基于近紅外光譜的柴油性質(zhì)基團(tuán)法預(yù)測模型.pdf

1、中國石油大學(xué)（華東）碩士學(xué)位論文基于近紅外光譜的柴油性質(zhì)基團(tuán)法預(yù)測模型姓名：王志芳申請學(xué)位級別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：夏伯鍇20090401PredictiveModelofGroupmethodforDieselOilPropertiesBasedonNearInfraredSpectroscopyWangZhifang(ControlTheoryandControlEngineering)DirectedbyProf

2、XiaBokaiAbstractAsarapid，nondestructive，safeandenvironmentprotectiveanalyticaltechnique，nearinfrared烈IR)spectroscopyanalyticaltechniquehasgotbroadapplicationinthefieldofpetrocheIllicalindustrySincetraditionallaboratoryan

3、alyticalmethodsfordieseloilproductspropertiesareusuallyhighcostandlongtimedelayedthisthesisgivesaresearchonthegroupmethodpredictivemodelfordieseloilproductspropertiesbasedontheNIRspectroscopyInthisthesis，thesuperiorityof

4、NIRspectroscopytechniqueonthepredictionofpetroleumproductsgroupstructureaswellaStherelationshipbetweenC—HgroupsofNIRspectroscopyanddieseloilproductspropertiesarediscussedandallideafordirectlypredictingthedieseloilproduct

5、sproperties謝tllcrucialCHgroupsbasedontheNIRspectroscopyisproposedFirstlytheNIRspectroscopydatafordieseloilispretreated，thenaccordingtOtherelativetheoryofgroupabsorbanceanddieseloilproperties，severalcrucialgroupsarechosen

6、asmodelinginputvariablesandanon—lineardieseloilpropertiespredictivemodelisdeveloped，whoseparametersaledeterminedbyGaussNewtonmethodThecharacterofthemodelisthatthecomputingamountdependsonnumberofinputvariablesamplesrather

7、thantheinputdimensionandCangethighmodelperformancewhiletreatingnonlinearproblemsUsingtheNIRspectroscopydataofdieseloilsamplesfromEigenvectorResearchCorporationNIRspectroscopypredictivemodelsfordieseloilproperties，suchasc

8、etanenumberhavebeenestablishedbyusingPLS，BPANN，SVMandthegroup—methodseparatelyThesimulationexperimentsshowthatthegroupmethodmodelgainshigllerpredictiveaccuracyandbettergeneralizationthantheothermodels，becauseofmechanismr