PCA—RBF方法在聚丙烯熔融指数建模中的应用

PCA-RBF method in polypropylene melt index modeling application

针对聚丙烯的生产过程是一个大滞后、时变、非线性的复杂系统,提出了基于主成分分析(PCA)的RBF神经网络聚丙烯熔融指数建模方法。该方法用主元分析对高维输入变量进行预处理,构造反应过程信息的低维主元变量,再经径向基函数神经网络对主元变量进行建模。该方法不仅简化了神经网络的结构,而且可以借助主元分析方法对过程故障和过失误差进行侦破,避免导致模型的错误输出。理论分析和实验结果表明,基于PCA和RBF网络方法的聚丙烯熔融指数建模具有精度高、鲁棒性强的优点,有利于工业生产应用。

In polypropylene production process is a big lag, time-varying, nonlinear complex system, is put forward based on the principal component analysis （ PCA ） and RBF neural network modeling method of polypropylene melt index. The method uses prin- cipal component analysis to high dimension input variables for pretreatment, structure of reaction process information of low dimen- sional principal component variables, and then by radial basis function neural network principal component variables for modeling. The method not only simplifies the structure of the neural network, but also by means of principal component analysis to process fault and negligence error is detected, to avoid the model error output. Theoretical analysis and experimental results show that, based on the PCA method and RBF polypropylene melt index modeling has the advantages of high precision, robust, conducive to industrial production and application.