AK-SYS: An adaptation of the AK-MCS method for system reliability
AK-SYS: An adaptation of the AK-MCS method for system reliability
复制标题
DOI：
10.1016/j.ress.2013.10.010
复制
发表时间：
2014-03
期刊：
影响因子：
--
通讯作者：
W. Fauriat;N. Gayton
中科院分区：
文献类型：
--
作者：
W. Fauriat;N. Gayton
在过去的二十年里，已经提出了许多研究工作来评估涉及非常耗时的力学模型的结构的失效概率。基于Kriging的代理模型方法，如高效全局可靠性分析(ERRA)或主动学习和基于Kriging的蒙特卡罗仿真(AK-MCS)方法，都是非常有效的方法，且各有优势。EGRA非常适合于评估小概率，因为代理可以用来对任何群体进行分类。AK-MCS是针对给定的群体而构建的，并且不需要执行主动学习过程的优化程序。因此，它更容易实现，并且更有可能将计算工作花费在具有显著概率内容的区域上。在评估系统可靠性时，文献中广泛使用解析法和一阶近似法。然而，在本文中，我们更侧重于抽样技术，并考虑到最近ERGRA方法对系统的适应，提出了一种将AK-MCS方法应用于系统可靠性的策略。提出了“主动学习和基于克里格的系统可靠性方法”的AK-sys方法。通过不同的算例说明了该算法的高效性和准确性。
A lot of research work has been proposed over the last two decades to evaluate the probability of failure of a structure involving a very time-consuming mechanical model. Surrogate model approaches based on Kriging, such as the Efficient Global Reliability Analysis (EGRA) or the Active learning and Kriging-based Monte-Carlo Simulation (AK-MCS) methods, are very efficient and each has advantages of its own. EGRA is well suited to evaluating small probabilities, as the surrogate can be used to classify any population. AK-MCS is built in relation to a given population and requires no optimization program for the active learning procedure to be performed. It is therefore easier to implement and more likely to spend computational effort on areas with a significant probability content. When assessing system reliability, analytical approaches and first-order approximation are widely used in the literature. However, in the present paper we rather focus on sampling techniques and, considering the recent adaptation of the EGRA method for systems, a strategy is presented to adapt the AK-MCS method for system reliability. The AK-SYS method, “Active learning and Kriging-based SYStem reliability method”, is presented. Its high efficiency and accuracy are illustrated via various examples.