Bagging-PLS的黄柏中试提取过程在线近红外质量监测研究

Study on Online NIR for Monitoring the Extraction Process of Cortex Phellodendri in Pilot Scale Based on Bagging-PLS

目的:采用近红外光谱技术,建立黄柏中试提取过程偏最小二乘法(PLS)定量模型,快速分析盐酸小檗碱含量,以Bagging-PLS集成建模方法提高所建模型的预测性能。方法:以HPLC为参考方法测定黄柏提取液中盐酸小檗碱含量,运用PLS建立与近红外光谱(NIR)预测值之间的多元校正模型,并建立Bagging-PLS模型提高模型的稳定性与预测性能。结果:盐酸小檗碱定量模型的校正均方根误差RMSEC为0.742 9,预测均方根误差RMSEP为0.961 4,校正集相关系数rc为0.986 6,预测集相关系数rp为0.977 0;Bagging-PLS模型预测均方根误差RMSEP均值为0.413 6,预测集相关系数rp均值为0.998 3,表现出了良好的稳健性与预测性能。结论:所建立的近红外定量模型拥有较好的预测性能,能够实现快速实时检测黄柏提取液中盐酸小檗碱的含量;所建立的Bagging-PLS模型稳健性与预测性能良好,可应用于中药近红外在线监测过程中。

Objective： To establish a rapid quantitative analysis method for the quality control of Cortex Phellodendri extraction by using near-infrared（ NIR） spectroscopy and improve the prediction performance and robustness of the model by using Bagging partial least square regression（ Bagging-PLS）. Methods： NIR spectroscopy was collected online,and a partial least square（ PLS）model was developed for determination the content of berberine hydrochlorid with HPLC method as referential method. Besides,bagging algorithm was adopted for improving the prediction performance and robustness of the model. Results： The RMSEC and RMSEP were 0. 7429 and 0. 9614,the correlation coefficient of the calibration model was 0. 9866,and the correlation coefficient of the prediction was 0. 9770. The mean PMSEP of the Bagging-PLS model was 0. 4136 and the correlation coefficient of the prediction was 0. 9983. Conclusion： In this study,NIR spectroscopy was successfully applied to Cortex Phellodendri extraction process,which can rapidly determine the content of berberine hydrochloride. In addition,the Baggting-PLS provided a better performance and prediction accuracy than single PLS model. This indicates that Baggting-PLS could potentially be used for a better predictive control of the extraction process.