Use of artificial neural networks for buffet environments

Abstract

An artificial neural network-based procedure for predicting empennage buffeting pressures and elastic response as a function of upstream flowfield and geometric conditions has been developed under a cooperative experimental and analytical research agreement between McDonnell Douglas Aerospace (MDA) and NASA Langley Research Center. This research program is a continuing MDA effort to develop a unified buffet design methodology. The current effort employs a hybrid cascading neural network and finite element modeling method to predict flexible tail response based on rigid test pressure information. This method is dependent on experimental data to train the neural network algorithms, but is robust enough to expand its knowledge base with additional aircraft data. Initial results show an incredible potential to predict accurate rms and frequencydependent tail pressures, as well as flexible response while providing the future capability to incorporate upstream computational fluid dynamics data for advanced design aircraft buffet pressure predictions.