Russian Federation
Russian Federation
Russian Federation
The article discusses the urgent problem of increasing the reliability and autonomy of distributed space systems (RCS), such as satellite groupings. An approach to predictive analysis of the state of devices based on a hybrid model combining long-term short-term memory (LSTM) and graph neural networks (GNN) is proposed. This architecture is designed for the joint analysis of multidimensional time series of telemetry and structural relationships in the grouping. The results of experiments on synthesized and real data are presented, demonstrating the effectiveness of the model in predicting anomalies and potential failures, which makes it possible to switch from reactive to preventive control and reduce the load on the mission control center (MCC).
failure prediction, predictive analytics, hybrid neural networks, LSTM, graph neural networks (GNN), distributed systems, telemetry analysis, time series, technical diagnostics, machine learning
1. Komp'yuternoe modelirovanie rabotosposobnosti finishnyh procedur i parazitnyh elementov v programmno-apparatnom komplekse proektirovaniya mikroshem / A. S. Yagodkin, N. N. Litvinov, P. S. Ivanin, A. S. Groshev // Modelirovanie sistem i processov. – 2024. – T. 17, № 3. – S. 106-115. – DOIhttps://doi.org/10.12737/2219-0767-2024-104-113. – EDN NICLOY.
2. Andryuhin, E. V. Prognozirovanie sboev i otkazov v raspredelennyh sistemah upravleniya na osnove modeley prognozirovaniya vremennyh ryadov / E. V. Andryuhin, M. K. Ridli, D. I. Pravikov // Voprosy kiberbezopasnosti. – 2019. – № 3(31). – S. 24-32. – DOIhttps://doi.org/10.21681/2311-3456-2019-3-24-32. – EDN EMVKAP.
3. Say, V. K. Prognozirovanie otkazov slozhnyh mnogoob'ektnyh sistem na osnove kombinacii neyrosetey: puti povysheniya tochnosti prognozirovaniya / V. K. Say, M. V. Scherbakov // Prikaspiyskiy zhurnal: upravlenie i vysokie tehnologii. – 2020. – № 1(49). – S. 49-60. – DOIhttps://doi.org/10.21672/2074-1707.2020.49.4.049-060. – EDN QIXRXX.
4. Minimizaciya emkosti dekady induktivnogo delitelya napryazheniya / A. I. Zarevich, S. V. Murav'ev // Izvestiya Tomskogo politehnicheskogo universiteta. – 2012. – T. 321, № 4. – S. 140–143.
5. Matematicheskaya model' induktivnogo delitelya napryazheniya s elektronnoy kompensaciey / A. I. Zarevich, S. V. Murav'ev // Izvestiya Tomskogo politehnicheskogo universiteta. – 2010. – T. 317, № 4. – S. 129–133.
6. Relativistic magnetron with distributed output of microwave radiation / I. I. Vintizenko, A. I. Zarevich, S. S. Novikov // Technical Physics Letters. – 2005. – Vol. 31, № 5. – P. 388–390. DOI: https://doi.org/10.1134/1.1931776
7. Avtomatizirovannyy kompleks dlya diagnostiki impul'sov relyativistskogo SVCh-generatora / A. I. Zarevich, E. V. Vegner, I. I. Vintizenko // Pribory i tehnika eksperimenta. – 2004. – № 3. – S. 78–82.
8. Regressionnyy analiz ocenki zagruzki platnogo otdeleniya stomatologicheskoy polikliniki / T. P. Novikova, A. A. Bodin, A. I. Zarevich // Nauchno-tehnicheskiy vestnik Povolzh'ya. – 2023. – № 3. – S. 77–80.
9. Effect of the external coupling of cavities on the stability and output power of a relativistic magnetron / I. I. Vintizenko, V. I. Guselnikov, A. I. Zarevich, S. S. Novikov // Technical Physics Letters. – 2003. – Vol. 29, № 4. – P. 294–296. DOI: https://doi.org/10.1134/1.1573295
10. Sintez bol'shih dannyh i korrekciya disbalansa klassov dlya uluchsheniya modeli klassifikacii lesnyh pozharov / A. I. Zarevich, F. V. Makarenko, A. V. Poluektov // Lesnoe hozyaystvo v usloviyah global'nyh vyzovov: novye paradigmy ustoychivogo razvitiya. Forestry – 2025 : materialy Mezhdunar. les. foruma, priuroch. k prazdnovaniyu 95-letiya VGLTU. – Voronezh, 2025. – S. 773–785.
