Abstract:

Turbulent fluid flows are ubiquitous in nature and engineering and are characterized by chaotic fluctuations spanning a wide range of interacting scales. A central challenge lies in understanding the small scales of turbulence, which are often presumed to be universal and are central to mixing and transport phenomena. A major tool in this regard is direct numerical simulations (DNS), whereby the governing equations are exactly solved on a supercomputer by resolving all the scales. While DNS offers essentially all possible information, it remains computationally prohibitive at high Reynolds numbers, and practical applications necessarily rely on modeling. In my talk, I will illustrate how big data from DNS can be utilized to study the small scales of turbulence—particularly through gradients—offering new insights into their scaling, structure, and universality. I will also demonstrate how data-driven deep learning techniques can learn from these gradient fields and predict flow dynamics at higher, unseen Reynolds numbers, paving the way toward the use of DNS data to inform the development of robust turbulence models.

 

Bio:

Dr. Dhawal Buaria is an Assistant Professor in Mechanical Engineering at Texas Tech University. Prior to joining Texas Tech, DB received his PhD in Aerospace Engg. from Georgia Tech, and thereafter held postdoctoral positions, first at Max Planck Institute of Dynamics and Self-Organization (Germany) and then at New York University. His work focuses on utilizing large-scale direct numerical simulations and data-driven approaches to understand and model turbulent flows and associated mixing and transport phenomena. The simulations are enabled using novel computational algorithms on some of the largest supercomputers in the world, providing high fidelity data to understand flow physics and also develop robust models based on modern deep learning paradigms.

---------------------------------------------------------------