【作者】 蒙杰丹；

【导师】 杨季冬；

【作者基本信息】 西南大学 ， 分析化学， 2012， 硕士

【摘要】 近红外光谱(Near Infrared,简称NIR)是位于紫外可见光谱区与中红外光谱区间的电磁波。通常将780～2526 nm(12820-3959cm-1)定义为近红外光谱区的范围。有机分子中C-H、O-H、N-H及S-H等含氢基团内部振动的倍频和组合频吸收是近红外光谱反应的基本信息,这些化学信息量相当丰富。由于绝大部分有机物都含有此类基团,因此适合近红外光谱测量的物质种类和场合相当广泛。分析速度快、样品预处理简单、可同时测定多组分、环保污染少等,是近红外光谱分析所具有的优点,这使得近红外光谱被广泛地应用于农业、石油化工、制药、环保、食品等领域中。中药材与中成药是中国传统药物,其药效成分复杂多样,外观形态各异,因此对其进行质量监控难度很大。近红外光谱结合化学计量学这一分析技术可以很快捷地对中药材与中成药的有效含量进行快速地测定,同时还可对药物的归属地进行鉴别。本文在重庆市科委科技攻关项目(CSTC 2010AC5170)的资助下,对三种剂型的中成药和两种中药材应用近红外分析手段做了以下几个方面的研究：1.近红外光谱法同时测定通天口服液的天麻素和芍药苷的含量,所建立的定量分析模型效果极佳,能应用于生产实践中,突破了传统方法只能测量单一组分的缺陷；2.用近红外漫反射光谱法建立益母草颗粒中盐酸水苏碱定量分析模型,同时还使用聚类分析方法对九个厂家的益母草颗粒进行定性分析；3.在没有光纤探头的情况下,对北豆根胶囊使用近红外光谱法进行探索性研究,分别建立了总生物碱定量分析模型、判别分析模型和聚类分析模型；4.使用近红外光谱法建立了银杏叶的定量分析与两种定性分析模型,并对各个模型进行应用；5.应用近红外光谱法对车前草的模式识别定性分析方法研究,分别用两种最经典的方法进行定性分析,均取得很好的分析效果。1.近红外光谱法同时测定通天口服液中的天麻素和芍药苷的含量基于近红外光谱结合偏最小二乘法建立了同时测定通天口服液中天麻素与芍药苷含量的方法。通过HPLC法测定通天口服液样品中天麻素和芍药苷的化学含量。随机抽取60个样本作为校正集,20个样本作为验证集。通过化学计量学方法将校正集样本的近红外光谱和与之对应样本的天麻素和芍药苷的含量分别相关联建立模型。最后得出,天麻素和芍药苷定量分析模型的决定系数分别为96.28%、94.55%,模型的交叉验证均方差分别为0.0336、0.00908,验证集的决定系数分别为94.23%、92.86%,预测集均方差分别为0.0453、0.00839。文章还做了模型的精密度实验,效果较佳。实验表明,该方法能用于大批量样品的快速分析。2.近红外漫反射光谱法对益母草颗粒的定量和定性分析本文首先利用近红外光谱结合偏最小二乘法建立了快速测定益母草颗粒的盐酸水苏碱的定量分析模型,模型的决定系数为0.9622,内部交叉验证均方差为0.0576；然后用10个预测集样本对模型的性能进行验证,决定系数为0.9495,预测集标准偏差为0.0547,相对分析误差为4.45,表明该方法能应用于生产实践。最后选取样品的近红外光谱的3999.6～6502.8 cm-1波段建立九个厂家的益母草颗粒聚类分析模型,结果不同厂家的样品能很好地进行分类,然后使用预测集样品对该模型进行检验,结果表明该方法准确快速。3.基于近红外光谱的北豆根胶囊的定量和定性分析应用近红外光谱技术建立了北豆根胶囊的总生物碱的定量和定性分析方法。收集了3个厂家85个批次的样品,将近红外光谱与相应的化学值相关联建立定量模型。模型的决定系数、内部交叉验证均方差、预测标准偏差分别为99.63%、1.04、4.49,相对分析误差为3.81,结果表明该定量分析模型能应用于生产实践中。聚类分析和K-最邻近法相结合对不同生产厂家的样品进行了鉴别,识别率分别为97%、100%。本实验在做判别分析时有了一个很大的突破：在一级谱库无法对一些样品进行正确归类时,可将这些样品归入二级子库,重新设定条件,进行进一步的定性鉴别。4.银杏叶的近红外定量分析与定性分析模型的建立及应用本文采用了近红外光谱分析方法对几个地区的银杏叶样品做了定量和定性分析研究与应用。选取7501.7～4246.5 cm-1做为建模的光谱范围,模型的预测值与标准值间的相关系数R2为98.62%,内部交叉验证均方差RMSECV为0.103,相对分析误差RPD为3.05,且5个样品的重复性实验的相对标准偏差都在2%以下,说明将该模型应用于实际检测中是可行的。在进行定性判别分析和聚类分析时,模型的识别率都达到了100%。5.基于近红外光谱法对车前草的模式识别定性分析方法研究本文采用近红外光谱技术,同时结合化学计量学方法,对车前草的归属地进行判别分类工作。在8003-4000cm-1光谱范围内,使用聚类分析方法建立了聚类分析模型,模型的识别率是100%,预测率也为100%。定性判别分析方法建模时,模型的识别率也达到了100%。研究表明,近红外漫反射光谱结合模式识别定性分析方法可作为一种可靠、准确、快速,能应用于生产实践的方法。通过实验得出这一结论：定性分析的质量很大程度上依赖光谱的预处理方法,合适的代表性样品的频率范围和定性分析方法。更多还原

【Abstract】 Near-infrared spectroscopy (for short NIR) is the electromagnetic wave between the utraviolet-visible region and mid-infrared region, which is in the range780-2526 nm (12820-3959cm-1). The overtones and combinations of vibrational modes involving C-H, N-H, O-H and S-H chemical bonds was comprised the absorption of near-infrared spectroscopy. As most of organics have these chemical bonds, the NIRS can be used in many fields. Fast, no pre-processing, simultaneous determination, unpolluted, non-destructive are the advantages of NIRS, that’s why NIRS can be used in Agriculture, Petrochemical, Pharmaceutical, Environment, Food Processing and so on.Herbal medicines and Chinese patent medicines are the tradition medicines in China. Because the drug effect component is rather complexity and multiplicity, meanwhile the appearance is hard to identify, it’s very difficulty to control the quality. However, near-infrared spectroscopy combined