文献类型：期刊文献

中文题名：基于混合核函数PCR方法的工业过程软测量建模

英文题名：Studies on Soft Sensor Modeling Using Mixtures of Kernels PCR

作者：王华忠[1];俞金寿[1]

机构：[1]华东理工大学自动化研究所,上海200237

年份：2005

卷号：32

期号：2

起止页码：23

中文期刊名：化工自动化及仪表

外文期刊名：Control and Instruments in Chemical Industry

收录：EI(收录号：2005279198126);Scopus

语种：中文

中文关键词：核函数;主元回归(PCR);软测量

外文关键词：kernel;PCR;soft sensing

摘要：针对核函数PCR方法中单个核函数的局限性,提出了混合核函数PCR方法并将其用于工业过程软测量建模,以提高模型的推广能力和非线性处理能力。混合核函数集中了多个核函数的特点,兼具局部和全局特性,并可以通过参数调节他们对混合核函数的作用,将过程的先验知识融入到核函数的确定,进而适合不同的应用。工业丙烯腈生产过程丙烯腈收率软测量建模的应用表明,基于混合核函数PCR的软测量模型要优于核函数PCR方法,满足了工业应用要求。
With regard to the limitations of kernel PCR with single kernel, mixtures of kernels principal component regression (PCR) is proposed to improve the generalization ability and nonlinearity. Mixtures of kernels are characterized by more global and local ability. The influence of different kernels can be turned by parameter to combine the prior knowledge in the determination of the kernels, which makes this method more suitable for different applications. Soft sensor modeling using this method to acrylonitrile process has shown that the method is superior to other methods, and meets the demands of field application better.