Bulletin of the American Physical Society
73rd Annual Meeting of the APS Division of Fluid Dynamics
Volume 65, Number 13
Sunday–Tuesday, November 22–24, 2020; Virtual, CT (Chicago time)
Session R07: Computational Fluid Dynamics: LES, DNS, Hybrid RANS/LES (5:00pm - 5:45pm CST)Interactive On Demand
|
Hide Abstracts |
|
R07.00001: A Stochastic Wall Model for Large-Eddy Simulation of Rough Channel Flows Livia S Freire, Marcelo Chamecki In Large-Eddy Simulation (LES) of high-Reynolds number channel flows, the near-wall parameterization remains a challenge due to computational cost. In this study, a stochastic wall model consisting of vertical lines within each wall-adjacent LES grid cell is tested. Each line corresponds to a one-dimensional turbulence (ODT) model, where the vertical diffusion equation is solved and the effect of turbulence is provided by stochastic eddies, which correspond to a mapping function that mixes the variables and redistributes energy. This approach provides two options for modeling the effects of roughness within the ODT: (i) via equilibrium assumption by imposing a log-law with a bulk roughness height parameter between the ODT lowest grid point and the wall, and (ii) via drag model represented as a body force resolved within the ODT domain. The first approach moves the log-law parameterization closer to the wall compared to its use in the LES directly, improving the LES near-wall spectra. The second approach provides canopy-flow statistics within the wall model, reproducing important features such as skewed velocity field. Both options provide an alternative for studies focused on near-wall turbulence, refining the flow field information and improving flow statistics in the LES. [Preview Abstract] |
|
R07.00002: A data-driven hybrid LES/RANS framework for wall-bounded turbulent flows Pasha Piroozmand, Patrick Jenny Large Eddy Simulation (LES) models are promising candidates for highly accurate simulations of turbulent flows. However, their computational cost for resolving near-wall regions can be very high, and thus for many applications, LES is unaffordable. Hybrid LES/RANS models are proposed to overcome this issue. As an instance, a dual-mesh hybrid LES/RANS framework has recently been introduced where LES and RANS simulations are performed on the same domain while different mesh resolutions are employed. To ensure consistency between the solutions relaxation forces are applied to RANS solution in the free shear flow region away from the walls and to the LES solution in the near-wall regions. To further improve the accuracy, we extended this framework by incorporating available sparse experimental data into the RANS simulation using a variational data assimilation technique. The eddy viscosity computed by the RANS model is corrected iteratively using the discrete adjoint method. Due to the tight coupling, also the LES solution is improved and the overall model uncertainty is significantly reduced. [Preview Abstract] |
|
R07.00003: Evaluation of Wall Models for Internal Combustion Engines using Direct Numerical Simulation Muhsin Ameen, Saumil Patel Accurate modeling of the mass, momentum and energy transfer processes in the near-wall boundary layer region are critical to capture the internal combustion engine processes such as combustion phasing, wall heat loss and wall film formation. In multi-dimensional engine simulations, the grid requirements for accurately resolving the boundary layers are too expensive and hence wall models are employed which specify the wall shear stress and wall heat transfer using empirical correlations. The accuracy of these correlations for realistic engine simulations have not been analyzed in detail due to lack of experimental or high-fidelity simulation results. In the current study, we performed direct numerical simulations (DNS) of the Transparent Combustion Chamber (TCC-III) optical engine under motored operating conditions. Nek5000, a leading high-order spectral element code, was used to perform these simulations. The simulation results were used to evaluate and improve the accuracy of traditionally used wall models. The analysis was performed across a range of near-wall grid sizes. The improved wall models developed as part of this work can be used to perform accurate wall-modeled large eddy simulations (LES) at significantly lower computational expense as compared to wall-resolved LES. [Preview Abstract] |
|
R07.00004: Large Eddy Simulations of Internal Combustion Engines to Understand the Origins of CCV Saumil Patel, Muhsin Ameen, Tanmoy Chatterjee, Sicong Wu Mitigating cycle-to-cycle variability (CCV) can improve the performance of an engine. This requires an in-depth understanding of the stochastic in-cylinder processes. The objective of this research is to understand the causes of CCV. Nek5000, a leading high-order spectral element, open source code was used to simulate turbulent flow in the engine combustion chamber. Multi-cycle, wall resolved large-eddy simulations (LES) were performed for the General Motors (GM), Transparent Combustion Chamber (TCC-III) optical engine. In this talk, we highlight the mechanisms that contribute to large-scale turbulent flow structures during the intake and compression strokes. In particular, we investigate the interaction of the fast-moving intake jet with the spark plug which enables vortex shedding and turbulent flow structures at the start of the intake stroke. Large-scale motion is characterized by integral quantities, i.e. tumble and swirl ratios, which were calculated during each state of the engine cycle. We attempt to quantify the tumble breakdown process during compression by investigating the evolution of kinetic energy. We also present statistical analysis, including mean and root mean-squared (rms) phased average velocity fields. Results are compared with experimental PIV data. [Preview Abstract] |
