21st Congress of International Council of the Aeronautical Sciences, Melbourne, Australia, 13-18 September, 1998
Paper ICAS-98-2.2.1

AN INVERSE DESIGN PROCEDURE FOR AIRFOILS USING ARTIFICIAL NEURAL NETWORKS

Hazarika N., Tuncer I. H.*, Lowe D.
Aston University, United Kingdom; *Middle East Techn. Univ., Turkey

Keywords: design, airfoils, artificial neural networks

In this paper, we investigate a novel method for the inverse design of airfoil sections using artificial neural networks (ANNs). Work on artificial neural networks has shown that ANN s can be used to emulate highly nonlinear relationships, such as that existing between surface pressures and the corresponding airfoil profiles in a flow. Surface pressure distributions generated by a panel code are used to train a neural network, which is then used to predict airfoil profiles for a given surface pressure distribution (the "inverse" problem), or to predict the pressure distribution for a given airfoil profile (the "forward" problem). The generalization capability of ANNs in the presence of noisy data is also studied. Results indicate that optimally trained artificial neural networks may accurately predict airfoil profiles and pressure distributions.

view full paper