近红外光谱（NIRS）广泛应用于生产过程分析与监测, 常需事先建立定量校正或定性判别模型, 并需在生产条件变化后调整模型, 使用较复杂。 本文从相异度和相似度两个对立互补的角度, 提出自适应移动窗口标准差法和过程光谱相似度法, 并以此为基础建立一种针对生产过程的无需校正模型的简易光谱在线监测方法。 论文以中药柱层析过程为例, 对监测过程作NIRS自适应移动窗口标准差趋势图和过程光谱相似度趋势图, 并通过HPLC离线分析所得的多指标成分含量变化趋势图进行对比验证, 发现可用于工艺状况实时监测, 指导收集起点、 终点、 溶液相变点的判断, 表明论文提出的方法合理可行。 该方法亦可用于紫外/可见、 红外、 拉曼、 荧光等光谱及色谱、 质谱等其他过程分析技术。

Near infrared spectroscopy(NIRS) is a process analysis and monitoring tool with many advantages, while it needs to set up quantitative or discriminative calibration models in advance, and needs to adjust these models when the process conditions are varied, which makes it difficult for ordinary user to take its full advantage of it. To tackle this problem, this paper presented a novel, simple and model-free methodology for online process monitoring based on two reciprocal viewpoints of measuring the variability of spectroscopy-both the similarity and dissimilarity of process spectrum, i.e., the adaptive moving window standard deviation function(AMWSW) and similarity function(S). The methodology was validated by a column chromatography process of traditional Chinese medicine using near infrared spectroscopy. The online trend curves of AMWSW and S obtained by proposed method were validated by a comparison with the content variation curves of multiple indicative components analyzed by high performance liquid chromatography (HPLC), and these trend curves demonstrated their potential for real-time process status monitoring, accurately determining the beginning point, the peak point, the end point of the elution, and the phase change from water solution to ethanol solution. The proposed methodology can also be used to other process analysis techniques, such as ultraviolet/visible, infrared, Raman, fluorescence, chromatograph and mass spectrum.