Efficient basic event orderings for binary decision diagrams

Significant advances have been made in methodologies to analyse the fault tree diagram. The most successful of these developments has been the binary decision diagram (BDD) approach. This approach has been shown to improve both the efficiency of determining the minimal cut sets of the fault tree and also the accuracy of the calculation procedure used to determine the top event parameters. To utilise the BDD approach the fault tree structure is first converted to the BDD format. This conversion can be accomplished efficiently but requires the basic events in the fault tree to be placed in an ordering. A poor ordering can result in a BDD which is not an efficient representation of the fault tree logic structure. The advantages to be gained by utilising the BDD technique rely on the efficiency of the ordering scheme. Alternative ordering schemes have been investigated and no one scheme is appropriate for every tree structure. Research to date has not found any rule based means of determining the best way of ordering basic events for a given fault tree structure. The work presented in this paper takes a machine learning approach based on genetic algorithms to select the most appropriate ordering scheme. Features which describe a fault tree structure have been identified and these provide the inputs to the machine learning algorithm. A set of possible ordering schemes has been selected based on previous heuristic work. The objective of the work detailed in the paper is to predict the most efficient of the possible ordering alternatives from parameters which describe a fault tree structure.