Learning Diagnostic Knowledge for Continuous-Variable Dynamic Systems
01 January 1989
Continuous-variable dynamic systems, such as chemical refineries, electrical power plants, and the human cardiovascular system, exhibit time varying behavior through continuous changes in parameter values. Diagnosis of such systems is difficult because they are dynamic, continuous in nature, and have feedback. These properties make it very difficult for a domain expert to enumerate all the manifestations of a fault or, more to the point, all the faults that could produce a given manifestation. We present a novel method for mechanically deriving diagnostic knowledge from models of a continuous-variable dynamic system.