You are on page 1of 83

Turbulent Heat Transfer

Scott Stolpa
April 30, 2004
ii
Preface
This paper is an independent research project that combines many different sources in the
field of heat transfer. Its goal is to collect and organize many of the turbulent heat transfer princi-
ples that are fundamental to heat transfer textbooks and combine those with state of the art find-
ings by some of the top researchers in the field. The ultimate aim is to connect everything to the
atmospheric boundary layer, a subject that is central to my own research. The audience is engi-
neering graduate students who have had some heat transfer and some turbulence. It is also neces-
sary to have some understanding of basic fluid mechanics.
iii
CONTENTS
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
CHAPTER 1: INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Principles of Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..1
1.2 Fundamental Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..2
1.3 Modes of Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..4
1.4 Laminar Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..5
1.5 Turbulent Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..6
1.6 External Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..6
1.6.1 Boundary Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6.2 Free Turbulent Flows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.7 Internal Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..9
1.7.1 Entrance Region. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7.2 Fully-Developed Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.8 Principles of Scaling and Non-Dimensionalization . . . . . . . . . . . . . . . . . . . ..11
CHAPTER 2: FORCED CONVECTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..15
2.2 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..15
2.3 Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..16
2.4 The Turbulent Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..17
2.4.1 Turbulence Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5 Boundary Layer Flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..22
2.5.1 Mixing Length Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5.2 Solving the Energy Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.5.3 Wall Heat Flux. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.6 Other External Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..32
2.6.1 Cylinder in Cross Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.6.2 Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.6.3 General Spheroid Body Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.6.4 Array of Cylinders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.7 Internal flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..36
2.7.1 Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
CHAPTER 3: NATURAL AND MIXED CONVECTION . . . . . . . . . . . . . . . . . . . . . . 42
3.1 Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..42
3.2 Boussinesq Approximation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..44
3.3 Natural Convection with a Vertical Wall . . . . . . . . . . . . . . . . . . . . . . . . . . . ..46
iv
3.4 Thermal Stratification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..49
3.5 Natural Convection Over a Flat Horizontal Plate. . . . . . . . . . . . . . . . . . . . . ..49
3.6 Vertical Channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..50
3.7 Mixed Convection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..52
3.7.1 Vertical Flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.7.4 Horizontal Flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
CHAPTER 4: ATMOSPHERIC BOUNDARY LAYER . . . . . . . . . . . . . . . . . . . . . . . . 55
4.1 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..55
4.2 Important Notes on Variable Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..56
4.3 Fluid Mechanics and Governing Equations for the ABL . . . . . . . . . . . . . . . ..56
4.4 Flat, Infinite, Uniform Terrain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..58
4.4.1 ABL Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.4.2 Stability Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.4.3 The Three States of the ABL. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.4.4 Properties of the Surface Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.4.5 Monin-Obukhov Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.4.6 Spectra and Cospectra for the Surface Layer. . . . . . . . . . . . . . . . . . 67
4.4.7 Beyond the Surface Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.5 Plant Canopies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..72
4.6 Changing Terrain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..72
4.7 Hills. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..73
CHAPTER 5: MEASUREMENTS TECHNIQUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.1 Velocity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..74
5.2 Temperature and Heat Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..77
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
1
CHAPTER 1
INTRODUCTION
1.1 Principles of Heat Transfer
Heat transfer is essentially made up of two different mechanisms. Radiation and conduc-
tion. A third mechanism, convection, is the conduction of heat between two objects with a rela-
tive velocity between them. There are many different scenarios under which convection occurs.
It can occur in horizontal flow or vertical flow. It can take place in a duct or in the open air. It can
be laminar or turbulent. Natural convection requires no external flow at all. Buoyancy effects
will create air flow and heat transfer will occur. Convection laws rely on the fundamental princi-
ples of both heat transfer and fluid flow. This includes obeying the laws of mass conservation,
momentum conservation, and energy conservation. Convection is also governed by the first and
second laws of thermodynamics. The simple equation for the convective heat transfer between a
flat plate and a fluid is Newtons Law of Cooling:
. (1.1)
The symbols q'', h, T
o
, and T

represent the heat wall heat flux, the heat transfer coeffi-
cient, the temperature of the flat plate and the temperature of the fluid respectively. Near the wall,
the fluid must obey the no-slip condition, and the velocity is essentially zero at the wall. Without
a relative velocity, the heat transfer is governed by pure conduction and the equation:
q'' h T
o
T

( ) =
2
. (1.2)
These two equations represent the instantaneous heat transfer between two points for conduction
and convection. In order to get the total heat transfer between a fluid and a flat plate, one must
solve the integral:
(1.3)
where q is the total heat transfer and W is the dimension of the plate perpendicular to the stream-
wise direction. The rest of this paper will deal with ways to adapt these equations to find the total
heat transfer between a surface and a fluid or between two fluids.
1.2 Fundamental Equations
The equations for all convective heat transfer are derived from the fundamental principles
that were stated in Section 1.1. This amounts to solving four nonlinear partial differential equa-
tions for the conservation of mass, momentum and energy. In their most general, incompressible,
Boussinesq (which will be discussed in subsequent chapters) forms, these equations read:
. (1.4)
(1.5)
(1.6)
One important thing that was neglected in the energy equation is a viscous dissipation term. The
viscous dissipation term is negligible at low Mach numbers. Since we are assuming incompress-
ible flow, we must also assume low Mach number since at M > 0.3 the fluid can no longer be
q'' k
cT
cy
------
\ .
| |
y 0 =
=
q q''W x d
0
L
}
=
c
i
v
i
0 =

Dv
i
Dt
-------- f
i
c
i
p c
i
c
j
v
i
+ =
c
p
DT
Dt
-------- kc
i
c
j
T =
3
assumed to be incompressible. The viscous dissipation does not disappear because of incom-
pressible effects, but because of low Mach number limits. In most cases considered however, the
two effects will be equivalent. Eqns. 1.4-1.6 are the building blocks for all subsequent heat trans-
fer analysis. They are general and valid for steady and unsteady incompressible flow with con-
stant properties. The material derivative, D/Dt, is defined as
. (1.7)
Assuming steady state flow would allow further simplification of the equations. In the case of
steady state flow, the partial time derivative (c/ct) vanishes. These are also subject to boundary
conditions that are unique to a given flow configuration. For example, a two dimensional bound-
ary layer would have the following boundary conditions:
, (1.8)
, (1.9)
Eqn. 1.8 is a consequence of the no-slip condition and Eqn. 1.9 is a condition of having a solid
wall. The other condition that is required at the wall is information on the temperature distribu-
tion or heat transfer. The most typical conditions are a uniform wall temperature or a uniform
wall heat flux. In the freestream, the following conditions typically apply:
, (1.10)
, (1.11)
. (1.12)
These three equations are the properties of the uniform flow in the freestream, that is the area infi-
nitely away from the wall in which there are no friction or wall effects. Each area of convective
D
Dt
------ c
o
u
j
c
j
+ =
u 0 =
v 0 =
u U

=
v 0 =
T T

=
4
heat transfer will have some variation of these equations that will be cultivated to the specific cir-
cumstances.

1.3 Modes of Heat Transfer
Conduction: Conduction is the most basic method of heat transfer. It involves direct heat trans-
fer from one object to another object. The objects can be gases, fluids, or solids.
Convection: The key to understanding convection is remembering that its is conduction between
two objects (or gases, fluids, or a mixture) in which the two substances have a relative velocity
between themselves. For example, with a gas moving over a hot plate, the plate transfers heat to
the gas and the heated molecules move on. Cooler molecules move into the vacated area, and
they are heated as well. Convection typically transfers more heat because the heated particles
move on and are replaced by cooler ones, allowing for a greater temperature difference for a
longer time.
Forced Convection: Forced convection is the transfer of heat by convection in which the
relative velocity is created by some outside source. For instance, a fan blowing air over a
Figure 1.1: Flow Over a Flat Plate
5
plate is forced convection. Any time there is a freestream velocity, there is going to be
forced convection. Chapter 2 deals with forced convection.
Natural Convection: Natural Convection has no freestream velocity, but there is move-
ment of air that is created by buoyancy effects. When air is heated it expands and
becomes less dense. Hot air will rise and cooler air will replace it. The cycle will con-
tinue as long as there is a temperature difference between the top of a room and the floor.
The movement that is created by this buoyant air displacement is the basis of natural con-
vection. Chapter 3 focuses on natural convection.
Mixed Convection: In a situation where there is both forced convection and natural con-
vection, there exists mixed convection. Usually one of the methods is the dominant mode
of heat transfer, and there are ways to determine how to handle these situations. Natural
convection usually serves to affect the up and downward components of velocity in these
cases.
Radiation: Radiation is a type of heat transfer between two objects that are not in contact with
each other. The transfer can occur through gas or fluid or a vacuum. The heat transfer is indepen-
dent of the medium through which it travels. The air can be heated by radiation if there are water
molecules in the air. The water is heated by the sun and then the air appears to be heated. The sun
heats the earth through the process of radiation. Radiation is not covered in this paper in detail,
but it is referred to on a couple of occasions.
1.4 Laminar Flow
The simplest type of convection is that which takes place in laminar flow. Laminar flow is
easy to predict and has very little fluctuation in it. There are several approaches to solving the
flow in a laminar boundary layer. Most situations begin as laminar flows and then later transition
6
to turbulence. Laminar flows are more structured than turbulent flows, but are still not fully
understood. In order to get solutions for laminar flow, assumptions must be made because the
equations cannot be analytically solved as they are. One must be careful to note when a flow tran-
sitions to turbulent, because it becomes more complicated and unpredictable. Although most real
flows are turbulent in nature, it is often helpful to understand the nature of the laminar flows
before attempting to tackle turbulent flows. Laminar flow principles are simpler and help one get
an idea of how the flow behaves before adding the complexities of turbulent effects.
1.5 Turbulent Flow
In contrast with laminar flow, turbulent flow is less structured and predictable. Structures
called eddies dominate the flow. The eddies can be any size, ranging from the entire width of the
boundary layer to microscopic structures. These eddies rotate and mix the flow in such a way that
there it is almost impossible to solve analytically. Turbulent flow problems involve systems of
equations in which there are more unknowns than equations. Assumptions and models reduce the
number of unknowns, but these assumptions are not perfect. Turbulence modelling software
relies on the accuracy of the models themselves. Researchers are always working to improve
these models in order to get a better understanding of turbulent flow. Typically all high Reynolds
number flows are turbulent. Low Reynolds number flow can be turbulent or laminar. It is easy to
make a flow transition to turbulence through methods such as tripping the flow, where roughness
is added to a surface in order to facilitate the transition. Turbulence is a phenomenon whose char-
acteristics are still being investigated by researchers. Until the turbulence problem can be solved,
researchers must rely on idealized laminar solutions and turbulence models.
1.6 External Flows
7
The most common type of external flow is flow over an object. The object is usually a flat
plate as the flat plate can simulate a number of different real applications such as the wing of an
airplane. Another common object is a cylinder. Boundary layers are important concepts in exter-
nal flows and they will be addressed in Section 1.6.1. Section 1.6.2 covers other types of external
flow that are called free turbulent flows. The effects of friction from walls are unimportant in
these situations. Although heat transfer is present in these situations, quantities are difficult to
measure because the transfer is from fluid to fluid (or gas to gas). They are also less important in
application and are only briefly covered.
1.6.1 Boundary Layers
Boundary layers are central to the idea of forced convection. Any flow that is bounded by
a surface will develop a region that is adjacent to the surface, in which the flow properties are dif-
ferent from the freestream. Friction is the primary reason for the existence of the boundary layer.
There are an infinite number of thin regions that can be considered a boundary layer. Each one is
based upon a property that varies from a value at the wall to a freestream value away from the
wall. The two most common types of boundary layers are velocity and thermal. In a velocity
boundary layer, the flow velocity is zero at the wall because of the no-slip condition. The velocity
increases as you move away from the wall until it eventually becomes a constant and is equal (or
close enough) to the freestream value. In a thermal boundary layer, the temperature varies from a
temperature T
o
at the wall to the freestream value T

at the edge of the thermal boundary layer.


