Bulletin of the American Physical Society
72nd Annual Meeting of the APS Division of Fluid Dynamics
Volume 64, Number 13
Saturday–Tuesday, November 23–26, 2019; Seattle, Washington
Session L19: CFD: Advanced Methods and Models II |
Hide Abstracts |
Chair: Robert Moser, University of Texas, Austin Room: 401 |
Monday, November 25, 2019 1:45PM - 1:58PM |
L19.00001: Numerical multiscale methods and effective boundary conditions Sean Carney, Bjorn Engquist, Robert Moser Numerical homogenization refers to the numerical extraction of the effective, ``macroscopic", or large scale behavior of a complex dynamical system at a reduced cost to resolving the full dynamics at all levels of detail. The no-slip boundary condition (BC) for viscous fluid flow over a solid surface can introduce asymptotically small scales that pose severe challenges for simulation; in this case it can be preferable to replace the no-slip condition with a homogenized BC. This talk discusses numerical techniques for generating slip BC, or wall laws, for laminar flows over rough boundaries, as well as turbulent boundary layer flows for constant favorable, zero, or adverse $\nabla p$. Guided by rigorous mathematics in the former case and recent empirical advances in the latter, numerical strategies are presented to overcome the high computational cost of resolving the full near wall dynamics. In both settings, the main idea consists of running high resolution simulations in a relatively small domain localized to the boundary. Numerical examples presented throughout validate the modeling approach. [Preview Abstract] |
Monday, November 25, 2019 1:58PM - 2:11PM |
L19.00002: Extending Partially-Averaged Navier-Stokes equations to Variable-Density Turbulent Flow Filipe Pereira, Fernando Grinstein, Daniel Israel, Rick Rauenzahn, Sharath Girimaji The mixing of distinct fluids is of importance to various areas of engineering. This class of problems is featured by its variable-density (VD) that leads to buoyancy effects, hydro-dynamical instabilities, transition and turbulent flow. All these phenomena turn the modeling of VD flows difficult. Whereas DNS and LES models are excessively demanding for practical problems, RANS tends to poorly predict such flows. Bridging methods, on the other hand, have the potential to surpass many of the limitations of the former models. By resolving only the phenomena that are not amenable to be modeled, these models can achieve a good compromise between accuracy and cost. Yet, their development for VD flows is rife with challenges. The aim of this work is to develop the Partially-Averaged Navier-Stokes (PANS) equations model for VD flow. To this end, the PANS framework proposed by Girimaji (2005) is extended to VD flow in order to derive a PANS-BHR2 closure. Particular attention is paid to the selection of the physical resolution (range of resolved scales) of the model. Thus, apriori testing is conducted to propose guidelines toward the efficient selection of the parameters determining the physical resolution of PANS-BHR2. The proposed model is then evaluated on two archetypal flow problem. [Preview Abstract] |
Monday, November 25, 2019 2:11PM - 2:24PM |
L19.00003: Identifying benefits of PANS-modeling over LES for engine flows Branislav Basara, Zoran Pavlovic Large-Eddy Simulation (LES) has been frequently used for engine flows providing reliable and accurate results. On the other hand, the limitations of Reynolds-Averaged Navier-Stokes (RANS) models are very well known and some of them are possible to compensate with e.g. modelling in combustion (knock), or they are just ignored like cycle-to-cycle variations etc. Most of limitations are removed by applying LES but leading to much higher computational costs. Up to now, it has not been clear if some of hybrid RANS-LES models could improve engine calculations with a moderate increase of computational costs. The Partially-Averaged Navier-Stokes (PANS) approach (Girimaji, 2006) is designed to resolve a part of the turbulence spectrum adjusting seamlessly from RANS to DNS. In other to optimize the basic PANS method for moving geometries, Basara, Pavlovic and Girimaji (2018) introduced an additional equation for the scale supplying variable. In this approach, the main resolution parameter, the unresolved to total kinetic energy ratio, can be calculated in-situ enabling simple and cost-effective calculations of engines. Cycle-to-cycle variations have been achieved with amplitudes of cycle relevant variables that are depending on the ratio between unresolved and total kinetic energy. Nevertheless, benefits in PANS calculations are also expected to come from the proper modelling of the wall including the heat transfer even for larger y$+$ values as RANS models are applied there. This should have a positive impact on the emission predictions. The work presented here will try to provide answers on all these questions. [Preview Abstract] |
Monday, November 25, 2019 2:24PM - 2:37PM |
