SBIR-STTR Award

Neural networks will be applied to monitor distributed systems on space station, with the eventual goal of developing hybrid neural network, expert system, and conventional software systems to monitor and control complicated systems. The Phase I effort will concentrate on using a network to monitor one distributed system and differentiate between normal and abnormal operating conditions and on analyzing the trained network to derive its strategies. The Central Thermal Bus will be simulated and the resulting model will provide simulated sensor data for network training and testing. Normal operating conditions will be simulated, as well as specific system failures. Several networks will be trained using various transforms of sensor input and the best network selected for analysis. Direct observation, and calculation of the response of each node in the network to various input patterns will be used to derive network strategies. The successful techniques for monitoring complicated systems can be extended to be distributed systems in space with human-like intelligence. Further applications abound in process control and military systems.A system composed of neural networks, expert systems, and conventional software will be capable of human-like performance in monitoring and controlling complex systems and will have wide commercial application.neural networks, artificial intelligence, distributed systems, sensor monitoringSTATUS: Phase I Only