Near-infrared Spectroscopy On-Line and Real-Time Monitoring of Extraction Process of Xuebijing Injection

JIN Ye,DING Hai-ying,WU Yong-jiang,LIU Xue-song,CHEN Yong

DOI: https://doi.org/10.16155/j.0254-1793.2012.07.014

2012-01-01

Abstract:Objective:To establish a rapid quantitative analysis model for quality control of Xuebijing Injection extraction with application of near-infrared(NIR) spectroscopy,and quickly determine the end point of extraction process,so as to realize on-line monitoring of the extraction process of Xuebijing Injection.Methods: Taking mixed extraction process of Ligusticum chuanxiong Hort.,Salvia miltiorrhiza Bge.and Angelica sinensis(Oliv.) Diels for an example,partial least squares regression(PLSR) models were developed for the parameters of interest: total solid content and ferulic acid.The established PLS calibration models were used for on-line and real-time monitoring of the extraction process,and the moving block of standard deviation(MBSD) was used to identify the extraction end point.Furthermore,a model updating method was adopted to improve the stability and prediction accuracy of the calibration models.Results: The correlation coefficients of total solid content and ferulic acid models have reached 0.986.When the established models were applied to on-line monitoring,the correlation coefficients of prediction(r) and the residual predictive deviations(RPD) were above 0.967 and 3,respectively,and the relative standard errors of predictions(RSEP) were less than 10%.The result of MBSD was generally coincided with the variation of actual-measured ferulic acid concentration.In addition,the predictive accuracy was significantly improved after model updating with new samples.Conclusions: In this study,NIR spectroscopy was successfully applied to predicting total solid content and ferulic acid concentration in real-time and rapid determination of the extraction end point.Moreover,to solve the problem of calibration invalidation,the method of model updating was utilized to extend the use of established models and obtain better prediction results.