The PUR-1 Cyber-Physical Digital Twin

The PUR-1 Digital Twin (DT) addresses the critical need for operational flexibility and responsiveness in nuclear systems. By creating a high-fidelity virtual replica of the Purdue University Reactor One (PUR-1), the authors developed a cyber-physical framework that maintains bidirectional synchronization with the physical reactor. This system supports advanced capabilities including real-time state estimation, anomaly detection, and predictive control, bridging the gap between static simulations and operational decision support.

The architecture is built on three layers: a data layer for signal processing, a modeling and simulation layer for physics-based and data-driven inference, and a visualization layer for operator interaction. The team coupled Monte Carlo neutron transport (OpenMC) and multiphysics thermal-hydraulic models with surrogate models (SMs) to overcome the computational bottlenecks of high-fidelity simulations. The system also incorporates a cyber-physical testbed (CERVEROS) to evaluate control actions and cybersecurity protocols, including quantum-key-distribution (QKD) for secure communication.

The PUR-1 DT successfully achieved full synchronization with the reactor at a 1 Hz frequency. Benchmarking against experimental gold foil activation and pool temperature measurements confirmed the accuracy of the physics-based models, with deviations below 10% for neutron flux mapping and approximately 2% for thermal fields. The integrated framework demonstrated an average predictive accuracy of 9.2% across full operational cycles, including startup, steady-state, and shutdown phases. The use of surrogate models allowed for real-time forecasting, effectively bypassing the long runtimes of traditional high-fidelity simulations while maintaining physical interpretability.

This work provides a validated, reference architecture for nuclear digital twins, moving beyond conceptual frameworks to a deployed, operational system. It demonstrates that hybrid modeling—combining physics-based rigor with AI-driven speed—is a viable path for enhancing the observability and safety of nuclear facilities. The framework lays the groundwork for future advancements in autonomous reactor control, predictive maintenance, and resilient cybersecurity in safety-critical infrastructure.