Professional Documents
Culture Documents
Second Edition
Advanced Heat Transfer
Second Edition
Greg F. Naterer
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made
to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all
materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all
material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not
been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any
future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized
in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying,
microfilming, and recording, or in any information storage or retrieval system, without written permission from the
publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA
01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For
organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Fundamental Concepts and Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Conservation of Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Thermophysical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.1 Thermodynamic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.2 Kinematic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.3 Transport Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Heat Conduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 Thermal Radiation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7 Phase Change Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.8 Mass Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2. Heat Conduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 One-Dimensional Heat Conduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.2.1 Heat Conduction Equation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.2.2 Thermal Resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.3 Fins and Extended Surfaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3 Multidimensional Conduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.1 Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.2 Orthogonal Curvilinear Coordinates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4 Method of Separation of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.5 Conformal Mapping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.6 Transient Heat Conduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.6.1 Lumped Capacitance Method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.6.2 Semi-Infinite Solid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.6.3 Unidirectional Conduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3. Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.2.1 Conservation of Mass (Continuity Equation) . . . . . . . . . . . . . . . . . . . . . . . . . 69
vii
viii Contents
Appendices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 471
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 497
Preface
Since the publication of the first edition of this book in 2002, heat transfer engineering has
expanded in many new and emerging technologies. This latest second edition was moti-
vated by a desire to broaden the scope of applications and enhance the depth and range
of analysis of thermal engineering systems. It was also updated based on input obtained
from many colleagues and readers over the years since the first edition. The first edition,
entitled Heat Transfer in Single and Multiphase Systems, provided a unique source of material
that covered each mode of multiphase heat transfer, as well as the fundamentals of heat
transfer. The title of the second edition was modified to Advanced Heat Transfer in order to
better reflect the focus on advanced methods of analysis and the broader range of applica-
tions, including new topics such as chemically reacting flows with heat transfer.
Traditionally, the advanced heat transfer topics of phase change and chemically reacting
flows are usually separated into separate sources which focus on a specific mode of multi-
phase heat transfer in depth. As a result, analogies among these modes are not explored, and
a systematic framework is not available in a single source. Also, advanced methods of anal-
ysis are normally covered in depth for a single specific mode of heat transfer, such as con-
duction or convection, rather than an overview of the primary solution methods for a wider
range of heat transfer modes. In this second edition, a single source of advanced solution
techniques is presented for a wide range of applications and modes of heat transfer.
Several new sections, figures, tables, example problems at the end of each chapter, and
graphs were added to the second edition. Solution methods by conformal mapping and
orthogonal curvilinear coordinates were added to the analysis of heat conduction (Chapter
2). In the treatment of convection (Chapter 3), new sections were added on entropy and the
second law of thermodynamics regarding an apparent entropy production difference as an
error indicator of approximate solutions. Also, a nondimensional entropy generation num-
ber was presented in terms of standard correlations for the skin friction coefficient and Nus-
selt number. The chapters on multiphase heat transfer were reorganized and renamed. In
Chapter 5 (Gas–Liquid Two-Phase Flows), several new sections were added, including
forced convection boiling, multiphase flow regimes in vertical tubes, horizontal two-phase
flows, forced convection condensation, and heat pipe fins.
The chapter on gas–liquid–solid systems was deleted from the first edition, its problems
were transferred to other chapters, and a new Chapter 6 (Multiphase Flows with Droplets
and Particles) was added in the second edition. New sections were also added to Chapter
6, including analysis of packed beds, impinging droplets on a freezing surface, particle for-
mation from evaporating droplets, external flow with droplets, and forced convection melt-
ing of particles.
Furthermore, a new chapter on chemically reacting flows (Chapter 8) was added. The
chapter includes sections on mole and energy balances, combustion, shrinking core and pro-
gressive conversion models of gas–solid reacting mixtures, and fluidized beds, including
hydrodynamics, solid conversion, and reacting flow models. In Chapter 9 (Heat Exchang-
ers), a new section was added on transient thermal response in heat exchangers. New sec-
tions on the finite volume method (including SIMPLE and SIMPLEC) and turbulent flow
modeling were added to the last chapter on computational heat transfer (Chapter 10). In
these updated chapters, numerous additional problems were added at the end of the
xiii
xiv Preface
chapters. Lastly, new tables were added to the appendices including convection equations
in various coordinate systems and thermodynamic properties of gases and liquids.
This second edition aims to cover a wide range of advanced heat transfer topics suitable
for both undergraduate and graduate level courses. It can serve both introductory and fol-
low-up courses in heat transfer, such as advanced topics courses or graduate-level heat
transfer. It would normally follow a first course in fluid mechanics. The student is expected
to have knowledge of vector calculus and differential equations.
The text is organized into six main parts: (i) introduction (Chapter 1); (ii) primary single-
phase modes of heat transfer (Chapters 2 to 4); (iii) multiphase heat transfer (Chapters 5–7);
(iv) chemically reacting flows (Chapter 8); (v) heat exchangers (Chapter 9); and (v) compu-
tational heat transfer (Chapter 10). The introduction provides the reader with fundamentals
of heat transfer. The modes of single phase heat transfer, including conduction, convection,
and radiation, are covered in the second part. Then, the reader may focus on all multiphase
systems (Chapters 5–7) or on any particular system, such as liquid–solid systems (Chapter
7), without a loss of continuity. Finally, chemically reacting flows, heat exchangers, and
numerical heat transfer are presented in the last parts, respectively. Again, either of these
three chapters can be studied independently of the others without a loss of continuity.
I’m grateful to numerous colleagues and students who have contributed in significant
ways to the development and preparation of the material in this second edition. It has
been a pleasure for me to serve on the Thermophysics Technical Committee of AIAA (Amer-
ican Institute of Aeronautics and Astronautics) and as the Editor-in-Chief of the AIAA Jour-
nal of Thermophysics and Heat Transfer. These roles have provided a source of valuable
inspiration and creativity. Also, special thanks to Brian McDonald, Nancy Chafe, Brandon
Howell, and Guofei Yan at Memorial University for their helpful contributions.
Greg F. Naterer
St. John’s, Canada
Author
Dr. Greg F. Naterer is presently the Dean of the Faculty of Engineering and Applied Science
and a Professor of Mechanical Engineering at Memorial University, Canada. He previously
held a Canada Research Chair in Advanced Energy Systems. Dr. Naterer has served in
prominent national and international leadership roles in education and research, including
as Chair of the National Council of Deans of Engineering and Applied Science of Canada
(NCDEAS) and the Thermophysics Technical Committee of AIAA (American Institute of
Aeronautics and Astronautics).
Dr. Naterer has made significant contributions to the fields of heat transfer, energy sys-
tems, and fluid mechanics. He led an international team that developed and constructed
a copper-chlorine cycle of thermochemical hydrogen production. In addition to this second
edition, he has coauthored two other books: Hydrogen Production from Nuclear Energy (with
I. Dincer, C. Zamfirescu; Springer, 2013) and Entropy-Based Analysis and Design of Fluids
Engineering Systems (with J.A. Camberos; CRC Press/Taylor & Francis, 2008).
Dr. Naterer is presently the Editor-in-Chief of the AIAA Journal of Thermophysics and Heat
Transfer. Among his awards and honors for teaching and research, Dr. Naterer has received
the EIC Julian C. Smith Medal, CNS Innovative Achievement Award, CSME Jules Stachie-
wicz Medal, and Best Professor Teaching Award. He is a Fellow of the Canadian Society for
Mechanical Engineering (CSME), American Society of Mechanical Engineers (ASME),
Engineering Institute of Canada (EIC), and Canadian Academy of Engineering (CAE).
Dr. Naterer received his PhD degree in Mechanical Engineering from the University of
Waterloo, Canada, in 1995.
xv
List of Symbols
xvii
xviii List of Symbols
Greek Symbols
Subscripts
Superscripts
′
turbulent fluctuating quantity
′′
flux quantity (per unit area)
* nondimensional quantity
1
Introduction
Until the mid-nineteenth century, heat was interpreted as an invisible form of matter called a
“caloric.” The caloric was understood as a fluid substance that was responsible for heat
phenomena. This perspective was held until about 1840, when the British physicist James
Joule showed that heat was not a material substance, but rather a form of energy. This
led to a new interpretation of heat as a mechanism of thermal energy transfer across the
boundaries of a system. This new insight led to a deeper understanding of the fundamental
modes of heat transfer, namely conduction, convection, and radiation.
Conduction heat transfer occurs from one part of a solid body or fluid to another, or
between bodies in contact, without any movement on a macroscopic level. Convection
occurs when heat is transferred between a solid surface and a fluid, or between different
fluid regions, due to bulk fluid movement. For forced convection, external processes
(such as pressure-induced forces) drive the fluid motion. These external processes may
result from devices such as pumps, fans, or atmospheric winds. In contrast, buoyancy
(rather than an external force) drives the fluid motion for natural convection (or free convec-
tion). Radiation is another fundamental mode of heat transfer. It occurs from the emission
of electromagnetic waves, or photons as packets of energy, by all surfaces above absolute
zero temperature (zero Kelvin, or −273.15◦ C). These processes will be described in detail
in individual chapters devoted to each mode of heat transfer.
Multiphase heat transfer, such as phase change in gas–liquid and liquid–solid flows, arises
in many practical applications. For example, predicting and controlling the operation of
condensers in thermal power plants in an efficient manner requires an understanding of
gas–liquid transport phenomena during condensation. In material processing technologies,
such as extrusion or casting, the liquid metal phase change has a significant role in the final
material properties, such as tensile strength, due to the alignment of grain boundaries and
solidification shrinkage. In upcoming chapters, various solution techniques and case studies
dealing with such multiphase systems will be presented.
A common design issue in heat transfer engineering is finding ways to reduce (or increase)
the heat transfer to a minimum (or maximum) value. For example, consider the directional
solidification of Ni-based superalloy turbine blade castings by liquid metal cooling. Several
complex interactions occur during the solidification process, such as shrinkage flow at the
phase interface, thermalsolutal convection in the bulk liquid, and radiative heat transfer.
Effective thermal control is important so that grain boundaries are aligned parallel to the
blade axes during solidification. In this way, the solidified material can most effectively
resist conditions of maximum stress during turbine operation.
Another common design issue in heat transfer engineering is how to achieve a specified
heat transfer rate as efficiently and economically as possible. For example, in microelectron-
ics assemblies, designers seek better ways of cooling the electronic circuits and more efficient
alternatives to conventional cooling with a fan. Another example is deicing of aircraft, wind
turbines, and other iced surfaces. Aircraft icing increases drag and weight, and it presents a
serious danger to air safety. It can damage downstream components if the attached ice
1
2 Advanced Heat Transfer
breaks off, and ingested ice can damage the jet engine. Several heating and cooling modes at
the surface affect the ice accretion such as surface heating, convection, conduction, and
incoming supercooled droplets which freeze on the surface. These combined processes
are complex and involve multiple modes of heat transfer simultaneously. This book will
present a range of advanced solution methods for the analysis of single and multiphase
thermal systems.
the thermal conductivity of polymers is usually about two orders of magnitude less than
that of metals and ceramics.
Unlike solids, there is a molecular freedom of movement with no fixed structure in liquids
and gases. From common everyday experience, liquids need a container for storage and
they cannot resist imposed shear stresses. However, they can resist compression. These
characteristics indicate some key differences between solids and liquids from a microscopic
point of view. Some materials, such as slurries, tar, toothpaste, and snow, exhibit multiple
characteristics. For example, tar resists shear at small stresses, but it flows at high stresses.
The study of these forms of hybrid materials is the subject of rheology.
In order to determine the macroscopic properties of a solid or fluid, such as density or ther-
mal conductivity, it is normally assumed that the substance is a continuous medium.
This approach is called the continuum assumption. It is an idealization that treats the
substance as continuous, even though on a microscopic scale, it is composed of individual
molecules. The continuum assumption is normally applicable to fluids beyond a minimum
of 1012 molecules/mm3. But in certain circumstances, the continuum assumption cannot be
used, for example, in rarefied gases at low pressure like the conditions experienced by space-
craft atmospheric reentry at high altitudes.
The continuum assumption considers macroscopic averaging rather than microscopic
properties arising from a varying spatial distribution of molecules. For example, consider
the definition of density by macroscopic averaging of the mass divided by the volume. In
this definition, a volume is chosen to be large enough so that the density is properly
defined. The mass of molecules is assumed to be distributed uniformly across the volume.
But the number of molecules varies within the volume if the volume size approaches the
scale of the mean free path of molecular motion. If the volume size is less than the mean
free path, then significant variations in density can arise due to the molecular fluctua-
tions. The molecules fluctuate randomly in and out of a selected control volume. On
the other hand, if the volume is large on a macroscopic scale, then variations associated
with the spatial density distribution would be observed. As a result, there is a specific lim-
ited range to be defined as the appropriate volume size for the continuum assumption
to be valid. The control volume size must be larger than the scale of the mean free
path, but less than the characteristic macroscopic dimensions, to properly define the local
fluid density.
Different techniques may be used to describe the motion of fluids. In an Eulerian frame of
reference flow quantities are tracked from a fixed location in space (or a control volume),
whereas in the Lagrangian framework, individual fluid particles are tracked along their trajec-
tories. In general, the Eulerian approach will be adopted throughout this book. However, it
should be noted that in some applications, a Lagrangian description may be more useful,
such as free surface flows with tagged particles for the tracking of wave motion on a
free surface.
As an example, consider a gas particle trajectory in a heated duct. If a thermocouple is
placed in the duct, then the temperature varies according to the selected position, as well
as time. This represents a fixed location, corresponding to the Eulerian approach. On the
other hand, if individual gas particles are tracked throughout the duct, then this represents
a Lagrangian approach. In this latter approach, the temperature of a specific particle is a
function of time along its trajectory. The particle is tracked over a trajectory so its velocity
has a functional dependence on both the trajectory and time, or in other words, spatial coor-
dinates of the pathlines that also vary with time. It would be impractical to trace all particle
trajectories within the duct so the Eulerian approach would be more suitable in this example.
In the Eulerian approach, the change of temperature and velocity with time would be
4 Advanced Heat Transfer
observed with a stationary control volume in the duct. The approaches are different but
ultimately both Eulerian and Lagrangian descriptions lead to the same results.
From left to right, the individual terms represent: (i) the rate of energy accumulation with
time in the control volume; (ii) the rate of energy inflow across the boundary surfaces;
(iii) the rate of energy outflow; and (iv) the rate of internal heat generation within the control
volume due to processes such as electrical resistive heating or chemical reactions. The over-
dot notation refers to the rate of change with respect to time. A “steady state” refers to con-
ditions which are independent of time, that is, negligible changes in the problem variables
with time.
The energy balance states that the rate of the increase of energy over time within the con-
trol volume equals the net rate of energy inflow plus any internal heat generation. The
energy inflow and outflow terms include heat and work flows across the boundary surfaces.
For example, work or power input (or output) may occur due to a protruding shaft across
the boundary of the control volume of a pump. A turbine shaft and blades would extract
power from a control volume encompassing a steam turbine in a power plant. Although
the above form of the energy balance indicates a single outlet and inlet, a more generalized
expression can be written for multiple inlets and outlets by taking a summation over all
inlets and outlets in the above energy balance.
Control
volume (CV)
.
Ein Control
volume (CV)
. Ein
. . Eout
Eg Ecv
Ecv
Eg
Eout
FIGURE 1.1
Schematic of energy balance for a control volume.
Introduction 5
For a control mass, an energy balance can be written to include both work and heat
modes of energy transfer across the boundary surface of the control mass. The first law of
thermodynamics over a finite period of time including work performed on the system
(such as compression/expansion of a gas in the closed system), denoted by W, can be
written as:
Ei + Q + W = Ef (1.2)
where the subscripts i and f refer to initial and final states, respectively, and Q refers to the
net inflow of heat into the control mass (note: negative Q represents a heat outflow).
At the edges of a control volume, a boundary condition can be established through an
energy balance for a control volume that shrinks to a zero thickness as it encompasses the
boundary. In this case, the transient energy accumulation term in the energy balance
becomes zero since the mass of the control volume approaches zero. The heat generation
may be nonzero as a result of processes such as friction between two different phases at
the interface or heat transfer due to latent heat evolved at a moving phase interface. The
energy balance at the edge of the control volume can be regarded as a boundary condition
that is used to solve the governing equations for variables internally within the domain.
on a plate as a result of fluid impact on the plate. The impulse (or change of momentum) of
a specific fluid particle near the plate is the change in momentum between an initial
point upstream of the plate and its final state (a zero velocity upon impact at the wall).
Summing over many molecules near the plate and taking the average normal velocity of
all molecules, an expression can be obtained for the average force per unit area exerted
by the molecules on the wall. For a gas at normal atmospheric conditions, the following ideal
gas equation of state can be used to relate pressure to the density and temperature.
p = ρRT (1.3)
Pressure is a scalar variable that acts perpendicular to a surface and whose magnitude
adjusts to conserve mass in the flow field. For example, consider an air gap in a window
cell with a buoyant internal flow arising from differential heating of both sides of the cell.
Due to buoyancy forces, warm air ascends near the hot wall until it reaches the top corner.
Conservation of mass dictates that the fluid cannot only ascend, but there must also be a bal-
ance of a descending flow to conserve mass within the cell. Therefore, an adverse pressure
gradient (i.e., increasing pressure in the flow direction) occurs as the fluid ascends toward
the top corner, thereby causing the airflow to change directions and descend down along
the other side to allow an overall conservation of mass within the cavity.
Another key thermodynamic property is energy. The total energy refers to the sum of
internal, kinetic, and potential energies. The internal energy of a system is characterized
by its temperature. Work and heat are the forms that energy takes to cross the boundaries
of a system. A force alone does not change the energy of a system, but rather a force acting
over a distance, leads to work and an energy change. At a visible or macroscopic scale, work
is a process that changes the potential and/or kinetic energy of a system. In contrast, heat
transfer leads to a change of internal energy at a microscopic scale. In other words, heat
transfer corresponds to work at a microscopic or sub-visible scale.
Every system above a temperature of absolute zero (zero Kelvin, or −273.15◦ C) has a state
of microscopic disorder. Entropy represents an uncertainty about a system’s microscopic
state. It characterizes the disorder at the molecular level and a statistical probability or
uncertainty of a particular quantum state. In a perfect crystal of a pure substance at absolute
zero temperature, the molecules are motionless and stacked precisely in accordance with
their crystal structure. Here there is no uncertainty about the crystal’s microscopic state
(called the third law of thermodynamics). The entropy at zero absolute temperature is
zero. The second law of thermodynamics requires that the entropy of a system, including
its surroundings (an isolated system), never decreases. So the entropy production of an iso-
lated system is equal to zero for reversible processes, or greater than zero for irreversible
processes. A process is irreversible if it is highly unlikely from a statistical probability per-
spective that the direction of energy conversion can be reversed. Examples of irreversible
processes are dissipation of kinetic energy to frictional heating in a boundary layer and
heat transfer from a higher to lower temperature.
Although entropy, s, cannot be measured directly, it can be determined indirectly from the
Gibbs equation. For a simple compressible substance, the Gibbs equation is given by:
where e and υ refer to the internal energy and specific volume, respectively. Entropy can
also be expressed in terms of the specific Gibbs free energy, g. The Gibbs free energy is a
Introduction 7
thermodynamic property which represents the maximum reversible work that can be
extracted from a closed system in an isothermal (constant temperature) and isobaric
(constant pressure) process. It is defined by:
g = h − Ts (1.5)
where h refers to specific enthalpy. If a system undergoes a reversible process, the decrease
in Gibbs free energy from the initial to final state equals the work performed by the system
on the surroundings, minus the work of pressure forces. In multicomponent systems, the
Gibbs free energy is minimized when the system reaches chemical equilibrium at constant
temperature and pressure. Therefore, a decrease of Gibbs free energy is required for the
spontaneity of processes to proceed at constant pressure and temperature.
For a given system, an increase in microscopic disorder (or entropy) results in a loss of
ability to perform useful work. For example, consider the expansion of an ideal gas from
a half-cavity into an adjacent evacuated side of the other half-cavity. A partition initially
divides the two sides of the entire cavity. When the partition is removed, the total energy
of the total cavity remains constant, but after the process, there was a loss of ability to per-
form useful work. If the first side had a piston instead of a partition, the initial state could
have performed some work on the other side. But now the final state cannot perform
work since the gas has expanded into both sections. In the final state, there is less certainty
about a particle’s location because it has moved randomly within the larger entire volume,
rather than only within the half-cavity in the initial state. Thus, the system entropy has
increased and there was a loss of ability to perform useful work.
Otherwise, if the index is not repeated, then it implies multiple values for each index, such as
a vector of quantities (tensor of rank 1) or a matrix (tensor of rank 2). For example, the var-
iable aij is a tensor of rank 2 that represents a 3 × 3 matrix, where i = 1, 2, 3 and j = 1, 2, 3.
Using tensor notation, the total derivative Equation 1.6 can be written more concisely as:
DB ∂B ∂B
= + ui (1.7)
Dt ∂t ∂xi
450
0.7
Water Silver
0.5 400
Hydrogen Copper
Thermal conductivity, λ (Wm–1K–1)
Ethylene glycol
Helium
0.2 300
Engine oil
Aluminium
0.1
200
Air
Brass
0.05
Ammonia Saturated
vapor steam
100
CO2 Platinum
0.02 Iron
Type 304 stainless steel
0.015
–100 0 100 200 300 400 500 T(°C) 0
–200 0 200 300 400 800 1,000 T(°C)
FIGURE 1.2
Thermal conductivities of fluids and solids. (Adapted from G.F. Hewitt et al. 1997. International Encyclopedia of
Heat and Mass Transfer. Boca Raton: CRC Press/Taylor & Francis.)
Introduction 9
arrangement. For example, Mg+ ions are attracted to a negative electron cloud. These
loosely held electrons in an electron cloud lead to high thermal conductivities, relative to liq-
uids and gases, since the electrons move readily through the solid.
Ionic bonds are formed between metal and nonmetal ions. A metal gives up its valence
electrons to the outer shell of the nonmetal. The positive metal ions and negative non-
metal ions are attracted to each another. For example, in sodium chloride, Na (one
valence electron) reacts with Cl (seven valence electrons). A Cl stable outer shell of eight
valence electrons is formed. The electrons are tightly held, thereby yielding a lower ther-
mal conductivity than solids with metallic bonds. Also, mechanical properties of the sol-
ids are affected by the nature of the ionic bonds. The electric fields of opposing ions in
different planes repel each other, leading to more brittle characteristics and crystal
fractures.
In covalent bonds, electrons are shared by participating atoms and held tightly together.
In van der Waals bonds, secondary bonding between molecules is formed due to the charge
attraction resulting from an asymmetrical charge distribution in the material. For example,
water molecules are attracted to each other by negatively and positively charged sides
of adjacent molecules. The type and nature of the atomic bonds within a substance have a
direct impact on the magnitude of transport properties such as thermal conductivity, as
well as their variations with temperature.
Fluid viscosity is a measure of the frictional resistance of a fluid element when applying a
shear stress across adjacent layers of fluid that move parallel to each other at different
speeds. Consider a layer of liquid between a lower, fixed plate and an upper plate, moving
at a constant velocity, u. As the top plate moves, each layer of fluid will move faster than the
layer below it, due to friction that is resisting their relative motion. Since the fluid applies on
the top plate a force opposite to the direction of fluid motion, an external force is required to
keep the top plate moving at a constant velocity. The magnitude of the force, F, divided by
the plate area, A, is found to be proportional to the rate of shear deformation, or local shear
velocity, ∂u/∂y, according to Newton’s law of viscosity,
F ∂u
τ= =μ (1.8)
A ∂y
where the proportionality constant, μ, is the dynamic viscosity, τ is the shear stress, and y
refers to the direction perpendicular to the plate (alternatively, ν = μ/ρ is the kinematic
viscosity). The viscosity expresses the resistance of the fluid to shear flows. For common
fluids, such as oil and water, experimental measurements of the applied force confirm
that the shear stress is directly proportional to the strain rate in the liquid. Fluids are called
Newtonian fluids when the shear stress varies linearly with strain rate. For non-Newtonian
fluids, such as paint films, water–sand mixtures, or liquid polymers, the applied shear stress
and resulting fluid strain rates are related in a nonlinear manner.
In the evaluation of transport properties, macroscopic relationships are often used to
approximate the microscopic transport phenomena. For example, heat conduction
involves molecular fluctuations. Tracking of individual molecules, through their rota-
tional, translational, and vibrational modes of motion, and assessing their energy
exchange during these intermolecular motions, would be impractical. Therefore, a phe-
nomenological law is used to relate the heat transfer rate, q, to the temperature gradient
via a proportionality constant, k (thermal conductivity), similarly to Newton’s law of vis-
cosity. Phenomenological laws refer to approximations based on experimental
10 Advanced Heat Transfer
observations that improve theoretical models but which cannot be proven rigorously from
mathematical or first principles. The thermal conductivity is defined through a phenom-
enological law as a macroscopic approximation of microscopic intermolecular interactions
which lead to the flow of heat by conduction.
Examples of other thermophysical properties are the specific heat, surface tension, coeffi-
cient of thermal expansion and saturation pressure. Further properties are usually required
when additional physical processes, such as a phase change, occur in a system. The specific
heat at constant volume, cv, and specific heat at constant pressure, cp, for simple compress-
ible substances can be represented by partial derivatives of the internal energy, e(T, v), and
enthalpy, h(T, p), respectively:
∂e
cv = (1.9)
∂T υ
∂h
cp = (1.10)
∂T p
where the subscripts υ and p refer to the variables, specific volume and pressure, held
fixed during differentiation. The specific heats represent the amount of required thermal
energy added by heat transfer to increase the temperature of the substance by 1 degree
Celsius.
Surface tension, σ, refers to the tension of a surface film or droplet of liquid caused by
the attraction of molecules in the surface layer with an adjoining fluid, tending to minimize
surface area. The coefficient of thermal expansion, β, describes the fractional change in the
size of a substance per degree of change of temperature at constant pressure. When a sub-
stance is heated, the molecular kinetic energy, vibrations, and movements increase, thereby
creating a larger intermolecular separation and expansion of the fluid. Another property is
the vapor saturation pressure, psat, which refers to the pressure at which liquid–vapor phase
change occurs in a pure substance. The vapor pressure increases with saturation tempera-
ture since more thermal energy is required to break molecular bonds in the phase transition
at higher temperatures.
T1
qx
T2
x T1 > T2 qx T2
FIGURE 1.3
Schematic of heat conduction by molecular motion.
of thermal energy from the hotter side to the colder side, without any net transport of mass
across the dividing plane.
As discussed earlier, it is impractical and difficult to track the energy exchange among
individual molecules during all of the intermolecular interactions. Therefore, the conduction
heat flow is typically approximated by macroscopic averaged quantities, such as tem-
perature and thermal conductivity, k. In a one-dimensional layer in Figure 1.3, the conduc-
tion heat transfer rate is governed by Fourier’s law, named after Jean-Baptiste Fourier
(1768–1830):
qcond ∂T ΔT
q′′cond = = −k ≈ −k (1.11)
A ∂x Δx
where A, q′′cond and qcond refer to the heat flow area, conduction heat flux (in units of W/m2)
and heat flow rate (in units of W), respectively. For steady-state one-dimensional problems,
the temperature gradient in Equation 1.11 can be approximated by the temperature dif-
ference divided by the thickness of the one-dimensional layer, Δx. In multidimensional
problems, the partial derivative in Equation 1.11 becomes a gradient of temperature and
q becomes a vector of heat flow components in the x, y, and z directions. Further detailed
analysis of heat conduction will be presented in Chapter 2.
The negative sign in Fourier’s law indicates that a negative temperature gradient (i.e.,
decreasing temperature in the positive x-direction) multiplied by a minus sign yields a pos-
itive heat flow in the positive x-direction. This sign convention is illustrated in Figure 1.4. It
can be observed that heat flows in the direction of decreasing temperature (as expected).
From a one-dimensional energy balance under steady-state conditions, the rate of heat
flow into the control volume balances the rate of outflow. Then based on Fourier’s law
for a homogenous material of constant thermal conductivity, the slope of the temperature
profile at the inlet boundary matches the temperature slope (or gradient) at the outlet sur-
face. The energy balance requires that both of these temperature gradients are constant
and equal to one another, assuming zero sources/sinks of energy within the control volume.
Therefore, the temperature profile must decrease linearly within the control volume to
match the temperature slopes at both the inlet and outlet. This confirms the prior
12 Advanced Heat Transfer
T T(x) T T(x)
+ΔT –ΔT
dT/dx dT/dx
is (+) is (–)
+Δx +Δx
x x
FIGURE 1.4
Sign convention for conduction heat flow.
1.5 Convection
Convective heat transfer refers to the combination of molecular diffusion (conduction) and
bulk fluid motion (or advection). Newton’s law of cooling states that the rate of convective heat
flow from an object is proportional to the surface area and difference between its tempera-
ture and the surrounding ambient temperature. The proportionality constant is called the
convective heat transfer coefficient, h. The name of this law is attributed to Sir Isaac Newton
(1642–1727), largely due to the associated Newton’s laws of motion in governing the motion of
the fluid which drives the process of convective heat transfer.
Consider a hot surface at a constant temperature, Ts, placed into contact with a uniform
fluid stream flowing at a temperature of Tf above the surface (see Figure 1.5). Then from
Newton’s law of cooling, the convective heat flux to the fluid stream from the surface of
area A, is given by:
qconv
q′′conv = = hA(Ts − Tf ) (1.12)
A
where the convective heat transfer coefficient is denoted by h. Also, q′′conv and qconv refer to the
convective heat flux (in units of W/m2) and convective heat flow rate (in units of W), respec-
tively. This coefficient depends on many complex factors, including the fluid properties,
Introduction 13
q = h A (Ts – Tf)
Convective heat
transfer rate (W/m2)
Area, A
h, Tf
Surface
temperature, Ts
FIGURE 1.5
Convective heat transfer from a heated surface.
geometrical configuration, fluid velocity, surface roughness, and other factors to be dis-
cussed in forthcoming chapters of this book.
Heat transfer near the surface occurs through a boundary layer. The boundary layer is a thin
layer of fluid close to the solid surface of the wall in contact with the moving fluid stream.
The fluid velocity varies from zero at the wall (called a no-slip condition) up to the freestream
velocity at the edge of the boundary layer, which approximately within 1% corresponds to
the freestream velocity. The boundary layer thickness is the distance from the surface up to
the point at which the velocity is 99% of the freestream velocity. To determine how the veloc-
ity and temperature fields change through a boundary layer, and more generally the entire
flow field, the governing differential equations of mass, momentum, and energy conserva-
tion must be solved. These three-dimensional equations in Cartesian, cylindrical, and spher-
ical coordinates are shown in Appendix C and will be discussed further in Chapter 3.
Convection can be induced by external means (e.g., fan, pump, atmospheric winds), called
forced convection, or else buoyancy induced motion in free convection (or natural convection).
Mixed convection includes both forced and free convection. The magnitude of the convec-
tive heat transfer coefficient usually indicates the type of problem. Examples of typical
values of convective heat transfer coefficients are listed below:
Thermophysical properties, phase change, and turbulence are common factors leading to
higher convective heat transfer coefficients. Thermodynamic and transport property tables
which are commonly used for convection problems involving various solids, liquids, and
gases are presented in Appendices D–F, respectively.
In convection problems, the fluid temperature in Newton’s law of cooling must be care-
fully specified. The selection of an appropriate temperature depends on the type of problem
under consideration. In the case of external flow, such as external flow over a circular cyl-
inder, the fluid temperature is given by the freestream temperature. However, for internal
14 Advanced Heat Transfer
flow, such as liquid flow in a pipe, the fluid temperature in Newton’s law of cooling becomes
the mean temperature of the fluid, since a freestream ambient temperature is not applicable.
The mean temperature is obtained by spatial averaging of the velocity multiplied by temper-
ature over the cross-sectional area of the pipe. It represents the mass flow weighted average
temperature of the fluid in the pipe at a particular axial location within the pipe.
The convective heat transfer coefficient varies with position along a surface. The local con-
vection coefficient is the value at a specific point on the surface, whereas the average or total
heat transfer coefficient refers to an integrated value across the surface. For example, con-
sider external flow over the top surface of the object in Figure 1.5. Define an x-coordinate
as the position from the leading edge of the surface. As the gas flows over the surface, a
thin boundary layer forms and grows. This affects the convective heat transfer process
and leads to a variation of the “local” convection coefficient with position, x. Averaging
of the local coefficient by integrating h(x) over the surface area yields an average convection
coefficient, which is more commonly used in convective heat transfer analysis. In Chapter 3,
further detailed analysis of convective heat transfer will be presented.
Surroundings
at Tsur
Emitted Gas
radiation, E T∞, h
Incident
radiation, G
Convection
heat flow, q
Surface of emissivity ε,
area A, and temperature Ts
FIGURE 1.6
Radiative heat transfer between and a surface and its surroundings.
Introduction 15
In heat conduction problems, the main governing equation was Fourier’s law, whereas
Newton’s law of cooling was adopted for convective heat transfer problems. In radiative
heat transfer, a primary governing equation is Stefan–Boltzmann’s law. For a surface at a tem-
perature of Ts (in absolute units, i.e., Kelvin or Rankine units) and a surface emissivity of ϵ,
the rate of radiative heat transfer from the surface is given by:
qrad
q′′rad = = εσTs4 (1.13)
A
where A is the surface area and σ = 5.67 × 10−8 W/m2 K4 is the Stefan–Boltzmann constant,
named after Jozef Stefan (1835–1893) and Ludwig Boltzmann (1844–1906). The radiative
heat flux (in units of W/m2) and radiation heat flow rate (in units of W) are represented
by q′′rad and qrad, respectively. The surface emissivity, ϵ, represents the surface’s ability to
emit radiation in comparison to an ideal emitter (i.e., ϵ = 1 for a blackbody). Radiative prop-
erties of various surfaces and gases are presented in Appendix G.
Radiation heat transfer between two surfaces or objects, at temperatures T1 and T2, respec-
tively, can be obtained based on the net radiation exchange between the surfaces,
q′′rad = εσ T14 − T24 (1.14)
where the effective radiation heat transfer coefficient, hr, is given by:
hr = εσ(T1 + T2 ) T12 + T22 (1.16)
In this way, radiation and convection coefficients can be combined together in a thermal
analysis involving both modes of heat transfer.
In addition to radiative heat transfer between surfaces in view of each other, Stefan–
Boltzmann’s law can also be used to predict radiation exchange among objects or groups
of surfaces. For example, consider the net radiation exchange between a small object at a
temperature of T1 and a surrounding cavity at T2, where T1 , T2. Since all of the radiation
emitted from the small object (object 1) is absorbed by the surrounding cavity (object 2),
the surroundings behave like a blackbody at T2. Then the net radiative heat gain by the small
object is given by:
qrad = εσA T24 − T14 (1.17)
where A refers to the surface area of object 1 and ϵ denotes the emissivity of the object.
The thermal physics of radiation is based on the random movements of atoms and mol-
ecules, composed of charged particles (protons and electrons), whose movement leads
to the emission of electromagnetic waves that carry energy away from the surface. In
Chapter 4, more detailed analysis, physical processes, governing equations, and methods
of analysis of radiation heat transfer will be presented and discussed.
16 Advanced Heat Transfer
l
Q Saturated Saturated
liquid vapor Q
g
(3) (5)
Q s+l (2)
Temperature, T
l (4) Gas
s+l
l+g
l+g Q
(1)
Solid (s) + gas (g)
Q s
Solid
(s)
Specific volume, υ
FIGURE 1.7
Phase diagram for solid–liquid–gas pure material.
Introduction 17
the saturated vapor point is reached, beyond which further heating is transferred as sensible
heating and temperature increase of the gas (region 5). If these steps are conducted at several
different pressures, then the resulting measurements and state points from each experiment
(i.e., temperature, pressure, specific volume, etc.) can be joined together. These resulting
data points would provide the entire phase diagram for phase change of a pure material
from a solid to a gas.
In region 4 of Figure 1.7, the boiling process begins when individual bubbles form and
grow along the heated surface. The thin liquid layer beneath a growing bubble is vaporized
after which the bubble detaches from the wall and ascends by buoyancy through the liquid.
During this process, heat is transferred by conduction from the heated surface to the liquid
layer in contact with the surface. The heat transfer process at the wall may be represented by
a combination of liquid and vapor periods in contact with the wall, which are characterized
by the frequency and sizes of detaching bubbles. The average heat flux over both periods
depends on the frequency of bubble departure. The changing liquid layer thickness beneath
a bubble is related to the heat conduction process in the liquid and latent heat absorbed
by the adjacent vapor phase. Further detailed analysis of boiling and condensation pro-
cesses will be presented in Chapter 5.
Melting occurs in region 2 of Figure 1.7. The importance of solidification and melting in our
everyday lives is evident in various ways. For example, consider water and its
density difference between the solid and liquid phases. Unlike many materials, its density
is higher in the liquid phase. This property arises from a unique angular arrangement of
hydrogen and oxygen atoms in a water molecule. As a result, in the natural environment,
ice freezes and floats on the top of oceans and lakes. Otherwise, if solidification of
water occurred with a higher density in the solid phase (like most metals), freezing ice would
descend. This could cause oceans, lakes, and rivers to potentially completely freeze, so that
life on Earth may not exist. So this relatively basic anomaly of the freezing process illustrates
the importance of solidification and melting in our everyday experiences. In Chapter 7,
further detailed investigation of solid-liquid phase change processes will be presented.
Recall that intermolecular energy exchange by heat conduction from a hot to cold side of a
layer was illustrated in Figure 1.3. Consider instead an analogous process, mass diffusion of
interstitial atoms (component B) through a solvent (component A), across the imaginary
plane in Figure 1.3 over a distance of Δx. In this case, Figure 1.3 illustrates a layer of fluid
with molecular motion and interstitial atoms that migrate across the imaginary plane. Let
ГB, N, N1, and N2 refer to the average number of interstitial atom jumps per second in ran-
dom directions, number of adjacent sites (i.e., n = 6 in three dimensions), number of B atoms
per unit area in plane 1 (immediately to the left of the imaginary plane), and the number of B
atoms per unit area in plane 2 (immediately to the right of the imaginary plane), respec-
tively. Then the net flux of interstitial atoms across the imaginary control surface can be writ-
ten as:
jB 1 1
j′′B = = j′′B,12 − j′′B,21 = ΓB N1 − ΓB N2 (1.18)
A N N
where the double prime notation (′′ ) represents a flux quantity (per unit area). A positive
expression for jB indicates that there is a net species mass flux from the left side to the
right side of the imaginary plane.
The concentration of B atoms at planes 1 and 2, CB,1 and CB,2, respectively, is:
N1 N2
CB,1 = ; CB,2 = (1.19)
Δx Δx
∂CB
CB,2 = CB,1 + Δx + ··· (1.20)
∂x
Combining these equations, neglecting higher order terms, and multiplying by the surface
area, A, yields the following total mass diffusion flow of interstitial atoms across the imag-
inary plane,
ΓB 2 ∂CB
jB = − Δx A (1.21)
6 ∂x
where the coefficient in parentheses is the mass diffusivity, or mass diffusion coefficient
(denoted by DAB) of constituents A and B. This equation can be generalized to the following
one-dimensional form of Fick’s law, named after Adolf Fick (1829–1901):
jB dC
j′′B = = −DAB (1.22)
A dx
This result is analogous to Fourier’s law of heat conduction earlier in this chapter.
Fick’s law describes the diffusion of mass down a species concentration gradient in a
manner analogous to Fourier’s law, which prescribes the diffusion of heat down a temper-
ature gradient. More detailed analysis of mass transfer processes is presented in a classic
book by Bird, Stewart, and Lightfoot (2007) and will be described in forthcoming chapters,
particularly for chemically reacting flows in Chapter 8.
Introduction 19
PROBLEMS
1.1 Consider touching and holding two different materials, such as a piece of steel and a
piece of wood, which were both initially at the same subzero temperature. Would
either material “feel” colder than the other? Explain your response.
1.2 The heat loss through a common brick wall in a building is 1,800 W. The height,
width, and thickness of the wall are 3, 8, and 0.22 m, respectively. If the inside
temperature of the wall is 22◦ C, find the outside surface temperature of the wall.
1.3 Heat flows in a one-dimensional direction through a layer of material of unknown
thermal conductivity. Explain a method to estimate the material’s conductivity based
on temperature measurements recorded by thermocouples at various positions
within the layer of material.
1.4 The heat loss through a brick wall with an outside temperature of 16◦ C is 800 W. The
height, width, and thickness of the wall are 4, 10, and 0.3 m, respectively. What is the
inside surface temperature of the wall?
1.5 A steel plate in a manufacturing plant is removed from a furnace at a temperature of
650◦ C. Heat loss from the plate to the surrounding air at 20◦ C occurs at a rate of 36
kW/m2. Find the average plate temperature at a location where the convection coef-
ficient is 30% higher than the initial value where the plate was initially removed from
the furnace.
1.6 A heat loss of 1,600 W is experienced through a glass window (1.4 m width × 2.6 m
height) at the outer wall of a building. The convection coefficient is 30 W/m2 K.
Estimate the window temperature when the outside air temperature reaches -10◦ C.
1.7 An aluminum sheet leaves a hot-roll section (called the entry point) of a mill at a tem-
perature of 620◦ C. The surrounding air temperature is 20◦ C. For a desired uniform
cooling rate of 38 kW/m2 across the sheet, estimate the aluminum surface tempera-
ture at a position where the convection coefficient is 40% higher than the value at
the entry point.
1.8 Air flows across a tube surface (emissivity of ϵ = 0.9) in a heat exchanger. The ambient
air temperature and convection coefficient are 20◦ C and 120 W/m2 K, respectively.
Compare the heat transfer by convection and radiation at surface temperatures of
(a) 60◦ C, (b) 400◦ C, and (c) 1,300◦ C.
1.9 A metal block (0.8 m × 1.2 m × 0.4 m) in a furnace is heated by radiation exchange
with the walls at 1,100◦ C. The block temperature is 420◦ C. The walls are approxi-
mated as blackbodies. Find the net rate of heat transfer to the block due to radiation
exchange with the walls. How would the result be altered if the block was placed in a
corner of the furnace?
1.10 Superheated steam at 240◦ C flows through an uninsulated pipe in a basement hall-
way at 20◦ C. The pipe emissivity is 0.7 and the coefficient of free convection is 20
W/m2 K. Find an expression for the total heat loss from the pipe (per unit length of
pipe) in terms of the pipe diameter.
1.11 Water is vaporized in a distillation unit and superheated steam exits from the heating
tank through a single tube. Once cooled, this steam is condensed and collected in a
second storage tank of purified water. If the electrical element provides a net heat
input of 400 W to the water, how many liters of distilled water are produced after
20 Advanced Heat Transfer
6 h of operation? Assume that external heat losses from the heating tank are negligible
and that heat input is directed entirely into phase change (i.e., 6 h refers to the time
taken once the water reaches the saturation temperature).
1.12 An ice-making machine uses a refrigeration system to freeze water with a net heat
removal rate of 11 kW. How much time is required to produce 1,200 kg of ice? It
may be assumed that water is cooled to slightly above 0◦ C before it enters the
refrigeration unit.
1.13 An industrial heat exchanger is used to vaporize water by electrical heating of 3,900
W of heat input to the storage tank of saturated water. How much vapor is generated
after 5 h of operation, assuming that the storage tank is well insulated, and all heat
input is directed to the vaporization process?
1.14 The diameter of the sun is approximately 1.39 × 109 m. A heat flux of about 1,353
W/m2 from the sun’s radiant heat reaches the outer atmosphere of the Earth. Estimate
the radius of the orbit of Earth’s trajectory around the sun (i.e., Earth’s distance from
the sun). The sun can be approximated as a blackbody at 5,800 K.
1.15 The wall of a walk-in freezer in a meat processing plant contains insulation with a
thickness of 10 cm and a thermal conductivity of 0.02 W/mK. The inside wall temper-
ature is -10◦ C. The outside air temperature and convection coefficient (outside) are
20◦ C and 5 W/m2 K, respectively. What additional thickness of insulation is required
to reduce the heat gain (into the freezer) by 10%?
1.16 The temperature of the indoor side of a stone concrete wall in a building is 20◦ C. The
wall thickness is 20 cm. The convection coefficient and ambient air temperature for
the outdoor air are 50 W/m2 K and -5◦ C, respectively. Determine the outer wall tem-
perature of the concrete wall.
1.17 A thin plate is exposed to an incident radiation flux of 2 kW/m2 on its top surface. The
bottom surface is well insulated. The exposed surface absorbs 80% of the incident
radiation and exchanges heat by convection and radiation to the ambient air at 290
K. If the plate’s emissivity is 0.8 and the wall temperature is 350 K, estimate the
convection coefficient, h, under steady-state conditions.
1.18 The gas temperature in an autoclave is measured with a thermocouple (a wire of
1-mm outer diameter). Composite material components for an aircraft are cured in
the autoclave. The junction of the thermocouple (ϵ = 0.6) is located at the end of
the thermocouple wire, which protrudes into the autoclave and exchanges heat
by radiation with the walls at 410◦ C. The freestream temperature and convection
coefficient of gas flow past the thermocouple are 180◦ C and 30 W/m2 K, respectively.
What error, arising from the difference between the thermocouple reading and gas
temperature, is expected from the measurement of the gas temperature? How can
this temperature measurement error be reduced? Conduction losses through the ther-
mocouple wire can be neglected.
1.19 Resistive heat generation occurs within an overhead power transmission cable. Air
flows past the cable with a freestream temperature and convection coefficient of
-6◦ C and 80 W/m2 K, respectively. For a specified copper cable (3-cm diameter), esti-
mate the required heat generation rate (per unit length of cable) to maintain the sur-
face temperature above 0◦ C.
1.20 The top cover of a solar collector plate is exposed to ambient air at 20◦ C with a con-
vection coefficient of 6 W/m2 K. A surface coating is developed and applied to
Introduction 21
modify the radiative properties of this absorbing cover plate. The incident solar radi-
ation is 860 W/m2 and the outer surface temperature of the absorber plate is 70◦ C.
a. What plate emissivity is required to provide a conduction heat flux of 500
W/m2 through the absorber plate?
b. What is the proportion of heat exchange by radiation between the cover plate
and surroundings (ambient air) relative to the convective heat loss from the
plate?
1.21 Heating, ventilating, and air conditioning (HVAC) equipment maintains a room in a
building at desired temperature conditions. Identify the heat transfer processes and
appropriate energy balance(s) to find the steady-state temperature of the room in
the building.
1.22 Aircraft icing may occur when an aircraft passes through a cloud containing super-
cooled droplets. Describe the heat flows to/from an iced aircraft surface and write
an energy balance for the surface based on these heat flows.
1.23 Particles of pulverized coal are injected and burned in a boiler of a thermal power
plant. Identify and briefly describe the relevant heat transfer modes that contribute
to an energy balance for a pulverized coal particle.
1.24 Identify the relevant heat transfer processes that arise in a thermal balance of the
human body. Explain how these processes are combined in an overall energy balance
to find the total heat losses from the body.
1.25 Various technical challenges are encountered by thermal engineers in the design of
more efficient systems for heating and cooling of buildings. Perform a technical liter-
ature review (including journals and magazines) to describe the main technological
advances and challenges facing thermal engineers in the development of next-gener-
ation HVAC systems.
References
R.B. Bird, W.E. Stewart and E.N. Lightfoot. 2007. Transport Phenomena, 2nd Edition, New York:
John Wiley & Sons.
R.C. Weast, Ed., 1970. CRC Handbook of Tables for Applied Engineering Science, Boca Raton: CRC Press/
Taylor & Francis.
2
Heat Conduction
2.1 Introduction
In this chapter, the governing equations and methods of analysis of conduction heat transfer
will be presented. One-dimensional and two-dimensional problems will be analyzed, as
well as steady-state and transient problems. Fundamental transport processes will be inves-
tigated. Also, advanced solution methods such as separation of variables, conformal map-
ping, and solutions in orthogonal curvilinear coordinates, will be outlined.
As discussed in the prior chapter, conduction heat transfer occurs at a molecular level
when thermal energy is exchanged among intermolecular interactions by rotational,
vibrational, and translational energies at a molecular scale. The temperature of the sub-
stance characterizes the thermal energy of this atomic motion. When collisions occur
among molecules, energy is passed from more energetic to less energetic neighboring
atoms, thereby allowing thermal energy to travel through the substance. Since the atomic
structure is more closely packed in solids, the atoms are closer together and unable to
move compared to liquids and gases, where atoms can more easily move past each other.
As a result, there are fewer collisions and a lower resulting thermal conductivity in liquids
and gases.
As introduced in the previous chapter, heat conduction is governed by Fourier’s law, which
relates the magnitude and direction of the heat flux to the temperature gradient. A negative
sign is placed in front of the temperature gradient to indicate that heat flows in the direction
of decreasing temperature. Or alternatively, a positive temperature gradient suggests that
temperature is increasing, and heat must then flow in the opposite (negative) direction.
The proportionality constant between the heat flux and temperature gradient is called the
thermal conductivity. This chapter will examine various formulations and solutions of Four-
ier’s Law and the heat equation for a range of coordinate systems.
The flow of heat by conduction depends on three key factors: the temperature gradient,
cross-sectional area of heat flow, and thermal conductivity. The temperature gradient pre-
scribes the rate and direction of temperature change at a particular location. The transfer of
heat continues until there is no longer any temperature difference within the domain at
which point a state of thermal equilibrium is reached. The second factor, cross-sectional
area, directly affects the amount of heat which can flow across the surface. Also, the prop-
erties of the material affect the heat flow, particularly the thermal conductivity. As discussed
in the previous chapter, solids are better conductors than liquids and gases, while materials
such as metals are better conductors than solids such as wood, paper, or cloth. Materials that
are poor heat conductors are called insulators. Air is an effective insulator because it can be
readily contained within any space so it is often used as an insulator in enclosed spaces, such
as double-pane glass windows.
23
24 Advanced Heat Transfer
From left to right, the terms represent the transient accumulation of energy within the
control volume, energy inflow and outflow from the control volume, and internal energy
generation. A sample control volume of the energy balance is illustrated in Figure 2.1.
The transient energy change, Ė cv, may be written in terms of a temperature change using
the definition of the specific heat (Chapter 1), multiplied by the volume size, Adx. The energy
inflow, Ė in, can be expressed by the conduction heat flux, multiplied by the surface area. The
energy outflow can be written similarly, along with a Taylor series expansion about the
influx value. Also, the total internal energy generation rate is expressed by the volumetric
· Insulated
Eg boundary
· ·
Ein = qx″ A Eout = qx″ A + (d(qx″A)/dx)dx
+ ...
·
Area, A Ecv
FIGURE 2.1
Schematic of one-dimensional control volume energy balance.
Heat Conduction 25
generation rate multiplied by the size of the volume. Substituting these individual terms into
the energy balance,
∂T ∂ ′′
ρcp Adx = qx A − qx A + (qx A)dx + · · · + q̇′′′ Adx
′′ ′′
(2.2)
∂t ∂x
∂T
q′′x = −k (2.3)
∂x
Substituting this expression into the energy balance, cancelling terms, and neglecting
higher order terms,
∂T ∂ ∂T
ρcp Adx = k A dx + q̇′′′ Adx (2.4)
∂t ∂x ∂x
Then, assuming a constant thermal conductivity and cross-sectional area, and dividing by
Adx, yields the following form of the one-dimensional heat conduction equation:
∂T ∂2 T
ρcp = k 2 + q̇′′′ (2.5)
∂t ∂x
In cylindrical coordinates, Fourier’s law and the heat conduction equation have a similar
form:
∂T
q′′r = −k (2.6)
∂r
∂T k ∂ ∂T
ρcp = r + q̇′′′ (2.7)
∂t r ∂r ∂r
A cylindrical volume size and surface area are used in the energy balance to obtain the
cylindrical form of the heat conduction equation.
The thermal diffusivity, α, may be interpreted in terms of a characteristic time, τc, required
to change the temperature of an object as a result of heating or cooling. For example, consider
dropping a small sphere of diameter D from the air into a container of water at a higher tem-
perature. Heat is transferred to the sphere over a characteristic time of D 2/α. At early periods
of time (t ≪ τc), much less than the characteristic time, the temperature of the sphere remains
approximately the same as its initial air temperature. But after a long period of time, after the
sphere has settled in the water (t ≫ τc), the sphere temperature approximately reaches the
water temperature. During the intermediate times (approximately of the order of τc),
the sphere temperature lies between the air and water temperatures. Thus, a characteristic
time of conduction can be represented by an appropriate length scale, squared, divided
by the thermal diffusivity of the material. This characteristic time will be shorter for better
conductors since the conduction process occurs more rapidly than poor conductors.
In addition to the above governing heat equation, a problem definition also requires
boundary conditions for completion of the problem specification. For heat conduction
26 Advanced Heat Transfer
problems, there are three main types of boundary conditions: Dirichlet, Neumann, and
Robin boundary conditions (see Figure 2.2). For a Dirichlet condition, a surface temperature
is specified at the boundary of the domain. For example, a Dirichlet boundary condition at
a wall (x = 0) can be represented by T(0, t) = Tw where Tw is the known or specified
wall temperature.
The second type of boundary condition is a heat flux or temperature gradient specified
condition (called a Neumann condition). For example, at a solid boundary (x = 0), a constant
specified wall heat flux balances the rate of heat flow into the solid. Applying Fourier’s law,
the Neumann condition becomes,
∂T q′′w
= − (2.8)
∂x 0 k
where q′′′
w refers to the known or specified wall heat flux. If the temperature gradient at the
boundary is zero, this represents an adiabatic (or insulated) boundary condition, as there is
no heat flow across the boundary. An insulated boundary is a special case of a Neumann
boundary condition.
The third type of boundary condition is a convection condition (called a Robin condition).
This type is a combination of the previous two types, where both the temperature and
the temperature gradient appear in the boundary condition. It is a commonly encountered
condition at a convective boundary. At the boundary between a solid and a fluid,
heat conduction within the solid (characterized by the temperature gradient in the
solid) balances the convective heat flow which is represented by Newton’s law of
cooling. For example, at a solid boundary (x = 0), a heat balance between the conduc-
tion heat flux in the solid from Fourier’s law and the convective heat flux in the fluid
yields,
∂T
−k = h(T(0, t) − Tf ) (2.9)
∂x 0
x x x
Tf , h
qw″ qx″
Tw
T(x,t) T(x,t)
T(x,t)
dT/dx|0
Fluid Solid
FIGURE 2.2
Types of boundary conditions.
Heat Conduction 27
V1 − V2
I= (2.10)
Re
where I, V1–V2, and Re refer to the electrical current, potential (voltage) difference, and elec-
trical resistance, respectively. From Ohm’s law, the current flow is proportional to the driv-
ing potential (a difference between the high and low applied voltages) and inversely
proportional to the resistance through the electrically conducting medium.
The thermal analogy of Ohm’s law may be written as follows:
T1 − T2
q= (2.11)
Rt
In this case, the analogue of the electrical current is the heat transfer rate, q. Rather than a
voltage difference, a temperature difference, T1–T2, represents the driving potential for the
heat flow. Also, the value Rt refers to the thermal resistance, instead of electrical resistance.
Then, this above thermal analogy and associated concept of a resistance network can be
used to analyze steady-state heat transfer problems. Like electrical circuits, a thermal circuit
can be formulated with resistances combined together in series or parallel to analyze each
stage of heat transfer. Using this approach, thermal circuit diagrams involving several
heat flows may be constructed to analyze heat transfer problems in a similar way to how
electrical circuits are used in electrical problems.
For one-dimensional heat conduction in Cartesian coordinates, Fourier’s law and Equation
2.11 imply that the thermal resistance of heat conduction through a plane layer of thickness
Δx can be written as:
Δx
Rt,cond = (2.12)
k
For convective heat transfer, the thermal resistance of a boundary layer is obtained from
Newton’s law of cooling. In this case, the convective heat transfer rate can be written as a
temperature difference (between the surface and fluid) divided by the convection thermal
resistance, Rt,conv, where,
1
Rt,conv = (2.13)
hconv A
28 Advanced Heat Transfer
Similarly, for radiation problems where the heat transfer coefficient is linearized
(as described previously in Chapter 1), the radiative thermal resistance becomes,
1
Rt,rad = (2.14)
hrad A
Similar expressions can be obtained for other geometrical configurations and combined
systems with multiple resistances in series and/or parallel. The following example demon-
strates how the thermal resistance can be calculated for a cylindrical geometry.
where r refers to the radial position. Integrating this equation twice yields the following
general solution,
dT T 1 − T2 T1 − T2
qr = −kA = = (2.18)
dr ln(r2 /r1 )/(2πLk) Rt,cond
T2
T1
r1 r1 r2
L T1 T2
In (r2/r1)
r2 Rt =
2πLk
FIGURE 2.3
Schematic of cylindrical heat transfer.
Heat Conduction 29
where L is the tube length. Adding this conduction resistance and the convection resis-
tance yields the following total thermal resistance per unit length of tube,
ln(r/r1 ) 1
R′t,tot = + (2.19)
2πk 2πrh
dR′ t,tot 1 1
= − =0 (2.20)
dr 2πkr 2πr2 h
Solving this equation yields a radius of r = k/h for which the total thermal resistance is a
minimum. The result may be also interpreted as a critical insulation radius, below which q’
increases with increasing radius, and above which q’ decreases with increasing radius.
This previous example presented a method to find the total thermal resistance in a
cylindrical geometry. The approach solved the heat equation and derived the temperature
distribution and then the heat flux by Fourier’s law. Then, this result was rewritten as a tem-
perature difference divided by the heat flux, which is defined as the thermal resistance.
Using this procedure, thermal resistances for various other geometrical configurations
can be readily obtained.
More complex systems such as composite walls can be represented by equivalent thermal
circuits. Composite walls of multiple types of thermal resistances and materials can
be assembled as a number of thermal resistances in series and/or parallel. Generalizing
the heat transfer rate and thermal resistance in Equation 2.11 to include multiple thermal
resistances,
ΔT 1
q= ; U= (2.21)
Rt Rtot A
where ΔT, Rtot and U refer to the overall temperature difference across the entire thermal
circuit, total thermal resistance (summation of all thermal resistances), and overall heat transfer
coefficient, defined by an expression analogous to Newton’s law of cooling,
q = UAΔT (2.22)
Consider heat flow through a composite wall consisting of multiple sections in both series
and parallel as illustrated in Figure 2.4. The total resistance to heat flow through the
q (2)
T1 (3)
T2 In parallel
FIGURE 2.4
Thermal circuit for a series-parallel composite wall.
30 Advanced Heat Transfer
composite wall, Rtot, can be expressed by a combination of resistances in series and parallel
as follows,
−1
1 1
Rtot = R1 + + + R4 (2.23)
R2 R3
This thermal circuit approach to approximate the heat flow is quasi one-dimensional in
the sense that isotherms are approximately perpendicular to the wall at the interface
between sections of the composite wall in parallel (sections 2 and 3). The boundary
between sections 2 and 3 is modelled as an adiabatic interface in this approach since
heat is assumed to flow only in the x-direction, leading to a temperature variation in the
x-direction only.
The method of parallel adiabats and parallel isotherms can be used to estimate the upper
and lower bounds of the overall thermal resistance of a composite material. In Figure 2.4,
if the heat flow was directionally unrestrained (a uniform conductivity throughout the
element), such that no cross-diffusion occurred against the principal heat flow direction,
then the thermal circuit would represent a series of one-dimensional resistances. In this
case, a series of adiabats parallel to the adiabatic boundaries would approximate the heat
flow lines and yield the maximum thermal resistance (an upper bound) of the composite
material. On the other hand, if the resistances between different material sections are
aligned in parallel, then this circuit yields the minimum thermal resistance. Thus, a parallel
isotherm construction leads to a lower bound on the material’s thermal resistance. Since
the upper and lower bounds on the thermal resistance, RUB and RLB, represent asymp-
totes, the actual thermal resistance lies between these two limit cases. An arithmetic
mean, R = (RUB + RLB)/2, is an approximation of the actual thermal resistance. If the upper
and lower bounds are far apart, then a geometric mean is a more accurate approximation of
the actual resistance.
The conduction shape factor can be obtained from the reciprocal of the thermal resistance
multiplied by the thermal conductivity. The shape factor and resulting heat flow rate are
expressed by:
1
S= (2.24)
Rt k
q = Sk(T1 − T2 ) (2.25)
TABLE 2.1
Conduction Shape Factors for Various Geometrical Configurations
Geometrical Configuration Shape Factor
A
(a) Plane wall with a width of L and cross-sectional S¼
L
area of A
2πL
(b) Cylindrical shell with inner and outer radii of r1 S¼
lnðr2 =r1 Þ
and r2, respectively
4π
(c) Spherical shell with inner and outer radii of r1 and S¼
1=r1 1=r2
r2, respectively
2πL
(d) Isothermal horizontal cylinder of length L buried S¼
with its axis a distance z below the surface (z=L ≪ 1 cosh1 (2z=D)
and D=L ≪ 1)
2πD
(e) Isothermal sphere of diameter D buried a distance S¼
1 D=(4z)
z below an isothermal surface (z . D=2)
2πD
(f) Isothermal sphere of diameter D buried a distance S¼
1 þ D=(4z)
z below an insulated surface (z . D=2)
2πz
(g) Vertical cylinder of diameter D buried in a semi- S¼
ln (4z=D)
infinite medium to a depth of z
h zi0:59 z0:078
(h) Isothermal rectangular parallelepiped of length L, S ¼ 1:685 L log 1 þ
b a
width b, and height a, buried in a semi-infinite
medium at a depth of z
2πL
(i) Cylindrical hole of diameter D through the center S¼
ln (1:08H=D)
of a square bar of height H and length L
2πL
(j) Eccentric cylinder of length L and diameter D1 in a S¼
cylinder of equal length and larger diameter D2 cosh1 ðD22 D21 4z2 Þ=(2D1 D2 )
2πL
(k) Horizontal cylinder of length L midway between S¼
ln (8z=πD)
parallel planes of equal length and infinite width
2πL
(l) Two isothermal cylinders of length L spaced a S¼
distance of W apart and buried in an infinite cosh1 ð4w2 D21 D22 Þ=(2D1 D2 )
medium
2πD2
(m) Two isothermal spheres spaced a distance of W S¼ h i
apart and buried in an infinite medium (D2 =D1 ) 1 ððD1 =2WÞ =ð1 ðD2 =2WÞ2 ÞÞ (D2 =W)
4
2πL
(n) Row of horizontal cylinders of length L and S¼
ln[(2W=πD) sinh (2π(z=W))]
diameter D spaced W apart in a semi-infinite
medium buried at a depth of z below the surface
(o) Thin disk of diameter D on a semi-infinite S ¼ 2D
medium
2πD
(p) Thin disk of diameter D buried in a semi-infinite S¼
(π=2) þ tan1 (D=4z)
medium with an adiabatic surface
between the adjoining materials is larger. Most thermal contact resistance correlations are
made as a function of contact pressure since this is the most significant factor. Another factor
is interstitial gaps, due to rough surfaces in contact, which influence heat flows through the
gases/fluids filling these gaps. Also, surface features such as roughness, waviness, and flat-
ness may be significant factors. Furthermore, surface cleanliness and the presence of dust
particles can also affect the thermal resistance.
Surface deformations, either plastic or elastic, also affect the contact resistance since the
deformation causes the contact area between adjoining surfaces to increase. In some cases,
the surface roughness at the contact interface may be quantified in terms of the mean
32 Advanced Heat Transfer
L
T1
L
z z
T1 z
T1 z
D D
a
T1 T1 D L b
FIGURE 2.5
Conduction shape factors for selected two-dimensional geometries [q = Sk(T1 − T2)].
series at the location of the interface between the two materials. A comprehensive
review of thermal contact resistance modeling and experimental data was presented by
Yovanovich (2005).
d ′′
q′′x Ac = q′′x Ac + (q Ac )dx + dAs h(T − T1 ) (2.26)
dx x
where Ac and As refer to the cross-sectional and outer surface areas, respectively, for a
differential control volume of thickness dx. The left side represents the heat inflow into
the control volume, while the heat outflows are shown on the right side of the equation.
Using Fourier’s law, a general fin equation becomes:
d2 T 1 dAc dT 1 h dAs
+ − (T − T1 ) = 0 (2.27)
dx2 Ac dx dx Ac k dx
FIGURE 2.6
Finned surfaces. (a) Spiral, (b) Axial, (c) Plate.
34 Advanced Heat Transfer
dAs
dq = dAs h (T – T∞)
Ac(x)
Fin
Base
qx qx+dx
x
Tb
CV (control
dx volume)
h, T∞
FIGURE 2.7
Schematic of a fin with a variable cross-sectional area.
where θ = T(x)–T∞, m 2 = hP/(kAc), and P, T∞, and Ac refer to the fin perimeter, ambient
fluid temperature, and cross-sectional area of the fin, respectively. The solution of this
equation is:
where x = 0 at the base of the fin and the coefficients D1 and D2 are obtained through
specification of the boundary conditions.
Applying the boundary conditions of an adiabatic fin tip with a specified base temper-
ature, the following results are obtained for the temperature excess, θ, and heat flow rate,
qf, respectively:
θ cosh(m(L − x))
= (2.30)
θb cosh(mL)
dθ
qf = −kAc = M tanh(mL) (2.31)
dx 0
Heat Conduction 35
where M = θb = (hPkAc)1/2. It can be shown that as the fin length becomes very long (i.e.,
L → ∞), the following results are obtained:
θ
= e−mx (2.32)
θb
qf = M (2.33)
Another type of boundary condition is a specified temperature at the tip of the fin,
θ(L) = θL. In this case, the results for temperature and heat flux become:
These results were obtained for the case of uniform fins with a constant cross-sectional
area. For nonuniform fins, the variation of cross-sectional and surface areas must be
included in the analysis and solution of the general fin equation.
Once the fin temperature and heat flux distributions are known, then the thermal effec-
tiveness or performance of the fin can be determined. The fin efficiency is defined as the ratio
of the actual fin heat transfer rate, qf, to the maximum heat transfer rate, qf,max. The maximum
heat transfer occurs if the base temperature is maintained throughout the entire fin, yielding
a maximum temperature difference between the fin and ambient fluid for convection. The
fin efficiency is written as:
qf qf
ηf = = (2.36)
qf ,max hAs θb
For the previous example of a uniform cross-sectional area fin with an adiabatic tip, the
expressions for the actual and maximum heat transfer rates can be substituted above to
yield:
tanh(mL)
ηf = (2.37)
mL
From this result, it can be observed that the fin efficiency decreases with longer fins, since
minimal heat transfer occurs when the temperature along the fin approaches the fluid tem-
perature in long fins.
The fin efficiency is often depicted in graphical form in terms of an abscissa, Xc, where,
1/2
h
Xc = L3/2
c (2.38)
kAp
which represents a dimensionless characteristic length of the fin. The fin efficiency, ηf, typ-
ically decreases with Xc. The efficiencies of various geometrical fin configurations can be
compared in terms of this abscissa under different operating conditions. Fin performance
curves are illustrated in Figure 2.8 for various fin geometries. All design parameters that
36 Advanced Heat Transfer
0.9 t
0.8
L
0.7
t
Fin efficiency, ηf
0.6 r2c/r1 = 2
L
0.5 r2c/r1 = 4
0.4
0.3
t
L r1 L = r2 - r1
0.2
Lc = L + t/2
r2c = r1 + Lc
0.1 r2 Ap = tLc
0
0 0.5 1 1.5 2 2.5
Lc 3/2 (h/kAp)1/2
FIGURE 2.8
Fin performance curves. (Adapted from J.P. Holman. 2010. Heat Transfer, 10th Edition, New York: McGraw-Hill.)
lead to a temperature distribution closer to the base temperature (i.e., higher fin conductiv-
ity, lower convection coefficient) also lead to a higher overall efficiency of the fin.
The geometrical configuration of a fin has a significant role in its performance. In
Figure 2.8, a triangular fin cross-sectional area shows better performance for a given surface
length than the rectangular profiles since it provides a surface temperature closer to the base
temperature. For a rectangular fin, the fin profile area is Ap = tLc and the characteristic fin
length is Lc = L + t/2, where L and t refer to the protruding fin width and base thickness,
respectively. On the other hand, Lc = L and Ap = tL/2 for a triangular fin. For purposes of
comparing different types of fins, the fin areas are usually matched.
In Figure 2.8 for shorter fins, Lc decreases, so the temperature profile remains closer to the
base temperature, thereby raising ηf. However, the heat transfer is reduced since the exposed
area of the fin is reduced. As a result, a common design objective is the minimization of Lc
while providing the required rate of heat transfer from the fin. In terms of conductivity
trends in the abscissa, Xc, a higher value of k (e.g., a copper rather than aluminum fin) allows
the fin temperature to become closer to the base temperature, thereby raising the
fin efficiency.
At a given Ap and Lc, while maintaining other parameters constant (such as the convection
coefficient), the fin efficiency for a triangular fin is higher than a rectangular fin. However,
fabrication of a triangular fin is generally more difficult and costly. Also, rectangular fins can
often be packed together more tightly than triangular fins. As a result, a trade-off exists
among various design criteria. In the limiting case of Xc → 0, the fin efficiency reaches its
maximum value of 1. But this case has a limited practical value in view of other design
limitations such as the implication of the convection coefficient approaching zero. In
practice, selecting the most suitable fin configuration involves a balance between both the
fin efficiency and other practical design considerations such as the cost of fabricating the
fin assembly.
Heat Conduction 37
Since heat transfer occurs through the fin and across the base surface, the associated ther-
mal resistance, Rf, and resistance of the base surface, Rb, are also useful in characterizing the
fin performance,
1
Rf = (2.39)
ηf hf Af
1
Rb = (2.40)
hb Ab
where the subscripts f and b refer to the fin and base, respectively. Heat transfer occurs in
parallel through the fin and base. The equivalent resistance, Re, can be written in terms of
a single overall surface efficiency, ηo, and an average convection coefficient, h, as follows:
1 1 −1 1 1
Re = + = = (2.41)
Rf Rb hb Ab + ηf hf Af ηo hA
where A refers to the total surface area (including fins and base). Equating the latter two
expressions for the equivalent resistance, substituting A = Af + Ab, and assuming the con-
vection coefficients are equal,
hb Ab + ηf hf Af Af
ηo = = 1 − (1 − ηf ) (2.42)
hA A
This overall surface efficiency can then be substituted back to obtain the overall equivalent
resistance of the fin, Re. Using this resistance in Equation 2.11, the overall heat transfer from
the fin to the surrounding ambient fluid can then be obtained by the temperature difference
(between the surface and ambient fluid) divided by the overall equivalent resistance. This
approach is particularly useful in the analysis of complex finned assemblies such as finned
heat exchangers (Chapter 9).
A thermal spreading resistance occurs when heat flows from a small heat source in contact
with the base of a larger heat sink. Then heat does not distribute uniformly throughout a
heat sink base and thus does not transfer efficiently to the fins for convective cooling. Razavi,
Muzychka, and Kocabiyak (2016) presented a comprehensive review of advances in thermal
spreading resistance problems. Further detailed analysis of extended surface heat transfer
was presented in a book by Kern and Kraus (1972).
where ∇ represents the vector gradient (or nabla) operator, and i, j, and k refer to unit vec-
tors in the x, y, and z directions, respectively. The bold font is used to designate a vector
quantity and distinguish the unit k vector from the thermal conductivity, k. In subsequent
notations in the book, the bold notation will be dropped as it is implicitly understood
that the heat flux and temperature gradient are vector quantities. The heat flux, q′′ , refers
to the vector components of heat flow per unit area (units of W/m 2) in the three coordinate
directions. In a similar manner as Fourier’s law, the vector of the diffusive mass flux may be
expressed in terms of the negative mass diffusion coefficient, D, multiplied by the concen-
tration gradient (called Fick’s law).
Consider a two-dimensional energy balance of conduction heat flows into/out of a differ-
ential control volume as illustrated in Figure 2.9. From the energy balance in Equation 2.1,
the transient energy accumulation within the control volume balances the net energy
inflow minus the energy outflow, plus any energy generated within the control volume.
Inflow and outflow terms may be evaluated as the heat flux multiplied by the area of the
face (unit depth) on the edge of the differential control volume. Across the upper and right
edges of the control volume, the heat fluxes may be expanded with a Taylor series in terms of
the corresponding heat fluxes at the inflow faces of the control volume. These heat fluxes are
depicted in Figure 2.9.
Similar to the procedure used to obtain the heat equation in Equation 2.5, by substituting
the individual heat flux terms into Equation 2.1 and simplifying, yields the following two-
dimensional heat equation:
2
∂T ∂ T ∂2 T
ρcp =k + + q̇ (2.44)
∂t ∂x2 ∂y2
where q̇ refers to a volumetric heat source. Under steady-state conditions without any heat
generation, the two-dimensional heat equation is reduced to the following Laplace equation:
∂2 T ∂2 T
+ = 0; ∇2 T = 0 (2.45)
∂x2 ∂y2
The divergence of the gradient of a function, ∇·∇, or ∇ 2, in the above equation, is called the
Laplace operator or Laplacian.
A steady-state heat conduction problem with constant thermophysical properties and
no internal heat generation is a linear system. The superposition principle can be applied to linear
y
(qy″ + (dqy″ /dy)dy + ...)dx
y 2-D control
volume y
qy″ dy
x
x x + dx x
FIGURE 2.9
Schematic of a two-dimensional control volume energy balance.
Heat Conduction 39
systems. The superposition principle states that the net response caused by the sum of two or
more solutions is the sum of the responses that would have been obtained by each response
individually. In other words, if a heat conduction problem, A, produces a solution X, and prob-
lem B produces a solution Y, then the problem (A + B) has a solution of (X + Y ).
For example, consider a square plate with each of the four sides maintained at four
separate temperatures, T1, T2, T3, and T4. Also, consider four individual problems with
the same square plate, but where one side is maintained at Ti (where i = 1, 2, 3, or 4), and
the remaining sides are held at 0◦ C. The side maintained at Ti is the same side corresponding
to the original problem involving four sides at different temperatures. Define the solution of
each individual problem, in order, by Ta, Tb, Tc, and Td. Each individual solution obeys
Laplace’s equation for steady-state heat conduction. Then, from the superposition principle,
the sum of the four individual problems is also a solution of the original overall problem.
The principle of superposition is a useful tool when analyzing complex linear problems
involving a combination of different types of boundary conditions.
z z
(x, y, z) (x, y, z)
y z y
θ (x, y, z)
θ
x Cylindrical
x Cartesian
z
y
r sin(θ )
φ
x
Spherical
FIGURE 2.10
Cartesian, cylindrical, and spherical coordinate systems.
40 Advanced Heat Transfer
The governing equations of heat conduction and Fourier’s Law can be generalized
and written in terms of curvilinear coordinates. Consider a set of coordinate transformation
equations, where ui = ui (x, y, z) are the curvilinear coordinates with inverses x = x(u1, u2, u3),
y = y(u1, u2, u3) and z = z(u1, u2, u3). For example, the coordinate transformations and
inverses for circular cylindrical coordinates, (r, θ, z), can be found as:
⎧ ⎧ ⎧
⎨ x = r cos θ ⎪
⎨ r = x +
y
2 2
⎨ u1 r
y = r sin θ y u θ (2.46)
⎩ ⎪ θ = tan−1 ⎩ 2
z=z ⎩ x u3 z
z=z
Substituting the individual arc length expressions, dxi, and rearranging, yields the follow-
ing arc length in the curvilinear coordinate direction si,
√
dsi = gi dui (2.50)
Q3 + dQ3
Q2 + dQ2
du3
u3
Q1
du1
u1 Q1 + dQ1
ds3
Q2
y ds2
u2
ds1
Q3
x du2
FIGURE 2.11
General coordinate system and heat flow through a differential flux tube in curvilinear coordinates.
Heat Conduction 41
Using this general form, the heat conduction equation can then be expressed in any of the
curvilinear coordinate systems. For example, it can be shown that the Lamé coefficients and
differential area elements in spherical coordinates are:
⎧ ⎧ ⎧ ⎧
⎨ x = r sin ϕ cos θ ⎨ u1 r ⎨ g1 = gr 1 ⎨ dA1 = r2 sin ϕdϕdθ
y = r sin ϕ cos θ u ϕ g2 = gϕ r2 dA2 = r sinϕdθ (2.58)
⎩ ⎩ 2 ⎩ ⎩
z = r cos ϕ u3 θ g3 = gθ r2 sin2 ϕ dA3 = rdϕdr
The resulting three-dimensional form of Fourier’s Law and the heat conduction equations
for Cartesian, cylindrical, and spherical coordinates are summarized as follows:
∂T ∂T ∂T
q′′ = −k ix − k jy − k kz (2.59)
∂x ∂y ∂z
42 Advanced Heat Transfer
∂T ∂ ∂T ∂ ∂T ∂ ∂T
ρcp = k + k + k + q̇ (2.60)
∂t ∂x ∂x ∂y ∂y ∂x ∂z
∂T k ∂T ∂T
q′′ = −k ir − jθ − k k z (2.61)
∂r r ∂θ ∂z
∂T k ∂ ∂T k ∂2 T ∂2 T
ρcp = r + 2 2 + k 2 + q̇ (2.62)
∂t r ∂r ∂r r ∂θ ∂z
∂T k ∂T k ∂T
q′′ = −k ir − j − kϕ (2.63)
∂r r ∂θ θ r sin θ ∂ϕ
∂T k ∂ 2 ∂T k ∂2 T k ∂ ∂T
ρcp = 2 r + 2 2 + sin θ + q̇ (2.64)
∂t r ∂r ∂r r sin θ ∂ϕ2 r2 sin θ ∂θ ∂θ
A similar approach as the previous section can be used to determine the conduction shape
factor for geometrical configurations in orthogonal curvilinear coordinates.
Consider a heat flow rate of Q1 in the u1 direction through a differential flux tube
bounded in ui coordinates (see Figure 2.11). The total thermal resistance through the flux
tube is given by:
where the superscript (3) designates a differential of order 3. Expressing the arc length and
differential surface area in terms of the metric coefficients,
1 g1 du1
d(3) R1 = √ (2.66)
k g du2 du3
Then the total thermal resistance of a flux tube between u1 = a and u2 = b becomes,
b
1 g1
d(3) R1 = √ du1 (2.67)
kdu2 du3 a g
The overall resistance and shape factor of the system can then be obtained by placing the
flux tubes “in parallel,” yielding,
1 du2 du3
S1 = = b (2.68)
kR1 u3 u2 √
g1 / g du1
a
Heat Conduction 43
This result represents the shape factor for regions bounded by curvilinear coordinate lines.
Teerstra, Yovanovich, and Culham (2005) extended this approach to calculate shape factors
for various three-dimensional geometries.
For example, consider the shape factor of a region between concentric cylinders of length L
and radii r1 and r2 from an angle of 0 to π/2. Using Equation 2.68, the shape factor in each of
the curvilinear coordinate directions is determined as follows:
⎧
⎧ ⎧ ⎪
⎪
⎨ u1 = r ⎨ g1 = 1 ⎪ ⎨ L π/2
dθ dz πL/2
Case (i): T = T(r) u2 = θ g2 = r 2
Sr = r2 = (2.69)
⎩ ⎩ ⎪ ln(r2 /r1 )
u3 = z g3 = 1 ⎪ ⎪
⎩
0 0
(1/r) dr
r1
⎧
⎧ ⎧ ⎪
⎪
⎨ u1 = z ⎨ g1 = 1 ⎪
⎨ r2 π/2
dθ dr π(r22 − r21 )
Case (ii): T = T(z) u =θ g2 = r2 Sz = L = (2.70)
⎩ 2 ⎩ ⎪
⎪ 4L
u3 = r g3 = 1 ⎪
⎩
r1 0
(1/r) dz
0
⎧
⎧ ⎧ ⎪
⎪
⎨ u1 = θ ⎨ g 1 = r2 ⎪
⎨ L r 2
dr dz L ln(r2 /r1 )
Case (iii): T = T(θ) u =r g =1 Sθ = π/2 = (2.71)
⎩ 2 ⎩ 2 ⎪
⎪ π/2
u3 = z g3 = 1 ⎪
⎩
0 r1
r dθ
0
The use of a particular coordinate system should be selected to closely approximate the
problem domain such that boundary conditions can be most directly and easily specified.
For example, in fluid flow through a pipe, it is more convenient to identify the pipe boun-
dary through cylindrical coordinates, rather than Cartesian coordinates. In numerical meth-
ods for heat transfer and fluid flow, Cartesian coordinates are typically adopted since a
variety of complex geometries are generally encountered and Cartesian coordinates provide
the most flexibility for a range of geometries.
The general application of the method of separation of variables involves two main steps.
Firstly, the temperature solution is assumed to be “separable” such that it can be written in
the following general form:
where X(x) depends only on x but not y, and Y( y) depends only on y but not x. In other
words, the spatial distributions in each coordinate direction can be separated into a product
solution. Since all solutions will not have this form, a reduced set of solutions, {Xi(x)Yi( y)},
will be found,
T(x, y) = Ci Xi (x)Yi (y) (2.73)
i
which is also a solution for suitable choices of the constants, Ci. The separable assump-
tion allows the governing equation to be reduced from a partial differential equation involv-
ing two variables to two ordinary differential equations which can be solved separately. The
goal of the solution method is to find the constants, Ci, so that the boundary conditions are
also satisfied.
Then, the second step of the solution procedure is to impose constraints on the solutions
based on the boundary conditions. Values of T(x, y) are specified on the boundaries. The fol-
lowing example further describes how the two steps of the method of separation of variables
are applied.
∂2 T ∂2 T
+ =0 (2.74)
∂x2 ∂y2
y Robin condition
h0, Tb –k dT/dy | b = hb (T(x,b) – Tf)
ha, Tf
b
Insulated a x
boundary
FIGURE 2.12
Schematic of two-dimensional heat conduction.
Heat Conduction 45
Along the bottom ( y = 0), top ( y = b), left (x = 0) and right (x = a) surfaces, Neumann
and Robin boundary conditions are applied, respectively, as follows:
∂T
=0 (2.75)
∂y y=0
∂T hb
= − (T(x, b) − Tf ) (2.76)
∂y y=b k
∂T h0
= − (Tb − T(0, y)) (2.77)
∂x x=0 k
∂T ha
= − (T(a, y) − Tf ) (2.78)
∂x x=a k
Using the method of separation of variables, assume that T(x, y) is separable and let
T(x, y) = X(x)Y( y). Substituting this separable form of temperature into the heat equation,
The boundary conditions along the y = 0 and y = b surfaces become Y′ (0) = 0 and Y′
(b) = –hY(b)/k, respectively. Substituting these boundary conditions into the Y(y) general
solution yields C4 = 0 and an equation for the eigenvalues (λ) of the Sturm–Liouville problem,
λb
= cot(λb) (2.83)
Bi
where Bib = hbb/k represents the Biot number. The roots of this equation are called the
eigenvalue solutions or separation constants. These roots lie approximately π apart. In
order to locate the exact roots, a Newton–Raphson root-searching algorithm can be used.
In order to satisfy the remaining boundary conditions along the x = 0 and x = a bound-
aries, an infinite series is formed by superposing all possible solutions to meet the boun-
dary conditions,
1
T(x, y) = (An cosh(λn x) + Bn sinh(λn x))cos(λn y) (2.84)
n=1
46 Advanced Heat Transfer
where λn refers to the set of n eigenvalues. Applying the right boundary condition along
the surface x = a yields
Bn = −An ψ n (2.85)
where,
At the asymptotes, Bia = 0 and Bia → ∞, the coefficients become ψn = tanh(λa) and ψn =
coth(λna), respectively. If λna . 2.65, then tanh(λa) ≈ coth(λa).
The temperature solution can then be rewritten in terms of only one unknown
coefficient:
1
T(x, y) = An (cosh(λn x) − ψ n sinh(λn x))cos(λn y) (2.87)
n=1
In order to obtain the final unknown coefficient, An, orthogonality relations are applied.
The eigenfunctions cos(λny) are multiplied by their orthogonal counterparts, cos(λmy),
and the resulting sum is integrated from y = 0 to y = b. This integrated sum of products
on the left side vanishes for all terms in the series, except for the case m = n, due to the
orthogonality relationship between cos(λmy) and cos(λny). Then, the remaining unknown
coefficient may be simplified as follows:
This coefficient then yields a final closed form solution of the temperature field.
1
(cosh(λn x) − ψ n sinh(λn x))cos(λn y)
T(x, y) = (Tb − Tf )sin(λn b)h0 (2.90)
n=1
kψ n λn (λn b + sin(λn b)cos(λn b))
In addition, the total heat flow into the solid (across the left boundary) may be obtained
from differentiation of the temperature field at x = 0 and then integration of the resulting
heat flux across the surface,
b 1
∂T
q=2 −k dy = 2k An ψ n sin(λn b) (2.91)
0 ∂y x=0 n=1
It was assumed at the start of the solution procedure that a separable solution could be
found. It is not normally apparent or obvious beforehand whether or not a given problem
will be separable. Sometimes it may be apparent to an experienced practitioner from the
nature of the boundary conditions. But in general, the method normally starts with an
assumption of a separable solution, after which the assumption is either validated through
the solution process, or found to be invalid through the subsequent analysis.
Separation of variables is a powerful solution method for problems governed by linear
partial differential equations with boundary and initial conditions (called boundary value
problems). In boundary value problems, boundary conditions are specified for the unknown
dependent variable(s), for example, temperature, along the external boundaries of the
domain. Conduction heat transfer is an example of a boundary value problem because
the temperature (or its derivative, in the case of the heat flux) is typically specified along
the physical boundaries.
The latter relationship between the exponential and trigonometric forms of the complex
number is called Euler’s formula. Like ordinary real numbers, also complex numbers can
be added, subtracted, and multiplied, as polynomials in the variable i.
