| Summary: | 碩士 === 國立海洋大學 === 航運技術研究所 === 89 === Comparison of aircraft landing performance based on different neural network controllers during wind disturbances encountered is studied. Five different network structures have been utilized in the controller design. They are the conventional back-propagation network (BPN), the improved back-propagation network (IBPN), the multilayer functional link network (MFLN), the counter-propagation network (CPN), and the radial basis function network (RBFN). To study this problem we needed reliable wind profiles. There are two wind disturbance models most common in aircraft flight paths. They are turbulence and wind shear. In this study the Dryden form was used to model the wind turbulence and a discretized approximation model was used to form the wind shear. A simplified model of a commercial aircraft that moves only in the longitudinal and vertical plane is used in the simulations for implementation ease. Currently, most conventional control laws generated by the Automatic Landing System (ALS) are based on the gain scheduling method. Control parameters are preset for specified flight envelope. If the flight conditions are beyond this preset envelope, the ALS is disabled and the pilot takes over. An inexperienced pilot may not be able to guide the aircraft to a safe landing. It is therefore desirable to develop an intelligent ALS that expands the operational envelope to include safe responses under a wider range of condition. In this study, computer simulations of the aircraft landing performance using different neural network controllers are presented. Comparison and analysis based on network structure and learning rule are sugested.
|
|---|
