Flight test flutter prediction using neural networks

Abstract

Abstract

Flight flutter testing is a crucial part in the certification of a prototype aircraft. To aid the clearance process, a number of different methods have been proposed to determine the speed at which flutter occurs on the basis of data obtained during the flight envelope expansion. However, the most commonly used approach is simply to extrapolate the estimated damping ratios. In this paper, a method is proposed for the prediction of damping ratios during a flight test using a neural network trained on model data. The proposed method is compared with a simple statistical extrapolation approach, and the effects of noise are investigated. For noise-free data, the neural network method shows improved accuracy compared with the statistical method. With noisy data, the accuracy of the statistical method is unacceptably poor, but the accuracy of the neural network method remains good as long as the network is trained with noisy data.