目的 采用近红外光谱（near infrared spectrum，NIRS）技术结合偏最小二乘法（partial least-square，PLS）建立香菊片提取液膜分离过程中没食子酸、鞣花酸、木犀草素3个成分的定量监测模型。方法 收集香菊片膜分离过滤液样本53个，其中预测集38个，验证集15个，用HPLC法测定所有样本中没食子酸、鞣花酸、木犀草素3个成分的含量，同时采集NIRS数据。将得到的光谱数据与3个化学成分含量数据应用PLS回归分析建立定量模型，采用模型的校正集相关系数（correlation coefficient of calibration，R_C）、预测集相关系数（correlation coefficient of prediction，R_P）、校正集误差均方根（root mean square error of calibration，RMSEC）、预测集误差均方根（root mean square error of prediction，RMSEP）和预测集相对偏差（relative standard of errors of prediction，RSEP）对定量预测模型进行评价。结果 没食子酸、鞣花酸、木犀草素近红外定量模型RMSEC分别为0.561、1.256、0.342，R_C分别为0.981、0.992、0.986；RMSEP分别为0.557、1.157、0.367，R_P分别为0.987、0.994、0.979；RSEP分别为5.73%、4.23%和3.78%，均小于10%。结论 建立的NIRS定量模型预测性良好，可用于香菊片提取液膜分离过程的成分含量测定和终点判断。

Objective The quantitative monitoring model of gallic acid, ellagic acid and luteolin in the membrane separation process of Xiangju Tablets extract was established by near infrared spectrum (NIRS) combined with partial least-square (PLS). Methods A total of 53 samples of filtrate separated by Xiangju Tablets membrane were collected, including 38 prediction sets and 15 validation sets. The contents of gallic acid, ellagic acid and luteolin in all samples were determined by HPLC, and the near-infrared spectral data were collected at the same time. The obtained spectral data and the content data of three chemical components were used to establish a quantitative model by PLS regression analysis. Correlation coefficient of calibration (R_C), correlation coefficient of prediction (R_P), root mean square error of calibration (RMSEC), root mean square error of prediction (RMSEP), and relative standard of errors of prediction (RSEP) were used to evaluate the quantitative prediction model. Results The RMSEC of gallic acid, ellagic acid and luteolin was 0.561, 1.256 and 0.342, respectively, and the R_C was 0.981, 0.992 and 0.986, respectively. The RMSEP was 0.557, 1.157 and 0.367, respectively, and the R_P was 0.987, 0.994 and 0.979, respectively. The RSEP was 5.73%, 4.23% and 3.78%, respectively, which were all less than 10%. Conclusion The established near-infrared quantitative model has good predictability and can be used for the determination of component content and terminus judgment in the membrane separation process of Xiangju Tablets extract.

R283.6

陕西省中医药管理局（2021-04-ZZ-007）；陕西省教育厅项目（21JC011）；国家重点研发计划（2019YFC1711204）