基于神经网络和遗传算法的红景天苷缓释微囊制备过程建模与优化

建立了红景天苷缓释微囊的人工神经网络模型及其遗传算法优化技术。结果表明,结构为5-12-3的神经网络模型能较为精确地拟合测试的样本数据,其最大相对误差不超过4%;遗传算法优化的红景天苷缓释微囊制作最佳工艺参数为:海藻酸钠与红景天苷的质量比为2,海藻酸钠浓度为30g/L,壳聚糖浓度为5g/L,氯化钙浓度为10g/L,壳聚糖溶液pH值为6.35,该工艺参数下的最大适应度较单因素及二次旋转组合试验中的最大适应度高14%,且最佳工艺参数下载药量、包埋率和决定系数的预测值和试验值基本相符。说明用神经网络模型描述微囊制作参数与性能之间的关系,用遗传算法优化微囊制作工艺参数,能设计出性能最佳的微囊制作工艺参数。

Modeling and optimization of release salidroside microcapsules manufacturing process based on artificial neural network and genetic algorithm

The Artificial neural network (ANN) model for salidroside microcapsules was studied and ANN model was optimized by using genetic algorithm (GA) in this paper.Results showed that the 5-12-3 structure of network had a highly generalization,the errors between the predicted and the real values were less than 4.The optimization process parameters,in which the fitness was 14 greater than that in experiments,the best technology to produce microcapsules was as follows:the ratio of alginate weight to salidroside weight was 2,alginate concentration 30 g/L,chitosan concentration 5 g/L,calcium chloride concentration 10 g/L,pH value of chitosan solation 6.35 and the predicted value fitted test result basically.The optimum process parameters could be obtained using ANN model to describe relationships between process parameters and performance and GA to optimize process parameters.