Application Of Feedforward Control Strategy Based On Spectra Of Raw Materials In Manufacturing Processes Of Traditional Chinese Medicines

In the production of the traditional Chinese medicines,how to reduce the fluctuation caused by raw material variation to the product quality is a key issue to be solved in the quality control process.There has been a lack of comprehensive and efficient acquisition methods for the unadjustable critical material parameters.Whereas we can quickly and comprehensively obtain raw materials’ quality information using spectroscopic techniques.With acknowledging quality variation of batches,quantitative models of important operating units are established to realize the feedforward control of the process,which can guide the adjustment of process parameters,reduce the fluctuation of product quality caused by raw material variation,thus improve the quality consistency.This thesis obtained raw materials’ quality information with spectroscopic techniques,taking the manufacturing processes of Panax notoginseng and Kushen as examples,and conducted researches on the application of feedforward control in manufacturing processes of traditional Chinese medicines.The main contents and achievements of this thesis are summarized as follows:1.Feedforward control strategy was applied in the alcohol extraction process of Panax notoginseng with near infrared spectroscopy.The principal component scores of near infrared spectra were used to represent the raw materials’ quality information.Quantitative calibration models of Critical Quality Attributes(CQA)were established including raw material variation and Critical Process Parameters(CPPs),respectively with Least Squares-Principal Component Analysis(LS-PCA)method and Partial Least-Squares Regression(PLSR)method.The results of the models were significant and could predict the alcohol extraction process properly.With new batch of the raw material introduced,the adjustment of the process parameters was guided and the optimal process parameters were calculated based on the LS-PCA models.Thereby the impact of raw material variation on product quality was reduced and the feedforward control of the extraction process was realized.2.The feedforward control strategy was applied with near infrared spectroscopy,taking the percolation process of Kushen-rhizoma heterosmilacis japonicae as an example.The quality information of the samples was represented by the principal component scores of near infrared spectra,then raw material variation and CPP were involved in Box-Behnken Design.The quantitative calibration models of the CQA in the percolation process were established.With new batch of the raw material introduced,the optimal process parameters were calculated based on quantitative calibration models.Thus,the CQA of the percolation process were adjusted to meet the target values so that the feedforward control could be realized.3.The feedforward control strategy was used to study the second alcohol precipitation process of Kushen.The effects of the primary alcohol sediment quality variation and CPP on the CQA of the second alcohol precipitation process were investigated.The principal component scores were used to represent the quality information of the primary alcohol solution.The development of quantitative calibration models was to predict the second alcohol precipitation process.The model obtained was significant and good prediction of the second alcohol precipitation process was realized.Based on the quantitative calibration models,new batch of the raw material was introduced,and then the adjustment of the process parameters was guided.Thereby the product quality fluctuations caused by raw material variation were reduced and the feedforward control of the second alcohol precipitation process was realized.4.Taking the percolation process of Kushen-rhizoma heterosmilacis japonicae as an example,the percolation process was monitored by online ultraviolet(UV)spectroscopy.The quantitative models of the CQA in the percolation process were established.The results of the models were significant with good robustness,and the predictive performance was also good,leading to accurate prediction of the penetrating endpoint,so as to provide a guide method for specified judgement of the endpoint.