An artificial neural network is trained using helicopter flight test data to predict rotor system component loads during high-speed maneuvering flight. Inputs to the network include control positions and aircraft state parameters. These parameters can be measured easily in the nonrotating system, i.e., the fuselage, and vary at a relatively low frequency. A network design sensitivity study is conducted and several networks are developed for three loads: the rotor blade pushrod load, blade normal bending moment, and main-rotor damper load. Prediction accuracy is evaluated using a validation data set consisting of symmetric pullout maneuvers, rolling pullout maneuvers, and climbing turns not contained in the training data set. A traditional statistical approach, stepwise multiple linear regression, also is utilized, and the two methods are compared and contrasted. Correlation coefficients from 84 % to 97 % are achievable using the neural network model for all three loads. Through a unified approach involving both neural network and statistical analysis, greater accuracy and understanding of the neural network is attained.
Prediction of helicopter component loads using neural networks
Die Vorhersage von Hubschrauber-Komponentenbelastungen unter Anwendung der Theorie neuronaler Netze
Journal of the American Helicopter Society ; 40 , 1 ; 72-82
1995
11 Seiten, 5 Bilder, 9 Tabellen, 9 Quellen
Aufsatz (Zeitschrift)
Englisch
Prediction of Helicopter Component Loads Using Neural Networks
Online Contents | 1995
|Prediction of helicopter component loads using neural networks
Tema Archiv | 1993
|Artificial neural networks for predicting nonlinear dynamic helicopter loads
Tema Archiv | 1994
|