Research on fault detection and principal component analysis for spacecraft feature extraction based on kernel methods

Abstract

Satellite anomaly is a process of evolution. Detecting this evolution and the underlying feature changes is critical to satellite health prediction, fault early warning, and response. Analyzing the correlation between telemetry parameters is more convincing than detecting single-point anomalies. In this article, principal component analysis method was adopted to downscale the multivariate probability model, T2{T}^{2} statistic was checked to determine the data anomaly, without the trouble of threshold setting. After an anomaly was detected, time-domain visualization and dimension reduction methods were introduced to visualize the satellite anomaly evolution, where the dimensions of telemetry or features were reduced and presented in two- or three-dimensional coordinates. Engineering practice shows that this method facilitates the early detection of satellite anomalies, and helps ground operators to respond in the early stages of an anomaly.

Keywords

• kernel principal component analysis
• pca dimension reduction
• feature extraction
• visualization