Simulating Turbulent Wakes without the Upstream Body

Zhicheng Wang 1 Theo Käufer 2 Khemraj Shukla 1 Michael Triantafyllou 2, George Em Karniadakis 1 · 2025

Read Simulating Turbulent Wakes without the Upstream Body by Zhicheng Wang 1 Theo Käufer 2 Khemraj Shukla 1 Michael Triantafyllou 2, George Em Karniadakis 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-24922.txt.

Introduction

Zhicheng Wang, Theo Kaufer, Khemraj Shukla, Michael Triantafyllou, and George Em Karniadakis run the wake of a round bar in flow with no bar in the grid at all. Wake runs are key for craft and sea work, but full runs are dear. Their frame solves the flow rules in a plain box, with the inflow set from lab reads or past full runs.

At a single down stream spot, lean inflow facts were enough to rebuild the main wake acts. Vortex shed, three dimensional traits, stress spread, and tone maps all came back true. Checks ran at three flow speeds, matched with full runs and lab clips. A small check on what inflow picks work is also given.

Catalog record

CREATORZhicheng Wang 1 Theo Käufer 2 Khemraj Shukla 1 Michael Triantafyllou 2, George Em Karniadakis 1
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-24922

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Simulating Turbulent Wakes without the Upstream Body Zhicheng Wang 1 Theo Käufer 2 Khemraj Shukla 1 Michael Triantafyllou 2, George Em Karniadakis 1 Abstract We present a simplified framework for simulating three-dimensional turbulent wakes without the upstream body that generates them. Instead of resolving the geometry, the incompressible Navier–Stokes equations are solved in a rectangular domain on which the inflow boundary condition is prescribed as either phase-averaged or time-averaged velocity profiles obtained from experimental measurements and direct numerical simulations. Remarkably, prescribing the inflow at a single downstream location is sufficient to reconstruct the entire wake, including coherent vortex shedding, Reynolds-stress distributions, and spectral content, for the two Reynolds numbers we investigate here: R ​ e = 500 Re=500 and 5,000 5,000 . Comparisons with corresponding full-body DNS and experiments show good agreement in mean velocity fields and turbulence statistics. Our results demonstrate that the essential dynamics of bluff-body wakes are induced by the instability of the near-wake profile, and do not require the explicit presence of the bluff body. This body-free simulation paradigm enables physically interpretable wake reconstruction from mean profiles that can be easily obtained from measurements or simple 2D simulations. Our approach reduces the computational cost of DNS by an order of magnitude, hence offering a new route for reduced-complexity modeling and control of turbulent separated flows. keywords absolute instability, PIV, spectral elements, flow reconstruction † † affiliation: 1 Division of Applied Mathematics, Brown University, Providence, RI 02906, USA 2 Massachusetts Institute of Technology,77 Massachusetts Avenue, Cambridge, Massachusetts 02139 1 Introduction The unsteady wake behind bluff bodies such as circular cylinders is a canonical phenomenon in fluid mechanics, revealing fundamental mechanisms of vortex shedding, instability, and transition. Despite its apparent geometric simplicity, the turbulent wake exhibits complex three-dimensional dynamics that depends sensitively on Reynolds number, body shape, and inflow conditions ( Gharib et al. 2006 ; Williamson 1996 ; Roshko 1954 ) . Conventional studies through direct numerical simulations (DNS) require inclusion of the body itself to generate the wake, leading to high computational and constraining efforts to isolate intrinsic instability mechanisms of the wake flow ( Kravchenko & Moin 2000 ) . Pioneering work leveraging linear-stability analysis established that the formation of the von Kármán vortex street is primarily governed by the absolute instability of the mean velocity profile in the near wake ( Triantafyllou et al. 1986 ) . Building upon this understanding, Triantafyllou & Karniadakis 1990 introduced a reduced-order model in which the Navier–Stokes equations were solved without the upstream cylinder, using as inflow a time-averaged velocity profile extracted from a pre-computed DNS at Re = 100. Remarkably, this minimal model reproduced the two-dimensional vortex street and its shedding frequency, showing that the essential wake dynamics are self-sustaining once the most unstable mean profile is prescribed at the inlet. Chan et al. 2006 numerically reproduced the dynamics of a vortex street in shallow wakes by prescribing only the time-averaged velocity profile at a