You are on page 1of 41

Advanced Fluid Mechanics

TURBULENT FLOW
Presented by:
Prof. D.Rashtchian

Sharif University of Technology


Chemical & Petroleum Engineering Department

Turbulent transport of momentum

Advanced Fluid Mechanics

Turbulence of random velocity fluctuations- Use statistical methods


Turbulent velocity

u~i

u~i = U i

Mean
Velocity

u i
Fluctuatin g
Component

*************************
Fig.1

Interpret Ui as a time averaged velocity defined by:

1
U i = lim
T T
u i = lim

1
T

1
u~i dt = lim
T T
0

u i dt =U i lim

1
T

(U i + u i )dt

u~i

(U )dt
T

=Ui Ui = 0
Sharif University of Technology
Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

i.e. the mean value (or time average) of the fluctuating quantity is zero. Assume that
Ui the mean flow is steady (Ui/t = 0)
Note: Time averaging commutes w.r.t. differentiation.

u~i
1
=
x j T

u~i

dt =
x j
x j

( )

~
U i
u~i dt =
ui
=
x
x

j
j

The time average of the fluctuation ui is zero, but the average of the square of the
ui2
is used as a convenient measure of the
fluctuation is not zero and the quantity
Ui
turbulent fluctuation-known as the "intensity of turbulence" and ranges from 0.01 to
0.1 for most turbulent flows.

(u ) r.m.s. velocity.
2
i

Mean K.E./unit volume =

KE =

1
1
(U i + u i ) = U i2 + u i2
2
2
mean flow + fluctuatio ns

Sharif University of Technology


Chemical & Petroleum Engineering Department

Equations for the mean flow

Advanced Fluid Mechanics

Consider the momentum and continuity equations. These apply to the instantaneous velocity
in a turbulent field.

2 u~i
u~i
1 ~
p
~
uj
+
=
x j x j
xi
x j

u~i
=0
xi

(1)

The equations must apply on average

u~i = U i + ui
Continuity

u~i
U i
1 T u~i

(
lim
=
=0
dt =
U i + ui ) =
T T 0 x

x
x
x
i
i
i
i

(2)

The mean value satisfies continuity. It is the mean value of velocity that we measure and
require in applications.

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Momentum:
The equations of motion for the mean flow Ui are obtained by taking the time average of all
terms in the resulting equation.
Consider each term:

u~ j
u~i
~~
~
~
(i) u j
(u j u i ) u i
=
x
x j x j
j

= U j + u j U i + u i
x
j

{(

)(

U jU i + u j u i + U i u j + U j u i
x j

U i

+
U iU j + u i u j = U j
(u j u i )
x j
x j x j

)}

(2.1)

1 ~
1
1 P
p
(ii)
( P + p i ) =
=
x i
x i
x i

(2.2)

2U i
2 u~i
2
(U i + ui ) =
(iii)
=
x j x j
x j x j
x j x j

(2.3)

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Hence

U i
2U i
1 P

Uj
+

(u j ui )
=
xi
x j x j x j
x j

(3)

Equation for mean flow has an additional term.(Drop the ^ u j ui = u j ui )

u j u i

U i
;
x j
x j
It represents the mean transport of fluctuating momentum by turbulent velocity
fluctuations.
If ui and u j uncorrelated i.e. u j ui = 0 - no turbulent momentum transfer but

Term

is analogous to the convective term U j

experience shows that u j ui 0 - momentum transfer is a key feature of turbulent


motion.

Term
(u j ui ) thus exchanges momentum between the turbulence and the mean
x j
flow (equation 2.1)even though the mean momentumof the turbulent velocity
fluctuations is zero ( u~i = 0 ).

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Because of the decomposition u~i = U i + u i , turbulent motion can be perceived as something which
produces stresses in the mean flow. For this reason, equation (3) may be rearrange so that all stress
can be put together.

