This research explores how Large Language Models (LLMs) can generate code for closed-source simulation software without compromising security or intellectual property.

• LLMs show promise in generating simulation models from natural language prompts
• Retrieval-Augmented Generation (RAG) enables effective code generation for proprietary software
• The approach balances engineering innovation with security requirements
• Creates new possibilities for automating complex engineering simulations

This work matters for engineering teams using proprietary simulation tools by potentially reducing model development time while maintaining security compliance.

Experiments with Large Language Models on Retrieval-Augmented Generation for Closed-Source Simulation Software