Вы находитесь на странице: 1из 18

Module 6

ISSUES IN PRACTICAL PROBLEMS, Part B:


Turbulent Flows

Dr. Sreenivas Jayanti


Department of Chemical Engineering
IIT Madras
Turbulent Flows
Turbulence is a three-dimensional unsteady viscous flow phenomenon that occurs at high
Reynolds numbers
For pipe flow, Recrit 2000
For boundary layer flow, Recrit 500 000
For flow over a sphere, Recrit 200 000
For stirred tank reactors agitated by impellers, Recrit 1000 to 20 000
For flow through a porous medium, Recrit 10 to 1000
For flow through mini or microchannels, compact heat exchangers, Recrit 500 to 1000

Turbulent flow is characterized by rapid and highly localized fluctuations in flow parameters
such as velocity components, pressure, temperature, species concentration etc.

These fluctuations are generated in regions of high shear (such as near a wall or in a mixing
layer) by an internal instability mechanism that is characteristic of the flow.

The fluctuations are thus generated, sustained and regulated by the flow itself. They persist
as long as the flow is maintained.
Turbulent Flows

Turbulent flow is characterized by rapid and highly localized fluctuations in flow parameters

Although the fluctuations are small in magnitude (typical amplitude of the velocity
fluctuations is in the range of 1 to 10% of the mean velocity), their effect on the flow can be
enormous.

Flow-dependent quantities such as the pressure drop and heat transfer rate can be an order
of magnitude higher in turbulent flow.

Transport properties such as friction factor, heat and mass transfer coefficients and reaction
rates can change by orders of magnitude compared to the value that would prevail if the
flow remained laminar at the same Reynolds number.
In many cases, the extent of enhancement (or suppression in some rare cases, e.g., drag on a
sphere around transition Reynolds numbers) of the transport coefficient depends on the
Reynolds number itself in a non-trivial way.
Friction factor ~ Ren where n ~ -0.25 to -0.2
Nusselt number ~ Re0.8

It is therefore important to get an accurate representation of these turbulent fluctuations


into flow analysis.
Turbulent Flows
Turbulent flow is characterized by rapid and highly localized fluctuations in flow parameters

0.8
velocity component

0.6

0.4

0.2

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (arbitrary units)

In turbulent flow through a pipe of, say, 5 cm diameter at a Reynolds number of 100 000
Typical time scales of turbulent fluctuations can be of the order milliseconds
Typical length scales are of the order of fractions of a millimeter

How to take account of these fluctuations: x 0.01 mm, t 0.001 s => huge grid, many time steps!
Time and Length Scales of Turbulent Pipe Flows
Turbulent flow is characterized by rapid and highly localized fluctuations in flow parameters
A discerning snap shot of turbulent flow would reveal islands of vorticity in a calmly
flowing river
The vorticity islands or whirlpools or eddies are constantly forming and disappearing and are
of different sizes and orientations
What are time scales and length scales of this churning that takes place in turbulent flow?
D = 0.05 m, Re = 105 1

component
0.8

velocity
Water, kin vis = = 10-6 m2/s 0.6
0.4
Uavg = 2 m/s, friction factor=f 0.005 0.2
0
|dp/dz| = f Ua *4/D = 800 Pa/m
2 0 0.2 0.4 0.6 0.8 1
Time (arbitrary units)
uf = Uavg* (f/2)0.5 = 0.1 m/s

Turb kinetic energy = ~ 0.3uf2 = 0.003 m2/s2


= dissipation rate ~ (|dp/dz|)*dz*Uavg*area/(*area*dz) = 800*2/1000 = 1.6 m2/s3

Eddy size:
largest ~ L/6 ~ 8 mm; smallest ~ (3/) ~ {(10-6)3/1.6}0.25 ~ 0.03 mm

Eddy lifetime:
largest ~ L/(6uf) ~ 0.08 s; smallest : ~ k/ ~ 0.002 s
Dynamic interaction between turbulent eddies
Large eddies are created by instabilities in large shear regions. These are stretched and get
broken up into smaller eddies, which in turn are stretched and produce smaller eddies.
Leads to the formation of an energy cascade which is a typical feature of turbulent flow
Accounts for the high diffusivity of turbulent flow, which is another characteristic feature

~ L/6 ~ (3/)
Energy spectrum in turbulent flow: turbulent kinetic energy (E) is produced by Taylor-scale instability
mechanisms and is cascaded down the inertial range to be dissipated in the Kolmogorov-scale eddies.