The lengthscale of the velocity boundary layer is typically given the symbol o while the thermal
boundary layer has the scale o
T
. In general these scales are not equal. The relative size of these
scales is a condition of a parameter known as the Prandtl number. It will be introduced later in
this chapter, and its specific function will be discussed in Chapter 3.
8
The boundary layer is an important concept because it is the region in which heat transfer
between a gas or a fluid and a surface take place. It is distinct from the rest of the flow, which is
usually considered to have constant freestream properties. The nature of the boundary layer is
determined by the surface fluid interaction. This includes the roughness and the shear stress at the
wall. The velocity distribution within a boundary layer is important because velocity plays an
important part in the mass momentum and energy equations. The temperature is the main compo-
nent of the energy equation and as such the thermal boundary layer is important to fluid mechan-
ics. Understanding the boundary layer is the first step in analyzing external fluid flow and
external convection.
1.6.2 Free Turbulent Flows
In a free turbulent flow, the flow is different from a typical external flow in that it is
located far enough away from all surfaces that there are no effects from the wall. Free turbulent
flows come in several different forms, but they all share the common theme of being unaffected
by walls. The flow is built on the interaction of two different fluids or gases mixing with each
other. The first fluid or gas has some velocity and it is released into a stagnant fluid or gas. Free
turbulent flows are built on the concept of shear layers. A shear layer is the area of interaction
that develops between the two fluids. The area of interaction usually grows at an angle from the
point where the first fluid leaves the pipe or duct in which it had been travelling. Laminar and tur-
bulent shear layers are both possible, with the laminar almost always becoming turbulent after
sufficient distance. A shear layer usually develops on all sides of the duct from which the fluid is
released. When these shear layers combined, the flow becomes what is known as a jet. The most
typical examples of a jet occur when the fluid is released from a slit rather than a large duct. The
9
shear layers combine almost instantaneously and the entire flow is a jet essentially. Often similar-
ity is used to find the solution to jet flow.
The final two examples of shear layers are the wake and the plume. When a fluid flows
over an object such as a cylinder, it creates a wake downstream from that object. Much like the
shear layer, the affected region expands at a certain angle from the object. The governing equa-
tions differ slightly from those of a jet. If the object is heated, there will be temperature effects
that complicate the situation more. A plume is a region of heated air that rises from a heated
source into cooler air. The source may be a cylinder or a heated plate. This is a prime example of
natural convection. The heated air near the source becomes less dense than the air surrounding it.
It rises into the air and in turn creates an angled area of interaction with the stagnant air around it.
These plumes are important in the atmospheric boundary layer and will be discussed further in
Chapter 4.
1.7 Internal Flows
In contrast to external flows, internal flows are completely encompassed by a pipe, a duct
or some other fluid carrier. Probably the most common type of internal flow is flow of water
through round pipes. In a situation involving round pipes, the easiest thing to do is to change the
coordinates to cylindrical. This is because the flow is virtually identical in all radial directions if
the pipe is full and the effects of gravity can be ignored. In internal flows, the boundary layers
grow to the point where they meet in the middle of the pipe, presenting a new type of flow called
fully developed.
10
1.7.1 Entrance Region
When a fluid leaves a reservoir and first enters a pipe, the flow cannot truly be considered
internal even though it is flowing inside of a pipe or duct. In this region, all of the principles of
external boundary layer flow apply. In a circular pipe, the boundary layers have not grown large
enough to interact, so there is no difference between flow that is bounded on all sides and flow
that is bounded on one side. It does help to change the coordinate system, but this is only to facil-
itate the analysis. There is also no rule for the entrance region being laminar or turbulent. If the
fluid only travels at a low speed, the flow may stay laminar. In most cases the fluid will reach a
certain transition Reynolds number and the flow will become turbulent. In either case, external
flow guidelines apply until the flow becomes fully developed.
1.7.2 Fully-Developed Region
The fully developed region of pipe flow refers to the area where the boundary layers com-
bine in order to form one flow in the entire pipe or duct. There is no longer any distinguishable
Figure 1.2: Internal Pipe Flow
11
freestream. Although internal, fully-developed flow retains some of the principles of boundary
layer and external flow, it is governed by additional theories. Most of the equations for fully
developed flow are empirical in nature. This is especially the case when the fully developed flow
is turbulent. The equations will be examined in depth in Chapter 2.
1.8 Principles of Scaling and Non-Dimensionalization
In order to get a better understanding of the relative sizes of each term in the fundamental
equations, one can identify scaling arguments. Not only does it give the researcher a better idea of
what to expect in terms of size, but certain terms can be eliminated in the equation. If L is the
streamwise length, then one can make the assumption that
(1.13)
indicating that the boundary layer is slender. With this assumption one can examine the terms in
Eqns. 1.4 to 1.6 in order to discover their relative sizes. Scaling uses one constant term that is rep-
resentative of the flow variable. It is often the freestream or sometimes the maximum value for
the variable. For instance, the y direction is scaled by o and the distance x is scaled by L, the total
streamwise length. If there were a third dimension, z would be scaled by W the width of the plate
or the duct. The temperature T and the velocity u are scaled by their freestream properties T

(or
some function of it) and U

. Anything that is a constant does not need to be scaled. Scaling


allows simplifications in equation because one term may dominate another term, making the sec-
ond term inconsequential. If we assume steady state flow (c/ct = 0), the mass continuity equation
(1.5) requires that
. (1.14)
o L
U

L
-------
v

o
------ ~
12
in order for the magnitudes of the two terms to be equal. This necessitates the conclusion that
both terms in the mass continuity equation are on the order of U

/L. This is important because it


allows some easier substitutions in the momentum equations. There are three different types of
forces in the momentum equation: inertia, pressure, and friction. The inertia forces are the terms
that have a squared velocity term in the numerator and scale as
. (1.15)
The pressure forces are the terms with a pressure term in the numerator and they scale as
. (1.16)
The final force term is a friction force. This case has a second derivative of velocity and they can
be found to scale as
. (1.17)
If we assume steady state flow, the x-momentum equation contains two inertia terms (once the
material derivative is expanded), a body force term, a pressure term, and two friction terms. Let
us consider a common case in which there is no external body force. The two inertia terms are of
the same order as a consequence of the continuity equation. They both scale as U
2

/L. Neither of
these terms are dominant. Looking at the friction terms on the right side of the equation, one term
scales as U

/L
2
and the other scales as U

/o
2
. If we continue with the assumption of Eqn. 1.13,
the second term is much larger than the first term. As such we can essentially ignore the second
derivative of x. This is an example of how scaling can be used in order to simplify equations. It
will be further used in subsequent chapters, and it is important to understand the process.
U

L
------- v
U

o
------- ,
P
L
------
v
U

L
2
------- v
U

o
2
------- ,
13
Often it is useful to work independent of units. This is helpful because solutions can apply
to any system of measurement one chooses to use in an analysis. The principle behind it is to turn
each and every variable into a new dimensionless variable. In most cases this means dividing by
a constant of the same units. In some cases one must divide by multiple constants in order to get
the units to cancel. All velocities are usually scaled by the same freestream velocity U

. Dis-
tances often have a choice in scaling. Normal distances are usually divided by a channel half
width or radius. Sometimes they are divided by the streamwise length. The streamwise distance
usually uses the streamwise length as its scaling variable, as long as the distance is not considered
to be an infinite pipe. In a case such as that, there is usually no variation in x anyway so those
terms are ignored. Temperature can often be trickier, as often it is scaled as
(1.18)
in order to have the term u vary from 0 to 1. Once all variables have been taken care of, the equa-
tion needs to be cleaned up in order to get rid of all dimensions. This gives rise to a number of
dimensionless constants that are a combination of other constants. Some of the most common
dimensionless constants are
Reynolds Number = (1.19)
Nusselt Number = (1.20)
Peclet Number = (1.21)
Prandtl Number = (1.22)
u
T T
o

T
o

------------------ =
Re
Lu

----------
Lu
v
------ = =
Nu
hL
k
------ =
Pe
U

L
o
----------- =
Pr
v
o
--- =
14
Many of these terms are important not only in nondimensionalization, but also in classifying
flows in all studies of heat transfer and fluid flow. The Reynolds number plays an important part
in wind tunnel tests when it comes to scaling models in order to apply work to real air and seac-
raft. Non-dimensionalization will be used in order to find some solutions later in this paper. The
principles outlined above remain the same no matter the situation.
15
CHAPTER 2
FORCED CONVECTION
2.1 Introduction
The most common type of convection is referred to as forced convection. As defined in
Chapter 1, it is the transfer of heat from one substance to another in which a relative velocity
exists between the two mediums. Most commonly this exchange is between a solid and a gas or a
liquid. Section 2.2 highlights the general equations that are used for forced convection and gives
a brief overview of the force convection in laminar flow. Section 2.3 introduces the reader to the
turbulent flow equations that become important for the study of turbulent heat transfer. Section
2.4 adapts the turbulent flow equations to boundary layer flow. This amounts to making some
assumptions in order to simplify the equations.
2.2 References
The information in this chapter is primarily taken from Bejan [1]. There are many other
good sources that one can investigate that will give them information about turbulence and about
heat transfer. Schlicting [17] and White [19] have written some of the pioneering texts on fluid
flow in a boundary layer. Schlicting was one of the first to develop boundary layer theory, while
White examined the effect of friction on flows. Some of the leading text on turbulence has been
written by Hinze [8], Pope [16], and Bernard and Wallace [2]. These authors focus on the concept
of turbulence primarily, though they do touch on heat transfer. Some of the better convective heat
16
transfer texts include Bejan, Kays and Crawford [11], and Burmeister [4]. A couple of introduc-
tory heat transfer textbooks include Janna [9] and Oosthuizen and Naylor [14]. For specific areas
of heat transfer, the reader is directed to Kutateladze and Leonitev [12] (boundary layer heat trans-
fer) and Petukhov and Polyakov [15] (turbulent mixed convection). The reader is directed to
Adrian et al. [20] for a journal article on a relative experiment involving heat transfer in turbulent
boundary layers.
2.3 Equations
It is often necessary to make certain assumptions about physical flows in order to facilitate
the analysis. The general equations 1.4 through 1.6 are written in cartesian notation, remember-
ing that incompressible (low Mach number) and constant property flow have already been
assumed. Another assumption that is commonly made about a flow of this nature is the Bouss-
inesq approximation. The Boussinesq approximation states that in a flow where the density
changes (not by compressibility effects, but by temperature differences) are small but not zero, the
density change is important only in relation to the body force. This body force is usually gravity,
and is typically important only in the vertical momentum equation. The Boussinesq approxima-
tion will be discussed further in the next chapter when talking about natural convection. All other
density gradients and fluctuations may be neglected. Recalling the equations in cartesian nota-
tion:
(2.1)
. (2.2)
cu
i
cx
i
------- 0 =

g
c
-----
Du
i
Dt
--------- F
i
cP
cx
i
-------
c
2
u
i
cx
j
cx
j
--------------- + =
17
. (2.3)
There is some further simplification that can be done if one wants to consider only steady flow. In
this case all partial derivatives with respect to time can be eliminated. However turbulent flows
are often unsteady and as such often contain time dependent terms. For completeness, we will not
ignore the unsteady terms at this time.
2.4 The Turbulent Equations
By nature turbulent flows are irregular. In the past turbulent flows have been considered
to be completely random and unpredictable. Some recent research suggests that there is an under-
lying structure to turbulent flow. Flow visualization has shown ordered structures composed of
eddies that provide the large scale order to the turbulent flow. It is still impossible to predict
instantaneous results in a turbulent flow, and as such time averaged methods are used to examine
turbulent flows. Transforming the equations from general to turbulent requires the following def-
initions for the flow:
, (2.4)
, (2.5)
, (2.6)
, (2.7)
, (2.8)
where the () indicates a time averaged quantity and () denotes an instantaneous fluctuation. The
general equations are valid for both turbulent and laminar flow. In order to make them specifi-
cally applicable to turbulent flow, Eqns. 2.4 to 2.8 are inserted into the general equations. The
c
DT
Dt
-------- k
c
2
T
cx
i
cx
i
--------------- =
u u u' + =
v v v' + =
w w w' + =
T T T ' + =
P P P' + =
18
equations are still valid for laminar flow by considering the fluctuations to be negligible. Further
manipulation of the equations can be done to compress some of the terms. The transformation is
done by substituting the instantaneous quantities and then integrating each term over time. We
recall the definitions:
(2.9)
(2.10)
Eqn. 2.9 is the rule for time averaging a quantity and Eqn. 2.10 states that all fluctuating quantities
time averaged over a period t are 0 by definition. Some of the other algebraic properties that will
help us arrive at the turbulent equations are:
(2.11)
(2.12)
(2.13)
(2.14)
(2.15)
(2.16)
(2.17)
u
1
t
-- - u t d
0
t
}
=
u' t d
0
t
}
0 =
u v + u v + =
uu' 0 =
uv uv u' v' + =
u
2
u
2
u'
2
+ =
cu
cx
------
cu
cx
------ =
cu
ct
------ 0 =
cu
ct
------ 0 =
19
These relations will be very important for the turbulent transformations that follow. The first
equation to transform is also the simplest. Taking the two dimensional mass equation (2.1) one
can expand the terms in the following manner:
(2.18)
After integrating each term over time and then applying Eqns. 2.10 and 2.15 the equation is
reduced to
(2.19)
which is almost identical to the original equation indicating that the turbulent fluctuations are not
important to the mass equation. Expanding the material derivative in the momentum equation,
one gets the following expression:
(2.20)
Adding Eqn. 2.20 to the mass conservation equation, the equivalent of adding 0:
, (2.21)
and then using the product rule to combine terms:
(2.22)
Substituting the instantaneous equations into the momentum equation and then averaging over
time using the previously stated rules gives:
. (2.23)
cu
cx
------
cu'
cx
-------
cv
cy
-----
cv'
cy
-------
cw
cz
-------
cw'
cz
-------- + + + + + 0 =
cu
cx
------
cv
cy
-----
cw
cz
------- + +
cu
i
cx
i
------- 0 = =

