An Integrated, Deep-shallow Expert System for )multi-leve~ Diagnosis of Dynamic Systems Deep-shallow Expert System for Multi-level Diagnosis of Dynamic Systems

ABSI1lACF An integrated e'Q>en system architecture and strategy for diagnosis of dynamic systems with feedback loops a¥ s~chronous or asynchronous state transitions is presented. The dynamic system under diagnosis is modeled' using structural and behavioral representations in multiple levels of abstraction I The diagnosis process inte~tes shallow ~d deep expertise. It recursively navigates through the structural hierarchy; and at each level tries the shallow expertise first If it fails it switches to deep, simulation based expertise. A multilevel simulator assists the diagnosis process in verification and elimination of hypothesized suspects. TIle simulator shifts on demand thrdugh several levels. from coarse qualitative modeling to detailed quantitative modeling. Knowledge of pathological behavior (failure modes) of.lower level cOmponents is incorporated in the simulator. Learning is exhibited as deep-to-shallow expertise transfer, as well as up the abstraction levels of the simulator itself, to improve future efficiency. In addition, knowledge about pathological behavior can be used for off-line training by artificially generating new cases for diagnosis.