基于SCIM的航天器遥测数据中心序列挖掘

Sequence Mining of Spacecraft Telemetry Data Center Based on SCIM

摘要: 随着航天技术的飞速发展，航天器遥测数据量呈爆炸式增长，对这些海量、高维、复杂的时间序列数据进行高效分析与挖掘，对于航天器健康状态监测、故障诊断及运行模式识别具有至关重要的意义。本文针对航天器遥测数据中存在的特点，提出了一种基于分段质心迭代法(Segment Centroid Iteration Method, SCIM)的中心序列挖掘方法。该方法首先通过特征表示和层次聚类对遥测时间序列进行预处理，有效减少了不同工作模式序列之间的相互影响；随后，通过迭代优化过程，利用序列段匹配和质心合并策略，实现了中心序列的快速近似求解。实验结果表明，SCIM方法在保留原始序列主要特征的同时，有效提升了中心序列计算的效率和表征能力。本研究为航天器遥测数据的深度分析提供了新的思路和技术支持。

Abstract: With the rapid development of aerospace technology, the volume of spacecraft telemetry data has been growing explosively. Efficient analysis and mining of these massive, high-dimensional, and complex time series data are crucial for spacecraft health monitoring, fault diagnosis, and operational mode identification. This paper proposes a Segment Centroid Iteration Method (SCIM) for central sequence mining, specifically addressing the characteristics of spacecraft telemetry data. The method first preprocesses telemetry time series through feature representation and hierarchical clustering, effectively reducing interference between sequences from different operating modes. Subsequently, an iterative optimization process, utilizing segment-based matching and centroid merging strategies, achieves rapid approximate solutions for central sequences. Experimental results demonstrate that the SCIM method retains the main features of the original sequences while significantly improving the efficiency and representational capability of central sequence computation. This research provides new insights and technical support for in-depth analysis of spacecraft telemetry data.