航天试验数据时间序列相似度分析  

Time Series Similarity Analysis of Space Test Data

作　　者：夏心蕾 刘俊阳[1] 彭汉章 韩翔宇[1] Xia Xinlei;Liu Junyang;Peng Hanzhang;Han Xiangyu(Beijing Aerospace Automatic Control Institute,Beijing 100854,China)

机构地区：[1]北京航天自动控制研究所,北京100854

出　　处：《航天控制》2023年第5期55-61,共7页Aerospace Control

摘　　要：针对航天试验曲线数据的特点,提出一种时间序列相似度分析方法,通过分析待测曲线与模板曲线的相似性,辅助曲线数据判读。首先,采用SAX和DTW融合综合分析时间序列的周期性,将时间序列数据划分为周期和非周期两类。再分别叙述两种类型的时间序列的相似度分析方法,判定曲线的正确性。本文中对非周期时间序列,提出采用带有双维度约束的DTW相似度分析技术。对周期时间序列,采用离散傅立叶变换,通过频域幅度的一致性来近似初步判断周期信号的相似性。最后经过实验验证本文提出的时间序列相似度分析方法的正确性。A time series similarity analysis method for space test data curve is proposed in this paper.By analyzing the similarity between measured curve and template curve,it can assist curve data interpretation.Firstly,SAX and DTW are used to comprehensively analyze the periodicity of time series,so that the time se⁃ries are divided into periodic and aperiodic types.Then,in order to determine the correctness of the curve,the similarity analysis methods of the two types of time series are briefly described.Regarding aperiodic time series,the DTW distance similarity analysis technology with double dimensional constraint is proposed.Re⁃garding periodic time series,it is translated into the frequency domain by using Discrete Fourier Transform.The consistency of amplitude in the frequency domain is approximately used to judge the correctness of the periodic signal.Finally,the correctness of the proposed time series similarity analysis method is verified through experiments.

关 键 词：航天器 飞行数据 时间序列相似性 符号集合近似 DTW距离 

分 类 号：V557.2[航空宇航科学与技术—人机与环境工程]