一种T-S模糊模型的自组织辨识算法及应用  

Self-organizing identification algorithm for T-S fuzzy model and its applications

作　　者：梁炎明[1] 刘丁[1] 

机构地区：[1]西安理工大学自动化与信息工程学院,西安710048

出　　处：《仪器仪表学报》2011年第9期1941-1947,共7页Chinese Journal of Scientific Instrument

基　　金：国家科技重大专项(No.2009ZX02011001)资助项目

摘　　要：提出了一种新的具有自适应学习能力的T-S模糊模型辨识算法。该算法通过使同一条规则的高斯函数的宽度参数彼此相等建立与支持向量机等效的T-S模糊模型,在此基础上,利用模糊聚类算法和支持向量机分别建立前后件辨识模型,并利用一种改进粒子群优化算法优化输出误差函数使前后件参数联合辨识,从而获得T-S模糊模型的结构和参数。仿真结果表明,相比其它方法,文中方法具有较高的逼近精度和较好的泛化能力,由此算法获得的直拉单晶炉热场模型具有0.1171的均方差,完全符合均方差小于0.5的要求。A new identification algorithm for T-S fuzzy model has been proposed, which has the properties of self- learning and self-adapting. This algorithm establishes a T-S fuzzy＇ model equivalent to support vector machine （SVM） model through making Gaussian function width parameters in the same rule equal each other. Based on this, the antecedent identification model and the consequent identification model are respectively established using fuzzy clustering algorithm and support vector machine （SVM）. An improved particle swarm optimization （PSO） algorithm is used to identify the antecedent parameters and consequent parameters simukaneously through optimizing the output error function. Thus, the structure and parameters of the T-S fuzzy model are obtained. Simulation results show that the proposed method demonstrates higher approximation accuracy and better generalization ability compared with other methods. The hot-zone model of a Czochralski crystal growing furnace, which was built based on the proposed identification algorithm, has the mean square error （MSE） of only 0.1171 which completely meets the requirements of the MSE within 0.5.

关 键 词：T—S模糊模型 自组织 支持向量机 模糊聚类 粒子群优化 直拉单晶炉热场 

分 类 号：TP14[自动化与计算机技术—控制理论与控制工程] TG232.5[自动化与计算机技术—控制科学与工程]