基于BP神经网络的模糊参数辨识  

Fuzzy parameters identification based on BP neural network

作　　者：韦卫星[1] 磨少清[1] 覃春芳[1] 廖义奎[1] 文勇[1] 

机构地区：[1]广西民族大学物理与电子工程学院,南宁530006

出　　处：《计算机工程与应用》2008年第18期44-47,131,共5页Computer Engineering and Applications

基　　金：广西自然科学基金(the Natural Science Foundation of Guangxi of China under Grant No.0575031)

摘　　要：提出一种与TSK模糊模型相似的模糊模型—M-2模型,证明了M-2模型与一个4层前向神经网络是等价的,在此基础上提出基于BP神经网络的模糊模型参数辨别算法,即通过BP神经网络对样本数据的学习,直接从样本数据获取模型参数,建立M-2模糊模型,通过仿真实例验证了该算法的有效性。A fuzzy model called M-2 model which is similar to TSK fuzzy model is proposed.The mathematical equivalence between M-2 model and a 4 layer feedforward neural network is verified.Then a new algorithm to identify parameters of fuzzy model based on BP neural network is put forward.All parameters required for establishing M-2 fuzzy model are extracted directly from a BP neural network trained by sample data.A simulation example is given to demonstrate the effectiveness of the algorithm.

关 键 词：TSK模糊模型 BP神经网络 模糊规则库 模糊参数辨识 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]