Research on Aerodynamic Optimization Design of High Lift Airfoil Based on Deep Learning and MOEA/D

Abstract

Purposes Aiming at the performance conflict between optimization parameters in prior optimization method, a hybrid optimization model based on MOEA/D is proposed, which integrates CNN and genetic algorithm into MOEA/D framework to balance the correlation and complexity between various objective functions. Methods First, the deep learning method is used as a supplement to the conventional fluid mechanics analysis method to establish a highly reliable CNN response prediction model for airfoil aerodynamic characteristics, which can be used to quickly evaluate the aerodynamic parameters of airfoil. Then, the response model and genetic operator are interpolated into the MOEA/D framework to construct a multi-objective hybrid optimization model based on MOEA/D. And the lift drag ratio and moment coefficient of a NACA high lift airfoil under cruise condition are taken as the optimization objectives for testing. Finally, through the analysis of aerodynamic performance and flow field structure of the airfoil on the Pareto front, the distribution law of different airfoil configurations on the front is studied, which further guides the designer to explore the potential basic airfoil in the airfoil selection.

Keywords

• aerodynamic optimization
• mixed method
• moea/d
• cnn
• cfd