Du
i
Dt
---------
cu
i
ct
------- = u
i
cu
i
cx
j
------- +

cu
i
ct
------- u
i
+
cu
i
cx
j
-------
cu
i
cx
i
------- + F
i
cP
cx
i
-------
c
2
u
i
cx
j
cx
j
--------------- + =

cu
i
ct
-------
c u
i
u
j
( )
cx
j
----------------- + F
i
cP
cx
i
-------
c
2
u
i
cx
j
cx
j
--------------- + =

c u
i
u
j
( )
cx
j
----------------- F
i
cP
cx
i
-------
c
2
u
i
cx
j
cx
j
--------------- + =
20
Although this form is helpful, we can use Eqn. 2.13 in order to show the effects of the instanta-
neous fluctuations. This transformation reads:
(2.24)
The term is called the Reynolds stress and is an important quantity in measuring turbulence.
Applying the mass continuity equation to the momentum equation yields the final form of the
momentum equation:
. (2.25)
The final form shows how the Reynolds stress affects the flow in a turbulent situation. The final
equation is the energy equation. In much the same way that the mass and momentum equations
were converted to turbulent form, the energy equation can be transformed into the following.
(2.26)
Analogous to the Reynolds stress is the turbulent heat flux which appears in Eqn. 2.26 and is writ-
ten . It is as important to turbulent heat transfer as the Reynolds stress is to turbulent flow.
Eqns. 2.19, 2.25, and 2.26 are the turbulent flow equations. They contain 17 unknowns, but only
represent 5 equations (since the momentum equation can be expressed in three different dimen-
sions). This is referred to as the closure problem that makes turbulent flow so difficult to analyze
and predict. Depending on the situations, flow-specific assumptions are often used to reduce the
number of unknowns. Although it is impossible to realistically reduce the number of unknowns
to be equal to the number of equations without introducing some inaccuracy, we can get close by

cu
i
u
j
cx
j
------------
cu'
i
u'
j
cx
j
-------------- +
\ .
|
| |
F
i
cP
cx
i
-------
c
2
u
i
cx
j
cx
j
--------------- + =
u'
i
u'
j
u
i
cu
i
cx
j
-------
cu'
i
u'
j
cx
j
-------------- + F
i
cP
cx
i
-------
c
2
u
i
cx
j
cx
j
--------------- + =
u
i
cT
cx
i
-------
cu'
i
T '
cx
i
------------- + o
c
2
T
cx
j
cx
j
--------------- =
u'
i
T '
21
making reasonable assumptions and estimations concerning the flow properties. Part of the prob-
lem is handled with turbulence models.
2.4.1 Turbulence Models
The use of models is situation dependent. There is no one model that is universally appli-
cable. Typically boundary layer flow (which will be discussed in the next section) relies on the
mixing length model for analytical solutions. Computer turbulence modeling often relies on com-
plicated models, but they are difficult to solve analytically. One of these models is known as the
k-c model. It is the most popular model used in computer turbulence solutions. The definition of
k, the turbulence energy, is:
, (2.27)
which is complemented by a model for the momentum eddy diffusivity that reads:
(2.28)
C

is an experimentally determined constant and L is a yet to be determined length scale similar


to the mixing length described above. The term c is the dissipation rate or the rate of k destruc-
tion. Details of this derivation can be found in Bejan [1] and other sources. Only the fundamental
equations will be presented here. The dissipation rate is defined as
. (2.29)
The length scale can be eliminated by using these two equations and setting C
D
to be 1.
(2.30)
Using this equation and the following two equations for k and c
k
1
2
--- u' ( )
2
v' ( )
2
w' ( )
2
+ + | | =
c
M
C

k
1 2
L =
c C
D
k
3 2
L
---------- =
c
M
C

k
2
c
----- =
22
, (2.31)
(2.32)
one can solve for the three unknowns (k, c, c
M
). The final values in these equations are empirical
constants that have typical values of
, , , , . (2.33)
which can be found in Bejan [1]. Similar constants are valid for most forms of flow. The number
of equations now matches the number of unknowns. The equation that completes the solution of
the energy equation involves the turbulent Prandtl number, which is defined as:
. (2.34)
This is the last equation needed to determine a full solution. The turbulent Prandtl number is typ-
ically assigned a value of 0.9, based on empirical data. One can see why this model is used only
in computer modelling. The mixing length model will be introduced in the next section and will
be used for boundary layer flow when computer modelling is not involved. Bejan [1] has pro-
vided much of the information for the modelling in this section, but the reader is also referred to
Launder [13] for more information on turbulence modelling. The next section will address the
specifics of the boundary layer flow configuration.
2.5 Boundary Layer Flow
The boundary layer was first introduced in Section 1.6.1. Its properties and many of its
assumptions were addressed in that section, but they will be quickly reviewed in this section. The
turbulent flow equations 2.19, 2.25, and 2.26 can be adapted to a boundary layer flow in order to
Dk
Dt
-------
c
cy
-----
c
M
o
k
------
ck
cy
-----
\ .
| |
c
M
cu
cy
------
\ .
| |
2
c + =
Dc
Dt
-------
c
cy
-----
c
M
o
c
------
cc
cy
-----
\ .
| |
C
1
c
M
cu
cy
------
\ .
| |
2
c
k
-- C
2
c
2
k
----- + =
C

0.09 = C
1
1.44 = C
2
1.92 = o
k
1 = o
c
1.3 =
Pr
t
c
M
c
H
------ =
23
reduce the number of unknowns. Consider a freestream flow with freestream properties U

, T

,
and P

. A boundary layer flow is usually symmetric. If the axes are chosen so that the x direction
is in the direction of the mean flow, and the y direction is the normal direction away from the wall
as shown in Fig. 2.1, the symmetric properties allow the elimination of all c/cz terms. Scaling
assumptions allow the neglect of several other quantities. The velocity fluctuations u and v are
both of the same magnitude because they are both products of the same mechanism, namely a
rotating eddy. While the eddy rotates it produces fluctuations of equal magnitude in the x and y
direction, while w fluctuations reduce to zero because of symmetric properties. Although eddys
are three dimensional in nature, they extend primarily in only the x and y directions. The relative
size of the w term is small enough comparatively to ignore.
The equal magnitudes of the u and v fluctuations mean that scaling laws allow partial
derivatives of the Reynolds stresses in the x-direction to be ignored. This goes back to the equa-
tion that assumed that o<< L. Scaling properties examined in Section 1.8 show that derivatives of
this type may be ignored because the relative change is very small compared to the relative
change in the y direction. This also includes the derivatives involving and in the
energy equation. The relative size of the y derivative is much larger than that of the x derivative.
Figure 2.1: Flow Over a Flat Plate
u' T' v' T '
24
Typically boundary layer flow also allows the assumption that pressure is a function of the x-
direction only, that is:
. (2.35)
Boundary layer flow considers only the x momentum equation along with the mass and energy
equations. Taking all of the preceding assumptions and assuming that there is no body force in
the x-direction and applying them to the turbulent flow equations, we get the following three
boundary layer equations:
, (2.36)
, (2.37)
. (2.38)
These are the important equations in turbulent boundary layer flow. They are similar to the equa-
tions that are used in laminar boundary layer flow. The laminar flow equations can be recovered
by assuming that all fluctuations are zero.
Another common way that the momentum and energy equations can be rewritten is:
, (2.39)
(2.40)
which can easily be seen using simple algebra and calculus rules. The importance of these
expressions lies in the final term of each category: and . The first term one should recog-
dP
dx
-------
cP
cx
------ =
cu
cx
------
cv
cy
----- + 0 =
u
cu
cx
------ v
cu
cy
------ +
1

---
dP
dx
------- v
c
2
u
cy
2
--------
cu' v'
cy
----------- + =
u
cT
cx
------ v
cT
cy
------ + o
c
2
T
cy
2
---------
cv' T '
cy
----------- =
u
cu
cx
------ v
cu
cy
------ +
1

---
dP
dx
-------
1

---
c
cy
-----
cu
cy
------ u' v' +
\ .
| |
+ =
u
cT
cx
------ v
cT
cy
------ +
1
c
P
---------
c
cy
----- k
cT
cy
------ c
P
v'T'
\ .
| |
=
u' v' v' T '
25
nize as the Reynolds stress, and the second term is the turbulent heat flux. If these terms vanish,
then the equation will be reduced to the boundary layer equations for laminar flow. For this rea-
son the importance of these two terms to the analysis of turbulent flow is evident. Typically both
of these terms have a negative value at all points in a boundary layer profile. As a step towards
reducing the number of unknowns and to give the reader a physical meaning to these terms, let us
introduce the definitions:
, eddy shear stress (2.41)
. eddy heat flux (2.42)
The terms c
M
and c
H
are known as the momentum eddy diffusivity and the thermal eddy diffusiv-
ity. They are empirical correlations and must be determined experimentally. They are not flow
properties independent of the situation, for each flow will have a different value although some
may have similar results. Substitution of the new terms into the previous turbulent flow equations
yields:
(2.43)
(2.44)
Despite giving physical meaning to our two new terms, this method has not reduced the number
of unknowns that we have been given. The closure problem still remains with 5 unknowns and
only 3 equations in the turbulent boundary layer. However we can now use another model to esti-
mate the diffusivities. Before we move on let us define an apparent shear stress, t
app
and an
apparent heat flux q
app
. Both quantities will be important in later analysis.
u' v' c
M
cu
cy
------ =
c
P
v' T ' c
P
c
H
cT
cy
------ =
cu
ct
------ u +
cu
cx
------ v
cu
cy
------ +
1

---
dP
dx
-------
c
cy
----- v c
M
+ ( )
cu
cy
------
\ .
| |
+ =
cT
ct
------ u +
cT
cx
------ v
cT
cy
------ +
c
cy
----- o c
H
+ ( )
cT
cy
------
\ .
| |
=
26
(2.45)
. (2.46)
These two terms can be found explicitly in the momentum equation (2.43) and the energy equa-
tion (2.44).
2.5.1 Mixing Length Model
This boundary layer specific model is known as the mixing length model. Using scaling
laws and boundary layer assumptions, an approximation for c
M
can be found in terms of the other
unknowns and the known constants. Details can be found in Bejan [1]. The expression for the
mixing length model is:
. (2.47)
k is an empirical constant that must be found by experiment. This is a minor complication in that
solving these equations still relies on some experimentation, but this is the best way to deal with
the closure problem. Other uses of scaling in the problem can advance the theoretical equation
further, but the mass and momentum equations can now be integrated numerically as a two equa-
tion, two unknown system. The reader is referred to books such as Bejan for further details.
Further analysis of the boundary layer makes use of the assumption that the left hand side
of the momentum can be neglected in the region near the wall due to small velocities and .
Let us assume that in an area near the wall, the velocities are small enough as to make the left side
of Eqn. 2.43 negligible, a reasonable assumption based on the no slip condition and wall impene-
t
app
v c
M
+ ( )
cu
cy
------ =
q''
app
c
p
o c
H
+ ( )
cT
cy
------ =
c
M
k
2
y
2
cu
cy
------
=
u v
27
trability. The apparent shear stress,t
app
, is then constant in this small region near the wall. Let us
redefine this constant as:
, (2.48)
with t
o
being the actual shear stress at the wall. The friction velocity u
t
is defined as:
. (2.49)
which is dependent only on x-position in the flow, and can often be taken as a constant every-
where in the flow. It becomes necessary at this point to introduce wall coordinate notation. Wall
coordinates are non-dimensional variables that are often used in boundary layers in order to uni-
versalize equations. They are as follows:
(2.50)
(2.51)
(2.52)
, (2.53)
(2.54)
Inserting the wall coordinates into Eqn. 2.48, the following expression is derived:
, (2.55)
v c
M
+ ( )
cu
cy
------
t
o

----- =
u
t
t
o

-----
\ .
| |
1
2
---
=
u
+ u
u
t
----- =
v
+ v
u
t
----- =
x
+
xu
t
v
-------- =
y
+
yu
t
v
-------- =
1
c
M
v
------ +
\ .
| |
du
+
dy
+
--------- 1 =
28
One can envision two different situations that factor into the result of this equation. The first is
the region immediately bordering the wall in which v>>c
M
. This region is known as the viscous
sublayer where Eqn. 2.55 reduces to
(2.56)
Upon integration with the initial condition u
+
(0)=0, the velocity distribution for the viscous sub-
layer reads:
. (2.57)
The other situation that arises is one in which the momentum diffusivity dominates the viscosity,
that is c
M
>>v. In this case Eqn. 2.55 reduces to
. (2.58)
The mixing length model transforms this equation to
(2.59)
Integration of this equation from the outer point of the viscous sublayer region where y
+
= y
+
VSL
,
and from Eqn.2.57 that says u
+
= y
+
VSL
, the velocity distribution in this region is
, (2.60)
where experimentally k, the von Karmann constant has been determined to be approximately 0.41
and B has been found to be approximately 5.5. This is known the law of the wall and extends
from the law of the wall out through the inner region of the boundary layer. The place where the
flow begins to deviate from the law of the wall is flow dependent. Other researchers have found
du
+
dy
+
--------- 1 =
u
+
y
+
=
c
M
v
------
du
+
dy
+
--------- 1 =
k
2
y
+
( )
2
du
+
dy
+
---------
\ .
|
| |
2
1 =
u
+ 1
k
--- y
+
B + ln =
29
different relations for the velocity distributions in the viscous sublayer and the inner region, but
this is the most common representation.
2.5.2 Solving the Energy Equation
Since the mass and momentum equations have been solved, we are left with one equation
and two variables: T and c
H
. The solution of this equation is an extension of the mixing length
model used for the momentum diffusivity. In the same way that we neglected the left hand side
of the momentum equation, let us neglect the left hand side of the energy equation as a result of
the small velocities near the wall. That is equivalent to saying that the apparent heat flux is not
dependent on the distance from the wall as long as the region is sufficiently close to the surface.
To put it another way,
(2.61)
Insertion of the wall coordinates into Eqn. 2.61 leads to the following expression:
. (2.62)
From this expression we can get the proper non-dimensionalization of the Temperature T in +
units:
. (2.63)
Placing T
+
into Eqn. 2.62 then gives the integral expression:
. (2.64)
o c
H
+ ( )
cT
cy
------
q''
o

