A Variable-Fidelity Approximate Modelling Method Based on Nested Design of Experiments

In the variable fidelity (VF) metamodeling method, an approximation model of high precision needs to be obtained with a few high fidelity (HF) sampling points available, so it´s important to conduct research on the experiment design. This paper carries out research into an improved nested design of experiment. Firstly, the improved successive local enumeration (ISLE) algorithm is proposed to build the low fidelity (LF) experiment design. Then the process of selecting the HF sample points from LF sample points is converted into the Maximum diversity problem (MDP) with constraint. Finally the HF sampling points are obtained by solving the MDP. To verify the validity of the sampling method presented for VF metamodeling, the optimal Latin hypercube is applied to produce equal sampling points for comparison.