卫星长期运行过程中积累的大量遥测数据蕴涵有宝贵的卫星状态变化信息,挖掘变化特征对发现卫星状态异常变化和诊断卫星运行过程故障具有重要意义.围绕在轨卫星遥测数据分析以及遥测的故障诊断相似性查找问题,提出了一种基于逐段回归近似(PRA)的卫星遥测时序数据相似性挖掘算法.算法首先通过对卫星遥测数据挖掘实现数据的降维处理,以减少数据量；其次通过建立多维空间索引树,并在索引树中通过相似阈值判断进行查找,以获得给定时序数据的相似序列.理论分析和仿真实测卫星电源分系统的时序遥测数据实验表明,该算法对遥测数据序列可有效地进行降维处理,并保持数据趋势性；建立的相似序列为提高在轨卫星数据分析和故障诊断的有效性和正确性提供了依据.%During the long operational time of a satellite,there is a large number of telemetry data which contain valuable information of satellite status change.Therefore,it is of great significance to mining the variation characteristics for detecting abnormal variations and performing fault diagnosis of a satellite.Aiming at the problems of in-or-bit satellite telemetry data analysis and the fault diagnosis similarity searching,this paper described a mining algorithm of telemetry time series similarity based on Piecewise Regression Algorithm (PRA).Firstly,this algorithm was used to realize the dimension reduction through mining satellite telemetry data to reduce the data volume.And then the index tree of multidimensional space was established.The time sequence data of similar sequences were obtained through the seeking of the similarity threshold detection.The theoretical analysis and the simulation test of satellite power subsystem telemetry data show that the algorithm can effectively reduce the dimension and keep the data trend.The establishment of the similar sequence proveds the basis for satellite orbit data analysis and fault diagnosis.
展开▼
