Abstract:
The predictive capabilities of computational fluid dynamics, which are critical for aerodynamic design, depend on the development of accurate closure models. However, no practical model has emerged as universally applicable across the wide range of flow regimes relevant to the industry. We introduce a closure model for large-eddy simulation, referred to as the Building-block Flow Model (BFM). This model is founded on the premise that a finite set of simple flows encapsulates the essential physics required to predict more complex scenarios. The BFM is implemented using artificial neural networks and introduces five unique advancements within the large-eddy simulation framework: (1) It is designed to predict multiple flow regimes, including laminar flow, wall turbulence under zero, favorable, and adverse mean-pressure gradients, separation, wall heat transfer, and wall roughness effects; (2) It leverages information-theoretic dimensional analysis to select the most relevant non-dimensional input/output variables; (3) It ensures consistency with numerical schemes and gridding strategies by accounting for numerical errors; (4) It is directly applicable to arbitrary complex geometries; and (5) It is scalable for future extensions to model additional flow physics (e.g., laminar-to-turbulent transition). The BFM is utilized to predict key quantities of interest across a wide range of cases, such as turbulent pipes, turbine blades with roughness, speed bumps, aircraft in landing configurations, and hypersonic flows in entry, descent, and landing vehicles.
Bio:
Adrian Lozano Duran is an Associate Professor at GALCIT, Caltech, and a Visiting Associate Professor at MIT AeroAstro. He received his Ph.D. in Aerospace Engineering from the Technical University of Madrid in 2015. From 2016 to 2020, he was a Postdoctoral Research Fellow at Stanford University's Center for Turbulence Research. He served as an Assistant Professor at MIT from 2021 to 2024. His research focuses on computational fluid mechanics and the physics of turbulence, including causal inference, modeling, and control of turbulence using information theory. He is also interested in developing closure models for large-eddy simulations of aerospace applications using artificial intelligence.