欢迎来到天天文库
浏览记录
ID:31988866
大小:3.19 MB
页数:88页
时间:2019-01-30
《近红外和拉曼光谱法测定蜂蜜掺假-.研究》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、张欣硕+}:学位论文目录ABSTRACTNearinfraredspectroscopyisefficientforthedetectionofreducingsugarinhoney,whichcostlittletimeandbewithhighaccuracy.Meanwhile,datapretreatmentmethods,suchasMSC,Norrissmooth,firstderivative,secondderivative.areusefultreatmenttoimprovetheaccuracyofthemodel.Thepredictionaccu
2、racyofthemodeliShigherthanthosebuiltbyoverseasresearchers.Thisarticlestudiedthetestmethodsoffructose,glucose,sucroseandmaltosebyNear.infraredspectroscopy.Theresultsshowthat,Kennard—StonemethodiSaneffectivealgorithmtodevisethecalibrationsetandthevalidationset.MonteCarlocrossvalidationwasuse
3、dtoexcludetheoutliers.Thedeterminationcoefficientsofpredictvalueandtherealvaluefructose,glucose,sucroseandmaltoseare0.945,0.927,O.90l,0.610and0.971,0.940,0.949,0.750separatelybeforeandafterexcludedtheoutlier.Itefficientlyimprovedthedatadistributionandtheaccuracyofthemodel.Themodelbuiltwith
4、NearinfraredspectroscopyaftertheWavelettransfoIllliSnosignificantchange.butitiSlesstimeconsuming.whichiS2.76sotherthan3.65sbeforetheWavelettransforill.Thehoneysamplesthatwereusedtobuildmodel,werecollectedfromteensofprovident.thegeographicandpollenofthehoneyhavesomeimpactsonthespectrum.Itis
5、widerimplicatedundercomplexconditions.Toadapttheneedoftheactualdetection,theadulterationsampleswereincludedinthemodel.Byusingthechmometricmethods,theresultspredictedbytheNearinfraredspectroscopyareaccurateandreliable.SoitiSfeasibilitytobuildarapidonlinetestingtechnologyforthequalitycontrol
6、ofhoney.ThecrossvalidationcoefficientofdeterrninationofHⅣ匝andtotalacidwere0.571and0.640separately.ThebestpredictionaccuracyoftheBPneuralnetworkisover90%inallthedifferentwaysofadulterationinhoneysamplesincludingadulteratesucrose,glucoseseparatelyandadulteratesucroseandglucosetogether.Butthe
7、BPneuralnetworkiswithpoorstability,andithadgottotallynI张欣硕I:学位论文目录differentdiscriminateresultsforthesamesetofdata.UsingRBFneuralnetworkinstead,47iterationswaschosen,andsettheMSEaSO.05,theerrorrateofvalidationsetandtestsetwerelowerthanthatofBPneuralnetwork,andt
此文档下载收益归作者所有