【作者】 陈民森；

【导师】 蒋诚志；

【作者基本信息】 天津大学 ， 测试计量技术及仪器， 2005， 硕士

【摘要】 在近红外光谱分析技术中,影响近红外光谱的因素很多,如被测样品的组分之间的相互作用、光谱仪的影响等,这些因素在很大程度上破坏样品的光谱,必须将这些影响光谱测量的背景因素进行扣除,以提高光谱测量的准确性,从而保证校正模型的适用性和稳健性。本文通过对葡萄糖水溶液、牛奶溶液以及人体血糖进行近红外光谱检测的实验分析,对扣除背景误差方法的适用性进行研究。论文的主要研究内容包括:1.扣除背景方法原理及分析。首先介绍了近红外光谱测量原理;然后从双光路补偿和扣标准液方法上对扣除背景方法的原理进行介绍;最后介绍了近红外光谱的多元数据分析方法:多元线性回归(MLR)、主成分回归(PCR)以及偏最小二乘(PLS),并给出了校正模型的评价指标。2.扣除背景方法在葡萄糖水溶液中的实验研究。实验结果表明:扣除水背景的影响后,模型的建模精度提高了37.9%,预测精度提高了22.5%。因此,在糖水溶液中采用扣除水背景的方法能够有效的消除一部分仪器漂移以及水的强吸收带来的影响,提高了近红外光谱测量的精度。3.扣除背景方法在牛奶溶液中的实验研究。实验结果表明:扣除背景的方法在测量牛奶中的脂肪和蛋白浓度时,并不能够有效地扣除牛奶中其他成分相互干扰误差的影响,提高测量的精度。但是,在测量牛奶中的葡萄糖浓度时,扣除纯牛奶背景的影响后,模型的建模精度提高了26.2%,预测精度提高了68.4%。因此,在牛奶溶液中采用扣除纯牛奶背景的方法测量牛奶中的葡萄糖浓度,能够有效的消除牛奶中其他成分干扰误差的影响,提高了近红外光谱测量的精度。4.扣除背景方法在人体血糖无创检测中的实验研究。OGTT实验结果表明:扣除背景后,模型的建模精度和预测精度都没有得到提高。因此,在人体血糖无创伤检测中,寻找一种合适的标准物作为背景以消除仪器漂移的影响,成为下一步工作的重点。更多还原

【Abstract】 During the near-infrared spectroscopy analysis, the reciprocity of thecomponent in the sample and the influence of the spectrum instrument willdestroy the spectrum of the sample. So they must be deducted to increase theveracity of the spectrum measurement. This will ensure the suitability androbustness of the calibration models. In the dissertation, the method of deductingbackground error is studied by the experiment of aqueous matrix、milk andblood glucose of the body.The main research content of the dissertation involves:Firstly, the principle of the method of deducting background error isintroduced. First of all, the principle of near-infrared spectrum measurement isintroduced;then the principle of the method of deducting background error isintroduced by the compensation of both light-paths and deducting the standardsolution;the methods of multi-date analysis are introduced: the MultivariateLinear Regression, Principle Component Analysis and Partial Least Squire. Theevaluating guidelines of the calibration models are also presented.Secondly, the suitability of the method of deducting background error in theaqueous matrix is studied. The results show that the modeling precisions ofanalytical model have raised 37.9% and the predictable precisions have raised22.5% after deducting the background error of water. So deducting thebackground error in the aqueous matrix to eliminate partially influences of theexcursion to the instrument and the strong absorption of water raises theprecision of near-infrared spectrum measurement.Thirdly, the suitability of the method of deducting background error in themilk is studied. The results show that the method of deducting background errorcannot deduct effectively the error of the reciprocity of the other components ofthe milk when measuring the fat and the protein in the milk. But when measuringthe glucose in the milk, the modeling precisions of glucose model have raised26.2% and the predictable precisions have raised 68.4% after deducting thebackground error of pure milk. So deducting the pure milk background errorwhen measuring the glucose in the milk can deduct effectively the error of thereciprocity of the other components of the milk and raises the precision of near-infrared spectrum measurement.Lastly, the suitability o the method of deducting background error in theblood glucose of the body measurement is studied. The OGTT result shows thatthe modeling and predictable precisions of analytical model aren’t raisedobviously after deducting from the background. So looking for a suitablestandard matter as background to eliminate partially influence of the excursion tothe instrument is important for the noninvasive monitoring of the blood glucose.更多还原