with chemometrics provide a wonderful technology for the quality control of Chinese medicine.Under the support of Chongqing Science and Technology Commission program (CSTC2010AC5170), this thesis mainly study on quality control of herbal medicines and Chinese patent medicines by Near Infrared Spectroscopy.1. Determinated simultaneously the content of Gastrodin and Paeoniflorin of tongtian oral liquid by near-infrared spectroscopy. Both of the models were excellent, and they can be used in practice.This method conquered the defect that the common method can’t determinated the couple matter simultaneously.2.The quantitative analysis model for the rapid determinantion of stachydrine hydrochloride of motherwort granules was based by the near infrared spectroscopy combining with the partial least square method. The clustering analysis model of motherwort granules from 9 manufacturers was established, and the samples from the different manufacturers can be primely classified.3.The total alkaloids’quantitative and qualitative analysis of Menispermi Capsule using near infrared spectroscopy were developed, without the fibre-optics probe.4. The total flavonoids’quantitative and qualitative analysis of Ginkgo Biloba Leaf using near infrared spectroscopy were developed, and it was valueable for productive practice.5. Qualitative analysis of pattern recognition for the plantain based on the Near-infrared Spectroscopy.1. Simultaneously Determination content of Gastrodin and Paeoniflorin of tongtian oral liquid by near-infrared spectroscopyA new, rapid analytical method for the analysis of Gastrodin and Paeoniflorin in tongtian oral liquid based on near infrared spectroscopy combined the partial least square method was developed. The content of Gastrodin and Paeoniflorin of the tongtian oral liquid samples was determined by HPLC, then randomly chose 60 samples for calibration set, while other 20 samples were used for validation set. The associated models of the near infrared spectroscopy in calibration set and the corresponding content of Gastrodin and Paeoniflorin were established by the chemometrics. The coefficient of determination of calibration models were 96.28%,94.55%; and the root mean squared error of corss-validation were 0.0336,0.00908, respectively. The coefficient of determination of the prediction models were 94.23%,92.86%; and the root mean squared error of prediction were 0.0453,0.00839, respectively. In the present paper, some precision experiment had been done. The method was accurate, rapid and simple. It maybe applied for the rapid determination of large numbers of samples.2. Quantitative and Qualitative Analysis of motherwort granules by Near Infrared Diffuse Reflectance SpectroscopyThe quantitative analysis model for the rapid determinantion of stachydrine hydrochloride of motherwort granules was based by the near infrared spectroscopy combining with the partial least square method. The coefficient of determination of calibration model was 0.9622, and the root mean squared error of corss-validation was 0.0576. Ten prediction samples were used to validate the performance of the calibration model. The coefficient