A satellite fault detection system based on telemetry data using statistical process control and time-domain feature extraction
Varsha Parthasarathy, Sajad Saraygord Afshari, Philip Ferguson
Abstract
• Introduction of a novel fault management system that synergistically integrates Statistical Process Control (SPC) and time-domain methods, enhancing satellite fault detection capabilities. • Demonstration of the integrated approach’s effectiveness using the power subsystem of ManitobaSat-1, a student-led CubeSat mission, successfully diagnosing three critical faults. • The proposed system offers a holistic view of spacecraft health by simultaneously capturing statistical variabilities and temporal deviations in system parameters, improving both reliability and accuracy. • Computational efficiency: Unlike traditional machine learning algorithms, the new system is based on mathematical and statistical algorithms that are less resource-intensive, making it suitable for real-time operations and smaller missions. • The model is designed to be scalable and universally applicable, thereby extending its utility from student-led initiatives like ManitobaSat-1 to potentially larger commercial and scientific satellite missions. In spacecraft operations, accurately detecting anomalies in telemetry is essential but often requires complex, time-consuming methods. With the growing number of low-earth orbit missions, there is an urgent need to streamline this process. In this paper, we introduce an efficient real-time fault detection system that specifically addresses three critical faults within a spacecraft’s power subsystem: loss of solar string(s), increase in the battery’s internal resistance, and excessive power consumption. We apply industrial statistical process control and time-domain feature extraction techniques to create algorithms for enhanced fault detection. Our approach involves extensive simulations using a dynamic model of the power subsystem, allowing us to develop a method that is both innovative and practical. This research represents a step forward in the field, as we utilize statistical process control for real-time health monitoring of spacecraft, providing a more efficient and accurate means of analysis.