本文選題：傅里葉變換紅外光譜 + 三七��； 參考：《光譜學(xué)與光譜分析》2017年08期

【摘要】：不同產(chǎn)地對(duì)中藥次生代謝產(chǎn)物有顯著影響,產(chǎn)地鑒別有助于中藥的科學(xué)合理利用;其次,有效成分含量檢測(cè)是評(píng)價(jià)中藥質(zhì)量的主要手段。通過傅里葉變換紅外光譜結(jié)合化學(xué)計(jì)量學(xué)建立快速鑒別三七產(chǎn)地及測(cè)定三七中四種主要皂苷的方法,為三七的科學(xué)、合理、規(guī)范使用以及對(duì)三七質(zhì)量進(jìn)行快速評(píng)價(jià)提供依據(jù)。采集5個(gè)區(qū)域12個(gè)產(chǎn)地117個(gè)三七樣本的紅外光譜。產(chǎn)地鑒別預(yù)處理數(shù)據(jù)采用離散小波變換除去噪音造成的部分高頻信號(hào),偏最小二乘判別對(duì)產(chǎn)地判別貢獻(xiàn)率大于1的數(shù)據(jù)進(jìn)行篩選,kennard-stone算法將117個(gè)個(gè)體分為70%訓(xùn)練集與30%預(yù)測(cè)集。訓(xùn)練集數(shù)據(jù)用于建立支持向量機(jī)判別模型,交叉驗(yàn)證法用于篩選支持向量機(jī)最優(yōu)參數(shù),預(yù)測(cè)集數(shù)據(jù)對(duì)支持向量機(jī)判別模型結(jié)果進(jìn)行驗(yàn)證。皂苷含量預(yù)測(cè)預(yù)處理數(shù)據(jù)采用標(biāo)準(zhǔn)正態(tài)變量變換、離散小波變換處理;處理的紅外數(shù)據(jù)設(shè)為X變量,三七樣品中通過高效液相色譜法測(cè)得的四種皂苷總量設(shè)為Y變量,采用正交信號(hào)校正去除紅外光譜中與四種皂苷總量無(wú)關(guān)的干擾數(shù)據(jù)。個(gè)體數(shù)據(jù)分為80%訓(xùn)練集與20%預(yù)測(cè)集,訓(xùn)練集建立偏最小二乘回歸模型,預(yù)測(cè)集數(shù)據(jù)對(duì)偏最小二乘回歸模型的預(yù)測(cè)結(jié)果進(jìn)行驗(yàn)證。結(jié)果顯示:(1)交叉驗(yàn)證法得到支持向量機(jī)判別模型的最優(yōu)參數(shù)為c=2.828 43,g=0.0625,訓(xùn)練集的產(chǎn)地判別最優(yōu)正確率為91.463 4%;(2)支持向量機(jī)判別模型參數(shù)設(shè)置為最優(yōu)參數(shù),代入預(yù)測(cè)集數(shù)據(jù),預(yù)測(cè)集的產(chǎn)地判別正確率為94.285 7%,判別正確率較高;(3)訓(xùn)練集建立偏最小二乘回歸模型的相關(guān)系數(shù)R2=0.941 8,校正均方差RMSEE=4.530 7;(4)代入預(yù)測(cè)集數(shù)據(jù),預(yù)測(cè)集的相關(guān)系數(shù)R2=0.962 3,外部檢驗(yàn)均方差RMSEP=3.855 9,皂苷預(yù)測(cè)值與高效液相檢測(cè)值接近,預(yù)測(cè)效果良好。傅里葉變換紅外光譜結(jié)合支持向量機(jī)能對(duì)三七進(jìn)行產(chǎn)地鑒別,正交信號(hào)校正結(jié)合偏最小二乘回歸能對(duì)三七中四種主要皂苷總量進(jìn)行準(zhǔn)確預(yù)測(cè),為三七質(zhì)量控制提供一種快速簡(jiǎn)便、無(wú)損、高靈敏度的檢測(cè)方法。
[Abstract]:The secondary metabolites of traditional Chinese medicine were significantly affected by different production areas, and the identification of origin was helpful to the scientific and rational utilization of traditional Chinese medicine. Secondly, the detection of effective component content was the main means to evaluate the quality of traditional Chinese medicine. By means of Fourier transform infrared spectroscopy (FTIR) combined with chemometrics, a rapid method for identifying the origin of Panax notoginseng and determining four major saponins in Panax notoginseng was established, which provides the basis for the scientific, rational, standardized use and rapid evaluation of the quality of Panax notoginseng. The infrared spectra of 117 panax notoginseng samples from 12 areas in 5 regions were collected. Partial high frequency signals caused by noise are removed by discrete wavelet transform. Partial least square discriminant is used to screen the data whose contribution rate is greater than 1. 117 individuals are divided into 70% training set and 30% prediction set. The training set data is used to establish the support vector machine discriminant model, the cross-validation method is used to screen the optimal parameters of the support vector machine, and the prediction set data is used to verify the result of the support vector machine discriminant model. The pretreatment data of saponin content prediction were processed by standard normal variable transform and discrete wavelet transform, the infrared data were set as X variable, and the total amount of four saponins measured by HPLC in Panax notoginseng samples was set as Y variable. Orthogonal signal correction was used to remove the interference data from infrared spectrum independent of the total amount of four saponins. The individual data are divided into 80% training set and 20% prediction set. The partial least square regression model is established in the training set and the prediction set data is used to verify the prediction results of the partial least squares regression model. The results show that: (1) the best parameter of SVM discriminant model is 2.828 43g / g 0.0625, and the optimal correct rate of training set is 91.463 ~ 4cm; (2) the parameter of SVM discriminant model is set as the optimum parameter, and the data of prediction set is added into the model. The correct rate of the prediction set is 94.285, and the correct rate is higher. (3) the correlation coefficient R _ (2) O _ (0.941) 8 of the training set is used to establish the partial least squares regression model, and the mean square deviation (RMSEEE) is 4.537. (4) the prediction data are substituted into the prediction set, and the correlation coefficient R _ (2) O _ (0.941 8) is corrected. The correlation coefficient of the prediction set was R2N 0.962 3, and the RMSEPN 3.8555 9. The predicted value of saponins was close to the value of high performance liquid phase detection, and the prediction effect was good. Fourier transform infrared spectroscopy combined with support vector function can identify the origin of Panax notoginseng. Orthogonal signal correction combined with partial least square regression can accurately predict the total amount of four main saponins in Panax notoginseng. It provides a rapid, simple, nondestructive and sensitive method for quality control of Panax notoginseng.
【作者單位】： 云南中醫(yī)學(xué)院中藥學(xué)院;云南省農(nóng)業(yè)科學(xué)院藥用植物研究所;云南省省級(jí)中藥原料質(zhì)量監(jiān)測(cè)技術(shù)服務(wù)中心;
【基金】：國(guó)家自然科學(xué)基金項(xiàng)目(81460581,81260610)資助
【分類號(hào)】：O657.33;R284.1
，

本文編號(hào)：2107331