论文部分内容阅读
在再生自适应子集模拟(RASS)法的基础上,提出了一种改进的再生自适应子集模拟(MRASS)法以用于结构系统的可靠性及可靠性灵敏度分析。MRASS法继承了RASS法中马尔可夫链再生、延迟拒绝、自适应马尔可夫过程及分量各自采样等优点,并改进了自适应采样过程。MRASS法通过对产生的样本点的接受率与最佳接受率进行比较,有针对性地寻找合适的建议分布方差,提高了抽样的效率。工程算例的数值结果表明:MRASS法相比于RASS法及传统的子集模拟(SS)法,在处理具有高维随机变量、小失效概率及高度非线性特点的结构系统时,有更好的适应性、稳健性及精度。
On the basis of regenerative adaptive subset simulation (RASS) method, an improved regenerative adaptive subset simulation (MRASS) method is proposed for structural system reliability and reliability sensitivity analysis. The MRASS method inherits the advantages of Markov chain regeneration, delay rejection, adaptive Markov process and component sampling in the RASS method, and improves the adaptive sampling process. The MRASS method compares the acceptance rate of the sample points generated with the best acceptance rate, finds the appropriate variance of the recommended distributions and improves the sampling efficiency. Numerical results of engineering examples show that compared with the RASS method and the traditional subset simulation (SS) method, the MRASS method has better performance in dealing with structural systems with high dimensional random variables, small failure probability and highly nonlinear characteristics Adaptability, robustness and accuracy.