AI Meets Nuclear Science: A Breakthrough Approach

Text-generating AI, like ChatGPT, is typically known for answering everyday queries, but its potential extends far beyond that. A recent study by nuclear engineering graduate student Zavier Ndum reveals how these AI technologies can propel the field of nuclear science forward, enhancing productivity and efficiency for health physicists.

The Promise of Large Language Models in Nuclear Engineering

Large-language models (LLMs) harness vast amounts of written data to generate text. While these models have rapidly found applications in areas like software engineering and research support, their involvement in nuclear science remains relatively rare. This limited use is largely due to the sensitive nature of proprietary data often involved in nuclear research, which can't be easily shared with general LLMs.

"In nuclear science, proprietary data often presents security challenges," Ndum noted. "You can't just input sensitive information into tools like ChatGPT without risks. However, if we develop a secure internal system, we can significantly enhance our workflow."

Introducing AutoFLUKA: The Future of Nuclear Simulations

Ndum's groundbreaking paper introduces AutoFLUKA, an AI tool designed to automate tasks in nuclear science research through simulations using a software called FLUKA. This innovative application not only runs simulations but also edits input files and analyzes results, generating insightful graphs to visualize findings.

Importantly, other researchers can utilize AutoFLUKA with their proprietary data, ensuring that sensitive information remains secure on the originating computer.

Speeding Up Research with AI

"With this technology, you can sift through vast amounts of information quickly, rather than manually searching through documents for answers," Ndum explained. By allowing the application to access pertinent documents, users can engage in rapid Q&A and receive tailored suggestions to enhance their specific projects.

Overcoming Challenges: From Health Physics to Cutting-Edge AI

Transitioning to a focus on AI presented its own challenges for Ndum, who had previously specialized in health physics and radiation dosimetry. Drawing from his experiences at Texas A&M and the guidance of esteemed professors, he was able to employ case studies that support the development of AutoFLUKA.

"Venturing into uncharted territory is always demanding, but perseverance can lead to groundbreaking innovations," he remarked.

Bridging AI and Health Physics: Envisioning the Future

Ndum's commitment to blending AI with health physics shines through as he plans to showcase his findings at the upcoming State of Texas Chapter of the Health Physics Society conference. Here, he’ll share insights on using LLMs as virtual assistants to expedite the retrieval of critical health physics information.

"For radiation safety officers, navigating complex regulations can be laborious; our application can transform a lengthy search into mere seconds," he highlighted.

Expanding Horizons: The Path Forward

Currently, Ndum continues enhancing his research by creating a sophisticated LLM application that answers complex, field-specific inquiries. This advanced tool not only analyzes various document types but also integrates real-time online information for comprehensive research support, ultimately streamlining nuclear science workflows.

"Exploring AI's potential in nuclear science is crucial, and I'm committed to pushing these boundaries further," Ndum asserted.

A Game-Changer for Nuclear Engineering

Dr. Yang Liu, director of the SMART group, echoes Ndum's vision, emphasizing the importance of integrating AI into nuclear research. "Zavier’s innovative approach is groundbreaking. Leveraging AI for secure, domain-specific automation represents a transformative shift, paving the way for more efficient, data-driven progress in areas like reactor modeling and nuclear safety."