Anda di halaman 1dari 20

The k Turbulence Model

ME 448/548 Lecture Notes


Gerald Recktenwald

August 3, 2009
1 Overview
This article gives a very brief overview of the basic k turbulence model used in CFD codes. The
major topics discussed are
The Reynolds Averaged Equations
The modied Boussinesq Eddy-Viscosity Concept
Prandtls Mixing Length Hypothesis
The k and equations
Boundary Conditions and Wall Functions
Good background reading for this material can be found in the books by Ferziger and Peric [1,
Chapter 9] Panton [4, Chapter 23], Pope [5, Chapter 10], and Tannehill et al. [7, 5.2, 5.4].
1.1 What Is Turbulence?
Turbulence is a uid ow phenomenon characterized by
unsteadiness,
uid motions that appear irregular or topologically complex,
uid motions occurring on a wide range of physical scales,
rapid mixing of passive contaminants (smoke, heat, concentrations of chemical species)
Unsteadiness Imagine using a probe to measure the uid velocity at a point in a turbulent ow.
To be specic, consider inserting a probe through a small hole in a duct so that the sensing element
of the probe is located near the centerline. Assume that the probe has a high frequency response,
meaning that it is capable of detecting any time variation in the velocity signal. Figure 1 shows the
output of the probe as it would be displayed on an oscilloscope. The velocity signal is characterized
by a nominal average U, and a superimposed uctuation, u

. The average value is representative


of the mass ow rate through the duct. The uctuating component is an essential feature of the
turbulence.

Associate Professor, Mechanical and Materials Engineering Department Portland State University, Portland,
Oregon, gerry@me.pdx.edu
1.1 What Is Turbulence? 2
0 0.5 1 1.5 2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
t
u
Figure 1: Turbulent velocity signal at a point. The Reynolds decomposition separates the signal
into an average, U, and a uctuating component, u

.
Flow Irregularity Let us continue with our imagined experiment with duct ow. Suppose there
was some way to make the ow patterns inside the duct visible. First of all, assume that the duct
walls were transparent. Then suppose that we were to inject smoke from a series of small pipes
having outlets at a small number of vertical locations on the center plane of the duct. Finally, in
order to keep the smoke from obscuring all the details, illuminate the center plane of the duct with
a sheet of laser light. The light has the eect of showing smoke that is only in the plane. All other
smoke in the duct is not illuminated and, therefore, invisible. Figure 2 is an example of such a ow
visualization.
The rst thing apparent from the smoke visualization is that the ow is extremely complicated.
The ow swirls and uctuates with no apparent pattern. It would be dicult, for example, to trace
out the trajectory of a particular pu of smoke.
Further complicating an identication of ow patterns is the realization that our visualization
system is illuminating the ow in a single plane. The smoke that we see is not conned to move in
the plane. In fact, any smoke that we see at one instant is likely to be swept out of the plane in
the next instant. Thus, although our view is two-dimensional, the ow features are fundamentally
three-dimensional as well as being unsteady.
Multiple Scales After staring at the smoke in the duct, we may begin to adjust our perspective.
In particular, if we allow our eyes to follow a billow of smoke as it is carried downstream, we will
see small whorls of smoke turning and tumbling. If we were to take photographs of the smoke, we
would be able to identify whorls of many dierent sizes. Fluid dynamicists call the whorls eddies.
An exact denition of a turbulent eddy is somewhat elusive. Pope [5, p. 183] describes an eddy
as a turbulent motion, localized within a region of size , that is at least moderately coherent over
this region. Although the shape of an eddy is not precise, we often think of it as being roughly
circular in cross-section. Because an eddy is embedded in an unsteady, three-dimensional ow, it
is in constant motion. Furthermore, the eddy will be stretched and twisted until it breaks up into
other smaller eddies. The notion that the eddy is coherent means that it exists for a period of time,
1.2 Why is Modeling Necessary? 3
Figure 2: Turbulent wake behind a cylinder at Re = 1770. White areas are oil fog illuminated by a
sheet of laser light. Image is from the compilation of Van Dyke [9].
and that while it exists, the uid that constitutes the eddy, has a temporary degree of organization.
For a high Reynolds number turbulent ow, there will be eddies of many dierent sizes
1
. In fact
a large scale eddy will likely be composed of many smaller scale eddies. All of these eddies, these
somewhat ill-dened, transient structures in the ow, are in constant motion.
In our imaginary duct ow, the largest eddy will be no larger than D/2, where D is the char-
acteristic size of the duct cross-section. The size of the smallest eddy depends on the Reynolds
number of the ow, but in general it will be much smaller than D/2. The size of the smallest eddy
is determined by the mechanism of viscous dissipation.
The largest eddies tumble and interact largely under the inuence of inertia, and to some extent
pressure. The largest eddies are virtually immune to viscous eects. As the large eddies break up
into smaller eddies, the role of inertia continues to dominate until the eddies become small enough
that viscous resistance converts their kinetic energy (of translation and rotation) to heat.
Rapid Mixing To complete our visualization of the turbulent duct ow, we turn o the laser
light sheet, and the streams of smoke. We illuminate the duct with a broad beam of light, and
introduce a small pu of smoke in the center of the duct. The pu of smoke is quickly dispersed as
it moves downstream. For a high Reynolds number ow, the dispersal happens so quickly that the
pu might seem to disappear almost as soon as it is introduced. The smoke disappears because it
has been diluted by rapid mixing with the rest of the air in the duct. The smoke particles remain,
but at such a low concentration that they are no longer visible.
The rapid mixing of the turbulent ows is due to bulk transport of uid. A blob of uid is
stretched and reshaped by the eddy motion. This mixing is order of magnitudes more eective
than the Brownian motion that is responsible for the transport properties that we call viscosity and
thermal conductivity.
1.2 Why is Modeling Necessary?
Suppose that to simulate a turbulent ow, we attempt to obtain a numerical solution to the unsteady
Navier-Stokes equations. This is the approach of a strategy called Direct Numerical Simulation
(DNS) of turbulence. DNS is only applicable as a research tool at relatively low Reynolds numbers.
The diculty with DNS is in the spatial grid and time scale requirements to model all the important
motions in a turbulent ow. Furthermore, any direct simulation of turbulence yields an unsteady
ow eld. Extraction of mean quantities (say overall drag, or local heat transfer rate) requires the
time (or ensemble) averaging of a vast amount of data.
1
The meaning of high is somewhat problem-dependent. Let us just assume that the velocity and length scales
of the problem are sucient that the ow is very far from being laminar.
2 REYNOLDS AVERAGED EQUATIONS 4
Table 1: Computational cost for direct numerical simulation of isotropic turbulence. Adapted from
Pope [5, Table 9.2, p. 349]. In November 2005, the Blue Gene/L computer was the fastest computer
in the world. Blue Gene/L has 131072 processors and achieved a sustained 280 TFlops on the
LINPACK benchmark. 1 Top = 10
12
oating point operations per second. A high end personal
computer has a theoretical performance of between 1 and 10 Gop. 1 Gop = 10
9
oating point
operations per second.
CPU Time
Re
L
N N
3
PC at 1 Gop
Supercomputer
at 1 Top
Blue Gene/L
at 280 Top
100 52 1.1 10
6
3 minutes 1.94 sec 0.2 sec
1000 291 1.0 10
7
47 hours 0.05 sec 0.2 msec
10,000 1640 1.2 10
8
5.3 years 47 sec 0.2 sec
100,000 9200 2.0 10
9
5300 years 5.3 years 6.9 days
1,000,000 52000 3.8 10
10
53,000 centuries 5300 years 19 years
Pope [5, 9.1.2] presents the following estimates of the computational cost of using DNS to
simulate a homogeneous turbulent ow. Homogeneous turbulence is an idealization of turbulent
ow far from any bounding walls, and free of any shear or other ow structure. For a homogeneous
turbulent ow in a region of size L, and having maximum mean velocity U, the Reynolds number
is Re = UL/. Let N be the number of points necessary to resolve all relevant ow scales in any
one coordinate direction. The total number of mesh points necessary to simulate the ow is
N
3
4.4Re
9/4
L
If the simulation is performed on a computer with a sustained throughput of one Gigaop (10
9
oating point operations per second) the time to complete the simulate in days is
T
G