c
P
---------- =
c
P
u
t
q''
o

--------------
c
cy
+
-------- T T
o
( )
1
o v c
H
v +
------------------------------ =
T
+
x
+
y
+
( , ) T
o
T ( )
c
P
u
t
q''
o
-------------- =
T
+ y
+
d
1 Pr 1 Pr
t
( ) c
M
v ( ) +
----------------------------------------------------------
0
y+
}
=
30
Unfortunately we still must deal with the c
M
and Pr
t
terms in order to make this equation solvable
analytically. The assumptions that we used to find two different situations for Eqn. 2.55 were that
in a region close to the wall viscous effects would dominate, and as one moved away from the
wall the momentum diffusivity would grow proportionately. Using these same assumptions, and
dividing the thermal boundary layer into two regions, one can see that the two terms in the
denominator of Eqn. 2.64 will dominate in different regions. When viscous effects are large, the
first term will dominate, making the second term negligible, and when momentum effects domi-
nate further away from the wall, the first term will be negligible. Instead of the viscous sublayer,
let us refer to this sublayer as a conductive sublayer, with its outermost point denoted by y
+
CSL
.
Introducing these terms and assumptions we then can transform the integral to
(2.65)
Integrating this equations gives the following piece-wise solution
and (2.66)
This solution relies on knowing three experimentally determined constants. We have already
given a typical value for k = 0.41. The outer point of the conduction sublayer is typically valued
around y
+
= 13.2. The final value is the turbulent Prandtl number which had been previously
defined as the ratio of the molecular diffusivity to the heat diffusivity. Studies have indicated that
the turbulent Prandtl number starts high near the wall and levels out around 0.85 as it progresses
away from the wall. One theory, the details of which will not be presented here is called the Rey-
nolds analogy. It states that the rate of thermal diffusivity is equal to that of the heat diffusivity.
T
+ y
+
d
1 Pr
-------------
y
+
d
1
Pr
t
-------
c
M
v
------
---------------
y
CSL
+
y
+
}
+
0
y
CSL
+
}
=
T
+
Pr y
+
= y
+
y
CSL
+
< T
+
Pr y
CSL
+
Pr
t
k
-------
y
+
y
CSL
+
----------- ln + = y
+
y
CSL
+
>
31
Although this is a somewhat reasonable assumption, it is not entirely accurate throughout the
flow. For this study we will use a value of Pr
t
= 0.9. The solution of the equation in the inner
region of the boundary layer (y
+
> y
+
CSL
) using these values is then
(2.67)
These solutions are valid as long as our previous assumptions of q
app
and t
app
remain constant
or reasonably so.
2.5.3 Wall Heat Flux
As long as this solution holds throughout the boundary layer, an expression for the wall
heat flux can be developed. Using the boundary condition that is equal to T

at y = o and
expanding some of the wall coordinates in Eqn. 2.66 gives the following expression:
. (2.68)
The thickness of the boundary layer, o, is still unknown, but let us use the law of the wall to obtain
the expression for the velocity at the edge of the boundary layer:
(2.69)
Also let us define the following quantity:
(2.70)
where C
f,x
is the coefficient of friction and is a function of x in the same way that the friction
velocity u
t
is. Combing the three previous equations we obtain the following expression:
T
+
2.195 y
+
13.2Pr 5.66 + ln =
T
c
P
u
t
T
o
T

q
o
------------------ Pr y
CSL
+
Pr
t
k
-------
ou
t
v
y
CSL
+
---------------
\ .
|
| |
ln + =
U

u
t
-------
1
k
---
ou
t
v
--------
\ .
| |
B + ln =
C
f x ,
2
u
t
U

-------
\ .
| |
2
=
32
(2.71)
The left side of the equation is an important quantity known as the local Stanton number. In rela-
tion to other dimensionless constants it is defined as:
. (2.72)
Eqn. 2.71 has a denominator of O(1) and is Prandtl number dependent, when experimentally
determined constants are substituted in the equation. Colburn simplified the situation further by
suggesting the following formula
. (2.73)
Experimental data has been found to confirm this relation for Prandtl numbers between 0.6 and
60. This formula can be rearranged using the previous relations to find that
(2.74)
. (2.75)
Other models for the velocity distribution can produce estimates and models for C
f,x
in these
equations. One example is the 1/7 power law developed by Prandtl. Further details can be found
in sources such as Bejan.
2.6 Other External Flows
Not all flows are best modeled by a flow over a flat plate. Other simple geometries have
their own heat transfer correlations that are better suited for that individual situation. These rela-
h
c
P
U

-----------------
1
2
-- -C
f x ,
Pr
t
1
2
---C
f x ,
\ .
| |
1 2
Pr y
CSL
+
BPr
t

Pr
t
k
------- y
CSL
+
ln
\ .
| |
+
----------------------------------------------------------------------------------------------------------------- =
St
x
h
c
P
U

-----------------
Nu
x
Pe
x
---------
Nu
x
Re
x
Pr
--------------- = = =
St
x
Pr
2 3 1
2
---C
f x ,
=
Nu
x
1
2
---C
f x ,
Re
x
Pr
1 3
= Pr 0.5 > ( )
Nu
L
1
2
---C
f 0-x ,
Re
L
Pr
1 3
= Pr 0.5 > ( )
33
tions will not be presented in detail here, but the important relations will be given. They have
been developed in a method similar to the flow over a flat plate, though portions of each equation
rely on empirical data.
2.6.1 Cylinder in Cross Flow
. (2.76)
This formula is valid as long as Pe
D
is greater than 0.2 and was developed by Churchill and Bern-
stein and listed in Bejan. If Pe
D
is less than 0.2 and the substance is air then the following rela-
tion, credited to Nakai and Okazaki and listed in Bejan, is more accurate
(2.77)
2.6.2 Sphere
Flow over a sphere has the following Nusselt number correlation:
. (2.78)
This relationship has been confirmed to be accurate for fluids with 0.71 < Pr < 380 and flows with
3.5 < Re
D
< 7.6 x 10
4
. The equation was developed by Whitaker and listed in Bejan.
2.6.3 General Spheroid Body Shapes
Yovanovich, referenced in Bejan, is credited with developing a formula that is applicable
to a general spheroid geometry. The semiaxis that is aligned with the freestream is C and the
other semiaxis is B. Representing the area of the objects surface as A, the length scale is written
as
Nu
D
0.3
0.62Re
D
1 2
Pr
1 3
1 0.4 Pr ( )
2 3
+ | |
-------------------------------------------- 1
Re
D
282000
------------------
\ .
| |
5 8
+
4 5
+ =
Nu
D
1
0.8237 0.5 Pe
D
ln
---------------------------------------------- =
Nu
D
2 0.4Re
D
1 2
0.06Re
D
2 3
+ ( )Pr
0.4

w
------
\ .
| |
1 4
+ =
34
. (2.79)
The Reynolds number and Nusselt Number are then represented as
and (2.80)
The Nusselt number equation, accurate for the ranges 0 < Re
A
<2 x 10
5
, Pr > 0.7, and C/B < 5, is
then represented by
. (2.81)
where p is the maximum perimeter of the object and is found from Table 2.1.
2.6.4 Array of Cylinders
A group of cylinders is usually positioned in an array of staggered cylinders or in an array
of aligned cylinders. Each arrangements has its own set of Reynolds number dependent Nusselt
Object
Sphere 3.545
Bi-Sphere 3.475
Cube 3.388
Cylinder 3.444
Prolate Spheroid
(C/B = 1.93
3.566
Oblate Spheroid
(C/B = 0.5)
3.529
Oblate Spheroid
(C/B = 0.1)
3.342
Table 2.1: Constants for General Shape Heat Correlation
A A
1 2
=
Re
A
U

A
v
------------ = Nu
A
hA
k
------- =
Nu
A
Nu
0
A
0.15
p
A
----
\ .
| |
1 2
Re
A
1 2
0.35Re
A
0.566
+ Pr
1 3
+ =
Nu
0
A
Nu
0
A
35
number correlations. Let us define the variables n and m as the number of rows and the number of
columns respectively. For an aligned array, the correlations are:
(2.82)
The Reynolds number in this case is defined using the maximum average velocity through the
cylinders. All terms in the equations are evaluated at the air temperature except for Pr
w
, which is
taken at the surface temperature of the cylinders. The term C
n
is evaluated based on Fig. xxxxxx,
which was created by Zukausakas and found in Bejan. The same can be said for Eqn. 2.82 and the
following staggered array correlations. For a staggered array of cylinders, the heat transfer rela-
tions are:
(2.83)
where X
t
is the transverse distance between cylinders and X
l
is the longitudinal distance between
cylinders.
Nu
D
0.9C
n
Re
D
0.4
Pr
0.36 Pr
Pr
w
---------
\ .
| |
1 4
= 1 Re
D
10
2
< <
Nu
D
0.52C
n
Re
D
0.5
Pr
0.36 Pr
Pr
w
---------
\ .
| |
1 4
= 10
2
Re
D
10
3
< <
Nu
D
0.27C
n
Re
D
0.63
Pr
0.36 Pr
Pr
w
---------
\ .
| |
1 4
= 10
3
Re
D
2 10
5
< <
Nu
D
0.033C
n
Re
D
0.8
Pr
0.4 Pr
Pr
w
---------
\ .
| |
1 4
= 2 10
5
Re
D
2 10
6
< <
Nu
D
1.04C
n
Re
D
0.4
Pr
0.36 Pr
Pr
w
---------
\ .
| |
1 4
= 1 Re
D
500 < <
Nu
D
0.71C
n
Re
D
0.5
Pr
0.36 Pr
Pr
w
---------
\ .
| |
1 4
= 500 Re
D
10
3
< <
Nu
D
0.35C
n
Re
D
0.6
Pr
0.36 Pr
Pr
w
---------
\ .
| |
1 4 X
t
X
l
-----
\ .
| |
1 5
= 10
3
Re
D
2 10
5
< <
Nu
D
0.031C
n
Re
D
0.8
Pr
0.36 Pr
Pr
w
---------
\ .
| |
1 4 X
t
X
l
-----
\ .
| |
1 5
= 2 10
5
Re
D
2 10
6
< <
36

2.7 Internal flow
Convection within a duct or pipe is very similar to boundary layer flow, mostly because it
begins as a boundary layer flow. The most practical application for turbulent duct flow is flow
within a circular pipe (see Fig. 1.2). Although ducts sometimes shapes other than circles for their
cross sectional areas, most real world fluid flows occur in circular pipes. This means changing
coordinates to cylindrical and making the same boundary layer assumptions that we made in Sec-
tion 2.5. The mass, momentum, and energy equations read as follows:
(2.84)
(2.85)
. (2.86)
Until the boundary layers merge and the flow becomes fully developed, internal flow is no differ-
ent than internal boundary layer flow. The Eqns. 2.84-2.86 are no different then Eqns.2.36-2.38
Figure 2.2: C
n
for an Array of Cylinders
cu
cx
------
1
r
---
c
cr
----- rv ( ) + 0 =
u
cu
cx
------ v
cu
cr
------ +
1

---
dP
dx
-------
1
r
---
c
cr
----- r v c
M
+ ( )
cu
cr
------ + =
u
cT
cx
------ v
cT
cr
------ +
1
r
---
c
cr
----- o v c
H
+ ( )
cT
cr
------ =
37
except that they are in different coordinate systems. Fully developed flow occurs at approxi-
mately
. (2.87)
The entrance length for turbulent flow is shorter than the entrance length in laminar flow. In fully
developed flow, all of the inertia terms are neglected. This means that all terms on the left hand
side of the momentum equation (2.85) vanish. There is also no room in the center for a wake
region. Let us now consider something called the control volume force balance that reads:
. (2.88)
Considering this equation and integrating Eqn. 2.85, the following relation is found:
(2.89)
where the y coordinate is defined as the distance away from the wall, y = r
o
- r. The apparent shear
stress is redefined slightly as
. (2.90)
The law of the wall is the same as it is in external flow, but it breaks down near the centerline in
fully developed flow. Unfortunately these equations produce a velocity profile that has a finite
slope at the channel centerline, when it should be zero for a continuity. An empirical relation that
does produce a zero slope was developed by Reichardt
. (2.91)
2.7.1 Heat Transfer
X
D
---- 10 ~
dP
dx
------- 2
t
o
r
o
----- =
t
app
t
o
---------- 1
y
r
o
---- =
t
app
v c
M
+ ( )
cu
cy
------ =
u
+
2.5
3 1 r r
o
+ ( )
2 1 2 r r
o
( ) + | |
------------------------------------y
+
5.5 +
\ .
| |
ln =
38
The momentum equation is rewritten for fully developed flow as
(2.92)
Integrate this equation two different ways to get:
(2.93)
(2.94)
Dividing these two equations and defining the term M to be
(2.95)
gives the following relation for the apparent heat flux
. (2.96)
In most practical cases, M is equal to 1, so let us assume that quantity for now. The following
analysis was developed by Prandtl and recounted in Bejan. Let us divide Eqn. 2.89 by Eqn. 2.96,
assume that M = 1, and substitute the values definitions for the apparent shear stress and the
apparent heat flux to get:
(2.97)
When we examined boundary layer flow, we noted that there was a region in which the viscosity
and conduction dominate in the area near the wall. Outside of that area, the momentum and heat
eddy diffusivities are dominant. In the same way let us imagine that region in the circular pipe as
c
P
u
cT
cx
------
1
r
---
c
cr
----- rq
app
( ) =
c
P
u
cT
cx
------r r d
0
r
}
rq
app
=
c
P
u
cT
cx
------r r d
0
r
0
}
rq
0
=
M
2
r
2
---- u
cT
cx
------r r d
0
r
}
2
r
0
2
---- u
cT
cx
------r r d
0
r
0
}
------------------------------ =
q
app
q
0
------------- M 1
y
r
0
----
\ .
| |
=
v c
M
+
t
0
----------------du
c
P
o c
H
+ ( )
q
0
---------------------------dT =
39
being bound by the point y
1
. The annulus inside of this point is analogous to the viscous sublayer.
The heat equation can then be integrated twice, once from 0 to y
1
and another time from y
1
to y
2
.
The point y
2
is the point in which (y
2
) is equal to
m
, where
(2.98)
and the subscripts represent the fluid temperature entering and exiting the pipe respectively. The
first integration in the near wall region which neglects c
M
and c
H
is
. (2.99)
The second integration yields
, (2.100)
where U is the mean velocity (defined in the same way that the mean temperature was) and is
approximately equivalent to the velocity at y
2
. Combining the previous two equations and recog-
nizing the definition of friction factor (f = ), we can get the Stanton number by elimi-
nating :
. (2.101)
From this point, most of the correlations are empirical. They can be found in Bejan and this text
will list the author who has developed them. For a smooth pipe surface Colburn found
(2.102)
T T
T
m
1
2
--- T
in
T
out
+ ( ) =
v
t
0
-----u
1
c
P
o
q
0

