Sei sulla pagina 1di 13

ME 223

Laminar Boundary Layer on a


Flat Plate
1. Introduction
In this notebook we solve the Blasius equation for the laminar boundary layer on a flat plate. We also
find the temperature distribution in the thermal boundary layer associated with the flow. The basic background,
which was given in detail in class, is summarized here. The x-y coordinate system is chosen so that x is along
the plate, and y is perpendicular to the plate. The leading edge of the plate is at x = 0, y = 0. The velocity
components -- u in the x-direction and v in the y-direction -- are expressed in terms of a stream function y(x,y):
(1)
u =
!y
!y
, v = -
!y
!x
.
The fundamental equation which determines y is the x-component of the momentum equation in the boundary
layer approximation. For the flat plate there is no pressure gradient, and the boundary layer approximation to the
x-momentum equation takes the form
(2) u
!u
!x
+ v
!u
!y
= n
!
2
u
!y
2
,
where n is the kinematic viscosity. (Note the unfortunate coincidence in the Times font of italicized v and the
Greek letter nu.) By substituting the representations (1) into (2), we get for y the partial differential equation
(3)
!y
!y
!
2
y
!x !y
-
!y
!x
!
2
y
!y
2
= n
!
3
y
!y
3
.
At the wall (y = 0) both velocity components must vanish, and far away from the plate, the horizontal velocity
must approach the given free stream velocity u

. These conditions translate into the following conditions on y:


(4)
y(x,0) = 0 ,
!y
!y
(x,0) = 0 , and
!y
!y
u

as y .
The absence of a length scale (the plate is semi-infinite in length) suggests a similarity solution, as
originally used by Blasius. The solution has the form
(5)
y = nxu

f(h) , with h = y/d, and d = nx u

.
The scale d is comparable with the boundary layer thickness. This substitution into the equation for y leads to
the following nonlinear ordinary differential equation for f:
(6) f ''' +
1
2
f f
"
= 0 , with f(0) = 0, f

(0) = 0 , and f

1 as h .
This equation is called the Blasius equation. We solve it numerically in the next section. Once f is known, the
velocity components may be computed as
(7) u =
!y
!y
= u

HhL, and v = -
!y
!x
=
1
2
nu

x
8hf

HhL - f HhL< .
2. Solution of the Blasius Equation
A direct attack on the Blasius equation requires some kind of iteration such as a shooting method,
because it is a two-point boundary value problem. Fortunately, there is a reformulation of the problem that
avoids an iteration. We begin this reformulation by introducing a new dependent variable g(h):
(8) g(h) = af(ah) ,
where a is a positive parameter to be determined. As one may show, the equation for g is then the same as that
for f :
(9) g ''' +
1
2
gg
"
= 0 .
From the conditions on f, we have g(0) = 0, g

(0) = 0, and g

(h)= a
2
f

(ah) a
2
as h. Instead of imposing
the condition on g at , we impose g
!
(0) = 1. Then the problem for g is an initial value problem with initial
conditions
(10) g(0) = 0, g

(0) = 0 , g
"
(0) = 1 .
The parameter a does not appear in this formulation, and we can solve for g numerically. After solving for g, we
choose
(11)
a = g

HL ,
Then f will satisfy the proper condition at . With this choice of a, we obtain f as
(12) f(h) =
1
a
g(
h
a
) .
We carry this out now using Mathematica's NDSolve to do the numerical integration.
blasol = NDSolve@8g'''@hD + H1 2L g''@hD g@hD 0,
g@0D 0, g'@0D 0, g''@0D 1<, g, 8h, 0, 15<D;
Mathematica returns an answer in the form of an interpolating function which in most ways can be manipulated
as an ordinary function. We assign this interpolating function to a function we call blag[h]:
blag@h_D = g@hD . First@blasolD;
We check our initial conditions.
blag@0D
0.
blag'@0D
-2.71051 10
-20
2 newblas.nb
blag''@0D
1.
Our numerical integration to get blag was carried out to h = 15. As we see from the plot below, this is more than
we need. The velocity function blag levels off well before h = 10.
Plot@blag

