Article
Engineering, Multidisciplinary
Linxiong Hong, Bin Shang, Shizheng Li, Huacong Li, Jiaming Cheng
Summary: Recently, many studies have focused on structural reliability analysis, and the Kriging-based active learning method has gained popularity. Various Kriging-based learning functions have been proposed and proven effective in different tasks. However, no single learning function consistently outperforms others in all tasks, posing a challenge in selecting the most appropriate function for engineering applications. This paper proposes a portfolio allocation approach inspired by the multi-armed bandit strategy to address the issue, where learning functions are selected online based on their past performance. Three numerical examples and two engineering applications are used to validate the effectiveness of the proposed method.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2023)
Article
Engineering, Civil
Changle Peng, Cheng Chen, Tong Guo, Weijie Xu
Summary: Reliability Analysis (RA) is critical in structural design and performance evaluation. This study proposes a novel learning function, SEUR, for surrogate model-assisted RA to improve efficiency and accuracy. The SEUR function is demonstrated to be more effective and efficient in dealing with nonlinear problems, small probabilities, and complex limit states.
Article
Engineering, Multidisciplinary
Yanjin Wang, Hao Pan, Yina Shi, Ruili Wang, Pei Wang
Summary: This paper proposes a new active learning Kriging-based method for estimating the failure probability in engineering structures. The method improves efficiency and accuracy through a penalty learning function and a distance constraint term, and convergence conditions are investigated using an error-based stopping criterion. Numerical examples demonstrate the accuracy and effectiveness of the proposed method.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2023)
Article
Computer Science, Interdisciplinary Applications
Hanyan Huang, Zecong Liu, Hongyu Zheng, Xiaoyu Xu, Yanhui Duan
Summary: This paper proposes a Co-kriging-based multi-fidelity sequential optimization method named proportional expected improvement (PEI), which aims to be more efficient for global optimization and to evaluate the costs and benefits of candidate points from different levels of fidelity more reasonably. Experiments show that the proposed method can better search for the global optimum, and the KL divergence can more significantly describe the relationship between high and low fidelity.
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION
(2023)
Article
Engineering, Aerospace
Yingshi Hu, Zhenzhou Lu, Ning Wei, Xia Jiang, Changcong Zhou
Summary: This paper proposes a Kriging surrogate model based method to estimate the safety lifetime of Multi-mode Time-Dependent Structural System (MTDSS). The method utilizes extremum learning function, Advanced First Failure Instant Learning Function (AFFILF), and Candidate Sample Pool (CSP) reduction strategy to accurately and efficiently search for the safety lifetime.
CHINESE JOURNAL OF AERONAUTICS
(2022)
Article
Computer Science, Interdisciplinary Applications
Zhendong Guo, Qineng Wang, Liming Song, Jun Li
Summary: The study introduces a new infill criterion called Filter-GEI for addressing sample assignment issue in multi-fidelity optimization. By considering correlations between HF and LF models and adding an adaptive filter function on top of the GEI acquisition function, Filter-GEI efficiently allocates HF and LF samples to achieve a good balance between local and global search, with further improvement in efficiency through infilling multiple HF and LF samples in each iteration along with parallel computing. Tests on mathematical toy problems and an engineering problem demonstrate the effectiveness of the proposed algorithm.
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION
(2021)
Article
Engineering, Industrial
Guofa Li, Tianzhe Wang, Zequan Chen, Jialong He, Xiaoye Wang, Xuejiao Du
Summary: This study proposes a new method, RBIK-SS, which combines the reliability analysis method based on importance sampling and k-medoids clustering with subset simulation, to estimate small failure probabilities. The method replaces the MCS sample pool in RBIK with a smaller SS population to overcome memory limitations. The results show that RBIK-SS can solve rare failure events with satisfactory accuracy and efficiency.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2023)
Article
Computer Science, Interdisciplinary Applications
Leshi Shu, Ping Jiang, Yan Wang
Summary: This work proposes a multi-fidelity Bayesian optimization approach that utilizes hierarchical Kriging to reduce optimization costs, quantifies the impact of high and low-fidelity samples based on expected further improvement, and introduces a novel acquisition function to determine the location and fidelity level of the next sample simultaneously. The proposed approach is compared with state-of-the-art methods for multi-fidelity global optimization and shows that it can achieve global optimal solutions with reduced computational costs.
