Applied Enhanced IMS “Inductive Monitoring System” on Cubesat Housekeeping Telemetry using One-class SVM
Nowadays Satellites are included in our daily life activities for several purposes. Some of them used for Remote Sensing, communications, navigation, weather broadcast and deep space telescopes. Once the satellite is launched and enters the orbit, it operates autonomously. During the orbital flight, it will be exposed to a variety of mechanical, thermal and electromagnetic effects, which may have a direct impact on its operations. Therefore, Satellite health monitoring is essential for the Satellite mission success. As a satellite mission became more sophisticated, accordingly the satellite health monitoring became more challenging to be handled. So new trend used by the satellite operators to use a Data-Driven technique for Satellite health monitoring which invented by David L. Iverson which called IMS “Inductive Monitoring system”. IMS is using Applied Data Mining and Machine Learning techniques on the archived satellite telemetry data to build a knowledge based model and using this model to be applied on the newer telemetry data coming from the satellite to monitor the health of the satellite. IMS usually uses K-means algorithm for the learning Algorithm in this paper we are proposing One Class SVM Algorithm to be used instead. We applied both learning Algorithms on a real CubeSat Telemetry data and One Class SVM showed a magnificent improvement for the ability to capture faulty data from the CubeSat telemetry data.
Keywords - CubeSat, Machine Learning, Data Mining, Clustering, K-means, one class SVM.