Improved Sequential Approximate Optimization for Aerodynamic Design Benchmark Problem

A modified surrogate-based optimization method based on Sequential Approximate Optimization (SAO) is proposed to purposively improve the efficiency of aerodynamic shape optimization. In this method, a specific initial sampling approach is proposed to obtain the initial sampling set of excellent properties of space-filling and orthogonality in the shape space, a field approximate model is presented to predict the flow field parameters of interest for the specific aerodynamic optimization problems. Moreover, a novel and efficient infill strategy is proposed, which uses the inaccurate search technique in cooperation with an elite archive to locate the promising region and improve the surrogate accuracy. Consequently, the optimization efficiency is well enhanced. Two benchmark aerodynamic optimization problemsare performed using the proposed method. Results reveal that the proposed method presents much better performances comparing to conventional SAO and the stochastic optimization methods, in terms of solution quality and convergence rate.