Contact

Operating AI-Assisted Nuclear Control Systems with COPA 500 and CPLANE at Texas A&M

“Essentially, if we want to demonstrate the system is working, we need to demonstrate on the physical facility.”
— Dr. Yang Liu, Assistant Professor, Texas A&M University

Nuclear engineering has long set the benchmark for safety-critical system design. Decades of experience — including hard lessons from past failures — have shaped a culture of rigor, discipline, and conservatism that has made nuclear one of the safest energy sectors in the world.

Nuclear systems have always demanded caution, because history has shown what’s at stake when systems fail.

That rigor, however, has also shaped control systems that were never designed for today’s realities. Traditional nuclear control platforms prioritize determinism and human oversight, but they lack the openness, compute headroom, and lifecycle flexibility required to support modern tools like artificial intelligence.

At the same time, the context around nuclear energy is changing:

For next-generation reactors, this raises a fundamental challenge. Future systems will increasingly require predictive intelligence — the ability to model complex physics in real time, detect deviations early, and support faster human decision-making.

But introducing AI into nuclear control environments raises an unavoidable question:

What kind of control system can support AI-assisted operation while meeting nuclear-grade requirements for safety, reliability, and trust?

At Texas A&M, this question became the mission. The research team set out to explore whether a physical nuclear facility could be co-piloted by AI — not as a replacement for operators, but as an augmentation — while preserving the rigor nuclear engineering demands.

Critically, this exploration could not remain in simulation. The system had to operate on real hardware, with real sensors, real control loops, and real consequences.

Solution

A production-grade OPA control system enabled by COPA 500 and operated through CPLANE

To meet this challenge, Texas A&M required more than a powerful controller. They needed a complete, operable control system — one that could integrate AI workloads, manage real-time control, and maintain continuous oversight in a high-stakes nuclear environment.

COPA 500: Open, AI-Ready Control at the Device Level

The foundation of the system was the COPA 500 controller, an OPA-aligned, high-performance industrial controller designed to support modern compute workloads.

The research team used the COPA 500 to:

“We’re getting the proven reliability of an industrial control system with the benefits of a modern IT architecture.”
— Bob Hagenau, Co-leader, COPA Team

This allowed the team to focus on what made the project unique: physics-informed AI.

Rather than using generic AI tools, the researchers augmented large language models with custom nuclear-physics subroutines, enabling the system to:

“We’re able to transform general-purpose models into physics-informed AI systems.”
— Zavier Ndum Ndum, Graduate Researcher, Texas A&M University

However, as the system grew more complex, another challenge emerged — one that COPA 500 alone was not designed to solve.

The Missing Piece: System-Level Management, Orchestration, and Oversight

Operating an AI-assisted nuclear control system is not just about executing control logic. It requires:

This is where CPLANE played a critical role.

CPLANE provided the system-level orchestration and management layer that allowed the COPA 500–based control system to operate as a coherent whole — not just a collection of advanced components.

From a system perspective, CPLANE enabled:

As the research team noted:

“For high-stakes environments like nuclear engineering, the system has to be absolutely perfect. COPA 500 just changes the landscape entirely, allowing the researcher or end user to focus on what they do best.”
— Timothy Triplett, Senior Control Systems Engineer

CPLANE ensured that this advanced control environment behaved like a production system, not an experimental integration — even while supporting cutting-edge AI workloads.

Importantly, CPLANE did not replace the openness of the OPA architecture. Instead, it productized the system layer — making multi-component, AI-assisted control operable, observable, and trustworthy.

Outcome

A first-of-its-kind AI-assisted nuclear control system operating safely on a physical facility

Over the past year, Texas A&M successfully demonstrated that AI-assisted nuclear control can operate safely, reliably, and securely when built on the right system foundation.

Key outcomes included:

1. Real-Time, AI-Augmented Operator Support

AI models continuously monitor reactor conditions and alert operators to emerging risks, increasing response speed without removing human authority.

2. Continuous Supervision and Robust Operation

By combining COPA 500 with CPLANE’s system-level management, the control environment ensured that data and system behavior were continually supervised.

“If the value starts to increase, the AI detects it and starts to warn the operator…The data is continually managed, continually supervised. It improves efficiency, accuracy, and speed.”
— Zavier Ndum Ndum, Graduate Researcher, Texas A&M University

3. Production-Grade Reliability from Day One

Despite being a novel control architecture, the system met industrial reliability expectations immediately — a requirement in nuclear environments.

“Industrial control systems have to be reliable from day one… and we’ve been able to do it because we’re using mature, proven components.”
— Bob Hagenau, Co-leader, COPA Team

4. A First-of-Its-Kind Facility

Texas A&M became the first lab to implement an AI-assisted nuclear control facility of this kind, operating seamlessly on a real physical system.

“Our lab is the first to implement this facility, and it’s working seamlessly.”
— Zavier Ndum Ndum, Graduate Researcher, Texas A&M University

5. A Path Toward Smarter, Safer Reactors

The work demonstrates how future SMRs could be deployed more predictably and operated more safely using AI — without sacrificing nuclear-grade rigor.

“Our recent work has demonstrated that AI can be a very useful tool to help nuclear reactors run more secure, safe, and more economical.”
— Dr. Yang Liu, Assistant Professor, Texas A&M University

Why This Matters

This case study shows that advanced AI and nuclear safety are not in conflict — but making them coexist requires more than powerful controllers.

It requires a system-level platform that:

With COPA 500 providing the control foundation and CPLANE providing the system layer, Texas A&M demonstrated what next-generation nuclear control systems can look like: open, intelligent, and trustworthy.