An Improved Hierarchical Kriging for Variable-Fidelity Surrogate Modeling

Abstract

Variable-fidelity (VF) modeling methods have been widely used in complex engineering system design to mitigate the computational burden. However, the existing VF modeling methods do not make full use of the mutual information between high-fidelity (HF) and low-fidelity (LF) models which may lead to low accuracy VF model. In this paper, a global VF modeling method named improved hierarchical kriging (IHK) is proposed, in which LF model is varied though a polynomial response surface (PRS) function to capture the tendency of HF model, and HF sample points are used to ensure the accuracy of the VF model. A numerical example and the modeling of aerodynamic coefficient for an aircraft are provided to demonstrate the approximation capability of IHK as well as other two metamodeling methods. Results show that IHK provides more accurate metamodel at the same simulation cost, that is very important in metamodel-based engineering design problems.