An adaptive sampling strategy for construction of surrogate aerodynamic model

Abstract

Surrogate model technology has gradually become an important branch and key technology of modern aircraft design in recent years. In this study, we present a novel adaptive sampling method in the process of surrogate aerodynamic model construction based on the hyper-volume iteration (HVI) strategy. The HVI strategy can be not only regarded an adaptive sampling method in the design of experiments but also used as a convergence criterion when updating the surrogate model. Compared with the traditional adaptive/non-adaptive approaches that focus on global approximation, the HVI method emphasizes more on the extrema capture ability. The most distinctive feature of the HVI method is the assigning of more sampling points where the function changes extensively while selecting fewer points where the function changes gently, as these are aspects that can represent the adaptability of the HVI method. The advantage of using this method is that we can arrange the sampling points intelligently according to the specific characteristics of the objective function to avoid problems caused by subjectivity.