近红外光谱技术在金线莲定性定量分析中的应用研究

发布时间：2018-05-22 10:29

本文选题：金线莲 + 近红外光谱　；参考：《福建中医药大学》2017年硕士论文

【摘要】：金线莲Anoectochilus roxburghii为兰科花叶开唇兰属的一种多年生珍贵中草药。目前市场上以栽培品为主,品种繁多,来源复杂,以次充好,缺乏科学的评价标准,掺伪掺假严重。近红外光谱技术具有无损、快速、绿色特点,本研究借鉴其在定性定量分析的优势,收集栽培品金线莲(尖叶、大圆叶、小圆叶)、台湾金线莲、红霞,野生品斑叶兰、血叶兰等样品,构建NIRS模型实现金线莲质量快速检测的目的,以期为日后品质评价研究提供科学依据。具体研究内容如下:1、基于NIR-PCA技术构建聚类分析模型快速区分金线莲及其伪品斑叶兰和血叶兰。以Q值、Cluster per Property、Spectra Residuals too big和Property Residuum为模型评价指标,选取4400-4800、5400-6600、7800-10000cm-1为建模波段,nle和dg1光谱预处理方式,建立金线莲及其伪品的定性判别模型,准确度为98%,模型预测性良好。2、基于NIR-PCA技术构建聚类分析模型快速区分不同栽培方式(3个月瓶苗、4个月瓶苗及入土种植6个月的种植苗)的金线莲,以Q值、Cluster per Property、Spectra Residuals too big和Property Residuum为模型评价指标,选取5000-1OOOOcm-1 为建模波段,nle和dg1光谱预处理方式,建立不同栽培方式金线莲的定性判别模型,准确度为80%。3、基于NIR-PCA技术构建聚类分析模型快速区分不同栽培方式的尖叶、大圆叶、红霞、及台湾金线莲,包括3个月瓶苗、4个月瓶苗及入土种植6个月的种植苗,以Q值、Cluster per Property、Spectra Residuals too big和Property Residuum为模型评价指标,4个模型Q值均大于0.9,准确度达100%,各模型预测性良好。4、采用UPLC法检测金线莲中黄酮类成分及含量,并基于NIR-PLS技术构建金线莲黄酮类成分定量分析模型。4.1建立UPLC法同时检测113批金线莲中水仙苷、芦丁、槲皮素、山奈酚和异鼠李素5种黄酮类成分。5种黄酮类成分在考察的浓度范围内呈较好的线性关系(R20.9999),平均加样回收率和RSD分别在97.06%-106.17%和3.17%-6.24%内。113批检测样品中水仙苷、芦丁、槲皮素、山奈酚和异鼠李素的含量范围分别为217.17-2243.26ug.g-1、15.95-1414.26ug.g-1、11.87-267.07ug.g-1、15.58-165.21ug.g-1、13.54-86.4ug.g-1。4.2基于NIR-PLS技术,Sa3+SNV光谱预处理方法,以全光谱为分析波段,构建金线莲水仙苷成分定量模型,其中SEC、SEP、Rc、Rv及Q值分别为0.134、0.147、0.974、0.969、0.792,定标模型的线性和相关系数分别为f(X)=0.9478X+0.0551,r=0.9736,并预测10批样品含量,平均相对偏差为6.09%,表明模型具有较好的预测性能。4.3基于NIR-PLS技术,SNV+MSC光谱预处理方法,以全光谱为分析波段,构建金线莲芦丁成分定量模型,其中SEC、SEP、Rc、Rv及Q值分别为0.024、0.022、0.978、0.969、0.866,定标模型的线性和相关系数分别为f(X)=0.9569X+0.0062,r=0.9782,并预测10批样品含量,平均相对偏差为4.50%。4.4基于NIR-PLS技术,MSC+Sa3光谱预处理方法,以全光谱为分析波段,构建金线莲槲皮素成分定量模型,其中SEC、SEP、Rc、Rv及Q值分别为15.508、16.732、0.962、0.950、0.430,定标模型的线性和相关系数分别为f(X)=0.9263X+7.8829,r=0.9625,并预测10批样品含量,平均相对偏差为8.15%。
[Abstract]:Anoectochilus roxburghii is a precious perennial herb of the genus Cymbidium of Orchidaceae. At present, the market is mainly cultivated in the market. There are many varieties, complex source, good inferior, lack of scientific evaluation standard and serious adulteration. The near infrared spectroscopy has the characteristics of non loss, fast and green. This study is used for reference to the qualitative determination. The advantage of quantitative analysis is to collect the samples of cultivated varieties of golden thread Lotus (apex leaf, large round leaf, small round leaf), Taiwan gold wire lotus, Hongxia, wild leaf and leaf blue, blood leaf blue and so on, to construct NIRS model to realize the rapid test of quality of the quality of gold thread, in order to provide scientific basis for the study of quality evaluation in the future. The specific contents are as follows: 1, based on the technology of NIR-PCA construction. Q, Cluster per Property, Spectra Residuals too big and Property Residuum are the model evaluation indexes, and 4400-48005400-66007800-10000cm-1 as the modeling band, NLE and spectral preprocessing formula are selected to establish the qualitative discriminant model of the gold thread lotus and its artifact. Type, the accuracy is 