Applying neural networks and other AI techniques to fault detection in satellite communication systems

A demonstration program has been completed to apply various artificial intelligence techniques including, neural networks, expert systems, and case-based reasoning to fault detection in satellite communications systems. The GMM program implemented these techniques for global military satellite communications maintenance. Neural networks were designed and trained to analyze incoming built-in-test (BIT) fault signatures from the satellite communications terminal. Expert systems were developed to embed diagnostic knowledge relating to equipment maintenance. The prototype hybrid system uses neural filters to detect faults, which are further processed by expert systems to classify the faults and provide repair directions.