Perspective on Turbulence Modeling using Reynolds...

Perspective on Turbulence Modeling using Reynolds Stress Models : Modification for pressure gradients

Tobias Knopp, Bernhard Eisfeld

DLR Institute of Aerodynamics and Flow Technology

Experimental data shown were obtained in cooperation withDaniel Schanz, Matteo Novara, Erich Schülein, Andreas Schröder DLR AS

Nico Reuther, Nicolas Buchmann, Rainer Hain, Christian Cierpka, Christian KählerUniv Bundeswehr München, Institute for Fluid Mechanics and Aerodynamics

IMPULSE progress meeting

Folie 2

Slide 2

Motivation: Digital aerospace products

Numerical simulation based future aircraft design

Challenges§A/δ99 extremely large: large Re and large A§#simulations large for design and optimization

=> RANS, not LES

Folie 3

Slide 3

Optimism to describe “first order” steady state aerodynamic flow phenomena within the RANS concept

Separation in the wing-body junction at APG

Interaction of wake flow and boundary layer at APG

Interaction of strake vortex and boundary layer at APG

DLR strategy for RANS model improvement

Turbulent boundary layers at APG Separation and reattachment Wake flow at APG

[28] Hoffenberg & Sullivan, 1998

[21] Liu et al., 2002

[19] Roos, 1997

[22] Tummers et al., 2007

[20] Driver & Mateer, 2002

Folie 4

Slide 4

High-quality data base (exp., DNS/LES)

Folie 5

Slide 5

(Empirical) laws of turbulence

Folie 6

Slide 6

Physics-based improvement of RANS

Folie 7

Slide 7

Physics based improvement of RANS

Experiments by Nikuradze

Surface roughness

Folie 8

Slide 8

kr+ = kr uτ / νΔu+=f(kr

Surface roughness

Folie 9

Slide 9

Physics-based improvement of RANS νt = κuτ (y + 0.03kr)

Surface roughness

Folie 10

Slide 10

Surface roughness

Folie 11

Slide 11

Turbulent boundary layer at APG

Folie 12

Slide 12

Relevant parameter space?

Turbulent boundary layer at APG

Folie 13

Slide 13

Reynolds number

Pressure gradientparameter

Data base set-up: The parameter space

Folie 14

Slide 14

Reynolds number

Pressure gradientparameter

A320 flap

Folie 15

Slide 15

A380 flap

A320 flap

Folie 16

Slide 16

A380 flap A380 wingA350 wing

Folie 17

Slide 17

DLR institute AS decision 2009:New data needed!Only experiments can give large Re!

Folie 18

Slide 18

Exp I.Assess PIV at moderate Re

Folie 19

Slide 19

Experiment #1 (2011): RETTINA I exp. (DLR + UniBw Munich)

Large scale overview 2D2C PIV

LR μPTV

2D3C PIV

U∞= 6 … 12m/s Reθ up to 10000 and δ99 =0.1mThick boundary layers enable PIV

Folie 20

Slide 20

Exp I.Assess PIVat moderate Re

Folie 21

Slide 21

Exp II.Higher Re

STELAR 2. Projekttreffen > DLR Göttingen > 23.04.2012

Experiment #2 (2015): RETTINA II exp. (DLR + UniBw Munich)

Flow relaxation

ZPGLog-law

APG region

Defined inflow condition. Defined

outflow condition

U∞=10m/s, 23m/s, 36m/s

Funded by DLR institute AS and by DFG

ZPGLog-law

APG region

U∞=36m/s

Reθ=41000

Large-scale 2D2C-PIV 9 cams2D3C PIVmicro 2D2C PTV3D3C PTV STB (shake the box)

Challenge: Large range of scale of turbulent boundary layer

Solution approach: Multi-resolution multi-camera PIV

U=36m/sReθ=41000

ν/uτ = 20μmδ99 = 0.2m

U=36m/sReθ=41000

ν/uτ = 20μmδ99 = 0.2m

U=36m/sReθ=41000

ν/uτ = 20μmδ99 = 0.2m

U=36m/sReθ=41000

ν/uτ = 20μmδ99 = 0.2m STB=„shake-the-box“

Particle-tracking-approachSchanz, Schröder, Novara@ DLR-AS

Measurement technique. Skin friction Clauser chart (y+-range depends on ZPG, FPG, APG)