In the method of conformal mapping, the solution of the heat equation is obtained by
transforming a given complicated geometry in the complex z-plane (z = x + iy) to a simpler
geometry in the w-plane (w = u + iv), or vice versa, via a conformal transformation, w = g(z).
A point (x, y) in the z-plane is mapped to a corresponding point (u, v) in the w-plane through
48 Advanced Heat Transfer
Im(z) Im(z)
y z = x + iy
y
r
r
θ
0
θ x Re(z)
θ
r
0 x
Re(z)
–y z = x – iy
FIGURE 2.13
Two-dimensional complex plane.
the mapping of w = g(z), where g′ (z) must be nonzero. Table 2.2 and Figure 2.14 illustrate a
number of common functions of conformal transformations.
In Table 2.1, the bilinear transformation involves four independent coefficients a, b, c, and
d. Assuming a ≠ 0, it can be rewritten in terms of three unknown coefficients as follows:
z + b/a
w = S(z) = (2.93)
cz/a + d/a
This form allows the bilinear mapping to be determined when only three distinct image
values are specified: S(z1) = w1, S(z2) = w2, and S(z3) = w3. Using these three points, it can
TABLE 2.2
Conformal Mapping of Various Geometrical Configurations
Geometrical Function Conformal Transformation
x u 1 a x 1 a u x u
(d) (e) (f )
y w = sin(z) v w = tan(z) w = exp(z) v
y v y
2 1 2 1
x u x 3 4 u x u
3 4
FIGURE 2.14
Schematic of conformal transformations of selected regions.
be shown that the bilinear mapping can be rewritten again in terms of an implicit transfor-
mation as follows:
z − z1 z2 − z3 w − w1 w2 − w3
= (2.94)
z − z3 z2 − z1 w − w3 w2 − w1
Therefore, substituting three mapped points into this expression yields the full general
form of the conformal mapping. A point at infinity can be introduced as one of the pre-
scribed points in either the z-plane or w-plane. For example, if w3 = ∞, then the bilinear
mapping becomes:
w 2 − w3
=1 (2.95)
w − w3
z − z1 z2 − z3 w − w1
= (2.96)
z − z3 z2 − z1 w2 − w1
50 Advanced Heat Transfer
Conformal mappings are useful because they allow heat conduction problems with com-
plicated boundaries to be mapped into simpler geometrical domains where Laplace’s equa-
tion can be more readily analyzed and solved. The temperature distribution that satisfies
Laplace’s equation in the z-plane also satisfies the heat equation in the w-plane so that cor-
responding points in both planes have identical temperatures. The temperature field in the
z-plane is determined based on temperatures in the simpler w-plane and the conformal
transformation between the z- and w-planes. The following example illustrates how a con-
formal mapping can be used to determine the temperature distribution in a region exterior
to two circles.
y
w = f (z)
2i
T=1 v T = –1
½i
i
q
x u
q –i
–½ i
T=1
T = –1
–2i
FIGURE 2.15
Conformal mapping of a region exterior to two circles.
Heat Conduction 51
Expanding the inverse function in terms of x and y variables, and multiplying and
dividing by (x –iy) to remove the imaginary number from the denominator, yields,
1 x − iy
= u + iv (2.99)
x + iy x − iy
Then, equating the real and imaginary parts to u and v, respectively, in the w-plane,
x −y
u= ; v= (2.100)
x2 + y2 x2 + y2
∂2 Ψ ∂2 Ψ
+ =0 (2.101)
∂u2 ∂v2
Solving this heat equation for the temperature field, Ψ (u, v), in the w-plane, subject
to the boundary conditions of Ψ = 1 at v = –1/2 and Ψ = –1 at v = 1/2, yields a one-
dimensional temperature distribution in the v-direction,
Therefore, by changing variables back to the original (x, y) coordinates in the above
result, the nondimensional temperature distribution, T(x, y) in the original z-plane
becomes,
2y
T(x, y) = (2.103)
x2 + y2
This example has shown how the temperature distribution in a complicated two-
dimensional z-plane can be determined based on a simpler one-dimensional solution in
the w-plane using a suitable conformal mapping.
The method of conformal mapping can also be applied to the transformation of a standard
coordinate system to other orthogonal curvilinear coordinates, for example, elliptic cylindrical
coordinates or bipolar coordinates (see Figure 2.16). An orthogonal coordinate system is a
y ψ = π/3 y
ψ = 2π/3 η=2 ψ = π/6 ψ = π/4
ψ = π/4 ψ = π/3
ψ = 3π/4
η = 3/2
ψ = 5π/6 ψ = π/6 η = –1/2 η = 1/2
η=1
η=0
ψ=π ψ=0 η = –1 η=1
x x
ψ = 2π
(–a, 0) η=1 (a, 0)
ψ = 7π/6 η = 3/2 ψ = 11π/6
ψ = 5π/4 ψ = 7π/4 ψ = 3π/2
η=2 ψ = 5π/3
ψ = 4π/3 ψ = 11π/6 ψ = 7π/4
ψ = 3π/2
Elliptic cylindrical coordinates Bipolar coordinates
FIGURE 2.16
Elliptic cylindrical and bipolar coordinate systems.
52 Advanced Heat Transfer
system of curvilinear coordinates wherein each family of surfaces intersects the others at
right angles. Elliptic cylindrical coordinates are an orthogonal coordinate system that results
from projecting the two-dimensional elliptic coordinate system in the perpendicular
z-direction. As a result, the coordinate surfaces become prisms of confocal ellipses and
hyperbolae. Orthogonal curvilinear coordinates are useful in solving the heat conduction
equation when boundaries are more closely aligned with these coordinates than Cartesian,
cylindrical, or spherical coordinates.
Heat conduction in elliptic cylindrical coordinates, (η, ψ), can be analyzed based on the
following conformal transformation from (x, y) coordinates in the z-plane to (η, ψ) coordi-
nates in the w-plane:
a
z = f (w) = a cosh(η + iψ) = (eη+iψ + e−(η+iψ) ) (2.104)
2
The real and imaginary parts are Re(z) = x and Im(z) = y, respectively. Dividing x on the
right side by a·cosh(η), dividing y by a·sinh(η), squaring both, and adding,
2 2
x y
+ = cos2 ψ + sin2 ψ = 1 (2.106)
a cosh η a sinh η
This represents a family of ellipses defined by η = constant. Similarly, dividing the x-term
on the right side by a·cos(ψ), y-term by a·sin(ψ), squaring both, and subtracting,
2 2
x y
− = cosh2 η − sinh2 η = 1 (2.107)
a cos ψ a sin ψ
Then the thermal resistance and shape factors for an elliptic cylindrical element can be
determined accordingly. For example, consider a region in Figure 2.16 that represents a
section bounded by two isothermal confocal ellipses (η1 ≤ η ≤ η2; 0 ≤ ψ ≤ 2π; 0 ≤ z ≤ L).
Using Equation 2.68, the shape factor for the region is given by:
1 du3 du2 dψdz 2πL
S= = b = η2 = (2.109)
kR u2 u3 √ 0 0 η2 − η1
(g1 / g)du1 dη
a η1
This represents the shape factor for one-dimensional steady-state heat conduction
through an elliptic cylindrical wall element of width η2 − η1 and a cross-sectional area of 2πL.
The temperature distribution can also be mapped from the one-dimensional linear solu-
tion in η-coordinates to the original (x, y) elliptic cylindrical domain. The one-dimensional
Heat Conduction 53
d2 Ψ
=0 (2.110)
dη2
Ψ − Ψ1 η
= (2.111)
Ψ2 − Ψ1 η2
Using this temperature distribution in Equation 2.56, the heat flux along on the boundary
η = 0 is determined as:
1 dΨ k Ψ2 − Ψ1
qη = −k √ = − (cosh2 η − cos2 ψ)−1/2 (2.112)
gη dη η=0 a η2 − η1
Alternatively, mapping back to the original (x, y) coordinates, the temperature distribu-
tion can be written as:
T − T1 cosh−1 (x/a)
= (2.113)
T2 − T1 cosh−1 (b/a)
where –a ≤ x ≤ a and 0 ≤ y ≤ b.
Another example of conformal mapping of curvilinear coordinates is bipolar coordinates
(η, ψ, z); see Figure 2.16.
ψ + iη
z = f (w) = ia cot (2.114)
2
The metric coefficients for the transformation from Cartesian to bipolar coordinates are
given by:
⎧ ⎧ ⎧
⎨ x = a sinh η/(cosh η − cos ψ) ⎨ u1 = η ⎨ g1 = a/(cosh η − cos ψ)
y = a sin ψ/(cosh η − cos ψ) u =ψ g2 = g1 (2.115)
⎩ ⎩ 2 ⎩
z=z u3 = z g3 = 1
Then the shape factor for the region bounded by two eccentric cylinders becomes:
L π
1 du3 du2 dψdz 2πL
S= = b = η2 = (2.116)
kR u2 u3 √ 0 −π η 2 − η1
(g1 / g)du1 dη
a η1
1996). The reader is referred to more detailed and comprehensive books on complex analysis
and conformal mappings by Mathews (1982) and Henrici (1986).
Th − Ts L/k hL
= = ; Bi (2.117)
Ts − Tf 1/h k
L
Fluid
Ts
Th Area, Th
A
Ts
Bi
Tf , h
Tf , h
x
Th L/kA Ts 1/hA Tf
FIGURE 2.17
Schematic of transient heat conduction through a wall.
Heat Conduction 55
where Bi refers to the Biot number. The Biot number (Bi) is a dimensionless parameter that
characterizes the relative significance of spatial temperature gradients within a body.
In this example, for small Biot numbers, either the wall thickness, L, is very small, and/or
the thermal conductivity, k, is large, meaning that the internal thermal resistance is much
smaller than the external resistance. In general, L refers to a characteristic length scale,
such as the wall thickness for a planar wall, or the diameter for a spherical droplet. If the
Biot number is large, then spatial gradients of temperature within the body become signifi-
cant and a lumped analysis cannot be used. Instead, the heat equation must be solved with
both transient and spatial derivatives in Equation 2.5. As a general guideline, if Bi ≤ 0.1, then
a lumped analysis may be used; otherwise, spatial temperature gradients should be retained
in the analysis.
Consider an arbitrarily shaped solid, initially at a temperature of Ti and suddenly exposed
and cooled in a convective environment at T∞ with a convection coefficient of h (see
Figure 2.18). The volume and surface area of the object are V and As, respectively. Defining
the solid object as the control volume, and applying an energy balance from Equation 2.1,
leads to,
∂T
ρcp V = −hAs (T − T1 ) (2.118)
∂t
where T . T∞. The factor ρcpV is called the lumped thermal capacitance. It represents the
energy stored (or released) per degree of temperature change for the mass of the object after
it is immersed in the fluid at T∞. The capacitance is “lumped” in the sense that temperature
is assumed to be constant throughout the solid object.
Define the temperature excess, θ, as the difference between the solid and fluid tempera-
tures, so that integrating both sides of the above energy balance, subject to an initial condi-
tion of θ(0) = θi yields:
θ T(t) − T1
= = e−t/τ (2.119)
θi Ti − T1
where τ refers to the thermal time constant,
ρcP L
τ= (2.120)
h
The characteristic length, L, may be regarded as the ratio between the object’s volume to its
surface area. The time constant, τ, characterizes the object’s rate of temperature change with
Ti
Eout
Ecv
Liquid
T∞ CV
FIGURE 2.18
Sudden cooling of an object immersed in a fluid.
56 Advanced Heat Transfer
time. If τ is small, the temperature change is rapid, and conversely, if the convection coeffi-
cient is small relative to the thermal capacitance, then τ is large such that the rate of temper-
ature change with time is slow.
Alternatively, the solution may be expressed as:
θ ht
= exp − = exp( − Bi · Fo) (2.121)
θi ρcp L
where Fo refers to the Fourier number (Fo = αt/L 2), representing nondimensional time, and
Bi is the Biot number (Bi = hL/k). In this lumped analysis, the temperature varies with time,
but not spatially within the object.
1 ∂T ∂2 T
= 2 (2.122)
α ∂t ∂x
x x T∞ , h x
qo″
TS
T(x,t) T(x,t)
T(x,t)
Solid
Fluid
Ti Ti Ti
FIGURE 2.19
Transient temperature distributions in a semi-infinite solid: (i) constant surface temperature; (ii) constant surface
heat flux; and (iii) surface convection condition.
Heat Conduction 57
Using this similarity variable, the heat equation can be reduced to an ordinary differential
equation in terms of η alone.
d2 T dT
= −2η (2.124)
dη2 dη
where erf(η) is the error function. Rewriting this result in terms of x and t, differentiating with
respect to x at the wall, and applying Fourier’s law, yields the following wall heat flux:
∂T k(Ti − Ts )
q′′0 (t) = −k = √ (2.126)
∂x 0 παt
The result shows that the heat flux decreases with time since the slope of temperature falls
with time when heat conducts further into the solid. Typical trends of the temperature and
heat flux profiles were illustrated in Figure 2.19. Temperatures within the solid approach the
surface temperature, Ts, over time and the wall heat flux decreases with t 1/2.
T(x, 0) = Ti (2.127)
∂T
− k = q′′o ; T(1, t) = Ti (2.128)
∂x 0
For this problem, it can be shown that the exact solution is given by:
2q′′o αt −x2 xq′′ x
T(x, t) = Ti + exp − o erfc √ (2.129)
k π 4αt k 2 αt
where erfc(w) = 1 − erf(w) is the complementary error function. Again the heat flux can be
determined based on differentiation of the temperature with respect to x at the wall. Refer-
ring to Figure 2.19 and the above analytical solution, it can be observed that the wall tem-
perature increases with t 1/2.
58 Advanced Heat Transfer
T(x, 0) = Ti (2.130)
∂T
−k = h(T1 − T(0, t)); T(1, t) = Ti (2.131)
∂x 0
In this case, it can be shown that the analytical solution is given by:
√
T(x, t) − Ti x hx h2 αt x h αt
= 1 − erf √ − exp + 2 1 − erf √ (2.132)
T1 − Ti 2 αt k k 2 αt k
Differentiating the temperature field and substituting x = 0 at the wall yields the wall heat
flux based on Fourier’s law. It can be observed that the wall temperature and internal tem-
peratures approach T∞ with increasing time. The wall heat flux decreases with time due to
the decreasing difference between the fluid and wall temperatures. Sample dimensionless
temperature profiles for this convection case (iii) are illustrated in Figure 2.20.
When h → ∞, the limiting case corresponds to case (i) with a sudden change of wall
temperature to Ts at the initial time. The last bracketed term in Equation 2.132 in case
(iii) becomes zero and the analytical solution approaches case (i) for a constant
wall temperature.
1
(T – Ti) /(T∞ – Ti)
0.5
0.1
0.2
0.1
h(αt)1/2/k = 0.05
0.01
0.0 0.5 1.0 1.5
x/(2(αt)1/2)
FIGURE 2.20
Temperature profiles in a semi-infinite domain with surface convection. (Adapted from P.J. Schneider. 1957. Con-
duction Heat Transfer, Cambridge, MA: Addison-Wesley.)
Heat Conduction 59
1.0
0.9 x/L = 0.2 0.9
r/ro = 2.0
0.8 0.8
0.4 0.4
(T – T∞ ) /(Ti – T∞ )
(T – T∞ ) /(Ti – T∞ )
0.7 0.7
0.6 0.6
0.5 0.6 0.6
0.5
0.4 0.4
0.3 0.3
0.8 0.8
0.2 0.2
0.1 0.1
x/L = 1.0 r/ro = 1.0
0.0 0.0
0.01 0.1 1 10 100 0.01 0.1 1 10 100
k/(hL) = Bi–1 k/(hL) = Bi–1
FIGURE 2.21
Temperature distributions in (a) a plane wall of thickness 2L and (b) long cylinder. (Adapted from M.P. Heisler.
1947. Transactions of ASME, 69: 227–236.)
sphere. Solutions are often illustrated in a graphical form through Heisler charts (Heisler
1947). A Heisler chart typically illustrates the variation of temperature at a particular posi-
tion(s) at different times.
For example, consider transient heat conduction in a plane wall of thickness 2L, initially at
a temperature of T(x, 0) = Ti, and suddenly immersed in a fluid at T∞ ≠ Ti. Since convection
conditions are the same on both sides of the wall, the temperature distribution is symmet-
rical about the midplane. Results from the Heisler charts for this case and an infinite cylinder
of radius ro can be obtained in terms of spatial temperature distributions within the solid (see
Figure 2.21).
Similar results are obtained for both the plane wall and long cylinder. As the Biot number
decreases (or Bi −1 increases), the temperatures throughout the solid approach the fluid
temperature as a result of a high convection coefficient and/or low thermal conductivity
of the solid. The results at x = 0 are expressed with respect to the centerline temperature.
Separate additional Heisler curves are available for the variation of the centerline tempera-
ture with time.
Additional results for a constant surface temperature instead of a convection boundary
condition can be obtained as special cases of the above results as Bi → ∞. Changes of the
dimensionless wall heat flux, q* = q′′wL/(k(Ts − Ti)), with time for a semi-infinite solid, plane
wall, infinite cylinder, sphere, and exterior objects are illustrated in Figure 2.22.
The characteristic length is L or ro for a plane wall of thickness 2L, or a cylinder (or
sphere) of radius ro, respectively. For exterior objects, the characteristic length is
L = (As/4π)1/2, where As is the surface area. All heat fluxes decrease initially with the Fourier
number but at some point decrease as they each approach a different equilibrium
temperature.
In addition to the previous solution methods and Heisler charts, transient heat conduction
problems are often separable, and as a result, can also be analyzed by the method of sepa-
ration of variables. Using separation of variables, the temperature solution can be sub-
divided into a product of functions of spatial coordinates and time. Further analysis
and results of transient and multidimensional heat conduction problems are presented by
Bergman et al. (2011), Kakac and Yener (1988), and Schneider (1957).
60 Advanced Heat Transfer
100
q*
10
Exterior objects
1 Semi-infinite
solid
Plane wall
0.1
Cylinder
Sphere
0.01
0.001 0.01 0.1 1 10 100
Fo = αt/L2
FIGURE 2.22
Transient wall heat flux for various geometrical configurations. (Adapted from T.L. Bergman et al. 2011. Fundamen-
tals of Heat and Mass Transfer, 7th Edition, New York: John Wiley & Sons.)
PROBLEMS
2.1 A system of mass m is initially maintained at a temperature of T0. Then the system is
suddenly energized by an internal source of heat generation, q̇o . The surface at x = L
is heated by a fluid at T∞ with a convection coefficient of h. The boundary at x = 0 is
well insulated. Assume one-dimensional transient heat conduction.
a. Write the mathematical form of the initial and boundary conditions.
b. Explain how the temperature varies with time and position within the sys-
tem. Will a steady-state condition be reached? Explain your response.
c. Explain how the net heat flux varies with time along the planes of x = 0 and
x = L.
2.2 A long metal slab is insulated along the left boundary (x = 0) and a surrounding fluid
is initially at a temperature of Tf at x = L (right boundary of slab). Then a resistance
heating element is turned on and heat is generated electrically within the slab. Write
the one-dimensional governing equation and boundary conditions for this problem.
Find the steady-state temperature at the fluid–solid boundary. Explain how the heat
fluxes at this boundary and the other insulated boundary vary with time.
2.3 One-dimensional heat transfer occurs in a system of mass m initially at a temperature
of T0. Suddenly the system is heated by an internal electrical heat source, qo. The
surface at x = L is heated by a fluid at T∞ with a convection coefficient of h. The
boundary at x = 0 is maintained at T(0, t) = T0.
a. Select an appropriate control volume and derive the differential heat equa-
tion for T(x, t). Identify the initial and boundary conditions for this problem.
b. How does the temperature distribution throughout the material vary with
time and position? Will a steady-state temperature be reached? Explain
your response.
Heat Conduction 61
c. Explain how the heat flux varies with time along the planes x = 0 and x = L.
d. Discuss how the relative magnitudes of the internal and external heat transfer
processes affect the results in parts (b) and (c).
2.4 A rectangular system of mass m involving one-dimensional heat transfer with cons-
tant properties and no internal heat generation is initially at a temperature of Ti. Sud-
denly the surface at x = L is heated by a fluid at T∞ with a convection coefficient of
h and a uniform radiative heat flux of qo. The boundaries at x = 0 and elsewhere
are well insulated.
a. Select an appropriate control volume and derive the governing differential
heat equation for T(x, t). Identify the initial and boundary conditions for
this problem.
b. Explain how the temperature varies with x at the initial condition (t = 0) and
for several subsequent times. Will a steady-state condition be reached?
Explain your response.
c. Explain how the heat flux varies with time at x = L.
d. How would the results in parts (a) and (b) change if only a portion of the
incoming radiation was absorbed by the surface at x = L and the remainder
was reflected from the surface?
2.5 An experimental apparatus for the measurement of thermal conductivity uses an
electrically heated plate at a temperature of Th between two identical samples that
are pressed between cold plates at temperatures of Tc. It may be assumed that the con-
tact resistances between the surfaces in the apparatus are negligible. Thermocouples
are embedded in the samples at a spacing of l apart. The lateral sides of the apparatus
are insulated (one-dimensional heat transfer). The thermal conductivity varies with
temperature approximately as k(T ) = k0(1 + cT 2), where k0 and c are constants and
T refers to temperature.
a. The measurements indicate that a temperature difference of T1,h − T1,c exists
across the gap of width l. Determine an expression for the heat flow per unit
area through the sample in terms of T1,h, T1,c, l, and the constants k0 and c.
b. Explain the advantage or benefit of constructing the apparatus with two iden-
tical samples surrounding the heater, rather than a single heater–
sample combination.
2.6 A composite wall in a house consists of drywall exterior sections and an inner
section equally divided between wood studs and insulation. Construct a thermal
circuit to depict the heat conduction through this composite wall. The outer bound-
aries are isothermal. Is your analysis a one-dimensional solution? Explain your
response.
2.7 A thin film-type resistance heater is embedded on the inside surface of a wind tunnel
observation window for defogging during testing of aircraft deicing systems. The
outside and inside air temperatures are 23 and −12◦ C, respectively, and the outside
and inside convective heat transfer coefficients are 12 and 70 W/m2 K, respectively.
The thermal conductivity of the observation window is 1.2 W/mK. The window is
5-mm thick. Find the outside window surface temperatures for the following cases:
(i) without a heater, and (ii) with a heater turned on and a heat input of 1.2 kW/m2.
Also, determine whether the film-type heater is sufficient to prevent frost formation
on the window.
62 Advanced Heat Transfer
thickness of an asbestos insulation layer (k = 0.1 W/mK) such that heat losses from
the pipeline are reduced by 50%.
2.15 A spherical storage tank with an inner diameter of 120 cm is maintained at a temper-
ature of 130◦ C. An insulation layer is required to cover the tank to reduce heat losses
from the tank and ensure that the outer tank surface temperature does not exceed
50◦ C. The tank is surrounded by an airstream at 20◦ C with a convection coefficient
of 20 W/m2 K. Determine the thermal conductivity and thickness of insulation
required to reduce the heat losses from the tank by 72%. It may be assumed that
the convection coefficient is unaltered by the insulation layer.
2.16 A cylindrical fuel rod within a nuclear reactor generates heat due to fission according
to q = qo − (r/R), where R and qo refer to the radius of the fuel rod and an empirical
coefficient, respectively. The boundary surface at r = R (2.2 cm) is maintained at a uni-
form temperature, To. Find the coefficient, qo, if the measured temperature difference
between the centerline and the outer surface of the rod is 400 K and the thermal con-
ductivity is k = 12 W/mK.
2.17 A plane layer of a fuel element within a nuclear reactor generates heat by fission at a
rate of qo′′′ (W/m3). The boundary surfaces at x = 0 and x = L are maintained at tem-
peratures of T1 and T2, respectively. The thermal conductivity of the element
increases with temperature based on k(T ) = k0 + k1T, where k0 and k1 are constants.
Assuming steady-state conditions, determine the heat flux across the boundary at
the surface x = L.
2.18 Brazing is a metal-joining process that heats a base metal to a high temperature and
applies a brazing material to the heated joint. The base metal melts the brazing alloy
and fills the joint, and after it cools, the metal solidifies in the joint. Consider a copper
alloy brazing rod with a thermal conductivity, diameter, and length of 70 Btu/h ft ◦ F,
1/4 in., and 5 ft, respectively. The end of the rod is heated by a torch and reaches an
average temperature of 1,400◦ F. A heat transfer analysis is required to determine
whether a machinist can grasp the rod. What is the rod temperature at a distance
of 2 ft from the heated end? The ambient air temperature and convection coefficient
are 70◦ F and 1 Btu/h ft ◦ F, respectively.
2.19 A team member in a race car competition has proposed to air-cool the cylinder of a
combustion chamber by joining an aluminum casing with annular fins (k = 240 W/
mK) to the cylinder wall (k = 50 W/mK). This configuration involves an inner cylin-
der (radii of 50 and 55 mm) covered by an outer cylinder (radii of 55 and 60 μm) and
annular fins. The outer radius, thickness, and spacing between the fins are 10 cm,
3 mm, and 2 mm, respectively. Air flows through the fins at a temperature of 308 K
with a convection coefficient of h = 100 W/m2 K. The temporal average heat flux at
the inner surface is 80 kW/m2, and the wall casing contact resistance is negligible.
a. If the fin efficiency is 84%, then find the steady-state interior wall temperature
(at r = 50 mm) for cases of a cylinder with and without fins.
b. Discuss what factors should be investigated prior to a final design with a spe-
cific fin configuration.
2.20 Pin fins are often used in electronic systems to cool internal components and support
devices. Consider a pin fin that connects two identical devices of width Lg = 1 cm and
surface area Ag = 2 cm2. The devices are characterized by a uniform heat generation
of q̇ ≈ 300 mW. Assume that the back and top or bottom sides of the devices are well
64 Advanced Heat Transfer
insulated. Also, assume that the exposed surfaces of the devices are held at a uniform
temperature, Tb. Heat transfer occurs by convection from the exposed surfaces to the
surrounding fluid at T∞ = 20◦ C with h = 40 W/m2 K. Select an appropriate control
volume and determine the base (surface) temperature, Tb. The fin diameter and
length are 1 and 25 mm, respectively, and the thermal conductivity of the fin is
400 W/mK.
2.21 Copper tubing is joined to a solar collector plate (thickness of t). Fluid flowing within
each tube maintains the plate temperature above each tube at T0. A net radiative heat
flux of qrad is incident on the top surface of the plate, and the bottom surface is well
insulated. The top surface is also exposed to air at a temperature of T∞ with a convec-
tion coefficient of h. Since t may be assumed to be small, the temperature variation
within the collector plate in the y direction (perpendicular to the plate) is assumed
negligible.
a. Select an appropriate control volume and derive the governing differential
equation that describes the heat flow and temperature variation in the x direc-
tion along the collector plate. Outline the boundary conditions.
b. Solve the governing differential equation to find the plate temperature half-
way between the copper tubes. The convection coefficient, thermal conduc-
tivity, plate thickness, T∞, T0, qrad, and tube-to-tube spacing are 20 W/m2 K,
200 W/mK, 8 mm, 20◦ C, 60◦ C, 700 W/m2, and 0.1 m, respectively.
2.22 A cylindrical copper fin is heated by an airstream at 600◦ C with a convective heat
transfer coefficient of 400 W/m2 K. The base temperature of the 20-cm-long fin is
320◦ C. What fin diameter is required to produce a fin tip temperature of 560◦ C?
Also, find the rate of heat loss from the fin.
2.23 Derive the three-dimensional heat conduction equation (including heat generation)
in the spherical coordinate system.
2.24 The temperature distribution within a solid material covering –0.1 ≤ x ≤ 0.4 and –0.1
≤ y ≤ 0.3 (m) is given by T = 20x 2 − 40y 2 + 10x + 200 (K). The thermal conductivity of
the solid is 20 W/mK.
a. Find the rate of heat generation and the heat flux (magnitude and direction) at
the point (x, y) = (0.1, 0.1) in the solid.
b. Find the total heat flows (W/m; unit depth of solid) across each of the four
surfaces of the rectangular region defined by 0 ≤ x ≤ 0.3 and 0 ≤ y ≤ 0.2
within the solid.
2.25 Find the upper and lower bounds on the thermal resistance of a composite cylindrical
wall consisting of an inner layer (thickness of Δ1, conductivity of k1) and two adjoin-
ing outer half-cylinders (thickness of Δ2, conductivities of k2 and k3). The contact resis-
tances at the interfaces between the cylindrical layers may be neglected.
2.26 The top and side boundaries of a long metal bar are maintained at 0◦ C, and the bot-
tom side is held at 100◦ C. If the square bar thickness is 1 cm, find the spatial temper-
ature distribution inside the plate.
2.27 During laser heating of a metal rod of height L, a uniform heating rate of qo is applied
along half of the upper surface of the bar, whereas the remaining half is well insu-
lated. The top end is also well insulated, and the other end is exposed to a fluid at
T∞ and a convection coefficient of h. Use the method of separation of variables to
the find the steady-state temperature distribution within the rod.
Heat Conduction 65
2.28 The outer wall of a hollow half-cylinder (r = ro, –90◦ , θ , 90◦ ) is subjected to a
uniform heat flux between –β ≤ θ ≤ β. It is well insulated over the remaining
angular range and the ends at θ = +90◦ . Along the inner boundary (r = ri), the
component is convectively cooled by a fluid with a temperature and convec-
tion coefficient of Tf and h, respectively. Use the method of separation of variables
to find the steady-state temperature within the half-cylinder metal component.
2.29 After heat treatment in a furnace, a cylindrical metal ingot is removed and cooled
until it reaches a steady-state temperature governed by two-dimensional heat con-
duction in the radial (r) and axial (z) directions. The top (z = H ), bottom (z = 0),
and outer surfaces (r = ro) are maintained at Tw, Tw, and Tc, respectively. Use the
method of separation of variables to find the temperature distribution and resulting
total rate of heat transfer from the ingot.
2.30 Consider an irregular region formed by the gap between concentric polygons. Use the
method of conformal mapping to establish an expression for the temperature distri-
bution in this region when Dirichlet (fixed temperature) boundary conditions are
applied along the surfaces of both polygons.
2.31 It is required to determine the temperature distribution inside a crescent spaced
region that lies inside the disk |z − 2| , 2 and outside the circle |z − 1| = 1. Deter-
mine a conformal mapping that allows the known temperature distribution within a
horizontal strip of unit width to be mapped to the crescent spaced region.
2.32 Find a conformal mapping which allows the temperature distribution in the upper
half-plane to be mapped to the upper half-portion of the circle |z| , 1 which lies
in the upper half-plane.
2.33 Material property issues have been observed after the quenching of steel ingots under
conditions when the surface cooling is not sufficiently fast to avoid the formation of
soft pearlite and bainite microstructures in the steel. Consider a small steel sphere
(diameter of 2 cm) that is initially uniformly heated to 900◦ C and then hardened by
suddenly quenching it into an oil bath at 30◦ C. The average convection coefficient
is h = 600 W/m2 K. What is the sphere surface temperature after 1 min of quenching
time has elapsed? The properties of the steel alloy are ρ = 7,830 kg/m3, cp = 430 J/kg
K, and k = 64 W/mK.
2.34 Water in a storage tank is heated by combustion gases that flow through a cylindrical
flue. In the water heater, gases flow upward through an unobstructed exit damper in
an open tube. A new energy storage device is proposed to improve the water heater
performance by constructing a damper that consists of alternating flow spaces and
rectangular plates (width of 0.01 m) aligned in the streamwise (longitudinal) direc-
tion. Each plate has an initial temperature of 25◦ C with ρ = 2,700 kg/m3, k = 230 W/
mK, and cp = 1,030 J/kg K. Under typical operating conditions, the damper is ther-
mally charged by passing a hot gas through the hot spaces with a convection coeffi-
cient of h = 100 W/m2 K and T∞ = 600◦ C.
a. How long will it take to reach 90% of the maximum (steady state) energy stor-
age in each central damper plate?
b. What is the plate temperature at this time?
2.35 A spherical lead bullet with a diameter of D = 5 mm travels through the air at a super-
sonic velocity. A shock wave forms and heats the air to T∞ = 680 K around the bullet.
The convection coefficient between the air and the bullet is h = 600 W/m2 K.
66 Advanced Heat Transfer
The bullet leaves the barrel at Ti = 295 K. Determine the bullet’s temperature
upon impact if its time of flight is 0.6 sec. Assume constant lead properties with
ρ = 11,340 kg/m3, cp = 129 J/kgK, and k = 35 W/mK.
2.36 A pulverized coal particle flows through a cylindrical tube prior to injection and com-
bustion in a furnace. The inlet temperature and diameter of the coal particle are 22◦ C
and 0.5 mm, respectively. The tubular surface is maintained at 900◦ C. What particle
inlet velocity is required for it to reach an outlet temperature of 540◦ C over a tubular
duct 5 m in length?
2.37 A cylindrical wall of a metal combustion chamber (ρ = 7,850 kg/m3, cp = 430 J/kgK) is
held at an initially uniform temperature of 308 K. In a set of experiments, the outer
surface is exposed to convective heat transfer with air at 308 K and a convection coef-
ficient of h = 80 W/m2 K. An internal combustion process exchanges heat by radia-
tion between the inner surface and the heat source at 2,260 K (the approximate
flame temperature of combustion). Use the lumped capacitance method to estimate
the time required for the cylinder to reach the melting temperature of 1,800 K. In prac-
tice, how would convective cooling be applied to prevent melting of the cylinder?
2.38 Consider a plane wall, initially at Ti, suddenly exposed to a convective environment
on both sides of the wall with a convection coefficient of h and an ambient tempera-
ture of T∞. The wall half-thickness is L, and x is measured from the midpoint of the
wall rightward. Find the temperature distribution within the wall, including the var-
iations with time and position.
References
T.L. Bergman, A.S. Lavine, F.P. Incropera and D.P. DeWitt. 2011. Fundamentals of Heat and Mass
Transfer, 7th Edition, New York: John Wiley & Sons.
H.S. Carslaw and J.C. Jaeger. 1959. Conduction of Heat in Solids, Oxford: Oxford University Press.
J.C. Han. 2011. Analytical Heat Transfer, Boca Raton: CRC Press/Taylor & Francis.
M.P. Heisler. 1947. “Temperature Charts for Induction and Constant Temperature Heating,” Transac-
tions of ASME, 69: 227–236.
P. Henrici. 1986. Applied and Computational Complex Analysis, New York: John Wiley & Sons.
J.P. Holman. 2010. Heat Transfer, 10th Edition, New York: McGraw-Hill.
S. Kakac and Y. Yener. 1988. Heat Conduction, Washington, D.C.: Hemisphere.
D.Q. Kern and A.D. Kraus. 1972. Extended Surface Heat Transfer, New York: McGraw-Hill.
J.H. Mathews. 1982. Basic Complex Variables, Boston: Allyn and Bacon.
G.E. Myers. 1971. Analytical Methods in Conduction Heat Transfer, New York: McGraw-Hill.
G.F. Naterer. 1996. “Conduction Shape Factors of Long Polygonal Fibres in a Matrix,” Numerical Heat
Transfer A, 30: 721–738.
M. Razavi, Y. S. Muzychka and S. Kocabiyik. 2016. “Review of Advances in Thermal Spreading
Resistance Problems,” AIAA Journal of Thermophysics and Heat Transfer, 30: 863–879.
P.J. Schneider. 1957. Conduction Heat Transfer, Cambridge, MA: Addison-Wesley.
P.M. Teerstra, M.M. Yovanovich and J.R. Culham. 2005. “Conduction Shape Factor Models for Three-
Dimensional Enclosures,” AIAA Journal of Thermophysics and Heat Transfer, 19: 527–532.
Y. Yener and S. Kakac. 2008. Heat Conduction, 4th Edition, Boca Raton: CRC Press/Taylor & Francis.
M.M. Yovanovich. 2005. “Four Decades of Research on Thermal Contact, Gap, and Joint Resistance in
Microelectronics,” IEEE Transactions on Components and Packaging Technologies, 28(2): 182–206.
3
Convection
3.1 Introduction
Convective heat transfer involves the combination of molecular diffusion of heat and bulk
fluid motion (also called advection). Molecular diffusion occurs by random Brownian
motion of individual particles in the fluid, whereas in advection, heat is transferred by
larger-scale motions of currents in the fluid. Since the bulk motion of the fluid signifi-
cantly affects the rate of heat transfer, detailed knowledge of the fluid velocity and pres-
sure distributions are essential in convection problems. Therefore, both fluid flow and
energy equations must be considered in the analysis. In addition to heat transfer, convec-
tion is also a major mode of mass transfer in fluids. These governing equations and asso-
ciated physical processes of convective heat and mass transfer will be examined in
this chapter.
As briefly described in Chapter 1, Newton’s law of cooling is commonly used in convection
problems. The rate of convective heat transfer is proportional to the heat transfer coefficient,
h, and temperature difference between the surface and fluid. Here the fluid temperature
must be carefully specified. For external flows, the fluid temperature is usually the ambient
or freestream temperature. However, for internal flows (such as flow in a duct or pipe), the
fluid temperature in Newton’s law of cooling is normally the mean temperature of the fluid
which varies with position. Proper selection of the fluid temperature is important in the
empirical correlations involving the heat transfer coefficient and calculation of thermophys-
ical properties.
This heat transfer coefficient depends on various problem parameters such as the fluid
thermophysical properties, problem geometry, and fluid velocity, among other factors. In
general, the type of convection problem can be categorized based on the magnitude of
the convective heat transfer coefficient. For example, a typical range of coefficients for com-
mon forms of convective heat transfer is: (i) free convection with air (5 , h , 25 W/m2 K);
(ii) forced convection with air (10 , h , 500 W/m2 K); (iii) forced convection with water
(100 , h , 15,000 W/m2 K); and (iv) condensation of water (5,000 , h , 100,000 W/m2 K).
Several excellent textbooks have been written with a broad and deep coverage of convective
heat transfer processes, for example, Bergman et al. (2011), Kakac, Yener, and Pramuanjar-
oenkij (2013).
Convection can be induced by a number of external forces leading to natural, forced, grav-
itational, granular, or thermomagnetic convection. Natural convection, or free convection,
occurs due to temperature differences which affect the density and hence lead to buoyancy
forces that drive the fluid motion. Free convection can only occur in a gravitational field. In
contrast, forced convection occurs from externally imposed forces such as a turbine or pump.
It is typically used to increase the rate of heat transfer and involves various modes of fluid
67
68 Advanced Heat Transfer
motion such as laminar and turbulent flow. Typical examples of forced convection include
fluid radiator systems, cooling of microelectronic assemblies, pipe flows, and aerodynamic
heating of aircraft.
Gravitational convection is similar to natural convection but arises from buoyancy
forces due to property variations other than temperature, such as concentration (also called
solutal convection). For example, variable salinity is a significant factor of convection in the
oceans. Thermomagnetic convection may occur when an external magnetic field in the pres-
ence of a temperature gradient leads to a nonuniform magnetic body force and fluid move-
ment. Another less common mode of convection, called granular convection, occurs in
powders and granulated materials in containers as a result of surface vibrations. When an
axis of vibration in a container is parallel to the gravitational force, a slow relative movement
of particles may occur along the sides when the container accelerates in the vertical
direction.
A wide range of physical processes may induce convection. The above examples
included buoyancy, external forces, concentration gradients, and a magnetic field. Another
common example is chemical reactions and combustion (to be further discussed in
Chapter 8). Capillary action is a process of inter-molecular attractive forces between a liquid
and solid surface where liquid spontaneously rises in a narrow space such as a thin tube or
porous material. Examples include capillary induced motion of a fluid in microdevices
or petroleum reservoirs. The Marangoni effect refers to convection of fluid along a phase
interface due to differences in surface tension caused by inhomogeneous composition of
the substances or temperature dependence of the surface tension forces (called thermocapil-
lary convection).
Convection has an important role in many engineering and scientific processes. It has
importance in the structure of the Earth’s atmosphere, oceans, and mantle—and on an
astronomical scale, also the sun, stars, and galaxies. This chapter will focus on forced
and natural convection in thermal engineering applications. It will present the governing
equations, solution methods, and correlations for a range of common geometrical config-
urations. This chapter will also provide a brief introduction to turbulent convection and
the second law of thermodynamics, as it relates to convective heat transfer.
where the subscripts cv, in, out, and g refer to control volume, inflow, outflow, and
generation term, respectively. Equation 3.1 states that the rate of accumulation (or decrease)
of B with time in the control volume is equal to the inflow of B, minus the outflow, plus the
rate of internal generation of B. This represents a conservation form of the governing
equation since it can be interpreted as a balance or conservation of B within the
control volume.
Convection 69
ṁx = ρu dy (3.2)
This expression considers a flow from left to right with a positive u-velocity in the
positive x-direction. Performing a Taylor series expansion at position x + dx (outflow
boundary),
∂ṁx
ṁx+dx = ṁx + dx + · · · (3.3)
∂x
Higher order terms can be neglected in the limit as the control volume size, dx, becomes
very small. Similar expressions can be obtained for the mass flow rates in the y-direction.
Substituting the various inflow and outflow terms into Equation 3.1,
∂ρ ∂(ρu) ∂(ρv)
dx dy = ρu dy + ρv dx − ρu dy + dx dy − ρv dx + dy dx (3.4)
∂t ∂x ∂y
By simplifying and rearranging this result, the following mass conservation equation is
obtained for two-dimensional flow:
∂ρ ∂(ρu) ∂(ρv)
+ + =0 (3.5)
∂t ∂x ∂y
pdx (τxy+d(τxy))dx
dy
(ρu)dy (ρu + d(ρu))dy (σxx)dy dy
(σxx+d(σxx))dy
(τxy)dy
(ρuu)dy (ρuu + d(ρuu))dy (p +dp)dx
dx dx
(x,y) (x,y)
Surface (τyx)dx pdy
(ρv)dx (ρvu)dx Direction
(σyy)dx
FIGURE 3.1
Mass flow, momentum flux, and force components of a control volume balance.
70 Advanced Heat Transfer
For incompressible flows with a uniform density (ρ ≈ constant), Equation 3.5 can be
reduced to:
∂u ∂v
+ =0 (3.6)
∂x ∂y
or more compactly as ∇·v = 0. The divergence of the velocity field, ∇·v, can be interpreted as
the net outflow of a fluid from a differential control volume, which is zero at steady state,
since any inflows are balanced equally by the mass outflows.
where the terms in brackets include two components: momentum fluxes due to advection;
and forces acting on the surfaces of the control volume. Both the x-momentum fluxes and the
sum of the forces on the control volume can independently alter the rate of change of
momentum in the control volume.
To perform the x-momentum balance, the various inflow and outflow terms will be
assembled like the previous derivation of the conservation of mass (refer again to Figure 3.1).
In this case, the conserved quantity refers to the x-direction momentum. The x-momentum
fluxes along the inlet and outlet surfaces of the control volume, per unit depth, at locations x
and x + dx, and y and y + dy, respectively, are given by:
∂Ṁx
Ṁx,x = ρu udy; Ṁx,x+dx = Ṁx + dx + · · · (3.8)
∂x
∂Ṁx
Ṁx,y = ρv udx; Ṁx,y+dy = Ṁx + dy + · · · (3.9)
∂y
The first and second subscripts refer to x-momentum and the surface location. Again,
higher order terms in the Taylor series expansions can be neglected and similar terms are
obtained for the x-momentum fluxes across the surfaces at locations y and y + dy.
In order to obtain the remaining force terms in the momentum balance, Equation 3.7, con-
sider the pressure, p, shear stress, τij, and normal stress, σij, acting on the surfaces of the con-
trol volume (see Figure 3.1). The two subscripts, ij, refer to the surface and direction,
respectively. In general, the stress that a surrounding fluid applies on the surface of a control
volume has two components: a normal stress (compression or tension) perpendicular to the
surface; and a shear stress parallel to the surface. The normal stress includes the pressure
perpendicular to the surface.
Convection 71
Consider the normal and shear stress components acting in the x-direction along the sur-
faces at locations x and y, respectively:
∂σ xx
Fn,x = σ xx dy; Fn,x+dx = σ xx + dx dy + · · · (3.10)
∂x
∂τyx
Fs,y = τyx dx; Fs,y+dy = τyx + dy dx + · · · (3.11)
∂y
where the subscripts n and s refer to normal and shear stresses, respectively. Higher order
terms in the Taylor series expansions can be neglected. Similar expansions are obtained for
the pressure terms.
Body forces such as buoyancy may also arise in the momentum balance. For example, in
natural convection along an inclined plate, the body force in the x-direction, Fbx, is the
force arising from buoyancy in the tangential x-direction along the plate, Fbx = ρ gx dx
dy, where gx is the gravitational component in the x-direction. Other examples of body
forces include centrifugal, magnetic, and/or electric fields. The body force acts at the center
of the control volume.
Assembling these individual expressions of the momentum fluxes and forces into
Equation 3.7 for both inflow and outflow terms, and cancelling terms using a similar proce-
dure as the previous continuity equation, yields the following x-momentum equation;
Alternatively, using the product rule on the left side and the continuity equation,
∂u ∂u ∂u ∂p ∂σ xx ∂τyx
ρ + ρu + ρv = − + + + ρFbx (3.13)
∂t ∂x ∂y ∂x ∂x ∂y
∂v ∂v ∂v ∂p ∂τxy ∂σ yy
ρ + ρu + ρv = − + + + ρFby (3.14)
∂t ∂x ∂y ∂y ∂x ∂y
These momentum equations cannot yet be solved directly in this form since there are
more unknowns than available equations. As a result, additional constitutive relations
between the stresses and velocities are required. In Newtonian fluids, the stresses are propor-
tional and linearly related to the rate of deformation (or strain rate) of a fluid element.
The proportionality constant is called the dynamic viscosity, μ. In contrast, non-Newtonian
fluids exhibit a nonlinear relationship between the fluid stress and velocity gradient. Exam-
ples of non-Newtonian fluids are Bingham plastics, Ostwald–de Waele fluids (exponential
relation), and Eyring fluids (hyperbolic relation) (see Figure 3.2).
For Newtonian fluids, it has been shown that the following two-dimensional constitutive
relations describe the normal and shear stresses in terms of the strain rate (deformation
rate, or velocity gradients):
∂u 2 ∂u ∂v
σ xx = 2μ − μ + (3.15)
∂x 3 ∂x ∂y
72 Advanced Heat Transfer
Pseudoplastic fluid
with yield value
0 τ0 Stress
FIGURE 3.2
Newtonian and non-Newtonian fluids.
∂v 2 ∂u ∂v
σ yy = 2μ − μ + (3.16)
∂y 3 ∂x ∂y
∂u ∂v
τyx = μ + = τxy (3.17)
∂y ∂x
where i = 1, 2, 3. Also, δij is the Kronecker delta (named after Leopold Kronecker, 1823–1891)
and equals 1 if i = j, and 0 otherwise. For incompressible flows, the last term in brackets on
the right side is zero.
Substituting the constitutive relations into the previous momentum equations leads to
the following two-dimensional incompressible form of the Navier–Stokes equations:
2
∂u ∂u ∂u ∂p ∂ u ∂2 u
ρ + ρu + ρv = − + μ + + ρFbx (3.19)
∂t ∂x ∂y ∂x ∂x2 ∂y2
2
∂v ∂v ∂v ∂p ∂ v ∂2 v
ρ + ρu + ρv = − + μ + + ρFby (3.20)
∂t ∂x ∂y ∂x ∂x2 ∂y2
The Euler equations represent a reduced form of the Navier–Stokes equations under the
idealized conditions of inviscid (frictionless) flow. An inviscid fluid refers to an idealized
fluid with a zero viscosity. Viscous effects are significant near a solid boundary but often
can be neglected at a sufficient distance away from the surface. The Euler equations are
obtained by neglecting the viscous and body force terms in the Navier–Stokes equations,
∂ui ∂ui ∂p
ρ + ρuj =− (3.22)
∂t ∂xj ∂xi
Inviscid fluid motion is also called a potential flow. In the absence of friction, inviscid
flows become irrotational such that the fluid vorticity, ω = ∇ × v (or curl v), is zero. The Euler
equations of inviscid flow can be further simplified in terms of a single scalar potential
function, φ(x, y). The irrotationality condition can be automatically satisfied by writing
the velocity field in terms of the potential function. For two-dimensional flows,
∂ϕ ∂ϕ
u= ; v= (3.23)
∂x ∂y
Substituting this result into the continuity equation yields Laplace’s equation, as pre-
sented earlier in Chapter 2, but in terms of the potential function. The continuity equation
for inviscid flow becomes:
∂2 ϕ ∂2 ϕ
+ =0 (3.24)
∂x2 ∂y2
Alternatively, for inviscid flows, the continuity equation can be written in terms of a
stream function, ψ(x, y), defined as follows:
∂ψ ∂ψ
u=− ; v= (3.25)
∂y ∂x
Through this definition, the stream function automatically satisfies the continuity equation.
The continuity equation can be removed from an inviscid flow analysis when this stream
function is used. The stream function is constant along each streamline (ψ = constant). At
each point along the streamline, the streamline is tangent to the velocity field. It can be shown,
based on integration of the mass flow between two adjacent streamlines, that the change in
streamline values between streamlines is the volume flow rate per unit depth between
those streamlines.
From left to right, the terms represent the energy change with time in the control volume;
energy inflow terms of advection (subscript adv), heat conduction (subscript cond), and flow
work done by pressure and viscous forces; three outflow terms; and internal heat
generation.
A similar procedure to derive the differential energy equation is adopted as the previ-
ous mass and momentum equations. The outflow terms at position x + dx are related
to inflow terms at position x based on a Taylor series expansion. For the heat conduction
and advection terms at the outflow surfaces of the control volume (position x + dx) in
Figure 3.3,
∂qx
qcond,x+dx = qx + dx dy + · · · (3.27)
∂x
V2 ∂ V2
qadv,x+dx = ρu e + + ρu e + dx dy + · · · (3.28)
2 ∂x 2
where V = (u 2 + v 2)1/2 is the total velocity magnitude. The flow stream in the advection
term carries both internal (e) and kinetic energy (V 2/2).
(ρv(e+V 2/2)+
(qy+dqy)dx (pv+d(pv))dx (σyyv+d(σyyv))dx
∂ (ρv(e+V 2/2)))dx
Advection (τyxu+d(τyxu))dx
Pressure work
FIGURE 3.3
Heat flux and work terms for a control volume.
Convection 75
The work terms of the surface forces (pressure, normal, and shear forces) along the
lower surface of the control volume, at position y, and upper surface at position y + dy,
are given by:
∂(pv)
Wp,y = pv dx; Wp,y+dy = pv + dy dx + · · · (3.29)
∂y
∂(σ yy v)
Wn,y = σ yy v dx; Wn,y+dy = σ yy v + dy dx + · · · (3.30)
∂y
∂(τyx u)
Ws,y = τyx u dx; Ws,y+dy = τyx u + dy dx + · · · (3.31)
∂y
Higher order terms in the Taylor series expansions will be neglected. Other similar terms
are constructed at the other faces of the control volume in Figure 3.3. Then each of the indi-
vidual terms are substituted into Equation 3.26 and the equation is further simplified and
rearranged. In three-dimensional tensor form, the total energy equation becomes:
∂ 1 ∂ 1 ∂qi ∂ ∂
ρ e + ui ui + ρuj e + ui ui = − − (pui ) + (τij uj ) + ρui Fb,i + q̇ (3.32)
∂t 2 ∂xj 2 ∂xi ∂xi ∂xi
where q̇ is the rate at which energy is generated per unit volume, i = 1, 2, 3 and j = 1, 2,
3. This result represents the first law of thermodynamics for a differential control volume.
It states that the rate of increase of total energy within the control volume equals the net
rate of energy input by heat conduction; plus work done by pressure, viscous, and external
forces on the control volume; and the rate at internal energy generation.
where i = 1, 2, 3 and j = 1, 2, 3. From left to right, the terms represent the total derivative of
kinetic energy of a fluid element with respect to time (temporal and convective components);
work of pressure and viscous forces on the fluid element; and the net rate at which body
forces perform work on the fluid to increase its kinetic energy within the control volume.
Using the product rule of calculus, the work of the pressure forces on the right side of
Equation 3.33 can be expanded as follows:
∂p ∂ ∂ui
ui = (pui ) − p (3.34)
∂xi ∂xi ∂xi
This term represents the net flow work done by pressure forces on the differential control
volume to increase its kinetic energy. It includes the total work done by pressure forces (first
term on the right side), minus the work done in fluid compression or expansion (second
76 Advanced Heat Transfer
term), which does not increase the kinetic energy of the fluid. The latter term becomes zero
for incompressible flows. It is an energy sink in the mechanical energy equation and there-
fore subtracted because the pressure work in the latter term is directed toward fluid com-
pression or expansion, rather than a change in the fluid’s kinetic energy.
Similarly, the work of the viscous forces on the right side of Equation 3.33 can be expanded
by the product rule as follows:
∂τij ∂ ∂ui
ui = (τij ui ) − τij (3.35)
∂xj ∂xi ∂xj
The left side is the net fluid work done by viscous stresses to increase the kinetic energy of
the fluid within the control volume. The difference on the right side represents the net rate at
which the surroundings perform work on the fluid through viscous stresses, minus the
portion of viscous work that is not transferred into kinetic energy. The last term on the right
side represents work lost through viscous dissipation, which does not change the fluid’s
kinetic energy.
Viscous dissipation is a degradation of mechanical energy into internal energy through
frictional effects. Using the constitutive relation, Equation 3.18, the viscous dissipation func-
tion, Φ, can be expressed as:
∂ui ∂ui ∂uj 2 ∂uk ∂ui
Φ = τij =μ + − δij (3.36)
∂xj ∂xj ∂xi 3 ∂xk ∂xj
Since this viscous dissipation function is a sum of squared terms, it is greater than or equal
to zero. It appears as a negative term in Equation 3.33. As expected, the conversion of
mechanical energy to internal energy through viscous dissipation is an energy sink in the
mechanical energy equation. Mechanical energy is not conserved since a portion is contin-
uously degraded and lost to internal energy through viscous dissipation. It is degraded in
the sense that the quality of thermal energy has a lower ability to perform useful work
than kinetic energy. This energy sink corresponds to an equivalent energy source term in
the internal energy equation. Both the energy sink and source terms should cancel each other
upon summation of the mechanical and internal energy equations. It can be verified that the
viscous dissipation function did not appear in the total energy equation, Equation 3.32, as
the energy sink and source terms cancelled each other.
where i = 1, 2, 3 and j = 1, 2, 3. From left to right, the terms represent the rate of thermal
energy change with time (temporal and convective derivatives); heat input by conduction;
pressure work done through fluid compression to increase the thermal energy (this term
vanishes for incompressible flows); viscous dissipation; and internal heat generation.
Using Fourier’s law in the heat conduction term, for incompressible flows, Equation 3.38
becomes:
∂e ∂e ∂ ∂T ∂ui
ρ + ρuj =k + τij + q̇ (3.39)
∂t ∂xj ∂xj ∂xj ∂xj
The viscous dissipation term is a positive energy source term that corresponds to the
energy sink previously obtained in the mechanical energy equation, Equation 3.33. The
magnitude is identical, but the sign has changed from negative to positive.
The internal energy equation may be written in a variety of other forms, including temper-
ature or enthalpy equations. Although temperature is not conserved like total energy, it is
often more useful to use a temperature form of the thermal energy equation because temper-
ature can be measured directly. Using the definitions of specific heat (Chapter 1), it can be
shown that the thermal energy equation, Equation 3.38, can be written as:
DT ∂ ∂T
ρcv =k + Φ + q̇ (3.40)
Dt ∂xj ∂xj
where D( )/Dt refers to the total (substantial) derivative, as defined in Chapter 1, including
transient and convective components. From left to right, the term on the left side represents
the transient accumulation and convective transport of thermal energy, while on the right
side, the terms represent conduction of heat, viscous dissipation, and source terms of ther-
mal energy such as from phase transition or chemical reactions.
Alternatively, using the definition of enthalpy, h = e + p/ρ, it can be shown that:
DT Dp ∂ ∂T
ρcp = βT +k + Φ + q̇ (3.41)
Dt Dt ∂xj ∂xj
For an ideal gas, the thermal expansion coefficient can be evaluated as:
1 p 1
β= = (3.43)
ρ RT2 T
spherical coordinates are shown in Appendix C. The equations may be derived in the same
manner as discussed in this section, except that the differential control volume is modified to
fit the coordinate system.
where,
y
p1 (y) = po − ρ1 (y)g dy (3.45)
yo
This definition will allow the buoyancy forces in the y-momentum equation to be written
in terms of a temperature difference through a linearized equation of state and thermal
expansion coefficient.
Control
U∞, T∞, ρ∞,
volume x
L g
Tw y
FIGURE 3.4
Isothermal object in a flow stream with forced and free convection.
Convection 79
The two-dimensional governing equations for mass, momentum (x and y directions), and
energy conservation can be written as follows:
∂u ∂v
+ =0 (3.46)
∂x ∂y
2
Du ∂pk ∂ u ∂2 u
ρ =− +μ + (3.47)
Dt ∂x ∂x2 ∂y2
2
Dv ∂pk ∂ v ∂2 v
ρ =− − ρ1 g + μ + − ρg (3.48)
Dt ∂x ∂x2 ∂y2
2
DT ∂ T ∂2 T
ρcv =k + +Φ (3.49)
Dt ∂x2 ∂y2
On the right side of Equation 3.48, the combined second and fifth terms represent the
buoyancy force. A natural convection flow arises when the density, ρ, varies with position,
y, thereby generating a body force term in Equation 3.48. The buoyancy force can be
expressed in terms of temperature using Equation 3.42 and the thermal expansion coeffi-
cient, β,
1 ρ − ρ1
β≈− (3.50)
ρ T − T1
t x y
t∗ = ; x∗ = ; y∗ = (3.51)
L/U L L
u v pk
u∗ = ; v∗ = ; p∗k = (3.52)
U U ρU2 /2
T − T1
θ= (3.53)
Tw − T1
where L is a characteristic length of the object, such as the square root of the surface area.
Also, U is the reference velocity, such as a freestream velocity. If the surrounding fluid is
motionless (U = 0), then another reference velocity is required. In free convection problems,
it can be shown that a scaling approximation of the momentum equation leads to a reference
velocity of Uref = (gβΔTL)1/2. Alternatively, using a timescale of t = L 2/α from conduction
heat transfer, another possible reference velocity is Uref = k/(ρLcp). The reference values
for the dimensionless variables should have approximately the same order of magnitude
as typical corresponding dimensional variables in the problem.
Rewriting the governing conservation equations in terms of the dimensionless variables,
∂u∗ ∂v∗
+ =0 (3.54)
∂x∗ ∂y∗
80 Advanced Heat Transfer
Du∗ ∂p∗k 1 ∂2 u∗ ∂2 u∗
= − + + (3.55)
Dt∗ ∂x∗ Re ∂x∗2 ∂x∗2
Dv∗ ∂p∗k 1 ∂2 v∗ ∂2 v∗ Gr
=− ∗+ + + θ (3.56)
Dt∗ ∂y Re ∂x∗2 ∂y∗2 Re2
2 ∗
DT ∗ 1 ∂ T ∂2 T ∗ Ec ∗
∗
= ∗2
+ ∗2
+ Φ (3.57)
Dt Re · Pr ∂x ∂y Re
where Re, Gr, Ec, and Pr refer to the Reynolds, Grashof, Eckert, and Prandtl numbers, respec-
tively. These numbers are named after Osborne Reynolds (1842–1912), Franz Grashof
(1826–1893), Ernst Eckert (1904–2004), and Ludwig Prandtl (1875–1953). Table 3.1 shows
the definitions of these nondimensional parameters as well as others which appear fre-
quently in heat and mass transfer problems.
Based on the form of these nondimensional equations, it can be observed that the momen-
tum and energy equations are coupled and the temperature solution has the following func-
tional form:
In the steady state, the time dependence vanishes, and also the dependence on spatial
coordinates vanishes when averaged quantities are determined by spatial integration
over the surface of the object.
The convective heat transfer coefficient, h, is commonly expressed in terms of the Nusselt
number, Nu, as follows:
hL (q′′ /ΔT)L
NuL = = (3.59)
k k
where n is the coordinate normal to the surface, and w refers to the wall. Therefore, the Nus-
selt number represents the nondimensional temperature gradient at the wall.
The average heat transfer coefficient is determined by integrating the local Nusselt num-
ber over the object’s surface, S,
Nu · dS
Nu = s
= Nu (Re, Pr, Gr, Ec) (3.61)
s
dS
Spatial integration and steady-state conditions allow the spatial and temporal dependen-
cies to be removed. Depending on the type of flow problem, certain dimensionless param-
eters can be dropped from the functional relationship if the physical processes underlying
the parameters are not relevant to the particular physical problem. For example, buoyancy
forces are insignificant for forced convection in high-speed flows so the Grashof number
dependence can be dropped.
Convection 81
TABLE 3.1
Selected Nondimensional Parameters of Heat and Mass Transfer
Group Definition Interpretation
Biot number (Bi); Jean-Baptiste Biot hL=ks Ratio of conduction resistance of the solid to
(1774–1862) the boundary layer resistance of the fluid
Coefficient of friction (cf) τw=(ρV 2=2) Nondimensional wall shear stress
Eckert number (Ec); Ernst Eckert V 2=(cpΔT ) Ratio of kinetic energy of the flow to the
(1904–2004) boundary layer enthalpy difference
Fourier number (Fo); Jean-Baptiste αt=L 2 Nondimensional time, or the ratio of the rate
Fourier (1768–1830) of heat conduction to thermal energy
storage in the solid
Grashof number (Gr); Franz Grashof gβ(Ts T∞)L 3=ν 2 Ratio of buoyancy to viscous forces
(1826–1893)
Colburn j number (jh); Allan Colburn St · Pr 2=3 Nondimensional heat transfer coefficient
(1904–1955)
Colburn j number (jm) Stm · Pr 2=3 Nondimensional mass transfer coefficient
Jacob number (Ja); Max Jakob (1879–1955) cpΔT=hfg Ratio of sensible to latent energy during
liquid–vapor phase change
Lewis number (Le); Warren Lewis α=D Ratio of thermal and mass diffusivities
(1882–1975)
Nusselt number (Nu); Wilhelm Nusselt hL=kf Nondimensional temperature gradient at the
(1882–1957) surface
Peclet number (Pe); Jean Péclet VL=α Heat transfer parameter characterizing the
(1793–1857) ratio of convection and diffusion effects
Prandtl number (Pr); Ludwig Prandtl ν=α Ratio of momentum and thermal diffusivities
(1875–1953)
Rayleigh number; John Rayleigh gβ(Ts T∞)L 3=να ¼ Ratio of buoyancy to viscous forces,
(1842–1919) Gr · Pr multiplied the ratio of momentum and
thermal diffusivities
Reynolds number (Re); Osborne VL=ν Ratio of inertial and viscous forces
Reynolds (1842–1912)
Schmidt number (Sc); Heinrich Schmidt ν=D Ratio of momentum and mass diffusivities
(1892–1975)
Sherwood number (Sh); Thomas hmL=D Nondimensional concentration gradient at
Sherwood (1903–1976) the surface
Stefan number (Ste); Jozef Stefan cp ΔT=hsf Ratio of sensible to latent energy during
(1835–1893) solid-liquid phase change
Stanton number (St); Thomas Stanton h=(ρVcp) ¼ Nu=Re · Pr Modified Nusselt number
(1865–1931)
Weber number; Moritz Weber ρV 2L=σ Ratio of inertia to surface tension forces
(1871–1951)
Based on the functional form of the Nusselt number in Equation 3.61, the relevant dimen-
sionless parameters for different cases of convection problems at steady state can be summa-
rized as follows.
Other dimensionless groups in Table 3.1 arise when other physical processes occur, such
as phase change, mass transfer, etc.
To demonstrate this procedure, consider the same example that was previously consid-
ered—convective heat transfer from an isothermal object in a surrounding flow stream. In
Step 1, all of the problem parameters are listed. Seven variables (k = 7) characterize a forced
convection problem: ρ, k, cp, μ, U, h, and L. These k parameters are then written in terms of r
primary dimensions. There are 4 primary dimensions (r = 4): time, mass, length, and tem-
perature (Step 2). Therefore, the number of pi terms is k − r = 3 (Step 3). Four independent
and repeating variables are then selected (Step 4): ρ, L, U, and k. The first pi term in Step 5 is
obtained by multiplying a non-repeating variable (h) by the product of repeating variables,
each raised to an exponent that makes the entire combination dimensionless.
a
kg b kg c kg · m d kg . 0 0 0 0
π 1 = (ρ L μ k )h =
a b c d
(m) = kg m s K (3.62)
m3 m·s s3 K s3 K
U, T∞ T∞ Temperature
U
Velocity boundary layer
boundary
δ(x) δ t(x)
layer Turbulent region
y
Buffer layer
Laminar sublayer
x xc L
Plate Wall temperature, Tw
Laminar Transition Turbulent
FIGURE 3.5
Velocity and thermal boundary layers on a flat plate.
about 3 cm at the end of the surface. This boundary layer growth displaces fluid momentum
in the y-direction.
The structure of a flat plate boundary layer involves a laminar portion from the leading
edge up to a transition point, xc (at Rex = Ux/ν ≈ 5 × 105), followed by a transition region
and turbulence thereafter. Laminar flow occurs when the fluid moves in parallel layers with-
out disruption between the layers. This typically occurs at low velocities where the fluid
moves without lateral mixing. In contrast, turbulence occurs at higher velocities when the
fluid motion is characterized by chaotic changes in pressure and flow velocity. Following
transition to turbulence, the enhanced mixing in the fluid leads to an increase of cross-stream
velocity fluctuations and the turbulent boundary layer grows more rapidly. There are three
distinct regions within the turbulent boundary layer—an inner viscous sublayer (where
molecular diffusion is dominant); an overlap or buffer layer; and an outer layer where tur-
bulence effects are dominant.
To formulate the governing equations, a number of simplifying boundary layer assump-
tions will be used. Two-dimensional incompressible flows under steady-state conditions
with constant thermophysical properties will be assumed. It is further assumed that the
velocity component, v, perpendicular to the wall (y-direction) is much smaller than the
velocity component, u, in the streamwise (x) direction. Also, since the boundary layer is nor-
mally very thin, it is assumed that spatial gradients of a flow quantity in the streamwise
direction are much less than the cross-stream direction, ∂/∂x ≪ ∂/∂y.
The pressure gradient in the x-direction can be determined based on the inviscid flow dis-
tribution outside the boundary layer from Euler’s equation. For a negligible velocity, v, in
the y-momentum equation, the y-direction pressure gradient becomes zero. Thus the pres-
sure inside the boundary layer at any position, x, matches the pressure at the same position,
x, outside the boundary layer in the inviscid freestream. Applying Equation 3.22 under
steady-state conditions, the x-momentum Euler’s equation becomes:
∂U ∂p
ρU =− (3.66)
∂x ∂x
where U(x) refers to the freestream velocity distribution (assumed to be known for a spec-
ified surface geometry). This expression for the x-direction pressure gradient will be used in
the x-direction momentum equation.
With these above boundary layer assumptions, the two-dimensional continuity, momen-
tum, and thermal energy equations become:
∂u ∂v
+ =0 (3.67)
∂x ∂y
Convection 85
∂u ∂u ∂U ∂2 u
ρu + ρv = ρU +μ 2 (3.68)
∂x ∂y ∂x ∂y
2
∂T ∂T ∂2 T ∂u
ρcp u + ρcp v =k 2 +μ (3.69)
∂x ∂y ∂y ∂y
∂u∗ ∂v∗
+ =0 (3.70)
∂x∗ ∂y∗
2 ∗
∂u∗
∗ ∗ ∂u
∗
∂p∗ 1 ∂ u
u +v =− ∗+ (3.71)
∂x∗ ∂y∗ ∂x Re ∂y∗2
2 ∗
∂T ∗ ∗ ∂T
∗
1 ∂T
u∗ ∗
+ v ∗
= (3.72)
∂x ∂y RePr ∂y∗2
These governing equations are subject to appropriate boundary conditions, such as a zero
velocity at the wall (no-slip condition) and a specified wall temperature. The dimensionless
parameters in the boundary layer equations are the Reynolds number (Re = UL/v) and
the Prandtl number (Pr = v/α). Recall that the Reynolds number represents a ratio of inertial
to viscous forces while the Prandtl number characterizes the ratio of momentum (viscous) to
thermal diffusivities.
The functional form of the solution of the boundary layer equations becomes:
∗ ∗ ∗ ∗ dp∗
u = u x , y , Re, ∗ (3.73)
dx
dp∗
T ∗ = T ∗ x∗ , y∗ , Re, Pr, ∗ (3.74)
dx
The dependence on the pressure gradient in the temperature relationship arises because of
the dependence of temperature on the velocities, which in turn depend on the pressure. The
pressure gradient is assumed to be known outside of the boundary layer for a given surface
geometry, based on a potential flow solution of the Euler’s equation outside the boundary
layer. For example, the pressure gradient is zero for a flat plate boundary layer due to a cons-
tant freestream velocity.
The slope of the velocity profile in the boundary layer at the wall is related to the shear
stress acting on the wall. As discussed earlier in Chapter 1, the shear stress for flat plate
boundary layer flow of a Newtonian fluid is represented by:
∂u
τw (x) = μ (3.75)
∂y0
The local skin friction coefficient, cf, is determined based on this wall shear stress,
τw (x)
cf (x) = (3.76)
ρU2 /2
86 Advanced Heat Transfer
where U refers to the freestream velocity (outside the boundary layer). Rewriting the skin
friction coefficient in terms of dimensionless variables, and observing the previous func-
tional form of the nondimensional velocity, implies:
2 ∂u∗ 2
cf = ∗
= f1 (Re) (3.77)
Re ∂y 0 Re
Once cf is known, the drag force on the surface due to friction can be obtained by integrat-
ing the local skin friction coefficient over the surface.
Similarly, within the thermal boundary layer, the local heat flux based on an energy bal-
ance at the wall is given by:
∂T
q′′w (x) = −kf = h(x)(Tw − T1 ) (3.78)
∂y w
where the subscripts w, f, and ∞ refer to the wall, fluid, and freestream, respectively.
Here the local convection coefficient, h(x), can be expressed in terms of the wall temperature
gradient, fluid conductivity, and temperature difference between the wall and the fluid.
Rewriting the energy balance in terms of the Nusselt number and dimensionless variables,
∗
hL ∂T
Nu = = = f2 (Re, Pr) (3.79)
kf ∂y∗ 0
where L is the plate length and the functional dependence on Re and Pr was obtained based
on the previously derived form of the nondimensional temperature.
Various methods can be used to determine the above functional dependencies, f1 and f2, of
the skin friction coefficient and Nusselt number, respectively, on the Reynolds and Prandtl
numbers. In upcoming sections, analytical solutions will be used to determine these func-
tional dependencies over a range of flow conditions. Alternatively, experimental methods
are commonly used to develop empirical correlations of the Nusselt number based on
physical measurements.
For example, consider a cold airstream which flows at a temperature of T∞ across a heated
plate. An experiment is performed with a heater embedded in the surface to determine the
heat transfer coefficient (see Figure 3.6). An electrical current is supplied through the heater
in order to maintain a constant surface temperature, Ts.
y Surface
U∞, T∞ temperature, Ts
Electrial
heating
Insulation
x x=L
FIGURE 3.6
Experimental setup for convective heat transfer coefficient measurements.
Convection 87
Based on an energy balance on the plate, the rate of heat transfer by convection to the air,
under steady-state conditions, will balance the electrical heat supplied within the plate.
The average convective heat transfer coefficient, h, can be determined based on the
applied electrical current, I, electrical resistance, R, and temperature difference between
the surface and the airstream. Several experiments are performed at various incoming veloc-
ities and working fluids, after which the measured values are combined into dimensionless
groups (Nu, Re, Pr) with a curve fitted correlation of the following form:
Nu = CRem
L Pr
n
(3.81)
For example, after performing several experiments with air, water, and oil, and also
changing the incoming velocities, the curve fits would indicate that n ≈ 1/3 and m ≈ 1/2
for laminar forced convection, but m ≈ 4/3 for turbulent forced convection.
Similar experiments and correlations can be obtained for mass transfer problems, but the
Prandtl and Nusselt numbers are replaced by the Schmidt number (Sc = v/DAB) and Sherwood
number (Sh = hm L/DAB), where DAB and hm refer to the diffusion and convective mass trans-
fer coefficients, respectively. A typical form of these heat and mass transfer correlations is
illustrated in Figure 3.7.
As shown in Figure 3.7, individual curves are obtained for each type of fluid, character-
ized by a distinct value of Pr (or Sc; mass transfer). The heat transfer coefficient increases
with Reynolds number as a result of the thinner boundary layer and steeper temperature
gradient at the wall at higher velocities. By combining both the Nusselt and Prandtl num-
bers on the vertical axis, the multiple curves can collapse into a single curve. Collapsing
the data in this way is a useful result that correlates a wide range of parameters in a
single curve.
NuL = CReLm Pr n
NuL
NuL = CReLm
Prn
Log (NuL) Pr3 Log
Pr n
Pr2 ShL
Log (ShL)
ShL = CReLm Scn Log ShL
Pr1 Pr n = CReLm
Scn
FIGURE 3.7
Nondimensional correlations for heat and mass transfer. (Adapted from T.L. Bergman et al. 2011. Fundamentals of
Heat and Mass Transfer, 7th Edition, New York: John Wiley & Sons.)
88 Advanced Heat Transfer
resulting nondimensional velocity and temperature solutions become identical too. Then f1
in Equation 3.77 and f2 in Equation 3.79 also become identical, so equating both functional
expressions, leads to the following Reynolds analogy between momentum and heat transfer
(with extension by analogy to mass transfer):
c
f
Re = Nu = Sh (3.82)
2
Thus, for boundary layer flows with a zero pressure gradient (flat plate) and Pr = 1, any
analytical, computational, or experimental results from the fluid mechanics problem can be
related (through the Reynolds analogy) to the corresponding heat and mass transfer
problems.
Alternatively, in terms of the Stanton numbers,
cf
= St = Stm (3.83)
2
Nu
St = (3.84)
RePr
Sh
Stm = (3.85)
ReSc
Even in cases with a pressure gradient in the freestream, the Reynolds analogy is still often
used since the results remain within the bounds of typical experimental errors associated
with heat transfer measurements. Corrections to account for cases where Pr ≠ 1 and Sc ≠ 1
can be determined through the following Chilton–Colburn analogy:
cf
= StPr1−n = Stm Sc1−n (3.86)
2
which is applicable for Pr ≠ 1 and Sc ≠ 1. A value of n = 1/3 is commonly assumed. The mid-
dle and last expressions in Equation 3.86 are called the Colburn j factors, jH and jm, for heat
and mass transfer, respectively. The Colburn factors are also widely used in the design
and analysis of heat exchangers (Chapter 9).
Alternatively, the Chilton–Colburn analogy, Equation 3.86, can be written as one convec-
tion coefficient, such as hm, in terms of the other coefficient, h, as follows:
hL/k hm L/DAB
= (3.87)
Pr1/3 Sc1/3
Alternatively, this relationship can be rearranged to express h/hm in terms of the Lewis
number, Le, where Le = α/DAB = Sc/Pr. Once the appropriate form of a correlation is deter-
mined, experimental data can be gathered and grouped based on this functional form to
reduce the number of independent parameters under investigation.
Convection 89
where hfg and m′′A refer to the latent heat of vaporization and evaporative mass flux of
component A of a mixture (such as water vapor in the previous example), respectively.
Under steady-state conditions, the latent energy lost by the liquid due to evaporation bal-
ances the rate of heat transfer from the surrounding gas to the liquid. For example, if heat
is supplied by convection, the heat gain balances the evaporative heat loss,
q′′evap = hfg hm (ρA (Ts ) − ρA,1 ) = h(T1 − Ts ) (3.89)
where hm, ρA(Ts), ρA∞, T∞ and Ts refer to the convective mass transfer coefficient, vapor den-
sity at the liquid surface, ambient vapor density, ambient temperature, and liquid surface
temperature, respectively.
Using the Chilton–Colburn analogy, Equation 3.87, to express the ratio of convective heat
to mass transfer coefficients, as well as the ideal gas law,
hfg pA,sat (Ts ) pA,1
T 1 − Ts = − (3.90)
ρcp Le2/3 RTs RT1
where Le refers to the Lewis number and the thermophysical properties (ρ and cp) are eval-
uated at the mean temperature of the thermal boundary layer.
The gas constant, R, is evaluated for the particular constituent, for example, water vapor.
Alternatively, the gas constant can be written in terms of the universal gas constant (Ru =
8.315 kJ/kmol K) divided by the molecular weight of the constituent of interest. Replacing
Ts and T∞ by an average temperature, Tav, the temperature difference can be rewritten as:
MA hfg
T 1 − Ts = (pA,sat (Ts ) − pA,1 ) (3.91)
ρcp Ru Tav Le2/3
where MA is the molecular weight of constituent A in the mixture. Thus, the evaporative
heat flux can be written as,
q′′evap = ωevap (pA,sat (Ts ) − pA,1 ) (3.92)
where,
MA h · hfg
ωeυap = (3.93)
ρcp Ru Tav Le2/3
90 Advanced Heat Transfer
A less accurate (but more convenient) approximation is obtained by assuming that the
partial pressure ( pA∞) is negligible relative to the saturation pressure ( pA,sat).
Similarly to evaporative cooling, sublimation is a phase change process where a material
is transformed from the solid to vapor phase without passing through a liquid state. All
solid materials will sublime below the triple point (the thermodynamic state where solid, liq-
uid, and gas phases all coexist in equilibrium). Heat transfer by sublimation can be calcu-
lated similarly as the previous analysis for evaporative cooling, except that the latent heat
of vaporization, hfg, is replaced by the latent heat of sublimation, hsg. The sublimation
heat flux is calculated by replacing ωevap with ωsub and replacing hfg with hsg.
Sublimation occurs in applications such as dry ice (solid carbon dioxide that is used as a
refrigerant for transporting perishables), antiseptics, fungicides, and light-sensitive materi-
als (silver iodide) in photography, as well as the production of dyes. Iodine forms black crys-
tals that readily sublime to violet vapor. In freeze drying, sublimation of ice occurs from
frozen foods under a vacuum in order to retain their texture and flavor.
∙ The order of magnitude of a sum of terms, each having the same order of magni-
tude, is the same as the order of magnitude of each individual term.
∙ The order of magnitude of a product (or quotient) of terms is the same as the product
(or quotient) of the orders of magnitude of the individual terms.
For example, consider external flow over a flat plate subject to the boundary layer equa-
tions of the previous section and the schematic in Figure 3.5. The quantities δ, L, vo, and U
will be used as the relevant characteristic scales for the boundary layer thickness, plate
length, cross-stream reference velocity, and freestream velocity, respectively.
By substituting the scaled quantities into the continuity equation, Equation 3.67, the
approximate order of the cross-stream velocity component becomes vo ≈ δU/L. Then substi-
tuting the scaled quantities into the x-momentum equation, Equation 3.68, leads to:
U δ U U
ρU + ρ U ≈μ 2 (3.94)
L L δ δ
L
δ ≈ √ (3.95)
Re
where Re = ρUL/μ is the Reynolds number. The symbol “≈” means an equivalent order of
magnitude only and the results cannot be relied upon with respect to the leading numerical
coefficients. More accurate methods of analysis are required to determine the leading coeffi-
cients and precise numerical values.
Using these scaled quantities, the order of magnitude of the skin friction coefficient
becomes:
τw μU/δ 2
cf = ≈ = √ (3.96)
ρU /2 ρU /2
2 2 ReL
In an upcoming section, an exact solution method yields the same functional form,
except that the coefficient in the numerator is 1.328 (rather than 2). The approximate scal-
ing analysis agrees fully with the functional form and order of magnitude of the skin fric-
tion coefficient, but the precise values of numerical coefficients require a more detailed
analysis.
The following Leibnitz rule is frequently used in the integral solution method:
b(x) b(x)
∂ ∂f (x, y) db da
f (x, y)dy = dy + f (b, y) − f (a, y) (3.97)
∂x a(x) a(x) ∂x dx dx
In convection problems, the variables a(x) and b(x) typically refer to a spatial range of
coordinates, or range of time, while f(x, y) represents a term in the governing equation.
Property variations in one direction may be well known, but changes in the other direc-
tion(s) are unknown. Therefore the integral method considers a control volume of finite
size in the direction of the known property variations and then substitutes this distribu-
tion during the integration procedure. By reducing the number of variables through the
integration process, a partial differential equation can be reduced to an ordinary
differential equation.
Consider again a flat plate boundary flow in Figure 3.5. The governing continuity and
momentum equations, Equations 3.67 and 3.68, are integrated across the boundary layer
thickness (δ) to find the spatial variations of dependent variables along the plate. Integrating
the continuity equation across the boundary layer from y = 0 → δ and using Leibnitz’s rule,
Equation 3.97,
δ
∂ ∂δ
ρu dy − ρU + ρvδ − ρvo = 0 (3.98)
∂x 0 ∂x
δ
∂ ∂δ ∂ δ ∂δ ∂u ∂u
ρuu dy − ρU2 + ρUvδ − ρUvo = − p dy + p + μ −μ (3.99)
∂x 0 ∂x ∂x 0 ∂x ∂y δ ∂y 0
The result from the integrated continuity equation can be substituted for the second
and third terms on the left side. On the right side, the velocity gradient at the edge of
the boundary layer becomes zero in the third term. The last term is the wall shear
stress, τw. Also, applying Euler’s equation, Equation 3.22, at y = δ (inviscid freestream)
allows the pressure gradient to be rewritten in terms of the freestream velocity, U,
yielding:
δ δ
d dU
ρu(U − u)dy+ ρ(U − u)dy = τw (3.100)
dx 0 dx 0
At this stage, an assumed profile for the velocity, u(y), must be provided across the boun-
dary layer in the y-direction so that the above integration can be completed. The resulting
equation will be an ordinary differential equation to be solved in the x-direction subject
to appropriate boundary conditions.
Convection 93
This equation can be solved to give the boundary layer thickness, which is then substi-
tuted into Equation 3.101 to establish the cubic velocity profile, and then differentiated at
the wall to determine the wall shear stress. Differentiating the velocity profile and evaluat-
ing at y = 0 yields:
τw 0.323
= √ (3.103)
ρU 2 Rex
In the next section, an exact solution produces this same result, except that the coefficient
0.323 is replaced by 0.332. Therefore, the integral analysis agrees within 3% of the exact sol-
ution. The correct functional dependence of the wall shear stress on the Reynolds number is
obtained. A similar procedure, involving the energy equation, can be used to determine the
heat flux from the plate.
Many problems of practical and industrial relevance can be analyzed effectively by the
integral method. For example, a thermally buoyant plume leaving the top of an industrial
stack can be examined with a control volume of finite width in the cross-stream direction
(assuming a symmetrical velocity profile) and resulting ordinary differential equation in
the axial direction. In upcoming chapters, applications of the integral method to various
other types of flows will be presented. For example, in two-phase (liquid–gas) flow in a
pipe, an integral analysis can be used for a control volume of finite width (pipe diameter)
in the cross-stream direction and resulting differential equation in the flow direction. The
integral analysis would require an assumed radial velocity distribution in order to find
the remaining profiles in the flow direction. The resulting temperature profile would then
be used to find the wall heat flux and Nusselt number.
River
Boundary layer Wall of channel edge
Inflow
y
x Wall
FIGURE 3.8
Similarity characteristics of various typical flows.
in the channel. If the y coordinate is stretched at each location x by a factor g(x), then all sol-
ution curves over the y direction would collapse onto a single curve. In other words, the sol-
ution becomes a function of a single variable, y · g(x), alone. This feature is a common
characteristic of self-similar flows. In a similar way, the free jet exhibits self-similarity in
terms of its flow structure. In contrast, self-similarity is not sustained with diverse flow
behavior in a river. The propagation of disturbances is not carried in a single flow direction.
The existence of similarity solutions may be determined from group theory. Group theory is
a powerful mathematical method for analyzing abstract and physical systems in which
some form of symmetry is present. It is based on invariance of the governing equations
under a group of transformations, where the same functional form of equations is retained
through the transformation. Examples of invariant transformations include stretching, rota-
tion, or translation of coordinates. A self-similar solution exists for flat plate boundary layer
flow because the governing equations and boundary conditions are invariant under the
transformations of x → a 2x, y → ay, u → u, and v → v/a. Similarity may be present whenever
one independent coordinate in the solution domain exists. As a result, the physical transport
processes will follow this single direction. Alternatively, if downstream disturbances do not
significantly affect the upstream flow structure, then also the flow profiles in the flow direc-
tion are usually self-similar.
Consider a similarity solution for the velocity and temperature profiles in boundary layer
flow along a flat plate. The solution procedure begins by assuming that a similarity profile
can be obtained. If this assumption is incorrect, then the analysis and equations would even-
tually indicate the incompatibility. Recall the steady incompressible boundary layer equa-
tions over a flat plate, Equations 3.67 through 3.69, are given by:
∂u ∂v
+ =0 (3.104)
∂x ∂y
2
∂u ∂u ∂p ∂u
ρu + ρv = − + μ (3.105)
∂x ∂y ∂x ∂y2
2 2
∂T ∂T ∂ T ∂u
ρcp u + ρcp v =k +μ (3.106)
∂x ∂y ∂y2 ∂y
Convection 95
u
= f ′ (η) (3.107)
u1
η = y · g(x) (3.108)
The derivative of the function, f, is used rather than the function itself since the analysis
will later involve integrations of the function. If the assumption in Equation 3.107 is incor-
rect, then the resulting analysis and equations will lead to some form of incompatibility. The
governing equations will be transformed from (x,y) to (x, η) coordinates.
