Fluidmechanik turbulenter Str omungen Turbulent ows - · PDF filePlane channel ow Wall roughness Conclusion Fluidmechanik turbulenter Str omungen Turbulent ows Markus Uhlmann Institut

Plane channel flowWall roughness

Conclusion

Fluidmechanik turbulenter StromungenTurbulent flows

Markus Uhlmann

Institut fur Hydromechanik

www.ifh.uni-karlsruhe.de/people/uhlmann

1 / 39


Conclusion

Summary of last lecture

Lecture 5 – The scales of turbulent motion

I The turbulent energy cascadeI hierarchy of eddies, downward transfer of energy

I dissipation determined by large scales,performed by small scales

I Kolmogorov’s theoryI building block of turbulence research

I valuable results for small scales (e.g. κ−5/3 spectrum)

I BUT: Problem of non-universality of large scales

2 / 39


Conclusion

LECTURE 6

Wall-bounded shear flows

3 / 39


Conclusion

Questions to be answered in the present lecture

What is the general structure of wall-bounded flows?

How does the presence of a solid boundary affect the turbulentmotion?

What is the effect of wall roughness?

4 / 39


Conclusion

Examples of tubulent wall-bounded flows (1)

Rivers

also: man-made canals, some ocean currents

5 / 39


Conclusion


Vehicle aerodynamics

Citroen DS, ONERA NASA aerodynamic truck

6 / 39


Conclusion


Atmospheric boundary layer:

storm clouds

7 / 39


Conclusion


Internal flows:

pipeline systems

water turbine

also: internal combustion engines, compressors, pumps,blood flow, . . .

8 / 39


Conclusion

Canonical wall-bounded flows

Idealized configurations

I plane channel flow

I pipe flow

I flat-plate boundary layer

CHAPTER 7: WALL FLOWS

Turbulent FlowsStephen B. Pope

Cambridge University Press, 2000

c©Stephen B. Pope 2000

h=2d

b z

y

flow

(a)

L

z

y

(c)

U0

flowy

rR=dD

(b)

x

x

x

Figure 7.1: Sketch of (a) channel flow (b) pipe flow and (c) flat-plate

boundary layer.

1

Pope “Turbulent flows”

9 / 39


Conclusion

Mean flowTurbulence statistics

Plane channel flow configuration

2hy,v

x,uz,w

I statistically stationary & fully developed (far from inlet)

→ homogeneous in x- & z-direction

I definition of “bulk velocity”: ub ≡∫ 2h

0 〈u〉dy/(2h)

→ bulk Reynolds number: Reb ≡ ubh/ν

I Reynolds at centerline: Re0 ≡ U0h/ν with U0 = 〈u〉(y =h)

10 / 39


Conclusion


Mean flow equations

Continuity equation (considering x-z-homogeneity)

��∂x〈u〉+ ∂y 〈v〉+��∂z〈w〉 = 0

⇒ mean wall-normal velocity 〈v〉(y) = 0, since 〈v〉 = 0 at walls

Mean wall-normal momentum equation

1

ρ∂y 〈p〉+ ∂y 〈v ′v ′〉 = 0

⇒ mean axial pressure gradient is constant: ∂x〈p〉 = cst.

11 / 39


Conclusion


Mean flow equations (2)

Mean streamwise momentum equation

1

ρ∂x〈p〉+ ∂y 〈u′v ′〉 = ν∂yy 〈u〉

I define total shear stress: τ ≡ ρν∂y 〈u〉 − ρ〈u′v ′〉

⇒ varies linearly: τ(y) = τw (1− y

h) where τw ≡ τ(y = 0)

I integral balance yields: ∂x〈p〉 = −τw/h

In laminar flow we obtain:

I 〈u〉(y) = τwρν

y2

(2− y

h

), ub = 2

3 U0

I friction factor: cf ≡ τwρu2

b/2= 6

Reb 12 / 39


Conclusion


Mean velocity profile

DNS data of Jimenez et al.

