identification and control of dynamical systems using neural networks

《identification and control of dynamical systems using neural networks 》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、II1IEEETRANSACTlONSONNEURALNETWORKS.VOL.I.NO.I.MARCH1990IdentificationandControlofDynamicalSystemsUsingNeuralNetworksKUMPATIS.NARENDRAFELLOW,IEEE.ANDKANNANPARTHASARATHYAbstract-Thepaperdemonstratesthatneuralnetworkscanbeusedarewellknownforsuchsystems[

2、13.Inthispaperoureffectivelyfortheidentificationandcontrolofnonlineardynamicalinterestisintheidentificationandcontrolofnonlinearsystems.Theemphasisofthepaperisonmodelsforbothidentificationdynamicplantsusingneuralnetworks.Sinceveryfewre-andcontrol.Stat

3、icanddynamicback-propagationmethodsforthead-justmentofparametersarediscussed.Inthemodelsthatareintro-sultsexistinnonlinearsystemstheorywhichcanbedi-duced,multilayerandrecurrentnetworksareinterconnectedinnovelrectlyapplied,considerablecarehastobeexerci

4、sedintheconfigurationsandhencethereisarealneedtostudytheminaunifiedstatementoftheproblems,thechoiceoftheidentifierandfashion.Simulationresultsrevealthattheidentificationandadaptivecontrollerstructures,aswellasthegenerationofadaptivecontrolschemessugge

5、stedarepracticallyfeasible.Basicconceptsandlawsfortheadjustmentoftheparameters.definitionsareintroducedthroughoutthepaper,andtheoreticalques-tionswhichhavetobeaddressedarealsodescribed.Twoclassesofneuralnetworkswhichhavereceivedconsiderableattentionin

6、theareaofartificialneuralnet-worksinrecentyearsare:1)multilayerneuralnetworksI.INTRODUCTIONand2)recurrentnetworks.MultilayernetworkshaveATHEMATICALsystemstheory,whichhasintheprovedextremelysuccessfulinpatternrecognitionprob-Mpastfivedecadesevolvedinto

7、apowerfulscientificlems[2]-[5]whilerecurrentnetworkshavebeenusedindisciplineofwideapplicability,dealswiththeanalysisassociativememoriesaswellasforthesolutionofopti-andsynthesisofdynamicalsystems.Thebestdevelopedmizationproblems[6]-[9].Fromasystemstheo

8、reticpointaspectofthetheorytreatssystemsdefinedbylinearop-ofview,multilayernetworksrepresentstaticnonlineareratorsusingwellestablishedtechniquesbasedonlinearmapswhilerecurrentnetworksarerepresentedbynonlin-algebra,complexvariabletheory,andthet