|
R07.00005: LES versus URANS for flow in a centrifugal pump Beomjun Kye, Haecheon Choi Turbulent flows in a centrifugal pump at design and off-design conditions are simulated using two different numerical methods, i.e., large eddy simulation (LES) and unsteady Reynolds-averaged Navier-Stokes equations (URANS), respectively, and we suggest from comparisons of two solutions which flow characteristics cannot be identified from URANS. URANS provides overall flow features within the impeller and volute at the design condition, but overestimates the pressure rise in the discharge pipe as it does not predict unsteady large-scale corner vortices which grows larger with increasing flow rate. This results in a performance mismatch between URANS and LES even at the design condition. At the off-design condition, URANS fails to accurately resolve unsteady flow features such as flow separation at the volute tongue and leakage to cavities which are caused by strong impeller-volute interaction. We will further investigate difference between two solutions to better understand unsteady flow characteristics in a volute-type centrifugal pump. [Preview Abstract] |
|
R07.00006: Integral length-scale approximation based modeling of subfilter-scale scalar flux for large-eddy simulation of stratified turbulent flows Reetesh Ranjan The integral length-scale approximation (ILSA) model for the subfilter-scale (SFS) stress is extended in this study for large-eddy simulation (LES) of stably stratified turbulent flows. The ILSA model is an algebraic eddy viscosity model, where for the model length scale an approximation of the integral length scale of turbulence is used instead of relating it to the grid size. The approximated length scale depends upon the local flow conditions and the SFS activity, thus allowing to obtain accurate results on relatively coarser grids compared to the other algebraic closures. In this study, two approaches relying on the eddy diffusivity based formulation are considered for modeling of the SFS scalar flux, which requires a closure while performing LES of stratified turbulent flows. The first approach utilizes the conventional dynamic eddy diffusivity model, and the second approach uses an approximation of the integral length scale for the scalar field. In both these approaches, the ILSA model is used for the SFS stress. These approaches are assessed for their accuracy by performing LES of a fully developed turbulent channel flow at frictional Reynolds number of 550 and frictional Richardson number of 0 and 60, corresponding to a neutral and stable stratification, respectively. [Preview Abstract] |
|
R07.00007: Requirements and sensitivity analysis of RANS-free wall modeled LES Michael Whitmore, Adrian Lozano-Duran, Parviz Moin Sensitivity analysis of wall modeled LES is performed through theoretical analysis and numerical simulations. Our focus is on the development of robust dynamic-slip RANS-free wall models which provide good performance for different LES subgrid-scale models and numerical discretizations. The study is motivated by the sensitivities observed in the dynamic slip wall models developed by our group (see Bose and Park, ARFM, 2018). Theoretical analysis is used to estimate the error propagation from the wall model inputs (LES velocities and derivatives) into the wall model predictions. We also investigate the impact of grid-scale models and numerics (discretization and grid topology) along with the ability of spatial and temporal filtering of the inputs to reduce these sensitivities. Our analysis is confirmed by actual wall modeled LES calculations of fully developed turbulent channels. [Preview Abstract] |
|
R07.00008: PANS Modeling for Variable-Density Flow Filipe Pereira, Fernando Grinstein, Daniel Israel, Sharath Girimaji The prediction of variable-density (VD) flows is crucial to various problems in engineering and nature. Yet, the accurate simulation of this class of flows is challenging due to the complex physics. Thus, traditional simulation strategies are either excessively demanding (DNS and LES) or inaccurate (RANS). An efficient alternative is often needed. We present a new PANS (Partially-Averaged Navier-Stokes equations) method designed to predict VD flows efficiently. The proposed method is evaluated through the simulation of two archetypal problems: the Taylor-Green vortex at Re$=$3000, and the Rayleigh-Taylor (RT) flow at At$=$0.50. The first represents a canonical case to study transition to turbulence driven by vortex stretching and reconnection mechanisms, while the second constitutes a mixing problem driven by buoyancy effects and the RT instability. These flows are computed at distinct physical resolutions (cut-off scale) and grid resolutions. The results show that the proposed PANS method can accurately predict the two flows. Yet, this is accomplished at a lower cost than with LES and DNS. For the same degree of accuracy, the results indicate that PANS can reduce the cost of LES computations by a factor of 16 for the problems considered. [Preview Abstract] |
