Conference paper Open Access

A Deep Learning Approach to Anomaly Detection in Nuclear Reactors

Francesco Caliva; Fabio De Sousa Ribeiro; Antonios Mylonakis; Christophe Demazière; Georgios Leontidis; Stefanos Kollias

MARC21 XML Export

<?xml version='1.0' encoding='UTF-8'?>
<record xmlns="">
  <datafield tag="041" ind1=" " ind2=" ">
    <subfield code="a">eng</subfield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">deep learning, convolutional neural networks, clustering trained representations, denoising autoencoders, signal processing, nuclear reactors, unfolding, anomaly detection</subfield>
  <controlfield tag="005">20200120172355.0</controlfield>
  <controlfield tag="001">1410084</controlfield>
  <datafield tag="711" ind1=" " ind2=" ">
    <subfield code="d">July 8-13, 2018</subfield>
    <subfield code="g">IJCNN</subfield>
    <subfield code="a">Joint Conference on Neural Networks</subfield>
    <subfield code="c">Rio de Janeiro</subfield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">University of Lincoln</subfield>
    <subfield code="a">Fabio De Sousa Ribeiro</subfield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">Chalmers University of Technology</subfield>
    <subfield code="a">Antonios Mylonakis</subfield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">Chalmers University of Technology</subfield>
    <subfield code="a">Christophe Demazière</subfield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">University of Lincoln</subfield>
    <subfield code="a">Georgios Leontidis</subfield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="u">University of Lincoln</subfield>
    <subfield code="a">Stefanos Kollias</subfield>
  <datafield tag="856" ind1="4" ind2=" ">
    <subfield code="s">2838336</subfield>
    <subfield code="z">md5:e2066544c8698653db7f809947fb1bff</subfield>
    <subfield code="u"></subfield>
  <datafield tag="542" ind1=" " ind2=" ">
    <subfield code="l">open</subfield>
  <datafield tag="260" ind1=" " ind2=" ">
    <subfield code="c">2018-09-06</subfield>
  <datafield tag="909" ind1="C" ind2="O">
    <subfield code="p">openaire</subfield>
    <subfield code="o"></subfield>
  <datafield tag="100" ind1=" " ind2=" ">
    <subfield code="u">University of Lincoln</subfield>
    <subfield code="a">Francesco Caliva</subfield>
  <datafield tag="245" ind1=" " ind2=" ">
    <subfield code="a">A Deep Learning Approach to Anomaly Detection in Nuclear Reactors</subfield>
  <datafield tag="536" ind1=" " ind2=" ">
    <subfield code="c">754316</subfield>
    <subfield code="a">Core monitoring techniques and experimental validation and demonstration</subfield>
  <datafield tag="540" ind1=" " ind2=" ">
    <subfield code="u"></subfield>
    <subfield code="a">Creative Commons Attribution 4.0 International</subfield>
  <datafield tag="650" ind1="1" ind2="7">
    <subfield code="a">cc-by</subfield>
    <subfield code="2"></subfield>
  <datafield tag="520" ind1=" " ind2=" ">
    <subfield code="a">&lt;p&gt;In this work, a novel deep learning approach to&lt;br&gt;
unfold nuclear power reactor signals is proposed. It includes a&lt;br&gt;
combination of convolutional neural networks (CNN), denoising&lt;br&gt;
autoencoders (DAE) and k-means clustering of representations.&lt;br&gt;
Monitoring nuclear reactors while running at nominal conditions&lt;br&gt;
is critical. Based on analysis of the core reactor neutron flux, it is&lt;br&gt;
possible to derive useful information for building fault/anomaly&lt;br&gt;
detection systems. By leveraging signal and image pre-processing&lt;br&gt;
techniques, the high and low energy spectra of the signals were&lt;br&gt;
appropriated into a compatible format for CNN training. Firstly,&lt;br&gt;
a CNN was employed to unfold the signal into either twelve or&lt;br&gt;
forty-eight perturbation location sources, followed by a k-means&lt;br&gt;
clustering and k-Nearest Neighbour coarse-to-fine procedure,&lt;br&gt;
which significantly increases the unfolding resolution. Secondly, a&lt;br&gt;
DAE was utilised to denoise and reconstruct power reactor signals&lt;br&gt;
at varying levels of noise and/or corruption. The reconstructed&lt;br&gt;
signals were evaluated w.r.t. their original counter parts, by way&lt;br&gt;
of normalised cross correlation and unfolding metrics. The results&lt;br&gt;
illustrate that the origin of perturbations can be localised with&lt;br&gt;
high accuracy, despite limited training data and obscured/noisy&lt;br&gt;
signals, across various levels of granularity.&lt;/p&gt;</subfield>
  <datafield tag="773" ind1=" " ind2=" ">
    <subfield code="n">doi</subfield>
    <subfield code="i">isVersionOf</subfield>
    <subfield code="a">10.5281/zenodo.1410083</subfield>
  <datafield tag="024" ind1=" " ind2=" ">
    <subfield code="a">10.5281/zenodo.1410084</subfield>
    <subfield code="2">doi</subfield>
  <datafield tag="980" ind1=" " ind2=" ">
    <subfield code="a">publication</subfield>
    <subfield code="b">conferencepaper</subfield>
All versions This version
Views 5252
Downloads 139139
Data volume 394.5 MB394.5 MB
Unique views 5252
Unique downloads 135135


Cite as