基于改进BP神经网络的复杂系统故障识别  

Complex System Fault Detection Based on Improved BP Network

作　　者：杜玮[1] 刘泉[1] 李景松[1] 

机构地区：[1]武汉理工大学信息工程学院,武汉430070

出　　处：《武汉理工大学学报》2011年第6期134-138,共5页Journal of Wuhan University of Technology

基　　金：国家自然科学基金(50935005;50775167)

摘　　要：复杂系统的故障种类多样,成因复杂,依靠传统的数学建模方式对复杂系统的故障进行识别和研究比较困难。研究了BP网络的非线性逼近能力和多分类能力,在此基础上分析了BP网络的设计方法和存在的缺陷,提出了一种基于变学习速率法与共轭梯度法相结合的BP网络性能改进算法,将其用于复杂系统的故障进行识别并进行了实验验证。实验的结果表明,改进后的BP网络缩短了训练时间,提高了故障识别的准确率,增强了网络的泛化能力,取得了良好的效果。There are different types of faults in a complex system,where causes of the faults are various.Thus,it is difficult to identify and study them relied on the traditional method of mathematical modeling for complex fault system.To solve that problem,this paper introduced an improved BP network method based on the variable learning rate and the conjugate gradient method which combines the nonlinear approximation ability and multi-classification ability of BP network.The method was tested by experiments,which demonstrated well that it can improve the accuracy of fault identification and enhance the network＇s ability of generalization.

关 键 词：BP网络 复杂系统 故障识别 变速率学习 共轭梯度法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]