红外和神经网络无损分析VC银翘片中二组分含量  

Two-component Nondestructive Analysis of VC Yinqiao Tablets with NIR And Bp Neural Network

作　　者：白英奎[1] 申铉国[2] 丁东[1] 

机构地区：[1]吉林大学通信工程学院 [2]吉林大学物理学院,吉林长春130025

出　　处：《激光与红外》2004年第5期354-356,共3页Laser & Infrared

基　　金：教育部高等学校骨干教师资助计划资助(2000931)。

摘　　要：应用近红外光谱和BP神经网络对VC银翘片进行非破坏定量分析,讨论了神经网络输入点数的选择和影响神经网络各参数的确定。采用了贝叶斯规范化的网络训练方法,解决了BP神经网络稳定性较差的问题,提高了网络的预测精度。实验及数据处理结果表明,应用本文方法对VC银翘片进行非破坏定量分析方法是可行和有效的。The application of NIR and artificial neural network for Vc Yinqiao tablet nondestructive quantitative analysis were investigated.The choice of ANN inputting node number and ANN's parameters affecting network were discussed,used Bayesian Regularization mode to train Bp Neural Network.With this mode,the question of Bp Neural Network's weak stalility were resolved,improved the forecasting precision of Neural Network.The experiment and the result of data process shows the application of Bp Neural Network In the nondestructive quantitative Analysis of Vc Yinqiao Tablets is effective and viable.

关 键 词：VC银翘片 无损分析 实验 影响 近红外光谱 定量分析 问题 BP神经网络 输入点 网络训练 

分 类 号：TN918.1[电子电信—通信与信息系统] R286.0[电子电信—信息与通信工程]