金银花醇沉多阶段多变量统计过程控制研究

Multi-phase and multivariate statistical process control for alcohol precipitation of water extract of Lonicerae Japonicae

目的:建立金银花醇沉加醇过程多阶段多变量统计过程控制模型,对醇沉加醇过程进行监控。方法:基于过程技术分析(PAT)理念,采用近红外技术对金银花醇沉过程进行实时测量,利用过程近红外光谱和有序样品聚类分析对加醇过程进行阶段划分,对全过程和过程的每一阶段建立多变量统计过程控制(MSPC)监控模型,分别计算Hotelling T2和SPE控制限,以实施对醇沉加醇过程的实时监控,并比较全段监控模型和分段监控模型的效果。结果:金银花醇沉加醇过程可划分为4个阶段,分段MSPC策略中Hotelling T2控制图可以更加灵敏的检测到过程微小变化,SPE控制图误报次数减少。结论:与全段MSPC模型相比,分段MSPC监控模型更加灵敏、稳健,适用于中药生产过程控制。

Objective： To establish the multi-phase and multivariate statistical process control strategy for alcohol precipitation of water extract of Lonicerae Japonicae.Methods： Based on the concept of PAT,NIR technology was applied to monitor the alcohol precipitation process in real time.Then the process was divided into several phases by sequential clustering analysis using the process NIR spectra.The MSPC models as well as the Hotelling T2 and SPE control limits were built for the whole process and each phases.The effects of the whole process MSPC model and phase based MPSC model were compared.Results： Alcohol precipitation process of water extract of Lonicerae Japonicae could be divided into four phases.In phase based MSPC,Hotelling T2 control chart could capture the subtle variation of batch process,and the false alarming in SPE control chart was reduced.Conclusion： Phase based MPSC strategy was more sensitive and robust than whole process MPSC,and was more suitable for process monitoring of Chinese medicine production.