The Design and Application of Fault Tree Benchmark

To cope with the problems related to benchmark size and diversity for performance evaluation of fault tree analysis methods, this paper defines a reliable benchmark and then proposes a two-step approach to create such a reliable benchmark by randomly generating a set of fault trees. As an illustration, two widely used variable ordering heuristics TD and DF are tested on 1200 random fault trees. Following conclusions are derived: (1) The impact of changing logic and structure parameters on performance is not significant. (2) With the increase of repeated events, TD has an increased chance to produce smaller BDDs than DF. Nevertheless, DF always has more than 80% chance to produce smaller BDDs.