非线性系统的模糊神经网络辨识方法

On Advantages of Fuzzy Neural Network Modelling of Nonlinear System

根据一种模糊动力学模型，提出一种非线性系统的模糊神经网络辨识方法。这种模型具有与线性系统ＤＡＲＭＡ模型类似的结构，证明了辨识算法的收敛性。最后结合实例进行了仿真。

The main advantage of fuzzy neural network modelling is that convergence, essential in any analysis of nonlinear system, can be proved. Eq.(1) describes a fuzzy neural network model for nonlinear system. The network weights denote the fuzzy subset parameters and the network runs in compliance with fuzzy inference. With the help of the gradient method, we propose eq.(5) as the learning algorithm for making the fuzzy subset parameters approach their correct values. At the end of section 3 of this paper, we give a theorem that stipulates the necessary conditions that ensure the convergence of subset parameters to correct values and ensure the satisfaction of two useful properties, inequality (Ⅰ) and eq.(Ⅱ). Simulation results shown in Figs.1 and 2 show the effectiveness of our algorithm.