红景天苷缓释微囊神经网络建模

本文在确定输入层和输出层单元、预处理网络数据、选择激活函数、选择训练方法的基础上,建立了红景苷缓释微囊制作参数与性能之间的神经网络模型,网络结构为5-12-3。该模型能较为精确的拟合输入的样本数据,其最大相对误差不超过4%,模型准确可信,可以代替真实试验,该模型的建立为工艺参数的优化打下基础。

Modeling for Salidroside Microcapsule Release Based on Artificial Neural Network

Artificial neural network model for chitosan-alginate microcapsules was established on the basis of selecting inputoutput units, pretreating data, determining transfer function and choosing training function. The network structure is five neurons input layer, three neurons output layer and hidden layers with the number of neurons 12. The trained network has a high generalization, and the error between the predictions and the real values is less than 4% and can take the place of real experiment. The works provide a basis for optimizing process parameters.