specific location downstream of the bluff body. The analysis of the secondary instability ( Henderson & Barkley 1996 ) indicates that the wake of flow past a cylinder transitions to a three-dimensional state as the Reynolds number exceeds 188.5. However, to the best of our knowledge, no study has yet examined whether a reduced model can reproduce this three-dimensional wake transition. In the present work, we extend this reduced-model concept to the three-dimensional turbulent regime, demonstrating that a properly defined inflow condition—obtained either from experimental particle image velocimetry (PIV) data or from high-fidelity DNS—can accurately generate the entire 3D wake in the absence of the upstream body. Specifically, prescribing only a one-dimensional inflow velocity profile at a single downstream position is sufficient to reproduce the complete wake structure, including coherent vortical motions, Reynolds stresses, and spectral energy distributions. The resulting flow fields closely match those of the corresponding full-body DNS and experiments, confirming that the wake is primarily governed by the mean velocity profiles rather than the presence of the generating body. This body-free wake simulation provides a powerful reduced-complexity framework for reconstructing and studying turbulent wakes directly from limited inflow data. It offers new opportunities for data-driven model reduction, flow control, and physical interpretation of self-sustained instabilities in separated flows. The remainder of this paper is structured as follows: In Sec. 2 we present the reduced model and the data generation, followed by comparisons with DNS and PIV data and a discussion of the results in Sec. 2 , highlighting its predictive accuracy and implications for future wake modeling, and a conclusion (Sec. 4 ). 2 Reduced Simulation Model for Flow Past a Cylinder In the reduced model, the three-dimensional incompressible Navier–Stokes equations are solved using a high-order spectral element method (SEM) ( Karniadakis & Sherwin 2005 ) within a rectangular computational domain, as illustrated in Figure 1 , where no physical cylinder is present. A time-averaged or phase-averaged velocity profile—e.g., obtained from experiments—is prescribed at the inflow (left) boundary, without introducing any additional perturbation. To demonstrate the concept, here we will employ both DNS data and experimental PIV data. In DNS, the phase-averaged field is computed by evenly sampling 32 snapshots over one vortex shedding period and subsequently averaging over multiple periods. At the outflow (right) boundary, homogeneous Neumann boundary conditions are applied to the velocity, while homogeneous Dirichlet conditions are imposed on the pressure. Periodic boundary conditions are used on the remaining four sides of the domain. The length of the computational domain in streamwise ( x ), crossflow ( y ), and spanwise ( z ) directions is 30D , 20D , and 10D , respectively, where D is the diameter of the imaginary circular cylinder. The domain in the x – y plane of both simulations is divided into 32\times 32 uniform quadrilateral elements, while in the z direction, 256 Fourier planes are used in the simulations at Re=500 , and 42\times 42 uniform elements are used in the simulations at Re=5,000 . In addition, the spectral element (polynomial) order for the simulations at the lower Re is 4, and 7 for the higher Re . The reduced model was implemented using both our in-house SEM code Nektar2.5D ( Newman & Karniadakis 1997 ; Wang et al. 2019 ), which employs Jacobi spectral element discretization in the x – y plane and Fourier expansion in the z direction, and the open-source SEM (Legendre) code nekRS ( Fischer et al. 2022 ) . Specifically, the simulation at Re=500 is carried out using Nektar2.5D, whereas the simulation at Re=5,000 was conducted with nekRS, which can significantly accelerate computations on modern graphical processing units (GPUs). All the reduced model simulations are initialized with a zero velocity field. To trigger the three-dimensional transition, a body force is applied during the time interval tu_{\infty}/D\in[0,20] , and the simulations are then continued for an additional period of tu_{\infty}/D\in[20,300] to reach a statistically steady state. The flow statistics are collected over the time ranges tu_{\infty}/D\in[200,300] and tu_{\infty}/D\in[150,300] for the lower and higher Re , respectively. The experimental data were obtained from 2D-PIV experiments in the MIT Intelligent Towing Tank facility ( Fan et al. 2020 ) . A cylinder of diameter 25.4 mm was towed at a velocity of 0.188 m/s. Particle images in

Read the complete text