Define the stream function, ψ, as follows:
∂ψ ∂ψ
u= ; v=− (3.109)
∂x ∂y
It can be observed that this definition allows the continuity equation to be removed from
the analysis since the stream function automatically satisfies the continuity equation by
its definition. Integrating Equation 3.109 and applying a no-slip boundary condition at
the surface,
where,
η = y · g1 (x) (3.111)
where dp/dx = 0 has been assumed for flat plate boundary layer flow. This result requires
that both sides equal a constant, denoted by C1, since the left side is a function of η alone,
whereas the right side is a function of x alone. If the original similarity assumption in Equa-
tion 3.107 was incorrect, then a separable constraint, Equation 3.112, could not be achieved.
Thus, the separability of Equation 3.112 indicates that flat plate boundary layer flows are
indeed self-similar with each other in the flow direction (as expected).
96 Advanced Heat Transfer
Equating both sides of Equation 3.112 with a constant, and solving subject to the boundary
conditions, yields the following Blasius equation:
1
f ′′′ (η) + f (η)f ′′ (η) = 0 (3.113)
2
where,
U
η=y (3.114)
vx
subject to,
f ′ (η 1) = 1 (3.115)
This system can be readily solved by numerical integration methods, such as the Runge–
Kutta method. Sample results are shown in Table 3.2.
The result for u/U is obtained by integrating Equation 3.113 twice.
η
η
u ′ exp − 0 f (η)dη) /2 dη
= f (η) =
1
η
0
(3.117)
U 0
exp − 0 f (η)dη) /2 dη
Based on this result, the boundary layer thickness can be found where the velocity reaches
99% of the freestream value, as follows:
5x
δ(x) = √ (3.118)
Rex
TABLE 3.2
Similarity Functions for Flat Plate Boundary Layer Flow
η f(η) f 0 (η) f 00 (η)
Also, the wall shear stress and skin friction coefficient can be determined,
∂u ∂f ′ ∂η 0.332ρU2
τw (x) = μ = μU = √ (3.119)
∂y y=0 ∂η ∂y y=0 Rex
τw (x) 0.664
cf (x) = = √ (3.120)
ρU2 /2 Rex
Similarly, for the energy equation, Equation 3.106 is also rewritten in terms of the similar-
ity variables, (x, η). Define the nondimensional temperature as:
T − Tw
θ(η) = (3.121)
T1 − Tw
Following a similar procedure as the momentum equation, the boundary layer energy
equation becomes:
Pr ′
θ′′ (η) + θ (η)f (η) = 0 (3.122)
2
subject to θ(0) = 0 and θ(η → ∞) = 1. Viscous dissipation effects can be neglected in the
energy equation except for highly viscous liquids or high speed flows. Since Equation
3.113 is decoupled from Equation 3.122, the solution of the Blasius equation for f(η) can be
substituted into Equation 3.122 to determine the temperature distribution. Then the result
for θ(η) is obtained by isolating θ’(η) and integrating Equation 3.122 twice to obtain:
η
η
exp −Pr 0 f (η)dη /2 dη
θ(η) =
10
η (3.123)
0
exp −Pr 0 f (η)dη /2 dη
where Pr is the Prandtl number. It can be observed that θ rises to θ∞ faster as Pr → ∞ (com-
pared with Pr ≪ 1) since the thermal boundary thickness is small relative to the velocity
boundary layer thickness at large Prandtl numbers.
Then the thermal boundary layer thickness, δt, wall heat flux, and local Nusselt number
can be determined as follows:
δt (x)
= Pr−1/3 (3.124)
δ(x)
∂T u1
q′′w (x) ′
= −k = kΔTθ (0) (3.125)
∂y 0 vx
hx (q′′ w /ΔT)x
Nux = = = θ′ (0)Re1/2
x (3.126)
k k
From the temperature results in Equation 3.123, it can be shown that the slope of the
dimensionless temperature profile varies with Pr according to θ’(0) ∼ 0.332 Pr 1/3. Therefore,
Nux = 0.332Re1/2
x Pr
1/3
(3.127)
98 Advanced Heat Transfer
TABLE 3.3
Correlations for Laminar and Turbulent Flow over a Flat Plate
Skin Friction Coefficient Nusselt Number Correlation
Boundary Layer Thickness τw hx
cf ¼ Nux ¼
Conditions δ ¼ δ(x) ρU 2 =2 k
Laminar δ(x) ¼ 5x Rex 1=5 cf (x) ¼ 0.664 Rex 1=2 Nu(x) ¼ 0.332 Rex 1=2 Pr 1=3
(Rex , 5 105) (Pr . 0.6)
Turbulent δ(x) ¼ 0.37x Rex 1=5 cf (x) ¼ 0.059 Rex 1=5 Nu(x) ¼ 0.0296 Rex 4=5 Pr 1=3
(Rex 5 105) (0.6 , Pr , 60)
where x refers to the position along the plate. This similarity solution agrees closely with
experimentally measured heat fluxes for flat plate boundary layer flows.
The results for the boundary layer thickness, skin friction coefficient, and Nusselt number
in flat plate boundary layer flow are summarized in Table 3.3. Results for laminar flow and
turbulent flow (Schlichting 1979) are listed. The wall shear stress and convective heat trans-
fer coefficient, h, decrease proportionally to x −1/2 in the positive x-direction (laminar
regime). As expected, the growth of the boundary layer leads to a decreasing velocity
and temperature gradient at the wall in the positive x-direction, thereby reducing the
wall shear stress and convection coefficient. In the turbulent flow region, τw and h increase
abruptly, but then start decreasing as x −1/5 when the velocity and temperature gradients at
the wall decrease upon growth of the boundary layer.
Since the boundary layer usually has both laminar and turbulent parts, the average heat
transfer coefficient over the entire plate length, L, is obtained by integrating both laminar
and turbulent correlations over the appropriate range,
L
xc
hL = 1 hlam (x)dx + hturb (x)dx (3.128)
L 0 xc
where xc refers to the critical point (transition to turbulence). The average Nusselt number
correlations are obtained after performing this integration (see Table 3.4). The laminar flow
results assume that more than 95% of the plate has laminar flow and similarly for turbulent
flow. The mixed-flow correlation uses a combination of laminar and turbulent flow correla-
tions when the transition to turbulence occurs past the front 10% or before the last 5% of
the plate.
In the previous similarity analysis, a particular form of similarity variable was initially
assumed. But finding a suitable similarity variable at the beginning of a similarity analysis
is usually difficult since the form of the variable is unknown. The method of unknown coefficients
may be used to determine a suitable function(s). In this method, a general similarity variable
TABLE 3.4
Average Heat Transfer Correlations for a Flat Plate Boundary Layer
Conditions Flow Heat Transfer
is defined based on a product of all independent problem variables, such as x and y, each
raised to an exponent (an unknown coefficient). The specific values of the unknown coeffi-
cients are then determined by proceeding in a manner similar to the previous example, but
instead by imposing constraints on the unknown coefficients in order to eliminate their
dependence on the independent variables in the boundary conditions. This approach can
provide a more general standard procedure for finding the similarity variables.
Laminar Transition to
boundary turbulence Turbulent
layer boundary layer
θsep θsep
U, T∞ U, T∞
Wake
Cylinder
Boundary layer
separation Separation
ReD ≤ 2 × 105 ReD > 2 × 105
FIGURE 3.9
Effects of separation and turbulence on external flow over a cylinder.
100 Advanced Heat Transfer
1000
100
10
Smooth cylinder
Cd
No separation Sphere
0.1
0.01
0.1 1 10 100 1,000 10,000 100,000 1,000,000
Re = VD/ν
FIGURE 3.10
Drag coefficient for a smooth circular cylinder and sphere in cross flow. (Adapted from H. Schlichting. 1979. Boun-
dary Layer Theory, New York: McGraw-Hill.)
The Nusselt number decreases with circumferential angle, θ, up to the separation point, as
the growing boundary layer thickness reduces the near-wall temperature gradient. Beyond
the separation point, NuD increases with θ since flow separation causes local flow reversal,
stronger mixing, and thus enhanced heat transfer. In Figure 3.11, similar trends are observed
for the cases of air flow past a long cylinder and a sphere. On the back side of the cylinder
between 90 and 140◦ , NuD decreases with θ since the increasing thickness of the boundary
layer reduces the near-wall temperature gradient and wall heat flux. Beyond the separation
point at about 140◦ , NuD also increases as a result of the enhanced fluid and thermal mixing
in the separated, recirculating flow.
800 100
Re = 219,000
186,000 80
600 Re = 150,000
170,000
120,000
60 89,000
h, W/m2K
58,000
Nuθ
400 44,000
40
Re =101,300
200
20
Re = 70,800
0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 180
Circumferential angle, θ Circumferential angle, θ
FIGURE 3.11
Heat transfer from a cylinder and sphere in cross flow. (Adapted from W.H. Giedt. 1949. Transactions of the ASME,
71: 375–381; G.F. Hewitt et al. Eds., 1997. International Encyclopedia of Heat and Mass Transfer, Boca Raton: CRC Press/
Taylor & Francis.)
Convection 101
TABLE 3.5
Zhukauskas Correlation Coefficients for External Flow Past a Circular Cylinder
ReD C m n (Pr 10) n (Pr . 10)
Based on experimental data for flow past a cylinder over a wide range of Reynolds num-
bers, the following Zhukauskas correlation can be used (Zhukauskas 1972; see Table 3.5).
1/4
Pr
NuD = CRem
D Pr
n
; 0.7 , Pr , 500 (3.129)
Prs
In Equation 3.129 and Table 3.5, all thermophysical properties are evaluated at the free-
stream temperature, except the Prandtl number, Prs, which is evaluated at the surface tem-
perature, Ts. Due to the inherent instability and unsteadiness of vortex shedding from the
cylinder, measurement uncertainties, and other factors, Equation 3.129 has an accuracy of
approximately +20%. The Zhukauskas correlation provides general applicability over a
wide range of Reynolds and Prandtl numbers. The reader is referred to other sources for
additional correlations, for example, Kakac et al. (2013), Bergman et al. (2011).
3.6.1 Sphere
External flow past a sphere exhibits similar characteristics as flow past a circular cylinder in
the previous section. Boundary layer separation and transition to turbulence involve similar
processes. For external flow at a freestream velocity of U and a temperature of T∞ past an
isothermal sphere at Tw of diameter D, the Whitaker correlation may be adopted:
1/4
0.4 μ1
NuD = 2 + 0.4Re1/2 + 0.06Re 2/3
Pr (3.130)
D D
μw
which is accurate within +30% in the range of 0.71 , Pr , 380, 3.5 , ReD , 76,000, and
1.0 , (μ∞/μw) , 3.2 (Whitaker 1972). Thermophysical properties are evaluated at the tem-
perature of T∞, except μw, which is evaluated at the surface temperature, Tw.
Variations of the heat transfer coefficient with circumferential angle from the stagnation
point for a sphere in an airstream are illustrated in Figure 3.11 (Hewitt, Shires, and
102 Advanced Heat Transfer
Polezhaev 1997). These trends are influenced by the processes of boundary layer growth,
separation, and turbulence. On the front side of the sphere facing the incoming flow, the
flow experiences a favorable pressure gradient (decreasing pressure) as it accelerates toward
the top and bottom sides of the sphere. The thickness of the boundary layer increases which
leads to a decrease of the wall temperature gradient and heat transfer coefficient. An adverse
pressure gradient (increasing pressure) occurs on the back side of the sphere. At sufficiently
high velocities, the increasing pressure along the back side of the cylinder causes the boun-
dary layer to separate from the wall and create a local flow reversal and shedding of vortices
from the surface. This flow separation leads to an abrupt increase of the heat transfer coef-
ficient. The adverse pressure gradient on the back side of the sphere causes the pressure to
act against the fluid motion and create the conditions of flow separation.
An alternative correlation for a sphere in cross flow was reported by Achenbach (1978):
1/2
ReD
Nu = 2 + + 3 × 10−4 Re1.6
D (3.131)
4
for 100 , ReD , 2 × 105. This correlation was determined experimentally for convective heat
transfer from isothermal sphere to air.
where the subscripts ∞ and w refer to evaluation of the Prandtl numbers at the freestream
and tube wall temperatures, respectively. This correlation is valid under the following for
SL
SD
D
ST
V, T
FIGURE 3.12
Schematic of flow across banks of tubes.
Convection 103
20 or more rows of tubes (NL ≥ 20), 0.7 , Pr , 500, and 1,000 , ReD,max , 2 × 106 (Zhukaus-
kas 1972). The thermophysical properties are evaluated at the arithmetic mean of the fluid
inlet and outlet temperatures, except Prw, which is evaluated at the wall temperature.
The tube bundle configurations are characterized by the tube diameter, D, transverse pitch
(measured between tube centers perpendicular to the flow direction), ST, longitudinal pitch
(parallel to the flow direction), SL, and the ratio of pitches, SR = ST/SL. Also, the constants C
and m depend on the flow conditions and geometrical configuration as listed below.
For aligned tubes:
For other flow configurations, such as NL , 20, additional correction factors can be
applied to these correlations (Zhukauskas 1972).
Correlations for various other configurations have been developed for forced convection.
Other common examples are impinging jets, packed beds, and various other forms of blunt
bodies. Appropriate correlations for other geometries are presented in comprehensive
books such as Kays and Crawford (1990) and Bergman et al. (2011).
r T(r,x)
D
x
T(x)
xc
Entry region qs
FIGURE 3.13
Schematic of developing flow in a tube.
developing since the velocity profile changes in the streamwise (x) direction as a result of the
viscous shear action on the developing flow. This critical distance is given by xc/D ≈ 0.05ReD
for laminar flow and 10 , xc/D , 60 for turbulent flow. Beyond the entry distance, xc, the
flow becomes fully developed. The boundary layers have merged together and the subsequent
downstream velocity profiles remain uniform. For fully developed flow, the velocity profile
is no longer changing in the x-direction.
For flow in tubes, it is convenient to write the governing equations in cylindrical coordi-
nates. The general form of the three-dimensional Navier–Stokes and thermal energy equa-
tions in cylindrical coordinates is shown in Appendix C. The boundary layer equations are
equivalent to Equations 3.67 through 3.69, except in cylindrical rather than Cartesian coor-
dinates. Using the boundary layer assumptions under steady-state conditions, the mass,
momentum (x-direction), and energy equations in cylindrical coordinates become:
∂u 1 ∂(rv)
+ =0 (3.133)
∂x r ∂r
∂u ∂u ∂p ∂ ∂u
ρu + ρv = − + μ r (3.134)
∂x ∂r ∂x ∂r ∂r
2
∂T ∂T ∂2 T k ∂ ∂T ∂u
ρcp u + ρcp v =k 2+ r +μ (3.135)
∂x ∂r ∂x r ∂r ∂r ∂r
Under the boundary layer assumptions, the radial velocity component, v, is assumed to be
zero. Then the continuity equation implies that the axial velocity is a function of radial
position, r, only. The momentum equation in the r-direction has not been written above.
Since the radial velocity component is zero, the equation has a pressure gradient term in
the r-direction, which is also zero. Therefore, pressure is a function of only x (axial coordi-
nate in the flow direction). Also, the viscous dissipation term in the energy equation may
be neglected except for highly viscous liquids.
Convection 105
For fully developed flow conditions, the changes of the velocity and temperature gradient
in the x-direction are zero. Then the reduced momentum and energy equations become:
∂p μ ∂ ∂u
0=− + r (3.136)
∂x r ∂r ∂r
∂T α ∂ ∂T
u = r (3.137)
∂x r ∂r ∂r
where α is the thermal diffusivity (k/ρcp). On the right side of the x-momentum equation, the
pressure gradient is a function of x only, while the viscous term is a function of r only. This
balance can only be satisfied when both terms are constant. Thus, the pressure gradient in
the axial direction remains constant.
The mean velocity, u , is defined as the average (integrated) velocity across the cross-
sectional area of the tube. Solving the above reduced equations for fully developed flow,
subject to no-slip and symmetry boundary conditions at r = ro and r = 0, respectively,
2
u(r) r
=2 1− (3.138)
u ro
This result is called the Poiseuille flow profile for laminar flow, named after Jean Poiseuille
(1797–1869). For turbulent flows, the power law profile is given by:
u(r) r 1/n
=2 1− (3.140)
u ro
−D(dp/dx)
f = (3.141)
u2 /2
ρ
The functional dependence of the friction factor on the Reynolds number, Re, and ratio of
the surface roughness to the pipe diameter, ϵ/D, is illustrated in the Moody chart (see Fig-
ure 3.14), in honor of Lewis Moody (1880–1953). For laminar flow, the friction factor
decreases with Re. In the range of transition to turbulence (2,100 , Re , 4,000), there is a
gap in values since the flow may be laminar or turbulent (or an unsteady mix of both)
depending on the specific flow and pipe conditions. The friction factor abruptly increases
for turbulent flow and then eventually reaches wholly turbulent flow where a laminar sub-
layer within the boundary layer is so thin that the surface roughness dominates the character
of the near-wall flow. As a result, the friction factor depends only on the relative roughness,
ϵ/D, at high Reynolds numbers.
106 Advanced Heat Transfer
Transition to wholly
turbulent flow
Laminar
flow ε/D = 0.04
0.05
ε/D = 10–2
Friction factor, f
ε/D = 4×10–3
ε/D = 10–3
ε/D = 4×10–4
ε/D = 10–4
ε/D = 4×10–5
Transition Smooth ε/D = 10–5
range pipe
0.005
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
Re = VD/ν
FIGURE 3.14
Moody chart for the friction factor in cylindrical pipes. (Adapted from L.F. Moody. 1944. Transactions of the ASME,
66: 671–684.)
Using the previous laminar flow solution and Equation 3.138, the Moody friction factor
can be expressed as:
64
f = (3.142)
ReD
For turbulent flow in a smooth tube, the following friction factor correlations may be used:
f = 0.316Re−1/4
D (ReD , 2 × 104 ) (3.143)
f = 0.184Re−1/5
D (ReD ≥ 2 × 104 ) (3.144)
From these correlations, it can be observed that the friction factor decreases with ReD, as a
result of the increasingly large velocity and denominator in the definition of the friction
factor.
With respect to heat transfer, the fully developed thermal condition may be expressed in
terms of the nondimensional temperature, θ, as follows:
T(r, x) − Ts (x) T(r, x) − Ts (x)
θ1 = − Ts = − Ts = θ2 (3.145)
T(x) 1 T(x) 2
where the subscripts 1, 2, and s refer to an upstream location, downstream location, and sur-
face, respectively. The mean temperature of the fluid is given by:
1
T(x) = u(r, x)T(r, x)dAc (3.146)
uAc Ac
where Ac is the cross-sectional area of the pipe. In fully developed thermal conditions, the
mean temperature gradient remains constant in the x-direction. Unlike the fully developed
Convection 107
condition of the velocity field, where the mean velocity remains constant, the mean temper-
ature changes in the downstream direction. Otherwise, if it is constant, then there is no heat
transfer. Therefore, zero streamwise derivatives of θ, rather than temperature, are assumed
for thermally fully developed conditions.
Boundary conditions are required in order to solve Equation 3.137 to determine the tem-
perature profile. Two different boundary conditions will be considered—(i) constant heat
flux; and (ii) constant surface temperature. These two cases are illustrated in Figure 3.15.
The case of a constant wall heat flux leads to a constant difference between the mean fluid
and wall temperatures in the fully developed region. For the case of an isothermal wall
boundary condition, this gap decreases because the mean fluid temperature increases
with position as a result of heat transfer to the fluid.
For a boundary condition of a constant wall heat flux, it can be shown that the reduced
energy equation, Equation 3.137, can be solved to obtain:
ur2o dT
2 3 1 r 4 1 r 2
T(r) = Ts − + − (3.147)
α dx 16 16 ro 4 ro
Substituting the temperature and velocity profiles into Equation 3.146 and integrating,
2
ro d T
= Ts − 11 u
T(x) (3.148)
48 α dx
Consider an energy balance over a control volume in the tube consisting of a disk of thick-
ness dx and a perimeter P around the pipe. The rate of change of thermal energy within the
control volume in the x-direction balances the rate of heat addition or removal (q′′s ) from the
boundary,
dT
ṁcp = q′′s P (3.149)
dx
T Developing T
ΔTout = Ts – Tout
Developed Ts(x)
Ts
Ts – T(x)
Constant
ΔT
ΔTin
T(x)
Ts,in (Non-linear)
T(x) (linear)
Tin
xc x L x
FIGURE 3.15
Temperature profiles for constant heat flux and isothermal boundary conditions.
108 Advanced Heat Transfer
where ṁ is the mass flow rate. Since the right side is constant, the energy balance can be
integrated to give the following linear variation of the mean temperature in the x-direction,
′′
q sP
T(x) = T in + x (3.150)
ṁcp
where the subscript in refers to inlet. From this result and Newton’s law of cooling, it can be
shown that the difference between the surface (wall) temperature and the mean temperature
remains constant, and so also q′′s /h is constant, leading to:
48 k
h= (3.151)
11 D
Alternatively,
hD
NuD = = 4.36 (3.152)
k
Therefore, both q′′s /h and the Nusselt number remain constant in fully developed flow
with a constant wall heat flux.
For a boundary condition of a constant wall temperature, Ts, the energy balance over the
same control volume of a disk of thickness dx and perimeter P becomes:
dT
ṁcp
= h(Ts − T)P (3.153)
dx
where ΔT = Ts − T(x). Integrating this energy balance and exponentiation of both sides of
the resulting expression involving ΔT leads to:
ΔT
Ts − T(x) −Pxh
= in = exp (3.155)
ΔTin Ts − T ṁcp
where the average convection coefficient along the length of pipe is given by:
x
h = 1 h(x)dx (3.156)
x 0
Also, consider an overall energy balance along the entire pipe, where the total heat input,
qnet, balances the net enthalpy change of the fluid between the inlet and outlet,
out − T
qnet = ṁcp (T in ) = ṁcp [(Ts − T
in ) − (Ts − T
out )] (3.157)
Convection 109
where the expression inside the bracketed term on the right side is called the log mean tem-
perature difference. Here ΔT refers to the difference between the surface and mean fluid
temperatures.
Also, the temperature profile can be obtained by directly solving the energy equation,
Equation 3.137, subject to the isothermal wall boundary condition. Substituting Equation
3.138 into Equation 3.137,
2
r T − Ts dT αd dT
1−
2u
ro − Ts dx = r dr r dr
T
(3.159)
The previous analysis can be extended to other, more complicated flows, such as a com-
bined internal flow with external flow over the outer surface of the tube. In this case, the
ambient fluid temperature (T∞) is known instead of the surface temperature of the pipe.
The previous results can be used except that Ts is replaced by T∞. Also, the convective
heat transfer coefficient of internal flow within the tube is replaced by the average (overall)
heat transfer coefficient, U, of the combined thermal resistances in series,
1 1 1
= + (3.161)
U hi ho
This result corresponds to convection resistances inside (subscript i) and outside (sub-
script o) of the pipe in series.
These heat transfer results and correlations are summarized in Table 3.6. The names attrib-
uted to the turbulent flow correlations are shown in parentheses. The subscript s implies
TABLE 3.6
Heat Transfer Correlations for Internal Flow in a Tube
Flow Conditions Heat Transfer (NuD ¼ hD== k)
evaluation of the viscosity at the surface temperature. In these correlations, the diameter and
length of the tube are D and L, respectively.
which represents turbulent flow. Using Table 3.6, the Seider–Tate correlation is selected,
and the convection coefficient becomes:
0.14
hi = k × 0.027Re4/5 Pr1/3 μ = 15 W/m2 K (3.163)
D D
μs
For external cross flow past the outside surface of the smokestack, ReD = 4.3 × 105 (tur-
bulent flow). The convection coefficient can be determined based on the Zhukauskas cor-
relation,
1/4
ho = k × 0.26Re0.6 Pr0.37 Pr = 22.7 W/m2 K (3.164)
D
D Prs
Here the temperature of the external airflow, T∞, is fixed rather than the smokestack
surface temperature, Ts. Thus, the analytical solution of internal flow in a tube may be
used except that Ts is replaced by T∞ and the convection coefficient, h, is replaced by
the overall heat transfer coefficient, U,
1 − T
out
T −PLU
in = exp (3.165)
T1 − T ṁcp
Thermal
plume Diameter,
0.6 m
10 m/s, 5 °C
Height, 8 m
FIGURE 3.16
Schematic of exhaust gases from an industrial smokestack.
Convection 111
where P, L, in and out refer to the perimeter and length of the smokestack, and inlet and
outlet temperatures, respectively. The overall heat transfer coefficient is:
1 1 1
= + (3.166)
U hi ho
Here the subscripts i and o refer to inner and outer, respectively. Substituting numerical
values,
in = 5 − (5 − 600)exp π × 0.6 × 8
T
1
(3.167)
1 × 1,104 1/15 + 1/22.7
which yields a base mean temperature of 678.1◦ C. Then, constructing a thermal circuit
with internal and external convection resistances in series,
i − T
T o i − T
T o
q′′s = = = 5.37 kW/m2 (3.168)
Rtot 1/hi + 1/ho
In order to refine this estimate, iterations are performed for the recalculation of thermo-
physical properties, based on the result obtained for the inlet temperature. From a practi-
cal perspective, these operating conditions should be well controlled so that discharge
gases do not condense within the smokestack. Also, gases discharged by buoyancy and
convection as thermal plumes from the smokestack must be well understood so as to
avoid any potential contamination or settling of harmful flue gases in nearby farmland
or residential areas.
This example has used SI units for thermophysical properties and problem parameters.
Conversions between Metric SI and English Imperial units are listed in Appendix B.
12
fRe√A = √ (3.169)
β(1 + β)(1 − 192β tanh π/2β /π 5 )
112 Advanced Heat Transfer
f ReL
Circular segment
Rectangle Circular sector
Trapezoid
Annular sector
100 100
10 10
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Aspect ratio, b/a Aspect ratio, b/a
FIGURE 3.17
Friction factor product, f·Re, for: (a) rectangular, elliptic and (b) other duct shapes. (Adapted from Y. Muzychka and
M.M. Yovanovich. 2016. Handbook of Fluid Dynamics, R. Johnson, Ed., Boca Raton: CRC Press/Taylor & Francis.)
Using the results of the previous circular tube analysis, the Nusselt numbers can be recal-
culated for noncircular cross sections such as triangles, squares, and so on, based on the
hydraulic diameter and square root of the area. Table 2.1 summarizes the Nusselt numbers
for fully developed Poiseuille flow using both the hydraulic diameter and the square root of
the cross-sectional area as the characteristic length scale.
For the rectangle cases, a range of different aspect ratios is presented. Two cases of isother-
mal and isoflux boundary conditions are shown. It can be observed that the Nusselt num-
bers for the different geometries of polygonal ducts approximately√collapse
onto a single
value in the latter case when the characteristic length scale is Lc = A.
Using the square root of area as the characteristic length, the Nusselt number for both the
isothermal (superscript T) and isoflux boundary conditions (superscript H) can be correlated
with respect to the Reynolds number as follows. For the constant wall temperature boun-
dary condition,
√
fRe
NuT√A = 3.01 √ Aγ (3.170)
8 πβ
where γ is a geometrical shape parameter. For the geometries shown in Table 3.7, γ = 1/10.
For other shapes such as isosceles, triangular, and trapezoidal ducts, γ = –3/10. Values of γ
establish the upper and lower bounds in the Nusselt number data. The lower bound consists
of all duct shapes with corner angles less than 90◦ , while the upper bound consists of all
ducts with rounded corners and/or right-angled corners.
For the constant wall heat flux boundary condition,
√
fRe A
√ = 3.66
NuH √ (3.171)
A 8 π βγ
The leading coefficient in each case (3.01 and 3.66) is the average value for the polygons in
Table 3.7 when the characteristic length scale is the square root of the cross-sectional area.
Convection 113
TABLE 3.7
Nusselt Numbers for Fully Developed Flow in Noncircular Ducts
Isothermal Isoflux
These models, when based on a characteristic length of the cross-sectional area, agree well
with experimental data typically within +10% or better. Based on the results of Muzychka
and Yovanovich (2016), the cross-sectional area represents an effective characteristic length
and alternative to the hydraulic diameter for convective heat transfer correlations.
Boundary layer y, v
thickness, δ
Boundary layer – laminar
Tw < T∞ Outer region
Quiescent fluid
Inner region T∞, g
u(y)
ym Turbulent
x, u
FIGURE 3.18
Schematic of a free convection boundary layer.
diffusion and buoyancy are dominant and an outer region where inertial effects and buoy-
ancy forces are dominant. The inner region extends from the wall (y = 0) to the position of
the maximum velocity (y = ym).
The governing continuity, momentum, and thermal energy equations for natural convec-
tion were presented earlier in Equations 3.46 through 3.49. Note that the y-direction is
aligned with the gravity vector in Equation 3.48, whereas in this case, the x-direction is
downward, so Equation 3.48 becomes the x-momentum equation in the vertical plate con-
figuration in Figure 3.18. Based on the definition of the thermal expansion coefficient, β, in
Equation 3.42, the buoyancy term in Equation 3.48, (ρ − ρ∞)g, can be rewritten in terms of
temperatures.
1 ∂ρ 1 ρ − ρ1
β=− ≈− (3.172)
ρ ∂T p ρ T − T1
Therefore, using Equations 3.46 through 3.49 and the boundary layer assumptions, the
reduced form of the two-dimensional steady-state equations of continuity, momentum,
and thermal energy for free convection on a vertical plate become:
∂u ∂v
+ =0 (3.173)
∂x ∂y
∂u ∂u ∂2 u
ρu + ρv = μ 2 − ρgβ(T − T1 ) (3.174)
∂x ∂y ∂y
∂T ∂T ∂2 T
ρcp u + ρcp v =k 2 (3.175)
∂x ∂y ∂y
The body force term of buoyancy in Equation 3.174 has been expressed in terms of tem-
perature differences through the Boussinesq approximation. This approximation assumes
that density differences are negligible in all terms, except in buoyancy forces, where the den-
sity difference drives the fluid motion. For free convection problems, it can be assumed that
viscous dissipation in the thermal energy equation can be neglected.
Convection 115
∂u∗ ∂v∗
+ =0 (3.176)
∂x∗ ∂y∗
2 ∗
∗ ∗∂u∗ ∗ ∗ ∂u
∗
1 ∂u
ρ u ∗
+ρ v ∗
= + GrL θ (3.177)
∂x ∂y ReL ∂y∗2
2
∂θ ∗ ∂θ 1 ∂ θ
u∗ + v = (3.178)
∂x∗ ∂y∗ ReL Pr ∂y∗2
where the reference velocity in the dimensionless velocity is ν/L and the nondimensional
temperature is θ = (T − T∞)/(Ts − T∞). It can be observed that the average Nusselt number
depends on the Reynolds number, ReL, Prandtl number, Pr, and Grashof number, GrL(or
Rayleigh number, RaL = GrL · Pr).
Note that the above form of the nondimensional equations is slightly different than other
forms derived earlier in this chapter because the reference scales in the nondimensionaliza-
tion are different. For example, the reference velocity for forced convection was U, whereas
the characteristic velocity for free convection in this section is ν/L.
Consider a similarity solution of the free convection boundary layer equations. Define the
similarity variable, η, and stream function, ψ, as:
y Grx 1/4
η= (3.179)
x 4
Grx 1/4
ψ(x, y) = f (η) 4ν (3.180)
4
2ν 1/2 ′
u= Grx f (η) (3.181)
x
The boundary layer equations are transformed to the following coupled ordinary differ-
ential equations for momentum and thermal energy, respectively,
η = 0; f = f ′ = 0; θ=1 (3.184)
η 1; f ′ 0; θ0 (3.185)
Using a similar solution procedure as the Blasius solution, the ordinary differential equa-
tions can be solved with a numerical integration method, such as the Runge–Kutta method.
116 Advanced Heat Transfer
Sample results of the velocity and temperature profiles at varying Prandtl numbers are
shown in Figure 3.19. It can be observed that the maximum velocity occurs inside the boun-
dary layer as previously illustrated in Figure 3.18. Also, the boundary layer thickness
increases when the Prandtl number, Pr, decreases.
Also, the temperature, heat flux, and Nusselt number distributions can be determined
from the similarity solution. For both the constant wall heat flux and constant wall temper-
ature cases,
1/4
Grx
Nux = g(Pr) (3.186)
4
4 GrL 1/4
NuL = g(Pr) (3.187)
3 4
where
0.75Pr1/2
g(Pr) = (3.188)
(0.609 + 1.221Pr1/2 + 1.238Pr)1/4
The following Churchill and Chu correlation, based on experimental data, extends over
both laminar and turbulent regions of a free convection flat plate boundary layer (Churchill
and Chu 1975):
2
9/16 −8/27
NuL = 0.825 + 0.387Ra1/6
L (1 + (0.492/Pr) ) (3.189)
Boundary layer transition to turbulence occurs at approximately Rax ≈ 109 for flow along a
vertical flat plate.
0.9
0.6
Pr = 0.01
0.8
Pr = 0.01
f ′(η) = ux Grx–1/2/(2ν)
θ = (T – T∞)/(Ts – T∞)
0.5 0.7
Pr = 0.72
0.4 0.6
Pr = 0.72 Pr = 1.0
0.5
0.3 Pr = 1.0
0.4
Pr = 10 Pr = 10
0.2 0.3
Pr = 100 Pr = 100
0.2
0.1
0.1
0.0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0
η = y (Grx/4)1/4/x η = y (Grx/4)1/4/x
FIGURE 3.19
Dimensionless velocity and temperature profiles in a free convection flat plate boundary layer. (Adapted from
S. Ostrach. 1953. “An Analysis of Laminar Free Convection Flow and Heat Transfer about a Flat Plate Parallel to
the Direction of the Generating Body Force”, National Advisory Committee for Aeronautics, Report 1111.)
Convection 117
Although the previous results were developed for a flat plate, the correlations may also be
applied to other surface configurations when surface curvature effects are negligible. For
example, the previous flat plate correlations can be applied to large vertical cylinders of
length L when surface curvature effects are negligible and the boundary layer thickness,
δ, is much less than the diameter; δ ≪ D and D . 35L/Gr1/4L .
where twice the boundary layer thickness at the base of the fin has been used as the value
for δ. This thickness represents the optimal height of the plate since it corresponds to
Vertical fin
Surface
temperature, Ts H Ambient air, T∞
s
t
FIGURE 3.20
Schematic of an array of vertical rectangular fins.
118 Advanced Heat Transfer
merging of the boundary layers at that location. Solving this equation yields an optimal
height of Hopt = 1.4 cm. Since the designers have selected H = 8.6 cm, it is recommended
to reduce this height to provide better thermal effectiveness of the fins, if the reduction
is feasible in terms of other design constraints such as costs and fabrication.
Subsequent calculations are based on the initially selected fin height of 8.6 cm. The total
rate of heat transfer from all fins excluding the base regions is given by:
q = 2N h(HL)(Ts − T1 ) (3.192)
where N = w/(s + t) is the number of fins, based on a spacing, s, between each fin and a
thickness, t, of each fin. The average convection coefficient, h, and fin thickness, t, can
then be determined based on the correlation developed from the similarity solution,
1/4
w k 4 GrHopt 0.75Pr1/2
t=2 Hopt L(Ts − T1 ) − s (3.193)
q Hopt 3 4 (0.609 + 1.221Pr1/2 + 1.238Pr)1/4
Based on the previously selected fin height, the Grashof number is calculated as 2.7 × 107
(laminar flow conditions). Substituting numerical values and suitable thermophysical
properties into this result, the fin spacing becomes s = 5.3 mm. It is anticipated that a fur-
ther increase of the total fin surface area from a reduction in fin spacing would increase the
heat transfer rate from the electronic enclosure. Optimization studies could consider the
trade-offs between more surface area and the resulting boundary layer structure to further
improve the design.
∂2 u
μ − gx β(T − T1 ) = 0 (3.194)
∂y2
Convection 119
∂2 T
k =0 (3.195)
∂y2
where gx refers to the component of the gravitational acceleration in the x direction. This
notation will be used to describe the component of gravity tangent to the surface. Through
this approach, the body gravity function method allows the analysis to be extended to other
surface configurations (e.g., inclined plates or axisymmetric surfaces oriented at any angle)
by using the same solution but modifying the body gravity function accordingly to fit the
surface geometry.
Solving Equation 3.195, subject to T( ym) = Tm, and substituting the result into the solution
of Equation 3.194,
ρβgx y2 y2m y y
u(y) = (T1 − Tw ) yym − − + (3.196)
μ 2 2Δ 6Δ
where Δ refers to the location where the extrapolated temperature from the inner region
equals the ambient fluid temperature. Integrating the velocity profile across the inner region
yields the following mass flow rate, Γi, per unit width of surface,
ym
ρ2 βgx y3 5M
Γi = ρu dy = (T1 − Tw ) m 1 − (3.197)
μ 3 8
0
where ym and M = ym/Δ are unknowns to be determined by matching conditions at the inter-
face between the inner and outer regions.
Consider a mass and energy balance for a differential control volume in the inner region
(see Figure 3.21), as follows,
dΓ i
m′′i + Γ i = Γ i + dx (3.198)
dx
where the subscripts i, o, m and x refer to inner region, outer region, location y = ym, and
position x. Also, m′′i , hm and h refer to the mass flow rate per unit width across the inner
T
Inner region
Tf
Outer region
(ΓiH)x
Tm
q qo
Tw
Wall (ΓiH)x+dx
ym Δ y
FIGURE 3.21
Energy balance and temperature profile in a free convection boundary layer.
120 Advanced Heat Transfer
region interface, enthalpy at y = ym, and average enthalpy across the surface at position x,
respectively. Also, qo and q are heat flow rates at the interface and wall, as illustrated in
Figure 3.21.
Combining the above mass and energy equations with the solution of Equation 3.195
and rearranging, the wall heat flux becomes:
Tm − Tw dΓ i
q=k = (hm − hi ) + qo (3.200)
ym dx
Similarly, in the outer region, the mass and energy balances can be written as:
dΓ o
m′′o − m′′i = (3.201)
dx
where the subscript ∞ refers to ambient conditions. Combining these results and rear-
ranging,
dΓi k
ym = (3.203)
dx cp B(x)
where,
i dΓ o /dx T1 − T
T1 − T o
B(x) = + (3.204)
Tm − Tw dΓ i /dx Tm − Tw
hx q′′ · x Mx
Nux = = = (3.206)
k (T1 − Tw )k ym
1/3 x −1/4
gx 1 gx
Nux = f (M) · Ra1/4
x · dx (3.207)
g x 0 g
Convection 121
where,
5M B 1/4
f (M) = M 1− (3.208)
8 4
Here, the Rayleigh number, Ra = Pr · Gr. Also, M can be correlated as a function of Prandtl
number by matching the above Nusselt number with the known exact (similarity) solution
of the free convection boundary layer on a vertical plate, thereby leading to:
1/4
Pr
f (M) = 0.48 (3.209)
0.861 + Pr
Integrating the local Nusselt number over the surface, S, yields the following average Nus-
selt number:
3/4
4 1 S gx 1/3
Nus = f (M) · Ras ·
1/4
dx (3.210)
3 S 0 g
This result is typically accurate within +1% of the exact similarity solution. Very good
agreement with experimental data is achieved for flat surfaces in laminar flow conditions.
At higher Rayleigh numbers and other geometrical configurations, discrepancies with
experimental data increase as a result of turbulence and surface curvature effects.
The last integral term in brackets on the right side of Equation 3.210 is called the body grav-
ity function. Lee et al. (1971) presented a generalized form of the body gravity function,
3/4
1 P(θ)sinθ 1/3
GL = dA (3.211)
A A A/L
where P(θ) and L represent the local perimeter associated and characteristic length scale,
respectively. Also sinθ represents the component of the local body force tangent to the
body. The square root of the surface area is an effective choice of the characteristic
length,
1/3
3/4
= 1 P(θ)sinθ
G√A √ dà (3.212)
A Ã A
Muzychka and Yovanovich (2016) presented a modified form of Equation 3.210 using this
body gravity function,
0.670
F(Pr) = 4/9 (3.214)
1 + (0.5/Pr)9/16
122 Advanced Heat Transfer
This generalized approach with the body gravity function provides a useful surface geom-
etry factor to determine the Nusselt number over a wide range of geometrical configura-
tions. The body gravity function depends on the body shape, aspect ratio, orientation
with respect to the gravity vector, and the characteristic length scale.
Consider a surface consisting of several segments with fluid flow sequentially over mul-
tiple sections of the surface. For example, a vertical cylinder consists of three separate surface
areas: a horizontal bottom surface, vertical side cylindrical area, and horizontal top surface.
The total body gravity function for multiple connected surfaces can be represented by a
series flow relationship (Muzychka and Yovanovich, 2016):
3/4
N
G√A = )4/3 Ai /A 7/6
(G√A i (3.215)
i=1
where A represents the total area of all surfaces (sum of the individual surface areas).
Instead of a sequential motion of the fluid over each surface, if the fluid flows over N
different surfaces with N different streams that are independent, then the body gravity func-
tion can be obtained from the following parallel flow expression:
N
=
G√A ) (Ai /A)7/8
(G√A (3.216)
i
i=1
0.589Ra1/4
NuD = 1 + D
4/9 (3.217)
1 + (0.469/Pr)9/16
which is applicable in the range of Pr ≥ 0.7 and RaD ≤ 1011 Churchill (1983). When a sphere
at a temperature of Tw is immersed in a fluid of temperature T∞, where Tw . T∞, an upward
moving boundary layer forms at the bottom of the sphere and grows in thickness along the
curved surface. The boundary layer thickness is a function of the angle around the surface.
This configuration represents an external flow since there are no obstructions by other sur-
faces in the flow field. The following configurations of concentric spheres and rectangular
Convection 123
TABLE 3.8
Body Gravity Functions for Various Surface Configurations
Surface Configuration Body Gravity Function Comments
1=8
P 1=4 P
(a) Vertical surface with a Gpffiffiffi
A
¼ p ffiffiffi
ffi ¼ Constant perimeter, height H, area of
A H
constant perimeter 1=8 A ¼ PH, sinθ ¼ 1
(b) Vertical circular cylinder ffiffiffi ¼ π 1=8 D
GpA Cylinder of height H, diameter D, area
H
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1=8 of A ¼ PH
ffiffiffi ¼ 4a
(c) Vertical elliptical cylinder GpA E 1 (b=a)2 Semi-major and semi-minor axes of a
H
and b, and E is the complete elliptic
1=8 integral of the second kind
ffiffiffi ¼ 21=8 W
(d) Vertical rectangular plate GpA Plate of width W, both sides cooled,
H
1=8 and P ¼ 2W
(e) Inclined rectangular plate ffiffiffi ¼ (sin θ)1=4 W
GpA Plate inclination angle of θ, and
H
bottom side cooled
3=4
ffiffiffi ¼ [0:389 þ 0:857(L=D)4=3 ]
(f) Horizontal cylinder GpA Length of L, diameter of D
(0:5 þ L=D)7=8
enclosures are internal flows that are constrained by the physical presence of solid bound-
aries on all sides of the flow field.
Free convection heat transfer in the region between concentric spheres involves a com-
bination of features observed in previous geometries. For this configuration, Raithby
and Hollands (1975) defined an effective thermal conductivity, keff, of free convection as
follows,
1/4
keff 1/4 Pr
= 0.74 Ra∗S (3.218)
k 0.861 + Pr
where,
L · RaL
Ra∗S = 5 (3.219)
D41 D42 D−7/5
1 + D −7/5
2
where D1 and D2 are the diameters of the inner and outer spheres, respectively, and L is
the radial gap width. In the range of 100 ≤ Ra∗S ≤ 10,000, the following expression can be
used to determine the free convection heat transfer rate between the surfaces of the
spheres at temperatures T1 and T2, respectively:
πD1 D2
q = keff (T1 − T2 ) (3.220)
L
Surface
temperature, Ts,1
T∞
Ts,2
FIGURE 3.22
Free convection in a tilted rectangular enclosure.
∗
∗
1708 1708(sin1.8τ)1.6 RaL cos τ 1/3
NuL = 1 + 1.44 1 − 1− + −1 (3.221)
RaL cos τ RaL cos τ 5830
where θp is the tilt angle with respect to the horizontal plane. Also, [ ]* means that if the quan-
tity in brackets is negative, then it is set equal to zero. These correlations provide close agree-
ment with experimental data below the following critical tilt angles: τ* = 25◦ for H/L = 1; 53◦
for H/L = 3; 60◦ for H/L = 6; 67◦ for H/L = 12; and 70◦ for H/L . 12.
+ B′
B=B (3.222)
where,
t+T
= 1
B B dt (3.223)
Δt
t
Here T represents a time scale that spans the range of time over which the turbulence
occurs. An example of mean and fluctuating components of a general scalar quantity is illus-
trated in Figure 3.23. The mean value is designated by an overbar notation and obtained by
integration of the quantity over a characteristic time period associated with the flow. The
fluctuating random component represents the deviation of the mean from the instantaneous
value.
B′
FIGURE 3.23
Mean and fluctuating components of a turbulent scalar quantity.
126 Advanced Heat Transfer
For example, the velocity components can be expressed as a sum of mean and fluctuating
velocities:
u=u + u′ ; v = v + v′ ; w = w
+ w′ (3.224)
1 ′2 2 2
k= u + v′ + w′ (3.226)
2
Large eddies
Inertial
Turbulent kinetic energy
sub-range
Energy
transfer
Dissipation
Energy range
containing
vortices
Frequency
Eddy size
FIGURE 3.24
Kolmogorov turbulence spectrum.
Convection 127
In Figure 3.24, turbulent kinetic energy is extracted from the mean flow by large eddies. The
large swirling motions occur at low frequencies. Kinetic energy is transferred from lower to
higher frequencies and smaller eddies until it is finally dissipated by viscous fluid action into
thermal energy which is later transferred as heat. In the higher frequency (smaller eddy)
range of the spectrum, the turbulent motion possesses local isotropy. In this local isotropic
region, the turbulence properties of the flow are uniquely determined by viscosity, v, and
the dissipation rate, ϵ (energy dissipation rate per unit mass), independently of direction.
As key features of the turbulence spectrum, define the following Kolmogorov scales of
turbulence for length (Lko), velocity (Vko), and time (Tko):
3 1/4
ν
Lko = (3.227)
ε
When the Reynolds number, based on these Kolmogorov scales, has an order of magni-
tude of approximately 1, the Kolmogorov scales are applicable to the small-scale eddy
motions in the viscous dissipation region of the spectrum.
∂
ui
=0 (3.230)
∂xi
∂ui i
∂u ∂p ∂ ∂u j
i ∂u
ρ j
+ ρu =− + μ + ′ ′
− ρu i u j (3.231)
∂t ∂xj ∂xi ∂xj ∂xj ∂xi
∂T ∂T ∂ ∂T
ρcp j
+ ρcp u =− −k + ρcp u′ j T ′ + Φ (3.232)
∂t ∂xj ∂xj ∂xj
where i = 1, 2, 3, and j = 1, 2, 3.
These equations appear the same as laminar flow, except for the second term in brackets in
each of Equations 3.231 and 3.232. Recall that the first terms in brackets in Equations 3.231
and 3.232 represent molecular diffusion of momentum and heat, respectively. Similarly, the
second terms in brackets can be interpreted as turbulent, rather than laminar (molecular),
diffusion of momentum and heat. Note that k in Equation 3.232, heat flux terms, and Four-
ier’s law, will refer to the thermal conductivity and should not be confused with the turbu-
lent kinetic energy.
128 Advanced Heat Transfer
∂T
qlj = −k ; qtij = ρu′ j T (3.234)
∂xj
where the superscripts l and t denote laminar and turbulent, respectively. The total stresses
and heat fluxes on a fluid element are composed of both laminar and turbulent components.
Turbulent stresses and fluxes can be viewed as a diffusion type of process arising from tur-
bulent mixing, rather than momentum and heat exchange at a molecular level. As a result of
the time averaging processes, many new variables (turbulent stresses and heat fluxes) have
been introduced. As a result, further closure equations, relating turbulent quantities to mean
flow variables, will be added in order to solve the governing equations.
In view of the analogous forms of laminar and turbulent components of diffusion in Equa-
tions 3.231 and 3.232, the concepts of turbulent viscosity, μt, and turbulent conductivity, kt, are
introduced as follows:
i ∂
∂u uj
τtij = −ρu′ i u′ j = μt + (3.235)
∂xj ∂xi
∂T
qtj = ρu′ j T = −kt (3.236)
∂xj
μt c p
kt = (3.237)
Prt
where Prt is the turbulent Prandtl number (usually taken as constant or 1). In order to
solve the governing equations, suitable models are needed for the turbulent viscosity. In
upcoming sections, zero-, one-, and two-equation models of the turbulent viscosity will
be examined.
With increasing computational speed and capabilities of computers, it is possible to
develop an exact turbulence model by discretizing the governing equations into very fine
scale equations. However, from a practical perspective, this is very difficult in view of the
wide range of turbulent length, time, and velocity scales. For example, about 1010 grid points
are needed to accurately resolve turbulent flow in a duct (10 cm × 10 cm × 2 m) at the scales
of turbulence down to about 0.1 mm. This represents a very large storage requirement for
computers. Also, designers are often more interested in average flow characteristics than
all of the detailed small-scale turbulence information. Therefore, turbulence modeling often
involves the development of relationships between the turbulent stresses, heat fluxes, and
mean flow quantities. In upcoming sections, models will be developed and presented for
turbulent stresses and heat fluxes in order to determine the turbulent viscosity and
conductivity.
Convection 129
where the subscript tot refers to the total of laminar (l ) and turbulent (t) components.
In a zero-equation model of the eddy viscosity, a single length scale is used to estimate the
turbulent viscosity. For example,
μt = ρVref L (3.240)
where Vref is a reference velocity and L is the mixing length, or characteristic length scale of
the turbulent eddies. It is usually an order of magnitude smaller than a characteristic flow
dimension. In upcoming sections, the mixing length will be related to mean flow quantities
through algebraic relations.
The characteristic velocity of the turbulence, Vref, can be determined based on reference
velocities in the problem definition. Alternatively, it can be selected as the square root of
the turbulent kinetic energy, k, as follows:
√
Vref ≈ k (3.241)
where the proportionality constants have been absorbed into a single mixing length, Lm, and
the modulus is needed to ensure a positive turbulent stress.
Thus, the mixing length hypothesis can be written as:
d
u
τtyx = μt (3.243)
dy
where,
d u
μt = ρL2m (3.244)
dy
The choice of a particular mixing length depends on position within the flow field. Near a
wall, the mixing length can be selected as the distance to the wall, whereas further away
from the wall, it depends on larger-scale turbulence structures that are problem dependent.
The method is generally limited to simple flow configurations since there is normally a lack
of knowledge of the mixing length beforehand in new problems. Another limitation is that
the model yields a zero-turbulent viscosity due to dū /dy, even though the velocity profile
around a maximum point may be nonzero or nonsymmetrical.
δκ yn
Lm = 1− 1− (3.247)
n δ
where δ is the boundary layer thickness and n = 5 for boundary layer flow.
Convection 131
where cD ≈ 0.09 and σk ≈ 1 are empirical constants. After solving Equation 3.248 subject
to appropriate boundary conditions, the spatial variation of turbulent kinetic energy and mix-
ing length can be obtained, after which the turbulent viscosity and conductivity are found.
From left to right in Equation 3.248, the rate of change of k with time equals molecular and
turbulent diffusion of k, plus the production of k by velocity gradients in the mean flow,
minus dissipation of k to internal energy in the small-scale eddy motion. This one-equation
model can be used successfully for shear layer flows and wakes where Lt/δ remains approx-
imately constant (δ refers to the wake thickness). In wall-bounded flows, the turbulent
length scale is approximately equal to the mixing length, as determined from the square
root of the turbulent kinetic energy in Equation 3.241.
In the above one-equation model, a differential transport equation was used to determine
the turbulent kinetic energy and hence a velocity scale based on k using the Prandtl–Kolmo-
gorov model. It remains to find the other characteristic scales of the turbulent flow. A two-
equation model includes an additional transport equation for another key dependent vari-
able, such as the rate of dissipation of kinetic energy, ϵ. As more transport equations are
introduced to improve the modeling of turbulence, the complexity of the problem increases,
and additional empirical coefficients and/or closure equations are required. Therefore trade-
offs exist in turbulent flow modeling involving accuracy, computational costs, complexity,
and robustness of the solution method.
A two-equation model involves differential transport equations for two key turbulence
quantities. In the previous models, the product of the reference velocity, Vref, and length,
L, was used to estimate the turbulent viscosity. Alternatively, a differential equation for k
can be used along with any product of k and L, such as k · L or k 3/2/L. Using the latter prod-
uct, the rate of dissipation of turbulent kinetic energy, ϵ, is defined as:
k3/2
ε = cμ (3.249)
L
where cμ = 0.09 is an empirical constant. Using this definition and Equation 3.241 for the ref-
erence velocity, Vref, the turbulent viscosity becomes:
k2
μt = ρVref L = ρcμ (3.250)
ε
132 Advanced Heat Transfer
The equation for the transport of the dissipation rate will be used to determine ϵ. This
equation consists of four key processes: convection; diffusion (involving viscous, turbu-
lent, and pressure terms); production; and dissipation. Consider two-dimensional incom-
pressible flow under steady-state conditions. The gradient diffusion hypothesis for the
dissipation rate will be assumed, as well as negligible effects of surface curvature. It
can be shown that the convection of ϵ balances the diffusion plus production of ϵ, minus
the destruction of ϵ, plus any additional terms, such as buoyancy (denoted by E), as
follows,
∂ε ∂ε ∂ μt ∂ε ε ε2
+ v =
u μ+ + cε1 P − cε2 + E (3.251)
∂x ∂y ∂y σ ε ∂y k k
where,
2
∂
u
P = ρμt (3.252)
∂y
Values of the empirical constants are σϵ = 1.3, cϵ1 = 1.44 and cϵ2 = 1.92. These values have
been established through numerical simulations and data fitting for a wide range of
turbulent flows.
Equations 3.248 and 3.251 represent the well-known k − ϵ model of turbulence. This model
is a two-equation closure that solves Equations 3.248 and 3.251 for k and ϵ, which are then
used to determine the turbulent viscosity and stresses in the momentum equations. The
turbulent Prandtl number is used to relate μt to the turbulent conductivity for subsequent
solutions of the turbulent heat transfer equations. The k − ϵ model is a widely used turbu-
lence model in commercial software and other design tools in industry for a wide range
of applications.
Although it widely used, there are significant limitations of the k − ϵ model. For example,
empirical correlations, such as the law of the wall, are usually required in the near-wall
region for numerical solutions. These correlations involve various limitations which con-
strain their applicability in practical applications. Another limitation of the k − ϵ model is
that turbulent stresses are assumed to be equal (isotropic) in all directions which is often
inaccurate. The reader is referred to other sources for more advanced turbulence models,
for example, Reynolds stress models (Rodi 1984), algebraic stress models (Hanjalic and
Launder 1972), and renormalization group theory-based models (Kirtley 1992).
total entropy cannot decrease over time for an isolated system. Entropy represents a mea-
sure of microscopic disorder of the system.
Through the minimization of entropy generation, energy losses and flow irreversibilities
can be minimized, thereby reaching the highest thermal efficiency. Entropy provides a
unique insight into reaching an optimal system performance. A range of entropy based
design methods and applications were analyzed and discussed in books by Naterer and
Camberos (2008) and Bejan (1996).
Consider B to represent specific entropy, s, in the general balance equation in Equation 3.1.
The second law requires that entropy is produced, but never destroyed, in an isolated system.
Therefore, in this case, Equation 3.1 is called a “transport equation” rather than a “conserva-
tion equation” since entropy is not a conserved quantity. The generation term in Equation 3.1
becomes the entropy production rate, Ṗs . According to the second law, the entropy production
rate is either positive (for irreversible processes) or zero (reversible) for an isolated system. Pro-
cesses such as friction, heat transfer, and chemical reactions are irreversible because the ther-
modynamic state of the system and its surroundings following the process cannot be returned
back to their initial state without a further expenditure of energy.
Entropy is a scalar quantity that is transported by fluid and heat flow, similar to the pre-
vious mass, momentum, and energy equations in this chapter. Performing the entropy bal-
ance on a differential control volume in Figure 3.1 yields:
∂s ∂s ∂ qj
ρ + ρuj =− + Ṗs (3.253)
∂t ∂xj ∂xj T
From left to right, the terms represent the total derivative of entropy of a fluid element
with respect to time (temporal and convective components); entropy flux due to heat
flow; and the entropy production rate. This result in Equation 3.253 is called the “transport
form” of the second law.
Entropy can also be defined with respect to the Gibbs equation in Chapter 1. Differentiating
the Gibbs equation for a simple compressible substance,
Ds De Dυ
T = +p (3.254)
Dt Dt Dt
where D()/Dt refers to the total (substantial) derivative, as defined in Chapter 1, including
transient and convective components.
The second law states that the entropy production rate is greater than or equal to zero
for an isolated system. Therefore another form of Equation 3.253 is needed to clearly indicate
a positive entropy production rate. To derive this alternative form, the Gibbs equation,
Equation 3.254, is combined with the continuity equation, Equation 3.5 and compared to
Equation 3.253. On the right side of Equation 3.253, the quotient rule of calculus and Four-
ier’s law are used to express the heat flux in terms of temperature. Then by comparing the
resulting terms with Equation 3.253, it can be shown that the entropy production rate is a
positive-definite sum of squared quantities:
k ∂T ∂T Φ
Ṗs = + ≥0 (3.255)
T 2 ∂xj ∂xj T
where the viscous dissipation function, Φ, was given in Equation 3.36. This form of the
entropy production rate is called the “positive-definite form” of the second law. The sum
of squares ensures a positive-definite quantity as required by the second law. Calculating
134 Advanced Heat Transfer
this local entropy production can be a powerful tool in engineering design because it gives a
quantitative measure of irreversibilities throughout the flow field. Regions of high entropy
production can be targeted for design modifications to improve the overall system efficiency
and performance.
The level or quality of energy in a flow stream is reduced when entropy is produced. For
example, the ability of a fluid stream to perform useful work is reduced when friction
irreversibilities convert mechanical energy to thermal energy through viscous dissipation.
Minimizing these flow irreversibilities can provide substantial benefits leading to direct
cost savings and more efficient use of limited energy resources. The method of entropy
generation minimization (Bejan 1996) is a powerful design tool that aims to reduce system irre-
versibilities through a systematic framework that uses entropy production as a key design
variable. An example of this method for optimizing a surface configuration in external flow
is provided below.
where Atube is the cross-sectional area of the stream tube. The energy balance and second
law can be written as follows:
q′′ As
Ṗs = ṁ(sout − sin ) − ≥0 (3.258)
Ts
U, T∞
Inflow Outflow
U, T∞
FIGURE 3.25
Schematic of external flow past an object.
Convection 135
where the subscripts in and out refer to the inlet and outlet, respectively. The subscript s
refers to the surface of the object, except in Ṗs where it refers to entropy. The overbar on h
distinguishes the average convective heat transfer coefficient from the enthalpy. Also,
using the Gibbs equation in Chapter 1, and replacing the internal energy with enthalpy,
h = e + p υ, yields:
pin − pout
hout − hin = T(sout − sin ) + (3.259)
ρ
From a balance of momentum on the stream tube, the latter pressure difference term
in the above equation can be rewritten in terms of the drag force, FD. The pressure dif-
ference can be expressed by the drag force on the body divided by the cross-sectional
area of the stream tube, since the inlet and outlet momentum fluxes are identical. As a
result,
Ts − T1 FD U
Ṗs = q′′ As + (3.261)
Ts T1 T1
Assuming that ΔT ≪ T, the first term on the left side can be simplified, yielding:
2
q′′ As F D U
Ṗs = h + T1 (3.262)
T1
On the right side, the first term represents entropy production due to heat transfer. This
thermal irreversibility decreases with fluid velocity, U. A larger convection coefficient due
to a higher velocity entails a lower surface–fluid temperature difference, ΔT, and reduced
entropy generation for a fixed heat flux, q", from the body. On the other hand, the latter
term on the right side in the above equation is the entropy generation due to fluid friction
(drag force), which increases with U.
The above entropy production rate expression can now be minimized in terms of the
surface geometry. Consider a flat plate of length L and width W and heated uniformly
at q". Integrating Equation 3.262 over a flat plate,
2 L L
q′′ Wdx U
Ṗs = + τw Wdx (3.263)
T1 0 h T1 0
The following heat transfer and fluid friction correlations can then be used for a flat
plate boundary layer flow.
hx
Nux = = 0.458Pr1/3 Re1/2
x (3.264)
k
τw
cf ,x = = 0.664Re−1/2
x (3.265)
ρU2 /2
136 Advanced Heat Transfer
This result shows the various irreversibilities of heat transfer and fluid friction which
contribute to the total entropy production rate. The plate surface can be optimized with
respect to the plate length, L, by setting the derivative of Ṗs with respect to ReL equal to
zero, yielding:
q′ 2
ReL,opt = 2.193Pr−1/3 (3.267)
U μkT1
2
where q’ refers to the heat transfer rate per unit length of plate.
This above result of the optimal Reynolds number can also be rewritten in terms of the
optimal plate length. Below the optimal length, to achieve a fixed heat flux, an increasingly
larger temperature difference and thermal irreversibility occur, whereas above the opti-
mal length, a larger surface area creates a higher overall drag force and friction irrevers-
ibility. Therefore, an optimal plate length exists which minimizes the overall entropy
production over the surface.
Exergy refers to the energy availability or maximum possible work that a system can
deliver as it undergoes a reversible process from a specified initial state to the state of the
environment at STP (standard temperature and pressure; 101 kPa, 25◦ C). Exergy has a
diverse range of applications in energy systems, environmental engineering, sustainable
development, natural resource utilization, and complex physical systems such as biological
evolution and ecosystems (Dincer and Rosen 2013). The term “exergy” was first used in 1956
by Zoran Rant (1904–1972), a Slovenian chemical engineer and professor.
In thermal engineering systems, exergy represents the capacity of energy to perform use-
ful work. For example, consider air in a tank, closed by a valve, at the same temperature and
pressure as the surrounding air at STP outside the tank. In this case, there would be zero
exergy because the system is in equilibrium with the environment and there is no potential
for work to be done on the surroundings. On the other hand, if an underground thermal
energy source has pressurized fluid at a temperature and pressure higher than STP, then
the energy availability is greater than zero since this fluid has a potential to perform
work. For example, the pressurized fluid could flow through a turbine to generate power.
A common design objective of exergy analysis is to minimize the destruction of exergy,
similarly to minimization of the entropy production earlier in this section (Dincer and
Rosen, 2013).
equation. On the other hand, the second expression is a positive-definite form with a sum of
squared terms and an inequality that represents the second law of thermodynamics. These
two separate forms of the entropy production rate will be denoted by subscripts “te” (trans-
port equation form) and “pd” (positive-definite form) as follows,
Ds ∂ qj
Ṗs,te = ρ + (3.268)
Dt ∂xj T
k ∂T ∂T Φ
Ṗs,pd = + ≥0 (3.269)
T 2 ∂xj ∂xj T
Although both forms should yield the same result, in practice, when numerical and
approximate methods are used to solve the governing equations, each expression for the
entropy production may yield different results since they are computed independently
and separately of each other. Ideally, the difference between each expression should be
zero, however, numerical or approximation errors may yield a nonzero difference that
can be used to characterize the solution error.
The Gibbs equations can be used to rewrite entropy in terms of temperature in Equation
3.253. Also, the heat flux can be expressed in terms of the temperature gradient through
Fourier’s law. Then, subtracting Equation 3.269 from Equation 3.268, and using the quotient
rule of calculus to combine and cancel terms with temperature derivatives, yields the follow-
ing “apparent entropy production difference”:
Ds k ∂ ∂T Φ
ΔṖs = Ṗs,te − Ṗs,pd = ρ − − (3.270)
Dt T ∂xj ∂xj T
In an exact solution, this difference should be precisely equal to zero. Using the Gibbs
equation, the right side can be rewritten as the energy equation, equaling zero. However,
in practice, approximation errors in the solution of the energy equation, for example through
numerical discretization, may yield a nonzero apparent entropy production difference. The
magnitude of this difference has significance in connection to numerical solution errors.
Adeyinka and Naterer (2002) analyzed the relationships between the apparent entropy
production difference and numerical solution errors.
Another significant insight can be obtained if Fourier’s law and the Gibbs equation are
used to rewrite Equation 3.268 in terms of temperature as follows:
ρcp DT ∂ k ∂T
Ṗs,T = − ≥0 (3.271)
T Dt ∂xj T ∂xj
where the right side must be equal or greater than zero, according to the second law of ther-
modynamics. However, numerical approximation errors may violate the required positivity
as a consequence of the nonzero entropy production difference. Approximation errors lead-
ing to a negative numerical entropy production are anticipated to lead to nonphysical trends
in the predicted results that violate the Second Law. This could provide valuable insight into
the physical realizability of numerical predictions in the absence of experimental data to val-
idate the simulation results.
Post-processing of the temperature results can be used to determine their physical plau-
sibility based on the sign computed in Equation 3.271 and magnitude of the computed
138 Advanced Heat Transfer
entropy production difference. If the computed entropy production rate is negative due to
approximation errors, and/or the computed entropy production difference is nonzero, then
this provides a useful indicator for the implausibility of the results due to components of the
numerical model, spatial discretization, or other factors. This methodology will be illus-
trated in the following example.
∂T ∂2 T
ρcp =k 2 (3.272)
∂t ∂x
Heat is conducted gradually into the bar and the temperature disturbances are propa-
gated into the positive x-direction. Define δ as the distance of thermal wave propagation.
A linear approximation of the temperature profile is illustrated in Figure 3.26.
Using a scaling approximation to estimate the characteristic temperature, time and
length scales, the linearized derivatives of temperature in the heat equation can be approx-
imated as:
T s − To Ts − To
ρcp
k (3.273)
t−0 δ2
This yields a characteristic length of δ 2 = αt which leads to the following wall heat flux,
∂T T s − To Ts − To
q′′ = −k
k
k √ (3.274)
∂x 0 δ αt
It can be shown that the exact solution using a similarity analysis is the same except with
√
an additional factor of π in the denominator. Thus, the scaling analysis provides the cor-
rect order of magnitude of the heat flux except for the leading coefficient.
(a) (b)
T x
Ts Entropy
Approximate T(x,t) compliant
region
Increasing t
x = x(t)
T0
δ x t
FIGURE 3.26
Heat conduction in a long bar: (a) approximate temperature profile and (b) entropy compliant region.
Convection 139
az2 + bz + c ≥ 0 (3.276)
√
where z = x/ t and,
√
a = To − T s ; b= α Ts ; c = 2α(Ts − To ) (3.277)
Therefore the region satisfying this inequality is the entropy compliant region and con-
sistent with the second law. A schematic of the entropy compliant region is illustrated in
Figure 3.26.
In the entropy compliant region, the approximate solution satisfies the second law of ther-
modynamics. It indicates the region where the linearized approximation of temperature is
physically viable. The approximate profile can be used within the entropy compliant region
otherwise it represents a solution that is nonphysical and may yield misleading trends of
variables, correlations, or processes based on the solution.
T − T1 ∗ Ṗs
θ= ; Ṗs = (3.280)
Ts − T1 k/L2
140 Advanced Heat Transfer
As shown earlier in this chapter, the Nusselt number (Nu) and skin friction coefficient (cf)
may be expressed in terms of the nondimensional velocity and temperature gradients at the
wall (y* = 0) as follows,
1 ∂u∗
ReL cf = ∗ (3.281)
2 ∂y 0
hL ∂θ
Nu = = ∗ (3.282)
k ∂y 0
Consider a shear layer region where these correlations can be applied using the boundary
layer assumptions. The cross-stream gradients of velocity and temperature are assumed to
be much larger than streamwise gradients in the flow direction.
The nondimensional entropy generation can be expressed in terms of the temperature and
velocity gradients. For two-dimensional incompressible flows in a boundary layer, the sim-
plified and reduced positive-definite form of the entropy production rate, Equation 3.255,
becomes:
μ ∂u 2 k ∂T 2
Ṗs = + 2 (3.283)
T ∂y T ∂y
V2
Ec = (3.286)
cp ΔT
Therefore, the nondimensional entropy generation number can be expressed as the sum of
frictional and thermal irreversibilities in the form of both skin friction and Nusselt number
correlations. Both forms of irreversibilities can be expressed in a single correlation. For cases
where analytical solutions are not available, existing correlations of the skin friction and
Nusselt number can be combined together and characterized uniquely in terms of the non-
dimensional entropy generation number.
Convection 141
PROBLEMS
3.1 Consider the processes of energy exchange between flat plates (Couette flow). In the
first case, the upper plate moves at a constant and steady velocity, U, in the positive
x-direction, and the lower plate is stationary (a wall). In the second case, consider a
deformable solid subjected to an applied force at its upper boundary, while maintain-
ing contact with a solid surface such that the wall shear stress is zero at the lower wall.
Determine the distributions of shear stress, internal energy, and kinetic energy within
the fluid between the plates.
3.2 In this problem, Bernoulli’s equation is derived by integrating the mechanical energy
equation (or total energy minus internal energy) along a streamline within a duct.
Two-dimensional steady flow through a duct of varying cross-sectional area per
unit area is considered. Start with the total energy equation and convert the external
work term (work done by gravity) to a potential energy form. Integrate the resulting
equation over a cross-sectional area, A(ds), where A refers to the local cross-sectional
area. Then integrate from the inlet (1) to the outlet (2), and subtract the analogous inte-
grated form of the internal energy equation to obtain Bernoulli’s equation for com-
pressible flow. The same result can be obtained by performing the integration of
the mechanical energy equation across A(ds) and from the inlet to the outlet instead.
Assume uniform velocity profiles at the inlet and the outlet. What expression is
obtained for the head loss term?
3.3 Moist air at 22◦ C flows across the surface of a body of water at the same temperature.
If the surface of the body of water recedes at 0.2 mm/h, what is the rate of evaporation
of liquid mass from this body? Explain how this evaporation rate can be used to deter-
mine the rate of convective heat transfer from the water to the moist air. It may be
assumed that the body of water is a closed system, with the exception of mass outflow
due to evaporation.
3.4 In a manufacturing process, a surface of area 0.7 m2 loses heat by convection (without
evaporative cooling) to a surrounding airstream at 20◦ C with an average convec-
tive heat transfer coefficient of 20 W/m2 K. The surface temperature of the component
is 30◦ C. The same manufacturing process is performed on another, more humid
day (50% relative humidity) at an ambient air temperature of 32◦ C. In this case, the
surface is saturated and a surface liquid film is formed at 36◦ C. Using the same con-
vection coefficient for both cases, what proportion of the total heat loss is due to evap-
orative cooling during the humid day? Use a diffusion coefficient (air −H2O) of
2.6 × 10−5 m2/s.
3.5 A thin liquid film covers an inclined surface in an industrial process. Evaporative
cooling of the surface film occurs and convective heating by air at 35◦ C flowing
past the surface. What is the saturated vapor pressure of the liquid when the mea-
sured surface temperature is 10◦ C? The film properties are given as follows: hfg = 180
kJ/kg, DAB = 0.6 × 10−5 m2/s, and R = 0.1 kJ/kg. Assume steady-state conditions.
3.6 Under what conditions of pressure and temperature will sublimation occur?
3.7 Determine the nondimensional functional relationship for the total heat transfer from
a convectively cooled solid material. Express your result in terms of the Biot number
and nondimensional time and temperature.
3.8 A similarity analysis is applied to a problem involving transient heat conduction.
Consider a semi-infinite body (or finite body at early stages of time) at a temperature
142 Advanced Heat Transfer
T − T1
θ(η) =
U2 /(2cp )
Give appropriate boundary conditions for this problem and find the
temperature rise in the boundary layer as a function of η and the Prandtl
number.
c. The dimensionless wall temperature rise in part (b) can now be evaluated
numerically by using an analytical approximation for f(η) from the Blasius
solution. Compare these results with the earlier scaling results in part (a).
Use the scaling analysis to show that θ has an order of magnitude of O(Pr)
and O(1) in the limit as Pr → 0 and Pr → ∞, respectively.
3.11 Consider a scaling analysis of transient heat conduction in a long bar. Assume one-
dimensional heat conduction in a bar, initially at T = To, with the left boundary (x = 0)
heated and maintained at T = Ts. Use the discrete scaling method to find the heat flow
across this left boundary.
3.12 Liquid propane at 27◦ C flows along a flat surface with a velocity of 1 m/s. At what
distance from the leading edge of the surface does the velocity boundary layer reach
a thickness of 6 mm?
3.13 Water at 27◦ C enters an inlet section between two parallel flat plates at a velocity of
2 m/s. If each plate length is 10 cm, then what gap spacing between plates is required
for the hydrodynamic boundary layers to merge in the exit plane?
3.14 Air at 17◦ C flows past a flat plate at 77◦ C with a velocity of 15 m/s.
Convection 143
a. Find the convective heat transfer coefficient at the midpoint along a plate of
length 1 m.
b. What plate width is required to provide a heat transfer rate of 1 kW from the
plate to the surrounding airstream?
3.15 A moist airstream with a velocity of U and a temperature of T∞ flows over a container
holding liquid at a temperature of Ts. The freestream concentration of water vapor
(subscript a) is ρa,∞ and the relative humidity is φ∞. Find an expression for the convec-
tive mass transfer coefficient, hm, in terms of the measured recession rate of the
liquid, dh/dt, in the container.
3.16 Heated air passes vertically through a vertical channel consisting of two parallel
plates. One plate is well insulated and the other plate is maintained at a temperature
of 20◦ C. The air enters the passage at 200◦ C with a uniform velocity of 8 m/s. The
width and height of each plate are 80 cm and 60 cm, respectively.
a. What spacing between the plates should be used to reach a fully developed
flow condition at the channel outlet?
b. At what position within the channel does the mean temperature fall 5◦ C
below its inlet value?
3.17 A hot-wire anemometer is a device that can measure gas velocities over a wide range
of flow conditions. The most common materials for anemometers are platinum and
nickel alloy wires as resistivity variations with temperature are very small for these
materials. Consider a fine platinum wire with a diameter, surface temperature, and
resistivity of 0.9 mm, 400 K, and 268.8 Ω/m, respectively (note: 1 W = 1 A 2 Ω). The
wire is exposed to an airstream at a temperature of 300 K.
a. If the gas velocity is 2 m/s, calculate the current flow through the wire.
b. Find a relationship between small changes in the gas velocity, ΔV, and the
current, ΔI. In other words, consider two experiments, namely, parts
(a) and (b). Describe a technique that determines the gas velocity difference
in terms of the current difference between both experiments.
3.18 Air at −5◦ C flows past an overhead power transmission line at 10 m/s. A 5 mm thick
ice layer covers the 1.9 cm diameter cable. The cable carries a current of 240 A with an
electrical resistance of 4 × 10−4 Ω/m of cable length. Estimate the surface temperature
of the cable beneath the ice layer.
3.19 A pipeline carries oil above the ground over a distance of 60 m where the surrounding
air velocity and temperature are 6 m/s and 2◦ C, respectively. The surface temperature
of the 1 m diameter pipeline is 17◦ C. What thickness of insulation (k = 0.08 W/mK) is
required to reduce the total heat loss by 90% (as compared with an uninsulated pipe
with the same surface temperature)?
3.20 Metallic spheres (diameter of 2 cm) at 330◦ C are suddenly removed from a heat treat-
ment furnace and cooled by an airstream at 30◦ C flowing past each sphere at 8 m/s.
Find the initial rate of temperature change of each iron sphere. Does this rate of
change vary significantly throughout the sphere? Neglect radiative heat exchange
with the surroundings.
3.21 In a reaction chamber, spherical fuel elements are cooled by a nitrogen gas flow
at 30 m/s and 210◦ C. The 2-cm-diameter elements have an internal volumetric
heating rate of 9 × 107 W/m3. Determine the surface temperature of the fuel
element.
144 Advanced Heat Transfer
3.22 Engine oil flows through the inner tube of a double-pipe heat exchanger with a mass
flow rate of 1 kg/s. An evaporating refrigerant in the annular region around the inner
tube absorbs heat from the oil and maintains a constant wall temperature of Tw. The
thermal conductivity of the inner tube is 54 W/mK, and the inner and outer radii are 3
and 3.2 cm, respectively. The oil enters the heat exchanger at a temperature of 320 K
and flows over a distance of 80 m through the tube. Find the required saturation tem-
perature of the refrigerant, Tw, to cool the oil to 316 K at the outlet of the tube.
3.23 A long steel pipe with inner and outer diameters of 2 and 4 cm, respectively, is heated
by electrical resistance heaters within the pipe walls. The electrical resistance ele-
ments provide a uniform heating rate of 106 W/m3 within the pipe wall. Water flows
through the pipe with a flow rate of 0.1 kg/s. The external sides of the electrically
heated walls are well insulated. Is the water flow fully developed at the pipe outlet?
If the mean temperatures of the water at the inlet and outlet are 20◦ C and 40◦ C,
respectively, then find the inner wall temperature at the outlet.
3.24 Oil transport through underground pipelines poses various challenges in the
petroleum industry. Pipe heat losses may significantly increase pumping power
requirements since the oil viscosity increases at lower oil temperatures. Consider
an oil flow within a long underground insulated pipe. Assume fully developed con-
ditions and constant thermophysical properties for oil (ρ = 90 kg/m3, cp = 2 kJ/kgK,
v = 8.5 × 10−4 m2/s, and k = 0.14 W/mK), insulation (k = 0.03 W/mK), and soil
(k = 0.5 W/mK).
a. Develop an expression for the mean oil temperature as a function of distance,
x, along the pipe. (Hint: shape factor for a cylinder buried in a semi-infinite
medium is 2πrL/ln(4z/D).)
b. Find the mean oil temperature at the pipe outlet, where x = L = 250 km. The
mean oil inlet temperature, mass flow rate, ground surface temperature
(above the pipe), pipe depth, and insulated pipe radii (inner and outer) are
110◦ C, 400 kg/s, 5◦ C, 6, 1.2, and 1.5 m, respectively.
3.25 Pressurized water at 200◦ C is pumped from a power plant to a nearby factory
through a thin-walled circular pipe at a mass flow rate of 1.8 kg/s. The pipe diameter
is 0.8 m. A layer of 0.05 m thick insulation with a conductivity of 0.05 W/mK covers
the pipe. The pipe length is 600 m. The pipe is exposed to an external air cross flow
with a velocity of 5 m/s and T∞ = –10◦ C. Develop an expression for the mean water
temperature as a function of distance, x, along the pipe. Also, find the mean water
temperature at the outlet of the pipe.
3.26 The purpose of this problem is to derive the Nusselt number for fully developed slug
flow through an annulus with inner and outer diameters of diand do, respectively. A
uniform heat flux, qwi, is applied at the inner surface and the outer surface is main-
tained at a constant temperature, Two. For slug flow, assume that the velocity is
approximately constant (um; mean velocity) across the pipe.
a. Explain the assumptions adopted to obtain the following governing energy
equation:
∂T k ∂ ∂T
ρcp um = r
∂x r ∂r ∂r
b. Explain how the Nusselt number (as a function of ro/ri) can be obtained
from the results in part (a) (without finding the explicit closed-form solution
for Nu).
3.27 Air flows at 0.01 kg/s through a pipe with a diameter and length of 2.5 cm and 2 m,
respectively. Electrical heating elements around the pipe provide a constant heat
flux of 2 kW/m2 to the airflow in the pipe. The air inlet temperature is 27◦ C. Assume
that thermally and hydrodynamically developed conditions exist throughout
the pipe.
a. Estimate the wall temperature at the pipe outlet.
b. Will additional pipe wall roughness increase or decrease the values of the
convection coefficient and friction factor? Explain your responses.
c. Various options are available for enhancing the heat transfer in this problem,
including longitudinal internal fins or helical ribs. What adverse effects might
these schemes have on the related fluid mechanics of this problem?
3.28 A vertical composite wall in a building consists of a brick material (8 cm thick with
k = 0.4 W/mK) adjoined by an insulation layer (8 cm thick with k = 0.02 W/mK)
and plasterboard (1 cm thick with k = 0.7 W/mK) facing indoors. Ambient air tem-
peratures outside and inside the wall are −20◦ C and 25◦ C, respectively. If the wall
is 3 m high, find the rate of heat loss through the wall per unit width of the wall.
3.29 An assessment of heat losses from two rooms on the side of a building is required. The
walls of rooms A and B have lengths of 12 and 8 m, respectively, and a height of 4 m
with a wall thickness of 0.25 m. The effective thermal conductivity of the wall is
approximately 1 W/mK. Outside air flows parallel to the walls at 8 m/s with an ambi-
ent temperature of −25◦ C. The air temperatures in rooms A and B are 25◦ C and 18◦ C,
respectively, and the air velocity in each room is assumed to be zero. Calculate the
rate of total heat transfer through both external walls of rooms A and B of the
building.
3.30 A section of an electronic assembly consists of a board rack with a vertically aligned
circuit board and transistors mounted on the surface of each circuit board. Four lead
wires (k = 25 W/mK) conduct heat between the circuit board and each transistor case.
The circuit board and ambient air temperatures are 40◦ C and 25◦ C, respectively. The
transistor case and wire lead dimensions are 5 mm (width) × 9 mm (height) × 8 mm
(depth) and 3 × 11 × 0.2 mm, respectively. The transistor case is mounted with a
1-mm gap above the board. The radiation exchange and heat losses from the board,
and case edges may be neglected.
a. Explain how the thermal resistances are assembled together in a thermal cir-
cuit for this problem.
b. The transistor temperature should not exceed 70◦ C for reliable and effective
performance. Consider the options of: (i) a stagnant air layer (ka = 0.028
W/mK); or (ii) a filler paste material (kp = 0.1 W/mK) in the gap between
the transistor case and the board. Which option would permit the highest
transistor heat generation without exceeding the temperature limit?
3.31 Water at a mean temperature of 70◦ C with a mean velocity of 1 m/s flows from a heat
exchanger through a long copper pipe (diameter of 1 cm) inside a building. The
poorly insulated pipe loses heat by free convection into the room air at 20◦ C. Find
the rate of heat loss per meter length of pipe. It may be assumed that the insulation
146 Advanced Heat Transfer
and pipe wall resistances are small in comparison to the thermal resistances associ-
ated with convection.
3.32 Perform a discrete scaling analysis for natural convection from a vertical plate. Con-
sider a flat plate maintained at T = Ts adjacent to a quiescent freestream at a temper-
ature of T = T∞. Find expressions for the boundary layer thickness, in terms of the
Prandtl number, as well as the average heat transfer coefficient and Nusselt number,
in terms of the Prandtl and Rayleigh numbers. Compare your result with other avail-
able correlations for natural convection.
3.33 Consider an integral analysis of free convection heat transfer from a sphere.
a. Derive expressions for the local and average Nusselt numbers, based on an
integral analysis, for natural convection heat transfer from a sphere.
b. How does surface curvature affect the results in part (a)?
c. Does your result approach the correct limits as Ra → ∞ and Ra → 0?
3.34 An inlet section of an air duct of total length L consists of a central duct carrying the
main (inner) inflow with a surrounding (outer) flow that independently carries a sep-
arate inflow stream. The inner velocity, u, and corresponding mass flow rate must be
found in order to transport a fixed outer flow across a specified pressure difference,
that is, Δp = p2 − p1 across the duct section. Use an integral analysis to find the
required inner velocity in terms of the inlet pressure, outlet pressure, and duct areas
(inner and outer). Assume uniform velocity profiles in the cross-stream (y) direction.
3.35 In a manufacturing process, a six-sided metal rod is heated by laminar free convec-
tion. The rod is oriented with the corner facing upward. The vertical sides have twice
the length (2L) of the other sides (L). The ambient air temperature is T∞ and the wall
temperature is Tw (where Tw , T∞).
a. Calculate the average Nusselt number for this configuration in terms of the
side length, L. What average heat transfer rate (per unit depth of rod) is
obtained when T∞ = 70◦ C, Tw = 20◦ C and L = 6 cm? Explain how and why
NuL will change if the vertical sides are lengthened (while maintaining the
same total surface area).
b. How would the heat transfer be altered if surface curvature effects were
included, that is, higher or lower NuL in comparison to part (a)? Explain
your response.
3.36 Small indentations along a vertically oriented surface are proposed for more effective
convective cooling of an electronic assembly. The lengths of the vertical and indented
surfaces are L1 and L2, respectively. The indented triangular cavity consists of both
surfaces inclined at 45◦ with respect to the vertical direction. The ambient air temper-
ature is T∞ and the wall temperature is Tw (where Tw . T∞).
a. Calculate the average Nusselt number for natural convection in this configu-
ration. The total side length, S, corresponds to N sets of vertical surface and
cavity sections. Find the total heat flow from the surface (per unit depth)
when T∞ = 17◦ C and Tw = 57◦ C. Express the answer in terms of L1, L2, and N.
b. Find the percentage increase of heat flow due to surface indentations with
L1 = 1 cm and L2 = 1 cm, in comparison to no surface indentations, while
maintaining the same gap spacing between the vertical sections. How does
the result depend on the relative magnitudes of L1 and L2? Explain
your response.
Convection 147
3.37 Liquid benzene is heated by free convection by a vertical plate of 10 cm height with a
surface temperature of 30◦ C. If the lowest allowed benzene temperature is 20◦ C, then
what minimum plate width is required to provide at least 300 W of heat transfer to the
liquid benzene?
3.38 An electrically heated rod of 4 cm height with an electrical resistance of 0.3 Ω/m is
immersed in a liquid acetic acid bath at 15◦ C. The rod’s diameter is 3 cm. Estimate
the electrical current required to provide an average surface temperature of 45◦ C
along the cylinder. Assume that heat transfer occurs predominantly by free
convection.
