基于主成分分析的在轨航天器故障诊断技术  

Application of Principal Component Analysis in Spacecraft Fault Diagnosis

作　　者：刘一帆 于忠良 吕建峰[2] LIU Yifan;YU Zhongliang;LYU Jianfeng(China Aerospace Science and Technology Corporation, Beijing 100190, China;Key Laboratory of Autonomous Intelligent Unmanned System, MITT, Harbin Institute of Technology, Harbin 150001, China)

机构地区：[1]中国航天科技集团公司五院总体部,北京100190 [2]哈尔滨工业大学自主智能无人系统工信部重点实验室,哈尔滨150001

出　　处：《兵器装备工程学报》2021年第S01期208-211,220,共5页Journal of Ordnance Equipment Engineering

基　　金：国家重点研发专项项目(2019YFB1312001)。

摘　　要：提出了一种基于主成分分析的多维遥测数据的故障诊断方法,主成分分析是非常经典的降维算法,在判断新的遥测数据是否为故障时,可以首先分析主成分,得到各个主成分方向的分量,判断是否为故障。通过对大量的航天器遥测数据的实验结果表明:采用本文所提出的方法提高了故障诊断效率。通过数据降维,简化并提高了故障诊断的识别速度,便于分析庞大的实验数据,快速从众多的实验数据中找到故障发生的原因,尤其适用于在轨航天器的故障检测。A fault diagnosis method of multi-dimensional telemetry data based on principal component analysis(PCA)was proposed.PCA is a classical dimensionality reduction algorithm.When judging whether the new telemetry data is a fault,the principal component can be analyzed first to get the components of each principal component direction to judge whether the data is a fault.The experimental results of a large number of spacecraft telemetry data show that the proposed method can improve the efficiency of fault diagnosis,and the identification speed of fault diagnosis can be simplified and improved through data dimensivity reduction.After dimensionality reduction,it is convenient to analyze huge experimental data and quickly find the cause of fault from numerous experimental data,which is especially suitable for the fault detection of spacecraft in orbit.

关 键 词：主成分分析 遥测数据 故障诊断 航天器 数据降维 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]