Modeling digital circuits for trouble-shooting: an overview

An overview of a model-based troubleshooting program that incorporates a domain-independent diagnosis engine based on J. de Kleer and B.C Williams´ General Diagnostic Engine (Artificial Intelligence, vol.32, no.1, p.97-130, April, 1987) is presented. The primary input to the program is a model of a digital circuit that is a network of components and connections. Each component has a description of its dynamic time-dependent behavior and each connection transmits signals between components. The secondary input to the program is a description of the stimuli presented to the circuit and observations of its actual responses. The model uses those stimuli to predict what the outcomes of observations ought to be. When discrepancies are discovered, the program produces a list of components that could be responsible for the discrepancies, ranked by their relative likelihood. The program interactively suggests what observations should be made next in order to discriminate among these possibilities, then uses the new observations to incrementally focus on the correct diagnosis. Eight modeling principles broken up into three sets are discussed. One set of principles concerns how the structure of a given circuit should be represented. A second set of principles concerns the representation of circuit behavior. The final set of principles concerns what knowledge about failures should be represented explicitly