Generative AI | IDEAS Lab at University of Michigan

OpenAeroStruct LLM Agent

Wed, 01 Jan 2025 00:00:00 +0000

Overview

OpenAeroStruct LLM Agent is a tool that leverages Large Language Models (LLMs) to automate aircraft wing design and analysis using the OpenAeroStruct framework. It allows users to input design specifications in natural language, and the system automatically handles meshing, analysis, optimization, visualization, and reporting. It currently supports main aerodynamic objectives of lift and drag, and geometric design variables of taper, sweep, dihedral, chord, and twist.

Key Features:

Natural language processing for wing design specifications
Automated mesh generation and refinement
Wing optimization based on specified objectives (drag, lift, etc.)
Automated visualization of results
Detailed report output

AI-Augmented Design

Sat, 18 Oct 2025 00:00:00 +0000

We are building a multi-faceted framework that incorporates Generative AI into multidisciplinary design. The goal is to introduce automation and traceability into complex design processes. Automation will allow for deeper design-space exploration while allowing human engineers to focus their efforts on less repetitive and potentially human-error-prone tasks.

Our recent work on generative design includes the creation of an Agentic-Based Aircraft Optimization Framework (Lee et al, 2025). This study utilized LLM to transform a design task formulated in common language into a structured script that could be run by OpenAeroStruct, a multidisciplinary wing design and optimization tool. The user inputs a wing design task with various requirements, and the LLM creates corresponding code for OpenAeroStruct to run and create an optimized wing. The LLM then creates a formal report based on the OpenAeroStruct output including optimization results and analysis.

Multi-agent LLM framework for OpenAeroStruct.

Full workflow illustrating subprocesses and information transfer.

Aerodynamic Design and Optimization via a Specialized Agentic Generative AI Framework

Thu, 31 Jul 2025 00:00:00 +0000