This is an unedited manuscript accepted for publication and provided as an Article in Press for early access at the author’s request. The article will undergo copyediting, typesetting, and galley proof review before final publication. Please be aware that errors may be identified during production that could affect the content. All legal disclaimers of the journal apply.
Shweta Zambare,
Shivani Shendage,
Pratiksha Surve,
Nilophar Shaikh,
Vaibhav Godase,
- UG Student, Department of Electronics and Telecommunication Engineering, SKN Sinhgad College of Engineering, Pandharpur, Maharashtra, India
- UG Student, Department of Electronics and Telecommunication Engineering, SKN Sinhgad College of Engineering, Pandharpur, Maharashtra, India
- UG Student, Department of Electronics and Telecommunication Engineering, SKN Sinhgad College of Engineering, Pandharpur, Maharashtra, India
- UG Student, Department of Electronics and Telecommunication Engineering, SKN Sinhgad College of Engineering, Pandharpur, Maharashtra, India
- Assistant Professor, Department of Electronics and Telecommunication Engineering, SKN Sinhgad College of Engineering, Pandharpur, Maharashtra, India
Abstract
Reliable fault detection is an essential requirement for safe and stable nuclear reactor operation because abnormal variations in temperature, pressure, coolant flow, neutron flux, and other process variables can develop rapidly during transient conditions. Conventional threshold-based monitoring methods are effective for clearly defined abnormal states but may provide limited discrimination when several sensor parameters change simultaneously. This paper proposes a machine-learning-based framework for detecting and classifying nuclear reactor faults using multivariate process-sensor data. A simulated dataset containing 1,200 operating samples is considered, with process variables including reactor coolant temperature, primary pressure, coolant flow rate, neutron flux, steam-generator pressure, pump speed, radiation level, and control-rod position. Five operating conditions are considered: normal operation, coolant-flow fault, pump fault, pressure anomaly, and sensor fault. Random Forest, Support Vector Machine, Artificial Neural Network, and XGBoost models are developed and compared. An illustrative XGBoost model achieves an accuracy of 97.2%, precision of 97.0%, recall of 96.8%, and F1- score of 96.9%. Feature-importance analysis indicates that coolant flow rate, primary pressure, neutron flux, and coolant temperature contribute strongly to fault classification. An illustrative multi-objective optimization process is further applied to identify a reduced sensor subset while maintaining diagnostic performance. The proposed framework demonstrates the potential of machine learning for intelligent reactor condition monitoring and operator-support applications. However, the reported numerical results are illustrative and require validation using a qualified reactor simulator or verified operational dataset before practical deployment.
Keywords: Nuclear Reactor, Fault Detection, Machine Learning, Multivariate Sensors, Random Forest, Reactor Safety.
References
1. Godase V, Pawar P, Nagane S, Kumbhar S. Automatic railway horn system using node MCU. Journal of Control & Instrumentation. 2024 Jan;15(1).
2. Godase V, Godase J. Diet prediction and feature importance of gut microbiome using machine learning. Evolution in Electrical and Electronic Engineering. 2024 Nov 6;5(2):214-9.
3. Godase V. A comprehensive study of revolutionizing EV charging with solar-powered wireless solutions. Advance Research in Power Electronics and Devices e-ISSN. 2025 Apr 18:3048-7145.
4. Salunkhe A, Pawar V, Pise P, Mule S, Survase A, Godase V, Zambre S. A review on real-time RFID-based smart attendance systems for efficient record management. Advance Research in Analog and Digital Communications. 2025 Aug;2(2):32-46.
5. Lyu M, Gong H, Chen Z, Wang J, Zhong M, Wang Z, Li Q, Pan Z. Deep learning-based dual monitoring system for power forecasting and fault detection in nuclear power applications. Energy and AI. 2025 May 1;20:100515.
6. Hu G, Zhou T, Liu Q. Data-driven machine learning for fault detection and diagnosis in nuclear power plants: A review. Frontiers in Energy Research. 2021 May 12;9:663296.
7. Gaber MI, Nechaevskiy AV. Enhancement of anomaly detection system using artificial intelligence for sensor signals in nuclear power plants. Journal of Instrumentation. 2026 Aug 1;21(08):P08016.
8. Yadav NR, Biswas K, Tamboli PK, Singh LK, Duttagupta SP. Sensor fault detection and Diagnostics in Small Modular Reactor and Advanced Analytical Perspectives: A Review. IEEE Sensors Journal. 2026 Aug 24.
9. Xu Y, Cai Y, Song L. Condition assessment of nuclear power plant equipment based on machine learning methods: A review. Nuclear Technology. 2023 Jul 3;209(7):929-62.
10. Choi J, Lee SJ. RNN-based integrated system for real-time sensor fault detection and fault-informed accident diagnosis in nuclear power plant accidents. Nuclear Engineering and Technology. 2023 Mar 1;55(3):814-26.
11. Xue Y, Zhou W, Lv Y, Lv J, Hu B. Research on fault monitoring and intelligent diagnosis of key equipment in nuclear power plant driven by knowledge and data. In2025 5th international conference on machine learning and intelligent systems engineering (MLISE) 2025 Jun 13 (pp. 90-98). IEEE.
Journal of Nuclear Engineering & Technology
| Volume | 16 | |
| 03 | ||
| Received | 22/09/2026 | |
| Accepted | 23/09/2026 | |
| Published | 02/11/2026 | |
| Publication Time | 41 Days |