Title: Advanced approach in satellite failure detection and predicting using machine learning techniques: Alsat-1B case

Authors: Ali Kaddouri; Saiah Bekkar Djelloul Saiah

Addresses: Department of Missions and Space Systems, Satellite Development Centre, POS 50 Ilot T12 Bir El Djir, Oran, 31130, Algeria ' Department of Missions and Space Systems, Satellite Development Centre, POS 50 Ilot T12 Bir El Djir, Oran, 31130, Algeria

Abstract: In this paper, we propose a novel approach for detecting and predicting satellite failures using machine learning, with a focus on the Algerian satellite Alsat-1B. Analysing five years of event data, comprising over seven million occurrences and 3,000 event types, we evaluate four sequence-to-sequence prediction models and eight classification models. Our key contribution combines a Markov chain for sequence prediction and a logistic regression model for classification, proving highly effective with 97.7% accuracy, precision of 1.0, recall of 0.84, and an f-score of 0.91. This approach showcases the potential of intelligent systems in satellite control, underscoring the imperative for further exploration and development in this promising field.

Keywords: satellite operations; decision support system; failure detection; event prediction; deep learning.

DOI: 10.1504/IJSPACESE.2024.139356

International Journal of Space Science and Engineering, 2024 Vol.7 No.1, pp.12 - 31

Received: 11 Jul 2023
Accepted: 16 Oct 2023
Published online: 01 Jul 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article