μPTV: Viscous sublayer mean velocityOil film interferometry

Folie 30

Slide 30

Mild separation and reattachment

Experiment #3 (August 2017) within DLR internal project

Folie 31

Slide 31

Folie 32

Slide 32

Empirical wall law at APG

Mean velocity profiles2D2C Reynolds stresses

Folie 33

Slide 33

U=36m/sΔpx

+=0.015

Aim: Empirical wall-law at APG for y<0.1δ99

y<10%δ99

Folie 34

Slide 34

U=36m/sΔpx

+=0.015

Aim: Empirical wall-law at APG for y<0.1δ99

y<10%δ99

Folie 35

Slide 35

U=36m/sΔpx

+=0.015

Hypothesis #1: Resilience of a small log-region

y<0.1δ99

Folie 36

Slide 36

U=36m/sΔpx

+=0.015

Hypothesis #1: Resilience of a small log-region

y<0.1δ99

Galbraith et al. (1977),Granville (1985), Bradshaw (1995), Perry & Schofield (1973), Durbin & Belcher (1992)Skare & Krogstad (1994)Spalart & Coleman …

Folie 37

Slide 37

Results and analysis

y+Ξ-1 =

Folie 38

Slide 38

U=36m/sΔpx

+=0.015

y<0.1δ99

Folie 39

Slide 39

U=36m/sΔpx

+=0.015

Hypothesis #2: square-root (sqrt)-law above the log-region

y<0.1δ99

Szablewski (1954), Townsend (1961), McDonald (1969),Brown & Joubert…

Folie 40

Slide 40

U=36m/sReθ=41000

Δpx+=0.015

Hypothesis #2: square-root (sqrt)-law above the log-region

y<0.1δ99

Folie 41

Slide 41

y+log,max

Composite profile Brown & Joubert 1969

Folie 42

Slide 42

Hypothesis #3: transition from log-law to sqrt-law depends on Δpx+

(probably more complex …)

y+log,max

Folie 43

Slide 43

y+log,max

Folie 44

Slide 44

y+log,max

Folie 45

Slide 45

ZPG Δpx

Pure log-law

Folie 46

Slide 46

Separation Δpx

+→ ∞

Pure sqrt-law(Stratford)

• SlopecoefficientK

„vonKarmanconstant“κ=0.40+/-0.1

Hypothesis #4: decrease of log-law slope parameter Ki at APG

Increasing Δpx+

Assumptions:§ Blended log-law + sqrt-law§ linear shear stress

Blending for eddy visc. νt

Idea: A composite wall-law for the turbulent viscosity

log-law at APG

sqrt-law at APG

(Galbraith et al. 1977)

Δpx+ = 0.012

Exp. by Skare & Krogstad

RANS modification

HGR01 airfoil at Rec=25Mio, incidence angle α=10°

RANS modification

HGR01 airfoil at Rec=25Mio, incidence angle α=10°

Idea #1: New terms with Δpx+-dependent local blending

Modifications should be activated only in parts of the boundary layer

Blending for y<0.15δ99 Activate only in the sqrt-region

Pressure diffusion at APG (Rao & Hassan)

Idea#2:RANSmodelcoefficientssensitizedtopressuregradientΔpx+

Standard calibration is for ZPG κ=0.41 = const

Coefficient γ controls the slope of the log-law

Idea#2:RANSmodelcoefficientssensitizedtopressuregradientΔpx+

• TheRANSmodelcoefficientwhichcontrolsthelog-lawslope(“Karman-constant”)becomesafunctiondependingonlocalflow

Data structure of wall-normal lines for Δpx+

Surface pointΔpx

+ from dp/dx and uτ

δ99, δ*, θ, H12

Field point

Extension of unstructured flow solver TAUWall-normal linesMethod to determine δ99, δ*, θ, H12

