Qualitative circuit models in failure analysis reasoning

Abstract

The engineering task of failure analysis involves reasoning about the behaviour of a system using appropriate models of system components and structure. This paper describes methods of qualitatively modelling electrical circuits that support the requirements for certain combinatorially demanding forms of failure analysis. Minimal models, based on zero-order quiescent conditions, are examined and a particular formulation, known as CIRQ, is shown to be an effective and efficient model with strong intuitive features. Theoretical background is given and simulation algorithms are described. These models have been used as the basis for successful failure analysis software packages that solve large-scale real applications involving repeated behaviour inference. The contributions of this work include the development of minimal qualitative circuit models and simulation algorithms, an understanding of their relationship to certain graph-theoretic properties of circuits and the relevance of such models for fault modelling in FMEA tasks. The limitations of the approach are discussed and its relation to other work is examined.