Details
| Original language | English |
|---|---|
| Article number | 113344 |
| Journal | Computer Methods in Applied Mechanics and Engineering |
| Volume | 372 |
| Early online date | 21 Aug 2020 |
| Publication status | Published - 1 Dec 2020 |
Abstract
The efficient estimation of the failure probability function of rare failure events is a challenging task in the structural safety analysis when the input variables are characterized by imprecise probability models due to insufficient information on these variables. The recently developed non-intrusive imprecise stochastic simulation (NISS) provides a general, yet competitive, framework for dealing with this type of problems, and it has been shown that many classical stochastic simulation techniques, with suitable adequations, can be injected into this framework for tackling different types of problems in uncertainty quantification. This work aims at investigating the rare failure event analysis based on the global version of NISS and line sampling. A new method, called global imprecise line sampling (GILS), is firstly proposed, to efficiently estimate failure probability function with the same computational cost as classical line sampling. By joint sampling from both the aleatory and epistemic spaces, the GILS provides elegant estimators for the functional components of the failure probability. Then, to further reduce the computational cost, and improve its suitability for nonlinear problems, an imprecise active learning line sampling procedure is established by combining GILS with Gaussian process regression (GPR) with the target of adaptively exploring the aleatory and epistemic spaces within the framework of line sampling. Two analytical examples and two engineering applications demonstrate the efficiency and accuracy of the proposed methods.
Keywords
- Active learning, Gaussian process regression, Imprecise probability, Line sampling, Sensitivity analysis, Uncertainty quantification
ASJC Scopus subject areas
- Engineering(all)
- Computational Mechanics
- Engineering(all)
- Mechanics of Materials
- Engineering(all)
- Mechanical Engineering
- Physics and Astronomy(all)
- General Physics and Astronomy
- Computer Science(all)
- Computer Science Applications
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: Computer Methods in Applied Mechanics and Engineering, Vol. 372, 113344, 01.12.2020.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - Adaptive reliability analysis for rare events evaluation with global imprecise line sampling
AU - Song, Jingwen
AU - Wei, Pengfei
AU - Valdebenito, Marcos
AU - Beer, Michael
N1 - Funding Information: This work is supported by the National Natural Science Foundation of China (grant number NSFC 51905430 ) and ANID (National Agency for Research and Development, Chile) under its program FONDECYT , grant number 1180271 . The first author is supported by the program of China Scholarships Council (CSC) . The second and third authors are both supported by the Alexander von Humboldt Foundation of Germany . The second author is also supported by the Top International University Visiting Program for Outstanding Young Scholars of Northwestern Polytechnical University of China .
PY - 2020/12/1
Y1 - 2020/12/1
N2 - The efficient estimation of the failure probability function of rare failure events is a challenging task in the structural safety analysis when the input variables are characterized by imprecise probability models due to insufficient information on these variables. The recently developed non-intrusive imprecise stochastic simulation (NISS) provides a general, yet competitive, framework for dealing with this type of problems, and it has been shown that many classical stochastic simulation techniques, with suitable adequations, can be injected into this framework for tackling different types of problems in uncertainty quantification. This work aims at investigating the rare failure event analysis based on the global version of NISS and line sampling. A new method, called global imprecise line sampling (GILS), is firstly proposed, to efficiently estimate failure probability function with the same computational cost as classical line sampling. By joint sampling from both the aleatory and epistemic spaces, the GILS provides elegant estimators for the functional components of the failure probability. Then, to further reduce the computational cost, and improve its suitability for nonlinear problems, an imprecise active learning line sampling procedure is established by combining GILS with Gaussian process regression (GPR) with the target of adaptively exploring the aleatory and epistemic spaces within the framework of line sampling. Two analytical examples and two engineering applications demonstrate the efficiency and accuracy of the proposed methods.
AB - The efficient estimation of the failure probability function of rare failure events is a challenging task in the structural safety analysis when the input variables are characterized by imprecise probability models due to insufficient information on these variables. The recently developed non-intrusive imprecise stochastic simulation (NISS) provides a general, yet competitive, framework for dealing with this type of problems, and it has been shown that many classical stochastic simulation techniques, with suitable adequations, can be injected into this framework for tackling different types of problems in uncertainty quantification. This work aims at investigating the rare failure event analysis based on the global version of NISS and line sampling. A new method, called global imprecise line sampling (GILS), is firstly proposed, to efficiently estimate failure probability function with the same computational cost as classical line sampling. By joint sampling from both the aleatory and epistemic spaces, the GILS provides elegant estimators for the functional components of the failure probability. Then, to further reduce the computational cost, and improve its suitability for nonlinear problems, an imprecise active learning line sampling procedure is established by combining GILS with Gaussian process regression (GPR) with the target of adaptively exploring the aleatory and epistemic spaces within the framework of line sampling. Two analytical examples and two engineering applications demonstrate the efficiency and accuracy of the proposed methods.
KW - Active learning
KW - Gaussian process regression
KW - Imprecise probability
KW - Line sampling
KW - Sensitivity analysis
KW - Uncertainty quantification
UR - http://www.scopus.com/inward/record.url?scp=85089604176&partnerID=8YFLogxK
U2 - 10.1016/j.cma.2020.113344
DO - 10.1016/j.cma.2020.113344
M3 - Article
AN - SCOPUS:85089604176
VL - 372
JO - Computer Methods in Applied Mechanics and Engineering
JF - Computer Methods in Applied Mechanics and Engineering
SN - 0045-7825
M1 - 113344
ER -