基于递阶分解聚类的非线性系统递推模糊辨识

提出一种基于递阶分解聚类的递推模糊辨识方法．采用半模糊化方法对论域内的样本进行归类，根据各子集“线性化”程度评判模糊聚类的有效性，通过对性能最差的子集进行分解并辨识新增子模型的参数，逐步完成整个样本空间的模糊划分和模型辨识过程．在线辨识时采用递推最小二乘算法对模糊规则进行修正，同时可根据建模精度的要求删除性能最差的规则，并确立新模糊规则．仿真研究表明了该方法的有效性．

Recursive fuzzy identification of nonlinear systems based on hierarchical decomposing clustering

A recursive fuzzy identification method based on hierarchical decomposing clustering is proposed. The samples of the region are classified by using a semi-fuzzy method, and then a judgment of the fuzzy clustering validity is made according to the linearizing level of each subset. Through decomposing the worst performance of subset and identifying these new models ' parameters, the fuzzy partition of the entire sample space and the process of model identification are gradually achieved. In process of online identification, the fuzzy rules can be rectified by using recursive least square algorithm. At the same time, the worst rules are deleted and new fuzzy rules are established to meet the demand of modeling accuracy. Simulation results show the effectiveness of this method.