Researchers Develop AI-Powered Framework to Automate Monte Carlo Simulations in Nuclear Engineering

Strategic Insights -

Researchers from Texas A&M University have developed a novel framework called AutoFLUKA that leverages large language models (LLMs) and AI agents to automate the entire FLUKA simulation workflow. This effort aims to enhance productivity, reliability, and accessibility in nuclear engineering analysis by eliminating labor-intensive and error-prone manual input file generation and post-processing. The AutoFLUKA framework integrates Retrieval-Augmented Generation (RAG) and a web-based user-friendly graphical interface, enabling users to interact with the system in real time. Preliminary results demonstrate substantial improvements in resolving FLUKA error-related queries, reducing resolution time from several days to under one minute, and mitigating human-induced simulation errors.

Key Takeaways:

  • AutoFLUKA is a novel framework that leverages domain knowledge-embedded LLMs and AI agents to automate the FLUKA simulation workflow from input file creation to execution management and data analysis.
  • The framework integrates RAG and a web-based user-friendly graphical interface, enabling real-time interaction with the system.
  • Benchmarking against manual FLUKA simulations, AutoFLUKA demonstrated substantial improvements in resolving FLUKA error-related queries, reducing resolution time from several days to under one minute.
  • Human-induced simulation errors were mitigated, and a high accuracy in key simulation metrics, such as neutron fluence and microdosimetric quantities, was achieved with uncertainties below 0.001% for large sample sizes.
  • The flexibility of AutoFLUKA was demonstrated through successful application to both general and specialized nuclear scenarios, and its design allows for straightforward extension to other simulation platforms.
  • The research was funded by the Office of Nuclear Energy and published in the journal Energy and AI.
  • The authors include Zavier Ndum Ndum, Jian Tao, John Ford, and Yang Liu from Texas A&M University's Department of Nuclear Engineering.

Statistics:

  • Resolution time for FLUKA error-related queries was reduced from several days to under one minute.
  • Uncertainties in key simulation metrics, such as neutron fluence and microdosimetric quantities, were below 0.001% for large sample sizes.
  • The AutoFLUKA framework demonstrated substantial improvements in resolving FLUKA error-related queries.

Sources:

  • Automating Monte Carlo simulations in nuclear engineering with domain knowledge-embedded large language model agents. Energy and AI, 2025, 21():100555. (Elsevier)