Auto-generating model-based reasoners through dynamic simulation

In recent years, a significant focus has been placed on the development and implementation of advanced prognostic and health management (PHM) technologies for air vehicles both in military and industrial applications. The term PHM encompasses anomaly, diagnostic and prognostic algorithms as well as higher level reasoning algorithms for isolating root causes of faults/failures and directing optimal operational or maintenance actions. In such systems, two current deficiencies exist. First, for a variety of reasons, component and subsystem interactions in such systems are poorly realized. The issue manifests itself as multiple dependent "boxes" indicating faults with shotgun tests or valuable domain expertise required to de-conflict and reduce ambiguity groups. Secondly, complex systems still largely rely on expert rule-bases for reasoning which are notoriously difficult to maintain over a life cycle and are prone to logical conflicts. This paper begins to address these deficiencies by outlining a simulation-based process for assisting the user in: 1) realizing complex system interactions for optimal PHM system design; and 2) building and maintaining model-based reasoning architectures where decisions and conclusions naturally precipitate out of a more manageable system dependency model. Concepts are reinforced and demonstrated with a Matlab/Simulink example of a complex dynamic system and an analysis test bench for reporting system wide dependencies, fault propagation and consolidated model-based reasoning architectures.