3.39 Thin metal plates are suspended in air at 25◦ C before processing in a manufacturing
operation. What initial plate temperature is required to produce an initial rate of tem-
perature change of 0.7 K/s for a single plate? The same length of each 1 mm thick
square iron plate is 40 cm. Neglect heat exchange by radiation.
3.40 Optimizing the parameters within a pipe flow has important implications in the
design of heat exchangers, underground oil pipelines carrying and many other appli-
cations. Consider steady, fully developed pipe flow with a fixed mass flow rate, ṁ,
and rate of heat transfer (per unit length), q’, to the pipe. Find the optimal pipe diam-
eter based on fixed values of ṁ and q’ and the method of entropy generation
minimization.
3.41 Determine the optimal sphere diameter (based on entropy generation minimization)
for external flow of a fluid at U and T∞ past a sphere of diameter D heated by a fixed
value of q. Express your answer in terms of q, D, U, T∞, and thermophysical proper-
ties. Use drag and heat transfer correlations of cD = 0.5 and NuD = C Re 1/2 for external
flow past the sphere, where C = C(Pr).
3.42 A heat exchanger design involves turbulent flow of water at 310 K (average fluid
temperature) through a tube while requiring a mass flow rate of 12 kg/s and a tem-
perature rise of 6 K/m. Find the optimal diameter based on entropy generation
minimization.
3.43 Air flows at 20 m/s and 400 K across a plate heated uniformly at 100 W/m. Find the
optimal plate length based on the method of entropy generation minimization.
3.44 Turbulent flow of air past a plate of length L and width W is encountered in an elec-
tronic assembly. Under these conditions, the following correlations for heat transfer
(Nusselt number) and friction coefficient are adopted:
τx
cf ,x = = 0.0576Re−1/5
ρ1 U 2 /2 x
where U and T∞ refer to the freestream velocity and temperature, respectively. The
plate is subjected to a uniform wall heat flux, q". Explain why an optimal plate length,
Lopt, exists for this external turbulent flow and find the optimal length based on
entropy generation minimization. Express your answer in terms of q" (or q’ = q"L),
U, T∞, Pr, and relevant thermophysical properties.
3.45 A brass cylindrical fin is joined to the top surface of an electronic component to
enhance convective cooling of the component.
148 Advanced Heat Transfer
Correlations for heat transfer and the drag coefficient, cd, in regard to external flow
past the fin, are known as follows:
where U, NuD, Pr, and ReD refer to the incoming velocity and Nusselt, Prandtl,
and Reynolds numbers, respectively. The freestream temperature is T∞ and
the fin is required to transfer a fixed rate of heat flow, q, from the base to the free-
stream air.
a. Find the optimal fin length, L, based on the method of entropy generation
minimization. Express your answer in terms of T∞, q, fin diameter (D), U,
and thermophysical properties.
b. Give a physical interpretation to explain how this optimum could be used to
deliver better convective cooling of the microelectronic system.
3.46 Consider one-dimensional transient fully developed laminar flow between two
parallel plates spaced a distance of L apart (Couette flow). The upper plate tempera-
ture is TL and it moves at a velocity of U. The lower plate temperature is stationary at a
temperature of T0. Determine the nondimensional entropy generation number based
on the Nusselt number and skin friction coefficient.
References
E. Achenbach. 1978. “Heat Transfer from Spheres up to Re = 6 × 106,” Proceedings of the Sixth Interna-
tional Heat Transfer Conference, Vol. 5, Hemisphere, Washington, DC.
O.B. Adeyinka and G.F. Naterer. 2002. “Apparent Entropy Production Difference with Heat and Fluid
Flow Irreversibilities,” Numerical Heat Transfer B, 42: 411–436.
A. Bejan. 1996. Entropy Generation Minimization, Boca Raton: CRC Press/Taylor & Francis.
T.L. Bergman, A.S. Lavine, F.P. Incropera, and D.P. DeWitt. 2011. Fundamentals of Heat and Mass Trans-
fer, 7th Edition, New York: John Wiley & Sons.
S.W. Churchill. 1983. “Free Convection around Immersed Bodies,” in Heat Exchanger Design Handbook,
E.V. Schlunder, Ed., Sec. 2.5.7, pp. 1–31. New York: Hemisphere.
S.W. Churchill and H.H.S. Chu. 1975. “Correlating Equations for Laminar and Turbulent Free Convec-
tion from a Vertical Plate,” International Journal of Heat and Mass Transfer, 18: 1323–1329.
I. Dincer and M.A. Rosen. 2013. Exergy: Energy, Environment and Sustainable Development, 2nd Edition,
Amsterdam: Elsevier.
W.H. Giedt. 1949. “Investigation of Variation of Point Unit Heat-Transfer Coefficient Around a Cylin-
der Normal to an Air Stream,” Transactions of the ASME, 71: 375–381.
K. Hanjalic and B.E. Launder. 1972. “Reynolds Stress Model of Turbulence and its Application to Thin
Shear Flows,” Journal of Fluid Mechanics, 52: 609–638.
G.F. Hewitt, G.L. Shires, and Y.V. Polezhaev, Eds., 1997. International Encyclopedia of Heat and Mass
Transfer, Boca Raton: CRC Press/Taylor & Francis.
J.O. Hinze. 1975. Turbulence, 2nd Edition, New York: McGraw-Hill.
K.G.T. Hollands, T.E. Unny, G.D. Raithby, and L. Konicek. 1976. “Free Convective Heat Transfer
across Inclined Air Layers,” ASME Journal of Heat Transfer, 98(2): 189–193.
S. Kakac, Y. Yener, and A. Pramuanjaroenkij. 2013. Convective Heat Transfer, 3rd Edition, Boca Raton:
CRC Press/Taylor & Francis.
W.M. Kays and M.E. Crawford. 1980. Convective Heat and Mass Transfer, New York: McGraw-Hill.
Convection 149
K. Kirtley. 1992. “Renormalization Group Based Algebraic Turbulence Model for Three-Dimensional
Turbomachinery Flows,” AIAA Journal, 30: 1500–1506.
F. Kreith, R.M. Maglik, and M.S. Bohn. 2010. Principles of Heat Transfer, 7th Edition, Pacific Grove:
Brooks/Cole Thomson Learning.
S. Lee, M.M. Yovanovich, and K. Jafarpur. 1971. “Effects of Geometry and Orientation on Laminar
Natural Convection from Isothermal Bodies,” AIAA Journal of Thermophysics and Heat Transfer,
5: 208–216.
L.F. Moody. 1944. “Friction Factors for Pipe Flow,” Transactions of the ASME, 66: 671–684.
Y. Muzychka and M.M. Yovanovich. 2016. “Convective Heat Transfer,” in Handbook of Fluid Dynamics,
R. Johnson, Ed., pp. 14‐1−14‐58 Boca Raton: CRC Press/Taylor & Francis.
G.F. Naterer and J.A. Camberos. 2008. Entropy Based Design and Analysis of Fluids Engineering Systems,
Boca Raton: CRC Press/Taylor & Francis.
S. Ostrach. 1953. “An Analysis of Laminar Free Convection Flow and Heat Transfer about a Flat Plate
Parallel to the Direction of the Generating Body Force,” National Advisory Committee for Aeronau-
tics, Report 1111.
G.D. Raithby and K.G.T. Hollands. 1975. “General Method of Obtaining Approximate Solutions to
Laminar and Turbulent Free Convection Problems,” in Advances in Heat Transfer, T.F., Irvine,
Jr. and J.P. Hartnett, Eds., Vol. 11, New York: Academic Press, pp. 265–315.
W. Rodi. 1984. Turbulence Models and Their Application in Hydraulics, Brookfield, VT: Brookfield
Publishing.
H. Schlichting. 1979. Boundary Layer Theory, New York: McGraw-Hill.
H. Tennekes and J. L. Lumley. 1972. A First Course in Turbulence, Cambridge: MIT Press.
S. Whitaker. 1972. “Forced Convection Heat transfer Correlations for Flow in Pipes, past Flat Plates,
Single Cylinders, Single Spheres, and Flow in Packed Beds and Tube Bundles,” AIChE Journal,
18: 361–371.
F. White. 2015. Fluid Mechanics, 8th Edition, New York: McGraw-Hill.
A. Zhukauskas. 1972. “Heat Transfer from Tubes in Cross Flow,” in Advances in Heat Transfer,
J.P. Hartnett and T.F. Irvine, Jr., Eds., Vol. 8, New York: Academic Press.
4
Thermal Radiation
4.1 Introduction
Radiative heat transfer is a form of energy emitted as electromagnetic waves or photons by
all matter above a temperature of absolute zero (−273.15◦ C or 0 K). This transport process
varies with temperature, wavelength, and direction. Thermal radiation occurs purely due to
the temperature of a source and therefore does not include nonthermal forms of radiation
such as gamma rays due to nuclear reactions or certain forms of X-rays.
Any object above a temperature of absolute zero contains molecules with electrons that
are situated at discrete energy levels (called quantized energy states). An object at a high
temperature has more electrons located at higher energy quantum levels than a cold object.
When these electrons fluctuate between different quantum states, due to temperature
disturbances of the surface, the fluctuations generate electromagnetic waves that are emitted
as a result of quantum disturbances. Changes in the electron configurations of constituent
atoms and molecules in the object lead to electromagnetic wave emissions, which transfer
heat in the form of radiation.
Radiative properties of a surface characterize how a surface emits, reflects, absorbs, and
transmits radiation. A blackbody is an ideal radiator that absorbs all incoming radiation, at
all wavelengths and directions, while not reflecting any radiation. The ratio of an actual sur-
face’s radiative absorption to a blackbody’s absorption is called the absorptivity, α. Similarly,
ϵ, ρ, and τ refer to the emissivity (fraction of radiation emitted relative to a blackbody), reflec-
tivity (fraction of radiation reflected by the object), and transmissivity (fraction transmitted
through the object), respectively. The radiative emission from a diffuse emitter is independent
of direction. This includes both surface-emitting radiation, equally in all directions, and
incoming radiation (called irradiation), absorbed equally from all directions at any point
on the surface. Also, an opaque medium refers to an object with a zero transmissivity (no
energy transmitted through the object).
A blackbody may refer to a single surface or an object or group of surfaces. For example,
an isothermal cavity containing a small hole, through which radiation passes or enters, can
be considered a blackbody also because it possesses the properties of a blackbody. There is
complete absorption of the incident radiation by the cavity, regardless of the condition of the
surfaces comprising the walls of the cavity. Since all irradiation is absorbed, the entire cavity
is seen by another object to behave like a blackbody, even though the surfaces within the
cavity may not be black. Also, irradiation within the cavity is diffuse and emissions from
the cavity through the hole are diffuse. As a result, the cavity exhibits all of the necessary
characteristics of a blackbody.
The structure of a material affects the radiative properties of its surface. For example, pol-
ished surfaces often have a higher reflectivity and lower absorptivity than painted and metal
151
152 Advanced Heat Transfer
e = hν (4.1)
c = λν (4.2)
The speed of light is 299,792,458 m/s in a vacuum. As an electromagnetic wave, the wave
has both electric and magnetic field characteristics. It is transmitted as synchronized oscil-
lations of electric and magnetic fields that propagate at the speed of light. The oscillations
of the electric and magnetic fields are perpendicular to each other. They are also perpendic-
ular to the direction of energy and wave propagation, thereby forming a transverse wave.
Radiation occurs across an electromagnetic spectrum over a wide range of wavelengths
and frequencies (see Figure 4.1). Thermal radiation generally refers to the range between
0.1 and 100 µm (encompassing ultraviolet, visible, and infrared regions) as the majority of
thermal engineering applications occur in this range. From very small wavelengths, below
Thermal Radiation 153
λ (µm)
10–5 10–4 10–3 10–2 10–1 100 101 102 103 104 105
Visible
0.4 0.7
λ (µm)
Violet
Blue
Green
Yellow
Orange
Red
FIGURE 4.1
Electromagnetic spectrum of radiation.
10−5 µm (or very high frequencies), to the longest wavelengths in the microwave region
(above 102 µm), the spectrum characterizes the properties of radiation. Common daily
experiences can be explained through this electromagnetic spectrum. For example, the color
of a rainbow is related to the radiative properties of the atmosphere in the visible range of the
spectrum. The human eye is remarkable in how it distinguishes incoming waves in the range
of visible light (between 0.4 and 0.7 µm) in the form of different colors.
Gamma rays are mainly of interest to astronomers and astrophysicists since these are gener-
ally encountered only in the transmission of signals from deep space. Between about 10−4 and
10−2 µm, X-rays are encountered in applications such as nuclear and medical applications. Fur-
ther up the scale, between about 0.01 and 0.4 µm, ultraviolet rays (UV) are encountered.
Chronic exposure to ultraviolet rays from the sun can be harmful to the human body, partic-
ularly the skin and eyes. Depletion of the ozone in the atmosphere, due to the release of chlo-
rofluorocarbons (CFCs) from refrigerants into the atmosphere, has a harmful effect since ozone
absorbs UV radiation. The next range is visible radiation, between 0.4 and 0.7 µm, which can be
further subdivided into regions interpreted by the human eye as white, violet, blue, green, yel-
low, and red. When an object is heated, its thermal energy increases and thus the frequency of
emitted radiation increases. This process explains why objects initially become red and even-
tually turn white when they are heated. The remaining ranges of the spectrum are the infrared
(between 0.7 and 100 µm) and microwave (above 100 µm) regions. Applications involving
radio wave propagation and electrical engineering often arise in the microwave region.
r r
dA
dl
dα = dl/r dω = dA/r2
z Emitted
(r,θ,Φ) z radiation
θ
θ
dA
dA1 dω
y
y
Φ
Φ
x x
FIGURE 4.2
Schematic of plane and solid angles.
Consider a beam of radiation that originates from a point on a surface and travels some
distance, r (see Figure 4.2). Upon arriving at another point, such as P, it forms a circular
cross-sectional plane, dAn, with a projected unit normal vector, n. The growing conical
(three-dimensional) beam is enclosed by a solid angle, dω, where:
dA
dω = (4.3)
r2
The units of the solid angle are steradians (sr). Consider an amount of radiant energy, d 2Q,
which travels one-way past dAn in the direction of dω. A second order differential, d 2Q, is
used because the radiant energy depends on both dAn and dω. The intensity of radiation
at r in the direction of d is defined as:
d2 Q
I(r, d) = (4.4)
dAn dω
The solid angle, dω, represents a useful parameter to identify the heat flux or radiation
intensity across a particular projected area before it arrives at a surface. A schematic of plane
and solid angles is illustrated in Figure 4.2.
For radiation emanating from a flat surface in the direction of angle θ, relative to
the direction perpendicular to the plate, and arriving at a second horizontal surface
Thermal Radiation 155
dA cos θ
dω = (4.5)
L2
where dA and θ refer to the surface area and directional angle of the incoming radiation mea-
sured with respect to the vertical direction. In general, if all outgoing beams of radiation
leaving a surface are considered, then integration over the hemispherical range of angular
directions leads to:
E = πIe (4.6)
This result indicates that the total emissive power, E, and total intensity of emitted radi-
ation, Ie, are equivalent within a factor of π. Note that the constant is π, not 2π, and it has units
of steradians.
Similarly, it can be shown that the radiation intensity can also be related to the total irra-
diation on a surface (incident radiation), G, and the total radiosity (the sum of emitted radi-
ation and reflected irradiation), J, as follows:
G = πIi (4.7)
J = πIe+r (4.8)
where Ii refers to the total intensity of incident radiation. Also, Ie+r is the total intensity of
both emitted and reflected components from a surface that is a diffuse reflector and
diffuse emitter.
A schematic of incident, emitted, reflected, absorbed and transmitted radiation is illus-
trated in Figure 4.3. Here “spectral” refers to a dependence on the wavelength, λ. For a trans-
parent medium in Figure 4.3, a portion of irradiation, αλGλ, may be absorbed as thermal
energy in the object, where αλ refers to the spectral absorptivity. The remaining components
of the incoming radiation may be reflected, ρλGλ, or transmitted, τλGλ, where ρλ and τλ refer
to the spectral reflectivity and transmissivity, respectively.
Radiosity Jλ = Eλ + ρλGλ
Incident
radiation ρλGλ Eλ
Gλ reflected emitted
irradiation
Transparent
medium αλGλ
absorbed
τλGλ
transmitted
FIGURE 4.3
Schematic of irradiation and radiosity.
156 Advanced Heat Transfer
The spectral radiosity, Jλ, is the sum of emitted radiation from the surface and the reflected
irradiation,
Jλ = E λ + ρ λ G λ (4.9)
The total spectral irradiation, Gλ, has reflected, absorbed, and transmitted components of
incident radiation as follows:
Alternatively, the individual components of the irradiation may be obtained by the total
irradiation multiplied by each respective radiative property,
Gλ = ρλ Gλ + αλ Gλ + τλ Gλ (4.12)
ρλ + αλ + τλ = 1 (4.13)
ρ+α+τ =1 (4.14)
Note that the emitted radiation, radiosity, and irradiation terms are based on an actual
surface area, whereas the intensity of radiation is based on a projected area.
C1
Eλ,T (λ, T) = = πIλ,b (λ, T) (4.15)
λ5 ( exp (C2 /λT) − 1)
where Iλ,b is the spectral radiation intensity. The subscripts b, λ and T denote a blackbody
and dependence on wavelength and temperature, respectively. The factor π is obtained
Thermal Radiation 157
from angular integration of the spectral distribution over the hemispherical range of radia-
tive emission by a surface. Also, C1 = 3.742 × 10−16 Wm2 and
hc
C2 = = 0.01439 [W/mK] (4.16)
B
where h is the Planck constant, B = 1.3805 × 10−23 J/K is the Boltzmann constant, and c is the
speed of light. The emissive power of a blackbody based on Planck’s law is illustrated in
Figure 4.4.
The spectral distribution of solar radiation can be approximated by blackbody radiation at
5,800 K with a maximum emissive power in the visible region of the spectrum. At a fixed
wavelength, the emissive power decreases with surface temperature. For surface tempera-
tures above approximately 700 K, a portion of the spectral emission lies within the visible
range. The spectra are shifted rightward outside the visible range below this temperature.
Also, the spectral distribution of emissive power decreases on both sides of the maxima.
Most electron activity is centered about a specific quantum level corresponding to a given
wavelength and frequency of radiation. Fewer fluctuations occur between energy states
at other quantum levels and so the spectral power decreases away from the peak value.
The locus of maxima of emissive power curves at different wavelengths is described by
Wein’s displacement law. As the surface temperature decreases, λmax increases. For example,
λmax ∼ 0.5 μm for solar radiation and it increases to 2.9 µm for a blackbody at 1,000 K. In order
to find the wavelength where the emissive power is maximized at a given surface temper-
ature, Planck’s law can be differentiated with respect to λ and the resulting expression is
equated to zero, yielding:
109
5,800 K Visible part
108
of spectrum
Blackbody emissive power, Ebλ (W/m2 µm)
107
Solar radiation
106
2,000 K
105
104 1,000 K
103
102
101
E bλ
(T
100
=c 3
/λ
10–1
)
10–2
300 K 100 K
10–3
10–4
0.1 0.2 0.4 1 2 4 10 20 40 100
Wavelength λ (µm)
FIGURE 4.4
Blackbody emissive power as a function of wavelength.
158 Advanced Heat Transfer
If a spectral distribution passes through only a part of the visible range (such as an object at
1,000 K), the color is discerned through the highest frequency photons visible to the eye. So
color is observed from the tail end of the spectral distribution. A tungsten filament in a light
bulb at 2,900 K appears white to the eye and λmax occurs at approximately 1 µm. In this case,
significant radiation emission occurs over the entire visible spectrum, leading to white light.
After the tungsten filament is turned off and it cools down below 1,300 K, its light becomes
barely visible.
A fundamental equation of thermal radiation is Stefan–Boltzmann’s law. It can be derived
from the integration of Planck’s law. Defining x = C2/(λT) and integrating Equation 4.15
over the entire range of wavelengths,
1
C1 T 4 x3 dx
Eb = 4 x − 1)
= σT 4 (4.18)
0 C 2 (e
where σ = 5.67 × 10−8 W/m2 K4 is the Stefan–Boltzmann constant. Here Eb represents the
total blackbody emissive power. Using this result, the radiation heat flux emitted from a sur-
face becomes:
This result represents Stefan–Boltzmann’s law. The surface emissivity, ϵ, is the ratio of the
radiant energy emitted from a surface to that radiated from a blackbody (a perfect emitter)
under the same viewing conditions.
The spectral distribution of emissive power for a blackbody in Planck’s law is smooth with
respect to varying wavelengths. However, for an actual surface, the distribution normally
has a lower magnitude and fluctuates with wavelength. Similar fluctuations are observed
for radiative surface properties of materials such as the emissivity. Various factors, such
as surface roughness, coatings, and material density, can lead to variations of surface scat-
tering of absorbed and reflected radiation, thereby leading to abrupt or fluctuating changes
with wavelength.
Define F(0λ) as the fraction of total blackbody radiation in the wavelength range from 0λ.
This fraction can be determined by integrating Eb,λ over the appropriate wavelength range,
λ λT
Eλ,b dλ Eb,λ
F(0λ) = 10 = d(λT) = f (λT) (4.20)
0
Eλ,b dλ 0 σT 5
Here Eb,λ is a function of λT only (using Planck’s law) so the band emission is written as a
function of this product alone.
Two important properties of the blackbody functions are given by:
Sample values of the blackbody function, F(0λ), at various products of λT are shown in
Table 4.1.
Using these blackbody functions, the fraction of emitted energy in any particular wave-
length region can be determined. A common example is finding the fraction of energy
Thermal Radiation 159
TABLE 4.1
Radiation Functions for a Blackbody
λT (μm·K) F(0!λ) λT (μm·K) F(0!λ)
emitted in the visible range (between 0.4 and 0.7 µm) by various light sources, such as a
tungsten filament (temperature of 2,900 K) or the sun (temperature of about 5,800 K). It
can be calculated known that about 7% of the energy emitted by a tungsten filament and
37% of emitted solar energy lies in the visible range of the spectrum.
To find the fraction of emitted energy in a specified range of wavelengths, λ1 ≤ λ ≤ λ2, and
temperature, first determine the products λ1T and λ2T. Then the band emission table
(Table 4.1) can be used to find the blackbody functions, F, for the ranges of 0 λ1 and
0 λ2, respectively. Then subtracting these two values gives the fraction of emitted energy
in the specified range of wavelengths.
properties with respect to the incident solar radiation. The surface could be designed to
absorb well in the direction of incoming radiation but poorly in other directions. This surface
would emit or lose less energy than a smooth surface emitting well in all angular directions.
As a result, more thermal energy could be retained by the solar collector, thereby increasing
its overall thermal efficiency.
The spectral emissivity, ϵλ, is defined as the actual emitted energy by the surface at a tem-
perature of T divided by the emitted energy by an ideal radiator (blackbody) at the same
temperature,
2π π/2
Eλ (λ, T) ελ,θ cos (θ) sin (θ) dθ dϕ
ελ (λ, T) = = 0 0
2π π/2 (4.23)
Eb,λ (λ, T) cos (θ) sin (θ) dθ dϕ
0 0
Not absorbed
by object 2
ρE1
Object
absorbed
2
FIGURE 4.5
Radiation exchange between two objects.
Thermal Radiation 161
The reflectivity, absorptivity and transmissivity are defined analogously to the emissivity,
but in terms of the incident radiation, Gλ,
1
ρλ Gλ dλ
ρ(T) = 0 1 (4.25)
0
Gλ dλ
1
αλ Gλ dλ
α(T) = 0 1 (4.26)
0
Gλ dλ
1
τλ Gλ dλ
τ(T) = 0 1 (4.27)
0
Gλ dλ
The energy emitted normally depends more strongly on surface temperature than
wavelength. However, in general, the spectral radiative properties of reflectivity, absorp-
tivity, and transmissivity are weak functions of temperature. These three latter properties
are more strongly dependent on wavelength than temperature. Also, the wavelength
dependence of the energy emitted or absorbed is closely related to the alignment of radi-
ative properties of the emitting surface and source of incident radiation. For example, the
energy absorbed over a particular wavelength range depends on the surface’s ability to
absorb radiation in the range of wavelengths corresponding to the incident radiation.
Consider some examples of spectral variations of radiative surface properties of vari-
ous materials, for example, tinted glass, aluminum film, brick materials, and painted
and polished surfaces. Tinted glass is transparent in the visible range, but opaque in
the infrared and ultraviolet ranges. An aluminum film is almost completely reflective
across all wavelengths, while white paint reflects well only between about 0.4 and
1 µm, and poorly otherwise. A black painted surface is a poor reflector in all wavelength
regions. Red paint alternates between good, moderate, and poor reflectivities across var-
ious wavelengths. Its properties are similar to brick materials consisting of clay, shale,
silicon carbide, and other materials. The decrease of reflectivity between 1 and 5 µm
occurs partly because clay and shale in the material exhibit a high absorptivity in that
range. A surface covered with black paint is a poor reflector but absorbs and emits
well. Conversely, polished aluminum reflects well but absorbs and emits weakly above
0.4 µm.
Kirchhoff’s law states that the emissivity at a given wavelength, λ, and direction, θ, must
equal the absorptivity at that given wavelength and direction,
Therefore, either of the following two conditions can be satisfied to establish the equiva-
lence between the total emissivity and absorptivity.
1. Irradiation arrives from a blackbody at the same temperature as the incident sur-
face. Therefore Gλ = Eb,λ and spectral integrations of emissivity and absorptivity
become identical.
2. Radiation exchange occurs with a gray surface, where ϵλ and αλ are both constant
and equal. If ϵλ is constant, then ϵ = ϵλ. Similarly, if αλ is constant, then α = αλ.
This requires that ϵ = α since ϵλ = αλ.
To summarize, Kirchhoff’s law requires that ϵλ,θ = αλ,θ for any surface. For a diffuse sur-
face, ϵλ = αλ. A gray surface has ϵλ and αλ which are independent of wavelength over the
dominant spectral regions of Gλ and Eλ. A diffuse gray surface has property characteristics
of both directional and wavelength (spectral) independence. In other words, ϵ and α are
identical at all angles and wavelengths, so ϵ = ϵλ = αλ = α. Under either condition 1 or 2
above, together with the diffuse property, ϵ = α for a diffuse gray surface.
Determine which diffuse surfaces at a temperature of 300 K are gray when exposed to
solar radiation.
From the electromagnetic spectrum, approximately 98% of solar irradiation occurs at
λ , 3 μm. Also, 96% of the energy emitted from a surface at 300 K occurs at λ . 6 μm,
with the peak occurring at approximately 10 µm. In comparing the four surfaces, a gray
surface occurs when the absorption of incoming energy over the spectral range of the irra-
diation matches the emission from the source of irradiation (sun) over the spectral range of
the source.
Therefore, the surfaces possess the following characteristics:
∙ αλ = 0.9 for Gλ and ϵλ = 0.9 for Eλ (gray);
∙ αλ = 0.7 for Gλ and ϵλ = 0.5 for Eλ (not gray);
∙ αλ = 0.2 for Gλ and ϵλ = 0.6 for Eλ (not gray);
∙ αλ = 0.1 for Gλ and ϵλ = 0.1 for Eλ (gray).
From this example, it can be observed that the radiative surface properties of gray surfaces
must match each other in the appropriate regions of the spectrum where the radiation
exchange occurs.
Thermal Radiation 163
Radiative surface properties are often used in temperature measurement devices. For
example radiation thermometry is a measurement technique for determining the surface tem-
perature of an object based on a radiance measurement at a particular wavelength(s) and the
surface emissivity. In dual-wavelength radiation thermometry (DWRT), measurements of
the radiance at two different wavelengths and the surface emissivity are used in a compen-
sation algorithm to determine the surface temperature. Wen and Lu (2010) used an emissiv-
ity model and multispectral radiation thermometry (MRT) to measure surface temperatures
in steel manufacturing processes. MRT uses radiance measurements at three or more wave-
lengths and an emissivity model to determine an object’s surface temperature.
where Ji is the radiosity of surface i. The view factor is also called the radiation shape factor or
the configuration factor. It will be used to determine the radiation exchange between surfaces
simultaneously emitting, absorbing, reflecting, and/or transmitting radiant energy.
Consider a surface of area dAi at a temperature of Ti emitting a beam of radiation toward
an element of surface area dAj at Tj and a distance L away. The normal vectors to each surface
are at angles of θi and θj, respectively, with respect to the line joining each of the differential
surface elements (see Figure 4.6). The solid angle formed by a beam of radiation leaving sur-
face i and spreading and arriving at surface j is given by:
dAi cos θi
dωij = (4.32)
L2
nj
ni, nj : unit normal vectors
ni θj
θi
L
Surface j
Aj, Tj
Surface i
Ai, Ti
FIGURE 4.6
Schematic of view factor between two surfaces.
164 Advanced Heat Transfer
Using Equation 4.6, the intensity of radiation, Ii, may be written as the radiosity, Ji (con-
sisting of energy emitted and reflected), divided by π.
Combining the previous equations, the total energy emitted from surface i that arrives at
surface j becomes:
cos θi cos θj
qij = Ji dAi dAj = Fij Ai Jj (4.34)
Ai Aj πL2
This result for the view factor is expressed in terms of geometric parameters only.
In a similar way, Fji represents the energy leaving surface j and arriving at surface i,
1 cos θi cos θj
Fji = dAi dAj (4.36)
Aj Ai Aj πL2
Comparing these two forms of the view factor yields the following reciprocity relation:
This allows either shape factor to be written in terms of the other factor based on the
respective area ratio between both surfaces.
View factors for basic two-dimensional geometries (e.g., parallel plates, inclined plates,
parallel cylinders, cylinder and parallel rectangle, etc.) and three-dimensional geometries
can be derived analytically. Common three-dimensional configurations are aligned parallel
rectangles and coaxial parallel disks (see Figure 4.7). View factors for perpendicular rectan-
gles with a common side are illustrated in Figure 4.8. A comprehensive source of radiation
view factors was provided by Howell (1982).
For example, the view factor between coaxial parallel disks (disk 1 of radius r1 at a distance
of h below another disk of radius r2) can be expressed as
1
F12 = B − B2 − 4(r2 /r1 )2 (4.38)
2
where,
1 + (r2 /h)2
B=1+ (4.39)
(r1 /h)2
The following example shows that view factors for basic geometries can often be obtained
by inspection.
Thermal Radiation 165
(a) 1 (b) 1
j 8
∞ j, rj
L 6 L
10 0.8
i Y i, ri
X 2 4
1 0.6
2
Fij
Fij
0.4
0.1 0.4
1
0.2
0.8
0.2
0.6
Y/L = 0.1
rj/L = 0.4
0.01 0
0.1 1 10 0.1 1 10
X/L L/ri
FIGURE 4.7
View factors of (a) aligned parallel rectangles and (b) coaxial parallel disks. (Adapted from J.R. Howell. 1982. Catalog
of Radiation Configuration Factors, New York: McGraw-Hill.)
0.6
Z
Aj Y
0.5 Ai
X
Y/X = 0.1
0.4
0.2
0.3 0.4
Fij
0.2 1.0
2
0.1
4
20
0
0.1 1 10
Z/X
FIGURE 4.8
View factors of perpendicular rectangles with a common side. (Adapted from J.R. Howell. 1982. Catalog of Radiation
Configuration Factors, New York: McGraw-Hill.)
is equally distributed among the upper surfaces due to symmetry, so F12 = 1/2 and F13 =
1/2. The other view factors can be inferred from these view factors based on the reciprocity
relation between respective surfaces and the ratio of the surface area of the base section to
the upper half dome.
None of the energy leaving surface 1 arrives back at that surface (excluding reflections
from other surfaces) since it is a flat surface. The view factors for flat and convex surfaces,
with respect to themselves, are zero. But concave surfaces may involve radiation arriving
back upon itself. If the base surface is concave, then the previous view factors of ½ would
be smaller since some of the radiation would be intercepted by surface 1.
A comprehensive list and analysis of radiation view factors for many geometrical config-
urations was reported by Howell (1982).
Jλ = Eλ + ρλGλ
Tj, Aj, εj
Surface J
Gj
Ei Ti, Ai, εi
Jj = radiosity
ρiGi Surface i
Jj
Heat
... Enclosure supply
(3)
ρiGi
Gj = irradiation
αiGi
T2, A2, ε2
(2) (n)
FIGURE 4.9
Radiation exchange in an enclosure.
Thermal Radiation 167
where qij denotes the radiative heat flow leaving surface i that is intercepted by surface j. The
first term on the right side may be nonzero if it is a concave surface (as discussed in the
previous example). The energy balance in Equation 4.40 states that the energy leaving
surface 1 balances the energy arriving at surface 1, surface 2, etc., up to surface n.
Dividing Equation 4.40 by JiAi and generalizing from surface 1 to surface i,
n
qij
1= (4.41)
j=1
A i Ji
where i = 1, 2, 3, … n. The lower limit in the summation represents the radiation leaving sur-
face i and arriving at surfaces j = 1, 2, 3, … n. Observing that the term in brackets is the view
factor yields the following summation relation:
n
Fij = 1 (4.42)
j=1
Alternatively, Equation 4.42 is called the enclosure relation. It is valid only for radiation
exchange within enclosures. Here Fii ≠ 0 for concave surfaces but Fii = 0 for plane or convex
surfaces. A convex surface does not intercept any of its outgoing radiation. The total heat
flow is obtained once the respective view factors are known.
In Figure 4.9, the temperature, area, and emissivity of surface i are denoted by Ti, Ai and ϵi,
respectively. Based on an energy balance for this surface,
qi = Ai (Ji − Gi ) (4.43)
where Ji, Gi, and qi refer to the radiosity, irradiation, and heat flow required to maintain sur-
face i at Ti, respectively. The energy balance requires that the energy needed to maintain sur-
face i at Ti is the net energy leaving the surface (radiosity minus irradiation).
The radiosity consists of the sum of emitted radiation, Ei, and reflected radiation, ρiGi. The
irradiation may be decomposed into a reflected component, ρiGi, and an absorbed portion,
αiGi, since none of the incident radiation is transmitted through the surface. Since each sur-
face is diffuse gray (αi = ϵi) and opaque (τi = 0),
ρi + αi = ρi + εi = 1 (4.44)
The blackbody radiation emitted, Ebi, can be evaluated by Stefan–Boltzmann’s law based
on a surface temperature of Ti.
168 Advanced Heat Transfer
Substituting the expression for Gi into Equation 4.43 and rearranging terms,
Ebi − Ji
qi = (4.46)
(1 − εi )/(εi Ai )
Therefore the net heat flow from surface i due to radiation exchange with other surfaces
can be expressed in terms of the radiosity at the surface, emissivity, surface area, and the
corresponding blackbody radiation emitted from the surface at a temperature of Ti.
This equality suggests that the irradiation arriving on surface i consists of radiation
leaving all other surfaces, j = 1, 2, … n. The fraction of radiation arriving from surface j is
multiplied by the area and radiosity of that surface.
Using the reciprocity relation to write the energy balance in terms of the area Ai and then
dividing by this area,
n
Gi = Fij Jj (4.48)
j=1
This result indicates that the irradiation arriving on surface i consists of the sum of view
factor-weighted radiosity contributions from all surfaces within the enclosure. Furthermore,
combining the previous equations, it can be shown that:
n
Ji = εTi4 + (1 − εi ) Fij Jj (4.49)
j=1
This result indicates that the radiosity leaving surface i includes the emitted radiation and
the reflected irradiation.
Using Equations 4.43 through 4.48, the heat transfer to/from surface i can then be
expressed in the following form,
n
qi = Ai Ji − Ai Fij Jj (4.50)
j=1
Thermal Radiation 169
Thus the heat required to maintain surface i at Ti balances the radiant energy leaving sur-
face i minus the energy arriving at surface i from all other surfaces. This energy balance
describes the same energy balance as Equation 4.43, but in terms of the radiosity and
view factors instead.
Using the summation relation with the previous result,
⎛ ⎞
n
n
n
n
qi = Ai ⎝ Fji Ji − Fij Jj ⎠ = Fij (Ji − Jj ) = qij (4.51)
j=1 j=1 j=1 j=1
Therefore the net heat loss or gain by surface i balances the sum of net radiative heat
exchanges between surface i and all other surfaces. It can be written in a more convenient
form as follows:
n
J i − Jj
qi = (4.52)
j=1
1/(Ai Fij )
This form indicates that the heat flow rate is given by the sum of potential differences
divided by individual resistances arising from radiation exchange with each surface. The
total resistance to radiation exchange among two surfaces includes the sum of resistances
of each surface due to their nonideal emissivities, based on Equation 4.46, and a spatial resis-
tance(s), based on Equation 4.52.
The form of Equation 4.52 allows a thermal circuit involving radiation exchange to be con-
structed (see Figure 4.10). The heat flow into surface i (node i of network) experiences a sur-
face resistance of (1 − ϵi)/(ϵiAi) with a radiosity of Ji, and then a group of spatial resistances,
(Ai Fij)−1, in parallel, corresponding to a heat flow of qij between surfaces i and j. If any of the
qi1
Tj, Aj, εj
Jj J1
Tk, Ak, εk
qi2
Jk
(AiFi→1)–1 J2
Gj (AiFi→2)–1
Ji
Ti, Ai, εi Ebi
qi
(AiFi→N)–1
(AiFi→3)–1
JN J3 qi3
...
qiN
FIGURE 4.10
Schematic of a thermal circuit for radiation exchange in an enclosure.
170 Advanced Heat Transfer
Surface
temperature, T2
q12 Enclosure
T1
FIGURE 4.11
Radiation exchange between two surfaces in an enclosure.
surfaces are blackbodies, then the same above equation may be used but with Ji replaced by
Ebi and ϵi = αi = 1.
The denominator includes the sum of both surface and spatial resistances. Using Stefan–
Boltzmann’s law and rearranging,
ε1 A1 σ T14 − T24
q12 =
(4.54)
(1 − ε1 ) + Fε121 + A 1 ε1
A2 ε2 (1 − ε2 )
A1, T1, ε1
A1, T1,
ε1
r1
A2, T2,
ε2 r2
A2, T2, ε2
FIGURE 4.12
Radiation exchange between (a) long parallel plates; (b) long concentric cylinders; (c) concentric spheres; and
(d) a small object in a large cavity.
Thermal Radiation 171
Using this general result, other special cases can be obtained as follows for two diffuse
gray surfaces forming an enclosure (see Figure 4.12).
a. For large parallel plates (surfaces 1 and 2), the view factor is F12 = 1 and A1 = A2.
Substituting these results into Equation 4.54,
A1 σ T14 − T24
q12 = (4.55)
(1/ε1 ) + (1/ε2 − 1)
b. For long concentric cylinders of radii r1 and r2, the view factor is F12 = 1 and
A1/A2 = r1/r2. The net heat exchange is obtained as:
A1 σ T14 − T24
q12 = (4.56)
1/ε1 + (r1 /r2 )(1 − ε2 )/ε2
c. For concentric spheres, F12 = 1 and A1/A2 = (r1/r2)2 so the next heat exchange
becomes:
A1 σ T14 − T24
q12 = (4.57)
1/ε1 + (r1 /r2 )2 (1 − ε2 )/ε2
d. For a small object (surface 1) in an enclosure (surface 2), A1 ≪ A2, F12 = 1, and
therefore:
q12 = A1 ε1 σ T14 − T24 (4.58)
A radiation shield is a protective layer constructed from a low emissivity (high reflectivity)
material to reduce the radiation exchange between the surfaces. The above results for a two-
surface enclosure can be extended to a radiation shield. The shield can be effectively
designed with different emissivities on both sides of the shield to reduce the radiation
exchange. If a radiation shield is placed between two objects, then the net radiation exchange
between the objects consists of the potential difference, Eb1 − Eb2, divided by the sum of sur-
face and spatial resistances. Using case (a) in Figure 4.12, the net radiative heat flow, q12,
between two parallel plates of equal area, separated by a radiation shield, is given by:
A1 σ T14 − T24
q12 = (4.59)
(1/ε1 ) + (1/ε2 ) + (1 − εs1 )/εs1 + (1 − εs2 )/εs2
where the subscripts 1, 2, s1, and s2 refer to plate 1, plate 2, shield (side 1), and shield (side 2),
respectively. It can be observed that smaller ϵs1 and ϵs2 values lead to an increased thermal
resistance in the denominator, thereby reducing the net heat transfer between the objects.
Thus the net radiative heat exchange can be minimized by controlling the shield emissivities.
In summary, the equations and procedure for the analysis of radiative heat exchange
between diffuse gray surfaces in an enclosure can be written as follows,
n
Ji = εi σTi4 + (1 − εi ) Fij Jj (4.60)
j=1
172 Advanced Heat Transfer
n
qi = Ai Fij (Ji − Jj ) (4.61)
j=1
where i = 1, 2, 3 … n. In Equation 4.60, the radiosity must balance the emission and reflection
of irradiation from the surface. In Equation 4.61, the net heat loss or gain balances the sum of
radiation exchanges between surface i and the other surfaces. This solution procedure allows
either the surface temperature or net heat exchange with each surface to be determined.
The solution procedures for three possible types of problems are summarized as follows:
1. All temperatures Ti are prescribed and all qi must be found. Using Equation 4.60, all
Ji can be found by a simultaneous solution of the linear algebraic equations. Then
Equation 4.61 is used to find the resulting qi values.
2. All qi are given and all Ti must be found. Using Equation 4.61, all Ji are computed
simultaneously. Then the temperatures, Ti, can be obtained from Equation 4.60.
3. Some qi and Ti are given. The remaining qi and Ti values must be found. The above
procedures for problem types 1 and 2 are then applied accordingly to a given surface.
Although the thermal network becomes increasingly complex with additional surfaces, it
is a useful approach for analyzing enclosures up to five surfaces or less. The resulting system
of linear algebraic equations may be solved directly by a numerical method or iteratively,
such as the Gauss–Seidel method.
where θ is the angle between the incident radiation and the tangent to the surface. The sur-
face would be nearly parallel to the sun’s incoming radiation at the north and south poles of
the Earth so cos θ ∼ 0 there, as expected, since the poles receive the least solar energy on
Earth. On the other hand, near the equator, cos θ ∼ 1, and the maximum solar radiation
passes through the top of the atmosphere.
The combined direct and diffuse radiation in Figure 4.13 is the total energy flux arriving
on the Earth’s surface, called the global irradiation. For the diffuse component, sky radiation
may be approximated as radiation emitted by a blackbody at a sky temperature between 230
and 285 K. Radiation emitted from the Earth’s surface, a temperature between 250 and
320 K, occurs predominantly in the long wavelength part of the spectrum, 4 ≤ λ ≤ 40 μm.
Using the Stefan–Boltzmann law, the Earth’s emitted radiation, E, is given by:
E = εσT 4 (4.62)
where typical values of the emissivity, ϵ, are 0.97 for water and 0.93–0.96 for soil. For the sky
radiation, Gsky,
Gsky = σTsky4
(4.63)
Scattered back to
space (1%–6%)
Absorbed by
atmosphere
(11%–23%)
Direct
(56%–83%)
Diffuse (5%–15%)
Surface of earth
Global
FIGURE 4.13
Components of solar radiation.
174 Advanced Heat Transfer
Since solar radiation has a significantly different spectral distribution than the Earth’s emis-
sions, the gray surface assumption cannot be used. The spectral distribution of solar irradia-
tion, Gλ,sky, is largely located below wavelengths of 4 μm and centered at about 0.5 μm,
whereas the Earth’s emissive power, Eλ,earth, is mainly located above 4 μm and centered at
about 10 μm. As a result, αs ≠ ϵ since the spectral distributions of solar irradiation and the
Earth’s energy emitted are positioned at different spectral regions. Therefore, instead of the
gray surface assumption, αs / ϵ ratios are usually tabulated and listed for different surfaces.
On the other hand, the gray surface assumption is reasonable for surfaces involving sky
irradiation, Gλ,sky, and the Earth’s emitted energy, Eλ,earth. Most sky irradiation is located
within the same wavelength range as the Earth’s emitted energy since the temperatures
in both cases are close in comparison to the previously described solar irradiation and
Earth’s emissive power. As a result, αsky = ϵ is a reasonable assumption when considering
the radiation exchange between the sky and the Earth’s surface.
where the subscripts s, a, and elec refer to surface, atmosphere, and electrical, respectively.
The individual terms represent (from left to right) solar irradiation, atmospheric irradia-
tion, electrical heat supplied, surface emission, and convective heat losses.
10-cm
square plate
Electrical
heater
Insulation
FIGURE 4.14
Schematic of a heated surface with solar irradiation.
Thermal Radiation 175
2 1
αλ Eb,λ dλ αλ Eb,λ dλ
αs = 01 + 2 1 = 0.8F(02μm) + 0.1(1 − F(02μm) ) (4.66)
0
Eb,λ dλ 0
Eb,λ dλ
Using the blackbody function table (Table 4.1) with λT = 2 × 5,800 = 11,600 μK, it is
observed that F(0λT) = 0.94. This yields an absorptivity of αs = 0.758. Thus, the surface
absorbs about 76% of the incident solar energy.
The absorptivity of atmospheric irradiation and the plate emissivity are calculated sim-
ilarly. For atmospheric irradiation,
1
αλ Gλ,atm dλ
αatm = 0 1 = 0.8F(02μm) + 0.1(1 − F(02μm) ) (4.67)
0
Gλ,atm dλ
From the blackbody table (Table 4.1), using λT = 2 × 345 = 690 μmK, it is found that
F(0λT) = 0.0 and ϵ = 0.1 = αatm. Thus, the plate is gray with respect to Gatm, but since
ϵ ≠ αs, it is non-gray with respect to Gsun.
For air at a film temperature of 320 K, ν = 17.8 × 10−6 m2/s, k = 0.028 W/mK, and Pr =
0.703. Then the Reynolds number based on the length of the plate, L, is
VL 5 × 0.1
ReL = = = 2.8 × 104 (4.69)
ν 19.9 × 10−6
The flow is laminar and the following correlation may be applied:
hL
NuL = = 0.664 Re1/2
L Pr
1/3
(4.70)
k
Substituting numerical values, the Nusselt number and convection coefficient become
99 and 27.7 W/m2 K, respectively.
Thus, the convective heat flux from the plate surface is estimated as:
Using this result and substituting the above values into Equation 4.64,
Solving this equation yields a required electrical heat input of approximately 821.9
W/m2.
From the previous example, it can be observed that a diffuse surface or a diffuse irradia-
tion assumption led to the results that ϵλ = αλ and ϵ = αatm.
176 Advanced Heat Transfer
∙ δ = declination angle (angle between the north pole and the axis normal to the sun’s
incoming rays);
∙ λ = latitude (degrees north or south of the equator);
∙ ω = hour angle (relative to the meridian of the plane of the sun’s incoming rays);
∙ θs = zenith angle of the sun (angle between the normal to the Earth’s surface and the
sun’s incoming rays at point P);
∙ α = solar altitude angle = π/2 − θs;
∙ θp = inclination (tilt) angle of the surface;
∙ θi = incident angle of the surface (between the normal to the inclined surface and the
sun’s incoming rays at point P on the surface);
∙ φ = azimuth angle.
The azimuth angle is defined with respect to a prescribed direction. For example, φs is the
azimuth angle formed between the direction south of point P and the sun’s incoming rays at
point P.
N
O = Center of earth
N = North pole
δ
O λ
Sun’s rays
Meridian of P
ω
N N
θs
θi
θp
P P
W E
φs
φp
S S
FIGURE 4.15
Solar angles.
Thermal Radiation 177
cos (θs ) = cos (λ) cos (δ) cos (ω) + sin (λ) sin (δ) (4.74)
360◦
ω = (tsol − 12 hr) (4.75)
24 hr
where the solar time, tsol, is based on the local civic time, tloc,civ, as follows:
Et
tsol = tloc,civ + (4.76)
60 min/hr
Et = 9.87 sin (2B) − 7.53 cos (B) − 1.5 sin (B) (4.77)
where,
n−1
B = 360◦ (4.78)
364
for day n of the year. The equation of time in Equation 4.77 gives a correction of the solar
time in units of minutes due to variations of the Earth’s orbit around the sun throughout
the year. It provides a correction of the local civic time in Equation 4.76.
The local civic time in Equation 4.76 differs from standard time by 4 min (1/15 h) for each
degree of difference in longitude from the reference meridian, such as the meridian dividing
Eastern Standard Time (EST) and Central Standard Time (CST). Some examples of standard
time zones are listed as follows:
cos (θi ) = sin (θs ) sin (θp ) cos (ϕs − ϕp ) + cos (θs ) cos (θp ) (4.80)
178 Advanced Heat Transfer
The declination angle varies throughout the year due to seasonal variations of the Earth’s
position in its orbit. For example, on June 21 in the summer solstice, the Earth’s declination is
23.5◦ . This corresponds to summer in the northern hemisphere. The autumnal equinox
begins on September 21. The declination angle becomes −23.5◦ (north pole facing away
from the sun’s incoming rays) on December 21 in the winter solstice (winter in the northern
hemisphere). On the vernal equinox (March 20/21) and autumnal equinox (September 22/
23), the sun appears directly overhead from an observer’s perspective at the equator.
During these various stages, transition of the declination angle occurs between a negative
angle (winter in the northern hemisphere) and a positive angle (summer in the northern
hemisphere). The following example illustrates how the declination angle is used to
calculate various solar angles.
Et = 9.87 sin (2B) − 7.53 cos (B) − 1.5 sin (B) = −8.2 min (4.81)
As a result, the local solar time becomes tsol = 10:01–0.08 = 9:53 a.m. which yields the
following hour angle:
where 15◦ /hr is based on 360◦ in 24 hours. The positive sign on the hour angle signifies
morning. Thus,
360(12 + 10)
sin (δ) = − sin (23.45) cos = −0.37 (4.83)
365.25
which yields a solar altitude angle of 13.45◦ at this time. Furthermore, the azimuth angle is
obtained by:
Thus, φs = 30◦ (east of south). Finally, the incidence angle is determined by:
cos (θi ) = cos (30) cos (13.45) sin (50) + sin (13.45) cos (50) = 0.795 (4.86)
It can be observed that the times of sunset and sunrise in the previous example could be
determined by setting α = 0 and solving the resulting hour angles. Substituting these hour
angles into Equation 4.75 would then give the corresponding times of sunset and sunrise.
The total daily extraterrestrial radiation is obtained by integrating cosθs from sunrise to sunset
and multiplying by the solar constant (I0).
Some incoming radiation from the sun is scattered within the atmosphere, both scat-
tered back to space and scattered eventually to reach the Earth’s surface (Idif,p). This
component is distinguished from the reflected portion (Iref,p), which refers to radiation
reflected off the ground and other surfaces on the Earth, rather than scattered within
the atmosphere. Various factors affect the diffuse (scattered) component of radiation,
Idif,p, including local weather conditions such as cloud cover, surface orientation, and
other factors.
The direct component of incident solar radiation can be expressed in terms of the incidence
angle, θi, and solar constant, I0, as follows:
where τatm refers to the transmissivity of the atmosphere. This transmissivity is dependent
on the path length of an incoming beam of radiation through the atmosphere.
For clear skies without pollution, the transmissivity can be approximated by:
patm (z)
m(z, α) = m(0, α) (4.90)
patm (0)
This coefficient, m, can be interpreted as the ratio of the actual distance traveled by the
incoming beam of radiation through the atmosphere to the distance traveled when the
180 Advanced Heat Transfer
sun is directly overhead of point P (solar noon). This ratio is equivalent to 1/sin(α),
where α is the solar altitude angle (as defined earlier). As α decreases toward sunset,
the transmissivity decreases since further scattering and absorption of radiation occurs
over the greater distance traveled by the incoming beam of radiation through the atmo-
sphere. For example, at sunrise or sunset, α = 0, which yields a minimum transmissivity
of τatm = 0.018.
The diffuse (scattered) component of solar radiation can be expressed in terms of the total
horizontal radiation, Ih, and extraterrestrial radiative flux on a horizontal surface, I0,h, as fol-
lows:
1 + cos (θp ) 1 + cos (θp )
Idif ,p = Idif ,h = (Ih − Idir,h ) (4.92)
2 2
where,
Ih
= 0.8302 − 0.03847m(z, α) − 0.04407(CC) + 0.011013(CC)2 − 0.01109(CC)3 (4.93)
I0,h
Here, CC refers to the cloud cover, where, 0 = clear sky and 1 = fully overcast.
For the diffuse component, the trigonometric factor in brackets in Equation 4.92 represents
a view factor between the sky and the inclined surface at a tilt angle of θp, which absorbs
radiation at point P on the Earth’s surface. For example, the view factor is F = 1 at θp = 0◦
(upward facing) and F = 0 at θp = 0◦ (downward facing). Also, the total incident radiation
on the surface when it is horizontally oriented, Ih, refers to the incident radiation after it
has passed through the atmosphere and arrives at the Earth’s surface, unlike the total solar
flux of I0,h which occurs prior to passage through the atmosphere. The extraterrestrial flux on
a horizontal surface, I0,h, is equivalent to I0 cos(θi,h), where θi,h refers to the incident angle for a
horizontal surface.
The remaining component in Equation 4.92 is the reflected component of solar radiation:
1
Iref ,p = ρg Ih (1 − cos (θp )) (4.94)
2
where ρg refers to the ground (or surrounding surface) reflectivity. The last expression, in
brackets, represents a view factor between the ground and the inclined surface at a tilt angle
of θp, which absorbs the incoming solar radiation.
360(38 + 10)
sin (δ) = − sin (23.45) cos = −0.27 (4.95)
365.25
Thermal Radiation 181
sin (α) = sin (49.8) sin (−15.7) + cos (49.8) cos (−15.7) cos (−40) (4.96)
cos (θi ) = cos (−48.6 − 0) cos (15.7) sin (60) + sin (15.7) cos (60) (4.98)
cos (θi ) = cos (−48.6 − 0) cos (15.7) sin (0) + sin (15.7) cos (0) = 0.27 (4.99)
m(0, α) = [(614 sin (15.7))2 + 1,229]1/2 − 614 sin (15.7) = 3.66 (4.100)
τatm ≈ 0.5[ exp (−0.095 × 3.65) + exp (−0.65 × 3.65)] = 0.4 (4.101)
Idir,p = Idir cos (θi ) = I0 τatm cos (θi ) = 1,353(0.4) cos (46.7) = 371.2 W/m2 (4.102)
For the second component (diffuse), under partial cloud cover (CC = 0.2),
Ih
= 0.8302 − 0.03847(3.65) − 0.04407(0.2) + 0.011013(0.2)2 − 0.01109(0.2)3
I0,h
= 0.68 (4.103)
where,
As a result,
Also,
Finally, the reflected portion of the incoming solar flux is given by:
1 − cos (60)
Iref ,p = 0.68 248.4 = 42.2 W/m2 (4.109)
2
Combining the direct, diffuse, and reflected portions of the incoming radiation, a total
hourly averaged incident solar flux of 490.2 W/m2 on the solar collector is obtained.
The accurate calculation of solar angles and the incident radiation heat flux is an important
element in the design of solar collectors. Further examples in solar energy applications were
reported by Howell, Bannerot, and Vliet (1982). The following section presents a thermal
analysis of solar collectors.
Ip Ac τs αs = qloss + qu (4.110)
Thermal Radiation 183
From left to right in this energy balance, the energy absorbed by the absorber plate bal-
ances the undesired heat losses to the surroundings due to reflected radiation and convec-
tive losses, qloss, and the energy gained by the working fluid, qu. The incident radiation at
point P, denoted by Ip, is multiplied by the collector surface area, Ap, transmissivity of the
glass cover τs, and absorptivity of the collector plate, αs. The incident radiation is first trans-
mitted through the glass covers and then absorbed by the collector plate. The heat losses,
qloss, and energy gained by the working fluid, qu, can be analyzed by a thermal network
involving the various thermal resistances.
The instantaneous collector efficiency, ηc, is the ratio of energy gained by the working fluid
to the total incoming solar energy,
qu
ηc = (4.111)
A c Ip
Integrating over a characteristic time period, T, such as an entire day, yields the following
average collector efficiency:
T
qu dt
ηc = T0 (4.112)
A I dt
0 c p
Consider a solar collector with two glass covers (top and lower covers), absorber plate,
collector tubes beneath the absorber plate, and insulation at the bottom side of the solar col-
lector (see Figure 4.16). The incident radiation arriving at the top glass cover, Ip, is either
reflected, absorbed by the top cover, or transmitted to the air gap beneath the top cover.
Similarly, the radiation that passes through the top cover, τsIp, is either reflected, absorbed
by the lower cover at Tg2, or transmitted through to an absorber plate at Tc. At the cover, the
incoming radiation of τsIp is either gained by the working fluid, qu, or lost back to
the surroundings.
A thermal circuit may be constructed to analyze the processes of heat transfer. Below the
collector plate, a series sum of resistances, includes R1 (between Tc and the base of the
Incident solar
radiation
Top glass cover, Tg1
Ambient air, Ta
Lower glass cover, Tg2
y
x
Insulation
Collector plate, Tc
FIGURE 4.16
Flat plate solar collector with two glass covers.
184 Advanced Heat Transfer
collector at Tb) and R2 (between Tb and the ambient air at Ta). In addition, there are
three resistances of combined convection and radiation: R3 (between Tc and the lower cover
at Tg2); R4 (between the top cover at Tg1 and Tg2); and R5 (between Tg1 and Ta). Then, the
overall thermal network at node Tc (collector plate) includes an incoming radiation flux, τsIp,
outgoing energy that is gained by the working fluid, and a total resistance, Rtot, between the
collector plate and the surrounding ambient air. This total resistance includes heat transfer
from both the top (convection to the air, and reflected and emitted radiation) and bottom
(convection to the air and working fluid, and conduction) sides of the collector.
Then the total conductance for the overall thermal network is given by:
1 1 1
Uc = = + (4.113)
Rtot R1 + R2 R3 + R4 + R5
In the first fraction on the right side, heat losses from the insulated bottom side of the col-
lector will be neglected. The heat losses to the surroundings are written as follows:
Based on the portion of the thermal network between the absorber plate and bottom glass
cover,
σAc Tc4 − Tg2
4
Tc − Tg2
qtop,loss = Ac hc2 (Tc − Tg2 ) + = (4.115)
(1/εp,i ) + (1/εg2,i ) − 1 R3
On the right side, the terms represent heat losses by natural convection and radiation
(based on view factors between two parallel plates facing each other), respectively.
Similarly, between the two glass covers,
σAc Tg2
4
− Tg1
4
Tg2 − Tg1
qtop,loss = Ac hc1 (Tg2 − Tg1 ) + = (4.116)
(1/εg2,i ) + (1/εg1,i ) − 1 R4
The first and second terms on the right side represent the heat losses by natural convection
and radiation, respectively. Alternatively, writing the heat loss from the top of the collector
based on the thermal network above the top glass cover,
T −T
qtop,loss = Ac hc,a (Tg1 − Tair ) + εg1,i σAc Tg1
g1 air
4
− Tsky
4
= (4.117)
R4
The three previous equations must be solved iteratively since values of R3, R4, and R5
depend on the unknown temperatures Tc, Tg1, and Tg2. The resistances include linearized
radiation coefficients which are dependent on temperature.
Alternatively, Klein’s method provides a direct empirical solution. Define N, θp, and ϵg, as
the number of glass covers, collector inclination with respect to the horizontal plane, and
emissivity of the glass cover and absorber plate, respectively. Also,
C = 365 1 − 0.00883 qp + 0.00013 q2p (4.118)
Thermal Radiation 185
f = 1 − 0.4 hc,a + 0.0005 h2c,a (1 + 0.091 N) (4.119)
Then the heat loss from the top cover of the solar collector is estimated by:
Ac (Tc − Ta ) σAc Tc4 − Ta4
qtop,loss =
−0.33 + 2N+f −1
(4.121)
εp +0.05N(1−εp ) + −N
NTc Tc −Ta 1
C N+f + 1/h c εg
Once the heat losses are determined, the energy gained by the fluid, qu, can be determined.
Heat is absorbed from incident solar radiation arriving directly above a tube, as well as heat
transfer along the absorber plate (effectively acting as a fin) between the tubes. Heat gained
by the absorber plate between the tubes is transferred by conduction through the plate to the
section of plate directly above the tubes carrying the working fluid. In the next section, qu
will be computed as the sum of heat transfer rates through the fin and solar energy absorbed
directly above each tube.
dTc
q′′x = −k (4.123)
dx
dy
D 2W D Collector plate
Side view
Tf,in Tube
FIGURE 4.17
Heat balance for a solar collector.
186 Advanced Heat Transfer
Expanding the heat flux at position x + dx about the value at x using a Taylor series
expansion,
dq′′ x
q′′x+dx = q′′x + dx + · · · (4.124)
dx
where higher order terms can be neglected for a differential thickness of the control volume,
dx. Substituting the heat fluxes into Equation 4.122 and rearranging,
d2 Tc Uc α s Is
= Tc − − T a (4.125)
dx2 kt Uc
where cosh(mW) = (emW + e −mW)/2 is the hyperbolic cosine function and m 2 = Uc/(kt). Dif-
ferentiating with respect to x, and evaluating at x = W, yields the following heat flux per
unit depth,
dTc α s Is
q′fin = −kt = −ktm T − T − tanh (mW) (4.127)
dx W
b a
Uc
where tanh(mW) refers to the hyperbolic tangent function. Alternatively, based on the def-
inition of the fin efficiency (Chapter 2),
where the factor 2 is introduced for heat transfer to a tube from two sides. The term in brack-
ets is the maximum heat transfer that would occur through the fin if the entire fin was held at
a temperature of Tb. With a fin entirely at a temperature of Tb, the resulting temperature dif-
ference (between the fin and air) would be maximized, thereby maximizing the convective
heat transfer between the absorber plate and surrounding air.
Comparing the above two equations of fin efficiency,
tanh (mW)
ηfin = (4.129)
mW
Thus the fin efficiency increases with a decreasing total conductance, larger plate conduc-
tivity, or thicker plate. Heat absorbed by the collector plate between the tubes is transferred
by conduction through the fin to the tube and the working fluid. This heat flow can be
increased through a reduced plate reflectivity (lower total conductance) of higher plate ther-
mal conductivity (lower m).
Combining the heat flows from the fin, Equation 4.128, and solar energy gained directly
above the tube of diameter, D, and unit depth,
Alternatively, based on an energy balance on the working fluid (inside each tube) in terms
of the convective heat transfer coefficient,
dTf
q′u = ṁcp = πDhc (Tb − Tf (y)) (4.131)
dy
where ṁ refers to the mass flow rate of working fluid through the tube.
The performance of solar collectors is often expressed in terms of a collector efficiency
factor and a collector heat removal factor. These factors are presented in the next
section.
Typical values of the total conductance, Uc, are 4 W/m2 K (two glass covers) and 8 W/m2
K (one glass cover). A working fluid of water with absorber plate materials of steel or copper
is a commonly used configuration. The collector efficiency factor, Fc, is given by:
1/Uc 1 1
Fc = + (4.133)
(D + 2W) Uc (D + 2Wηfin ) hc πDi
On the right side, the terms within square brackets refer to the fin resistance and convec-
tion resistance to fluid flow in the tubes, respectively. This collector efficiency factor repre-
sents a ratio of the thermal resistance between the collector and the environment to the
thermal resistance between the working fluid and the environment.
The fluid temperature in Equation 4.132 varies from Tf,in at the tube inlet to Tf,out at the tube
outlet. In order to find the total heat absorbed, the full variation of Tf with position y is
required. Then the heat absorbed by the working fluid between the inlet and outlet can
be determined. Applying an overall energy balance for the working fluid from the inlet to
the outlet of a single tube,
The temperature difference of the fluid between the inlet and outlet is obtained by first
deriving the variation of fluid temperature with position within the tube.
Consider a control volume in Figure 4.17 of differential thickness dy in the y-direction
(along the tube axis). Performing an energy balance for this control volume,
dTf
ṁcp = q′u = (D + 2W)Fc [αs Is − Uc (Tf − Ta )] (4.135)
dy
188 Advanced Heat Transfer
ṁ
G= (4.137)
L(D + 2W)
The fluid temperature difference can be isolated by adding and subtracting Tf,in in the
numerator. Then, substituting the temperature difference into Equation 4.134,
The heat removal factor characterizes the ability of the working fluid to remove heat
through the fin and collector.
Using Equations 4.111 through 4.138, an alternative expression for the solar collector effi-
ciency becomes:
qu Tf ,in − Ta
ηc = = Fr αs τs − Fr Uc (4.140)
Ac Ip Ip
Ta = 15 °C
Collector plate, 2.2 m × 3.1 m
Ip = 500 W/m2
Collector plate
Tube
10 cm
Tf,in = 47 °C Tube
1 cm
h = 1,100 W/m2K
FIGURE 4.18
Schematic of a solar collector and tube flow.
required water flow rate to achieve a heat removal factor of 0.74. Also, find the required
absorptivity of the collector plate for a collector efficiency of 0.13.
The fin efficiency of the collector plate is calculated as follows:
Uc 8
m= = = 9.19 (4.141)
kt 237(0.0004)
Then the collector efficiency factor and heat removal factors are given by:
−1
1 1 1
Fc = + = 0.934 (4.143)
8 × 0.1 8(0.01 + 0.09 × 0.95) 1,100π(0.01)
ṁ × 4,200 8 × 0.934
Fr = 0.74 = 1 − exp (4.144)
2.2(3.1)8 4,200ṁ/(2.2 × 31)
which can be solved iteratively to give a mass flow rate of 0.025 kg/s.
Also, from the collector efficiency relation,
qu Tf ,in − Ta
ηc = = Fr αs τs − Uc (4.145)
Ac Ip Ip
320 − 288
0.13 = 0.74 αs (0.92) − 8 (4.146)
400
The previous sections have described selected elements of solar energy systems. A more
complete treatment and analysis of solar energy engineering including other applications is
provided in other excellent sources such as Duffie and Beckman (1974), Hsieh (1986), Garg
(1982), and Howell and Bereny (1979).
190 Advanced Heat Transfer
PROBLEMS
emissivity is 0.94 and its surface temperature is 1,100 K. Find the absorber surface
temperature and the radiative heat transfer rate from the heater.
4.16 A radiation shield is placed between two large plates of equal area. Find the steady-
state temperature of the shield in terms of the temperatures and emissivities of
both plates.
4.17 The cross-sectional area of a solar collector consists of a right-angled triangle with two
equal-sided absorber plates and a glass cover along the longest side. A heat flux of
300 W/m2 passes through the glass cover to yield a cover temperature of 30◦ C. The
surface emissivity is 0.8. The absorber plates are approximated as blackbody surfaces
that are insulated on both back sides. The temperature of the upper absorber plate is
70◦ C. What is the surface temperature of the other plate resulting from radiation
exchange between all surfaces in the enclosure? It may be assumed that the cover
plate is opaque with respect to radiation exchange with the absorber plates.
4.18 Solar collectors are used to deliver heated water in a building. Consider a single solar
collector that is facing S 20◦ E during a clear day in Toronto, Canada, on April 20 (79◦
30′ W, 43◦ 40′ N). What tilt angle (with respect to the horizontal plane) is required
to provide a direct solar radiation flux of 800 W/m2 on the inclined collector at
11:00 a.m., EST (note: CST at 90◦ , EST at 75◦ )?
4.19 A solar collector is located in Winnipeg, Canada (latitude 49◦ 50′ N, longitude
97◦ 15′ W). It is tilted 45◦ up from the horizontal plane and it is facing S 15◦ E at
9:30 a.m., CST, on August 17.
(a) Find the solar altitude and azimuth angles.
(b) Estimate the total incident solar radiation on a clear day.
4.20 A solar collector with a single glass cover (surface area of 3 m2) and insulated back
and bottom sides is tilted at 50◦ with respect to the horizontal plane. The spacing
between the plate (ϵp = 0.25) and glass cover (ϵg = 0.85) is 4 cm. The average plate
temperature is 70◦ C and the air temperature and wind speed are 15◦ C and 5 m/s,
respectively. Using the method of Klein and the exact (iterative) method, find the
heat loss from the solar collector.
4.21 The copper tubes of a single-glazed solar collector are connected by a 2 mm thick cop-
per plate with a conductivity of 400 W/m2 K. Water flows through the tubes (0.5 mm
thick wall) with a heat transfer coefficient of 440 W/m2 K. The total conductance is
Uc = 10 W/m2 K. The tube-to-tube distance between inner edges of adjacent tubes
is 15 cm. What tube diameter is required to give a collector efficiency factor of 0.87?
4.22 A solar collector is designed with the following characteristics: one glass cover; tilt
angle of 60◦ C; overall conductance of Uc = 6 W/m2; Ta = 18◦ C; and Ip = 760 W/m.
The cover transmissivity is τ = 0.8 and the plate absorptivity is α = 0.9. The copper
plate thickness is 0.06 cm. Each tube has an inner diameter and wall thickness of
1.5 and 0.06 cm, respectively. The tube-to-tube centered distance is 10 cm. What is
the plate temperature above each tube, Tb when the collector efficiency reaches 0.35?
4.23 Performance testing of a single-glazed flat plate solar collector yields the following
result for collector efficiency, ηc, in terms of incident radiative flux, Ip, ambient tem-
perature, Ta, and incoming fluid temperature (within the collector), Tf,in:
Tf ,in − Ta
ηc = 0.82 − 7.0
Ip
Thermal Radiation 193
The transmissivity of the glass cover is 0.92 and the collector heat removal factor is 0.94.
a. Find the collector surface absorptivity, αs, and heat loss conductance, Uc, for this
solar collector.
b. The net energy absorbed by a solar collector over a surface area of 3 m2 is qu = 1,600 W.
◦
If the incident radiative flux and ambient air temperature are 800 W/m2 C and 4◦ C,
respectively, calculate the incoming fluid temperature, Tf,in, into the solar collector.
Find the collector efficiency factor when the mass flow rate of water is 8 kg/min through
the collector tubes.
References
J.A. Duffie and W.A. Beckman. 1974. Solar Energy Thermal Processes, New York: John Wiley & Sons.
H.P. Garg. 1982. Treatise on Solar Energy, New York: John Wiley & Sons.
J.R. Howell. 1982. Catalog of Radiation Configuration Factors, New York: McGraw-Hill.
J.R. Howell, R.B. Bannerot, and G.C. Vliet. 1982. Solar-Thermal Energy Systems: Analysis and Design,
New York: McGraw-Hill.
Y. Howell and J.A. Bereny. 1979. Engineer’s Guide to Solar Energy, San Mateo: Solar Energy Information
Services.
J.R. Howell, M.P. Menguc, and R. Siegel. 2015. Radiation Heat Transfer, 6th Edition, Boca Raton: CRC
Press/Taylor & Francis.
J.S. Hsieh. 1986. Solar Energy Engineering, Upper Saddle River, NJ: Prentice Hall.
F. Kreith, R.M. Manglik, and M.S. Bohn. 2010. Principles of Heat Transfer, 7th Edition, Stamford: Cen-
gage Learning.
J.H. Lienhard IV and J.H. Lienhard V. 2000. A Heat Transfer Textbook, 3rd Edition, Cambridge: J.H.
Lienhard IV and J.H. Lienhard V.
M. Planck. 1959. The Theory of Heat Radiation, New York: Dover Publications.
C.D. Wen and C.T. Lu. 2010. “Suitability of Multispectral Radiation Thermometry Emissivity Models
for Predicting Steel Surface Temperature”, AIAA Journal of Thermophysics and Heat Transfer, 24:
662–665.
5
Gas–Liquid Two-Phase Flows
5.1 Introduction
Gas–liquid two-phase flows are characterized by the presence of a moving and deforming
gas–liquid phase interface. The heat released or absorbed during a gas–liquid phase change,
per unit mass, is represented by the latent heat of vaporization. This chapter examines the
physical processes of boiling, condensation, and two-phase flows. Phase change in gas–liq-
uid flows occurs in many scientific and industrial systems such as boilers in thermal power
plants, cooling systems in nuclear reactors, heat pumps, refrigeration systems, quenching in
materials processing, among many others.
Boiling occurs by rapid vaporization of a liquid in the presence of a heated surface. There
are several types or modes of boiling processes. Pool boiling occurs when the heating surface
is submerged in a large body of stagnant fluid. Buoyancy forces cause a relative motion of
the vapor produced in the surrounding liquid near the heating surface. In nucleate boiling,
small bubbles form along a surface, while critical heat flux boiling occurs when a surface is
heated above a critical temperature after which a film of vapor forms on the surface. Tran-
sition boiling is an intermediate, unstable form of boiling with characteristics of both nucleate
and critical heat flux boiling.
In nucleate boiling, vapor bubbles initially form within cavities along the heated surface.
The bubbles grow when liquid is vaporized around the bubble. Heat is also transferred from
the wall around the cavity to liquid beneath the bubble (called a liquid microlayer) and even-
tually the bubble detaches from the surface cavity. Then additional liquid comes into contact
with the heating surface to sustain the boiling process. The local heat transfer rate is
enhanced when liquid comes into contact with the surface due to a higher fluid thermal con-
ductivity. After a bubble departs from the heated surface, a microconvection process occurs
in which liquid is entrained into the cavity where the vapor previously emerged. As further
heating occurs, the number of nucleation sites increases and isolated bubbles can merge into
continuous columns or channels of vapor.
Surface geometry and orientation are important factors in the boiling process. More
frequent bubble sweeps along the surface lead to more vigorous heat transfer in a down-
ward-facing orientation. Inclined surfaces can promote more active microlayer agitation
and mixing along the surface in comparison to upward facing horizontal surfaces. Bubbles
forming in a cavity begin to slide along and away from the cavity’s upper edge during the
boiling process. After a bubble expands and departs from the cavity, the lower edge of the
trailing liquid–vapor interface moves into the cavity. Other ascending bubbles approach
the cavity and merge with the remaining gas phase in the cavity. A new bubble is formed
in this merging process. The wake of passing bubbles mixes with the microlayer forming
195
196 Advanced Heat Transfer
over the cavity. This sequence of stages of bubble growth, detachment, and formation again
is repeated throughout the boiling process.
In contrast to boiling heat transfer, condensation occurs when vapor is brought into con-
tact with a cooled surface below the saturation temperature. When a quiescent vapor comes
into contact with a cooled surface, two types of condensation may occur—dropwise conden-
sation or film condensation. The liquid does not fully cover or wet the surface in dropwise con-
densation, whereas the liquid film covers the entire surface in film condensation. A lower
rate of heat transfer between the vapor and wall usually occurs in film condensation, due
to the additional thermal resistance of the liquid film.
The heat transfer coefficients of boiling and condensation depend on thermodynamic and
flow conditions, thermophysical properties, and the geometrical configuration, among
other factors. For condensation of steam on surfaces of 3–20◦ C below the saturation point,
the average heat transfer coefficient typically ranges between 11,000 and 23,000 W/m2 K
for horizontal tubes and between 5,700 and 11,000 W/m2 K for vertical surfaces. For conden-
sation of ethanol along a horizontal tube, the heat transfer coefficient typically varies
between 1,700 and 2,600 W/m2 K.
In this chapter, boiling, condensation, and related two-phase flows will be examined,
including physical processes and advanced solution methods for nucleate boiling, forced
convection boiling, two-phase flows in vertical and horizontal tubes, laminar and turbulent
film condensation, forced convection condensation, and heat pipes and thermosyphons.
Free
surface
Liquid q
q Vapour
Inflow
bubbles
Outflow
Heater block Tube containing hot fluid
FIGURE 5.1
Electrical and thermal heating.
Gas–Liquid Two-Phase Flows 197
Alternatively, in thermal heating, the surface temperature can be set independently of the
heat flux.
Saturated pool boiling arises when the temperature of the liquid pool is maintained at or
close to the saturation temperature, Tsat. Bubbles formed at the heated surface are propelled
upward through the liquid by buoyancy. In subcooled pool boiling, the temperature of the
liquid pool is lower than the saturation temperature and bubbles formed near the heater
surface may later condense in the liquid.
In pool boiling problems, the fluid is initially quiescent near the heating surface. Sub-
sequent fluid motion arises from free convection and circulation induced by bubble
growth and detachment. A key parameter in boiling analysis is the degree of wall
superheating, ΔT = Tw − Tsat, or the difference between the wall and bulk liquid satura-
tion temperature at the local pressure. As ΔT increases, the process of boiling proceeds
more rapidly.
The boiling curve illustrates the various stages of boiling from the early stages of free con-
vection in the liquid prior to phase change, to nucleate, transition, and finally film boiling
modes along a surface (see Figure 5.2). Up to point A in Figure 5.2, free convection occurs
in the liquid as there is insufficient vapor to cause active boiling. Small temperature dif-
ferences exist in the liquid and heat is removed by free convection to the free surface.
At point A, isolated bubbles initially appear along the heating surface. This point is the
onset of nucleate boiling (ONB). Nucleate boiling occurs between points A and
C. Isolated bubbles appear and heat is transferred mainly from the surface to the liquid.
As ΔT increases (B–C), more nucleation sites become active and bubbles coalesce, mix, and
ascend as merged jets or columns of vapor. At point C, the maximum heat flux, or critical
heat flux (CHF), occurs.
Transition boiling occurs between points C and D. An unstable (partial) vapor film forms
on the heating surface and conditions oscillate between nucleate and film boiling. Intermit-
tent vapor formation blocks the liquid of higher thermal conductivity from contacting the
surface, thereby lowering the surface heat flux. Film boiling occurs beyond point D. In addi-
tion to conduction and convection, heat transfer by radiation is important at high wall
superheating levels. A stable vapor film covers the surface in this region.
Columns Stable
107 and slugs film
Free Nucleate boiling
CHF
convection boiling
106 C Transition
boiling
qs (W/m2)
FIGURE 5.2
Boiling curve.
198 Advanced Heat Transfer
where q′′w , hfg, L, σ, C, a, b, c, and d refer to the wall heat flux, latent heat of vaporization, sur-
face length, surface tension, and coefficients to be determined from the Buckingham Pi anal-
ysis, respectively. The subscripts l and v refer to the liquid and vapor phases, respectively.
Using experimental data for pool boiling, Rohsenow (1952) obtained the following form of
correlation for the wall heat flux:
g(ρl − ρv ) 1/2 cp,l ΔT 3
q′′w = μl hfg (5.2)
σ cs,f hfg Prn
where cp,l is the liquid specific heat and cs,f and n are empirical coefficients which depend on
the fluid–surface combination. Figure 5.3 shows good agreement between this correlation
and measured data of pool boiling of water over a range of operating pressures.
The empirical coefficients in Rohsenow’s correlation were determined based on ex-
perimental data over a range of operating conditions. Typical values of the coefficients
for various liquid–surface combinations are given by: cs,f = 0.013, n = 1.0 (water–copper);
cs,f = 0.006, n = 1.0 (water–brass); cs,f = 0.0132, n = 1.0 (water–mechanically polished stain-
less steel); cs,f = 0.101, n = 1.7 (benzene–chromium); cs,f = 0.027, n = 1.7 (ethyl alcohol–
chromium); cs,f = 0.00225, n = 1.7 (isopropyl alcohol–copper); and cs,f = 0.015, n = 1.7
(n-pentane–chromium).
Gas–Liquid Two-Phase Flows 199
100
Rohsenow (1952)
0.1
0.005 0.05 0.5
cp ΔT/(Prl hfg)
FIGURE 5.3
Rehsenow’s correlation of pool boiling heat transfer for water. (Adapted from W.M. Rohsenow. 1952. Transactions of
ASME, 74: 969–976.)
The conditions of the surface, including its roughness, surface coating, oxidation, and
fouling, affect the bubble formation and dynamics. These effects and others have been inves-
tigated in past studies, such as Vachon, Nix, and Tanger (1968), including modified empir-
ical coefficients and exponents which account for various surface modifications.
The addition of other constituents into the working fluid also significantly affect the
heat transfer characteristics of the liquid–surface combination. For example, at atmo-
spheric pressure, the ratio of the average heat transfer coefficient for diluted water to
pure water, h/hw, can vary widely with constituent concentration: h/hw = 0.61 (24%
NaCl, 76% water), h/hw = 0.87 (20% sugar, 80% water), and h/hw = 0.53 (100% methanol,
no water).
The maximum heat flux (or critical heat flux; CHF) occurs at the transition between nucle-
ate boiling and film boiling. Kutateladze (1948) presented a dimensional analysis of the
dependence of the CHF on various operating parameters. Accurate prediction of the CHF
is critical in various industrial systems because operations beyond the CHF may pose major
safety risks. If the heat flux exceeds the CHF, then a large sudden increase in system temper-
ature can lead to system damage, such as melting of components in a nuclear reactor, or
overheating of a chemical reactor. Even operating near the CHF has safety risks due to boil-
ing in the transition region which may cause unstable and abrupt changes in the wall
heat flux.
By applying a dimensional analysis with the Buckingham Pi theorem at the maximum
heat flux, it can be shown that five variables and four primary dimensions yield 5 – 4 = 1
pi group. As a result, the critical heat flux at the wall can be expressed as:
σg(ρl − ρv ) 1/4
q′′max = 0.149hfg ρv (5.3)
ρ2v
200 Advanced Heat Transfer
where the coefficient 0.149 is based on comparisons with experimental data of Lienhard,
Dhir, and Riherd (1973). This result holds for pure liquids (saturated) on horizontal, up-
facing surfaces. The maximum heat flux varies with pressure through the dependence of
thermophysical properties on the saturation pressure.
Subcooled pool boiling occurs when the liquid pool temperature is less than the
saturation temperature. The maximum heat flux for subcooled conditions, q′′max,sub, can be
expressed in terms of the maximum heat flux for a saturated liquid, q′′max,sat, as follows:
where
Here Tpool refers to the bulk fluid temperature of the liquid pool.
The Leidenfrost point is the point of the minimum heat flux (see Figure 5.2). There are sim-
ilarities in approaching the region of transition boiling from the side of cooling to the min-
imum heat flux from film boiling by reducing the wall superheat, or heating to the CHF from
nucleate boiling by increasing the temperature difference. When reducing the wall super-
heat below the minimum heat flux, the rate of vapor generation is not high enough to sustain
a stable vapor film. Then Equation 5.3 can be modified as follows (Zuber 1958):
1/4
σg(ρl /ρv )
q′′min = Chfg ρv (5.6)
(ρl + ρv )2
where C ≈ 0.09 (Berenson 1961) was obtained from experimental data for horizontal sur-
faces. The result is accurate within approximately +50% at moderate pressures for
most fluids.
where C ≈ 0.62 (horizontal cylinders) or 0.67 (spheres) and the corrected latent heat is
given by:
The subscripts v, l, w, and sat refer to vapor, liquid, wall, and saturation, respectively. The
vapor properties are evaluated at the film temperature, (Tw + Tsat)/2, and the liquid
properties are evaluated at Tsat. Berenson (1961) and Duignam, Greene, and Irvine (1991)
applied this correlation to film boiling on horizontal surfaces, based on a modification of
the diameter, D.
In film boiling, radiative heat transfer often becomes significant. In cases with a large-wall
superheating level, the effective heat transfer coefficient, including both convection and
radiation, can be expressed as:
h4/3 = h4/3
conv + hrad h
1/3
(5.10)
For the case of hrad , hconv, the heat transfer coefficients can be approximated by:
where ɛs refers to the surface emissivity and σ is the Stefan-Boltzmann constant. This
modeling approach for radiative heat transfer in film boiling assumes that the wall
is parallel to the phase interface and radiation exchange approximates the wall as
a blackbody.
∂T ∂2 T
=α 2 (5.13)
∂t ∂y
where α is the thermal diffusivity. Initial and boundary conditions are T( y, 0) = Tsat and
T(0, t) = Tw. Using a similarity solution, it can be shown that the following temperature
and wall heat flux profiles are obtained during the liquid period:
T − Tw y
= erf √
(5.14)
Tsat − Tw 2 αt
202 Advanced Heat Transfer
k(Tw − Tsat )
q′′w = √
(5.15)
παt
where erf(w) is the Gaussian error function. This result is used to specify an initial temperature
profile in the next vapor period.
During the period of bubble formation and growth, a vapor bubble covers the heated sur-
face. Define the function δ(t) as the diameter of the growing bubble. Transient conduction in
the vapor phase is simplified by the following heat conduction equation:
∂T ∂2 T
=α 2 (5.16)
∂t ∂y
A conformal mapping (using a bilinear function; Chapter 2) can be used to transform the
two-dimensional region of a growing bubble to a one-dimensional region with a planar
phase interface at a position of δ(t). Using this change of variables, heat transfer through
the microlayer beneath the bubble is approximated by one-dimensional heat conduction
across a planar layer. The thickness of the microlayer decreases as liquid in the microlayer
is vaporized.
The initial and boundary conditions in the vapor period are approximated by:
y
T(y, tl ) = Tw − λ(Tw − Tsat ) ; T(0, t) = Tw ; T(δ, t) = Tsat (5.17)
δ
where
δ
λ = √
(5.18)
παtl
− erf
+λ (5.19)
Tsat − Tw n=0 2 α(t − tl ) 2 α(t − tl ) δ
The heat balance at the phase interface involves heat conduction to/from the interface and
a latent heat of vaporization absorbed by the liquid,
∂T dδ ∂T
− kl −ρhfg = −kv (5.20)
∂y δ− dt ∂y δ+
where the last term on the right side in the vapor phase is assumed to be negligible.
Using the temperature distribution from Equation 5.19 at late stages of bubble growth
(λ ∼ 1), and neglecting higher order terms, the interface position becomes:
πρhfg α kΔT t
δ(t) = παtl − 1 exp −2 − 1 − 1 + πα(t − tl ) (5.21)
2kΔT ρLπα tl
Gas–Liquid Two-Phase Flows 203
Also, using Fourier’s law to determine the wall heat flux from Equation 5.19,
√
kΔT παtl − δ −δ2 kΔT
q′′w =2
exp + √
(5.22)
πα tl (t − tl ) 4α(t − tl ) παtl
The average heat flux during a cycle throughout a liquid–vapor period is obtained by inte-
grating the heat fluxes over their respective liquid and vapor periods. Naterer et al. (1998)
presented comparisons of this predictive model with experimental data and found reason-
able agreement over a range of boiling conditions.
