摘要
依据工艺机理和先验知识 ,初选了乙烯氧化反应器的收率神经网络预测模型的输入变量 ,用主元分析方法对输入变量进行主元分解 ,消除了变量之间的相关性 ,减少了输入变量数 ,简化了RBF神经网络的结构。仿真结果表明 ,PCA RBF网络模型结构具有很强的预测能力。
A PCA RBF neural network model is established to predict yield of ethylene oxidization reactor on basis of its technology mechanism and pre knowledge.The principal components analysis is applied in the data processing in training sets. The result shows that the correlation of variables could be eliminated and the structure of neural network could be simplified,so that its training time will be shortened largely by this method.The simulation results show that the excellent performance of model memory and prediction are obtained by the PCA RBF neural network model.
出处
《石油化工自动化》
CAS
2001年第5期14-16,33,共4页
Automation in Petro-chemical Industry
