文献类型：期刊文献

中文题名：基于多神经网络模型的酯化反应软测量

英文题名：Soft Sensor of Polyester Reaction Based on Multiple Neural Network Model

作者：张宇[1];李柠[1];黄道[1]

机构：[1]华东理工大学自动化研究所,上海200237

年份：2005

卷号：31

期号：2

起止页码：208

中文期刊名：华东理工大学学报（自然科学版）

外文期刊名：Journal of East China University of Science and Technology

收录：CSTPCD;;EI(收录号：2005209110093);Scopus;北大核心:【北大核心2004】；CSCD:【CSCD2011_2012】；

基金：国家863资助项目(2002AA412120)

语种：中文

中文关键词：多模型;模糊C-均值聚类(FCM);神经网络;软测量

外文关键词：multi-model; fuzzy C-means clustering (FCM); neural network; soft senor

摘要：对模糊C-均值聚类算法加以改进,将系统输入数据进行模糊划分,分成具有几个不同聚类中心的子集;继而引入到多模型建模过程中,针对每个子集建立相应的径向基函数(RBF)网络模型。而全局模型则由各个子模型的输出加权组合。最后通过对聚合釜反应器软测量建模的研究,表明该方法具有拟合精度高和泛化能力强的特点,验证了此多模型建模方法的有效性和快速性。
Based on fuzzy C-means clustering (FCM) algorithm, a modified clustering algorithm is proposed. Using this algorithm, the input data set of a system can be quickly divided into several fuzzy clusters with distinct centers. Then, the multiple neural network modeling is introduced to the determination process. Corresponding to different clusters, each subset can be trained by radial basic function networks (RBF), and the global model is a certain combination of these multiple models. Finally, the performance of this method is evaluated by a practical case of the soft sensor of polyester reactor, which demonstrates that it has a higher approaching precision and a stronger generalization capacity. The obtained results prove its accuracy and validity.