@hD, 8h, 0, 10<,


AxesLabel 8"h", "g'HhL"<, PlotRange 80, 3<D
0 2 4 6 8 10
h
0.5
1.0
1.5
2.0
2.5
3.0
g'HhL
We can now determine a:
a = Sqrt@blag'@10DD
1.44409
Finally we map back to the original function f.
blaf@h_D := H1 aL * blag@h aD
We check to see that this function satisfies the conditions at zero and infinity.
blaf@0D
0.
blaf'@0D
-1.29975 10
-20
newblas.nb 3
blaf'@10D
1.
We see that blaf[h] is the desired solution of the Blasius equation.
3. Velocity Components
The horizontal velocity component is given by
xcomp@h_D := u* blaf'@hD
We plot the ratio of u to u

.
graphxcomp = PlotBblaf

@hD, 8h, 0, 6<, AxesLabel :"h", "


u
u

">,
ImageSize 360, PlotLabel "Scaled Horizontal Velocity"F
1 2 3 4 5 6
h
0.2
0.4
0.6
0.8
1.0
u
u

Scaled Horizontal Velocity


The vertical velocity component is given by
ycomp@x_, h_D :=
1
2
n u
x
yfunc@hD
where
4 newblas.nb
yfunc@h_D := h blaf'@hD - blaf@hD
Because blaf is asymptotic to h for large h, the leading terms in the above difference cancel out. If the first two
terms in the asymptotic expansion of blaf are h + c, then yfunc will approach -c as h. We let the numerical
solution tell us if this is correct.
Plot@yfunc@hD, 8h, 0, 10<, PlotRange 80, 2<, AxesLabel 8"h", "v"<,
PlotLabel "Scaled Vertical Velocity", ImageSize 400D
0 2 4 6 8 10
0.5
1.0
1.5
2.0
v
Scaled Vertical Velocity
We see indeed that there is a constant vertical velocity at the outer limit of the boundary layer. This is caused by
the retardation in the boundary layer, which in turn deflects some of the oncoming fluid upward. The constant
that yfunc approaches is
c = yfunc@10D
1.72079
Thus at the outer edge of the boundary layer,
(13) v =
1
2
n u

x
c

= u

1
2
1
Re
x
c

= 0.8604 (u

/ Re
x
) ,
where Re
x
= Hu

xL n is the Reynolds number based on the distance from the leading edge of the plate. Because
Re
x
is large, we see that the vertical velocity at the edge of the boundary layer is small.
newblas.nb 5
4. Shear Stress on the Plate
The shear stress on the plate is given by
(14)
t = m
!u
!y
y=0
=
m
d
u

f
"
(0) =
rmu

3
x
f
"
(0) .
We get f
!
(0) from the numerical solution:
c0 = blaf''@0D
0.332057
The local coefficient of friction C
f
Idenoted by C
f ,x
in the text) is the local shear stress divided by
1
2
ru

2
, and the average of this over a plate of length L is C
f
. These are given by
(15)
C
f
= 0.664 Re
x
-12
and C
f
= 1.328 Re
L
-12
, where Re
L
=
u
L
n
.
5. Thickness of the Boundary Layer
There is no sharp outer edge of the boundary layer. Rather it merges smoothly and gradually with the
outer free stream. However, there are several rational ways of defining a boundary layer thickness, and we
consider four of them here. The first two methods are based on the same idea -- namely, determining where the
horizontal velocity is nearly equal to the free stream velocity. The first thickness, which we call d
95
, is the y-
value at which the velocity is 0.95 of the free stream velocity. This happens for h = h
95
where
h95 = h . FindRoot@blaf'@hD 0.95, 8h, 3<D
3.91804
Thus d
95
= h
95
d = 3.92 nx u

. The second thickness, d


99
, corresponds to the velocity being equal to 0.99 of
the free stream velocity. This happens for h = h
99
, where
h99 = h . FindRoot@blaf'@hD 0.99, 8h, 3<D
4.90999
Thus d
99
= h
99
d = 4.91 nx u