----------- T
1
T
0
( ) =
c
M
t
0
------ U u
1
( )
c
P
c
H
q
0

------------ T
M
T
1
( ) =
t
0
1
2
---U
2
\ .
| |

T
1
St
f 2
Pr
t
u
1
U ( ) Pr Pr
t
( ) +
---------------------------------------------------------- =
Nu
D
0.023Re
D
4 5
Pr
1 3
= 2 10
4
Re
D
10
6
s s
40
Other authors have improved this formula by being more specialized. Dittus and Boelter pro-
posed
(2.103)
where n is 0.3 if the fluid is cooled by the wall and 0.4 if the wall is heating the fluid. If the fluid
properties are heavily influenced by the temperature, Sieder and Tate proposed using:
(2.104)
Gnielinskis developed a correlation that is accurate within 10%:
(2.105)
Eqns. 2.102 - 2.105 are valid for constant wall heat flux and constant wall temperature. Eqn.
2.105 can be simplified to the following two equations:
(2.106)
(2.107)
All of the above equations are valid only for liquids and gases. One must find different correla-
tions for liquid metals. They will not be covered here.
The total heat transfer rate in internal flow is defined as:
(2.108)
Nu
D
0.023Re
D
4 5
Pr
n
=
0.7 Pr 120 s s
2500 Re
D
1.24 10
5
s s ( )
L D 60 >
Nu
D
0.027Re
D
4 5
Pr
1 3

0
-----
\ .
| |
0.14
=
0.7 Pr 16700 < <
Re
D
10
4
>
Nu
D
f 2 ( ) Re
D
10
3
( )Pr
1 12.7 f 2 ( )
1 2
Pr
2 3
1 ( ) +
--------------------------------------------------------------------- =
0.5 Pr 10
6
< <
2300 Re
D
5 10
6
< <
Nu
D
0.0214 Re
D
0.8
100 ( )Pr
0.4
=
0.5 Pr 1.5 < <
10
4
Re
D
5 10
6
< <
Nu
D
0.012 Re
D
0.87
280 ( )Pr
0.4
=
1.5 Pr 500 < <
3 10
3
Re
D
10
6
< <
q hA
w
AT
lm
=
41
where A
w
is the total wall area, perimeter times length (A
w
= pL). The term must be found
to make this analysis complete. Let us define the terms:
. (2.109)
Details of this analysis will be skipped, but the result for an isothermal wall is:
. (2.110)
For a wall that provides a uniform heat flux, the solution is:
(2.111)
AT
lm
AT
in
T
0
T
in
= and AT
out
T
0
T
in
=
AT
lm
AT
in
AT
out

AT
in
AT
out
-------------
\ .
| |
ln
------------------------------- =
AT
lm
AT
in
AT
out
= =
42
CHAPTER 3
NATURAL AND MIXED CONVECTION
Natural Convection is governed by the same general principles as forced convection does.
However it must be treated slightly differently than forced convection because there is no external
flow to carry heat away from a surface. Instead, natural convection induces flow in a fluid by
heating a portion of that fluid. The heated fluid expands, becoming less dense than the cooler air
around it. The heated air rises due to buoyancy effects, and the cooler air descends to replace it.
Now if the surface that is heating the fluid is horizontal for example, the cold air will then heat up
and the hot air cool, and the flow will continue. Another area of question is whether this flow is
laminar or turbulent. Typically low speed flow is considered laminar. In this case however, there
is little predictable about the flow. Turbulent flow mist be considered as well, especially when it
combines with forced convection to form mixed convection. This chapter will primarily focus on
laminar natural convection since it is the most common due to the low induced speed. This chap-
ter serves mostly to give the reader an idea of natural convection so that the subsequent chapters
can be properly understood. As such it is not of primary importance to learn about turbulent nat-
ural convection.
3.1 Equations
The objective of natural convection heat transfer problems is to solve the equation:
43
, (3.1)
that is to solve for the total heat transfer when the wall and fluid temperatures are known. HW is
the wall area. The conservation equations that are used for heat transfer begin with the equations
that are found in Chapter 1 (1.4-1.6). The equations are expanded and slightly modified as fol-
lows:
(3.2)
(3.3)
(3.4)
(3.5)
The only major difference is that the equations have been simplified to two dimensions only, and
the body force has been modified. The x-momentum equation neglects the body force while the
y-momentum equation retains the term and defines it as g, the gravity force. As was mentioned
above in the description of natural convection, the gravity is the driving force behind the motion.
Much like forced convection, boundary layer assumptions can be made for natural convection to
simplify the problem. In previous boundary layer assumptions, the motion was horizontal. The
motion in natural convection is vertical. Instead of neglecting the y-momentum equation, the x-
momentum equation becomes negligible because of the minimal movement in the x-direction.
The boundary layer assumptions we used for forced convection are then applied to natural con-
vection, except now the flow is in the vertical direction. Consider that the scale of the boundary
layer, o

or o
T
are much smaller than the vertical scale H. This allows us to neglect the
Q HW ( )h
0 h
T
0
T

( ) =
cu
cx
------
cv
cy
----- + 0 =
u
cu
cx
------ v
cu
cy
------ +
\ .
| |
cP
cx
------ V
2
u + =
u
cv
cx
----- v
cv
cy
----- +
\ .
| |
cP
cy
------ V
2
u g + =
u
cT
cx
------ v
cT
cy
------ + oV
2
T =
c
2
cy
2

44
term in the momentum and energy equation due to scaling laws that were developed previously.
The velocity boundary layer in this case does not vary from 0 at the wall to some freestream
velocity. It starts at 0 at the wall and then rises to reach a maximum before falling to zero again at
the outer edge of the boundary layer. The final general equations for natural convection are then
reduced to:
(3.6)
(3.7)
(3.8)
Note that the variables still remain general and have not been converted to their turbulent forms.
3.2 Boussinesq Approximation
One common approach to natural convection is to simplify the density term. This simpli-
fication is called the Boussinesq approximation. The y-momentum equation (3.4) can be simpli-
fied by removing the variable , the density. In a boundary layer, the pressure is effectively a a
function of the primary direction of motion only. We can write
. (3.9)
Also a fluid of density

has a hydrostatic pressure gradient


. (3.10)
This eliminates all pressure terms from the equation and allows us to work with only density.
This may not seem like it has simplified the problem much, but the Boussinesq approximation
cu
cx
------
cv
cy
----- + 0 =
u
cv
cx
----- v
cv
cy
----- +
\ .
| |
cP
cy
------
c
2
v
cx
2
-------- g + =
u
cT
cx
------ v
cT
cy
------ + o
c
2
T
cx
2
--------- =
cP
cy
------
dP
dy
-------
dP

dy
---------- = =
dP

dy
----------

g =
45
allows a simplification of the equations further. It allows us to rewrite the momentum equation
as:
. (3.11)
Remembering the ideal gas law,
, (3.12)
we can rewrite the density terms in Eqn. 3.11 as:
, (3.13)
which can be restated as
. (3.14)
In the limit that (T-T

) << T

, the Taylor series expansion reveals


. (3.15)
This can be rewritten in terms of the volume expansion coefficient at constant pressure, |, as:
. (3.16)
Rearranging this equation allows us to rewrite the momentum equation as:
. (3.17)
The pressure term has been completely eliminated from the conservation equations. We now
have three differential equations, with three unknowns, u, v, and T. These are the equations that
are important for natural convection analysis.
u
cv
cx
----- v
cv
cy
----- +
\ .
| |

c
2
v
cx
2
--------

( )g =
P RT =

1
T
T

------
\ .
| |
=

----------------
\ .
| |
1

----------------
\ .
| |
1
T T

---------------- =

1
1
T

------ T T

( ) . + ~

1 | T T

( ) . + | | ~
u
cv
cx
----- v
cv
cy
----- +
\ .
| |

c
2
v
cx
2
-------- g| T T

( ) =
46
3.3 Natural Convection with a Vertical Wall
The flow induced by natural convection is considered to be laminar as long as the wall
height is not great enough to make the Rayleigh number surpass a critical value. The Rayleigh
number is defined as:
. (3.18)
The critical Ra
y
is dependent on the Prandtl number of the fluid. Defining the Grashof number as
being Ra
y
/Pr, the Grashof number (Gr
y
) must not exceed 10
9
as long as the Prandtl number is
between 10
-3
and 10
3
. The Grashof number is defined as:
. (3.19)
Beyond this point the flow will become turbulent.
Scaling analysis (details can be found in Bejan [1]) reveals that natural convection is influ-
enced by three terms: inertia, friction and buoyancy. In high Prandtl number flow, the boundary
layer is governed by a friction-buoyancy balance. In low Prandtl number fluids, the boundary
layer will feature buoyancy balanced by inertia.
High Pr fluids have a different set of relations that characterize their flow. For a fluid with
a high Pr,
(3.20)
(3.21)
(3.22)
Ra
y
g|ATy
3
ov
------------------- =
Gr
y
g|ATy
3
v
2
-------------------
Ra
H
Pr
---------- = =
o
T
HRa
H
1 4
~
v
o
H
----Ra
H
1 2
~
o HRa
H
1 4
Pr
1 2
~
47
(3.23)
The velocity boundary layer o is larger than the temperature boundary layer because the moving
fluid will drag some fluid with it outside of the region with a nonzero temperature gradient. So
although the temperature gradient is the driving force behind natural convection, friction effects
will cause fluid flow beyond the area where there is a temperature difference. The maximum
velocity is indicated by the term v, which is also the velocity scale. The boundary layer in high Pr
fluids is dominated by friction and buoyancy in the area up until the velocity peak, also indicated
by the edge of the thermal boundary layer. Outside of this region, until the end of the boundary
layer, the flow is driven by a friction and inertia balance.
For low Pr fluids, the resulting balances of the conservation equations lead to the follow-
ing observations
(3.24)
(3.25)
(3.26)
(3.27)
The term o
v
is the viscous sublayer where friction is important, and it is the area that exists in low
Prandtl number fluids where the velocity has not yet reached its maximum. It is smaller than both
the thermal and velocity boundary layers. Low Pr numbers also have a boundary layer divided
into two subsections. Near the wall, in the boundary layer o
v
the flow is governed by a friction
buoyancy balance. Outside of this layer, until the edge of the thermal boundary layer, the flow is
Nu
hH
k
------- Ra
H
1 4
~ =
o
T
H Ra
H
Pr ( )
1 4
~
v
o
H
---- Ra
H
Pr ( )
1 2
~
o
v
HGr
H
1 4
~
Nu
hH
k
------- Ra
H
Pr ( )
1 4
~ =
48
dominated by an inertia-buoyancy force balance. Since the quantity Ra
H
Pr seems to be prevalent
in this analysis, let us define the Boussinesq number:
. (3.28)
The dimensionless quantities are very useful in giving us an idea of how thick the boundary layer
is. The following three relations show an interesting effect of these numbers:
(3.29)
These numbers show just how slender the boundary layer is in relation to the total wall height.
Churchill and Chu developed an empirical relation for the Nusselt number if the wall is
isothermal. It holds for all Prandtl numbers and 10
-1
< Ra
y
< 10
12
.
. (3.30)
All fluid properties are evaluated at the film temperature, defined as (T
w
+ T

)/2.
The case of a uniform heat flux led Vliet and Liu to produce the following relations for
turbulent flow:
. (3.31)
This equation is valid for 10
13
< Ra
*y
< 10
16
. Ra
*y
is a Rayleigh number based on the heat flux:
Bo
H
Ra
H
Pr
g|ATH
3
o
2
--------------------- = =
wall height
thermal boundary layer thickness
------------------------------------------------------------------------------- Ra
1 4
~ Pr 1 >
wall height
thermal boundary layer thickness
------------------------------------------------------------------------------- Bo
H
1 4
~ Pr 1 <
wall height
wall shear layer thickness
------------------------------------------------------------- Gr
H
1 4
~ Pr 1 <
Nu
y
0.825
0.387Ra
y
1 6
1 0.492 Pr ( )
9 16
+ | |
8 27
-------------------------------------------------------------- +
)
`

2
=
Nu
y
0.568Ra
*y
0.22
=
Nu
y
0.645Ra
*y
0.22
=
49
. (3.32)
Chuechill and Chu also have a relation that is very similar to their equation for an isothermal wall
and is valid for all Prandtl numbers and Rayleigh numbers:
. (3.33)
All of the above correlations can be found in Bejan.
3.4 Thermal Stratification
It is always simple to consider a vertical wall that is infinitely high. Like in many other
cases, assuming an infinite lengthscale simplifies things. This section will briefly discuss the
effects of having a wall of finite height. As the fluid is heated, it continues to move up the wall.
Eventually it will reach the top of the wall and run into a ceiling. The fluid cannot rise any higher
and so it will discharge horizontally in the room. It will not be able to drop because it is warmer
than the other fluid that it is above. If the room is small enough, the eventual effect will be ther-
mal stratification. This is the situation where there are different layers of air temperature on top of
one another. They become gradually cooler as one progresses from the ceiling to the floor.
3.5 Natural Convection Over a Flat Horizontal Plate
Consider a situation with stagnant air situated over or under a flat horizontal plate. Two
separate flow situations arise. If the air is above a warmer wall or if it is below a cooler wall, the
flow will release from the wall and go up in the first case and down in the second case. It will
release in a plume that is centered in the middle of the wall. Small balls of heated fluid will
release from the wall intermittently. These balls are called thermals. In the other case, where the
Ra
*y
g|y
4
q
ovk
------------------ =
Nu
y
0.825
0.387Ra
y
1 6
1 0.437 Pr ( )
9 16
+ | |
8 27
-------------------------------------------------------------- +
)
`

