摘要
通过对输入变量进行主元分析 ,简化网络结构 ,提高网络训练速度和外推能力 ,为软测量技术的在线应用提供了更大的方便 .仿真结果表明 :改进后的方法改善了网络训练速度和外推能力 .
In application of soft sensing technique based on neural network, the choice of input variable is very important for success of application. Whereas,in continuous producing process ,it is inevitable for linear correlation between input variable selected. In this way, training speed of network can become slow and generalization of network can become bad. In this paper, PCA(principle component analysis) method is incorporated into network, which not only solve the linear correlation of inputs, but also simplify the network structure and improve the training speed and generalization. This brings great convenience to on line application. Simulation results demonstrate that: technique improved is very valid.
出处
《沈阳化工学院学报》
2000年第2期81-86,共6页
Journal of Shenyang Institute of Chemical Technolgy
关键词
神经元网络
软测量技术
主元分析
过程控制
neural network
\ soft sensing technique
\ PCA(principle component analysis)