. This is often approximated as d


99
= 5 nx u

.
The other two commonly used definitions of boundary layer thickness are more closely tied to the
physics of the flow. The displacement thickness (as discussed in ME 225) is a measure of the mass flux defect
in the boundary layer. It is given by d
d
= h
d
d , where
hd = NIntegrate@1 - blaf'@hD, 8h, 0, 10<D
1.72079
6 newblas.nb
Thus d
d
= h
d
d = 1.72 nx u

.
The last boundary layer thickness is the momentum thickness. As discussed in ME 225, it is a measure of the
momentum flux defect carried by the actual flow. It is given by d
m
= h
m
d, where
hm = NIntegrate@blaf'@hD * H1 - blaf'@hDL, 8h, 0, 10<D
0.664115
Thus d
m
= h
m
d = 0.66 nx u

.
6. Table of Values of the Blasius Function
We construct here a table of values of the Blasius function and its first two derivatives. We do this for h
running from 0 to 8 at intervals of 0.25.
newblas.nb 7
TableForm@Table@8PaddedForm@Chop@0.25 * nD, 84, 2<D,
PaddedForm@Chop@blaf@0.25 * nDD, 88, 4<D,
PaddedForm@Chop@blaf'@0.25 * nDD, 88, 4<D,
PaddedForm@Chop@blaf''@0.25 * nDD, 88, 4<D<,
8n, 0, 32<D, TableHeadings ->
8None, 8" h", " f@hD", " f

@hD", " f
!
@hD"<<D
h f@hD f

@hD f
!
@hD
0.00 0.0000 0.0000 0.3321
0.25 0.0104 0.0830 0.3319
0.50 0.0415 0.1659 0.3309
0.75 0.0933 0.2483 0.3282
1.00 0.1656 0.3298 0.3230
1.25 0.2580 0.4096 0.3146
1.50 0.3701 0.4868 0.3026
1.75 0.5011 0.5605 0.2866
2.00 0.6500 0.6298 0.2668
2.25 0.8156 0.6936 0.2434
2.50 0.9963 0.7513 0.2174
2.75 1.1906 0.8022 0.1897
3.00 1.3968 0.8460 0.1614
3.25 1.6131 0.8829 0.1337
3.50 1.8377 0.9130 0.1078
3.75 2.0691 0.9370 0.0844
4.00 2.3057 0.9555 0.0642
4.25 2.5464 0.9694 0.0474
4.50 2.7901 0.9795 0.0340
4.75 3.0360 0.9867 0.0236
5.00 3.2833 0.9915 0.0159
5.25 3.5316 0.9948 0.0104
5.50 3.7806 0.9969 0.0066
5.75 4.0300 0.9982 0.0040
6.00 4.2796 0.9990 0.0024
6.25 4.5294 0.9994 0.0014
6.50 4.7793 0.9997 0.0008
6.75 5.0293 0.9998 0.0004
7.00 5.2792 0.9999 0.0002
7.25 5.5292 1.0000 0.0001
7.50 5.7792 1.0000 0.0001
7.75 6.0292 1.0000 0.0000
8.00 6.2792 1.0000 0.0000
7. The Thermal Boundary Layer
As we showed in class, the similarity solution of Blasius for the velocity in the boundary layer may be
extended to include a calculation of the temperature distribution in the thermal boundary layer. We outline the
calculation here, and use Mathematica to calculate the heat transfer coefficient as a function of Prandtl number
and position along the plate.
8 newblas.nb
As we showed in class, the similarity solution of Blasius for the velocity in the boundary layer may be
extended to include a calculation of the temperature distribution in the thermal boundary layer. We outline the
calculation here, and use Mathematica to calculate the heat transfer coefficient as a function of Prandtl number
and position along the plate.
We start by introducting a dimensionless version q of the temperature T(x,y).
(16) q =
T -T
s
T

