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 H16: Focus Session: Exascale Computations of Complex Turbulent Flows I |
Hide Abstracts |
Chair: Ramesh Balakrishnan Room: 4c3 |
Monday, November 25, 2019 8:00AM - 8:13AM |
H16.00001: Thoughts on Very Large Turbulence Simulations Invited Speaker: Philippe Spalart The Reynolds number in channel DNS rose by a factor of 29 since the KMM paper of 1987. The imminent step from about 10petaFLOPS to 1exaFLOPS allows a factor of about 3, if we elect to simulate the same flow. Turbulence created three challenges: its results conflict with our semi-theory of wall-bounded flows. First, even with this factor of 29, a true logarithmic layer is not found. Second, the Reynolds stresses do not have a plateau in the (near) log layer. Third, they are not independent of the flow Reynolds number at a fixed $y^+$. All these properties are implied by classical theory and satisfied by turbulence models of conventional type. This has greatly limited the contribution of DNS to RANS models. The complaint in engineering is over smooth-body separation and reattachment. DNS of such flows at two Reynolds numbers are most desirable, but defining “the same flow, at different Reynolds numbers” is non-trivial because of the incoming boundary-layer thickness. A strategic community decision is whether simulations can be free-standing, or experimental confirmation is needed. The use of simulation data, be it physical understanding, engineering tools built with classical thinking, or tools built with Machine Learning, is challenging. For this third option, a sound definition of the mission is sorely needed. Also in high demand are non-DNS turbulence-resolving simulations, including Wall-Modeled LES and DES, as are simulations over complex geometries, say a Formula-1 car. Such simulations have had poor grids, and automatic adaptation is needed. Grid generation, even non-adaptive, could create a bottle-neck on massively-parallel machines. Grid convergence is not achieved over full geometries, even for RANS. Research can be improving SGS and Wall Models using DNS flow fields or exercising the non-DNS tools at high Reynolds numbers. It is difficult to define challenges that are clear, are attractive in an “academic” sense, and will illustrate the value of having 1 exaFLOPS. [Preview Abstract] |
Monday, November 25, 2019 8:13AM - 8:26AM |
H16.00002: Tractability of LES for complex flows Sanjeeb Bose, Kan Wang, Chris Ivey, Frank Ham To date, the use of large-eddy simulations (LES) has often been restricted due to relatively large computational expense (leading to intractably long wall clock times) and challenges in the generation of suitable grids for complex geometries. Advances in scalable mesh generation, nonlinearly stable numerical methods and wall modeling have substantially reduced the computational cost of performing LES of flows with either complex physics and/or geometries. We will discuss the computational performance and scalability of the flow solver charLES on petascale architectures and the prospects of using LES within engineering design environments where higher simulation throughput is required. The cost and fidelity will be assessed in the context of some benchmark LES, including the prediction of transonic compressor stall (NASA Rotor 37). By leveraging massively parallel petascale computing platforms, it will be shown that such calculations can be completed within hours of wall clock time. [Preview Abstract] |
Monday, November 25, 2019 8:26AM - 8:39AM |
H16.00003: Exascale Turbulence Simulations: From Fundamental Flows to Flight Scale Aerodynamics Kenneth Jansen, Riccardo Balin, John Evans, Philippe Spalart This talk will provide an update on two Argonne Early Science Program (ESP) projects. The first is a Simulation ESP where prior delayed detached eddy simulations of flow control on a vertical tail at a chord Reynolds number of 325k were validated against experiments are being extending to flight scale (53 times higher Reynolds number). The second is Data and Learning ESP focused on: 1) data compression, 2) turbulence modeling improvement from machine learning, 3) uncertainty quantification and multi-fidelity modeling, and 4) in situ data analytics. Progress made towards these ambitious goals will be shared as well as future plans and needed developments. [Preview Abstract] |
Monday, November 25, 2019 8:39AM - 8:52AM |
H16.00004: A second-order consistent time-resolved database of turbulent channel flow up to $Re_\tau \approx 5000$ Alberto Vela-Martin, Miguel P. Encinar, Javier Jimenez Wall-bounded flows play an important role in numerous common applications, and have been intensively studied for over a century. Understanding the fundamental mechanisms of the logarithmic and outer regions is essential for the development of effective control strategies and for the construction of a complete theory of wall-bounded flows. A proper analysis of the logarithmic and the outer layers requires time-resolved simulations at high Reynolds numbers in large domains, which makes the storage of the time series impractical. We present a novel low-storage method for time-resolved simulations. By retaining only the large and intermediate scales, while taking care to keep all the variables needed to fully reconstruct the flow at the level of second-order statistics, we reduce the storage requirements by a factor of $10^3$. This new methodology is efficiently implemented by using a new high-resolution hybrid CUDA-MPI code, which exploits the advantages of GPU co-processors on distributed memory systems, and allows to run for physically meaningful times. Databases for channel flows at up to $Re_\tau=5300$ in large boxes $(8\pi h\times 3\pi h)$ for over 30 turnover times, are presented. [Preview Abstract] |