_
Re
L
800
_
3
Evaluating these formulas for a range of Re
L
gives the results in Table 1. Clearly, DNS is not
applicable for engineering design work.
2 Reynolds Averaged Equations
The governing equations for ow in Cartesian coordinates are the continuity equation

t
+

x
(u) +

y
(v) +

z
(w) = 0 (1)
and the momentum equations (in conservative form, no body forces)

t
(u) +

x
(uu) +

y
(vu) +

z
(wu) =
p
x
+

xx
x
+

xy
y
+

xz
z
(2)

t
(v) +

x
(uv) +

y
(vv) +

z
(wv) =
p
y
+

yx
x
+

yy
y
+

yz
z
(3)

t
(w) +

x
(uw) +

y
(vw) +

z
(ww) =
p
z
+

zx
x
+

zy
y
+

zz
z
(4)
2.1 Reynolds Decomposition 5
where the velocity vector is u = e
x
u +e
y
v +e
z
w, and the shear stresses are

xx
=
_
2
u
x

2
3
( u)
_

yy
=
_
2
v
y

2
3
( u)
_

zz
=
_
2
w
z

2
3
( u)
_

xy
=
yx
=
_
u
y
+
v
x
_

xz
=
zx
=
_
u
z
+
w
x
_

yz
=
zy
=
_
v
z
+
w
y
_
2.1 Reynolds Decomposition
Figure 1 shows a short sample of a turbulent signal. The Reynolds decomposition identies an
average value U and a uctuating component u

such that the instantaneous signal is


u = U + u

(5)
The average of a signal is also designated with an overbar, i.e.
u = U
By denition of the average
u

= 0
The governing equations are averaged with the Reynolds averaging rules which apply to any two
functions f and g
1. f + g =

f + g
2. af = a

f (a is a constant)
3.
f
s
=


f
s
4.

fg =

f g
It is possible to dene averages that do and do not satisfy these rules.
A common form of averaging is time-averaging, which is dened by
U = lim
T
(1/T)
_
t0+T
t0
u(t) dt (6)
In general both the average and uctuating component are functions of space. For time averaging,
as dened by Equation (6) the average velocity will not be a function of time, i.e., the average ow
is steady. Some turbulent ows are unsteady in the mean, for example the ow inside the cylinder
of an IC engine. In these cases the ensemble average (see below) is used.
The averaging process has the important property that although u

