Print Email Facebook Twitter Application of Deep Learning for Spacecraft Fault Detection and Isolation Title Application of Deep Learning for Spacecraft Fault Detection and Isolation Author Voss, Sander (TU Delft Aerospace Engineering) Contributor Guo, Jian (mentor) Degree granting institution Delft University of Technology Programme Aerospace Engineering Date 2019-11-20 Abstract Spacecraft require high availability, autonomous operation, and a high degree of mission success. Spacecraft use sensors, such as star trackers and GPS, and actuators, such as reaction wheels, to reach and maintain a correct attitude and position. Failures in these components will have a significant negative impact on the success of the mission, or may even cause total loss of mission. Fault Detection, Isolation and Recovery (FDIR) aims to detect and isolates faults and recover them before they develop into failures. This makes it an important factor in the success of a satellite’s mission. It is also a determining factor in the level of autonomy if a system does not require ground intervention to perform FDIR. Development of FDIR methods is a difficult task, of which the success depends largely on the knowledge of the system and suffers under noisy environments. This research aims to explore the use of Deep Learning for fault detection and isolation in spacecraft. In a case study the proposed method is used to detect and isolate reaction wheel, GPS, Star Tracker, and magnetometer faults as well as two simultaneous faults. The research suggest successful classification of faults in the star trackers, GPS and magnetometers but a lack in performance in misalignment faults. Tachometer faults are often not isolated to the correct wheel. There is a high degree of false alarms and missed detection and preliminary results suggest separating detection and isolation may resolve this. Dataset size has also been shown to be a large contributor to the accuracy and above all loss performance of the network. Subject FDIFDIRFault DetectionFault IsolationDeep LearningRecurrent networksRecurrent Neural Networklong short-term memory networksLSTM To reference this document use: http://resolver.tudelft.nl/uuid:7c308a4b-f97b-4a83-b739-4019ad306853 Part of collection Student theses Document type master thesis Rights © 2019 Sander Voss Files PDF SanderVossMscThesisReport.pdf 13.07 MB Close viewer /islandora/object/uuid:7c308a4b-f97b-4a83-b739-4019ad306853/datastream/OBJ/view