Stochastic efficient global optimization with high noise variance and mixed design variables

Engineering design and optimization commonly require the minimization of expected value functions with high noise variance and mixed/discrete design variables. To solve such problems, we extend the stochastic efficient global optimization (SEGO) method of [Carraro et al., Struct Multidiscipl Optim 60(1):245–268 (2019)]. To address high noise variance, we propose two additional stopping criteria for the Monte Carlo integration that is required to approximate the objective function. Moreover, the active learning algorithm within SEGO is adapted to handle discrete design variables. The method is investigated on two numerical examples and the results highlight the efficiency of the proposed method, especially for cases with low computational budget.