Abstract
To improve the safety record of the private general aviation sector, the Advisory System for Training and Safety (ASTRAS) was conceived. The ASTRAS software provides timely information to the pilot, assisting him in properly configuring the aircraft for different phases of the flight. In order to perform this task, the ASTRAS system is endowed with an artificial intelligence engine or Situation Recognizer (SR) which is able to discern the flight mode from sensor readings. The current SR is based on fuzzy logic membership functions. Although functional, the limitations of this method have prompted the development of an artificial neural network based SR (ANNSR). The goal of the ANNSR was to provide more accurate mode inferences, particularly during off nominal flight conditions. The ANNSR performed better than the fuzzy logic based SR in flight tests on the Engineering Flight Simulator (EFS). The ANNSR returned correct mode inferences which were slightly more accurate than the fuzzy logic SR during nominal flight conditions. The most important result was that the AN-NSR was able to infer the correct flight mode even when the aircraft was off nominal conditions, a task which the fuzzy logic SR failed to accomplish.
Nguyen, Thinh Xuan (1996). A neural network mode inference engine for the advisory system for training and safety. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1996 -THESIS -N47.