航空学报 > 2021, Vol. 42 Issue (6): 624376-624376   doi: 10.7527/S1000-6893.2020.24376

基于不确定传感器状态的机载系统多层故障诊断方法

石健1,2, 王少萍1,2, 罗雪松1   

  1. 1. 北京航空航天大学 自动化科学与电气工程学院, 北京 100191;
    2. 北京航空航天大学 宁波创新研究院, 宁波 315800
  • 收稿日期:2020-06-08 修回日期:2020-09-04 出版日期:2021-06-15 发布日期:1900-01-01
  • 通讯作者: 石健 E-mail:shijian@buaa.edu.cn
  • 基金资助:
    国家自然科学基金(51875014,51875015);专项基金(2017-V-0011)

Multi-layer fault diagnosis of airborne system based on sensor uncertainty

SHI Jian1,2, WANG Shaoping1,2, LUO Xuesong1   

  1. 1. School of Automation Science and Electrical Engineering, Beihang University, Beijing 100191, China;
    2. Ningbo Institute of Technology, Beihang University, Ningbo 315800, China
  • Received:2020-06-08 Revised:2020-09-04 Online:2021-06-15 Published:1900-01-01
  • Supported by:
    National Natural Science Foundation of China (51875014,51875015); Special Fund (2017-V-0011)

摘要: 准确的机载系统故障诊断是保证飞机安全飞行和实现经济效益最大化的重要途径。然而传感器受到内外部环境条件的影响而不可避免的存在检测状态的不确定性,因此基于单个传感器或局部区域传感器综合检测结果的方法难以完全保证故障诊断的有效性和正确性。针对飞机机载系统的结构和工作原理,充分利用系统中不同层级、不同区域传感器检测特征之间的关联关系,考虑单个传感器本身存在的不确定性,构建了传感器信息前向融合与反向校验相结合的分层诊断决策方法,实现了对系统状态和传感器状态的双重估计与更新,克服了单一传感器故障对系统诊断推理准确度的影响。该方法较传统故障诊断模型,不再依赖某一个或某一类传感器信息的绝对可靠,在实现系统级的准确故障诊断同时,还能判断具体某一传感器本身是否发生虚拟警。在飞机液压系统故障诊断案例中,新方法成功将系统故障诊断的虚警率降低了96%,传感器的不确定度降低了84%。

关键词: 传感器虚警, 传感器信息融合, 机载系统, 交叉校验, 贝叶斯网络

Abstract: Accurate fault diagnosis of airborne system is an important way to ensure safe aircraft flight and maximize economic benefits. However, due to impacts of internal and external environmental conditions, the sensor has detection uncertainties, and is thus difficult to guarantee the validity and correctness of fault diagnosis based on detection of single sensor or of local sensors. Based on the structure and working principle of the airborne system, this paper proposes a hierarchical diagnosis decision-making method based on a combination of evidence forward fusion and reverse verification. The correlation between the detection features of sensors in different levels and different regions in the airborne system, and the uncertainty of a single sensor itself is considered. Estimation of both of the system state and sensor state can be realized, and the influence of single sensor fault on the accuracy of system diagnosis can be overcome. Compared with the traditional fault diagnosis model, the proposed method no longer depends on the absolute reliability of one or a kind of sensor information. At the same time, it can realize the accurate fault diagnosis at the system level, and can also judge whether a specific sensor itself has virtual alarm. In the case of fault diagnosis of aircraft hydraulic system, it can be seen that with the method proposed, the false alarm rate of system fault diagnosis is successfully reduced by 96% and the sensor uncertainty by 84%.

Key words: sensor false alarm, evidence fusion, airborne system, cross check between layers, Bayesian network

中图分类号: