Presentation
Presentation
Presentation
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
www.numeca.com<br />
Towards a succesful implementation of<br />
DES strategies in industrial RANS solvers<br />
L. Temmerman, Temmerman,<br />
Ch. Hirsch<br />
Acknowledgements<br />
The DESider project (Detached Eddy Simulation for Industrial Aerodynamics) is a collaboration between<br />
Alenia, ANSYS-AEA, Chalmers University, CNRS-Lille, Dassault, DLR, EADS Military Aircraft,<br />
EUROCOPTER Germany, EDF, FOI-FFA, IMFT, Imperial College London, NLR, NTS, NUMECA, ONERA,<br />
TU Berlin, and UMIST. The project is funded by the European Community represented by the CEC, Research<br />
Directorate-General, in the 6th Framework Programme, under Contract No. AST3-CT-2003-502842.
www.numeca.com<br />
Aim:<br />
To adapt a RANS solver to perform DES/LES computations<br />
Code overview<br />
Modeling<br />
Calibration<br />
Applications<br />
Concluding remarks<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Outline<br />
2
www.numeca.com<br />
Compressible, co-located finite-volume<br />
Diffusive term: second order central<br />
Inviscid term:<br />
Central scheme with scalar dissipation (Jameson et al, 1981)<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Code overview<br />
Central scheme with matrix dissipation (Swanson and Turkel, 1992)<br />
Upwind scheme (Roe, 1981)<br />
Low Mach number pre-conditionning<br />
4 th order Runge-Kutta smoother<br />
Acceleration with local time-stepping, residual smoothing and<br />
multi-grid<br />
Time-marching: implicit second-order backward differencing<br />
Turbulence models: Spalart-Allmaras, k-ε Yang-Shi, Menter SST<br />
3
www.numeca.com<br />
Two DES models:<br />
Spalart-Allmaras based (Shur et al, 1999)<br />
Menter SST based (Travin et al, 2002)<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
min ( d C Δ)<br />
Δ = max(<br />
Δx,<br />
Δy,<br />
Δz)<br />
l DES = w,<br />
DES<br />
l RANS<br />
= k<br />
0.<br />
5<br />
/<br />
CDES<br />
= 0.<br />
65<br />
( l C Δ)<br />
l DES = min RANS , DES<br />
( C ω )<br />
C = F C )<br />
DES<br />
1<br />
μ<br />
k −ω<br />
k −ε<br />
DES + ( 1−<br />
F1<br />
CDES<br />
( Δx,<br />
Δy<br />
Δz)<br />
Δ = max ,<br />
k −ω<br />
k −ε<br />
C = 0.<br />
78 C<br />
= 0.<br />
61<br />
DES<br />
DES<br />
DES modelling<br />
4
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Calibration of LES mode<br />
Purpose: ensure that the code returns a LES solution whenever<br />
required<br />
Calibration tool: decay of isotropic turbulence<br />
Benchmark data from Comte-Bellot and Corrsin (1971)<br />
Initial conditions:<br />
Velocity fields obtained from inverse<br />
Fourier transform<br />
Initial fields for turbulent quantities<br />
obtained from freezing the velocity fields<br />
Comparisons of energy spectra and the<br />
energy decay allows to calibrate model<br />
constants and to evaluate the level of<br />
numerical dissipation<br />
5
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Influence of the numerical scheme<br />
Calibration of LES mode<br />
6
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Calibration of LES mode<br />
Influence of the modeling and effect of varying the model and<br />
numerical parameters<br />
7
www.numeca.com<br />
List of applications:<br />
NACA0021 aerofoil with high angle of attack<br />
Flow over a descending bump in a closed duct<br />
Ahmed car body with 25 o slant<br />
All these cases have been considered within the DESider project<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Applications<br />
8
www.numeca.com<br />
Experiment of Katrina Swalwell<br />
(Monash University):<br />
Mean static pressure<br />
distribution on aerofoil<br />
Time histories of forces<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 1: NACA0021 aerofoil<br />
Low mach number flow (M = 0.11) at 60 o angle of attack (Re = 270,000<br />
based on freestream velocity and chord length)<br />
Grid made of 141 x 101 x 36 (approx. 510,000 cells - L z = 1 c - Δz =<br />
0.0278 c)<br />
Δt = 4.554 10 -5 s (Δt U o /c = 0.0125)<br />
Averaged sample: T U o / c = 177<br />
Modelling: SA-DES<br />
9
www.numeca.com<br />
Pressure coefficient distribution<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 1: NACA0021 aerofoil<br />
C L<br />
C D<br />
Exp. 0.921 1.547<br />
Comp. 1.020 1.679<br />
Lift and drag coefficients<br />
10
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 1: NACA0021 aerofoil<br />
PSD for the lift coefficient PSD for the tangential force<br />
coefficient<br />
11
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 1: NACA0021 aerofoil<br />
