基于混合SVR-PLS方法的丙烯腈收率软测量建模  

Soft sensor modeling of acrylonitrile yield based on hybrid SVR-PLS approach

作　　者：王华忠[1] 俞金寿[1] 

机构地区：[1]华东理工大学自动化研究所,上海200237

出　　处：《控制与决策》2005年第5期549-552,共4页Control and Decision

摘　　要：为了更有效地处理过程非线性、多输入和数据共线性等复杂特性,提高模型的推广能力和精度,提出了混合支持向量回归机-偏最小二乘法(SVR-PLS)方法.该方法兼具SVR和PLS的优点,用PLS进行特征提取,用SVR建立PLS的内部模型.对工业丙烯腈生产过程丙烯腈收率软测量建模的应用表明,采用该方法建立的软测量模型,在模型精度、推广能力等方面明显优于一些传统软测量建模方法,满足工业应用要求.A hybrid SVR-PLS method is proposed to deal with complicated process with nonlinearity and a large number of correlated inputs. The SVR-PLS method, which has merits of both SVRs and PLS, is an integration of support vector regression machine and partial least squares. The PLS outer projection is used as a dimension reduction tool to remove collinearity and the SVRs are trained to capture the nonlinearity in the projected latent space. Soft sensor modeling of acrylonitrile yield is established using SVR-PLS method. The generalization ability and accuracy of the soft sensor using the method proposed is superior to traditional methods.

关 键 词：支持向量回归机 偏最小二乘法 丙烯腈 软测量 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TQ063[自动化与计算机技术—控制科学与工程]