Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
where
thus
1
T
(7.1)
t T
u ( x , y, z, t ) dt
(7.2)
u 0
(7.3)
It is important to remember that turbulence flows are three dimensional. Thus, even if u is twodimensional (u u ( x, y ) ), the velocity fluctuations in the x, y, and z directions (u , v, w)
are not zero. Moreover, although the average of each flow property fluctuation is zero, averages
of products of various fluctuations are not zero, i.e. u v 0, u w 0 .
2
The boundary layer equations in two-dimensions for incompressible or constant density
flow can be written as
u
v
0
x
y
(7.4)
xy
u
u
dP
u
u
u
, where xy
x
y
dx
y
y
(7.5)
xy
u 2
(uv )
dP
x
y
dx
y
(7.6)
Setting
(7.7)
Substituting into Eqs. (7.4) and (7.5), and averaging, the following equations that govern the
mean flow result.
u
v
0
x
y
(7.8)
dP
[(u ) 2 (u ) 2 ]
u v u v
xy
x
y
dx
y
(7.9)
u 2
dP
(u ) 2
uv
( xy u v)
x
y
dx
x
y
(7.10)
or
u 2
and
u v
are called the turbulent normal and shear stresses, respectively. In boundary
layers, the normal stresses are usually much less than shear stresses. Moreover
x
y
Therefore, the term involving normal turbulent stress is a small term and will be neglected.
Expanding the left hand side of Eq. (7.10) and substituting for xy , Eq. (7.10) reduces to
u
u
dP
u
v
dx y
u
y u v
(7.11)
The resulting turbulent boundary layer equations cannot be solved without a closure
assumption which relates the turbulent shear stress to the mean flow variables. As a first attempt
in this direction and using the analogy of laminar flow, Boussinesq proposed that
u v t
u
y
(7.12)
where t is the turbulent shear viscosity or eddy viscosity. There is a major difference between
and t . is a property of the fluid and is a function of the temperature, while t is a
function of the flow and its value depend on the initial and boundary conditions of the problem
under consideration.
The first theory for predicting t is the mixing length theory developed by Prandtl. He
used a combination of dimensional analysis and experiments on a flat plate to develop his theory.
For laminar flows on a flat plate all velocity profiles reduce to a single curve given by u/ue =
f(). Prandtl discovered that such similarity does not exist in a turbulent boundary and he had to
consider two regions, one close to the wall, the inner region, and one away from the wall, the
outer region. He reasoned that, in the near region, the velocity u should depend on the y location,
wall shear stress, shear viscosity and density but does not depend on the freestream conditions.
On the other hand, the velocity in the outer region is independent of the direct effects of
viscosity. Noting that w can be written as
w u *2
(7.13)
where u* is called the friction velocity and using the methods of dimensional analysis, one can
show that
u y u *
f , /
u*
(7.14)
ue u
g ( y / )
u*
(7.15)
where u e is the edge velocity, and is the boundary layer thickness, in the outer region. It
should be noted that dimensional analysis does not provide expressions for the function f and g.
Equation (7.14) is referred to as the law of the wall and Eq. (7.15) is referred to as the velocity
defect law or Clauser defect law.
Prandtls measurement in the near wall region produced velocity profiles similar to those
shown in Fig. 7.1. It is seen from Fig. 7.1 that it consists of three distinct regions: the viscous
4
sublayer, the log layer and the defect layer. The viscous sublayer lies in the range 0 y+ 30-50,
where y+ = yu*/. In this region, the inertia terms are negligible. Moreover, the turbulent
fluctuations are negligible with the result that the turbulent shear stress is much less than the
laminar shear stress. Within the viscous sublayer there is a region very near the wall (y +<5),
called the laminar sublayer, where u+=y+. The width of log layer is Reynolds number dependent
and typically lies in the range 30-50<y+ < 350-500 and .01-.02<y/<0.1-0.2. In this region, the
inertia terms are also negligible but the turbulent shear stress dominates the laminar shear stress.
Finally, the defect layer lies between the log layer and the edge of the boundary layer.