Correct prediction of C p distribution. Mean lift and drag<br />
coefficients are over-estimated<br />
Forces PSD: main frequencies are captured at the correct<br />
locations<br />
Animation shows nice unsteadiness in the aerofoil wake<br />
Results are satisfactory although:<br />
Spanwise extent is too small (4 chords recommended)<br />
Sampling duration is too short (at least 400 convective<br />
time units)<br />
12
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 2: Bump in closed duct<br />
Flow over a bump, inside a duct (rectangular section)<br />
Experimental data (ONERA) from the DESider Consortium:<br />
Pressure distribution along different lines and pressure spectra<br />
on the lower wall<br />
LDV and PIV measurements at different locations of the flow<br />
Inflow conditions<br />
Flow conditions:<br />
Incompressible flow<br />
Reynolds = 1,236,000 (based<br />
on a bump height 0.138 m and a<br />
reference velocity of 8 m/s)<br />
Modelling: SA-DES<br />
Grid made of 320 x 120 x 148 cells (3,907,200)<br />
Averaged sample: 3540 Δt (T Uo / c = 205.2)<br />
Δt = 10-3 s (Δt Uo /c = 5.797 10-2 )<br />
13
www.numeca.com<br />
Mean streamlines and<br />
separation zone<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 2: Bump in closed duct<br />
Iso-contours Iso contours of λ2 criterion<br />
Exp. Comp.<br />
Reattachment point ≈ 0.625 ≈ 0.47<br />
14
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 2: Bump in closed duct<br />
Streamwise velocity and shear stress profiles at<br />
x = 0.35 m in the centre plane<br />
15
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 2: Bump in closed duct<br />
Streamwise velocity and shear stress profiles at<br />
x = 0.92 m in the centre plane<br />
16
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 2: Bump in closed duct<br />
The agreement between the experiment and the computation is<br />
not too good with the flow reattaching too early<br />
The main features of the flow are however captured<br />
DDES modification may improve the solution<br />
17
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 3: Ahmed body with 25 o slant angle<br />
Ahmed body: generic car body with slant angle of 25 o<br />
Experimental data from Becker et al (2000):<br />
available from Ercoftac database (case C. 82 )<br />
LDA and hot wire measurements available<br />
Pressure measurements on the body rear<br />
Incompressible flow<br />
Reynolds number 768,000, based on a reference velocity of 40 m/s<br />
and a body height of 288 mm<br />
Modeling: SA-DES and SST-DES<br />
Grid:<br />
3,000,000 cells (SA-DES)<br />
4,500,000 cells (SST-DES)<br />
Δt = 10 -3 s<br />
18
www.numeca.com<br />
z (mm)<br />
380<br />
360<br />
340<br />
320<br />
300<br />
280<br />
260<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 3: Ahmed body with 25 o slant angle<br />
Experiments<br />
SA-DES<br />
SST-DES<br />
240<br />
-263 -243 -223 -203 -183 -163 -143 -123 -103 -83 -63 -43 -23 -3 17 37<br />
x (mm)<br />
Streamwise velocity profiles along the slant<br />
19
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 3: Ahmed body with 25 o slant angle<br />
Instantaneous snapshot of the wake (iso ( iso-contours contours<br />
of the vorticity magnitude – obtained with SST- SST<br />
DES)<br />
20
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Application 3: Ahmed body with 25 o slant angle<br />
Computations are unsteady with many structures in the wake<br />
Both simulations fail to predict the mean flow on the slant:<br />
SA-DES: the flow remains re-attached<br />
SST-DES: too strong recirculation zone on the slant<br />
… we see a strong influence of the underlying model in this<br />
case<br />
Failure to predict correctly the flow may be due to several<br />
factors: grid inadequacy, case more suitable for DDES, too much<br />
numerical dissipation?<br />
21
www.numeca.com<br />
DESider workshop, Corfou, Corfou,<br />
17-18 17 18 June ‘07<br />
Concluding remarks<br />
Once calibrated, the RANS solver is able to return LES-type<br />
solutions<br />
NACA0021 results are good, considering the limitations of the<br />
computations<br />
DESider Bump:<br />
Experimental and computational data do not agree too well<br />
Main flow features still correctly predicted<br />
Ahmed body:<br />
Failure to correctly predict the flow re-circulation on the slant<br />
Strong influence of the underlying RANS model of the DES on<br />
the results for this particular case<br />
Results are good for the case for which DES has been designed, not<br />
so for the others => some improvements needed (DDES, additional<br />
modifications)<br />
22