基于近红外光谱的红参提取过程动态预测模型研究

目的：利用近红外光谱技术，建立红参提取过程中关键组分的定量模型，实现快速检测功能；以近红外光谱为基础，结合动力学方程，建立提取过程动态趋势模型，实现全过程预测功能。方法：在线采集红参提取液近红外光谱，以HPLC获取关键成分数据，使用最小二乘法（PLSR）建立红参总皂苷的定量模型；通过定量模型以及近红外光谱，结合传质动力学方程，拟合建立提取过程随时间的动态关系模型，实现提取过程预测。结果：红参总皂苷定量模型的校正集相关系数r、校正均方根误差RMSEC、预测均方根误差RMSEP分别为0.996 09，0.018 9，0.016 8；以红参提取一阶动力学方程结合NIR定量模型建立提取过程趋势预测模型，模型显示趋势预测性能良好，具有较高的精度。结论：近红外法获得的定量模型拥有较好的检测精度，能实现快速在线检测功能；所建立的全过程提取动力学方程与实际提取过程趋势较为契合，满足预测需求。

Dynamic predictive modeling of extraction process for red ginseng using near-infrared spectroscopy

It is the objective of this study to develop dynamic predictive model for the extraction process of red Ginseng using NIR spectroscopy. NIR spectroscopy was collected online and PLSR models were developed for total quantity of ginsenosides. The performance of NIR prediction model achieved R, RMSEC, RMSEP of 0.996 09, 0.018 9, 0.016 8, respectively. A first order dynamic mass transfer model was combined with NIR prediction of the quality indicator to predict the trajectory of the extraction process based upon the initial 3 or 4 data points. The results showed good agreement with actual measurements indicating reasonable accuracy of the predictive model. It could potentially be used for advanced predictive control of the extraction process.