Bridging the gap: the interface of experimental and computational data in deep learning for fluid mechanics

Over the past several years, deep learning methods have seen widespread adoption in fluid mechanics research. While these methods have been applied successfully in both computational and experimental settings, there has been relatively little work bridging the gap between these domains. In this talk, we demonstrate how using a combination of numerical and experimental data may enable accelerated training for real-world applications of machine learning in fluid mechanics. Training deep learning models to predict the time evolution of a turbulent cylinder wake, this work explores potential applications for "digital twin" modeling in the prediction and control of fluid flows.