Monday, November 25, 2019 8:52AM - 9:05AM |
H16.00005: Toward Simulating Turbulent Wall-Bounded Flows at High Reynolds Numbers on Exascale Platforms Ramesh Balakrishnan, Paul Fischer Given that wall resolved LES (WRLES) of high Reynolds number flows will continue to be intractable on even exascale computing platforms, there is considerable effort to make hybrid RANS/LES (HRLES) methods the tool of choice for predictive simulations of high-Re separated flows. The fidelity of HRLES depends on the ability of the subgrid model to account for the effects of the filtered scales on the resolved scales, and on the numerical schemes that are used to evolve the Navier-Stokes equations. While higher-order numerical schemes, with low dissipation/dispersion errors (and higher arithmetic intensity), are commonly used for canonical flow simulations (on simpler geometries), the bulk of the high-Re flow simulations on complex geometry still employ nominally second-order accurate schemes in structured/unstructured flow solvers. Hence, there is a need for better sub-grid models that can improve the predictive capability of both the higher-order flow solvers and existing second-order accurate flow solvers, for simulating turbulent separated flows. This talk is about ongoing efforts to develop HRLES subgrid models from DNS/WRLES simulations of canonical flows over curved and sharply discontinuous surfaces and their ability to predict high-Re flows on current petascale platforms. [Preview Abstract] |
Monday, November 25, 2019 9:05AM - 9:18AM |
H16.00006: Wall-resolved LES of a complex turbulent flow Mujeeb Malik, Ali Uzun Implicit, wall-resolved large eddy simulation, using a fourth-order compact difference scheme, is performed for a relatively high Reynolds number flow involving shock-induced flow separation. This is perhaps one of the most ambitious such simulations employing 24 billion grid points, pushing the boundary of flow simulations on high performance computing hardware. The particular case selected for this simulation is that of the well-known Bachalo-Johnson experiment conducted on a cylindrical body with an axisymmetric bump, which involves transonic shock-induced boundary layer separation with subsequent reattachment downstream of the bump. The Reynolds number based on the hump chord is 2.763 million. The relatively high Reynolds number of the test case makes the wall-resolved simulation very challenging, requiring billions of grid points. We will discuss the various issues and challenges encountered during the course of this research. Comparison of the simulation results is made with the available experimental measurements, with a view toward assessing the predictive capability of the simulations. Work supported by NASA's Transformational Tools and Technologies Project. [Preview Abstract] |
Monday, November 25, 2019 9:18AM - 9:31AM |
H16.00007: Wall-Modeled Large Eddy Simulation (WMLES) of High-Lift Aircraft Konrad Goc, Sanjeeb Bose, Parviz Moin Wall-Modeled Large Eddy Simulations with an equilibrium wall model (WMLES) are carried out on the JAXA high-lift model at various angles of attack. The configuration features the geometric complexity of deployed slats, flaps, and associated bracketry with flow in the low Mach, high Reynolds number regime (e.g., takeoff conditions). Salient flow features arising from smooth-body and geometrically-imposed separation are captured. The lift curve slope predicted by WMLES is well-characterized near the stall flight condition. Good agreement with experimental sectional pressure measurements is obtained. Calculations that include the wind tunnel walls and half-model mount are compared to uncorrected experimental data to mitigate wind tunnel correction uncertainty. Simulations were performed using the CharLES code at core-hour costs comparable to some RANS calculations of the same configuration (Rumsey, 2018). [Preview Abstract] |
Monday, November 25, 2019 9:31AM - 9:44AM |
H16.00008: Harnessing Exascale Platforms to Predict Shock-Wave / Boundary-Layer Interaction Jonathan Poggie Separated shock-wave / boundary-layer interactions occur in a broad flight regime, from transonic commercial aircraft to hypersonic space-access vehicles. These interactions tend to be intensely unsteady, with frequency content from fine-grained turbulence to low-frequency ($\sim 10$ Hz) separation bubble breathing. The presence of this unsteadiness can cause problems with flight control and structural fatigue on aircraft. Simulation of this important phenomenon is extremely challenging because of the need to capture such a broad range of scales. Identification of the physical mechanisms underlying the unsteadiness may help guide the development of simplified models that capture the essential features of the interactions. To this end, our research group has been investigating separated compression ramp flows, and studied their selective response to disturbances in the incoming turbulent boundary layer. In implicit large-eddy simulations, conditional averaging identified a perturbation velocity profile associated with separation motion, and forcing in the upstream boundary layer with this particular form was found to drive large-scale separation unsteadiness downstream. [Preview Abstract] |
