On deep-learning-based geometric filtering in aerodynamic shape optimization

Abstract

Geometric filtering based on deep-learning models has been shown to be effective to shrink the design space and improve the efficiency of aerodynamic shape optimization. However, since the deep-learning models are trained by existing airfoils, it is criticized that geometric filtering would prevent optimization from finding innovative aerodynamic shapes. This work is conducted to address the concern. By performing 216 airfoil design optimization and several wing design optimization of a conventional wing-body-tail configuration and a blended-wing-body configuration, we find that using the geometric filtering with a lower bound of ∼0.7 does not exclude innovative aerodynamic shapes that maximize cruise efficiency. The results strengthen the confidence of applying deep-learning-based geometric filtering in aerodynamic shape optimization. Then, two applications of geometric filtering in aerodynamic shape optimization are showcased: the geometric validity constraint and global modal shape derivation. The former is shown to enable aerodynamic shape optimization in a large design space, and the latter provides an efficient parameterization approach to aerodynamic modeling of three-dimensional aircraft configurations.