Intelligent fault isolation and diagnosis for communication satellite systems

Abstract

NASA-Lewis Research Center recently completed the design of a Ka-band satellite transponder system, as part of the Advanced Communication Technology Satellite (ACTS) System. To enhance the reliability of this satellite, NASA funded The University of Akron to explore the application of an expert system, in order to provide this satellite with autonomous diagnosis capability. The result of this research was the development of a prototype diagnosis expert system, called Fault Isolation and Diagnosis EXpert (FIDEX). FIDEX is a frame-based system that uses hierarchical structures to represent such items as the satellite's subsystems, components, sensors, and fault states. This overall frame architecture integrates these hierarchical structures into a lattice that provides a flexible representation scheme and facilitates system maintenance. To overcome limitations on the availability of sensor information, FIDEX uses an inexact reasoning technique based on the incrementally acquired evidence approach that was developed by Shortliffe during his MYCIN project. The system is also designed with a primitive learning ability through which it maintains a record of past diagnosis studies. This permits it to search first for those faults that are most likely to occur. And finally, FIDEX can detect abnormalities in the sensors that provide information on the transponder's performance. This ability is used to first rule out simple sensor malfunctions. The overall design of the FIDEX system, with its generic structures and innovative features, makes it an applicable example for other types of diagnostic systems. This article discusses these aspects of FIDEX, and illustrates how they can be applied to fault diagnostics in other types of space systems.