Aerodynamic lift coefficient prediction of supercritical airfoils at transonic flow regime using convolutional neural networks (CNNs) and multi-layer perceptions (MLPs)

Designing an aircraft involves a lot of stages, however, airfoil selection remains one of the most crucial aspects of the design process. The type of airfoil chosen determines the lift on the aircraft wing and the drag on the aircraft fuselage. When a potential airfoil is identified, one of the first steps in deciding its optimality for the aircraft design requirements is to obtain its aerodynamic lift and drag coefficients. In the early stages of trying to select a candidate airfoil, which a whole part of the design process rests on, the conventional method for acquiring the aerodynamic coefficients is through Computational Fluid Dynamics Simulations (CFDs). However, CFD simulation is usually a computationally expensive, memory-demanding, and time-consuming iterative process; to circumvent this challenge, a data-driven model is proposed for the prediction of the lift coefficient of an airfoil in a transonic flow regime. Convolutional Neural Networks (CNNs) and Multi-Layer Perceptrons (MLPs) were used to develop a suitable model which can learn a set of usable patterns from an aerodynamic data corpus for the prediction of the lift coefficients of airfoils. Findings from the training revealed that the models (MLPs and CNNs) were able to accurately predict the lift coefficients of the airfoil.