APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION

B. Andersen; A. Godfrey; J. Hou; D. Kropaczek

doi:10.13182/M&C21-33930

APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION

B. Andersen, A. Godfrey, J. Hou, D. Kropaczek

Nuclear Energy and Fuel Cycle Division

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

2 Scopus citations

Abstract

CRUD-induced power shift and CRUD-induced localized corrosion are two operational challenges faced by pressurized water reactors. This work details the development of a neural network surrogate model, crUdNET, for predicting the CRUD distribution based on the power distribution and core-wide parameters. Current methods for determining core susceptibility to CRUD deposition cannot be adequately used with optimization algorithms due to the computational burden or manner of output for the method. This provided the motivation for the development of crUdNET, which is based on the U-NET neural network architecture, a deep-learning convolutional architecture. CrUdNET was trained by using 6,000 unique power distributions calculated by SIMULATE3 calculated and CRUD distributions calculated by CTF and MAMBA. CrUdNET can accurately reproduce the CRUD distributions determined by MAMBA with 90% while using only 1% of MAMBA's computational resources.

Original language	English
Title of host publication	Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021
Publisher	American Nuclear Society
Pages	1782-1791
Number of pages	10
ISBN (Electronic)	9781713886310
DOIs	https://doi.org/10.13182/M&C21-33930
State	Published - 2021
Event	2021 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021 - Virtual, Online Duration: Oct 3 2021 → Oct 7 2021

Publication series

Name	Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021

Conference

Conference	2021 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021
City	Virtual, Online
Period	10/3/21 → 10/7/21

Funding

This research was supported by the Consortium for Advanced Simulation of Light Water Reactors (www.casl.gov), an Energy Innovation Hub (http://www.energy.gov/hubs) for Modeling and Simulation of Nuclear Reactors under US Department of Energy (DOE) contract no. DE-AC05-00OR22725. This research used resources of the Compute and Data Environment for Science at the Oak Ridge National Laboratory, which is supported by the DOE Office of Science under contract no. DE-AC05-00OR22725. This research used the resources of the High Performance Computing Center at Idaho National Laboratory, which is supported by the DOE Office of Nuclear Energy and the Nuclear Science User Facilities under contract no. DE-AC07-05ID14517.”

Keywords

CNN
CRUD
deep learning
optimization
surrogate modeling

Access to Document

10.13182/M&C21-33930

Cite this

Andersen, B., Godfrey, A., Hou, J., & Kropaczek, D. (2021). APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION. In Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021 (pp. 1782-1791). (Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021). American Nuclear Society. https://doi.org/10.13182/M&C21-33930

Andersen, B. ; Godfrey, A. ; Hou, J. et al. / APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION. Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021. American Nuclear Society, 2021. pp. 1782-1791 (Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021).

@inproceedings{9ab60ad11c1947de88c6b81ed6e27db4,

title = "APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION",

abstract = "CRUD-induced power shift and CRUD-induced localized corrosion are two operational challenges faced by pressurized water reactors. This work details the development of a neural network surrogate model, crUdNET, for predicting the CRUD distribution based on the power distribution and core-wide parameters. Current methods for determining core susceptibility to CRUD deposition cannot be adequately used with optimization algorithms due to the computational burden or manner of output for the method. This provided the motivation for the development of crUdNET, which is based on the U-NET neural network architecture, a deep-learning convolutional architecture. CrUdNET was trained by using 6,000 unique power distributions calculated by SIMULATE3 calculated and CRUD distributions calculated by CTF and MAMBA. CrUdNET can accurately reproduce the CRUD distributions determined by MAMBA with 90\% while using only 1\% of MAMBA's computational resources.",

keywords = "CNN, CRUD, deep learning, optimization, surrogate modeling",

author = "B. Andersen and A. Godfrey and J. Hou and D. Kropaczek",

note = "Publisher Copyright: Copyright {\textcopyright} 2021 AMERICAN NUCLEAR SOCIETY, INCORPORATED, LA GRANGE PARK, ILLINOIS 60526.All rights reserved.; 2021 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021 ; Conference date: 03-10-2021 Through 07-10-2021",

year = "2021",

doi = "10.13182/M\&C21-33930",

language = "English",

series = "Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021",

publisher = "American Nuclear Society",

pages = "1782--1791",

booktitle = "Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021",

}

Andersen, B, Godfrey, A, Hou, J & Kropaczek, D 2021, APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION. in Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021. Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021, American Nuclear Society, pp. 1782-1791, 2021 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021, Virtual, Online, 10/3/21. https://doi.org/10.13182/M&C21-33930

APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION. / Andersen, B.; Godfrey, A.; Hou, J. et al.
Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021. American Nuclear Society, 2021. p. 1782-1791 (Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021).

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

TY - GEN

T1 - APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION

AU - Andersen, B.

AU - Godfrey, A.

AU - Hou, J.

AU - Kropaczek, D.

PY - 2021

Y1 - 2021

N2 - CRUD-induced power shift and CRUD-induced localized corrosion are two operational challenges faced by pressurized water reactors. This work details the development of a neural network surrogate model, crUdNET, for predicting the CRUD distribution based on the power distribution and core-wide parameters. Current methods for determining core susceptibility to CRUD deposition cannot be adequately used with optimization algorithms due to the computational burden or manner of output for the method. This provided the motivation for the development of crUdNET, which is based on the U-NET neural network architecture, a deep-learning convolutional architecture. CrUdNET was trained by using 6,000 unique power distributions calculated by SIMULATE3 calculated and CRUD distributions calculated by CTF and MAMBA. CrUdNET can accurately reproduce the CRUD distributions determined by MAMBA with 90% while using only 1% of MAMBA's computational resources.

AB - CRUD-induced power shift and CRUD-induced localized corrosion are two operational challenges faced by pressurized water reactors. This work details the development of a neural network surrogate model, crUdNET, for predicting the CRUD distribution based on the power distribution and core-wide parameters. Current methods for determining core susceptibility to CRUD deposition cannot be adequately used with optimization algorithms due to the computational burden or manner of output for the method. This provided the motivation for the development of crUdNET, which is based on the U-NET neural network architecture, a deep-learning convolutional architecture. CrUdNET was trained by using 6,000 unique power distributions calculated by SIMULATE3 calculated and CRUD distributions calculated by CTF and MAMBA. CrUdNET can accurately reproduce the CRUD distributions determined by MAMBA with 90% while using only 1% of MAMBA's computational resources.

KW - CNN

KW - CRUD

KW - deep learning

KW - optimization

KW - surrogate modeling

UR - https://www.scopus.com/pages/publications/85130530327

U2 - 10.13182/M&C21-33930

DO - 10.13182/M&C21-33930

M3 - Conference contribution

AN - SCOPUS:85130530327

T3 - Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021

SP - 1782

EP - 1791

BT - Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021

PB - American Nuclear Society

T2 - 2021 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021

Y2 - 3 October 2021 through 7 October 2021

ER -

Andersen B, Godfrey A, Hou J, Kropaczek D. APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION. In Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021. American Nuclear Society. 2021. p. 1782-1791. (Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021). doi: 10.13182/M&C21-33930

APPLICATION OF DEEP LEARNING NETWORKS TO SURROGATE MODELING OF CRUD DEPOSITION

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this