Fast regression of the tritium breeding ratio in fusion reactors
Fast regression of the tritium breeding ratio in fusion reactors
The tritium breeding ratio (TBR) is an essential quantity for the design of modern and next-generation D-T fueled nuclear fusion reactors. Representing the ratio between tritium fuel generated in breeding blankets and fuel consumed during reactor runtime, the TBR depends on reactor geometry and material properties in a complex manner. In this work, we explored the training of surrogate models to produce a cheap but high-quality approximation for a Monte Carlo (MC) TBR model in use at the UK Atomic Energy Authority. We investigated possibilities for dimensional reduction of its feature space, reviewed 9 families of surrogate models for potential applicability, and performed hyperparameter optimization. Here we present the performance and scaling properties of these models, the fastest of which, an artificial neural network, demonstrated R 2 = 0.985 and a mean prediction time of 0.898 μ s , representing a relative speedup of 8 × 10 6 with respect to the expensive MC model. We further present a novel adaptive sampling algorithm, Quality-Adaptive Surrogate Sampling, capable of interfacing with any of the individually studied surrogates. Our preliminary testing on a toy TBR theory has demonstrated the efficacy of this algorithm for accelerating the surrogate modelling process.
adaptive sampling, fast approximation, nuclear fusion, regression, surrogate model, tritium breeding
Mánek, P.
3615d46c-e225-40f9-90ab-eabde49f85b1
Van Goffrier, G.
18877be8-d9be-4c90-a625-8f1c11b9cb84
Gopakumar, V.
1b663b9f-bdcb-4d79-be7f-fa718227c55f
Nikolaou, N.
ed61ff7c-4b80-408b-b8de-f0d38f148a76
Shimwell, J.
976b0e1e-7bd7-425d-bdd7-148a5547450c
Waldmann, I.
8149a917-e0dd-473c-bbec-68bc4fe70c2f
31 January 2023
Mánek, P.
3615d46c-e225-40f9-90ab-eabde49f85b1
Van Goffrier, G.
18877be8-d9be-4c90-a625-8f1c11b9cb84
Gopakumar, V.
1b663b9f-bdcb-4d79-be7f-fa718227c55f
Nikolaou, N.
ed61ff7c-4b80-408b-b8de-f0d38f148a76
Shimwell, J.
976b0e1e-7bd7-425d-bdd7-148a5547450c
Waldmann, I.
8149a917-e0dd-473c-bbec-68bc4fe70c2f
Mánek, P., Van Goffrier, G., Gopakumar, V., Nikolaou, N., Shimwell, J. and Waldmann, I.
(2023)
Fast regression of the tritium breeding ratio in fusion reactors.
Machine Learning: Science and Technology, 4 (1), [015008].
(doi:10.1088/2632-2153/acb2b3).
Abstract
The tritium breeding ratio (TBR) is an essential quantity for the design of modern and next-generation D-T fueled nuclear fusion reactors. Representing the ratio between tritium fuel generated in breeding blankets and fuel consumed during reactor runtime, the TBR depends on reactor geometry and material properties in a complex manner. In this work, we explored the training of surrogate models to produce a cheap but high-quality approximation for a Monte Carlo (MC) TBR model in use at the UK Atomic Energy Authority. We investigated possibilities for dimensional reduction of its feature space, reviewed 9 families of surrogate models for potential applicability, and performed hyperparameter optimization. Here we present the performance and scaling properties of these models, the fastest of which, an artificial neural network, demonstrated R 2 = 0.985 and a mean prediction time of 0.898 μ s , representing a relative speedup of 8 × 10 6 with respect to the expensive MC model. We further present a novel adaptive sampling algorithm, Quality-Adaptive Surrogate Sampling, capable of interfacing with any of the individually studied surrogates. Our preliminary testing on a toy TBR theory has demonstrated the efficacy of this algorithm for accelerating the surrogate modelling process.
Text
Mánek_2023_Mach._Learn.__Sci._Technol._4_015008
- Version of Record
More information
Accepted/In Press date: 12 January 2023
Published date: 31 January 2023
Additional Information:
Funding Information:
This work has been carried out within the framework of the EUROfusion consortium and has received funding from the Euratom research and training programme 2014–2018 and 2019–2020 under Grant Agreement No. 633053. The views and opinions expressed herein do not necessarily reflect those of the European Commission.
Funding Information:
This project was supported by the EU Horizon 2020 research & innovation programme [Grant No. 758892, ExoAI]. N Nikolaou acknowledges the support of the NVIDIA Corporation’s GPU Grant.
Funding Information:
PM and GVG were supported by the STFC UCL Centre for Doctoral Training in Data Intensive Science (Grant No. ST/P006736/1). GVG was funded by the UCL Graduate Research and Overseas Research Scholarships.
Funding Information:
This work has been partly funded by the Institutional support for the development of a research organization (DKRVO, Czech Republic).
Publisher Copyright:
© 2023 The Author(s). Published by IOP Publishing Ltd.
Keywords:
adaptive sampling, fast approximation, nuclear fusion, regression, surrogate model, tritium breeding
Identifiers
Local EPrints ID: 482261
URI: http://eprints.soton.ac.uk/id/eprint/482261
ISSN: 2632-2153
PURE UUID: 9b5bd7cf-4e20-4b88-a034-ad19c1ed1813
Catalogue record
Date deposited: 22 Sep 2023 16:38
Last modified: 18 Mar 2024 04:16
Export record
Altmetrics
Contributors
Author:
P. Mánek
Author:
G. Van Goffrier
Author:
V. Gopakumar
Author:
N. Nikolaou
Author:
J. Shimwell
Author:
I. Waldmann
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics