FT-NIR光谱仪测定烟草化学成分不同谱区范围对数学模型影响的研究

应用傅里叶近红外技术建立快速定量分析烟草化学成分的数学模型 ,使用BrukerOptics公司的MPA型傅里叶变换近红外光谱仪 ,谱区扫描范围 12 0 0 0～ 4 0 0 0cm- 1 ,选择不同的谱区范围对烟碱、总糖和总氮三种组分的数学模型进行优化 ,结果显示不同谱区范围对同一组分数学模型影响有明显的差异 ,而且不同组分所选择的最佳建模谱区范围是不一样的 ,说明在模型优化过程中选择最佳谱区范围是非常关键的 ,烟碱、总糖和总氮三种组分模型的最佳谱区范围分别是 95 0 0～ 4 2 31 2cm- 1 ,75 0 2 1～ 4 2 4 6 7cm- 1 ,75 0 2 1～4 5 97 7cm- 1 ,三种组分最佳模型交叉检验的均方差 (RMSECV)分别为 0 0 815 ,0 80 8,0 0 5 6。

Determination of Chemical Components in Tobacco Leaves by FT-NIR Spectroscopy:Study of Influence of Spectral Ranges on PLS Modeling

NIR spectra of tobacco leaves were measured in the range of 12 000 to 4 000 cm -1 using a Bruker MPA FT-NIR spectrometer. PLS calibration models were developed and optimized for rapid quantitative analysis of nicotine alkaloids, total sugar and total nitrogen contents in tobacco leaves. It was found that the prediction errors of the same component were significantly different when different spectral regions were used for PLS modeling, and the best spectral range is also different for each component. The study demonstrated that wavelength range selection is one of the important keys to optimizing the NIR calibration model. In this study it was found that the optimized calibration ranges for nicotine alkaloids, total sugar and total nitrogen are 9 500-4 231.2 cm -1, 7 502.1- 4 246.7 cm -1 and 7 502.1-4 597.7 cm -1, respectively. The Root Mean Square Error of Cross Validation (RMSECV) of the three calibration models are 0.081 5, 0.808 and 0.056, respectively.