Combining adjoint-based and surrogate-based optimizations for benchmark aerodynamic design problems

In recent years, surrogate-based modeling and optimization have received much attention in the area of aerodynamic design optimization (ADO). However, for high-dimensional problems with large number of design variables, surrogate-based optimization (SBO) is suffering from the prohibitive computational cost associated with evaluating a large number of sample points by high-fidelity and expensive computational fluid dynamics (CFD) simulations. In this paper, we propose to use gradient-enhanced kriging (GEK) to combine the adjoint-based and surrogate-based optimizations, to greatly improve the efficiency of global optimization. The GEK model is integrated to a surrogate-based optimizer and demonstrated for Benchmark Case 1, Case 2 and Case 4 developed by the AIAA Applied Aerodynamics Discussion Group (ADODG), with the number of design variables in the range from 18 to 42. It is observed that, GEK model is much more efficient than the traditional kriging model, indicating that the proposed method has great potential for breaking or at least ameliorating the “curse of dimensionality” for higher-dimensional engineering design problems.