Е-Публикации
Технически университет - София

Abstract: The increasing complexity of nuclear reactor operations and the growing availability of heterogeneous operational data have motivated the development of advanced artificial intelligence (AI) and digital twin frameworks for enhanced monitoring, diagnosis, and decision support. However, existing approaches remain limited by static modeling assumptions, post-hoc uncertainty quantification, and externally applied explainability mechanisms, which restrict their applicability in safety-critical environments. This paper proposes a conceptual adaptive digital twin framework for nuclear reactor operations that integrates physics-informed machine learning, meta-learning-inspired adaptation, embedded uncertainty quantification, and built-in explainable AI within a unified closed-loop architecture. The framework combines real-time sensor data, simulation outputs, and operational records through a hybrid modeling layer that supports both physical consistency and data-driven inference. Unlike conventional static digital twins or task-specific machine learning models, the proposed system is designed to continuously adapt to evolving reactor conditions, including distribution shifts and rare operational events. Uncertaintyaware prediction and integrated explainability are embedded directly into the inference pipeline, enabling transparent and reliable decision support. The proposed framework provides a structured foundation for next-generation nuclear digital twins with potential applications in anomaly detection, predictive maintenance, and operational optimization, while aligning with safety, reliability, and regulatory requirements in nuclear engineering systems.

References

1. Q. Huang, A review of the application of artificial intelligence to nuclear reactors: Where we are and what's next, Heliyon, 2023. https://www.sciencedirect.com/science/article/pii/S240584402301090 3.
2. International Atomic Energy Agency (IAEA), Considerations for Deploying Artificial Intelli-gence Applications in the Nuclear Power Industry, IAEA Technical Document, 2025. https://www.iaea.org/publications/15866/ considerations-for-deploying-artificial-intelligence-applications-in-the-nuclear-power-industry.
3. OECD Nuclear Energy Agency (NEA), Benchmark on Artificial Intelligence and Machine Learning for Scientific Computing in Nuclear Engineering, Working Paper, 2024. https://www.oecdnea.org/upload/docs/application/pdf/2024-01/nea_wkp_2023_1_2024-01-18_16-23-40_237.pdf.
4. J. Daniell, K. Kobayashi, A. Alajo, S. B. Alam, Digital Twin-Centered Hybrid Data-Driven Multi-Stage Deep Learning Framework for Enhanced Nuclear Reactor Power Prediction, En-ergy and AI, Volume 19, 2025, 100450, ISSN 2666-5468, https://doi.org/10.1016/j.egyai.2024.100450.
5. K. Kobayashi et al., Deep neural operator-driven real-time inference to enable digital twins for reactor power prediction, PMC, 2024. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10808608/.
6. Bei, X., Cheng, M., Zuo, X., Yu, K., Dai, Y. (2023). Surrogate Models Based on Back-Propagation Neural Network for Parameters Prediction of the PWR Core. In: Liu, C. (eds) Proceedings of the 23rd Pacific Basin Nuclear Conference, Volume 2. Springer Proceedings in Physics, vol 284. Springer, Singapore. https://doi.org/10.1007/978-981-19-8780-9_107
7. M. D. Wang et al., Abnormal event detection, identification and isolation in nuclear systems, 2021. https://www.sciencedirect.com/science/article/abs/pii/S014919702100 2882.
8. Z. Chen, Research on Fault Prediction of Nuclear Safety-Class Signal Conditioning Module, Energies, 2024. https://www.mdpi.com/19961073/17/16/4063.
9. G. Ioannou / T. Tasakos et al., Intelligent techniques for anomaly detection, EPJ/PHYSOR conference paper, 2021. https://www.epjconferences.org/articles/epjconf/pdf/2021/01/epjconf_physor2020_21 011.pdf.
10. A. Chaudhary et al., Anomaly Detection and Analysis in Nuclear Power Plants, Electronics, 2024. https://www.mdpi.com/20799292/13/22/4428.
11. Z. Guo et al., Defect detection of nuclear fuel assembly based on deep neural network, Annals of Nuclear Energy, 2020. https://doi.org/10.1016/j.anucene.2019.107078.
12. C. He et al., A data-driven adaptive fault diagnosis methodology based on NSGAII-CNN, Annals of Nuclear Energy, 2021. https://doi.org/10.1016/j.anucene.2021.108326.
13. A. Deleplace et al., Ensemble Learning-based Fault Detection in Nuclear Systems, 2020. Sci-enceDirect, https://www.sciencedirect.com/science/article/pii/S240589632033536 9.
14. M.H. Saad, Said, A. Machine learning-based fault diagnosis for research nuclear reactor me-dium voltage power cables in fraction Fourier domain. Electr Eng 105, 25–42, 2023. https://doi.org/10.1007/s00202-022-01649-7.
15. E. Kautz, A. Hagen, J. Johns, D. Burkes, A machine learning approach to thermal conductivi-ty modeling: A case study on irradiated uranium-molybdenum nuclear fuels, OSTI, 2019. https://arxiv.org/abs/1901.00722https://www.osti.gov/pages/biblio/1515844.

Issue