= 0,
u

= 0
Terms like u

i
u

j
are called Reynolds stresses, and arise from the to process of Reynolds averaging of
the momentum equations.
2.2 Ensemble Averaging 6
0 1 2 3 4 5 6 7 8
2
4
6
Sample 1
0 1 2 3 4 5 6 7 8
2
4
6
Sample 2
0 1 2 3 4 5 6 7 8
2
4
6
Sample 3
0 1 2 3 4 5 6 7 8
2
4
6
Ensemble
Figure 3: Ensemble averaging of a velocity at a point in a turbulent ow with a periodic forcing
function.
2.2 Ensemble Averaging
Consider an unsteady ow with a periodic forcing function (e.g. sinusoidally varying inlet pressure
or inlet velocity, or a boundary that moves with simple harmonic motion). Figure 3 depicts three
samples (top three plots) and the ensemble average of the velocity measured at a point. If the
velocity vector is u = u(x, t), then the ensemble averaged velocity is
U(x, t) = lim
n
1
n
n

i=1
u
i
(x, t) (7)
where n is the total number of samples and i is the index of the sample. U(x, t) is the ensemble
average of u at a xed location, x.
Ensemble averaging satises the Reynolds averaging rules. By applying the Reynolds averag-
ing rules to the governing equations, a new set of equations is obtained. These equations have
ensemble averaged velocities (U, V, W) and second order turbulence correlations as dependent vari-
ables. Note that if ensemble averaging is used, the average velocities will be functions of time, e.g.,
U = U(x, y, z, t). The ensemble averaging has achieved a separation of the uctuations from the
average quantities.
2.2.1 Example of Reynolds Averaging
Each component of the velocity eld is assumed to uctuate so that we can identify three pairs of
average velocity (components) and uctuating components
u = U + u

v = V + v

w = W + w

2.2 Ensemble Averaging 7


Apply the Reynolds averaging rules to the left side of Equation (2). For simplicity, assume that
there are no uctuations of pressure or density, i.e., p

= 0 and

= 0. Consider each term

t
(u) =

t
(u) rule 3
=

t
( u) rule 2, with = constant
=

t
(U) by denition of U

x
(uu) =

x
_
(U + u

)(U + u

)
_
rule 3
=

x
_

_
UU + 2Uu

+ u

_
rule 2, with = constant
=

x
_
UU + u

2
_
by denition of U, and since Uu

= Uu

= 0
Similarly,

y
(vu) =

y
_
V U + u

z
(wu) =

z
_
WU + u

_
Thus,

t
(u) +

x
(uu) +

y
(vu) +

z
(wu) =

t
(U) +

x
(UU) +

y
(V U) +

z
(WU)
+

x
_
u

_
+

y
_
u

_
+

z
_
u

_
The second line of the preceding equation consists of second order correlations that are a direct result
of the averaging process. These second order correlations of the uctuating velocity components are
called the Reynolds stresses.
Applying the Reynolds averaging rules to the right hand side of Equation (2) gives

p
x
+

xx
x
+

xy
y
+

xz
z
=
p
x
+

xx
x
+

xy
y
+

xz
z
Notice that no extra correlation terms appears due to averaging of the right hand side.
Putting the averaged x-direction momentum equation back together and moving the Reynolds
stresses to the right hand side gives.

t
(U) +

x
(UU) +

y
(V U) +

z
(WU) (8)
=
p
x
+

x
_

xx
u

+

y
_

xy
u

+

z
_

xz
u

Equation (8) suggests how the Reynolds stresses get their name. Averaging the governing equations
(to simplify the analysis) results in additional terms that appear as stresses.
2.3 Flow That Is Steady in the Mean 8
Repeating the averaging process for the continuity and the y and z momentum equations gives

t
+

x
(U) +

y
(V ) +

z
(W) = 0 (9)

t
(V ) +

x
(UV ) +

y
(V V ) +

z
(WV ) (10)
=
p
y
+

x
_

yx
v

+

y
_

yy
v

+

z
_

yz
u

t
(W) +

x
(UW) +

y
(V W) +

z
(WW) (11)
=
p
z
+

x
_

zx
w

+

y
_

zy
w

+

z
_

zz
w

2.3 Flow That Is Steady in the Mean


In the preceding section the governing equations were transformed by performing an ensemble
average. Now consider a ow that lacks any periodic forcing function and that does not exhibit
large scale periodic motion of any kind. This is the case in many practical situations, for example,
in turbulent ow in ducts. One expects that while the unsteady turbulent uctuations exist, the
average velocity eld appears to be steady. Such is the case depicted by the velocity signal in
Figure 1.
When time averaging of the velocity eld at each point in the ow yields constant value of U,
V , W and p, then the ow is said to be steady in the mean. For such ows it is appropriate to use
Equation (6) to perform Reynolds averaging of the governing equations. The result is the same as
Equations (8) through (11) with the exception of the time derivatives, which vanish.
If the ow is steady in the mean then the governing equations are simplied. Although the
Reynolds stresses are still present, at least the average velocity eld is steady. If appropriate models
for the Reynolds stresses can be introduced, a steady numerical solution can be obtained even
though the ow is characterized by a high degree of uctuation in the velocity signal. This is the
heart of turbulence models used in CFD codes.
2.4 The Closure Problem
The Reynolds averaging process has created new dependent variables the Reynolds stresses
without adding to the set of governing equations. This is called the turbulence closure problem. For
three-dimensional, incompressible ow there are six Reynolds stresses, which can be represented by
the following matrix.
_
_
u

