S. Reed

An artificial neural network based parametric fatigue monitor for the empennage (tailplane region) of a military trainer aircraft has been developed. Multilayer perceptron artificial neural networks were used to predict tailplane low-frequency strains and high frequency fatigue damage rates throughout the flight, using low-bandwidth flight parameters, such as accelerations and control positions. The system was trained and tested using Operational Loads Measurement data from the RAF Tucano trainer aircraft. Fourteen highly damaging flights were selected for use as the training set and the models produced were tested on 38 unseen flights. The average correlation coefficient between predicted strain and measured strain for the low-frequency strain model, for an unseen test data set, was 0.991. Furthermore, the combined low and high frequency models were able to predict accumulated fatigue damage over the test data set to within 4% of that calculated from the measured strain time histories.

23rd ICAF Symposium of the international Committee on Aeronautical Fatique, 2005, Hamburg

Conference Poster

englisch

A5, 12 Seiten

DGLR-Bericht, 2005, 2005-03, 23rd ICAF Symposium of the international Committee on Aeronautical Fatique - Proceedings; S.693-704; 2005; Deutsche Gesellschaft für Luft- und Raumfahrt - Lilienthal-Oberth e.V., Bonn

NA

artificial neural networks, fatigue monitoring

Bestellbar

2005