基于神經(jīng)網(wǎng)絡(luò)的電液伺服地震模擬振動臺的控制研究.pdf

1、湖南大學碩士學位論文基于神經(jīng)網(wǎng)絡(luò)的電液伺服地震模擬振動臺的控制研究姓名：夏玲瓊申請學位級別：碩士專業(yè)：結(jié)構(gòu)工程指導教師：劉一江易偉建20090901AbstractThestmcturalseismictestisanecessarystepinresearchofquakeproofanddisasterpreventingWiththedevelopmentofthecontroltheoryandcomputertechn0109

2、Ythetendofseismictestmethodsistobelargescaled，digitized，accuratedandintelligentApplicationofnewtestingmethodsonQuasi。statictesting，PseuIdo—dynamictestingandtheShakingtabletestisthehotspotoftheresearchoftheseismictestingo

3、fstructures，especiallytheshakingtabletestbasedonintelligentcoll仃olmethods，whichevenbecametheleading—edgeofthestructuralseismictestTheidiographicworkisreconstructingtheservohydrauliccontrolsystemoftheSha虹ngtable，whichisim

4、portanttotheresearchoftheseismicsimulatedshakingtabletestmethodInthispaper，thebasiccontrolmethodsofelectrohydraulicshakingtableandthetrendofuseofintelligentcontroltechnologyareanalysed，anditisfocusingonapplicationofitera

5、tivelearningcontrolalgorithmforneuralnetworksintheshakingtablemotionaccelerationwaveformreproductiontechnologyFortheneedofStmctu】reLaboratoryofHunanUniversityshakingtableaccelerationwaveformreproductionbasedonthePIDcontr

6、olloop，theapplicationofneuralnetworkson。lineiterativelearningcontrolandmultiparametercontroltheorybuildanexternalaccelerationlooptoformthedoubleclosedloopcontr01One‘useofMTScontrolsystemwhichisPIDcontrolmethodoftheexisti

7、ngTheouterloopbyconfiguringtheaccelerationsensors，velocitysensors，displacementsensorsandthepreparationoftheⅥ，avefo冊conversionandloadingprocess，usingneuralnetworkiterativelearningcont∞lalgorithmtoachieveanexternalclosedlo

8、opcontroltomodifythePIDparametersThisouterloopthroughtheRBFnetworktecognizetheinformationofthesensitivityonthevmiateofcontrolinputandfeedback，andonlinetraintheRBFnetworkAcertainalgorithmwasimprovedbyaddingordeletingnodes

9、inhiddenlayerneuralnetworktoacceleratethespeedofonlinelearningalgorithmThisarticlealsoattemptstogeneticalgorithmtooptimizetheneuralnetworkinitialparameters，increasingthenetworksabilityofglobaloptimizationThefinaladoption