Wang Yu, Wang Ya-Jing, Qi Dong-Li, Gao Di, Wang Xiao-Yu, Huang Zan-Yang, Wang Yan-Wen, Ye Xiang-Yin
Engineering Research Center of Modern Chinese Medicine Discovery and Preparation Technique, Ministry of Education,Tianjin University of Traditional Chinese Medicine Tianjin 301617, China.
Zhongguo Zhong Yao Za Zhi. 2022 Jun;47(11):2955-2963. doi: 10.19540/j.cnki.cjcmm.20210707.303.
In this paper, a flavonoid extract powder properties-process parameters-granule forming rate prediction model was constructed based on design space and radial basis function neural network(RBFNN) to predict the formability of flavonoid extract gra-nules. Box-Behnken experimental design was employed to explore the mathematical relationships between critical process parameters and quality attributes. The design space of critical process parameters was developed, and the accuracy of the design space was verified by Monte Carlo method(MC). Design Expert 10 was used for Box-Behnken design and mixture design. Scutellariae Radix mixed powder was prepared and its powder properties were measured. The mixed powder was then subjected to dry granulation and the granule forming rate was determined. The correlations between powder properties were analyzed by variance influence factor(VIF), and principal component analysis(PCA) was performed for the factors with strong collinearity. In this way, a prediction model of powder properties-process parameters-granule forming rate was established based on RBFNN, the accuracy of which was evaluated with examples. The results showed that the model had a good predictive effect on the granule forming rate, with the average relative error of 1.04%. The predicted value and the measured value had a high degree of fitting, which indicated that model presented a good prediction ability. The prediction model established in this study can provide reference for the establishment of quality control methods for Chinese medicinal preparations with similar physical properties.
本文基于设计空间和径向基函数神经网络（RBFNN）构建了黄酮提取物粉末性质 - 工艺参数 - 颗粒成型率预测模型，以预测黄酮提取物颗粒的成型性。采用Box - Behnken实验设计探究关键工艺参数与质量属性之间的数学关系。开发了关键工艺参数的设计空间，并通过蒙特卡罗方法（MC）验证了设计空间的准确性。使用Design Expert 10进行Box - Behnken设计和混合设计。制备黄芩混合粉末并测定其粉末性质。然后对混合粉末进行干法制粒并测定颗粒成型率。通过方差影响因子（VIF）分析粉末性质之间的相关性，对共线性较强的因素进行主成分分析（PCA）。以此为基础，基于RBFNN建立了粉末性质 - 工艺参数 - 颗粒成型率预测模型，并通过实例评估了其准确性。结果表明，该模型对颗粒成型率具有良好的预测效果，平均相对误差为1.04%。预测值与实测值拟合度高，表明该模型具有良好的预测能力。本研究建立的预测模型可为具有相似物理性质的中药制剂质量控制方法的建立提供参考。
