28.07.2013 Views

Elliptic-blending k - ε model

Elliptic-blending k - ε model

Elliptic-blending k - ε model

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Elliptic</strong>-<strong>blending</strong> k −<strong>ε</strong> <strong>model</strong><br />

Validation and verification<br />

Sylvain Lardeau


Introduction<br />

Assessing performance of the newly implemented elliptic<br />

<strong>blending</strong> <strong>model</strong> from University of Manchester, Billard &<br />

Laurence (2012), in canonical and complex flow<br />

configurations.


Introduction<br />

Assessing performance of the newly implemented elliptic<br />

<strong>blending</strong> <strong>model</strong> from University of Manchester, Billard &<br />

Laurence (2012), in canonical and complex flow<br />

configurations.<br />

1. Formulation of the baseline <strong>model</strong> (comments and<br />

adaptations needed)<br />

2. Wall-treatment for irregular meshes<br />

3. Results


Formulation<br />

Original <strong>model</strong>, Billard & Laurence (2012)<br />

Dρk<br />

Dt = Pk +E −ρ<strong>ε</strong>+ ∂<br />

<br />

µ µt ∂k<br />

+<br />

∂xj 2 σk ∂xj<br />

Dρ<strong>ε</strong> 1 <br />

= C<strong>ε</strong>1<br />

Dt τ<br />

Pk −ρC ∗ <strong>ε</strong>2<strong>ε</strong> + ∂<br />

<br />

µ µt ∂<strong>ε</strong><br />

+<br />

∂xj 2 σ<strong>ε</strong> ∂xj<br />

Dρϕ<br />

Dt = ρ(1−α3 )fw +ρα 3 ϕ<br />

fh −Pk<br />

k +CDϕ + ∂<br />

<br />

µ µt ∂<br />

+<br />

∂xj 2 σϕ ∂<br />

L 2 ∂2 α<br />

∂xj∂xj<br />

= α−1


Formulation<br />

Equation-by-equation modifications<br />

Turbulent kinetic energy k:<br />

Modifications:<br />

Dρk<br />

Dt = Pk +Pb +E −ρ<strong>ε</strong>+ ∂<br />

◮ Adding buoyancy source term Pb,<br />

∂xj<br />

<br />

µ µt ∂k<br />

+<br />

2 σk ∂xj<br />

◮ What alternative formulation for E ? Expensive to compute,<br />

and only active very near the wall.<br />

◮ Definition of time-scale τ and τlim:<br />

<br />

k2<br />

τ =<br />

<strong>ε</strong><br />

+C 2 t<br />

ν<br />

<strong>ε</strong> ; τlim<br />

<br />

CT<br />

= max τ, √<br />

3CµϕS<br />

with S = 2SijSij. Where is CT = 0.6 coming from?


