基于模糊最小二乘支持向量机的软测量建模  

Soft sensor modeling based on fuzzy least squares support vector machines

作　　者：张英[1] 苏宏业[1] 褚健[1] 

机构地区：[1]工业控制技术国家重点实验室浙江大学先进控制研究所,浙江杭州310027

出　　处：《控制与决策》2005年第6期621-624,共4页Control and Decision

基　　金：国家"十五"科技攻关项目(2001BA204B07);国家863计划项目(2001AA413020).

摘　　要：将模糊隶属度概念引入最小二乘支持向量机,提出一种基于支持向量数据域描述的模糊隶属度函数模型,将输入空间中的样本映射到一个高维的特征空间;然后根据其偏离数据域的程度赋予不同的隶属度.该方法提高了最小二乘支持向量机的抗噪声能力,尤其适用于未能完全揭示输入样本特性的情况.将提出的方法用于催化裂化分馏塔轻柴油凝固点的软测量建模,仿真结果表明,该模糊隶属度函数模型能够提高最小二乘支持向量机的预测精度.A fuzzy membership model based on support vector data description is proposed to fuzzify all the training data. Then the model is introduced into least square support vector machines (LS-SVM). Data samples in input space are mapped into a high dimensional feature space, and then the smallest enclosing hypersphere is found. The fuzzy membership value to each input point is computed according to the distances to the center of hypersphere. The proposed soft sensor model based on fuzzy least squares support vector machines is applied to predict the frozen point of light diesel in fluid catalytic cracking process. Simulation result shows that the proposed method actually increases the accuracy of LS-SVM.

关 键 词：支持向量机 数据域描述 模糊隶属度 软测量 建模 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]