Turbulence Closure

Turbulence closure aims to accurately represent the effects of unresolved small-scale turbulent motions on larger scales in fluid simulations, a crucial challenge in computational fluid dynamics. Current research heavily utilizes machine learning, particularly neural networks (including Bayesian and recurrent architectures), to develop data-driven closure models and quantify their inherent uncertainties, often leveraging high-fidelity simulations for training. These advancements improve the accuracy and reliability of simulations across various applications, from predicting turbulent flames to designing more efficient turbomachinery, by providing more robust and physically-constrained uncertainty estimates.