Sunteți pe pagina 1din 120

FLUID DYNAMICS

MAM 3040W, MODULE 3FD, 2010


Department of Mathematics and Applied Mathematics,
University of Cape Town
By
T1 T 111 (H1^O//
Department of Mathematics and Applied Mathematics
.and.
Center for Research in Computational & Applied Mechanics
August 4, 2010
Contents
1 Introduction 3
1.1 Denition of a uid . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Flow Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Streamlines . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Pathlines . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.3 Streaklines . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Orthogonal Curvilinear Coordinates . . . . . . . . . . . . . . . . . 8
2 Governing Equations I 12
2.1 Conservation of mass . . . . . . . . . . . . . . . . . . . . . . . . . 13
3 Governing Equations II 17
3.1 Conservation of momentum . . . . . . . . . . . . . . . . . . . . . 17
3.1.1 Newtonian Fluids . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.2 Deformation law for a Newtonian uid . . . . . . . . . . . 21
3.1.3 The Navier-Stokes equations . . . . . . . . . . . . . . . . . 25
4 Governing Equations III 26
4.1 Conservation of energy . . . . . . . . . . . . . . . . . . . . . . . . 26
2
CONTENTS 3
4.1.1 Energy equation . . . . . . . . . . . . . . . . . . . . . . . . 30
4.1.2 Mechanical energy equation . . . . . . . . . . . . . . . . . 31
4.1.3 Thermal energy equation . . . . . . . . . . . . . . . . . . . 31
4.1.4 Non-conservative energy equation . . . . . . . . . . . . . . 32
5 Dimensionless Equations 33
5.1 Incompressible ow . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.1.1 Parameters for Natural convection ow . . . . . . . . . . . 35
5.2 Compressible ow . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
6 Steady Couette Flows 39
6.1 Flow between a moving and a xed plate . . . . . . . . . . . . . . 39
6.2 Axially moving concentric cylinders . . . . . . . . . . . . . . . . . 41
6.3 Flow between rotating concentric cylinders . . . . . . . . . . . . . 42
7 Steady Poiseuille Flows 45
7.1 Flow Between Parallel Planes . . . . . . . . . . . . . . . . . . . . 45
7.1.1 Non-isothermal case . . . . . . . . . . . . . . . . . . . . . 47
7.2 Flow in through a duct . . . . . . . . . . . . . . . . . . . . . . . . 48
7.2.1 Flow in a circular pipe: Hagen-Poiseuille ow . . . . . . . 48
7.2.2 Temperature distribution in Hagen-Poiseuille ow . . . . . 50
8 Combined Couette-Poiseuille ow 52
9 Unsteady Flows With Moving Boundaries 55
9.1 Suddenly accelerated at plate . . . . . . . . . . . . . . . . . . . . 56
9.1.1 Unsteady ow between two innite plates . . . . . . . . . 57
CONTENTS 4
10 Flows With Suction and Injection 60
10.1 Uniform suction on a plane . . . . . . . . . . . . . . . . . . . . . . 60
10.2 Flow between plates with bottom injection and top suction . . . . 61
11 Vorticity & Bernoullis Equation 64
11.1 Vorticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
11.1.1 Velocity potential . . . . . . . . . . . . . . . . . . . . . . . 64
11.2 Bernoullis equation . . . . . . . . . . . . . . . . . . . . . . . . . . 65
11.2.1 Boundary conditions for potential ows . . . . . . . . . . . 66
12 2-D Flows & the Stream Function 67
12.1 Stream function & streamlines . . . . . . . . . . . . . . . . . . . . 67
12.2 2-D momentum equations . . . . . . . . . . . . . . . . . . . . . . 68
13 Compressible Flow 70
13.1 Introduction to thermodynamics & compressible ow . . . . . . . 70
13.1.1 The rst law of thermodynamics . . . . . . . . . . . . . . 71
13.2 Specic heats of a perfect gas . . . . . . . . . . . . . . . . . . . . 71
13.2.1 Functions of state . . . . . . . . . . . . . . . . . . . . . . . 73
13.2.2 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
13.3 Flow of an inviscid gas . . . . . . . . . . . . . . . . . . . . . . . . 77
13.3.1 1-D steady ow in a nozzle . . . . . . . . . . . . . . . . . . 78
14 Sound Waves 81
14.1 Speed of sound in a gas . . . . . . . . . . . . . . . . . . . . . . . . 81
14.2 Mach waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
CONTENTS 5
15 Boundary layer theory 91
15.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
15.1.1 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . 91
15.2 Boundary layer thickness . . . . . . . . . . . . . . . . . . . . . . . 92
15.2.1 Displacement thickness . . . . . . . . . . . . . . . . . . . . 93
15.3 Boundary layer equations . . . . . . . . . . . . . . . . . . . . . . 94
16 Exact Solutions 98
16.1 Blasius solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
16.2 Flat plate with varying outer ow . . . . . . . . . . . . . . . . . . 106
17 Similarity Solutions 108
17.1 Falkner-Skan equation . . . . . . . . . . . . . . . . . . . . . . . . 108
17.2 Stagnation point ow . . . . . . . . . . . . . . . . . . . . . . . . . 112
17.3 Flow over a wedge . . . . . . . . . . . . . . . . . . . . . . . . . . 113
18 Approximate Solutions 114
18.1 The momentum integral . . . . . . . . . . . . . . . . . . . . . . . 114
18.2 von Karman - Polhausen method . . . . . . . . . . . . . . . . . . 117
Lecture 1
Introduction
1.1 Denition of a uid
Denition 1.1 Continuum Hypothesis: We can associate with any innites-
imal (but nonzero) volume of uid those macroscopic properties (e.g. velocity,
density, pressure, temperature) that we normally associate with the bulk uid.
Hence, despite the fact the uids actually consist of billions of individual molecules
or atoms, we will proceed to treat uids as continuous media and thus will only
be interested not in the behavior of individual constituent molecules/atoms but
in those macroscopic properties of the bulk uid.
In uid mechanics, the distinction between uids and nonuids (solids) is the
ability of solids to resist deformation under an applied shear force (tangential
force per unit area) and remain at rest while uids cannot. For example, if we
were to impose a shear stress on a block of wood or steel, the block would not
begin to change shape until an extreme amount of stress has been applied. On
the other hand, if we apply a shear stress to a uid, e.g. water, we observe that
no matter how small the stress is, the uid deforms, see illustration in gure 1.1.
6
LECTURE 1. INTRODUCTION 7
- -

FLUID
SOLID
Shear Stress
Shear Stress
Shear Stress Shear Stress
1
Figure 1.1: Distinction between uids and nonuids
Denition 1.2 A uid is any substance that deforms continuously when sub-
jected to a shear stress, no matter how small.
We remark that the continuous deformation is only partially obtainable in solids.
Solids will deform (and change shape) until either a balance is reached between
the applied forces and internal forces or the solid breaks up end of continuous
medium! Fluids on the other hand will deform continuously (and keep changing
shape) under the action of a shearing force.
Remark 1.1 This distinction between uids and solids is however not completely
clear-cut, when one considers for example uids like toothpaste, ketchup, may-
onnaise etc. Such uids are classied as non-Newtonian uids. We will however
limit our attention to Newtonian uids which conform nicely with the given de-
nition. Thus, unless otherwise stated, the term uid hereafter will be synonymous
with Newtonian uid.
LECTURE 1. INTRODUCTION 8
1.2 Flow Visualization
The uid velocity is the basic description of how a uid moves in space and
time, but in order to visualize the ow pattern it is useful to describe some
other properties of the ow. these denitions correspond to the most popular
experimental methods of visualizing uid ow.
1.2.1 Streamlines
Denition 1.3 A streamline, at any given time (t = constant) is a curve
which has the same direction as the velocity vector at every point in space. Equiv-
alently, the tangent to the curve at each point in space is the direction of the
velocity vector at that point.
The ow velocity is thus tangential to a streamline at all points along it. Mathe-
matically, suppose the velocity vector in cartesian coordinates x = (x, y, z) with
corresponding velocity components u, v, w is given by
v(t, x) = (u, v, w).
From elementary physics, the velocity at t = s is
u(s, x(s)) =
dx
ds
, v(s, x(s)) =
dy
ds
, and w(s, x(s)) =
dz
ds
. (1.1)
For two-dimensional ows (say w 0 hence dz = 0) we have z = constant
thus the streamlines will be parallel to the x-y plane. In the general 3-D case,
eliminating ds in equations (1.1) leads to
dx
u
=
dy
v
=
dz
w
. (1.2)
Hence if the velocity eld is known, the streamlines would be obtained by solving
the dierential equations (1.2) or (1.3) for each constant value of t:
dy
dx
=
v
u

t=const
,
dz
dx
=
w
u

t=const
,
dz
dy
=
w
v

t=const
. (1.3)
LECTURE 1. INTRODUCTION 9
Remark 1.2 Streamlines display a snapshot of the ow eld at a given instant in
time with each streamline starting from a dierent selected point in the ow eld.
Thus, in time-dependent ow, a ow visualization based on streamlines will be
constantly changing.
Since streamlines are in the direction of the velocity, they cannot cross. If they
were to cross at any point in a ow eld, there would have to be two dierent
velocity vectors at that point in the ow eld and hence the velocity would not be
well dened.
1.2.2 Pathlines
Denition 1.4 A pathline or particle path is the trajectory of an individual
uid element.
We obtain the particle paths as a solution to the dierential equations (1.4):
dx
dt
= u(t, x),
dy
dt
= v(t, x), and
dz
dt
= w(t, x). (1.4)
subject to a specied initial condition x(t
0
) = x
0
. In a steady ow (velocity
independent of time) the pathlines and streamlines coincide.
1.2.3 Streaklines
Denition 1.5 A streakline is the locus of all the uid elements that have
previously passed through a given point.
A streakline is thus the set of points at some xed time t = constant consisting of
uid elements that, at some previous time

t, have all passed through a common
point x. We obtain the streaklines from the dierential equations for particle
paths (1.4) subject to the relevant initial conditions. Once again, in steady ow
streaklines, streamlines an pathlines all coincide.
LECTURE 1. INTRODUCTION 10
Example 1.1 Obtain and describe the streamlines & pathlines for the 2D time
dependent ow:
v(t, x, y) = u
0
_
_
_
_
_
_
_
1
0
t
_
_
_
_
_
_
_
, u
0
= constant.
Use t
0
= 0 and x(t
0
) = (x
0
, 0, y
0
).
Also, describe the streaklines that have passed through (i) the point x(t
0
) =
(x
0
, 0, y
0
), & (ii) the point x(

t) = ( x, y, z).
Streamlines:
At each time t = s the streamlines are found from the dierential equations:
dz
dx
= s, dy = 0.
Thus the streamlines are parallel to the x-z plane and are described by the equa-
tions:
z = s x + c
1
, c
1
= arbitrary constant.
At each time t = s, the streamlines are then straight line with slope s and each
value of c
1
denes a dierent streamline
Pathlines:
We need to solve
dx
dt
= u
0
,
dy
dt
= 0, and
dz
dt
= u
0
t,
_
_
_
_
_
_
_
x(0)
y(0)
z(0)
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
x
0
0
z
0
_
_
_
_
_
_
_
.
The spatial positions of the uid particles thus evolve in time according to:
x = x
0
+ u
0
t, z = z
0
+
1
2
u
0
t
2
, y = 0.
LECTURE 1. INTRODUCTION 11
Eliminating t gives
z = z
0
+
1
2u
0
(x x
0
)
2
, y = 0.
Thus the pathlines are parabolas in the x-z plane.
Streaklines:
(i) These are exactly the same parabolas as the pathlines above:
z = z
0
+
1
2u
0
(x x
0
)
2
, y = 0.
(ii) Similarly these are parabolas located in the plane y = y:
x = x + u
0
(t

t), z = z +
1
2
u
0
(t
2

t
2
), y = y,
hence
z = z +
1
2u
0
(x x)((x x + 2u
0

t), y = y.
1.3 Orthogonal Curvilinear Coordinates
Suppose we write x
i
, i = 1, 2, 3 to represent the cartesian coordinates and y
i
to
denote coordinates in any other system. If a unique representation can be found
such that
x
i
= x
i
(y
1
, y
2
, y
3
), y
i
= y
i
(x
1
, x
2
, x
3
), i = 1, 2, 3, (1.5)
then y
i
are called curvilinear coordinates. Now consider the metric tensor,
g
ij
=
3

k=1
x
k
y
i
x
k
y
j
. (1.6)
If, further to condition (1.5), only the diagonal terms g
ii
are non-zero and g
ij
=
0, i ,= j, then y
i
are called orthogonal curvilinear coordinates. In this case it is
convenient to dene:
h
2
i
:= g
ii
=
3

k=1
_
x
k
y
i
_
2
,
where h
i
are called scale factors.
LECTURE 1. INTRODUCTION 12
Example 1.2 For the cylindrical polar coordinates (r, , z), we have y
1
= r, y
2
=
and y
3
= z and it immediately follows that
h
1
= 1, h
2
= r, and h
3
= 1.
Similarly, for the spherical coordinates (, , ), we have y
1
= , y
2
= and
y
3
= , where x = sin cos , y = sin sin and z = cos it follows that
=

x
2
+y
2
+z
2
, = tan
1
(y/x), = cos
1
(z/

x
2
+ y
2
+ z
2
) and that
h
1
= 1, h
2
= sin , and h
3
= .
The equations of uid ow are usually derived in vector/tensor form so that they
are independent of any particular coordinate system. To recast these equations in
any particular orthogonal curvilinear coordinate system, we utilize the following
denitions of div, grad, curl and Laplacian based on the scale factors.
:=
3

i=1
1
h
i

y
i
e
i
, (1.7)
div v :=
1
h
1
h
2
h
3
3

i=1

y
i
_
h
1
h
2
h
3
h
i
v
i
_
, (1.8)
v :=
1
h
1
h
2
h
3

h
1
e
1
h
2
e
2
h
3
e
3

y
1

y
2

y
2
h
1
v
1
h
2
v
2
h
3
v
3

, (1.9)

2
:=
1
h
1
h
2
h
3
3

i=1

y
i
_
h
1
h
2
h
3
h
2
i

y
i
_
, (1.10)
where e
i
are the unit vectors in the corresponding coordinate directions, v
i
are
components of the vector v and is a scalar function.
The convective and viscous terms respectively dened by:
(v )v, and
2
v,
LECTURE 1. INTRODUCTION 13
are encountered in the equations of uid mechanics.
Exercise 1.1 Use the vector identities
(v )v =
[v[
2
2
v (v),
and

2
v = ( div v) (v)
to write out the components of the vectors v v and
2
v in polar coordinates.
Note that in cartesian coordinates we simply have
(v )v =
_
_
_
_
_
_
_
_
_
u
u
x
+ v
u
y
+ w
u
z
u
v
x
+ v
v
y
+w
v
z
u
w
x
+v
w
y
+ w
w
z
_
_
_
_
_
_
_
_
_
, and
2
v =
_
_
_
_
_
_
_