2
=
50
air is above a cooler wall or below a hotter wall, gravity will push down on the wall and a bound-
ary layer is created over the entire wall. The flow splits and spills over the side in all directions.
All of these motions are the results of the buoyancy effects that have been discussed. Different
densities create motion as the less dense fluids (the warmer fluid) try to move upwards.
The heat transfer correlations for this situation are based on which type of flow we have.
A hot surface facing upward or a cool surface facing downward has the following correlation:
, (3.34)
. (3.35)
The opposite situation, a cold surface facing upward or a hot surface facing down uses the corre-
lation:
. (3.36)
All of these equations are based on a length scale defined as:
, (3.37)
where A is the area of the plate, and p is the perimeter of the plate.
3.6 Vertical Channels
If two vertical plates are close enough that the flow created by natural convection causes
their boundary layers to interact, the flow will become fully developed. This can be found in a
circular pipe or in a rectangular duct, or just two walls that are close together. The principle is the
same, though the most suitable coordinate system may vary depending on the circumstances. Just
as in forced convection, fully developed flow differs slightly from external flow. If the air is
heated from the bottom of the channel, density difference will cause the air to rise and create a
Nu
L
0.54Ra
L
1 4
= 10
4
Ra
L
10
7
< <
Nu
L
0.15Ra
L
1 3
= 10
7
Ra
L
10
9
< <
Nu
L
0.27Ra
L
1 4
= 10
5
Ra
L
10
10
< <
L
A
p
--- =
51
boundary layer on both walls. Eventually those boundary layers will merge to form a fully devel-
oped vertical flow. The governing equations for this situation are
(3.38)
(3.39)
which are the mass and vertical momentum equations for boundary layers. Fully developed flow,
assumptions (u=0 and =0) combined with the previously discussed Boussinesq approxima-
tion and the mass equation reduces the momentum equation to:
. (3.40)
Next we must solve either the velocity profile or the temperature profile. If one is solved, the
other will be simple. If we assume an isothermal wall, then we can make the following assump-
tion about the fully developed flow:
(3.41)
The right side of Eqn. 3.40 is then a constant, and the equation can be integrated to get:
(3.42)
The mass flow rate is:
. (3.43)
The total heat transfer rate between the fluid and the channel walls is:
(3.44)
cu
cx
------
cv
cy
----- + 0 =
u
cv
cx
----- v
cv
cy
----- +
\ .
| |
cP
cy
------
c
2
v
cx
2
-------- g + =
cv cy
d
2
v
dx
2
--------
g|
v
------ T T

( ) =
T
0
T ( ) T
0
T

( ) < <
v
g|D
2
T
0
T

( )
8v
------------------------------------- 1
x
D 2
-----------
\ .
| |
2
=
m

g|D
3
T
0
T

( )
12v
----------------------------------------- =
Q m

c
P
T
0
T

( ) =
52
The average heat flux is
. (3.45)
The Overall Nusselt number is then:
. (3.46)
The heat transfer solution is valid for all Prandtl numbers, but it is impacted by the relative length
of the channel, reflected in the following two conditions:
. (3.47)
3.7 Mixed Convection
3.7.1 Vertical Flow
It is a very rare case where there is natural convection on its own with no external flow.
Even within a building, there is constantly air being circulated by the heating and cooling vents.
Heat is moved by a combination of both effects in most cases. In any case, the heat transfer prob-
lem is the heat flux from the wall to the fluid in the room. There is a way to determine which kind
of heat transfer is taking place. The dominant mechanism of heat transfer is found by determining
the thickness of the thermal boundary layer, o
T
. The first step is to determine the magnitude of the
boundary layer under each heat transfer mechanism. Then use the following rule to determine the
type of convection:
(3.48)
q
0 H
Q
2H
------- =
Nu
q
0 H
H
T
0
T

( )k
-------------------------- =
Ra
D
24
---------- =
Ra
D
1 4
2
H
D
----
\ .
| |
1 4
< Pr 1 < ( )
Bo
D
1 4
2
H
D
----
\ .
| |
1 4
< Pr 1 < ( )
o
T
( )
NC
o
T
( )
FC
< Natural Convection
o
T
( )
NC
o
T
( )
FC
> Forced Convection
53
The thickness of the boundary layer depends on the flow.
3.7.2.1 Pr > 1
A flow in which there is natural convection has a thermal boundary layer on the order of
. (3.49)
The case where the flow has an external flow has a boundary layer on the order of:
. (3.50)
In summary, the type of convection can be found with the following rule:
(3.51)
3.7.3.2 Pr < 1
The same argument can be used for fluids which have Prandtl numbers that are less than 1.
The summary of these equations is:
. (3.52)
3.7.4 Horizontal Flow
Many cases see mixed convection in a horizontal flow. In this case the mixed convection
is more of a buoyancy effect in the flow. For the most part the heat transfer is done by the forced
convection. The buoyancy or natural convection effects create some turbulence in the flow. If the
system is unstable then the increases the vertical velocity in the flow. If the flow is stable then
there will be some downward velocity in the fluid as the fluid cools near the surface. The second
o
T
( )
NC
yRa
y
1
4
---
~
o
T
( )
FC
yRe
y
1
2
---
Pr
1
3
---
~
Ra
y
1 4
Re
y
1 2
Pr
1 3
----------------------------
> O(1) Natural Convection
< O(1) Forced Convection

Bo
y
1 4
Pe
y
1 4
--------------
> O(1) Natural Convection
< O(1) Forced Convection

54
case will create much less flow because the cooler air is presumably already at the bottom. In the
unstable situation the less dense fluid will want to rise upwards. In the second case the cooler air
is already on the bottom. It has less reason to move. The stable situation can be said to actually
decrease the turbulent motion in the flow.
55
CHAPTER 4
ATMOSPHERIC BOUNDARY LAYER
The atmospheric boundary layer (ABL) transfers heat in all of the aforementioned meth-
ods. If there is wind, it transfers heat based on forced convection. In the absence of wind, natural
convection will transfer heat to the atmosphere, or it will transfer heat to the ground. The makeup
of the atmosphere has two main classifications. These are effectively distinguished by stability
parameters which will be discussed later in this chapter. The local thermal stability of the atmo-
sphere is characterized as stable or unstable. If the ground is warmer than the air, the system is
unstable. If the air is warmer, the system is stable. This chapter will take a look at the stable and
unstable classifications and how they affect turbulence and heat transfer in the atmosphere. The
method of heat transfer is usually mixed convection, with the dominant transfer method found
using stability parameters. The natural convection creates a buoyancy effect due to the tempera-
ture differences in the atmosphere. This will lead to more turbulence in the atmosphere, but will
not make a significant contribution to heat transfer.
4.1 References
Information found in this chapter was taken primarily from Kaimal and Finnigan [10].
This text focuses on the general equations, and the makeup of the entire boundary layer. For the
specific conservation equations, Garratt was used [7]. These two books address the atmospheric
56
boundary layer from the side of fluid mechanics perspective. There are several other books that
are good for learning more about the physics of the ABL, and these are Curry and Webster [5],
Stull [18], and Dutton [6]. These texts are primarily interested in the meteorology of the atmo-
sphere and the ABL, but the fluid mechanics are addressed as well. If the reader wished to read
more about the ABL in the context of actual experiments, they are directed to the following jour-
nal articles: Kanechika et al. [21] and Qian et al. [22].
4.2 Important Notes on Variable Changes
The author has decided to try and keep consistent with books on the ABL, so that if the
reader chooses to seek other sources of information, he or she will not be confused. This means
that a couple of variables had to be changed in order to remain consistent with ABL texts. In this
chapter, the z direction will be normal from the surface. Previously this had been the y direction,
which now becomes the crossflow coordinate. Previously u had been used to describe the nondi-
mensional temperature, but now it will be used to stand for the potential temperature. This will be
discussed more when it is introduced.
4.3 Fluid Mechanics and Governing Equations for the ABL
Like any situation, a set of equations can be derived from the general turbulent equations
that are specifically tailored for ABL flow. Some of the modifications to these equations are
based on changes in density (thermal stratification) and the effects of the earths rotation. Further
discussion of these equations can be found in Stull [18], but the main points will be presented
here. We start with the general turbulence equations with the Boussinesq approximation assumed.
The continuity equation is unchanged
57
. (4.1)
The momentum equation has a couple of changes to it. Both the streamwise and normal compo-
nents are non-negligible because of the effects of gravity. The cross flow momentum can be
neglected but will be kept here for simplicity. The equation for conservation of momentum in the
ABL is:
. (4.2)
Several changes have been made to the general equations. The body force has been replaced with
gravity, which acts only in the z direction as indicated by the o. The f
c
term reflects the effect of
the earths rotation, or Coriolis effects. The Coriolis parameter defined as
(4.3)
where O is the earths rotation rate and | is the latitude. This acts only in the horizontal direc-
tions, and can often be neglected close to the surface. The energy equation contains two terms
that are added to tailor the equations to the ABL. These terms account for radiation, and for latent
heat release. Also temperatures have been replaced by the potential temperatures, and allowances
for unsteady flow have been made. The conservation of energy equation is:
, (4.4)
where Q
j
is the net radiation in the j direction, L
v
is a constant dependent on temperature and
phase change, and E is the mass of water vapor per unit volume per unit time created by a phase
change. The ABL has many more variables than a typical boundary layer, and the conservation
equations must be modified to account for these circumstances.
cu
i
cx
j
------- 0 =
cu
i
ct
------- u
i
+
cu
i
cx
j
------- o
i3
g f
c
c
ij3
u
j
1

---
cP
cx
i
------- u
c
2
u
i
cx
j
cx
j
---------------
cu'
i
u'
j
cx
j
-------------- + + + =
f
c
2tO | sin =
cu
ct
------ u
j
cu
cx
j
------- + o
c
2
u
cx
j
--------
c u'
j
u' ( )
cx
j
------------------
1
c
P
---------
cQ
j
cx
j
---------
L
v
E
c
p
--------- =
58
4.4 Flat, Infinite, Uniform Terrain
The primary focus of this chapter will be on a boundary layer that is formed from flow
over flat, uniform terrain. Another assumption is that the terrain is infinite in all directions. Real-
istically this is not achievable, but the solution can approximate a non-infinite, slightly non-uni-
form region. The closest the earth comes to the ideal situation is flow over the ocean. A situation
like this allows us to assume horizontal homogeneity (partial derivatives of mean quantities in the
x and y directions, often called the advection terms, can be neglected). The flow over land can be
locally homogeneous if the surface is relatively flat with little vegetation. We can also assume
stationarity in most cases, that is the flow does not vary with time. Although this is not an entirely
accurate assumption because of the nature of the atmosphere, the flow can often be viewed as a
series of periods that are stationary.
4.4.1 ABL Structure
The ABL is typically divided into two sections. The first is a surface layer that extends
approximately 50-100 m from the earth. Here the shearing stress is essentially constant, there is
no effect of the earths rotation on the flow, and the flow structure is dependent on surface friction
and the vertical temperature gradient. Above that layer is a region that extends another 500-1000
m. The shearing stress is no longer constant, and the flow structure is based not only on the sur-
face friction and the temperature gradient, but also on the earths rotation. Above these regions
there is no influence from surface friction.
The depth of the ABL, the height at which the atmosphere becomes geostrophically bal-
anced is approximated by:
, (4.5) z
h
t
2K
m
f
c
----------
\ .
| |
1 2
=
59
where K
m
is the exchange coefficient for momentum and f
c
is the aforementioned Coriolis param-
eter. This definition is only an estimation and it assumes neutral stability (no net buoyancy
forces). More common than the neutral boundary layer are the states represented by stable and
unstable conditions. The ABL, taken as a whole is stable in nature, meaning that there is always a
cap to the ABL that is stable. During most night times, the stability stretches all the way down to
the surface. During the daytime, the stable layer forms a cap over a convective boundary layer
that is marked by instability.
The fluid mechanics of the ABL are governed by the transformation of mean kinetic
energy to turbulent kinetic energy to the energy of the smaller eddies and finally to heat. This pro-
cess is governed by the turbulent kinetic energy (TKE) equation, or the rate of change of kinetic
energy:
. (4.6)
There are five terms on the right side of the equations with meanings as follows: The first term is
the shear production. The second term is the buoyant production. The third term is the pressure
transport term. The fourth term is the turbulent transport term, and the final term is the viscous
dissipation term. All of these terms are with respect to the time rate of change of the kinetic
energy. In a steady state, these terms balance each other to equal 0.
4.4.2 Stability Parameters
The thermal stability of a boundary layer is based on the ratio of the buoyant turbulence
production to the shear turbulence production. This ratio can be expressed in different ways, but
the meaning of the ratio is consistent. Several methods are available for determining the thermal
De
Dt
-------
ce
ct
----- u'w' ( )
cu
cz
------
\ .
| |

