基于近红外光谱技术的甘草提取过程最优建模方法研究

发布时间：2018-11-03 19:01

【摘要】：通过实时采集正常操作条件下及发生异常工况时的甘草提取液的动态近红外光谱数据,结合主成分分析法(PCA)、偏最小二乘回归法(PLSR)和平行因子-偏最小二乘回归联用法(PARAFAC-PLSR)建立3种甘草提取过程的实时监测模型,并分析各种模型的特点。结果表明,基于3种方法建立的模型均能在一定程度上预测异常加热工况的发生,但同时也存在一定误判。其中,PCA方法建立的模型出错率最高,在60 min之前就出现3次"故障"误判,不适用于该过程的分析应用。而PLSR和PARAFAC-PLSR模型基本效果相似,校正集相关系数分别高达0.934 2,0.928 1,验证集相关系数也分别达到了0.856 7,0.828 3;并且这2种方法建立的预测模型误判率较低,首次成功预测的故障均发生于75 min。此外,PLSR和PARAFAC-PLSR模型均能在一定程度上预测出系统状态的走势。说明基于动态近红外光谱动态数据建立的PLSR和PARAFAC-PLSR模型均具有良好的在线监测和预测功能,为中药提取过程动态监测方法的优化选择提供了参考依据。
[Abstract]:The dynamic near infrared spectrum data of Glycyrrhiza uralensis extract under normal operating conditions and abnormal operating conditions were collected in real time, combined with principal component analysis (PCA),). Partial least square regression (PLSR) and parallel factor-partial least squares regression (PARAFAC-PLSR) were used to establish real-time monitoring models for the extraction process of Glycyrrhiza uralensis, and the characteristics of the models were analyzed. The results show that the models based on the three methods can predict the abnormal heating condition to a certain extent, but there are also some misjudgments at the same time. Among them, the PCA method has the highest error rate and "fault" misjudgment occurs three times before 60 min, which is not suitable for the analysis and application of this process. The basic effect of PLSR and PARAFAC-PLSR models is similar, the correlation coefficient of correction set is as high as 0.934 ~ (22) 928 1, and the correlation coefficient of verification set is 0.856 ~ 7 ~ 0.828 ~ 3 respectively. The error rate of the two prediction models was low, and the first successful prediction occurred at 75 min.. In addition, both PLSR and PARAFAC-PLSR models can predict the trend of system state to some extent. It is concluded that both PLSR and PARAFAC-PLSR models based on dynamic near infrared spectrum data have good on-line monitoring and prediction functions, which provides a reference for the optimization and selection of dynamic monitoring methods in the process of traditional Chinese medicine extraction.
【作者单位】：天津中医药大学中药制药工程学院;天津中医药大学现代中药产业技术研究院;天津市现代中药省部共建国家重点实验室(培育);
【基金】：天津市科技创新体系及条件平台建设计划项目(14TXZYJC00440);天津市科技创新体系及条件平台建设计划项目(15PTCYSY00030)
【分类号】：R284.2;O657.33

【相似文献】