|
R07.00009: A Compressible Formulation of the Active Model Split (AMS) Hybrid RANS/LES Model Clark Pederson, Todd Oliver, Robert Moser Hybrid RANS/LES models show promise for accurate prediction of flows with large-scale unsteadiness, such as separated flow. However, the predictive accuracy of typical hybrid models is limited by several shortcomings, such as modeled-stress depletion and a dependence on scalar grid measures. The "active model-split" (AMS) hybrid RANS/LES model was developed to address these shortcomings. In this model, the mean and fluctuating portions of the stress are modeled separately; this allows for a consistent treatment of the mean flow even in the presence of resolved fluctuations. The AMS has previously shown improved accuracy in several incompressible test cases. In this presentation, a compressible extension of the AMS is outlined. Models are presented for the additional closures needed, and the accuracies of various alternative models are compared. Hybrid RANS/LES modeling concerns specific to compressible flows are also addressed. Results are shown for both transonic and supersonic test cases. [Preview Abstract] |
|
R07.00010: An Analysis of LES models in Stably Stratified Flows Jeremy Melvin, Robert D. Moser Large eddy simulations (LES) of complex turbulent flows, such as for the simulation of wind farms, are limited by resolution requirements in boundary layers and modeling assumptions for subgrid stress (SGS) models which are inconsistent with flow characteristics. In this talk, we discuss limitations of standard SGS models based on a scalar eddy viscosity and their performance in anisotropic turbulence. A set of DNS simulations of homogeneous stratified turbulence is used to inform current efforts to develop an improved SGS model for anisotropic turbulence. A priori and a posteriori error analysis for a set of LES models is conducted using the DNS data. As expected, it is found that these models are insufficient, especially under stronger levels of stratification. Anisotropic filtering, which is necessary as a result of both the natural meshing of boundary layers as well as the dynamics of the flow, interacts poorly with these models. This emphasizes the need for the formulation of LES models to naturally handle anisotropies present in both the grid and the underlying turbulence. We provide a brief overview of current modeling efforts using a tensorial eddy viscosity and other design choices to develop a more robust LES model. [Preview Abstract] |
|
R07.00011: Length-Scale Dependence of Dynamics of Homogeneous Variable Density Turbulence Juan Saenz, Denis Aslangil, Daniel Livescu We investigate the length-scale dependence of filtered dynamical quantities that are important for modeling variable density turbulence, namely the Reynolds stress $\mathcal{T}_{ij}$, mass-flux velocity $a_i$ and density-specific volume covariance $b$, using theory and diagnostics from DNS of homogeneous variable density turbulence. From the perspective of length-scales resolved by filter-width $w$, $\mathcal{T}_{ij}, a_i, b$, and terms in their transport equations vary smoothly between DNS and their classical RANS definitions at the small and large filter width limits, respectively. Further, the generalized central moments in the filtering approach (Germano '92) are expressed as inner products of generalized fluctuating quantities, $q'(\xi,x)=q(\xi)-\overline q(x)$, which represent fluctuations of a field variable $q$ at points $\xi$ with respect to its filtered value at a point $x$. At large $w$ values, the generalized fluctuations become the RANS fluctuations, and realizability conditions for $\mathcal{T}_{ij}, a_i, b$ become the realizability conditions for their RANS counterparts. This work supports the notion of a generalized, length-scale adaptive model that converges to DNS at high resolutions, and to classical RANS statistics at coarse resolutions. [Preview Abstract] |
|
R07.00012: DNS of Flow Over Smooth and Rough Wavy Walls at $\mathit{Re}_\lambda=4760$ Vedant Puri, Ramesh Balakrishnan Direct Numerical Simulations of turbulent flows over smooth and rough wavy walls has been conducted. The \textit{Smooth Wavy Wall} is described by a sinusoidal wave in the streamwise direction with amplitude to wavelength ratio $a/\lambda=0.05$. Small-amplitude sinusoidal roughness elements are superimposed on to the \textit{Smooth Wavy Wall} to obtain the \textit{Rough Wavy Wall}. The smaller undulations on the \textit{Rough Wavy Wall} represent undulations that a Large Eddy Simulation may not be able to resolve, but whose effects should be reflected on the resolved flow field. The flows are characterised by Reynolds number of $2390$ based on bulk velocity and channel half-height. The budget terms of the Reynolds Stress Transport Equations reveal strong coupling between wall-topography and turbulence dynamics near the wall. Flow separation at the crest of the \textit{Smooth Wavy Wall} and the formation of a persistent recirculation zone near the trough are observed along with a shear layer spanning multiple wavelengths. [Preview Abstract] |