_
_
The matrix is symmetric because u

= v

and v

= w

. The Reynolds stresses are eld


variables, i.e., each term in the matrix is a function of space in the ow.
3 Basic Turbulence Modeling
Consult Tannehill [7, 5.4] and Wilcox [11] for overviews of turbulence modeling applied to CFD
codes. Pope [5, Chapter 10] explains the fundamental deciencies of this modeling approach.
3.1 Eddy-Viscosity Concept 9
3.1 Eddy-Viscosity Concept
The viscous stress tensor is

ij
=
__
U
j
x
i
+
U
i
x
i
_

2
3

ij
U
k
x
k
_
(12)
By direct analogy we propose (guess, assume, hope) that the Reynolds stresses have the same
form as the viscous stresses. This model introduces a turbulence viscosity
t
or eddy viscosity such
that [2, 5, 7]
u

i
u

j
=
t
__
U
j
x
i
+
U
i
x
i
_

2
3

ij
U
k
x
k
_

2
3

ij
k (13)
where k is called the turbulence kinetic energy
k =
1
2
u

k
u

k
=
1
2
_
u

+ v

+ w

_
(14)
Equation (13) eectively replaces the Reynolds stresses by two new unknowns, k and
t
. Both

t
and k vary from point to point in the ow. It is important to realize that Equation (13) is used
because it is convenient, not because it is an accurate model of the Reynolds stresses.
Substituting Equation (12) and Equation (13) into the Reynolds averaged momentum equations,
and simplifying for steady, incompressible ow gives

x
(UU) +

y
(V U) +

z
(WU)
=
p
x
+

x
_

e
U
x
_
+

y
_

e
U
y
_
+

z
_

e
U
z
_
+ S
U
(15)

x
(UV ) +

y
(V V ) +

z
(WV )
=
p
y
+

x
_

e
V
x
_
+

y
_

e
V
y
_
+

z
_

e
V
z
_
+ S
V
(16)

x
(UW) +

y
(V W) +

z
(WW)
=
p
z
+

x
_

e
W
x
_
+

y
_

e
W
y
_
+

z
_

e
W
z
_
+ S
W
(17)
where

e
= +
t
(18)
p = p
1
3

e
u +
2
3
k (19)
and S
U
, S
V
, and S
W
are additional source terms due to the non-uniform viscosity. For example,
S
U
=

e
y
V
x

V
y

e
x
+

e
z
W
z

W
z

e
x
.
Use of the eddy viscosity concept in a CFD code involves replacing the true viscosity, , by
e
.
In general
e
. We now turn to the job of computing
t
.
3.2 Prandtl Mixing Length Hypothesis 10
Table 2: Comparison of the molecular model of viscosity from the kinetic theory of gases with
Prandtls mixing length hypothesis.
molecular viscosity of gases turbulence viscosity
=
1
3

f
V
m

t
=
m
V
t

f
= mean free path
m
= mixing length
V
m
= velocity of molecules V
t
= velocity scale of turbulence
3.2 Prandtl Mixing Length Hypothesis
The Prandtl mixing length hypothesis is based on a direct analogy with the molecular theory of
viscosity from the kinetic theory of gases. In general the (true) viscosity is a macroscopic manifes-
tation of the microscopic behavior of a uid. For a gas, the viscosity arises because momentum is
transferred across velocity gradient by the random motion of gas molecules (see, e.g., Panton [4,
Chapter 6] and Tennekes and Lumley [8]). Prandtl hypothesized that the turbulent eddies enhance
the exchange of momentum (and passive scalars) by augmenting the molecular diusion process.
The relationship between Prandtls mixing length model and the viscosity model from the kinetic
theory of gases is summarized in Table 2.
The mixing length hypothesis has two parts:
1. A model for the turbulence viscosity

t
=
u

i
u

j
U/y
where U/y is the mean velocity gradient. (The mixing length was rst developed for bound-
ary layer ows where the only important velocity gradient is U/y.)
2. A model for the turbulent velocity uctuations
V
t
=
m

U
y

One problem with the mixing length hypothesis is that


t
is not dened wherever U/y = 0,
e.g. at the centerline of a pipe. Another problem is that the mixing length hypothesis does not
include the inuence of upstream events, i.e. turbulence cannot be convected downstream. Yet
another problem is that the mixing length model requires specication of the mixing length. For
some ows, e.g. boundary layers, and the far eld of isolated turbulent jets and wakes, experimental
data is available so that specication of a mixing length is possible. For a general ow, especially
one involving strong recirculation, identication of a single mixing length is not possible. The idea
of the mixing length is still useful as a conceptual model, and as a point of reference for more
sophisticated models.
3.3 Constant Eddy Viscosity Model
A simplistic model of turbulence involves asserting that
e
is constant throughout the ow. The
trick, of course, is to specify a value of
e
that is meaningful. The constant eddy viscosity model
3.3 Constant Eddy Viscosity Model 11
will not allow accurate prediction of the smaller scale features of the ows. An estimate of
e
is
obtained by assuming
m
is uniform throughout the ow.
3.3.1 Example: Flow in a Pipe
We can easily estimate the magnitude of the eddy viscosity in a pipe. Prandtls model is

t
V
t

m
In general, the magnitude of the turbulent uctuations are small compared to the mean velocity in
the pipe.
0.01
u

U
0.15
To estimate
t
, take u

/U 0.1 and V
t
u

so that
V
t
0.1U
The ow in the pipe contains eddies ranging from half the diameter of the pipe down to the scale
at which kinetic energy is dissipated by viscosity. To make a somewhat arbitrary choice of a single
length scale that characterizes the turbulent mixing, take

m

D
4
Combining the preceding expressions gives the following estimate of turbulence viscosity

t
= (0.1U)(0.25D) = 0.025UD
From the estimate of
t
we can compute an eective Reynolds number
Re
e
=
UD