How to take account of these fluctuations: x 0.01 mm, t 0.001 s => huge grid, many time steps!
Time averaging of Fluctuations
Turbulent flow fluctuations can be
smoothed out through time-averaging
1
known as Reynolds decomposition

component
0.8

velocity
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
Time (arbitrary units)

0.8

velocity component
0.6

0.4
u'(t)
0.2

Use these operators to time average


0
governing equations 0.4 0.42 0.44 0.46 0.48 0.5
Time (arbitrary units)

<u/x + v/y + w/z> = <0>


=> <u/x> + <v/y> + <w/z> = 0
or <u>/x + <v>/y + <w>/z = 0
Time averaging of Momentum Balance Equations
Consider x-momentum balance in 3-d:
u/t + (u2)/x + (uv)/y + (uw/z = -(1/) p/x + (2u/x2 + 2u/y2+ 2u/z2)
Time-average both sides:
< u/t + (u2)/x + (uv)/y + (uw/z> = <-(1/) p/x + (2u/x2 + 2u/y2+ 2u/z2)>

<u/t> + <(u2)/x> + <(uv)/y> + <(uw)/z> = <-(1/) p/x>+ (<2u/x2 >+ <2u/y2>+ <2u/z2>)
Or
<u>/t + (<u2>)/x> + (<uv>)/y + (<uw>)/z = -(1/) <p>/x + (2<u>/x2 + 2<u>/y2+2<u>/z2)

Apply the rule that <fg> = <f> <g> + <fg> to all products

<u>/t + (<u>2+ <u2>)/x + (<u><v>+<uv>)/y + (<u><w>+<uw>)/z


= -(1/) <p>/x + (2<u>/x2 + 2<u>/y2+2<u>/z2)

Similarly, y- and z-momentum balance equations, upon time-averaging, become


<v>/t + (<u><v>+ <uv>)/x + (<v>2+<v2>)/y + (<v><w>+<vw>)/z
= -(1/) <p>/y + (2<v>/x2 + 2<v>/y2+2<v>/z2)

<w>/t + (<u><w>+ <uw>)/x + (<v>,w>+<vw>)/y + (<w>2+<w2>)/z


= -(1/) <p>/z + (2<w>/x2 + 2<w>/y2+2<w>/z2)
The Turbulence Closure Problem
The time-averaged conservation equations for constant property flow can be written, using the
overbar to indicate time-averaging and Einsteins summation of repeated index convention, as

Reynolds stresses = =

Note <fg> = 0 only if variation of f and g are statistically independent.

If f = temperature of the day and g = no of ice creams sold at a shop, <fg> will be +ve
If f = temperature of the day and h = no of coffees sold at a shop, then <fh> will be ve

Data show that <ui uj> 0 => coherent structures and eddies

Then, time-averaging momentum balance equations introduces six additional variables


=> more number of variables (10) than the number of equations (4) available!

Closure problem of turbulence


Reynolds Stresses
The time-averaged conservation equations for constant property flow can be written as

An insight into these additional terms can be obtained by rewriting the equation as

Reynolds stresses = =

Time-averaged velocity profile Reynolds stresses in pipe flow


The Turbulence Closure Problem in 1-D Flow
Consider steady, fully developed turbulent flow between two infinitely long and wide parallel
plates

The pressure gradient is constant and <v> = 0 = <w> and d(anything else)/dx = 0

The time-averaged continuity equation is d/dx (<u>) = 0

The time-averaged x-momentum balance equation is


<u>/t + (<u>2+ <u2>)/x + (<u><v>+<uv>)/y + (<u><w>+<uw>)/z
= -(1/) <p>/x + (2<u>/x2 + 2<u>/y2+2<u>/z2)

Or (1/) d<p>/dx = C = (d2<u>/dy2) d(<uv>)/dy

Thus, there is only one equation and two unknowns, namely, <u> and <uv>
=> closure problem

