文献类型：期刊文献

中文题名：BP神经网络模型预测白芷超临界萃取结果的研究

英文题名：Forecasting the Extraction of Coumarins in Angelica dahurica by Supercritical CO_2 with BP Neural Network

作者：刘红梅[1]

第一作者：刘红梅

机构：[1]北京联合大学生物化学工程学院

第一机构：北京联合大学生物化学工程学院

年份：2006

卷号：17

期号：2

起止页码：176-177

中文期刊名：时珍国医国药

外文期刊名：LiShiZhen Medicine and Materia Medica Research

收录：北大核心:【北大核心2004】；CSCD:【CSCD_E2011_2012】；

基金：山东省医学科学院课题(No.2001-89)

语种：中文

中文关键词：神经网络;超临界萃取;白芷;正交实验设计

外文关键词：BP Neural network; Supereritieal Extraction; Angelica dahurica; Orthogonal Design

摘要：目的用BP神经网络对白芷中香豆素类成分的超临界CO2萃取结果进行预测。方法以萃取压力、温度、时间、分离压力、药材粉碎度为网络的输入向量,以总香豆素、氧化前胡素和欧前胡素作为网络的输出,建立BP神经网络模型,利用正交实验数据对网络进行训练。结果得到一个结构为5×4×3的2层BP网络,测试样本的实际测量值与网络的输出吻合很好。结论可利用正交实验数据对BP网络进行训练,利用训练好的网络对中药药效物质基础的超临界萃取结果进行预测。
Objective To predict the extraction results of coumarins in Angelica dahurica by supercritical CO2 using back - propagation neural network. Methods BP neural network model was established by taking extraction pressure, temperature and time as well as separation pressure and pulverized degree as the inputs of network and the contents of oxyimperatorin, imperatorin and the overall coumarins in the extract as outputs of neural network. BP network was trained with Orthogonal experimental results. Results A 5×4×3 neural network has been set up. The forecasted profiles based on BP - NN model were closely similar to the target values. Conclusion Orthogonal experimental results could be used to training BP - NN and the trained BP - NN could foreeast the extraction results of active components in Chinese herb mcdicines.