63 巻 (2015) 4 号 p. 119-128
We propose using dimensionality reduction techniques for analyzing House-Keeping (HK) data of satellites for condition inspection of faults. HK data are high-dimensional data and recorded for a long time, which makes it difficult for analyzers, especially when the persons are not specialists of the satellite. Utilizing hidden data structures and power of computers, dimensionality reduction techniques can be used for visualizing and classification of high-dimensional HK data and that provides useful information. In this paper we treated image degradation problem occurring in an ultra-small satellite XI-V, which is under operation in space. We showed that dimensionality reduction techniques are actually effective in the real spacecraft problem.