〈u〉U0

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1 Re ↑

laminar

y/h

visc

ous

stre

ss/τ

w

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

Re ↑

laminar

y/h

I mean velocity profiles depend upon Reynolds number(contrary to free shear flows!)

I viscous stress ρν∂y 〈u〉 is always important near the wall

13 / 39


Conclusion


Characteristic scales in the near-wall region

Viscosity and wall shear-stress are important parameters

I define friction velocity and viscous length scale:

uτ ≡√τwρ, δν ≡ ν

√ρ

τw=

ν

uτ

→ friction Reynolds number: Reτ ≡ uτh

ν=

h

δνI wall-distance in viscous lengths (“wall units”):

y + ≡ y

δν=

uτy

ν

→ y + equivalent to a local Reynolds number

14 / 39


Conclusion


Importance of viscous and turbulent effects

Data shows near the wall:

I stress curves collapse whenplotted against y +

I viscous contributiondecreases with y + st

ress

con

trib

uti

on

DNS data of Jimenez et al, Reτ = 180 . . . 2000

0 10 20 30 40 50 60 700

0.2

0.4

0.6

0.8

1

−ρ〈u′v ′〉τ(y)

ρντ(y)∂y 〈u〉

y +

I y + < 5: viscous sublayer – Reynolds stress negligible

I y + < 50: viscous wall-region – direct effect of viscosity

I y + > 50: outer layer – no direct viscous effect

15 / 39


Conclusion


Scaling of the mean velocity profile

Selecting the relevant parameters – dimensional considerations

I channel flow is fully determined by: ρ, ν, h, uτ

I wall-distance y is the only relevant coordinate

→ two non-dimensional numbers (y/h & Reτ or y/h & y +)

⇒ we can write the ansatz:d〈u〉dy

=uτy· Φ(

y +,y

h

)I Φ is a general non-dimensional function

16 / 39


Conclusion


The inner layer – the law of the wall

For y/h� 1 velocity profile is independent of h

I limy/h→0

Φ(

y +,y

h

)= ΦI

(y +)

→ this yields:du+

dy +=

1

y +· ΦI

(y +), with: u+ ≡ 〈u〉/uτ

I integration gives: u+ = fw (y +) “law of the wall”

I experiments show:

fw (y +) is universal function (BL, channel, pipe . . .)

17 / 39


Conclusion


Mean flow behavior derived from the law of the wall

Viscous sublayer (y+ < 5)

I Taylor expansion around y =0:

u+ = y + “linear law”

Logarithmic layer (y+ > 30)

I for large y +, the functionΦI (y +) becomes constant:

u+ =1

κlog(y +) + B “log-law”

I κ ≈ 0.41, B ≈ 5.2

u+

——DNS by Jimenez et al, Reτ = 180 . . . 2000

0 5 10 15 200

5

10

15

y +

u+

100

101

102

103

0

5

10

15

20

y +

• exp. Wei & Willmarth Reτ = 1655

18 / 39


Conclusion


Outer layer – velocity defect law

For y+ > 50: velocity profile independent of viscosity

I limy/h→∞

Φ(

y +,y

h

)= ΦO

(y

h

)→ this yields:

du

dy=

uτy· ΦO

(y

h

)I integration from y to h:

U0 − 〈u〉uτ

= FD

(y

h

)I note: function FD is not

universal!(different in boundary layer)

U0−〈u〉

uτ

——DNS by Jimenez et al., Reτ = 2000

10−1

100

0

2

4

6

8

10

y/h

for comparison: – – – –log-law

19 / 39


Conclusion


Summary of regions in near-wall flow

I inner layer: 〈u〉 independent of U0

and h

I viscous wall region: viscouscontribution of shear stresssignificant (y+ < 50)

I viscous sublayer: Reynolds shearstress negligible

I buffer layer: region between viscoussublayer and log-region

I logarithmic region: the log-law holds

I outer layer: direct effects of viscosityon 〈u〉 negligible

I overlap region: overlap betweeninner and outer layers

y

h

Reb

(from Pope “Turbulent flows”)

