Improved efficiency in qualitative fault tree analysis

Abstract

The fault tree diagram itself is an excellent way of deriving the failure logic for a system and representing it in a form which is ideal for communication to managers, designers, operators, etc. Since the method was first conceived, algorithms to derive the minimal cut sets have worked directly with the fault tree diagram using either bottom-up or top-down approaches. These conventional techniques have several disadvantages when it comes to analysing the fault tree. For complex systems an analysis may produce hundreds of thousands of minimal cut sets, the determination of which can be a very time-consuming process. Also, for large fault trees it may not be possible to evaluate all minimal cut sets, so methods to identify those event combinations which provide the most significant contributions to the system failure are evoked. Such methods include probabilistic or order culling to reduce the problem to a practical size, but they can also create considerable inaccuracies when it comes to evaluating top event probability parameters. This paper describes how the binary decision diagram method can be employed to evaluate the minimal cut sets of a fault tree efficiently and without the need to use approximations such as order culling.