t
=
UD
0.025UD
= 40
Therefore, the constant viscosity model has the eect of modeling the uid as if it were very viscous.
At rst this may seem counter-intuitive, but deeper examination shows that the large apparent
viscosity is consistent with Prandtls mixing length hypothesis.
Recall that the mixing length hypothesis is based on an analogy with the model of viscosity
obtained from the kinetic theory of gases. There, it is assumed that the small scale random motion
of individual molecules gives rise to a transport of momentum whenever there is a velocity gradient.
In the mixing length model, uid blobs transported by turbulent uctuations take the role of the
individual molecules in the kinetic theory. Thus, turbulent uctuations enhance the transport of
momentum. The increase in momentum transport is reected by the increase in the apparent
viscosity.
Though this model is intuitively compelling, the reality of turbulent mixing is not so simple.
Pope [5, 10.1] uses experimental data and scaling arguments to show that the mechanics of molec-
ular diusion and turbulent mixing happen on vastly dierent time scales. Molecular mixing happens
very quickly because the velocity of molecules on the microscopic scale is very high
2
. Turbulent
eddies rotate on a time scale comparable to the mean ow velocities. As a consequence, models of
turbulent mixing that rely on mean velocity gradients only work when the mean ow eld changes
relatively slowly in the ow direction.
2
From the kinetic theory of gases (see, e.g. Wark [10]), the RMS velocity of a molecule of gas is vrms =
p
3kT/m,
where k = 1.3810
23
J/K/molecule is Boltzmanns constant, T is the absolute temperature of the gas, m = M/N
A
is the mass per molecule, Mis the molar mass of the gas, and N
A
= 6.02410
26
atoms/k-mol is Avogadros number.
For Nitrogen (M= 28 kg/k-mol) at 294 K, vrms = 511 m/s.
4 THE K MODEL 12
3.3.2 Constant Turbulence Viscosity Models in CFD Codes
The constant viscosity model attempts to simulate the eect of turbulent uctuations by increasing
the apparent viscosity of the uid. This is a crude interpretation of the mixing length model, for it
assumes that the mixing length is uniform throughout the ow. In fact, turbulent transport occurs
due to eddies of many dierent sizes. Near walls, the eddies are smaller. Downstream protruding
objects such as cylinders or other blu bodies, the eddy size is strongly inuenced by the size of the
object. Shear layers, e.g. where two streams of dierent velocity mix, create high levels of turbulent
uctuations. All of these ow features are not accounted for by a model that assumes a uniform
turbulence viscosity.
Though the constant turbulence viscosity model cannot be expected to yield accurate simulation
of many ow features, it is still useful in CFD codes. The constant viscosity model is especially help-
ful for preliminary analysis, and for developing an initial guess at the ow eld. More sophisticated
turbulence models require signicantly greater computational eort. In addition, more complex
models may also introduce convergence problems for the calculations. Therefore, when beginning
a new CFD simulation, it is helpful to use the constant turbulence viscosity model to generate a
rst guess at the ow eld. The computations can then be restarted with a more sophisticated
turbulence model after the large scale features of the ow have been established.
4 The k Model
The k model was rst proposed by Jones and Launder [3]. It is now consider the standard
turbulence model for engineering simulation of ows.
The modied Boussinesq eddy viscosity model overcomes the rst problem of the mixing length
hypothesis, viz. that
t
is not dened in regions of zero shear (U/y = 0). To relate
t
to the
Reynolds stresses, and assume that
V
t

k
so that

t
= C
m
k
1/2
(20)
where C is a constant. Using this model
t
is nonzero everywhere in the ow that k is nonzero.
The new independent variables of the turbulence model are
m
and k.
4.1 The k Equation
An exact equation for k is obtained by taking the inner product of the velocity vector and the
momentum equation (in vector form). The result after some algebra is a conservation equation for
k (see [2]).
Dk
Dt
=

x
i
_
u

i
_
p

+ k
_
_
. .
convective diusion
u

i
u

j
U
j
x
i
. .
production by deformation
+

x
i
_
u

j
_
u

i
x
j
+
u
j
x
i
_
_
. .
production by viscous shear

_
u

i
x
j
+
u
j
x
i
_
u

j
x
i
. .
dissipation
4.2 The Equation 13
Like the Reynolds averaged equations this equation also has higher order correlations. The solution
is to model these correlations. The standard form of the model is (see [7])
Dk
Dt
=

x
i
_

k
k
x
i
_
. .
diusion
+
_

t
_
U
i
x
j
+
U
j
x
i
_

2
3

ij
k
_
U
j
x
i
. .
production
c
D
k
3/2

m
. .
dissipation
(21)
Note that the model equation for k includes the mixing length,
m
in the dissipation term.
Equation (21) is a conservation equation for k. The left hand side has terms for storage and
convection. These terms are balanced on the right hand side by diusion, and source terms due to
turbulence production (source) and turbulence dissipation (sink).
If
m
were known, then we could obtain a numerical solution to Equation (21) using the same
approximations applied to the momentum equations. The result would be a spatial distribution of
k on the computational mesh. From the discrete k(x, y, z) eld, Equation (20) would be used to
compute the local
t
, which in turn would be used in the discrete form of the momentum equations.
The net eect is that solution of the momentum equations would require simultaneous solution of
the k equation. This would be an improvement over the constant viscosity model.
The question remains, How can
m
be specied?
4.2 The Equation
In a turbulent ow, some mechanism or device is responsible for adding energy to the uid. In a duct
ow, a fan converts electrical energy to work, and imparts momentum to the uid. In atmospheric
ows, e.g. the weather in the troposphere, solar heating causes large columns of air to rise, while
precipitation induces downdrafts. These external energy inputs cause large scale motions in the
uid. Motions on a large scale correspond to large Reynolds numbers. The organization of these
large scale motions is not stable, i.e. a large swirling vortex tends to break up into smaller vortices.
High Reynolds number turbulent ows have motions that range over many length scales. The
largest scales correspond to the scale of the energy input mechanism or to the size of the bounding
walls. At the large scales, viscosity is unimportant. In other words, the large vortex structures exist
and breakup into smaller scale vortices under the inuence of pressure and inertia alone. Viscous
eects are important predominantly at the smallest scales of motion. At the smallest scales the
velocity gradients are smoothed out by the dissipative eect of viscosity.
In a typical turbulent ow, the kinetic energy of the largest scale motions is transferred to
successively smaller scale motions without loss. It is only at the small scales that energy dissipation
occurs.
We can use a simple scaling argument to estimate the dissipation rate (see Panton [4, 22.10]).
Consider a ow where u
0
is linear velocity associated with the largest eddy. The kinetic energy
of this eddy is proportional to u
2
0
. If the diameter of the eddy is L, then the eddy completes a
revolution in time L/u
0
. Let be the rate at which energy is dissipated. An upper estimate of is

