用近红外光谱法实现嫩, 中, 老三种炼蜜含水量的快速测定及不同炼制程度蜂蜜的快速分级。 过程化控制是中药质量稳定的关键因素, 但由于检测手段的滞后而无法实现, 因而影响中药质量。 蜜丸是具广阔应用市场的剂型; 炼蜜工艺是影响蜜丸质量的关键工序; 在炼蜜过程中, 蜂蜜从嫩蜜炼制到中蜜仅需几至几十秒, 变化迅速, 传统检测手段不能对蜂蜜炼制过程进行实时监测, 进而无法保证炼蜜质量的均一性。 力求采用近红外光谱法(NIRS)对炼蜜过程进行实时检测, 准确控制嫩蜜、 中蜜的炼制程度, 并对其含水量进行定量测定。 实验通过Rudolph折光仪测定不同炼蜜的含水量, 并快速测定样品的近红外光谱, 结合偏最小二乘法(PLS)建立并优化88批炼蜜水分的定量模型, 利用校正模型决定系数( R²)、 校正模型均方根误差(RMSEE); 内部交叉验证决定系数( R²), 内部交叉验证均方根误差(RMSECV); 预测模型的相关系数( r)、 预测均方根误差(RMSEP)对校正模型进行评价, 其中RMSECV主要用来筛选建立模型的最佳主成分数, RMSEE和RMSEP分别对所建立校正集和验证集进行误差分析。 同时采用鉴别分析法建立定性模型用来区分炼蜜等级。 经不断优化, 图谱经矢量归一化法(SNV)预处理后, 在7 201.25 446.2 cm^-1波段内选取9个主成分建立水分模型。 建立的最佳模型中, 校正集 R², RMSEE分别为99.43, 0.299, 内部交叉验证 R², RMSECV分别为99.05, 0.348; 预测集 R², RMSEP分别为98.19, 0.347; 定量模型显示, NIRS可快速、 准确、 无损的对炼蜜含水量进行测定。 同时按照含水量测定结果对炼蜜嫩、 中、 老3个等级进行划分, 进而采用鉴别分析法对炼蜜进行定性分析, 结果显示嫩、 中、 老3种炼蜜有明显的聚类现象, 表明因子化法是鉴别炼蜜程度的一种有效方法; 综上所述, NIRS可望实现对炼蜜过程实时监测, 准确控制嫩蜜, 中蜜的炼蜜程度并对炼蜜含水量进行快速实时定量测定, 为不同炼蜜等级划分提供参考, 保证炼蜜质量的均一性, 最终保证蜜丸质量的稳定性。

Near-infrared spectroscopy(NIRS) was used to quantify moisture content about refined-honey, and the practicability to classify the samples degree through refining was investigated. Process Analysis Technology is vital to affecting the quality stability of traditional Chinese medicines, with the lagging nature of detection methods, accurate real-time control of them cannot be achieved, ultimately affecting the homogeneity and stability of quality about traditional Chinese medicines. Sweetpill has a wide application market. The honey refining process is a critical process to affect the quality of it. However, the transition from tender to medium honey takes only a few to several tens of seconds. The traditional detection method cannot monitor it, leading to the fact that the uniformity of honey quality cannot be achieved. In order to identify and predict moisture content for tender, medium and old honey at the same time, the 88 different honeies were refined, and the qualitative and quantitative models were established based on Near Infrared Spectroscopy (NIRS). Using refractometer to measure the moisture content, and near infrared spectra of samples were also measured. Combined with partial least squares(PLS) algorithm were conducted on the calibration of regression model; the qualitative prediction model was used to establish the level of the refined honey by the factorization method. For the quantitative calibration model, the R², RMSEE, RMSECV were 99.43, 0.299, 0.34, respectively. For the prediction model, the R², RMSEP were 98.19, 0.347, respectively. Through the discriminant analysis, the qualitative model showed obvious clustering phenomenon. NIRS has good application prospects to be applied to real-time monitoring and moisture determination of refined honey.