Coding efforts vs. Improvements in predictive accuracy

RANSmodelsensitizedtopressuregradientΔpx+

Surfacequantities Δpx+fromdp/dxanduτ

Maptocorrespondingfieldpoints

FielddistributionofΔpx+

RANSmodelsensitizedtopressuregradientΔpx+

RANSmodelcoefficientγ becomesafunctionofΔpx+

Validation of modified RANS

U=36m/s

Validation of modified RANS

U=36m/s

Validation of RANS

U=36m/s

Questions to the audience

Turbulent Boundary Layer at ZPG

-www.DLR.de • Folie 71

Turbulent Boundary Layer at ZPG

-www.DLR.de • Folie 72

Δcf=5%!

Conclusions

- DLR strategy: Physics based improvement of RANS using new laws of turbulence- First step: Wall law at APG- Composite wall law at APG (Brown & Joubert)

- Small log-region around y+~100- Sqrt-law region above log-region

- Machine learning for further tuning of first theoretical ideas

- Optimism to improve RANS models for aerodynamic flows during next decades- Great potential for future research- DNS/LES at relevant Re possible

- New smart DNS flows (e.g., work of Spalart & Coleman, Soria & Jimenez)- Improved measurement techniques (e.g. advances in PIV/PTV)

End of the presentation§ Experimental data shown in cooperation with

§ Daniel Schanz, Matteo Novara, Erich Schülein, Andreas Schröder DLR AS

§ Nico Reuther, Nicolas Buchmann, Rainer Hain, Christian Cierpka, Christian Kähler, UniBw München

§ Funding of measurement campaign by DFG within „Investigation ofturbulent boundary layers with pressure gradient at high Reynolds numbers with high resolution multi-camera techniques“

Validation RETTINA II

U=36m/s

U=23m/s

Folie 79

Slide 79

Measuring technique

Folie 80

Slide 80

Modification of the k-omega model

Modified consistent model

Modification of the k-equation:Pressure diffusion term (Rao & Hassan)

Modification of the ω-equation:Pressure diffusion term

Negative cross-diffusion term

Folie 81

Slide 81

Consistency with wall-law

Summary of boundary layer analysis

Consistency with log-law at APG

Consistency with sqrt-law at APG

k-equation No No

ω-equation Yes No

Folie 82

Slide 82

Consistency with log-law at APG

Consistency with sqrt-law at APG

Modified k-equation Yes Yes

Modified ω-equation Yes Yes

Folie 83

Slide 83

Blending active in log-law region at APG

Blending active in sqrt-law region at APG

Modified k-equation Yes Yes

Modified ω-equation No Yes

Folie 84

Slide 84

The ω-equation in the sqrt-law region

Following ideas by Rao & Hassan, Catris & Aupoix

Folie 85

Slide 85

Following ideas by Rao & Hassan, Catris & Aupoix

Folie 86

Slide 86

Folie 87

Slide 87

Conclusion: The ω-equation is not consistent with the assumed solution in the sqrt-law region at APG

Folie 88

Slide 88

Measurement technique combining different PIV systems

1.5m0.78m

0 x [m]7.88 8.66 10.166.38 11.66

2D2C-PIV§ Large scale PIV§ 8 PCO 4000 cams side-by-side 2D3C stereo PIV overview

2D3C stereo PIV detail

Folie 89

Slide 89

Requirements/Design criteria on a new flow experiment

1. Large Reθ for sufficiently thick overlap region2. Log-law established before entering the APG-section3. Thick BL at low flow velocity so that PIV in viscous sublayer possible4. Obtain large values for Δpx

+ , i.e. Δpx+>0.06

U∞= 9m/s and U∞= 12m/s

At x=6mReθ up to 8000 and δ99 =0.1m

Design of the test case using RANS-CFD with the DLR TAU code by DLR

Folie 90

Slide 90

Data for mean velocity down to y+=1 using LR-PIV with PTV

1.5m6.38m

1.5m0.78m

x [m]7.88 8.66 10.166.38 11.66

2D2C PIV Long-range microscopic PIV with PTV

Particle-tracking velocimetry algorithm (PTV) applied to the LR-PIV data see also Kähler et al., Exp. Fluids (2012)

