基于简约集向量的Takagi–Sugeno模糊模型

A Takagi-Sugeno fuzzy model based on the reduced set-vector

中文关键词  模糊建模  核函数  简约集向量

英文关键词  fuzzy modeling  kernel  reduced-set vector

中文摘要

      利用支持向量学习机制建立模糊模型时, 过多的支持向量将导致复杂的模糊模型. 为此提出了一种基于简约集向量的Takagi-Sugeno模糊模型. 该模型抽取简约集向量产生模糊规则, 规则前件的乘积型多维模糊隶属度函数直接由Mercer核构成, 而规则后件则采用非线性函数. 模型的结构和参数可通过自下而上的简化规则以及不敏感学习进行有效地辨识. 最终得到的模糊模型具有良好的推广能力与精确性, 同时拥有高透明度的模糊规则库. 通过二维sinc函数的逼近及球棍系统的模糊控制的仿真实例, 说明了提出模型的有效性.

英文摘要

      In applying the support-vector learning mechanism to fuzzy modeling, we find that excessive support-vectors will lead to a complicated fuzzy model. Based on the reduced set-vector, we present a Takagi-Sugeno fuzzy model(RV–TSFM) which alternatively extracts reduced set-vectors for generating fuzzy rules. The product-type multidimensional fuzzy membership functions of rule-antecedents can be directly generated by Mercer kernels; and the nonlinear functions represent the rule-consequents. By utilizing the bottom-up simplification algorithm and insensitive learning, we can effectively identify the model structure and parameters. The final model derived provides a desirable accuracy and generalization performance, as well as a highly transparent rule base. Finally, the proposed fuzzy model is successfully applied to the approximation of a two-dimensional sinc function, and the fuzzy control of a ball-and-beam system.