目的 建立一种基于近红外光谱（near infrared spectroscopy，NIRS）快速检测维C银翘片（Vitamin C Yinqiao Tablets，VCYT）引湿率与包衣增重的方法，用于包衣过程的实时检测。方法 参考《中国药典》2020年版测量引湿率，将其作为包衣防潮效果的指标。通过在线NIRS技术实时采集包衣过程中的片剂光谱，比较了区间偏最小二乘法（interval partial least squares，iPLS）、随机蛙跳（rand frog，RF）、蒙特卡罗-无信息变量消除（Monte Carlo-uninformative variable elimination，MC-UVE）、竞争自适应重采样法（competitive adaptive reweighted sampling，CARS）4种波长选择算法对偏最小二乘回归（partial least squares regression，PLSR）的影响。结果 所得定量模型可准确、稳定地预测引湿率与包衣增重，其决定系数（Rp2）分别为0.891 8和0.939 6，预测均方根误差（root mean square error of prediction，RMSEP）分别为0.175 4和0.274 2，相对分析误差（relative prediction errors，RPD）分别为3.121 9和4.148 4。另外，在1批包衣过程实时检测的结果显示，预测模型对引湿率的预测效果良好（RPD=4.199 1），但对包衣增重的预测效果较差（RPD=1.815 2）。结论 NIRS对维C银翘片包衣过程的实时检测是可行的。
Objective A method based on near-infrared spectroscopy (NIRS) was developed for rapid detection of moisture absorption rate and coating weight gain for Vitamin C Yinqiao Tablets (维C银翘片, VCYT) during the coating process, and was used for real-time monitoring of the coating process. Methods The moisture absorption rate was determined according to Chinese Pharmacopeia and used as an indicator for the moisture-proof effect of a coating. Real-time spectral data was collected using online NIRS technology during the coating process. Four wavelength selection algorithms, specifically interval partial least squares (iPLS), random frog (RF), Monte Carlo-uninformative variable elimination (MC-UVE) and competitive adaptive reweighted sampling (CARS), were tested to determine their impact on partial least squares regression (PLSR). Results The developed quantitative model can predict moisture absorption rate and coating weight gain accurately and precisely during the coating process. The model exhibited determination coefficients (Rp2) of 0.891 8 and 0.939 6 and prediction root mean square error values (RMSEP) of 0.175 4 and 0.274 2 for moisture absorption rate and coating weight gain, respectively. Moreover, the relative prediction errors (RPD) for moisture absorption rate and coating weight gain were 3.121 9 and 4.148 4, respectively. The real-time detection results of the developed model for a batch of products during the coating process confirmed the precision and accuracy of the model's capability to predict moisture absorption, with RPD value of 4.199 1, while, for weight gain, the model's prediction capability was less precise, with RPD value of 1.815 2. Conclusion The study confirms the feasibility of NIRS as a real-time monitoring tool for the coating process of the VCYT.