15:00
Large Eddy Simulation 2
Chair: Richard Stevens
15:00
15 mins
|
Global and local energy dissipation in a turbulent Von Kármán flow
Denis Kuzzay, Davide Faranda, Bérengère Dubrulle
Abstract: We use PIV measurements to study local and global energy transfer in a Von Kármán flow. First, we use a Large Eddy Simulation (LES) approximation to model and compute the injected and dissipated power. This procedure involves a free parameter that is calibrated using angular momentum balance. We then estimate the local and global mean injected and dissipated power for several types of impellers, for various Reynolds numbers and for various flow topologies. These PIV-estimates are then compared with direct injected power estimates, provided by torque measurements at the impellers. The agreement between PIV-estimates and direct measurements depends on the flow topology. In symmetric situations, our estimates capture 30 to 70% of the actual energy dissipation. However, our results become increasingly inaccurate as the shear layer responsible for most of the dissipation is approaching one of the impeller, where it cannot be resolved by our PIV set-up. Finally, we show that a very good agreement between PIV-estimates and direct estimates of the dissipated power is obtained using a new method based on the work of Duchon and Robert that generalizes Kármán-Horwath equation to non-isotropic, non homogeneous flows. This method provides parameter-free estimates of the energy dissipation as long as the smallest resolved scale lies in the inertial range.
|
15:15
15 mins
|
CAN REYNOLDS STRESS TRANSPORT MODELS BE USED FOR LARGE EDDY SIMULATION?
J. Blair Perot, Jason Gadebusch
Abstract: This work explores a route to unify Reynolds averaged (RANS) and large eddy simulation (LES). The approach is to use a slightly modified Reynolds stress transport model for any mesh resolution. The model is formulated in terms of both total kinetic energy and modeled kinetic energy in such a way that the RST model correctly reproduces RANS results, LES results, and even DNS results (by turning itself off). The model equations do not contain functions of the mesh size within any of the model terms or constants. It is demonstrated that this approach works at any mesh resolution. In addition, the model naturally transitions between mesh resolutions, either coarse to fine or vice-versa. It is shown that for LES mesh resolutions the model returns a turbulent length scale that is proportional to the mesh size (the classic LES turbulent length scale).
|
15:30
15 mins
|
Large-Eddy Simulation of Turbulent Supersonic Cold Flow in Ramp-Cavity Combustor With Injector
Zia Ghiasi, Dongru Li, Jonathan Komperda, Farzad Mashayek, Arnab Chaudhuri, Gustaaf Jacobs
Abstract: Development of efficient supersonic propulsion systems requires a deep understanding of the flow within such combustors. Performing numerical simulation of flow in a scramjet engine is challenging due to the concurrent presence of shock waves and turbulence. In this work, we employ a high-order discontinuous spectral element method (DSEM). An entropy-based artificial viscosity method is used to capture the shock and the turbulence model is the standard Smagorinsky-Lilly model in conduction with two new sensors that prevent addition of turbulent viscosity in non-turbulent regions. Results are presented for the simulation of supersonic turbulent cold flow in a ramp-cavity combustor with an injector at the ramped side of the cavity. The Reynolds number based on the cavity height and mean inlet flow velocity is Re=25,288 and the Mach number is Ma=2.
|
15:45
15 mins
|
The influence of subgrid-scale modelling on the performance of a new non-equilibrium wall-model for large-eddy simulation
William Sidebottom, Olivier Cabrit, Ivan Marusic, Charles Meneveau, Andrew Ooi, David Jones
Abstract: The computational cost of wall-resolved large-eddy simulations (LES) rapidly becomes prohibitive with increasing Reynolds number. Wall-modelled LES attempts to significantly reduce the computational cost of simulating wall-bounded turbulent flows by modelling the effect of the near-wall small-scale motions, rather than fully or partially resolving them. The present study concentrates on a new wall-model that is able to predict fluctuating wall-shear stress given a large-scale velocity input. The velocity input for the model is affected by the choice of subgrid-scale (SGS) model. Therefore, this study also focusses on the impact of the SGS-model on the distribution of quantities at the wall. Results show that the new wall-model is able to resolve more of the wall shear-stress variance than a standard wall-model; and that the SGS-model affects the distribution of fluctuations of both wall-shear stress and wall-pressure.
|
16:00
15 mins
|
Energy flux in isotropic turbulence under large variations of external forcing
Haitao Xu, Fabio Di Lorenzo, Alain Pumir, Eberhard Bodenschatz
Abstract: We investigate the response of energy flux in isotropic turbulence to step-function like perturbation in external forcing at large length scales.
From both physical experiments and direct numerical simulations, we measured the evolution of the Eulerian velocity structure functions, such as $D_{LL}(r)$, $D_{NN}(r)$, before and after the perturbation in forcing.
In both cases, we observed the cascade of the energy excess at large scales cascade through scales to the dissipative range, which can be used to study the dynamics of the cascade, and in particular, to estimate the relevant time scales.
|
16:15
15 mins
|
"A Dynamic Subfilter-scale Stress Model for Large Eddy Simulations Based on Physical Flow Scales"
Amirreza Rouhi, Ugo Piomelli, Bernard Geurts
Abstract: We propose a new definition of the length scale in an eddy-viscosity model for large-eddy simulations (LES). This formulation extends and generalizes a previous proposal [Piomelli, Rouhi and Geurts, Proc. ETMM10, 2014], in which the LES length scale was expressed in terms of the integral length-scale of turbulence determined by the flow characteristics and explicitly decoupled from the simulation grid; this approach was named Integral Length-Scale Approximation (ILSA). As in the original ILSA, the model coefficient was determined by the user, and required to maintain a desired contribution of the unresolved, subfilter scales (SFS) to the global transport. We propose a local formulation (local ILSA) in which the model coefficient is local in space, allowing a precise control over SFS activity as a function of location. This new formulation preserves the properties of the global model; application to channel flow and backward-facing step verifies its features and accuracy.
|
16:30
15 mins
|
A Lagrangian sub-grid model for the dispersion of clouds of tracers
Federico Toschi, Irene M. Mazzitelli, Alessandra S. Lanotte
Abstract: Turbulence models are expected to satisfy the conflicting requirements
of accuracy and computational efficiency. Here we discuss a new model
that was recently developed in order to accurately and efficiently
describe the dynamic of a clouds of tracers particles in Large Eddy
Simulations of homogeneous and isotropic turbulent flows. The models
incorporates the multi-scale nature of time and space turbulent
velocity correlations that are essential in order to correctly
reproduce the relative dispersion of multi-particle clouds. The model
can be seen as an off-grid solver for the Eulerian velocity field at
the positions of a given number of Lagrangian tracers that
self-consistently move with it. Extensions to non homogeneous and
isotropic turbulence as well as to the dynamics of particles will be
discussed.
|
16:45
15 mins
|
Stochastic Reynolds transport theorem and generalized subgrid tensor
Valentin Resseguier, Etienne Memin, Bertrand Chapron
Abstract: We propose a representation that allows decomposing the flow velocity in terms of a smooth component and a highly oscillating
random component. This decomposion leads through a stochastic representation of the Reynolds transport theorem to a large-scale
expression of the Navier-Stokes equations. In this work we show the benefit of such a representation to construct low order dynamical
systems that include naturally a dissipative term related to the action of the small-scale random component.
|
|