Characterizing non-intermittent faults

Abstract

A faulty component that behaves consistently over time is said to behave non-intermittently. For any given set of inputs, such a component will always generate the same output. Assuming that components fail non-intermittently is a common simplifying strategy used by diagnosticians, because (1) many real-world devices often fail this way, (2) this strategy removes the need to repeat experiments, and (3) this strategy allows information from independent examples of system behavior to be combined in relatively simple ways. This paper extends the formal framework for diagnosis developed in [7, 12] to allow nonintermittency assumptions. In addition we show how the definitions can be easily integrated into ATMS-based diagnosis engines. Within our formulation, components can be individually assumed to be intermittent or nonintermittent.