u u*
u
y y u * /
Figure 7.1. Typical velocity profile for flat plate turbulent-boundary layer
u
C1 ln( yu * / ) C2
u*
(7.16)
where C1 and C2 are constants. Now Prandtls task reduced to finding an expression for t
that reproduces Eq. (7.16). For a flat plate in the near region, where the inertia terms are
negligible, Eqs. (7.11) and (7.12) reduce to
d
dy
du
( t ) dy 0
or
( t )
du
w
dy
(7.17)
du
w u *2
dy
(7.18)
with
du
,
dy
mix y
(7.19)
where mix is called the mixing length, and = 0.4 is the Karman constant. Using Eq. (7.19)
into Eq. (7.15),
2mix
du du
u *2
dy dy
(7.20)
du
du
u*
dy
(7.21)
u 1
ln y const
u*
1 y u*
l n C , C 5 .0
(7.22)
As may be seen from Fig. (7.1) and Eq. (7.22), the mixing length expression indicated in
(7.19) can not produce the entire velocity field. As a result, two different approaches are
employed in further developing turbulent boundary layer theory: one suited for use in
conjunction with integral methods while the other is suited for use with differential methods
which solve the turbulent boundary layer equations numerically
In the first approach Coles developed the composite law of the wall-law of the wake,
which describes the velocity profile in the log-law region and outer region. This law can be
written as
u 1 yu* 2I 2 y
ln C sin
u* 2
(7.23)
where II is Coles wake-strength parameter and is the boundary layer thickness. The quantity
II varies with the pressure gradient and can be estimated using the relation
II 0.6 0.51 T 0.01 2
T
(7.24)
where
* dP
w dx
(7.25)
* 1 2II
* 2 6.4II 6 II 2
2 2
(7.26)
(7.27)
The local skin friction coefficient, C f , may be related to II and local Reynolds number
Re ue / by evaluating the wall-wake law, i.e., Eq. (7.23), at the edge of the boundary layer
ue
1
(2 / C f )1/ 2 ln(Re / ) C 2II /
u*
(7.28)
In order to calculate the flow properties as a function of x, use has to be made of the
momentum integral equation. The momentum integral equation for turbulent flows has the same
form as that developed for laminar flow, i.e.
7
Cf
d
1 due
(2 *)
dx ue dx
2
(7.29)
Although the momentum integral equation for laminar and turbulent flows are the same, there is
no length scale in turbulent boundary layers that leads to the collapse of the complete set of
velocity profiles into a single set as was the case for laminar flows. Therefore, relations amongst
, * and C f depend on the assumed velocity profile. And, as may be seen from Eq. (7.23),
their dependence on x will also depend on the assumed velocity profile. Thus, if Eq. (7.23) is
chosen as the velocity profile, * and can be expressed in terms of and II from Eqs. (7.27),
and (7.27) and equation (7.28) provides an implicit relation for C f in terms of . Because of
this, a numerical solution of Eq. (7.29) is required. Such a solution will yield = (x) and this,
in turn, will provide *(x), (x), and C f (x) .
A simple example of historical interest is due to Prandtl who assumed the velocity profile
over a flat plate to be
u
y
ue
1/ n
n7
(7.30)
Above relation is not valid in the near wall region and thus cannot provide information on the
skin friction coefficient. However, it can provide a relationship between and x which can be
used to calculate C f (x) .
For a flat plate, Eq. (7.29) reduces to
d
Cf /2
dx
and give
*
1/ 8 ,
For
Cf
(7.31)
/ 7 / 72
(7.32)
(7.33)
Substituting Eqs. (7.32) and (7.33) into Eq. (7.31) and integrating
0.375 / Re1x/ 5
x
(7.34)
Although the above development is crude, it nevertheless indicates that a turbulent boundary
layer grows much faster than a laminar boundary layer where / x 5 / Re1x/ 2 .
8
The approach based on the solution of the turbulent boundary layer is discussed next. As
is seen from Fig. 7.1, there is a need to develop an expression for t which is valid in the
laminar sublayer region. Moreover, another expression is required for the outer region.
The near wall modification of t is due to van Driest. He modified the expression
derived in the log layer by a damping factor, with the result that the mixing length is now given
by
yu *
(7.35)
A for flat plates is a constant and is chosen so as to reproduce the velocity profile in the near
wall region and log-law region. Since in the near wall region the laminar shear stress, , is not
negligible, Eq. (7.17) gives
du
y
w
2 y 2 1 exp
dy
du du
dy dy
(7.36)
w u *2 ,
yu*
(7.37)
y
1
( y ) 1 exp
dy
du / dy .
du
dy
1 4 2
2 2
du
dy
du / dy
(7.38)
gives
(7.39)
where
y 1 exp y / A
1 4 2
1 4 2
(7.40)
y 0 .
To
dy
1
Note that
du
dy
1 at
y 0 .
2
1 4 2
A 26
dP/ dx
A 26 1 y
2
u*
(7.41)
To fit
1/ 2
(7.42)
One of the widely used expressions for the eddy viscosity in the outer region is that
developed by Clauser
t outer 0.018 ue *
(7.43)
It has been found by Corrsin and Kistler and Klebanoff that approaching the freestream
from within a turbulent boundary layer that the flow is not always turbulent; it is intermittent,
i.e., it fluctuates between laminar and turbulent To account for this behavior, Eq. (7.43) is
multiplied by
y
Fkleb 1 5.5
6 1
(7.44)
While carrying out numerical calculations of the turbulent boundary layer, one starts by
the near wall expression for t and switches to the outer region expression when the two
quantities are equal.
Since t 0 at y = 0, w retains its traditional definition
w
du
dy y 0
(7.45)
Above equation does not mean that the wall shear stress is the same for laminar and turbulent
du
flows. The turbulent shear stress is higher because dy y 0 is higher in turbulent boundary
layers.
10
The above turbulence model, where t is given by algebraic expressions works well as
long as the flow is attached. More complicated theories are needed in the presence of flow
separation.