目的 应用近红外光谱（NIRS）技术快速预测天舒片薄膜包衣终点。方法 首先，采用组合间隔偏最小二乘法（synergy interval partial least square，siPLS）优选建模区间，并对预处理方法进行筛选；其次，根据优选的建模条件，基于建立的参考光谱库，采用一致性检验和相似度匹配建立校正模型；随后，通过对比测试光谱与参考光谱间的相似度大小，设定阈值，确定相似度、阈值与包衣终点的关系；最后，采用验证集光谱对模型性能进行验证。结果 在包衣过程中，当连续多条测试光谱的相似度超过阈值时，指示为包衣终点。结论 一致性检验和相似度匹配模型性能良好，能够敏锐地反映包衣薄膜的均匀性，并准确预测包衣终点。该研究对减小包衣的批间差异性和降低原料损耗，提高中药固体制剂的包衣生产效率和提升质量均一性具有重要意义。

Objective To investigate how to rapidly predict coating terminal point in the thin film coating of Tianshu Tablets by near-infrared spectroscopy (NIRS). Methods Firstly, synergy interval partial least square (siPLS) was used to optimize the modeling intervals and the pretreatment methods were screened. Secondly, according to the optimal modeling parameters, the conformity test and similarity matching were used to establish calibration models based on the established reference spectra library. Subsequently, by comparing the similarities between the test spectra and the reference spectra, the threshold value was set. The relation among the similarities, threshold value and the coating end point was obtained. Finally, the performance of the calibration model was verified by the validation set. Results During the coating process, the similarities between the test spectra and the reference spectra were gradually increasing until the similarities of continuous test spectra exceeded the threshold, thus, indicating the end point. Conclusion Conformity test and similarity matching can sensitively monitor the change of similarities among spectra and the corresponding trend. Both models have high performance, which can accurately predict the coating terminal point. It's of great significance to reduce batch-to-batch variation in film coating and the loss of coating materials, and improve the coating efficiency and the quality uniformity of solid preparation.

国家“重大新药创制”科技重大专项“国家工信部智能制造综合标准化与新模式应用项目”（2018ZX09201010）