U j

U i

=
x j x j

U j
P + U i +

u
u
(T ji ) - mean stress tensor.(~ = T + )

ji
j i =

x j
xi
x j

T ji = P ji + ji u j ui

U U j
; ji = i +

x
x

i
j

(normal) (shear)
The contribution of the turbulent motion to the mean stress tensor is Tji = u j ui called the
Reynolds stress tensor. Define, ji = ji + Tji

Sharif University of Technology


Chemical & Petroleum Engineering Department

Turbulent shearing stresses

Advanced Fluid Mechanics

Time averaging of the equations of motion leads to the Reynolds stress tensor, u j ui .

u i and u j are the velocity fluctuations in the i j directions at one point and ui u j is a
measure of the "correlation" between the fluctuations.

Correlated variables
u~i u~ j = (U i + ui )(U j + u j ) = U iU j + ui u j
If u i u j 0 , ui and u j are said to be correlated i.e. dependent.
If ui u j = 0 , uncorrelated i.e. ui and u j are independent.

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Fig2(a)

Fig2(b)

Fig2(c)

u 1u 2 > 0

u 1u 2 < 0

R12 = 1

R12 = 1

u 1u 2 0
R12 1

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

A measure of the degree of correlation between u1 and u 2 is obtained from:

u1u 2

{u .u }

( )

2
1

1 T u1u 2
1 T 2

R12 = lim
dt : ui = ui = lim ui dt
T T 0 u u
T T 0

1 2
uu
R12 = 1 2
u1u 2
1 2
2
2
N.B. (a b) 0 (a + b ) ab
2
Hence

2
2

1 T u12 u 22
R12 lim + dt 1
T T 0
u12 u 22

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Pure shear flow


Consider a turbulent shear flow with U1(x2) the only non-zero velocity component.
U
12 is the only component of the mean stress tensor, 12 = 1 u 2 u1
x2
12 stress in 1 direction on face, normal in 2 direction and must result from
molecular transport of momentum in the x2 direction, and turbulent transport.
U 1
Assume
>0.
x2
A fluid particle with positive u 2 is being carried by turbulence in positive x2
direction. It is coming from a region where the mean velocity is smaller i.e. is likely
to be moving downstream more slowly than its new environment. Thus u1 is negative.
Similarly negative u 2 associated with positive u1 .
************************
Fig3

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

{Momentum/unit volume of flow at A in 1-direction} = u~1 = (U 1 + u1 )


The x1-momentum is transported in the x2-direction if u1 and u2 are correlated.
{Flux of x1-momentum in x2-direction} = (U 1 + u1 )u 2
{Average flux of x1-momentun in x2-direction} = u1u 2
T
= u 2 u1
u1 and u 2 are negatively correlated: 12T = 21

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Turbulent Channel Flow


The Navier-Stokes equations in rectangular
coordinates are

2U i
U i
1 P

Uj
(u i u j )
=
+v

x j
x j x j x j
xi
For parallel, fully developed, 2 D flow

U2 = U3 = 0

U i
U i

=0;
= 0
x1
x3

(u i u j ) = 0 ;
x1

L . H .S . = 0

(u i u 3 ) = 0
x 3

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Hence the equations can be written in the simplified form,

2U
1 P
+ v 2 (uv)
0=
y
y
x

0=

1 P 2
(v )
y y

At the walls v 2 = 0, P = P0(x)


P0

+ v2

P P0 dP0
=
=
dx
x
x

(1)

(2)
. Hence form (2)
(3)

(4)

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Hence (1) can be integrated from y=0 to y with uv

0=

y =0

=0

y dP0
U
U
(
)+v
v
dx
y
y

uv
y =0

At y=h, uv=0, U / y = 0 (zero velocity gradient, no correlation)

w
h dP
U
= ( 0)=

dx
y

y =0

Defining a friction velocity u*

w = u2
Substituting in (5)

uv + v

U
y
= u *2 (1 )
y
h
Sharif University of Technology
Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Equation (8) may be written in dimensionless form in 2 ways.

(I)

uv v d (U / u* )
y
+
= (1 )
2
u*
u*h d ( y / h)
h

R* =u* h/v. As R* becomes large, (R* is of course a Reynolds number), the


viscous stress is suppressed. Such a limit will not applied because viscous
forces must always dominate near solid boundaries.

