NEURAL NETWORK HARDWARE FOR HELICOPTER CONTROL SYSTEM MONITORING

A SIGNIFICANT CAUSE OF CATASTROPHIC HELICOPTER FAILURE IS FAULTS IN THE CONTROL SYSTEM. A NOVEL MEANS FOR DIAGNOSING THE HEALTH OF THE CONTROL SYSTEN IS TO MONITOR THE DYNAMIC AND CONTROL PARAMETERS OF THE HELICOPTER. IN ORDER TO COMPUTE THE DYNAMIC AND CONTROL PARAMETERS, A PARAMETRIC MODEL OF THE HELICOPTER IS REQUIRED. RECOVERY OF THE PARAMETERS IS VERY COMPUTATIONALLY EXPENSIVE AND TRADITIONAL METHODS USING LINEAR ANALYSIS ARE UNRELIABLE. NEURAL NETWORKS OFFER THE POTENTIAL OF PROVIDING MORE ROBUST SOLUTIONS TO THE PARAMETER IDENTIFICATION PROBLEM. THE ADVENT OF NEURAL NETWORK INTEGRATED CIRCUITS CREATES A NEW OPPORTUNITY TO IDENTIFY THE PARAMETERS IN THE FIELD. THE FEASIBILITY OF THIS APPROACH WILL BE PROVEN BY DEVELOPING A HELICOPTER MODEL, TRAINING A NEURAL NETWORK TO RECOVER THE DYNAMIC AND CONTROL SYSTEM PARAMETERS, AND DEVELOPING A CONCEPTUAL DESIGN OF A HARDWARE IMPLEMENTATION USING NEURAL NETWORK INTEGRATED CIRCUITS.