模糊基函数网络辨识算法稳定性

Identification Stability on Fuzzy Basis Function Network

模糊基函数网络能很好地以“if…then…”形式利用专家的知识，且以这种模型进行辨识，其辨识误差方程与辨识参数成线性关系，因而非常适用于实时辨识系统中．本文研究了这种网络模型辨识的持续激励条件和基于梯度法的辨识系统稳定性，给出了它们存在的充要条件．

Linguistic fuzzy IF-THEN rules can often be obtained from human experts who are familiar with the system under consideration. A fuzzy basis function relates to a linguistic fuzzy IF-THEN rule. In addtion, this fuzzy basis function expansion is linear in its adjustable parameters. Therefore, The fuzzy basis function network is fit to be used as an adaptive identifier. This paper deals with the identification stability. A persistently exciting condition must be satisfied in order to guarantee parameter convengence in the adaptive identifier. Moreover,a sufficient and necessary test between the persistently exciting condition and fuzzy basis functins is given.