文献类型：期刊文献

中文题名：基于结构无损检测出裂纹概率下的可靠性计算方法

英文题名：Reliability analysis method based on the probabilistic fracture detection

作者：刘长虹[1];杨存攀[1];李勇[1];王少楠[1];程世玮[1];申一淇[1]

机构：[1]华东理工大学机械与动力工程学院承压系统安全科学教育部重点实验室,上海200237

年份：2017

期号：1

起止页码：51

中文期刊名：制造技术与机床

外文期刊名：Manufacturing Technology & Machine Tool

收录：CSTPCD;;Scopus;北大核心:【北大核心2014】；

基金：国家自然基金(11372111)

语种：中文

中文关键词：马尔科夫链;蒙特卡洛方法;裂纹;检出概率;Metropolis-Hastings方法

外文关键词：Markov chain; Monte Carlo sampling; fracture; probability of detection; Metropolis-Hastings algorithm（MH）

摘要：针对结构的无损检出概率密度函数,提出一种可靠性计算方法。首先定义以方差固定,均值随机游走策略的正态分布函数,作为Metropolis-Hastings(MH)抽样方法的提议分布函数,从而建立了一条马尔可夫数据链。再根据无损检出概率分布函数的均值与方差数值积分值,并结合马尔可夫数据链的平均值曲线作为马尔科夫链收敛性准则。从数据链中抽取出一条稳定的数据链,就可以利用蒙特卡洛抽样法计算出结构的失效概率。此外通过对重要性抽样、接受-拒绝抽样以及MH法3种算法比较,得出MH方法是一种有效的抽样算法。最后讨论了在不同的无损检出概率下,对压力容器贝叶斯可靠性分析结果的影响。
In this paper,a reliability analysis method is proposed for the nondestructive detection probability density function of structures.First,we define the normal distribution function with the variance fixed and mean random walk strategy,as a proposed distribution function for the Metropolis-Hastings（MH）sampling method,thus a Markov data chain is established.Then,according to the mean value and the variance integral value of the probability distribution function of non-destructive detection,combined with the mean value curve of the Markov chain as the Markov chain convergence criterion.By extracting a stable data link from the Markov chain,the Monte Carlo method can be used to calculate the failure probability of the structure.In addition,by comparing the importance sampling,the acceptance-rejection sampling and the MH algorithm,it is concluded that the MH method is the most effective sampling algorithm.Finally,the influence of different nondestructive detection probability function on the Bayesian reliability analysis of pressure vessels is discussed.