NuD = AF Re0.67
D Pr
0.4
(5.23)
′′
The boiling Reynolds number is expressed in terms of the heat flux, q , and latent heat of
vaporization, hfg, as follows:
q′′ D
ReD = (5.24)
μf hfg
204 Advanced Heat Transfer
The coefficients A and F are functions of the critical pressure ratio, pr = p/pc, as follows:
F = 1.8p0.17
r + 4p1.2
r + 10pr
10
(5.25)
A = 9.7p0.5
c (5.26)
where pc is the critical pressure of the fluid in bars (10−5 N/m2). The correlation was devel-
oped for a wide range of common fluids (water, organics, and refrigerants) but excludes
liquid metals and cryogenic fluids.
Typical trends data of the heat transfer coefficient for boiling of R-113 at a pressure of 1
atm at various approach velocities, U (m/s), past a horizontal tube are shown in Figure 5.4.
The variables h and U refer to the heat transfer coefficient and liquid velocity, respectively,
where U = 0 corresponds to pool boiling over a horizontal tube. The experimental results in
Figure 5.4 were obtained for a 27 mm diameter tube with a tube wall heat flux of q′′ = 25
kW/m2. As expected, the value of the heat transfer coefficient increases from the base due
to the increasing vapor velocity.
3.4
3.2
3.0
h (kW/m2k)
2.8
2.6
U = 0.0 m/s
2.4 U = 0.1 m/s
U = 0.2 m/s
2.2 U = 0.3 m/s
U = 0.37 m/s
2.0
0 45 90 135 180 225 270 315 300
Angle from top (degrees)
FIGURE 5.4
Heat transfer coefficient for boiling with forced convection at q′′ = 25 kW/m2. (Adapted from K. Cornwell and S.D.
Houston. 1994. International Journal of Heat and Mass Transfer, 37: 303–309.)
Gas–Liquid Two-Phase Flows 205
where hc, hb, Tsat, Tw, and Tm refer to the heat transfer coefficients for single phase convection
and nucleate boiling, respectively, and the saturation, wall, and mean bulk fluid tempera-
tures, respectively.
The combined effects of both forced convection and nucleate boiling on the overall heat
transfer coefficient can be expressed through the following model of Kutateladze (1963):
1/2
h2
h = hc 1 + b2 (5.28)
hc
It can be observed that the overall heat transfer coefficient approaches the single phase
convection limit as the boiling coefficient diminishes to zero.
FIGURE 5.5
Two-phase flow patterns in vertical flow.
206 Advanced Heat Transfer
a churn flow occurs when the slug flow bubbles break down and give an oscillating
motion of liquid upward and downward in the tube. Then an annular flow occurs
when the liquid flows as a film along the walls with some internal liquid entrainment
in the core, while the vapor flows through the center of the tube. Entrainment may
occur with some droplets in the inner gas core and potentially some bubbles entrained
in the liquid film. At high mass flow rates, a wispy/annular flow occurs where the con-
centration of droplets in the core increases sufficiently to form coalescence of droplets
and large lumps or streaks (or wisps) of liquid.
Transition between various flow regimes often involves flow instabilities. In vertical
flow, the bubble to plug flow transition occurs during the process of bubble coalescence
when bubbles grow, increase in frequency and eventually join together to occupy a large
section of the pipe cross section. Typically, this transition to plug flow occurs at a vapor
fraction of 25%–30%. The effects of turbulence may disrupt this process of bubble coales-
cence. As the fluid velocity increases, chaotic eddy motion in the turbulent flow acts to
break up the bubbles and create “void waves” in the flow. The transport of these waves
in the flow direction can lead to the packing of bubbles and coalescence as a result of the
induced packing.
In the transition between plug and annular flow, another flow regime may occur,
called churn flow. A developing plug or slug flow leads to a sharp increase of the pres-
sure gradient at the onset of churn flow. Flooding-type waves are formed which are
typically absent in both slug flow and annular flow, but which characterize the liquid
motion in churn flow. Between successive flooding waves, liquid flow in the film
region along the wall reverses in direction and becomes entrained by the next upward
moving wave.
Another transition point is the transition from churn to annular flow. The pressure gradi-
ent decreases and reaches a minimum as the gas velocity increases and a more stable inner
core of gas flow is formed. The intermittent flooding waves disappear as well as their pul-
sating gas–liquid fluctuations which lead to the large pressure gradients. Eventually as the
gas flow rate continues to increase, the pressure gradient again rises. Beyond the point of a
pressure gradient minimum, there is no longer a flow reversal within the liquid film.
Another final transition occurs from annular flow to wispy annular flow when droplets
in the inner core are broken into distinct droplets, destabilized, and stretched into wispy
droplet streams.
surrounding liquid which must by moved to allow bubble growth; and second the need for
thermal diffusion (heat conduction) from the surrounding liquid to the interface to cause
vaporization.
For inertia controlled growth of a bubble, conservation of mass requires that the outward
flow rate of liquid balances the velocity of the bubble, u(r), at the boundary multiplied by its
surface area. The outward flow rate of liquid at every radius, r, is the same, so therefore:
drb
4πr2b = constant = 4πr2 u (5.29)
dt
The kinetic energy of the moving liquid, supplied by the expanding bubble, can be
written as:
1 2
u2 drb
4πr ρl 2
dr = 2πrb ρl
3
(5.30)
rb 2 dt
The corresponding vapor pressure of the bubble as it expands against the external pres-
sure is given by:
4
p dV = (pg − p1 ) πr3b (5.31)
V 3
2
drb 2 pg − p1
= = constant = A2 (5.32)
dt 3 ρl
rb = At (5.33)
The constant A can be determined by replacing the pressures with temperature using the
Clausius–Clapeyron equation,
2 ρg T1 − Tsat
A = hfg
2
(5.34)
3 ρl Tsat
For thermal diffusion controlled growth of the bubble, the latent heat required to expand
the bubble of radius rb must be transferred by conduction over a distance proportional to rb.
The time taken to diffuse over this distance can be determined, from which the following
Plesset–Zwick solution (Plesset and Zwick 1954) is obtained for the growth of the bubble
radius with time,
rb = Bt1/2 (5.35)
208 Advanced Heat Transfer
Alternatively, the following model can be used for either the inertia controlled or thermal
diffusion controlled regions,
2 +
r+
b = [(t + 1)
3/2
− (t+ )3/2 − 1] (5.37)
3
where
Arb A2
r+
b = ; t+ = t (5.38)
B2 B2
This solution is an inertia controlled model for t + ≪ 1 and a thermal diffusion controlled
model for t + ≫ 1.
During the growth of bubbles in the bulk liquid flow, another simultaneous process is
occurring by the coalescence of adjacent and nearby bubbles. Consider the formation and
thinning of a liquid film between two bubbles of radii rb1 and rb2, and velocities ug1 and
ug2, respectively. Chesters and Hofman (1982) showed that the point of coalescence of
two bubbles occurs at a Weber number of:
where
1 1 1
= + (5.40)
req 2rb1 2rb2
The velocity difference (ug1 − ug2) represents the relative velocity between the two
deformed bubbles.
As the heat flux increases, the number of active nucleation sites increases and bubbles start
to coalesce close to the wall. This flow regime is called fully developed boiling. The wall
becomes covered by a thin liquid-rich layer through which thin stems of vapor are con-
nected to a cloud of coalescing bubbles. Heat transfer occurs primarily by conduction across
the unsteady layer, causing evaporation at the base of the bubble cloud and the vapor stems
feeding the bubbles. The process of fully developed boiling can be either thermally con-
trolled through the wall heat flux, or hydrodynamically controlled, through the mass
flow rate of the bulk liquid flow.
The following Saha–Zuber correlation (Saha and Zuber 1974) can be used to determine the
heat transfer coefficient at the onset of fully developed boiling (called the OFDP point). In the
thermally controlled region,
hDh q′′w Dh
NuD = = = 455 (5.41)
kl kl (Tsat − Tf )
Gas–Liquid Two-Phase Flows 209
where
GDh cp,l
Pe = ≤ 70,000 (5.42)
kl
Here Pe, G, Dh, cp,l and kl refer to the Peclet number, phase-averaged velocity (mass flow rate
divided by cross-sectional area), hydraulic diameter, liquid specific heat, and liquid thermal
conductivity, respectively.
In the hydrodynamically controlled region, for Pe . 70,000,
NuD
St = = 0.0065 (5.43)
Pe
These correlations can also be used to predict the rate of vapor generation in bubble flows
subsequent to the onset of fully developed boiling.
∂ ∂ 2 dp τ w P τ i Pi
[ρl ul (1 − ζ g )] + ρl ul (1 − ζg ) = −(1 − ζ g ) − gρl (1 − ζg ) sin θ − + (5.44)
∂t ∂z dz A A
∂ ∂ dp τ i Pi
(ρ ζ ug ) + (ρg ζ g u2g ) = −ζ g − gρg ζg sinθ − (5.45)
∂t g g ∂z dz A
where the subscripts l, g, w, and i refer to liquid, gas, wall, and interface, respectively.
Also, A, τw, τi, P, Pi, θ, and ζg refer to the tube cross-sectional area, wall shear stress, inter-
facial shear stress (rate of momentum transfer from the gas to liquid phase per unit inter-
facial area), tube perimeter, interfacial perimeter, and the fraction of the tube volume that
is occupied by the gas phase.
210 Advanced Heat Transfer
For steady-state flow in a vertical tube, the gas phase equation reduces to:
∂ dp τ i Pi
ρg ζ g u2g = −ζg − gρg ζ g − (5.46)
∂z dz A
Assume that the inner gas core of the annular flow is a homogeneous mixture of gas and
droplets occupying a fraction ζc of the channel cross-sectional area. Then the fraction ζc can
be related to the film thickness, δ, and tube radius, ro, as follows:
(ro − δ)2
ζc = (5.47)
r2o
ṁle + ṁg
ρc = (5.48)
ṁle /ρl + ṁg /ρg
where ṁle and ṁg refer to the mass flow rate of entrained liquid and the gas phase,
respectively.
Also, the gas core fraction can be written in terms of the fraction of the total cross section
occupied by droplets, ζd, where,
ṁg /ρg + ṁle /ρl
ζc = ζg + ζd = ζg (5.49)
ṁg /ρg
Then the momentum balance in the gas phase, Equation 5.46 can be rewritten as:
∂ dp τ i Pi
ρc ζc u2c = −ζc − gρc ζc − (5.50)
∂z dz A
Here χg and χe are the gas mass fraction and fraction of the liquid phase that is entrained.
The interfacial shear stress can be determined from the momentum balance based on the
total pressure gradient and above mean core velocity.
If the total pressure gradient is not known, then the interfacial shear stress can be esti-
mated based on an approximation of the shear stress in the absence of phase change using
an equivalent laminar film method. In this method, it is assumed that a laminar boundary layer
is formed in the gas phase adjacent to the interface and the velocity changes from the mean
value of the gas phase, ug, to the liquid interfacial velocity, ul. Then the shear stress, τ, in the
liquid film can be determined through a force balance on an annular ring of axial length of
Δz, outer radius of r, and inner radius of ri, where ri = r − δ, as follows,
2 dp 2
2πrτΔz = 2πrτi Δz + ρl gπΔz ri − r + π p − p + Δz
2
r − r2i (5.52)
dz
Gas–Liquid Two-Phase Flows 211
From left to right, the terms represent the shear stress at the outer radius, interfacial
shear stress, gravitational force, and net pressure force on the annular ring. Rearranging
this force balance,
ri 1 dp r2i − r2
τ = τi + ρg+ (5.53)
r 2 l dz r2
This result expresses the shear stress as a function of radial position in the film.
Using this shear stress distribution, the velocity profile in the liquid film can also be
determined by integration of the constitutive relation for a Newtonian fluid across the
film, yielding:
1 1 dp 2 ro 1 dp 2
u= τ i ri + ρl g + ri ln − ρl g + ro − r2i (5.54)
μl 2 dz ri 4μl dz
Further integration of this velocity profile across the film yields the following film
flow rate:
2πρl 1 dp 2 1 2 1 ro πρ dp 2 2
ṁf = τ i ri + ρl g + ri ro − r2i − r2i ln − l ρl g + ro − r2i (5.55)
μl 2 dz 4 2 ri 8μl dz
where the subscripts f, i and o refer to the liquid film, interface (or inner) and outer, respec-
tively. For turbulent flow, the effective viscosity can be calculated from the sum of the
laminar and turbulent viscosities (Chapter 3). Further details on turbulent viscosity model-
ing in the liquid film of annular flows are presented by Hetsroni (1982).
The liquid film thickness, δ, can be determined based on the mass flow rate ratio, ṁf /ṁg ,
and the interfacial shear stress, τi, which in turn is related to the pressure gradient in the flow
field, dpf/dx. This three-way relationship between δ, ṁf /ṁg , and δ, is known as a “triangular
relationship.” Given any two of the parameters, the remaining third parameter can be deter-
mined based on the shear stress distribution in the film. For thin films, a constant shear stress
can be assumed. Also, an “interfacial roughness relationship” is used to relate the effective
roughness of the interface and the liquid film thickness.
For example, the mass flow rate in the liquid film can be calculated based on the triangular
relationship in the following four steps:
1. The interfacial shear stress, τi, is determined based on the pressure gradient.
2. Then the shear stress distribution in the liquid film is calculated from a force balance
and the shear stress obtained in the previous step.
3. The velocity profile in the liquid film is then obtained from the shear stress profile
and the effective viscosity.
4. Finally, the mass flow rate, ṁf , is obtained by integration of the velocity profile.
These relations apply to concurrent flow with both the liquid and gas phases moving in
the same direction. If there is a counterflow with the liquid film flowing downward through
the channel and upward gas counterflow, then as the gas velocity increases, waves are gen-
erated on the film surface which are carried upward by the gas stream. This transition point
212 Advanced Heat Transfer
is called “flooding.” Eventually as the gas velocity increases, the liquid film reverses direc-
tion and both the liquid and gas motion becomes concurrent in the same upward direction.
The total heat transfer coefficient, htot, can be obtained in terms of the convective and boil-
ing coefficients, hc and hb, respectively. Define the mass velocity, G, and density-weighted
phase quality (or Martinelli parameter), X, as follows:
ṁ
G= (5.56)
πD2 /4
0.9
1 − χg ρv 0.5 μl 0.1
X= (5.57)
χg ρl μv
where χg is the quality (vapor fraction). Also, the dynamic transition factors, F and S, are
defined as follows:
2.35(0.213 + 1/X)0.736 if 1/X . 0.1
F= (5.58)
1 if 1/X ≤ 0.1
⎧
⎪
⎪ 0.1 if ReT . 70
⎨
S = 1/ 1 + 0.42Re0.78 if 32.5 , ReT , 70 (5.59)
⎪
⎪
T
⎩
1/ 1 + 0.12Re1.14
T if ReT , 32.5
where,
GD(1 − χ g ) −4 1.25
ReT = 10 F (5.60)
μl
Then the single phase heat transfer coefficient, hc, can be determined based on ReT using
previous convection correlations in Chapter 3 with saturated liquid properties.
hc D GD(1 − χ g ) 0.8 0.4
NuD,c = = 0.023F Pr (5.61)
kl μl
Also, the boiling coefficient, hb, can be estimated based on the form of previously devel-
oped boiling correlations in this chapter:
p,l ρl
c0.45 0.49
hb D
NuD,b = = 0.00122 · D · S 0.5 0.21 0.29 0.24 0.24 ΔT 0.24 Δp0.75
sat (5.62)
kl σ kl μl hfg ρv
Using these two correlations, the overall heat transfer coefficient, htot, for the combined
convection–boiling flow is approximated by:
htot = F · hc + S · hb (5.63)
Gas–Liquid Two-Phase Flows 213
This result includes the contributions of both the convection coefficient, hc, and the boiling
heat transfer coefficient, hb. Further correlations of annular boiling flow over a wide range of
operating conditions were presented by Chen (1963).
FIGURE 5.6
Flow regimes in horizontal two-phase flows.
214 Advanced Heat Transfer
including the geometrical configuration and turbulence in the liquid phase. A transition
between stratified and slug flow occurs with the onset of Kelvin–Helmholtz instabilities
(an instability in the flow of two horizontal parallel streams of fluids at different velocities
and densities with one stream above the other). In the transition of slug to annular flow, clas-
sical theories assume that the point of transition occurs where the equilibrium liquid level
at the onset of the Kelvin–Helmholtz instability is less than half of the channel diameter.
The classification of two-phase flows into different flow regimes is often illustrated by
two-phase flow maps. Typical two-phase flow maps for vertical and horizontal flows in tubes
are illustrated in Figure 5.7. The graphs are normally plotted in terms of the velocity, mass
flux and/or phase fraction. In the two-phase flow map for horizontal tubes in Figure 5.7, the
map is plotted in terms of the pressure gradient parameter (Xp) and superficial velocity
parameter (F ) as follows:
(dp/dx)l
Xp2 = (5.64)
(dp/dx)g
ρg ug
F=
(5.65)
ρl − ρg Dg cos θ
where (dp/dx)l and (dp/dx)g are the pressure gradients in the tube for the liquid and gas
phases, respectively; D is the tube diameter; and θ is the angle of inclination of the tube.
The superficial velocity, ug, is a hypothetical (artificial) flow velocity in the gas phase that
would be obtained if the gas was the only phase flowing or present in the given cross-
sectional area.
The overall pressure drop of the two-phase flow mixture, Δpm, in the tube can be
expressed by:
where the terms on the right side, from left to right, represent the pressure losses due to flow
acceleration of the liquid and vapor phases due to changes of vapor fraction or channel
cross-sectional area; friction; and hydrostatic effects. For one-dimensional flows in the axial
Annular/ Dispersed/
Vapour momentum (ρgUg2)
FIGURE 5.7
Two-phase flow maps for (a) vertical and (b) horizontal tubes.
Gas–Liquid Two-Phase Flows 215
Here ρm, θ, χg and D refer to the mixture density of the two-phase flow, angle of incli-
nation of the tube with respect to the horizontal plane, vapor mass fraction, and tube
diameter, respectively. Also, C and ftp refer to the resistance coefficient and two-phase
flow friction factor, respectively. Both factors are functions of the Reynolds number, deter-
mined empirically, and which depend on the flow rate, pressure, and inclination angle of
the tube. These factors include the effects of bubbles along the wall, which increase
frictional forces.
Two types of predictive models are normally used for the various flow regimes—either
homogeneous or separated flow methods. In a homogeneous model of two-phase flow,
mass weighted and averaged parameters are used such that the velocity of the liquid
and gaseous phases are equal. The bubble and dispersed phases are most effectively
described by a homogeneous model due to the nearly uniform distribution of the dis-
persed phase in the carrier stream. The model is also effective in two-phase flows at
high pressures when the densities of the liquid and vapor phases approach each other.
In contrast, a separated flow model allows for differences in phase velocities and energy
interactions at the interfacial boundaries. This model is generally more computationally
intensive, but more accurate in resolving flow structures in annular, annular/dispersed,
and wave flows.
U = Ul + Ug (5.70)
where Ul and Ug refer to the liquid and gas superficial velocities. Consider further an
imaginary plane moving along the axial direction in the tube at a velocity of U. The drift
velocity for each phase with respect to this plane is defined as:
ugU = ug − U; ulU = ul − U (5.71)
The drift velocities, ugU and ulU, refer to the velocities of each phase, ug and ul (gas and
liquid phases), respectively, relative to the total superficial velocity.
Also, define the drift fluxes, jgl and jlg, for the flux of each phase through the plane moving
at the velocity U. The drift fluxes are given by the drift velocities multiplied by the area
216 Advanced Heat Transfer
where ζg refers to the fraction of the tube volume that is occupied by the gas phase.
From conservation of mass, there is no drift through the plane moving at U,
When the void fraction approaches zero, also the above drift fluxes approach zero. At very
high void fractions, the drift flux also approaches zero and therefore the drift flux passes
through a maximum point between zero and unity.
The rise velocity of bubbles, u∞, is directly related to the drift flux, jgl, and the fraction of
the tube cross-sectional area occupied by bubbles, ζg. At a low void fraction, the drift flux can
be expressed by the bubble rise velocity multiplied by the bubble fraction. As the number of
bubbles increases, the interference among bubbles also increases and the drift flux
approaches zero as ζg → 1. A suitable functional form of this relationship is given by:
jgl = u1 ζ g (1 − ζg )n (5.75)
2ρl u1 rb gμ4l
Reb = ; Gal = (5.76)
μl ρl σ 3
where rb and σ refer to the bubble radius and surface tension, respectively.
In the bubble flow regime, Rohsenow and Griffith (1955) estimated the total wall heat
flux as follows:
TABLE 5.1
Expressions and Coefficients for the Bubble Rise Velocity
Range Bubble Rise Velocity, u∞ Coefficient, n
The total heat flux at the wall (subscript w, tot) is the sum of the heat fluxes due to single
phase convection (subscript w, conv) and nucleate boiling (subscript w, nuc). Correlations
from Chapter 3 for forced convection, and nucleate boiling correlations such as Equation
5.2, can be used in this model. This approximation can be applied to convection boiling,
where the near-wall fluid is at or above the saturation temperature, but the bulk portion
of the remaining liquid temperature is subcooled.
The Miropolski correlation (Miropolski 1963) for the dispersed bubble flow regime is given
by:
where,
0.4
ρl
Y = 1 − 0.1 −1 (1 − ζg )0.4 (5.79)
ρg
ρmix uD ρg
Remix = χ g + (1 − χ g ) (5.80)
μg ρl
Here Remix and χg are the mixture two-phase flow Reynolds number, based on the tube
diameter D, and vapor fraction, respectively.
As bubbles move through the liquid phase and ascend under buoyancy, mass transfer
occurs from the liquid to gas phase, thereby affecting the bubble size and shape. The Morton
number (Mo) characterizes the bubble shape and the Eötvos number (Eö) characterizes the
ratio of the buoyancy force on a bubble to the surface tension, as follows:
gD2b (ρl − ρg )
Eö = (5.81)
σ
gμ4 (ρl − ρg )
Mo = (5.82)
ρ2l σ 3
FIGURE 5.8
Schematic of parameters of stratified flow.
The liquid phase flows along the bottom of the tube to a height of hl and a gas phase exists at
the top. Neglecting inertial terms, the integrated form of the momentum equations in the
axial direction, x, for the liquid and gas phases, can be written as:
dp
−Al − τ0l Pl + τi Pi − ρl Al g sin θ = 0 (5.84)
dx
dp
−Ag − τ0g Pg + τi Pi − ρg Ag g sin θ = 0 (5.85)
dx
where Al, Ag, τ0l, τ0g, τi, Pi, Pl, and Pg are the cross-sectional flow areas of the liquid and
gas phases; wall shear stresses for sections of the tube in contact with the liquid and gas;
interfacial shear stress; and perimeters in contact with the interface, liquid and gas,
respectively.
Combining these equations and eliminating the pressure gradient,
Pg Pl 1 1
τ0g − τ0l + τi Pi + − (ρl − ρg )g sin θ = 0 (5.86)
Ag Al Al Ag
The wall and interfacial shear stress terms can be expressed as:
1 1
τ0l = fl ρl u2l ; τ0g = fg ρg u2g (5.87)
2 2
1
τi = fi ρg (ug − ul )2 (5.88)
2
where fl and fg are friction factors in the liquid and gas phases, respectively. These factors can
be expressed in terms of the equivalent diameters, Dl and Dg, for the liquid and gas phases,
−n −m
Dl ul ρl Dg u g ρ g
fl = C l ; fg = C g (5.89)
μl μg
4Al 4Ag
Dl = ; Dg = (5.90)
Pl Pg + Pi
where the constants are Cg = 16 = Cl and n = 1.0 = m for laminar flow. For turbulent flow,
Cg = 0.046 = Cl and n = 0.2 = m.
Gas–Liquid Two-Phase Flows 219
Dl Dg Al Ag
D∗l = ; D∗g = ; A∗l = ; A∗g = (5.91)
D D D2 D2
Pl Pg Pi
P∗l = ; P∗g = ; P∗i = (5.92)
D D D
ul ug
u∗l = ; u∗g = (5.93)
Ul Ug
where D is the tube diameter and Ug and Ul are the superficial gas and liquid velocities,
respectively. Using these dimensionless variables and friction factors in Equation 5.86, the
momentum balance becomes:
−n ∗2 P∗l −m ∗ P∗g P∗i
∗2 Pl
X 2
u∗l D∗l ul ∗ ∗
− ug Dg ug + + − 4Y = 0 (5.94)
A∗l A∗l A∗g A∗g
where X is the Martinelli parameter. The pressure gradient parameter, Y, represents the
effect of tube inclination as follows:
(ρg − ρl )g sin θ
Y= (5.95)
(dpf /dx)g
where (dpf/dx)g is the frictional pressure gradient in the gas phase. The nondimensional
areas and perimeters can be related to the dimensionless liquid height, h∗l = hl /D, as follows
(refer to Figure 5.8):
Al 1 ∗ ∗
∗ 2
π
A∗l = = π − cos−1
2hl − 1 + 2hl − 1 1 − 2hl − 1 = − A∗g (5.96)
D2 4 4
Pi 2 Pi
P∗i = = 1 − 2h∗l − 1 ; P∗l = = π − cos−1 2h∗l − 1 = π − P∗g (5.97)
D D
Also, the dimensionless equivalent diameter for the liquid phase and superficial velocities
can be expressed as:
u g A l + Ag u l A l + Ag
u∗g = = ; u∗l = = (5.99)
Ug Ag Ul Al
When these relationships are substituted into Equation 5.94, for given values of X and
Y corresponding to the known flow conditions, the nondimensional liquid height, h∗l ,
can be determined. Then, the corresponding gas fraction can also be determined based on
the above geometrical relationships between the liquid area, gas area, and total cross-
sectional area of both liquid and gas phases.
220 Advanced Heat Transfer
ug = C0 U + us (5.100)
where us is the rise velocity of a single slug flow bubble in a static liquid and U = Ug + Ul
is the total superficial velocity. The coefficient C0 is an empirical distribution parameter
(C0 = 1.2 is commonly used). The fraction of the tube volume occupied by the gas phase,
ζg, can be calculated from:
Ug
ζg = (5.101)
ug
The rise velocity can be expressed in terms of the Viscosity number, Vi, and Eötvos num-
ber, Eö, as follows:
us 3.37 − Eö
D3 g(ρl − ρg )ρl
Vi = (5.103)
μl
The Viscosity and Eötvos numbers represent the effects of the liquid velocity and surface
tension on the slug motion. The empirical coefficient m varies with Vi as follows: m = 25 (for
Vi , 28); m = 69 Vi−0.35 (for 18 , Vi , 250); and m = 10 (for Vi . 250).
For low fluid viscosities and large tube diameters, the expression for the rise velocity can
be simplified as follows:
us = 0.345 gD (5.104)
This expression for the slug rise velocity can also be obtained from a potential flow solu-
tion based on inviscid flow theory.
For annular flow, models in the previous section for inclined and vertical tubes may also
be applied to horizontal tubes. Due to the diverse flow conditions and geometrical config-
urations which may occur, universal relations for all annular flow conditions are not avail-
able, or those available have limitations that should be used with caution.
Solutions can be obtained by first establishing the void fraction, followed by the interfacial
stress, velocity in the liquid film, and then the entrained liquid mass flow rate. Alternatively,
the liquid film thickness can be determined from the liquid film thickness and pressure
gradient using the triangular relationship, after which the void fraction can be determined.
Then the friction factor is calculated and the governing equations for mass, momentum,
and energy transport are solved to determine the velocity and temperature fields. Further
Gas–Liquid Two-Phase Flows 221
details on the modeling of laminar and turbulent annular flows are described by
Hetsroni (1982).
hD b
NuD = = c Re2D F (5.105)
kf
in the range of 109 , Re 2 F , 0.7 × 1012 where the Reynolds number, Re = GD/μl, is based on
the tube diameter, D, and mass velocity, G, in Equation 5.56. For incomplete vaporization,
c = 0.0009 and b = 0.5, whereas for complete vaporization, c = 0.0082 and b = 0.4. Also, the
dimensionless load factor, F, is defined by:
hfg Δζg
F = 102 (5.106)
L
where L is the tube length (m), hfg is the enthalpy of evaporation (kJ/kg), and Δζg refers to
change in the vapor fraction over length of the tube. The correlation was successfully
applied to refrigerants flowing in horizontal tubes over a range of boiling flow regimes.
The fraction occupied by the gas phase in the inner core of the tube, ζg, may be determined
by Armand’s correlation as follows:
4 + (8/7)b
ζg = 1 − (5.107)
5 + b(β/(1 − β) + 8/7)
where,
0.5
ρg ṁg /ρg
b= 4aRe0.125 ; β= (5.108)
L
ρl ṁg /ρg + (ṁlf + ṁle )/ρl
Here ṁg , ṁlf and ṁle refer to the flow rates of the gas (vapor), liquid in the wall film, and
liquid entrained in the core. Also, the empirical coefficient, a, and Froude number, Fr, are
defined by:
a = 0.69 + (1 − β) 4 + 21.9 Frl (5.109)
ṁ2l
Frl = (5.110)
ρ2l gD
The mass flow rate of liquid, ṁl , is the sum of the flow rates of liquid in the wall film and
entrainment in the core.
Alternatively, Kandlikar (1990) presented a single general correlation for a range of
flow regimes which is applicable to both vertical and horizontal tube orientations. The ratio
of the heat transfer coefficient of two-phase flow, htp, to the corresponding single phase
222 Advanced Heat Transfer
TABLE 5.2
Coefficients for the Kandlikar Correlation
C0 , 0.65 C0 . 0.65
(Convective Region) (Boiling Region)
C1 ¼ 1.1360 0.6683
C2 ¼ –0.9 –0.2
C3 ¼ 667.2 1,058.0
C4 ¼ 0.7 0.7
C5 ¼ 0.3 0.3
htp
= C1 CC0 2 (25Frl )c5 + C3 BC0 4 Ffl (5.111)
hl
where the convection number, C0, and boiling number, B0, are defined as follows:
0.8
1 − χg ρv 0.5
C0 = (5.112)
χg ρl
qw
B0 = (5.113)
AGhfg
The single phase convection coefficient, hl, can be determined based on the Dittus–Boelter
correlation (Chapter 3). The coefficient Ffl is dependent on the type of working fluid, for
example, Ffl = 1 (water), Ffl = 4.7 (nitrogen), Ffl = 1.24 (R-114), Ffl = 1.1 (R-113, R-152a),
and Ffl = 1.63 (R-134a).
Recommended values of the empirical coefficients are summarized in Table 5.2. For
vertical tubes with Frl . 0.04 and horizontal tubes, it is recommended that C5 = 0.
Kandlikar’s correlation is applicable over the range of 0.001 , χg , 0.95. The correlation
was developed based on experimental data obtained from 24 sources involving many dif-
ferent working fluids including refrigerants and water.
gravity. In general, there are few and limited comprehensive models and experimental data
of dropwise condensation (Takeyama and Shimizu 1974).
In film condensation, a liquid film covers the entire surface. The rate of heat transfer is
usually at least an order of magnitude smaller than dropwise condensation, particularly
at small temperature differences between the wall and saturation temperature. Typical val-
ues of the condensation heat transfer coefficient, h, for steam range between 10,000 and
30,000 W/m2 K for horizontal tubes and between 5,000 and 15,000 W/m2 K for vertical
surfaces if the surface is about 3–20◦ C below the saturation temperature. This section will
primarily focus on film condensation as dropwise condensation is not usually sustained
over long periods of time.
y
δ
y
x Plate at Tw
Tw Liquid
Tsat
Vapour
mv Tsat T(y)
x
L
x + dx mv (vapour)
x
mx
δ(x)
x + dx
u(y)
mx+dx
dx
(liquid)
Vapour
Liquid Wall
FIGURE 5.9
Film condensation on a vertical plate—velocity and temperature profiles.
224 Advanced Heat Transfer
terms in the energy equation. The reduced form of the mass, momentum, and energy equa-
tions becomes:
∂ul ∂vl
+ =0 (5.114)
∂x ∂y
∂2 u l
μl = −gx (ρl − ρv ) (5.115)
∂y2
∂2 T
=0 (5.116)
∂y2
where gx refers to the component of gravitational acceleration in the direction along the
surface. For example, gx = g for a vertical plate, whereas gx = 0 for a horizontal plate.
At the wall ( y = 0), no-slip and specified temperature conditions are applied. Boundary
conditions at the interface ( y = δ) are T = Tsat, a zero-velocity gradient in the liquid film,
and the following interfacial heat balance:
∂T
kl = ṁ′′v hfg (5.117)
∂y y=δ
Define the liquid mass flow rate along the surface, per width W, as Γ = ṁl /W. From a
mass balance within the liquid film, the rate of change of the liquid mass flow rate in the
x-direction must balance the vapor supply rate. Using this mass balance, along with the
above temperature profile and interfacial heat balance,
dΓ kl (Tsat − Tw )
= ṁ′′v = (5.120)
dx δhfg
Thus the liquid film growth and heat transfer are controlled by the thermal resistance of
the film, which depends on the local thickness of the liquid layer.
The mass flow rate, Γ, can also be determined by integration of the velocity profile,
Equation 5.118, across the film from y = 0 → δ, as follows,
δ
ρ (ρ − ρv )gx 3
Γ = ρl ul dy = l l δ (5.121)
0 3μl
The film thickness can be determined from Equation 5.121 based on the result in
Equation 5.122. A correction to Equation 5.122 to include thermal advection effects was
recommended by Rohsenow (1956):
Since the temperature profile in the liquid is linear, the heat flux is proportional to
ΔT (= Tw − Tsat) divided by the film thickness, δ. Then the local Nusselt number can be
obtained and integrated along the surface to give the following average Nusselt number:
⎡ s ⎤3/4
hS 4 gρl (ρl − ρv )h′ fg S3 1/4 1 gx
1/3
hL 1/4
g(ρl − ρv )h′ fg L3
NuL = = 0.943 (5.125)
kl kl (Tsat − Tw )νl
The effects of the Prandtl number on the heat transfer from a laminar film of condensate
are illustrated in Figure 5.10. Additional restraining effects of vapor drag on the downward
2.0
Pr → ∞
Pr = 10
NuL [0.943 L3 hfg′ g(ρl –ρv)/(kl νl ΔT)]–1/4
Pr = 1
0.3
0.1
0.03
0.01
0.003
Pr = 0.001
0.02
0.002 0.02 0.2 2
Ja = cp ΔT/h′fg
FIGURE 5.10
Heat transfer of laminar film condensation on a vertical plate. (Adapted from M.M. Chen. 1961. ASME Journal of
Heat Transfer, 83: 48–54.)
226 Advanced Heat Transfer
acceleration of the liquid film lead to lower Nusselt numbers at small Prandtl numbers. In
general, the Nusselt number decreases at higher Jacob numbers except above Prandtl and
Jacob numbers of about 1 (see Figure 5.10).
Good agreement is achieved between this model and experimental data for problems
involving laminar film condensation. Discrepancies arise due to factors such as wall rough-
ness, turbulence, or non-condensable gases in the vapor. Chin, Ormiston, and Soliman
(1998) reported the effects of non-condensable gases on the boundary layer formation and
growth during film condensation.
′ 3 1/4
h ≈ 0.725 gρl (ρl − ρv )h fg kl (5.126)
μl D(Tsat − Tw )
This correlation is also illustrated in Figure 5.10 by replacing the coefficient 0.943 and
plate length, L, with a coefficient 0.725 and tube diameter, D, respectively.
For N tubes, replace D by ND in Equation 5.126. For example, the coefficient is reduced
from 0.725 to 0.56 for the case of N = 10. Alternatively, for a horizontal tube bank with N
tubes placed directly above each another in the vertical direction, the convection coefficient
for the N-tube arrangement, hN, can be approximated in terms of the single tube coefficient,
h, as follows (Kern 1958):
For surfaces such as plates or cylinders inclined at an angle of θ with respect to the hori-
zontal plane, the resulting heat transfer coefficient can be approximated by replacing the
gravitational acceleration, g, by g sin θ and integrating the modified result around the sur-
face. This body gravity function method reflects the modified component of gravity in the
direction of condensate flow along the surface.
A similar result is obtained for film condensation around a sphere (diameter D),
′ 3 1/4
h ≈ 0.815 gρl (ρl − ρv )h fg D (5.128)
μl kl (Tsat − Tw )
The heat transfer coefficient decreases at larger values of μl and temperature excess
(Tsat − Tw). For a higher liquid viscosity or temperature excess, a thicker liquid film is
formed, which leads to a larger thermal resistance to heat transfer between the surface
and surrounding saturated vapor.
Gas–Liquid Two-Phase Flows 227
Chen, Gerner, and Tien (1987) presented the following correlation of the average heat
transfer coefficient through both the wavy and turbulent regions.
This correlation as well as the laminar flow correlation in Equation 5.125 are illustrated in
Figure 5.11. The heat transfer coefficient reaches a minimum and then increases with both
Prandtl and Reynolds numbers in the turbulent region.
In general, these correlations agree within +10% of experimental data when the vapor
is motionless or slow enough that shear effects at the gas–liquid interface are negligible.
Once the heat transfer coefficient is determined, the rate of heat transfer from the surface
can be obtained by multiplying the heat transfer coefficient by the exposed surface area
and temperature difference (between the surface and ambient temperatures).
where νt and αt represent the turbulent eddy viscosity and eddy diffusivity, respectively.
1
Pr = 10
6
hL (νl2/g)1/3/kl
Laminar Wavy
region region Turbulent region
0.1
10 100 1,000 10,000 100,000
ReL
FIGURE 5.11
Average heat transfer coefficient for laminar, wavy, and turbulent film condensation. (Adapted from A. Bejan. 2013.
Heat Transfer, 4th Edition, New York: John Wiley & Sons.)
Gas–Liquid Two-Phase Flows 229
δ θ x
g
τw
τδ ro
Condensate film
Sphere surface
FIGURE 5.12
Schematic of turbulent film condensation on a sphere.
Integrating the energy equation across the condensate layer leads to:
δ
ρl d kl dT
( sin θ)udy = ro (5.137)
sin θ dθ hfg dy y=0
0
This result indicates a balance among the wall shear stress, gravitational force on the con-
densate film, and interfacial shear stress. The shear stress at the liquid–vapor interface, τδ,
can be expressed in terms of the local friction coefficient, fθ, as follows:
τδ
fθ = = 2CRev−0.2 (5.139)
ρv u2v /2
where C is a constant depending on the flow configuration. The vapor velocity at the edge of
the boundary layer is estimated by a potential flow solution of a uniform vapor flow over a
sphere,
3
uv = U sin θ (5.140)
2
9
τw − Cρv U2 Re−0.2
v sin2 θ − gδ(ρl − ρv ) sin θ = 0 (5.141)
4
230 Advanced Heat Transfer
yu∗ δu∗
y+ = ; δ+ = (5.142)
νl νl
τw u T − Tw
u∗ = ; u+ = ; T+ = (5.143)
ρ u∗ Tsat − Tw
Re+ u ∗ ro
Re∗ = ; Re+ = (5.144)
Gr1/3 νl
U2 gr3o ρl − ρv cp (Tsat − Tw )
Fr = ; Gr = 2 ; Ja = (5.145)
gro νl ρl hfg
The turbulent Prandtl number is assumed to be unity so the momentum eddy diffusivity,
νm, is assumed to be equal to the heat transfer eddy diffusivity, αh.
The elemental force balance can be rewritten in nondimensional form as follows:
9
Re∗3 = Re∗2 ϕFr0.9 sin2 θ + δ+ sin θ (5.146)
8π
Rearranging this result yields the following expression for the local dimensionless thick-
ness of the condensate layer,
Results of this turbulence model were presented and discussed by Hu and Chen (2005).
It was found that the Nusselt number is significantly influenced by the vapor velocity
and its role in the onset of turbulence in the condensate film. Laminar condensate flow mod-
els are accurate only at low vapor velocities. At high vapor velocities, the Nusselt number
increases with higher values of the interfacial shear parameter, whereas for low vapor veloc-
ities, the influence of the shear parameter on the heat transfer coefficient is less significant.
FIGURE 5.13
Flow regimes for condensation with forced convection in a horizontal tube.
232 Advanced Heat Transfer
A condensed liquid flows along the inner surface of the tube wall concurrently with vapor
flow in the inner core of the channel (annular/dispersed flow region). Droplets are entrained
into the vapor core flow if the vapor velocity is high due to interfacial shear forces along
the liquid–gas interface. The role of gravitational forces increases with a higher liquid
fraction. The flow pattern becomes wavy and separated with more liquid accumulating
in the bottom of the tube (stratified/wavy flow region). Eventually a slug/plug flow is
formed with vapor regions confined along the upper wall of the tube. With further
cooling the entire vapor stream condenses and a single phase liquid flows near the end of
the tube.
Heat transfer correlations depend on the flow regime and phase distribution in Figure 5.13.
For the single-phase region of superheated vapor, prior correlations from Chapter 3 may be
used, such as the Dittus–Boelter correlation.
In the annular flow region, a condensate film grows along the walls while vapor
and dispersed droplets occupy the inner core of the tube. This flow resembles forced con-
vection with vapor inside a horizontal tube and a liquid thermal resistance due to the
condensate film along the wall. Fujii et al. (1977) presented the following heat transfer
correlation:
Nu 0.45f
St = =
(5.154)
Re Pr 1 + 5 f /2[Prv − 1 + ln(1 + 5(Prv − 1)/6)]
where St, Prv, and f represent the Stanton number, Prandtl number of the vapor, and friction
factor, given by:
0.046ζ2.5
g X
2
f = (5.155)
(ṁv D/μv )0.2
Here ζg, X, D and μv are the fraction of the tube occupied by the gas phase, Martinelli
parameter, tube diameter, and kinematic viscosity of the vapor, respectively. The gas vol-
ume fraction can be expressed in terms the mass fraction of vapor, χg, as follows:
1
ζg = 1/2
(5.156)
1 + (ρg /ρl ) (1 − χ g )/χ g
Then the total wall heat flux, including the effects of condensation and forced convection,
is given by:
2ri
qw = (qi + ṁc hfg ) (5.157)
D
where qi, hfg, ri and ṁc represent the interfacial condensing heat flow rate determined by
the heat transfer coefficient in the Stanton number correlation, Equation 5.154, latent heat
of vaporization, average radius of the vapor-liquid interface, and mass flow rate of the
condensate, respectively.
In the condensing saturated vapor region, heat transfer occurs by condensation of vapor
onto the growing liquid condensate layer. The wall heat flux can be determined from the
same above expression for the wall heat flux excluding the interfacial heat flux, qi.
Gas–Liquid Two-Phase Flows 233
Jaster and Kosky (1976) reported various criteria for transition between annular and strat-
ified flows based on the following transition factor, F:
fl u2l
F= (5.158)
2gδ
where ul refers to the condensate velocity and δ is the condensate film thickness. The liquid
friction factor, fl, is calculated as though liquid alone is flowing through the pipe. The con-
densate velocity can be approximated by:
Vl
ul = (5.159)
1 − ζg
where Vl and ζg refer to the liquid velocity in the pipe (based on the same mass flow rate,
but in the absence of vapor) and fraction of vapor flow area divided by the total area,
respectively. Based on these definitions, it was observed that F ≥ 29 exhibits annular flow,
F ≤ 5 yields stratified flow, and 5 , F , 29 is a transition regime between stratified and
annular flow.
Due to the complexities of flow instabilities associated with transition between various flow
regimes, two-phase flow maps are commonly used to characterize the flow patterns (such as
Figure 5.7). Mandhane et al. (1974) and Travis and Rohsenow (1973) presented flow maps for
forced convection condensation in a horizontal tube of various working fluids. These flow
maps include the entire range of flow regimes and typically use the superficial liquid velocity,
Vsl, and superficial gas velocity, Vsg, for classification of the various regions,
(1 − χ g )G χgG
Vsl = ; Vsg = (5.160)
ρl ρv
The heat transfer coefficient varies along the periphery of the tube, while having its largest
value at the bottom.
An external vapor flow within or past the condensation surface alters the rate of heat
transfer at the surface. With reference to Figure 5.13, in a vertical tube, the condensing fluid
usually flows downward. In this case, the flow regime is mostly annular flow through the
tube and it changes to slug flow just prior to the end of the tube.
The following Carpenter–Colburn correlation can be used to find the heat transfer coefficient
for pure vapors of steam and hydrocarbons up to Gm ≈ 150 m/s,
where,
2
Gm = G1 + G1 G2 + G22 /3 (5.165)
f = 0.046 Re−1/5
v (5.166)
and Rev = GmD/μv. The values G1 and G2 refer to the mass velocities of the inlet and exit
vapor flows, respectively, for example, G1 = χ 1 ṁ/A, where A is the tube cross-sectional area.
where the subscripts g and s refer to gravity and shear stresses alone, and the following
correlation may be used:
hs D ρl uv D
NuD = = C Re1/2
D ; ReD = (5.168)
kl μl
where D is the tube diameter. Here C = 0.9 for ReD , 106, whereas C = 0.59 is recommended
when ReD . 106.
Gas–Liquid Two-Phase Flows 235
For vertical vapor downflow over a tube with a condensate surface film, the Nusselt
number can be approximated by:
0.9 + 0.728F1/2
NuD = 1/4
Re1/2
D (5.169)
(1 + 3.44F1/2 + F)
where,
Grl μl hfg
F= (5.170)
ReD kl (T − Tsat )
2
ρl (ρl − ρv )D3
Grl = (5.171)
μ2l
The subscripts l and v refer to liquid condensate and vapor, respectively. The results agree
reasonably well with experimental data at low vapor velocities. At higher vapor velocities,
the correlation tends to overestimate the mean heat transfer coefficient for steam, while
underestimating it for refrigerants when F , 1. This may occur as a result of the variation
of pressure of steam, and hence saturation temperature, around the tube, and effects of tur-
bulence in the condensate layer.
The quantity F indicates the relative magnitudes of the gravitational force and vapor shear
stress at the liquid–gas interface. For large values of F above 10, the effects of the vapor
velocity become negligible. Then the above correlation can be simplified as follows:
1/4
μl hfg
NuD = 0.728F1/4 Re1/2
D = 0.728 Grl (5.172)
kl ΔT
For values of F less than 0.1, it can be shown that gravitational effects become negligible
and the correlation can be reduced to a form corresponding to the case of condensate flow
and heat transfer dominated by the vapor shear stress (F = 0), yielding:
Further extensions of this model have been presented by Rose (1988) including effects of
the direction of the vapor flow relative to gravity, pressure variations in the condensate film
due to the vapor flow around the surface, and improved estimates of the vapor shear stress
and vapor boundary layer separation. The shear stress exerted by the vapor on the surface
film affects the rate of condensation since the flow of vapor past the surface reduces the
thickness (and thus thermal resistance) of the liquid film. This leads to an enhancement of
the heat transfer rate.
The previous results for a single tube can be extended to the configuration of external flow
of saturated vapor past a number of tube bundles. The vapor velocity, uv, in Equation 5.168,
is replaced by uv → uvo/ζv, where ζv is the void fraction of the tube bundle (ratio of the free
volume to the total volume), and uvo is the vapor velocity that would be obtained without
the presence of the tubes. The gravity component of the heat transfer coefficient is based
on a previous correlation obtained for N tubes in a vertical alignment, Equation 5.127.
236 Advanced Heat Transfer
where A, Ab, and Af refer to the total area and areas of the base and fin, respectively, and η
refers to the fin efficiency, which is related to the areas as follows,
A = Ab + ηAf (5.175)
The fin efficiency can be calculated from conventional fin relations (Chapter 2). Also, the
variable L refers to the average fin height over the circumference of the tube,
π D2o − D2i
L= (5.176)
4 Do
where Do and Di are the outer and inner diameters of the finned tube, respectively.
Beatty and Katz (1948) reported that the fin efficiency of copper tubes with short fin
heights (less than 1.6 mm) is generally larger than 0.96. The result in Equation 5.174 has
assumed that the temperature difference between the wall and surrounding vapor is the
same throughout the fin. Young and Ward (1957) raised the last expression in brackets in
Equation 5.174 to a three-fourths power, rather than the first power, to accommodate the
varying temperature difference between the vapor and wall.
The fins lead to a thinner condensate film, thereby reducing the thermal resistance in com-
parison to an unfinned tube. The last term in brackets in Equation 5.174 may be interpreted
as an equivalent diameter to the power of −1/4. This effectively leads to a smaller equivalent
diameter of a finned tube than an unfinned tube of diameter D, thereby leading to a larger
heat transfer coefficient. Even low-finned tubes can have a significantly larger heat transfer
coefficient than an unfinned tube of equal diameter. Various types of extended surfaces and
fins on tubes are widely used in industrial condenser equipment, including fins of a saw-
tooth shape, as well as wires loosely attached to the tubes.
Heat Condensing
output section
Heat Boiling
Input Section
FIGURE 5.14
Schematic of a thermosyphon.
various energy and industrial systems, including heat exchangers, solar energy devices,
thermal control of food storage, and other applications.
Heat pipes are similar devices but they operate in any orientation and use capillary forces
in a wick, rather than gravitational forces, to return liquid from the condenser to the evap-
orator. A heat pipe uses the wick structure instead of gravity to return the liquid flow from
the condenser to the evaporator. It operates by vaporization of liquid in the evaporator,
transport of vapor through the core of the pipe to the condenser, heat transfer by condensa-
tion, and liquid return flow by capillary action in the wick back to the evaporator (see
Figure 5.15). The adiabatic section is designed to fit external geometrical requirements
such as space limitations. At the evaporator section, heat input from an external source is
transferred to the working fluid. A buffer volume may be constructed at the end of the
heat pipe to enclose a non-condensable gas (such as helium or argon) and control the oper-
ating temperature by controlling the pressure of the inert gas. Vapor flow through the core
interior region is transported at high velocities to the condensing section.
Evaporator
(heat source)
Heat in
Adiabatic
section
Condenser Liquid return flow
(heat sink)
Heat out
FIGURE 5.15
Schematic of a heat pipe.
238 Advanced Heat Transfer
The porous wick material has small, random interconnected channels that are constructed
along the inner wall of the container of the heat pipe. The pores in the wick act like a capillary
pump where the word “pump” is used because of its analogous role to regular pumping
action on fluids. It provides an effective means of transporting liquid back to the evaporator
through surface tension forces within the wick. Also, the wick serves as an effective separa-
tor between the vapor and liquid phases, thereby allowing more heat to be carried over
longer distances than regular pipes.
Heat pipes are also used in many types of applications, such as heat recovery systems,
microelectronics cooling, and spacecraft thermal control. For example, a series of heat
pipes in an air-to-air heat recovery system in a building allows effective storage of thermal
energy of exiting flue gases. Heat pipes can be up to 1,000 times more conductive than
metals at the same weight. Therefore laptops and other computers, as well as telecommu-
nications equipment, have adopted heat pipes for thermal management of electronic
assemblies. Also, heat pipes are used in satellites to transfer heat generated by electronic
equipment to radiation panels that dissipate heat into space. They provide effective con-
trol of temperatures required for reliable performance of the electrical components in the
satellite. Heat pipes have key advantages over other conventional thermal enhancement
devices, including low maintenance with no moving parts, passive operation and a
long lifespan.
Desirable characteristics of the working fluid include a high latent heat of vaporization,
high thermal conductivity, high surface tension, low dynamic viscosity, and suitable
saturation temperature. Also, the working fluid should effectively wet the wick material.
Examples of effective working fluids include water or ammonia for operation at moderate
temperatures, or liquid metals, such as sodium, lithium, or potassium, at high temperatures
(above 600◦ C). A heat pipe with water as the working fluid and a wick vessel material of
copper–nickel can provide a surface heat flux exceeding 146 W/cm2 at temperatures of
about 200◦ C. Typical heat flux values of common heat pipes, working fluids, and wick
vessels, are listed in Table 5.3.
An important design element is a fluid circulation criterion for the heat pipe. Proper
liquid circulation is maintained within the heat pipe as long as four forces are effectively
controlled—capillary forces; liquid pressure drop; vapor pressure drop; and the gravity
head. The driving pressure (capillary forces) within the wick must exceed the sum of fric-
tional pressure drops in the liquid and vapor and the gravitational head of the liquid in
the wick structure.
Capillary action occurs from surface tension forces in the wick. To understand this pro-
cess, consider the capillary rise in a tube that is partially submerged beneath a liquid surface.
An internal rise of the liquid level, due to surface tension, σ, occurs within the tube. It can be
shown that the pressure difference in the liquid due to this capillary rise can be determined
TABLE 5.3
Typical Heat Fluxes of Heat Pipes
Axial Flux Surface Flux
Range (K) Fluid Vessel (kW== cm2) (W== cm2)
230–400 Methanol Copper, nickel, stainless steel 0.45 at 373 K 75.5 at 373 K
673–1,073 Potassium Nickel, stainless steel 5.6 at 1,023 K 181 at 1,023 K
773–1,173 Sodium Nickel, stainless steel 9.3 at 1,123 K 224 at 1,033 K
Gas–Liquid Two-Phase Flows 239
where rp and θ refer to the radius of the capillary tube (corresponding to the pore radius in a
wick) and the angle subtended by the liquid rise along the capillary wall, respectively. A
value of θ = 0 yields the maximum capillary pressure, which corresponds to perfect wetting
within the tube (or wick). In an actual wick structure, liquid flow is generated by
this capillary action due to liquid entrainment within the wick. The spatial differences in
local capillary pressure differences (due to different curvatures of liquid menisci) induce
the capillary flow.
Secondly, a liquid pressure drop occurs when liquid flows through grooves in the wick
from the condenser back to the evaporator. Darcy’s law relates the mass flow rate, ṁ, through
a porous medium to the pressure drop, Δpl, as follows:
ρl Kw Aw Δpl
ṁ = (5.178)
μl Leff
where Kw, Aw and Dh refer to the wick permeability (wick factor), wick cross-sectional area,
and hydraulic diameter (four times the liquid flow area divided by the wetted perimeter),
respectively. Also, Leff refers to the length over which the pressure drop occurs, namely
the effective length between the condenser and evaporator sections (evaluated from the
midpoint of the condenser to the midpoint of the evaporator). Rearranging Darcy’s law,
ρl Vl2 Leff 64
Δpl = (5.179)
2 Dh ReD
Typical wick structures have a wrapped screen along the inner wall of the heat pipe or
screen-covered grooves. Typical wick parameters for common heat pipes are shown in
Table 5.4.
Thirdly, a vapor pressure drop also occurs since vapor drag in the core region may impede
liquid flow in the grooves of the wick at high vapor velocities. From Moody’s chart, an
expression for the vapor pressure drop can be determined. However, in practice, this
term is often a small component of the overall force balance since the vapor density is
much smaller than the liquid density. Also, vapor drag is often reduced by covering the
grooves in the wick structure with a screen.
Lastly, the gravity head can be positive (gravity assisted) or negative, although the latter
case defeats the purpose of the wick. A positive head implies that the evaporator is above the
condenser. The pressure term arising from this factor can be written as:
where φ refers to the inclination angle of the heat pipe with respect to the horizontal plane.
Assembling the four previous pressure terms, while neglecting the vapor pressure drop,
leads to the following equilibrium design condition:
2σ cos θ μl Leff ṁ
= + ρl gLeff sin ϕ (5.181)
rp ρl Kw Aw
240 Advanced Heat Transfer
TABLE 5.4
Typical Wick Parameters of Heat Pipes
Material Pore Radius (cm) Permeability (m2)
Setting θ = 0 on the left side yields the maximum capillary pressure. Operating beyond
this point in the design condition can dry out the wick (called a “burnout” condition) and
therefore the right side should remain below the maximum capillary pressure.
σρl hfg Aw Kw 2 ρl gLeff sin ϕ
qmax = ṁmax hfg = − (5.182)
μl Leff rp σ
The maximum mass flow rate can be determined based on given parameters and ther-
mophysical properties. The maximum flow rate is found as ṁmax = 7.11 × 10−6 kg/s. Also,
multiplying this result by the latent heat of vaporization, as indicated in the equation,
yields qmax = 16 W. The corresponding axial heat flux is obtained by dividing this value
by the heat pipe area (π × 0.0122 m2), yielding q′′ max ∼ 3.55 W/cm2. It is noted that the
heat transport capability of the heat pipe can be significantly increased by adding extra
layers of mesh screen within the wick material.
Gas–Liquid Two-Phase Flows 241
σρl hfg
M= (5.184)
μl
Another key design parameter is the heat transfer factor, HTF, defined as follows:
2MAw Kw
HTF = qmax Leff ϕ=0 = (5.185)
rp
This heat transfer factor in heat pipes can be improved by selecting wicks with large Kw
values or small rp values. In the intermediate range of operating temperatures (between
400 K and 700 K), water is one of the most suitable and widely used working fluids in
heat pipes.
In the entrainment limitation, the vapor velocity may become sufficiently high to produce
shear force effects on the liquid return flow from the condenser to the evaporator. In this
case, waves can be generated on the liquid surface and droplets may be entrained by the
1,000
1–2 : sonic limit Sodium
2–3 : entrainment limit
3–4 : wicking limit 100
4–5 : boiling limit
M (MW/cm2)
q″ (W/cm2)
3 4 10
Methanol
2 Water
5
1
1
0.1
T (k) 0 200 400 600 800 1,000
T (k)
FIGURE 5.16
(a) Heat pipe limitations and (b) figure of merit.
242 Advanced Heat Transfer
vapor flow as there would be inadequate restraining forces of liquid surface tension in the
wick. The relevant condition at the onset of entrainment is expressed in terms of the Weber
number (We). This dimensionless parameter is defined as the ratio of inertial effects in the
vapor to surface tension forces in the wick. The entrainment limitation is defined at the point
of onset of entrainment:
ρv Vv2 Lh
We = =1 (5.186)
σ
where Lh refers to the hydraulic diameter of the wick surface pores. The actual vapor veloc-
ity should remain lower than the above value in the entrainment limit. Otherwise, entrain-
ment of droplets into the vapor flow may cause starvation of liquid return flow from the
condenser (called dryout).
During startup from near-ambient conditions, a low vapor pressure within the heat pipe
can lead to a high resulting vapor velocity. In addition to the previous entrainment limit, if
the vapor velocity becomes choked (called a sonic limit), this condition limits the axial heat
flux in the heat pipe. The axial heat flux at the sonic limit is defined as:
ṁhfg
q′′axial = = ρv Vv hfg (5.187)
Av
The sonic limit, as well as the other limits, are often shown graphically in terms of the fluid
operating temperature (see Figure 5.16).
The heat flux limits generally increase with evaporator exit temperature due to the effect
of temperature on the speed of sound in the vapor. For example, for sodium, the heat flux
limit increases from 0.6 kW/cm2 at 500◦ C to 94.2 kW/cm2 at 900◦ C. For a working fluid of
potassium, the heat flux limit is 0.5 kW/cm2 at 400◦ C (evaporator exit temperature), and
it increases to 36.6 kW/cm2 at 700◦ C. For high-temperature applications, lithium can be
used since its heat flux limit ranges between 1.0 kW/cm2 at 800◦ C and 143.8 kW/cm2
at 1,300◦ C.
The previous three cases involved limitations on the axial heat flux in the direction of
the vapor flow in the heat pipe. A fourth case is a boiling limitation that involves the
radial heat flux through the container wall and wick. In particular, the onset of boiling
within the wick interferes with and obstructs the liquid return flow from the condenser.
Boiling within the wick may cause a burnout condition by drying out the evaporator con-
tainment. As a result, this situation places an additional limit on the design of the
heat pipe.
The four limitations are illustrated in Figure 5.16. Each limitation spans across a given
temperature range and the maximum heat flux occurs in the range of the wicking limit.
Additional features and analysis of heat pipes are provided in other sources including
Drolen (2017) and books by Dunn and Reay (1994) and Chi (1976).
well as outside the fin, were assumed constant. Two cases were considered: flush mounted
against the object; and inserted or extended into the cooled object (see Figure 5.17).
Consider a one-dimensional steady-state energy balance of a differential element of the
heat pipe,
∂2 T
kAw − ho Po (T − T1 ) − hi Pi (T − Tv ) = 0 (5.188)
∂x2
where k is thermal conductivity, Aw is the cross-sectional area, T is the wall temperature, ho is
the outside convection heat transfer coefficient, Po is the outside perimeter, T∞ is the sur-
rounding temperature, hi is the inside convection heat transfer coefficient, Pi is the perimeter
of the vapor space, and Tv is the interior vapor temperature. Radiation effects have been
neglected. The energy balance may be rewritten in the following nondimensional form,
∂2 θ
− Z2 θ = −N 2 θv (5.189)
∂X2
where
x T − T1
X= ; θ= (5.190)
L Tb − T1
ho Po L2 hi Pi L2
M2 = ; N2 = ; Z2 = M2 + N 2 (5.191)
kAw kAw
The length of the heat pipe is L and the base fin temperature is Tb. Solving this fin equation
yields:
N 2 θv cosh (Z(X − 1)) N 2 θv
θ(X) = 1 − 2 + 2 (5.192)
Z cosh (Z) Z
Using this result, the fin efficiency can be determined based on the rate of heat transfer
from the fin divided by the maximum heat transfer rate if the entire fin was maintained
at the fin’s base temperature,
"L 1
ho Po (T − T1 )
η= 0 = θ dX (5.193)
ho Po L(Tb − T1 ) 0
FIGURE 5.17
Heat pipe fin configurations.
244 Advanced Heat Transfer
Evaluating this efficiency for the case of a heat pipe that is flush mounted (subscript fm)
against the object yields:
tanh (Z)
ηfm = (5.194)
Z + (N 2 /Z2 )( tanh (Z) − Z)
Another approach to incorporate the heat pipe into the fin is to insert the heat pipe
evaporator into the object. This would increase the area for conduction to the evaporator.
Figure 5.17 illustrates the difference between the flush mounted and inserted configurations
of the heat pipe fins. For the case of the inserted heat pipe, the same above expression for the
fin efficiency is obtained except that the temperature distribution along the heat pipe will be
different. The insertion of the heat pipe will affect the temperature distribution of the heat fin
and vapor.
For X ≤ 0 (inserted region), assume that the heat pipe wall temperature is the same as the
base temperature (θ = 1). Then, for X . 0, the temperature remains the same as the previous
result for the flush mounted case because the wall temperature is assumed to be the base
temperature at X = 0. But the vapor temperature, θv, will be different. Applying an energy
balance in the vapor space,
L 1
hi Pi (T − Tv ) dx = (θ − θv ) dX = 0 (5.195)
−Le −Xe
where Xe refers to the length of the inserted section of the heat pipe fin. Substituting the
temperature distribution for θ, solving for the resulting vapor temperature, and substituting
the results into the prior expression for fin efficiency, yields the following result for the
inserted (subscript i) heat pipe fin:
tanh (Z) + Xe Z
θv = (5.196)
(N 2 /Z2 )(tanh (Z)
− Z) + (1 + Xe )Z
tanh(Z) N 2 tanh(Z) ηfm (Xe Z + tanh(Z))
η= + 2 1− (5.197)
Z Z Z ηfm Xe Z + tanh(Z)
From these results, it can be shown that heat pipe fins have a significantly higher efficiency
than standard fins. Bowman et al. (2000) showed that for comparable heat fins, in either
vertical or horizontal orientations, the efficiency of a typical case is 0.64 for a standard fin,
but 0.8 – 0.85 for a heat pipe fin. The insertion of the heat pipe into the object also signifi-
cantly improves the efficiency over a flush mounted heat pipe case. For example, with N =
30 and M = 3.0, the efficiency rises when the level of insertion increases: 0% inserted (0.77);
2% (0.84); 4% (0.88); 6% (0.90); and 8% (0.92). This increase arises as a result of the improved
heat transfer into the evaporator section of the heat pipe.
The above results can also be applied to the case of evaporation from the end of the vapor
space (not considered in the previous analysis). In this case, the inserted length is increased
by an amount equal to the area of the end of the vapor space divided by the interior perim-
eter of the heat pipe. This approach is analogous to increasing a standard fin’s length so as to
extend the results of an insulated fin tip result to the case of convection from the tip of the fin.
Gas–Liquid Two-Phase Flows 245
PROBLEMS
5.1 A rigid tank initially has a specified mass of liquid and vapor in equilibrium with a
relief valve at the top of the tank to maintain a constant pressure during phase
change. Consider a process of phase change either in view of condensation (heat
lost from a saturated vapor) or boiling (heat gained by a saturated liquid). Derive
an expression for the release of latent heat during the change of phase.
5.2 Consider a phase change process with boiling and bubble detachment from a heated
horizontal surface. Alternating periods of a liquid- and vapor-covered surface are
observed after the bubble detachment and during the bubble growth, respectively.
An important parameter is the fraction of time with a particular phase covering the
surface, for example, fv (fraction of time that a bubble covers the surface at a specific
position). If fv is known or measured, explain how the heat flux to the liquid can be
determined based on fv and solution methods of heat conduction developed in Chap-
ter 2. Neglect the effects of convective motion and the region of influence encompass-
ing the vortices and wakes of departing bubbles.
5.3 During a boiling process, bubbles grow outward from small indentations or cav-
ities along a heated horizontal wall and detach from the surface once their size
becomes sufficiently large. Beneath a bubble surrounding the cavity, a thin micro-
layer of liquid is vaporized gradually as the bubble expands. This process may be
approximated by one-dimensional downward movement of a gas–liquid interface
until the vapor front reaches the wall. Using a scaling analysis, estimate the tran-
sient variation of the heat flux by conduction from the wall to the liquid during
the microlayer evaporation. Express your answer in terms of qw (wall heat flux),
Tw (wall temperature), thermophysical properties, time, and the initial microlayer
thickness, δ0.
5.4 A bubble grows spherically outward along a heated wall during a boiling process.
The rate of phase change with time is assumed to be constant (the bubble radius
grows linearly with respect to time). The pressure within the bubble is denoted by
pv, whereas the pressure in the liquid surrounding the bubble is denoted by pl.
a. Use the one-dimensional continuity equation to determine the liquid velocity
surrounding the bubble (note: due to bubble expansion alone, not including
advection).
b. Perform an energy balance on an expanding bubble to find the pressure dif-
ference between the phase interface and the surrounding liquid over time.
c. Perform a force balance to estimate the pressure difference between the vapor
(within the bubble) and the phase interface over time.
5.5 Steam flows through a pipe (3.3 cm inner diameter, 3.8 cm outer diameter) with a
mean velocity, temperature, and convection coefficient of 2 m/s, 105◦ C, and 400 W/
m2 K, respectively. Heat losses to the surrounding air at 20◦ C lead to condensation
of steam along the inner pipe walls. The convection coefficient at the outer surface
is 15 W/m2 K. The thermal conductivity of the 1.3 cm thick layer of insulation around
the pipe is 0.03 W/mK.
a. Over what length of pipe will the rate of condensation reach 0.1 g/s?
b. What is the change of phase fraction of steam over this distance of pipe
length?
246 Advanced Heat Transfer
5.6 The wall of a tank containing liquid is heated electrically to provide a constant heat
flux during boiling of the liquid along the wall. Due to alternating periods of liquid
and vapor covering the heated wall, the temperature of the wall, Tw, varies with time.
Explain how the various stages of phase change along the wall contribute to this var-
iation of wall temperature with time. How would thermocapillary effects influence
this variation with time?
5.7 During phase change with boiling, a bubble grows spherically outward from a small
indentation along a heated wall. Two distinct regions of fluid motion are observed: (i)
viscous boundary layer in the liquid microlayer beneath the bubble; and (ii) motion
outside the boundary layer ahead of the outer bubble edge.
a. Give the governing equations and boundary conditions for laminar flow in
each of these regions.
b. Under suitable simplifying assumptions (state the assumptions), solve the
one-dimensional form of these equations to find the radial component of
velocity outside the boundary layer. Assume that the phase interface move-
ment is inversely proportional to the square root of time, V ≈ At −1/2 where A
depends on thermophysical properties and other factors.
5.8 Liquid methanol at its saturation temperature (65◦ C) is suddenly exposed to a heated
vertical wall at 75◦ C. Measurements are taken during boiling and bubble formation
during phase change, indicating: (i) an average microlayer thickness of 0.025 mm
beneath the bubble; and (ii) characteristic bubble diameter of 1 mm after 15 ms
from its onset of formation. Can this data provide sufficient information to determine
the relative significance of viscous forces acting on the liquid, associated with bubble
expansion? The following properties may be used for methanol: ρl = 751 kg/m3, ρv =
1.2 kg/m, and μl = 3.3 × 10−4 kg/ msec.
5.9 Surface roughness can have a substantial effect on various forces, including surface
tension, which affect the processes of phase change at a wall. How could the surface
roughness be included in the thermal analysis and characterized? Give an example
and physical explanation of how its effects could be included within the phase
change predictions.
5.10 Heat conduction from a wall through the liquid beneath a growing bubble affects the
phase change process during boiling. Consider the two-dimensional cross-sectional
area of the liquid region between a flat, smooth wall and an adjoining stationary bub-
ble. This region may be transformed to a planar channel region by the following
Mobius transformation:
z − z1 z2 − z3 w − w1
· =
z − z3 z2 − z1 w 2 − w1
In the conformal mapping, the complex variables are z = x + iy and w = u + iv, where
z1, z2, and z3 refer to points on two concentric circles (joined at a point) that are
mapped to corresponding points w1, w2, and w3 in the channel region.
a. Use this conformal mapping to find the temperature distribution beneath the
bubble at a given time. The wall can be represented by a very large radius of
the outer circle.
b. Based on the result in part (a), develop an expression for the wall heat flux.
Gas–Liquid Two-Phase Flows 247
5.11 In the previous problem, conformal mapping was used to transform the temperature
solution from a simplified geometry (planar channel) to a more complicated domain
(between the wall and a bubble). Assume that the liquid temperature ahead of the
phase interface is Tsat (saturation temperature) and the vapor temperature is linear
in the simplified domain. Express your answer in terms of Tw, Tsat, δ0 (initial micro-
layer thickness) and thermophysical properties.
a. Perform a heat balance at the phase interface to estimate the time required to
vaporize the liquid beneath the bubble entirely.
b. What value of time is obtained in part (a) for water heated by a wall at a tem-
perature of 20◦ C above Tsat? The initial microlayer thickness is 0.01 mm. The
vapor density and thermal conductivity are 0.6 kg/m3 and 0.025 W/mK,
respectively.
5.12 A liquid–vapor phase change process is encountered in a porous medium. Assume
one-dimensional, diffusion-dominated conditions. The liquid temperature is initially
at T0 (below the phase change temperature, Tsat). The wall temperature at x = 0 is sud-
denly raised to Ts (above Tsat), vapor is generated, and the phase interface moves into
the medium over time. Use a similarity solution to find the change of interface posi-
tion with time. Define a similarity variable of η = x/(4αt)1/2. Assume that the phase
interface moves according to δ = 2 γ (αt)1/2, where γ must be determined.
5.13 A balance of pressure and surface tension forces on a droplet can be used to find
the pressure difference across the phase interface. An elongated droplet of satu-
rated liquid with radii of curvature of r1 and r2 in the xz and yz planes, respec-
tively, is in thermal and mechanical equilibrium with the surrounding subcooled
vapor. Show that the pressure difference across the liquid–vapor interface can be
written as follows:
1 1
pv − psat =σ +
r1 r2
where σ, pv, and psat refer to the surface tension, pressure of subcooled vapor out-
side the droplet, and the pressure of saturated liquid, respectively. The term in
brackets on the right side of the equation is called the mean curvature of
nonspherical surfaces.
5.14 A liquid–vapor phase change process in microgravity conditions occurs sufficiently
slowly so that a one-dimensional linear profile of temperature can be assumed
throughout the vapor. Liquid (initially at the saturation temperature, Tsat) is suddenly
exposed to a heated wall at Tw, where Tw . Tsat. A planar vapor film is formed and it
propagates outward over time. Find the variation of vapor temperature and phase
interface position with time. Assume that the liquid ahead of the interface remains
at Tsat. Neglect convective motion in the analysis.
5.15 Repeat the previous problem while using an integral method to account for time-
dependent conduction in the vapor region. Derive the governing differential equa-
tion for the change of interface position with time and describe how this equation
can be used to find the wall heat flux. Assume a quadratic profile of temperature
in the vapor for the integral solution.
5.16 After a bubble departs from a heated wall during boiling, heat conduction into the
liquid yields a certain temperature profile which is present at the onset of the next
248 Advanced Heat Transfer
vapor period. Use the profile slope obtained from one-dimensional conduction into a
semi-infinite liquid domain as the initial condition for subsequent heat conduction
into the vapor (bubble). Solve the heat conduction equation in the vapor, subject to
a fixed wall temperature, Tw, to find the dependence of vapor temperature on x (posi-
tion), t (time), and thermophysical properties.
5.17 In a series of boiling experiments, a copper sphere with a diameter of 8 mm and
an initial temperature of 500◦ C is immersed in a water bath at atmospheric pres-
sure. The sphere cools along the boiling curve from point 1 (film boiling) to point 4
(nucleate boiling). Values at these points given as (temperature excess in K; heat
flux in W/m2): (1) 350, 105; (2) 120, 1.89 × 104 (minimum heat flux); (3) 30, 1.26 × 106
(critical heat flux); and (4) 5, 104. Use the lumped capacitance method to estimate the
time for the sphere to reach a temperature of 200◦ C in the transition boiling regime.
Assume that the boiling curve may be approximated by a set of piecewise curves with
a heat flux of cΔTn in each region (i.e., nucleate, transition, and film boiling regions),
where c and n are constants estimated from a curve fit. Use properties of copper at
standard temperature and pressure (STP) conditions.
5.18 Brass tubes are designed to boil saturated water at atmospheric pressure. The tubes
are operated at 80% of the critical heat flux. The diameter and length of each tube are
19 mm and 0.5 m, respectively. How many tubes and what surface temperature (each
tube) are required to provide a vapor production rate of 10 kg/min?
5.19 Pool boiling occurs along the outer surface of a copper pipe (2 cm outer diameter)
submerged in saturated water at atmospheric pressure. Up to what temperature
excess, ΔT (difference between the surface and saturation temperatures), will the
heat transfer remain within the nucleate boiling regime? Find the heat transfer coef-
ficient when ΔT = 12◦ C.
5.20 Heat transfer to water at atmospheric pressure occurs by nucleate boiling along a
heated steel surface. Estimate the amount of subcooling of a stationary liquid pool
to yield a maximum heat flux that is 20% greater than the maximum heat flux for a
saturated liquid pool. What heat flux is obtained under the subcooled conditions?
5.21 At copper bar at 700◦ C is immersed in a water bath at 60◦ C. The rate of heat transfer
from the bar is measured based on transient thermocouple measurements within the
bar. It may be assumed that the Biot number is small (Bi , 0.1) and the bath and liquid
free surfaces within the container do not appreciably affect the boiling process along
the surface of the bar.
a. Explain how the surface cooling rate varies with temperature during this
quenching process. Discuss the physical mechanisms that distinguish the dif-
ferent regimes of heat transfer and dominant modes of heat transfer in
each regime.
b. How would different initial water temperatures and different surface rough-
ness affect the results in part (a)?
c. How would agitation (fluid circulation) within the bath affect the maximum
cooling rate? Explain your response.
5.22 An electrical current passes through a polished 2.5 cm diameter conductor rod
immersed in saturated water at 2.64 MPa. Find the maximum rod temperature and
heat transfer coefficient if the heat input must not exceed 70% of the critical heat
flux during boiling. Determine the rate of vaporization over 0.8 m of conductor length.
Gas–Liquid Two-Phase Flows 249
5.23 After heat treatment in a furnace, a carbon steel bar is quenched in a water bath to
achieve specified surface hardness requirements. The cylindrical bar of 16 cm height
is removed from the furnace at 500◦ C and then submerged in a large water tank at
atmospheric pressure. If the surface emissivity of the bar is 0.85, what bar diameter
is required to produce a desired total heat transfer rate of 5 kW initially during
film boiling?
5.24 Forced convection with boiling occurs in an 8 cm heated vertical pipe. The water
temperature and mass flow rate are 430 K and 360 kg/h, respectively. If the quality
(vapor fraction) is 40% at a position where the wall temperature is 440 K, find the
rate of heat transfer to the two-phase flow at that location.
5.25 Laminar film condensation occurs within an industrial apparatus consisting of
small square ducts located above one another with baffles separating each duct.
The sides of the square duct of length L are inclined at 45◦ with respect to the hor-
izontal plane. A baffle is a plate that connects the base of a duct with the top cor-
ner of the duct below it. The condensate flows downward from the top duct to
lower ducts. The wall temperature of the condensing surface is 60◦ C. The space
adjacent to the wall is occupied by stagnant saturated steam at atmospheric
pressure.
a. Find the average Nusselt number in this geometry for N sets of ducts and baf-
fles. What Nusselt number is obtained for the case of N = 10 with L = 1 cm?
b. Find an expression for the ratio of convective heat transfer coefficients, hN/h1,
where hN and h1 refer to coefficients for N sets and one duct/baffle set, respec-
tively. Explain the functional dependence on N. In other words, explain why
the ratio increases (or decreases) with N.
5.26 A liquid film (representing condensate) flows steadily down along a flat plate
inclined at an angle of θ with respect to the horizontal plane. It is assumed that the
film thickness, δ, remains nearly uniform. Also, the film, wall, and air temperatures
are To everywhere upstream of x = 0 (note: x measured along the direction of the
plate). The surface temperature increases abruptly to Tn downstream of x = 0 and a
thermal boundary layer, δT, develops and grows in thickness within the film.
a. Start with the relevant film momentum equation and derive the following
result for the velocity distribution in the film:
u y y2
=2 −
U δ δ
where U refers to the velocity at the film–air interface. Assume that the termi-
nal film velocity is attained (i.e., a zero-velocity change in the x-direction
along the plate).
b. Assume that the temperature profile within the thermal boundary layer
can be approximated linearly, that is, T = a + by, where the coefficients a
and b are determined from appropriate boundary conditions. Perform an
integral analysis to estimate the location where δT reaches δ by integrating
the relevant energy equation and finding δT in terms of x, δ, U, and α.
c. Explain how the Nusselt number can be obtained from the results in part (b)
without finding the explicit closed-form solution for Nu.
250 Advanced Heat Transfer
5.27 A laminar film of condensate covers the outer surface of a horizontal tube (radius of
ro) in a heat exchanger. Suddenly, the wall temperature of the tube is slightly lowered
to Tw. Assume this temperature change occurs without generating any additional
condensate. Use the similarity method with a similarity variable of η = r 2/(4αt) to
solve the following heat conduction equation for the temperature in the condensate
layer (r . ro):
∂2 T 1 ∂T 1 ∂T
+ =
∂r2 r ∂r α ∂t
References
M. Altman, R.H. Norris, and F.W. Staub. 1960. “Local and Average Heat Transfer and Pressure Drop
for Refrigerants Evaporating in Horizontal Tubes,” ASME Journal of Heat Transfer, 82: 189–196.
K.O. Beatty and D.L. Katz. 1948. “Condensation of Vapors on the Outside of Finned Tubes,” Chemical
Engineering Progress, 44: 55–70.
A. Bejan. 2013. Heat Transfer, 4th Edition, New York: John Wiley & Sons.
P.J. Berenson. 1961. “Film Boiling Heat Transfer from a Horizontal Surface,” ASME Journal of Heat
Transfer, 83: 351–358.
W.J. Bowman, T.W. Moss, D. Maynes, and K.A. Paulson. 2000. “Efficiency of a Constant-Area, Adia-
batic Tip, Heat Pipe Fin,” AIAA Journal of Thermophysics and Heat Transfer, 14: 112–115.
L.A. Bromley. 1950. “Heat Transfer in Stable Film Boiling,” Chemical Engineering Progress, 46: 221–227.
V.P. Carey. 2007. Liquid–Vapor Phase Change Phenomena, 2nd Edition, Boca Raton: CRC Press/Taylor &
Francis.
J.C. Chen. 1963. “Correlation for Boiling Heat Transfer to Saturated Fluids in Convective Flow,” ASME
Paper 63-HT-34, 6th ASME-AIChE Heat Transfer Conference, Boston, MA.
M.M. Chen. 1961. “An Analytical Study of Laminar Film Condensation: Part 1 - Flat Plates,” ASME
Journal of Heat Transfer, 83: 48–54.
S.L. Chen, F.M. Gerner, and C.L. Tien. 1987. “General Film Condensation Correlations,” Experimental
Heat Transfer, 1: 93–107.
A.K. Chesters and G. Hofman. 1982. “Bubble Coalescence in Pure Liquids,” Applied Scientific Research,
38: 353–361.
S.W. Chi. 1976. Heat Pipe Theory and Practice, Washington, D.C.: Hemisphere.
Y.S. Chin, S.J. Ormiston, and H.M. Soliman. 1998. “A Two-Phase Boundary Layer Model for Laminar
Mixed Convection Condensation with a Noncondensable Gas on Inclined Plates,” Heat and Mass
Transfer, 34(4): 271–277.
J.G. Collier and J.R. Thorne. 1999. Convective Boiling and Condensation, 3rd Edition, Oxford: Oxford Uni-
versity Press.
K. Cornwell and S.D. Houston. 1994. “Nucleate Pool Boiling on Horizontal Tubes: A Convection Based
Correlation,” International Journal of Heat and Mass Transfer, 37: 303–309.
B.L. Drolen. 2017. “Performance Limits of Oscillating Heat Pipes: Theory and Validation,” AIAA Jour-
nal of Thermophysics and Heat Transfer, 31: 920–936.
M.R. Duignam, G. Greene, and T. Irvine. 1991. “Film Boiling Heat Transfer to Large Superheats from a
Horizontal Flat Surface,” ASME Journal of Heat Transfer, 113: 266–268.
P.D. Dunn and D.A. Reay: 1994. Heat Pipes, 4th Edition, New York: Pergamon.
T. Fujii, H. Honda, S. Nozu, and S. Kawakaml. 1977. “Condensation of Superheated Vapor inside a
Horizontal Tube,” Refrigeration, 52: 553–575.
G. Hetsroni, Ed., 1982. Handbook of Multiphase Systems, New York: McGraw-Hill.
H.P Hu and C.K. Chen. 2005. “Turbulent Film Condensation on an Isothermal Sphere,” AIAA Journal of
Thermophysics and Heat Transfer, 19: 81–86.
H. Jaster and P.G. Kosky. 1976. “Condensation Heat Transfer in a Mixed Flow Region,” International
Journal of Heat and Mass Transfer, 19: 95–99.
O.A. Jianu, M.A. Rosen, G.F. Naterer, and Z. Wang. 2015. “Two-Phase Bubble Flow and Convective
Mass Transfer in Water Splitting Processes,” International Journal of Hydrogen Energy, 40: 4047–
4055.
S.G. Kandlikar. 1990. “General Correlation for Saturated Two-Phase Flow Boiling Heat Transfer Inside
Horizontal and Vertical Tubes,” ASME Journal of Heat Transfer, 112: 219–228.
D.Q. Kern. 1958. “Mathematical Development of Tube Loading in Horizontal Condensers,” AIChE
Journal, 4: 157–160.
S.S. Kutateladze. 1948. “On the Transition to Film Boiling under Natural Convection,” Kotloturbostroe-
nie, 3: 10–12.
S.S. Kutateladze. 1963. Fundamentals of Heat Transfer, New York: Academic Press.
Gas–Liquid Two-Phase Flows 253
D. Labunstov. 1957. “Heat Transfer in Film Condensation of Pure Steam on Vertical Surfaces and Hor-
izontal Tubes,” Teploeneroetika, 4: 72–80.
J.H. Lienhard, V.H. Dhir, and D.M. Riherd. 1973. “Peak Pool Boiling Heat Flux Measurements on Hor-
izontal Finite Horizontal Flat Plates,” ASME Journal of Heat Transfer, 95: 477–482.
J.M. Mandhane, C.A. Gregory, and K. Aziz. 1974. “Flow Pattern Map for Gas–Liquid Flow in Horizon-
tal Pipes,” International Journal of Multiphase Flow, 1(4): 537–554.
Z.L. Miropolski. 1963. “Heat Transfer in Film Boiling of a Steam-Water Mixture in Steam Generating
Tubes,” Teploenergetika, 10: 49–52.
G.F. Naterer, W. Hendradjit, K. Ahn, and J.E.S. Venart. 1998. “Near-wall Microlayer Evaporation
Analysis and Experimental Study of Nucleate Pool Boiling on Inclined Surfaces,” ASME Journal
of Heat Transfer, 120: 641–653.
M.S. Plesset and S.A. Zwick. 1954. “Growth of Vapor Bubbles in Superheated Liquids,” Journal of
Applied Physics, 25: 693–700.
W.M. Rohsenow. 1952. “Method for Correlating Heat Transfer Data for Surface Boiling of Liquids,”
Transactions of ASME, 74: 969–976.
W.M. Rohsenow. 1956. “Heat Transfer and Temperature Distribution in Laminar Film Condensation,”
Transactions of ASME, 78: 1645–1648.
W.M. Rohsenow and P. Griffith. 1955. “Correlation of Maximum Heat Flux Data for Boiling of Satu-
rated Liquids,” AIChE-ASME Heat Transfer Symposium, Louisville, KY.
J.W. Rose. 1988. “Fundamentals of Condensation Heat Transfer: Laminar Film Condensation,” JSME
International Journal, Series II, 31: 357–375.
P. Saha and N. Zuber. 1974. “Point of Net Vapor Generation and Vapor Void Fraction in Subcooled
Boiling,” 5th International Heat Transfer Conference, 4: 175–179.
