Abstract
The practical implementation of dynamic digital twins in aerospace demands rapid multidisciplinary design analysis and optimization (MDAO), yet this process is currently limited by slow, manual workflows. This paper introduces an agentic Generative Artificial Intelligence framework to automate the synthesis of complex MDAO workflows from high-level natural language commands. The framework employs a multi-agent architecture that autonomously generates, executes, and analyzes aerodynamic optimization scripts using the OpenAeroStruct tool. This process establishes a robust digital thread from requirement to result, providing a foundational component for the near-real-time model updates demanded by the digital twin paradigm. Experiments demonstrate that the framework successfully converges on optimized aircraft wing designs for given sets of constraints, showcasing accuracy and efficiency superior to single-shot language model prompting. This work presents a foundational enabling technology for the digital twin paradigm, providing a method to significantly accelerate the design cycle and make the exploration of novel aircraft concepts more efficient and accessible.
Conan Lee
Undergraduate Research Assistant
Conan Lee is a third-year undergraduate exchange student from the Hong Kong University of Science.
