Behaviour-Based Methodology for Fault Tree Generation
Andrew John Rae · The University of Queensland · 2007
This thesis presents a new theory of fault trees for complex systems, and a method for their generation. The theory treats faults as behaviours, and fault-tree gates as operations on those behaviours. Fault tree analysis is an important and widely used technique for understanding the behaviour of safety-critical systems. The development of new technologies with more sophisticated interactions between components, and hence more complicated causal relationships between failure modes, has made traditional methods of fault tree synthesis difficult if not obsolete. Consequently, this thesis addresses a matter of ongoing concern in the safety analysis community. The thesis argues that it is feasible and effective to automate the generation of fault trees by describing systems using hierarchically structured models, with component interactions described in terms of behaviours. A detailed methodology for generating fault trees is presented. The methodology includes consideration of design faults, hardware failures, and operator errors. A prototype tool called Eucalypt is used to demonstrate the methodology on four realistic case studies. Eucalypt provides semi-automated support for system modelling and fault generation, and fully automatic fault tree synthesis.