传感器故障检测、分离与恢复的神经网络方法  

Sensor Fault Detection, Isolation, and Accommodation Using Recurrent Neural Networks

作　　者：刘勇[1] 沈毅[1] 胡恒章[1] 

机构地区：[1]哈尔滨工业大学控制工程系,327信箱哈尔滨150001

出　　处：《传感技术学报》1999年第4期310-315,共6页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金!(69674019);国家863高技术资助

摘　　要：传感器是测控系统不可缺少的部件，传感器数据的高可靠性是系统正常工作的重要保证．本文基于递归神经网络具有优良的动态系统建模能力和时间数据序列预报能力，融合时空信息，构造出具有传感器故障检测、分离和故障恢复能力的智能传感器系统，理论分析和仿真结果表明了所研究系统的优良性能．Sensor is one of the vital parts in monitoring systems and control systems, it is im portant for these systems to work with more reliable sensor-data. In this paper, Intelligent sensor systems with fault detection, isolation,and accommodation have been constructed based on recurrent neural networks which fuse spatio-temposral information of the sensors. Recur- rent neural networks hold the abilities of modeling systems and predicting time-series data with excellent performance. The results of simulation are given to show that the smart sensor sys- tem can validate the sensor values and recover the fault sensors' signals.

关 键 词：传感器 故障检测 神经网络 

分 类 号：TP212[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]