I.G. Shekriladze and V.I. Gomelauri. 1966. “Theoretical Study of Laminar Film Condensation of Flow-
ing Vapor,” International Journal of Heat and Mass Transfer, 9: 581–591.
T. Takeyama and S. Shimizu. 1974. “On the Transition of Dropwise-Film Condensation,” Proceedings
of the 5th International Heat Transfer Conference, Tokyo, 3: 274.
L.S. Tong and Y.S. Tang. 1997. Boiling Heat Transfer and Two-Phase Flow, 2nd Edition, Boca Raton: CRC
Press/Taylor & Francis.
D.P. Travis and W.M. Rohsenow. 1973. “Flow Regimes in Horizontal Two-Phase Flow with Conden-
sation,” ASHRAE Transactions, 7: 31–34.
P.B. Whalley. 1987. Boiling, Condensation and Gas–Liquid Flow, Oxford: Clarendon Press.
R. Vachon, G. Nix, and G. Tanger. 1968. “Evaluation of Constants for the Rohsenow Pool-Boiling
Correlation,” ASME Journal of Heat Transfer, 90: 239–247.
E.H. Young and D.J. Ward. 1957. “How to Design Finned Tube and Shell-and-Tube Heat Exchangers,”
The Refining Engineer, 29: 32–36.
N. Zuber. 1958. “On the Stability of Boiling Heat Transfer,” Transactions of ASME, 80: 711–720.
6
Multiphase Flows with Droplets and Particles
6.1 Introduction
Multiphase flows with droplets and particles arise in many engineering systems such
as combustion of coal particles, spray coatings, pollutant dispersion in the atmosphere,
aerosol deposition in spray medication, fluidization in combustion systems, among
many others. These systems involve the transport of suspended droplets or particles in
a gas flow stream. Often heat transfer by convection, radiation, and/or phase change
has a significant role. This chapter will examine the physical processes, solution methods,
and applications related to heat transfer in multiphase flows with droplets and particles.
Throughout the chapter, references are made to droplets or particles. Due to significant
similarities, in practice, the analysis of one can often be applied to the other, particu-
larly for small particles or droplets where the lumped capacitance approximation can
be used.
The number and distribution of particles or droplets in the continuous (or carrier; gas)
stream affect the flow patterns. In a dilute flow (or dispersed flow), the dynamics of heat
and fluid flow are governed primarily by the carrier stream since generally the particles
are very small and occur in low concentrations. The particle motion is controlled by drag
and lift forces on the particle. Individual particles typically avoid significant particle–parti-
cle interactions since there is a relatively large distance between particles. Thus particle–gas
interactions are the dominant transport processes in dilute flows.
Particle–particle collisions become a significant transport process among particles or
droplets in dense flows. Dense flows can be further subdivided into collision-dominated flows
or contact-dominated flows. In collision-dominated flows, the particles collide in pairs and
then move to the next collision. This process leads to a formation of clusters of particle
clouds, such as particle cloud formations in fluidized bed reactors. The detailed modeling
of all collisions is not necessary to resolve overall key macroscopic features of the flow
stream. However, in contact-dominated flows, details of the collisions are needed to accu-
rately predict the main features of the flow patterns. Here a particle comes into contact
with many other particles simultaneously. This type of flow arises in applications such as
horizontal pipe flows with dense packing of particles.
The coupling between the particle and gas streams can be either a one-way coupling or a
two-way coupling. In one-way coupling, the gas stream affects the temperature and motion
of the particles, but a particle does not significantly affect the velocity or temperature of
the gas. On the other hand, an interdependent interaction between the gas and solid phases
occurs in two-way coupled flows. Although all flows are two-way coupled to some extent,
the effects of particle dynamics on the gas phase become negligible when the particle con-
centration is sufficiently low. This critical concentration of transition between one-way
255
256 Advanced Heat Transfer
and two-way coupling depends on various factors, such as the particle size, turbulence in
the gas stream, and other factors.
In pneumatic transport, a gas stream is used to transport solid particles through a pipe-
line, such as pulverized coal particles, grain, flour, and plastic. There are two modes of
pneumatic transport involving either a dilute phase (particles are fully suspended with sol-
ids less than 1% by volume) or a dense phase (particles not suspended at more than 30% by
volume). A transition phase occurs between the two modes. In vertical pneumatic trans-
port, the gas velocity must be sufficiently larger than the settling velocity of the solid par-
ticles to retain the particle transport. In horizontal pneumatic transport, there are four
distinct flow regimes (see Figure 6.1): (i) homogeneous flow; (ii) dune flow; (iii) slug flow;
and (iv) packed-bed flow. The transition between these regimes depends on the gas velocity
and particle concentration.
In homogeneous flow in horizontal tubes, the gas velocity is high enough to maintain the
solid particles in suspension. As the gas velocity decreases or the particle concentration
increases, some particles begin to settle along the bottom of the tube (similar to sand
dunes). This flow pattern is called dune flow. The velocity when particles begin to settle
is called the saltation velocity. A slug flow occurs with further reduction in the gas velocity
since the flow pattern then resembles alternating regions between suspended particles and
slugs, similarly to slugs arising in vapor–liquid flows with forced convection boiling. A
packed bed flow occurs at low gas velocities where the particles behave similarly to a
porous medium and the gas moves through interstitial regions between the particles.
The gas flow may also produce a slow net movement of the packed bed along the bottom
of the tube.
Two main groups of models have been developed for gas–solid flows (Crowe 1999;
Hewitt, Shires, and Polezhaev 1997): particle trajectory and two-fluid models. In particle tra-
jectory models, momentum and heat equations are solved for both the particle velocity and
temperature fields. Also, the energy and momentum equations are solved for the carrier
phase. A two-way coupling is established with the particle trajectories. Although the trajec-
tory model is effective for dilute flows, a two-fluid model is more suitable for dense flows. In
the two-fluid method, the particulate phase is treated as a second fluid with corresponding
mass-weighted thermophysical properties. Cross-phase interactions in the momentum and
energy equations are included in the coupling between the phases. In both the particle tra-
jectory and two-fluid models, a lumped capacitance approximation is often used for a thermal
analysis of particles which are small enough to be assumed isothermal.
FIGURE 6.1
Gas–solid flow regimes.
Multiphase Flows with Droplets and Particles 257
This chapter will examine the physical processes, governing equations, and solution
methods for multiphase flows with droplets and particles. Topics include the formulation
of dispersed and carrier phase equations, gas–particle interactions, packed bed flow in
tubes, and external flows with droplets. Also, the topics of droplet evaporation, particle for-
mation, melting of particles, radiation exchange involving particles, and liquid–particle and
slurry flows will be presented and discussed.
dup 1 πD2p
m = ρ g cd (ug − up )|ug − up | + mg (6.1)
dt 2 4
where ug and up refer to the gas and particle velocities, respectively. Assume a spherical par-
ticle of mass m. Then the equation of motion becomes:
dup cd 18μ
= (ug − up ) + g (6.2)
dt 24/Rep ρp D2p
ρ g u p Dp
Rep = (6.3)
μ
The factor 24/Rep in Equation 6.2 is called the Stokes drag since it arises from the solution
of the Navier–Stokes equations for laminar flow around a sphere at low Reynolds numbers,
Rep , 1 (called Stokes flow).
The reciprocal of the second factor in brackets in Equation 6.2, ρp D2p /(18μ), has dimensions
of time. As a result, this term is often interpreted as a particle velocity response time, τv, or a
measure of the time of response of a particle to a change in the gas velocity. This interpre-
tation is analogous to the thermal response time (Chapter 2), when a particle is suddenly
subjected to an environment at a different temperature. At times much less than the
response time, the body remains largely near its initial temperature, whereas when the
actual time far exceeds the characteristic response time, the particle approaches the temper-
ature of the surroundings. In between these limits, the particle’s temperature is in transition
between its initial temperature and the surrounding environment temperature. Response
258 Advanced Heat Transfer
times, such as the velocity and thermal response times, are effective ways of characterizing
the various transport mechanisms associated with particle interactions with the surround-
ing gas stream.
The density of particle packing in the gas stream affects the resulting pressure loss within
the tube. At high velocities in the homogeneous flow regime, the pressure drop varies nearly
with the square of the gas velocity, similarly to single phase flows. The pressure drop
increases with particle concentration because particles lose momentum upon contact with
the wall. Performing an integral balance of the momentum equation over a section of
tube of length Δx, it can be shown that the pressure drop, Δp, over this section can be
expressed as:
Δp ρ′ d f
≈ (ug − up ) (6.4)
Δx τv
where ρ′p is the apparent or bulk particle density (mass of particles per unit volume). The
pressure loss increases at a higher particle concentration and relative velocity between the
particle and gas stream. The increase of pressure loss with particle concentration demon-
strates the two-coupling effect between particles and the gas stream.
The Stokes law represents a solution of the Navier–Stokes equations to determine the force
on a stationary sphere in a slowly moving fluid stream. This configuration is equivalent to a
spherical particle moving at a relative velocity (ug − up) with respect to the gas stream.
At higher particle Reynolds numbers, the Stokes law accuracy decreases and alternative
correlations for the drag coefficient are needed. The following particle drag force, Fd, and
Schiller–Naumann correlation can be used at higher Reynolds numbers up to Rep = 800:
π
Fd = ρg u2p D2p cd (6.6)
8
24
cd = 1 + 0.15Re0.687
p (6.7)
Rep
Here, the drag force is proportional to the square of velocity and the drag coefficient
decreases with particle Reynolds number.
For very small particles, a correction factor, C, is used to modify the Stokes law when the
particle diameter becomes comparable to the mean free path of the gas molecule. This prox-
imity of scales leads to “slip effects” where microscopic momentum exchange occurs due to
interactions between the molecular and particle motions. The drag force, Fp, on a particle is
modified through a correction factor in the following manner:
3π
Fp = μup Dp (6.8)
C
Multiphase Flows with Droplets and Particles 259
where Kn = λ/Dp and λ is the mean free path of the gas molecules. Extensions of this
correlation to account for nonspherical particles were formulated by Clift, Grace, and
Weber (1978).
In some applications, such as electrostatic precipitators, cohesive forces between particles
(electrostatic and van der Waals forces) have significant effects on the gas–particle interac-
tions. Particles may exhibit an attractive positive–negative charge combination for grouping
and collection. Electrostatic forces among particles arise from a surplus or deficit of elec-
trons, thereby leading to a charged particle. For example, fly ash particles from coal combus-
tion processes are cohesive and charged particles. Van der Waals forces arise from the
attraction between dipoles at the atomic and molecular levels. These forces act over much
shorter ranges, at atomic and molecular scales, than electrostatic forces. Electrostatic forces
are more effective in separating particles from gases since they act across a larger spatial
range than van der Waals forces. The Hamaker constant (A) characterizes the particle–par-
ticle interactions as a result of van der Waals cohesion.
If a liquid film resides on the surface of a particle or joined particles in a gas stream, then
capillary forces also arise. Capillary forces are typically several orders of magnitude larger
than the van der Waals forces. Surface tension in the film between the two coalesced parti-
cles creates an attractive force. The capillary force between two particles due to a surface liq-
uid film can be represented by (Hewitt et al. 1997):
πDp σ
Fc = (6.10)
1 + tan(ϕ/2)
where φ refers to the angle subtended by the line joining the particle midpoints and the edge
of the film. Also, σ refers to the surface tension at the gas–solid interface. The cohesive force
increases to a maximum value of πDpσ when φ approaches 0◦ .
Separation of particles from a gas stream occurs in many engineering systems, for exam-
ple, sawmills, removal of particulates from smokestacks, oil refineries, cement plants, and so
on. Furnaces use precipitators to remove dust, dirt, and other particles from the air prior to
ventilation of return air through a house or building. Two common devices are used for
gas–solid separation: the cyclone separator and the electrostatic precipitator. In a cyclone sepa-
rator, centrifugal acceleration separates solid particles from the gas stream. In electrostatic
precipitators, particles are charged by Coulomb forces and later removed from the gas
stream by a collecting surface that attracts the charged particles. The gas–solid flow passes
through a region enclosed by an array of vertically suspended and charged metal plates.
High-voltage wires between the plates produce an electric field between the walls and wires.
The resulting magnetic effects on each particle act as forces driving particles toward the
collector plates.
The collector efficiency, η, of an electrostatic precipitator is defined as the ratio of
particles entering to particles collected by the precipitator. It can be expressed by the
260 Advanced Heat Transfer
where ud, ug, A and k refer to the deflection velocity (or drift velocity) of particles toward the
wall, gas velocity, plate surface area and a collector coefficient, respectively. The collector
efficiency decreases exponentially with deflection velocity.
where the , . brackets denote spatial averaging over the volume of phase k, denoted as Vk.
Also, define the mass and volume fractions, χk and ζk, respectively, as the respective frac-
tions of mass, m, and volume V, of phase k within a multiphase control volume,
mk
χk = (6.13)
m
Vk
ζk = (6.14)
V
This volume fraction can be written in terms of the mass fraction after multiplying by
the density.
Consider the conservation of the general scalar quantity, φk, in a multiphase mixture
within a two-dimensional control volume. For example, consider two phases occupying a
differential control volume in Figure 6.2. For multiphase flows with droplets and particles,
the two phases represent the carrier phase (gas; k = 1) and the dispersed phase (droplets or
particles; k = 2).
Using a similar procedure as presented earlier in Chapter 2 to derive the governing con-
servation equations for mass, momentum, and energy, it can be shown that a conservation
balance for the general scalar quantity, φk, in tensor form, leads to:
∂ ∂ ∂
(ρ ϕ ) + (ρ ϕ uk,j ) = − (jk,j ) + Sk (6.15)
∂t k k ∂xj k k ∂xj
where j = 1, 2, 3 and jk,j and Sk refer to the diffusion flux of φk in the xj direction and
source term for phase k, respectively. From left to right in Equation 6.15, the terms represent
the transient accumulation of φk in the control volume occupied by phase k, net advection
Multiphase Flows with Droplets and Particles 261
dx
Gas
n1
n2
Droplet nx
FIGURE 6.2
Multiphase control volume.
of φ across the phase k portion of the control surface, diffusion into phase k across the inter-
facial surface, and the source or sink of φk in the phase k portion of the control volume,
respectively.
Using the Leibnitz and Gauss rules of calculus, Equation 6.15 can be volume averaged by
integration of the governing equation over Vk, leading to:
∂ ∂ 1
(ζ ρ ϕ ) + (ζ (ρ ϕ uk,j + jk,j )) = − ṁ′′k ϕk + jk,m nk,m dS + ζ k (Sk ) (6.16)
∂t k k k ∂xj k k k V S
where the tensor subscripts are j = 1, 2, 3, m = 1, 2, 3 and nk,m refers to the component of the
unit outward normal vector to phase k in the m coordinate direction. In Equation 6.16, the ṁ′′k
term refers to the interphase mass flux (units of mass flow per unit time and area). This term
is generally dropped upon evaluation along the walls due to the no-slip condition of velocity
along the wall. The integration on the right side of Equation 6.16 refers to integration over
the total surface area, including the interfacial area, per unit volume (Si along the boundary
separating two distinct phases within the control volume), and any area of phase k in contact
with the external walls, Sw, or boundaries of the system.
∂ ∂ ∂
ζ kρ l + ζ k kρk uk l + ζ k kρk vk l + kṁ′′k lint = 0 (6.17)
∂t k k ∂x ∂y
where the subscript int refers to evaluation along the interfacial area of the control vol-
ume. The last term is the interfacial mass flux, such as phase change at the interface
which leads to accumulation or destruction of mass in one phase at the expense of
the other. For example, evaporation of droplets will increase the volume fraction in
the vapor phase at the expense of liquid of the droplet. In Equation 6.17, uk with k = 2
refers to the droplet(s) velocity in the x-direction. Since the governing differential equa-
tion was spatially averaged over the control volume, this velocity may involve more
than a single droplet.
262 Advanced Heat Transfer
Alternatively, Equation 6.17 can be written in terms of the mass of liquid droplets per unit
volume of a gas–liquid mixture, called the liquid water content, as follows,
ml
ρ̃l = = ζl kρl l (6.18)
V
where the subscript l refers to liquid. This definition is analogous to the concept of specific
humidity, except that it includes discrete droplets rather than a continuous mixture of mois-
ture in the air.
Using the above liquid water content in Equation 6.17, the mass conservation equation
becomes:
The last term on the left side represents droplet coalescence and/or evaporation processes.
This equation is similar to the continuity equation for single phase flow and therefore can be
interpreted as a form of conservation of the “concentration” of droplets (or phase fraction) in
the flow field. It is solved to determine the volume fraction of droplets, ζk, occupying the
droplet–air control volume.
∂
∂
ζk kρk uk,i l + ζ k kρk uk,i uk,j l + ρk kτk,ij nj l
∂t ∂xj
∂
1
′′
=− ζk kpk l + ṁk uk,i − τk,im nm dS + ζ k kFk,i lint (6.20)
∂xi V S
where the subscripts g and d refer to gas and droplet, respectively, and ξ is an empirical
coefficient which depends on flow conditions and thermophysical properties. From Equa-
tion 6.2, a suitable form of the interfacial force due to cross-phase interactions between
the dispersed phase (droplets) and the gas stream can be written as:
cd 18μg
Fk,int = (ug − ud ) (6.22)
24/Red D2d
Multiphase Flows with Droplets and Particles 263
where Red and Dd are the droplet Reynolds number (based on the relative velocity) and
mean droplet diameter in the flow field. Assuming a spherical droplet, the drag coefficient
can be related to the droplet Reynolds number as follows,
24 4.73
cd = + + 0.00624Re0.38
d (6.23)
Red Re0.37
d
∂ ∂ ∂ ∂τij ∂
ζg kρg ug,i l + ζg kρk ug,i ug,j l = ζ g kpl + ζ g + ζg kFg,i lint − ρg ζ g ku′ i u′j l
∂t ∂xj ∂xi ∂xj ∂xj
(6.24)
where i = 1, 2, 3, j = 1, 2, 3 and u′i refers to the turbulent fluctuating deviation from the vol-
ume-averaged mean velocity component in the xi direction. The cross-phase interaction
term, Fg,i, for the xi-direction momentum equation represents the net averaged force on
the carrier (gas) flow due to interactions with the droplets (or particles). The last term in
Equation 6.24 is the averaged Reynolds turbulent stress in the gas phase. The presence of
droplets (or particles) in the gas stream alters the length scales of turbulence due to gas-
droplet interactions.
Performing a similar procedure of substituting the droplet phase (subscript d ) for the
phase subscript, k, in Equation 6.20, leads to:
∂
∂
∂
∂τij ∂
ζd kρd ud,i l + ζd kρd ud,i ud,j l = ζ d kpd l + ζd + ζ d kFd,i lint − ρd ζ u′ d,i,n u′ d,j,n
∂t ∂xj ∂xi ∂xj ∂xj n d,n
(6.25)
where j = 1, 2, 3 and i = 1, 2, 3. The pressure, droplet velocities and shear stresses are mass-
averaged quantities. The last term in Equation 6.25 represents the turbulent Reynolds
stresses due to the collective droplet (or particle) interactions with the gas stream. This
model has assumed dilute flow in the dispersed phase, where interactions among droplets
are negligible and turbulent stresses arise mainly from turbulence in the carrier phase. The
last term in Equation 6.25 can be formulated in terms of a turbulent eddy viscosity including
an additional near-wall formulation to account for momentum exchange due to particle–
wall contact.
Alternatively, rather than an eddy viscosity in the carrier phase, constitutive relations
based on kinetic theory can be used instead for turbulent gas–particle flows. In this
approach, an additional equation is introduced for the kinetic energy associated with the
turbulent fluctuations of droplet velocities. This resulting velocity is related to the bulk vis-
cosity of the gas/droplet mixture. The model is primarily applicable to vertical gas–particle
flows with negligible particle–particle and particle–wall interactions. Further details are
provided by Gidaspow (1994). A scale analysis of transient gas–solid flows with particles
was presented by Martin (2011).
264 Advanced Heat Transfer
FD μV
= k1 + k2 ρV 2 (6.26)
As rh
where k1 and k2 are empirically determined constants, As is the total surface area of
particles, and rh is the average hydraulic radius of particles. The bulk velocity, V, can be
approximated by the ratio of the superficial gas velocity, Vs, to the void fraction, ζv, of the
packed bed:
Vs
V= (6.27)
ζv
The area is represented by the product of the number of particles, Np, and the surface area
of a single particle, sp,
A s = N p sp (6.28)
The number of particles can be calculated from the ratio of the total volume of solids to the
volume of a single particle, Vp, as follows:
S0 L(1 − ζv )
Np = (6.29)
Vp
Multiphase Flows with Droplets and Particles 265
where S0 is the cross-sectional area of the empty bed and L is the bed depth. Also, the
hydraulic radius can be determined from:
ζ v Vp ζ v Dp Φ
rh = = (6.30)
(1 − ζv )sp 6(1 − ζ v )
150
fp,E = + 1.75 (6.31)
ΦRep
which is valid for 1 , Red , 1,000, where the Reynolds number for a packed bed, Red, is
defined by:
Dp V s
Red = (6.32)
ν(1 − ζ v )
At low Reynolds numbers below 1 (Stokes flow; or “creeping flow”), viscous effects are
dominant and the Navier–Stokes equations can be solved to obtain the following Stokes law:
FD /As 6μζ v
cd = = (6.34)
ρV 2 /2 ρΦDp Vs
Rewriting this result in terms of the friction factor for Stokes flow (subscript S),
24μζ v
fp,S = 4cd = (6.35)
ρΦDp Vs
A transition region (1 , Rep , 20) occurs where the pressure drop and friction factor
change between the Ergun and Stokes flow formulations. A composite relation can be
used to create a smooth transition with correct asymptotic trends at both low and high
Reynolds number limits. The friction factor of the transition region (fp,T) is correlated by:
m m
Rel,T Rep
fp,T = fp,S + fp,E (6.36)
Rep Reu,T
where Rel,T and Reu,T are lower and upper limits of the transition Reynolds numbers, and m
is an empirical coefficient. The composite solution (subscript C) between the Ergun and
266 Advanced Heat Transfer
where n is an empirical coefficient. Values of n between 1 and 5, and m between 1.3 and 5,
provide good agreement with experimental data over a range of particle diameters (Pope
et al. 2011). This composite solution can be used to predict the friction factor over a range
of packed bed flow conditions, including the transition region between low and high
Reynolds numbers.
where NuD is the Nusselt number of flow past a sphere (Chapter 3) and Tg and Tp refer to the
gas and particle temperatures, respectively.
Although the gas velocity does not explicitly appear in Equation 6.38, an inter-equation
coupling between velocity and temperature exists through the Nusselt number. The convec-
tive heat transfer coefficient depends on the relative velocity between the particles and gas
flow. Also, the physical properties in the Reynolds number are temperature dependent. An
iterative procedure may be required to resolve inter-equation couplings between velocity
and temperature. Assuming a one-way coupling between the gas and particle phases, the
temperature-dependent property variations in the Reynolds number may be neglected.
Dense packing of solid particles in a packed bed can strongly affect the convective
heat transfer between the gas flow and particles. Packed bed convection correlations are
typically based on a superficial fluid velocity, Vs, which would exist if the packed bed
was empty (without particles). Whitaker (1972) recommended the following correlation
for convective heat transfer in packed-bed flows:
hDp 1 − ζ
NuD = = v
0.5Re1/2 2/3
D + 0.2ReD Pr1/3 (6.39)
k ζv
Here the void fraction, ζv, is the fraction of the packed bed that is empty (occupied by gas).
The equivalent diameter of the packed particles, Dp, is defined as six times the volume of a
particle divided by its surface area. For spherical particles, this relationship reduces to the
diameter of the particle. Whitaker’s correlation is accurate to within +25% for the above
indicated range of flow parameters.
Multiphase Flows with Droplets and Particles 267
Consider external flow with droplets in a boundary layer over an airfoil. The functional
form of Hilpert’s correlation can be extended to this configuration but with different
empirical coefficients, c and m, determined experimentally. At a Reynolds number of Rec ≥
6 × 105, transition to turbulence occurs in the boundary layer for an airfoil. The transition
point occurs at a smaller Reynolds number, ReD = 2 × 105, for a cylinder in cross flow. It
is expected that the liquid water content has a significant effect on the heat transfer coeffi-
cient. Wang, Naterer, and Bibeau (2008) reported a number of heat transfer correlations
for a NACA airfoil in cross flow at different angles of attack between 0◦ and 25◦ , with
and without droplets in the air stream. For the case without droplets at Re ≤ 6 × 105,
For Re . 6 × 105,
For the case of impinging droplets on the airfoil, a multiphase Reynolds parameter, Rem =
Re(1 + w), is defined, where w is the non-dimensional liquid water content. The Nusselt
number correlation is expressed as:
W
w= (6.45)
W0
where W0 is a reference value of the liquid water content, determined empirically in a similar
manner as the other coefficients (c and m). The resulting Nusselt number correlation can then
be used for both single phase (without droplets) and two-phase flows (with droplets), pro-
vided the Reynolds number is replaced by the multiphase Reynolds number.
For convective heat transfer from a NACA airfoil with impinging droplets at Re(1 + w) ≤
6 × 105,
Using the multiphase Reynolds number, a consistent trend can be obtained which col-
lapses data onto a single curve for a range of cases with and without droplets in the airstream.
The definition of Re(1 + w) provides a useful normalization of measured data and incorpo-
rates the effects of both the single-phase Reynolds number and the liquid water content.
External flows with droplets occur in a range of other applications, such as impinging
droplets on airfoils (icing of aircraft and wind turbines), spray coatings, manufacturing pro-
cesses of melt particularization, and powder production from the injection of a melted metal
stream and molten droplets into a gas stream.
In the melt particularization process, an injected molten stream is disintegrated by fluid
interaction with a liquid or gas. Initially, breakup of the injected molten stream into droplets.
The droplet dynamics can be analyzed by a momentum and heat balance involving surface
tension along the droplet and gravity (see Figure 6.3). The disintegrated droplets come into
contact with a cross-stream fluid to initiate secondary breakup into smaller droplets. The
processes of particle formation include liquid evaporation, droplet coalescence, and lastly
solidification of the droplets.
The droplet flow processes are significantly affected by changes of the Weber number,
WeD (ratio of droplet inertia to surface tension), and Reynolds number, ReD. At high ReD
and WeD numbers, the droplets are atomized and broken into distinct, separate droplets.
Destabilization and disintegration into wispy droplet streams occurs at lower values. The
rate of heat transfer from the solidifying droplets affects their structural properties following
impact on the surface. The following Mehrotra correlation (Mehrotra 1981; Sahm and Hansen
1984) can be used for the melt particularization processes,
hmp kg
NuD = = 2 + 0.6Re1/2
D Pr
1/3
(6.48)
Dd
where Dd is the droplet diameter and kg is the thermal conductivity of the gas stream. Con-
vective heat transfer from the droplets is a key factor affecting the solidification process and
Metal stream
We
Fluid (ρV2D/σ)
100
10
Primary
breakup
Destabilize
Secondary Distintegrate
breakup Atomize
Powder/ Re
Solidification particles (ρVD/μ)
104 105
FIGURE 6.3
Schematic of melt particularization.
Multiphase Flows with Droplets and Particles 269
∂B
ρi = ηc FWV (6.49)
∂t
where ρi, B, ηc, F, W and V, refer to the ice density, ice thickness, collection efficiency, view
factor, liquid water content, and freestream air velocity, respectively. The mass balance
states that the rate of increase of ice on the surface balances the net rate of impinging droplets
which freeze upon impact (rime ice).
The collection efficiency, ηc, is the ratio of mass flux of impinging droplets on the curved sur-
face to the impinging mass flux that would occur if the droplets were not deflected by
the airstream. Also, the view factor, F, is the planar projection of the curved surface in the direc-
tion of the droplet influx relative to the full perimeter of the surface. It represents a geometrical
factor that accounts for the inability of incoming droplets to directly reach an obstructed back
side of the surface. For example, the view factor is F = 1/π for a circular conductor. The liquid
water content, W, represents the mass of water within a specified volume of air (kg/m3).
Solving Equation 6.49 subject to an initial condition of B = 0 at t = 0, yields:
ηc FWV
B= t (6.50)
ρi
(a) (b)
Viscous heating Liquid—air interface
FIGURE 6.4
Schematic of (a) rime and (b) glaze ice.
270 Advanced Heat Transfer
For circular conductors, this mass balance is called Goodwin’s model (Goodwin et al. 1982).
This model accurately predicts both rime and glaze (wet) ice growth over a range of flow
conditions. It was developed particularly for applications to unheated cylindrical conduc-
tors (representing overhead power transmission lines).
At temperatures near the freezing point, transition occurs from rime to glaze ice. For glaze
ice, an additional energy balance is required to predict the rate of ice growth. The energy
balance involves several components as follows at the ice surface (see Figure 6.4).
where ξ, ca, hsl, Tw, and Ta refer to the recovery factor (an empirical coefficient accounting for
viscous dissipation in the boundary layer), specific heat of air, latent heat of fusion, and
water droplet and air temperatures, respectively. The droplet and air temperatures will
be assumed equal prior to droplet impact on the surface. Also, the effects of surface curva-
ture are assumed to be negligible for the thin water layer.
When sufficient energy is imparted into the ice surface during freezing to sustain an unfro-
zen water layer, including the release of latent heat from the impinging droplets, then tran-
sition to glaze ice occurs. Performing an energy balance in the unfrozen water layer on a
glaze ice surface,
∂B ξhV 2 1
qw + ρi hsl + + ηc V 3 W = h(Tw |B+b − Ta ) + ηc WVcw (Tw − Ta ) (6.51)
∂t 2ca 2
where b and cw refer to the thickness of the unfrozen water layer above the ice surface and
specific heat of water, respectively. From left to right in this energy balance, the terms
represent the wall heat input, release of latent heat of freezing droplets, viscous heating
in the boundary later, kinetic energy of incoming droplets, convective heat transfer, and sen-
sible cooling of the impacted droplets to the surrounding ambient air temperature.
From a scale analysis in the thin unfrozen water layer, it can be shown that the dominant
mechanism of heat transfer in the thermal energy equation is lateral conduction across the
layer in the y-direction, perpendicular to the surface, yielding,
∂2 T w
=0 (6.52)
∂y2
subject to T(B) = Tf (phase change temperature). At the liquid/air interface ( y = B + b), the
energy balance in Equation 6.51 provides the remaining boundary condition for the solution
of the thermal energy equation.
Solving the heat equation and differentiating the resulting temperature profile to find the
temperature gradient,
∂Tw C2 − C1 (Tf − Ta )
=− (6.53)
∂y 1 + C1 b
Multiphase Flows with Droplets and Particles 271
where,
Then substituting the water temperature into Equation 6.51 and rearranging,
∂B qw C1 C2 b
=− − + C3 − C2 (6.55)
∂t ρi hsl 1 + C1 b
where,
ηc WV 3
C3 = − (6.56)
2ρi hsl
Thus, the ice growth initially follows Equation 6.50 during rime ice growth and then
Equation 6.55 after the transition to glaze (wet) ice.
In Equation 6.55, the thickness of the unfrozen water layer, b(t), is required. From a mass
balance at y = B (ice/water interface),
dB db
ρi + ρw = ηc FWV (6.57)
dt dt
where B = Bw at t = tw at the onset of the glaze ice. The specific values of Bw and tw can be
determined by setting b→0 in Equation 6.55, equating this result with Equation 6.50 since
both results must match each other at the transition point, and then solving the resulting
values of ice thickness (Bw) and time (tw).
After Bw and tw are found, then the water layer thickness can be obtained by solving
Equation 6.57,
ηc FWV ρi
b= (t − tw ) − (B + Bw ) (6.58)
ρw ρw
The nonlinear first-order ordinary differential equations in Equations 6.55 and 6.57 can be
solved numerically to yield solutions for B(t) and b(t).
Sample model results of ice accretion on a circular conductor of radius 1.05 cm due to
freezing rain precipitation are shown in Figure 6.5. For the rime ice results, the liquid water
content and ambient air temperature are W = 0.001 kg/m3 and Ta = 263 K, respectively. The
surface is unheated and droplets freeze immediately upon impact to form rime ice. For the
glaze ice results, the surface is heated at Q = 520 W/m (heat input per unit length of cable)
and the other problem parameters are G = 0.00056 kg/m3 and Ta = 270.3 K.
In Figure 6.5, the dimensionless ice thickness, B*, and dimensionless time, t*, are defined as
B* = B/ro and t* = 2 (ρw/ρi) (PFt/ro) where ro and P represent a reference length (conductor
radius) and precipitation rate (P = WV/ρw), respectively. In the glaze ice results, the dry
growth limit (rime ice) is shown for reference purposes. Comparisons with experimental
data of Lu et al. (1998) shown close agreement between the predictive model and measure-
ments of ice growth due to freezing precipitation on a circular conductor. Further detailed
modeling of impinging supercooled droplets on ice surfaces has been reported by Poots
(1996), Myers and Hammond (1999), and Naterer (2011).
272 Advanced Heat Transfer
(a) (b)
0.015
Rime ice limit
Rime ice model Experimental—glaze (Lu et al.1998)
0.8
Experimental (Lu et al. 1998) Glaze ice model
0.010
0.6
B (m)
B*
0.4
0.005
0.2
0 0
0 0.4 0.8 1.2 1.6 2.0 0 2,000 4,000 6,000 8,000 10,000
t* t (s)
FIGURE 6.5
Thickness of an ice layer during freezing precipitation on a heated circular conductor for (a) rime ice and (b) glaze ice
conditions. (Adapted from G.F. Naterer. 2011. Cold Regions Science and Technology, 65: 5–12.)
Convective heat transfer and evaporation of droplets to particles occurs in many industrial
processes such as the production of spray-dried products, pharmaceutical powders, laun-
dry detergents, and spray pyrolysis. A common industrial device for particle formation is
a spray dryer. Slurry droplets are typically sprayed into a mixing chamber and then dried
by a hot gas stream entering from below the chamber in an opposite direction to the drop-
lets. The dried particles fall through the bottom of the chamber, after which they are later
collected, processed, and packaged as a final powdered product. In these processes, the final
quality of the drying product is highly sensitive to temperature. Cooling of the drying gases
by the downward particle flow reduces the drying effectiveness so the gas temperatures and
velocities are key design parameters. The gas temperature cannot be too high as overheating
or burning of the powder can adversely affect the product quality.
Consider a process of evaporative spray drying of droplets and particle evolution in a gas
stream. Figure 6.6 shows various stages of the drying process. The spray dryer takes a liquid
stream and separates the solute or suspension as a solid from the solvent. The liquid or slurry
stream is sprayed through a nozzle into a heated gas stream and vaporized. Solids form as mois-
ture leaves the droplets. Detailed heat and mass transfer models of these evaporation and par-
ticle formation processes have been reported by Eslamian, Ahmed, and Ashgriz (2009).
The process of droplet evaporation to particle formation consists of three main steps: a
shrinkage period; transition from the shrinkage to a constant diameter period; and finally
a constant diameter period. In the initial stage, a slurry solution is atomized and moved
into a reactor or spray chamber where the volatile species of droplets (solvents) evaporate
and mix with the carrier gas. The droplet is heated to the wet-bulb temperature of the sur-
rounding gas. The wet-bulb temperature refers to the temperature that a parcel of air would
have if it was cooled to the saturation temperature by the process of evaporation of water
into it using latent heat supplied by the parcel.
Multiphase Flows with Droplets and Particles 273
FIGURE 6.6
Droplet drying and shrinkage processes.
During the shrinkage period, the solvent evaporates from the droplet, leaving a higher
concentration of solute in the droplet. As more volatiles evaporate from the droplet, satura-
tion conditions develop on the droplet surface. Once the solute concentration on the surface
reaches a point of critical supersaturation (CSS), the solute starts to precipitate on the droplet
surface. CSS is a state that occurs when the solution contains more of the dissolved material
than can be dissolved by the solvent.
At the CSS point, the solute concentration within the droplet is equal or above the equilib-
rium saturation (ES). The solute begins to precipitate from nucleation sites and crystals
through the droplet (called volume precipitation). Eventually the droplet is filled with
crystals if the amount of solute available can fill the volume of the droplet.
For low initial solution concentrations, at the onset of precipitation on the droplet surface,
the solute concentration in a portion of the droplet may be less than the equilibrium satura-
tion. In this case, solution precipitation occurs where the local solute concentration is higher
than the equilibrium concentration, resulting in a thin layer of solute on the droplet surface.
As the evaporation continues and the droplet shrinks, this layer grows until the thickness
reaches a critical value and a rigid hollow particle or shell is formed.
Eventually the outer crust becomes a full outer shell of the solid particle. As the crust forms
and the shrinkage process ends, water is brought to the surface through pores which are
formed in the crust. At this stage, the third period begins, called the constant diameter
period. The solvent evaporation continues until the particle is entirely dried. After all water
leaves the droplet, a solid particle is formed and the particle is heated to the surrounding
gas temperature.
the solvent (denoted as Dls) and the concentration gradient, according to Fick’s law
(Chapter 1). Therefore, the mass conservation equation in spherical coordinates can be
rewritten as:
∂χ 1 ∂ ∂χ
ρl l − 2 ρl r2 Dls l = 0 (6.60)
∂t r ∂r ∂r
where ρl and χl refer to the liquid density and mass fraction of solvent, respectively.
Another important process during droplet evaporation is called Stefan flow. Although it
will be neglected in this analysis, Stefan flow becomes significant at high evaporation rates.
It can be included by adding it to above equation and/or correcting the heat and mass trans-
fer coefficients. Stefan flow is a transport process involving the movement of chemical
species by the gas flow due to production or removal of species at an interface. The Stefan
flux is different than Fick’s law (Chapter 1) but has the same dependence on a concentration
gradient in the gas stream. When some of the liquid droplet evaporates into vapor at the sur-
face of the droplet, it flows away from the droplet when it is displaced by additional vapor
evaporating from the droplet. At high evaporation rates, the total transport of species during
the evaporation process is a sum of both Fickian diffusion and Stefan flow.
Define the following variables,
χl
w= (6.61)
1 − χl
r
ms 4πr2 ρl χ s dr
y= = r0o (6.62)
m0 0
4πr2 ρl χ s dr
where w represents a ratio of the mass fraction of solvent (χl) to the solute (χs = 1 – χl). Also,
ms and mo represent the mass of solute within radius r and the total mass of solute in the
droplet, respectively.
Then the equation of mass conservation becomes:
2
∂w 16π 2 ∂ ρ l r2 ∂w
= 2 Dls (6.63)
∂t mo ∂y 1 + w ∂y
∂w
=0 (at y = 0) (6.65)
∂y
2
dmd 16π 2 ρl r2o ∂w
= Dls (at y = 1) (6.66)
dt mo 1 + w ∂y
Solving the mass conservation equation for w( y, t), subject to the initial and boundary
conditions, leads to the following droplet radius:
3mo 1 1 + w
ro =
3
dy (6.67)
4π 0 ρl
Multiphase Flows with Droplets and Particles 275
Once the solute concentration on the droplet surface reaches the CSS point, there is a
time delay between the onset of solute precipitation on the droplet surface and the time
when solute is precipitated to create a thin outer shell on the particle (called an induction
period). After this induction period, the droplet/particle diameter remains constant. Since
the diameter and volume remain constant, the further reduction of liquid due to evaporation
leads to an increasing saturated vapor space within the interior of the surface shell. As evap-
oration of the trapped liquid in the inner core continues, more solute precipitates in the wall
of the particle. Conservation of mass for the solute mass within the droplet, ms, can be writ-
ten as:
4π 3
ms = ρl rp,o − r3p,i − ρl Vp (6.68)
3
where rp,i, rp,o and Vp refer to the particle inner and outer radii, and the total void volume
within the particle, respectively.
Sample results for drying of zirconium oxychloride droplets are shown in Figure 6.7.
The solvent is water with a small percentage of hydrochloric acid. Problem parameters
include a droplet number density of 5 × 106 cm−3, reactor wall temperature of 600◦ C, and car-
rier gas flow rate of 2 L/min. The gas and droplet temperatures at the reactor inlet are 25◦ C.
The results show different droplet evolution stages in the shrinkage, induction, and constant
diameter periods. In Figure 6.7, for a 2.5 μm droplet, the duration of the constant diameter and
shrinkage periods are about 35% and 61% of the total drying time, respectively. The relative
duration of the constant diameter period increases with smaller initial droplet sizes. Also, it
can be observed that the evaporation rate is nearly constant during the constant diameter
period. The evaporation rate decreases with time in the shrinkage and induction periods
due to diffusion and accumulation of water vapor in the surrounding gas stream.
The droplet temperature has a significant effect on the evaporation process. As the tem-
perature of a droplet increases, the thermal energy of molecules also increases at the surface
(a) (b)
1.0 10
0.9
0.8
(dmd/dt)×109 (g/s)
Shrinkage period
0.7 ro = 1.5 μm Shrinkage
period Constant
(r/ro)2
FIGURE 6.7
(a) Droplet diameters and (b) evaporation rates during different drying periods. (Adapted from M. Eslamian, et al.
2009. Drying Technology, 27: 3–13.)
276 Advanced Heat Transfer
of the droplet, therefore increasing the rate of evaporation. Hoffman and Ross (1972) pre-
sented the following correlation for evaporating droplets and 1 , ReD ≤ 400:
During a spray drying process, the liquid stream is sprayed through a nozzle to
separate the solute or suspension as a solid from the solvent. A common form of
correlation describing the jet atomization process and droplet size is given by the ratio of
the Sauter mean diameter, D32, to the initial diameter, D0, of droplets, as follows:
D32
= C · Weα · Ohβ (6.70)
D0
The Sauter mean diameter, D32, represents an average droplet size defined by the diam-
eter of a sphere with the same volume to surface area ratio as a droplet of interest. The Ohne-
sorge number, Oh, is a ratio of the square root of the Weber number, We (defined in
Chapter 3), to the Reynolds number, Re,
√
μ We
Oh = √ = (6.71)
ρσL Re
The coefficients α and β depend on the atomizer design and must be determined
experimentally. For pressure atomizers, due to challenges with making droplet size
measurements in dense sprays with plain orifice nozzles, few correlations for the
mean droplet size have been published. For rotary atomizers, correlations for the
mean droplet size include the effects of variations in disc or cup diameter and rota-
tional speed, in addition to liquid properties and the liquid flow rate. The mean droplet
sizes produced by twin-fluid atomizers are typically correlated in terms of the Weber
and Ohnesorge numbers.
1/3
6
Dp = V (6.72)
π
Multiphase Flows with Droplets and Particles 277
The sphericity, Φ, of the particle is the ratio of the surface area of a sphere (with an equiv-
alent volume as the particle) to the actual surface area of the particle,
where Ap is the actual surface area of the particle. For Φ = 1, the particle is a sphere.
Prior to melting of the particles, the energy balance of a particle in an air stream can be
written as:
dTp,m πD2p
mp c p = h(Ta − Tp,s ) (6.74)
dt Φ
where Dp is the particle diameter, mp is the particle mass, cp is the particle specific heat, Tp,m is
the mean particle temperature, hm is the convective heat transfer coefficient, Tp,s is the mean
surface temperature of the particle, and Ta is the air temperature. The heat transfer coeffi-
cient can be determined by the Frossling correlation,
hDp
Nup = = 2 + 0.552Re1/2
p Pr
1/3
(6.75)
ka
hDp √
Nup = = 2 Φ + 0.552Re1/2
p Pr Φ
1/3 1/4
(6.76)
ka
h m Dp √
Shp = = 2 Φ + 0.552Re1/2
p Sc Φ
1/3 1/4
(6.77)
Dv,a
where hm, Dv,a, Sc and Sh refer to the mass transfer coefficient, diffusion coefficient of vapor
in a vapor/air mixture, Schmidt number, and Sherwood number, respectively. Based on the
analogy of heat and mass transfer (Chapter 3), the Nusselt and Sherwood numbers are iden-
tical when Pr = Sc.
The melting process of the particles can be divided into two stages: first, the ice particle is
heated from its initial temperature to the melting temperature, Tmelt, equal to 0◦ C; and
second, the particle melts at a constant temperature of 0◦ C until all solid becomes liquid
water. During the first stage, a lumped capacitance model can be used for the particle
temperature.
In the second stage during the phase transition process, the heat balance can be written as:
Nup
πDp ka (Ta − Tmelt ) = ṁev hv + ṁf hsf (6.78)
Φ
where ṁev is the evaporation rate, ṁf is the melting rate and hv and hsf refer to the latent heat
of vaporization and fusion of ice, respectively.
278 Advanced Heat Transfer
The mass of the particle ice core surrounded by liquid water, mp,i, can be determined by:
Also, the evolution rate of the total particle mass is given by:
dmp Shp
= −ṁev = −πDp ρ Dv,a (χ v,s − χ v,a ) (6.80)
dt Φ a
where ρa, χv,s and χv,a represent the density of air, vapor mass fraction at the particle surface,
and free stream vapor mass fraction, respectively.
Then a mass balance of the total liquid water, mp,w, is given by:
Also, the mass of the particle can be found based on the mass of the liquid layer,
π 3
mp = mp,i + ρw Dp − D3p,i (6.82)
6
Here ρw is the density of liquid water. This predictive model has shown good agreement
with experimental data for the melting times of particles, ratios of initial ice particle mass to
the final liquid droplet mass, and changes of surface shapes of non-spherical ice particles
(Hauk et al. 2016).
X
Gas absorption
Iλ Iλ + dIλ
Gas layer
Iλ0
dx x=L
FIGURE 6.8
Radiative absorption in a planar gas layer.
radiation intensity, Iλ, across a layer of thickness dx decays in proportion to the magnitude of
the intensity as a result of the participating medium,
where the proportionality constant, aλ, is called the absorption coefficient. It is usually tab-
ulated as a function of wavelength, λ, gas temperature, Tg, partial pressure, and the total
pressure of the gas. The intensity of radiation varies with wavelength, λ, and decays as a
result of absorption by a gas–particle layer.
Dividing by Iλ yields the following radiation equation of transfer for a homogeneous, non-
scattering gas layer:
dlλ
= −aλ dx (6.85)
Iλ
This equation indicates the decay of radiation intensity due to a participating medium.
Integrating the equation of transfer from x = 0, where I = Iλ0, to x = L, where I = IλL, yields
Beer’s law as follows:
Therefore the intensity of radiation decays exponentially due to absorption of the radia-
tion by particles and/or constituents in the gas layer between the surfaces.
The spectral absorptivity of the gas can be determined from the differences in radiative
intensities at the different locations,
Iλ0 − IλL
αλ = = 1 − exp(−aλ L) (6.87)
Iλ0
Based on Kirchkoff’s law for a diffuse gas (recall αλ = ϵλ; Chapter 4),
1 1
Eb,λ ελ dλ 1
εg ; 01 = 4 Eb,λ (1 − exp(−aλ L))dλ (6.88)
0
Eb,1 dλ σTg 0
280 Advanced Heat Transfer
For participating gases (water vapor and carbon dioxide) in radiation exchange at high
temperatures, the following correlations can be used:
0.45
Tg
αw = cw εw (6.89)
Ts
0.65
Tg
αc = cc εc (6.90)
Ts
where the subscripts w and c refer to water vapor and carbon dioxide, respectively, while Tg
and Ts refer to the gas temperature and particle (or surface) temperature, respectively. The
coefficients cw and cc are correction factors that are typically determined from empirical
correlations based on experimental data. The water vapor and carbon dioxide emissivities,
ϵw and ϵc, are also tabulated through empirical correlations as functions of the gas and sur-
face temperatures, mean beam length, and partial pressures. The mean beam length can be
interpreted as the radius of a hemispherical gas region exhibiting the same effective emissiv-
ity as the region enclosing the radiation exchange.
For a gas mixture involving multiple constituents, individual radiative properties are
combined appropriately. For example, in a mixture of water vapor and carbon dioxide,
where the subscript g refers to a gas. Here ϵg and αg are functions of the shape of the gas region
as a result of their dependence on the container in which the partial pressures and concentra-
tions of each gas constituent are calculated. Correlations are typically presented in terms of
the partial pressure of each gas constituent and an equivalent mean hemispherical beam
(a) (b)
0.7 0.3
pH L = 6.1 atm m pCO L =
2
2O
1.52 atm m
0.3
0.3
0.15
0.003 0.03
2
2O
εCO
0.07 0.03
εH
0.0015
0.03
0.015 pCO L =
2
0.0003 atm m
pH L=
2O
0.003 atm m
0.007 0.003
240 640 1,040 1,440 1,840 2,240 2,640 240 640 1,040 1,440 1,840 2,240 2,640
Temperature, K Temperature, K
FIGURE 6.9
Emissivities of (a) water vapor and (b) carbon dioxide in a gas mixture at 1 atm total pressure in hemispherical
domain. (Adapted from H.C. Hottel. 1954. “Radiant Heat Transmission,” in Heat Transmission, W.H. McAdams,
Ed., 3rd Edition, New York: McGraw-Hill.)
Multiphase Flows with Droplets and Particles 281
(a) (b)
0.07 0.07
T = 400 K T = 1,200 K
0.06 0.06 p L + p L = 1.52 atm m
c w
pc L + pw L = 1.52 atm m
0.05 0.05
0.91
0.04 0.04
0.91
Δε
Δε
0.46
0.03 0.03
0.46
0.02 0.02
0.23 0.23
0.01 0.01
0.091 0.091
0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
pw/ (pc + pw) pw/ (pc + pw)
FIGURE 6.10
Emissivity correction factors of mixtures of water vapor and carbon dioxide at temperatures of (a) 400 K and
(b) 1,200 K. (Adapted from H.C. Hottel. 1954. “Radiant Heat Transmission,” in Heat Transmission, W.H. McAdams,
Ed., 3rd Edition, New York: McGraw-Hill.)
length, L (approximately 3.4 × volume/surface area). The gray gas assumption (ϵg = αg) is
generally not valid for radiation exchange involving participating gases.
Tabulated values of cw, cc, ϵw, ϵc, Δϵ, and Δα, as well as correlations for various other gases,
are available in sources such as Hottel (1954) and Bergman et al. (2011). The emissivities of
water vapor and carbon dioxide in a mixture on nonradiating gases at atmospheric pressure
in a hemispherical domain are shown in Figure 6.9. These emissivities increase at higher
beam lengths and partial pressures of the gas. Also, emissivity correction factors at temper-
atures of 400 and 1,200 K are shown in Figure 6.10. In a gas mixture of water vapor and car-
bon dioxide, the emissivity and absorptivity correction factors are the same, Δα = Δϵ.
The following example considers how soot particles affect the processes of radiation
exchange. The radiative properties of particles emitting, absorbing, and scattering radiation
in a participating gas have been reported by various authors, for example, Lynch, Krier, and
Glumac (2010).
When soot particles are not present in the gas region, these emissivity and absorptivity
coefficients are readily available.
282 Advanced Heat Transfer
qnet
Typical gas
ελ, α λ
qe
qa
FIGURE 6.11
Radiation exchange in a closed chamber.
G = τg J + εg Ebg (6.94)
ρs εg Ebg + αg εs Ebs
J = ρs G + εs Ebs = (6.95)
1 − ρs τg
εs Aσ 4
qrad = εg Tg − αg Ts4 (6.98)
1 − ρs τ g
where ρs = 1 – ϵs and τg = 1 – αg. This result assumed gray walls, but still without soot par-
ticles in the gas mixture. Another correction is required to account for the soot particles.
Soot particles in the reacting flow are active participants in the radiative exchange.
Unlike H2O and CO2, soot particles emit continuous radiation. The sooty gas emissivity
depends on the operating pressure, quality of fuel atomization, particle distribution
within the chamber, air/fuel ratio, and other factors.
Multiphase Flows with Droplets and Particles 283
Using Beer’s law and a functional form of the absorption coefficient based on experi-
mental data,
βζs
aλ = (6.99)
λ
where ζs and β refer to the soot particle volume fraction and an empirical factor accounting
for the soot composition (usually between 4 and 10).
Integrating the spectral distribution of the soot emissivity with Beer’s law would be
impractical. Instead the emissivity of the soot–gas mixture can be approximated directly
through Beer’s law as follows:
where L, ag and as refer to the length of the chamber, and absorption coefficients for the gas
and soot, respectively. Then the mixture emissivity becomes:
ελ,mix = (1 − exp( − ag L)) + (1 − exp( − as L)) − (1 − exp( − ag L))(1 − exp( − as L)) (6.101)
Treating the soot particles as a gray substance (ϵs = αs) leads to:
εmix εg + εs − εg εs (6.103)
This result can be used to account for soot particles in the radiation exchange during
gas–particle flows with combustion.
The previous example showed that solid particles can significantly affect the process of
radiation exchange in gas–particle flows. Radiation through participating media of combus-
tion gases with particles was reported by Baker and Miller (2013). Further details on radia-
tive heat transfer in gas–particle flows were presented in books by Klinzing (1981) and
Kreith, Manglik, and Bohn (2010).
uniform solid distribution is often observed in vertical transport of liquid–solid flows in the
presence of gravity. However, in horizontal two-phase flows, such as the transport of solid
particles by a liquid flow in a pipe, asymmetrical forces lead to segregation and settling of
particles in the bottom section of a pipe (called settling slurries). The solid concentration pro-
file, which describes the spatial distribution of particle density across the pipe, becomes an
important design factor in horizontal slurry flows.
In Newtonian slurries, colloidal dispersion refers to the transport of very fine particles by
a liquid flow (Rep , 10−6, where Rep is the particle Reynolds number). Particles are main-
tained in suspension in the liquid. A homogeneous flow regime occurs with larger particles
(10−6 , Rep , 0.1). Low levels of turbulence are required to maintain the particles in a homo-
geneously suspended state. A pseudo-homogeneous flow regime occurs when particle sizes
increase to the range of 0.1 , Rep , 2. In horizontal two-phase flow in a pipe, segregation
is normally obtained as solid particles collect in a denser region along the bottom section
of the pipe. For larger particles with Rep . 2, a heterogeneous flow regime is observed with a
higher degree of segregation of solid particles throughout the mixture. At the critical deposi-
tion velocity, the solid particles begin to settle out of the mixture. The pressure drop through
the flow stream is reduced as a result of less drag and resistance of individual solid particles
suspended in the mixture flow.
Similarly to two-phase flow maps in the previous chapter, phase diagrams of liquid–solid
flows are also useful to determine various characteristics and features of the flow regimes
(Hewitt et al. 1997). Phase diagrams include factors such as pipe diameter, friction velocity,
particle size, and others. A phase diagram for horizontal liquid–solid flows would typically
illustrate the particle settling velocity, uso, in terms of a friction velocity, uo. The laminar sub-
layer near the wall of the pipe corresponds to a low value of uo and higher values occur
toward the inner turbulent core of the pipe flow. At high uo and low uso, homogeneous
flow occurs. Transition to longitudinal standing particle waves and transverse waves (con-
sisting of a grouping of solid particles) occurs when uo decreases and uso increases. Limits
involving the minimum transport for homogeneous flow with particles in suspension are
also usually included in these two-phase maps.
where Dp,50 is the particle diameter at a 50% passing sieve. It represents the median diameter
of the particle size distribution and the value of the particle diameter at 50% in the
Multiphase Flows with Droplets and Particles 285
cumulative distribution. Also, cd is the drag coefficient, which can be correlated as:
4 24
cd = √ + + 0.4 (6.106)
Res Res
where,
uso Dp,50
Rep = (6.107)
νl
These results can be derived based on approximate solutions of the settling and inertial
effects of particles during transport by the liquid–solid flow.
For horizontal two-phase flows in pipes, the solid particles are transported horizontally
without deposition when the fluid velocity, ul, exceeds the critical deposition velocity, uc.
Based on experimental data of Wasp, Kenne, and Gandhi (1977),
1/6
0.234 Dp,50 ρs − ρl 1/2
uc = 3.525ζs 2gD (6.108)
D ρl
where D refers to the tube diameter. The first factor on the right side, exponentiated by a
factor of 0.234, denotes the delivered fraction of particles through the flow stream.
Nonsettling slurries with a nonlinear dependence between the shear stress and strain rate
in the fluid are non-Newtonian flows called Binghamian slurries. These slurries obey the fol-
lowing form of constitutive relation:
du
τ = τo + μb (6.109)
dy
where μb and τo refer to the Bingham viscosity and slurry yield stress, respectively. The
form of the Navier–Stokes equations (Chapter 3) is altered for non-Newtonian fluids. Often
some form of linearization is required for the analysis of non-Newtonian flows.
Binghamian slurries can be effectively transported over short piping distances in laminar
flow if a sufficient flow rate is delivered to prevent settling out of solids in the mixture.
Higher frictional resistance with turbulence generally leads to higher pumping power for
the transport of liquid–solid flows in pipes. Similar to the transition to turbulence in single
phase flows, the following Durand–Condolios correlation gives the critical Reynolds number,
Recrit, below which the flow remains laminar:
⎡ 1/2 ⎤
ρ um,c D τo D ρm
2
Recrit = m = 1,000⎣1 + 1 + ⎦ (6.110)
μb 3,000μ2b
where um,c is the critical (transition) velocity. The subscripts b and m refer to Binghamian and
mixture, respectively.
At a certain particle size, the tendency of settling exceeds the tendency of the particle to
remain suspended in the flow. If the particle’s weight exceeds its net upward force from
dynamic effects in the flow, settling out of solid particles will occur. Particles with a suffi-
ciently small diameter, less than Ds,c, will remain in stable laminar flow (Dedegil 1986),
3πτo
Ds,c = (6.111)
2(ρs − ρl )g
286 Advanced Heat Transfer
6.11 Nanofluids
A nanofluid is a fluid containing nanometer sized particles which have an enhanced thermal
conductivity and heat transfer coefficient in comparison to the base fluid. Due to their
enhanced thermal properties, nanofluids have many promising applications in engineering
devices and systems such as heat exchangers, microelectronics cooling, vehicle thermal
management, pharmaceutical processes, among others. However, despite their advantages,
nanofluid mixtures also have drawbacks as they are inherently unstable due to sedimenta-
tion of particles. Also, solid particles increase the pumping power and may erode the surface
of a tube or wall.
Nanoparticles are colloidal suspensions of very small particles (1–100 nm) in the base fluid.
Typical nanoparticle materials are metals, oxides, and carbon nanotubes. Common base flu-
ids are water, oil, and ethylene glycol. This section will provide a brief overview of transport
mechanisms, governing equations, and thermophysical properties of nanofluids. Further
comprehensive reviews were presented by Kakaç and Pramuanjaroenkij (2009) and Wang
and Mujumdar (2008).
manipulation of species with selective polarity using nanoparticles to influence the fluid
motion in ways not possible in microscale or larger flow structures.
Two other significant transport processes occur in nanofluids—liquid layering at the
liquid/particle interface; and nanoparticle clustering. Liquid molecules may form a nano-
layer around the particles and hence increase the ordering of the atomic structure around
the particle. Also, nanoparticles can form clusters which become more conductive due to
the changes of packing fraction and effective volume as a result of clustering.
2
cp,m = ζk cp,k = (1 − ζp )cp,f + ζp cp,p (6.115)
k=1
2
(ρcp )m = ζk (ρcp )k = (1 − ζp )(ρcp )f + ζp (ρcp )p (6.116)
k=1
where the subscripts k, m, p, and f refer to the phase (base fluid or nanoparticle), mixture,
particle, and base fluid, respectively. The effective mixture viscosity can be determined
(Drew and Passman 1999) as follows:
μm = μf (1 + 2.5ζp ) (6.117)
where ζp is the particle volume fraction. Using these properties, the energy equation for fully
developed flow of a nanofluid in a heated tube becomes:
2 2
∂T ∂ T 1∂ ∂T ∂u
(ρcp )m u = (km + kd ) + r + μm (6.118)
∂x ∂x 2 r ∂r ∂r ∂r
where kd is a thermal dispersion coefficient (or dispersion conductivity) which accounts for
the effects of hydrodynamic dispersion and irregular movement of the nanoparticles due to
Brownian diffusion and thermophoresis. Often the thermal dispersion effects are neglected
or else included within the mixture conductivity so there is only one effective mixture ther-
mal conductivity that includes dispersion and other effects of intermolecular interactions.
The drift velocity, vdr, refers to the average fluid velocity when following a specific fluid
element as it travels with the flow stream. The drift velocity, vpf, is the velocity of the particle
relative to the base fluid,
2
ρ
vdr,p = vp − vf = vm − ζk k vf ,k (6.119)
k=1
ρm
288 Advanced Heat Transfer
where the bold font refers to a vector quantity. The slip velocity, vfp, can be determined from:
τp ρp − ρm
vpf = am (6.120)
fdrag ρp
where the particle relaxation time, τp, drag factor, fdrag, and acceleration, am, are given by:
ρp D2p
τp = (6.121)
18μf
1 + 0.15Re0.687 ; for Rep ≤ 1,000
fdrag = p (6.122)
0.0183Rep ; for Rep . 1,000
am = g − (∇ · vm )vm (6.123)
Here Rep = ρmVmDp/νm is the particle Reynolds number and Dp is the particle diameter.
Nanoparticles fluctuating randomly due to dispersion experience this net drift velocity in
the direction of the net bulk motion of the fluid.
kp ζp (∂T/∂x)p + kf ζf (∂T/∂x)f
km = (6.124)
ζ p (∂T/∂x)p + ζ f (∂T/∂x)f
Various correlations have been developed based the classical work of Maxwell (1881) on
fluid–particle mixtures. Maxwell’s model for the mixture thermal conductivity, km, of solid-
liquid mixtures with relatively large particles (above 1 μm diameter) provides reasonable
agreement with experimental data at low solid volume fractions.
kp + 2(kp − kf )ζ p + 2kf
km = kf (6.125)
kp − (kp − kf )ζ p + 2kf
Maxwell’s model predicts that the mixture thermal conductivity depends on the thermal
conductivity of the spherical particle and base fluid, as well as the volume fraction of the
solid particles.
Multiphase Flows with Droplets and Particles 289
More recent developments have attempted to include other effects such as intermolecular
interactions among randomly distributed particles, nonspherical particles, clustering of par-
ticles, the nanolayer between the nanoparticles and base fluid, Brownian motion, and ther-
mophoresis. Given the complexity of these processes, there have been no generally
applicable universal relations for all types of nanofluids. Some of the commonly adopted
correlations are listed below in Table 6.1.
TABLE 6.1
Thermal Conductivity Models for Nanofluids
Investigator(s) Model of Thermal Conductivity
Yu–Choi model !
(2003) kp þ 2(kp kf )(1 þ β)3 ζp þ 2kf
km ¼ kf
kp (kp kf )(1 þ β)3 ζp þ 2kf
(includes the effects of particle size, volume fraction, temperature, and properties
of the base fluid and particle subject to Brownian motion)
Xue–Xu model (2005) " #
kp kf (kp kf )ζ p þ 2kp ζp ln ((kp þ kf )=(2kf ))
km ¼ k
kp kf (kp kf )ζp þ 2kf ζ p ln ((kp þ kf )=(2kf )) f
(includes the effects of the nanolayer thickness, h, particle radius, Dp, particle volume
fraction, ζp, and temperature, T, in K)
290 Advanced Heat Transfer
TABLE 6.2
Correlations of the Heat Transfer Coefficient for Nanofluids
Investigator(s) Correlation of Heat Transfer Coefficient and Nusselt Number
Xuan–Li correlations
(2002) NuD ¼ 0:4328 1:0 þ 11:285ζ 0:754
p Pe0:218
D Re0:333
D Pr0:4 (laminar)
NuD ¼ 0:0059 1:0 þ 7:6286ζ 0:6886
p Pe0:001
D Re0:9238
D Pr0:4 (turbulent)
(laminar and turbulent flow in a tube including the effects of the volume fraction,
ζp, and Peclet, Reynolds, and Prandtl numbers, based on the particle diameter, Dp)
Pak–Cho correlation
(1998) NuD ¼ 0:021Re0:8
D Pr
0:5
(correlation for pool boiling of Al2O3-water nanofluids, where c and m are empirical
coefficients which depend on the particle volume fraction)
Yang et al.
correlation (2005) NuD ¼ cRem
D Pr
0:333 1=3
α (μf =μ1 )0:14
(turbulent flow of nanofluids in a tube, where f and δφ are the friction factor and
dimensionless thickness of the laminar sublayer, respectively)
In general, these models are extensions of classical correlations but with additional or
modified empirical parameters to account for various nanoscale processes. As a result,
the correlations are generally valid only for certain nanofluids over a selected range of
parameters. The development of more accurate and robust correlations for nanofluids is
an active area of continued research and development.
PROBLEMS
where,
∂T 1
q′′1 = −k ; q′′2 = WVcw (Tw − T1 ); q′′3 = − WV 3
∂y 0 2
Here V, W, and cw refer to the impacting droplet velocity, liquid water content
(mass of droplets per unit of volume of air–droplet mixture) and droplet spe-
cific heat, respectively. Evaporation and formation of a thin liquid film along
the wall, due to the droplet influx, are not considered in this approximation of
the wall heat flux.
b. Perform a scaling analysis of the flat plate boundary layer equations for single
phase flow (air) to estimate the thermal boundary layer thickness, δ1, at the
end of a plate of length L. Neglect viscous dissipation and assume steady-
state conditions.
c. Find the average Nusselt number based on the heat flux in part (a) and δ1 in
part (b). Use this result to determine the required ranges of G and ΔT so that
only q1 and q2 contribute appreciably to the wall cooling.
6.3 The quality of solidified layers obtained in a plasma spray deposition process is sig-
nificantly affected by solidification of the droplets impinging on a cold surface. Ini-
tially, the droplets are solidified upon impact on the surface. After a certain time
has elapsed, latent heat released by the solidified droplets generates a thin liquid
film above the solid layer. Assume that this liquid layer, b(t), grows over time accord-
ing to:
db a2
= a1 −
dt B(a3 + b)
where a1, a2, and a3 are constants and B(t) refers to the thickness of the solidified layer.
Perform a one-dimensional analysis to estimate the time when the liquid film first
appears above the solid layer. Express your answer in terms of the aforementioned
constants, V (velocity of incoming droplets), W (liquid content of droplets in the
air), and thermophysical properties.
6.4 Supercooled droplets are sprayed onto a plate (length of L, thickness of W ) tilted at an
angle of θ with respect to the horizontal plane. The droplets are partially solidified
and the remaining impacting liquid accumulates as a liquid film along the solidified
layer. The rate of film growth with distance along the plate, x, is assumed to be cons-
tant. The incoming droplet velocity is V and the liquid content in the air (kg/m3) is G.
a. Find the rate of growth of the solidified layer over time, in terms of L, θ, V,
and G.
b. How does the solid formation change if the film growth varies nonlinearly
with distance along the plate? Explain how the film thickness can be com-
puted under these conditions.
292 Advanced Heat Transfer
6.5 Consider the previous problem, but without the surface liquid film, so droplets
solidify immediately upon impact on the plate. Also, the thin plate is heated inter-
nally by electrical resistive elements so that the wall temperature beneath the solidi-
fied layer on the plate varies with time. The ambient temperature and convective heat
transfer coefficient of the surrounding airflow, above the solidified layer, are Ta and h,
respectively. Find the variation of plate temperature with time in terms of thermo-
physical properties, h, Ta, plate dimensions and the electrical parameters (current,
I, and resistance, R).
6.6 Metal powders and components with fine dimensional tolerances, such as gears, are
produced in a multistage melt particularization process. Atomization of the liquid
alloy through an impinging cool air jet separates individual liquid droplets. Subse-
quent solidification of droplets creates a temperature fluctuation, ΔTu. The droplet
temperature continually decreases as it is cooled until solidification, when a slight
temperature rise is observed. During this time interval of thermal fluctuation, assume
radiative heat transfer is negligible.
a. What physical process generates the temporary thermal fluctuation?
b. Consider a liquid metal stream with a density of ρ, melt temperature of Tm,
latent heat of fusion of hsl and a final spherical particle diameter of D.
The phase transformation occurs at the rate of ρhsl(dχl/dt) per unit volume,
where χl refers to the liquid fraction. Find an expression for the minimum con-
vective cooling coefficient, hmin, required to ensure that the thermal fluctua-
tion is avoided. Give your result in terms of ρ, hsl, air–melt temperature
difference, and the solidification rate, dχl/dt.
6.7 Spray deposition (or spray casting) is a manufacturing process where liquid metal
droplets are sprayed and deposited on a substrate. Consider a spherical droplet of liq-
uid metal with an initially uniform temperature that is suddenly subjected to convec-
tive and radiative cooling on its outer surface. The extent of partial solidification prior
to impact on the substrate is an important factor in assessing the properties of the final
spray-formed material. Find the rate of solidification of the droplet prior to impact on
the substrate in terms of the convection coefficient, h, droplet diameter, ambient gas
temperature, T∞, and surface temperature of the droplet, Ts.
6.8 In the previous question, the rate of solidification of droplets was expressed in terms
of the convection coefficient for an accelerating or decelerating droplet. In this ques-
tion, a closed form (analytic) solution is sought for the transient temperature distribu-
tion in the droplet when the terminal velocity of the droplet is attained. Express your
answer for the droplet temperature in terms of the drag coefficient for a spherical
droplet, cd, droplet diameter, droplet solidification rate (assumed constant), ṁ, ambi-
ent gas temperature, and thermophysical properties.
6.9 During a melt particularization process, a molten metal stream is injected into nitro-
gen gas at 120◦ C. Assume a Weber number of We = 3. If the metal stream disinte-
grates at 1,200◦ C when the fluid velocity is 22 m/s, estimate the rate of heat loss
from the disintegrated metal stream at the transition to secondary breakup.
6.10 Find expressions for the interfacial mass, momentum, and heat balances at the liquid–
vapor interface of an evaporating droplet in a microgravity environment. Assume
that the liquid and gas have constant thermophysical properties and the droplet
evaporates with a constant mass flux, ṁ. The surrounding gas has a uniform temper-
ature of T∞.
Multiphase Flows with Droplets and Particles 293
6.11 Spherical particles at a temperature of 30◦ C are injected into a cross flow of air
within a furnace. The relative velocity between the airstream and particles is 5 m/s.
Processed sandstone particles have properties of ρ = 2,200 kg/m3, cp = 920 J/kgK,
k = 1.7 W/mK and an emissivity of ϵ = 0.9. The mean particle diameter is 3 mm
and the freestream air temperature is 100◦ C. What is the initial rate of heating
(K/s) of the particles by the wall and airstream as they enter the furnace? Assume
that the particles do not appreciably participate in the radiation exchange with the
furnace wall at 600◦ C.
6.12 In a particulate removal process, solid particles migrate at 4 m/s relative to an air-
stream at 500 K. Find the drag force of the gas on an individual particle when the par-
ticle diameter is 0.2 mm.
6.13 Spherical particles of pulverized coal are injected into a cross flow of air in a furnace.
The initial temperature difference at the point of injection between the air and parti-
cles is ΔT0. The particles are heated by convection and radiation, at a heat transfer
coefficient of h, during their trajectory over a specified distance, L. If the particle veloc-
ity, V, is assumed to remain constant throughout the trajectory, then find the increase
of temperature of a particle in terms of ΔT0, L, V, h, D (particle diameter) and
thermophysical properties.
6.14 A packed bed with a 60% void fraction consists of spherically shaped particles
(0.4 mm in diameter) flowing through air at 600 K. Estimate the convective heat trans-
fer coefficient when the superficial fluid velocity is 12 m/s.
6.15 A horizontal airstream (carrier phase, subscript c) carrying small particles flows past
an inclined surface. Under certain conditions when the particle temperature and
velocities match airstream values, assume the thermal boundary layer equation can
be reduced to:
∂T ∂T ∂2 T
ζ c ρcp u + ζ c ρcp v = keff 2
∂x ∂y ∂y
where ζc refers to the volume fraction of the air phase (assumed constant) and keff
is the weighted average of particle and air conductivities. The velocity field is pre-
dicted by:
∂ψ ∂ψ
u= ; v=−
∂y ∂x
where,
1/2
U
η=y
vx
ψ = (Uvx)1/2 f (η)
The freestream velocity is U = Cxm, where the exponent, m, is related to the surface
inclination angle. Use a similarity solution method to find the temperature profile
within the boundary layer. Explain how this solution can be used to find the local
Nusselt number.
294 Advanced Heat Transfer
6.16 Spherically shaped particles of diameter D are dispersed uniformly in a gas flow and
heated primarily by radiative exchange with furnace walls at a temperature of Tw.
Find the time taken for the particles to reach a specified temperature, Tspec, when
the particle emissivity is ϵ. Express your result in terms of ϵ, thermophysical proper-
ties, D, Tw, and T0 (initial particle temperature).
6.17 Radiative heat exchange involves a participating gas at 1,600 K containing finely dis-
persed particles. The mixture has emission bands between 1 and 3 µm (where ϵ = 0.7)
and between 6 and 9 µm (where ϵ = 0.5). Find the total emissivity and emissive power
(W/m2) of the gas–particle mixture.
6.18 A gaseous mixture containing pulverized coal particles at 1,200◦ C has a radiative
absorption coefficient of κλ = 4 – exp(–λ/6) where λ refers to wavelength. A mono-
chromatic beam of radiation at λ = 5 μm enters a 20 cm thick layer of the mixture
with an intensity of 9 kW/m2 μm·sr. What beam intensity emerges from the layer?
The solid particles emit and absorb radiation as they pass through the layer.
6.19 A monochromatic beam of radiation at a wavelength of 4 m and intensity of 8 kW/m2
μm·sr enters a sooty gas layer (10 cm thick) containing finely dispersed particles. The
particles absorb and emit radiation so that the beam intensity is reduced by 20% as it
leaves the layer. What is the absorption coefficient of the gas–particle layer at the
given wavelength of radiation?
6.20 How can the absorption coefficient be used to characterize the optical thickness or
degree of transparency of a sooty gas with particles?
6.21 How can Beer’s law be generalized to an equation of transfer for a participating gas–
particle mixture including emission and absorption of radiation by the particles?
Assume that the radiant intensity becomes independent of position as the gas layer
thickness becomes sufficiently large. Does your result approach the correct asymp-
totic trends in free space, without emission or absorption?
6.22 Generalize the result from the previous question to a vector form of the equation of
transfer for a participating medium.
6.23 Consider radiative heat exchange in a closed chamber between a participating gas
with particles at Tg and surfaces at a temperature Ts. Find an expression for the net
rate of radiative heat transfer between the gas and walls.
6.24 For radiative absorption by solid particles in a sooty gas layer, the radiation
equation of transfer can be used to determine how the intensity of radiation is
reduced due to the absorption. How can the radiative heat flux be determined at a
particular location, r, when the intensity of the beam of radiation is known at that
position? Provide an integral expression that shows the appropriate range of
solid angles.
6.25 Extend the result from the previous question to derive the governing equation for
combined convective–radiative transport through the gas–particle mixture. Show
how the thermal energy equation can be written in terms of a radiative heat flux vec-
tor, qr, and alternatively, in terms of the intensity of radiation, Iλ.
6.26 Convection of a gas stream with dispersed particles is encountered in a channel of
height H and length L. The wall temperature, Tw, is constant. Assume that heat
released by particle combustion can be approximated by a constant heat source, q̇,
in the energy equation. Also, the wall heat flux is qw(1 + x/L), which increases with
distance, x, along the channel.
Multiphase Flows with Droplets and Particles 295
∂(uT) ∂(vT) ∂2 T
ρcp + ρcp = k 2 + q̇
∂x ∂y ∂y
References
T.M. Baker and T.F. Miller. 2013. “Ultraviolet Radiation from Combustion of a Dense Magnesium
Powder Flow in Air,” AIAA Journal of Thermophysics and Heat Transfer, 27: 22–29.
T.L. Bergman, A.S. Lavine, F.P. Incropera, and D.P. DeWitt. 2011. Fundamentals of Heat and Mass
Transfer, 7th Edition, New York: John Wiley & Sons.
J. Buongiorno. 2006. “Convective transport in Nanofluids,” ASME Journal of Heat Transfer, 128: 240–
250.
R. Clift, J.R. Grace, and M.E. Weber. 1978. Bubbles, Drops and Particles, New York: Academic Press.
C. Crowe. 1999. “Modeling Fluid–Particle Flows: Current Status and Future Directions,” AIAA Paper
99–3690, 30th AIAA Fluid Dynamics Conference, Norfolk, VA.
S.K. Das, N. Putra and W. Roetzel. 2003. “Pool Boiling Characteristics of Nanofluids,” International
Journal of Heat and Mass Transfer, 46(5): 851–862.
C.N. Davies. 1945. “Definitive Equations for the Fluid Resistance of Sphere,” Proceedings of the Royal
Society, A83(57): 259–270.
M.Y. Dedegil. 1986. “Drag Coefficient and Settling Velocity of Particles,” International Symposium on
Slurry Flows, ASME, FED, Anaheim, CA.
D.A. Drew and S.L. Passman. 1999. Theory of Multicomponent Fluids, Berlin: Springer.
M. Eslamian, M. Ahmed, and N. Ashgriz. 2009. “Modeling of Solution Droplet Evaporation and
Particle Evolution in Droplet-to-Particle Spray Methods,” Drying Technology, 27: 3–13.
D. Gidaspow. 1994. Multiphase Flow and Fluidization: Continuum and Kinetic Theory Descriptions,
Cambridge: Academic Press.
296 Advanced Heat Transfer
E.J. Goodwin, J.D. Mozer, A.M., Di Gioia Jr., and B.A. Power. 1982. “Predicting Ice and Snow Loads for
Transmission Lines,” Proceedings of 1st IWAIS, Hanover, NH, pp. 267–273.
T. Hauk, E. Bonaccurso, P. Villedieu, and P. Trontin. 2016. “Theoretical and Experimental Investiga-
tion of the Melting Process of Ice Particles,” AIAA Journal of Thermophysics and Heat Transfer,
30: 946–954.
G.F. Hewitt, G.L. Shires, and Y.V. Polezhaev, Eds. 1997. International Encyclopedia of Heat and Mass
Transfer, Boca Raton: CRC Press/Taylor & Francis.
R. Hilpert. 1933. “Heat Transfer from Cylinders,” Forsch. Geb. Ingenieurwes, 4: 215.
T.W. Hoffman and L.L. Ross. 1972. “Theoretical Investigation of the Effect of Mass Transfer on Heat
Transfer to an Evaporating Droplet,” International Journal of Heat and Mass Transfer, 15: 599–617.
H.C. Hottel. 1954. “Radiant Heat Transmission,” in Heat Transmission, W.H. McAdams, Ed., 3rd Edi-
tion, New York: McGraw-Hill.
S. Kakaç and A. Pramuanjaroenkij. 2009. “Review of Convective Heat Transfer Enhancement with
Nanofluids,” International Journal of Heat and Mass Transfer, 52: 3187–3196.
G.E. Klinzing. 1981. Gas-Solid Transport, New York: McGraw-Hill.
J. Koo and C. Kleinstreuer. 2005. “Laminar Nanofluid Flow in Micro Heat Sinks,” International Journal
of Heat and Mass Transfer, 48(13): 2652–2661.
F. Kreith, R.M. Manglik, and M.S. Bohn. 2010. Principles of Heat Transfer, 7th Edition, Pacific Grove:
Brooks/Cole Thomson Learning.
M.L. Lu, N. Popplewell, A.H. Shah, W. Barrett, and A. Au. 1998. “Mass of Ice Accretion from Freezing
Rain Simulations,” Proceedings, 8th IWAIS Conference, Reykjavik, Iceland.
P. Lynch, H. Krier and N. Glumac. 2010. “Emissivity of Aluminum-Oxide Particle Clouds: Application
to Pyrometry of Explosive Fireballs,” AIAA Journal of Thermophysics and Heat Transfer, 24:
301–308.
M.J. Martin. 2011. “Scaling of Transient Particle-Fluid Heat Transfer in Brownian Motion,” AIAA
Journal of Thermophysics and Heat Transfer, 25: 177–180.
A.D. Maude and R.L. Whitmore. 1958. “A Generalized Theory of Sedimentation,” British Journal of
Applied Physics, 9: 477–452.
J.C. Maxwell. 1881. A Treatise on Electricity and Magnetism, 2nd Edition, Oxford: Clarendon Press.
S.P. Mehrotra. 1981. “Mathematical Modeling of Gas Atomization Process for Metal Powder Produc-
tion,” Powder Metallurgy International, 13(3): 132–135.
T.G. Myers and D.W. Hammond, 1999. “Ice and Water Film Growth from Incoming Supercooled
Droplets,” International Journal of Heat and Mass Transfer, 42: 2233–2242.
G.F. Naterer. 2011. “Multiphase Transport Processes of Droplet Impact and Ice Accretion on Surfaces,”
Cold Regions Science and Technology, 65: 5–12.
B. Pak and Y. Cho. 1998. “Hydrodynamic and Heat Transfer Study of Dispersed Fluids with Submi-
cron Metallic Oxide Particles,” Experimental Heat Transfer, 11(2): 151–170.
G. Poots. 1996. Ice and Snow Accretion on Structures, New York: John Wiley & Sons.
K. Pope, G.F. Naterer, and Z. Wang, 2011. “Pressure Drop of Packed Bed Vertical Flow for Multiphase
Hydrogen Production,” International Journal of Hydrogen Production, 36: 11338–11344.
S.A. Putnam, D.G. Cahill, P.V. Braun, Z. Ge, and R.G. Shimmin. 2006. “Thermal Conductivity of Nano-
particle Suspensions,” Journal of Applied Physics, 99(8): 084308.
P.R. Sahm and P. Hansen, 1984. Numerical Simulation and Modelling of Casting and Solidification Processes
for Foundry and Cast House, CIATF, CH-8023, International Committee of Foundry Technical
Associations, Zurich, Switzerland.
D.S. Udawattha, S. Dilan, and M. Narayana. 2018. “Development of a Model for Predicting the Effec-
tive Thermal Conductivity of Nanofluids: A Reliable Approach for Nanofluids Containing
Spherical Nanoparticles,” Journal of Nanofluids, 7(1): 129–140.
X. Wang, G.F. Naterer and E. Bibeau, 2008. “Multiphase Nusselt Correlation for the Impinging Droplet
Heat Flux from a NACA Airfoil,” AIAA Journal of Thermophysics and Heat Transfer, 22: 219–226.
X.Q. Wang and A.S. M.S. Mujumdar. 2008. “A Review of Nanofluids - Part I. Theoretical and Numer-
ical Investigations,” Brazilian Journal of Chemical Engineering, 25(4): 613–630.
Multiphase Flows with Droplets and Particles 297
E.J. Wasp, J.P. Kenne, and R. Gandhi. 1977. “Solid-Liquid Flow, Slurry Pipeline Transportation,” Trans
Tech Publications, Zurich, Switzerland, pp. 9–32.
S. Whitaker. 1972. “Forced Convection Heat Transfer Correlations for Flow in Pipes, past Flat Plates,
Single Cylinders, Single Spheres, and for Flow in Packed Beds and Tube Bundles,” AIChE Journal,
18: 361–371.
Y. Xuan and Q. Li. 2002. “Investigation of Convective Heat Transfer and Flow Features of Nanofluids,”
Journal of Heat Transfer, 125: 151–155.
Q. Xue and W.M. Xu. 2005. “A Model of Thermal Conductivity of Nanofluids with Interfacial Shells,”
Materials Chemistry and Physics, 90(2): 298–301.
Y. Yang, Z.G. Zhang, E.A. Grulke, W.B. Anderson, and G. Wu. 2005. “Heat Transfer properties of
Nanoparticle-in-fluid Dispersions (Nanofluids) in Laminar Flow,” International Journal of Heat
and Mass Transfer, 48(6): 1107–1116.
W. Yu and S.U. Choi. 2003. “The Role of Interfacial layers in the Enhanced Thermal Conductivity of
Nanofluids: A Renovated Maxwell Model,” Journal of Nanoparticle Research, 5: 167–171.
7
Solidification and Melting
7.1 Introduction
Solidification and melting occur in many engineering systems such as materials process-
ing (casting solidification, extrusion, and injection molding, among others), phase change
materials (PCMs), and molten salts for thermal energy storage of solar thermal or
advanced nuclear power plants. Two methods are commonly used in the analysis of
solid–liquid phase change problems—either an interface tracking method or a mixture formu-
lation. In the interface tracking method, the moving phase interface is explicitly tracked by
analytical methods or a computational grid. For example, the interface movement is
determined by adjusting the position of nodes on the interface. In a mixture model, the
governing equations are written in terms of mixture quantities within a control volume,
including both the liquid and solid phases, such as, mixture velocity, mixture density, and
so on. This approach reduces the resulting number of equations and overall solution
complexity but the spatial averaging procedure loses detailed features of the interfacial
transport processes. As a result, further supplementary relations, typically in the form
of algebraic equations, are added back into the formulation to resolve the interfacial
transport processes.
Various types of microstructures are formed during solidification processes. Allotropy
refers to the existence of two or more molecular or crystal structures within a solidified
material. Interstitial diffusion refers to the migration of atoms through microscopic (intersti-
tial) openings in the lattice structure of a solid. A peritectic reaction is a process in which a
solid changes to a new solid phase and liquid upon heating and the reverse on cooling. A
peritectoid refers to a transformation from two solid phases to a third different solid phase
upon cooling. In a polyphase material, two or more phases are present. The primary phase
appears first upon cooling. Also, the solubility of a multicomponent mixture is an impor-
tant concept in the solidification of binary and multicomponent alloys. It is defined as the
maximum concentration of atoms of a particular constituent that can dissolve into
the mixture.
A solid–liquid phase interface during solidification can be characterized as faceted or
non-faceted. A faceted interface is relatively smooth at an atomic scale but microscopically
jagged at a macroscopic scale. Comparatively few bonds are exposed to atoms in the liquid
and there is a tendency to close atomic gaps at the phase interface. It is usually formed at
high undercooling levels (at a temperature below the freezing point without becoming a
solid). Minerals usually have faceted interfaces during phase transition. A large difference
in atomic structures in both phases leads to a diminished tendency to incorporate new
atoms into the solidified crystal structure, thereby leading to a microscopically smooth
phase interface. On the other hand, non-faceted interfaces are usually formed under
299
300 Advanced Heat Transfer
low undercooling levels and materials with a low entropy of fusion such as metals and metal
alloys. Anisotropy of material properties occurs due to dendritic growth along preferred
crystallographic directions.