2
u

2
v

2
w
_
_
_
_
_
_
_
,
where we have written v = (u, v, w).
Exercise 1.2
Consider the velocity eld
v(t, x) =
x
1 + t
e
1
+y e
2
.
1. Determine and describe the particle paths.
2. Determine and describe the streamlines.
Exercise 1.3
1. Find the parametric equations, x = x(t) & y = y(t), of the particle paths
for the ow:
v = U
0
y i +U
0
(bt x) j, U
0
= constant, b = constant,
passing through the point (x
0
, y
0
) at t = 0.
LECTURE 1. INTRODUCTION 14
2. Find the streamlines for the ow,
v = a cos(t)i + a sin(t)j,
passing through the origin at s = 0.
Lecture 2
Governing Equations I
The fundamental equations of uid mechanics are derived from the basic conser-
vation laws for physical systems, namely:
Conservation of mass (continuity)
Conservation of momentum (Newtons second law)
Conservation of energy (rst law of thermodynamics)
In the subsequent analysis, we will utilize the particle, material or substantive
derivative
D
Dt
=

t
+v
which is appropriate for the so called Eulerian description of uid ow.
We derive the equations governing 3-D, unsteady, compressible and viscous uid
ow. These equations are called the Navier-Stokes equations and were developed
by Navier in 1831, and more rigorously be Stokes in 1845. These equations still
stand to date with no modications.
In three dimensions, we have six ow properties that are unknowns: the three
velocity components u, v, w; density ; temperature T; and pressure p. Therefore,
15
LECTURE 2. GOVERNING EQUATIONS I 16
we need six equations linking them. One of these six equations is the equation
of state, given by
p = p(, T). (2.1)
At moderate temperatures that arise in subsonic and supersonic ows without
chemical reactions, this equation of state may be simplied to the following form:
p = RT. (2.2)
Here R is a gas constant, given by R/M, where R is the universal gas constant
and M is the molecular weight of the gas (or the gas mixture). For air, the gas
constant is given by R = 2817 Joules/kg/

K.
The other ve equations are:
a) Conservation of mass, known as continuity,
b) Conservation of u- momentum,
c) Conservation of v- momentum,
d) Conservation of w- momentum, and
e) Conservation of energy.
2.1 Conservation of mass
We consider a small control volume (CV) of (horizontal) length x, (vertical)
width y, (height z perpendicular to the plane of the paper).
LECTURE 2. GOVERNING EQUATIONS I 17
6
-
+
x
y
z
(x, y, z)
x
y
z
Figure 2.1: 3-D control volume
The principle of conservation of mass states that
The rate at which mass increases within the control volume = The rate at which
mass enters the control volume through its six boundaries
Let be the average density of the uid within the control volume. Then,
Mass of uid within CV = Density of uid Volume of CV
= x y z (2.3)
Time rate of change of mass within CV =

t
( x y z)
= xyz

t
(2.4)
LECTURE 2. GOVERNING EQUATIONS I 18
Next, consider the rate at which mass enters through the six faces of the CV.
On each face, this equals the product of the density, the area of the face and
the velocity perpendicular to that face (check that this products has units of
mass/time). If we use the subscripts i and o to represent inlet and outlet quan-
tities respectively, then we have:
x-direction;
(uyz)
i
(uyz)
o
,
y-direction;
(vxz)
i
(vxz)
o
,
z-direction;
(wxy)
i
(wxy)
o
.
Summing up the contributions from the six faces, and equating the result to the
time rate of change of mass within the CV, we get

t
=
(u)
i
(u)
o
x
+
(v)
i
(v)
o
y
+
(w)
i
(w)
o
z
. (2.5)
Now, consider the limits of the above equation as x, y and z tend to zero
(keeping in mind, say that, (u)
i
(u)
o
:= (u)
x
(u)
x+x
, etc.) Applying
these limits, and bringing all the terms to the left hand side, we get

t
+
(u)
x
+
(v)
y
+
(w)
z
= 0. (2.6)
The above equation, in vector form is given by:

t
+ div (v) = 0, (2.7)
where v = (u, v, w) is the velocity vector. The vector form is more useful than
it would rst appear. If we want to derive the continuity equation in another
coordinate system such as the polar, cylindrical or spherical coordinate system,
LECTURE 2. GOVERNING EQUATIONS I 19
all we need to know is (a) look up the Del operator in that system, (b) look up
the rules for the dot product of Del operator and a vector in that system, (c)
perform the dot product. Note that the rules (a - c) depend on the scale factors
h
i
for the particular coordinate system as described in Lecture 1.
The most general form of the continuity equation for a compressible uid is
D
Dt
+ div v = 0, (2.8)
which is simply a restatement of equation (2.7). If the density is constant
(incompressible ow), the continuity equation reduces to the simpler form:
div v = 0. (2.9)
Exercise 2.1
Given the velocity eld:
v(t, x) =
_
_
_
_
_
_
_
x sin t y cos t
x cos t + y sin t
tan(xy)
_
_
_
_
_
_
_
.
Show that this represents a possible motion of an incompressible uid.
Lecture 3
Governing Equations II
3.1 Conservation of momentum
Before we can proceed any further, we need to get a rm understanding of terms
such as viscosity, viscous stresses, conductivity, etc.
Consider the left face of the control volume considered earlier. The uid molecules
to the left of this CV can interact with our CV in one of three ways:
(i) organized motion from left to right. While the molecules are constantly
moving about back and forth, over a small period of time, the majority of
these molecules either enter the control volume (u > 0) or leave the control
volume. This average over a period of time is called the ow velocity
component u, and is measured by probes such as LDVs (Laser Doppler
Velocimetry) and hot wires.
20
LECTURE 3. GOVERNING EQUATIONS II 21
y
y
z
z
q
z
(x, y, z)
LHS Face
Figure 3.1: Orderly motion from left to right, convection.
(ii) Exchange of u- momentum between the molecules on the left and those on
the right by collisions. In this case, there is no net gain in mass, but there
is a gain (or loss) in momentum. These collision eects may be averaged
over a suciently small period of time, and may be viewed as a pressure
force exerted by the uid on the left on our CV. Again, only this average
eect is felt or measured by pressure probes, and barometers. The individual
collisions occur far too rapidly and far too frequently to be sensed by probes
or measuring devices.
LECTURE 3. GOVERNING EQUATIONS II 22
y
y
z
(x, y, z)
LHS Face
w
y
y
z
y
q
i
q
i
Figure 3.2: Exchange of u- momentum between the molecules on the left and
those on the right by collisions, pressure.
(iii) Exchange of u-, v- and w- momentum by random linear motion of molecules
jumping in and out of our control volume, across the left face. In this case,
all the molecules that jumped in also jump out over a suciently small time
period. Thus, this random motion does not add mass to our control volume
(and was not considered in our continuity equation). They however bring
u-, v- and w- momentum in or out (associated with their random motion).
The time averages of these rates at which u-, v- and w- momentum is
brought into the CV across a face are called viscous forces. The forces per
unit area are called viscous stresses. The viscous stresses that bring in/out
u- momentum across the left face are called normal viscous stresses, while
those that bring in v- and w- momentum (by entering the face at an angle)
are called tangential viscous stresses.
LECTURE 3. GOVERNING EQUATIONS II 23
By convention, pressure forces are considered positive, if they act towards the
uid element, or control volume. The normal viscous stresses (following solid
mechanics conventions) are considered positive if they act away from the control
volume, producing a tension.
(x, y, z)
-

-

-
Normal Stress
Normal Stress
Pressure
Pressure
S
B
S
T
S
F
S
Bot
S = Shear stress
Figure 3.3: F=Front, T=Top, B=Back, Bot=Bottom.
These stresses (normal, and tangential or shear) are given the symbol T. They
are identied by two subscripts.
(i) The rst subscript indicates the plane on which they act. For example, if a
plane is normal to the x- axis (the LHS and RHS faces), the rst subscript
will be x.
(ii) The second subscript identies the direction of the force associated with
the force. For example, if a shear force is pointing in the y- direction, the
second subscript will be y. Hence for example T
yz
represents a stress on a
face that is normal to the y- axis and acting in the direction of the z- axis.
Thus, a viscous stress is a tensor quantity, and requires three pieces of information
(its magnitude, its direction and the plane on which it acts) to completely specify
LECTURE 3. GOVERNING EQUATIONS II 24
it. This separates a tensor from a vector (magnitude and direction), and a scalar
(magnitude only).
3.1.1 Newtonian Fluids
Because our primary unknowns are the ow properties (u, v, w, p, , T) there is
a need to link the stresses T with these physical variables. In solid mechanics
(Hookes law) stress is set proportional to strain. This works for solids because
a solid undergoes only a nite amount of deformation when a force or stress
is applied to it. In uid mechanics, this approach does not work because uid
continuously deforms when a shear stress is applied. It is this characteristic that
distinguishes a uid from a solid.
3.1.2 Deformation law for a Newtonian uid
Stokes (1845) postulated that for a nonelastic uid, the viscous stress varies
linearly with strain rate. The resulting deformation law is satised by all gasses
and most of the common uids (e.g. water). Stokes postulate was based on the
following assumptions:
1. The uid is continuous, and its stress tensor T is at most a linear function
of the deformation tensor (strain rates) D,
D
ij
=
1
2
_
du
i
dx
j
+
du
j
dx
i
_
or D =
1
2
_
v + (v)
T
_
.
2. the uid is isotropic; i.e., its properties are independent of direction, and
therefore the deformation law is independent of the coordinate axes in which
it is expressed.
3. When there is no deformation (i.e. the strain rates are zero), the stress T
is zero and the overall stress is hydrostatic;
ij
= p
ij
, where
ij
is the
LECTURE 3. GOVERNING EQUATIONS II 25
Kronecker delta function.
The most general deformation law for a Newtonian viscous uid hence gives the
total stress as:

ij
= p
ij
+ T
ij
= p
ij
+ 2D
ij
+
ij
div v, (3.1)
where is the coecient of dynamic viscosity and , which is associated with
volume expansion, is the coecient of bulk viscosity. Equation (3.1) can be
written in the alternative non-index vector-tensor notation:
= pI +T, with T = 2D+ ( div v)I, (3.2)
where I is the unit tensor.
Both
tr D = div v,
and
tr = 3p + (3 + 2) div v,
are tensor invariant. For an incompressible uid we see that the pressure is the
mean of the principal stresses:
1
3
tr = p, div v = 0.
For a compressible uid, a similar quantity ( p) dened by the mean of the prin-
cipal stresses:
p =
1
3
tr = p + ( +
2
3
) div v,
is not equal to the thermodynamic property called pressure. We call p the me-
chanical pressure. The bulk viscosity
+
2
3
,
LECTURE 3. GOVERNING EQUATIONS II 26
is a subject of much controversy. Stokes hypothesis
+
2
3
= 0,
has been found reasonable for monatomic gases but is certainly not true for
polyatomic gases or liquids. The precise value becomes unimportant for incom-
pressible or nearly incompressible uids.
We now return to the derivation of the u- momentum equation. This equation is
a generalization of Newtons second law of motion, and may be verbally stated
as:
The rate of change of u- momentum within a control volume is equal to the
net rate at which u- momentum enters the control volume plus forces (pressure,
viscous and body) acting on the control volume in the x- direction
As before we consider the control volume of gure 2.1.
Rate of change of umomentum within CV = mass acceleration
=

t
( u x y z) (3.3)
Rate at which u- momentum enters through:
LHS Face : u
2
y z,
RHS Face : u
2
y z, (enters in negative x direction),
Bottom Face : uv x z,
Top Face : uv x z, (enters in negative y direction),
Back Face : uwx y,
Front Face : uwx y, (enters in negative z direction).
LECTURE 3. GOVERNING EQUATIONS II 27
In the x- direction, pressure forces and normal viscous stresses only act on the
left and right hand faces:
Pressure force acting on LHS Face : p y z,
Pressure force acting on RHS Face : p y z,
Normal viscous stress acting on LHS Face : T
xx
y z,
Normal viscous stress acting on RHS Face : T
xx
y z,
On the other hand, the shear stresses in the x- direction only act on the top,
bottom, front and back faces:
Tangential viscous stress acting on top Face : T
yx
x z,
Tangential viscous stress acting on bottom Face : T
yx
x z,
Tangential viscous stress acting on front Face : T
zx
x y,
Tangential viscous stress acting on back Face : T
zx
x y,
We collect the body forces (per unit volume in the CV) such as gravity, electrical
and electromagnetic eects in the term:
body forces = bx y z.
Summing up these contributions, dividing through x y z and taking the
limits as x, y and z go to zero, we get:
(u)
t
+
(u
2
)
x
+
(uv)
y
+
(uw)
z
=
p
x
+
T
xx
x
+
T
yx
y
+
T
zx
z
+ b
1
. (3.4)
The v- and w- momentum equations may be derived using a logic identical to
that used above, and are left as an exercise. The nal form is:
v
t
+
(uv)
x
+
(v
2
)
y
+
(vw
2
)
z
=
p
y
+
T
xy
x
+
T
yy
y
+
T
zy
z
+ b
2
, (3.5)
w
t
+
(uw)
x
+
(vw)
y
+
(w
2
)
z
=
p
z
+
T
xz
x
+
T
yz
y
+
T
zz
z
+ b
3
. (3.6)
LECTURE 3. GOVERNING EQUATIONS II 28
For general viscous compressible ow, the momentum equations can be written
in compact tensor notation as:
(u
i
)
t
+
(u
i
u
j
)
x
j
=
p
x
j
+
T
ji
x
j
+ b
i
, (3.7)
where T
ji
= T
x
j
x
i
, and a repeated index denotes summation over that index:
(u
i
u
j
)
x
j
= ( u
i
v),
T
ji
x
j
= T
i
.
For incompressible ow, the momentum equations simplify:

u
t
+ u
u
x
+ v
u
y
+ w
u
z
=
p
x
+
T
xx
x
+
T
yx
y
+
T
zx
z
+b
1
, (3.8)

v
t
+ u
v
x
+ v
v
y
+ w
v
z
=
p
y
+
T
xy
x
+
T
yy
y
+
T
zy
z
+b
2
, (3.9)

w
t
+ u
w
x
+v
w
y
+ w
w
z
=
p
z
+
T
xz
x
+
T
yz
y
+
T
zz
z
+ b
3
, (3.10)
and can be written in vector/tensor form:

Dv
Dt
= p + div T+b. (3.11)
3.1.3 The Navier-Stokes equations
The desired momentum equation for a general Newtonian viscous incompressible
uid is now obtained via the stress relations (3.2). The result is the famous
Navier-Stokes equations (Navier-1823 & Stokes-1845)

Dv
Dt
= p + div (2D) +b. (3.12)
For constant viscosity, these reduce to:

Dv
Dt
= p +
2
v +b. (3.13)
Lecture 4
Governing Equations III
4.1 Conservation of energy
The energy equation is a generalized form of the rst law of thermodynamics,
which states that the time rate of change of a material regions total energy (i.e.
internal + kinetic) is equal to the rate of heat transferred to the material region
less the rate of work done by the material region, i.e.
dE
dt
=
dQ
dt

dW
dt
. (4.1)
The only dierence to our current case is that we are studying an open system
(i.e. control volume) that can gain and lose mass. The classical form studied
in courses on thermodynamics is applicable only for closed systems - i.e. xed
collection of particles. The version of the rst law of Thermodynamics that will
be applicable to our situation hence states that:
(A) The rate at which the total energy (i.e. internal + kinetic) increases within
a control volume =
(B) the rate at which total energy enters the control volume +
29
LECTURE 4. GOVERNING EQUATIONS III 30
(C) the rate at which work is done on the control volume boundary by surface
forces +
(D) the rate at which work is done on the CV by body forces +
(E) the rate at which heat is added to the control volume at the surfaces by
heat conduction +
(F) the rate at which heat is released within the CV due to chemical reactions.
The body forces (D) include, say, gravity, electrical and electromagnetic forces
and term (F) corresponds, say, to chemical reactions and thermal radiation. In
some applications, e.g. modeling of weather, term (D) is important, while in
others (e.g. modeling combustors) term (F) is important.
We dene the specic total energy (i.e. total energy per unit mass) E as
E = e +
1
2
v
T
v, (4.2)
where e is specic internal energy. The word specic means per unit mass.
Term A
The rate at which the total energy, E, increases within a control volume may be
expressed as:

t
( E xyz) or

t
( E) xyz (4.3)
Term B
This involves mass entering and leaving the six faces of the CV, carrying with it,
the total energy E. For each face, this term equals density normal velocity
total energy area of face and hence can be expressed as:

_
uE[
lhs
uE[
rhs
x
+
vE[
bot
vE[
top
y
+
wE[
back
wE[
front
z
_
, (4.4)
LECTURE 4. GOVERNING EQUATIONS III 31
where = xyz. As x, y, z 0, the terms within the square brackets
will approximate the continuous partial derivatives and hence we can write:
_

(uE)
x

(vE)
y

(wE)
z
_
xyz (4.5)
Term C
At the surfaces of the control volume, there are pressure and viscous forces that
do work on the uid, as the uid particles cross the boundary. For examples, on
the left and right hand faces, the pressure forces, p, and normal viscous forces,
T
xx
, act parallel to the u velocity component whereas the shear viscous forces T
xy
& T
xz
act parallel to v and w respectively. The rate at which work is done by
these surface forces is then:
yz [pu T
xx
u T
xy
v T
xz
w
lhs
pu T
xx
u T
xy
v T
xz
w
rhs
]
=
_
pu T
xx
u T
xy
v T
xz
w
lhs
pu T
xx
u T
xy
v T
xz
w
rhs
x
_


x
pu T
xx
u T
xy
v T
xz
w xyz. (4.6)
Similarly, the work done by the pressure and viscous forces (i.e. surface forces)
on the bottom and top faces is:


y
pv T
yx
u T
yy
v T
yz
w xyz, (4.7)
and the work done by the pressure and viscous forces on the back and front faces
is:


z
pw T
zx
u T
zy
v T
zz
w xyz. (4.8)
LECTURE 4. GOVERNING EQUATIONS III 32
Term D
Taking the body forces per unit volume as b. The work done by these body
forces on the control volume is
(ub
1
+vb
2
+ wb
3
) xyz. (4.9)
Term E
The heat conduction eects are associated with the random motion of gas molecules
across the control volume. As they move in and out, they bring energy into and
out of the control volume. When integrated over a small but nite period of time,
a net exchange of heat energy occurs at the boundary, without any exchange in
mass. Since this process is a random, chaotic process, it must somehow be empir-
ically modeled. We adapt Fouriers law used to model conduction of heat through
solids, which states
The rate at which heat ows across a surface of unit area is proportional to the
negative of the temperature gradient normal to this surface.
The constant of proportionality is called conductivity , a property of the solid
(or uid, in the present context). Note that the heat ux is proportional to the
negative of the gradient, because heat ows from hot to cold. On the left and
right hand faces, the net heat ux is

T
x
yz

lhs
+
T
x
yz

rhs
=
_
_

T
x

lhs
+
T
x

rhs
x
_
_
xyz


x
_

T
x
_
xyz. (4.10)
LECTURE 4. GOVERNING EQUATIONS III 33
Similarly, the net heat ux in the y- direction is

y
_

T
y
_
xyz, (4.11)
and that in the z- direction is

z
_

T
z
_
xyz. (4.12)
Term F
The rate at which heat is released per unit volume within the CV due to chemical
reactions and/or thermal radiation will be represented by:
r xyz. (4.13)
4.1.1 Energy equation
Summing up all the terms (A) through (F), and dividing through by the common
factor xyz, we get the energy equation:

t
( E) +
(uE)
x
+
(vE)
y
+
(wE)
z
+
(pu)
x
+
(pv)
y
+
(pw)
z
=

x
_

T
x
_
+

y
_

T
y
_
+

z
_

T
z
_
+

x
T
xx
u + T
xy
v + T
xz
w +

y
T
yx
u + T
yy
v + T
yz
w

z
T
zx
u + T
zy
v +T
zz
w + (ub
1
+vb
2
+ wb
3
) + r (4.14)
This can be written in compact vector form

t
( E) + (vE) + (pv) = (T) + (T v) + v b + r, (4.15)
where
(T v) =

x
i
(T
ij
v
j
) .
LECTURE 4. GOVERNING EQUATIONS III 34
4.1.2 Mechanical energy equation
The mechanical energy equation is obtained by taking the inner (dot) product of
the velocity vector with the momentum equation:
v
T
momentum equation.
This leads to

t
_

v
T
v
2
_
+
_
v
v
T
v
2
_
= v (p) +v ( T) + v b. (4.16)
The mechanical energy equation (4.16) shows that the kinetic energy (v
T
v/2)
increases due to three eects:
uid motion in the direction of decreasing pressure,
uid motion in the direction of increasing viscous stress, or
uid motion in the direction of a body force.
Note that body forces themselves aect mechanical energy, while it is imbalances
in the surface forces which aect mechanical energy.
4.1.3 Thermal energy equation
If we subtract the mechanical energy equation (4.16) from equation (4.15), we
obtain an equation for the evolution of thermal energy:

t
( e) + (ve) + p v = (T) +T: v + r, (4.17)
where the notation (:) denotes the double dot product of two tensors,
T: v =
3

i=1
3

j=1
T
ij

i
v
j
=
3

i=1
3

j=1
T
ij
v
j
x
i
.
LECTURE 4. GOVERNING EQUATIONS III 35
= T: v is called the viscous dissipation function. Since the stress tensor is
symmetric, we have
= T: v = T: D,
where D = (v + (v)
T
)/2.
The thermal energy equation (4.15) shows that the thermal energy (or internal
energy) increases due to:
pressure force accompanied by a mean negative volumetric deformation,
(i.e., a uniform compression; note that v is the relative expansion rate),
positive gradients in temperature (or negative gradients in heat ux, more
heat enters than leaves), or
viscous forces associated with deformation.
Note that in contrast to mechanical energy, thermal energy changes do not require
surface force imbalances; instead they require kinematic deformation. Moreover,
body forces have no inuence on thermal energy. The work done by body forces
is partitioned entirely to the mechanical energy of a body.
4.1.4 Non-conservative energy equation
We can use the conservation of mass (continuity equation) to obtain the com-
monly used non-conservative form of the energy equation:

De
Dt
= p v + (T) +T: v + r. (4.18)
The (non-conservative) temperature equation can then be obtained from e = c
v
T,
where c
v
is the specic heat at constant volume:
c
v
DT
Dt
= p v + (T) +T: v + r. (4.19)
Lecture 5
Dimensionless Equations
As given earlier, the continuity, momentum and energy equations are all di-
mensional. These can be made dimensionless if we redene all variables to be
dimensionless, by dividing them by constant reference properties appropriate to
the ow:
5.1 Incompressible ow
reference velocity U
0
(the freestream) or
0
/(
0
L) if there is no freestream,
as in natural (free) convection ow. For compressible (gas) ow, the speed
of sound a
0
is the ideal reference.
Other freestream properties p
0
,
0
, c
p
0
, T
0
,
0
,
0
.
Reference length L (body length, duct diameter, channel (half) width etc.)
Residence time L/U
0
, say in steady ows with no characteristic time of
their own.
It is essential to use constants for nondimensionalizing and not to mix variables
together; for example dene x/L and y/L as new dimensionless variables and not
36
LECTURE 5. DIMENSIONLESS EQUATIONS 37
x/y.
The most commonly encountered dimensionless variables are listed below and
denoted by an asterisk.
t

=
U
0
L
t, x

=
x
L
, u

=
u
U
0
,

0
,
p

=
p p
0
U
2
0
, T

=
T T
0
T
w
T
0
,

0
,

= L, T

=
L

0
U
0
T, D

=
L
U
0
D,
On substituting the dimensionless variables into the basic equations, we notice
that the continuity equation results only in a change of variables:

= 0, (5.1)
showing that continuity is not directly inuenced by any ow parameters. The
momentum and energy equations on the other hand will now contain dimen-
sionless ow parameters. In particular, the dimensionless momentum equation
(without body forces) now reads:
Du

Dt

+
1
Re

(2

), (5.2)
which contains a dimensionless parameter ( Re ) called the Reynolds number:
Re =

0
U
0
L

0
.
In very high speed ows (say supersonic ows) we can neglect viscous terms from
the momentum equation and the resulting ow is classied as inviscid.
LECTURE 5. DIMENSIONLESS EQUATIONS 38
The dimensionless energy equation is given by:
DT

Dt

= Ec
Dp

Dt

+
1
Re Pr

) +
Ec
Re

, (5.3)
which in addition to the Reynolds number, now contains two additional dimen-
sionless parameters, the Prandtl ( Pr ) and Eckert ( Ec ) numbers:
Pr =

0
c
p
0

0
, Ec =
U
2
0
c
p
0
(T
w
T
0
)
.
In high speed ows (U
0
1), all three parameters are important, whereas for
low speed ow (in which case U
2
0
1) we an usually neglect both the pressure
terms and dissipation terms from the energy equation.
5.1.1 Parameters for Natural convection ow
Free convection ows contain a body force due to buoyancy, to account for the
density changes that arise due to temperature changes in the ow eld. The co-
ecient of thermal expansion for a uid, , is related to the density, temperature
and pressure via:
=
1

T
_
p
,
where the subscript
p
shows that the derivatives are evaluated at constant pres-
sure. Thus

0
T,
and we can approximate the density changes by
=
0
+
0

0
T.
The total body force for Natural convection ows (gravity plus buoyancy) is thus
usually approximated as
LECTURE 5. DIMENSIONLESS EQUATIONS 39

0
gk +
0
(T T
0
)gk,
where g is the acceleration due to gravity and k is the unit vector in the positive
z direction. The corresponding Navier-Stokes equations are given as:

0
Dv
Dt
= (p +
0
g z) +
0
(T T
0
) g k + div (

T). (5.4)
In dimensionless terms, this reads
Dv

Dt

+
Gr
Re
2

k +
1
Re

, (5.5)
where
P

=
p +
0
g z p
0

0
U
2
0
,
and Gr is the Grashof number:
Gr =
g
0

2
0
L
3
(T
w
T
0
)

2
0
.
For natural convection ow, there is no free stream and the ow is driven by
temperature dierences. The proper reference velocity is then
U
0
=

0

0
L
,
thus
1 =

0
U
0
L

0
= Re ,
so that the Grashof number is the single dominant parameter.
The full equations of motion are highly nonlinear and analytic solutions rarely can
be found except in a few special cases. The solutions for real physical problems
thus necessarily have to be found mostly via computational uid dynamics (CFD).
LECTURE 5. DIMENSIONLESS EQUATIONS 40
In the following lectures, we focus attention on particular cases where analytical
or nearly analytical solutions are obtainable. As expected, almost all the known
particular solutions are for the case of incompressible Newtonian ow with con-
stant transport properties for which the nonlinear convective terms either vanish
or reduce to very simple forms. In such cases, the basic equations reduce to
div v = 0, (5.6)

Dv
Dt
= b p +
2
v, (5.7)
c
p
DT
Dt
=
2
T + . (5.8)
In the case of simple body forces like gravity, the fact that is constant shows here
that the continuity and momentum equations are uncoupled from the tempera-
ture, T, and thus can be solved for the velocity, v, and pressure, p, independently
of T. The temperature can then be found from the energy equation based on the
already computed velocity and pressure.
Because of the alluded to uncoupling, it is usually common practice in incom-
pressible ows with constant properties to simply ignore the energy equation!
5.2 Compressible ow
The ambient speed of sound a
0
plays an important role in compressible ow. As
we shall see later, this is dened as
a
0
=

p
0

0
=
_
RT
0
, (5.9)
where R is the gas constant, p
0
is the ambient pressure,
0
the ambient density,
T
0
the ambient temperature and the adiabatic constant is the ratio of specic
LECTURE 5. DIMENSIONLESS EQUATIONS 41
heats:
=
c
p
c
v
. (5.10)
The dimensionless quantities for compressible ow thus include the adiabatic
constant () and the Mach number (M) dened as the ratio of the ambient
velocity to the ambient speed of sound:
M
2
=
v
2
0
a
2
0
=

0
v
2
0
p
0
. (5.11)
The non-dimensionalisation process for the pressure will thus remain as given for
the incompressible case:
p

=
p p
0

0
U
2
0
.
Exercise 5.1 Perform the nondimensionalization process for the continuity, mo-
mentum and energy equations (2.7), (3.13) and (4.17) respectively. Neglect the
body force in the momentum equations.
Repeat the procedure for the case where the momentum equation has the form
given in equation (5.4), i.e. where there is now a body force due to natural con-
vection. Only consider the case of an incompressible uid with constant viscosity.
Lecture 6
Steady Couette Flows
Due to Couette (1890), who investigated ows between a xed and moving con-
centric cylinder.
6.1 Flow between a moving and a xed plate
-
-
-
-
-
6
-
x
y
u = 0, T = T
0
-
u = U
0
T = T
1
u = u(y), T = T(y)
1
Figure 6.1: Steady shear ow between parallel plates
Consider two parallel plates of innite extent in the x direction as shown in gure
6.1. The lower plate is stationary and is located along y = 0 and the upper plate
42
LECTURE 6. STEADY COUETTE FLOWS 43
at y = h moves with constant velocity U
0
in the x-direction. The pressure is
assumed constant and the motion is driven by the moving upper wall. Suppose
also that the lower and upper walls are maintained at temperatures T
0
and T
1
respectively.
It follows from the above that u = u(y) and T = T(y). The governing equations
(5.6 - 5.8) thus reduce to:
u
x
= 0, (6.1)
0 =
d
2
u
dy
2
, (6.2)
0 =
d
2
T
dy
2
+
_
du
dy
_
2
. (6.3)
The boundary conditions are
u(0) = 0, and u(h) = U
0
, no slip, (6.4)
T(0) = T
0
, and T(h) = T
1
, no temperature jump. (6.5)
The velocity distribution is thus
u(y) =
U
0
h
y, (6.6)
and it follows that the corresponding solution for the temperature is
T(y) = T
0
+ (T
1
T
0
)
y
h
+
U
2
0
2 h
2
y (h y). (6.7)
The shear stress at any point in the ow follows from Newtons law of viscosity:
T
xy
=
_
u
y
+
v
x
_
=
du
dy
=
U
0
h
, (6.8)
showing that the shear stress for this ow is constant throughout the ow eld,
as is the strain rate U
0
/h. In engineering, the shear stress is usually dened as
the skin friction coecient, C
f
,
C
f
=
2
U
2
0
T
xy
= 2

U
0
h
=
2
Re
, (6.9)
LECTURE 6. STEADY COUETTE FLOWS 44
where Re is the Reynolds number based on the velocity of the upper plate and
the channel width. The rate of heat transfer at the walls is
q
w
=

