PERFORMANCE ENHANCED NAVIGATION USING NEURAL NETWORK TECHNOLOGY (PENANT)

SYNETICS PROPOSES A NEW METHOD OF FAULT DETECTION AND IDENTIFICATION (FDI) FOR AIRCRAFT NAVIGATION SYSTEMS, TAKING ADVANTAGE OF BOTH NEURAL NETWORK TECHNOLOGY AND ADVANCED FILTERING TECHNOLOGY. NEURAL NETWORKS OFFER A GREAT DEAL OF PROMISE IN THE GENERAL AREA OF FAULT DETECTION, SINCE THEIR HIGHLY PARALELLIZED STRUCTURE AVOIDS MOST OF THE COMPUTATIONAL DIFFICULTIES ASSOCIATED WITH A TIME-TAGGED FAULT TREE. IN PARTICULAR, RECENT WORK INDICATES THAT THE MULTI-LAYER PERCEPTRON ARCHITECTURE CAN GIVE GOOD FDI PERFORMANCE FOR SOME APPLICATIONS. IN ORDER TO TAKE ADVANTAGE OF NEURAL NETS FOR THE AIRCRAT FDI APPLICATION, IT IS NECESSARY TO FIRST COMPENSATE FOR THE MISSION -DEPENDENT, TIME-VARYING, NON-STATIONARY NATURE OF TYPICAL NAVIGATION ERROR PROCESSES. ALSO, IT IS NECESSARY TO CREATE NEURAL NET INPUTS OVER A LONG ENOUGH TIME HORIZON TO SUCCESSFULLY DIAGNOSE FAULTS, WHILE PRESERVING THE ABILITY TO ELIMINATE THE FAULTY INFORMATION. SYNETICS PROPOSES TO ACCOMPLISH THIS BY PRE-FILTERING THE SENSOR DATA THROUGH A SYNTHETIC MEASUREMENT GENERATOR, BASED ON SYNETICS NET INFORMATION APPROACH (NIA) DISTRIBUTED KALMAN FILTER ALGORITHM. PHASE I WILL INCLUDE ACTUAL TRAINING AND TESTING OF A PROTOTYPE HYBRID NEURAL NET/SMG FOR A REDUCED-ORDER FDI PROBLEM.