An efficient adaptive kriging refinement method for reliability analysis with small failure probability

Estimating the small failure probability of highly reliable structures is often computationally expensive in reliability analysis. The adaptive kriging-based reliability analysis methods have been widely used to solve this issue. However, the kriging refinement phase of these methods may not achieve...

Full description

Saved in:
Bibliographic Details
Published in:Structural and multidisciplinary optimization 2023-10, Vol.66 (10), p.222, Article 222
Main Authors: Shi, Luojie, Xiang, Yongyong, Pan, Baisong, Li, Yifan
Format: Article
Language:English
Subjects:
Accuracy
Computational Mathematics and Numerical Analysis
Design of experiments
Efficiency
Engineering
Engineering Design
Limit states
Methods
Monte Carlo simulation
Random variables
Reliability analysis
Research Paper
Strategy
Theoretical and Applied Mechanics
Training
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Estimating the small failure probability of highly reliable structures is often computationally expensive in reliability analysis. The adaptive kriging-based reliability analysis methods have been widely used to solve this issue. However, the kriging refinement phase of these methods may not achieve adequate efficiency due to a large candidate sample pool (CSP) size and unnecessary limit state function (LSF) evaluations. In this work, an efficient adaptive kriging refinement method (EAKRM) is proposed to alleviate the computational burden. First, a CSP generation strategy is developed using the radius sequence to gradually generate uniform samples along the direction vector until the failure region appears. Considering points with a high risk of misjudgment and located outside of CSP, a two-stage training point selection strategy is then proposed based on the learning function U to determine the most valuable training point. Finally, a correlation-based stopping criterion is presented by quantifying the consistency of the kriging predictive signs between two successive refinement processes. Three mathematical examples and three engineering examples are employed to illustrate the effectiveness of the proposed EAKRM.
ISSN:1615-147X
1615-1488
DOI:10.1007/s00158-023-03672-z