LEARNING OPTIMAL CONTROL SYSTEM FOR FIBER PLACEMENT PROCESSING

SIGNIFICANT PROGRESS HAS BEEN ACHIEVED IN THE DEVELOPMENT OF NEW EQUIPMENT AND MODIFICATION OF EXISTING FILAMENT WINDING EQUIPMENT TO ACCOMMODATE IN SITU FABRICATION OF THERMOPLASTIC PARTS. ADVANTAGES OF ON-LINE CONSOLIDATION THROUGH FIBER PLACEMENT ARE CONTINUOUS WINDING OF THICK-WALLED PARTS, ELIMINATION OF OVEN OR AUTOCLAVING CURING, LOW VOID CONTENT, AND LESS MATERIAL SCRAP. ONE OF THE CHALLENGES TO WIDESPREAD USE OF FIBER PLACEMENT IS THE REDUCTION OF PRODUCTION COSTS TO BE COMPETITIVE WITH THAT OF EPOXY SYSTEMS. A KEY TO REDUCING MANUFACTURE COSTS IS THE INCREASE OF FABRICATION SPEEDS WHILE MAINTAINING PROCESS QUALITY. A ROBUST CONTROL SYSTEM FOR RAPID CONSOLIDATION OF COMPOSITE MATERIAL REQUIRES THE ABILITY TO OBTAIN ACCURATE FEEDBACK INFORMATION FROM THE PROCESS AND THE ABILITY TO ADAPT TO UNFORESEEN CHANGES IN THE PROCESS DYNAMICS. NEURODYNE WILL APPLY INTERNALLY DEVELOPED INNOVATIONS IN NEURAL NETWORK BASED LEARNING OPTIMAL CONTROL TO LEARN AND OPTIMIZE FIBER PLACEMENT PROCESS DYNAMICS ON-LINE. NEURODYNE WILL ALSO USE NEW INNOVATIONS IN SIGNAL PROCESSING TO OPTIMALLY FILTER PROCESS FEEDBACK TO PROVIDE ACCURATE LINE NON-DESTRUCTIVE EVALUATION (NDE) MEASURES TO ENSURE THE QUALITY MANUFACTURE OF COMPOSITE PARTS. ANTICIPATED BENEFITS/POTENTIAL APPLICATIONS - MODEM MANUFACTURING PROCESSES REQUIRE INCREASING FLEXIBILITY TO ACCOMMODATE CHANGES IN PROCESSES DUE TO IMPROVED MATERIALS, DEGRADED SENSORS, AND MACHINE WEAR. THE CONTROL SYSTEM DEVELOPED IN PHASE I WILL BE CAPABLE OF LEARNING COMPLEX PROCESS CONTROL, ROBUST IN THE PRESENCE OF NOISE, AND BE ABLE TO ADAPT ON-LINE TO CHANGES IN PROCESS DYNAMICS. THE APPLICATIONS FOR THIS SYSTEM INCLUDE ANY NONLINEAR MANUFACTURING PROCESS WHICH REQUIRES ROBUST ADAPTIVE CONTROL.