Studies On Key Techniques Of Near Infrared Spectroscopy In Traditional Chinese Medicine Analysis

As the main tool of process analytical technology(PAT),near infrared(NIR)spectroscopy has attracted more and more attention in traditional Chinese medicine(TCM)analysis because it is fast,nondestructive,and of high potential for on-line analysis.Rich information can be extracted from near infrared spectra to make qualitative and quantitative analysis on TCM with the help of chemometrics.However,many problems have occurred during the usage of near infrared spectroscopy,such as bad robustness,low accuracy,and lack of enough experience about on-line applications.This thesis tries to explore solutions through experiments for specific objects.(1)Maintenance method of NIR model.This part took Paeoniae Alba Radix as research object,and explored the maintenance methods to improve the model precision.PLSR models of Paeoniae Alba Radix of 2014 were used to predict samples recently collected.RSEP was used to evaluate the accuracy,and the results showed that it was worse than the original ones.Then,HCA-RA and SSC were used to fix the models.HCA-RA applied hierarchical cluster analysis on the data combined with original and new samples,then a new calibration set was constructed.SSC calculated the residual between the average of original and new spectra,then the spectra of new samples werecorrected by the residual.The results showed the models were significantly improved by HCA-RA,and both original and new samples could be accurately predicted.While SSC had little effects on the models,and it is unsuitable for the model maintenance of Paeoniae Alba Radix.(2)Study on the spectral fusion of near infrared and ultraviolet.This part acted experiments on concentration process of Andrographis paniculata extract to seek the feasibility of using spectra fusion to improve model accuracy.An on-line spectra collection device was set up for the concentration process.Near infrared and ultraviolet spectrometers were connected to collect the spectrum of each sample.Spectra of near infrared,ultraviolet and the fused were used to build quantitative models of andrographolide and dehydroandrographolide.The validation samples of two batches were used to test the accuracy of these models.The results showed ultraviolet models were better than near infrared ones.The models of fused spectra performed best with the highest correlation coefficient and lowest RSEP.(3)Establishing on-line NIR detection system combined with automatic control.This part acted experiments with a concentration tank in the workshop.Near infrared detection pipeline was set up at the concentration tank.Flow cell in the pipeline was connected with workstation in the analysis lab with optical fibers.And the workstation communicated with the control system by OPC protocols.A near infrared on-line detection system was finished.It could monitor the manufacturing process automatically and continuously in real-time.Key information was displayed on the surface of control system.Ganmaoling concentration process was taken to test this system.The samples of several batches were collected,and the contents of chlorogenic acid,linarin,solid content and relative density were used to build quantitative models.The results showed that it is feasible to use these models to monitor this process.Real time contents were displayed on the surface in the form of trend chart.And the prediction results were close to actual values.The near infrared models combined with the on-line detection system could work successfully in the manufacturing.