Skip to content

Neural networks for displacement analysis in an advanced gas cooled reactor core model

Research output: Contribution to journalArticle

Original languageEnglish
Pages (from-to)252-266
Number of pages15
JournalNuclear Engineering and Design
Volume332
Early online date30 Mar 2018
DOIs
DateAccepted/In press - 26 Mar 2018
DateE-pub ahead of print - 30 Mar 2018
DatePublished (current) - 1 Jun 2018

Abstract

This paper presents a Neural Network (NN) approach for displacement analysis with applications in modelling the seismic response of the UK's Advanced Gas Cooled Reactors (AGRs). A quarter sized physical model of a reactor core was developed at the University of Bristol to provide experimental validation to the existing numerical models that support the seismic resilience assessments of the AGRs. The physical model outputs include displacement and acceleration datasets of considerable size and complexity, collected for a range of seismic inputs and postulated component damage scenarios. Rich sets of displacement data were employed in training two NN models that can predict displacement at user-defined locations in the core physical model and can map the correlation between the component relative displacements. Understanding component displacements is particularly important, as such displacements may affect the channel shapes and can cause local and general distortion of the core. This paper presents the development, testing and performance of the NN models. The NNs yield predictions that compare well with the experimentally obtained parameters. As more experimental test data become available, the NN's prediction capability will benefit from accumulated training. In the future, the NNs will be incorporated into a multi-layered framework for dynamic response prediction and analysis.

    Research areas

  • Advanced gas cooled reactor, Displacement prediction, Neural network, Seismic testing

Documents

Documents

  • Full-text PDF (accepted author manuscript)

    Rights statement: This is the author accepted manuscript (AAM). The final published version (version of record) is available online via Elsevier at https://www.sciencedirect.com/science/article/pii/S0029549318303558 . Please refer to any applicable terms of use of the publisher.

    Accepted author manuscript, 2 MB, PDF-document

    Licence: CC BY-NC-ND

DOI

View research connections

Related faculties, schools or groups