Comparison of Kriging and Radial Basis Function surrogate models applied to a global optimization framework

Resumo

A literature survey reveals that many optimization problems present objective and/or constraints func-
tions that demand high computational effort. Optimization algorithms which are able to solve these problems with

just a few evaluations of such functions become necessary, in order to avoid prohibitive computational costs. In

this context, there are a lot of surrogate models that can be employed to replace objective and/or constraints func-
tions whenever possible, which are much faster to be evaluated than the original functions. In the present paper, a

global optimization framework based on surrogate models is investigated, and two different surrogate models are
considered: Radial Basis Function and Kriging. The framework consists of three search strategies, which may take

place in each iteration of the optimization process: a local search, a global search and a refinement step. This opti-
mization procedure is applied in benchmark problems from the literature and the results obtained by each surrogate

model are compared. As a result, the framework was found to be considerably stable and to achieve satisfactory
responses with both surrogates. Overall, among the cases analyzed, the framework based on Radial Basis Function
showed better performance.