近红外光谱运用于中成药口服液多糖的快速分析

发布时间：2018-05-22 10:58

本文选题：中成药口服液 + 多糖　；参考：《暨南大学》2017年硕士论文

【摘要】：中成药口服液(PCM-OS)是一类以中药材为主要原料,兼具治疗与保健功能的中药制剂。在生产过程中,为了确保口服液的品质,需要对主要活性成份进行测定,但既有的测定方法需要消耗试剂,操作繁琐,不便于生产过程中的快速测定分析。近红外光谱作为一种快速简便,无需试剂的分析技术,已经应用于多个领域。本文以一个中成药口服液产品(活性成份为多糖)为例,采用近红外光谱结合化学计量学,研究建立PCM-OS中多糖的快速分析方法。收集到某品牌的1533份PCM-OS样品,使用XDS Rapid Content TM型近红外光栅光谱分析仪测定其光谱值(扫描范围400-2498 nm,波长间隔为2 nm)。此外,采用高锰酸钾滴定法测定多糖含量,作为NIR建模、检验的参考值。随机选择693样品用于检验,余下840样品用于建模,进一步随机划分为定标(420份)和预测(420份)共30次。对所有划分建立定标预测模型,依据所有划分的平均预测效果优选参数,使得模型具有稳定性。为了消除光谱基线漂移和倾斜,首先采用Savitzky-Golay(SG)校正方法,对全谱建立偏最小二乘(full-PLS)模型,依据预测效果优选出最佳SG校正模式为导数阶数d=1,多项式次数p=2,平滑点数m=39。此外,为了避开水分子导致的近红外光谱的饱和吸收,再采用吸光度上界优化偏最小二乘(AUO-PLS)方法选出适当的吸光度上界,同时确定波段组合为400~(-1)8802088-2346 nm。为了进一步提取信息,提升模型预测性能,基于400~(-1)8802088-2346 nm组合波段的SG校正光谱,分别采用移动窗口(MW),等间隔组合(EC)和重复率优先组合(RRPC)等波长筛选方法建立SG-AUO-MW-PLS,SG-AUO-EC-PLS和SG-AUO-RRPC-PLS模型。NIR预测值与标准方法实测值的误差(RMSEPV)分别为28.18、27.40、27.19(mg·L~(-1));预测相关系数(RP,V)分别为0.864、0.875、0.887。SG-AUO-RRPC-PLS方法达到最好的预测效果,且使用的波长数最少(N=94)。结果表明,近红外分析技术可以运用于PCM-OS多糖的快速定量分析,对于其质量监控具有重要意义。所提出的PLS结合SG校正、饱和吸收剔除(AUO)、波长筛选(MW、EC、RRPC)的集成优化方法,可以有效提升光谱分析的精度。为近红外技术的深入应用提供了有价值的参考。
[Abstract]:Chinese patent medicine oral liquid (PCM-OS) is a kind of traditional Chinese medicine preparations with Chinese medicinal materials as main raw materials and both therapeutic and health care functions. In order to ensure the quality of oral liquid, it is necessary to determine the main active components in the production process, but the existing determination methods need to consume reagents, and the operation is cumbersome, which is not convenient for rapid determination and analysis in the production process. Near-infrared spectroscopy (NIR), as a rapid and simple analytical technique without reagent, has been applied in many fields. A rapid method for the analysis of polysaccharides in PCM-OS was established by using near infrared spectroscopy (NIR) combined with chemometrics, taking a Chinese patent medicine oral liquid product (active ingredient as polysaccharide) as an example. A brand of 1533 PCM-OS samples were collected and their spectral values were measured by XDS Rapid Content TM near infrared grating spectrometer (scanning range 400-2498 nm, wavelength interval 2 nm). In addition, the content of polysaccharide was determined by potassium permanganate titration, which was used as the reference value for NIR modeling. 693 samples were randomly selected for test, and the remaining 840 samples were used for modeling, further divided randomly into 420 calibration samples and 420 predicted samples for 30 times. The calibration prediction model is established for all partitions, and the model is stable according to the optimal parameters of the average prediction effect of all partitions. In order to eliminate the drift and tilt of spectral baseline, Savitzky-Golay-SGG) correction method is used to establish a partial least square full-PLS model for the whole spectrum. According to the prediction results, the optimal SG correction model is selected as derivative order dn1, polynomial degree p2, and smooth number mN39. In addition, in order to avoid the saturated absorption of the near infrared spectrum caused by water molecules, the optimum partial least square method is used to select the appropriate absorbance upper bound, and the band combination is determined to be 400 ~ (1) ~ (-) 8802088 ~ (8-2346) nm. in order to avoid the saturated absorption of the near infrared spectrum caused by water molecules. In order to extract further information and improve the prediction performance of the model, the SG correction spectrum based on the combined wavelength band of 400,0002088-2346 nm is used to improve the performance of the model. The error of SG-AUO-MW-PLSU SG-AUO-EC-PLS and SG-AUO-RRPC-PLS model. NIR was 28.1827.4027.4027.19mg / L, respectively, and the correlation coefficient of prediction was 0.8640.875U 0.887.SG-RPC-PLS, respectively, by using the wavelength screening methods such as moving window MMW, equal interval combination (ECC) and repetition rate priority combination (RRPC-RPC-PLS). The RPV was 0.8640.875U 0.887.SG-RPC-PLS respectively. The error between the predicted values of SG-AUO-MW-PLSU and the measured values by the standard method was 28.1827.4027.4027.19mg / L; and the correlation coefficient of prediction was 0.8640.875U 0.887.SG-RPC-PLS square, respectively. Method to achieve the best prediction results, And the number of wavelengths used is the least. The results show that the near infrared analysis technique can be applied to the rapid quantitative analysis of PCM-OS polysaccharides and has important significance for its quality control. The proposed integrated optimization method of PLS combined with SG correction, saturation absorption elimination and wavelength screening can effectively improve the accuracy of spectral analysis. It provides a valuable reference for the further application of near infrared technology.
【学位授予单位】：暨南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O657.33;R286.0

【相似文献】