of determination of the prediction model was 0.9495, and the root mean squared error of prediction was 0.0547. The relative prediction deviation was 4.45. It indicated that this method can be applied to the production and practice. Ultimately, the 3999.6～6502.8 cm-1 wave band of the near infrared spectroscopy was chosen to establish the clustering analysis model of motherwort granules from 9 manufacturers, and the samples from the different manufacturers can be primely classified. Then the clustering analysis model was validated by the prediction set and it indicated that this method was surely much more rapid and accuracy.3. Quantitative Identification and Qualitative Detection of Menispermi Capsule Based on Near- Infrared SpectroscopyThe total alkaloids’quantitative and qualitative analysis of Menispermi Capsule using near infrared spectroscopy were developed.85 batchs samples with three manufacturers were collected. The near-infrared spectroscopy correlated the relevant values to establish the quantitative model. The coefficient of determination of the model, the root mean squared error of corss-validation and the root mean squared error of prediction were 99.63%,1.04,4.49, respectively. The relative prediction deviation was 3.81. It indicated that this method can be applied to the production and practice. Clustering analysis combined with K-nearest neighbour method to distinguish the samples of different manufacturers, and recognition rate were 97%,100%, respectively. From this experiment, some samples can’t be distinguished in first-degree spectroscopy storeroom, and put them into the second-degree, then them can be distinguished by anther condition.4. Establishment and Application of Ginkgo biloba leaf’s quatitative and qualitative analysis Model for NIR SpectroscopyThe total flavonoids’quantitative and qualitative analysis of Ginkgo Biloba Leaf using near infrared spectroscopy were developed. Chose the spectroscopy range from 7501.7 to 4246.5 cm-1 to establish the quantitative model. The coefficient of determination of the model, and the root mean squared error of corss-validation were 98.62%,0.103. The relative prediction deviation was 3.05, and the relative standard deviation of the repeatability experiment for five samples were all below 2%. Therefore, this method can be applied to the production and practice. Two kinds of qualitative analysis were used here, and both of recognition rate were 100%.5. Qualitative analysis of pattern recognition for the plantain based on the Near-infrared Spectroscopy To distinguish the product area of plantain, the near-infrared spectroscopy combined chemometrics can be used. Chose the spectroscopy range from 8003 to 4000cm-1 to establish the cluster analysis quanlitative model. Both the classific rate and prediction rate are 100%. The classific rate of qualitative discriminant analysis was 100%. This method is quick, accurate, and can be applied to the production and practice. The point of view form this experiment is that, the pretreatment method, spectroscopy range of model and appropriate method of quantitative identification is quite exent to the qulity of model.更多还原