98%, the model predictability is good.2. Based on the NIR-PCA technology, the cluster analysis model is constructed to distinguish different cultivation methods (3 months bottle, 4 months bottle and 6 month planting seedlings), with Q value, Cluster per Property, Spectra Residuals too big and Property Residuum as the model evaluation index, select 5000 -1OOOOcm-1 for the modeling band, NLE and DG1 spectrum pretreatment methods, the qualitative discriminant model of different cultivation methods was established, the accuracy was 80%.3. Based on NIR-PCA technology, the cluster analysis model was constructed to quickly distinguish the tip leaves of different cultivation methods, big round leaves, Hongxia, and Taiwan Golden Lotus, including 3 months bottle seedlings, 4 months bottle seedlings and 6 soil planting. Q value, Cluster per Property, Spectra Residuals too big and Property Residuum are the model evaluation indexes for the planting seedlings of the month. The 4 model Q values are both greater than 0.9, the accuracy is 100%, the models are well predicate.4. The quantitative analysis model.4.1 established the UPLC method for the simultaneous determination of the 5 flavonoids of the 113 batches of Narcissus, rutin, quercetin, ananolol and isomanrin, which had a better linear relationship (R20.9999) in the concentration range (R20.9999). The average recovery rate and RSD were detected in.113 batch in 97.06%-106.17% and 3.17%-6.24%, respectively. The content range of narcissuin, rutin, quercetin, kaempferol and isosamin is 217.17-2243.26ug.g-1,15.95-1414.26ug.g-1,11.87-267.07ug.g-1,15.58-165.21ug.g-1,13.54-86.4ug.g-1.4.2 based on NIR-PLS technology and Sa3+SNV spectrum pretreatment method. The quantitative model of Narcissus is constructed with full spectrum as the analytical band. The SEC, SEP, Rc, Rv and Q values are 0.134,0.147,0.974,0.969,0.792 respectively. The linear and correlation coefficients of the calibration model are f (X) =0.9478X+0.0551, r=0.9736, respectively, and the average relative deviation is 6.09%. It shows that the model has good predictive performance based on NIR-PLS technology, spectral preprocessing method, with full spectrum. The quantitative model of rutin was constructed for the analysis band, in which the values of SEC, SEP, Rc, Rv and Q were 0.024,0.022,0.978,0.969,0.866 respectively. The linear and correlation coefficients of the calibration model were f (X) =0.9569X+0.0062, r=0.9782, and the 10 batch of samples were predicted. Ping Junxiang's deviation was 4.50%.4.4 based on NIR-PLS technology. The quantitative model of quercetin was constructed with full spectrum, in which the values of SEC, SEP, Rc, Rv and Q were 15.508,16.732,0.962,0.950,0.430 respectively. The linear and correlation coefficients of the calibration model were f (X) =0.9263X+7.8829 and r=0.9625 respectively, and the content of 10 samples was predicted, and the average relative deviation was 8.15%..
【学位授予单位】：福建中医药大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R284.1;O657.33

【相似文献】