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 P16: Focus Session: Exascale Computations of Complex Turbulent Flows III |
Hide Abstracts |
Chair: Philipp Schlatter Room: 4c3 |
Monday, November 25, 2019 5:16PM - 5:29PM |
P16.00001: A massively parallel unstructured overset method for large-scale simulation of moving bodies in turbulent flows Wyatt Horne, Krishnan Mahesh We present an unstructured overset method capable of performing direct numerical simulation (DNS) and large eddy simulation (LES) of many ($O(10^{5})$) moving bodies, utilizing many computational cores ($O(10^{5})$) as shown in Horne \& Mahesh [J. Comput. Phys.(2019) \textit{In Press}]. A dynamic overset assembly is conducted to connect mesh solutions. To establish communication patterns a parallel master/slave algorithm is used. A parallel flood-fill algorithm is used for cutting. For searches, k-d tree data structures are used. Often the connectivity between overset meshes remains the same between time steps. The temporal coherence of objects is directly used to only update necessary information with time, resulting in substantial cost savings. A non-dissipative finite volume method is used for the fluid flow. An interpolant is used which has superior kinetic energy properties compared to local reconstructions. To solve pressure, a penalty constraint formulation is used, resulting in a symmetric, positive definite system. Strong scaling is demonstrated for 100,000 particles in a turbulent channel flow up to 492,000 cores. Detailed flow results illustrating the method are presented. [Preview Abstract] |
Monday, November 25, 2019 5:29PM - 5:42PM |
P16.00002: High-order discontinuous Galerkin methods for multi-physics flow simulations on exascale computing systems Matthias Ihme, Kihiro Bando, Eric Ching, Alex Aiken High-order discontinuous Galerkin (DG) methods have emerged as attractive techniques for simulating complex flows. In particular, these methods combine features of variational finite-element methods with finite-volume discretizations, thereby (i) allowing for arbitrarily high order of accuracy, (ii) enabling the discretization of complex geometries on irregular meshes, (iii) providing advanced refinement strategies, and (iv) the large degree of structured computations and data locality introduce a high level of parallelism, making these methods particularly suitable for high-performance computing on exascale machines. This presentation will discuss recent advancements on developing high-order DG-methods for complex flows that involve chemical reactions and multiphase flows. Following a theoretic consideration of performance gains of high-order schemes on exascale systems, we will discuss algorithmic developments and programming techniques for enabling the application of these high-order methods to turbulent reacting flows and hypersonic particle-laden flows. The presentation will close by discussing open research needs on programming paradigms and the utilization of emerging architectures with heterogeneous processors and complex memory hierarchies. [Preview Abstract] |
Monday, November 25, 2019 5:42PM - 5:55PM |
P16.00003: High fidelity turbulence simulations with adaptive mesh refinement in Nek5000 Nicolas Offermans, Adam Peplinski, Philipp Schlatter The design of an adequate mesh is a complex and time-consuming task for users of computational fluid dynamics codes, which typically requires some a priori knowledge of the developing physics. We use adaptive mesh refinement, which combines error estimators and grid adaptation, as a tool for automatic mesh optimization. This strategy allows for better error control and easier mesh generation. Our framework is Nek5000, a highly-scalable code based on the spectral element method. The $h$-refinement technique is chosen for mesh adaptation, where selected elements are split via an octree structure, and two types of error estimators are considered. The first estimate relies on the local spectral properties of the solution. The second one is goal-oriented and based on the dual-weighted residual method, which requires the computation of an adjoint problem. Applications include direct numerical simulations of 3D turbulent cases such as the flow in a periodically constricted channel, also called periodic hill, and the flow around a NACA4412 wing profile at $Re_c = 850,000$. We ensure that the adaptive simulations are reliable and stable, and we compare the choice of error estimators on the refinement patterns and other relevant flow quantities. [Preview Abstract] |
Monday, November 25, 2019 5:55PM - 6:08PM |
P16.00004: Algorithmic grid selection in LES Siavash Toosi, Johan Larsson, Ivan Bermejo-Moreno Given the recent progress in LES modeling and numerical schemes the computational grid has now become the single most important factor determining the quality of an LES; and yet the current state-of-the-art is to rely fully on user expertise to build the grid. While this is a workable process for academic problems in relatively simple geometries, it becomes untenable going forward towards more complex flows in complex geometries with multi-physics effects at increasing computational scales. The present work is aimed at developing an algorithmic process for how to select a nearly optimal grid (maximal accuracy at minimal cost) for LES. Two error indicators are used to drive an iterative grid selection process, where the solution from a previous LES run is used to select a more optimal grid for a subsequent run. The resulting method is highly systematic, with minimal dependence on user input, and can be operated both in a free mode that results in unstructured grids and in a constrained mode that results in structured grids. The process is tested on a variety of test cases, including wall-resolved and wall-modeled LES of both canonical flows (channels and boundary layers) and more complex flows (backward-facing step and smooth-body separation) with excellent results in all cases. [Preview Abstract] |
Monday, November 25, 2019 6:08PM - 6:21PM |