The phase diagram for solid–liquid mixtures illustrates the effects of species concentration
on the phase change processes. Figure 7.1 illustrates a typical phase diagram for a binary
alloy. When a liquid alloy is cooled, it first solidifies at the liquidus temperature correspond-
ing to the concentration of constituent A in the mixture of A and B. The solid has a compo-
sition determined by the solidus line at the same temperature. A two-phase mixture of
solid and liquid coexist in equilibrium in the region between the liquidus and solidus
lines (also called a mushy region). The mixture is cooled further until it reaches the eutectic
temperature. This temperature refers to the minimum temperature of the two-phase region
between the solidus and liquidus lines that contains unfrozen liquid. The eutectic point refers
to the composition of the minimum freezing point, or the temperature at which a liquid of
eutectic composition freezes. The eutectic temperature is the temperature at which a liquid
of eutectic composition freezes to form two solids simultaneously under equilibrium
conditions.
Below the eutectic temperature, simultaneous growth of two or more solid phases from
the liquid are obtained. A material with a composition below the eutectic point is called a
hypoeutectoid, whereas a material above the eutectic composition is called a hypereutectoid.
These concepts are applicable to binary constituent materials and analogous concepts
are used for a higher number of constituents in the mixture. In multiphase systems, the
Gibbs phase rule identifies the number of phases, P, present, by P = C + N − F where C, N,
and F refer to the number of constituents, number of noncompositional variables (normally
2; p and T ) and number of degrees of freedom (or the number of independent intensive
variables), respectively.
In this chapter, phase change heat transfer in the solidification and melting processes will
be examined. Topics include the formulation of governing equations, one- and two-dimen-
sional problems in various geometries, and advanced solution methods based on integral,
similarity, and quasi-stationary models.
solidification is complete)
Eutectic line
Solid (α + β )
FIGURE 7.1
Phase diagram for a solid–liquid binary mixture.
Solidification and Melting 301
g = h − Ts (7.1)
G = H − TS (7.2)
where H, T, and S refer to the total enthalpy, temperature, and total entropy, respectively.
Using the Gibbs free energy, an important parameter in solidification processes (called the
entropy of fusion) can be determined. The entropy of fusion characterizes the degree of change
of molecular disorder in the transition from the microscopic structure of a liquid to a
crystalline solid. Define the Gibbs free energy of the liquid and solid phases as follows,
Gl = Hl − TSl (7.3)
Gs = Hs − TSs (7.4)
where the subscripts l and s refer to liquid and solid, respectively. Thermodynamic curves of
the Gibbs free energy intersect at the phase change temperature, Tm. The change in Gibbs free
energy of a mixture of two phases, ΔGmix, at any temperature is given by:
ΔG = 0 = ΔH − Tm ΔS (7.6)
Therefore the entropy of fusion, per unit mass, can be approximated by:
hsl
Δsf = (7.7)
Tm
where hsl represents the latent heat of fusion. Richard’s rule states that the entropy of fusion is
approximately constant, 8.4 J/mol K, for most metals. The entropy of fusion is a useful
parameter to characterize the nature of solidification and melting processes of materials.
particle impurities preexist to assist in generating nucleation sites for the solidifying crystals.
This chapter will focus on heterogeneous nucleation as the onset of solidification processes.
The critical condition for the onset of nucleation occurs when the driving force for solid-
ification is equal to the respective force for melting. This occurs when G is minimized, or
alternatively ΔG from the initial liquid state is maximized. This extremum corresponds to
the activation energy that must be overcome in order to form a crystal nucleus from the liq-
uid. The critical condition for the onset of homogeneous nucleation occurs when the
decrease in the Gibbs free energy of the crystal becomes (Kurz and Fisher 1984):
ΔG = σA + ΔgΔV (7.8)
where Δg and ΔV refer to the Gibbs free energy difference between the liquid and solid, per
unit volume, and the spherical volume enclosed by a crystal of surface area of A, character-
ized by a radius of ro, respectively. Also, σ is the specific interface energy. It refers to the
excess energy at the interface compared to the bulk liquid, or alternatively, the work
required to maintain crystal bonds together at the interface without melting. It can be
expressed in terms of the Gibbs–Thomson coefficient, Гgt, as follows,
where Δsf is the entropy of fusion per unit volume. The Gibbs–Thomson coefficient is 1 × 10−7
for most metals.
Using Richard’s rule in Equation 7.7 at the phase change temperature,
hsf
Δg = Δh − TΔs = (Tm − T) (7.10)
Tm
where ΔT = Tm − T is the undercooling level below the phase change temperature. Combin-
ing this result in Equation 7.8 with a characteristic spherical volume of radius ro which is
encompassing the nucleating crystal,
4
ΔG = σ4πr2o + πr2o Δsf ΔT (7.11)
3
Therefore, at the onset of nucleation, the decrease in the Gibbs free energy due to phase
change balances the work required to keep the initial crystal bonds in the lattice structure
from melting back to the liquid, plus the change in the Gibbs free energy due to the transition
from liquid to solid phases.
Another important concept to be used in this chapter for solidification processes involving
multiple constituents is the chemical potential. The chemical potential of species A in a multi-
component mixture, μA, represents the rate of change of the Gibbs free energy of species A
with respect to a change in the number of moles of A in the mixture, NA,
∂gA
μA = (7.12)
∂NA T,p
Here the derivative is evaluated at constant pressure and temperature. Uniformity of the
Gibbs free energy (i.e., dg = 0) is required at thermodynamic equilibrium.
Solidification and Melting 303
where ΔGon and κ refer to the Gibbs activation energy (a material dependent property) and
Boltzmann’s constant (κ = 1.38 × 10−23 J/K), respectively. The exponential term in brackets
indicates the amount of activation energy required for the onset of nucleation of a critical
number of clustered atoms in a crystal lattice. The rate of nucleation, or rate of forming
new crystals on a lattice structure, increases exponentially with the level of undercooling
below the phase change temperature.
Alternatively, Equation 7.13 can be written as follows:
ΔGon
Nr = No exp − (7.14)
κT
where Nr and No refer to the average number of newly formed spherical clusters of radius
r and the total number of atoms along the interface, respectively. The Gibbs activation
energy, ΔGon , represents a thermal activation barrier over which a solid cluster must
pass to become a stable nucleus. The probability of forming a cluster from the liquid phase
decreases rapidly as the cluster size increases. A typical limiting number for sustaining a
stable solid crystal within a liquid is 100 atoms. At small undercooling levels, the critical
radius of nucleation is large, meaning a small probability of forming a stable nucleus from
the liquid.
A typical process of crystal formation and movement of a solid–liquid interface is
illustrated in Figure 7.2. The structure of the phase interface can be characterized by the
nature of the surface area bonding at the interface. Figure 7.2 illustrates a simplified process
of attachment of cube-shaped crystals onto a lattice structure at an interface. The interface
(0) (4)
(2) (3)
(1)
(5)
(6)
FIGURE 7.2
Stages of crystal and solid formation.
304 Advanced Heat Transfer
growth is determined by the probability that a molecule will reach the interface and remain
there until it bonds to the interface, without melting back to the liquid. This probability
increases with a higher number of neighbors in the crystal.
In Figure 7.2, a type 3 atom is halfway in the solid (three sides bonded to the solid) and
halfway in the liquid (three sides exposed to liquid). Type 3 atoms are added until a row
along the interface is completed. Then type 2 atoms (two sides bonded to the solid) are
needed to start the next row, which is adjacent to the previous row at the solid–liquid
interface. This requires more undercooling than a type 3 formation. Finally, an entirely
new row is initiated by a type 1 atom (only one side attached to the solid). This formation
requires the most undercooling since the remaining five sides are exposed to liquid and
possible remelting.
The morphological stability of the interface affects the resulting final structure of the solid–
liquid interface. For a pure material, a stable interface is called columnar, whereas an unsta-
ble interface is equiaxed. Consider the onset and advancement of a wavy phase interface in
the positive x-direction in a stable, columnar interface (see Figure 7.3). The peaks and valleys
of the wavy interface, where the peak refers to the farthest extent into the liquid, are denoted
by planes A and B, respectively, in Figure 7.3. The corresponding heat flows are qA (through
the solid) and qB (into the solid from the liquid), respectively. The temperature gradient in
the liquid is higher along plane A since the phase change temperature is reached at the edge
of the solid perturbation. As a result, qA is larger than qB at that location, so that the pertur-
bation is melted and the interface remains stable (planar).
On the other hand, consider a wavy interface, specifically an interface formed along the
outer edge of a spherically shaped crystal that grows radially outwards. In this case, latent
heat is released radially outward from the solidified crystal into the liquid. The crystal
grows radially until it impinges on other nuclei. The heat flows along the peak (plane A)
and valley (plane B) of the wavy interface are denoted by qA and qB, respectively. In this
case, the temperature gradient is again higher along plane A, which suggests that the tip
of the wavy interface rejects more heat. As a result, the perturbation at this tip grows faster
and the waviness of the interface increases (an unstable interface). The resulting structure is
equiaxed due to the sustained growth of interface perturbations along certain crystal
axis directions.
Constitutional supercooling during solidification occurs due to compositional changes in the
solid as a result of cooling of a liquid below the freezing point ahead of the solid–liquid
A qA B T Solid
TM TL
B qB Solid Liquid
Solid V A Liquid
Stable (columnar) Stable (planar)
qB V T Solid TL
TM
A Solid
Liquid
qA B
Liquid
FIGURE 7.3
Stability of the phase interface.
Solidification and Melting 305
GV
=c Equiaxed
(C on
o dendritic
mi nstan sta
nt
cro t s
str c a l
uc e o
tur f
e)
Dendritic
V
nstant
GV = co Cellular
e of
ant typ
(Const ructure)
t
micros Planar
G = ∂T/∂x
FIGURE 7.4
Interface structures at varying temperature gradients.
interface. As a result of constitutional supercooling, various transitions can occur among the
phase interface structures, for example, planar, cellular, dendritic, or globulitic (a formation
of isolated globule-type structures resembling dendrites). A range of common interface
structures during solidification is illustrated in Figure 7.4.
Figure 7.4 depicts the variation of structures in terms of the interface velocity, V(mm/s),
and liquid temperature gradient at the phase interface, G = ∂T/∂x(K/mm). For low temper-
ature gradients, cellular, dendritic, and equiaxed dendritic structures are observed. How-
ever, at high liquid temperature gradients, planar, cellular, dendritic, and equiaxed
dendritic structures are observed. Transitions between each type of interface structure usu-
ally occur across lines of constant G/V ratios. Also, transitions between different scales of
interface structures are observed across lines of constant GV products. Further detailed anal-
ysis of various interface structures was presented by Chalmers (1964).
Unlike solidification, which typically requires undercooling for the onset of nucleation,
melting processes do not require superheating. Phase transition from a solid to a liquid
phase typically occurs at the melting temperature. Superheating is not required in melting
since no additional interface energy is required to hold a particular molecular structure
together in the liquid phase. In melting of materials, a newly melted solid particle forms
in an array of voids surrounded by loose regions of disordered crystals. The crystallographic
order of the solid disappears when a series of dislocations breaks up the closely packed
structure during a melting process.
where c1 and c2 are empirical constants. For example, c1 = 4.3 and c2 = 0.12 for an Al 7% Si
alloy. Also, the gradient acceleration parameter, GAP, is given by:
V∂T/∂x|i
GAP = (7.16)
tsol
where V, ∂T/∂x|i and tsol refer to the interfacial velocity and temperature gradient, and solid-
ification time (time to pass from the liquidus to solidus temperature), respectively. The ulti-
mate tensile strength represents the capacity of a material to withstand tensile loads of
elongation, as opposed to the compressive strength which withstands loads of compression.
Another example is the yield stress. The yield stress is defined as the stress at which a mate-
rial begins to deform plastically and nonlinear (elastic and plastic) deformation begins. It can
be related to a reference yield strength, σo, as follows:
σ y = σ o + cV 1/4 (7.17)
where χk refers to the mass fraction of phase k in the multiphase mixture. For example, the
mixture density and velocity can be expressed as:
ρ = χ l ρl + χ s ρs (7.21)
u = χ l ul + χ s us (7.22)
Solidification and Melting 307
(a) (b)
Liquid TM Liquid
Mushy
Temperature
dy
Solidus
Phase k Liquidus
(solid)
Solid
TE
Control
volume CS CL
(Eutectic)
dx Composition
FIGURE 7.5
(a) Multiphase control volume and (b) phase diagram.
Consider the transport of the general scalar quantity for phase k, φk, in a differential control
volume comprising the two-phase mixture (see Figure 7.5). For multicomponent mixtures, a
phase equilibrium diagram will also be required. In Figure 7.5, a binary eutectic phase dia-
gram is illustrated including the liquidus and solidus lines, which are functions of temper-
ature and solute concentration.
The general conservation equation for φk may be written as:
∂ ∂ ∂
(χ ρ ϕ ) + (χ ρ uk,j ϕk ) = − (χ jk,j ) + χ k S˙k (7.23)
∂t k k k ∂xj k k ∂xj k
where j = 1, 2, 3 and k = 1, 2. Also, χk and jk refer to the phase mass fraction and diffusive flux
in phase k, respectively. From left to right in Equation 7.23, the terms represent the transient
accumulation of φk in the control volume occupied by phase k; the net advective flow of φk
across the phase k portion of the control surface; the net diffusive flow into phase k across the
interface; and the source of φk in the phase k portion of the control volume, respectively. It
will be assumed that the differential surface area of phase k, denoted as dAk, within the con-
trol volume is equal to χkdA.
The mixture multiphase equation will be obtained by summing the individual phase
equations over all phases within the control volume and rewriting the variables in terms
of mixture quantities. In addition to these mixture conservation equations, interfacial bal-
ances will also be developed in upcoming sections.
∂ρ ∂
+ (ρuj ) = 0 (7.24)
∂t ∂xj
At the solid–liquid phase interface, the difference between the liquid and solid flows into/
out of the interface balance the rate of mass change due to differences in phase densities.
In one-dimensional conditions:
dn
ρs Vs − ρl Vl = (ρs − ρl ) (7.25)
dt
where the solid and liquid phases are designated by subscripts s and l, and n, Vs and Vl
denote the normal to the interface and individual phase velocities normal to the interface,
respectively. The interface velocity, dn/dt, is based on movement of the interface over a dis-
tance of dn over a time increment of dt. If there is little or no density difference between the
solid and liquid phases, the interfacial mass balance reduces to equal velocities on both sides
of the interface. In solidification processing of materials, the density change effect has signif-
icant implications on shrinkage voids and resulting defects in materials.
For the conservation of momentum, substitute φk,i = uk,i, jk,ij = pkδij − τk,ij and Sk,i = Fk,i,b +
Fk,i,int in Equation 7.23 where the source term represents the xi component of the body force
on phase k, as well as interfacial momentum exchange and cross-phase interactions. Then
adding the individual phase momentum equations over both liquid and solid phases, yields
the following mixture momentum equation:
∂ ∂ ∂p ∂ ∂
(ρui ) + (ρ χ uk,i uk,j ) = − + μ (χ uk,i ) + Fk,i,b + Fk,i,int (7.26)
∂t ∂xj k k ∂xi ∂xj k ∂xj k
where i = 1, 2, 3 and j = 1, 2, 3. The body forces and momentum exchange due to inter-phase
interactions require additional supplementary relations. For example, a significant body
force in metallurgical solidification problems is the combined thermal and solutal buoyancy
in the y-direction,
Fb = gβT (T − T0 ) + gβC (C − C0 ) (7.27)
where βT and βC refer to thermal and solutal expansion coefficients, respectively. The values
of T0 and C0 refer to a reference temperature and concentration, respectively.
The interfacial momentum exchange term in the xi direction, Fk,i,int, represents the change
of momentum in phase k due to interactions between both phases along the interfacial boun-
dary. Supplementary relations, for flow through a porous medium may be used to specify
the inter-phase momentum exchange terms. For example, Darcy’s law for liquid flow
through a permeable dendritic solid matrix may be written as:
Kx Fx,int = νl (χ l ur ) (7.28)
where the coefficient Kx denotes the permeability of the porous solid matrix in the
x-direction and ur = ul − us represents the x-component relative phase velocity. Similar
expressions are obtained in the y-direction.
The following Blake–Kozeny equation may be used to calculate the solid permeability in
solid–liquid phase change problems:
χ 3l
K = K0 (7.29)
(1 − χ l )2
This model is based on a physical analogy between interdendritic flow and Hagen–
Poiseuille viscous flow through a noncircular tube with an equivalent hydraulic radius
Solidification and Melting 309
based on the local liquid fraction. As χl→0, then K→0 and the effective viscosity becomes
large so Fint→∞. As a result, the fluid velocity is essentially damped to zero, as expected.
Also, as χl→1, then K→∞ and Fint→0. In this case, the momentum equations have zero
interphase interaction forces and single-phase flow equations are obtained, again
as expected.
Alternatively, to account for interdendritic flow both parallel and perpendicular to inter-
dendritic arms during solidification, the following isotherm gradient model (Naterer and
Schneider 1995) may be used to calculate the interdendritic permeability:
where the flow alignment factor, ξ, represents the component of the velocity vector, v, in
the direction of the temperature gradient, ∇T, relative to the total magnitude of the product
of both quantities,
|v · ∇T |
ξ= (7.31)
|v||∇T |
The permeability components in the axial and cross-flow directions, FA and FC, respec-
tively, are given by:
χ 3l χl
FA = ; FC = (7.32)
(1 − χ l ) 2
1 − χl
Since the primary dendrite arms grow in the direction of the local temperature gradient,
the flow alignment weighting factor represents a scale factor between the axial and cross-
flow permeabilities, respectively, based on the interdendritic velocity direction relative to
the local temperature gradient. For example, ξ = 0 corresponds to a cross-flow condition
and ξ = 1 represents an axial flow.
where the ranges of subscripts are i = 1, 2, 3 and j = 1, 2, 3. It will be assumed that viscous
dissipation and heat source terms are negligible. In general, for multicomponent systems
involving a mushy two-phase region in Figure 7.1, the subscript k refers to the solid (k = 1),
mushy (k = 2) or liquid (k = 3) region. For pure materials, phase change occurs at a single
discrete temperature and there is no two-phase mushy region.
The internal energy of phase k as a function of temperature, ek(T ), consists of two
components—a latent energy term related to the change of phase between the solid and
liquid; and a sensible heat term related to changes of temperature.
T
ek (T) = er,k (T) + cr,k (T)dT (7.34)
T0
310 Advanced Heat Transfer
Here cr,k(T ) represents the effective specific heat of phase k, also a function of temperature,
T. A supplementary equation of state is required so that Equation 7.34 can be written in
terms of temperature alone. Using a piecewise linear approximation,
where the subscript r refers to a reference value. Also, cr,k represents the regular specific
heats in the solid and liquid, but in addition, includes the latent heat of fusion in the two-
phase mushy region (k = 2). The reference energies and temperatures reflect a linear rise
of er,1 in the solid up to er,2 at the phase change temperature, a step change of the latent
energy from er,2 to er,3 in the liquid, and a further linear rise of er,2 with temperature due
to sensible heating in the liquid.
At the solid–liquid phase interface, the rate of heat transfer from each of the liquid and
solid phases into the phase interface consists of conduction and advection components.
The change of energy that accompanies the advance of the interface arises due to the energy
difference between an initially liquid volume, occupying dAdn, and a final solid volume. The
interfacial heat balance can be written as:
dn dT dT
(ρl el − ρs es ) · = −kl + ks + ρl Vl el − ρs Vs es (7.36)
dt dn l dn s
where n, Vs, Vl and dn/dt denote the normal to the interface, individual phase velocities nor-
mal to the interface, and interface velocity, respectively. Alternatively, using Equation 7.25,
dT dT
− kl + ks = −ρs Vs hsl + ρs Vi hsl (7.37)
dn l dn s
where hsl is the latent heat of fusion and vi = dn/dt refers to the interface velocity. If the solid
phase is stationary, then the first term vanishes on the right side of the above interfacial
heat balance.
where μc,k is the chemical potential of constituent c in phase k. For a multi-component mix-
ture of N components, c = 1, 2 … N. In tensor form, Equation 7.23 becomes:
∂ ∂ ∂ k(∂T/∂xj ) ∂ ρμc Dc (∂Cc /∂xj )
(ρs) + (χ k ρk uk,j sk ) = + + Ṗs,k (7.39)
∂t ∂xj ∂xj T ∂xj T
entropy to temperature in each of the phases, including a discrete change between the solid
and liquid phases due to the entropy of fusion.
Also, the Gibbs equation relates the entropy in phase k with other thermodynamic vari-
ables as follows,
N
Tdsk = dek + pdυ − μc,k dCc,k (7.40)
c=1
where υ represents the specific volume. Differentiating the Gibbs equation for an incom-
pressible substance in each phase,
D(χ k ρk sk ) D(χ k ρk ek ) N
D(χ k ρk Cc,k )
T = − μc,k (7.41)
Dt Dt c=1
Dt
where D( )/Dt refers to the total (substantial) derivative, as defined in Chapter 1, including
transient and convective components.
Using a similar procedure as Chapter 3 to derive the positive-definite form of the entropy
production rate, by comparing the differentiated Gibbs equation with the transport form of
the second law in Equation 7.38, it can be shown that the mixture entropy production rate in
phase k can be written as:
∂ ∂T Φ ∂ ∂Cc
Ṗs = k /T 2 + + ρμc Dc /T 2 (7.42)
∂xj ∂xj T ∂xj ∂xj
where Φ refers to the viscous dissipation function. It can be observed that this result is a sum
of squared terms, which ensures positivity of the entropy production, as required by the sec-
ond law of thermodynamics.
At the solid–liquid interface, entropy is produced due to irreversibilities of heat and mass
transfer as well as phase transition. The entropy flux from each of the liquid and solid phases
into the interface consists of diffusion and advection components. The difference between
the interfacial solid and liquid entropy fluxes, together with the entropy produced at the
moving phase interface, balances the entropy change in an interfacial control volume,
dAdn (area multiplied by distance normal to the interface). For a single component system,
the interfacial entropy balance can then be written as:
dn kl dT ks dT dn
(ρl sl − ρs ss ) = − + + ρ Vl sl − ρs Vs ss + ρl Ps,i (7.43)
dt Tl dn l Ts dn s l dt
The interfacial entropy production, Ps,i, designates the entropy produced due to heat
transfer and shear action along the dendrite arms as the dendrite and other microscopic
solidified structures move a distance of dn during the time interval dt.
Using Equation 7.25, the interfacial entropy production rate becomes:
ρs hsf kl dT 1 1
Ps,i = Δsf − + − (7.44)
ρl T ρl Vi dn l Tl Ts
where Δsf = sl − ss is the entropy of fusion. As mentioned previously, the entropy of fusion
for most metals and alloys is approximately constant (Δsf = 8.4 J/mol K) and equal to the
heat of fusion divided by the phase change temperature (Richard’s rule).
312 Advanced Heat Transfer
∂Ts ∂2 T s
ρs cp,s = ks 2 (7.45)
∂t ∂x
∂Tl ∂2 T l
ρl cp,l = kl 2 (7.46)
∂t ∂x
where the subscripts s and l refer to the solid and liquid phases, respectively.
The boundary and interface conditions are given by:
Ts (0, t) = Tw (7.47)
where X refers to the position of the solid–liquid interface. At the phase interface, the heat
balance requires that the difference between the solid–liquid conduction heat fluxes must
balance the latent heat released by the liquid as it solidifies. At x = X,
∂Tl ∂Ts dX
−kl + ks = ρhsl (7.49)
∂x ∂x dt
T(x,t)
T(x,0) = Ti
Ti
Solid
t Liquid
Tf
TW x
FIGURE 7.6
Schematic of phase change in a semi-infinite domain.
Solidification and Melting 313
The phase change problem becomes nonlinear due to the latent heat term in this interfacial
heat balance and the unknown position of the moving interface. It can be shown that the
transient temperature profiles are self-similar so a similarity solution can be constructed.
Define a similarity variable, η, as follows:
x
η = √ (7.50)
2 αs t
where αs is the thermal diffusivity of the solid. Using this similarity variable, all temperature
profiles in time and space can be collapsed onto a single profile in terms of the similarity
variable. The similarity analysis simplifies the solution procedure by introducing a new
variable that reduces the combined transient–spatial dependencies into a single dependence
on the similarity variable alone.
Assume that temperature becomes a function of η alone. Also, the derivatives with respect
to x and t in Equation 7.45 become derivatives in terms of η through the chain rule, thereby
yielding:
d2 Ts dTs
+ 2η =0 (7.51)
dη dη
where C1 is a constant of integration and the error function, erf(η), is given by:
η
2
e−s ds
2
erf (η) = √ (7.53)
π 0
Based on a similar procedure in the liquid region, Equation 7.46 is solved to yield:
At the solid–liquid interface (x = X ), this liquid temperature must balance the solid tem-
perature in Equation 7.52. Equating both temperatures, it can be observed that the numer-
ators in the error function arguments must be proportional to the square root of time, so as to
eliminate the appearance of time for self-similar temperature profiles to match each other at
all values of time. Thus X is proportional to the square root of time.
Alternatively, based on a scale analysis of the heat equation, the functional form of the
moving interface position can be written as:
X = 2β αs t (7.55)
After substituting the derivatives of temperature with respect to x into Equation 7.49,
three equations are obtained for the three unknown coefficients, C1, C2, and β. Solving these
equations and substituting C1, C2, and β back into Equations 7.52 through 7.54,
Tf − Tw x
Ts = Tw + erf √ (7.57)
erf (β) 2 αs t
Ti − Tf x
Tl = T i − √ 1 − √ (7.58)
1 − erf β αs /αl 2 αl t
√ √
hsl β π exp (−β2 ) kl αs /αl (Ti − Tf ) exp −αs β2 /αl
− + √ = 0 (7.59)
cs (Tf − Tw ) erf (β) ks (Tf − Tw ) 1 − erf β αs /αl
2
erf (β) ≈ √ β (7.60)
π
TABLE 7.1
Values of β for Water Where ΔTi = Ti − Tf
(Tf Tw) C ΔT ¼ 0 C ΔT ¼ 1 C ΔT ¼ 2 C ΔT ¼ 3 C ΔT ¼ 4 C
cs (Tf − Tw )
Ste = (7.63)
hsl
The undercooling level is denoted as ΔTw = Tf − Tw. These results were first presented by
Stefan (1891). For problems involving freezing and thawing of the ground (water and
soil), the values of β at different undercooling levels are obtained as: β = 0.054 at ΔTw = 0◦ C;
0.076 at 1◦ C; 0.094 and 2◦ C; and 0.108 at 3◦ C.
For problems with small values of β where the Taylor series in Equation 7.60 was trun-
cated after one term, such as freezing or thawing in the ground, the approximate solution
in Equation 7.62 yields good accuracy. For larger values of β such as metals where β ≈ 1,
the accuracy decreases. However additional terms in the Taylor series can be included to
improve the solution accuracy and therefore extend the applicability of the Stefan solution
to a wide range of other materials.
δ
d dδ dX ∂Tl ∂Tl
Tl (x, t) dx − Ti + Tf − αl + αs = 0 (7.67)
dt X dt dt ∂x δ ∂x X
where the fourth term becomes zero since the temperature is unchanged at x = δ.
Assume the spatial distributions of temperature in the solid and liquid vary linearly and
quadratically with x, respectively, where x refers to the distance from the wall,
Ts = a1 + a2 x (7.68)
Tl = b1 + b2 x + b3 x2 (7.69)
At least a second order interpolation is required in the liquid to match the required slope of
temperature at the interface, x = X, as well as a zero temperature slope in the liquid at x = δ.
316 Advanced Heat Transfer
The unknown coefficients, from a1 to b3, can be determined from the interfacial and boun-
dary conditions, where, Ts = Tw at x = 0, Ts = Tf = Tl at x = X, and Tl = Ti at x = δ. Finding
these unknown coefficients and substituting back into the temperature profiles,
x
Ts = Tw + (Tf − Tw ) (7.70)
X
x−X x−X 2
Tl = Tf + 2 (Ti − Tf ) − (Ti − Tf ) (7.71)
δ−X δ−X
Tf 2Ti dX dX 2αl (Ti − Tf ) dX
+ (p − 1) − Ti p + + Tf =0 (7.72)
3 3 dt dt (p − 1)X dt
where p = δ/X.
Based on a scale analysis of the heat equations in the liquid and solid regions, the functional
forms of the moving interface position and thermal penetration depth can be written as:
X = 2β αs t (7.73)
δ = 2ψ αl t (7.74)
where β and ψ are unknown constants. Substituting this functional form of the interface posi-
tion into Equation 7.72, it can be shown that:
1 1 dX αs 2kl (Ti − Tf )
+ = 1− (7.75)
Ste 2 dt X ks (Tf − Tw )(p − 1)
Solving this equation subject to X = 0 at t = 0 yields the same functional form of the inter-
face position as Equation 7.73, which is a useful verification of the initial assumption of
Equation 7.73. The parameter β is given by:
p2
p22 − 4p1 (Ste)2
β2 = − − (7.76)
2p1 2p1
and,
2kl αs (Ti − Tf )Ste
p1 = (2 + Ste) 2 + + Ste (7.77)
ks αl (Tf − Tw )
4kl2 αs (Ti − Tf )2 Ste2 kl αs (Ti − Tf )Ste
p2 = − − 2Ste 2 + + Ste (7.78)
3ks2 αl (Tf − Tw )2 ks αl (Tf − Tw )
Solidification and Melting 317
Once the parameter β is known, then the interface position, X, can be determined, fol-
lowed by the temperature profiles in both the liquid and solid regions.
This approximate integral solution agrees well with the exact solution using the β param-
eter in Equation 7.59. The exact and approximate β values differ by less than 1% for Stefan
numbers of less than 0.1, with somewhat larger errors at higher Stefan numbers. The heat
balance integral method provides a useful approximate method of analyzing phase change
problems over a range of Stefan numbers. Higher accuracy in the integral solution can be
obtained by using higher order interpolations of the assumed temperature profiles in the
integral solutions. However, there is a trade-off between accuracy and solution complexity
since higher order profiles lead to additional complexity of equations that may no longer
allow closed-form solutions.
d2 T U dT
− =0 (7.79)
dx2 α dx
where α is the thermal diffusivity. In the liquid region, α represents the liquid thermal dif-
fusivity, while in the solid, it represents the solid thermal diffusivity. The solution of the heat
equation in the solid, subject to Dirichlet conditions, T(x1) = T1 and T(x2) = T2, may be
expressed as follows:
where x1 ≤ x ≤ xm. The subscript m denotes the melting point at the phase interface. In the
liquid,
where xm ≤ x ≤ x2. The coordinate of xm represents the (unknown) position of the phase
interface.
To determine the interface location, the following heat balance at the phase interface is
applied:
dT dT
− k + k = ρUhsl (7.82)
dx x+m dx x−m
T2 − Tm Tm − T1 hfg
+ = (7.83)
1 − exp(U(x2 − xm )/α) 1 − exp(−U(xm − x1 )/α) cp
Solving this equation for the unknown interface position, xm, leads to:
T2 − T1 1 2 1 T2 − T1 exp (Pe)
− ξ + (1 + exp (Pe)) − exp (Pe) − +1 ξ− = 0 (7.84)
Tm − T1 Ste Ste Tm − T1 Ste
where,
U(xm − x1 )
ξ = exp (7.85)
α
U(x2 − x1 ) cp (Tm − T1 )
Pe = ; Ste = (7.86)
α hsl
Here Pe refers to the Peclet number. The root solution of Equation 7.84 lies in the interval of
0 ≤ ξ ≤ 1. Results and applications of this model were presented by Pardo and Weckman
(1990) in welding applications. The results indicate a sharp difference in the temperature
gradients across the phase interface, particularly for low Stefan numbers. Sharp changes
in curvature of the temperature profile occur at the phase interface.
The left side represents the incoming solute fluxes due to diffusion at a position x and
advection at location x + dx. The control volume moves rightward and thus solute in the liq-
uid is advected into the control volume due to the relative velocity of the domain and sta-
tionary liquid. Similarly, the right side represents the outgoing solute fluxes. Expanding the
terms at x + dx using a Taylor series,
d d ′′
j′′x ′′
+ UC + (UC)dx + . . . = j x + (j x )dx + . . . + UC (7.88)
dx dx
Using Fick’s law for the diffusion fluxes and neglecting higher order terms,
d2 C U dC
+ =0 (7.89)
dx2 D dx
where the mass diffusivity is D. Solving this equation for the solute concentration in the
liquid, CL, ahead of the phase interface (i.e., xi measured outward from the interface),
C0 U
CL = exp − xi (7.90)
κ D
where C0 is the initial composition and κ represents the solute–solvent partition coefficient
of the phase diagram, or ratio of the solidus to liquidus slopes in Figure 7.1. The result indi-
cates that the liquid concentration decreases exponentially ahead of the phase interface. The
characteristic decay distance is D/U, or in other words, D/U is the distance in which the liq-
uid concentration falls to 1/e of its interfacial value.
Effective control of a moving solidification or melting front to achieve a constant or near-
uniform velocity has significance in a number of materials processing technologies, such as
metallurgical casting, welding, zone refining, and others. For example, Figure 7.7 illustrates
various types of casting processes where effective control of the interface velocity is a critical
factor in the resulting thermomechanical properties of the solidified materials.
Wire casting is commonly used in the production of wire-type products. Molten metal is
supplied from a chamber as a thin liquid stream that is cooled and solidified in the form
of a thin wire or similar product. In thin sheet or foil casting processes, medium- to high-
thickness sheet castings of metal alloys are produced. Here a molten stream is supplied
from a ladle and forced to flow through two counter-rotating rollers maintained at a
specific temperature. The quality of the solidified sheet or foil is closely related to the roller
surface characteristics, homogeneity of the supplied molten liquid, and rate of nucleation
and interface movement in the solidifying material. In continuous casting solidification in
the production of various types of components such as bars and slabs, the molten material
is poured into open-ended cooled molds. This process differs from centrifugal casting
(used for pipes with molten material poured into a rapidly rotating mold by a centrifuge)
and die casting (molten material forced under pressure into a die).
During a solidification process, contraction upon phase change often occurs due to the dif-
ferences in density between the solid and liquid phases. The onset of shrinkage voids can be
identified by the following Niyama factor (Niyama et al. 1982):
∂T/∂n
Ny = √ , Nycrit (7.91)
∂T/∂t
320 Advanced Heat Transfer
Molten metal
Molten
metal
Downflow
Heat
flow
Liquid
stream
Liquid
Heat
flow Double
roller Solid
Wire
Thin sheet,
Wire casting foil casting Continuous casting
FIGURE 7.7
Wire, foil, and continuous casting processes.
where n is the direction perpendicular to the moving solid–liquid interface. Here Nycrit refers
to a critical Niyama number, below which shrinkage voids may likely form. For example,
Nycrit = 1.0 (soC)1/2/mm for high-nickel alloys. The local time and spatial gradients of tem-
peratures are evaluated near the end of a solidification process when shrinkage voids are
formed. Metals and alloys usually contract upon solidification and shrinkage voids are cre-
ated when interdendritic channels are closed to liquid inflow at the phase interface. Shrink-
age voids are generally undesirable since they often lead to defects in the material’s
mechanical properties.
Zone refining is another important solidification technology for crystal growth and mate-
rial purification (Pfann 1958). Movement of a molten zone along the length of an initially
solid material transfers impurities to an opposite end of the original material. For example,
as a heater moves upward with a metal bar in a cylindrical container, a finite molten zone
forms along the bar and moves upward with the heater while the lower section resolidifies.
Due to different constituent solubilities in the solid and liquid phases, solute is transferred
into the molten region at the phase interface thereby creating a more purified material in the
solid. The net effect is a solute transfer from the original end to the opposite end of the metal
bar. This zone refinement process can be repeated several times to further purify the
material.
Effective control of a melting front is also critical in gas metal arc welding. Liquid metal is
deposited into a weld pool from an electrode, usually moving at a constant velocity, U (see
Figure 7.8). An important design parameter is the weld geometry, which consists of the weld
profile shape and size. This geometry can be determined by first solving the heat equation
for the weld temperature and then adjusting the weld geometry so that the heat balance is
satisfied for the newly solidified weld pool. An iterative procedure is required until conver-
gence is achieved between the heat balance, heat losses by convection and radiation, and
resulting weld geometry.
Solidification and Melting 321
Electrode
U
Liquid metal
deposited into weld
y pool from electrode
FIGURE 7.8
Schematic of gas metal arc welding.
The welding speed, U, and resulting velocity of the melting front have significant effects
on the melt pool and weld properties. For example, equiaxed dendrites and columnar struc-
tures are usually formed under high-velocity conditions. Also, a high-temperature gradient
occurs within the material if the temperature of the molten metal in the weld pool is much
larger than the base metal temperature. This often leads to cellular or columnar microstruc-
tures in the welded materials.
where Xsol , x , Xliq and χe refers to the solid fraction at the eutectic composition. The liquid
fraction, χl, is 100% minus the solid fraction. The linear approximation assumes that the liq-
uid fraction increases linearly with position between the solidus and liquidus interfaces.
This model is reasonably accurate for metal alloys. But other materials may require different
functional forms. For example, in soil mixtures, an exponential change of solid fraction
through the two-phase range more closely approximates soil–water data.
The heat conduction equations in the solid and liquid regions are the same as prior
Equations 7.45 and 7.46. However, an additional heat equation is required in the
two-phase (mushy) region:
where the subscript sl refers to the two-phase region (i.e., solid and liquid coexisting simul-
taneously in equilibrium). This heat equation will be solved subject to the interfacial condi-
tions of Ts = Tsol = Tsl at x = Xsol and Tsl = Tliq = Tl at x = Xliq (see Figure 7.9).
322 Advanced Heat Transfer
T(x,t)
Two-phase region T(x,0) = Ti
Ti
Solid
t Liquid
Tsol
Xliq(t) = liquidus interface position
Tliq
FIGURE 7.9
Schematic of solidification in a two-phase region.
Also, the interfacial heat balance, Equation 7.49, is modified to the following two heat
balances at the solidus and liquidus interfaces, respectively:
∂Tsl ∂Tl
−ksl = −kl (7.95)
∂x ∂x
The latent heat absorbed at the liquidus interface is modeled within the two-phase region
by Equation 7.93 and thus does not appear in the interfacial heat balance in Equation 7.95.
Unlike a pure material with a single phase interface, a binary mixture requires tracking of
both the solidus and liquidus interfaces. Extending Equation 7.73 to two interfaces,
Xsol = 2βsol αs t (7.96)
Xliq = 2βliq αs t (7.97)
Using a similarity solution method (Cho and Sunderland 1969) or heat balance integral
method (Tien and Geiger 1967), the following results are obtained in the solid, two-phase
(mushy), and liquid regions, respectively:
√
Ts − Tw erf x/ 2 αs t
= (7.98)
Tsol − Tw erf (βsol )
√
Tsl − Tw hsl χ e −βsol + x/2 αs t
=1−
Tsol − Tw cl (Tsol − Tw )(βliq − βsol )
⎡ √ √ ⎤
Tliq − Tsol + hsl χ e /cl erf x/2 αl t − erf βsol αs /αl
⎣ √⎦
+
Tsol − Tw √ (7.99)
erf β αs /α − erf β
liq αs /α l sol l
Solidification and Melting 323
√
Tl − Ti 1 − erf (x/2 αl t)
= √ (7.100)
Tliq − Ti 1 − erf (βliq αs /αl )
√
kl αm /αl exp (−β2liq αs /αl ) Tliq − Tsol + exp (−β2liq αm /αs )hsl χ e /cm
(Tliq − Ti ) √ + √ √
km 1 − erf −βliq αs /αl erf βliq αs /αl − erf βsol αs /αl
√
hsl χ e παm /αl
= (7.101)
2cm (βliq − βsol )
Further cooling below the eutectic temperature in a binary system leads to simultaneous
growth of two or more phases through solid state transformations. Eutectics are com-
posed of more than one solid phase. These phases can exhibit a range of geometrical
arrangements.
∂2 T ∂2 T
+ =0 (7.102)
∂x2 ∂y2
∂2 T ∂2 T
+ + εf (T) = 0 (7.103)
∂x2 ∂y2
where f(T ) is a nonlinear function of temperature and 0 ≤ ϵ ≤ 1. It can be shown that solu-
tions of Equation 7.103 may be obtained by adding small perturbations to the solution of
Equation 7.102 through a power series expansion:
The individual coefficients are then obtained by substituting this temperature series into
the heat equation and equating the same powers of coefficients to each other, thereby lead-
ing to a sequence of problems that is solved successively.
Consider a perturbation solution for solidification due to wall cooling and freezing of a
flowing liquid along the wall. The flowing liquid comes into contact with a chilled surface
324 Advanced Heat Transfer
at a specified temperature, Tw, below the phase change temperature, Tf (see Figure 7.10). The
heat conduction equation in each of the resulting solid and liquid phases is given by:
∂T ∂2 T
=α 2 (7.105)
∂t ∂x
A Dirichlet boundary condition is applied at the wall, T(0,t) = Tw. At the phase interface
(x = X ), the temperature is T(X,t) = Tf and the interfacial heat balance is given by:
∂T dX
k − h(T1 − Tf ) = ρhsl (7.106)
∂x X dt
x X h(T1 − Tf )
x∗ = ; X∗ = ; t∗ = t (7.108)
Xs Xs ρhsl Xs
T − Tw cp (Tf − Tw )
θ= ; Ste = (7.109)
Tf − Tw hsl
Rewriting Equation 7.105 and the boundary and interface conditions in terms of these
dimensionless variables,
∂2 θ ∂θ dX∗
= Ste (7.110)
∂x∗2 ∂X∗ dt∗
Tf Liquid
FIGURE 7.10
Schematic of freezing with a flowing liquid.
Solidification and Melting 325
dX∗ ∂θ
= ∗ −1 (7.111)
dt∗ ∂x X∗
Using the perturbation solution method, the temperature and non-dimensional interface
velocity are expanded in the form of a power series about the perturbation parameter, Ste, as
follows:
Setting the terms with the same powers of Ste equal to each other, the following sequence
of equations is obtained:
∂2 θ0
=0 (7.116)
∂x∗2
∂2 θ 1 ∂θ0 ∂θ0
= −1 (7.117)
∂x∗2 ∂X∗ dx∗
∂2 θ 2 ∂θ1 ∂θ0 ∂θ0 ∂θ1
∗2
= ∗ ∗
−1 + ∗ ∗ (7.118)
∂x ∂X dx ∂X ∂x
Additional higher order equations are obtained for higher order terms involving Ste. Each
problem can be solved successively based on the solution from the previous problem. For
example, Equation 7.116 and its boundary conditions are solved to give θ0. Then, substitut-
ing the differentiated θ0 into Equation 7.117, the next solution of θ1 is obtained. The differ-
entiated θ0 and θ1 expressions are then substituted into Equation 7.118 to yield the solution
for θ2, and so forth.
Also, individual terms in the power series in Equation 7.114 involving temperature are
equated to corresponding terms in the power series of the interface position terms,
dX0∗ ∂θ0
= −1 (7.119)
dt∗ ∂x∗ X∗
326 Advanced Heat Transfer
dX1∗ ∂θ1
= −1 (7.120)
dt∗ ∂x∗ X∗
x∗
θ0 = (7.121)
X∗
X∗ − 1 ∗2
θ1 = (x − X∗2 )x∗ (7.122)
6X∗3
1 − X∗ 3 − 2X∗ x∗5 x∗3 X∗2 19
θ2 = + − − X x∗
∗
(7.123)
12X∗3 X∗2 10 3 5 6
The following results are obtained for the first few cases:
1
t∗1 = t∗0 (7.126)
3
1
t∗2 = − (3X∗2 + 2X∗ + 2 ln (1 − X∗ )) (7.127)
90
Once individual terms are substituted and assembled into Equation 7.124, a final resulting
equation is obtained for the interface position, X*, with respect to time, t*, after which the
temperature can also be determined. Further details, results, and discussion of this pertur-
bation solution and others involving forced convection freezing are presented by Seeniraj
and Bose (1982) and Lunardini (1988).
∂2 Ts
=0 (7.128)
∂x2
∂Ts dX
k = ρhsl (7.129)
∂x X dt
∂Ts
k = h(Ts (0, t) − T1 ) (7.130)
∂x 0
Also, at the solid–liquid phase interface and initially, Ts(X,t) = Tf = Ts(x,0). In the quasi-
stationary approximation, transient effects are included in the interfacial heat balance, Equa-
tion 7.129, but not the heat equation in the solid, Equation 7.128.
Solving Equation 7.128 subject to Equations 7.129 and 7.130,
x + k/h
Ts = T1 + (Tf − T1 ) (7.131)
X + k/h
T(x,t)
Solid
Tf
Liquid
t
FIGURE 7.11
Schematic of a solidified layer with convective cooling.
328 Advanced Heat Transfer
The temperature in the solid can then be obtained by substituting this result into Equation
7.131. Sample results of temperature profiles and depth of the freezing front with time over a
range of Stefan numbers are shown in Figure 7.12. The surface temperature decreases and
the freezing front advances further at higher Stefan numbers.
Two thermal resistances affect the heat flow to the moving phase interface: a convection
resistance which remains constant; and a conduction resistance which increases with time as
the frozen layer grows. After a sufficient time period has elapsed, the solidification proceeds
essentially as a constant surface temperature case since the relative effect of the surface resis-
tance approaches zero. The results can also be used for melting if the properties of the
thawed material are used instead of the liquid and the sign of the latent heat term in the
interfacial heat balance is changed (to negative) since latent heat is absorbed, rather than
released at the phase interface.
The quasi-stationary approximation provides good accuracy at small Stefan numbers. But
the accuracy declines at higher Stefan numbers. The lower magnitude of the latent heat, rel-
ative to sensible cooling, increases the interface velocity, thereby allowing less time for ther-
mal equilibrium in the solid, compared to cases with low Stefan numbers. Further detailed
analysis and results of freezing of liquids by convective cooling were reported by London
and Seban (1943).
1.0 1.0
(Tw–Ta)/(Tf –Ta)
Ste = 10
h X/kl
Ste = 4
0.8 Ste = 2
Ste = 0.04 Ste = 1
Ste = 0.5
0.6
0.1 Ste = 0.2
1 Ste = 0.1
0.2 Ste = 0.04
0.4
2
10
0.2
0.0 0.1
0.01 0.1 1 10 100 1000 0.001 0.01 0.1 1 10
α h2t/kl2 αl h2t/kl2
FIGURE 7.12
(a) Surface temperature and (b) depth of solidification in a semi-infinite domain with a convection boundary con-
dition. (Adapted from F. Kreith and F.E. Romie. 1955. Proceedings of the Physical Society B, 68: 277–291.)
Solidification and Melting 329
in semi-infinite domains or early stages of time where end effects in a finite domain have not
yet propagated back to the phase interface. This section will apply the frozen temperature
approximation to a solidification problem with convection in a semi-infinite domain.
Consider freezing of a liquid at a temperature of Tf in contact with a wall subjected to con-
vective cooling from a surrounding gas at T∞, where T∞ , Tf. Foss and Fan (1972) and
Lunardini (1988) have analyzed this problem and other similar configurations using both
the quasi-stationary and frozen temperature approximations. As the phase interface moves
into a semi-infinite domain of liquid, the temperature gradient at the liquid side of the phase
interface is assumed to remain constant in the frozen temperature approximation.
The governing, boundary, initial, and interfacial conditions in the solid are given by:
∂2 Ts
=0 (7.134)
∂x2
∂Ts
−ks = h(Ts (0, t) − T1 ) (7.135)
∂x 0
∂Ts ∂Tl dX
ks − kl = ρhsl (7.136)
∂x X ∂x X dt
where the subscripts s and l refer to solid and liquid regions, respectively. At the solid–liquid
phase interface and initially, Ts(X,t) = Tf = T(x,0). Using the frozen temperature approxima-
tion, the second term is assumed to be constant in Equation 7.136,
∂Tl
q′′w = −kl (7.137)
∂x X
Differentiating this expression with respect to x and substituting the result along with the
constant interfacial heat flux into Equation 7.136 yields:
dX h(Tf − T1 ) q′′
= − w (7.139)
dt ρhsl (1 + hX/ks ) ρhsl
Solving this equation subject to the initial condition, X(0) = 0, yields (Lunardini 1988):
′′
hX hq w h(Tf − T1 ) h(Tf − T1 ) − q′′ w
+ t= ln (7.140)
ks ks ρhsl q′′ w h(Tf − T1 ) − q′′ w (1 + hX/ks )
This result gives an implicit closed-form solution for the movement of the phase interface,
X, with time. Sample results of the depth of solidification for constant and sinusoidal
ambient temperatures are illustrated in Figure 7.13. The dimensionless heat flux is defined
as q* = q/(h ΔT). Once the interface position is known, the temperature distribution can be
obtained from Equation 7.138.
330 Advanced Heat Transfer
(a) (b)
6 12
h X/kl
4 8
0.0627
3 3
Ta = –15(1 – cos(0.0112t*))
2 +2.5(1 – cos(1.0828t*))
4
0.0314 qw* = 0.0127
1 2
0 0
0 100 200 300 400 500 0 100 200 300 400 500
t* = h2 t/(kl ρl cl) t* = h2 t/(kl ρl cl)
FIGURE 7.13
Depth of solidification for (a) a constant and (b) sinusoidal ambient temperature. (Adapted from S.D. Foss and S.S.T.
Fan. 1972. Water Resources Research, 8: 1083–1086.)
Setting the left side of Equation 7.139 to zero and solving for X on the right side yields the
following steady-state position of the solid–liquid interface:
Tf − T1 ks
Xmax = ks − (7.141)
q′′ w h
As expected, this steady-state position is larger at lower ambient fluid temperatures, T∞,
since more convective cooling occurs from the boundary of the ambient fluid.
T(r,t)
Tf Solid
ri Tf
t
Ti
R(t) r
FIGURE 7.14
Schematic of outward cylindrical phase change.
r2
ξ= (7.144)
4αt
d2 T dT
ξ 2
+ (1 + ξ) =0 (7.145)
dξ dξ
In terms of the transformed variable, ξ, the initial condition becomes T (ξ→∞) = Ti.
The solution of this problem was reported by Carslaw and Jaeger (1959) and Lunardini
(1988). Solving Equation 7.145 in terms of the new variable, ξ,
where the constant of integration, C, is determined from boundary conditions. The exponen-
tial integral, E1(ξ), is defined by
1
e−t dt
E1 (ξ) = (7.147)
ξ t
332 Advanced Heat Transfer
Using the expected functional form of the moving phase interface as determined previ-
ously in Equation 7.55,
√
R = 2β αt (7.148)
Differentiating Equations 7.148 and 7.146, and substituting the results into the interfacial
heat balance, Equation 7.143, leads to:
Once β is obtained, the interface position and temperature distribution can be determined
and rewritten in terms of the original coordinates, r and t. In this problem, the latent heat
released by the solidifying liquid is transferred to the liquid at the phase interface, rather
than the solid, thereby raising the temperature of the subcooled liquid to Tf. This heat trans-
fer into the subcooled liquid sustains the movement of the phase interface into the remainder
of the subcooled liquid.
The initial, boundary, and interfacial conditions are given by: T(r . ro, 0) = Tf ; T(ro, t) =
Tw; T(R, t) = Tf and
∂T ∂R
k = ρL (7.151)
∂r R ∂t
Using the frozen temperature approximation, this equation can be solved subject to the
initial condition of R = ro at t = 0,
k ∂T
R(t) = t + ro (7.152)
ρL ∂r R
Assume a logarithmic temperature profile in the solid, rather than a linear or quadratic
profile used in previous integral solutions for planar geometries, since the area in the
path of heat transfer is increasing with radial position. A logarithmic profile that satisfies
the boundary and interfacial conditions is given by:
T − Tw ln (r/ro )
= (7.154)
Tf − Tw ln (R/ro )
where Tw is the constant tube wall temperature. If this wall temperature varies with time,
Kreith and Romie (1955) presented a series solution to represent the temperature in the solid.
Substituting the temperature profile in Equation 7.153 and solving for the interface posi-
tion, R (Lunardini 1988),
Ste 1
2n ( ln R∗ )n 1 1 1
∗2
R −1− + R∗2 ln R∗ − R∗2 + = t∗ (7.155)
4 n=1
nn! 2 4 4
where R* = R/ro and t∗ = α(Ste)t/r2o are the dimensionless interface position and time,
respectively. Using this solution for the interface position, the temperature can then
be determined.
This solution method can also be extended to inward freezing of a liquid. The governing
equations are again given by Equations 7.150 and 7.151 but the interface moves in the neg-
ative r-direction (inward) and the temperature gradient in Equation 7.151 becomes a nega-
tive constant. Also, for inward phase interface movement, the radial position of the interface,
r = R, decreases with time. An example of inward moving phase change is provided below.
The time required to completely freeze the liquid can be estimated by setting the right
side to zero (i.e., phase interface reaches the center of the pipe). Then solving for the result-
ing solidification time in terms of the interfacial temperature gradient and thermophysical
properties,
ρhsl ro
tf = (7.157)
k(dT/dr)R
As expected, the solidification time increases for materials with a larger latent heat of
fusion or lower thermal conductivity.
334 Advanced Heat Transfer
ρhsl r2o
tf = (7.159)
4k(Tf − Tw )
It can be observed that this result is similar to Equation 7.157 but with a specific expres-
sion for the temperature gradient based on the quasi-stationary approximation of the heat
conduction equation.
Also, kl, αl, ks and αs are the thermal conductivity and thermal diffusivity of the liquid and
solid phases, respectively. The constant λ is determined from the following transcendental
equation:
∂Tl
q′ = −2πrkl (7.164)
∂r
The expected functional form of the moving phase interface, based on earlier examples, is
given by:
R(t) = 2λ αs t (7.165)
Duan and Naterer (2010) analyzed this problem configuration and presented results over
a range of thermal conditions in relation to a cylindrical battery cell or heater surrounded by
phase change material (PCM) in an electric vehicle. The heating strength of the line heat
source represents a battery cell and can be calculated based on the total heat generation
rate, Q, divided by the battery length L. The PCM is initially at a solid state and then starts
melting when the battery cell generates heat.
ln (r/ro )
Ts = Tw + (Tf − Tw ) (7.167)
ln (R/ro )
ln (r/δ)
Tl = Ti + Tf (7.168)
ln (R/δ)
where the subscripts s, l, f, w, and i refer to solid, liquid, fusion (phase change temperature),
wall, and initial, respectively.
336 Advanced Heat Transfer
The total heat flow through the wall, per unit length of pipe, consists of two components:
sensible cooling due to temperature changes between the liquid and wall temperatures; and
latent heat released at the interface to freeze the material.
R
δ
Q′tot = ρcs (Ts − Tf )2πr dr + ρcl (Tl − Tf )2πr dr + ρπ R2 − r2o (hsl + cl (Tf − Ti )) (7.169)
ro R
Also, the rate at which heat is released by the freezing front balances the rate of heat loss
through the wall,
′ dTs
Q̇tot = −ks (2πro ) (7.170)
dr ro
The differentiated temperature from Equation 7.167 is substituted into Equation 7.170.
Also, using the above heat balances, the following implicit solution can be obtained for
the interface position (Lunardini 1988),
2
R 2 R R R 2 R R
Ste + γ − 1 + ln ln − Ei 2 ln +M 2 ln +1−
ro ro ro ro ro ro
4αSte
= t (7.171)
r2o
where,
2
cl Tf − Ti δ − R2 − 2R2 ln (δ/R)
M = 1 + Ste + 1 (7.172)
cs Tw − Tf 2R2 ln (δ/R)
Then the resulting temperature distributions can be determined from Equations 7.167 and
7.168. The current analysis was based on the quasi-steady approximation in Equation 7.166
for low Stefan numbers. Better accuracy at higher Stefan numbers can be obtained through a
heat balance integral solution method but at the cost of additional complexity of equations
that may no longer allow a closed-form solution.
The one-dimensional heat equation in spherical coordinates in the radial direction (see
Appendix C) is given by:
1 ∂T 1 ∂ 2 ∂T
= r (7.173)
α ∂t r2 ∂r ∂r
Define the following change of variables, similarly as Equation 7.144 in cylindrical coor-
dinates,
r
ξ = √ (7.174)
2 αt
T = A + BF(ξ) (7.176)
where A and B are constants of integration, determined from the boundary and initial con-
ditions. Also, F(ξ) is the spherical function defined by:
√
1 π
F(ξ) = exp (−ξ ) −
2
(1 − exp (ξ)) (7.177)
2ζ 2
Specific forms of Equation 7.176 for the temperature field can be obtained once appropri-
ate boundary and initial conditions are applied.
Assume the center nucleation site and solid phase remain at Tf. The interfacial and initial
conditions are given by: T(R, t) = Tf ; T(r, 0) = Ti and,
∂T dR
− kl = ρhsl (7.178)
∂r R dt
The constants of integration in Equation 7.176, A and B, can be determined based on the
initial and interfacial temperatures, Ti and Tf, respectively, thereby yielding the
temperature profile.
Also, using the expected functional form of the interface position as obtained previously in
Equation 7.55,
√
R = 2β αt (7.179)
The coefficient β can be determined by substituting the interface position and temperature
into the interfacial heat balance, Equation 7.179, yielding (Lunardini 1988):
F(ξ)
T = Ti + (Tf − Ti ) (7.180)
F(β)
338 Advanced Heat Transfer
where,
√ √ 1
β2 exp (β2 ) exp (−β2 ) − π β + π βerf (β) = Ste (7.181)
2
TABLE 7.2
Melting Temperatures of Common PCMs
Non-Congruent PCMs Congruent (Eutectic) PCMs
PCMs have also been applied to deicing of structures. For example, concrete slabs in brid-
ges can use PCMs to prevent freezing on the bridge before it occurs on adjacent roads lead-
ing to the bridge. The phase change material may be stored in a pellet form in the concrete
such that it stores daytime heat and releases thermal energy at freezing temperatures. Icing
on the bridge would be delayed relative to icing on the adjacent road. In addition to reduc-
ing potential accidents due to iced roads, the PCMs could also lead to other benefits such as
fewer potholes on the bridge and less cracking due to repeated freezing and thawing cycles.
PROBLEMS
7.1 Find the number of phases that coexist in equilibrium for a binary component mixture
containing solid and liquid phases. Use the Gibbs phase rule to find the number of
independent thermodynamic variables in the pure phase and two-phase regions.
7.2 A liquid is cooled to a temperature below its equilibrium phase change temperature.
Use the Gibbs free energy to determine the critical radius of a crystal that initiates
homogeneous nucleation in the solidifying liquid.
7.3 A hemispherical-shaped solid nucleus with a spherical radius of curvature, r, is
formed in contact with a smooth mold surface. An angle of contact, θ, is formed
between the tangent to the surface and the solid–liquid interface. Using appropriate
force balances at the interfaces between the solid and liquid, explain how to find the
critical undercooling required for the onset of heterogeneous nucleation of a crystal in
terms of the angle of contact.
7.4 Find the governing equation for solute diffusion and redistribution at the phase inter-
face during solidification in a binary component mixture. Consider diffusion in the
liquid and solute redistribution at the phase interface according to the phase equilib-
rium diagram. Solid diffusion and convection may be neglected. Express the result in
terms of a uniform interface velocity, U, and the liquid mass diffusivity.
7.5 Consider the towing of an iceberg across the ocean to provide a fresh water
supply. An iceberg has a relatively flat base 700 m long, 600 m wide and a volume
of 9 × 107 m3. What power is required to tow this iceberg over a distance of
7,200 km if at least 90% of the ice mass is retained (not melted) at its destination?
Assume that the average temperature of ocean water during the voyage is 13◦ C. Con-
sider only friction and heat transfer along the flat base in the analysis. State the
assumptions used in the solution.
7.6 The nucleation rate can be interpreted as the probability of adding atoms to a stable
cluster multiplied by the number of clusters reaching a critical size to sustain the
phase change. Derive Equations 7.13 or 7.14 by using the statistical definition of
entropy to find the number of equally probable quantum states of an atom in the
stable cluster. Entropy, S, may be defined as S = −κΣi Ωi ln(Ωi) where Ωi is the prob-
ability of quantum state i and κ is Boltzmann’s constant.
7.7 What differences are encountered between the solidification of metals as compared
with polymers?
7.8 In the liquid region of a binary component mixture ahead of an advancing solid–
liquid interface during phase change, the mean concentration of solute is C0. Define
K as the ratio of solute concentrations in the solid and liquid phases. Show that the
340 Advanced Heat Transfer
solute concentration decreases exponentially with distance ahead of the phase inter-
face. Apply suitable boundary conditions and express your answer in terms of C0, K,
interface velocity, R, and the liquid mass diffusivity, Dl. How would this analysis
change if a ternary mixture was used instead?
7.9 The mixture form of the conservation of mass equation was obtained in this chapter
as a special case of the general scalar conservation equation. Derive this mixture equa-
tion of mass conservation for a control volume occupied by two phases (solid and liq-
uid) by performing a mass balance in each phase individually and then summing
over both phases. In the derivation, use an averaged interfacial velocity, Vi, and a vol-
ume occupied by phase k, Vk, based on the phase fraction, χk, multiplied by the total
volume, V.
7.10 For a binary alloy, the energy and entropy equations of state in the two-phase region
depend on temperature and species concentration. Derive the energy and entropy
equations of state of a binary component mixture by integrating the Gibbs equation
through the phase change region. Assume that the liquid fraction varies linearly
with temperature.
7.11 Liquid metal at an initial temperature of Ti in a large tank is suddenly exposed to a
chilled surface at Tw, below the phase change temperature, Tf. Solidification begins
at this surface and the solid–liquid interface moves into the liquid over time. Assume
that spatially linear profiles may be used to approximate the temperature profiles in
the solid and liquid and the ratio of profile slopes is approximately constant in time.
Derive an expression for the change of interface position with time in terms of ther-
mophysical properties, Tf, Tw and the ratio of slopes of temperatures in the liquid
and solid.
7.12 A large tank contains liquid at a temperature of Ti, above the phase change tem-
perature, Tf. At t = 0, the surface temperature of the tank drops to Tw, where Tw , Tf.
Solidification begins and the solid–liquid interface moves linearly in time. Can
a similarity transformation such as Equation 7.50 be used to find the interface
velocity in terms of thermophysical properties and the above temperatures? Explain
your response by reference to a one-dimensional transient solution of the problem.
7.13 A liquid alloy undergoes phase change over a range of temperatures between the
liquidus and solidus temperatures. During solidification, a layer of material releases
latent heat that can be characterized by a heat source that varies linearly across the
layer, equal to βx, where β is a constant. A temperature of T1 is imposed at the left
edge (x = 0). Also, a specified heat flux, q = qw, is applied at the right edge of the layer
(x = L).
a. Define the heat conduction equation and boundary conditions for this one-
dimensional problem.
b. Solve the governing equation to find the temperature variation across the
layer of material. Express your answer in terms of β, L, qw, and the effective
conductivity of the material, k.
c. What values of β are required to maintain a unidirectional flow of heat at the
left boundary?
7.14 Consider a region of liquid initially at the phase change temperature (Tf) that sud-
denly begins freezing outwards from a wall due to cooling at a uniform rate through
the wall at x = 0.
Solidification and Melting 341
7.20 A tube is submerged in liquid at an initial temperature of Ti, above the phase change
temperature, Tf. Then coolant is passed through the tube and outward freezing of the
liquid is observed over time. The wall temperature of the tube and the temperature
throughout the solid can be assumed to be uniform at Tf. Outward directional solid-
ification occurs in the r-direction. Assume that the frozen temperature approximation
can be used with a constant temperature gradient in the liquid at the phase interface
over time.
a. Use a one-dimensional, quasi-stationary approximation to find the tempera-
ture distribution in the liquid. Express your answer in terms of Tf, R, thermo-
physical properties, and the interface velocity.
b. At what position does the liquid temperature exceed Tf by 10◦ C?
7.21 A liquid surrounding a pipe of radius ri is initially held at a temperature of Ti
(above the phase change temperature, Tf). Then the wall of the pipe is cooled and
its temperature is kept constant at Tw, where Tw , Tf. The liquid begins to freeze
progressively over time and heat is conducted inward radially through the liquid.
Use a one-dimensional, quasi-stationary analysis to obtain the rate of interface
advance with time, R(t), as well as temperatures in the solid and liquid phases.
7.22 Inward phase change from a cylindrical surface occurs during freezing of water in an
annular region between two pipes. The initial liquid temperature is Ti (above the
phase change temperature, Tf). The outer pipe surface of radius ro is cooled at a uni-
form rate to initiate freezing of water between the pipes. Assume the heat flux from
the liquid at the phase interface changes with inverse proportionality to the interface
position, R(t).
a. Derive expressions for the spatial variation of temperature in the solid using a
one-dimensional, quasi-stationary approximation.
b. Find the interface position, R(t).
c. Find the highest cooling rate from the outer pipe below which freezing cannot
occur under the specified conditions. Express your answer in terms of Ti, ther-
mophysical properties, outer radius ro, and the constant of proportionality.
7.23 Water inside a pipe is initially at a temperature of Ti. It begins to freeze inward when
the outer wall temperature of the pipe is lowered to Tf (phase change temperature).
Use a one-dimensional quasi-stationary analysis in this problem.
a. Find the time required for the position of the phase interface to reach one half
of the outer radius of the pipe. Express your answer in terms of thermophys-
ical properties, pipe radius, and the interface velocity. Assume that the inter-
face moves inward at a constant velocity and the temperature throughout the
solid is uniform.
b. Find the spatial temperature distribution in the liquid at the time calculated in
part (a).
7.24 A solid material at a temperature of Tf (phase change temperature) in a spherical con-
tainer is heated externally. Assume that the melting front moves inward at an approx-
imately constant phase change rate with a constant heat flux at the liquid side of the
phase interface. Also, assume that the temperature of the solid remains nearly uni-
form at Tf.
a. Using a one-dimensional, quasi-stationary analysis, find the time required to
melt the spherically shaped solid entirely. Express your answer in terms of
Solidification and Melting 343
the interfacial heat flux, thermophysical properties, and the initial (unmelted)
sphere radius.
b. Determine the rate of change of wall temperature with time.
7.25 The wall of a solar energy collector consists of a phase change material of thickness L
and height H. Temperatures at the left and right boundaries are TH and TM (phase
change temperature). The top boundary is insulated and a material with known
properties (kw, ρw, and cp,w) is located beneath the PCM. A buoyant recirculating
flow arises in the liquid as a result of the differentially heated boundaries. What
are the dimensionless parameters affecting the Nusselt number for two-dimensional
heat transfer across the PCM.
7.26 By imposing a prescribed pressure gradient, a non-Newtonian slurry of a liquid–sand
mixture flows through a channel of height 2H with a mass flow rate of ṁ. The fully
developed laminar flow in the x-direction is unheated up to x = a, but then heated
uniformly by a wall heat flux, qw, between a ≤ x ≤ b. The shear stress of the slurry
is τ = τb + μb(du/dy), where τb refers to the Bingham yield stress and μb is the Bingham
viscosity. Write the governing equations and boundary conditions which define the
two-dimensional heat and fluid flow of this slurry mixture. Non-dimensionalize
these equations to derive the parameters affecting the temperature and Nusselt num-
ber distributions. Assume that the thermophysical properties, such as cp of the mix-
ture, remain constant, and the inlet temperature is Ti.
7.27 Laser heating of a metal initiates melting at the surface of the metal. At an initial time,
t = 0, the solid is initially at a temperature of To. A uniform heat flux, qw, is applied
and held at the wall (x = 0) where the laser heat is experienced. The resulting temper-
ature disturbance propagates into the metal. At x . δ(t), where δ varies with time, the
solid temperature is approximately To. The purpose of this problem is to use a one-
dimensional integral analysis to estimate the time elapsed, tm, before the surface
reaches the melting temperature, Tm (where To , Tm).
a. Write the integrated form of the energy equation.
b. Assume a temperature profile of the following form:
x x2
T = a0 + a1 + a2
δ δ
matrix. Assume that a constant pressure gradient is experienced by the flow through
the solid matrix and each fluid element travels an equal distance through the
porous matrix.
7.30 A compartment of PCM is used for thermal control of a satellite. The satellite gener-
ates heat uniformly throughout its orbit from internal power. When the satellite is
shaded from solar radiation by the Earth, latent heat is released by the PCM during
solidification. On the bright side of its orbit around the Earth, heat is absorbed by the
melting PCM and the satellite is heated by an incident solar heat flux. Estimate the
heat flux due to phase change of PCM that should be maintained so that the satellite
temperature remains approximately uniform throughout its orbit. Express your
answer in terms of the uniform incoming solar flux on the bright side of the orbit.
List all assumptions adopted in the analysis.
7.31 Consider the use of PCMs in a solar thermal power generation system. A solar con-
centrator consists of several mirrors focusing solar radiation on a PCM heat
exchanger. It receives solar energy and transfers it to a PCM as a heat source for sub-
sequent power generation. Define U, Tc, Tm, and To as the total thermal conductance
of the PCM heat exchanger, surface temperature of the solar collector, PCM melt tem-
perature and ambient (heat sink) temperature, respectively. Of the total incoming
solar energy, Q, a fraction QH is absorbed by the PCM, and the remaining fraction,
Qe, is reflected. The power generation system has an incoming heat supply of QH,
power output of W, and heat rejection to the environment of Qo. Determine the opti-
mal phase change temperature, Tm, based on the method of entropy generation min-
imization (yielding a maximum power output) of the power station.
References
H.W. Carslaw and J.C. Jaeger. 1959. Conduction of Heat in Solids, 2nd Edition, Oxford: Clarendon Press.
B. Chalmers. 1964. Principles of Solidification, New York: John Wiley & Sons.
S.H. Cho and J.E. Sunderland. 1969. “Heat Conduction Problems with Melting or Freezing,” ASME
Journal of Heat Transfer 91: 421–426.
J. Darkwa, O. Su, and T. Zhou. 2015. “Evaluation of Thermal Energy Dynamics in a Compacted
High-Conductivity Phase-Change Material,” AIAA Journal of Thermophysics and Heat Transfer,
29: 291–296.
X. Duan and G.F. Naterer, 2010. “Heat Transfer in Phase Change Materials for Thermal Management
of Electric Vehicle Battery Modules,” International Journal of Heat and Mass Transfer, 53: 5176–5182.
S.D. Foss and S.S.T. Fan. 1972. “Approximate Solution to the Freezing of the Ice-Water System with
Constant Heat Flux in the Water Phase,” Water Resources Research, 8: 1083–1086.
G.R. Jackson and T.S. Fisher. 2016. “Response of Phase-Change-Material-Filled Porous Foams Under
Transient Heating Conditions,” AIAA Journal of Thermophysics and Heat Transfer, 30: 880–889.
F. Kreith and F.E. Romie. 1955. “Study of the Thermal Diffusion Equation with Boundary Conditions
Corresponding to Solidification or Melting of Materials Initially at the Fusion Temperature,” Pro-
ceedings of the Physical Society B, 68: 277–291.
W. Kurz and D.J. Fisher. 1984. Fundamentals of Solidification, Switzerland: Trans Tech Publications.
G.A. Lane. 1996. Solar Heat Storage: Latent Heat Materials, Boca Raton: CRC Press/Taylor & Francis.
A.L. London and R.A. Seban. 1943. “Rate of Ice Formation,” Transactions of ASME, 65: 771–779.
V.J. Lunardini, 1988. “Heat Conduction with Freezing or Thawing,” CRREL Monograph 88-1, U.S.
Army Corps of Engineers.
Solidification and Melting 345
G.F. Naterer and G.E. Schneider. 1995. “PHASES Model of Binary Constituent Solid-Liquid Phase
Transition, Part 2: Applications,” Numerical Heat Transfer B, 28: 127–137.
E. Niyama, T. Uchida, M. Morikawa, and S. Saito. 1982. “A Method of Shrinkage Prediction and its
Application to Steel Casting Practice,” AFS International Journal of Metalcaasting, 7: 52–63.
E. Pardo and D.C. Weckman. 1990. “Fixed Grid Finite Element Technique for Modeling Phase Change
in Steady-State Conduction-Advection Problems,” International Journal for Numerical Methods in
Engineering, 29: 969–984.
W.G. Pfann. 1958. Zone Melting, New York: John Wiley & Sons.
P.R. Sahm and P. Hansen. 1984. Numerical Simulation and Modelling of Casting and Solidification Processes
for Foundry and Cast House, CIATF, CH-8023, International Committee of Foundry Technical
Associations, Zurich, Switzerland.
R.V. Seeniraj and T.K. Bose. 1982. “Planar Solidification of a Warm Flowing Liquid under Different
Boundary Conditions,” Warme Stoffubertragung 16: 105–111.
J. Stefan. 1891. “Uber die Theorie des Eisbildung, Insbesonder uber die Eisbildung im Polarmere,”
Annual Review of Physical Chemistry, 42: 269–286.
R.H. Tien and G.E. Geiger. 1967. “Heat Transfer Analysis of the Solidification of a Binary Eutectic Sys-
tem,” ASME Journal of Heat Transfer 89: 230–234.
8
Chemically Reacting Flows
8.1 Introduction
Heat transfer is a major element of the design and analysis of chemical reaction engineering
systems. Chemically reacting flows are characterized by chemical changes of reactants that
yield one or more products in solid, liquid, and/or gas phases. These reactions often consist
of a sequence of individual sub-steps, called elementary reactions, which occur at a character-
istic reaction rate, temperature, and chemical concentration.
Typically, reaction rates increase with temperature because a higher thermal energy can
more readily overcome the activation energy (minimum energy required to break molecular
bonds and start a chemical reaction). Chemical reactions may proceed in a forward or
reverse direction until they move to completion or reach equilibrium. Forward proceeding
reactions are called spontaneous reactions, requiring no free energy input to proceed, whereas
nonspontaneous reactions (such as charging a battery with electricity input) require input
energy to proceed.
There are two main types of classifications of chemical reactions—the first is a division
between homogeneous and heterogeneous systems; and the second is a division between non-
catalytic and catalytic reactors. A reaction is homogeneous if it involves one phase only,
whereas heterogeneous reactions involve at least two phases. Catalytic reactions have for-
eign materials, called catalysts, which are neither reactants nor products, but which alter
the rate of a chemical reaction. In general, catalysts increase the rate of reaction because
they require less activation energy to begin the reaction. Most gas phase reactions are non-
catalytic, while liquid phase reactions are usually catalytic.
In chemical reaction engineering, two common types of reactors are batch and conti-
nuous reactors. A batch reactor is a tank with a mixer and heating or cooling system.
Continuous reactors have inflow and outflow streams that typically bring in reactants
and move the resulting products through an exit stream. An exothermic reactor releases
heat and therefore requires a cooling system to maintain a uniform temperature, whereas
an endothermic reactor requires heat input to drive the chemical reaction. The residence
time is an important parameter for incoming reactants in continuous reactors. It charac-
terizes the amount of time a reactant spends inside a reactor before it reacts to the
product(s).
Heterogeneous reaction systems involve multiphase flows with gas, liquid, and/or solid
phases. Often the liquid and solid phases appear in the form of droplets and particles,
respectively. An important group of chemically reacting flows is combustion. Combustion
of coal particles involves multiphase gas–solid flows, whereas combustion of fuel droplets
involves gas–liquid flows. Soot in the form of black particles consists of carbon particles
formed during incomplete combustion of hydrocarbon fuels. Soot particles absorb, emit,
347
348 Advanced Heat Transfer
and scatter radiation within a combustion chamber and therefore have a significant role in
heat and fluid flow processes.
In this chapter, the fundamentals of heat transfer and energy exchange in chemically react-
ing flows will be presented. Topics to be covered include fundamental concepts (such as
reaction rates and mole balances), combustion reactions and various modes of multiphase
reacting flows (gas–solid, gas–liquid, and gas–solid–liquid). Also, fluidized beds will be
introduced, as well as advanced solution methods such as shrinking core and progressive
conversion models. For a more detailed analysis of chemical reaction engineering, refer to
other sources such as Levenspiel (1999) and Fogler (2016).
∙ Mass of mixture: m = m1 + m2 + ⋯ = Σ mi
∙ Mass fraction: χi = mi/m
∙ Total number of moles of mixture: N = N1 + N2 + ⋯ = Σ Ni
∙ Mole fraction: yi = Ni/N
mi = N i M i (8.1)
n
m= Ni Mi (8.2)
i=1
The units of the molecular weight (or atomic weight) are kg/kmol. The molecular weights
of elements are presented in Appendix H.
The mixture molecular weight is given by:
m n
M= = yi Mi (8.3)
N i=1
Also, the mass fraction and mole fraction are related by:
χ i mi /m mi N Mi
= = = (8.4)
yi Ni /N Ni m M
Chemically Reacting Flows 349
A molar analysis involves the number of moles of each component in a mixture, whereas a
gravimetric analysis specifies the mass of each component.
For a mixture of ideal gases in a reacting flow, the mole fraction can be written in terms of
the volume and pressure ratios as follows (see end-of-chapter problem):
Vi pi
yi = = (8.5)
V p
where pi and p refer to the partial pressure of component i and total pressure, respectively. The
partial pressure is the pressure that the individual component would have if it occupied the
entire volume of the mixture by itself. Also, Vi refers to the volume occupied by component i,
or the volume occupied if the individual component was isolated from the other
components.
Mole fractions are frequently used in the analysis of reacting flows since the fractions of
each reactant (by volume) are often known. Also, the properties of reacting mixtures are typ-
ically based on mole (or mass) fraction-weighted sums of individual component properties.
For example, the specific heat of an air (subscript A)–fuel (subscript F ) mixture is given by:
where yA and yF are the mole fractions of air and fuel, respectively. Also, χA and χF are the
respective mass fractions. Once the mixture properties are obtained, the energy and heat bal-
ance equations can be solved in the regular manner as previous chapters in terms of these
mixture properties.
The rate of reaction is the rate at which the chemical species loses its chemical structure
through decomposition, combination, or isomerization. The rate of reaction (in units of
mol/m3) of a species component, for example species A, or product P, can be expressed
as the rate of disappearance of the reactant, –rA, or the rate of formation (generation) of
the product, rP.
For example, consider an isomerization reaction, A→B. Define rA as the rate of formation
of species A per unit volume. The rate of disappearance of species A per unit volume is –rA
whereas the rate of formation of species B per unit volume is rB. These rates are functions of
350 Advanced Heat Transfer
concentration, temperature, and pressure. They are independent of the type of chemical
reactor. The rate of reaction is usually expressed as an algebraic function of species concen-
tration, for example, –rA = kCA, where k is a reaction rate coefficient and CA is the concentra-
tion of species A in the mixture.
The reaction rate represents the speed of the reaction or how quickly a reactant is con-
verted to a product. Consider a typical reaction of the following form:
where the lowercase letters (a, b, d, e) represent stoichiometric coefficients and the capital let-
ters represent the reactants (A, B) and products (D, E).
Assuming the stoichiometric amount of reactants are converted to products, without
excess reactants or by-products as a result of an incomplete reaction, then the reaction
rate (r) in a closed system of constant volume can be expressed as:
where Ci refers to the molar concentration of species i. The molar concentration represents
the amount of a constituent in moles divided by the volume of the mixture. The units of the
reaction rate are typically mol/Ls. The reaction rate is always positive. A negative sign
above is used when a reactant is consumed and hence its concentration is decreasing. It
can be observed that the rate of reaction is inversely proportional or normalized by the stoi-
chiometric number so that it becomes independent of which reactant or product species is
used in the analysis.
The rate equation for a chemical reaction is used to relate the rate of the reaction to the
concentration of each reactant. It is typically expressed as follows:
r = kCnA Cm
B (8.10)
where k is the reaction rate coefficient or rate constant, and the exponents n and m are called
reaction orders. For gas phase reactions, the reaction rate is often expressed instead in terms
of partial pressures of each constituent. The rate constant includes the effects of all param-
eters (except concentration) on the reaction rate. Usually, temperature is the most significant
factor.
For elementary single-step reactions, the order with respect to each reactant is equal to its
stoichiometric coefficient or the number of molecules participating in the reaction. For a
unimolecular reaction or step, the number of collisions of molecules is proportional to the
concentration of molecules of reactant. The rate equation is first-order. For a bimolecular
reaction or step, the number of collisions is proportional to the product of the two reactant
concentrations. In this case, the reaction is second order. Similarly, a trimolecular reaction is
third order, and so forth.
There are numerous factors which influence the rate of reaction. For example, the nature of
the reaction, such as the number of reacting constituents and physical state, affect the com-
plexity of the reaction. The reaction rate normally increases with concentration and pressure
since the frequency of collisions among reactants increases at higher concentrations and
pressures. Also, temperature increases the reaction rate due to more energetic colliding par-
ticles with higher activation energies to initiate the chemical reactions. The presence of a cat-
alyst increases the rate of reaction as it provides an alternative pathway for the chemical
Chemically Reacting Flows 351
reaction at a lower activation energy. As mentioned earlier, the activation energy is the min-
imum energy which must be available in the reactants for the chemical reaction to proceed,
or in other words, the minimum energy required to start the chemical reaction.
The reaction rate coefficient, k, normally has a strong dependence on temperature. This
dependence is described by the Arrhenius equation,
Ea
k = A exp − (8.11)
RT
where Ea and R refer to the activation energy and gas constant, respectively. Also, the pre-
exponential factor, A, is a constant that is unique for each chemical reaction and character-
izes the frequency of collisions of reactant molecules. The Arrhenius equation indicates
that the number of collisions resulting in a reaction per second is equal to the number
of collisions (both leading to a reaction and not leading to a reaction) per second, multi-
plied by the probability that any given collision results in a reaction.
The rate of reaction can be expressed in terms of the mass, volume, and/or number of
moles consumed or released in the reaction. For example, in the gas–solid reaction of Equa-
tion 8.8, the mass of gaseous reactant, A, decreases when it flows through the reactor as it is
consumed in the reaction. The mass of solid product increases as particles of constituent B
are converted to the solid product, D. The change in solid mass with time can be determined
by the accumulation of solid product, less the consumption of reactant solid mass, as
follows,
dms dND
= (dMD − bMB ) (8.12)
dt dt
where the subscript s refers to solid. The variables MB and MD represent the molar masses of
constituents B and D, respectively.
The reaction rate, r, of species i, based on the unit volume of reacting solid (constituent B)
can then be expressed as:
The volume of the reactant solid, VB, as a function of time, can be written as:
MB
VB (t) = NB (8.14)
ρB
where ρB and NB represent the density and molar quantity of constituent B. The molar
quantity of B is calculated by,
dNB
NB = N0 − (8.15)
dt
where N0 represents the initial quantity of B in the reactor when the reaction is initiated, or
alternatively, the initial mass of B divided by its molar mass.
For a reaction of the stoichiometric amount of reactant gas and solid, the rate of consump-
tion of solid reactant, dNB/dt, balances the rate of formation of product gas, dNC/dt.
352 Advanced Heat Transfer
Then, combining the previous two equations allows the volume of reactant solid to be deter-
mined by:
MB m0,B dNC
VB = − (8.16)
ρB MB dt
This expression can be used determine the reaction rate based on the volume of the reac-
tant solid and rate of molar formation of products. If many chemical reactions with multiple
constituents occur simultaneously, the reaction rates involve coupled equations involving
the individual reaction rate coefficients, constituent concentrations, and their interdepen-
dencies, for example, as presented by Sinha and Reddy (2011).
dNA
= FA0 − FA + ZA (8.17)
dt
Here, F and the subscript 0 refer to the molar flux (kmol/s) and inlet value, respectively.
From left to right, the terms in the mole balance represent the molar rate of accumulation of
species A with time, molar flow rate of species A into the control volume, molar outflow rate
of species A, and molar rate of generation of species A due to chemical reactions. The units of
each term are mol/s. For a reaction rate of rA, which may vary throughout the control vol-
ume, V, the total molar generation of species A is given by:
ZA = rA dV (8.18)
V
For spatially uniform molar generation of species A, the molar generation rate becomes
ZA = rAV.
Four common types of reactors are normally encountered in chemical reaction engineer-
ing systems (see Figure 8.2): (1) a batch reactor (BR); (2) a continuous stirred tank reactor
Control
volume (CV)
Reactants GA
Products
FA, CA, TP
FA0, CA0, TR
FIGURE 8.1
Control volume with a reacting mixture.
Chemically Reacting Flows 353
FA0, CA0
FA, CA
FIGURE 8.2
Schematic of batch, stirred, plug flow, and pack bed reactors.
(CSTR); (3) a plug flow reactor (PFR); and (4) a packed bed reactor (PBR). Using the previous
general form of the mole balance for chemically reacting flows, a specific formulation can be
developed for each type of chemical reactor.
8.4.2 Batch Reactor
For a batch reactor, FA0 = 0 = FA, which is well mixed. Then Equation 8.17 yields:
dNA
= rA V (8.19)
dt
Integrating this expression from t = 0 when NA = NA0 to a final time, t = tf, when NA = NAf,
yields the time necessary, tf, to reduce the number of moles of A from NA0 to NAf,
NAf
−dNA
tf = (8.20)
NA0 rA V
FA0 − FA + rA V = 0 (8.21)
Rearranging this result yields the following CSTR volume necessary to reduce the molar
flow rate from FA0 to FA,
FA − FA0
V= (8.22)
rA
354 Advanced Heat Transfer
where the terms represent the molar inflow, molar outflow, and generation of species A
over the volume ΔV. Rearranging this result and taking the limit as the volume becomes
small, ΔV → 0,
FA |V+ΔV − FA |V
lim = rA (8.24)
ΔV0 ΔV
which implies,
dFA
= rA (8.25)
dV
FA
dFA
V= (8.26)
FA0 rA
This represents the volume required to reduce the incoming molar flow rate (mol/s) of
species A from FA0 to the exit flow rate of FA.
