摘要
在定量化风险评估中,针对变量分布参数存在依赖时的混合不确定性传播问题,提出一种考虑分布参数完全依赖、部分依赖以及独立传播情形的双层混合不确定性刻画与传播框架,内外层分布参数不确定性分别用概率分布与可能性分布刻画,采用蒙特卡罗模拟法与模糊扩展原则相结合的数值求解方法。针对认知不确定性分布参数依赖性,构建了统一的认知不确定性分布参数依赖性模型,并给出依赖性系数的概念。为实现认知不确定性分布参数独立性采样,设计了基于D-S证据理论与随机集相结合的不确定性传播算法,相比于概率刻画下的双层蒙特卡罗方法,计算代价有效降低。以某型氢氧发动机贮箱共底漏气率为算例,验证本文方法的有效性与可行性。
In view of hybrid uncertainty propagation with parameters dependency of variables in quantified risk assessment,a two-level hybrid uncertainty presentation and propagation framework considering parameters total dependency,partial dependency and independency was proposed,in which inner and outer parameters were specified with probability and possibility,and the numerical values were calculated by Monte Carlo simulation and fuzzy extension principle. Based on the epidemic uncertainty parameter dependency,a model with epidemic uncertainty parameters dependency and a dependency coefficient were constructed. An uncertainty propagation algorithm based on the D-S evidence theory and random set theory for parameters sampled independently was built,which,compared with the two-level Monte Carlo method presented with probability,reduced the time costs largely. Leakage rate of the hydrogen and oxygen co-bottom tank was taken as an example,and the effectiveness and feasibility of the proposed method were validated.
出处
《北京航空航天大学学报》
EI
CAS
CSCD
北大核心
2017年第12期2439-2448,共10页
Journal of Beijing University of Aeronautics and Astronautics
