Diagnosis of condition systems using causal structure

In a system composed of multiple subsystems, each subsystem imposes constraints on the behavior of the system through its interactions with other subsystems and with the external environment. A failure occurs when the subsystem no longer imposes its constraints, thus permitting unexpected observed behaviors from the system. In this paper, we consider systems composed of interacting condition systems, a form of Petri net with input and output signals defining its interactions with other subsystems. We present a method of transforming the system model into a causal model defining which subsystems can potentially affect other subsystems. When observed system outputs are not consistent with expected behaviors, then the causal model is analyzed to present a set of diagnostic hypotheses. It is shown that this set of hypotheses is a superset of the subsystems which could account for the failure.