Establishment of Feedforward Control Model for Dry Granulation of Polysaccharide Components Based on Quality by Design / 中国实验方剂学杂志

ObjectiveA feedforward control model for dry granulation of polysaccharide components was established to guide the adjustment and optimization of critical process parameters （CPPs） in the design space， so as to reduce the impact of fluctuations in raw materials properties on the quality of medicines. MethodTaking Astragali Radix extract powder as the model drug， the design space of dry granulation CPPs was determined by Box-Behnken design. Astragali Radix mixed powder with different powder properties were prepared by mixture design， the variance inflation factor （VIF） was used to diagnose the multicollinearity of the powder properties， and principal component analysis （PCA） was used to extract the characteristic data of the model. Radial basis function neural network （RBFNN） was used to establish a feedforward control model for reflecting the relationship between the powder properties of polysaccharide components， dry granulation CPPs and one-time molding rate. ResultThe design space for dry granulation CPPs of polysaccharide components was 16-35 Hz for feeding speed， 10-23 Hz for roller speed， and 10-46 kg·cm-2 for roller pressure. The established RBFNN feedforward control model had a good predictive effect on the one-time molding rate of dry granulation of polysaccharide components， which could be used to guide the adjustment and optimization of CPPs in the design space， the relative error was 0.38%-6.73%， and the average relative error was 3.42%. ConclusionThe established feedforward control model can well reflect the relationship between the powder properties of the polysaccharide components， the dry granulation CPPs and the one-time molding rate of the granules， which can be used to guide the adjustment and optimization of CPPs in the design space， reduce the impact of material property fluctuation on product quality， and provide ideas for promoting the quality of traditional Chinese medicine from passive control to active control.

Prediction of Disintegration Time of Traditional Chinese Medicine Tablets Based on Generalized Regression Neural Network / 中国实验方剂学杂志

Objective:This paper constructs a generalized regression neural network （GRNN） model to predict the disintegration time of traditional Chinese medicine （TCM） tablets. Method:Taking Astragali Radix as a model drug， the mixed Astragali Radix powders with different powder properties were prepared by mixing Astragali Radix extract powders with microcrystalline cellulose and lactose， which were made to Astragali Radix tablets by direct compression method. The powder properties of mixed Astragali Radix powders and the disintegration time of Astragali Radix tablets were determined， respectively. The correlation between the original data was eliminated by principal component analysis （PCA）. The principal component factors were used as the input layer of the GRNN model， and the disintegration time was used as the output layer for network training. Finally， the verification group data was used to predict the disintegration time， and the network prediction accuracy was calculated by comparing with the actual value. Result:Three principal component factors were obtained through PCA by analyzing the original nine variables that were correlated with each other （Hausner ratio， true density， tap density， compression degree， angle of repose， bulk density， porosity， water content and total dissolved solids）， which reduced the complexity of the network. The prediction value of the disintegration time based on this prediction method was in good agreement with the actual value， the error of disintegration time was 0.01-1.34 min and the average relative error was 3.16%. Conclusion:Based on the GRNN mathematical model， the physical properties of Astragali Radix extract powders can be used to accurately predict the disintegration time of Astragali Radix tablets， which provides a reference for studying the disintegration time of TCM tablets.

Research on Bayesian fault diagnosis model of traditional Chinese medicine dry granulation based on failure model and effect analysis (FMEA) / 中国中药杂志

This paper aims to construct a Bayesian(BN) fault diagnosis model of traditional Chinese medicine dry granulation based on the failure model and effect analysis(FMEA), effectively control risk factors and ensure the quality of granules.Firstly, the risk ana-lysis of dry granulation process was carried out with FMEA, and the selected medium and high risk factors were taken as node variables to establish corresponding BN network with causality.According to the mathematical reasoning method of probability theory, the model was accurately inferred and verified by Netica, and the granule nonconformance was used as the evidence for reversed reasoning to determine the most likely cause of the failure that affected the granule quality.The BN fault diagnosis model of traditional Chinese medicine dry gra-nulation was established based on the medium and high risk factors of process, prescription and equipment screened out by FMEA, such as roller pressure, raw material viscosity, clearance between rollers in the paper.The fault diagnosis of traditional Chinese medicine dry granulation process was then carried out according to the model, and the posterior probability of each node under the premise of nonconforming granule quality was obtained.This method could provide strong support for operators to quickly eliminate faults and make decisions, so as to improve the efficiency and accuracy for fault diagnosis and prediction, with innovation in its application.

Material properties and tensile strength prediction model of traditional Chinese medicine tablets based on PCA-RBF neural network / 中国中药杂志

This paper constructs a prediction model of material attribute-tensile strength based on principal component analysis-radial basis neural network( PCA-RBF),in order to predict the formability of traditional Chinese medicine tablets. Firstly,design Expert8. 0 software was used to design the dosage of different types of extracts,the mixture of traditional Chinese medicine with different physical properties was obtained,the powder properties of each extract and the tensile strength of tablets were determined,the correlation of the original input layer data was eliminated by PCA,the new variables unrelated to each other were trained as the input data of RBF neural network,and the tensile strength of the tablets was predicted. The experimental results showed that the PCA-RBF model had a good predictive effect on the tensile strength of the tablet,the minimum relative error was 0. 25%,the maximum relative error was2. 21%,and the average error was 1. 35%,which had a high fitting degree and better network prediction accuracy. This study initially constructed a prediction model of material properties-tensile strength of Chinese herbal tablets based on PCA-RBF,which provided a reference for the establishment of effective quality control methods for traditional Chinese medicine preparations.