超声强化超临界流体萃取人工神经网络模拟  

The simulation on ultrasound-enhanced supercritical fluid extraction with the artificial neural network

作　　者：杨日福[1] 丘泰球[2] 丁彩梅[2] 

机构地区：[1]华南理工大学物理科学与技术学院,广东广州510640 [2]华南理工大学轻化工研究所,广东广州510640

出　　处：《计算机与应用化学》2007年第9期1201-1204,共4页Computers and Applied Chemistry

基　　金：广州市科技计划项目(穗科条字[2002]27号)

摘　　要：将人工神经网络技术用于超声强化超临界流体萃取过程的模拟,以香椿叶中黄酮类化合物为提取对象,系统地研究萃取温度、萃取压力、流体流量、夹带剂、萃取时间、超声电功率对超声强化超临界流体萃取的影响,建立结构为7-10-1的三层BP网络模型,确定输入层—隐含层、隐含层—输出层之间的传递函数的最优化组合形式,选择较好的L-M算法,可以用一定量的萃取实验数据对网络进行训练,能够对同类实验结果进行模拟,更真实反映超声强化超临界流体萃取实验规律。The artificial neural network was taken to simulate the ultrasound-enhanced supercritical fluid extraction （USFE）. Effects of extraction temperature, pressure, fluid flow rate, modifier, time and ultrasonic power on flavonoids extraction yield from Toona sinensis leaves with USFE were discussed. Based on a reasonable model of BP artificial neural network, the structure of network was three-layer of 7-10-1 and the optimized combination form of transfer function between hidden layer and input layer or output layer was determined, and with L-M method （arithmetic function： trainlm）, it is feasible that a proper amount of extraction experimental data is selected to train the network and simulate the results of similar experiments. It can reflect the experimental rules of USFE more actually.

关 键 词：人工神经网络 超声强化超临界流体萃取 黄酮类化合物 模拟 

分 类 号：TQ028.32[化学工程]