基于模糊聚类和矩阵分解的模糊辨识方法

Fuzzy Identification Based on Fuzzy Clustering and Decomposition of Matrices

通过改进模糊聚类方法确定模糊模型的前件结构 ,然后对模糊推理关系矩阵进行QR分解 ,通过分析秩的亏损来确定聚类规则的有效性 ,然后采用基于矩阵UD分解最小二乘确定模糊模型的后件参数 ,实现模糊模型的结构和参数的优化 .该方法成功地应用于Box

First,the premise structure of fuzzy model is obtained by the improved fuzzy clustering. Fuzzy inference matrices are decomposed on the basis of QR decomposition. According to analyzing the order of matrices,the validity of clustering is acquired. The recursive least squares based on matrices UD decomposition is utilized to estimate the conclusion parameters of fuzzy model. The structure and parameters of fuzzy model are optimized in the paper. The model of the gas furnace data of Box and Jenkins is successfully built.