Mining diagnostic knowledge from spacecraft data based on Spark cluster

• Published in: IEEE Conference on Industrial Electronics and Applications (Online) pp. 1762 - 1767
• Main Authors: Wang, Haoran, Yu, Jinsong, Tang, Diyin, Han, Danyang, Tian, Limei, Dai, Jing
• Format: Conference Proceeding
• Language: English
• Published: IEEE 09.11.2020
• Subjects: association rules; big data mining; Classification algorithms; Clustering algorithms; Data mining; distributed computing; Knowledge based systems; Software algorithms; Space vehicles; Spark cluster; Telemetry; Telemetry data
• ISSN: 2158-2297
• DOI: 10.1109/ICIEA48937.2020.9248423

Compared with the data obtained from ground simulation experiments, the spacecraft telemetry data can better reflect the real working state of the spacecraft. How to effectively utilize telemetry data and extract effective information is an important issue. This paper uses diagnostic data from real spacecraft telemetry to mine diagnostic knowledge and build a diagnostic knowledge base. Compared with the traditional fault diagnosis method based on expert knowledge, the diagnostic knowledge mined can enrich the existing expert knowledge base. In this paper, the FP-Growth algorithm is used to mine the association rules of the parameters to obtain the diagnostic knowledge, and a satellite telemetry data diagnosis knowledge base is constructed. Mined diagnostic knowledge includes association rules among parameters, and the relationship between parameters and faults. In addition, due to the large number of telemetry parameters, the amount of data reaching TB level, the Spark distributed computing cluster is used to implement distributed and efficient computing of the algorithm. Finally, building a spacecraft telemetry data mining diagnostic platform with the Django architecture.