多变量系统模糊动态模型的辨识

介绍了一种新型的基于模糊神经网络的多变量模糊动态模型的辨识方法 ,该方法是通过将输入空间进行直接划分 ,而不是在输入空间的每一维上进行划分来得到模糊规则的。这样所形成的隶属函数为多维隶属函数 ,并使模糊规则的数目大为减少。在模糊聚类算法的基础上 ,提出了一个衡量聚类有效性的函数 ,以确定模糊规则的数目。以二级倒立摆系统为应用背景 ,取得了较好的辨识效果。

Identification of Dynamic Fuzzy Model for Multivariable System

In this paper, based on the fuzzy neural network, a new method of the dynamic fuzzy model identification for multivariable system is proposed. The fuzzy rule is based on the fuzzy partitions of input space directly rather than that of the each dimension of input space. Thus the membership functions are multidimensional ones and less number of rules is required. A function for measuring clustering validity based on the fuzzy clustering algorithm is defined with which the number of fuzzy rules can be determined. The method is proved that it has better effect through a double inverted pendulum by experiments.