Aerodynamic optimization based on multi-fidelity surrogate

Constructing surrogate models need lots of high-fidelity sample data in complex multivariable simulation-based design optimization problems, and the calculation is very costly. Thereby constructing higher accuracy surrogate model with lower computational cost has great engineering significance. In this paper Co-Kriging method is used to construct multi-fidelity surrogate model based on two independent sample datasets. Thus a greater quantity of low-fidelity information can be used to enhance the accuracy of the high-fidelity surrogate model by formulating a form of correction process which reflects the differences between the low and high fidelity model, and the total computational cost of sample data can be greatly reduced. Two function examples demonstrated the influence of approximate accuracy of sample number, and the feasibility of this methodology in engineering design problem is validated by a comparison of different surrogate model based aerodynamic optimization design of RAE2822.