Bayesian optimization for robust engineering design with noise factors

Robust parametric engineering design with noise is critical due to the biasing effect that noise factors can have on the system during the actual process of engineering design. Most Bayesian algorithms for robust parameter design use a quadratic loss function as a measure of quality characteristics, which is not always realistic. In this paper, two Bayesian optimization algorithms are developed for general loss functions for the first time, and Monte Carlo sampling is used to implement the computation of the proposed acquisition functions for the optimization process. The effectiveness of the two proposed algorithms is demonstrated in three arithmetic cases.