神经网络自适应非线性滤波器及在线故障检测  

Neural Networks-Based Nonlinear Adaptive Filters and On-line Fault Detection

作　　者：董选明[1] 裘丽华[1] 王占林[1] 

机构地区：[1]北京航空航天大学自动控制系,100083

出　　处：《控制与决策》1997年第1期78-82,共5页Control and Decision

摘　　要：提出一种BP神经网络在线自学习方式及用该方式建立自适应非线性滤波器的方法,并提出了基于该BP滤波器对机器人系统进行在线实时状态跟踪、预报和突变性故障检测的方法。为了快速、正确地检测出故障,设计了一复合故障检测策略。对所提方式、方法及策略进行了仿真验证。The paper presents a on -line self- learning scheme and a novel method for constructing BP neural networks - based nonlinear adaptive filters (BPNNNAFs) using the scheme. A method based on the BPNNNAFs tor on - line real-time tracking, predicting the states and detecting jump faults of a robot system is also proposed. To detect repidly and correctly the faults, a compound decision for detecting faults is designed. Finally simulation experiments are given on the above scheme. methods and detecting decision proposed.

关 键 词：BP神经网络 非线性系统 滤波器 在线故障检测 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置] TP18[自动化与计算机技术—控制科学与工程]