Towards the prediction of wall-pressure fluctuations for space launchers using ZDES-based convolutional neural networks
Vers la prévision des fluctuations de pression à la paroi des lanceurs spatiaux à l’aide des réseaux de convolutions basés sur des calculs ZDES
Résumé
Convolutional neural networks (CNNs) are developed to predict the RMS wall-pressure coefficient and single-point wall-pressure spectra in the separating/reattaching flow region of a space launcher in the transonic regime. Trained on a generic launcher afterbody configuration (NASA model 11 hammerhead), the present models are found to accurately predict the evolution of the aforementioned quantities for the tested launcher geometry. A scaling method is proposed to retrieve correct magnitudes for a given configuration. It is also demonstrated that good performances are achieved with the models using data from RANS simulation as input, paving the way for a drastic reduction in the computational cost for predicting wall-pressure fluctuations around space launchers.
Origine | Fichiers produits par l'(les) auteur(s) |
---|