Hybrid neural network systems for NASA ground operations

This paper describes work involving hybrid neural network systems for use by NASA ground controllers at Marshall Space Flight Center, Huntsville, Alabama. First, the authors discuss a prototype system employing a conceptual graph knowledge representation front end interfacing with a counterpropagation neural network for Space Shuttle subsystem anomaly detection. Second, the authors discuss a planned architecture in development for interfacing a neural network front end preprocessing system with a commercially available expert system for Space Station User Operations Facility (UOF) ground operations