模糊神经网络在列车制动控制中的建模及应用

《模糊神经网络在列车制动控制中的建模及应用》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、北京交通大学硕士学位论文模糊神经网络在列车制动控制中的建模及应用姓名：吴海俊申请学位级别：硕士专业：系统理论指导教师：毛保华;孙全欣20080601ABSTRACTABSTRACT：Safetyandefficiencyareeternalthemesforrailwaytransportation，especiallyastherapiddevelopmentofChina’Srailway．Inrecentyears，therailwaytransportationhasshownheavy'high—speed，high-densitycharacteristics，allwhi

2、chneedahighreliabletraincontrolsystemasaguarantee．Brakingcontrol，asapartofthetraincontrol，playsakeyroleintrainsafetyButtherearesomeerrorsbetweenthebrakingmodecurvebasedontrain··traction-·calculationtheoryandtheactualoperationofthetrain，whichresultsinthattheeffectofthecontrolisnotgood．Therefore，h

3、owtoconstructamoreaccuratetrainbrakingcontrolmodelbecomesahotresearchinrecentyears．Complexsystemsmodelingandsimulationisakeyissueinsystemsmodelingfield．FuzzyNeuralNetwork(FNN)integratedbothadvantagesoffuzzysystemsandneuralnetworks，canbetterrealizecomplexsystemsmodeling．Inthispaper,themethodofFNN

4、modeling,FNNstructurewhichissuitableformodelingofthetrainbrakingcontrolanditsapplicationshavebeenstudied，whichisbasedonthebackgroundofcomplexsystemsmodeling．Thispaper,basedontheresearchathomeandabroad，isontheissueoftrainbrakingcontrolmodeling，andthemainresearchcontentsareasfollows：1．Thefuzzyprob

5、lemsexistedinthetrainbrakingcontrolanditscauseshavebeenanalyzedfromtheperspectiveofcomplexsystemsmodeling，anditisdrawnthattheFNNusedinthetrainbrakingmodelingisfeasible．2．Inthispaper，itisfocusedonhowtoconstructasuitableFNNmodelfortrainbrakingcontrolunderthepremisethatthecharacteristicsofthebrakin

6、gcontrolsystemisnotclearenough．ThenthestandardFNNhasbeenimprovedandanimprovedfourlayersfuzzyneuralnetworkWasacquiredwithitslearningalgorithmdeduced，whichisprovedthatthismodelhasthecharacteristicsoftheoverallapproachintheory．Afterwards，weusedtheimprovedmodeltothetrainbrakingprocessanddeterminethe

7、inputandoutputvariablesfrompracticalpointofviewonthemanipulation，whichmadethemodelismoreadaptedtotheactualoperatingenvironment．Thisisdifferentfromthepreviousmodelstructureandthevariablesselectedarealsodifferent．3．Traditional