Formulation<br />

Equation-by-equation modifications<br />

Turbulent dissipation rate <strong>ε</strong>:<br />

Dρ<strong>ε</strong><br />

Dt<br />

Modifications:<br />

1 <br />

= C<strong>ε</strong>1<br />

τ<br />

(fcPk +Pb)−ρC ∗ <strong>ε</strong>2<strong>ε</strong> + ∂<br />

∂xj<br />

◮ Adding buoyancy source term Pb,<br />

<br />

µ µt ∂<strong>ε</strong><br />

+<br />

2 σ<strong>ε</strong> ∂xj<br />

◮ Testing curvature correction fc, as suggested by Durbin for<br />

k −ω <strong>model</strong>s:<br />

◮ Option for C ∗ <strong>ε</strong>2 :<br />

fc = 1+α1|η3|+3α1|η3|<br />

C ∗ <strong>ε</strong>2 =<br />

⎧<br />

⎨ C<strong>ε</strong>2<br />

⎩ C<strong>ε</strong>2 +α3 (0.4−C<strong>ε</strong>2 )tanh<br />

<br />

∂ νt ∂k<br />

∂xj σk ∂xj<br />

<strong>ε</strong>−1 <br />

<br />

3/2


Formulation<br />

Equation-by-equation modifications<br />

Reduced normal stress ϕ:<br />

Dρϕ<br />

Dt = ρ(1−α3 )fw+ρα 3 fh−ρCµτlimϕ 2 S 2 + ∂<br />

with<br />

fw = − ϕ<strong>ε</strong><br />

2k ; fh = − 1<br />

τ<br />

Modifications:<br />

<br />

C1 −1+C2<br />

<br />

µ µt ∂ϕ<br />

+<br />

∂xj 2 σϕ ∂xj<br />

<br />

Pk +Pb<br />

ϕ−<br />

ρ<strong>ε</strong><br />

2<br />

<br />

3<br />

◮ Buoyancy source term added here Pb (is it correct?),<br />

◮ Cross-diffusion term neglected: found to pose significant<br />

problems in many cases with coarse mesh.<br />

◮ Dissipation term Pk ϕ<br />

k by ρCµτlimϕ 2 S 2 to avoid division by<br />

zero when k → 0.


Formulation<br />

All-y + wall-treatment<br />

Those terms needs to be defined at the first cell away from the<br />

wall:<br />

Remarks:<br />

Pk =<br />

<br />

(U + uk) 2 νt 0.95(1−νt)<br />

2 classical u∗<br />

u4 ∗ dU<br />

ν<br />

+<br />

dy +<br />

<br />

1− dU+<br />

dy +<br />

<br />

iterative u∗<br />

<strong>ε</strong> = Γ νk<br />

+(1−Γ)<br />

y2 E = − u2 k<br />

µ 0.0875exp<br />

Cµu∗k<br />

2κy<br />

<br />

−20.66<br />

◮ U + is taken from Reichardt’s law:<br />

U + = 1<br />

κ ln(1+κy+ )+C<br />

<br />

<br />

ln(y + )<br />

ln(10) −1.01<br />

<br />

2<br />

1−e<br />

with C = 1<br />

κ ln <br />

E 1 y +<br />

κ and b = mκ 1<br />

2 C +<br />

y + m<br />

−y +<br />

y + m + y+<br />

y + m<br />

<br />

.<br />

e −by+


Formulation<br />

All-y + wall-treatment<br />

Those terms needs to be defined at the first cell away from the<br />

wall:<br />

ϕ + ⎧<br />

⎪⎨ 0.0015y<br />

=<br />

⎪⎩<br />

+2<br />

if y + < y + 1<br />

Fi(t) if y +<br />

i ≤ y+ < y +<br />

i+1<br />

0.30766 log(y+ )<br />

log(10) −0.2775 if y+ ≥ y + 5<br />

Fi(t) = (2t 3 −3t 2 +1)f +<br />

i +(t3 −2t 2 +t)m +<br />

i ∆y+ +(3t 2 −2t 3 )f +<br />

i+1<br />

⎧<br />

˜y<br />

⎨<br />

˜y+ 1 if y<br />

˜y<br />

+ > 17<br />

α =<br />

⎩<br />

with uk = νy + /y.<br />

What about Pb?<br />

<br />

−y +<br />

1−exp<br />

24.52<br />

if y + < 17


Results: convergence


Validation<br />

Convergence<br />

Best convergence rate obtained when initial flow condition from:<br />

Turbulence Specification:<br />

turbulence intensity + turbulent viscosity ratio<br />

with Tu = 5% and TVR = 200.<br />

All cases computed in Double Precision.<br />

Residual<br />

1<br />

0.01<br />

1E−4<br />

1E−6<br />

1E−8<br />

1E−10<br />

1E−12<br />

1E−14<br />

1E−16<br />

1E−18<br />

0 1000 2000 3000 4000 5000<br />

Iteration<br />

Continuity<br />

X−momentum<br />

Y−momentum<br />

Tke<br />

Tdr<br />

Phi<br />

Alpha<br />

1E−12<br />

0 200 400 600 800<br />

Iteration<br />

1000 1200 1400<br />

Channel flow backward-facing step<br />

Residual<br />

10<br />

1<br />

0.1<br />

0.01<br />

0.001<br />

1E−4<br />

1E−5<br />

1E−6<br />

1E−7<br />

1E−8<br />

1E−9<br />

1E−10<br />

1E−11<br />

Continuity<br />

X−momentum<br />

Y−momentum<br />

Tke<br />

Tdr<br />

Phi<br />

Alpha


Validation<br />

Convergence<br />

For all cases shown below, the best convergence rate was obtained<br />

when the initial flow condition were computed from: Turbulence<br />

Specification: turbulence intensity + turbulent<br />

viscosity ratio with Tu = 5% and TVR = 200. The following<br />

cases were also run in Double Precision to validate the linearization<br />

method.<br />

Residual<br />

10<br />

0.001<br />

1<br />

0.1<br />

0.01<br />

1E−4<br />

1E−5<br />

1E−6<br />

1E−7<br />

1E−8<br />

1E−9<br />

1E−10<br />

1E−11<br />

1E−12<br />

1E−13<br />

1E−14<br />

1E−15<br />

1E−16<br />

1E−17<br />

1E−18<br />

1E−19<br />

0 2000 4000 6000 8000 10000<br />

Iteration<br />

12000 14000 16000 18000 20000<br />

1E−17<br />

0 400 800 1200 1600 2000 2400 2800<br />

Iteration<br />

Diffuser 2d naca 4412<br />

Residual<br />

10<br />

0.1<br />

0.001<br />

1E−5<br />

1E−7<br />

1E−9<br />

1E−11<br />

1E−13<br />

1E−15<br />

Continuity<br />

X−momentum<br />

Y−momentum<br />

Tke<br />

Tdr<br />

Phi<br />

Alpha


Results: validation of all-y + wall-treatment


Validation — Wall-treatment<br />

Channel flow<br />

7 different meshes, with different y + values but also different grid<br />

ratio above the first cell,<br />

for Reτ =550 but similar conclusion where obtained for Reτ = 2000.<br />

Mesh1 and LowRe Mesh2 Mesh3 Mesh4<br />

y + =1 y + =14 y + =28 y + =34<br />

Mesh5 Mesh6 Mesh7<br />

y + =66 y + =40 y + =14


Validation — Wall-treatment<br />

Channel flow<br />

Realizable k −<strong>ε</strong> two-layer <strong>model</strong>:<br />

U/uτ<br />

20<br />

15<br />

10<br />

5<br />

0<br />

DNS<br />

Mesh 1<br />

Mesh 2<br />

Mesh 6<br />

1 10 100<br />

y +


Validation — Wall-treatment<br />

Channel flow<br />

SST k −ω <strong>model</strong>:<br />

U/uτ<br />

20<br />

15<br />

10<br />

5<br />

0<br />

DNS<br />

Mesh 1<br />

Mesh 2<br />

Mesh 6<br />

1 10 100<br />

y +


Validation — Wall-treatment<br />

Channel flow<br />

<strong>Elliptic</strong>-<strong>blending</strong> k −<strong>ε</strong> <strong>model</strong>:<br />

U/uτ<br />

20<br />

15<br />

10<br />

5<br />

0<br />

DNS<br />

Mesh 1<br />

Mesh 2<br />

Mesh 6<br />

1 10 100<br />

y +


Validation — Wall-treatment<br />

Channel flow<br />

Realizable k −<strong>ε</strong> two-layer <strong>model</strong>:<br />

k/u 2 τ<br />

5<br />

4<br />

3<br />

2<br />

1<br />

0<br />

1 10 100<br />

y +


Validation — Wall-treatment<br />

Channel flow<br />

SST k −ω <strong>model</strong>:<br />

k/u 2 τ<br />

5<br />

4<br />

3<br />

2<br />

1<br />

0<br />

1 10 100<br />

y +


Validation — Wall-treatment<br />

Channel flow<br />

<strong>Elliptic</strong>-<strong>blending</strong> k −<strong>ε</strong> <strong>model</strong>:<br />

k/u 2 τ<br />

5<br />

4<br />

3<br />

2<br />

1<br />

0<br />

1 10 100<br />

y +


Validation — Wall-treatment<br />

Channel flow<br />

Realizable k −<strong>ε</strong> two-layer <strong>model</strong>:<br />

νt/ν<br />

100<br />

10<br />

1<br />

0.1<br />

0.01<br />

0.001<br />

1 10 100<br />

y +


Validation — Wall-treatment<br />

Channel flow<br />

SST k −ω <strong>model</strong>:<br />

νt/ν<br />

100<br />

10<br />

1<br />

0.1<br />

0.01<br />

0.001<br />

1 10 100<br />

y +


Validation — Wall-treatment<br />

Channel flow<br />

<strong>Elliptic</strong>-<strong>blending</strong> k −<strong>ε</strong> <strong>model</strong>:<br />

νt/ν<br />

100<br />

10<br />

1<br />

0.1<br />

0.01<br />

0.001<br />

1 10 100<br />

y +


Results: performance compare to current default<br />

<strong>model</strong>s<br />

(Realizable k −<strong>ε</strong> two-layer and SST k −ω)


Validation — Performance<br />

Channel flow - Comparison with defaults <strong>model</strong>s<br />

Simulation on fine mesh (y + = 1), forced dp/dx, C ∗ <strong>ε</strong>2 =cst:<br />

U/uτ<br />

20<br />

15<br />

10<br />

5<br />

0<br />

DNS<br />

Rke 2-layer<br />

SST k −ω<br />

EB k −<strong>ε</strong><br />

1 10 100<br />

y +<br />

k/u 2 τ<br />

5<br />

4<br />

3<br />

2<br />

1<br />

0<br />

1 10 100<br />

y +


Validation — Performance<br />

Channel flow - Comparison with defaults <strong>model</strong>s<br />

Simulation on fine mesh (y + = 1), forced dp/dx, C ∗ <strong>ε</strong>2 =cst:<br />

