Investigating Feature Selection for Machine Learning Turbulence Models

What is it about?

Modern machine-learning approaches hold promise to improve current models in aerodynamics with large amounts of high-accuracy simulation data. In this paper, we use a high-accuracy dataset with challenging flow conditions to investigate an approach for selecting inputs for machine learning-based turbulence models. This work is a first step toward more systematic approaches to selecting model inputs.

Featured Image

Photo by Susan Wilkinson on Unsplash

Photo by Susan Wilkinson on Unsplash

Why is it important?

Accurately modeling turbulence is a critical aspect of simulations in the aerospace industry and machine learning has opened up new paths for improvement. Selecting input features is an important part of developing machine learning turbulence models, but there is little published research on feature selection for turbulence modeling specifically. This research is a step towards filling that gap.