Abstract
Real time condition-based evaluation of system health must not only be efficient, but also produce usable and expressive results. To this end, this paper presents a set-theoretic formulation of fault trees. Such a formulation provides a usable scaffolding on which ensembles of machine health measurement techniques may eventually operate. Initially, we present a negation free formulation of a forest of fault trees as a set of 2-level sum-of-products expressions. Given this formulation, we express measures for certainty and specificity, both of which further qualify the various well-studied measures for severity. This formulation is subsequently refined to represent multi-state systems, non-coherent valuations, and node sharing-all necessary for practical monitoring solutions. Finally, we present an evaluation rule embodying these refinements and analyze its complexity. Examples pertaining to unattended ocean systems illustrate these concepts.