- T
s
.
Here T
s
is the surface temperature of the flat plate, and T

is the upstream temperature in the fluid. The energy


equation satisfied by q is
(17) u
!q
!x
+ v
!q
!y
= a
!
2
q
!x
2
+
!
2
q
!y
2
.
Here u and v are the x and y velocity components, and a is the thermal diffusivity. Using the usual arguments of
boundary layer theory, we may drop the second x-derivative of q compared with the second y-derivative. For u
and v we have the representation in terms of the Blasius function f(h), and we assume that q = q(h), where h is the
Blasius similarity variable given in equation (5). Using all of this information, we get the following equation for
q(h):
(18)
2
d
2
q
dh
2
+ Pr f HhL
dq
dh
= 0.
The Blasius function f(h) is known from our previous integration. The equation for q is linear, and it is first
order in
dq
dh
. In solving the equation, we have the following endpoint conditions:
(19) q(0) = 0 and q() = 1.
By treating equation (18) as a first-order equation in
dq
dh
we get the following solution:
(20)
dq
dh
= C
-
Pr
2
F HhL
, where F HhL =

0
h
f HsL s .
We integrate once more to get q, using the fact that q vanishes at h = 0. The result is
(21) q HhL = C

0
h

-
Pr
2
F HrL
r .
We can evaluate C by using the condition that q should go to 1 as h goes to infinity. This gives
(22)
C =
1

-
Pr
2
F HrL
r .
Substituting the expression for C into (21), we get the answer for q:
(23) q HhL =

0
h

-
Pr
2
F HrL
r

-
Pr
2
F HrL
r
.
While this is a valid expression, it is somewhat inconvenient to work with because it involves iterated numerical
integrations - first to get F from f, and then to get q from F. We can do better. From the Blasius equation (6) we
get
newblas.nb 9
(24) f = -2 f ''' f '' , hence f = -2 ln H f "L ' , so F HrL = -2 ln H f ' ' HrL f ' ' H0LL.
Then
(25) Exp@-HPr 2L F HrLD = H f ' ' HrL f ' ' H0LL
Pr
.
Substituting this into (23) we get
(26) q HhL =

0
h
H f ' ' HrL f ' ' H0LL
Pr
r

H f ' ' HrL f ' ' H0LL


Pr
r
We can use this to calculate the local heat transfer coefficient h (denoted by h
x
in the text). As shown in
class, it is given by
(27) h = k
u

n x
dq
dh
H0L.
We finish the calculation of h by using (26) to calculate
dq
dh
H0L. The result is
(28)
dq
dh
H0L =
1

