基于Box-Behnken响应面法结合BP神经网络多指标优化汉桃叶微丸的制备工艺

Multi-index optimization of Hantaoye micro-pills based on Box-Behnken response surface method combined with BP neural network

目的采用Box-Behnken响应面法结合BP(back-propagation)神经网络多指标优化汉桃叶微丸的制备工艺。方法利用挤出滚圆法制备汉桃叶微丸,在单因素试验的基础上,以辅料微粉硅胶与微晶纤维素(microcrystalline cellulose,MCC)的比例、润湿剂、载药量、滚圆频率和滚圆时间为考察因素,以收率(%)、圆整度、豪斯纳比(hausner ratio,HR)和脆碎度(%)为评价指标,基于G1-熵权法对各评价指标进行组合赋权并计算综合评价结果,从而优化处方组成及其制备工艺;建立BP神经网络模型,选取合理数据进行学习和验证并预测汉桃叶微丸的最佳制备工艺。结果采用Box-Behnken响应面法及BP神经网络预测的汉桃叶微丸的最佳制备工艺为微粉硅胶与MCC的比例为1∶3(g∶g)、载药量为25%、滚圆频率为22 Hz。验证试验的结果表明,Box-Behnken响应面法的综合评价结果均值与响应面优化理论值的绝对误差为0.0077,相对误差为0.79%。结论基于Box-Behnken响应面法结合BP神经网络多指标优化的汉桃叶微丸的制备工艺稳定可行、较为合理。

Objective:The Box-Behnken response surface methodology combined with BP neural network was used to optimize the preparation process of Hantaoye micropills.Methods:preparation of micropills of Hantaoye by extrusion rounding method was utilized,based on the one-way test,the ratio of micronized silica gel to MCC,wetting agent,loading capacity,rounding frequency and rounding time of excipients were examined as factors,and yield(%),roundness,HR(hausner ratio)and brittleness(%)are used as evaluation indexs,based on G1-entropy weighting method,the evaluation indexs are combined and the comprehensive evaluation results are calculate-d,so as to optimize the prescription composition and its preparetion process,and a BP neural netw-ork model was established to select reasonable date for learning and validating and predicting theoptimal preparation process of Hantaoye micropills.Results:The optimum process for the preparation of Hantaoye micropills predicted by Box-Behnken response surface methododology and BP neural network was 1∶3(g∶g)ratio of micronized silica gel to MCC,25%drug loading,and 22 round-ing frequency.The results of the validation test show that the absolute error of the mean value of the comprehensive evaluation results of the Box-Behnken response surface method with the theoretical value of the response surface optimization is 0.0077,and the relative error is 0.79%.nse surface method combined with BP neural network multi-indicator optimization is stable,feasi-ble and more reasonable.Conclusion:The preparation process of Hantao leaf micro-pills based on Box-Behnken response.