Efficient Design of Airfoil Shapes Using Surface Pressure–Informed Bayesian Optimization

Aerodynamic shape optimization (ASO) of airfoil shapes is an essential part of the aerodynamic design process. Optimizing the shape of an airfoil is a challenging and time-consuming task when computationally expensive numerical simulations are required. As an alternative, surrogate-based optimization methods, such as machine learning models and Bayesian optimization (BO), have been explored to optimize the shapes of airfoils. However, most currently used methods do not utilize information from surface pressure distributions in creating the surrogate model or the optimization search. Surface pressure distributions are essential aerodynamic features that influence the design of an airfoil. As such, this work proposes surface pressure-informed Bayesian optimization (SPINBO), a novel BO methodology that incorporates surface pressure distribution information into the surrogate modeling methodology and relates the pressure distribution to the aerodynamic coefficients. The proposed approach is characterized using the drag minimization problem of the RAE 2822 airfoil in transonic viscous flow at a Mach number of 0.734 and a Reynolds number of 6.5 million. A comparison is made between the performance of the proposed approach and a standard BO approach. The results demonstrate that the best airfoil shape obtained from both approaches has similar characteristics and achieves a drag reduction of close to 43% over the RAE 2822 airfoil. However, the proposed approach shows an enhanced rate of convergence to the best airfoil shape, achieving a lower objective function value than the standard BO method in fewer sampling iterations. This illustrates the benefits of SPINBO as an optimization methodology. It also demonstrates that incorporating the surface pressure distribution information into the model has the potential to enhance the convergence of the optimization and reduce the computational expense of performing ASO for airfoils.