Proceedings of the

The 33rd European Safety and Reliability Conference (ESREL 2023)
3 – 8 September 2023, Southampton, UK

An Active Learning Reliability Analysis Framework Based on Multi-Fidelity Surrogate Model

Ning Lua, Yan-Feng Lib, Song Baic, Tudi Huangd and Hong-Zhong Huange

School of Mechanical and Electrical Engineering, University of Electronic Science and Technology of China, China. Center for System Reliability and Safety, University of Electronic Science and Technology of China, China.


On the one hand, the limit state function (LSF) can be used to define structural reliability, while complex structures frequently correspond to LSFs with high computational costs, necessitating numerous calls to LSFs for the reliability analysis of such structures. On the other hand, detailed paradigms and coarse paradigms are generally considered high-fidelity (HF) models with low model uncertainty and low-fidelity (LF) models with low computational cost, respectively. To effectively address both of these common challenges, this paper proposes an active learning multi-fidelity surrogate modeling framework for structural reliability analysis (SRA). The multi-fidelity (MF) surrogate model has received widespread attention in the performance evaluation of complex structures by fusing models with different accuracies to reduce the computational demand and effectively balance the prediction performance and modeling cost of the surrogate model. In three numerical examples and one engineering example in different dimensions, two multi-fidelity surrogate models with four learning functions are tested and compared with the corresponding single-fidelity (SF) models. All the results demonstrate that the MF model based on this framework is more efficient than the SF model at reducing computational costs without compromising accuracy.

Keywords: Reliability analysis, Active learning, Multi-fidelity, Surrogate model, Aero engine gear.

Download PDF