Instability localization with artificial neural networks (ANNs) T.Tambouratzis*,M.Antonopoulos-Domis Institute of Nuclear Technology, Radiation Protection, NCSR ‘Demokritos’, Aghia Paraskevi 153 10, Athens, Greece Received 19 February 2001; received in revised form 14 March 2001; accepted 17 March 2001 Abstract Theaimofthispieceofresearchistoinvestigatethepotentialofartificialneuralnetworks (ANNs) for tackling the problem of instability localization. The instability is modeled by a variablestrengthabsorber(point-source)inatwo-dimensionalbarereactormodelwithaone neutron-energygroup.Theproposedapproachconstitutesanexerciseinsimplicityinthat:(1) anarbitrarilysimplifiedmodelisemployedforANNtrainingandvalidation;(2)fewtraining andvalidationpatternsoflowcomplexityareutilized;(3)theANNinputsarederiveddirectly from the neutron noise signals, the proposed location of instability is given on-line via an uncomplicatedcombinationofANNoutputs;(4)theANNarchitectureisindependentofthe numberofpossiblelocationsofinstability.Infact,unlikepreviousapproacheswhichemploy hundreds of outputs (one for each fuel assembly), only two ANN outputs are employed representingtheX-andY-coordinates(location)ofinstability;(5)theresponsesofonlyafew detectors are employed; (6) a measure of confidence in the prediction is assigned. The results of ANN testing, which is performed on patterns from both actual and simplified models, are reported and analyzed. # 2001ElsevierScienceLtd.Allrightsreserved. 1. Introduction Local density wave oscillations in a BWR core give rise to neutron flux fluctua- tionsinthenearbylocalpowerrangemonitors(LPRMs).Suchoscillationsinduced by local channel thermal hydraulic instabilities (probably self-sustained coolant densitywaveoscillations)havebeenobservedinpractice,forinstanceattheFinnish Annals of Nuclear Energy 29 (2002) 235–253 www.elsevier.com/locate/anucene 0306-4549/02/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved. PII:S0306-4549(01)00036-6 * Corresponding author. Tel.: +30-1-65-03-748; fax: +30-1-65-33-431. E-mail address: tatiana@ipta.demokritos.gr (T. Tambouratzis).