P16.00005: Multifidelity ensemble-based prediction of turbulent flows at the Exascale Lluis Jofre, Manolis Papadakis, Alex Aiken, Gianluca Iaccarino The study of complex multiphysics turbulent flows is commonly based on intensive computational high-fidelity simulations. To build confidence and improve their prediction accuracy, very large computational budgets are typically required to characterize the impact of uncertainties on the quantities of interest. In this regard, multifidelity methods have become increasingly popular in the last years as acceleration strategies. Exascale computing resources promise to facilitate the use of these approaches on larger scale problems by providing 1-10k times augmented floating-point capacity, but at expenses of requiring more complex data management as memory is expected to become more heterogeneous and distributed. The objective of this work, therefore, is to explore the performance of multifidelity ensemble-based strategies in large-scale multiphysics applications using an Exascale-ready computational framework. [Preview Abstract] |
Monday, November 25, 2019 6:21PM - 6:34PM |
P16.00006: High-fidelity simulations for wind farm control co-design: evaluation of individual blade pitch control for turbine arrays and utility-scale wind farms Fotis Sotiropoulos, Xiaolei Yang, Peter Seiler With the exponential growth of computer power, high-fidelity simulations are playing an increasingly important role enabling for the first time control co-design of wind farms, which can dramatically increase the annual energy production (AEP) and reduce the levelized cost of energy (LCOE). Individual blade pitch control (IBPC), which can effectively reduce the load fluctuations caused by the non-uniform incoming wind speed, has the potential to significantly reduce the LCOE of wind farms. IBPC, however, has been mostly evaluated for stand-alone individual wind turbines and its promise has yet to be demonstrated for turbine arrays especially at utility scale. We employ herein the VFS-Wind code to carry out large-eddy simulation (LES) with actuator-based parametrizations of turbine blades to explore the potential of IBPC in large wind farms. Two types of cases are investigated computationally: 1) a canonical turbine array with three different spanwise turbine spacings with the downwind turbine spacing fixed at seven rotor diameters; and 2) the XCEL Energy utility-scale wind farm in Pleasant Valley, Minnesota, United States, which consists of 100 wind turbines with generation capacity up to 200MW. Significant fatigue load reduction is observed for all the simulated cases. [Preview Abstract] |
Monday, November 25, 2019 6:34PM - 6:47PM |
P16.00007: Blade Resolved Wind Turbine Simulation with the Hybrid Time-Averaged Model Split Turbulence Model Jeremy Melvin, Marc Henry de Frahan, Shreyas Ananthan, Ganesh Vijayakumar, Robert Moser Blade resolved numerical simulations of wind turbines are an essential tool to help guide turbine design and placement in wind farms. Due to the high Reynolds number of these flows and the complex geometry of the turbine, a computational approach utilizing Reynolds Averaged Navier Stokes (RANS) or hybrid RANS/Large Eddy Simulation (LES) turbulence models are needed for computational feasibility. However, typical turbulence modeling approaches struggle with the complex flow characteristics present in flow fields around wind turbines. The newly developed Time-Averaged Model Split (TAMS) hybrid RANS/LES approach described by Haering et al. (AIAA Scitech 2019 Forum, AIAA 2019-0087, 2019) has shown the potential to resolve many of the issues with existing hybrid approaches for these complex flows. In this talk, we discuss efforts to conduct a blade resolved wind turbine simulation with the TAMS model. We integrate TAMS with a base SST RANS model, and conduct both airfoil and turbine simulations to compare performance and efficiency with standard DES hybrid approaches. We highlight the advantages of TAMS and discuss a path forward for additional improvements. [Preview Abstract] |
Monday, November 25, 2019 6:47PM - 7:00PM |
P16.00008: Nek5000 LES of realistic urban geometries initialized from weather models Aleksandr Obabko, Gökhan Sever, Rajeev Jain, Yu-Hsiang Lan, Paul Fischer, Haomin Yuan, Robert Jacob, Charlie Catlett, Misun Min In the atmospheric modelling community, urban boundary layers have been generally treated using mesoscale models where a presence of obstacles are taken into account by paramaterizations of the urban (or vegetation) canopy. There are not enough CFD-quality observations to validate the momentum and heat transfer in these models for all cases of interest. We seek to improve parametrization and complement observations with CFD reference solutions using an open-source CFD solver Nek5000. High scalability combined with high-order discretization and low count of degrees of freedom per processor allows an efficient exploitation of high-fidelity approaches like large-eddy simulation (LES) in complex geometries. With initial and boundary conditions derived from a High-Resolution Rapid Refresh (HRRR) initialized WRF-urban model, we have performed wall-resolved Nek5000 LES of realistic urban geometries including Lake Point Tower and Goose Island regions of Chicago. These preliminary simulations suggest we have the capability to start modelling the flow in any city geometry assuming a proper mesh can be built. Extending these reference simulations to larger domains will be possible with upcoming exascale supercomputers while longer simulation times may still remain a challenge. [Preview Abstract] |
Monday, November 25, 2019 7:00PM - 7:13PM |
P16.00009: BCM-LES application to wind gust disaster under extreme meteorological events. Tetsuro Tamura, Masaharu Kawaguchi Recently in Japan, people tend to realize so frequent occurrences of tornado and many attacks of typhoon to Japanese islands. In 2011, Tsukuba tornado arose in the Kanto plain based on a supercell. It causes the wind gust disaster on houses built on city area. In 2018, the typhoon Jebi attacked at Osaka area on September 4. High wind collapsed the wooden houses and the claddings of buildings and structures. This study applies LES based on BCM (Building Cube Method) to flow field for several km square area of disaster occurrence. This numerical model is formulated on the very fine Cartesian mesh system utilizing the IBM which can solve unsteady wind flows around actual complicated building shapes. Also, this research presents LES method for generating broad-banded turbulence flow that is able to regenerate high frequency components for existing meteorological model output. We performed the hybrid analysis of the meteorological model and engineering LES, in order to investigate near-ground turbulent wind under extreme meteorological events. This study numerically estimates the maximum wind velocity and the peak pressures on the building. Mechanism for process to failure of buildings will be discussed for establishment of safety at city. [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