[2608.30186] The PUR-1 Cyber-Physical Digital Twin

The PUR-1 Cyber-Physical Digital Twin is a high-fidelity, real-time virtual replica of the Purdue University Reactor Number One (PUR-1), developed to serve as a cyber-physical testbed for advanced nuclear technologies. Authored by Vasileios Theos and colleagues, with the paper submitted to arXiv in August 2026, the system integrates physics-based models (neutronics, thermal-hydraulics) with AI-driven data-driven approaches to enable closed-loop diagnostics, forecasting, and predictive control.

Key capabilities and features include:

• Real-Time Synchronization: The twin maintains two-way communication with the physical reactor, achieving state estimation and short-term forecasting with latency lower than the reactor's operational cycle.
• Autonomous Control: It supports remote, automatic real-time adjustments of reactor power, as demonstrated in collaborations with Idaho National Laboratory and the University of Illinois, using reinforcement learning to minimize power fluctuations without manual intervention.
• Cybersecurity Research: The platform facilitates testing of AI/ML algorithms for detecting cyberattacks and simulating quantum encryption protocols to secure remote reactor communications.
• High Accuracy: The underlying machine learning algorithms have demonstrated the ability to predict reactor power output fluctuations with 99% accuracy, validating its potential for improving the safety and efficiency of future Small Modular Reactors (SMRs).

The PUR-1 Cyber-Physical Digital Twin presents a real-time virtual replica of a physical nuclear reactor system, positioned at the intersection of process modeling, sensing, and cyber infrastructure. Rather than treating a digital twin as an offline simulation or retrospective analytics tool, the material frames the twin as a continuously synchronized computational mirror of the reactor’s operating state. It emphasizes the coupling of physical reactor dynamics with real-time data streams and a cyber layer for state estimation, visualization, scenario analysis, and decision support. This framing is significant because it treats the reactor not merely as a physical plant, but as a cyber-physical system whose performance and safety depend on the integrity, latency, and fidelity of the information loop.

A key contribution is the concrete instantiation of digital-twin concepts in a nuclear domain, where the value comes from closing the loop between high-fidelity reactor models and live telemetry. The twin can support use cases such as operator training, anomaly detection, control validation, and resilience testing without exposing the physical system to unnecessary operational risk. The work also highlights the engineering challenges specific to nuclear cyber-physical systems: safety constraints, model uncertainty, regulatory scrutiny, and the need for trustworthy, low-latency data pipelines. In other words, the twin is not just a visualization tool, but a decision-support and validation environment that must remain tightly aligned with the physical plant.

The material matters because nuclear facilities are among the most safety-critical and operationally constrained industrial systems, where direct experimentation is limited and the cost of misdiagnosis is high. A real-time digital twin can provide a safer, repeatable, and scalable environment for monitoring, prediction, and control development, while also serving as a testbed for integrating modern cyber capabilities into legacy safety-critical infrastructure. In that sense, PUR-1 is less a single application than a reference point for how digital twins can be made operationally meaningful in domains where fidelity, timing, and trust are non-negotiable.