An algorithm for diagnostic reasoning using TFPG models in embedded real-time applications

Embedded diagnostic reasoners require compact modeling representations and efficient reasoning algorithms given limited available computational resources. Timed failure propagation graphs (TFPG) are compact representations used to model failure causes and progressions of conditions that are symptoms of failure occurrence, together with the temporality and likelihood of these symptom progressions and the observation of some of the aberrant conditions. Algorithms for design-time diagnosability analysis using TFPG models have previously been reported, but these algorithms have different design objectives leading to different computational strategies and optimization criteria. This paper presents and discusses an algorithm specifically designed for efficient failure isolation based on reported observations of abnormal conditions that is suitable for use in embedded real-time applications.