03.1 - Aerodynamics – CFD Methods and Validation

EXPLAINABLE KNOWLEDGE DRIVEN LLM AGENTS FOR TRUSTWORTHY AIRFOIL AERODYNAMIC OPTIMIZATION DESIGN

Y. Fan¹, Y. Sun¹, J. Xue¹, Y. Wang¹, B. Mi¹; ¹NORTHWESTERN POLYTECHNICAL UNIVERSITY, China

This work introduces an explainable, knowledge-driven LLM-agent framework for airfoil aerodynamic optimization. A pyramid knowledge model fuses data regularities with expert experience, guiding geometry updates and recording traceable evidence. Initial experiments show consistent gains over genetic algorithms indicating improved performance and optimization stability.

View Paper

DOI: 10.71945/icas2026_0490