Metamodels for aerothermodynamic design optimization of hypersonic spiked blunt bodies

The use of spikes in the design of hypersonic vehicles yields a considerable reduction in drag and aerodynamic heating. However, their effectiveness relies strongly on the parameters of the actual designs. In this work, surrogates generated from high fidelity models are used in the context of design optimization of a spiked blunt body in hypersonic flow conditions. Four different surrogate models including a quadratic response surface and three Kriging surrogates were constructed based on the values of drag and maximum temperature responses. The evolutionary genetic algorithm is applied to find the optimum design based on each of the surrogates for these two objectives. The structure of the various surrogate models was investigated and the main differences were addressed. For the cases investigated in this work, the Kriging surrogate based on exponential correlations produced a relatively better overall performance. It was concluded that, instead of adding more training samples, eliminating some samples that show undesired aerodynamic performance yield a remarkable improvement in surrogates' performance. The optimized designs produced 92% and 13% reduction in drag and wall temperature responses, respectively.