五传感器配置的MW-GLRT故障检测算法  

MW-GLRT fault detection algorithm based on five-sensor configuration

作　　者：黄帅[1] 刘海颖[2] 华冰[2] 

机构地区：[1]南京航空航天大学自动化学院,江苏南京210016 [2]南京航空航天大学高新技术研究院,江苏南京210016

出　　处：《传感器与微系统》2013年第10期121-125,共5页Transducer and Microsystem Technologies

基　　金：江苏省自然科学基金资助项目(KB2011729);南京航空航天大学自主科研专项基金资助项目(NS2013092)

摘　　要：针对多传感器的冗余配置的导航组件故障检测问题,广义似然比检测(GLRT)法是一种有效的方法。但该方法对于特征向量的提取不够精确,在故障检测中经常会出现误警和漏检的现象,并且随着采样时间的累积,样本奇偶向量对故障的敏感程度下降。针对五传感器配置系统,采用了移动窗口GLRT(MW-GLRT)方法,并结合信息贡献率法,解决了一般GLRT方法的算法缺陷。分别用阶跃故障和斜坡故障2种模式,对算法进行了仿真验证,结果表明:改进的MW-GLRT方法相比一般GLRT方法性能得到很大提高。Aiming at problem of fault detection of navigation components of multi-sensors redundant configuration,generalized likelihood ratio test （GLRT）is an effective fault detection method. However, as its extraction on feature vector is not so accurate, false alarm and missing often occur in fault diagnosis, and with accumulation of sampling time, sensitivity of the parity vector of sample to fault becomes weak. Aiming at fivesensor configuration system, use moving window-GLRT （MW-GLRT） method, combine with information contribution rate method to overcome the shortcoming of GLRT method. Step fault and slope fault are used, and simulation verification is carried out. The simulation results show that the detection performance is greatly improved by using this method.

关 键 词：传感器冗余 故障检测 奇偶校验 移动窗口法 信息贡献率 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]