Fault-class-aware fault tree generation and analysis

Abstract

Over decades, fault-tree has remained as one of the most widely used fault management (FM) mechanisms by practitioners as a visualization/communication media, as well as a quantitative analysis tool for building dependable systems. Nonetheless, to date fault tree analysis is often unable to provide us with high-confidence results due to potential misrepresentation of the relationships among failure events. In retrospect, catastrophic system failures were almost always caused by a chain of mixed fault manifestation and interaction. Moreover, faults in FM itself may become a critical element in a fault-manifestation chain. The 1996 Ariane-5 rocket self-destruction provides us with a lesson on the devastating consequence of inappropriate FM application. Such prior lessons have inspired us to propose a fault-class-aware and FM-capability-aware fault tree analysis framework.