L19.00004: Enhancement of PANS with Non-Linear Eddy Viscosity Closure Sagar Saroha, Krishnendu Chakraborty, Sawan Suman Sinha, Sunil Lakshmipathy In recent years partially-averaged Navier-Stokes (PANS) method has emerged as a promising bridging method for simulating turbulent flows. However, most PANS simulations reported in the literature have been performed employing linear constitutive equation for unclosed stresses. While PANS inherently addresses the limitations of Reynolds-averaged Navier-Stokes (RANS) by reducing the overly diffusive effects by choosing appropriate sub-unity values of its filter parameters, the limitations of a linear eddy viscosity model inherited from a parent RANS model still bottlenecks its performance. Indeed experimental evidence suggest that in flow past bluff bodies, there are regions where the Reynolds stress tensor is substantially misaligned with the local resolved strain rate, and a simple linear eddy viscosity assumption is not realistic. We present an enhanced version of the PANS methodology which employs a non-linear eddy viscosity model. We show that this enhancement substantially improves the prediction of various hydrodynamic as well as heat transfer statistics in flows past two representative bluff bodies: a square cylinder and a sphere. This enhanced nonlinear PANS methodology is presented as a more potent bridging method than its linear counterpart. [Preview Abstract] |
Monday, November 25, 2019 2:37PM - 2:50PM |
L19.00005: Sensitized RANS modeling of turbulence: a scale-resolving ERM-based eddy-viscosity model Suad Jakirlic, Benjamin Krumbein, Robert Maduta, Cameron Tropea A near-wall URANS (Unsteady Reynolds-Averaged Navier-Stokes) eddy-viscosity model based on ellipticrelaxation methodology (ERM) is sensitized to resolve fluctuating turbulence by introducing an appropriately formulated production term into the scale-supplying equation governing the turbulent inverse time scale. The latter term, inspired by the scaleadaptive simulation concept (Menter and Egorov, FTaC 85, 2010), enables an adequate suppression of the modeled turbulence intensity toward the respective sub-scale level. It implies the model's self-balancing between the resolved and modeled (unresolved) contributions to the turbulence kinetic energy. The feasibility of this grid-spacing-free model formulation is checked by computing a series of internal heat and fluid flow configurations featuring boundary layer separation, flow impingement, thermal mixing of flow-crossing streams as well as flows over rough and porous walls. Comparison to under-resolved Large-Eddy Simulations (LES) applying the dynamic Smagorinsky model and to a hybrid RANS/LES model based on the same eddy-viscosity scheme, denoted by VLES (Very LES, Chang et al., IJHFF 49, 2014) indicates an advantage in terms of the predictive capabilities of the presently proposed model, especially on relatively coarser meshes. [Preview Abstract] |
Monday, November 25, 2019 2:50PM - 3:03PM |
L19.00006: The use of experimental data as the inlet to direct numerical simulations of turbulent channel flow Callum Atkinson, Ezhilsabareesh Kannadasan, Adrian Lozano-Duran, Peter Schmid, Javier Jimenez, Julio Soria In the direct numerical simulation (DNS) of turbulence it is the large scales that require large computational domains and long simulation times to attain the correct statistically stationary state. In contrast experimental measurements struggle to resolve down to the dissipative flow scales, but have far less trouble capturing the large scales. In this work we demonstrate how large scale information from planar experimental measurements can be fed into a turbulent channel flows DNS to reduce the required computational time and domain for a given friction Reynolds number. The effect of inlet resolution and required spanwise extent are examined by generating synthetic experimental fields from streamwise periodic channel flow DNS at Re$\tau =$180 and 550 and using this data as the inlet to a channel flow DNS with inlet-outlet boundary conditions. When fully resolved inlet data is used the streamwise domain of the inlet-outlet DNS can be reduced to 1/16 of the periodic domain with minimal influence on the flow statistics and can even withstand a halving of the spanwise domain. [Preview Abstract] |
Monday, November 25, 2019 3:03PM - 3:16PM |
L19.00007: Turbulent Flow Simulations with the Julia Programming Language Manuel F. Schmid, Marco G. Giometto, Marc B. Parlange In turbulent flow simulations, there is little room for computational inefficiency. The resolution of a simulation and its time to completion are often limiting factors for the problems that can be studied numerically. At the same time, problem-specific extensions to the simulation code are often necessary. The Julia programming language promises to enable quick, iterative development in a friendly, high-level language while achieving a performance comparable to Fortran and C. We present a new code for direct numerical simulation of turbulent channel flows written in Julia, scaling to thousands of CPU cores. We compare its performance to a Fortran code with the same numerical approach and discuss advantages and drawbacks of the new code. [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