Alternative cokriging model for variable-fidelity surrogate modeling

An alternative approach for the construction of the cokriging covariance matrix is developed and a more practical cokriging method in the context of surrogate-based analysis and optimization is proposed. The developed cokriging method is validated against an analytical problem and applied to construct global approximation models of the aerodynamic coefficients as well as the drag polar of an RAE 2822 airfoil. The kriging and cokriging models for the moment coefficient show that the poor space-filling properties of the quasi Monte Carlo sampling of the RANS simulations leaves a noticeable gap in the data and, consequently, inaccuracies in the kriging model in this area. For a given number of RANS samples, the accuracy can be dramatically improved by using cokriging, because additional information from Euler computations is used. The overall efficiency can be improved compared to a situation where additional high-fidelity samples rather than low-fidelity samples are included.