简介：Accordingtothemultipleresearchesinthelastcoupleofyears,laser-inducedbreakdownspectroscopy(LIBS)hasshownagreatpotentialforrapidanalysisinsteelindustry.Nevertheless,theaccuracyandprecisionmaybelimitedbycomplexmatrixeffectandself-absorptioneffectofLIBSseriously.Anovelmultivariatecalibrationmethodbasedongeneticalgorithm-kernelextremelearningmachine(GA-KELM)isproposedforquantitativeanalysisofmultipleelements(Si,Mn,Cr,Ni,V,Ti,Cu,Mo)inforty-sevencertifiedsteelandironsamples.First,thestandardizedpeakintensitiesofselectedspectralinesareusedastheinputofmodel.Then,thegeneticalgorithmisadoptedtooptimizethemodelparametersduetoitsobviouscapabilityinfindingtheglobaloptimumsolution.Basedonthesetwostepsabove,thekernelmethodisintroducedtocreatekernelmatrixwhichisusedtoreplacethehiddenlayer’soutputmatrix.Finally,theleastsquareisappliedtocalculatethemodel’soutputweight.InordertoverifythepredictivecapabilityoftheGA-KELMmodel,theR-squarefactor(R2),Root-mean-squareErrorsofCalibration(RMSEC),Root-mean-squareErrorsofPrediction(RMSEP)ofGA-KELMmodelarecomparedwiththetraditionalPLSalgorithm,respectively.TheresultsconfirmthatGA-KELMcanreducetheinterferencefrommatrixeffectandself-absorptioneffectandissuitableformulti-elementscalibrationofLIBS.

简介：Theaccuracyoflaser-inducedbreakdownspectroscopy(LIBS)quantitativemethodisgreatlydependentontheamountofcertifiedstandardsamplesusedfortraining.However,inpracticalapplications,onlylimitedstandardsampleswithlabeledcertifiedconcentrationsareavailable.Anovelsemi-supervisedLIBSquantitativeanalysismethodisproposed,basedonco-trainingregressionmodelwithselectionofeffectiveunlabeledsamples.Themainideaoftheproposedmethodistoobtainbetterregressionperformancebyaddingeffectiveunlabeledsamplesinsemi-supervisedlearning.First,effectiveunlabeledsamplesareselectedaccordingtothetestingsamplesbyEuclideanmetric.Twooriginalregressionmodelsbasedonleastsquaressupportvectormachinewithdifferentparametersaretrainedbythelabeledsamplesseparately,andthentheeffectiveunlabeledsamplespredictedbythetwomodelsareusedtoenlargethetrainingdatasetbasedonlabelingconfidenceestimation.Thefinalpredictionsoftheproposedmethodonthetestingsampleswillbedeterminedbyweightedcombinationsofthepredictionsoftwoupdatedregressionmodels.Chromiumconcentrationanalysisexperimentsof23certifiedstandardhigh-alloysteelsampleswerecarriedout,inwhich5sampleswithlabeledconcentrationsand11unlabeledsampleswereusedtotraintheregressionmodelsandtheremaining7sampleswereusedfortesting.Withthenumbersofeffectiveunlabeledsamplesincreasing,therootmeansquareerroroftheproposedmethodwentdownfrom1.80%to0.84%andtherelativepredictionerrorwasreducedfrom9.15%to4.04%.

简介：Anexperimentalsetuphasbeendesignedandrealizedinordertooptimizethecharacteristicsoflaser-inducedbreakdownspectroscopysystemworkinginvariouspressureenvironments.Anapproachcombinedthenormalizationmethodswiththepartialleastsquares(PLS)methodaredevelopedforquantitativeanalysisofmolybdenum(Mo)elementinthemulti-componentalloy,whichisthefirstwallmaterialintheExperimentalAdvancedSuperconductingTokamak.Inthisstudy,thedifferentspectralnormalizationmethods(totalspectralareanormalization,backgroundnormalization,andreferencelinenormalization)areinvestigatedforreducingtheuncertaintyandimprovingtheaccuracyofspectralmeasurement.TheresultsindicatesthattheapproachofPLSbasedoninter-elementinterferenceissignificantlybetterthantheconventionalPLSmethodsaswellastheunivariatelinearmethodsinthevariouspressureformolybdenumelementanalysis.