目的 建立热毒宁注射液青蒿金银花醇沉浓缩过程主要药效成分的定量校正模型，实现生产过程在线监控。方法 采用近红外光谱技术（NIRS）结合偏最小二乘法（PLSR）分别建立新绿原酸、绿原酸、隐绿原酸的定量校正模型。结果 新绿原酸、绿原酸、隐绿原酸定量校正模型的决定系数（R²）分别为0.954 5、0.975 2、0.969 1；校正集误差均方根（RMSEC）为0.213、0.676、0.225，交叉验证集误差均方根（RMSECV）分别为0.233、0.692、0.258。采用所建模型进行在线分析，新绿原酸、绿原酸、隐绿原酸的预测值与实测值的决定系数分别为0.984 2、0.983 7、0.987 0，预测相对误差（RPD）分别为4.77、5.29、4.37，预测相对偏差（RSEP）分别为3.519%、3.778%、3.895%。结论 所建的模型可以用于热毒宁注射液青蒿金银花醇沉浓缩过程中新绿原酸、绿原酸、隐绿原酸的在线定量测定。

Objective The calibration models were developed in the concentration of alcohol precipitation proee for Artemisiae Annuae Herba (AAH) and Lonicerae Flos (LF) in Reduning Injection (RI) to realize the on-line monitoring of production process. Methods Based on the near infrared reflectance spectroscopy (NIRS), partial least regression (PLS) models were developed to fast measure the contents of neochlorogenic acid, chlorogenic acid, and 4-O-caffeoylquinic acid in the concentration of the alcohol precipitation proee for AAH and LF. Results In the quantitative models of neochlorogenic acid, chlorogenic acid, and 4-O-caffeoylquinic acid, the coefficient of determination (R²) of cross validation sets were 0.954 5, 0.975 2, and 0.969 1; The root mean square errors of calibration (RMSEC) were 0.213, 0.676, and 0.225; The root mean square errors of cross-validation (RMSECV) were 0.233, 0.692, and 0.258. When the established models were applied to on-line monitoring, the coefficient of determination of neochlorogenic acid, chlorogenic acid, and 4-O-caffeoylquinic acid were 0.984 2, 0.983 7, and 0.987 0, the residual predictive deviation (RPD) were 4.77, 5.29, and 4.37; The relative standard errors of prediction (RSEP) were 3.519%, 3.778%, and 3.895%. Conclusion The models above are proved to fast measure the contents of neochlorogenic acid, chlorogenic acid, and 4-O-caffeoylquinic acid in the concentration of alcohol precipitation proee for AAH and LF in RI.

国家科技部“国家重大新药创制”项目：现代中药创新集群与数字制药技术平台（2013ZX09402203）