故障树随机生成及变量排序

摘要

To cope with the problems related to benchmark size and diversity for performance evaluation of fault tree analysis methods, based on the idea of random generation of fault trees, this work proposed six important structural characteristic parameters of various fault trees, a top-down generation algorithm ac-cording to these parameters, which included a random tree skeleton generation algorithm and a randomly repeated leaf sequence generation algorithm, then illustrated the benchmark application with variable orde-ring heuristic performance evaluation. The evaluation shows that; deep (DEEP) heuristic is far superior to the wide (WIDE) heuristic; the superiority of the WIDE heuristic depends on the number of repeated leav-es in the fault tree; improved weighting deep (WDEEP) heuristic is a complementary strategy for DEEP heuristic and not able to replace it.%针对故障树分析方法性能评价研究中存在的测试基准规模和多样性问题,基于故障树样本随机生成的思想,确定了故障树6个主要结构特征,并依据这些结构特征给出了自顶向下地生成算法,包括随机树骨架生成算法和随机重复度叶子序列生成算法.通过变量排序策略性能评价给出了测试基准的示范应用.性能评价表明:深度策略(DEEP)性能远优于广度策略(WIDE)；WIDE策略的优越性依赖于重复叶子节点数量；改进带权值深度策略(WDEEP)是DEEP策略的互补策略并不能够代替DEEP策略.