kinetic energy of an eddy
time for one rotation
=
u
2
0
L/u
0
=
u
3
0
L
(22)
Ultimately all of this energy is dissipated. In a steady ow the amount of energy dissipated is equal
to the amount of energy that must be continually supplied to the ow.
In a turbulent ow the energy is dissipated at the smallest scales. Thus, although Equation (22)
provides an estimate for the magnitude of the dissipation rate, it does not correspond to the actual
dissipation mechanism. The turbulence kinetic energy, k is a measure of the energy in the velocity
4.3 Calculation of Eective Viscosity 14
uctuations. If the turbulence is isotropic, i.e. if |u

| = |v

| = |w

| then
3
k =
3
2
u

isotropic turbulence
and we can estimate the velocity of the uctuations with |u

k. Furthermore if we dene

as
the dissipation length scale, i.e. the length scale on which the viscous dissipation mechanism occurs,
then we can write the dissipation rate as
C
D
k
3/2

(23)
where C
D
is to be a constant that is assumed adjust the magnitude of the right hand side to the true
magnitude of . Equation 23 is a scaling relationship, not the equation that allows us to compute
the distribution of .
We can develop a relationship between
t
and as follows. Solve Equation (23) for

C
D
k
3/2

(24)
Use

as an indication of the scale for


m
. We do not assert that

and
m
are equal. Rather
assume that their ratio is constant. Thus, if
m
= C

, where C

is a constant, then we use the


following estimates in Prandtls mixing length model

t
V
t

m
Prandtl
V
t

k isotropic turbulence

C
D
k
3/2

scale estimate
to get

t
= C
_
C
D
k
3/2

_
_
k
1/2
_
or

t
= C

k
2

(25)
The standard form of the modeled conservation equation for is (check this?)
D
Dt
=

x
i
_

x
i
_
. .
diusion
+c
,1
_

t
_
U
i
x
j
+
U
j
x
i
_

2
3

ij
k
_
U
j
x
i
. .
production
c
,1

2
k
. .
dissipation
(26)
4.3 Calculation of Eective Viscosity
A typical turbulent ow would involve the following iterative loop:
1. Set-up and solve equations for the velocities and pressure, using the current guess at the
eective viscosity.
2. Solve the k and equations
3
|u

| = |v

| = |w

| is a consequence of isotropy, not the denition of it.


4.4 Wall Functions 15
u
e

~ 0.1
~ 0.1
viscous sublayer
0 y
+
8
<
~
buffer layer
8 y
+
30
overlap layer
(a.k.a. log region,
inerial sublayer)
30 y
+
<
~
<
~
<
~
Figure 4: Velocity prole in a turbulent boundary layer.
3. At each point in the ow compute

t
= C

k
2

4. Use
t
as appropriate to compute the eective diusion coecient for each dependent variable.
5. Return to step 1 until convergence.
4.4 Wall Functions
In the basic CFD model of turbulent ow, the boundary conditions for the dependent variables (u, v,
w, T, k, ) are implemented with wall functions. Wall functions are derived from a semi-empirical
model of turbulent boundary layer ow called the law-of-the-wall. More sophisticated near-wall
treatments are available.
As depicted in Figure 5 the k model applies to the central part of the calculation domain.
Because velocity and temperature gradients are very steep near the wall, it is often impractical to
resolve all the details of the ow in the near-wall region. Wall functions are an economical (in terms
of mesh points and CPU eort) way to bridge the region between the true wall boundary values and
the turbulent core ow. Before discussing the implementation of wall functions we will rst review
the features of turbulent boundary layers and introduce the law-of-the-wall.
4.4.1 Law of the Wall
See Panton [4] for a clear and concise introduction to the scaling laws for turbulent boundary layers.
Panton also gives a nice introduction to turbulent ow in general. More detailed discussions are
provided by Hinze [2] and Pope [5].
Figure 4 depicts the prole of the mean velocity in a turbulent boundary layer. The boundary
layer is a region of thickness over which the velocity varies from 0 to u
e
, where u
e
is the local
velocity of the external ow. The external velocity is assumed to vary slowly with position, i.e.
u
e
is not expected to change much over distances of the order of, say, 10. When this is true,
the boundary layer will have a universal structure that is the same in many dierent ows. This
assumption of universal structure is at the heart of the near wall model of turbulence used in the
common CFD models of turbulent ow.
4.4 Wall Functions 16
In the left half of Figure 4 the velocity prole is divided into two regions: an inner region between
the wall and roughly 0.1, and an outer region between approximately 0.1 and the free stream.
The inner region is further divided into three layers as shown in the right half of Figure 4. Wall
functions are computational models used to bridge the numerical approximation in the turbulent
core ow (the edge of which corresponds to the outer region), and the wall without resolving the
details of the inner region. In other words, wall functions are models of the turbulent momentum
transport in the region 0 y 0.1.
The extent of the sublayers of the inner region are characterized by a dimensionless variable
y
+
=
yu

