R. Laakso, A. Siljander, J. Tikka, M. Bäckström

The structural integrity of aging aircraft can in principle be managed by two alternative approaches, either by estimating structural life using analytical methods or by observing the growth of structural damage. Parallel to these, various research efforts towards embedded structural intelligence are made. Neural network applications form one potential way of embedded structural intelligence applications, in which the overall goal is e.g. to seek optimal solutions in view of usability, perforn1ance, and life cycle costs. Among the neural network application efforts that are currently under work is the SelfOrganizing Map (SOM) aimed at the needs of the aircraft structural integrity. The SOM is an unsupervised neural network approach and a highly visual Data Mining (DM) tool, which offers some new possibilities to the flight data analysis. The DM tools become particularly interesting the more data there exists. The Finnish Air Force (FiAF) Operational Loads Measurement (OLM) programs offer a convenient test and development environment of the neural network tools. The SOM is used to find e.g. correlations between structural strains and flight parameters and to find out how the flight data is clustered (natural groups such as a data error cluster). As the correlation between the various structural strains was known from previous Hawk Mk.Sl/SlA OLM analysis experience without the SOM, these data from approximately 1000 flights could be used to validate the SOM-based approach.

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

Conference Poster

englisch

A5, 11 Seiten

DGLR-Bericht, 2005, 2005-03, 23rd ICAF Symposium of the international Committee on Aeronautical Fatique - Proceedings; S.519-529; 2005; Deutsche Gesellschaft für Luft- und Raumfahrt - Lilienthal-Oberth e.V., Bonn

NA

neural networks, loads measurement

Bestellbar

2005