Fault tree analysis and binary decision diagrams

Abstract

Fault tree analysis is now commonly used to assess the adequacy, in reliability terms, of industrial systems. For complex systems, an analysis may produce thousands of combinations of events which can cause system failure (minimal cut sets). The determination of these minimal cut sets can be a very time consuming process even on modern high speed digital computers. Also, if the fault tree has many minimal cut sets, calculating the exact top event probability will require extensive calculations. For many complex fault trees this requirement is beyond the capability of the available machines, thus approximation techniques need to be introduced resulting in loss of accuracy. This paper describes the use of a binary decision diagram for fault tree analysis and some ways in which it can be efficiently implemented on a computer. The work to date shows a substantial improvement in computational effort for large, complex fault trees analysed with this method in comparison to the traditional approach. The binary decision diagram method has the additional advantage that as approximations are not required, exact calculations for the top event parameters can be performed.