<strong>ε</strong>ν/u 4 τ<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

1 10 100<br />

y +<br />

νt/ν<br />

90<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0<br />

1 10 100<br />

y +


Validation — Performance<br />

Backward facing step - Comparison with defaults <strong>model</strong>s<br />

Set-up:<br />

◮ Periodic boundary conditions,<br />

◮ Variable C ∗ <strong>ε</strong>2 ,<br />

◮ Iterative u ∗


Validation — Performance<br />

Backward facing step - Comparison with defaults <strong>model</strong>s<br />

Cf<br />

0.003<br />

0.002<br />

0.001<br />

0<br />

-0.001<br />

-0.002<br />

-0.003<br />

-0.004<br />

expe.<br />

Rke 2-layer<br />

SST k −ω<br />

EB k −<strong>ε</strong><br />

0 5 10<br />

x/H<br />

15 20<br />

Skin friction coefficient


Validation — Performance<br />

Backward facing step - Comparison with defaults <strong>model</strong>s<br />

y/H<br />

2<br />

1<br />

0<br />

-4 -2 0 2 4 6<br />

x/H<br />

8 10 12 14<br />

Streamwise velocity


Validation — Performance<br />

Backward facing step - Comparison with defaults <strong>model</strong>s<br />

y/H<br />

2<br />

1<br />

0<br />

-4 -2 0 2 4 6<br />

x/H<br />

8 10 12 14<br />

Turbulent kinetic energy


Validation — Performance<br />

Periodic 2d hill - Comparison with defaults <strong>model</strong>s<br />

Comparison with case from Fröhlhich et al. Two meshes used:<br />

fine, wall-resolved, mesh, and coarse mesh.<br />

Set-up:<br />

Coarse mesh<br />

◮ Periodic boundary conditions,<br />

◮ Variable C ∗ <strong>ε</strong>2 ,<br />

◮ Iterative u ∗


Validation — Performance<br />

Periodic 2d hill - Comparison with defaults <strong>model</strong>s<br />

LES Realizable k −<strong>ε</strong> two-layer<br />

SST k −ω EB k −<strong>ε</strong>


Validation — Performance<br />

Periodic 2d hill - Comparison with defaults <strong>model</strong>s<br />

Cf<br />

Comparison with Fröhlhich et al. (2001) LES data:<br />

0.01<br />

0<br />

LES<br />

Rke 2-layer<br />

SST k −ω<br />

EB k −<strong>ε</strong><br />

0 1 2 3 4 5 6 7 8 9<br />

x/H<br />

Skin friction coefficient, bottom wall


Validation — Performance<br />

Periodic 2d hill - Comparison with defaults <strong>model</strong>s<br />

y/H<br />

Comparison with Fröhlhich et al. (2001) LES data:<br />

1<br />

0<br />

0 1 2 3 4 5 6 7 8 9<br />

x/H<br />

Streamwise velocity profiles


Validation — Performance<br />

Periodic 2d hill - Comparison with defaults <strong>model</strong>s<br />

y/H<br />

Comparison with Fröhlhich et al. (2001) LES data:<br />

1<br />

0<br />

0 1 2 3 4 5 6 7 8 9<br />

x/H<br />

Turbulent kinetic energy


Validation — Performance<br />

2d diffuser - Comparison with defaults <strong>model</strong>s


Validation — Performance<br />

2d diffuser - Comparison with defaults <strong>model</strong>s<br />

Realizable k −<strong>ε</strong> two-layer<br />

SST k −ω<br />

<strong>Elliptic</strong>-<strong>blending</strong> k −<strong>ε</strong>


Cf<br />

Validation — Performance<br />

2d diffuser - Comparison with defaults <strong>model</strong>s<br />

0.002<br />

0.001<br />

0<br />

-0.001<br />

Exp.<br />

Rke 2-layer<br />

SST k −ω<br />

EB k −<strong>ε</strong><br />

0 10 20 30<br />

x/H<br />

40 50 60<br />

Skin friction coefficient


Validation — Performance<br />

2d diffuser - Comparison with defaults <strong>model</strong>s<br />

y/H<br />

5<br />

4<br />

3<br />

2<br />

1<br />

0<br />

y/H<br />

0<br />

0 5 10 15 20 25 30<br />

x/H<br />

5<br />

4<br />

3<br />

2<br />

1<br />

0 5 10 15 20 25 30<br />

x/H<br />

Streamwise velocity Turbulent kinetic energy


Validation — Performance<br />

3d diffuser - Comparison with defaults <strong>model</strong>s


Validation — Performance<br />

3d diffuser - Comparison with defaults <strong>model</strong>s<br />

Realizable k −<strong>ε</strong> two-layer SST k −ω<br />

<strong>Elliptic</strong>-<strong>blending</strong> k −<strong>ε</strong>


Where <strong>model</strong> should be improved:<br />

1. Rotating and strong curvature<br />

⇒ frame-rotation independent solution,<br />

