IMPLEMENTATION OF FAULT TOLERANT CONTROL ALGORITHMS USING NEURAL NETWORKS

SYSTEMATIX, INC. PROPOSES TO DEMONSTRATE THE FEASIBILITY OF APPLYING RECENT ADVANCES IN NEURAL COMPUTING TO THE PROBLEM OF DESIGNING HIGHLY RELIABLE SPACE POWER SYSTEMS BY IMPLEMENTING A SET OF CONVENTIONAL CONTROL ALGORITHMS USING NEURAL NETWORK TECHNIQUES. THIS SET WILL INCLUDE THE CLASSICAL PID (PROPORTIONAL INTEGRAL DERIVATIVE) CONTROL ALGORITHM, LATCHING AND NONLATCHING ALARM BLOCKS AND CURRENT-TIME FUNCTIONS. CONCEPTUALLY THESE NEURAL-BASED CONTROL SYSTEM BUILDING BLOCKS WILL PERFORM LIKE THEIR TRADITIONAL COUNTERPARTS WHICH ARE WELL-UNDERSTOOD AND COMMONLY USED BY SYSTEM DESIGNERS. THE UNDERLYING IMPLEMENTATION, HOWEVER, WILL BE FOUNDED ON HIGHLY DISTRIBUTED NEURAL NETWORKS DESIGNED TO RUN AS SOFTWARE IN A NETWORK OF MICROPROCESSORS, WITH THE POTENTIAL FOR ADAPTATION TO DEDICATED NEURAL PROCESSORS. THE ULTIMATE GOAL IS THE INTEGRATION OF HIGHLY DEVELOPED AND WELL-UNDERSTOOD POWER CONTROL TECHNIQUES WITH NEURAL SYSTEMS, WHOSE INHERENT FAULT TOLERANCE, SPEED AND LIGHT WEIGHT ARE ADVANTAGEOUS IN SPACE APPLICATIONS.