T
y

0,h
=
k
h
(T
1
T
0
)
U
2
0
2 h
, (6.10)
where the () refers to lower and upper wall, respectively.
6.2 Axially moving concentric cylinders
-
-
?
*
-
6
w = W
1
w = W
0
z
r
r
0
r
1
Fluid
Fluid
1
Figure 6.2: Steady shear ow between concentric cylinders
Consider the isothermal (constant temperature) and isobaric (constant pressure)
ow of an incompressible viscous uid enclosed between two concentric cylinders
whose axis is along r = 0 as shown in gure 6.2. The inner and outer cylinders,
(r = r
0
) and (r = r
1
) respectively, move axially at speeds W
0
and W
1
. The
velocity distribution satises w = w(r), in cylindrical coordinates (r, , z) with
the corresponding velocity components (u, v, w). The only nontrivial equation of
motion is the momentum equation:

2
w =
1
r
d
dr
_
r
dw
dr
_
= 0. (6.11)
LECTURE 6. STEADY COUETTE FLOWS 45
The boundary conditions are derived from no-slip:
w(r
0
) = W
0
, and w(r
1
) = W
1
. (6.12)
the velocity eld is then
w(r) = W
0
+
(W
1
W
0
)
ln r
1
ln r
0
(ln r ln r
0
). (6.13)
6.3 Flow between rotating concentric cylinders
?
*
-
6
z
r
R
0
R
1
Fluid
Fluid
1
3

1
T = T
1
T = T
0
1
Figure 6.3: Steady ow between rotating concentric cylinders
Consider the steady motion of a uid between two innite concentric cylinders
with radii R
0
& R
1
(R
1
> R
0
), rotating about their axis with steady angular
velocities
0
&
1
as shown in gure 6.3. The inner cylinder is maintained at a
temperature T
0
, and the outer cylinder is maintained at T
1
.
Take cylindrical polar coordinates (r, , z) with the z-axis along the axis of the
cylinders. Suppose the corresponding velocity components in the (r, , z) direc-
tions are respectively (u, v, w), then it is evident from the geometry that the
velocity, temperature and pressure must be functions only of radial distance r:
LECTURE 6. STEADY COUETTE FLOWS 46
v = v(r), u = w = 0, T = T(r), p = p(r).
The equations of motion thus reduce to
v

= 0, (6.14)

v
2
r
=
dp
dr
, (6.15)
0 =
_
1
r
d
dr
_
r
dv
dr
_

v
r
2
_
, (6.16)
0 =
1
r
d
dr
_
r
dT
dr
_
+
_
dv
dr

v
r
_
2
, (6.17)
with boundary conditions
v = R
0

0
, T = T
0
, p = p
0
, at r = R
0
,
v = R
1

1
, T = T
1
, at r = R
1
.
Equation (6.16) is a Cauchy-Euler dierential equation
d
2
v
dr
2
+
1
r
dv
dr

v
r
2
= 0,
and thus admits solutions of the form r
n
. We have n = 1, leading to
v = C
1
r +
C
2
r
,
so that the velocity distribution is
v =
1
R
2
1
R
2
0
_
(
1
R
2
1

0
R
2
0
) r + (
0

1
)
R
2
0
R
2
1
r
_
. (6.18)
The (dimensionless) temperature distribution is
T T
0
T
1
T
0
= Pr Ec
R
4
1
(
0

1
)
2

2
0
(R
4
1
R
4
0
)
(1 f(r))
_
1
R
2
0
r
2
_
+ f(r). (6.19)
LECTURE 6. STEADY COUETTE FLOWS 47
where
f(r) =
ln r ln R
0
ln R
1
ln R
0
, and Pr Ec =
R
2
0

2
0
(T
1
T
0
)
.
The pressure distribution can be written in the form
p = p
0
+
_
C
1
2
(r
2
R
2
0
) + 2C
1
C
2
(ln r ln R
0
)
C
2
2
_
1
r
2

1
R
2
0
__
, (6.20)
with
C
1
=
1
R
2
1
R
2
0
(
1
R
2
1

0
R
2
0
), and C
2
=
1
R
2
1
R
2
0
(
0

1
)R
2
0
R
2
1
.
Exercise 6.1 For the ow considered in section (6.2), suppose the inner cylinder
is maintained at a temperature T
0
and the outer cylinder at T
1
. Neglecting natural
convection, compute the temperature distribution, T(r), of the uid enclosed in
between.
Lecture 7
Steady Poiseuille Flows
7.1 Flow Between Parallel Planes
6
-
x
y
u = 0, T = T
0
u = u(y), T = T(y), p = p(x)
-
-
-
-
-
-
-
-
u = 0, T = T
1
1
Figure 7.1: Steady pressure driven ow in a channel.
Consider steady, pressure driven ow between two xed parallel planes as shown
in gure 7.1. Limit the analysis to the x-y plane and so that the xed planes are
parallel to the x-axis and the ow is in the direction of the positive x-axis. We
can deduce from the geometry that
u = u(y) and v = 0,
48
LECTURE 7. STEADY POISEUILLE FLOWS 49
in which case the Navier-Stokes equations reduce to
u
x
= 0, (7.1)
0 =
p
x
+
d
2
u
dy
2
, (7.2)
0 =
p
y
. (7.3)
We then have that p = p(x), hence
d
2
u
dy
2
=
1

dp
dx
. (7.4)
This clearly shows that dp/dx = constant say dp/dx = G with G > 0. Applying
this and the no-slip boundary conditions for the velocity
u(0) = 0 and u(h) = 0,
leads to the parabolic velocity distribution:
u =
G
2
y(y h). (7.5)
The shear stress at any point in the ow follows from Newtons law of viscosity:
T
xy
=
du
dy
=
G
2
(h 2y), (7.6)
so that the shear stress at the wall is
[T
xy
]
0,h
=
Gh
2
. (7.7)
where the refers to the bottom and top wall respectively.
The volume ow rate, Q, is
Q =
_
A
u(y) dA =
_
h
0

G
2
y(y h) dy =
G
12
h
3
, (7.8)
LECTURE 7. STEADY POISEUILLE FLOWS 50
hence the mean velocity, u, is
u =
Q
h
=
G
12
h
2
. (7.9)
The wall shear stress can then be expressed in terms of the mean velocity as
[T
xy
]
0,h
=
6
h
u,
and hence the skin friction coecient, C
f
, is
C
f
=
2
u
2
[T
xy
]
0,h
=
12
Re
(7.10)
where Re is the Reynolds number based on the mean velocity and the channel
width.
7.1.1 Non-isothermal case
The assumption of constant viscosity uncoupled the temperature eects from
the Navier-Stokes equations if natural convection is neglected. Now with the
velocity distribution known from relation (7.5), suppose the lower wall is kept at
a temperature T
0
and the upper wall is maintained at T
1
, then the temperature
distribution, T(y), can be found from the energy equation:
0 =
d
2
T
dy
2
+
_
du
dy
_
2
, (7.11)
with boundary conditions
T(0) = T
0
and T(h) = T
1
. (7.12)
We obtain
T = T
0
+ (T
1
T
0
)
y
h
+
G
2
24
y(h
3
2y
3
+ 4hy
2
3h
2
y). (7.13)
LECTURE 7. STEADY POISEUILLE FLOWS 51
The rate of heat transfer at the walls is
q
w
=

T
y

0,h
=
k
h
(T
1
T
0
)
G
2
24
h
3
(7.14)
where, as before, the () refers to lower and upper wall, respectively.
7.2 Flow in through a duct
Suppose we extend the above analysis to 3D so that the ow in now in a duct
whose walls are parallel to the x-z plane but the ow is still only in the direction
of the positive x-axis. We now have
u = u(y, z) and v = w = 0,
and the Navier-Stokes equations are modied to
u
x
= 0, (7.15)
0 =
p
x
+
_
d
2
u
dy
2
+
d
2
u
dz
2
_
, (7.16)
0 =
p
y
, (7.17)
0 =
p
z
. (7.18)
The pressure is thus still only dependent on x, p = p(x), and as before we can
write
d
2
u
dy
2
+
d
2
u
dz
2
=
2
v =
1

dp
dx
, (7.19)
where again dp/dx = G is constant.
7.2.1 Flow in a circular pipe: Hagen-Poiseuille ow
Consider the steady pressure driven ow of an incompressible uid in a circular
pipe as shown in gure 7.2. In fact this is a 3-D ow which, due to its axi-
symmetric nature, reduces to a unidirectional ow along the z-direction.
LECTURE 7. STEADY POISEUILLE FLOWS 52
*
-
6
z
r R
-
-
-
-
-
-
-
-
-
w = 0, T = T
R
w = 0, T = T
R
w = w(r), T = T(r), p = p(z)
1
Figure 7.2: Steady pressure driven ow in a circular pipe.
Taking the origin at the center of the pipe and use cylindrical polar coordinates
(r, , z) with the ow in the positive z direction, we have u = v = 0 and w = w(r).
Here u, v & w are the velocity components in the r, and z directions respectively.
The relevant Navier-Stokes equations (7.19) are given in cylindrical coordinates
as
1
r
d
dr
_
r
dw
dr
_
=
G

,
with no-slip velocity conditions
w(R) = 0,
where R is the pipe radius. We obtain
w =
G
4
(R
2
r
2
). (7.20)
The volume ow rate (discharge) through any cross-section, A, of the pipe is
Q =
_
A
w(r) dA,
which for the circular pipe is
Q =
_
R
0
w(r) (2 r dr) =
R
4
8
G.
LECTURE 7. STEADY POISEUILLE FLOWS 53
The uid maximum velocity obtains at the centerline of the pipe, r = 0, and is
given by
w
max
=
G
4
R
2
. (7.21)
The mean velocity w is half of the maximum velocity:
w =
Q
A
=
R
2
G
8
=
1
2
w
max
, (7.22)
and the wall shear stress is constant
(T
12
)
wall
=
_

dw
dr
_
wall
=
1
2
GR =
4 w
R
. (7.23)
The skin friction coecient, C
f
, can then be calculated from
C
f
=
2
w
2
(T
12
)
wall
= 8

wR
=
8
Re
, (7.24)
where Re = wR/ is the Reynolds number based on the mean velocity and the
pipe radius.
7.2.2 Temperature distribution in Hagen-Poiseuille ow
Suppose we impose a constant temperature,
T = T
R
, on r = R,
then, as with the velocity, the temperature only depends on radial distance,
T = T(r), and the energy equation in cylindrical coordinates reduces to
0 =

r
d
dr
_
r
dT
dr
_
+
_
dw
dr
_
2


r
d
dr
_
r
dT
dr
_
=
16
R
4
w
2
r
2
. (7.25)
LECTURE 7. STEADY POISEUILLE FLOWS 54
The condition that the uid temperature remains bounded at the center of the
pipe, r = 0, leads to
T = T
R
+ w
2

_
1
r
4
R
4
_
. (7.26)
The wall heat transfer, q
R
, is
q
R
=

T
r

r=R
= 4 w
2

R
. (7.27)
Exercise 7.1 Consider the ow of an incompressible viscous uid enclosed be-
tween two concentric and stationary cylinders each of whose axis is along r = 0.
The inner and outer cylinders located respectively at (r = R
0
) and (r = R
1
), are
maintained at temperatures T
0
and T
1
. If the uid is driven only by a constant
pressure gradient dp/dz = G, G > 0 (i.e. neglect eects of natural convec-
tion) obtain the velocity, w = w(r), and temperature, T = T(r), distributions in
cylindrical coordinates (r, , z).
Also, obtain the volume ow rate, Q, the mean velocity, w, the wall shear stress,
[T
12
]
R
0
,R
1
, the skin friction coecient, C
f
, and the rate of heat transfer at the
walls, q
w
R
0
,R
1
.
Lecture 8
Combined Couette-Poiseuille ow
6
-
x
y
u = 0, T = T
0
-
u = U
0
T = T
1
u = u(y), T = T(y)
-
-
-
-
-
-
-
1
Figure 8.1: Steady combined Couette-Poiseuille ow in a channel.
Consider two parallel plates of innite extent in the x direction. The lower plate
is stationary and is located along y = 0 and the upper plate at y = h moves with
constant velocity U
0
in the x-direction as shown in gure 8.1.
A constant pressure gradient dp/dx = G is imposed on the ow so that the
uid motion is driven by the combined eects of the moving upper wall and the
pressure drop.
Suppose also that the lower and upper walls are maintained at temperatures T
0
55
LECTURE 8. COMBINED COUETTE-POISEUILLE FLOW 56
and T
1
respectively. If we neglect natural convection, then u = u(y), v = w = 0,
T = T(y) and the equations of motion reduce to
u
x
= 0, (8.1)
0 =
dp
dx
+
d
2
u
dy
2
, (8.2)
0 =
d
2
T
dy
2
+
_
du
dy
_
2
, (8.3)
with boundary conditions
u(0) = 0, and u(h) = U
0
, no slip, (8.4)
T(0) = T
0
, and T(h) = T
1
, no temperature jump. (8.5)
The velocity distribution is thus
u(y) =
U
0
h
y +
G
2
y (h y), (8.6)
and it follows that the corresponding temperature distribution is
T(y) = T
0
+ (T
1
T
0
)
y
h
+

_
U
2
0
2h
2
(hy y
2
)
+
U
0
G
6h
(h
2
y 3hy
2
+ 2y
3
)
+
G
2
24
2
(h
3
y 3h
2
y
2
+ 4hy
3
2y
4
)
_
(8.7)
The wall shear stress is
(T
xy
)
0,h
=
U
0
h

Gh
2
, (8.8)
where the refers to the bottom and top wall respectively. The volume ow
rate, Q, is
Q = U
0
h +
G
12
h
3
. (8.9)
LECTURE 8. COMBINED COUETTE-POISEUILLE FLOW 57
If U
0
= 0 (stationary walls) all the quantities reduce to those for pure Poiseuille
ow between parallel plates. If on the other hand G = 0 (no driving pressure
gradient, hence constant pressure throughout the ow eld) then all the above
quantities reduce to those for Couette ow between a xed (lower plate) and a
moving (upper plate).
Exercise 8.1 Consider the ow of an incompressible viscous uid enclosed be-
tween two concentric cylinders each of whose axis is along r = 0. The inner and
outer cylinders located respectively at (r = R
0
) and (r = R
1
), are maintained at
temperatures T
0
and T
1
and move axially (in the z-direction) with constant speeds
W
0
and W
1
respectively.
A constant pressure gradient dp/dz = G, G > 0 is also imposed on the ow.
Neglecting eects of natural convection, obtain the velocity, w = w(r), and tem-
perature, T = T(r), distributions in cylindrical coordinates (r, , z).
Also, obtain the volume ow rate, Q, the mean velocity, w, the wall shear stress,
[T
12
]
R
0
,R
1
, the skin friction coecient, C
f
, and the rate of heat transfer at the
walls, q
w
R
0
,R
1
.
Exercise 8.2 Consider a layer of uid of constant thickness h owing steadily
due to gravity down a xed plane inclined at an angle . If the liquid lm is
bounded above by a free surface, obtain the velocity and pressure distribution for
the ow.
Take the y-axis to be perpendicular to the plane and the ow to be in the direction
of the x-axis, down the inclined plane.
Lecture 9
Unsteady Flows With Moving
Boundaries
A variety of solutions are known for unidirectional non-isothermal ows (motion
is only parallel to a single coordinate axis and velocity components in remaining
coordinate directions vanish identically) with constant properties and moving
boundaries.
If we limit our attention to the rectangular coordinate system (x, y, z), then this
parallel ow assumption usually takes the form
u = u(t, y, z), v = 0, w = 0.
The momentum equations under such a framework reduce to
u
x
= 0, (9.1)
u
t
=
1

p
x
+
_

2
u
y
2
+

2
u
z
2
_
, (9.2)
0 =
p
y
, (9.3)
0 =
p
z
, (9.4)
where = / is called the coecient of kinematic viscosity. Here equations
58
LECTURE 9. UNSTEADY FLOWS WITH MOVING BOUNDARIES 59
(9.1), (9.3) and (9.4) simply rearm that the velocity depends only on y, z &
t and also show that the pressure only depends on x & t. However equation
(9.2) implies that the pressure gradient does not depend on x and hence can be
absorbed into the velocity by a change of variables:
u

