Adaptive Sampling Strategies for Surrogate-Based Aerodynamic Optimization

Abstract

The chapter proposes the application of surrogate-based optimization to the efficient design of aeronautical configurations. The surrogate model consists of the Proper Orthogonal Decomposition of computed aerodynamic flow fields and Radial Basis Functions interpolation to reconstruct the aerodynamic flow at any unknown design vector. The surrogate model is coupled to an evolutionary algorithm to globally explore the design space. Several adaptive sampling strategies are proposed, either objective-driven (i.e. aimed at improving the fitness function) or error-driven (i.e. aimed at reducing the prediction error of the surrogate model globally). The proposed methodology is applied to the design optimization of a two-dimensional airfoil in multi-point transonic conditions. The results of different training strategies are critically discussed and compared.