20 / 39


Conclusion


The friction law for channel flow

I friction factor:cf ≡ τw

ρu2b/2

= 2u2τ

u2b

I transition occurs aroundReb = 1000

I integrating log-law:

Reτ ≈ 0.166 Re0.88b

cf

– – – – laminar (cf = 6/Reb), • experiments (Dean 1978)

—— turbulent approx. (cf = 0.055Re−0.24b

)

CHAPTER 7: WALL FLOWS

Turbulent FlowsStephen B. Pope

Cambridge University Press, 2000

c©Stephen B. Pope 2000

102 103 104 105 1060.000

0.002

0.004

0.006

0.008

0.010

Re

cf

Figure 7.10: Skin friction coefficient cf ≡ τw/(12ρU 2

0 ) against Reynolds

number (Re = 2Uδ/ν) for channel flow: symbols, experimental

data compiled by Dean (1978); solid line, from Eq. (7.55); dashed

line, laminar friction law cf = 16/(3Re).

9

2Reb

I lengthscale ratio h/δν =Reτ increases almost linearly with Reb

→ very small viscous scales at high Reynolds numbers

21 / 39


Conclusion


Profiles of the Reynolds stress components

Observations from exp. &DNS

I peak TKE in buffer layer

→ signatures of coherentstructures (cf. lecture 7)

I anisotropy bij

I strong changes inviscous wall region

I plateau in log-region

I towards isotropy in coreflow

〈u′i u′j 〉u2τ

DNS by Jimenez et al., Reτ = 2000

0 50 100 150 200−2

0

2

4

6

8

〈u′u′〉

〈w′w′〉

〈v′v′〉〈u′v′〉

y+

bij

0 400 800 1200 1600 2000

0 0.2 0.4 0.6 0.8 1

−0.4

−0.2

0

0.2

0.4

b11

b33

b22b12

y/h

22 / 39


Conclusion


Near-wall asymptotics of velocity fluctuations

Taylor expansion for small wall-distance y :

u′(y) = �a1 + b1 y + c1 y2 + . . .

v ′(y) = �a2 +��b2 y + c2 y2 + . . .

w ′(y) = �a3 + b3 y + c3 y2 + . . . ⇒〈u′u′〉 = 〈b2

1〉 y2 + . . .

〈v ′v ′〉 = 〈c22 〉 y4 + . . .

〈w ′w ′〉 = 〈b23〉 y2 + . . .

〈u′v ′〉 = 〈b1c2〉 y3 + . . .

I no-slip, impermeability: a1 = a2 = a3 = 0

I continuity: ∂y v ′(y = 0) = b2 = 0

I taking mean of products of Taylor series:

⇒ Reynolds-stress components have different near-wall slopesO(y 2), O(y 3), O(y 4)

23 / 39


Conclusion


Near-wall asymptotics of Reynolds-stresses

I 〈u′u′〉 = O(y 2)

I 〈v ′v ′〉 = O(y 4)

I 〈w ′w ′〉 = O(y 2)

I 〈u′v ′〉 = O(y 3)

〈u′i u′j 〉u2τ

DNS by Jimenez et al., Reτ = 950

10−4

10−3

10−2

10−10

10−5

100

〈u′u′〉〈w′w′〉

〈u′v′〉

〈v′v′〉

y/h

I this knowledge can be used for building consistent models

24 / 39


Conclusion


Turbulent kinetic energy bugdet

Transport equation for statistically stationary channel flow

P −ε +νd2k

dy2−1

2

d

dy〈v ′u′j u′j 〉 −1

ρ

d

dy〈v ′p′〉 = 0

I II III IV V

I pseudo-dissipationε ≡ ν〈u′i ,ju′i ,j〉

I peak production occurs inbuffer layer

I buffer layer exports energy

I dissipation not zero at wall,balanced by visc. diffusion

bu

dg

et

DNS by Jimenez et al.,Reτ = 2000

0 10 20 30 40 50 60−0.3

−0.2

−0.1

0

0.1

0.2

0.3I

II

III

IV

V

y+

25 / 39


Conclusion


Length scales in plane channel flow

Observations from DNS by Hoyas & Jimenez (Reτ = 2000)

