ABSTRACT
Objective To develop recovery prediction models for danshensu,salvianolic acid B,and salvianolic acid D in danshen ethanol precipitation.Methods Recorded the operation parameters of 15 normal batches of danshen ethanol precipitation in one Chinese medicine plant and measured five active component contents in danshen concentrated liquid and danshen ethanol precipitated liquid,the Stepwise-MLR recovery prediction models of dashensu,salvianolic acid B,and salvianolic acid D were developed based on an approach which combined the process parameters and the active component contents in danshen concentrated liquid.The importance of variables in the prediction models was analyzed later.Results The coefficient correlations of dashensu,salvianolic acid B,and salvianolic acid D recovery prediction models were all above 0.95.Conclusion The models developed for active component recovery prediction show a good predictability.Models can be used to help manufacturers to understand the ethanol precipitation process and improve the quality control ability of danshen ethanol precipitation.