可靠性(半导体)
计算机科学
采样(信号处理)
非线性系统
替代模型
极限(数学)
代表(政治)
灵敏度(控制系统)
过程(计算)
功能(生物学)
随机变量
随机过程
不确定度量化
数学优化
可靠性工程
算法
数学
工程类
统计
机器学习
进化生物学
政治学
电子工程
法学
功率(物理)
计算机视觉
量子力学
数学分析
生物
政治
物理
操作系统
滤波器(信号处理)
作者
Weifei Hu,Jiquan Yan,Feng Zhang,Chao Jiang,Hongwei Liu,Hyunkyoo Cho,Ikjin Lee
摘要
Abstract A mature digital twin (DT) is supposed to enable engineers to accurately evaluate the real-time reliability of a complex engineering system. However, in practical engineering problems, reliability analysis (RA) often involves nonlinear, implicit, and computationally expensive relationships between the performance and uncertain parameters, which makes it very challenging to conduct time-dependent reliability analysis (TRA) instantly and accurately for a DT. This article proposes a new surrogate-based time-dependent reliability analysis (STRA) method for a DT, specifically making the following three contributions: (i) the number of discrete time nodes used to convert the stochastic processes into a series of random variables in the expansion optimal linear estimation process is dynamically selected, leading to a good tradeoff between the accurate representation of stochastic processes and fast reliability evaluation; (ii) based on Voronoi partition sampling and a modified leave-one-out cross-validation procedure, multiple sensitive subdomains in each iteration are selected simultaneously to guide adaptive sampling at the insufficiently fitted vicinity of the limit state function, which helps accurately calculate the probability of failure and reduce the number of design-of-experiment (DoE) samples; and (iii) an improved weighted expected feasibility function is proposed considering the importance of each sample and the sensitivity of the subdomain to which it belongs, which further improves the sampling efficiency. The proposed STRA method is applied to the TRA of a numerical model, a corroded beam structure, and a cutterhead of a tunnel boring machine to demonstrate its effectiveness for realistic DT applications.
科研通智能强力驱动
Strongly Powered by AbleSci AI