Leveraging Machine Learning Techniques for Health Monitoring of Nuclear Systems: A Case Study Using Real Plant Data from Surry Power Station
Conference
Mohamed, A, Tansel, I, Tosunoglu, S. (2025). Leveraging Machine Learning Techniques for Health Monitoring of Nuclear Systems: A Case Study Using Real Plant Data from Surry Power Station
. 592-601. 10.13182/NPICHMIT25-46834
Mohamed, A, Tansel, I, Tosunoglu, S. (2025). Leveraging Machine Learning Techniques for Health Monitoring of Nuclear Systems: A Case Study Using Real Plant Data from Surry Power Station
. 592-601. 10.13182/NPICHMIT25-46834
In recent years, there has been a growing focus on artificial intelligence technology in general, and specifically within the nuclear industry. Efforts have been made to leverage AI and related technologies such as machine learning, digital twins, and neural networks, to enhance safety and reduce costs for both traditional and advanced nuclear reactors. These technologies offer significant potential for improving the efficiency and reliability of nuclear systems by enabling more sophisticated monitoring and control mechanisms. This paper introduces a novel method for incorporating machine learning into the online monitoring processes of nuclear systems, aim-ing to improve safety by increasing efficiency and automation. The proposed tool employs a combination of supervised learning (Counter-propagation Artificial Neural Networks, Supervised Kohonen Networks, and XY-Fuse) and unsupervised learning (self-organizing maps) in different phases. By integrating these advanced techniques, engineers can better detect anomalies, pre-dict potential failures, and optimize operational parameters in real-time. Real plant data from the Surry Power Station were utilized to demonstrate and validate the effectiveness of the proposed method. The dimensionality of the system data was reduced, and distinct clusters were identified, allowing for more precise monitoring and analysis. The algorithm successfully identified labels in the supervised learning phase to an acceptable level, proving the method's efficacy. Importantly, while leveraging these cutting-edge technologies, the approach still keeps human experience cen-tral to the system monitoring process. By combining human expertise with the benefits of new technology, the proposed method ensures that critical decisions are informed by both advanced analytics and human judgment, enhancing overall safety and reliability.
