资源描述：
《应用信度分配的模糊CMAC实现非线性系统的容错控制》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、Vol.32,No.3ACTAAUTOMATICASINICAMay,2006Fault-tolerantControlofNonlinearSystemUsingCreditAssignFuzzyCMAC')ZHUDa-QiKONGMin(ResearchCen￡二ofControlScienceandEngineering,SouthernYangtzeUniversity,Wuxi214122)(E-mail:zdg367Qyahoo.com.cn,zhizi0708Qsohu.com)AbstractTheadaptivefaul
2、t-tolerantcontrolschemeofdynamicnonlinearsystembasedonthecreditassignedfuzzyCMACneuralnetworkispresented.Theproposedlearningapproachusesthelearnedtimesofaddressedhypercubesasthecredibility,theamountsofcorrectingerrorsareproportionaltotheinversionofthelearnedtimesofaddress
3、edhypercubes.Withthisidea,thelearningspeedcanindeedbeimproved.BasedontheimprovedCMAClearningapproachandusingtheslidingcontroltechnique,theeffectivecontrollawreconfigurationstrategyispresented.Thesystemstabilityandperformanceareanalyzedunderfailurescenarios.Thenumericalsim
4、ulationdemonstratestheeffectivenessoftheimprovedCMACalgorithmandtheproposedfault-tolerantcontroller.KeywordsCreditassignedfuzzyCMAC,faultdiagnosis,fault-tolerantcontrol,nonlinearsystem1IntroductionDuringthelasttwodecades,extensiveresearchactivitieshavefocusedonfault-toler
5、antcontrol(FTC)tomaintainthesystemstabilityandtoavoidlossesundervariousfailurescenarios.Reference[1-3]providedexcellentoverviewsofrecentresearchworkonFTC.However,linksbetweenfaultdiagnosisandFTCtechniquesarestilllackinglll，andsomerecentresultsontheintegrationoffaultdiagno
6、siswithFTCcanbefoundin[4-刘.Fromtheserecentresearchresults,twokeycomponentsmustberesearchedfurther:1)theonlinefaultdiagnosismoduleconsistingofthefaultdetectionmechanismandonlinefaultestimator;and2)thecontrollermoduleconsistingofnominalcontrollerandafault-tolerantcontroller
7、.ThefaultinformationgeneratedbydiagnosisprocedurecanbeveryusefultoFTC,andaccuracyandspeedoffaultdiagnosisareveryimportantforactivefault-tolerantcontrol.Withdevelopmentofartificialneuralnetwork,ithasstrongsuperiorityinfaultdiagnosis,especiallyforthecasesthatcannotbeexpress
8、ed饰formula,andincomplexnonlinearsituationwithstrongaccommodation.Invariousneuralnetworks,theBP(b