H f ' ' HrL f ' ' H0LL


Pr
r
.
Now we define this for Mathematica. We call the integral denom.
denom@Pr_D := ReANIntegrateAHblaf''@rD blaf''@0DL
Pr
, 8r, 0, 10<EE
Finally, we define qd as
dq
dh
H0L.
qd@Pr_D := 1 denom@PrD;
Let's check this for the one case where we know the answer. When Pr = 1, we know that q =
df
dh
, and from that
we know that qd = 0.3321.
qd@1.0D
0.332057
The answer looks good. We construct a table of values of qd versus Prandtl number. We also include in the
table the values of 0.332 Pr
13
, which the book recommends as an approximation for qd for Pr > 0.6. We will
construct the table for Pr varying from 0.2 to 2, in increments of 0.1.
qdbook@Pr_D := 0.332 Pr
13
10 newblas.nb
TableFormB
Table@8PaddedForm@0.1 * n, 86, 3<D, PaddedForm@qd@0.1 * nD, 86, 3<D,
PaddedForm@qdbook@0.1 * nD, 86, 3<D<, 8n, 2, 20<D, TableHeadings ->
:None, :" Pr", "Exact Slope
dq
dh
H0L ", " 0.332 Pr
13
">>F
Pr Exact Slope
dq
dh
H0L 0.332 Pr
13
0.200 0.185 0.194
0.300 0.215 0.222
0.400 0.239 0.245
0.500 0.259 0.264
0.600 0.277 0.280
0.700 0.293 0.295
0.800 0.307 0.308
0.900 0.320 0.321
1.000 0.332 0.332
1.100 0.343 0.343
1.200 0.354 0.353
1.300 0.364 0.362
1.400 0.373 0.371
1.500 0.382 0.380
1.600 0.391 0.388
1.700 0.399 0.396
1.800 0.407 0.404
1.900 0.415 0.411
2.000 0.422 0.418
We see that the approximation 0.332 Pr
13
is good for the range of Prandtl numbers in the table, even
below the value of 0.6. The best fit is in the vicinity of Pr = 1, which we would expect because the approxima-
tion is exact at Pr = 1.
We conclude this notebook by calculating the thickness of the thermal boundary layer as a function of
the Prandtl number. We have seen earlier that the thickness of the velocity boundary layer (the d
99
thickness)
corrresponds to h = 4.90999. We calculate here the value of h at which the temperature has a value of 0.99 of
the free stream temperature, and then we will divide that value of h by 4.90999 to get the thickness of the
thermal boundary layer relative to the thickness of the velocity boundary layer. We start with equation (26) for
q(h).
This is the expression which we will use to calculate q and to find the thickness of the thermal boundary layer.
We define the expressions for Mathematica now.
num@h_, Pr_D := NIntegrate@Hblaf''@rD blaf''@0DL^Pr, 8r, 0, h<D;
denom@Pr_D := NIntegrate@Hblaf''@rD blaf''@0DL^Pr, 8r, 0, 10<D;
theta@h_, Pr_D := Re@num@h, PrD denom@PrDD;
We take the real part of the answer for theta (Re) because Mathematica sometimes returns spurious small
complex parts when evaluating the integrals that define theta. We test the definitions by evaluating theta for Pr
= 1 and h = 4.90999. In that case we should get theta = 0.99.
newblas.nb 11
We take the real part of the answer for theta (Re) because Mathematica sometimes returns spurious small
complex parts when evaluating the integrals that define theta. We test the definitions by evaluating theta for Pr
= 1 and h = 4.90999. In that case we should get theta = 0.99.
theta@4.90999, 1.0D
0.99
It checks. Now we define a function that gives the thermal boundary layer thickness as a function of Pr. We use
the method of bisection to find the root.
therm@Pr_D := ModuleA8sleft, sright, smid<, sleft = 0.1; sright = 10.0;
smid = H0.5L * Hsleft + srightL; WhileAIsright - sleft > 10
-5
M,
HIf@Htheta@smid, PrD - 0.99 < 0L, Hsleft = smidL, Hsright = smidLD;
smid = H0.5L * Hsleft + srightLLE; smidE
We try this for Pr = 1.
therm@1.0D
4.90999
This is correct. We now define the thermal boundary layer thickness relative to the velocity boundary layer
thickness.
reltherm@Pr_D := therm@PrD 4.90999;
This should give 1 for Pr =1:
reltherm@1.0D
1.
We use this to construct a plot of the relative thickness for Pr running from 0.2 to 2.
12 newblas.nb
Plot@reltherm@PrD, 8Pr, 0.2, 2<, PlotRange 880, 2<, 80, 2<<,
AxesLabel 8"Pr", "d
therm
d
vel
"<, ImageSize 350D
0.0 0.5 1.0 1.5 2.0
Pr 0.0
0.5
1.0
1.5
2.0
d
therm
d
vel
We see that as the Prandtl number increases, the thermal boundary layer becomes thinner than the velocity
boundary layer. This makes sense physically, because the Prandtl number is the ratio of the viscosity
(momentum diffusivity) to the thermal diffusivity, and the smaller the diffusivity, the thinner the corresponding
layer. For large Prandtl numbers (water, or especially oil and glycerine) the velocity boundary layer is thicker
than the thermal boundary layer. For small Prandtl numbers ( for example for liquid metals), the thermal
boundary layer is thicker than the velocity boundary layer. For very small or very large Prandtl numbers, the
calculations as defined above break down, and we have to use approximations based on the size of the Prandtl
number. Some asymptotic results that can be shown by scale analysis are given below. The dimensionless
constants C
1
and C
2
can be determined by a detailed analysis.
(29) d
therm
d
vel
= C
1
Pr
-12
for Pr << 1 , and d
therm
d
vel
= C
2
Pr
-13
for Pr >> 1 .
newblas.nb 13

Potrebbero piacerti anche