= u +
_
1

p
x
dt. (9.5)
The momentum equations then reduce to
u

t
=
_

2
u

y
2
+

2
u

z
2
_
, (9.6)
which is simply the homogeneous heat conduction equation for which a wealth
of unsteady solution methods are known. Our preferred solution method will be
as follows. Dene
u

= U
steady
+ f(t, y, z),
where U
steady
is the steady ow velocity. Substitute this into the momentum
equations and determine the unknown function f(t, y, z) via separation of vari-
ables and Fourier series techniques, subject to appropriate initial and boundary
conditions.
9.1 Suddenly accelerated at plate
Consider ow in the x-y plane. Suppose uid rests on a at plate of innite
extent. Let the plane be at y = 0 so that u = u(t, y). The relevant boundary
conditions (no slip) and initial conditions (zero initial notion) are:
u(t, 0) = U(t) t 0, y = 0,
u(t, y) = 0 t < 0, y 0,
u U

y .
(9.7)
LECTURE 9. UNSTEADY FLOWS WITH MOVING BOUNDARIES 60
9.1.1 Unsteady ow between two innite plates
Incompressible uid of kinematic viscosity is bounded by two innite at planes
y = 0 and y = h. The uid is initially at rest at time t = 0, and after that the
plane y = 0 is suddenly caused to move in its own plane with constant speed U
0
.
Assuming zero pressure gradient, derive the governing equation for the velocity
component u = u(t, y) together with the boundary and initial conditions.
Assuming a solution of the form u(t, y) = U
steady
+ f(t, y) where U
steady
is the
steady state velocity for this ow, use separation of variables and Fourier series
to nd the function f(t, y) and hence the velocity u(t, y).
In this case the momentum equation reads
u
t
=

2
u
y
2
, (9.8)
with boundary and initial conditions
u(t, 0) = U
0
t > 0, y = 0,
u(t, h) = 0 t 0, y = h,
u(0, y) = 0 t = 0, y 0.
(9.9)
It follows that the steady state velocity distribution is
U
steady
= U
0
_
1
y
h
_
, (9.10)
hence we will assume the transient velocity
u(t, y) = U
0
_
1
y
h
_
+f(t, y). (9.11)
The basic dierential equation (9.8) becomes
f
t
=

2
f
y
2
, (9.12)
LECTURE 9. UNSTEADY FLOWS WITH MOVING BOUNDARIES 61
with boundary and initial conditions
f(t, 0) = 0 t > 0, y = 0,
f(t, h) = 0 t 0, y = h,
f(0, y) = U
steady
t = 0, y 0.
(9.13)
Equation (9.12) yields to a separation of variables f(t, y) = T(t)Y (y), leading to

T
T
=
Y

Y
=
2
, (9.14)
It follows that
T(t) = C
1
e

2
t
, Y (y) = C
2
sin
_
y

_
+ C
3
cos
_
y

_
,
where Y (0) = 0 and Y (h) = 0. The rst condition implies that C
3
= 0 and hence
to avoid trivial solutions, we will require that C
2
,= 0 (say C
2
= 1), in which case
the second condition leads to

n
=
n

h
, n = 1, 2, 3, . . . .
The Fourier components of the solution can then be written
f
n
(t, y) = T
n
(t)Y
n
(y) = A
n
exp(
2
n
t) sin
_

n
y

_
,
so that the full Fourier series solution is in the form
f(t, y) =

n=1
f
n
(t, y) =

n=1
A
n
exp
_

n
2

h
2
t
_
sin
_
n
h
y
_
. (9.15)
The initial condition is then required to satisfy the Fourier sine series
f(0, y) =

n=1
A
n
sin
_
n
h
y
_
= U
steady
= U
0
_
1
y
h
_
, (9.16)
and thus the Fourier coecients A
n
can be computed from the orthogonality
condition
A
n
=
2
h
_
h
0
U
0
_
1
y
h
_
sin
_
n
h
y
_
dy =
2
n
U
0
. (9.17)
LECTURE 9. UNSTEADY FLOWS WITH MOVING BOUNDARIES 62
The nal velocity distribution for this start-up Couette ow is thus
u(t, y) = U
0
_
1
y
h
_
U
0

n=1
2
n
exp
_

n
2

h
2
t
_
sin
_
n
h
y
_
, (9.18)
or in dimensionless form
u

= (1 y

n=1
2
n
exp(n
2

2
t

) sin(ny

), (9.19)
where
u

=
u
U
0
, y

=
y
h
, t

=

h
2
t.
Exercise 9.1 Suppose we now impose a constant pressure gradient
dp
dx
= G,
on the start-up Couette ow considered above. Obtain the modied governing
equations, boundary and initial conditions for this ow. Compute the velocity
distribution for this ow.
Lecture 10
Flows With Suction and Injection
All the exact solutions discussed so far had vanishing convective terms in the
equations of motion. We thus did not have to worry about nonlinear terms in
the momentum equation. We now consider a case where the convective terms in
the momentum equations take a simple but non-vanishing form, i.e. ows with
porous boundaries which allow for suction or injection.
10.1 Uniform suction on a plane
-
-
-
-
-
-
-
-
-
-
-
-
u = u(y)
u = U

x
6
y
? ? ? ? ? ? ? ? ? ? ? ?
u = 0, v = V
0
1
Figure 10.1: Uniform suction on a plane.
63
LECTURE 10. FLOWS WITH SUCTION AND INJECTION 64
Consider steady ow of a viscous uid past an innite porous plate (y = 0) where
there is a constant suction v = V
0
< 0. The uid is otherwise unbounded far
away from the plate, in which range it moves with uniform velocity U

, see gure
10.1. If the ow is conned to the x-y plane, then u = u(y), v = V
0
= constant
and the only nontrivial equation of motion is
V
0
du
dy
=
d
2
u
dy
2
, (10.1)
where we have assumed the ow to be isothermal and subjected to a constant
pressure. The relevant boundary conditions are
u = 0 at y = 0, and u U

as y . (10.2)
Since V
0
is constant, the dierential equation (10.1) is linear and can be readily
solved to obtain
u = U

_
1 exp
_
V
0
y

__
. (10.3)
10.2 Flow between plates with bottom injection
and top suction
Consider the ow between two porous at plates at y = h and y = h. Let the
main ow be generated by a constant pressure gradient dp/dx = G and let the
porous walls be such that a uniform vertical crossow (from bottom to top) is
generated, see gure 10.2,
v = V
0
= constant, V
0
> 0. (10.4)
We then have, as before, that u = u(y) and the only nontrivial momentum
equation is
V
0
du
dy
=
dp
dx
+
d
2
u
dy
2
, (10.5)
LECTURE 10. FLOWS WITH SUCTION AND INJECTION 65
6
-
x
y
u = u(y), p = p(x)
-
-
-
-
-
-
-
-
u = 0, v = V
0
u = 0, v = V
0
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
1
Figure 10.2: Uniform suction/injection ow in a channel.
with boundary conditions
u = 0 at y = h, and u = 0 at y = h. (10.6)
Equation (10.5) can be simplied (number of parameters reduced) by introducing
dimensionless variables
u

=
u
V
0
, x

=
x
h
, y

=
y
h
, p

=
p
V
2
0
,
leading to the dimensionless equation
du

dy

=
dp

dx

+
1
Re
d
2
u

dy
2
, (10.7)
where
Re =
V
0
h

,
is the Reynolds number based on the suction/injection velocity and the channel
half width.
The dimensionless boundary conditions are
u

= 0 at y

= 1, and u

= 0 at y

= 1, (10.8)
LECTURE 10. FLOWS WITH SUCTION AND INJECTION 66
and hence the dimensionless velocity distribution is
u

=
G

sinh Re
_
e
Re
e
Re y

_
+ (y

+ 1) G

, (10.9)
where G

= dp

/dx

is the dimensionless pressure gradient.


Exercise 10.1 Solve equation (10.5) directly without resorting to dimension-
less variables and compare your resulting velocity distribution with that given
in (10.9).
Lecture 11
Vorticity & Bernoullis Equation
11.1 Vorticity
The vorticity vector, w, is a measure of rotational eects in uid ow and is
dened as
w = v, (11.1)
where u as usual represents the velocity vector. Irrotational ows are by deni-
tion those in which the vorticity vector vanishes identically.
11.1.1 Velocity potential
For irrotational ows in which w = 0 identically, the velocity vector must, by
denition, be a potential function. In other words, there should exist a scalar
function (called the velocity potential ) such that
v = . (11.2)
The vorticity vector can be introduced into the Navier-Stokes equations via the
67
LECTURE 11. VORTICITY & BERNOULLIS EQUATION 68
following vector identities:
(v )v =
1
2
U
2
v w, (11.3)
( )v =
2
v = ( div v) w, (11.4)
where U = [v[.
11.2 Bernoullis equation
Consider the incompressible Navier-Stokes equations with constant properties
and with gravity as the sole body force:
div v = 0 (11.5)

v
t
+ (v )v = p gk +
2
v. (11.6)
Introduce the vorticity into both the convective and viscous terms, using the two
vector identities (11.3) and (11.4),

v
t
+
_
1
2
U
2
v w
_
= p gk [w]. (11.7)
Since k is the unit vector in the positive z-direction, we can write gk = (gz),
and thus reduce the above equation to the following form,

v
t
+
_
p + gz +
1
2
U
2
_
= v ww. (11.8)
For irrotational ow (w = 0 and v = ) we have:

t
+ p +gz +
1
2
U
2
_
= 0, (11.9)
which leads to Bernoullis equation

t
+ p +gz +
1
2
U
2
= f(t), (11.10)
LECTURE 11. VORTICITY & BERNOULLIS EQUATION 69
where f(t) is sone arbitrary function of time. For steady ows f(t) constant
and Bernoullis equation reads
p + gz +
1
2
U
2
= const. (11.11)
11.2.1 Boundary conditions for potential ows
The full Navier-Stokes equations admit two velocity boundary conditions at solid
walls, the no slip condition (for tangential velocity) and the vanishing of the
normal velocity (due to wall impermeability).
For potential (irrotational) ows, the assumption of zero vorticity eliminates the
second order derivatives from the equations of motion, leaving only rst order
derivatives, so that only one velocity condition can be satised at a solid wall.
The no slip condition results from uid viscosity and since the eliminated second
order derivatives are precisely the viscous terms, potential ows are not expected
to satisfy the no-slip condition.
In potential ows, therefore, we only require the vanishing of the normal velocity
at a solid wall.
Exercise 11.1 Starting from the Euler equations for inviscid uid ow,

v
t
+(v )v = p gk,
and the additional vector identity
(v w) = (w )v (v )w +v( w) w( v),
Derive a time derivative equation for the vorticity. Hence show that for two-
dimensional v = (u, v, 0) incompressible ow with conservative body forces,
w
3
t
+ u
w
3
x
+ v
w
3
y
= 0,
where w = (w
1
, w
2
, w
3
).
Lecture 12
2-D Flows & the Stream Function
Consider a steady, two-dimensional ow in any coordinate system. For ease of
notation, we will consider rectangular coordinates and assume incompressibility.

In this case, the continuity equation reduces to


u
x
+
v
y
= 0, v = (u, v, 0). (12.1)
If we dene a smooth scalar function , called the stream function, such that
u =

y
, and, v =

x
, (12.2)
then direct substitution shows that the continuity equation (12.1) is identically
satised.
12.1 Stream function & streamlines
To obtain the relationship between streamlines and the stream function, consider
the total derivative:

The conclusions arrived at are however valid in any other orthogonal curvilinear coordinate
system and also for steady compressible ows.
70
LECTURE 12. 2-D FLOWS & THE STREAM FUNCTION 71
d =

x
dx +

y
dy,
= vdx +udy,
= v dA,
= d

V .
It follows that
d = 0, (12.3)

dx
u
=
dy
v
, (12.4)
and d

V = 0. (12.5)
We notice that the lines of constant (12.3) are precisely the streamlines (12.4)
of the ow and hence also the lines across which there is no volume ow (12.5).
By integrating (12.3) we also notice that the dierence between the values of of
any two streamlines is numerically equivalent to the volume ow between those
two streamlines.
12.2 2-D momentum equations
The vorticity for a 2-D ow v = v(t, x, y) is w = (0, 0, w
3
) where
w
3
=
v
x

u
y
=

x
_

x
_


y
_

y
_
=
2
. (12.6)
Now consider the momentum equations for a 2-D incompressible ow in the
following form

Dv
Dt
= (p + gz) +
2
v. (12.7)
LECTURE 12. 2-D FLOWS & THE STREAM FUNCTION 72
Taking the curl of equation (12.7) and noting that (p + gz) is a conservative
eld (has zero curl) gives
Dw
Dt
=
2
w, (12.8)
of which the only nontrivial equation is
w
3
t
+u
w
3
x
+v
w
3
y
=
2
w
3
. (12.9)
Using the relation (12.6), equation (12.9) can be rewritten as a fourth order
partial dierential equation with the stream function as the only variable:

t
(
2
) +

y

x
(
2
)

x

y
(
2
) =
4
. (12.10)
The boundary conditions for u (usually no slip) would be in terms of /y and
those for v would be in terms of /x.
Exercise 12.1 Show that the stream function for a 2-D irrotational ow satises
Laplaces equation.
Exercise 12.2 Using the stream function formulation, determine the streamline
pattern for a 2-D ow with velocity components:
u = U
(x
2
y
2
)
(x
2
+ y
2
)
2
, v = U
(2xy)
(x
2
+y
2
)
2
.
Exercise 12.3 Introduce

= /(U
0
L) and other relevant dimensionless vari-
ables from Lecture 2 to recast equation (12.10) in dimensionless form, identifying
the only dimensionless parameter that arises.
Lecture 13
Compressible Flow
13.1 Introduction to thermodynamics & com-
pressible ow
If a given mass of material undergoes volumetric change when subjected to ex-
ternally applied forces, the material is said to be compressible. Compressibility
is exhibited much more markedly in gases since the mean free molecular paths
are much greater in gaseous media.
The pressure p, density, and temperature T are related through an equation of
state which for a perfect

or ideal

gas is simply
p = RT or pV = RT, (13.1)
where V = 1/ is the volume of a unit mass of the gas and R is the universal gas
constant.

