1 EDF Electricité de France Materials Performance Centre Seminars, 12/09/2006 Application of Large Eddy Simulation to thermal-hydraulics in the Nuclear Power Generation Industry. [email protected]School of Mech, Aero & Civil Eng. Fluids AIG / CFD group [email protected]EDF R&D Chatou tions: d, I. Afgan, S. Benhamadouche, S. Berrouk, N Jarrin C. Moulinec, T. The University of Manchester
Materials Performance Centre Seminars, 12/09/2006. Application of Large Eddy Simulation to thermal-hydraulics in the Nuclear Power Generation Industry. [email protected] School of Mech, Aero & Civil Eng. Fluids AIG / CFD group [email protected] EDF R&D Chatou. - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Contributions: Y. Addad, I. Afgan, S. Benhamadouche, S. Berrouk, N Jarrin C. Moulinec, T. Pasutto,
The
Uni
vers
ity
of M
anch
este
r
2
EDFElectricitéde France
Osborne Reynolds 1868, became a professor of engineering at Owens College (now University of Manchester)
Reynolds tank, G. Begg building
Low speed jet
Higher speed jet
Turbulence, Reynold Number Re = UD/visc.
3
EDFElectricitéde France
Kolmogorov Energy Cascade
2 212
12
Assume : ( ) sin( ) 2 /
. sin( )cos( ) sin(2 )
thus energy transfer from 2 / to 2 2 /( )
i i i i
i i i i i i i
i i i i
u x u k x k L
duu u k k x k x u k k xdx
k L k L
Андре́5й Никола́5е́вич Колмого5ров
Moscow State Uni. 1939
4
EDFElectricitéde France
Solve the standard (Navier Stokes) flow equation on a very fine mesh- All scales are resolved on that mesh (down to Kolmogorov scale)- No model needed- Results considered as valid as experimental data- Enormous computer resources, for even modest speed-domain size (Reynolds number)
Sergio Hoyas and Javier Jimenez, (2006) "Scaling of velocity fluctuations in turbulent channels up to Re_tau = 2000", Phys. of Fluids, vol 18,
Luckily, only Large Scales matter (most of the time)
Large Scales & Human activity• Drag, mixing, heat transfer, • Large Scales dictate flow physics• Generated by/scale with obstacle• Impose dissipation rate
Exceptions: noise, combustion,
Weather forecast (we are the small scales !)
6
EDFElectricitéde France
Large Eddy Simulation = Filtering
1( , ) ( , ') '
2
t T
t Tu x t u x t t dt
T
drrxGtrutxuV
)(),(),(
CFD codes naturally induce filter = 2 dx
Space Filter
Time Filter
7
EDFElectricitéde France
DHIT: Decay of Hom. Iso. Turb.
MANDATORY test case for first time LES !
Reveals numerical dissipation, stability,
G rid
S ta tio n4 2
S ta tio n9 8
U = 1 0 m s0
-1
8
EDFElectricitéde France
• Direct Numerical Simulation (DNS) databases <=> Experiments
= “Costly” Fluid Dyn., exceptional, limited to zoom effect,
- 100% accurate, back to 1st principles, NO modelling hypothesis
• Large Eddy Simulation (LES)= “Colourful” Fluid Dyn., much detail,
fluctuations, spectra,- Applicable Eng. problems, at some cost- Almost as reliable as DNS, but know-how
required, not well established
• Reynolds Averaged (RANS)= “Conventional” Fluid Dyn., used daily in
Eng., only mean values (B&W)- Economical, full reactor or sub-component
design (parametric) possible- Problem: wide range of models to choose
from, - needs improvement & validation for new
range of applications (high temperature, buoyancy, conjugate heat
transfer, )
Future : Coupling of RANS and LES, using DNS for insight & validation
3 Levels of CFD approaches to turbulent flow
9
EDFElectricitéde France
Industrial LES applications to reactor thermal hydraulics
LES is mostly about numerical methods
Grid able to capture most turbulent scales Easier in Power Industry (confined, non-streamlined geometry) Local/embedded grid refinement, polyhedral scales
Boundary conditions Walls => quasi DNS (wall functions not ideal) Some real periodic geometry. pb. in Power Gen (tube bundles) Synthetic inlet turbulence
Target values Order of Mag. (within 10%), not 0.01 on Cd Thermal mixing & loading, spectra, vibrations….
Need Fast Unstructured FV Solver
EDF code Saturne & Star-CD very similar
High Accuracy Numerical Scheme No numerical dissipation
( Central differencing, Second order in time) Avoid any mesh distortion
Addad Y. , D. Laurence and S. Benhamadouche. The Negative Buoyant Wall Jet: LES Results, I.J. Heat and Fluid Flow, 25, 795-808, 2004
22
EDFElectricitéde France
Grid generation (buoyant case)
● Pre- k-eps simulation
● cell Volume near jet inlet (V)1/3=0.002● y+=1
● In mixing region (V)1/3=0.007
● NCELLS=770 000
●StarCD code
IntegralLength scalefrom k-epsilon
23
EDFElectricitéde France
Horizontal Velocity Comparison
24
EDFElectricitéde France
LES DB => Analytic Wall Function development
(from A. Gerasimov)
25
EDFElectricitéde France
Thermal hydraulics of reactors
Study the physics of the flow in the decay heat inlet pen
Examine the LES solution of the code Star-CD for the natural/mixed convection cases.
Validate further the analytical wall functions developed at UMIST by Gerasimov et al.
Mixed convection in co-axial pipes(Y. Addad PhD, M. Rabitt British Energy)
26
EDFElectricitéde France
K-eps pre-study
Streamlines coloured by temperature
Cold Inlet Pipe in vessel => stratification trap
27
EDFElectricitéde France
Coaxial heated cylinder study
• LES validation and parametric test cases: Case1-Natural convection in square cavity (Ra=1.58 109) Case2-Natural convection in annular cavity (Ra=1.8109) Case3- annular cavity single coaxial cylinder (Ra=2.381010) Case4- annular cavity with 3 coaxial cylinders (Ra=2.381010) Case5- Flow in actual penetration cavity (bulk Re=620,000).
reconstruct temperature fluctuations at solid wall
Link with materials ageing research
38
EDFElectricitéde France
Research: Mesh strategy for LES
a)
Possible FV near-wall refinements: a) dichotomy, b) non-conforming, c) & d) polyhedral & zoom.
- Non dissipative Finite Volume Methods
- Optimal meshing strategy for LES
- Quality criteria for LES General but essential issue. Collaboration with:- CD – Adapaco (STAR-CD code)- EDF R&D (Saturne code)- Health & Safety Labs, CFD for Nuclear Reactor licensing ?
39
EDFElectricitéde France
DNS at Re* N Cells
395 10 Million
640 28 Million
720 84 Million
2000 17,921 Million
RANS LES
Under-resolved LESRANS – LES coupling
U
Wall distance
J. Uribe, Manchester
Research: RANS – LES coupling
LES: ~ 0.1 Million cells
40
EDFElectricitéde France
Conclusions – Industrial LES
• LES of Industrial flow• Much more information:
Thermal stresses, fatigue, Acoustics, FIV (vibrations)• Cost-wise accessible when limited to subdomain
(synthetic turbulence for inlet)• Complex geometry possibly easier than smooth channel flow (academic overkill ?)
• Flexible flexibility with professional/commercial software:
• Opens new range of applications for LES• Medium Re number : DNS near wall resolution possible• Greater breakthrough than elaborate SGS models?
• Further developments: • More meshing control (total cell size control from pre-simulation)• High Re : RANS –LES coupling, embedded LES