A VARIABLE ORDERING HEURISTIC BASED ON ZERO-SUPPRESSED BINARY DECISION DIAGRAMS

Two approaches have been proposed to solve the large-scale fault trees or event trees for Probabilistic Safety Assessment in a nuclear power plant. The first one consists in MCS/ZBDD, which uses ZBDDs (Zero-suppressed Binary Decision Diagrams) to implement classical MCS (Minimal Cut Sets) algorithm. The second consists in designing heuristics and strategies to reduce the complexity of the BDDs (Binary Decision Diagrams) construction. This paper was motivated to combine the MCS/ZBDD and designing heuristics for ZBDDs together. A heuristic, which took the failure rate of basic event into account and utilized that truncation could be implemented on ZBDDs during the calculating process, was proposed. This heuristic accelerated the analysis progress by bringing forward the truncation and reducing the complexity of the intermediate ZBDDs. RiskA, a Zero-suppressed Binary Decision Diagram package extended to safety and reliability analysis, has adopted this heuristic. RiskA's truncation strategies, which had some relations with the ordering scheme, were also introduced. The correctness and efficiency of this new heuristic were verified by some practical models' analyses.