When the molecules of a gas have negligible volume and there are no intermolecular forces,
we have the case of a perfect gas
73
LECTURE 13. COMPRESSIBLE FLOW 74
13.1.1 The rst law of thermodynamics
The mathematical statement of the rst law of thermodynamics is
dE = dQdW, (13.2)
where dE is the change in internal energy of the system, dQ is the quantity of
heat added to the system and dW is the work done by the system. Note that
dW has units of force times distance, i.e. pressure times area times distance or
simply pressure times volume:
dQ = dE + pdV. (13.3)
13.2 Specic heats of a perfect gas
Equation (13.1) implies that T = T(p, ) or T = T(p, V ). In general, any of p,
and T may be expressed in terms of the other two, thus we may take:
E = E(, T) or E = E(V, T), (13.4)
so that
dE =
_
E
V
_
T
dV +
_
E
T
_
V
dT, (13.5)
and
dT =
_
T
p
_
V
dp +
_
T
V
_
p
dV. (13.6)
Equation (13.5) reads:
dE =
_
E
V
_
T
dV +
_
E
T
_
V
__
T
p
_
V
dp +
_
T
V
_
P
dV
_
, (13.7)
so that the rst law of thermodynamics now gives:
dE =
_
E
V
_
T
dV +
_
E
T
_
V
__
T
p
_
V
dp +
_
T
V
_
P
dV
_
+pdV. (13.8)
LECTURE 13. COMPRESSIBLE FLOW 75
We can now obtain the specic heat at constant pressure, c
p
:
c
p
=
_
Q
T
_
p
, (13.9)
and the specic heat at constant volume, c
v
:
c
v
=
_
Q
T
_
V
. (13.10)
Equations (13.8) and (13.10) give:
c
v
=
_
E
T
_
V
_
T
p
_
V
_
p
T
_
V
, (13.11)
i.e.
c
v
=
_
E
T
_
V
. (13.12)
Similarly, we can show that
c
p
= c
v
+
_
p +
_
E
V
_
T
_ _
V
T
_
p
. (13.13)
Equation (13.13) show that for a compressible substance, c
p
,= c
v
, whereas for an
incompressible one, c
p
= c
v
, (since in this case and hence V is constant so that
(V/T)
p
= 0).
For a perfect gas, E = E(T) and hence:
c
v
=
dE
dT
, (13.14)
and
c
p
= c
v
+ p
_
V
T
_
p
. (13.15)
Furthermore, the equation of state for a perfect gas (13.1) i.e. (PV = RT) shows
that keeping p constant leads to
p
_
V
T
_
p
= R,
LECTURE 13. COMPRESSIBLE FLOW 76
hence equation (13.15) becomes
c
p
= c
v
+ R. (13.16)
Additionally, the kinetic theory of gases shows that for a perfect gas, E(T) varies
linearly with T so that equation (13.14) implies that
c
v
= constant.
Equation (13.16) then shows that c
p
is also constant so that
c
p
c
v
= , (13.17)
where = 1 + R/c
v
is called the adiabatic constant. We thus have
c
v
=
R
1
, (13.18)
c
p
=
R
1
. (13.19)
For air and most diatomic gases like oxygen (O
2
), hydrogen (H
2
) and nitrogen
(N
2
), = 1.4. Observe that, for a perfect gas, equation (13.14) implies that
E(T) = c
v
T,
if E(0) = 0. Under these circumstances, there is no internal energy when T = 0.
The temperature T = 0 is then called the absolute zero and the corresponding
temperature scale is called the absolute or Kelvin scale.
13.2.1 Functions of state
Let be some thermodynamic function expressible in terms of any two of the
measurable state variables p, T & V (= 1/). If in changing from an initial state
LECTURE 13. COMPRESSIBLE FLOW 77
A to another state B, the change in depends only on the conditions at A and
B and not at all on the paths joining them, then is called a function of state.
Suppose, say, that is expressible as = (p, V ) and that
d = M(p, V )dp +N(p, V )dV. (13.20)
If
B

A
is independent of the path taken in changing from the conditions at
A to those of B, then d must be an exact dierential and hence (from ODEs)
the necessary condition for this is
_
M
V
_
p
=
_
N
p
_
V
. (13.21)
If equation (13.21) holds, then (p, V ) is a function of state. From the rst law
of thermodynamics (13.3) we have
dQ = dE + pdV.
For a perfect gas, dE = c
v
dT. Since pV = RT then dT = d(pV )/R, and so
dE = c
v
/R d(pV ), which in turn implies that
dQ = pdV +
c
v
R
(pdV + V dp)
=
_
1 +
c
v
R
_
pdV +
c
v
R
V dp
=
c
p
p
R
dV +
c
v
V
R
dp. (13.22)
Since c
p
,= c
p
, then it immediately follows from test (13.21) that dQ is not an
exact dierential for a perfect gas. Thus the quantity of heat, Q, that must be
added to a unit mass of perfect gas to eect a change of state from A to B is
not a function of state but instead does depend on the path taken as well as the
values of p, V, T at A and B.
LECTURE 13. COMPRESSIBLE FLOW 78
13.2.2 Entropy
For theoretical discussions, it is desirable to nd another thermodynamic variable
closely related to Q and which is a function of state. Consider the function S
such that
dS =
dQ
T
, (13.23)
then
dS =
c
p
p
RT
dV +
c
v
V
RT
dp
=
c
p
V
dV +
c
v
p
dp. (13.24)
In this case
_

p
_
c
p
V
_
_
V
= 0 =
_

V
_
c
v
p
__
p
, (13.25)
showing that dS = dQ/T is an exact dierential of a function of state S. In fact,
equation (13.24) implies in moving from A to B that
S
B
S
A
= c
p
ln
_
V
B
V
A
_
+ c
v
ln
_
p
B
p
A
_
. (13.26)
If in particular, say ln V
0
= ln p
0
= 0, then
S S
0
= c
p
ln V + c
v
ln p,
which simplies as
ln (V
c
p
p
c
v
) = S S
0
ln
_
(V
c
v
)
c
p
c
v
p
c
v
_
= S S
0
ln [(pV

)
c
v
] = S S
0
LECTURE 13. COMPRESSIBLE FLOW 79
ln (pV

) =
S S
0
c
v
pV

= exp
_
S S
0
c
v
_
(13.27)
or
p

= exp
_
S S
0
c
v
_
. (13.28)
The quantity S is called the entropy per unit mass of gas and is a function of
state for a perfect gas. Fluid ows for which S is constant are called isentropic
ows.
Isentropic ows satisfy
pV

= constant or
p

= constant. (13.29)
Proof: Consider the general case given by equation (13.26):
S
B
S
A
c
v
= ln
_
V
B
V
A
_
+ ln
_
p
B
p
A
_
.
Isentropic implies that S
A
= S
B
which in turn leads to
ln
_
V

B
V

p
B
p
A
_
= 0,
showing that
p
B
V

B
= p
A
V

A
,
or in other words
pV

= constant.

LECTURE 13. COMPRESSIBLE FLOW 80


13.3 Flow of an inviscid gas
The equation of continuity of a compressible uid reads:

t
+ (v) = 0, (13.30)
and in the absence of viscosity (i.e. inviscid ow), the Euler equations of motion
are
v
t
+ (v )v =
1

p +F. (13.31)
In the case of steady motion with no body forces, these equations reduce to
(v) = 0, (13.32)
(v )v =
1

p. (13.33)
However,
(v )v =
1
2
U
2
v w,
where U = [v[ is the magnitude of the velocity vector and w = v is the uid
vorticity, hence for irrotational ow, equation (13.33) becomes
1
2
U
2
+
1

p = 0,
and taking the dot product with dr = (dx, dy, dz) gives the Bernoulli equation
1
2
dU
2
+
1

dp = 0, (13.34)
or
1
2
U
2
+
_
dp

= constant. (13.35)
In the special case of isentropic ow, the entropy of each uid particle stays
constant along any streamline and for each such particle p = k

so that equation
LECTURE 13. COMPRESSIBLE FLOW 81
(13.35) reduces to
1
2
U
2
+
_
k
(1)

d = constant

1
2
U
2
+
k
1

(1)
= constant

1
2
U
2
+
1
1
k
(1)

= constant

1
2
U
2
+
a
2
1
= constant, (13.36)
where we have made use of the speed of sound, a, dened by:
a
2
=
k
(1)

=
p

or a
2
=
dp
d
.
For motion started from rest, equation (13.36) implies that along a stream line:
1
2
U
2
+
a
2
1
=
1
2
0
2
+
a
2
0
1
, (13.37)
where a
0
is the speed of sound at the point where v = 0 (i.e. speed of sound at a
stagnation point). Thus Bernoullis equation for steady, irrotational, compressible
isentropic ow may be written:
1
2
( 1)U
2
= a
2
0
a
2
. (13.38)
13.3.1 1-D steady ow in a nozzle
-
-
-
-
x
LECTURE 13. COMPRESSIBLE FLOW 82
Suppose A(x) is the cross-sectional area of the nozzle. Then for one-dimensional
steady ow of a compressible uid in the nozzle, the conservation of mass gives:

x
(uA) = 0
uA = constant (13.39)
ln(uA) = constant
ln + ln u + ln A = constant

d +
1
u
du +
1
A
dA = constant. (13.40)
From Bernoullis equation (13.34), we have
udu +
dp

= 0.
But for isentropic ow, dp = a
2
d hence
udu +
a
2

d = 0,
or
1

d =
u
a
2
du.
Also
a =
u
M

1
a
2
=
M
2
u
2
,
so
1

d =
M
2
u
u
2
du =
M
2
u
du,
where M = u/a is the local Mach number. Substituting the results in equation
(13.40) leads to:
(1 M
2
)
du
u
=
dA
A
. (13.41)
If M < 1, equation (13.41) shows that a decrease in A (i.e. dA < 0) produces
an increase in u (i.e. du > 0) and conversely. Thus to accelerate subsonic ow
LECTURE 13. COMPRESSIBLE FLOW 83
through a duct, it is necessary to decrease the duct section area, A, downstream
of the ow.
If M > 1, equation (13.41) shows that
[1 M
2
[
du
u
=
dA
A
,
and hence A and u increase or decrease together. Thus to accelerate supersonic
ow through a duct, it is necessary to widen the duct downstream of the ow.
Summary
-
-
-
-
x
M < 1 M = 1
M > 1
Exercise 13.1 For a steady one-dimensional ow of a compressible uid under
no body forces, what is the for of the relation between the pressure and the velocity
at two points in the ow if
(a) the ow is isothermal
(b) the ow is isentropic
Exercise 13.2 Show that if a compressible uid ows steadily through a tube of
slowly varying cross-section A = A(x), then neglecting external forces:
1
A
dA
dx
=
u
2
a
2
a
2
u
2
du
dx
,
where x is the distance along the tube and a is the local speed of sound.
Lecture 14
Sound Waves
We shall see shortly that a small disturbance in a compressible uid is prop-
agated throughout the uid as a wave motion and at a nite speed called the
speed of sound. In an incompressible uid these features are not realized and
instead a small disturbance applied at any point of such a uid is transmitted
instantaneously throughout the entire uid.
14.1 Speed of sound in a gas
We suppose that a small disturbance is generated within an inviscid uid in
accordance with the following assumptions:
1. The disturbance is propagated as a wave motion, known as a sound wave.
2. Prior to the disturbance, the uid is at rest with velocity v = 0. The
motion is therefore irrotational so that a velocity potential exists at each
point of the uid such that v = .
3. The squares and products of all disturbance quantities can be neglected.
Also U = [v[ is so small that U
2
is negligible.
84
LECTURE 14. SOUND WAVES 85
4. The isentropic law p/

= constant = p
0
/

0
holds as a consequence of
assuming that heat exchanges and hence also entropy changes take place
so rapidly as to be negligible.
We write the disturbance to the density as
=
0
(1 + s), (14.1)
where the disturbance to the density s = s(t, x) is assumed small. The continuity
equation

t
+ (v) = 0,
thus becomes:
s
t
+ [(1 + s)] = 0. (14.2)
In equation (14.2) [s[ is so small as to be negligible hence
s
t
+
2
= 0. (14.3)
The appropriate equations of motion can be obtained from the inviscid Euler
equations with no body forces:
v
t
+ (v )v =
1

p. (14.4)
According to assumption (3) we can neglect the term (v )v thus:

t
() =
1

p.
Taking the dot product with dr leads to

t
(d) =
1

dp or d
_

t
_
=
1

dp,
On integration, this result yields

t
+
_
p
p
0
dp

= 0. (14.5)
LECTURE 14. SOUND WAVES 86
From assumption (4) we have that p = p
0

0
hence:
dp
d
=
p
0

1
=
p
0

0
_

0
_
1
= a
2
_

0
_
1
where a
2
=
p
0

0
. (14.6)
From equation (14.1) /
0
= 1 + s hence equation (14.6) implies that
dp
d
= a
2
(1 + s)
1
a
2
, (14.7)
where the last approximation in equation (14.7) follows from s 1. We thus
have
dp

=
a
2

d,
and hence
_
p
p
0
dp

=
_

0
a
2

d = a
2
ln
_

0
_
= a
2
ln(1 + s) a
2
s.
The nal approximation in the above result was obtained from the Taylor expan-
sion of ln(1 + s). Equation (14.5) thus becomes:

t
+ a
2
s = 0, (14.8)
so that
s
t
=
1
a
2

t
2
,
and hence equation (14.3) leads to

t
2
= a
2

2
. (14.9)
LECTURE 14. SOUND WAVES 87
Equation (14.9) represents a wave type motion with wave speed a. This equation
thus shows that small disturbances are propagated in a gas with speed a, where
a
2
=
dp
d
=
p
0

0
. (14.10)
a is called the speed of sound in the gas and we usually write
a
2
=
_
dp
d
_
S
,
to emphasize that equation (14.10) is obtained under isentropic conditions. For
ideal gases under isentropic conditions, we have
p

= constant say p = C

,
hence
dp
d
= C
1
=
p

= RT,
where the last equality follows from the ideal gas law: p = RT. For the case of
an ideal gas under isentropic conditions, the speed of sounds is thus written as:
a =
_
RT. (14.11)
14.2 Mach waves
Let u be the speed of a gas at a certain location and let a be the local speed of
sound. The local Mach number is the dimensionless parameter dened by:
M =
u
a
. (14.12)
If:
u a then M 1 and we have incompressible ow,
u < a then M < 1 and we have subsonic ow,
LECTURE 14. SOUND WAVES 88
u a then M 1 and we have transonic ow,
u = a then M = 1 and we have sonic ow,
u > a then M > 1 and we have supersonic ow,
u a then M 1 and we have hypersonic ow.
The Mach number plays a dominant role in aerodynamics because, for example,
subsonic and supersonic ows have markedly dierent features.
First consider a source O emitting spherical sound waves in a gas at rest:
r
+
?
O
at
2at
S
1
S
2
Figure 14.1: Sound waves emitted by a gas at rest.
Spherical wave fronts centered at O travel outwards from O. At time t after
emission from O, the disturbance is spread uniformly over the surface S
1
of the
LECTURE 14. SOUND WAVES 89
sphere of center O and radius at. At successive times 2t, 3t, etc. the disturbance
lies on the surfaces S
2
, S
3
, etc. of concentric spheres centered at O and with radii
2at, 3at, etc.
Now suppose that the gas ows with uniform velosity u past the source (or the
source moves through the gas with speed [u[). Then
(i)
t t t
t
O O
1
O
2

at
2at
3at

OO
1
= ut

OO
2
= 2ut

OO
3
= 3ut
etc.