L(1)αα

h

streamwise integral length scale

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

L11

L22

L33

y/h

L(3)αα

h

spanwise integral length scale

0 0.2 0.4 0.6 0.8 10

0.05

0.1

L33

L22

L11

y/h

I integral scale depends strongly on component and direction

→ turbulence structure highly anisotropic and inhomogeneous

26 / 39


Conclusion


Length scales in plane channel flow (2)

Definition of isotropic length scale

I L ≡ k3/2

εI no direct physical

interpretation

I but: often used inturbulence models

I slope in log-layer:L/h ∼ 2.8y/h

I in outer region:L/h ≈ 0.8

k3/2

ε

1

h

DNS by Jimenez et al.,Reτ = 2000

y+

0 400 800 1200 1600 2000

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

2.8 yh

y/h

27 / 39


Conclusion

Flow over rough walls

I s is characteristic length ofprotrusions

I roughness: s . h/20(else: “obstacles”)

I details of flow in cavities iscomplicated

I consequence: ansatz for meanvelocity modified

d〈u〉dy

=uτy· Φ(

y +,y

h, s+)

s

h

28 / 39


Conclusion

Logarithmic law for rough walls

Inner layer (y/h < 0.1):

limy/h→0

Φ(

y +,y

h, s+)

= ΦR

(y +, s+

)I integration for y + � 1:

u+ =1

κlog(y

s

)+ B(s+)

⇒ κ = 0.41 as for smooth wall

⇒ B depends on s+

I “fully rough” for s+ & 70

B

fully rough: B = 8.5

← smooth

s+

(Nikuradse’s data, from Pope “Turbulent flows”)

29 / 39


Conclusion

Transitional roughness – influence of shape

Roughness function ∆U+

I equivalent form of log-law:

u+ = 1κ log (y +) + B −∆U+

I with:∆U+ = 1

κ log (s+) + B − B

I ∆U+ measures departurefrom smooth-wall behavior

I measurements show:friction can be decreased!

∆U+

(from Jimenez, Ann. Rev. Fluid Mech. 2004)

10 Nov 2003 22:16 AR AR203-FL36-08.tex AR203-FL36-08.sgm LaTeX2e(2002/01/18)P1: IBD

ROUGH-WALL BOUNDARY LAYERS 183

Figure 3 Roughness function for several transitionally rough surfaces, as a functionof the Reynolds number based on the fully rough equivalent sand roughness.e, uni-form sand (Nikuradse 1933);O, uniform packed spheres (Ligrani & Moffat 1986);N,triangular riblets (Bechert et al. 1997);· · · · · ·, galvanized iron; – – – –, tar-coated castiron; — ·—, wrought-iron (Colebrook 1939); ——, Equation 12.

Bradshaw (2000) revived the question, noting that a minimum transitional heightwas unlikely for sparse roughness because the drag of the roughness elements in ashear is proportional tok2 even in the low Reynolds number limit, and this shouldbe reflected in1U+. In recent years the matter has become topical because some ofthe experiments undertaken to clarify the high Reynolds number behavior of flowsover smooth walls have surfaces that would be hydrodynamically smooth or roughdepending on which criterion is used (Barenblatt & Chorin 1998, Perry et al. 2001).Figure 3 shows that there is no “true” answer, and that each surface has to be treatedindividually.

