TELEMETRY MONITORING BY DIMENSIONALITY REDUCTION AND LEARNING HIDDEN MARKOV MODEL

This paper proposes a data-driven health monitoring / anomaly detection method for spacecraft systems. Especially, we focus on some common properties spacecraft telemetry data has, such as highdimensionality, multi-modality and periodicity. The proposed method first monitors the static relationships among a number of variables contained in the telemetry by hybrid of clustering and dimensionality reduction. Then it monitors macroscopic state transition patterns of spacecraft systems using semi-Markov model. In the experiment, we applied this method to past artificial satellite data and verified its validity.