Closure problem can be solved only by bring EMPIRICAL information or by making simplifying
assumptions.
Boussinesq Hypothesis and Prandtls Mixing Length Model
Boussinesq (1877 => about 140 years ago!) suggested the following closure:

e.g.,

where t = turbulent or eddy viscosity

For steady, fully developed turbulent flow between two parallel plates

But how to specify t ? Specifying a constant value that would match the pressure gradient
would give a laminar-like velocity profile for turbulent flow which is not correct

t has to vary with y in order to have a different velocity profile while match pressure gradient

About 50 years after Boussinesq, Prandl (1925) suggested a model for t which came to be
known as the mixing length model
Prandtls Mixing Length Model
For steady, fully developed turbulent flow between two parallel plates

with

Prandtl (1925) proposed that

or is the mixing length

The mixing length is the characteristic distance over which turbulent eddies are mixing fluids
layers of different momentum or velocities.
This causes a velocity fluctuation which is roughly the difference in the mean velocities between
the layers, i.e., urms ~ U ~ (dU/dy) lm
Also in turbulent flow, urms ~ vrms => <uv> ~ [(dU/dy) lm] [(dU/dy) lm] which leads to the model

A simple model for the mixing length is that where y is the normal
distance from the wall. This incorporates the idea that very close to the wall, the eddy size is
regulated by the proximity to the wall
Note that with this model, t varies with y; the model is highly successful in predicting the
logarithmic variation of velocity in the near wall region
Several prescriptions for mixing length are available which is both an advantage and a
disdavantage of the model!
More Generic Formulation
Mixing length model is relatively simple but cannot work in 3-D flows and flows with significant
convective and diffusive effects
It also requires the specification of a mixing length which is difficult to conceive of in 3-D

A parameter-free framework for turbulence closure is provided the k- turbulence model.

Here where = turbulent kinetic energy

= turb kin energy dissipation rate

Both k and are field variables and can vary with position and time. They are flow
properties and are non-zero in turbulent flow.

Scalar transport equation-type conservation equations can be derived EXACTLY from the
Navier-Stokes equations (see the books by Warsi (1993) or Wilcox (1993)). But these bring in
more unknowns just as time-averaging of the momentum equations has brought in Reynolds
stresses

Approximate equations for k and are derived based partly on the kind of processes that the
terms in the scalar transport equations represent.
Steps in the derivation of the k-equation
A. Write down the instantaneous ith-direction momentum balance equation
B. Write the down the time-averaged ith direction momentum balance equation
C. Subtract B from A to get the momentum balance equation for ui
D. Similarly derive the momentum balance equation for uj
E. Multiply eqn. C with uj and eqn D with ui and add the two
F. Time-average eqn E to get the equation for Reynolds stressses <umun>
G. Contract the indices of eqn. F to get the equation for k which is <umum>

Eqn. F

Eqn. G
Steps in the derivation of the -equation
C. Get the momentum balance equation for ui as before
D. Differentiate each term with respect to xm, i.e., take /xm of eqn. C
E. Take the inner product (product of two tensors leading to a scalar, i.e., SpqTpq) of ui/xm
with each term of eqn. D
F. Time-average eqn E to get the equation for

Eqn. F

Note that
More Generic Formulation
Turbulence model equations:

where

These equations are solved together with the time-averaged continuity and momentum
balance equations:

Thus, in the model, we have to solve six pdes for six variables, namely, <u>, <v>, <w>, <p>, k
and .
All these equations are converted using discretization schemes into A =b and are solved
sequentially
Course Plan & Closure

Week 1: Introduction : calculation of flow in a rectangular duct


Week 2: Calculation of fully developed flow in a triangular duct
Week 3: Derivation of equations governing fluid flow
Week 4: Equations for incompressible flow and boundary conditions
Week 5: Basic concepts of CFD: Finite difference approximations
Week 6: Basic concepts of CFD: Consistency, stability and convergence
Week 7: Solution of Navier Stokes for compressible flows
Week 8: Solution of Navier Stokes equations for incompressible flows
Week 9: Solution of linear algebraic equations: basic methods
Week 10: Solution of linear algebraic equations: advanced methods
Week 11: Basics of finite volume method including grid generation
Week 12: Turbulent flows and turbulence modelling