一种多传感器故障诊断系统的设计  

Design of a fault Diagnose Scheme in Redundant-Sensors Systems

作　　者：黄帅[1] 刘海颖[2] 吴风喜 

机构地区：[1]南京航空航天大学自动化学院,江苏南京210016 [2]南京航空航天大学航天学院,江苏南京210016

出　　处：《计算机仿真》2014年第7期59-62,111,共5页Computer Simulation

基　　金：江苏省自然科学基金(KB2011729);中央高校自主科研专项基金(NS2013092)

摘　　要：在多传感器系统故障诊断优化问题的研究中,常用的奇偶校验法的缺陷是检测残差向量对于故障特征的提取能力随时间推移而下降,产生误警和漏检;在传感器数量不足时,只能判断系统的完好性而无法隔离故障传感器。针对上述问题,引入移动窗口及故障信息贡献率法弥补一般奇偶校验法的算法缺陷,同时采用单传感器独立诊断的算法,选择合理的小波基及小波分解层数,使信号对传感器故障的敏感程度区别于载体机动,实现故障传感器的隔离。上述方法互补,组成一个多传感器故障诊断系统。通过对阶跃和斜坡故障仿真表明,故障诊断的实时性、准确性显著提高。In the research of optimization in muhi-sensors system fault diagnosing, the common method, called parity check, has a great defect, the capability of parity residual vector to seize the characteristics of fault become weak as time goes by, this may cause false and missing alarm. When the redundant sensors are not enough, this method can only used to judge whether the system is intact. In order to solve this problem, the idea of moving-win- dow and rate of contribution to fault as well as the single-sensor diagnosing method are applied to the normal parity check method. By choosing the proper wavelet basis and the scale of wavelet decomposition, it makes the signals sen- sitive to the sensor faults rather than to the movements of the airframe. In this way, the faulty sensors can he isolated. These two methods are relatively complemented and comprise a multi-sensors fault diagnosing system. By using step bias and slope faults in simulation, the result shows that the system has better real-time performance and the accuracy is greatly improved.

关 键 词：奇偶向量 小波分析 传感器故障诊断 移动窗口法 

分 类 号：TH873.7[机械工程—仪器科学与技术]