(II)

uv d (U / u* )
yu* v
.
2+
= 1
u* d ( yu* / v)
v hu*

In this case as R* becomes large the change in total stress becomes small.
Defining appropriate dimensionless variables

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

yu *
y+ =
;
v

u+ =

U
u

y
h

Then

uv 1 du +
+ *
= 1
2
u* R d

(11)

uv du +
y+
+
=

1
u*2 dy +
R*

(12)

Law of wall
For large R* (from 12)

uv du +
+
=1
2
u* dy +

(13)

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

The solution of this equation must be of the form,

uv
= g ( y +) ; u + = f ( y +)
2
u*

(law of the wall)

(14)

For sufficiently small y+, turbulent stress negligible.

du +
=1
dy +

with

u+(0)=0

(15)

u+ = y+
Core region
For large R* (from 11)

uv
= (1 )
2
u
Sharif University of Technology
Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

This equation gives no information, about U itself. However h and u* are the only feasible
length and velocity scales, we can write

dU u * dF
=
dy
h d

Where F( ) is some function of .

(17)

Integration from the center where U=U0

U U0
= F ( )
u*

(18)

From equation (14),

U
= f ( y +) ;
u*

dU u *2 df ( y +)
=
dy
v dy +

(19)

Matching (17) & (19),

u dF u2 df
=
.
;
h d dy +

dF
df
1
= y+ + =
d
dy
K

(20)

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

F ( ) =

1
ln + const.
K

f ( y +) =

1
ln y + + const.
K

Hence
U U0 1
= ln + const.
u*
K

U
1
= ln y + + const.
u* K

Discussion
To simplify (12) to (14) requires
To simplify (11) to (16) requires
Matching only possible if

In practice it is found that

y+
= 1
R*
1 du +
1
R * d

(a)
(b)

y+
0
y + > 100

< 0.1

are sufficient

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Now < 0.1 ;


And y + > 100

y+
< 0.1
R*
R* > 100
R* > 100 /

(cf.(a))

< 0.1
R* > 1000
Experimentally

Hence

du + 2.5
=
d

1 du +
2.5
=

R * d
R *

1 du +
2.5
=
<< (1 )
R * d
R *

Sharif University of Technology


Chemical & Petroleum Engineering Department

( cf . (b))

Advanced Fluid Mechanics

df
1
=
dy + K
1
f ( y + ) = ln( y + ) + const.
K

y+

Also from (20)

(21)

Experimentally

U U0
= 2.5 ln 1.0
u*
U
= 2.5 ln y + + 5.0
u*

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Application
1. For Engineering Purposes these
equations have been used for > 0.1 ,
i.e. to describe the core region, and
also for 0 . Note as
0 , u + = U / u*
2. Sometimes the Universal Velocity
profile is used.
Equn. (15) u+=y+ for y+ 5
Equn. (21) u+=2.5lny++5.0for y+ 30.
Limits determined experimentally.
A curve fit for 5<y+<30
Is u+=5.0 ln y+ - 3.05

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Example of use of turbulent velocity profiles.


momentum transfer
Friction factor

f =

1
U
2

2u *2
= 2
U

Using the velocity defect law for flow in a tube

1
h 2

y =h

U U 0
U U0
=
.
2

rdy
y =0 u*
u*
=1

Now

2r
h{2.5 ln 1}d

2
= 0 h

r = h - y ; = y / h dy = hd

r = h(1 )
Sharif University of Technology
Chemical & Petroleum Engineering Department

=1

Hence

(U U 0 )
= 2(1 ){2.5 ln 1}d
u*
=0

Advanced Fluid Mechanics

Also from experimental results

U0 U
=
2.5 ln + 1.0
u* u*

= 2.5 ln y + 2.5 ln + 1.0 + 5.0


= 2.5 ln R * +6.0
hu * h
R* =
=
v
v

fu 2 Re
=
2
2

f
2

U0
Re f
= 2.5 ln[
]+6
u*
2 2
U
Re f
5 2
5 2 2 1
= 2.5 ln[
] + 6 + [5 ln 5 2
ln +
+
]0
u*
4
2
2 2
2

Re
= 4.07 log 10
f
2

f
+ 0.53
2
Sharif University of Technology
Chemical & Petroleum Engineering Department

Mass transfer: Turbulent Taylor Analysis. Proc. Royal. Soc. (1954), A223, P446, for

Advanced Fluid Mechanics

Axial Dispersion in turbulent pipe flow.


