基于自适应模糊网络的在线辨识

研究了基于一阶Sugeno的自适应网络模糊推理系统（ANFIS）进行在线辨识的方法。给出了该自适应网络的结构，在此基础上给出了网络权值的修正算法，即综合最陡下降法和最小二乘法得到的一种混合学习算法。对一个非线性模型进行了数字仿真，得到的在线辨识的结果优于采用反传算法的普通神经网络辨识方法。由此证明，一阶Sugeno模糊推理模型和混合学习算法的采用，使得该辨识方法具备网络结构简单、收敛速度快的优势，便于工程实现。

On-line Identification of Control System Based on ANFIS

Nonlinear system modeling using adaptive-network-based fuzzy inference system (ANFIS) based on one degree Sugeno is discussed. The structure of ANFIS is proposed. Then a mixed learning arithmetic based on back promulgate and least-square arithmetic is presented to modify the network parameters. The simulation results show that the identification method can get better results than the common neural networks. Because this method adopts one degree Sugeno fuzzy reasoning structure and the mixed learning arithmetic, it has some predominance. Its simply structure and rapid learning speed can be applied to the project realizations.