基于RBF网络辨识的模型参考自适应控制系统分析与仿真

《基于RBF网络辨识的模型参考自适应控制系统分析与仿真》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、西南交通大学硕士研究生学位论文第1l页AbstractModelreferenceadaptivecontrol(MRAC)isallLmportantadaptivecontr01．Ithadmorematuretheoriesandmethodsofanalysisandsynthesis．Andinpractice，itwasbecomingusedwidely,suchasautopilotoftheaircraft，autopilotsystemoftheship，servosystemofthephotoelectricopticaltrackingtelescope，spe

2、ed—governingsystemofthecontrolledsilicon,controlsystemofthemanipulatorandSOon．However,thetechnicalmeansoftraditionalalgorithmformodelreferenceadaptivecontrolisverylimited．Whenitencountercomplexnonlinearsystems，thedesignandimplementationisverydifficult．Inordertogivefullplaythesuperiorperformanceof

3、modelreferenceadaptivecontrol，andincreasetherobustness，real-time，faulttoleranceofcontrolandtheadaptiveandlearningabilityofcontrolparameters，thepeoplemademodelreferenceadaptivecontrolandneuralnetworks(NNs)upappropriatlytothesystemofmodelreferenceadaptivecontrolbasedontheneuralnetworks．硼舱thesisanal

4、yzedthedesignofthemodelreferenceadaptivecontrolsystem．andusedtheradialbasisfunction(RBF)neuralnetworktothenonlinearsystemidentificationasanidentifierinthecontrolsystem．Firstly,thethesisintroducedthebasictheoriesofthemodelreferenceadaptivecontrolsystem．Then,thethesisintroducedthreeclassicdesignpla

5、nsoftheadaptivecontrollawsandstudiedthesimulationofthreeclassicexamples．ThethesisanalyzedtheRBFneuralnetworkasanidentifierinthecontrolsystem．Furthermore，thethesisimprovedthelearningalgorithmoftheRBFneuralnetworkandobtainedanimprovedK-meansalgorithm．Then,asimulationexampleofsystemidentificationhad

6、beenshown．Finally,theRBFn．euralnetworkidentifierbasedonimprovedK-meanslearningalgorithmhadbeenusedinthesingleneuronPIDmodelreferenceadaptivecontrolsystem．Andasimulationexamplehadbeenanalyzed．Thethesisfocusedonthethesimulationandanalysisofthreekindsofclassicdesignplansofthemodelreferenceadaptiveco

7、ntrolandthesummaryofadvantagesanddisadvantages．Onthisbasis，thethesisintroducedtheneuralnetworkmodelreferenceadaptivecontrolsystemsandimprovedthetrainingalgorithmoftheneuralnetworksidentifier．Finally,thethesisgaveasimul