On a new measure of system diagnosability in presence of hybrid faults

This paper deals with the problem of testing the diagnosability of a system when the units of the system may fail permanently or intermittently. It takes into account the fact that different units of a system have different probabilities of failing permanently and intermittently. For each unit, these probabilities are computed assuming a continuous parameter Markov model. It has also been shown that, with this choice of model, the incompleteness of diagnosis (i.e., an intermittently failing unit classified as faultfree) can be avoided up to any level of confidence.