Prediction of transonic wing buffet pressure based on deep learning

Abstract

In the present study, a deep learning approach based on a long short-term memory (LSTM) neural network is applied for the prediction of transonic wing buffet pressure. In particular, fluctuations in surface pressure over a certain time period as measured by a piezoresistive pressure sensor, are considered. As a test case, the generic XRF-1 aircraft configuration developed by Airbus is used. The XRF-1 configuration has been investigated at different transonic buffet conditions in the European Transonic Wind tunnel (ETW). During the ETW test campaign, sensor data has been obtained at different local span—and chordwise positions on the lower and upper surface of the wing and the horizontal tail plane. For the training of the neural network, a buffet flow condition with a fixed angle of attack alpha and a fixed sensor position on the upper wing surface is considered. Subsequent, the trained network is applied towards different angles of attack and sensor positions considering the flow condition applied for training the network. As a final step, the trained LSTM neural network is used for the prediction of pressure data at a flow condition different from the flow condition considered for training. By comparing the results of the wind tunnel experiment with the results obtained by the neural network, a good agreement is indicated.