Machine Learning to Predict Aerodynamic Stall

A convolutional autoencoder is trained using a database of airfoil aerodynamic simulations and assessed in terms overall accuracy interpretability. The goal to predict the stall investigate ability distinguish between linear non-linear response pressure distribution changes angle attack. After sensitivity analysis on learning infrastructure, we latent space identified by targeting extreme compression rates, i.e. very low-dimensional reconstructions. We also propose strategy use decoder generate new synthetic geometries solutions interpolation extrapolation representation learned autoencoder.