Adaptive multi-fidelity metamodel for CFD-based optimization via radial basis functions and preliminary results on analytical benchmark problems

The report presents a multi-fidelity metamodel to combine data produced by CFD computer simulations with different accuracy, such as potential ow and Reynolds averaged Navier-Stokes equations solvers. The multi-fidelity (MF) metamodel is based on stochastic radial basis functions and is built as the sum of a low-fidelity-trained metamodel and an error metamodel, based on the difference between high- and low-fidelity simulations. The MF metamodel is adaptively refined using dynamic sampling criteria, based on the prediction uncertainty in combination with the objective optimum and the computational cost of high- and low-fidelity evaluations. Four adaptive sampling methods are demonstrated by four analytical benchmark problems. The performance of the adaptive sampling methods is assessed via objective function convergence. The presented results are also published in Serani et al. (2019).