Sie sind auf Seite 1von 58

Implicit LES of Low and High Speed Flows Using High-Resolution and High-Order Methods

Dimitris Drikakis
Cranfield University Aerospace Sciences Department Fluid Mechanics & Computational Science Group d.drikakis@cranfield.ac.uk

Classical & Implicit LES


Classical LES (filtering of the Navier-Stokes equations)

u j ij ui u i + + (ui u j + p ij ) = x x j t x j x x i j j
U( x, t ) = U ( x) + U( x, t )

U ( x, t ) =

G( x , t t )U( , t )dt d

SGS : ij = uiuj = ui u j ui u j

Implicit LES (no explicit filtering and dissipation)

u j ui u i + + (ui u j + p ij ) = t x j x j x x j i

Some issues about classical LES


Commutation errors: Masking of the SGS terms by the truncation errors Aliasing errors in high-order methods used in classical LES, e.g., spectral schemes Design of SGS models in high-Re (wall-bounded) number flows Quantification of errors associated with SGS terms (Geurts & Frlich, POF, 2002) Coupling with RANS: Mathematically and numerically inconsistent

LES : U ( x, t ) =

G( x , t t )U( , t )dt d
T

1 RANS : U ( x) = lim U( x, t )dt T T 0

Remarks
SGS models provide the dissipation for unresolved scales and numerical stability but the models and numerics cannot be decoupled unless the flow is fully resolved Are LES (and DNS) fully resolved? This always requires explicit verification Does this occur in practice? Most often not

What about under-resolution (intermittency!)? Less resolution means non-smooth solutions locally this motivates the use of high-resolution (nonlinear) methods Not all high-resolution methods are good for ILES

LES of Bifurcating and Transitional Suddenly Expanded Flows

Numerical Methodology

Implicit Large-eddy Simulation


 The physical link: The Taylor series expansion of the subgrid stress tensor (Approximate Deconvolution) is exactly the same as the conversion from a cell averaged to a cell centred value Thus a 3rd order accurate MUSCL scheme contains inherently a 2nd order subgrid model  This is a structural model dissipation is dependant on the resolved scales

Shock-Capturing Methods for Turbulence: Theoretical Basis


Kolmogorov (1941):

5 3 K t L = (u ) 4

Bethe (1942):

G c 3 S = (u ) 3 6cs T
3 2

S= entropy V=specific volume

K=kinetic energy L=length scale

G= curvature of the isentrope

If your theory is found to be against the second law of thermodynamics, I give you no hope; there is nothing for it but to collapse in deepest humiliation. Sir Arthur Eddington

Shameless advertising Examples of Highresolution methods:


Fi +1/ 2 = 1 1 (FL + FR ) A (U R U L ) 2 2

high low Fi +1/ 2 = Filow + ( F F i +1 / 2 i +1 / 2 ) +1 / 2

Initial data

Averaging (Godunov-type)

Characteristics-based solution

Flux Calculation

Implicit Model
 This presentation will focus on two reconstruction methods: 2nd order Minmod reconstruction

3rd order limiting method

 Dissipation can be determined by writing the fluxes as

Implicit Model U Momentum Eqn.


 The truncation error of the Minmod limiter is

 For the third order limiter

Implicit Model U Momentum Eqn.


 The kinetic energy dissipation for the third order limiter is

 The kinetic energy dissipation for the expansion of the exact subgrid tensor gives

 Very good agreement

Homogeneous decaying turbulence in a triply periodic cube


 The flow field is initialised from a kinetic energy spectrum  Mach = 0.1, Four resolutions investigated 323,643,1283 and 2563  If KE decay rate is not in the correct range then the simulation will rapidly decorrelate from reality  Generally accepted that KE decays as:  The decay exponent p varies from 1.2-1.37 in recent experimental results, standard LES from 1.17 2.0 were quoted

Isosurfaces of Vorticity at 2563

Mesh Resolution 323 643 1283 2563

Minmod 0.83 1.26 1.34 1.28

3rd order 1.1 1.39 1.34 1.22

Velocity Derivative Skewness


 Calculated as:

 Most recent high Re experiments ~ -0.34, typical experimental values are -0.5, DNS ~ -0.4 to -0.5

Mesh Resolution 323 643 1283 2563

Minmod -0.08 -0.22 -0.31 -0.34

3rd order -0.14 -0.31 -0.34 -0.36

The Experiment
 Half-height experiment reported by Holder and Barton (IWPCTM, 2004)

Video of the experiment

(Courtesy of Holder and Barton, IWPCTM, 2004)

Numerical Method
 Finite volume Godunov scheme using a characteristics based solver  THCM gas mixture model (5 additional equations)  2nd order MUSCL reconstruction in space, third order Runge Kutta in time  Mesh resolution 600 x 160 x 320  Slip wall boundary conditions  1-D extended domain for the drain hole  Ratio x/M = 1E(-04)

SF6 density (b/w)

SF6 density (colour)

Comparison with experiment

Volume fraction

Shock / SF6 Position

 Excellent agreement with SF6 and shock position  Slight discrepancy in shock angle due to diffusion in experiment

Shock / SF6 Position

 Good overall matching of experiment and results  No oscillations at shock front or SF6 boundaries

