Machine Learning to Predict Aerodynamic Stall

TL;DR

This paper presents a convolutional autoencoder trained on airfoil simulation data to predict stall, analyze pressure responses, and generate new geometries through latent space interpolation, enhancing aerodynamic understanding.

Contribution

It introduces a novel autoencoder approach for stall prediction, interpretability of pressure responses, and synthetic airfoil generation via latent space manipulation.

Findings

Autoencoder accurately predicts stall behavior.

Latent space captures key aerodynamic features.

Synthetic geometries can be generated through interpolation.

Abstract

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