基于PCA和LS-SVM的软测量建模与应用  

Soft Sensor Modelling and Application Based on PCA and LS-SVM

作　　者：俞佩菲[1] 吴燕玲[1] 卢建刚[1] 孙优贤[1] 

机构地区：[1]浙江大学工业控制技术国家重点实验室,浙江杭州310027

出　　处：《江南大学学报（自然科学版）》2006年第2期182-186,共5页Joural of Jiangnan University (Natural Science Edition)　

基　　金：国家自然科学基金项目(20206027);国家973计划项目(2002CB312200)

摘　　要：针对工业过程中某些重要过程变量难以实现实时在线检测和高维数据处理的问题,提出了将主元分析与最小二乘支持向量机相结合的软测量建模方法,并利用该方法建立了工业阿维菌素发酵过程中的菌丝浓度软测量模型.主元分析方法的引入,有效地提高了最小二乘支持向量机软测量模型的精度和泛化能力.应用结果表明,该方法与基于径向基函数神经网络软测量模型相比具有有效性和优越性.To solve the problems of real-time on-line measurements of some important process variables and of data handling with high dimension, a novel method of soft sensor based on the integration of both PCA and LS-SVM is proposed and applied in the modelling of biomass estimation in the fermentation process of Avermectin in this paper. The introduction of the method of PCA contributes to the distinct improvement of precision and generalization ability of the soft sensor model based on LS-SVM. Industrial application has shown that the proposed method is superior to the soft sensor model based on RBF neural network.

关 键 词：最小二乘支持向量机 主元分析 软测量 RBF神经网络 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]