Nuclear Expertise Guides Global Nonproliferation Innovation

ORNL has helped further this crucial IAEA mission for decades, but the mission has never been more critical. With recent advances in nuclear technology – and a concurrent global resurgence of interest in nuclear power as a carbon-free energy source – the lab is exploring new approaches to ensure that nuclear materials and technologies are safeguarded against both existing and emerging threats.

Advancing nonproliferation, verification amid new reactor designs

As new nuclear energy technologies are developed, including advanced reactors and their associated fuel cycles – like the design and use of smaller, modular reactors with newer, more robust, built-in safety features by design – nonproliferation verification technology must constantly redefine itself to remain not only effective, but relevant. 

ORNL researchers such as Senior Research and Development Scientist Louise Evans are working to ensure the early integration of new safeguards approaches and verification technologies into advanced reactor designs before they’re finalized for manufacture. “Part of realizing the potential benefits of these designs is making sure that they are safe, secure and safeguardable, when it comes to nuclear material,” said Evans. New reactor designs – including molten-salt reactors and those that use new fuels – will likely drive innovation in both detection tools and verification methodologies.

“The different operational environments and unique fuels associated with some of these advanced reactors, as well as the wide variety of designs, pose new challenges for nuclear materials tracking and data analysis,” Evans said. “There is no single, ‘one size fits all’ nonproliferation-verification technological solution for every advanced reactor design. Our current nonproliferation verification technologies were designed for our current reactor materials and processes. They likely won’t apply directly to these advanced reactor designs without further development.”

Focused on “safeguards by design,” Evans and colleagues within ORNL’s advanced reactor safeguards program engage early with reactor developers, discussing plant layouts, nuclear material locations, potential design options and opportunities for integration of verification technologies, while targeting key measurement points for tracking, detecting, characterizing and protecting nuclear material as it moves through preliminary facility designs. 

ORNL’s experts are integrated into each step of the verification technology development life cycle, over the full term of reactor development, which ideally occurs in parallel – from verification technology equipment design, prototyping and testing, through its field trials, commercialization and regulatory approval – well before designs are approved by the Nuclear Regulatory Commission and finalized for the build.

Augmenting international nuclear safeguards with machine learning

ORNL researchers Scott Stewart and Nathan Martindale hope to equip IAEA nuclear inspectors with advanced, interactive machine learning, or IML, tools for more efficient data analysis. As more nuclear facilities are built or expanded to meet global energy demands and carbon reduction targets, the amount of data needed to verify nuclear activities within a country continues to grow exponentially – which is where textual data analysis can help. Finding key indicators amid huge data streams can be a real challenge for researchers in many fields, and more accurate analysis can help enable stronger safeguards while increasing efficiencies. 

For Stewart and Martindale, better textual data analysis tools, in this case facilitated by IML, can help improve international safeguards by empowering IAEA open-source analysts or inspectors with more precise information when they need it most. As a result, the IAEA investigators are better informed – before, during and after a nuclear site inspection, or while reviewing a member state’s literature and declarations – as they strive to determine whether a certain site is adhering to global standards for nuclear safety and security. Aided by IML, these inspectors and analysts can sift through increasingly massive datasets more effectively, as they look for anomalies within various nuclear systems that can help highlight potential violations of a member state’s nonproliferation commitments.

“Maybe it’s not a unique problem, but we do have a problem. Relevant examples within massive datasets for us are sparse,” said Stewart. “Even if the overall amount of data that we’re looking at is very large, you’re stuck with sort of a needle-in-a-needle-stack situation. Another thing is the consequences of incorrect analysis can be extremely high.” When analysts are experiencing what Martindale called a “huge dump of textual data,” the IML used in Stewart and Martindale’s Interactive Corpus Analysis Tool, or ICAT, enables the average IAEA analyst (or inspector) to sift through far more data, faster, and with more relevant results. 

One key point for users at the IAEA and elsewhere is that ICAT requires no prior experience in machine learning. “You do not need to be a machine-learning expert to interact with, train and benefit from our underlying IML model,” Martindale said. Another is that although the interface itself is a research-enabling priority that relies on artificial intelligence, keeping a human decision-maker in the loop remains essential. 

“The main idea behind this tool is that it is very much still a human-in-the-loop type of approach,” Martindale said. “The idea for the tool is that it helps make the human more effective in making the determinations that they need to make, rather than just trusting that the model is going to make a solid determination based on what it sees.” Emerging tools like ICAT offer additional sources of finely tuned information, providing faster, explainable textual data analysis for the publications and records from a country’s nuclear program. 

Looking ahead with nuclear

As the nuclear field rapidly expands to meet both energy demand and climate commitments, ORNL researchers are leaders in a never-ending mission to advance nuclear nonproliferation, at the forefront of related efforts to anticipate new threats and facilitate more effective, world-class verification techniques. At ORNL, research by Evans, Stewart and Martindale directly assists the IAEA’s Department of Safeguards, where it makes a powerful impact on international nuclear program verification, contributing significantly to nonproliferation efforts across the globe.

According to the IAEA, a “net-zero” future – when carbon emissions are virtually eliminated – simply isn’t possible without widespread adoption of nuclear power. Helping ensure that this essential source of power and versatile resources is kept safe, secure and safeguardable – by reducing risks, deterring threats, advancing security, furthering innovative approaches and responding to emergencies – is all part of ORNL’s mission. 

Christopher J Driver is Science & Technical Writer at Oak Ridge National Laboratory (ORNL). The article was originally poste to the website of ORNL.