g
u
--- w'u' ( )
1

---
c
cz
----- w'p' ( )
c
cz
----- ew' ( ) c + = =
60
stability of the ABL, and this section will briefly list some of these methods. The first is the Rich-
ardson flux number, R
f
, whose definition is:
(4.7)
The flux Richardson number comes directly from the TKE equation. It is the ratio of the second
term in the right hand side to the first term on the left hand side. It is an accurate representation of
the effect of thermal stratification, but is not used very often because of the combination of mean
quantities and fluctuations. We introduce here the potential temperature which is defined as:
(4.8)
where P
o
is a reference temperature which is usually set to 100 kPa or to the surface temperature.
It can be approximated as:
(4.9)
Specifically it is the temperature that a small portion of air would have, if it were brought to a
pressure of 100 kPa adiabatically. T is the absolute temperature of the molecules and Az is the
height difference between the airs current position and the level where it would have a pressure
of 100 kPa. When the Richardson number is less than 0, the atmosphere is unstable and if it is
greater than 0, it is stable. For a small arbitrary region around zero, the atmosphere can be consid-
ered neutral, that is, the effects of buoyancy are essentially zero.
There are two other forms of the Richardson number that are sometimes used to character-
ize the thermal stability of the atmosphere. The gradient Richardson number, or simply the Rich-
ardson number is defined as
R
f
g u ( )w'u'
u'w' cu cz ( )
-------------------------------- =
u T
P
o
P
------
\ .
| |
0.286
=
u T
g
c
P
-----Az + ~
61
. (4.10)
This definition is found by using proportionality substitutions for the terms in the flux Richardson
number. It is the most common Richardson number, which is why the gradient prefix is often
dropped. Stability definitions are categorized in the same way that they were for the flux Richard-
son number. The advantage of using this term to measure stability is that it uses mean quantities
that can be measured rather easily with common instruments. However it is not known how the
Richardson number varies with height, making it difficult to use in the surface layer.
The bulk Richardson number is defined as:
(4.11)
where is the potential temperature at height z, and is the potential temperature at the sur-
face. It is a good indicator of the stability in light winds, but for the most part it is unreliable
because it is only an estimation of the Richardson number in the first place. It is also rarely used.
More on these Richardson numbers can be found in Kaimal and Finnigan and also in Stull.
The most widely recognized stability parameter is based on the Obukhov length, L. It
comes from nondimensionalizing the TKE equation. The physical meaning is that it is propor-
tional to the normal distance from the surface at which buoyancy production is equal to shear pro-
duction. The stability parameter is a combination of the Obukhov length and the height z, and is
written:
(4.12)
Ri
g u ( ) cu cz ( )
cu cz ( )
2
------------------------------------ =
R
b
g u ( ) u
z
u
0
( ) 2
u
z
z ( )
2
------------------------------------------- =
u
z
u
0
z
L
---
g u ( ) w'u' ( )
0
u
t
3
kz ( )
--------------------------------- =
62
where is the value at the surface and k is the von Karman constant. This quantity is not
difficult to measure, and L can be assumed constant throughout the surface layer.
4.4.3 The Three States of the ABL
The three possible states of the ABL, as mentioned in the section above, are stable, unsta-
ble, and neutral. This section will provide a description of each of these states to give the reader a
better understanding of the terms. The unstable boundary layer is accompanied by a convective
boundary layer often called the CBL. The unstable layer is found near the surface while the CBL
refers to the entire boundary layer which can stretch as high as 1 or 2 km into the atmosphere.
The unstable and CBL layers begin to grow in the morning after sunset. As the sun heats the
earth, the CBL grows. It reaches its peak in the afternoon and stays relatively constant in height
for a while. The unstable layer is found near the surface during this time period. Heat transfer in
the CBL occurs by way of thermal plumes. The thermal plumes carry heat upwards in the CBL.
The plumes are sometimes small, and often these smaller plumes merge into larger plumes. The
heat is mixed in a circular manner. Air from above the CBL can penetrate the top layer often
called the capping inversion. This term is used to show that at the top of the CBL, the heat is no
longer transported upward and heat is contained within the CBL. Air penetrating the inversion
layer then causes the circular motion within the CBL.
Towards sunset, the inversion layer begins to weaken as the earth begins to cool. Air in
close proximity to the surface also cools. Turbulent motions begin to lose much of their energy
and the flow begins to settle. This heat inversion grows constantly to a height of 100-200 m. At
the surface, a stable surface layer is formed. Again it is does not stretch very high, and it is in fact
smaller than the unstable layer in the day. The eddies that were large in the CBL are now rather
w'u' ( )
0
63
small in the stable boundary layer (SBL). The energy is much less by nature, as the energy that
was contained in the thermal plumes during the day has been lost to radiation. Buoyancy forces
are driving the flow down and suppressing turbulence. The surface layer transition, as the unsta-
ble layer shrinks to zero and the stable surface layer begins to build is the period of neutral stabil-
ity. Buoyancy forces are absent in the vicinity of the surface. The flow is pure shear, that is, the
wind is the only source of motion in the surface layer. Buoyancy forces are zero or negligible.
4.4.4 Properties of the Surface Layer
Horizontal homogeneity and stationarity are most closely realized in the surface layer.
The surface layer is dominated by vertical gradients that provide the transportation of heat, mass
and momentum. The turbulent transport of these quantities is analogous to molecular diffusion
that governs laminar flows. This allows us to write:
(4.13)
(4.14)
(4.15)
where t, H, and E are the fluxes of momentum, heat, and moisture respectively. K
m
, K
h
, and K
q

are the turbulent exchange coefficients for momentum, heat, and moisture. The mean quantity in
Eqn. 4.15, , is the mean specific humidity. Horizontal homogeneity allows us to assume that
these mean quantities vary only in the z direction. These fluxes do not tell us much by them-
selves, but we will see how important they are to the profiles of the surface layer.
Let us define the term K
m
as:
t K
m

cu
cz
------ =
H K
h
c
P
cu
cz
------ =
E K
q

cq
cz
------ =
q
64
. (4.16)
The term t is a vector quantity with components t
x
and t
y
. Aligning the wind direction with the x
direction means that t
y
is 0. It is reasonable to assume that the change in t
x
with height is negligi-
ble in the surface layer so that t
x
= t
0
. Recalling the definition:
, (4.17)
allows us to make substitutions into Eqn. 4.13 to get the following relation:
(4.18)
Integration of this equation leads to:
, (4.19)
where k is the von Karman constant and z
0
is the constant of integration often called the rough-
ness length. Closer examination of this equation reveals that it is exactly the same as Eqn. 2.60
after some rearrangement of terms. This should not be surprising because of the analogy that was
used to construct the flux equations. It is valid only for a neutral surface layer. Buoyancy forces
cause the mean velocity to deviate from a logarithmic profile in a stable or unstable boundary
layer. Using the same reasoning and methods used to derive the velocity profile, we can derive
logarithmic profiles for the other quantities as well. These expressions are:
, (4.20)
(4.21)
K
m
ku
t
z =
t
0
u
t
2
=
cu
cz
------
u
t
kz
----- =
u z ( )
u
t
k
-----
z
z
0
----
\ .
| |
ln =
t

---
k
z
2
z
1
( ) ln
------------------------
\ .
| |
2
u
2
u
1
( )
2
=
H
c
P
---------
k
z
2
z
1
( ) ln
------------------------
\ .
| |

2
u
2
u
1
( ) u
2
u
1
( ) =
65
. (4.22)
Again these equations are only truly valid in a neutral boundary layer. The more that the thermal
stability deviates from neutral, the more error will be introduced. The equations are also depen-
dent on the fluxes being constant with height. Introducing the instantaneous definitions which are
identical to those that were referenced earlier when turbulent flow was introduced, the flux equa-
tions can be written:
(4.23)
(4.24)
. (4.25)
These flux equations require no assumptions.
4.4.5 Monin-Obukhov Similarity
An interesting phenomenon that has been experimentally found to exist in the ABL is the
Monin-Obukhov similarity. Many different flow properties in the atmosphere exhibit a universal
functionality of z/L. Quantities are first normalized by the characteristic velocity and tempera-
ture. The characteristic velocity is simply the friction velocity, u
t
, although it is cast in another
form based on the previously defined shear stress:
(4.26)
The characteristic temperature is the previously undefined T
*
:
(4.27)
E

---
k
z
2
z
1
( ) ln
------------------------
\ .
| |

2
u
2
u
1
( ) q
2
q
1
( ) =
t u'w' =
H c
p
w'u' =
E w'q' =
u
t
u'w' ( )
0
| |
1 2
=
T
*
w'u' ( )
0

u
t
--------------------- =
66
Although these two terms are based on the fluxes at the wall, the no slip condition states that the
velocities must be zero at the wall. The actual fluxes are actually measured at some small dis-
tance close to the wall where the fluxes can be considered constant. The following nondimen-
sional parameters result from using these characteristic values:
(4.28)
(4.29)
(4.30)
(4.31)
(4.32)
The variables o
w
and o
u
are the standard deviations of w and u, k is the von Karmans constant
again, and c is the rate of dissipation of turbulent kinetic energy. The empirical similarity rela-
tions that have been deduced are as follows:
(4.33)
(4.34)
(4.35)
(4.36)
|
m
kz u
t
( ) cu cz ( ) = wind shear
|
h
kz T
*
( ) cu cz ( ) = thermal stratification
|
w
o
w
u
t
= variability in w
|
u
o
u
T
*
= variability in u
|
c
kzc u
t
3
= dissipation of turbulent kinetic energy
|
m
1 16 z L + ( )
1 4
2 z L 0 s s
1 5z L + ( ) 0 z L 1 s s

=
|
h
1 16 z L + ( )
1 2
2 z L 0 s s
1 5z L + ( ) 0 z L 1 s s

=
|
w
1.25 1 3 z L + ( )
1 3
2 z L 0 s s
1.25 1 0.2z L + ( ) 0 z L 1 s s

=
|
u
2 1 3 z L + ( )
1 3
2 z L 0 s s
2 1 0.5z L + ( )
1
0 z L 1 s s

=
67
(4.37)
In a neutral boundary layer we can assume:
(4.38)
because of the balance between the shear production of turbulence and its viscous dissipation.
There are no M-O similarity equations for u and v.

4.4.6 Spectra and Cospectra for the Surface Layer
|
q
1 0.5 z L
2 3
+ ( )
3 2
2 z L 0 s s
1 5z L + ( ) 0 z L 1 s s

=
c u
t
2
u
t
kz
-----
\ .
| |
=
68
Turbulence is not an entirely random process. Patterns can be found in the velocity and
other flow properties. This is the result of the organization of the coherent structures in the flow:
eddies and vortices for example. The eddies derive most of the energy from the interacting with
the mean flow, with most of the energy being contained in the largest eddies. These large eddies
also are significant in transporting most of the turbulence. Instabilities in the eddies giving them a
limited time before they are broken into smaller eddies. These eddies are in turn broken into
smaller eddies, until they are small enough to be viscously dissipated to heat.
Fourier spectra and cospectra are a method of separating the scales of motion in a turbu-
lent flow. It shows how much energy is contained in each scale, and how that changes. When the
flow parameters are appropriately normalized, the spectra and cospectra exhibit functional depen-
dence only on z/L in the surface layer. These curves are universal and is very helpful in under-
standing the structure of turbulence. The examination of the energy spectrum requires the
definition of the two point covariance tensor:
(4.39)
The Fourier transform of this term is the energy spectrum tensor, E , wherek represents the
wavenumber vector. The energy spectrum tensor presents a picture of the energy distribution in
the surface layer. Three major regions can be identified in the flow. The first is the energy-con-
taining range, where the majority of the turbulent energy is held and produces. The second region
is the inertial subrange where the energy is transferred to smaller scale motions. The final region
is the dissipation range in which the kinetic energy is converted into heat or internal energy. Each
of these terms has its own length and time scales, with the scales varying in direction as well. In
the energy containing region, the time scale in the u direction is defined as:
R
ij
x
k
r
k
, ( ) u'
i
x
k
( )u'
j
x
k
r
k
+ ( ) =
x
i
k
j
, ( )
69
(4.40)
where T
u
is the corresponding timescale and is the time lag from one reading to the next. The
maximum energy occurs at a wavenumber corresponding to k ~ 1/A. The dissipation range has a
length scale of:
, (4.41)
where v is the kinematic viscosity of dry air. To put it in perspective, A ranges from 10 to 500 m,
while q is on the order of 0.001 m.
The one dimensional spectrums are the forms that are most commonly used to describe the
energy in a boundary layer. Subsequent spectra will be a function of the streamwise wave number
k
1
. The inertial subrange has a spectrum that is governed by the similarity equations:
(4.42)
(4.43)
where o
1
is the Kolmogorov constant with a value between 0.5 and 0.6. It is believed that the
flow in the inertial subrange exhibits local isotropy. The second equation is a consequence of this
isotropy. There is also a temperature spectrum available with the equation:
, (4.44)
where |
1
has a value of around 0.8 and N
u
is the dissipation rate for half of the temperature vari-
ance. More information on the spectra in the energy containing range can be found in Kaimal and
Finnigan [10]. For now however we will focus on the spectra and cospectra of the surface layer.
A
u
uT
u
u
u
( )
0

