Fault tree (FT) is a simple, visual, popular and standardized notation for representing relationships between a fault in a system and the associated events. FTs are widely used for supporting products and systems in diverse industries like process control, avionics, aerospace, nuclear power systems, etc. where they are used to capture specialized and experiential knowledge for diagnosis and maintenance. FTs are also used to represent safety requirements of a system, obtained during the hazard analysis phase of the system development cycle. However, a problem that prevents more analytical use of FT is their lack of rigorous semantics. Users' understanding of an FT depends on the clarity and correctness of the natural language annotations used to label and describe various parts. Moreover, it is not clear how to adapt the FT notation to represent temporal relationships between faults and events in dynamic systems. We propose to augment the FT notation by adding simple temporal gates to capture temporal dependence between events and faults. We propose techniques to perform qualitative analysis of such temporal fault trees (TFT) to detect the causes of the top event fault by matching the TFT with the trace (or log) of the system activities. We present two algorithms for depth-first traversal and cut-set computations for a given TFT that can be used for diagnosis based on TFTs. (C) 2002 Elsevier Science B.V. All rights reserved.
