基于SOMK算法的T-S模糊系统建模方法  

T-S fuzzy system modeling based on hybrid of SOM and K-means

作　　者：杨印生[1] 孙赵华[1] 马萍[2] 陶跃 司瑾 

机构地区：[1]吉林大学生物与农业工程学院,长春130022 [2]黑龙江八一农垦大学食品学院,黑龙江大庆163319 [3]长春理王大学计算机科学技术学院,长春130022

出　　处：《吉林大学学报（工学版）》2008年第3期658-661,共4页Journal of Jilin University:Engineering and Technology Edition

基　　金：吉林省科技发展计划项目(20060529)

摘　　要：针对T-S模糊系统建模中的结构辨识和参数辨识问题,以自组织特征映射SOM算法与K-means算法相结合的SOMK算法对输入样本进行聚类,同时以Davies-Bouldin(DB)指标来验证聚类的有效性,得到最佳聚类数即为T-S模糊系统模型的规则数。然后用所得模糊规则数和聚类中心来辨识T-S模糊系统模型前件参数进而采用最小二乘算法来确定模糊推理规则的后件参数。最后,用仿真实例验证了文中所提出的建模方法与相应算法。仿真结果表明,该方法逼近精度高,泛化能力强,具有较好的实际应用价值。Focused on the structure and parameter identificantions in T-S fuzzy system modelling, a hybrid clustering method SOMK, which combines the Self-Organizing Map （SOM） and K-means approaches, is used to cluster the input samples, and the Davies-Bouldin （DB） validity criterion is used to determine the optimal number of clusters, which is taken as the optimal number of rules of the T-S fuzzy system. Then the premise parameters are determined by the number of rules and the cluster centers. Furthermore, a least square algorithm is applied to obtain the consequent part of each rule. A case simulation is used to illustrate the proposed medolling method and the algorithm. Simulation results prove that the method has high modelling precision and high ability of power generalization, and it is valuable for practical application.

关 键 词：系统工程 SOM算法 K-MEANS算法 FCM算法 T-S模糊系统 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]