Application of Extreme Learning Machines to inverse neutron kinetics

Paolo Picca, Roberto Furfaro

Research output: Research - peer-reviewArticle

Abstract

The paper presents the application of Extreme Leaning Machines (ELMs) for inverse reactor kinetic applications. ELMs were proposed by Huang and co-workers (2004, 2006a,b, 2015), which showed their enhances capabilities in terms of training speed and generalization with respect to classical Artificial Neural Networks (ANNs). ELMs are here implemented for reactivity determination as an alternative to ANNs (e.g. Picca et al. (2008)) and Gaussian Processes (Picca and Furfaro, 2012). After a review of the main features of ELMs, their application to inverse kinetic problems is proposed. The ELMs performance is tested on a typical accelerator drive system configuration (Yalina reactor) and the inversion is carried out on an accurate kinetic model (multi-group transport).

LanguageEnglish (US)
Pages1-8
Number of pages8
JournalAnnals of Nuclear Energy
Volume100
DOIs
StatePublished - Feb 1 2017

Fingerprint

Learning systems
Neutrons
Kinetics
Neural networks
Particle accelerators

Keywords

  • Accelerator-driven system
  • Artificial Neural Network
  • Extreme Learning Machines
  • Inverse neutron kinetics

ASJC Scopus subject areas

  • Nuclear Energy and Engineering

Cite this

Application of Extreme Learning Machines to inverse neutron kinetics. / Picca, Paolo; Furfaro, Roberto.

In: Annals of Nuclear Energy, Vol. 100, 01.02.2017, p. 1-8.

Research output: Research - peer-reviewArticle

@article{c90e6d29224e49e38b2fe6be8d2ec03e,
title = "Application of Extreme Learning Machines to inverse neutron kinetics",
abstract = "The paper presents the application of Extreme Leaning Machines (ELMs) for inverse reactor kinetic applications. ELMs were proposed by Huang and co-workers (2004, 2006a,b, 2015), which showed their enhances capabilities in terms of training speed and generalization with respect to classical Artificial Neural Networks (ANNs). ELMs are here implemented for reactivity determination as an alternative to ANNs (e.g. Picca et al. (2008)) and Gaussian Processes (Picca and Furfaro, 2012). After a review of the main features of ELMs, their application to inverse kinetic problems is proposed. The ELMs performance is tested on a typical accelerator drive system configuration (Yalina reactor) and the inversion is carried out on an accurate kinetic model (multi-group transport).",
keywords = "Accelerator-driven system, Artificial Neural Network, Extreme Learning Machines, Inverse neutron kinetics",
author = "Paolo Picca and Roberto Furfaro",
year = "2017",
month = "2",
doi = "10.1016/j.anucene.2016.08.031",
volume = "100",
pages = "1--8",
journal = "Annals of Nuclear Energy",
issn = "0306-4549",
publisher = "Elsevier Limited",

}

TY - JOUR

T1 - Application of Extreme Learning Machines to inverse neutron kinetics

AU - Picca,Paolo

AU - Furfaro,Roberto

PY - 2017/2/1

Y1 - 2017/2/1

N2 - The paper presents the application of Extreme Leaning Machines (ELMs) for inverse reactor kinetic applications. ELMs were proposed by Huang and co-workers (2004, 2006a,b, 2015), which showed their enhances capabilities in terms of training speed and generalization with respect to classical Artificial Neural Networks (ANNs). ELMs are here implemented for reactivity determination as an alternative to ANNs (e.g. Picca et al. (2008)) and Gaussian Processes (Picca and Furfaro, 2012). After a review of the main features of ELMs, their application to inverse kinetic problems is proposed. The ELMs performance is tested on a typical accelerator drive system configuration (Yalina reactor) and the inversion is carried out on an accurate kinetic model (multi-group transport).

AB - The paper presents the application of Extreme Leaning Machines (ELMs) for inverse reactor kinetic applications. ELMs were proposed by Huang and co-workers (2004, 2006a,b, 2015), which showed their enhances capabilities in terms of training speed and generalization with respect to classical Artificial Neural Networks (ANNs). ELMs are here implemented for reactivity determination as an alternative to ANNs (e.g. Picca et al. (2008)) and Gaussian Processes (Picca and Furfaro, 2012). After a review of the main features of ELMs, their application to inverse kinetic problems is proposed. The ELMs performance is tested on a typical accelerator drive system configuration (Yalina reactor) and the inversion is carried out on an accurate kinetic model (multi-group transport).

KW - Accelerator-driven system

KW - Artificial Neural Network

KW - Extreme Learning Machines

KW - Inverse neutron kinetics

UR - http://www.scopus.com/inward/record.url?scp=84994593667&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84994593667&partnerID=8YFLogxK

U2 - 10.1016/j.anucene.2016.08.031

DO - 10.1016/j.anucene.2016.08.031

M3 - Article

VL - 100

SP - 1

EP - 8

JO - Annals of Nuclear Energy

T2 - Annals of Nuclear Energy

JF - Annals of Nuclear Energy

SN - 0306-4549

ER -