Consider diffusion equation in rectangular coordinates for simplicity.

Sharif University of Technology


Chemical & Petroleum Engineering Department

Turbulence in pipe flows

Advanced Fluid Mechanics

Scope of Turbulence
Most flows in nature: rivers, the atmosphere
Engineer: pipe flow, packed and plate column
Pipe Flow

Laminar sublayer - viscous forces dominate, very thin


Transition region - region of damped turbulence because of nearby wall,
eddy size y.
Turbulent core
- region of fully developed turbulence, eddies of size d,
velocity nearly constant.
Sharif University of Technology
Chemical & Petroleum Engineering Department

Turbulent Velocities

Advanced Fluid Mechanics

local downstream velocity fluctuates due to turbulent eddies .decompose

u t = u + u
Instantane us localmean

=
velocity

velocity

definition of u (mean velocity)


u=

eddy

+
velocity

1
T

u t dt

clearly the average of the eddy velocity is zero

1
u=
T

(
u
+
u
)
dt
=
0
T
1
T

1
udt
+
0
T

udt
0

udt = 0
0

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

the magnitude of turbulent velocities is characterized by the RMS


1

1 2 2
u = u dt
T 0

(RMS fluctuating or eddy velocity.)


T

the turbulence intensity is defined by,


u
turbulent intensity = (typically up to 0.1) i.e. the average eddy velocity
u
may be 1/10 of the mean velocity.

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Properties of turbulent flows


(with particular reference to pipe flows)
1) Irregularity local velocities fluctuate in random manner. But all
turbulent flows are irregular. E.g. smoke plume.
2) 3D Nature pipe flows are normally considered as 1 dimension in
that downstream velocity depends only on radius. However in
turbulent flows normal velocity components, though zero on average,
have fluctuating components, ( V and W ). These give rise to
turbulent stresses (remember the mail bag example) and are important
in turbulent energy processes. This 3D nature adds the mathematical
difficulty.

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

3) Turbulence is a property of the flow not of the fluid writing Newtons


law for a flow involving turbulent stresses.

du
= ( + T )
; [divided by ]
dy

2
[
L]

Where = is the kinematics viscosity .

[T ]
T = eddy viscosity
In laminar sub layer T <<
T ~
Transition region
Turbulent core
T >>
Thus T varies with environment and is a flow property.

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

4) Mixing in turbulent flows-diffusivity


Rewrite Newtons law in the form, explicit in shear stress.

= ( + T )

d ( u )
dy

[M ][U ] 3
[M ][U ] = [L] .
[L]
[L]
[L]2 [T ] T
2

Dimensions :

i.e. (momentum flux) = (diffusivity) * (gradient of momentum / volume)


- this fundamental relation shows how transport (here of momentum) is related
to the driving force (momentum gradient).the coefficient, , is the momentum
diffusivity. It shows how large a flux is produced by a given gradient. Exactly
analogous laws apply for heat transfer (Fourier's Law) and mass transfer
(Flick's law).

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

- Now T ff in turbulent flows.


Turbulent is a very effective mixer of momentum which accounts for
the almost constant velocity of the core region will usually be of almost
constant temperature and composition.
But ff T in laminar sublayer.

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

In laminar flow it is the molecular motion which transports momentum.


(Remember mail bag example). Hence lower rates of transport for a given
driving force. Alternatively if we consider heat transfer from the wall to
bulk, heat conduction across the laminar sub layer dominates the process
(Heat transfer comes later).
- Pictorially
Eddy gives rise to normal velocity V .
This transports x directional momentum
in the y direction gives rise to a
momentum flux, or shear, .

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

5) Dissipative Nature of Turbulence.


