复杂体系近红外光谱定量分析建模新方法研究.pdf

摘要 组烟草样品用于测试该方法的预测性能，计算结果显示与传统PLS和MB-PLS 相比结合小波变换(DWT)的加权MB．PLS方法可以给出更好的预测精度和模 型可解释能力。 关键词：近红外光谱；化学计量学；定量分析；多模块偏最小二乘 Ⅱ ABSTRACT ABSTRACT Near isa withlittleor infrared(NIR)spectroscopyrapid，nondestructiveanalysis no andhasbeen in and samplepreparation applied isbasedonmeasurementofthe petrochemicalindustry,etc．However,NIRspectrum overtoneandcombination ofthevibrationsofchemical frequencies bonds，and recordsthe informationof and makeit mainly absorption C—H，0-H，N-HS-H，which weakand are relatively highly methods usedtoextractthe information．Inthis commonly requiredquantitative thesis，new and selectionmethodswere forthe wavelength modeling investigatedquantitative of contentsofthisthesisinclude： analysiscomplexsamples．The 1．Latent in selectionforN瓜 graph(LPG)Waswavelength projective adopted onthe thatcollinear inthe spectralanalysis．Basedassumption wavelengths calibration havethesamecontributiontothe shouldbe spectramay modeling，it touse thevariablesthatlocatedattheinflexionsinLPGTwoNIR enough only datasetsoftobacco for thenicotineandreducedcontents samplespredicting