基于KPCA和最小二乘支持向量机的软测量建模  

Soft Sensor Modeling Based on KPCA and Support Vector Machines

作　　者：周林成[1] 杨慧中[1] 

机构地区：[1]江南大学通信与控制工程学院,江苏无锡214122

出　　处：《计算机仿真》2008年第10期94-97,共4页Computer Simulation

基　　金：国家自然科学基金资助项目(No60674092);江苏省高科技研究项目(工业)(BG2006010)

摘　　要：为了解决化工领域数据建模小样本、不适定性、非线性等问题,提出了一种基于核主元分析(KPCA)和最小二乘支持向量机的软测量建模方法,用核主元分析对输入变量进行数据压缩,消除变量之间的相关性,运用PSO算法对核参数进行了寻优,通过交叉验证的方法对支持向量机进行参数选择。将其用于双酚A(BPA)软测量建模的研究结果表明:方法具有学习速度快、泛化能力强等优点,为BPA软测量建模的在线实施提供了方便。In order to solve the problem of the small and ill - posed sample and nonlinearity in chemical industry processes,a soft measure method based on Kernel Principal Component Analysis （KPCA） and least square support vector machine is presented. The Kernel Principal Component Analysis method could not only solve the linear correlation of the input but also compress the data. Kernel parameters of KPCA are optimized by particle swarm optimization. Cross validation method is used to select parameters of least square support vector machine model. The measure method is applied to predict BPA. Results indicate that this method features high learning speed, good approximation and well generalization ability. It provides evidences for on - line BPA measurement.

关 键 词：最小二乘支持向量机 核主元分析 软测量 泛化能力 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]