Monday, November 25, 2019 9:44AM - 9:57AM |
H16.00009: Shock-induced transition and heating in hypersonic boundary layers Lin Fu, Michael Karp, Sanjeeb T. Bose, Parviz Moin, Javier Urzay The interaction of an incident shock wave with a Mach-6 undisturbed laminar boundary layer is addressed using DNS and equilibrium wall-modeled LES (WMLES). The wall temperature is cold compared to the free-stream stagnation temperature, such that the mean temperature profile develops a peak near the wall due to viscous heating. The consequences of the interaction are that the boundary layer transitions to turbulence downstream of the shock impingement point, and that transition causes a localized significant increase in the Stanton number and skin-friction coefficient. The peak thermomechanical load increases approximately linearly with the incidence angle. WMLES prompts transition and peak heating, delays separation, and shortens the separation bubble. WMLES provides predictions of DNS peak loads within $10\%$ at 150 times lower computational cost. In the fully-turbulent boundary layer, WMLES agrees well with DNS for the Reynolds-analogy factor (Chi and Spalding, 1966), the mean velocity and temperature profiles, including the temperature peak, and the temperature/velocity correlations. [Preview Abstract] |
Monday, November 25, 2019 9:57AM - 10:10AM |
H16.00010: Using DNS to improve wall-modeled LES of turbomachinery flows Koen Hillewaert, Michel Rasquin, Thomas Toulorge Argo is developed for scale-resolving simulations in urbomachinery in view of generating reference data for RANS models as well as the prediction of off-design performance. Based on the Discontinuous Galerkin Method, low dispersion and dissipation commensurate with DNS and LES is maintained on an unstructured mesh. Recently, work is ongoing on wall modeling in view of full machine computations. The development of turbulence models in turbomachinery is hindered by the lack of detailed reference data, as usually only experimental data are available. Due to the continuous increase in computational power, it has become possible to generate detailed numerical databases in fully controlled conditions for industrially relevant conditions. The generation and exploitation of such data sets is the aim of the European project HiFiTurb. This contribution focuses on a variant of the axisymmetric transonic bump of Bacchalo and Johnson, storing all of the terms relevant to RANS, as well as time-resolved data in the boundary layer in view of improving modeling for shock-boundary layer interaction conditions. [Preview Abstract] |
Monday, November 25, 2019 10:10AM - 10:23AM |
H16.00011: Large-Eddy Simulations of turbomachinery flows: from wall-resolved academic configurations to wall-modeled industrial geometries. Florent Duchaine, Luis Segui, Jerome De Laborderie, Nicolas Odier, Jerome Dombard, Laurent Gicquel LES has been shown to be a promising tool to tackle turbomachinery challenges induced by high Reynolds and Mach numbers and complex flow physics. The CPU cost is however identified as the reason why existing LES of turbomachinery flows concern simplified configurations. CERFACS has extended the capability of the reactive LES solver AVBP for turbomachinery applications. The developments and validations have concerned the application of accurate boundary conditions at inlets and outlets as well as the numerical treatment of the rotor/stator interface compliant with LES requirements. After a brief description of the flow solver, the presentation will focus on the results and insights obtained on two configurations. The first one is the high-pressure turbine cascade LS89 for which the MUR239 operating point is still today a challenge to simulate accurately. To address this academic aerothermal case, a wall-resolved approach is used with a high-order numerical scheme. The second configuration of interest is the 3.5 stages high-pressure axial compressor CREATE. This simulation, corresponding to one of the first wall-modeled LES of such a complex machine, has shown very promising results by comparison with experimental data. [Preview Abstract] |
Monday, November 25, 2019 10:23AM - 10:36AM |
H16.00012: Challenges and progress in adaptation of RANS models for Scale Resolving Simulations (SRS) of Turbulence Pedram Tazraei, Sharath Girimaji Even with the advent of exascale computing, many complex turbulent flows require some degree of modeling to enable predictive calculations of real-life applications. Scale resolving simulations (SRS) capable of yielding the best possible results at various and varying degrees of resolution are ideally suited for these computations. In general, SRS models can be broadly classified into zonal and bridging methods depending on the manner in which scale resolution is achieved. While the need for SRS approach is compelling, there is no clear consensus on closure model development thus far. In this presentation, we present a formal framework for adapting well-tested RANS (Reynolds-Averaged Navier-Stokes) models for SRS sub-grid stress computations. Various challenges such as commutation error, appropriate fixed-point behavior, near-wall closures are identified. Reasonably rigorous theoretical techniques to address the above issues in SRS modeling context are proposed. Asymptotic approach of SRS toward DNS (direct numerical simulations) in the limit of cut-off length scale approaching Kolmogorov length scale is also examined. [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