Improved Methodology for the Prediction of the Empennage Maneuver In-Flight Loads of a General Aviation Aircraft using Neural Networks
EMBRY-RIDDLE AERONAUTICAL UNIV DAYTONA BEACH FL
Pagination or Media Count:
This research presents and documents an improved methodology for the prediction of empennage maneuver loads of a general aviation aircraft. Previous research using Neural Networks and angular accelerometers achieved predictions of strains to within - 50 psi of the measured strains in 100 of the cases for the vertical tail and 93 for the horizontal tail. The improved methodology is based on maneuver recognition using Neural Networks with the focus on both cost saving and improving the horizontal tail predictions. The recorded data are first classified by data clusters corresponding to known maneuvers then analyzed using Neural Networks trained for those maneuvers. This weighted sum approach successfully predicted strains to within - 50 psi of the measured strains in 100 of the cases for the horizontal tail. Angular accelerometer signals were replaced by numerically differentiated rate-gyro signals for the Neural Networks, and the predictions were found to be comparable resulting in considerable cost savings for the required minimum instrumentation.