Development of automated computer-aided diagnostic systems using FMECA-based knowledge capture methods

Abstract

This paper describes the application in an industrial domain (commercial automotive design/maintenance) of a technique being developed at NASA Ames Research Center for building automated diagnostic tools for embedded (i.e., combined hardware/software) systems. The technique involves integrating a "real-time" sensor-information-monitoring computer process together with a static knowledge base (KB) that contains specific information about a system's architecture, its nominal behavior, and its behavior in the presence of failures and/or anomalies. The monitoring program samples status information from the system under test. The KB is then consulted by an inference engine (IE) component of the monitoring program which, based on the system's sampled status information and the system's architectural and behavioral information contained in the KB, diagnoses the potential cause(s) of any observed anomalous symptoms indicated in the system. The automated diagnosis technique described is being developed at NASA Ames Research Center for for use aboard NASA's new Stratospheric Observatory For Infrared Astronomy (SOFIA) airborne astronomy observatory. This paper demonstrates that the same technology (FMECA-based derivation of a diagnostic KB, automated computer-assisted diagnosis of complex failure situations, and computer-based repair advisory to reduce repair-time and personal-expertise requirements of repair technicians) is also applicable for industrial applications which need to reduce cost and improve service to customers. We conclude with a summary of plans for future work.