Toward Explainable AI in Spacecraft Health Monitoring: Comparative Benchmarking of Anomaly Detection Models and Open-Source Datasets
IEEE Access · 2025 · 13 · 209368-209398
DOI: 10.1109/access.2025.3639551 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2025. Verfasst von Hassam Ahmed Tahir und 5 weiteren Autor:innen. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 187 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: SmartSat CRC, Australian Government’s CRC Program.
Abstract (Originalsprache)
The growing autonomy and complexity of spacecraft systems demand intelligent, interpretable, and resilient onboard health monitoring. This paper presents a dual benchmarking study of anomaly detection methods and telemetry datasets, evaluating both classical and deep learning models across real spacecraft telemetry and synthetic data generated via the Basilik simulator. Extensive experiments assess model performance across diverse fault types, data distributions, and operational scenarios. Furthermore, the study systematically compares datasets in terms of anomaly coverage, fidelity, and operational relevance, providing critical insights into their suitability for training and validating machine learning models. While the ultimate objective is to develop a resilient explainable AI (XAI) system for anomaly detection, this study focuses on benchmarking representative anomaly detection models and datasets to identify robust and generalizable backbones suitable for future XAI integration, rather than presenting an end-to-end XAI implementation. Beyond performance evaluation, the study underscores the importance of integrating anomaly detection with engineering frameworks such as Model-Based Systems Engineering (MBSE) and Digital Twins to enhance fault traceability, state validation, and system-level interpretability. The findings highlight that effective deployment in safety-critical space environments requires not only high-performing models and representative data but also tight integration with formal systems engineering paradigms.
Förderung
- SmartSat CRC
- Australian Government’s CRC Program
Zitieren
APA
Tahir, H. A., Jabbarpour, M. R., Vo, B. Q., Kowalczyk, R., Barr, J., & Bessell, T. (2025). Toward Explainable AI in Spacecraft Health Monitoring: Comparative Benchmarking of Anomaly Detection Models and Open-Source Datasets. IEEE Access, 13, 209368-209398. https://doi.org/10.1109/access.2025.3639551
Vancouver
Tahir HA, Jabbarpour MR, Vo BQ, Kowalczyk R, Barr J, Bessell T. Toward Explainable AI in Spacecraft Health Monitoring: Comparative Benchmarking of Anomaly Detection Models and Open-Source Datasets. IEEE Access. 2025;13:209368-209398. doi:10.1109/access.2025.3639551
BibTeX
@article{Tahir2025,
author = {Tahir, Hassam Ahmed and Jabbarpour, Mohammad Reza and Vo, Bao Quoc and Kowalczyk, Ryszard and Barr, James and Bessell, Travis},
title = {Toward Explainable AI in Spacecraft Health Monitoring: Comparative Benchmarking of Anomaly Detection Models and Open-Source Datasets},
journal = {IEEE Access},
year = {2025},
volume = {13},
pages = {209368--209398},
doi = {10.1109/access.2025.3639551}
}
EUVIMED