|
R07.00013: DNS-data-driven subgrid-scale scalar flux model for turbulent spatially evolving flows Oriol Lehmkuhl, Guillermo Araya The evolution of passive scalars (Pr = 0.2 and 0.71) in spatially-developing turbulent boundary layers (SDTBL) is numerically studied in zero-pressure gradient flows. Direct Numerical Simulation (DNS) is performed at high Reynolds numbers (Re$_{\theta}$ $\approx$ 2500). Turbulent inflow information is generated via the dynamic multiscale rescaling-recycling approach (J. Fluid Mech., 670, pp. 581-605, 2011). Furthermore, the Integral Length-Scale Approximation with Subfilter-Scale Stresses (hereafter, ILSA SFS) model, as proposed by Rouhi et al. (PHYSICAL REVIEW FLUIDS 1, 044401, 2016), is extended to consider passive scalar transport in a suite of Large Eddy Simulation (LES) for the first time. Similar to the momentum equation treatment, an SFS activity (based on thermal fluxes) is proposed, which is related to the thermal integral length scale. The model is explored and assessed by direct comparison with the DNS database and the classical turbulent Prandtl number approach. The present LES modelling effort is assessed via low/high order statistics computation, energy budget analysis of thermal fluctuations and coherent structure study. [Preview Abstract] |
|
R07.00014: Towards optimal numerics for the simulation of boundary-layer flows Manuel Schmid, Marco Giometto, Marc Parlange Due to the strict performance requirements, turbulence-resolving flow simulations are usually implemented in Fortran or C++. In these highly optimized codes, it can be challenging to explore the impact of modeling decisions such as the choice of numerical methods. The Julia programming language has been developed specifically for scientific computation, with a focus on providing high flexibility without sacrificing performance. We present a Julia code for direct numerical simulation as well as large-eddy simulation of turbulent channel flows and use it to explore the impact of modeling decisions such as the choice of time-integration scheme and the dealiasing of non-linear terms. [Preview Abstract] |
|
R07.00015: Direct numerical simulation (DNS) study of the effect of wake structure on drag coefficient of thin flexible cylinder at low Reynolds numbers Harika Gurram, Chelakara Subramanian, Maytal Dahan, Tracy Brown Previous experimental and numerical study conducted by the present author [1,2] shows the occurrence of a drag crisis at low Reynolds numbers (less than 250) for a thin flexible cylinder of diameters in the range of 0.7mm to 1mm. The transition and unsteadiness in the wake structure cause the drag reduction and a periodic vortex shedding as also observed in the far-field downstream. The reduction of wake width in the downstream indicates that low Re, suggests that transitional unsteadiness increases in the near-field wake which tends to decrease the pressure drag. In order to explain the wake transition structure, the present work focuses on investigating the fluid-structure interaction of the thin flexible cylinder for a lower Re ranging from 50 to 250 for the same cylinder diameters. The direct numerical simulations (DNS) are conducted using ICEM CFD grid generator and ANSYS solver on the Stampede2 supercomputer. The preliminary results capture the laminar, transitional, and turbulent regions of the wake in the downstream. This study has some implication on the drag of underwater tethered systems.\newline $[1]$ C. Subramanian and H. Gurram, "Drag coefficient of thin flexible cylinder," APS Division of fluid dynamics, Boston, 2015.\newline $[2]$ Chelakara S. Subramanian, Harika Gurram, “A Computational and Experimental Study of Wake of Thin Flexible Wires”, Journal of Fluid Flow, Heat and Mass Transfer (JFFHMT, Volume 7, 2020, DOI: 10.11159/jffhmt.2020.004 [Preview Abstract] |
|
R07.00016: Subgrid-scale Modeling and Resolution Sensitivities in Wall Modeled LES of High-Lift Aircraft Configurations Konrad Goc, Sanjeeb Bose, Parviz Moin The prediction of the onset of stall in high-lift aircraft configurations is of paramount importance in the design of aircraft. Prior simulations (Goc et al., 2020) have demonstrated that aerodynamic performance of the JAXA Standard Model, including at maximum lift conditions, could be predicted with tractable computational cost using wall modeled LES. This investigation assesses the sensitivity of those calculations with respect to subgrid-scale modeling closures and grid resolution across all angles of attack in the lift curve. The geometry includes the nacelle installation, wind tunnel mounting system and tunnel sidewalls. Two subgrid closures (Vreman and dynamic Smagorinsky) on grids ranging in size from approximately 10 to 200 million control volumes are considered for the sensitivity analysis. We find that the solutions exhibit weakening sensitivity to resolution as the grid is refined, suggesting that quantities of interest are approaching a grid converged state in the wall modeled limit as the number of points spanning the trailing edge boundary layer nears 20 points, particularly in the linear region of the lift curve. Sensitivity to SGS closures is weaker and more uniform across the lift curve, with the DSM outperforming the static coefficient Vreman model. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700