Selected-Eddy Simulations (SES): a novel approach for turbulence simulations at extreme scales
ORAL
Abstract
Direct Numerical Simulations (DNS) resolve all dynamically relevant scales of a turbulent flow. Because of the detail they provide, they have
become indispensable to fundamental turbulence research despite its high computational cost at high Reynolds numbers. At realistic conditions, though, the computational power needed to resolve all spatial and temporal scales in flows of practical interest is beyond today's capabilities. Thus, other methods have been developed which are less expensive but, as expected from well-known trade-offs, less accurate. One such widely-used approach is Large Eddy Simulation (LES) which resolves only the large scales and models the small scales. This approach presents challenges in flows where small-scale physics is dominant, like fast reactions or shocks.
We present a novel method, namely Selected Eddy Simulations (SES), which resolve a subset of modes across the entire spectrum of scales and models the complementary set. The key ingredient in this approach is the proper selection and modeling of unresolved modes which can generally be tailored to particular problems. We show recent results using SES and compare them with well-resolved DNS of isotropic turbulence at different Reynolds numbers and show that SES can produce the entire spectrum and a number of important statistics with high degree of accuracy with as low as 10% of the modes. Effects of modeling methods, and the fraction and distribution of unresolved modes, and their variation with Reynolds number will be discussed. We will also discuss aspects of implementation of SES at extreme scales and the challenges expected on next-generation computational systems.
become indispensable to fundamental turbulence research despite its high computational cost at high Reynolds numbers. At realistic conditions, though, the computational power needed to resolve all spatial and temporal scales in flows of practical interest is beyond today's capabilities. Thus, other methods have been developed which are less expensive but, as expected from well-known trade-offs, less accurate. One such widely-used approach is Large Eddy Simulation (LES) which resolves only the large scales and models the small scales. This approach presents challenges in flows where small-scale physics is dominant, like fast reactions or shocks.
We present a novel method, namely Selected Eddy Simulations (SES), which resolve a subset of modes across the entire spectrum of scales and models the complementary set. The key ingredient in this approach is the proper selection and modeling of unresolved modes which can generally be tailored to particular problems. We show recent results using SES and compare them with well-resolved DNS of isotropic turbulence at different Reynolds numbers and show that SES can produce the entire spectrum and a number of important statistics with high degree of accuracy with as low as 10% of the modes. Effects of modeling methods, and the fraction and distribution of unresolved modes, and their variation with Reynolds number will be discussed. We will also discuss aspects of implementation of SES at extreme scales and the challenges expected on next-generation computational systems.
–
Presenters
-
Diego A Donzis
Texas A&M University
Authors
-
Diego A Donzis
Texas A&M University