(27)
where y is the distance normal to the wall, u

is called the friction velocity and is the kinematic


viscosity of the uid. Note that y
+
has the same form as a Reynolds number where the characteristic
velocity is u

instead of u
e
.
The delineation of an inner region (y < 0.1) and outer region (y > 0.1) reects a fundamental
truth about turbulent boundary layers. The outer region is characterized by the ow outside the
boundary layer, and the inner region is characterized by the conditions at the wall. Near the wall
the velocity prole has the functional form (see, e.g., [4])
u = f
in
(y, , ,
w
,
w
) (28)
where
w
is the wall shear stress, and
w
is a length scale characterizing the wall roughness
4
. In the
outer region the velocity prole has the functional form
u u
e
= f
out
_
y, , ,
dp
dx
,
w
_
(29)
Comparing Equation (28) and (29) we see that the near wall region is insensitive to the external
pressure gradient, the boundary layer thickness, , and the external velocity, u
e
. Also note that the
wall shear stress,
w
, aects both the inner and outer proles.
The dimensionless form of the inner part of the prole (Equation (28)) is
u
+
= F
in
(y
+
,
+
w
) (30)
where
u
+
=
u
u

(31)
and the friction velocity is dened by
u

=
_

(32)
Note that
w
has the units of pressure (stress), and (by denition of u

)
w
= u
2

. The dimensionless
wall roughness is

+
w
=

w
u

(33)
Equation (30) is called the law-of-the-wall. This so-called law is a semi-empirical model which
describes the near-wall behavior of boundary layers and ow in ducts. The law-of-the-wall covers
the three sublayers depicted in the right half of Figure 4.
4
Do not confuse the roughness scale, w, with the turbulence dissipation rate, , used in the k model.
4.4 Wall Functions 17
k- model is used to compute
eff
in the central region of the flow
Wall functions are used to compute
eff
near solid surfaces.
Figure 5: Wall functions apply to thin layers between solid walls and the central part of the ow
domain.
In the viscous sublayer the shear stress is dominated by viscous forces
5
. The velocity prole in
the viscous sublayer is
u
+
= y
+
or u = y (34)
i.e., the mean velocity prole in the viscous sublayer is nominally linear. Equation (34) is hard to
verify experimentally because the region y
+
< 5 is very close to the wall. It is dicult to build a
probe that is small enough to accurately measure the velocity in this region. We do know, however,
that all velocity uctuations are zero at the wall so the importance of the Reynolds stresses vanish
as y 0.
The middle part of the inner layer is called the buer layer. It has no simple correlating equation.
The outer part of the inner layer is called the overlap layer or the log-law layer. The log-law layer
is correlated by an equation of the form
u
+
= c
1
lny
+
+ c
2
(35)
where c
1
and c
2
are constants.
The dimensionless form of the outer part of the prole (Equation (29)) is
u u
e
u

= F
out
_
y

,

u
2

dp
dx
_
(36)
In a CFD code the outer part of the velocity prole is not used.
4.4.2 Wall Function Boundary Conditions
Wall functions provide a computational glue layer that allow the turbulent core ow to be linked
to the physical boundary conditions at the wall of the domain. Wall functions are used because it
is often not economical to resolve the structure of the turbulent boundary layer, especially in the
sublayers close to the wall. Simulation of the full turbulent boundary layer would require many
5
Some older texts refer to this as the laminar sublayer. It is misleading to consider this a laminar layer because
the ow in the sublayer is characterized by slower motions driven by the turbulent eddies outside of the viscous
sublayer. Although viscous forces are dominant in this sublayer, the ow in the sublayer is not necessarily smooth
and sheet-like the dening characteristic of laminar ow.
4.4 Wall Functions 18
I
B
y
I
u
res
Figure 6: Wall functions model the variation of dependent variables between nodes adjacent to a
solid boundaries. Node B is on the boundary. Node I is the nearest interior node close to node B.
nodes (cells) to be placed very close to the wall. This consumes memory (to store element and node
data) and increases the CPU time to achieve a converged solution
6
. Instead of resolving the details
of the wall boundary layers, wall functions attempt to impose the boundary condition information
on the core ow in such a way that the correct ux of momentum (drag) and heat at the wall is
obtained. One way to look at wall functions is to consider them as a computational device for
specifying
e
for nodes on and near the wall. The
e
near the wall needs to be modied by the
presence of the wall, and wall functions achieve approximate values of
e
from the law-of-the-wall.
Figure 5 depicts the ow very close to a solid wall. The circles labeled B and I represent
a node on the boundary, and the nearest internal node on the computational mesh. In the wall
function model, node I is assumed to lie outside of the log-law region, i.e. y
+
I
> 30. The purpose
of the wall function is to model the wall shear stress from the value of the mean velocity at node I.
The model enables a prediction of the wall shear stress (and wall heat ux if there energy transport
is being simulated) without resolving the detailed structure of the boundary layer with many nodes
between node B and node I.
The log-law part (cf. Equation (35)) of the law-of-the-wall can be written (see e.g. Rodi [6])
u
res
u