curvature neglected<br />

2. Buoyancy-driven flows<br />

⇒ laminar solution, poor convergence,<br />

algebraic-heat flux


Part 1: Rotation and strong curvature


Failure — need for improvement<br />

Rotating channel flow<br />

Rotating channel flow DNS study from Kristofferson & Andersson<br />

(1993), for Rossby number varying as 0 < Ro < 0.5:<br />

U/U0<br />

45<br />

40<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0<br />

Ro = 0.1<br />

Ro = 0<br />

DNS<br />

Baseline<br />

CC<br />

Ro = 0.5<br />

-1 -0.5 0<br />

y/H<br />

0.5 1<br />

k/U 2 0<br />

10<br />

5<br />

0<br />

5<br />

0<br />

5<br />

Ro = 0.5<br />

Ro = 0.1<br />

Ro = 0<br />

0<br />

-1 -0.5 0<br />

y/H<br />

0.5 1<br />

Streamwise velocity Turbulent kinetic energy


Failure — need for improvement<br />

NACA 0012 at 10 ◦ incidence<br />

Experimental results from Chow, Zilliac and Bradshaw (1997)<br />

Conclusions from previous studies:<br />

1. T.E. vortex diffuses too quickly with all classical 2-eq. <strong>model</strong>s,<br />

2. lag-<strong>model</strong> and RSM improves on results,<br />

3. CC very beneficial in predicting correct position/intensity of<br />

vortex


Failure — need for improvement<br />

NACA 0012 at 10 ◦ incidence<br />

Axial velocity:<br />

Without CC<br />

With CC


Failure — need for improvement<br />

NACA 0012 at 10 ◦ incidence<br />

Axial velocity:<br />

Without CC<br />

With CC


Failure — need for improvement<br />

NACA 0012 at 10 ◦ incidence<br />

U/Uin<br />

2<br />

1.8<br />

1.6<br />

1.4<br />

1.2<br />

1<br />

Exp.<br />

Baseline<br />

CC<br />

-0.4-0.2 0 0.20.40.60.8 1<br />

x/C<br />

Axial velocity in the core of the<br />

vortex<br />

Comments:<br />

◮ Without Curvature<br />

correction, vortex bursts<br />

too quickly → not good<br />

for F1,<br />

◮ Current curvature<br />

correction probably not<br />

ideal: needs bound to<br />

get current results, and<br />

changes results for some<br />

of the baseline case → it<br />

is a new <strong>model</strong>.


Part 2: Buoyancy-driven flows


Failure — need for improvement<br />

Buoyancy driven flow — Tall cavity<br />

Case from Betts and Bokhari (2001)<br />

◮ Symbols: Exp.<br />

◮ Yellow line: v 2 −f<br />

◮ Red line: EB k −<strong>ε</strong><br />

Comments:<br />

Too large decay of k close to<br />

the wall (laminar-state like)<br />

⇒ over-prediction of velocity<br />

and over-prediction of<br />

temperature gradient<br />

y/H<br />

Turbulent kinetic energy<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 0.01 0.02 0.03<br />

x/H


Failure — need for improvement<br />

Buoyancy driven flow — Tall cavity<br />

Case from Betts and Bokhari (2001)<br />

Temperature Velocity<br />

y/H<br />

308<br />

306<br />

304<br />

302<br />

300<br />

298<br />

296<br />

294<br />

292<br />

290<br />

0 0.01 0.02 0.03<br />

x/H<br />

y/H<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 0.01 0.02 0.03<br />

x/H


Failure — need for improvement<br />

Vertical channel — Kasagi & al<br />

Normalized velocity profiles (local scaling):<br />

Hot aiding wall Cold opposing wall<br />

U/U0<br />

20<br />

18<br />

16<br />

14<br />

12<br />

10<br />

8<br />

6<br />

4<br />

2<br />

0<br />

DNS<br />

Std k −<strong>ε</strong><br />

v2−f<br />

EB k −<strong>ε</strong><br />

1 10<br />

x/H<br />

100<br />

U/U0<br />

20<br />

18<br />

16<br />

14<br />

12<br />

10<br />

8<br />

6<br />

4<br />

2<br />

0<br />

DNS<br />

Std k −<strong>ε</strong><br />

v2−f<br />

EB k −<strong>ε</strong><br />

1 10<br />

x/H<br />

100


Failure — need for improvement<br />

Vertical channel — Kasagi & al<br />

U/U0<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

Velocity Turbulent kinetic energy<br />

DNS<br />

Std k −<strong>ε</strong><br />

v2−f<br />

EB k −<strong>ε</strong><br />

-1 -0.5 0<br />

x/H<br />

0.5 1<br />

U/U0<br />

6<br />

5<br />

4<br />

3<br />

2<br />

1<br />

0<br />

-1 -0.5 0<br />

x/H<br />

0.5 1

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!