1、南京理工大學(xué)碩士學(xué)位論文數(shù)據(jù)校正技術(shù)研究姓名：于永軍申請學(xué)位級別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：周凌柯20100620AbstractEverydaymodementerprisesarecollectingalargeamountofmeasurementsThesemeasurementscanreflecttheoperatingconditionsofproductiondevices，andisalsothebasis

2、ofvariousworkProcessmeasurementsareofteninaccurate，andevencontaingrosserrorsIfthesemeasurementswithgrosserrorsaredirectlyused，theenterprisesCannoteffectivelycarryouttheworkandevenhaveagreateconomiclossSoitisnecessarytoco

3、rrectmeasurementstoobtainhighqualityandreliabilitydataforprocessindustriesInthisdissertation，somemethodsofgrosserrordetectionareanalyzedandstudiedThenthemethodswithsomedefcctsaleimprovedThemaincontributionsofthisdisserta

4、tionaredescribedasfollows：(1)SomeproblemsofMTNTmethodarediscussed，thenanimprovedmethodisproposedAfteragrosserrorisdetected，itsvarianceinsteadofitsvalueiscorrected，andthenitsoriginalvarianceisreplacedwithcorrectedvalueThe

5、improvedmethodCanovercomethedecreaseofcoefficientmatrixrankandalseducetheprobabilityofmisidentificationThesimulationresultsshowthatthismethodiSeffective(2)TosolvethemisjudgmentofGeneralizedlikelihoodratio(GLR)method，noda

6、ltestmethodisintroduced，andthenacombinedmethodisproposedAfteragrosserrorisdetectedbyGLRmethod，nodaltestmethodwilltestthegrosserrordetectedagainThecombinedmethodCanavoidthemisjudgmenttoacertainextentAndthesimulationresult

7、sshowitseffectiveness(3)TheproblemsofserialidentificationwithcollectivecompensationmethodareanalyzedanddiscussedAfterthatthelinearcombinationtechniqueisintroduced，andthenailimprovedmethodispresentedThemethodcailreducethe

8、numberofvariablesinthesetofcandidategrosserrors，SOthatitCanreducetheprobabilityofmisidentificationBesides，thismethodCanalsoavoidthecomputationofprojectmatrix(4)Severalcommonmethodsofdataclassificationareintroduced，andthe