=
1

ln(y
+
E) (37)
where u
res
is the resultant velocity parallel to the wall, = 0.4 is von Karmans constant, and E
is the roughness parameter (E = 9.0 for smooth walls). Use a nite-dierence model to relate the
wall shear stress to the velocity gradient in the y direction

w
=
e
u
y

w

e
u
I
u
B
y
I
(38)
where
e
is the eective viscosity. Note that the velocity prole between node B and node I is
assumed to look like the right half of Figure 4. The role of the wall function is to provide the estimate
of
e
that compensates for the crude estimate of the velocity gradient given by Equation (38).
Since the no-slip condition requires u
B
= 0, Equation (38) can be rearranged as

e
=
w
y
I
u
I
Using the denitions of y
+
, u
+
, and u

the preceding equation becomes

e
=
w
(y
+
I
/u

)
u
+
I
u

=

w
u
2

y
+
u
+
=
y
+
I
u
+
I
(39)
6
Because the solution is a superlinear function of the total number of cells in the mesh.
5 SUMMARY 19
where y
+
I
and u
+
I
are evaluated at the interior point, I.
Evaluation of
e
from Equation (39) depends on the value of y
+
I
. Ideally the rst interior point
is far enough from the wall so that y
+
I
> 30. In that case, Equation (37) is used to relate y
+
I
to u
+
I
,
i.e.
u
+
I
=
1

ln(y
+
I
E) =
y
+
I
u
+
I
=
y
+
I
ln(y
+
I
E)
If y
+
I
< 30 the wall function model breaks down because there is no universal functional relationship
for the buer layer (8 y
+
30). When y
+
I
< 30 most CFD codes print a warning message. Some
codes apply ad-hoc analytical corrections to the wall function boundary condition. If y
+
I
< 5
Equation (34) may be used, i.e,

e
= if y
+
I
< 5
Otherwise

e
=
y
+
ln(Ey
+
)
if y
+
I
> 30
5 Summary
1. Reynolds averaging leads to equations governing the mean or ensemble averaged ow eld
2. Reynolds averaging also introduces the Reynolds stresses and creates the closure problem
3. Closure in the k model is achieved by
a. relating u

i
u

j
to
t
b. using model equations to evaluate
t
For the constant eddy-viscosity model,
t
is assumed to be uniform throughout the ow eld.
For the k model
t
is a local function of the k and elds. Wall functions are used to
compute the value of
t
near solid boundaries for the k model.
4. Eddy-viscosity models are not accurate for ows with anisotropic Reynolds stresses, and ows
for which the local production and dissipation of turbulence kinetic energy is not in balance.
5. k is the turbulence kinetic energy
k has a direct physical signicance
there is an exact equation for k, but it has higher order correlations that cannot be
directly computed
In the k model, the k is an approximation to the true k
turbulence is assumed to be isotropic
k represents energy contained in eddies of many sizes
k is governed by a transport equation with terms that are models of the exact terms
6. is the turbulence dissipation rate
represents the action of the small eddies that are responsible for dissipating the kinetic
energy of turbulence into heat.
an exact dierential equation for can be derived, but it contains higher order correlations
that cannot be directly computed
6 ADVICE 20
In the k model, the equation is an approximation to the true equation
7. Boundary conditions in a k model are often implemented by wall functions
The turbulent boundary layer model is used to avoid resolving steep gradients near walls
6 Advice
Turbulence modeling is an inexact art. Most commercial codes oer a variety of turbulence models.
Here we have described the standard k model. We have barely scratched the surface of this
important topic.
It is important that you do additional research on your application to determine which turbulence
model should be used. You should not expect the k model to give good results in ow with
rapid streamwise changes in the mean ow variables, or in ows with strong swirling components.
Other models to investigate are the k model and Reynolds stress models. Many commercial
CFD codes include these models as options. The k model has some advantage over the standard
k model for simulating near-wall ows and ows with strong streamwise pressure gradients. The
Reynolds stress models are theoretically more sound because they do not rely on the eddy viscosity
model to relate the Reynolds stresses to the mean ow gradients.
Research in this area is continuing, and you should expect to be faced with more choices in
turbulence modeling in the future. Above all, remember this: just because you have obtained a
simulation of a turbulent ow with a turbulence model, in no way guarantees that this simulation
is accurate.
References
[1] Joel H. Ferziger and Milovan Peric. Computational Methods for Fluid Dynamics. Springer-
Verlag, Berlin, third edition, 2001.
[2] J.O. Hinze. Turbulence. McGraw-Hill, New York, second edition, 1975.
[3] W.P. Jones and B. E. Launder. The prediction of laminarization with a two-equation model of
turbulence. International Journal of Heat and Mass Transfer, 15:301314, 1972.
[4] Ronald L. Panton. Incompressible Flow. Wiley, New York, second edition, 1996.
[5] Stephen B. Pope. Turbulent Flows. Cambridge University Press, Cambridge, UK, 2000.
[6] Wolfgang Rodi. Turbulence Models and Their Application in Hydraulics. International Associ-
ation for Hydraulic Research, Delft, Netherlands, second edition, 1984.
[7] John C. Tannehill, Dale A. Anderson, and Richard A. Pletcher. Computational Fluid Mechanics
and Heat Transfer. Taylor and Francis, Washington, D.C., second edition, 1997.
[8] H. Tennekes and J.L. Lumley. A First Course in Turbulence. MIT Press, Cambridge, MA,
1972.
[9] Milton Van Dyke, editor. An Album of Fluid Motion. The Parabolic Press, Stanford, CA, 1982.
[10] Kenneth Wark. Thermodynamics. McGraw-Hill, New York, third edition, 1977.
[11] David C Wilcox. Turbulence Modeling for CFD. DCW Industries, Inc., La Ca nada, CA, 1993.

Anda mungkin juga menyukai