Very high-order (5th & 9th order) ILES: Richtmyer-Meshkov Turbulent Mixing

Richtmyer-Meshkov

Time evolution comparison of experiment (Jacobs) with WENO5 ILES code

Time evolution comparison of experiment (Jacobs) with WENO5 ILES code

Time evolution comparison of experiment (Jacobs) with WENO5 ILES code

Time evolution comparison of experiment (Jacobs) with WENO5 ILES code

Time evolution comparison of experiment (Jacobs) with WENO5 ILES code

Time evolution comparison of experiment (Jacobs) with WENO5 ILES code

Incompressible Lid-Driven Cavity


 Standard benchmark case  Detailed experimental data available for laminar, transitional and turbulent flows  Re= 3,200 (laminar), 5,000 (transitional), 10,000 (fully turbulent)  64x64x64
Fully-developed turbulent flow

Validation (Laminar)

Validation (Turbulent)

Validation (Transitional)

Compressible, Open Cavity Flows


Strouhal Number

Comparison of DNS and ILES

Coarse grid ILES: 15 times coarser grid than DNS

Turbulent Heat Transfer in Cooling Devices


JET Fusion Facility

Transitional and turbulent vortical flow Re=250,000, low speed ILES using 3M grid points Experiment Manchester Univ.

Comparison with the experiment

No forcing

Forcing (prescribed spectrum)

Synthetic jet configuration

Re =

U jd

Average velocity during expulsion phase of cycle in experiments

U j = 14.0 m/s (PIV) Re = 1150

Sine wave as inflow boundary condition at cavity base

u = A sin( 2 pi f t )

Experimental data provided by NASA Langley Research Centre.

Max amplitude

Frequency of oscillation

Synthetic Jets in Quiescent Air

Streamwise velocity contours with streamlines at various phase angles

45

90

135

180

225

270

Synthetic Jets in Quiescent Air

Isosurfaces of vorticity at various phase angles

45

90

180

270

360

Comparison with experiments


Velocity profiles across the slot exit at 1mm above the slot exit at phase angle of 90

20

15

CFD PIV data

u-velocity m/s

v-velocity m/s

CFD PIV data

10

-1

5
-2

0 -2 -1 0 1 2

-3

position across slot mm

-2

-1

position across slot mm

Streamwise velocity

Cross streamwise velocity

Comparison with experiments


Velocity profiles across the slot exit at 1mm above the slot exit at phase angle of 270
4
-1

3 2

-2

u-velocity m/s

v-velocity m/s

1 0 -1 -2 CFD PIV data

-3

CFD PIV data

-4

-5

-3 -4

-2

-1

-2

-1

position across slot mm

position across slot mm

Streamwise velocity

Cross streamwise velocity

Work in Progress  Various strategies of Hybrid ILES and RANS  Multi-phase flows  Contaminant dispersion (indoor and outdoor configurations)  Coupling ILES with acoustic models: high-order BEM and Ffowcs-Williams-Hawkings

Hybrid ILES-RANS FLOWer code DLR & ILES Method (Cranfield)

2nd-order RANS model

High-Resolution Scheme (Cranfield)

2nd-order RANS model DLR, Germany

High-Resolution Model (Cranfield)

Ventilation and Contaminant Dispersion in Aircraft  Problem: Ventilation flow analysis in Airbus A380 aircraft cabin

Experimental Data
 DLR experiment

RANS Computations
Experiment RANS computation
1.5
Velocity 1.55 1.3 1.05 0.8 0.55 0.333169 0.170712 0.05

Y
0.5 0

0.5

1.5

High-speed flows: Schematic view of foreshock/aftershock interactions on regions of a re-entry vehicle

Instabilities in high-speed flows and their control

Pulsation mode around a spiked cylinder (M=6.0).

Oscillation mode (M=6.0).

Contaminant Propagation
Wind

10000 ppm emission

900 800 700 600 500 400 300 200 100 0 0 4 9 13 17 22 27 31 35.5 40

C o n cen tartio n (p p m )

53.6

Concentration (ppm)

53.5 53.4 53.3 53.2 53.1 53


0 9 18 27 35.5 44 53 62 71 80 89 98 107 115.5

Distance (ft)

Distance (ft)

Critical level

Acceptable level

Coupling ILES with Acoustic Models


High-order BEM, FW-H

Acknowledgements
Fluid Mechanics & Computational Science Group, Cranfield: Ben Thornber Andrew Mosedale Marco Hahn Sanjay Patel Zeshan Malick Evgeniy Shapiro Chara Papachristou Anthony Milonas Bowen Zhong David Youngs (AWE) Fernando Grinstein, Bill Rider (LANL) Thorsten Schwartz, Klausdieter Phalke (DLR)
Sponsors:

MOD EPSRC AWE UKAEA Eaton Aerospace BAE Systems European Union DLR Eurocopter Turbomeca

Short Course
Large Eddy Simulation of Transitional and Turbulent Flows: Numerical, Physical & Implementation Issues

Cranfield University 6-8 November 2006 http://www.cranfield.ac.uk/soe/cpd/ large_eddy_simulation.htm

Das könnte Ihnen auch gefallen