基于目标函数的矩阵分解非线性系统模糊建模

提出了基于新的目标函数的模糊聚类建模方法．改进的模糊聚类方法把模糊模型结构辨识和参数辨识融为一体．首先，通过新的目标函数的最小化确定模糊模型的输入空间，即确定模糊规则和规则数、参数．然后对经模糊聚类得到的模糊前件推理矩阵进行QR分解，通过分析秩亏损确定重要的聚类规则．为了证明这种建模方法的性能，对非线性系统进行了仿真建模研究，仿真结果证明所提出方法是一种有效的、精确的模糊建模方法.

Fuzzy modeling of nonlinear system with matrices decomposition based on objective function

The modeling method is proposed by fuzzy clustering based on a new objective function. The improved fuzzy clustering can blend the integral body with the structure identification and parameter identification of fuzzy model. First, the input space of fuzzy model, rules and parameters are confirmed by minimizing the new objective function. Then fuzzy inference matrix confirmed by using the proposed fuzzy clustering is decomposed by using matrix decomposition of QR. According to the analyses of the rank loss of matrix, the important cluster is confirmed. To demonstrate the performance of this modeling method, simulation example in a non-linear system is studied. The results show that the proposed method provides effective and accurate fuzzy modeling method.