Research On The Design And Analysis Method Of NIR Model Of Traditional Chinese Medicine Based On System Modelin

For a long time,Chinese medicine pharmaceutical process detection lag,the quality fluctuation of raw materials and production process complexity,it is difficult to guarantee the quality of traditional Chinese medicine products stable and uniform,so the establishment of online monitoring technology for effective Chinese medicine pharmaceutical process control is the key to ensure the quality of traditional Chinese medicine.Near infrared spectroscopy(NIRS)online monitoring technology is widely used in online monitoring and quality control of Chinese medicine manufacturing process based on its fast,non-destructive and reliable characteristics.However,NIR online monitoring technology is an indirect analysis technology.We must rely on chemometrics to establish a robust and reliable mathematical model to ensure the reliability of online detection technology in Chinese medicine manufacturing process.Therefore,based on the above problems,this article from the NIR model to identify the key factors of modeling and analysis,model design three levels to carry out research work,based on quality by design(Quality by,design,QbD)the concept of near infrared model method to identify the key factors of key factors in the process of identification and control sites;combined process the analytical method of traditional Chinese medicine to understand the process of NIR modeling parameters including sample set error division ratio,latent variable factor selection,spectral preprocessing and variable selection method produced in the transmission of the link,at the same time quantization error transfer process parameters influence degree of the quality of the final model;design space based on the QbD concept of TCM based on near infrared model the parameters to select the best model parameters,provide a reference method for establishing quantitative models for robust and reliable It provides the application support for the near infrared spectroscopy analysis technology to realize the on-line monitoring reliability of traditional Chinese medicine pharmaceutical process.Based on the above research purposes,the main contents of this paper include the following three parts:(1)Identification of key factors in the modeling process of traditional Chinese medicine near infraredThe first part mainly honeysuckle,licorice and gardenia herbs;NIR data of cisplatin,carboplatin and oxaliplatin for anticancer drugs as the research object,the results were compared with the traditional method of modeling the first screening of different chemometric methods variables on the characteristic information of honeysuckle,licorice and gardenia,find the optimal variable selection method and the establishment of NIR stable quantitative model,at the same time the effects of variables,screening and screening method for modeling band model prediction performance of different pretreatment methods results;while using two-dimensional correlation spectroscopy analytical method,NIR spectrum generation of cisplatin,carboplatin and oxaliplatin for anticancer drugs,and by two dimensional correlation spectroscopy combined with chemometrics analysis of the auto correlation peak,by NIR features the auto correlation peak determine platinum anticancer drug absorption peak.The results showed that the main NIR absorption bands of cisplatin,carboplatin and oxaliplatin were three 1380-1409 nm,1510-1565 nm,1857-1911 nm,2016-2217 nm and 2231-2500 nm.(2)Understanding of the process of establishing NIR quantitative model based on the theory of multivariable detection limit.The second part mainly in open source corn NIR data and NIR data extraction of Gardenia pilot online as the research object,multivariate error transfer theory and multi variable detection limit theory formula based on study NIR quantitative model in the process from latent variables to the spectral pretreatment to the variable selection method selection error three steps from a selection of parameters to the next parameter selection generated and quantified,results show that while the impact of several different factors;transfer multivariate error is discussed,including sample set division ratio,noise level,add different near infrared bands and other factors.The model of covariance corn and glutinous rice was optimized through the error transmission and screening of various modeling steps.The best modeling conditions for corn samples were:SG9 pretreatment method,6 latent variable factors,and VIP variable screening method.The RMSEC and Real2 of the model were 0.04569 and 0.9908,respectively,and the RMSEP and Rpre2 were 0.03827 and 0.9873,respectively.The best modelling conditions of wolfberry samples were:SG9+1D pretreatment method,five latent variable factors,and SiPLS variable selection method.The RMSEC and Rcal2 of the obtained model were 0.02049 and 0.9978,and the RMSEP and Rpre2 were 0.02825 and 0.9692,respectively.(3)Establishing NIR robust quantitative model of traditional Chinese medicine based on QbD conceptThe third part is to extract NIR data from Gardenia and Glycyrrhiza online,and optimize the establishment process parameters of licorice and gardenia near infrared model with experimental design and design space to improve the robustness of the model.Firstly,through experience analysis and selection of calibration set decision coefficient Rcal2,root mean square error of calibration and prediction sets RMSEC decision coefficient Rpre2,root mean square error of prediction(the root mean square error of prediction,RMSEP),the multi variable detection limit of MDL as a key model parameter evaluation index,and then the D-optimal test design selection of key model parameters,namely the sample set division(K-S)ratio,the number of latent variables(latent,variables,LVs)spectral pretreatment methods(pretreaments,P)and variable selection method(variable selection V),a total of 56 times in the modeling experiment,and establish the mathematical model between key parameters and evaluation model of the key modeling parameters.The design space of Gardenia near infrared model is adopted to estimate the reliability of the confidence interval method,in the design space to get the best parameters for the modeling of Geniposide by Baseline pretreatment method,selecting 8 latent variables,establishing method of Geniposide NIR quantitative model by SiPLS variable selection.The model of RMSEC and Rcal2 were 0.02230%and 0.9773,RMSE is 0.02616 and RMSEC is close to that of MDL and Rpre2 were 2.013*10-3 and 0.9678;calculate the design space of Bayesian probability licorice near infrared model based on Monte by using Caelo,and finally determine the optimal combination of parameters for modeling,the result shows that D-optimal test design the feasibility of extraction process application in traditional Chinese medicine.