Folie 91

Slide 91

Boundary layer theory.An approximative model for the

shear stress

Folie 92

Slide 92

Turbulent boundary layer equation

Consider the equation for wall-parallel mean velocity component U

Integration in wall-normal direction from y‘=0 to wall-distance y

Or in viscous inner units

Folie 93

Slide 93

Review: Modeling the total shear stress (van den Berg 1973)

Aim: Relate the total shear stress and the mean velocity profile for U

by approximating the integrated convective (or: mean inertial) terms

Van den Berg makes the modeling assumption that

Folie 94

Slide 94

Modeling assumption

For the V-velocity and using the continuity equation

This gives for the mean inertial term

Folie 95

Slide 95

This gives the model by van den Berg (1973)

Short notation:

Pressure gradient parameter

Wall shear-stress gradient parameter

Folie 96

Slide 96

Extended modeling for the total shear stress

Modeling assumption

For the chain rule of differentiation we need

Then we obtain, e.g.

With an additional parameter taking into account d²P/dx²

Folie 97

Slide 97

Extended modeling for the total shear stress

This gives the extended model

With integrals

and with parameters

Pressure gradient parameter

Wall shear-stress gradient parameter

Cross parameter

d²P/dx² -parameter

Folie 98

Slide 98

Characteristic non-dimensional parametersComparison of the characteristic non-dimensional parameters

Folie 99

Slide 99

Turbulent shear stress in non-equilibrium flow

Mean inertial terms cause a significant reduction of the turbulent shear stress τ+=1+Δpx+y+

Present exp.at x=8.18mΔpx+=0.0473

Folie 100

Slide 100

Systematic trend/quantitative formula for variation of Ko

Folie 101

Slide 101

Conclusion

Folie 102

Slide 102

Conclusion.

Design of a flow experiment suitable to study the law-of-the-wall for flows at a substantial adverse pressure gradient

Sufficiently large overlap-layer due to large Reθ

Significant adverse pressure gradient Δpx+ up to Δpx

+ =0.065

Combination of different PIV techniques gives high-quality data setLarge number of velocity profilesHigh quality gradients and slope diagnostic functionsComparison of direct and indirect method for uτ locally

Folie 103

Slide 103

Conclusion.

The log-law does no longer describe the overlap layer 300 < y+ < 0.2 δ99+

A generalized half-power law (or modified log-law) gives a good description

Idea of a composite wall-law by Brown & Joubert (1969) is supportedLog-law-fit region is thin 60 < y+ < 130

1/slope Ki decreasing with increasing Δpx+ as devised by Nickels (2004)

Modified log-law region for 300-400 < y+ < 0.2 -0.4 δ99+

1/slope Ko decreasing with increasing Δpx+

But mean inertial terms need to be accounted for Ko

Folie 104

Slide 104

Characterization of the flow

ZPGlog-law

Folie 105

Slide 105

Favourable dp/dsLFPG ~ 5 δref

Folie 106

Slide 106

Focus region:

Adverse dp/dsLAPG ~ 5 δref

Folie 107

Slide 107

Universal log-law in zero-pressure gradient region

Universal log-law at the inlet of the adverse pressure gradient section

At x=6mReθ=8000

U∞=12m/s

Folie 108

Slide 108

Systematic trend/quantitative formula for variation of Bi

Folie 109

Slide 109

Systematic trend/quantitative formula for variation of Bi

Folie 110

Slide 110

Measurement technique using PIV

Quanta Ray 400

Innolas Spitlight 1000

8 PCO 4000

Superelliptical nose

Flat plate of length 6.38m

APG region

Folie 111

Slide 111

Large-scale 2D2C PIV. Instantaneous flow field

Cam 5Cam 6

Cam 7Cam 8

Folie 112

Slide 112

Motivation: Digital aerospace products

Numerical simulation based future aircraft design

Challenges§A/δ99 extremely large: large Re and large A§#simulations large for design and optimization

=> RANS, not LES

Folie 113

Slide 113

Motivation: Turbulent boundary layers at adverse pressure gradient

During take-off and landingSignificant adverse pressure gradient (APG)

Goals:Þ Wall-laws at APGÞ Improvement RANS models

Perspective on Turbulence Modeling using Reynolds...

