摘要
针对航天器动态特性分析过程中遇到的建模过程复杂、处理速度较慢、分析结果不易进行直观解释等问题,通过借鉴符号化近似思想,提出了一种基于时序形态的航天器动态模式提取方法。该方法首先采用分段线性化算法获取一系列分割线段以近似表示遥测序列形态,然后使用系统聚类分析技术对分割线段进行分类,最后基于分类结果将遥测时间序列数据转化为蕴含形态信息的符号序列。验证结果表明该方法简单、有效,所提取的动态模式符号序列不仅可体现遥测时序数据的显著变化,也可表征遥测序列的细微变化,可作为遥测序列识别与异常检测的依据。
To solve problems such as complex modeling procedure, low processing rate and no intuitional results while analyzing spacecraft dynamic characteristics, a method of spacecraft dynamic mode extraction based on telem- etry time series shape pattern is proposed in this paper. With reference to the thought of Symbolic ApproXimation (SAX) method, the method features acquisition of a series of split lines to represent telemetry shape pattern ap- proximately by using Piecewise Linear Representation (PLR) algorithm at first. Then, Hierachical Cluster Analysis (HCA) technology is applied to classify split lines. Finally, based on classification results, telemetry time series are converted to symbol sequence containing shape pattern information. Tests show that the method proposed is simple and effective. The symbol sequence extracted can represent both marked changes and slight changes of time series, which can be used as a basis of series recognition and anomaly detection.
出处
《飞行器测控学报》
CSCD
2016年第3期193-199,共7页
Journal of Spacecraft TT&C Technology
关键词
航天器
时间序列
动态模式
分段线性化
聚类分析
符号序列
spacecraft
time series
dynamic mode
Piecewise Linear Representation (PLR)
cluster analysis
symbol sequence