卫星异常模式挖掘方法  

Method to Mine Satellite Abnormal Patterns

作　　者：徐宇航[1] 皮德常[1] 

机构地区：[1]南京航空航天大学计算机科学与技术学院,南京210016

出　　处：《小型微型计算机系统》2015年第9期1988-1992,共5页Journal of Chinese Computer Systems

基　　金：中央高校基本科研业务费(NZ2013306)资助;江苏省"青蓝工程";"333工程";技术基础预研项目(JSJC2013605C009)资助

摘　　要：及时发现在轨卫星的异常状态是卫星监控的一项重要工作,它不仅关系到卫星的正常使用,也影响卫星的寿命.为了能够及时发现卫星的故障,并且从中提取异常模式,提出一种基于Prefix Span算法的卫星异常模式挖掘方法.首先通过灰关联分析剔除冗余参数,然后利用各参数阈值提取异常数据集并且通过信息熵对各参数进行离散化,最后使用Prefix Span算法对异常数据集进行模式挖掘.该方法在数据量大、信息复杂的前提下,实现了对卫星异常模式的挖掘.通过对某卫星数据的实验分析,验证了所提方法的可行性和有效性.The effective and timely discovery of abnormal states of in-orbit satellites is an important task in satellite management,the abnormality not only influence the normal use of satellites,but also shorter the life expectancy of satellites. In order to discover the satellite failures promptly and mine the abnormal patterns,this paper proposes a method to mine abnormal patterns in satellites based on Prefix Span. Firstly,redundant parameters are eliminated through grey correlation analysis. Secondly,outlier dataset is extracted according to the threshold value of each parameter and each parameter is discretized via information entropy. Finally,Prefix Span method is employed to mine patterns in outlier dataset. The method effectively mined the abnormal patterns in satellites although the amount of data is very large and the information is of high complexity. M oreover,w e verify the feasibility and effectiveness of the proposed method through an experimental analysis of a certain satellite data.

关 键 词：在轨卫星 灰关联分析 信息熵 PREFIXSPAN 异常模式 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]