摘要
介绍了将主元分析方法（ＰＣＡ）与径向基函数神经网络（ＲＢＦＮ）相结合，用主元分析对高维输入变量进行预处理，构造反映过程信息的低维主元变量，再用径向基函数神经网络对主元变量建立自校正模型。这种方法不仅简化了神经网络模型的结构，而且可以借助主元分析方法对仪表和过程故障引起的数据过失误差进行侦破，避免导致模型的错误输出。用这种方法建立可侦破故障的反应器温度自校正模型，取得了良好的效果。
A new modeling method based on principle component analysis(PCA) and radical basis function neural network(RBFN)is presented.The PCA is worked as a preprocessor,which projects the high dimensions input variables into a low dimensions principle components,while the RBFN is used to establish a self tuning model with these principal components,the PCA RBFN method can not only simplify the structure of neural networks and improve the accuracy of model ,but also avoid wrong result caused by fault of sensor or process based on multivariable statistics ,which is very important in optimal control.An on line temperature self tuning model which can detect faults for a reactor was established by utilizing the PCA RBFN method ,and results were satisfied.
出处
《石油化工自动化》
CAS
1998年第1期23-25,共3页
Automation in Petro-chemical Industry
