MooseAgent: A LLM Based Multi-Agent Framework for Automating Moose Simulation
a year ago
- #Finite Element Method
- #Automation
- #Machine Learning
- MooseAgent is a multi-agent framework that automates MOOSE simulations using LLMs.
- The framework understands natural language requirements and generates MOOSE input files automatically.
- It employs task decomposition and multi-round verification to improve accuracy.
- A vector database of annotated MOOSE input cards and documentation reduces model hallucinations.
- Experiments show high success rates for simple single-physics problems.
- The framework is open-sourced and aims to lower the barrier for finite element simulations.