Advances in Active Learning Kriging Surrogate Models for Reliability Assessment

Abstract

Reliability assessment is an important link to ensure product quality. However, both the approximate analytical method and the simulation method have shortcomings in applicability. At present, active learning Kriging surrogate model has become a hot spot in reliability assessment methods owing to its high calculating effectiveness and accuracy. The composition and structure for the Kriging theories, the methods for samples generation, together with the theories related to active learning are described in detail. Several kinds of classical active learning Kriging algorithms are analyzed. This paper emphasizes the status of research on Kriging algorithms with active learning processes for solving small failure probability, system reliability, time-dependent reliability and hybrid variable problems. Finally, the development prospect of active learning Kriging algorithm is discussed.