At FPCRL, we delve into the intricate world of turbulent flows through high-fidelity numerical simulations. Our simulations capture the essence of turbulent motion at both slow and rapid speeds, traversing rough surfaces, and within spinning frames of reference. We also explore this chaotic dance of fluids at transitional Reynolds numbers, seeking a deeper physical insight into these phenomena. Our mission extends beyond just simulations; we aim to unravel the secrets of turbulent flows in practical scenarios. We endeavor to craft models that mirror real-world conditions, thus enhancing our grasp of these complex behaviors. In our pursuit, we employ a comprehensive toolkit that encompasses theoretical analysis, computational techniques, and hands-on experiments. Furthermore, we embrace the cutting-edge capabilities of machine learning technology, adding another dimension to our investigative arsenal.
High-fidelity numerical simulations of turbulent flows
Our group conducts direct numerical simulations and large-eddy simulations of fluid flows in various engineering contexts. A few examples are shown below: top left: DNS, contours of streamwise velocity over cube arrays of various arrangements; top right: DNS, contours of the longitudinal velocity in a stratified wake; bottom left: LES, isosurfaces of Q2 criteria colored by pressure for prolate spheroid at 10 and 30 degrees angle of attack; bottom right: DNS, contours of cooling effectiveness for shaped hole film cooling.
Machine-learning turbulence models
Our group leverages machine learning tools for turbulence modeling. Below is an illustration of the rubber-band approach. Following this approach, we preserve the basic calibrations of a baseline model when developing data-enabled augmentations. This is in direct contrast with the conventional machine learning method, where machine-learned augmentations are often detrimental outside the training dataset.
Near-wall turbulence modeling
Wall modeling reduces the cost of high Reynolds number boundary-layer flows. Below is our “best practice” for wall-modeled LES.
Rough-wall models
From the earliest seminal works by Nikuradse and Colebrook and the consolidated engineering interpretation of the data in Moody, the field of rough-wall bounded turbulence has evolved remarkably to its present. We continue to pursue roughness modeling. Below is a collection of the rough walls we consider.
We acknowledge ONR, AFOSR, NSF, and Elliott Group for financial support.