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION
(2021)
Article
Computer Science, Interdisciplinary Applications
Ying Huang, Jian-Guo Zhang, Lu-Kai Song, Xue-Qin Li, Guang-Chen Bai
Summary: In this paper, a novel unified reliability evaluation framework is proposed to address the high-efficacy computing and correlation quantification issues in evaluating complex structures with multiple dangerous sites. The proposed framework includes an optimized Kriging surrogate-based improved importance sampling method for efficient computing, and a novel failure correlation analysis strategy for quantifying failure correlations among the multiple dangerous sites. The effectiveness of the framework is validated through a reliability evaluation of a high-pressure turbine rotor.
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION
(2023)
Article
Computer Science, Hardware & Architecture
Nan Ye, Zhenzhou Lu, Xiaobo Zhang, Kaixuan Feng
Summary: In this article, a metamodel-based directional importance sampling method (Meta-DIS-AK) is proposed to improve the efficiency of reliability analysis. The main novelty of Meta-DIS-AK lies in constructing the quasi-optimal DIS density (DIS-D) using the Kriging model and accurately extracting DIS-D samples through a simple rejection sampling algorithm. Compared with existing methods, Meta-DIS-AK overcomes the difficulties of constructing DIS-D and extracting important direction vector samples, while maintaining the advantages of DIS in dealing with high-dimensional and small failure probability problems.
IEEE TRANSACTIONS ON RELIABILITY
(2023)
Article
Engineering, Industrial
Seonghyeok Yang, Hwisang Jo, Kyungeun Lee, Ikjin Lee
Summary: This paper proposes a new active learning function, called expected system improvement (ESI), for system reliability analysis to predict how updates on component reliability affect system reliability. It shows that the proposed method outperforms existing methods in terms of the number of function evaluations and computational time.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2022)
Article
Computer Science, Interdisciplinary Applications
Haizhou Yang, Seong Hyeong Hong, Yi Wang
Summary: This paper presents a novel computation-aware multi-fidelity surrogate-based optimization methodology and a new sequential and adaptive sampling strategy based on expected improvement reduction. It improves the exploration and convergence rate of the optimization process under a fixed computational budget.
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION
(2022)
Article
Computer Science, Interdisciplinary Applications
Tong Zhou, Tong Guo, You Dong, Yongbo Peng
Summary: This paper proposes a computationally-cheap integral learning function called EMVR for structural reliability analysis. It derives a closed-form expression for the inner integral based on the tractable definition of margin volume and defines a confined integral domain for the outer integral by exploiting the locality of the integrand. The results show that EMVR outperforms existing learning functions in terms of both computational accuracy and efficiency.
COMPUTERS & STRUCTURES
(2023)
Article
Engineering, Mechanical
Jiaqi Wang, Zhenzhou Lu, Lu Wang
Summary: This paper proposes an efficient method to estimate the FP-GS using reliability updating, avoiding the time-consuming double-loop structure analysis. By utilizing the likelihood function and adaptive Kriging model, the unconditional FP and all conditional FPs can be estimated simultaneously.
PROBABILISTIC ENGINEERING MECHANICS
(2024)
Article
Engineering, Civil
Cheng Chen, Yanlin Yang, Hetao Hou, Changle Peng, Weijie Xu
Summary: This study explores the use of Co-Kriging metamodeling for global response prediction under the presence of structural uncertainties in real-time hybrid simulation (RTHS). The results demonstrate that Co-Kriging can effectively reduce the number of RTHS tests in the laboratory and significantly improve metamodel accuracy for global prediction of structural response under uncertainties.
EARTHQUAKE ENGINEERING & STRUCTURAL DYNAMICS
(2022)