- Turbulence comprises eddies of all sizes.
- The largest eddies are as big as the flow field. They extract energy from the
flow but are not efficient at dissipating energy. In the absence of an energy
source, however, turbulence dies away .
- There is an energy cascade from the large eddies, through eddies of
progressively smaller size until a lower limit is reached. This lower limit is
controlled by viscous dissipation of energy and Kinematics viscousity and the
rate of energy supply are the important quantities. Based on dimensional analysis
this lower limit of eddy size is given by:


=
d u d

Where = size of small eddies;


d = size of largest eddies;

= Re 3 4

= kinematics viscosity
u' = RMS turbulent velocity

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

dissipatio n rate

(i )

s 6
=
v d

(ii )

surface
volume

s
6d 2
12
= 8

d3 =
4
d
v

s

v

(iii )

s
6d 2
= 64
16
v

d3 =

Sharif University of Technology


Chemical & Petroleum Engineering Department

24

Advanced Fluid Mechanics

High Reynolds number phenomenon


-

Express Newton's Law of viscosity in dimensionless form.

(u )

+
d

d
du
du
1
1
u =
+ =
= 2
=
2
Re d y
Re dy +
u
u d dy

Reynolds's number arises in dimensionless form of Newton's Law.


- Similarity: compare two flows in similar geometries(same shape but
different size)i.e. flows exhibiting geometrical similarity. Suppose
Reynolds numbers of each flow are the same though d,u, and of
each flow may be individually different. Then as a consequence of the
above equation each flow will have the same dimensionless distribution
of stress and velocity gradient as a function of position,

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Stress and velocity gradient as a function of position, provided each has the same
Reynolds number;
i.e.
u+ = f(y+)
kinetic similarity
+ = g(y+)
dynamic similarity.
The consequence is that friction factor (dimensionless wall shear stress) can be
considered a unique function of Re.
Consider a cylindrical element of diameter d and length of x

du
. d x
dr
du
d 2
e
x
4
dt

Viscous forces
Interia forces

d dr
InteriaFor ces

Re
dt
ViscousFor ces
High Re-interia forces dominate Turbulent flow
Low Re-viscous forces dominate Laminar flow

Sharif University of Technology


Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Summery Notes on Turbulence

Most flows are turbulent both in nature and engineering.


A turbulent pipe flow can be divided into three regions:
a) Laminar sublayer - no eddies.
b) Transition region damped eddies (size y)
c) Turbulent core undamped eddies (size d)

Turbulent velocities:
ut = u + u

{Instantaneous = local mean + fluctuant}

1
u = ut dt
T 0

{T is a time long enough to include many eddies}

1 2 12
u = [ u dt ]
T 0
u v w
u / u 0.1

{RMS velocity characterizes turbulence}


{Turbulence is homogenous}
{Turbulence intensity}

Newton's Law in turbulent flows. It is tempting to write

/ = ( + t )

du
dy

Kinematic Vis cosity for Molecular

t Eddy Vis cosity for Turbulence


Sharif University of Technology
Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

is a fluid property and constant.


t is a flow property and depends on environment (eddy size)
Rewrite above equation as

= (v + vt )

d ( u )
dy

dimensions

[M ] [L]
[ L ] 2 [T ] 2

(momentum flux)=(momentum diffusivity)(gradient of mom/vol)


Large implies rapid mixing. Diffusivity has dim. [L]2/[T]
T >> : Turbulent flows are rapidly mixed due to eddies.

Energy in turbulent flows: turbulent dissipates considerable energy.


Large eddies take energy from mean floe, but are not efficient in
dispersing energy. Small eddies do dissipate energy efficiently. There
is a transfer of energy to the small eddies, which appears as heat due to
frictional effects.
Smallest eddy size, , si given by ( dimensional analysis)
3

u d 4
)
{ is also a good estimate of laminar sub-layer thickness}
=(
d
v
Sharif University of Technology
Chemical & Petroleum Engineering Department

Advanced Fluid Mechanics

Rynolds Number arises in Newton's Law in dimensionless form

1 du +

du
d (u / u )
=
=
=
=
2
2
2
Re dy +
u
u dy
u d d ( y / d )
+

It may be interpreted as the ratio (interia forces / viscous forces).


Large Re implies dominance of interia forces which promote turbulence.
Small Re will dominance of friction (viscous) forces gives laminar flows.
Similarity (Consider different flows of same Reynolds Number) If we have
geometric similarity (e.g. two different pipe flows) then we will have
kinematic similarity (same du+/dy+) and dynamic similarity (same +).

Result

f = w+ = f(Re)

Sharif University of Technology


Chemical & Petroleum Engineering Department

You might also like