Reliability analysis with stratified importance sampling based on adaptive Kriging

•Stratified importance sampling and adaptive Kriging are combined for reliability analysis.•Importance sampling density is constructed through an adaptive Kriging method.•Variable for stratification is chosen based on Kriging surrogate to decrease the cost.•Numerical and practical examples show the...

Full description

Saved in:
Bibliographic Details
Published in:Reliability engineering & system safety Vol. 197; pp. 106852 - 12
Main Authors: Xiao, Sinan, Oladyshkin, Sergey, Nowak, Wolfgang
Format: Journal Article
Language:English
Published: Barking Elsevier Ltd 01.05.2020
Elsevier BV
Subjects:
Adaptive learning
Adaptive sampling
Adaptive systems
Carbon dioxide
Carbon sequestration
Coefficient of variation
Computer applications
Computing costs
Failure
Gaussian process
Importance sampling
Kriging
Mathematical analysis
Probability
Regression analysis
Reliability analysis
Reliability engineering
Sampling
Sampling methods
Statistical analysis
Stratified importance sampling
Variance-based sensitivity measure
Adaptive learning
Reliability analysis
Stratified importance sampling
Kriging
Variance-based sensitivity measure
ISSN:0951-8320, 1879-0836
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•Stratified importance sampling and adaptive Kriging are combined for reliability analysis.•Importance sampling density is constructed through an adaptive Kriging method.•Variable for stratification is chosen based on Kriging surrogate to decrease the cost.•Numerical and practical examples show the efficiency of the proposed method. In reliability engineering, estimating the failure probability of a system is one of the most challenging tasks. Since many applied engineering tasks are computationally expensive, it is challenging to estimate failure probabilities using acceptable computational costs. In this paper, to reduce computational cost, we combine a stratified importance sampling method with an adaptive Kriging strategy to estimate failure probabilities. Compared to the importance sampling method, stratified importance sampling needs fewer samples to get an estimate of failure probability with the same coefficient of variation. In the proposed method, we improve the importance sampling density and determine the best input variable for stratification through a Kriging-based model surrogate technique (like a Gaussian process regression). Then, the Kriging surrogate is further adaptively improved to get an accurate estimate of failure probability. The efficiency of the proposed method is demonstrated using several analytic examples and then transferred to a carbon dioxide storage benchmark problem.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0951-8320
1879-0836
DOI:10.1016/j.ress.2020.106852