Kindratenko, Volodymyr

AU - Cui, Jiahuan

N1 - Funding Information: Foundation’s Major Research Instrumentation program, grant #1725729, as well as the University of Illinois at Urbana-Champaign. Funding Information: Acknowledgments. This work utilizes resources supported by the National Science Publisher Copyright: © 2020, Springer Nature Switzerland AG.

PY - 2020/10/20

Y1 - 2020/10/20

N2 - Identifying the appropriate parameters of a turbulence model for a class of flow usually requires extensive experimentation and numerical simulations. Therefore even a modest improvement of the turbulence model can significantly reduce the overall cost of a three-dimensional, time-dependent simulation. In this paper we demonstrate a novel method to find the optimal parameters in the Reynolds-averaged Navier–Stokes (RANS) turbulence model using high-fidelity direct numerical simulation (DNS) data. A physics informed neural network (PINN) that is embedded with the turbulent transport equations is studied, physical loss functions are proposed to explicitly impose information of the transport equations to neural networks. This approach solves an inverse problem by treating the five parameters in turbulence model as random variables, with the turbulent kinetic energy and dissipation rate as known quantities from DNS simulation. The objective is to optimize the five parameters in turbulence closures using the PINN leveraging limited data available from costly high-fidelity DNS data. We validated this method on two test cases of flow over bump. The recommended values were found to be Cϵ1 = 1.302, Cϵ2 = 1.862, Cμ = 0.09, σK = 0.75, σϵ = 0.273; the mean absolute error of the velocity profile between RANS and DNS decreased by 22% when using these neural network inferred parameters.

AB - Identifying the appropriate parameters of a turbulence model for a class of flow usually requires extensive experimentation and numerical simulations. Therefore even a modest improvement of the turbulence model can significantly reduce the overall cost of a three-dimensional, time-dependent simulation. In this paper we demonstrate a novel method to find the optimal parameters in the Reynolds-averaged Navier–Stokes (RANS) turbulence model using high-fidelity direct numerical simulation (DNS) data. A physics informed neural network (PINN) that is embedded with the turbulent transport equations is studied, physical loss functions are proposed to explicitly impose information of the transport equations to neural networks. This approach solves an inverse problem by treating the five parameters in turbulence model as random variables, with the turbulent kinetic energy and dissipation rate as known quantities from DNS simulation. The objective is to optimize the five parameters in turbulence closures using the PINN leveraging limited data available from costly high-fidelity DNS data. We validated this method on two test cases of flow over bump. The recommended values were found to be Cϵ1 = 1.302, Cϵ2 = 1.862, Cμ = 0.09, σK = 0.75, σϵ = 0.273; the mean absolute error of the velocity profile between RANS and DNS decreased by 22% when using these neural network inferred parameters.

KW - Neural network

KW - Physics embedded machine learning

KW - Turbulence modeling

UR - http://www.scopus.com/inward/record.url?scp=85096418029&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85096418029&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-59851-8_9

DO - 10.1007/978-3-030-59851-8_9

M3 - Conference contribution

SN - 978-3-030-59850-1

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 137

EP - 149

BT - High Performance Computing - ISC High Performance 2020 International Workshops, Revised Selected Papers

A2 - Jagode, Heike

A2 - Anzt, Hartwig

A2 - Juckeland, Guido

A2 - Ltaief, Hatem

PB - Springer

ER -

Parameter Identification of RANS Turbulence Model Using Physics-Embedded Neural Network

Abstract

Publication series

Keywords

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this