AI-Driven Complexity Reduction and Multi-Physics Digital Twins: From Theory to Engineering Implementation in Nuclear Reactors

Résumé 

To meet the rigorous demands of Best-Estimate Plus Uncertainty (BEPU) in modern nuclear engineering, it is essential to characterize safety margins and system dynamics with both high fidelity and high efficiency. Building upon foundational complexity reduction methods—such as the Generalized Empirical Interpolation Method (GEIM) and Reduced Basis methods—this talk presents the recent advancements in applying these mathematical tools to real-world nuclear engineering practices.

By integrating Model Order Reduction (ROM) with Artificial Intelligence (AI) and Data Assimilation, we have developed a data-enabled, physics-informed digital twin framework. This approach effectively resolves high-dimensional multi-physics coupling problems and allows for ultra-real-time state estimation and parameter identification. Furthermore, the presentation will highlight the engineering implementation of these methodologies, demonstrating how theoretical reduced-order models are deployed into industrial software and platform architectures (e.g., AI-Enhanced Digital Twin Engineering Platform) for the online monitoring and predictive simulation of commercial nuclear reactor cores.

Helin Gong

Helin Gong is an Associate Professor and Doctoral Supervisor at the Paris Elite Institute of Technology (SPEIT), Shanghai Jiao Tong University (SJTU). He holds a Bachelor's degree in Nuclear Engineering and Technology from Tsinghua University, a Master's degree in Nuclear Energy Science and Engineering from the Nuclear Power Institute of China (NPIC)—where he also served as a Senior Engineer from 2010 to 2022—, and a Ph.D. in Mathematics from Sorbonne University. Dr. Gong’s research focuses on complex system engineering for advanced nuclear energy systems. He has made systematic academic contributions by integrating Artificial Intelligence (AI), Reduced Basis (RB) methods, and Data Assimilation (DA) to build real-time digital twins and intelligent support technologies. His work systematically advances physics-informed modeling and specialized software development, providing critical solutions for the safety, economics, and operational flexibility of next-generation nuclear systems. He has served as a principal investigator for numerous high-level research grants, authored over 60 peer-reviewed papers, and received multiple prestigious science and technology awards.