-
body
gas
Figure 14.2: Sound wave pattern under subsonic conditions.
when u < a (i.e. M < 1) every particle of say S
1
is displaced a distance
LECTURE 14. SOUND WAVES 90
ut and the disturbance which was initially at O now lies on the surface S

1
of a sphere of center O
1
and radius at where

OO
1
= ut. Thus if at times
t = 0, t, 2t, 3t, . . . the body (source) is at positions O, O
1
, O
2
, O
3
, . . . then,
since the waves spread radially outward from their point of origin at the
speed of sound, the wave pattern at time 3t will be as shown in gure 14.2.
(ii) when u > a (and hence M > 1), it is observed that all the waves lie within
the cone indicated in gure 14.3
s r s s

-
body
gas
ut ut
ut
3at
2at
at
O O
1
O
2 O
3

Figure 14.3: Sound wave pattern under supersonic conditions, Mach cone.
Figure 14.3 has its vertex at the body (source) and only the gas that lies
within the Mach cone is aware of the presence of the body. In other
words, the disturbance when M > 1 is conned to the interior of the Mach
LECTURE 14. SOUND WAVES 91
cone. The cone has a half angle where
sin =
at
ut
=
2at
2ut
=
3at
3ut
= . . . =
a
u
,
hence
sin =
1
M
,
and is called the Mach angle. The Mach angle is real only when M 1
and non-existen for subsonic ow. In two-dimensions, the spheres become
circles and the cone becomes a pair of lines called Mach lines or Mach waves.
The basic dierences between subsonic and supersonic ow can thus be summa-
rized as follows:
In subsonic ow, an innitesimal disturbance (sound wave) will eventually
be felt throughout the entire ow. For example, in the case of an aircraft
ying overhead at subsonic speed, any observer on the ground will hear
the disturbance (while the aircraft is still directly overhead) once the sound
waves have spread out to the observers position.
In supersonic ow, a disturbance is only felt over a potion of the ow (the
region inside the Mach cone). Thus one may spot a supersonic aircraft or
missile travelling overhead but only hear the sound waves a little afterwards
after the aircraft/missile has moved signicantly away from the overhead
position and the Mach cone has eventually enclosed the observers position.
Aerodynamical calculations involving sound speed invariably require knowledge
of the variation of atmospheric conditions like temperature with, say, altitude.
In such calculations, a standard atmosphere is usually invoked:
US Standard Atmosphere: In this case, the temperature, T, in the inner
LECTURE 14. SOUND WAVES 92
potion of the atmosphere is related to the altitude, H, via:
T =
_

_
288.16 0.0065H, 0 H 11019m,
216.66, dH > 11019m,
where H is measured in meters and T in Kelvins.
where H = 0 is understood to mean sea level.
Example 14.1
t

Aircraft
Observer

5km
12km
Mach wave
Mach wave
An observer on the ground nds out that an aircraft ying horizontally at an
altitude of 5000m has traveled 12km from the overhead position before the sound
of the aircraft is rst heard. Estimate the speed at which the aircraft is ying
Since temperature varies through the atmosphere, the speed of sound will vary as
the sound waved move through the atmosphere. We will use the sound speed at
the average temperature between the ground and the aircraft to describe the Mach
wave.
LECTURE 14. SOUND WAVES 93
At H = 2500m, the temperature is T = 288.16 0.0065(2500) = 271.9K. The
mean speed of sound is thus given by:
a =
_
RT =

1.4 287.04 271.9 = 330.6m/s.


If is the Mach angle based on the mean speed of sound, then
tan =
5000
12000
=
5
12
.
But since sin = 1/M we deduce that
tan =
1

M
2
1
,
hence
M =

_
12
5
_
2
+ 1 = 2.6,
from which it follows that the speed of the aircraft u is
u = Ma = 330.6 2.6 = 859.6m/s.
Exercise 14.1 An aeroplane is ying at 2000km/h at an altitude where the tem-
perature is 50

C. Calculate the Mach number at which the aeroplane is ying.


Exercise 14.2 An aeroplane can y at a speed of 800km/h at sea level where
the temperature is 15

C. If the aeroplane ies at the same Mach number at


an altitude where the temperature is 44

C, calculate the speed at which the


aeroplane is ying at this altitude.
Exercise 14.3 Air at 10

C is owing at a Mach number of 1.9. Calculate the


air velocity and the Mach angle.
Exercise 14.4 An aircraft is ying at an altitude of 6km at a Mach number of
3. Find the distance behind the aircraft at which the soundwaves created by the
aircraft reach sealevel. The temperature at 6km is 249.16K.
Lecture 15
Boundary layer theory
15.1 Background
The concept of drag is important in uid dynamics especially in the design of
ecient aerodynamical objects (say race cars and aircrafts) or submersibles (e.g.
submarines). The drag results from the friction between the surface of the solid
body and the uid layers. Frictional eects are however only important close to
the surfaces of the solid bodies. The boundary layer concept was introduced to
explain this phenomenon. In other words, the eects of viscosity is signicant
only in thin layers of uid close to rigid boundaries - the so called boundary layers.
Further from the rigid body, viscosity loses its potency and the uid motion is
eectively inviscid.
15.1.1 Terminology
The following terminology is closely connected with boundary layer ow:
1. Stagnation points: points in the ow where the uid is stationary.
2. Inner ow: regions in the ow where large velocity gradients and strong
94
LECTURE 15. BOUNDARY LAYER THEORY 95
viscous eects exist.
3. Outer ow: region outside the boundary layer in which the ow can be
considered inviscid (and irrotational if incident ow is uniform).
4. Separation points: points on the surface of the body where tangential
stresses (skin friction) vanish and the boundary layer detaches from the
body leading to a
5. Wake region: a region in the ow characterized by adverse pressure gra-
dients and in which velocity gradients become smaller and viscous eects
cease to be of importance.
-
-
1
2
2
3
3
4
4
5
-
-
-
15.2 Boundary layer thickness
The exact thickness of the boundary layer cannot of course be determined but
we could dene the boundary later thickness (x) to be the value of y at which
the component of the velocity parallel to the body reaches about 99% of the
free-stream velocity, say U, i.e.
(x) = y when u(x, y) = 0.99U.
LECTURE 15. BOUNDARY LAYER THEORY 96
-
6
-
-
-
-
6
?
y
x
(x)
U
U
Alternatively, for a laminar ow over a at plate, the boundary layer thickness
is usually dened as:
(x) =
5x

R
x
,
where R
x
= Ux/ is the local Reynolds number of the ow along the at plate
and is the uid kinematic viscosity.
15.2.1 Displacement thickness
The displacement thickness (

) is a measure by which the streamlines just out-


side the boundary layer are displaced from their original paths by the action of
viscosity within the boundary layer. The displacement thickness is derived from
consideration of mass conservation.
Consider a uniform stream parallel to the x-axis and suppose that the height of
the uid above the plane y = 0 is h.
A at plate is inserted into the stream at y = 0, x = 0 creating a boundary layer
and hence displacing the uid by a distance

. The new height of the uid above


the plate is now:
(x) =

(x) + h.
LECTURE 15. BOUNDARY LAYER THEORY 97
To satisfy conservation of mass, the mass ux must remain constant:
_
h
0
U(x)dy =
_
(x)
0
u(x, y)dy. (15.1)
The velocity u(x, y) within the boundary layer satises u(x, y) < U(x) and hence
equation (15.1 implies that:
U(x)h =
_
(x)
0
[U(x) + u(x, y) U(x)]dy
= U(x)(x) +
_
(x)
0
[u(x, y) U(x)]dy,
h = (x) +
_
(x)
0
_
u
U
1
_
dy,
(x) = h +
_
(x)
0
_
1
u
U
_
dy = h +

(x),

(x) =
_
(x)
0
_
1
u
U
_
dy =
_

0
_
1
u
U
_
dy, (15.2)
15.3 Boundary layer equations
The boundary layer equations are derived from the Navier-Stokes equations by
order of magnitude arguments. Consider a uniform stream with uniform speed
U over a at plate of length L over which the boundary layer has formed. The
steady Navier-Stokes equations for 2-D ow of a viscous and incompressible uid
are:
u
x
+
v
y
= 0, (15.3)
u
u
x
+ v
u
y
=
1

p
x
+
_

2
u
x
2
+

2
u
y
2
_
, (15.4)
u
v
x
+ v
v
y
=
1

p
y
+
_

2
v
x
2
+

v
y
2
_
. (15.5)
LECTURE 15. BOUNDARY LAYER THEORY 98
If we denote the boundary layer thickness by , then
u = O[U], x = O[L], y = O[], (15.6)
and equation (15.6) then implies
u
x
= O
_
U
L
_
,
u
y
= O
_
U

_
. (15.7)
For all terms of equation (15.3) to be of the same order of magnitude, we require
that:
v
y
= O
_
U
L
_
. (15.8)
At y = 0, the no-slip & impermeablity conditions imply that u
y=0
= v
y=0
= 0 so
that
v
y
= O
_
v
y=
v
y=0

_
= O
_
U
L
_
, (15.9)
hence
v
y=
= O
_
U
L
_
. (15.10)
The magnitudes of the second order derivatives are:

2
u
y
2
= O
_
U

2
_
,

2
u
x
2
= O
_
U
L
2
_
, (15.11)
and since L, we have from equations (15.11) that

2
u
y
2


2
u
x
2
, (15.12)
and hence, when only the dominant viscous term is retained in equation (15.4),
we have:
u
u
x
+ v
u
y
=
1

p
x
+

2
u
y
2
. (15.13)
LECTURE 15. BOUNDARY LAYER THEORY 99
For all terms in equation (15.13) to be of the same order of magnitude, we require
that
p = O
_
U
2
_
,
v

2
= O
_
U
L
_
, (15.14)
which in turn implies that

2
= O
_
L
U
_
, or

2
L
2
= O
_
1
Re
_
, (15.15)
where Re = UL/ is the Reynolds number. A similar analysis applied to the
y-component (equation 15.5) gives:
O
_
U
2

L
2
_
+ O
_
U
2

2
L
2
_
=
1

p
y
+ O
_
U
2

3
L
4
_
+ O
_
U
2

L
2
_
, (15.16)
where

p
y
= O
_
U
2

_
,
is the dominant term. Neglecting the small terms (those terms that are multiples
of
n
, n 1) from equation (15.16) yields:

p
y
= 0 and hence p = p(x). (15.17)
The equations which apply in the boundary layer when only the dominant terms
are retained are therefore:
u
x
+
v
y
= 0, (15.18)
u
u
x
+ v
u
y
=
1

dp
dx
+

2
u
y
2
, (15.19)
subject to the boundary conditions
u = v = 0 on y = 0, (15.20)
and the asymptotic conditions
u U, v 0 as y . (15.21)
LECTURE 15. BOUNDARY LAYER THEORY 100
The boundary layer merges smoothly with the outer inviscid ow without any
discontinuities in pressure and velocity. We can therefore use Bernoullis equation
to obtain the pressure in the boundary layer:
p

+
1
2
U
2
(x) = constant, (15.22)
or
1

dp
dx
+ U(x)
d
dx
U(x) = 0. (15.23)
Here U(x) is a general function of x. For the ow over a at plate, U(x) = U =
constant and hence the pressure is constant over a at plate.
The boundary layer equations to be solved for the velocity distributions u(x, y)
and v(x, y) are therefore:
u
x
+
v
y
= 0, (15.24)
u
u
x
+ v
u
y
= U(x)
d
dx
U(x) +

2
u
y
2
, (15.25)
subject to the boundary conditions
u = v = 0 on y = 0, (15.26)
and the asymptotic conditions
u U(x), v 0 as y . (15.27)
The outer ow velocity U(x) is assumed to be known from the analysis of the
outer inviscid (potential) ow. It must be borne in mind that these equations
are only valid for ows at high Reynolds numbers (i.e. 0).
Lecture 16
Exact Solutions
We examine some exact solutions of the boundary layer equations.
16.1 Blasius solution
-
6
-
-
y
x
U
6
?
(x)
boundary layer
-
-
-
U
u(x, y)
at plate
O
Consider a at plate placed along the direction of a uniform stream of velocity
v = U

i where U is constant. Since dU/dx 0, the boundary layer equations for


101
LECTURE 16. EXACT SOLUTIONS 102
the present problem reduce to:
u
x
+
v
y
= 0, (16.1)
u
u
x
+ v
u
y
=

2
u
y
2
, (16.2)
subject to the boundary conditions
u = v = 0 on y = 0, u U, v 0 as y . (16.3)
Equations (17.1) is satised by dening the stream function so that
u =

y
, v =

x
. (16.4)
We introduce a dimensionless coordinates dened by
=
y

2
=
_
U
2x
_
1
2
y. (16.5)
A corresponding dimensionless stream function f() is dened by:
= (2Ux)
1
2
f(), (16.6)
where the factor 2 is purely for convenience. We have:

y
=
_
U
2x
_
1
2
,

x
=

2x
,

= (2Ux)
1
2
f

(),
hence
u =

y
=

y
=
_
U
2x
_
1
2
(2Ux)
1
2
f

()
= Uf

(), (16.7)
LECTURE 16. EXACT SOLUTIONS 103
Similarly
v =

x
=

x
_
(2Ux)
1
2
f()
_
=
_
(2Ux)
1
2
f

()

x
+
1
2
(2U)
1
2
x

1
2
f()
_
=
_
(2Ux)
1
2
f

()

2x
+ (2Ux)
1
2
f()
2x
_
=
_
U
2x
_
1
2
(f

() f()). (16.8)
The velocity derivatives appearing in the boundary layer equations can thus be
calculated:
u
x
=
U
2x
f

(), (16.9)
u
y
= U
_
U
2x
_
1
2
f

(), (16.10)

2
u
y
2
=
U
2
2x
f

(). (16.11)
Substituting equations (16.7) - (16.11) into boundary layer x- direction equation
(16.2) leads to:
Uf

U
2x
f

_
+
_
U
2x
_
1
2
(f

f) U
_
U
2x
_
1
2
f

=
U
2
2x
f

,
which simplies as:
U
2
2x
f

+
U
2
2x
ff

= 0,
and hence the function f() satises the equation:
f

() + f()f

() = 0. (16.12)
LECTURE 16. EXACT SOLUTIONS 104
Since
y = 0 = 0 and y ,
the boundary conditions (16.3) require that:
0 = u = Uf

(0) at = 0,
0 = v =
_
U
2x
_
1
2
(0 f(0)) at = 0,
Uf

() U as
and hence the boundary conditions on f() are:
f(0) = f

(0) = 0, f() = 1. (16.13)


The boundary value problem (16.12, 16.13) for f() is a third order nonlinear
ordinary dierential equation and has no analytic solution. The nature of the
boundary conditions (imposed at both 0 and ) complicates numerical tech-
niques. If we however suppose that
f

(0) = , (16.14)
where is to be found, then the series expansion of f() in the neighborhood of
= 0 can be developed using equations (16.12) and (16.13). For example since
f

(0) = f(0)f

(0) = 0,
and
f
(iv)
() + f()f

() + f

()f

() = 0,
it follows that
f
(iv)
= 0.
LECTURE 16. EXACT SOLUTIONS 105
Similarly, since
f
(v)
() + f()f
(iv)
() + f

()f

() + f

()f

() + (f

())
2
= 0,
it follows that
f
(v)
=
2
,
etc. In general, only the terms f
(2+3n)
(0), n = 0, 1, 2, . . . are dierent from zero.
We then have the power series expansion for f():
f() =

2
2!

