Overview

• The newly developed Deep Operator Neural Networks (DeepONet) enable real-time monitoring of nuclear reactors, overcoming the limitations of physical sensors in harsh environments.
• The AI-driven virtual sensors predict critical thermal-hydraulic parameters at speeds 1,400 times faster than traditional Computational Fluid Dynamics (CFD) simulations.
• This breakthrough addresses long-standing challenges in nuclear energy safety, allowing early detection of system degradation and potential failures.
• The research heavily utilized high-performance computing resources, including NCSA's Delta and NVIDIA A100 GPUs, to train and deploy the models efficiently.
• Interdisciplinary collaboration between nuclear engineers, AI specialists, and HPC experts was crucial to achieving this transformative advancement in reactor monitoring.