Distributed and Stacked Neural Network for Anomaly Detection in Small Satellites

TitleDistributed and Stacked Neural Network for Anomaly Detection in Small Satellites
Publication TypeConference Paper
Year of Publication2018
AuthorsSun, F., A. Dubey, C. Kulkarni, and A. Guarneros
Refereed DesignationRefereed
Conference Name15th Annual Cubesat Developer Workshop
Conference LocationSan Luis Obispo

The International Space Station (ISS) plans to launch 100+ small-sat missions for different science experiments in the next coming years. At present these missions are limited to couple of months but in the future these will last longer and it becomes crucial to monitor and predict future health of these systems as they age to prolong the usage time. This paper describes a hierarchical architecture which combines data-driven anomaly detection methods with a fine-grained model based diagnosis and prognostics architecture. At the core of the architecture is a distributed stack of deep neural network that detects and classifies the data traces from nearby satellites based on prior observations. Any identified anomaly is transmitted to the ground, which then uses model-based diagnosis and prognosis methods. In parallel, periodically the data traces from the satellites are transported to the ground and analyzed using model-based techniques. This data is then used to train the neural networks, which are run from ground systems and periodically updated. This collaborative architecture enables quick data-driven inference on the satellite and more intensive analysis on the ground where often time and power consumption are not key concerns. We demonstrate this architecture through an initial battery data set. In the future we propose to apply this framework to other electric and electronic components onboard the small satellites.