2

5
5!
+ 11
3

8
8!
. . . . (16.15)
This may be written as
f() =

n=0
(1)
n
C
n

n+1

3n+2
(3n + 2)!
, (16.16)
where for example the rst six coecients are:
C
0
= 1; C
1
= 1; C
2
= 11; C
3
= 375; C
4
= 27, 897; C
5
= 3, 817, 137.
We next determine the value of . Suppose F() is a solution when = 1, i.e.
F() =

2
2!


5
5!
+ 11

8
8!
. . . . (16.17)
Equation (16.16) can then be written in the form:
f() =
1
3
F(
1
3
). (16.18)
Since F() is a solution, equation (16.13) implies that
F(0) = F

(0) = 0, F

(0) = 1.
Also, since
lim

() = lim

[
2
3
F

(
1
3
)] = 1,
LECTURE 16. EXACT SOLUTIONS 106
it follows that
=
_
1
lim

()
_3
2
. (16.19)
Equation (16.12) may be integrated numerically to give F, F

and F

, the in-
tegration being started from the origin and continued until F

is constant to a
sucient approximation. In this manner, more correct values of may be found.
For a boundary layer on a at plate, the nal value of = f

(0) is
= 0.4696, (16.20)
to four decimal places. We can obtain the behavior of f() for both small and
large values of as follows:
For small values of , Taylor series expansions of f and f

are:
f() = f(0) + f

(0) +

2
2!
f

(0) + O(
3
),
(16.21)
f

() = f

(0) + f

(0) +

2
2!
f

(0) + O(
3
),
and from the boundary conditions (16.13), we have
f()

2
2!
f

(0),
(16.22)
f

() f

(0).
For large values of (i.e. ), boundary conditions (16.13) also imply
that
f

() 1 as . (16.23)
We may therefore write
f

() 1 +

(), (16.24)
LECTURE 16. EXACT SOLUTIONS 107
where

() 0 as .
Equation (16.24) implies that
f() + () + , (16.25)
where is some constant. Equations (16.25) and (16.12) imply that

+ [( ) + ]

= 0. (16.26)
For large values of , and its derivatives are small so the term

may be
neglected in equation (16.26) leading to:

+ ( )

= 0. (16.27)
Equation (16.27) can be regarded as a rst order ordinary dierential equation
in

, in which case we have:

= c
0
exp
_

1
2
( )
2
_
, (16.28)
where c
0
is an integration constant. Consequently,

= c
1
_

2
d + c
2
, (16.29)
where c
1
and c
2
are constants, = and = /

2. Since

= 0 as ,
the limit must be and c
2
= 0. Thus

= c
1
_

1
2
()
2
d, (16.30)
and hence
= c
1
_

d
_

1
2
()
2
d,
LECTURE 16. EXACT SOLUTIONS 108
leading to
f() = + c
1
_

d
_

1
2
()
2
d. (16.31)
The constants , and c
1
may be determined by joining the solutions (16.16)
and (16.31) at a point =
1
where both solutions are serviceable. f, f

and f

from the power series and the asymptotic solution are brought to agreement.
Having thus found the velocity distribution, we can then proceed to obtain the
boundary layer thickness (

(x) or (x)), shear stress distribution along the solid


surface (
w
) and the drag acting on the surface (D):
u = Uf

()

(x) =
_

0
_
1
u
U
_
dy (16.32)
=
_
2x
U
_
1
2
lim

( f) (16.33)
=
_
2x
U
_
1
2
. (16.34)
(x) = (x) =
5x
_
Ux/
= 5
_
x
U
_1
2
. (16.35)

w
=
_
u
y
_
y=0
=
_
U
3
2x
_1
2
f

(0) =
_
U
3
2x
_1
2
. (16.36)
D =
_
x
0

w
()d = 2
_
U
3
2
_1
2

x. (16.37)
The drag coecient or skin friction coecient for the plate is:
C
D
=
(D/x)
(U
2
/2)
= 4
_

2Ux
_1
2
. (16.38)
LECTURE 16. EXACT SOLUTIONS 109
16.2 Flat plate with varying outer ow
Consider a steady 2-D ow of an incompressible uid past a semi-innite at
plate x 0, y = 0 and suppose that outside the boundary layer, the velocity is
v = U(x)

i. The equations governing the ow within the boundary layer are:


u
x
+
v
y
= 0, (16.39)
u
u
x
+ v
u
y
= U(x)
d
dx
U(x) +

2
u
y
2
. (16.40)
These equations can be manipulated as follows:
2u
u
x
u
u
x
+ v
u
y
= U
dU
dx
+

2
u
y
2
,
which on employing the continuity equation leads to
u
2
x
+ u
v
y
+ v
u
y
= U
dU
dx
+

2
u
y
2
,
or

u
2
x
+
uv
y
= U
dU
dx
+

2
u
y
2
. (16.41)
Integration of equation (16.41) from y = 0 to y = (x) leads to:
_

0

u
2
x
dy +U[v]
y=
= U
dU
dx
+
_
u
y
_
y=0
.
The skin friction on the plate is thus given by:

0
=
_
u
y
_
y=0
=
dU
dx
_

0
(U u)dy +
d
dx
_
_

0
u(U u)dy
_
. (16.42)
Equation (16.42) is known as the momentum integral equation. The upper limit
of the integrals can in fact be replaced by since u = U for y > . Equation
(16.42) is usually written as

0
=
dU
dx
(U
1
(x)) +
d
dx
_
U
2

1
(x)
_
, (16.43)
LECTURE 16. EXACT SOLUTIONS 110
where
1
(x) is the displacement thickness:

1
(x) =
_

0
_
1
u
U
_
dy, (16.44)
and the momentum thickness
1
(x) is dened as:
U
2

1
(x) =
_

0
u(U u)dy. (16.45)
Lecture 17
Similarity Solutions
17.1 Falkner-Skan equation
Consider the boundary layer equations for steady ow with U(x) the component
of the outer ow velocity in the direction of increasing x.
u
x
+
v
y
= 0, (17.1)
u
u
x
+ v
u
y
= U
U
x
+

2
u
y
2
, (17.2)
A similarity solution of the boundary layer equations (17.1 - 17.2) is one for
which the velocity prole at each value of x is the same except for a change of
scale. We look for a similarity solution of the form
u(x, y) = U(x) f

(), (17.3)
where = y/(x), with some (x) some unspecied function, with the dimensions
of length, which has to be found. Furthermore, since the stream function satises:
u =

y
=
1

,
it follows from equation (17.3) that
=
_
u d = U(x) (x) f(). (17.4)
111
LECTURE 17. SIMILARITY SOLUTIONS 112
It follows from equation (17.3) that:
u
x
= U

u
y
= U
f


2
u
y
2
= U
f

2
Equation (17.4) implies that
v =

x
= [U

f + U

Uf

],
Equation (17.2) then gives
f

+ ff

+ [1 (f

)
2
] = 0, (17.5)
where
=

d
dx
(U), =

2

dU
dx
. (17.6)
The boundary conditions for f are:
f(0) = f

(0) = 0, (17.7)
f

() 1 as . (17.8)
Since is dimensionless and x is dimensional, it follows from equation (17.5)
that and must at most be constants. Thus once the constants and are
selected, equation (17.6) determines the function (x) and the outer ow velocity
U(x).
With the determination of and U together with f() (the solution of the BVP)
the stream function for the ow in the boundary layer is thus known through
equation (17.4).
LECTURE 17. SIMILARITY SOLUTIONS 113
Equation (17.5) is known as the Falkner-Skan equation. Included within its set of
solutions is the Blasius ow past a at plate which arises when = 1 and = 0.
In this case, equation (17.6) yields

2
dU
dx
= 0,
d
dx
(U) =

.
Since ,= 0, it follows that
U = constant,
and (ignoring constant of integration)
(x) =
_
2x
U
_
1
2
.
Since = 1 and = 0, the Flakner-Skan equations read
f

() + f()f

() = 0,
with
f(0) = f

(0) = 0, f() = 1,
which expectedly is the same as the Blassius BVP.
The external velocity prole for the Falkner-Skan boundary layers is of the form
U(x) = c x
m
,
where c and m are constants. To see this, consider
d
dx
(U
2
) =
2
dU
dx
+ 2U
d
dx
= 2
2
dU
dx

2
dU
dx
+ 2U
d
dx
= 2
_

dU
dx
+ U
d
dx
_

2
dU
dx
= 2
d
dx
(U)
2
dU
dx
LECTURE 17. SIMILARITY SOLUTIONS 114
= 2
_

d
dx
(U)
_

dU
dx
_
= 2
= (2 ). (17.9)
Integrating equation (17.9) leads to
U(x)
2
= (2 ) x. (17.10)
Equations (17.6) and (17.10) lead to
1
U
dU
dx
=
m
x
, where m =

2
.
This leads to
ln U = mln x + ln c,
and thus
U(x) = c x
m
. (17.11)
The function can now be found from equation (17.6):
=
_
1
c
(2 ) x
1m
_
1
2
. (17.12)
The stream function is thus
= Uf = c x
m
_
1
c
(2 ) x
1m
_
1
2
f()
=
_
(2 )c x
m
x
(1m)/2
f()
=
_
(2 )c x
/(2)
f() (17.13)
where
=
y
(x)
=
y x
(m1)/2
_
1
c
(2 )
_1
2
. (17.14)
LECTURE 17. SIMILARITY SOLUTIONS 115
17.2 Stagnation point ow
Setting = = 1 gives m = 1 and
U(x) = c x,
(x) =
_

c
,
(x) =

c f(),
u(x, y) = c f

(), and
v(x, y) =

c f(),
where f() is the solution of the Flakner-Skan equation:
f

+ ff

+ 1 (f

)
2
= 0,
with boundary conditions
f(0) = f

(0) = 0, f() = 1.
LECTURE 17. SIMILARITY SOLUTIONS 116
17.3 Flow over a wedge
-
>
~
>
~
]

If we set = 1 but allow to be arbitrary, then m = /(2 ) and the outer


potential ow is
U(x) = c x
/(2)
. (17.15)
The Falkner-Skan equation of interest becomes:
f

+ ff

+ [1 (f

)
2
] = 0,
with boundary conditions
f(0) = f

(0) = 0, f() = 1.
The velocity prole (17.15) describes the ow over a wedge of angle . To prove
this would require solving directly for the potential ow past a wedge of angle
, details of which are not of particular interest to the direction of the course.
Lecture 18
Approximate Solutions
Similarity solutions are exact in the sense that the boundary layer equations are
reduced to the non-linear Falker-Skan equation which is an ODE solvable by
numerical methods to any desired degree of accuracy.
Where similarity solutions do not exist, an approximate solution to the boundary
layer equations may be found by assuming the velocity as a polynomial in y whose
coecients are functions of x and utilizing the momentum intergal equation.
18.1 The momentum integral

0
=
dU
dx
_

0
(U u)dy +
d
dx
_

0
u(U u)dy
=
dU
dx
(U
1
) +
d
dx
(U
2

1
), (18.1)
where

1
=
_

0
_
1
u
U
_
dy,
117
LECTURE 18. APPROXIMATE SOLUTIONS 118
is the displacement thickness and

1
=
1
U
2
_

0
u(U u) dy,
is the momentum thickness. As an example, let us consider the steady ow over
a at plate. We assume a representation of u of the form:
u(x, y)
U(x)
= a
0
+ a
1
y
(x)
+ a
2
_
y
(x)
_
2
, (18.2)
where a
0
, a
1
and a
2
are possibly functions of x and (x) is the boundary layer
thickness which is itself unknown. The boundary conditions for u are:
u(x, 0) = 0, u(x, ) = U(x), u
y
(x, ) = 0. (18.3)
Substituting these in equation (18.2) leads to
a
0
= 0, a
1
= 2, a
2
= 1, (18.4)
and hence the expression for u becomes:
u(x, y)
U(x)
= 2
y
(x)

_
y
(x)
_
2
. (18.5)
Substituting equation (18.5) into the momentum integral (18.1) leads to a dier-
ential equation for (x).
For the ow over a at plate, we have that U(x) is constant, say U, and hence
equation (18.1) reduces to

0
= 0 (U
1
) +
d
dx
(U
2

1
)
=
d
dx
_
_

0
u(U u) dy
_
=
d
dx
_
U
2
_

0
u
U
_
1
u
U
_
dy
_
= U
2
d
dx
_
_

0
u
U
_
1
u
U
_
dy
_
. (18.6)
LECTURE 18. APPROXIMATE SOLUTIONS 119
Using equation (18.5), we have:

0
= U
2
d
dx
_
_

0
_
2
y


y
2

2
__
1 2
y

+
y
2

2
_
dy
_
= U
2
d
dx
_
2
15
(x)
_
=
2
15
U
2
d
dx
(x) (18.7)
However,

0
=
_
u
y
_
y=0
= U
_

y
_
u
U
_
_
y=0
= U
_

y
_
2
y


y
2

2
__
y=0
= 2
U
(x)
. (18.8)
Equations (18.7) and (18.8) yield the dierential equation
d
dx
(x) = 15

U (x)
, (18.9)
with boundary condition:
(0) = 0. (18.10)
The solution to this boundary value problem is:
(x) =
_
30
x
U
_1
2
. (18.11)
The shear stress at the wall is thus

0
= 2U
_
U
30x
_
1
2
.
We also note that

0
U
2
=

4
15
_

2Ux
_1
2
= 0.5164
_

2Ux
_1
2
.
LECTURE 18. APPROXIMATE SOLUTIONS 120
This solution compares favorably with the exact Blassius solution:

0
U
2
=
_

2Ux
_1
2
f

(0) = 0.4096
_

2Ux
_1
2
.
Higher degree polynomials in y can be used, which will give better accuracy.
The additional boundary conditions which must be imposed to determine the
coecients, are derived from the boundary layer momentum equation and its
derivatives with respect to y.
A generalization of this method which allows for a variable U(x) and utilizing a
fourth order polynomial in y is known as the von Karman - Polhausen method.
18.2 von Karman - Polhausen method
u
U(x)
= c
0
+ c
1
y
(x)
+ c
2
_
y
(x)
_
2
+c
3
_
y
(x)
_
3
+ c
4
_
y
(x)
_
4
. (18.12)
The functions c
i
, i = 0, 1 . . . 4 are determined from the boundary conditions:
u(x, 0) = 0, u(x, ) = U(x), u
y
(x, ) = 0, (18.13)
u
yy
(x, ) = 0, u
yy
(x, 0) =
U

dU
dx
. (18.14)
It is usually common practice to introduce the dimensionless parameter:
=
y
(x)
,
in which case, the dimensionless velocity u/U is given only in terms of :
u
U
= c
0
+c
1
+ c
2

2
+c
3

3
+ c
4

4
. (18.15)

S-ar putea să vă placă și