Documents

Reynolds-number dependency in homogeneous, stationary two-dimensional turbulence · 2009. 11. 24. · Under consideration for publication in J. Fluid Mech. 1 Reynolds-number dependency

A k LOW REYNOLDS NUMBER TURBULENCE MODEL FOR …fpinho/pdfs/CI1-AERC2010_Turb_k_omega_v2.pdf · A k-! LOW REYNOLDS NUMBER TURBULENCE MODEL FOR TURBULENT CHANNEL FLOW OF FENE-P FLUIDS

High Reynolds Number Wall Turbulence · High Reynolds Number Wall Turbulence: ... •Understand behavior of mean properties as the Reynolds number becomes ... experiment, Manchester,

Outline I - thomas-gomez.netthomas-gomez.net/images/Lectures/Turbulence/turbulence_lectures... · Reynolds decomposition Mean NS equations Reynolds stress tensor Kinetic energy Fluctuating

The cockpit perspective on turbulence

Two-Equation Low-Reynolds-Number Turbulence · PDF fileTurbulence Modeling of Transitional Boundary Layer Flows Characteristic ... and Suhas V. Patankar ... Number Turbulence Modeling

Turbulence: Challenges for theory and experiment.mixing.coas.oregonstate.edu/people/jmoum/courseinfo/Frisch&Orszag90.pdf · TURBULENCE: CHALLENGES FOR THEORY AND EXPERIMENT High-Reynolds-number

Atmospheric Turbulence - lib.iszf.irk.rulib.iszf.irk.ru/Atmospheric Turbulence. A Molecular... · Atmospheric Turbulence: A Molecular Dynamics Perspective Adrian F. Tuck 1. 3 Great

Turbulence Reynolds Averaged Numerical Simulations …gtryggva/CFD-Course2010/2010-Lecture-18.pdf · A wire hoop ahead of the equator ... Reynolds Averaged Numerical Simulations!

Turbulence Characteristics and Dissipation Estimates … · Turbulence Characteristics and Dissipation Estimates in the Coastal ... Estimates of Reynolds stresses ... The platform

Turbulence - Kapteyn Astronomical Instituteweygaert/tim1publication/astro... · 2015-04-07 · 4/7/2015 1 Turbulence Instability 1) All flows become unstable above a certain Reynolds

NASA€¦ · The Langley Low- Turbulence Pressure Tunnel is a uniquefacility that provides flight Reynolds number tests capability for two-dimensional airfoils and a low turbulence

Turbulence Models and Reynolds Analogy for Two'Dimensional

Introduction to Turbulence Modeling - nuaa.edu.cnaircraftdesign.nuaa.edu.cn/aca/Slide/28-Turbulence Modeling.pdf · Reynolds equations, ... – is to introduce the methodology commonly

Turbulence statistics in fully developed channel flow at ......Turbulence in channel flow at low Reynolds number ! Flow 135 FIGURE 1. Coordinate system in channel. 2. Computational

Validation of a Low-Cost Transitional Turbulence Model for Low-Reynolds-Number External Aerodynamics

Measurement of High Reynolds Number Turbulence in the ... · 10th International Symposium on Turbulence and Shear Flow Phenomena (TSFP10), Chicago, USA, July, 2017 Measurement of

Turbulence Model Behavior in Low Reynolds Number Regions of Aerodynamic Flowfields

TSF experiment for comparison of high Reynolds number turbulence in both HeI … · 2017-01-25 · TSF experiment for comparison of high Reynolds number turbulence in both HeI and

University of Michigan · Web viewCase3 Clustered Fine Mesh With Laminar Flow The Reynolds-averaged approach to turbulence modeling requires that the Reynolds stresses in Equation