The solid symbols in Figure 3 correspond to triangular riblets measured byBechert et al. (1997). The drag-reducing property of streamwise-aligned ribletsis a transitional roughness effect (Tani 1988). When they exceedk+ ≈ 10 theyloose effectiveness, and their behavior whenk+ À 1 is that of regulark-surfaces.Their drag-reducing mechanism is reasonably well understood. Luchini, Manzo &Pozzi (1991) showed that, in the limitk+ ¿ 1, the effect of the riblets is to imposean offset for the no-slip boundary condition which is further into the flow for the

s+

◦ Nikuradse sand grainsO Ligrani & Moffat spheresN Bechert et al. riblets

30 / 39


Conclusion

Friction law for rough pipes

Considering pipe flow

I friction factor:

cf ≡ τwρu2

b/2=

2u2τ

u2b

I using velocity defect law:

(U0 − ub)/uτ = 3/(2κ)

I using rough-wall log-law:

cf = 2

(1

κlog(R/s) + B(s+)− 3

2κ

)−2

4cf

(Nikuradse’s data, from Pope “Turbulent flows”)

smooth

Reb

31 / 39


Conclusion

Physical insight from DNSFurther reading

Summary

Main questions of the present lecture

I What is the general structure of wall-bounded flows?I inner layer/outer layer: linear/log-law, defect law

I How does the presence of a solid boundary affect theturbulent motion?

I stronger anisotropy than free shear flows

I wall has selective effect on velocity components

I What is the effect of wall roughness?I shift of the log-law compared to smooth walls

32 / 39


Conclusion


Additional effects on wall-bounded flows

Important effects which are not treated here:

I imposed streamwise pressure gradient

I surface curvature

I system rotation

I geometrical complexities (secondary flows,. . .)

33 / 39


Conclusion


Wall turbulence – visualization

Channel flow at Reτ = 590 2hy,v

x,uz,w

I visualizing streamwise velocity fluctuations u′

x-y slice

→ flow direction

z-y slice

⊗ flow direction

34 / 39


Conclusion


Wall turbulence – visualization (2)

Channel flow at Reτ = 590, wall-parallel planes, u′

x-z slice, wall-distance y+ = 45

→ flow direction

x-z slice, wall-distance y+ = 170

→ flow direction

I typical structures: streamwise velocity “streaks”

→ found in all boundary-layer type flows

35 / 39


Conclusion


Facts about velocity streaks in the buffer layer

Statistically speaking:

I lateral spacing of streaks:

∆`+z ≈ 100

I how do we know?

⇒ two-point correlations:minimum of Ruu athalf of the streak spacing

=⇒ flow

Ruu

0 100 200

−0.2

0

0.2

0.4

0.6

0.8

1

r+z

(Moser et al. 1999)

36 / 39


Conclusion


Streamwise vortices

•• Streamwise vortices (ω′x)

`+x ≈ 200

I associated with streaks

(from Jeong et al. 1997)198 J. Jeong, F. Hussain, W. Schoppa and J. Kim

(b)

(a)

(c)

Low-speed streak

Sections in figure 9 (a)–(e)

E

H

SN

G

SP

F

Q4

Q2 B

E

HD

SPU

W

U

W

A

C

θ

θ

x

z

y

x

y

z

(d)

Q2

H

Q4

SP

E

x

z

Top view

y

E H

Q3

Q1

x

x

z

Time

t1

θ1

t2

θ2

t3

θ3

(e)

Figure 10. Conceptual model of an array of CS and their spatial relationship with experimentallyobserved events discussed in the text: (a) top view; (b) side view; (c) structures at cross-section FGin (a); (d) expanded views of structures C and D in (a,b), showing the relative locations of Q1, Q2,Q3, Q4, E and H. A schematic demonstrating the counteracting precession of SN in the (x, z)-planedue to background shear is shown in (e). The arrows in (b) denote the sections of figure 9(a–e).

=⇒ flow

37 / 39


Conclusion


Complex vortex tangles at different Reynolds numbers

Reτ = 180

Reτ = 1900

(from del Alamo et al. 2006)

(movie)

38 / 39


Conclusion


Further reading

I S. Pope, Turbulent flows, 2000→ chapter 7

I H. Schlichting and K. Gersten, Grenzschicht–Theorie, 10thedition, 2006→ part D

39 / 39

Documents

Fluidmechanik turbulenter Str omungen Turbulent ows - · PDF filePlane channel ow Wall roughness Conclusion Fluidmechanik turbulenter Str omungen Turbulent ows Markus Uhlmann Institut