基于神经网络在线建模的非线性动态系统中传感器故障检测方法

本文提出一种基于神经网络在线建模的动态非线性系统中传感器故障检测方法，它首先利用神经网络在线建立动态非线性系统的超前一步预测模型，然后利用神经网络对传感器的预测输出和传感器实际输出之差与一预定阈值比较以检测传感器故障。本文的优点是可以检测多个传感器故障，同时由于采用在线学习方式，非常适于航天器自主系统传感器故障检测的需要。此外，故障检测阈值的选取也比较简单。为了验证本文方法，仿真了一控制系统中同时发生漂移故障的两个传感器故障检测过程。结果表明，方法十分有效。

MULTIPLE SENSOR FAILURE DETECTION IN NONLINEAR SYSTEM BASED ON SYSTEM IDENTIFICATION METHOD USING ON-LINE LEARNING NEURAL NETWORK

This paper presents an innovative approach for multiple sensor failure detection in a nonlinear system based on system identification using an on line learning neural network The approach builds an one step ahead prediction model for the nonlinear system using an on line learning neural network firstly,and then the discprepancy between the on line estimations of the sensors using the built neural system model and the actual values of them is compared with a predetermined threshold to detect sensor failures One advantage of the approach is the ability to detect multiple sensor failures Another advantage is that it may be suitably used in autonomous systems with the neural network learning and working on line The method is proved to be very effective by a simulation result of detecting two sensor failures for a space robot system