神经网络自适应控制器仿真.研究

《神经网络自适应控制器仿真.研究》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、太原理工大学硕士研究生学位论文RESEARCHOFNNSELF—ADAPTIVECONTROLERABSTRACTTheprocessofindustrialproductionissometimesnonlinearanduncertainandisdifficulttoestablishaccuratemathematicmodule，SOusingroutinePIDcontrollerishardtoachieveperfectcontrollingeffect．Neuralnetwork(NN)hasfavorab

2、lenonlinearmappingperformanceandhigherparallelinformationprocessingcapability,andithasbeenthemostengagingtheoryandmethodfornonlinearsystemmodeling，recognitionandcontr01．Onthebaseofconsultingliteraturesbothhereandabroad，themodelofNNidentification，PIDcontrollerbaseds

3、ingle—neuron，PIDadaptivecontrollerandtheNNadaptivecontrollerarediscussed．AimedattheshortagesofPID，theconnectionbetweenNNandadaptivecontrollerareresearchedinordertoacceleratingconvergence-speedandtheaccumcy．Thispaperarebasedontheresearchingofthemodel，realizationsand

4、theperformanceofthealgorithms．ThemaintaskCanbeC0nCludedasfcIllows(1)AimedattheasortofNon—linearsystem，thepaperproposedanIV太原理上大学硕士研究生学位论文stateobserverbasedontheBPNNinverse—model，thismodelcanobservethestateofsysteminrealtime．Andtheanalysisoftheoryandsimulationshowed

5、thatthestateobserverisverygood(2)Therehavesomeproblemsinthesingle-neuronadaptivecontroller：①thechoosingofweightcoefficient．Theinitialvalueoftheparameterhasstrongeffectonthecontrolperformance@Theslicingproblemofweightcoefficient．whentheparameterisimpregnated，thestud

6、yabilitywilllost．⑨Therisetimeislong，anti-jammingcapabilityisinfirmness．Inordertosolvetheproblem，twoimprovedmethodareproposed：thefirstisusingGAtooptimizetheparameter,andthesimulationshowedthewaycanfindthebestparameterandeliminatingtheeffectofinitialvalue．Thesecondis

7、acompoundedadaptivePIDcontrollerbasedonCMAC．thestudyprocessofCMACinvolvedthetotalcontrolprocessofsystem，thesimulationshowedthemethodhadthestrongadaptiveabilityandanti-jammingcapabilityetc．(3)Analyzedthemodel，algorithmandcharactersoftheadaptivecontrollerbasedonBPNN．

8、Inorderimprovetheeffectoncontrol,theappropriatedmodelmustbeseted．UsingtheideaofNNcontrarymodelrecognition，thetextputforwardadesigningschemethatNN