Database self-expansion based on artificial neural network: An approach in aircraft design

Aircraft design today requires large amount of CFD calculation. For example when Natural Laminar Flow technique is applied to reduce aircraft skin friction drag by extending laminar length over surface, flowfield calculation related with airfoil laminar transition is computationally intense. Situations like this make iterative trial-and-error approach very inefficient. In order to improve this, this paper aims to exploit airfoil database of geometry and aerodynamic performance (from accumulated experiment and CFD calculation results) based on Artificial Neural Network to develop the approach of database self-expansion. It can generate airfoils with better aerodynamic performance from original database, so that the new airfoils can be applied to improve local aerodynamic performance of aircraft. The motive of the approach is to utilize the resource of accumulated optimization products in order to aid aircraft design. In this paper, we will discuss its application in laminar length extension over the surface of nacelle and wing. Geometry description in preparation of database establishment, configuration of network training, and workflow will be described in the paper.