01.1 - Aircraft Design and Integrated System (Basics and Theory)
DEEP LEARNING TECHNIQUES FOR HIGH-DIMENSIONAL SURROGATE-BASED AERODYNAMIC DESIGN
M.A. Hariansyah¹, K. Shimoyama¹; ¹Institute of Fluid Science, Tohoku University, Japan
A large-scale wing shape optimization is characterized by its high dimensionality, making gradient-free population-based methods such as genetic algorithm (GA) unpopular in the field. We address this issue by incorporating deep learning techniques to assist a GA called NSGA-II and demonstrate the method by solving lift-constrained drag minimizations of the Common Research Model wing.