A novel learning function for adaptive surrogate-model-based reliability evaluation

Abstract

The classical reliability analysis methods, due to the ever-increasing complexity of engineering structure, may lead to higher and higher calculation errors and costs. The adaptive surrogate-model-based reliability evaluation method strikes a desirable balance between computational efficiency and accuracy, making it a prevalent technique in the domain of reliability evaluation. Learning function is the core of this reliability evaluation method. In this study, a novel learning function is proposed to adaptively choose the best update sample. This learning function does not depend on the prediction variance provided by the Kriging model. Therefore, this learning function is not limited to the Kriging model. In theory, it can be combined with different surrogate models. Four comparative cases are used to illustrate the computational efficiency and accuracy of the proposed method, including series system case with four branches, highly nonlinear two-dimensional numerical example, and two practical engineering case. This article is part of the theme issue 'Physics-informed machine learning and its structural integrity applications (Part 2)'.