采样点个数对T-S模糊建模精度的影响  

The influence of sampling points on the descriptive performance of T-S fuzzy modeling

作　　者：刘福才[1] 陈超[1] 张彦柳[1] 

机构地区：[1]燕山大学工业计算机控制工程河北省重点实验室,河北秦皇岛066004

出　　处：《智能系统学报》2008年第6期541-547,共7页CAAI Transactions on Intelligent Systems

基　　金：国家863计划资助项目

摘　　要：在模糊建模中所取的采样点个数会对辨识出的模型精度产生影响,在只给出有限个数据采样点且数据分布不能人为控制的情况下怎样选取最优的采样点个数是模糊辨识中要解决的问题之一.通过采样点个数变化的模糊辨识算法来研究模糊建模中采样点个数对模型描述性能的影响.基于T-S模糊模型,采用对称三角形模糊划分和"网格对角线法"提取模糊规则,通过对DISO系统和Mackey-Glass无序时间序列进行建模,给出模糊模型训练性能指标和检验性能指标随采样点个数增加的变化趋势曲线.The number of sampling points in fuzzy modeling has a substantial influence on the accuracy of models. If the sampled data is limited and its distribution not properly controlled, choice of the optimal number of sampling points creates significant problems in fuzzy identification. The author proposed a fuzzy identification algorithm with varied sampling points to investigate the influence of the number of sampling points on descriptive performance. Based on the T-S fuzzy model, we extracted the fuzzy rules by using the symmetrical triangular fuzzy division and the net-diagonal method. By modeling the DISO system and the Mackey-Glass chaotic time-series, we concluded that training and testing performance indexes in fuzzy models will increase with increased numbers of sampling points.

关 键 词：模糊辨识 样本点个数 三角形隶属函数 网格对角线法 T—S模糊模型 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]