}
u
u' t ( )u' t + ( )
o
u
2
--------------------------------- d
0

}
= = =
q
u
3
c
-----
\ .
| |
1 4
=
F
u
o
1
c
2 3
k
1
5 3
=
F
v
k
1
( ) F
w
k
1
( ) 4 3 ( )F
u
k
1
( ) = =
F
u
|
1
c
1 3
N
u
k
1
5 3
=
70
Most measurements are made in frequency space, rather than wavenumber space. For this
reason we need to convert between wavenumber space and frequency space, and back again
depending on the situation. The conversion is this:
(4.45)
and
(4.46)
where S
u
represents the frequency spectrum. Let us also define the nondimensional frequency:
. (4.47)
Remembering the equation for the dimensionless turbulent dissipation, we have:
. (4.48)
The dimensionless spectra for the inertial subrange can be expected to follow the following
trends:
(4.49)
(4.50)
(4.51)
. (4.52)
k
1
2tf u =
k
1
F
u
k
1
( ) fS
u
f ( ) =
n
fz
u
---- =
|
c
kzc
u
t
3
-------- =
fS
u
f ( )
u
t
2
|
c
2 3
---------------- 0.3n
2 3
=
fS
v
f ( )
u
t
2
|
c
2 3
---------------- 0.4n
2 3
=
fS
w
f ( )
u
t
2
|
c
2 3
---------------- 0.4n
2 3
=
fS
u
f ( )
T
*
|
N
|
c
1 3
-------------------------- 0.3n
2 3
=
71
It becomes more difficult to produce equations for the entire surface layer. One set of spectra that
is valid for the surface layer under neutral conditions is:
(4.53)
(4.54)
(4.55)
Just as important as the spectra in the surface layer are the cospectra, which provide infor-
mation about the relationship between multiple turbulent quantities. Two of the most important
cospectra are the relationships between streamwise velocity u and the normal velocity w, and
between the normal velocity w and the potential temperature u. These frequency cospectra, repre-
sented by C are given as:
(4.56)
(4.57)
with
(4.58)
(4.59)
4.4.7 Beyond the Surface Layer
fS
u
f ( )
u
t
2
--------------
102n
1 33n + ( )
5 3
------------------------------- =
fS
v
f ( )
u
t
2
--------------
17n
1 9.5n + ( )
5 3
-------------------------------- =
fS
w
f ( )
u
t
2
---------------
2.1 102 ( )n
1 5.3 33 ( )n + ( )
5 3
-------------------------------------------- =
fC
uw
u
t
2
G z L ( )
------------------------- 0.05n
4.3
=
fC
wu
u
t
T
*
H z L ( )
------------------------------ 0.14n
4 3
=
G z L ( )
1 2 z L 0 s s
1 7.9z L + 0 z L 2 s s

=
H z L ( )
1 2 z L 0 s s
1 6.4z L + 0 z L 2 s s

=
72
The equations that have been discussed previously are valid only in the surface layer.
Beyond the surface layer are other regions known as the mixed layer and the stable outer layer.
Further information on these two regions can be found in Kaimal and Finnigan [10].
4.5 Plant Canopies
The next few sections will briefly describe the flow over other types of terrain. There will
not be as much depth in these sections as in the previous flat terrain section, but if the reader
wishes to learn more, they are referred to Kaimal and Finnigan [10]. The air flow over a canopy
shows significant differences from flow over flat terrain. In flow over flat terrain, momentum is
lost to the ground. For flow over plants, especially tall plants, momentum is lost in the vegetation
as well. The log dependence of the velocity profile does not have its origin at the top of the
plants, but at some height above the plants. The exact height is dependent on the makeup of the
vegetation and the drag force it exerts on the flow. Turbulence intensities are much greater within
the canopy itself than above the vegetation, or in flow over flat terrain. The M-O relations are still
valid at a certain height above the canopy, but within the canopy there is no universal set of equa-
tions to describe the flow. There are too many non-uniform features that affect the flow to predict.
The canopy also releases gases and water vapor in the air in a way that is a different from flat ter-
rain. Flow over plant canopies can thus not be viewed as a simple application of flow over flat
terrain.
4.6 Changing Terrain
Most real world applications involve changing terrain. Most studies have been done over
the change from one uniform surface to another, so this section will address that situation. There
are two types of possible changes that can affect the surface layer flow. The first is a change in
73
surface roughness which will affect the momentum flux, and the second is the change in some
surface quantity such as the heat or the moisture. There is a noted difference in whether the air
flows from a smooth surface to a rough surface, and the same can be said for differently heated
surfaces. Situations in which the flow is over a changing terrain are no longer one dimensional.
The flows are two dimensional and heterogeneous. Monin-Obukhov similarities are no longer
valid in a two-dimensional flow. A boundary layer within a boundary layer develops after a
change in surface. It slowly grows with distance. The profiles and fluxes are no longer in equilib-
rium with the new surface once the change is made. Much work is still being done in these areas,
but these are the basic effects of a change in terrain.
4.7 Hills
Flow over a hill will cause a large scale change in the pressure field, as compared to the
small scale change that would occur over changing terrain. Since flow over hills is very common,
they are important in the study of the ABL. The question is how much drag force is exerted on
the airflow by the hills, This is an effect of both the size and shape of the hill or hills involved.
Higher hills. Buoyancy is an important factor when considering flow over hills. If the hills are
high enough, gravity will exert forces on the flow in order to push down the flow. In a stable
boundary layer, buoyancy effects on the flow patterns are present even in smaller scale hills. One
must determine how much the hill or hills in question disturbs the boundary layer in order to con-
sider the effects of buoyancy. Often in flow over hills, several layers of the boundary layer seem
to develop. Near the surface the M-O similarities still hold, but as one works away from that sur-
face, local equilibrium assumptions no longer apply. Mathematical modelling is the primary way
to deal with flow over hills, because of the difficulty in measuring the flow over hills.
74
CHAPTER 5
MEASUREMENTS TECHNIQUES
Practical experiments need a way to collect data in order to perform detailed analysis on
heat transfer and turbulence. This section will illustrate some of the techniques and equipment
that are used to collect data, both in the ABL and in the wind tunnel. Information in this chapter
was taken from Kaimal and Finnigan [10] primarily. Hotwire information was taken from Bruun
[3] and the reader is directed there for an in-depth discussion of hotwires.
5.1 Velocity
The most common device to make velocity measurements is the hotwire. It can come in
many different forms, such as X-wires and rakes in order to get directional and spatial measure-
ments. However its spatial resolution is still minimal, and it is rather fragile. The measurements
taken by a hotwire are inherently temperature dependent. This causes a problem when working in
a temperature gradient. If the change in temperature is not accounted for, then the velocity read-
ings will be inaccurate. According to Bruun [3], this can be accounted for in a couple of different
manners. It is possible to build a temperature corrector into a wheatstone bridge in order to auto-
matically account for changes in the temperature. In some cases, exterior devices can be used to
measure the temperature and then incorporate that into the data analysis. Thermistors are com-
monly used to do this, but they have a couple limitations. The first is that the acquisition fre-
75
quency is a couple of seconds long. Any temperature variations less than two seconds long must
be ignored. If they are significant, then this can cause errors in the data. The other problem is the
proximity that the thermistor must be placed close to the hotwire. The technique works if the
thermistor is measuring the same temperature variations that the hotwire is feeling. If these two
are feeling different temperatures, there will be inaccuracies in the measurements.
Particle Image Velocimetry (PIV) is also a common method for determining velocity mea-
surements. It can provide good directional and spatial resolution, although 3-D PIV is still being
perfected. It requires much more equipment that the hotwire and the flow must be seeded. Seed-
ing is an added complication, that is at best, nontrivial and at worst, difficult to execute well. The
temporal capabilities of PIV are also much less than the hotwire. A PIV system captures images
at a significantly lower frequency than a hotwire. There are more variables to producing good
data, but the spatial resolution and directional capabilities provide more information about the
flow than the hotwire.
Freestream velocities in the wind tunnel can be computed with pitot probes. These are not
good for collecting turbulent or instantaneous data, but they provide important mean velocity data
to supplement the instantaneous measurements that are taken by the other equipment.
The three most common instruments for measuring the mean velocity in the atmospheric
boundary layer are the propeller anemometer, the cup anemometer and the vane. The propeller
anemometer consists of a propeller that rotates about the streamwise axis which is then attached
to a structure that is allowed to rotate about the vertical axis. The propeller is able to determine
the speed at which the air is flowing, and the structure allows the discernment of flow direction.
The cup anemometer has multiple shapes that it can employ. The most common is a three-cup
system that rotates about the vertical axis in order to determine the air velocity. The cup aneme-
76
ometer is usually coupled with a vane that can determine the wind direction. The two instruments
are situated so as not to interfere with each other, usually on the same horizontal plane. These
instruments are usually too large to collect data inside of a wind tunnel, but they devices are usu-
ally calibrated in a wind tunnel before they taken out into the atmosphere. The drawback to using
these instruments is that they have a very low dynamic response time. It takes a little time for the
device to reach equilibrium with the air flow and report the proper air speed. They are useful for
measuring the mean air velocity, but they cannot measure the turbulent fluctuations that are
present in a flow.
The two methods of choice for measuring instantaneous velocity in the wind tunnel have
already been described as hotwires and PIV. Hotwires have been extensively used outdoors, but
they are very fragile and susceptible to damage from atmospheric flows. PIV has only recently
been used in the ABL to collect data and still has a long way to go before it is perfected. Another
useful instrument for measuring instantaneous velocity in the ABL is the sonic anemometer. The
principle behind a sonic anemometer is that a sensor releases either a sonic pulse or a continuous
sound wave. The time it takes for the pulse or wave to travel from one sensor to another, which is
positioned a certain distance away, determines the velocity of the air flowing between the sensors.
Although there are many different arrangements for a sonic anemometer, one of the most common
ways to structure the instrument is to align three of these pairs of sensors so that each one records
the wind velocity for each of the three directions. The advantage of using the sonic anemometer
is that has a very good dynamic response time, has good directional discernability, and is durable
and reliable in the outdoors. The disadvantages include the fact that the ability to measure the
velocity is based on averaging the velocities across the distance between the sensors. If the sen-
sors are too far apart, then the measurements will be inaccurate. The size of the arrangement also
77
does not allow for good spatial resolution. The hotwire is still better able to get achieve spatial
resolution when used with a rake. Each individual situation must be evaluated to select the best
possible instruments to produce the desired results.
5.2 Temperature and Heat Flux
Measurements of vertical gradients of mean temperature (an indicator of heat flux) are
collected typically using one or more of four different devices: platinum resistance thermometers,
thermocouples, thermistors, and quartz thermometers. When these are placed in the ABL espe-
cially, they are placed in metal or glass for protection from the outdoor elements. This does not
provide any problem when only the mean temperatures are being measured. The best of these
instruments is the quartz thermometer, but it is also the most expensive.
Computing the instantaneous temperature data is also done using the previously men-
tioned thermistors and thermocouples. Fine-wire platinum resistance thermometers and sonic
thermometers can also be used, and often are better suited for the instantaneous measurements.
Platinum wires can be made smaller than thermistors and thermocouples, and they are thus more
responsive and better suited for making flux measurements near the surface in both a wind tunnel
and in the ABL. The glass and metal shields that are used in the devices that are used to measure
mean temperatures are not used in these instruments. They are more likely to damage in the out-
doors from dirt and other elements. The sonic thermometers have a good frequency response.
The results are however affected by the humidity in the air, and are sometimes obscured by the
wind direction normal to the sensing path.
78
REFERENCES
[1] Bejan, Adrian. Convection Heat Transfer. Second Edition. John Wiley & Sons Inc.: New
York, New York. 1995.
[2] Bernard, Peter S. and Wallace, James M. Turbulent Flow. John Wiley & Sons, Inc.: Hobo-
ken, New Jersey. 2002.
[3] Bruun, H. H. Hot-Wire Anemometry. Oxford University Press: New York, New York.
1995.
[4] Burmeister, Louis C. Convective Heat Transfer. John Wiley & Sons, Inc.: New York, New
York. 1993.
[5] Curry, Judith A. and Webster, Peter J. Thermodynamics of Atmospheres & Oceans. Aca-
demic Press: New York, New York. 1999.
[6] Dutton, John A. Dynamics of Atmospheric Motion. Dover Publications, Inc.: New York,
New York. 1986.
[7] Garratt, J.R. The atmospheric boundary layer. Cambridge University Press: New York, New
York. 1992.
[8] Hinze, J.O. Turbulence. McGraw-Hill, Inc.: New York, New York. 1975.
[9] Janna, William S. Engineering Heat Transfer. CRC Press: New York, New York. 2000.
[10] Kaimal, J.C. and Finnigan, J.J. Atmospheric Boundary Layer Flows. Oxford University
Press: New York, New York. 1994.
[11] Kays, W.M. and Crawford, M.E. Convective Heat and Mass Transfer. McGraw-Hill, Inc.:
New York, New York. 1993.
79
[12] Kutateladze, S.S. and Leontiev, A.I. Heat Transfer, Mass Transfer, and Friction in Turbulent
Boundary Layers. Hemisphere Publishing Corporation: New York, New York. 1990.
[13] Launder, B.E. and Spalding, D.B. Lectures in Mathematical Models of Turbulence. Aca-
demic Press: New York, New York. 1972.
[14] Oosthuizen, Patrick H. and Naylor, David. Introduction to Convective Heat Transfer Analy-
sis. McGraw-Hill, Inc.: New York, New York. 1999.
[15] Petukhov, B.S. and Polyakhov, A.F. Heat Transfer in Turbulent Mixed Convection. Hemi-
sphere Publishing Corporation: New York, New York. 1988.
[16] Pope, Stephen B. Turbulent Flows. Cambridge University Press. New York, New York.
2000.
[17] Schlicting, Dr. Hermann. Boundary-Layer Theory. McGraw-Hill, Inc.: New York, New
York. 1979.
[18] Stull, Roland B. An Introduction to Boundary Layer Meteorology. Kluwer Academic Pub-
lishers: Boston, Massachussetts. 1988.
[19] White, Frank M. Viscous Fluid Flow. McGraw-Hill, Inc.: New York, New York. 1974.
[20] Adrian, R.J., Ferreira, R.T.D.S., and Boberg, T. Turbulent Thermal Convection in Wide Hor-
izontal Fluid Layers. 1986. Experiments in Fluids. Vol 4. 121-141.
[21] Kanechika, O., Kondo, J., and Yasuda, N. Heat and Momentum Transfers under Strong Sta-
bility in the Atmospheric Surface Layer. 1978. Journal of the Atmpospheric Sciences.
Vol. 35, 1012-1021.
[22] Qian, M.W., Longhetto, A., and Cassadro, C. Heat Energy Balance in the Convective Atmo-
spheric Boundary Layer at Xianghe (Beijing Area), China. (2000). Journal of the Atmo-
spheric Sciences. Vol. 57. 3881-3891.

You might also like