AK-SYSi: an improved adaptive Kriging model for system reliability analysis with multiple failure modes by a refined U learning function

Yun, Wanying; Lü, Zhenzhou; Jiang, Xian

doi:10.1007/s00158-018-2067-3

Cited by 125 publications

(16 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This active learning process relies on a merit function that quantifies the expected merit of choosing each of the MC samples as the next simulation point. AK-SYSi is an improved version of AK-SYS specifically for the composite criterion approach by proposing a refined learning function to avoid false identification of min=max failure modes (Yun et al 2019). Hu et al (2017) proposed the efficient kriging surrogate modeling approach (EKSA) for SSR analysis by using a composite surrogate built upon individual component and singular value decomposition (SVD) to efficiently represent the high-dimensional correlated vectors (Palmer et al 2012).…”

Section: Methods For General Systemsmentioning

confidence: 99%

See 1 more Smart Citation

Structural System Reliability: Overview of Theories and Applications to Optimization

Song

Kang

Lee

et al. 2021

ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng.

View full text Add to dashboard Cite

This paper provides an overview of theories and applications of structural system reliability (SSR). The paper defines SSR problems and discusses the growing needs for SSR analysis and technical challenges. Detailed literature reviews are provided for three subtopics: SSR methods for Boolean system events, SSR methods for sequential failures, and SSR-based design/topology optimization. Discussions of each subtopic define the target problem using mathematical formulations and categorize existing SSR methods in terms of the characteristics of the problems and approaches. The paper summarizes SSR methods that are considered critical in the history and have introduced notable technological developments in recent years. In each subtopic or category, the reviewed methods are compared with each other in terms of accuracy, computational efficiency, and implementation issues to allow identifying apposite methods for SSR applications. The paper concludes with remarks on future research needs and opportunities.

show abstract

Section: Methods For General Systemsmentioning

confidence: 99%

“…Number of the initial design of experiments needs to be chosen based on experience. AK-SYSi (Yun et al 2019) Tested for up to 10 failure modes with 11 random variables.…”

Section: Ak-sys (Fauriat and Gayton 2014)mentioning

confidence: 99%

Structural System Reliability: Overview of Theories and Applications to Optimization

Song

Kang

Lee

et al. 2021

ASCE-ASME J. Risk Uncertainty Eng. Syst., Part A: Civ. Eng.

View full text Add to dashboard Cite

show abstract

“…Because y 3 ( X ) has no contribution to system failure, new added training points should be as few as possible. We perform the reliability and sensitivity by ALK-TCR method combined with different learning functions, as well as AK-SYSi (Yun et al , 2019) method to compare the efficiency of the proposed method. The added training points of ALK-TCR-U (learning function is U function), ALK-TCR-ERF (learning function is ERF), ALK-TCR-EFF (learning function is EFF), AK-SYSi are shown in Figure 3(a)–3(d), and the initial training samples are same.…”

Section: Numerical Examplesmentioning

confidence: 99%

“…ALK model combining with an adaptive size of candidate points is proposed by Yang et al (2019). Yun et al (2019) proposed an improved AK-SYS by using a refined U learning function. Wei et al (2018) proposed an active learning procedure, which can efficiently and adaptively produce surrogate models for the failure surfaces of systems.…”

Section: Introductionmentioning

confidence: 99%

A global sensitivity analysis approach for multiple failure modes based on convex-probability hybrid uncertainty

Zhang

Liu

Guo

2020

View full text Add to dashboard Cite

Purpose This paper aims to develop a method for evaluating the failure probability and global sensitivity of multiple failure modes based on convex-probability hybrid uncertainty. Design/methodology/approach The uncertainty information of the input variable is considered as convex-probability hybrid uncertainty. Moment-independent variable global sensitivity index based on the system failure probability is proposed to quantify the effect of the input variable on the system failure probability. Two-mode sensitivity indices are adopted to characterize the effect of each failure mode on the system failure probability. The method based on active learning Kriging (ALK) model with a truncated candidate regions (TCR) is adopted to evaluate the systems failure probability, as well as sensitivity index and this method is termed as ALK-TCR. Findings The results of five examples demonstrate the effectiveness of the sensitivity index and the efficiency of the ALK-TCR method in solving the problem of multiple failure modes based on the convex-probability hybrid uncertainty. Originality/value Convex-probability hybrid uncertainty is considered on system reliability analysis. Moment-independent variable sensitivity index based on the system failure probability is proposed. Mode sensitivity indices are extended to hybrid uncertain reliability model. An effective global sensitivity analysis approach is developed for the multiple failure modes based on convex-probability hybrid uncertainty.

show abstract

“…Therefore, in order to improve the computational efficiency and to avoid a large of simulation cost in engineering reliability analysis, more recent attention has focused on surrogate model-based method, which aims to replace original performance function with approximate numerical model. Up to now, various surrogate models are widely used to balance accuracy and efficiency, such as response surface method (RSM), 17,18 neural network (NN), 19 radial basis function (RBF), 20,21 support vector machine (SVM), 22,23 Kriging model, 24–28 polynomial chaos expansion (PCE), 29 polynomial chaos kriging (PCK), 30 and deep neural network (DNN), 31 etc. Generally, to construct a surrogate model, one-shot sampling and sequential sampling are two typical methods.…”

Section: Introductionmentioning

confidence: 99%

An active learning method combining adaptive kriging and weighted penalty for structural reliability analysis

You

Zhang

Tang

et al. 2021

Proceedings of the Institution of Mechanical Engineers, Part O:

View full text Add to dashboard Cite

Reducing the surrogate model-based method computation without loss of prediction accuracy remains a significant challenge in structural reliability analysis. The unbalanced probability density, important information in critical region and information redundancy of added sample points are ignored in most of traditional surrogate-based methods, resulting in heavy computational burden. In this work, an active learning combining adaptive Kriging method and weighted penalty (AK-WP) is proposed to analyze the reliability of engineering structures. Firstly, an active learning and weighted penalty function (WPLF) is the result of integrating active learning method, weighted function and penalty function, which is proposed to find the most probable point (MPP). Meanwhile, to avoid redundant information, the best suitable MPP is determined by a proposed distance law established between the found MPP and the existing design of experiment (DoE). Secondly, the Kriging model is refined according to best suitable MPP in each iteration. Thirdly, the failure probability is estimated by the Monte Carlo sample points from the n-ball domain until the convergence condition is satisfied. The accuracy and efficiency of the proposed method are demonstrated by some numerical examples including the highly nonlinear, the small probability problems and implicit function, as well as a real engineering application.

show abstract

AK-SYSi: an improved adaptive Kriging model for system reliability analysis with multiple failure modes by a refined U learning function

Cited by 125 publications

References 32 publications

Structural System Reliability: Overview of Theories and Applications to Optimization

Structural System Reliability: Overview of Theories and Applications to Optimization

A global sensitivity analysis approach for multiple failure modes based on convex-probability hybrid uncertainty

An active learning method combining adaptive kriging and weighted penalty for structural reliability analysis

Contact Info

Product

Resources

About