Unlike the previous two cases of reactors with well-mixed conditions (rA assumed cons-
tant throughout the volume), for this case of a plug flow reactor, the reaction rate varies
throughout the volume. Therefore, the mole balance in Equation 8.17 under steady-state
conditions becomes:
FA0 − FA + rA dV = 0 (8.27)
The rates of reaction affect the formation of products from the reactants. Chemical
reactions that are characterized by zeroth-order kinetics can be written as:
Ṅ A = k0 (8.28)
where Ṅ A = dNA /dt is the molar accumulation rate of species A. A first-order reaction can be
written as:
Ṅ A = k1 CA (8.29)
Similar expressions are written for higher order reactions. In Equations 8.28 and 8.29,
k0 and k1 are the reaction rate constants. The units of k0 and k1 are kmol/s m3 and 1/s, respec-
tively. The chemical reaction may occur at a constant rate (e.g., Equation 8.28 is zero-order)
or a rate that is proportional to the local concentration (e.g., Equation 8.29 is first-order). If
Ṅ A is positive, then the reaction leads to the production of constituent A. On the other hand,
if Ṅ A is negative, then the reaction is characterized by a consumption of constituent A.
or alternatively,
For the energy balance, the incoming and outgoing enthalpies of reactants and products
are needed. The specific enthalpy of a compound can be determined by the enthalpy of
formation of the compound at standard temperature and pressure conditions (STP; 25◦ C,
1 atm), plus the specific enthalpy change between the state of interest and the standard state
(STP; designated by a superscript o),
where the overbar designates a molar quantity (units in a per mole basis). Reactions leading
to new constituents or compounds are called formation reactions because a substance is
formed from its elements in their respective natural states (e.g., gas, liquid, or solid) at STP.
The enthalpy at the reference state of STP, h(Tref, pref), is denoted by ho. For a formation
reaction, a subscript f is used for the enthalpy of formation and the superscript o denotes
STP. In other words, the enthalpy of a compound consists of the enthalpy of formation of
the compound from its elements, plus the change of enthalpy associated with its change
of state (temperature, pressure) at a constant composition. An arbitrary choice of datum
can be used to determine the enthalpy change since it is a difference at a constant
356 Advanced Heat Transfer
where the first term on the right side is the chemical (formation) enthalpy and the sec-
ond term is sensible enthalpy. The chemical enthalpy is set to zero for elements in their
natural state. For a constant specific heat, the integral can be evaluated as follows for
ideal gases:
Variations of specific heat with temperature can be included in the evaluation of enthalpy
using this expression.
Performing an energy balance over the control volume in Figure 8.1,
2 o
2 o
Ṅ R,i hf + h − h Ṅ P,j hf + h − h
o o
− + Q̇ = 0 (8.35)
R,i P,j
i=1 j=1
or alternatively,
2 o
2 o
Ṅ R,i hf + Δh − Ṅ P,j hf + Δh + Q̇ = 0 (8.36)
R,i P,j
i=1 j=1
where Q̇ is the rate of heat addition or removal from the reactor to maintain steady-state
conditions. The reaction is called endothermic if Q̇ is positive (heat added) in the energy bal-
ance and exothermic if Q̇ is negative (heat removed). A negative Q̇ indicates that heat is
removed to lower the products to the temperature of TP.
In the energy balance, generally the temperatures of the products are unknown so the
enthalpies must be computed as functions of temperature. Values can be retrieved from
thermodynamic tables (see Appendices E, F, and I). For ideal gases, enthalpies are a function
of temperature alone. An iterative solution procedure is normally required.
If the rate of heat transfer from the reaction chamber is given, then the following iterative
procedure can be used:
1. Estimate the outlet temperature and then calculate the enthalpies of the products of
the reaction based on this temperature.
2. Check and determine whether the energy balance is satisfied.
3. Repeat steps 1 and 2 until the energy balance is satisfied.
For a nonreacting mixture, the fluid constituents that enter a control volume also come
out. Thus a reference enthalpy, href, is often not required in single phase problems since
Chemically Reacting Flows 357
only the enthalpy differences between the same constituents at the outlet and inlet are
required. However, for a reacting mixture, the compounds entering the control volume
do not all come out. In this case, href becomes more important.
For example, consider the heat released by the following formation reactions.
The heat flow per kmol of product is determined by the total heat transfer, Q, in units of
Joules, divided by the number of moles of products of the reaction. The negative sign
indicates that the products of the reaction are at high temperatures, so that heat must
be removed from the control volume to return the temperature of exiting products down
to STP. The above q values are called the standard heats of formation (or enthalpies of
formation).
In some cases, the enthalpy of formation of a product may not be readily or directly avail-
able if the reactants tend to produce other products. For example, consider the following
reaction:
1
C(s) + O2 (g) CO(g) (8.37)
2
Here the enthalpy of formation of CO is difficult to measure because carbon (C) tends to
burn to carbon dioxide (CO2) rather than carbon monoxide (CO). However, the following
summation law of reactions can be used:
1
CO(s) + O2 (g) CO2 (g) (8.39)
2
with an enthalpy of formation of –283 MJ/kmol. Subtracting Equation 8.39 from the prior
equation, together with their respective enthalpies of formation,
1
C(s) + O2 (g) CO(g) (8.40)
2
The enthalpy of formation of CO becomes –110 MJ/kmol. It should be noted that CO is not
an element in Equation 8.39. Thus –283 MJ/kmol does not represent the enthalpy of forma-
tion of CO2 on the right side of the equation.
In Appendix I, thermodynamic properties of relevance to reacting mixtures of gases and
liquids at varying temperatures are presented. This includes values of specific enthalpy, h,
internal energy, u , and entropy, s, at varying temperatures; standard molar entropy, so , and
specific heat, cp , of one mole of substance at the standard state; enthalpy and entropy of
o
formation, hf and sof , respectively; and the enthalpy of combustion, Δhc .
o o
To perform a second law analysis, the entropy of individual constituents and mixture
entropy are required. Using the ideal gas law and Gibbs equation (Chapter 3), it can be
358 Advanced Heat Transfer
shown that the entropy of component i in the mixture can be written in terms of the temper-
ature, total pressure and mole fraction, as follows:
T p
si = si,0 + cp,i ln − Ri ln − R ln(yi ) (8.41)
T0 p0
where the subscripts i and 0 refer to component i and reference value, respectively. If there is
only one component in the mixture (e.g., i = 1 for a pure gas), then the last term becomes
zero since yi = 1. In general for reacting flows, reference values are typically taken at stan-
dard temperature and pressure (STP; 298 K and 1 atm).
The chemical equation for the reaction of a hydrocarbon fuel in air can be written as:
b b b
Ca Hb + 4.76 a + (0.21 O2 + 0.79 N2 ) aCO2 + H2 O + 3.76 a + N2 (8.43)
4 2 4
The leading coefficients of each term can be determined based on conservation of atoms
for carbon, hydrogen, oxygen, and nitrogen on the left side. The second term on the left side
is the stoichiometric air (or theoretical air), which refers to the minimum amount of air required
for complete combustion of the hydrocarbon fuel.
The above products of combustion are formed under conditions of complete combustion
when all CaHb is burned in air without any excess air remaining from the reaction. Other-
wise, with incomplete combustion, by-products such as carbon monoxide and others arise
if not all of the CaHb is burned. Incomplete combustion is generally undesirable for many
reasons such as the undesired release of harmful pollutants into the atmosphere and
reduced fuel efficiency. Thus, the amount of air supplied should be equal to or greater
than the stoichiometric amount in Equation 8.43.
Consider the combustion of methane in air. Using a = 1 and b = 4 in Equation 8.43 for
methane, the chemical reaction for the combustion of methane in air becomes:
It can be observed that the sum of moles of reactants on the left side does not balance the
sum of moles of products on the right side. However, the total mass of reactants must bal-
ance the total mass of products.
The mass of reactants or products, m, can be determined by the number of moles, N, mul-
tiplied by the molecular weight, M. Since air consists mainly of oxygen and nitrogen, the
Chemically Reacting Flows 359
where the subscript i refers to constituent i and yi denotes the mole fraction of constituent i.
Substituting the molecular weights of oxygen and nitrogen yields Mair = 28.84 kg/kmol.
The air–fuel mass ratio (AF) is defined by:
where MC = 12 kg/kmol and MH = 1 kg/kmol refer to the molecular weights of carbon and
hydrogen, respectively.
Excess air (EA) is defined as more air than the stoichiometric amount. For example, 25%
excess air corresponds to 1.25 times the stoichiometric amount of air. Adapting Equation
8.43 to chemical reactions with excess air,
b
Ca Hb + (1 + EA) 4.76 a + (0.21O2 + 0.79N2 )
4
b b b
aCO2 + H2 O + (1 + EA)4.76 a + (0.79N2 ) + (EA)4.76 a + (0.21O2 ) (8.47)
2 4 4
The last term arises since the reaction is not stoichiometric and hence not all oxygen reacts.
For a compound containing carbon, hydrogen, and oxygen, the combustion reaction is
given by:
b c b
Ca Hb Oc + a + − O2 aCO2 + H2 O (l) (8.48)
4 2 2
During this reaction, the heat of combustion (or enthalpy of combustion) is the amount of heat
released during the combustion process. The standard heat of combustion, Δhoc , is the neg-
ative of the enthalpy change for the reaction, which is given by:
b
Δhoc = −aΔhof (CO2 , g) − Δhof (H2 O, l) + Δhof (Ca Hb , Oc ) (8.49)
2
Δhoc = 393.51a + 142.915b + Δhof (Ca Hb , Oc ) (8.50)
Values of the heat combustion for various substances are presented in Appendix I, for
example, inorganic substances, hydrocarbons, alcohols, ethers, carbonyl compounds, acids,
esters, and nitrogen compounds. The above equation for the heat of combustion applies if
the reactants start at STP and the products return to the same conditions. The same equation
also applies if the compound contains other elements if that element ends in the standard
reference state, however, the products containing the other elements must be known to cal-
culate the heat of combustion.
The adiabatic flame temperature, Tad, refers to the highest temperature that will be obtained
by burning a fuel under a specified set of conditions. This temperature is obtained when the
reactants enter a well-insulated mixing chamber and the products of combustion leave the
360 Advanced Heat Transfer
chamber at Tad. Thus all heat generated by the combustion reaction is transferred to thermal
energy of the combustion products. The adiabatic flame temperature is normally around
2,300 K for several hydrocarbon fuels. It varies with the air–fuel ratio, AF, and excess air,
EA. The adiabatic flame temperature decreases if more excess air is used since more thermal
energy is required to raise the temperature of the nonreacting air. Conversely, if less air than
the stoichiometric amount is supplied, Tad decreases due to incomplete combustion. Thus
the maximum adiabatic flame temperature is reached at the air–fuel ratio corresponding
to the stoichiometric amount of air.
The heating value of a fuel is defined as the heat of combustion for 1 mole of the fuel in
stoichiometric air when the reactants and products are at STP. The higher heating value
(HHV) refers to the heating value when H2O in the products of combustion is in the liquid
phase. Conversely, the lower heating value (LHV) is the heating value when H2O in the prod-
ucts is in a gaseous phase. Further detailed analysis of combustion reactions is available in
thermodynamic textbooks, for example, Moran and Shapiro (2014).
The conversion of solid reactant depends on the rate of reaction and residence time of a
particle. A progressive conversion model would be used if the diffusion of gaseous reactant,
A, into the particle is much faster than the chemical reaction. The solid reactant B would be
Chemically Reacting Flows 361
Time Time
Unreacted
core
concentration
Solid reactant
Radial
ro 0 ro ro 0 ro ro 0 ro position
FIGURE 8.3
Schematic of the shrinking core model.
consumed nearly uniformly throughout the particle. On the other hand, a shrinking core
model can be used if the diffusion of gaseous reactant, A, is much slower. As illustrated
in Figure 8.3, this restricts the reaction zone to a thin layer that moves inward into the par-
ticle over time during the reaction.
In the shrinking core model, several simplifications and assumptions are made to deter-
mine the rates of reaction and solid particle conversion. The reaction is assumed to occur
at the interface between the outer shell and the unreacted core of the solid. The unreacted
core shrinks in size as the reaction proceeds. Also, the particle is assumed to be nonporous
and spherical. Isothermal conditions, equimolar counter-diffusion of gases/reactants and
products (opposite direction concentration gradients leading to diffusion in counter direc-
tions), and a first-order reaction are also assumed.
Define CAs, CAi, CAb, κg, κs, ri, rc, and Ds to represent the concentration of gaseous reactant
at the particle surface (subscript As), solid interface (subscript Ai) and bulk gas phase
(subscript b); mass transfer coefficient based on the concentration change in the gas (sub-
script g) and solid (subscript s); radius of the initial particle (subscript i) and unreacted
core (subscript c); and diffusion coefficient, respectively. Then the decomposition steps
and corresponding reaction rates in a shrinking core model can be analyzed as follows:
1. Diffusion of gaseous reactant, A, through a gas film surrounding the solid particle.
4πri rc Ds
rA = (CAs − CAi ) (8.53)
ri − rc
The rate of reaction at a given time is obtained by combining these equations to eliminate
the unknown intermediate compositions, CAi and CAs, yielding
−1
1 ri − rc 1
rA = + + CAb (8.55)
4πri κ g 4πri rc Ds 4πr2i κs
2
The rate of movement of the interface can be related to the rate of reaction of solid through
a mole balance and stoichiometric balance for constituent B, yielding:
d 4 3 ρB
− πrc = brA (8.56)
dt 3 MB
Substituting for rA and integrating Equation 8.56, then rewriting in terms of the solid con-
version rate,
ri r2 ri bCAb MB
χ s + c [1 − 3(1 − χ s )2/3 + 2(1 − χ s )] + (1 − (1 − χ s )1/3 ) = t (8.58)
3κ g 6Ds κs ρB
This represents the conversion of solid with time, including the processes of the gas film,
diffusion through the product layer, and the chemical reaction. The time required for com-
plete conversion of the reactant particle to the product can be obtained by setting χs = 1,
yielding:
ri r2i ri ρB
t= + + (8.59)
3κ g 6Ds κs bCAb MB
In addition to the reaction time, this analysis can also be used to evaluate the magnitudes
of individual resistances, as well as the overall resistance of the reaction rate. Daggupati,
Naterer, and Dincer (2011) applied this shrinking core model to analyze a hydrolysis reac-
tion of cupric chloride particles with steam to produce copper oxychloride solid and hydro-
gen chloride gas in a process of thermochemical hydrogen production.
The previous reaction in Equation 8.51 involved a combination reaction with two reac-
tants (solid and gas). Consider a decomposition reaction instead, involving a single reactant
of a solid particle that decomposes into gaseous and solid products, as follows,
Unlike the previous reaction, the decomposition reaction is primarily controlled by the
rate of desorption of gas from the interface. There is no chemical reaction among reactants
and the rate of desorption of gas is mainly controlled by a gas film and product layer around
the solid. The rate of desorption of product gas in the gas film and product ash layer can be
Chemically Reacting Flows 363
Substituting for rE in this equation, integrating, and following a similar procedure as the
combination reaction, the following equation for the conversion of solid, χs, is obtained:
ri r2i 2/3 K bMB
χs + [1 − 3(1 − χ s ) + 2(1 − χ s )] = − CEb t (8.63)
3κ g 6Ds Rg Tb ρB
where K, Rg and Tb refer to the chemical equilibrium constant, gas constant, and temperature
of the bulk phase, respectively. This result gives the time of decomposition of the solid with a
gas film and desorption of gas from the interface. It can be used to predict the rate of decom-
position of solids for isothermal systems without bulk diffusion of gases.
Time Time
concentration
Solid reactant
Radial
ro 0 ro ro 0 ro ro 0 ro position
FIGURE 8.4
Schematic of the progressive conversion model.
364 Advanced Heat Transfer
particle. The following example demonstrates this progressive conversion of a solid and esti-
mates the time elapsed for complete conversion of a metal pellet in a gas–solid reaction.
′′
Ṅ A = −k1 C (8.65)
′′
where k, C and Ṅ A refer to the reaction rate constant, concentration of calcium carbonate in
the pellet, and its rate of formation (per unit area and time), respectively. Conservation of
species concentration requires that:
d(VC)
= −k1 AC (8.66)
dt
where V and A refer to the volume and surface area of the pellet, respectively. Assuming
that the concentration of calcium carbonate remains constant over time during decompo-
sition of the pellet,
dL3
C = −k1 (6L2 )C (8.67)
dt
dL
= −2k1 (8.68)
dt
which indicates that the reaction front moves back into the pellet at a constant velocity of
2k1 during the thermal decomposition. Solving the equation subject to the conditions of
L = L0 at t = 0 (initial condition) and L = 0 at t = tf (final time),
L0 = 2k1 tf (8.69)
This result represents the initial side length of the cube-shaped pellet. This length char-
acterizes the size of a pellet that is completely decomposed by the chemical reaction with
the CO2 gas over a time duration of tf.
Chemically reacting systems involving gas flow and solid particles occur in numerous
industrial processes. An example is purification of metals using carbon-based gases. Pellets
of metal oxides (such as oxides of zinc, magnesium, and lead) are initially purified using car-
bon monoxide or carbon dioxide gases (Guthrie 1993). Oxide ores are typically obtained
Chemically Reacting Flows 365
from siliceous rocks by grinding, crushing, or flotation operations. The impurities in these
ores are removed by chemical reactions with gases such as CO or CO2 (called topochemical
reduction).
Another common industrial process involving gas–solid reactions is smelting of ores (see
Figure 8.5). Blast furnaces are used in smelting processes and refining of metals including
the production of iron ore, tin, and lead. High-temperature and high-pressure gases are
used to induce combustion within a vertically oriented furnace. The combustion zone is
called a bosh and the region where the molten material is processed is called a hearth. In mod-
ern production facilities, blast furnaces can often exceed 30 m in height, 10 m in diameter,
and 1,700 tons of production per day.
During the production process in the furnace, pellets of materials such as limestone, coke,
or iron oxide ore are supplied at the top of the furnace. As pellets slowly descend through
the furnace, they are heated by hot ascending gases. These gases have a higher carbon mon-
oxide content due to combustion lower in the furnace. The descending pellets are decom-
posed as a result of the chemical reaction and combined heat–mass transfer of carbon
dioxide and hydrogen from the gas phase into the solid pellets. The converted iron oxide
pellets then react further with carbon monoxide as follows:
The products of composition are wustite (FeO) and carbon dioxide (CO2). As the wustite
descends further and mixes with carbon dioxide and carbon monoxide gases, the following
final reactions occur:
Gas uptake
Solids flow
Steel shell
Gas flow
Ore (pellets/sinter)
Coke layer
Active coke zone
Water cooling
Hot air blast
Slag + coke
Slag/metal flow
Hearth
Carbon block
(iron + coke)
FIGURE 8.5
Schematic of internal processes within a blast furnace.
366 Advanced Heat Transfer
One of the final products of combustion is solid iron. During these combustion reactions, a
waste material called slag is formed as an upper, molten layer. The slag consists of oxides,
ash, and impurities such as silica, magnesia, and sulfur. Solidified slag is used in various
applications, such as phosphate fertilizers and aggregate material for road construction.
Slag is formed within the blast furnace when limestone decomposes and combines with
ash and other impurities.
The molar balance for the chemically reacting flow is given by:
Neglecting heat losses through the walls, the steady-state energy balance can be
written as:
2 o
2 o
Ṅ R,i hf + Δh − Ṅ P,j hf + Δh + Q̇ = 0 (8.76)
R,i P,j
i=1 j=1
or alternatively,
where H denotes the total enthalpy flow rate, the subscripts A − D refer to the reactant/
product constituents, and o and i refer to outlet and inlet, respectively. If there are uncon-
verted reactants, additional molar flow rates and enthalpy values are added to the above
equations in the product stream to account for unconverted reactants.
Chemically Reacting Flows 367
The enthalpy values of input reactants and output products are given as follows. For the
reactants,
Tp
HA,i = Ṅ A,i Δhof,A(g) + cp,A(g) dT (8.78)
298
Tp
HB,i = Ṅ B,i Δhof,B(s) + cp,B(s) dT (8.79)
298
Recall Δhf denotes the enthalpy of formation. The right side reflects the enthalpy change
due to the chemical reaction and the amount of heat required to raise the temperature of
the reactant from 25◦ C (STP) to the reaction temperature of the products.
For the products and unconverted reactants:
Tp
HC,o = Ṅ C,o Δhof,C(s) + cp,C(s) dT (8.80)
298
Tp
HD,o = Ṅ D,o Δhof,D(g) + cp,D(g) dT (8.81)
298
The enthalpy change due to the chemical reaction depends on the rate of reaction, r. The
unconverted reactants and product molar flow rates can be written in terms of the rate of
reaction as follows:
where ν is the stoichiometric coefficient. It is negative for the reactants and positive for the
products. Also, Ṅ A , Ṅ B , Ṅ C , and Ṅ D are the molar flow rates of reactants and products,
respectively.
The above energy balance and heat transfer rate, Q̇, were applied to the overall system and
control volume. To effectively supply or remove heat by an inert or participating gas that
moderates the vessel’s temperature, internal heat transfer between the solid particles and
gas stream must be determined. Therefore, convection correlations for heat transfer between
particles or a packed bed and a coflowing gas stream (Chapters 3 and 6) and models of radi-
ation exchange in participating media (Chapter 4) are needed.
The following Frössling–Ranz–Marshall correlation can be used to predict the rate of heat
transfer between a spherical solid particle and coflowing gas stream,
hDp
Nup = = 2 + 0.6 Re0.5 0.33
p Prg (8.86)
kg
368 Advanced Heat Transfer
where Rep is the particle Reynolds number based on the particle diameter, Dp, and relative
velocity between the particle and gas stream. Also, Nup is the particle Nusselt number and
Prg is the gas Prandtl number.
A transition occurs between flow regimes for gas flow past a single particle and a fixed bed
of particles. For Rep . 100, the Nusselt number falls between values for a single particle and
a fixed bed. For 0.1 ≤ Rep ≤ 100, Kunii and Levenspiel (1991) modified the Frössling–Ranz–
Marshall correlation for a fixed bed of particles as follows,
Another correlation including a Prandtl number dependence in the Nusselt number for a
fixed bed was presented by Chen (2003):
where hb, Ap, Tb, and Tp are the bed heat transfer coefficient, particle surface area, fixed bed
temperature, and particle temperature, respectively. This heat input is then added to the
heat of reaction between the fixed bed and the gas stream based on previous energy balances
in order to determine the net heat addition or removal that maintains a prescribed temper-
ature of the reaction vessel.
Al Purified Al
Salt layer
Tearing of removed or
surface film floats out
generates
salt droplets
Formation of AlCl3 gas bubbles:
3Cl2(g) + 2Al(l) → 2AlCl3(g)
FIGURE 8.6
Refinement of molten aluminum.
In these reactions, Cl2 gas bubbles react with liquid aluminum to form gas bubbles of
AlCl3, which subsequently react with Mg impurities in liquid aluminum to form purified
liquid aluminum and MgCl2 droplets. The gas bubbles generate a fine distribution of salt
particles in the diffusion layer around the bubbles. These particles must be removed during
the process.
or alternatively,
Ṅ F MF + Ṅ A MA = Ṅ P MP (8.96)
3 o
4 o
NR,i hf + h − h NP,j hf + h − h
o o
Q= − (8.97)
R,i P,j
i=1 j=1
For ideal gases, the enthalpies are a function of temperature alone. The values
of reactant and product enthalpies as a function of temperature can be obtained from
thermodynamic tables (see Appendices E, F, and I). Substituting the values of enthalpy
into Equation 8.97, the rate of heat released by combustion of the fuel droplets becomes:
ṁ
Q̇ = Q (8.98)
M
O2
CO/CO2
Lance
CO bubbles Iron droplets
FIGURE 8.7
Schematic of gas–liquid reactions in a basic oxygen furnace.
Chemically Reacting Flows 371
other products of combustion that are removed as scrap slag. The gas–liquid reactions for
carbon and sulfur impurities in the iron are given by (Guthrie 1993):
2C(l) + O2 (g) 2CO(g) (8.99)
In the former reaction, dissolved carbon within the liquid is transported to the gas–
liquid interface and subsequently removed as carbon monoxide through its chemical reac-
tion to form CO with the impinging oxygen jet. During this process of impurity removal
to produce carbon monoxide (CO) and slag (such as SiO2 and others), a substantial
amount of heat is transferred during the chemical reactions. In order to prevent overheat-
ing of the BOF, extra steel scrap (often up to 30% of the molten steel mass) is added as
a coolant. This additional scrap metal leads to other difficulties in the processing of the
molten steel, for example, by obstructing the heat and mass transfer processes in the
main region of the molten steel.
The molten steel is typically solidified into slabs or billets. The oxygen content in the mol-
ten steel is carefully controlled to minimize defect voids in the solidified material by provid-
ing a sufficiently high partial pressure of gas ahead of the phase interface during
solidification. Then the rate of formation of gas bubbles can be matched with the rate of vol-
umetric shrinkage of the metal to reduce or eliminate the formation of defects voids. The
larger voids are replaced with smaller dispersed voids, which can then be readily removed
by subsequent hot rolling and other operations.
where F refers to the element of the liquid metal droplet, such as zinc in zinc refinement pro-
cesses. During the fuming process, a diffusion layer of metallic vapor is formed around the
droplet in the presence of the surrounding oxygen flow. If the partial pressure of oxygen in
the gas flow is sufficiently high, then gases react to form an oxide fume of particles around
the droplet.
The nucleation of particles in the gas layer around a droplet leads to submicrometer size
particles that are difficult to filter out of the exhaust gases from the furnace. In some steel-
making operations, up to 1 ton of iron oxide dust can be produced for every 100 tons of steel.
This can lead to a significant environmental challenge and risk to safety of personnel.
372 Advanced Heat Transfer
Evaporating
Boundary layer
droplet of
of gas
liquid metal
Oxygen and
argon mixture
FIGURE 8.8
Fuming process in metallurgy.
In the fuming process of Equation 8.101, an outward diffusion flux of metallic vapor from
the droplet surface occurs simultaneously with an inward diffusion flux of oxygen from the
freestream gas flow. These species concentration lead to diffusion of F in the opposite direc-
tion to the diffusion of O2 (called equimolar counter-diffusion). It is assumed that the molar
fluxes of each component are equal, but opposite in direction, and the total pressure is cons-
tant. Therefore, on a per mole basis, this balance of diffusion fluxes under steady-state con-
ditions requires that:
′′ ′′
Ṅ F Ṅ O2
+ =0 (8.102)
a b/2
Define the thicknesses of the diffusion layers of both counter-diffusing gases (metallic
vapor and oxygen) as δF and δO2 , respectively. Also, assume that the oxygen concentration
reaches zero at some location near the surface of the droplet inside the metallic vapor diffu-
sion layer. This location corresponds with the reaction plane where vapor and oxygen react
to form solid oxide particles. Then using the ideal gas law, Fick’s law and Equation 8.102, the
net diffusion flux of metallic vapor becomes:
′′ 2a ρO2 DF
Ṅ F = − (8.103)
b δF − δO2
A fluidized bed is a two-phase apparatus in which a fluid flows at a sufficient velocity through
a fixed bed of particles such that the bed becomes loosened and the particle–fluid mixture
behaves like a fluid. When a fixed bed of particles is fluidized, the entire bed can be trans-
ported like a fluid if needed. A fixed bed is normally fluidized to allow better mixing of the
solids in contact with a fluidizing gas thereby allowing excellent heat transfer. A fluidized
bed is often used for chemically reacting flows because the surface reactions and heat trans-
fer can be significantly enhanced, in comparison to other types of reactors, since the gas is
exposed to a large effective solid surface area.
In a fixed bed, immobile solid particles are packed in layers on top of each another. They
exert gravitational forces involving individual particles and the entire bed weight at contact
points joining each solid material in a layer. These forces spread in all directions from the
contact points. A fluidized state occurs when a sufficiently high velocity of the fluid pene-
trates through the bed to separate parts of the fixed bed. When the fluid stream exceeds a
certain critical velocity (called the minimum fluidization velocity), the solid particles begin
moving and interacting with each other. A transition point occurs from a fixed bed to a flu-
idized bed. The mean distance between particles increases since the particles move to
occupy a larger portion of the volume enclosing the fluidized bed. The pressure drop
remains approximately constant (bed weight divided by an appropriate surface area) within
the fluidized bed. Homogeneous fluidization refers to a bed fluidized with liquids, whereas het-
erogeneous fluidization refers to a bed fluidized with a gas. Although this section will focus on
heterogeneous fluidization and the resulting gas–solid flows, similar correlations and solu-
tion methods are applicable to homogeneous fluidization.
A fluidized bed typically consists of a vertical cylinder or tank loaded with solid particles
and supplied with a gas through a perforated distributor plate at the bottom of the cylinder
(see Figure 8.9). Various engineering systems, such as coal gasification, polyethylene
Gas
Solid
Gas –
liquid
interface
Gas
Bed
bubble
height,
H
Solid
particle Solid
Gas Distributor
plate
FIGURE 8.9
Schematic of a fluidized bed.
374 Advanced Heat Transfer
A schematic of a fluidized bed and diffusion resistances are illustrated in Figure 8.10. The
bed consists of two regions—a bubble phase with a bulk concentration of constituent A of
CA,b; and an emulsion phase at a concentration of CA,e. It is assumed that gas–solid reactions
occur only in the emulsion phase. Heat and mass transfer resistances include the bubble to
CAo
Particle
Reaction
resistance Film
Solids Gas flow
gas
layer Exchange of
reactant gas Bubble
CAs CAe between phase
Dense phase
phases
(emulsion)
CAi
FIGURE 8.10
Diffusion resistances in a fluidized bed.
Chemically Reacting Flows 375
emulsion resistance, external film resistance around a solid particle, and an interparticle
resistance.
Define the concentration efficiency for constituent A as follows:
CA,i − CA,o
ηc,A = (8.105)
CA,i − CA,e
where the gaseous reactant, A, enters the inlet of the fluidized bed at a concentration of CA,i
and exits from the outlet at CA,o. From a mass transfer analysis (Haseli, Dincer, and Naterer
2008), it can be shown that:
NTU
ηc,A = 1 − β exp − (8.106)
β
where,
Kbe ζ b
NTU = (8.107)
Uo /Hb
Uo − Umf
β= (8.108)
Uo
Here ζb, Kbe, Uo, Umf, β, and Hb refer to the bed void fraction, overall mass transfer coeffi-
cient of gas exchange between the bubble and emulsion phases (in units of s−1), superficial
gas velocity (m/s), minimum fluidized velocity (m/s), dimensionless flow excess parameter,
and bed height (m), respectively.
Also, define the average fraction of conversion of solid reactant in a particle in the bed, xc,b,
as follows:
χ c,i − χ c,b
xc,b = (8.109)
χ c,i (1 − χ c,b )
where χc,i and χc,b represent the mass fraction of solid reactant, C, in the feed (inlet) and bed,
respectively. It will be assumed that all particles enter the bed with the same proportion of
conversion, χc,i, and exit at χc,b (average conversion of perfectly mixed particles in the bed).
The gas velocity at the minimum fluidization condition, Umf, can be estimated from the
Wen–Yu correlation as follows (Wen and Yu 1966):
ρg Umf Dp
Remf = = (33.7 + 0.0408Ar )1/2 − 33.7 (8.110)
μg
gD2p ρg (ρp − ρg )
Ar = (8.111)
μ2g
Here the subscripts g and p refer to the gas and particle, respectively. The Reynolds num-
ber at the minimum fluidization velocity corresponds to the condition where all particles in
the fixed bed begin floating.
376 Advanced Heat Transfer
The pressure drop across the bed, Δpb, can be determined based on the overall particle
weight (Hewitt, Shires, and Polezhaev 1997),
where ζmf is the porosity of the particulate (emulsion) phase at the state of minimum fluid-
ization. Porosity or void fraction is a measure of the empty void spaces in a fluidized bed. It
represents the fraction of the volume of voids over the total volume. It can be shown that the
pressure drop becomes nearly constant beyond the minimum fluidization velocity.
A splash zone is generated above the free surface in the fluidized bed as bubbles rise,
grow, and burst across the free surface. Particles are ejected into the space above the free sur-
face. As a result, the free surface is often wavy and chaotic rather than smooth and horizon-
tal. A large particle concentration exists below the free surface and this concentration drops
abruptly across this surface. Large particles with sufficient weight may drop back into the
bed. The remaining lighter particles are carried away from the bed by the gas stream.
When bubbles of particles depart from the surface, the presence of other bubbles causes
lateral motion and coalescence of bubbles within the bed. Various forces acting on a large
bubble may break it up prior to its arrival at the free surface. Bubbles are usually faster in
fluidized beds with small particles than beds with larger particles. Fluidized beds with
slower bubble velocities are often more desirable than fast flows because the entire gas
flow can mix fully during the chemical reaction. With fast bubbles, the gas may pass through
the fluidized bed without reacting with any particles.
The bubble rise velocity can be estimated by analogy to bubbles ascending in a liquid con-
tainer during pool boiling. The bubble rise velocity, Vb, can be approximated in terms of the
bubble diameter, Db, as follows (Hewitt et al., 1997):
Vb = 0.71 gDb (8.113)
Unlike bubbles that rise in pool boiling, upward gas flow in a fluidized bed keeps the par-
ticles located primarily along the upper section of the bubble. Bubbles interact with the gas
flow, creating a wake behind the bubble. Since overall mass is conserved, the upward par-
ticle movement in one section is accompanied by downward flow in another section.
The rise of the fluidized bed height due to this bubble motion is related to the volume frac-
tion occupied by the bubbles and the fluidized bed porosity. The ratio of the bed height, Hb,
to the minimum fluidization height, Hmf, can be expressed as:
Hb 1 − ζmf
= (8.114)
Hmf 1 − ζb
where the bed porosity, ζb, can be determined from the following Todes correlation,
0.21
Rep + 0.02Re2p
ζ b = ζ mf (8.115)
Remf + 0.02Re2mf
Since the gas velocity at the free surface may exceed the mean fluidization velocity, larger
particles may be ejected far from the free surface. Single particles or clusters of particles may
be ejected from the free surface. These particle clusters may break apart into smaller individ-
ual particles that later descend back down into the fluidized bed surface. Only small
Chemically Reacting Flows 377
particles will ascend out of the bed. The specific size and dynamic factors affecting this
ascent include forces such as the particle weight relative to the gas inertia.
Fewer particles are observed further from the surface because many particles have fallen
back to the free surface below those locations. The particle cluster density above the surface
decreases with height. There exists a critical or maximum height in the fluidized bed
whereby only particles with a terminal velocity less than the gas velocity are observed.
Above this height—called the transport disengaging height (TDH)—the particle concentration
becomes zero. The TDH can be determined from the following Geldart equation:
Re1.55
p
TDH = 1,200Hmf (8.116)
A1.1
r
This correlation is generally applicable for particle diameters between 0.075 and 2 mm.
The gas-to-particle heat transfer coefficient increases with Rep/ζb (as expected) since the
intensity of mixing is increased.
With chemical reactions between active particles and the fluidizing gas, the heat transfer
becomes strongly coupled with mass transfer between the particles and the bed. Mass trans-
fer occurs when the reacting gas diffuses toward the surface of the active particle. Both dif-
fusion and convection processes contribute to mass transfer. A widely used correlation for
estimating the mass transfer coefficient, hm, in a fluidized bed is the following La Nuaze–Jung
correlation (La Nuaze and Jung 1982):
1/2
Rep
Shp = 2ζmf + 0.69 Sc0.33 (8.119)
ζb
378 Advanced Heat Transfer
where Shp and Sc refer to the Sherwood and Schmidt numbers (based on the particle diam-
eter) for mass transfer, respectively. This correlation incorporates the movement of clusters
of particles carrying fresh gas from the fixed bed toward the active particles.
Although the above correlations are useful, their limitations should not be overlooked,
considering the complexity of the transport processes during the heterogeneous fluidiza-
tion. The processes involve complicated interactions involving chemical reactions, turbu-
lence, radiation, phase change, and others. As a result, experimental data is often
scattered and the accuracy of the previous correlations generally lies within a broad range
of +50%. In addition to the flow complexities, this wide scatter arises because the correla-
tions usually do not include contributing parameters such as bubble sizes, tube arrange-
ments, geometrical factors, among others. Further detailed analysis of transport processes
in fluidized bed systems was presented by Doraiswamy and Majumdar (1989).
CA,o
xg = 1 − (8.120)
CA,i
CA,e n
ηp = (8.121)
CA,i
where the subscripts i, e and o refer to inlet, emulsion phase, and outlet, respectively. The gas
conversion can be determined by:
xg = 1 − η1/n
p ηc (8.122)
where n denotes the order of the reaction and ηc is the concentration efficiency in
Equation 8.105.
The reactor contains particles that have spent different times inside the fluidized bed so
there is a distribution of conversion rates. Population balance equations (PBEs) are often
used to describe this evolution of a population of particles in a fluidized bed. PBEs may
be derived as an extension of the Smoluchowski coagulation equation (Smoluchowski
1916),which describes the coalescence of particles. However PBEs are more general as
they define how populations of separate particles develop over time. They are given by a
set of integral/differential equations that identify the mean-field behavior of a population
of particles from the analysis of the behavior of a single particle. The particle systems are
characterized by various stages of a particle transformation from its initial formation to
its subsequent decomposition from the chemically reacting flow. The PBE represents a bal-
ance of the number of particles of a particular state. Monte Carlo methods, discretization
methods, and moment methods have been used to solve population balance equations. Fur-
ther detailed analysis of population balances was presented by Ramkrishna (2000).
It can be shown that a population balance over the reactor yields the following distribu-
tion of solid conversion within the fluidized bed,
1 1 1 − χ c,i xc Θ(xc )
pb (xc ) = exp − (8.123)
Das F(xc ) 1 − χ c,i xc,i λ
Chemically Reacting Flows 379
kr,e Wb
Das = = kr,e tres (8.125)
Fi
kr,e Wb kr,e Wb
λ= = (8.126)
Fo Fi − rc,bed
Here kr,e, tres, Wb, and rc,bed represent the kinetic coefficient (units of s−1), mean residence
time (average time spent by a particle in the reactor in units of s), bed inventory (total
mass in kg), and the overall rate of reaction within the bed (kg/s). The inlet and outlet
flow rate of solids, in units of kg/s, are represented by Fi and Fo, respectively. The
Damköhler number (Da), named after Gerhard Damköhler (1908–1944), relates the chem-
ical reaction timescale (reaction rate) to the transport phenomena rate occurring in the
chemical reaction system.
Define F(xc) as a function that expresses the dependence of the conversion rate of a single
particle on xc and the particle effectiveness factor, ηp.
In order to determine the distributions of Θ(xc) and F(xc) in Equations 8.124 and 8.127, the
evolution of particle mass must be determined for the decomposition process during the
gas–particle reaction. Recall two models were developed earlier in this chapter to analyze
this decomposition process with diffusion of gaseous reactant into a particle: (1) Progressive
Conversion Model; and (2) Shrinking Core Model. These two models represent the limiting
extreme cases of mass transfer rates during the decomposition process. Table 8.1 shows the
resulting kinetic models for Θ(xc) and F(xc) in these two limiting cases.
The kinetic coefficient, kr,e, accounts for the concentration of the gaseous reactant and tem-
perature in the emulsion. It is calculated by:
Mp
kr,e = bki CnA,e (8.128)
ρp
where ki is the kinetic coefficient based on the particle volume, determined at the inlet con-
ditions, Mp is the molecular mass of the solid reactant, and b is the stoichiometric factor, or
mole ratio. The stoichiometric factor is the ratio of the coefficients of products and reactants
in the stoichiometric reaction.
TABLE 8.1
Limiting Cases of Two Kinetic Models
Progressive Conversion Model Shrinking Core Model
The overall rate of reaction within the bed, denoted by rc,bed, may be determined by the
product of the bed mass, reactivity, and solid reactant conversion in a particle, integrated
from the value of the conversion of solid reactant at the inlet, xc,i, to 1, as follows:
1
rc,bed = Wb R(s)pb (s) ds (8.129)
xc0
Das
rc,bed = 1− Fi (8.131)
λ
Also, performing a mass balance of nonreacted material of solid particles, S, and combin-
ing it with the expression for the conversion distribution of solids, pb(xc), it can be shown
that:
where,
1
1−s Θ(s)
f1 (xc,i , λ) = exp − ds (8.133)
xc0 F(s) λ
The conversion of solid particles and gaseous reactant can be related through an
overall mass balance on the solid particles and gaseous reactant and stoichiometry of
the reaction. The following expression is then obtained for the gas conversion, xg, by
equating the rate of consumption of solid particles, with the rate of consumption of gaseous
reactant,
1 Das
xg = 1− (8.134)
α λ
where Ac,bed is the bed cross-sectional area. Taking into account the average conversion of
solid particles within the bed, xc,b, and solving the distribution function, pb(xc), over the
bed, leads to:
f1 (xc,i , λ)
xc,b = 1 − (8.136)
λ
Chemically Reacting Flows 381
Combining this equation with the prior expression for f1(xc,i, λ) yields the following result
for the conversion of gaseous reactant to solid particles,
xc,b − xc,i
xg = (8.137)
α(1/χ c,i − xc,i )
Das rc,bed
=1− (8.138)
λ Fi
The overall rate of reaction, rc,bed, is determined from Equations 8.123, 8.129, and 8.130,
yielding:
1
rc,bed = Wb R(s)pb (s) ds (8.139)
xc,i
where,
1
Θ(s)
f2 (xc,i , λ) = exp − ds (8.141)
xc,i λ
Furthermore, by combining this result with Equation 8.140, and substituting χc,i = 1 and
xc,i = 0, the Damköhler number can be expressed as a function of λ as follows:
1
Das Θ(s)
= 1 − f2 (λ) = 1 − exp − ds (8.142)
λ 0 λ
382 Advanced Heat Transfer
From Table 8.1, the kinetic functions for the progressive conversion model are:
Fp (xc ) = 1 − xc (8.143)
1
ln(1 − xc ) λ
f2 (λ) = exp ds = (8.145)
0 λ λ + 1
Hence, the Damköhler number is obtained by substituting this result into Equation 8.142,
Das 1
= (8.146)
λ λ+1
The overall rate of the reaction can be determined by combining this result with Equation
8.138, yielding:
λ
rc,bed = Fi (8.147)
λ+1
In summary, the following solution procedure is used to evaluate the conversion of solid
particles and gas based on the progressive conversion model.
Through this trial-and-error procedure, the conversion of reactants and spatial distribu-
tion of the conversion of reacting particles in the fluidized bed can be determined.
The shrinking core model (SCM) is the other limiting case. Recall the gas–solid reaction
occurs first at the outer surface of the particle in the shrinking core model. Then the reaction
zone moves into the solid and leaves behind a completely converted “ash” material and an
unreacted core that shrinks in size during the reaction. Assuming that the particle decom-
position follows this process, the kinetic functions in Table 8.1 are used as follows:
Evaluating this integral and combining the result with Equations 8.142 and 8.125 yields:
2 2
λ 2 λ λ 3
rc,bed = Fi λ 1− + −2 exp − (8.151)
3 3 3 λ
A solution procedure is also needed for this model to evaluate the conversion of solid and
gas reactants. In this case, the procedure uses the following steps.
Unlike the solution for the Progressive Conversion Model, the solution procedure for the
Shrinking Core Model requires two trial-and-error loops.
Sample results of solid/gas conversion and reaction rates for a noncatalytic reaction of
cupric chloride particles with superheated steam in a fluidized bed are shown in Figure 8.11.
The bench-scale fluidized bed reactor has a diameter of 2.66 cm and height of 16 cm. The
average particle diameter is 2.5 mm. The fluidized bed inventory is 15 g and the inlet flow
rate of solids of 0.9 g/s. The predictions of two kinetic models (shrinking core model and
progressive conversion model) are compared against each other under identical process
conditions. Figure 8.11 indicates that the predicted conversion of particles and fluidizing
steam are higher with the shrinking core model. Also, the shrinking core model predicts a
higher overall reaction rate.
After these parameters are determined, the overall effectiveness of the fluidized bed, ηbed,
can be determined. The bed effectiveness represents the overall performance of the bed in
converting reactants to products. The best performance of the bed occurs when all moles
of reactants are converted to products so there are no unreacted particles or gas at the
384 Advanced Heat Transfer
0.6
(a) (b)
Shrinking core model
Shrinking core model
Progressive conversion model Progressive conversion model
0.6 0.5
0.2 0.3
Gas conversion
0.0 0.2
0 0.2 0.4 0.6 0.8 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Superficial velocity, U (m/s) Superficial velocity, U (m/s)
FIGURE 8.11
(a) Solid/gas conversion and (b) reaction rates for a noncatalytic reaction of cupric chloride particles with super-
heated steam in a fluidized bed. (Adapted from Y. Haseli et al. 2008. Chemical Engineering Science, 63: 4596–4604.)
exit. The bed effectiveness can be represented by a ratio of the total number of moles of reac-
tants that are converted to products to the total reactant moles at the inlet stream:
i (NR,i )inlet − i (NR,i )outlet
ηbed = (8.152)
i (NR,i )inlet
Np,o Cp,o
= = 1 − xc,b (8.153)
Np,i Cp,i
Ng,o Cg,o
= = 1 − xg (8.154)
Ng,i Cg,i
Thus the net effectiveness can be determined after calculating the conversions of reactants
within the fluidized bed reactor.
As the previous analysis has shown, fluidization processes are complex and depend on
many factors. Although the analysis has captured most of the key parameters, there are still
several others not considered here such as the detailed physical characteristics of the solid
particles. For example, the particle shapes and sizes affect the aerodynamic properties of
the solids in the gas stream. The terminal velocity of a group of particles is related to the
size distribution of particles and relative interactions between gravitational and drag forces
on the particles. The solid particles may be floating, motionless, or actively moving in the
fluidized bed. Particle movement can be free to move within the central core of the bed
but mechanical devices such as perforated plates are often added, which obstruct the parti-
cle motion in the outer regions. The reader is referred to other sources for more detailed anal-
ysis of fluidized bed reactors such as Grace, Knowlton, and Avidan (2012).
Chemically Reacting Flows 385
PROBLEMS
8.1 Derive Equation 8.5 by using the ideal gas law to show that the mole fraction is equal
to the volume fraction (Amagat–Leduc law of additive volumes) as well as the pressure
fraction (Dalton’s law of additive pressures).
8.2 For an ideal gas mixture, show that R = Σi χiRi. Also, show that yi/χi = Ri/R, where Ri
and R refer to the gas constants of component i and the entire mixture, respectively.
8.3 For an ideal gas mixture, show how the specific entropy of component i can be
computed in terms of the mole fraction of component i, temperature and total
pressure.
8.4 A mixture of 60% nitrogen and 40% propane (by volume) flows through a well-insu-
lated compressor. The inlet conditions of the gas mixture are 30◦ C and 0.4 MPa. Find
the input power, per unit mass of mixture, required to compress the steady, nonreact-
ing flow to outlet conditions of 100◦ C and 1.2 MPa. For propane, cp = 1.68 kJ/kgK and
M = 44.1 kg/kmol, and for nitrogen, cp = 1.04 kJ/kgK and M = 28 kg/kmol.
8.5 For the previous question, find the specific entropy change of each gas component
separately (i.e., propane and nitrogen separately) between the inlet and outlet of
the compressor. Which gas component in the mixture undergoes a larger specific
entropy change? Give a physical explanation of this difference of computed specific
entropy values.
8.6 A rigid tank is initially divided by a partition into two sides. The left side contains
methane (component A, mass of mA and NAmoles) and the right side contains nitro-
gen (component B, mass of mBand NB moles). Both sides are initially at a temperature
of T and pressure of P. Then the partition is removed and both sides of the insulated
tank are completely mixed. Find the total entropy change of the gases due to the mix-
ing process. Verify that this entropy change complies with the second law of
thermodynamics.
8.7 Explain how the adiabatic flame temperature, Tad, can be calculated for the combus-
tion of methane (CH4) at STP in air. Show the relevant chemical reaction and energy
balance equations. Describe the steps required to calculate Tad.
8.8 Fuel and air enter a mixing chamber and react to form products of combustion. Find
the amount of heat released by combustion while raising the products of combustion
to an outlet temperature of TP. Express the result in terms of the enthalpies of the reac-
tants and products of combustion.
8.9 Significant variations can be observed between adiabatic flame temperatures
obtained from actual experiments, tabulated values, and calculations based on a the-
oretical analysis in this chapter. Explain why these variations may occur and how
more accurate calculations can be carried out in the previous question.
8.10 A stoichiometric mixture of liquid hydrogen and oxygen react in a low temperature
combustion chamber. Exhaust gases are released at a rate of 60 kg/s from the cham-
ber. The cylindrical reactor has a length of 60 cm and diameter of 50 cm. Find the reac-
tion rates of hydrogen and oxygen for the complete stoichiometric combustion
reaction.
8.11 The average minimum daily requirement of food energy intake is about 1,800 kilocal-
ories (7,500 kJ) per person, according to the Food and Agriculture Organization of the
United Nations. Assume the food intake is sugar in the form of glucose with a
386 Advanced Heat Transfer
on how CO2 gas can diffuse through the gas boundary layer surrounding the pellet.
Use Fick’s law to estimate the required species concentration of CO2 in the bulk gas
phase (called C∞) to provide a specified steady-state diffusive flux, Ṅ s , at the surface
of a spherical pellet (at r = R).
8.20 During thermal decomposition of spherical limestone pellets (CaCO3), solid crystals
of CaO and CO2 gas are formed. Using first-order kinetics of the chemical reaction
with a proportionality constant of 0.02 mm/s, estimate the time taken for a pellet to
be decomposed to one half of its initial radius of 1 cm.
8.21 Heat is released in a process of desulfurization of a liquid iron alloy. Sulfur is sepa-
rated out and removed from a molten iron alloy during the magnesium injection
operation. During the process, magnesium gas bubbles react with sulfur in the mol-
ten alloy to form particles of magnesium sulfide which rise to the surface of the con-
tainer. Derive an expression for the amount of heat released by this desulfurization
process in terms of the enthalpies of formation and sensible enthalpies of the product
and reactants.
8.22 In a dehydration process, water is removed from a substance. Under what conditions
of pressure and temperature can water be separated into ice and water vapor? Give
an example of a basic procedure to remove water from a substance by phase transfor-
mations to solid and gas phases.
8.23 An experimental apparatus has been proposed for the evaluation of the effective
thermal conductivity of a fluidized bed at elevated temperatures. The apparatus
consists of a furnace enclosure with internal heating elements above the packed
bed; a packed bed of polished alumina spheres at the bottom of the enclosure; a
series of thermocouples inside the packed bed; and a lower heat exchanger
below the packed bed with a coolant water flow. The heat exchanger directly
contacts the packed bed with a surface area of 0.3 × 0.3 m. It extracts heat
from the radiant heaters by circulating cold water through internal channels.
Measurements of the water temperature rise between the exchanger inlet and
outlet and the mass flow rate indicate that ΔT = 1.3◦ C and ṁ = 0.05 [kg/s],
respectively. Also, the following thermocouple measurements at vertical locations
within the bed were recorded: 300 K at the wall, 376 K at 2 cm, 444 K at 4 cm,
517 K at 6 cm, and 590 K at 8 cm.
a. Estimate the effective thermal conductivity of the packed bed. State the
assumptions in your analysis.
b. Discuss the factors that may lead to experimental errors and suggest
approaches to minimize these errors.
8.24 A catalytic reactor contains an airflow that passes through a packed bed of spherical
nickel pellets in a hydrogenation process for petroleum refining. Consider a typical
packed-bed flow that operates under the following conditions: void fraction of 0.4;
superficial gas velocity of 5 m/s; pellet diameter of 4 mm; and an incoming gas
(air) temperature of 350◦ C. The initial temperature of pellets in the packed bed is
30◦ C. What temperature will these pellets reach after the air flows through the packed
bed for 10 seconds?
8.25 What thermophysical properties of solid particles affect the characteristics and
parameters of fluidized beds? Describe how these properties affect the relevant
transport processes.
388 Advanced Heat Transfer
8.26 In a fluidized bed, how can the transition from the fixed-bed regime to the fluidized
bed regime be detected? Explain your response by referring to specific measurements
that can be taken within a fluidized bed.
8.27 What similarities exist between particle movement in fluidized beds and two-phase
flows of liquid–gas or solid–liquid mixtures?
8.28 Thermal processes in a fluidized bed are controlled so that the temperature difference
between points at two selected heights does not exceed 5◦ C. Heat transfer coefficients
to particles in the bed are generally less than 25 W/m2K. What mechanism(s) are
primarily responsible for the rate of heat transfer between particles and the
fluidizing gas?
References
J.C. Chen. 2003. “Heat Transfer,” Chapter 10, in Handbook of Fluidization and Fluid Particle Systems, W.C.
Yang, Ed., Boca Raton: CRC Press/Taylor & Francis, pp. 257–286.
V.N. Daggupati, G.F. Naterer, and I. Dincer. 2011. “Convective Heat Transfer and Solid Conversion
of Reacting Particles in a CopperII Chloride Fluidized Bed,” Chemical Engineering Science, 66:
460–468.
L.K. Doraiswamy and A.S. Majumdar. 1989. Transport in Fluidized Particle Systems, Amsterdam: Elsev-
ier Science Publishers.
H.S. Fogler. 2016. Elements of Chemical Reaction Engineering, 5th Edition, Upper Saddle River, NJ: Pren-
tice Hall.
D. Geldart. 1973. “Types of Gas Fluidization,” Powder Technology, 7: 285–292.
H.I. Gelperin and V.G. Einstein. 1971. “Heat Transfer in Fluidized Beds,” in Fluidization, J.F. Davidson
and D. Harrison, Eds., Cambridge: Academic Press, pp. 471–450.
A. Gómez-Barea, B. Leckner, D. Santana, and P. Ollero. 2008. “Gas-Solid Conversion in Fluidized Bed
Reactors,” Chemical Engineering Journal, 141: 151–168.
J.R. Grace, T.M. Knowlton, and A.A. Avidan, Eds. 2012. Circulating Fluidized Beds, Berlin: Springer
Science and Business Media.
R.I.L. Guthrie. 1993. Engineering in Process Metallurgy, Oxford: Oxford University Press.
Y. Haseli, I. Dincer, and G. F. Naterer. 2008. “Hydrodynamic Gas-Solid Model of Cupric Chloride
Particles Reacting with Superheated Steam for Thermochemical Hydrogen Production,”
Chemical Engineering Science, 63: 4596–4604.
G.F. Hewitt, G.L. Shires, and Y.V. Polezhaev, Eds. 1997. International Encyclopedia of Heat and Mass
Transfer, Boca Raton: CRC Press/Taylor & Francis.
J.R. Howard, Ed. 1983. Fluidized Beds Combustion and Applications. London: Applied Science
Publishers.
D. Kunii and O. Levenspiel. 1991. Fluidization Engineering, 2nd Edition, London: Butterworth-
Heinemann.
R.D. La Nauze and K. Jung. 1982. “The Kinetics of Combustion of Petroleum Coke Particles in a
Fluidized Bed Combustor,” 19th International Symposium on Combustion, Haifa, Israel,
1087–1092.
O. Levenspiel. 1999. Chemical Reaction Engineering, 3rd Edition, New York: John Wiley & Sons.
M.J. Moran, H.N. Shapiro, D.D. Boettner, and M.B. Bailey. 2014. Fundamentals of Engineering Thermo-
dynamics, 8th Edition, New York: John Wiley & Sons.
D. Ramkrishna. 2000. Population Balances: Theory and Applications to Particulate Systems in Engineering,
Cambridge: Academic Press.
Chemically Reacting Flows 389
K. Sinha and D.S. Reddy. 2011. “Effect of Chemical Reaction Rates on Aeroheating Predictions of
Reentry Flows,” AIAA Journal of Thermophysics and Heat Transfer, 25: 21–33.
M. Smoluchowski. 1916. “Drei Vorträge über Diffusion, Brownsche Molekularbewegung und
Koagulation von Kolloidteilchen” in German, Zeitschrift für Physik, 17: 557–571.
C.Y. Wen and Y.H. Yu. 1966. “A Generalized Method for Predicting the Minimum Fluidization
Velocity,” AIChE Journal, 12: 610–612.
9
Heat Exchangers
9.1 Introduction
Heat exchangers are engineering devices that transfer thermal energy between two or more
fluids, or between a solid object and a fluid, at different temperatures. The fluids may be sep-
arated by walls to prevent mixing or they have direct contact with each other. Heat exchang-
ers are used widely in space heating, power plants, refrigeration systems, air conditioning,
and chemical plants, among many others. Most energy systems in industry include one or
more heat exchangers. Empirical correlations and numerical methods are normally required
for a detailed analysis of heat exchangers. In some limited cases (to be presented in this chap-
ter), semi-analytical methods can be used to predict the temperature distributions, heat
transfer rates and thermal effectiveness of heat exchangers.
The most common types of heat exchanger configurations are concentric tube (or tubular),
cross-flow, and shell-and-tube heat exchangers. A concentric tube heat exchanger consists of
two fluid streams whereby an internal fluid flows through the inner tube and an external
flow passes through the annular region between the inner and outer tubes. If the outer fluid
flows in the same direction as the inner flow, then the configuration is called a parallel flow,
otherwise it is a counterflow heat exchanger (see Figure 9.1).
Cross-flow heat exchangers typically consist of an outer flow moving across tubes carry-
ing fluid that flows in a direction perpendicular to the cross flow. In many cases, the tubes
are covered with fins or other annular attachments to enhance the rate of heat transfer
between the different fluid streams. If the cross-flow streams are separated from one another
(e.g., fins separating fluid streams), then the configuration is unmixed. Otherwise a mixed
configuration permits complete mixing of the fluid streams in the external cross flow.
A shell-and-tube heat exchanger consists of an outer shell pipe where fluid enters through
one end, passes across internal tubes carrying a fluid at a different temperature, and exits
through the other end. A shell-and-tube heat exchanger with one shell and one tube pass is
illustrated in Figure 9.2 and a configuration with one shell and two tube passes is shown in
Figure 9.3. Baffles are usually placed perpendicular to the inner tubes to enhance mixing
and turbulence of the outer fluid stream. Baffles are perforated plates that obstruct some region
of the outer flow while directing the inner flow around the remaining uncovered sections. A
common example of a shell-and-tube heat exchanger is a condenser. The outer flow is steam
that condenses and leaves as water while transferring heat to the inner tubes carrying
cold water.
The packing of tubes within heat exchangers involves a wide range of possible surface
area density configurations (i.e., number and diameter of tubes) which vary by application.
For noncircular tubes, the hydraulic diameter is used; Dh = 4 As/P, where As and P refer to
the surface area and perimeter, respectively. For example, a typical range is 0.8 , Dh , 5 cm
391
392 Advanced Heat Transfer
Cross-flow Cross-flow
FIGURE 9.1
Concentric tube and cross-flow heat exchangers.
for shell-and-tube heat exchangers, 0.2 , Dh , 0.5 cm for automobile radiators, and 0.05 ,
Dh , 0.1 cm for gas turbine regenerators. In biological systems such as human lungs, heat
exchange occurs in the range of 0.01 , Dh , 0.02.
In this chapter, the design and analysis of various types of heat exchangers will be inves-
tigated. Both single and multiphase flows will be considered. Topics to be covered include
the governing equations and solution methods for tubular, cross-flow and shell-and-tube
heat exchangers; effectiveness–NTU method; thermal response to transient temperature
changes; and condensers and evaporators.
Tube Shell
outlet inlet
FIGURE 9.2
Shell-and-tube heat exchanger (one shell pass and one tube pass).
Heat Exchangers 393
Tube
Tube
inlet
inlet
Tube
outlet
Tube
outlet Baffle
One shell/two tube
Shell outlet Shell outlet
passes
FIGURE 9.3
Shell-and-tube heat exchanger (one shell pass and two tube passes).
r = ro Convection
Convection
Th,2
Tc,2 ΔT1 ΔT1
Th,2
Tc,1 Tc,1 ΔT2
Tc,2
FIGURE 9.4
Thermal resistances in a concentric tube heat exchanger.
394 Advanced Heat Transfer
the axial x-direction and heat transfer occurs in the radial r-direction. Hot fluid at a
temperature of Tf(x) in the outer tube annulus transfers heat to a cold fluid at Tc(x) in the
inner tube.
Thermal resistances to heat transfer include convection, fouling (due to fluid impurities,
rust, and/or deposit formation on the tubes), and conduction through the pipe wall. Based
on all thermal resistances in series in the radial direction (see Figure 9.4), the overall heat
transfer coefficient, U, can be written as:
1 1 Rf ,h ln(ro /ri ) Rf ,c 1
= + + + + (9.1)
UA hh A A 2πkL A hc A
where the subscripts c and h refer to the cold and hot streams, respectively. Also, h, A, L, Rf,h,
and Rf,c refer to the convection coefficient, surface area of the inner pipe, pipe length, and
fouling resistances of the hot and cold sides of the inner pipe wall, respectively. On the right
side of Equation 9.1, the component resistances represent convection (outer hot fluid
stream), fouling (outer surface), conduction through the inner pipe wall, fouling (inner sur-
face), and convection (inner cold fluid stream). Periodic cleaning of the heat exchanger sur-
faces should be performed to reduce and minimize the adverse effects of fouling. Surface
fouling leads to a higher pressure drop and lower heat transfer effectiveness.
Consider a steady-state energy balance for a differential control volume of width dx in
Figure 9.4. In the control volume of the outer tube,
where the temperatures represent mean temperatures at position x. The differential heat
flow, dq, from the outer tube to the inner tube balances the enthalpy change of the fluid
over the distance dx. Performing a similar heat balance within the inner tube, combining
with Equation 9.2, and using the overall heat transfer coefficient from Equation 9.1, leads to:
dq dq U(Th − Tc ) dA U(Th − Tc ) dA
− − = dTh − dTc = − − (9.3)
ṁh ch ṁc cc ṁh ch ṁc cc
By rearranging, the total rate of heat transfer from the hot fluid stream to the cold stream
between the inlet and outlet can be rewritten as:
q = UAΔTlm (9.6)
Heat Exchangers 395
ΔT2 − ΔT1
ΔTlm = (9.7)
ln(ΔT2 /ΔT1 )
This result was obtained for a parallel flow configuration. A similar result can be derived
for a counterflow heat exchanger. The same result in Equation 9.6 may be used in a counter-
flow configuration, except that the variables are replaced by ΔT1 = Th,1 − Tc,1 and
ΔT2 = Th,2 − Tc,2.
In the previous energy balances, thermal resistances were required, including the con-
vection and fouling coefficients. If these are unknown and instead the mass flow rates
and inlet/outlet temperatures are given, an alternative approach can be used by performing
heat balances on the entire tube. For the inner tube,
qc = ṁc cc Tc,o − Tc,i (9.8)
where cc is the specific heat of the cold stream. This energy balance states that the total heat
transfer to the cold fluid stream in the inner tube equals the net change of enthalpy flow rates
between the inlet and outlet. Similarly, for the outer tube,
where the subscripts h, o, and i refer to hot, outlet, and inlet, respectively. Assuming that heat
losses to the surroundings are negligible, then qh = qc = q. Equations 9.8 and 9.9 represent
heat balances performed for the cold and hot fluid streams, respectively, over the entire
heat exchanger from the inlet to the outlet.
For a parallel flow arrangement in Figure 9.4, the highest temperature difference occurs
between the two incoming fluid streams. Heat transfer from the hot stream to the cold
stream reduces the temperature difference between the fluids in the axial flow direction.
In contrast, for the counterflow arrangement, the temperature difference increases in the
flow direction (as viewed by the cold fluid stream). The temperature of the incoming cold
fluid stream increases due to heat transfer from the hot stream flowing in the opposite direc-
tion. If the same inlet and outlet temperatures are used, then the log mean temperature dif-
ference of the counterflow arrangement exceeds the value for a parallel flow heat exchanger.
Thus, a counterflow heat exchanger is usually more effective since a smaller surface area is
required to achieve the same heat transfer.
Often finned surfaces are used for heat transfer enhancement. Finned surfaces can be rep-
resented in Equation 9.1 by multiplication of the denominators of the convection terms by an
overall surface efficiency of the finned surface, ηo. This efficiency includes the efficiency of
heat exchange through the fin as well as heat transfer through the base surface between
the fins (Chapter 2),
Af
ηo = 1 − (1 − ηf ) (9.10)
A
where ηf, Af and A refer to the fin efficiency, total fin area, and total surface area of the finned
surface, respectively. For example, the fin efficiency for a uniform fin with an insulated tip
396 Advanced Heat Transfer
tanh(mL)
ηf = (9.11)
mL
Ti − To
R= (9.13)
t o − ti
t o − ti
P= (9.14)
T i − ti
Ti
F
ti
R = (Ti–To)/(to–ti)
decreasing
R1 R2 < R1 to
Baffle
P = (to–ti)/(Ti–To)
To
FIGURE 9.5
Correction factors for shell-and-tube heat exchangers.
Heat Exchangers 397
where lowercase and uppercase values of temperature refer to inner flow (tube) and outer
flow (shell), respectively. The subscripts o and i refer to outlet and inlet, respectively. For
example, ti refers to the inlet temperature of the tube flow and To denotes the outlet temper-
ature of the shell flow. As (to − ti) increases, at a fixed value of P, R decreases, F increases, and
thus q also increases. Correction factors for a variety of heat exchanger configurations have
been presented in several sources, for example, Bowman, Mueller, and Nagle (1940) and
Bejan (2013). Also, the Standards of the Tubular Exchange Manufacturers Association (Harrison
1999) presents a wide range of correction factors.
The correction factors are available in both graphical form and algebraic relations. For
example, the correction factor for a one-pass shell side, with any multiple of two tube passes,
can be expressed by:
√
R2 + 1/(R − 1) · log[(1 − P)/(1 − PR)]
F= √ √ (9.15)
log a + R2 + 1 / a − R2 + 1
For two shell passes and any multiple of four tube passes,
√
R2 + 1/(2R − 2) · log[(1 − P)/(1 − PR)]
F= √ √ (9.16)
log a + b + R2 + 1 / a + b − R2 + 1
where,
2 2
a= − 1 − R; b= (1 − P)(1 − PR) (9.17)
P P
From these relations, as expected, a higher temperature drop within the tube flow
corresponds to enhanced heat transfer to the external flow outside the tube. Various other
physical trends can be observed based on the relationships among F, R, P, and the heat trans-
fer rate. The following example demonstrates how the correction factors are applied.
1.58 × 106
Tc,o = 20 + = 145.5◦ C (9.20)
3(4,195)
398 Advanced Heat Transfer
The outlet temperature can then be used to compute the heat exchange factors as
follows:
145.5 − 20
P= = 0.39 (9.21)
340 − 20
340 − 180
R= = 1.27 (9.22)
145.5 − 20
which together yield a correction factor of F ≈ 0.98. Then using Equation 9.12, the surface
area can be obtained as:
Since the total surface area of tubes is A = 2πDL and L = 60 m, the required tube diam-
eter is 7.02/(2π60) = 1.9 cm, which corresponds to a tube diameter of approximately 3/4 in.
In the calculations of the heat transfer rate based on the log mean temperature difference,
axial heat conduction was neglected in the wall. In practice, axial temperature gradients
within the wall of the tube lead to some axial heat conduction and a reduced mean temper-
ature difference. Depending on the axial temperature gradient relative to the radial conduc-
tance, this effect may lead to a difference of up to +5% between the hot and cold fluid streams.
The pressure drop associated with cross flow across a group of uniformly spaced finned-
tube banks in a heat exchanger can be expressed as follows:
2
1 G2 Aff ρi ρi A
Δp = 1+ −1 +f (9.24)
2 ρ2i Afr ρo ρm Aff
where the subscripts i, o, and m refer to inlet, outlet, and mean (average) values, respectively.
Also, G, f, A, Aff, and Afr refer to the maximum mass velocity (density multiplied by maxi-
mum velocity), friction factor, total heat transfer surface area, minimum free-flow area of the
finned passages (cross-sectional area perpendicular to the flow direction), and frontal area of
the heat exchanger, respectively. Friction factors and Colburn j factors were discussed in
Chapter 3 and have been documented extensively by Kays and London (1984) for a variety
of heat exchangers, including finned and various tubular configurations.
In heat exchanger design, often the pressure drop and heat transfer enhancement have
competing influences. Heat transfer rates can be enhanced by increasing the packing of
tubes within the heat exchanger or using baffles or other heat enhancement devices. How-
ever, this occurs at the expense of a higher pressure drop, which requires more pumping
power to move the fluid through the system at a prescribed mass flow rate. Conversely,
fewer heat exchange tubes can lead to a smaller pressure drop, but often at the expense of
lower heat transfer, in comparison to a design with a high surface area density. An optimal
solution is desirable to provide the most effective balance between heat exchange and
pressure losses. This optimal condition can be determined based on the second law of
thermodynamics through entropy generation minimization (Chapter 3). The total entropy
generation can be minimized with respect to a key design parameter, such as the packing
density or diameter of tubes, yielding the most effective balance between heat exchange
and pressure losses.
Heat Exchangers 399
q
ε= (9.25)
qmax
As outlined earlier, the maximum heat transfer rate is achieved in a counterflow arrange-
ment, when one of the fluids experiences the maximum possible temperature difference,
where c = cv = cp refers to the fluid specific heat. From Equations 9.8 and 9.9, the fluid with
(ṁc)min has a larger temperature change.
Define the heat capacity rates are follows:
Cc = (ṁc)c (9.27)
Ch = (ṁc)h (9.28)
where the subscripts c and h refer to the cold and hot fluid streams, respectively. Then
combining Equations 9.8, 9.9, and 9.26,
Tc,o − Tc,i
εc = (9.30)
Th,i − Tc,i
Th,i − Th,o
εh = (9.31)
Th,i − Tc,i
The maximum heat capacity rate is Cmax = (ṁc)max and the minimum heat capacity rate
is Cmin = (ṁc)min. Furthermore, define the number of transfer units, NTU, as:
UA
NTU = (9.32)
Cmin
400 Advanced Heat Transfer
where U is the overall heat transfer coefficient. The effectiveness of a particular heat
exchanger, ɛ, is often tabulated and graphically illustrated as a function of Cmin/Cmax
(denoted by Cr) and NTU.
Sample curves of the effectiveness for a number of different heat exchanger configurations
are illustrated in Figure 9.6. Also, algebraic expressions involving ϵ, NTU, and Cr have been
(a) 1 (b) 1
Cr = 0 0.5
0.8 0.8
Cr = 0 0.99
0.5
0.6 0.6
ε
ε
1.0
0.4 0.4
0.2 0.2
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
NTU NTU
Parallel flow heat exchanger Parallel flow heat exchanger
Cr = Cmin/Cmax Cr = Cmin/Cmax
(c) (d)
1 1
Cr = 0.01
0.8 0.8 0.5
Cr = 0 0.5 1.0
0.6 0.6
ε
1.0
0.4 0.4
0.2 0.2
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
NTU NTU
Shell-and-tube flow heat exchanger Single pass, cross-flow heat
(1 shell pass, 2, 4, ... tube passes) exchanger (both fluids unmixed)
(e) 1 (f ) 1
Cr = 0.01 Cr = 0.01
0.8 0.8
0.5 0.5
0.6 0.6
1.0
ε
1.0
0.4 0.4
0.2 0.2
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
NTU NTU
Single pass, cross-flow heat exchanger, Single pass, cross-flow heat exchanger,
Cmax(mixed), Cmin(unmixed) Cmax(unmixed), Cmin(mixed)
FIGURE 9.6
Effectiveness curves of various heat exchangers. (Adapted from T.L. Bergman et al. 2011. Fundamentals of Heat and
Mass Transfer, 7th Edition, New York: John Wiley & Sons.)
Heat Exchangers 401
developed and summarized for a variety of heat exchangers by Kays and London (1984).
The following example demonstrates these concepts further.
q = ṁc cc (Tc,o − Tc,i ) = ṁh ch (Th,i − Th,o ) = 0.4(2,100)(55 − 35) = 16,800 W (9.33)
16,800
Tc,o = 10 + = 14◦ C (9.34)
1(4,181)
55 − 35
P= = 0.44 (9.35)
55 − 10
14 − 10
R= = 0.2 (9.36)
55 − 35
UA 525.9
NTU = = = 0.83 (9.41)
Cmin 630
402 Advanced Heat Transfer
Based on the parameters of Cr and NTU, the effectiveness curves (Figure 9.6)
yield ɛ = 0.55.
Then this effectiveness can be used to determine the oil inlet temperature,
Th,i − Th,o
ε = εh = (9.42)
Th,i − Tc,i
or alternatively,
Th,o − εh Tc,i 25 − 0.55(10)
Th,i = = = 43.3◦ C (9.43)
1 − εh 1 − 0.55
Temperatures above this maximum value will likely yield an outlet oil temperature
above the prescribed 25◦ C under the modified operating conditions.
The number of transfer units, NTU, can be interpreted as the ratio of the total heat conduc-
tance (or reciprocal of the total thermal resistance, R) to the minimum heat capacity rate.
Also, a timescale can be defined to explain the meaning of NTU. If a specified amount of
mass, m, resides in the heat exchanger at some stage of time and the rate of mass flow
through the heat exchanger (minimum heat capacity fluid) is ṁ, then the ratio of m/ṁ rep-
resents a mass (or fluid) residence time, tres, in the heat exchanger. Based on this interpreta-
tion, Equation 9.32 can be rewritten as:
UA UA 1 tres
NTU = = = = (9.44)
Cmin (ṁc)min (1/UA)(mc/tres )min (Rmc)min
The units of the denominator are units of time. Therefore, NTU can be interpreted as the
ratio of the fluid resistance time to a representative time constant of the Cmin fluid. This time
constant indicates a representative time required for the fluid to experience a fixed change in
temperature when a unit amount, such as 1 J, of heat is transferred to the fluid. For example,
values of 0.45 , NTU , 1 are representative of automotive radiators, and other ranges of
NTU can be used to identify the range of other heat exchanger applications.
The following algebraic expressions corresponding to the heat exchanger configurations
in Figure 9.6a–f have been presented by Kays and London (1984). The Cr is given
by Cmin/Cmax.
c. Shell-and-tube (one shell pass; two, four, … tube passes): Figure 9.6c
⎡ ⎤−1
1 + exp −NTU 1 + C2r
ε = 2⎣1 + Cr + 1 + C2r × ⎦ (9.47)
1 − exp −NTU 1 + C2r
Heat Exchangers 403
Fins can substantially improve the heat transfer rates from heat exchanger surfaces. The
following example demonstrates how fins can be incorporated into the previous analysis.
ṁa 1
G= = = 20 kg/s m2 (9.51)
Ac 0.05
Dh G 0.01(20)
Re = = = 11,173 (9.52)
μ 1.79 × 10−5
Then based on the specified Colburn factor and Chilton–Colburn analogy (Chapter 3),
jGcp 0.02(20)1,007
h= = = 511 W/m2 K (9.53)
Pr2/3 0.72/3
Since the fin efficiency is 0.94, the total thermal resistance on the air side of the heat
exchanger becomes:
1 1
Ro = = = 0.0013 (9.54)
ηo hA 0.94(511)1.6
For water flow within the tubes, the following thermophysical properties are
evaluated at 90◦ C or 365 K: cp = 4,209 J/kgK, ρ = 963 kg/m3, k = 0.677 W/mK, Pr = 1.91,
404 Advanced Heat Transfer
NuD = 0.023Re0.8
D Pr
0.33
(9.56)
1 1
Rw = = = 0.0003 K/W (9.58)
hw A 6,577π(0.02)(0.2)40
π
ṁw = ρVAN = 963(1) 0.022 (40) = 12.1 kg/s (9.59)
4
As a result, the overall heat transfer coefficient, including fouling, as well as the air and
water side convection resistance can be calculated as follows:
1 1
Utot = = = 390.4 W/m2 K (9.60)
ARtot 1.6(0.0013 + 0.0003 + 10−6 )
which yields a ratio of Cmin/Cmax = 0.02. Then the number of thermal units becomes:
Utot A 390.4(1.6)
NTU = = = 0.62 (9.64)
Cmin 1,007
Based on these Cmin/Cmax and NTU values, the effectiveness charts yield ɛ ≈ 0.5,
Ta,o − Ta,i
ε= = 0.5 (9.65)
Tw,i − Ta,i
Thus the air outlet temperature becomes Ta,o = 52.5◦ C. The resulting heat gained by the
air can be equated with the heat removed from the water within the tubes to also obtain
the water outlet temperature.
Heat Exchangers 405
This example involved a finned tube cross-flow configuration for heat exchange between
liquid and gas flows. Plate–fin heat exchangers are also commonly used particularly for heat
exchange between gas–gas streams. For example, air–air heat exchangers typically use
plate–fin arrangements. Various arrangements include plain fins, strip fins, pin fins, and
perforated fins. Detailed design data for finned heat exchanger surfaces was presented by
Kays and London (1984).
∂Ts ∂Ts
ms cps + ṁs cps L = hs (x)As (Tw − Ts ) (9.67)
∂t ∂x
x dx
Constant
temperature
(phase change)
fluid
Tc CV3
Tw CV1 Wall
q = h(x) (Tw–Tc)
Ts Single-phase
fluid
Ts,in CV2 Ts, out
FIGURE 9.7
Schematic of a heat exchanger control volume.
406 Advanced Heat Transfer
∂χ g
ṁc Lhfg = hc (x)Ac (Tw − Tc ) (9.68)
∂x
where L, hfg, and χg refer to the heat exchanger length, latent heat of vaporization, and vapor
phase fraction, respectively. The heat transfer coefficient, hc, depends on the two-phase flow
regime (Chapter 5). The two-phase fluid at Tc undergoes a change of phase at a constant tem-
perature. Under a quasi-stationary approximation, the net enthalpy difference across the
streamwise edges of the control volume equals the convective heat transfer across the wall.
Define the following dimensionless variables:
∗ x ∗ ṁs T − Ts,i
x = ; t = t; θ= (9.69)
L ms Tc,1 − Ts,i
N2 N3 cps (Tw − Ts )
NTU = ; Ja = (9.71)
N 2 + N3 hfg
where Ja refers to the Jacob number and the subscripts i and ∞ refer to the inlet and final
conditions, respectively.
Using the above dimensionless variables, the energy equations in both fluids and the wall
become:
∂θw
N1 + N2 (x)(θw − 1) + N3 (x)(θw − θs ) = 0 (9.72)
∂t∗
∂θs ∂θs
+ = N3 (x)(θw − θs ) (9.73)
∂t∗ ∂x∗
∂χ g
N4 = N2 (x)(θw − 1) (9.74)
∂x∗
Integrating over the length of the heat exchanger (from x* = 0 to 1), these equations
become:
1
1
1 1
∂
N1 ∗ θw dx∗ + N2 (θw − 1) dx∗ + N3 θw dx∗ − θs dx∗ =0 (9.75)
∂t 0 0 0 0
1
1 1
∂ ∗ ∗ ∗
θs dx + θs (1) − θs (0) = N3 θw dx − θs dx (9.76)
∂t∗ 0 0 0
1
N4 (χ g (1) − χ g (0)) = N2 (θw − 1) dx∗ (9.77)
0
The initial and steady-state temperatures can be determined by solving the previous
governing equations without the transient terms, since the initial and steady-state
Heat Exchangers 407
conditions do not change during a given time interval (the period between step-changes of
the fluid temperature). Initially, a step-change of fluid temperature Tco Tc1 is applied.
Using the steady-state and initial temperature profiles, it can be shown that the fluid and
wall temperatures become:
θs = 1 − e−NTUx∗ f̃ (t∗ ) (9.78)
NTU −NTUx∗
θw = 1 − e g̃(t∗ ) (9.79)
N2
where,
Substituting these profiles into the integral equations and integrating from x* = 0→1
leads to:
df̃ (t∗ )
= C1 f̃ (t∗ ) − C1 g̃(t∗ ) (9.82)
dt∗
dg̃(t∗ )
= D1 f̃ (t∗ ) + D2 g̃(t∗ ) + D3 (9.83)
dt∗
where,
N2 + N3 N2
D2 = − ; D3 = (9.86)
N1 N1 + (N1 /N2 )e−NTU − N1 /N2
∗ ∗ C1 D3
g̃(t∗ ) = aeβ1 t + beβ2 t + (9.88)
C 2 D 1 − C 1 D2
where,
D1 + D2 − β2 0 D3 (C1 β2 − C1 D2 − C2 D1 )
a= θc + (9.89)
β1 − β 2 (β1 − β2 )(C1 D2 − C2 D1 )
408 Advanced Heat Transfer
D1 + D2 − β1 0 D3 (C1 β1 − C1 D2 − C2 D1 )
b=− θc − (9.90)
β1 − β2 (β1 − β2 )(C1 D2 − C2 D1 )
1
β1,2 = C1 + D2 + (C1 + D2 )2 + 4(C2 D1 − C1 D2 ) (9.91)
2
The heat transfer coefficient of the constant temperature fluid, hc, as defined in N2(x),
depends on the local phase fraction. It varies with position as a result of the varying two-
phase flow regime and vapor fraction in the x-direction of the constant temperature fluid.
The solution first requires a boundary condition to be applied at the inlet of the heat
exchanger. A suitable correlation of single-phase forced convection is then used up to the
point of onset of phase change. This transition point is identified when the phase fraction
first becomes nonzero, based on the temperature (with respect to the phase change temper-
ature) or enthalpy (with respect to an enthalpy equation of state). At the first point of phase
change, the phase fraction is computed, after which the type of two-phase flow regime is
identified for the appropriate heat transfer correlation. Near the saturation points, a suitable
convection correlation is used with property values evaluated along the saturated liquid
and vapor lines. Since an integral method has been used, only the inlet and outlet values
of the vapor phase fraction are required. Once the mixture becomes saturated vapor, further
heat input is transferred to the vapor in a superheated state.
Sample results are presented in Figure 9.8 for the transient response to step-wise temper-
ature changes in an aircraft two-phase heat exchanger with working fluids of refrigerant
(R134a) and air. At the initial time, the fluid undergoes a step-change of inlet temperature
and the fluid eventually stabilizes to a new asymptotic temperature after a sufficient period
of time has elapsed. Sample results for two cases are presented in Figure 9.8—(1) N2 = 1,
N3 = 1 (NTU = 0.5), x* = 1 (outlet) and θoc = 0; and (2) N1 = 300, N2 = N3 = 1 (NTU = 0.5)
and θoc = 0.4. In the simulations, a steady-state condition is reached initially throughout
the entire system after which the constant temperature fluid undergoes an instantaneous
step-change in temperature.
From the results, both wall and fluid temperatures stabilize faster at lower values of N1
due to a lower thermal inertia of the wall. A larger mass of fluid promotes heat exchange
θs (t* = 240)
0.2
FIGURE 9.8
Changes of fluid and wall temperatures with (a) time at varying N1 values; and (b) position at varying times.
(Adapted from G.F. Naterer and C.H. Lam. 2006. ASME Journal of Heat Transfer, 128: 953–962.)
Heat Exchangers 409
more rapidly as a result of a larger volume of heat exchanger or larger mass flow rate
through the heat exchanger. Lower stabilization times are observed upstream since a certain
time is required for thermal disturbances to propagate downstream.
In the spatial variations of temperatures, both fluid and wall temperatures rise nearly lin-
early from the inlet to the outlet. Higher nonlinearity is observed for the cases with larger
changes of the convection coefficient throughout the heat exchanger. Increasing linearity
of the temperature profile is observed for cases with smaller Jacob numbers and larger val-
ues of the latent heat of vaporization. These results indicate that faster cooling of the single-
phase fluid stream occurs in those cases since more heat is absorbed from the phase-change
fluid for a fixed difference of phase fraction between the inlet and outlet.
where h and ṁc denote the fluid enthalpy and mass flow rate, respectively. Also, Ai refers to the
inner area of the tube based on the inner diameter, di. The upcoming energy balances will pri-
marily involve enthalpy, rather than temperature, in order to include both sensible heat (relat-
ing to a temperature change) and latent heat (associated with phase change) portions of the
heat exchange. Heat balances in the previous sections, such as Equations 9.2 and 9.7, must
be modified to include both latent and sensible heat for condensation and boiling processes.
The overall heat transfer coefficient, U, can be expressed in terms of the heat transfer coef-
ficients for the inner and outer tube surfaces, Ui and Uo, respectively, as follows:
1 1 1
= = (9.93)
UA Ui Ai Uo Ao
410 Advanced Heat Transfer
The heat transfer coefficient on the inner side can be determined by:
−1
1 1
Ui (πDi ) = + (9.94)
hc (πDi ) hh (πDo )
where hc and hh refer to convection coefficients for the hot and cold streams, respectively,
along the surface, dx, separating both fluid streams.
Since the fluid may pass through several phase change regimes, the tube length is
subdivided into discrete elements. Consider a uniform subdivision of the tube length into
N elements of width Δx. Then the previous energy balances are applied individually over
each element. For the jth element, Equation 9.92 may be discretized as follows:
Ui,j πDi Δx
Hj+1 = (Th,j − Tc,j ) + Hj (9.96)
ṁc
Therefore, the enthalpy of each element can be determined based on previous (upstream)
values of enthalpy and temperatures of the cold and hot fluid streams.
Once the enthalpy in an element is calculated, its value may exceed the saturated vapor
enthalpy at the flow pressure (a superheated vapor state). The enthalpy can then be used
to determine the temperature of the superheated vapor based on thermodynamic tables
(see Appendix E). Alternatively, since there is no longer phase change in the superheated
vapor state, the change of enthalpy between elements can be used to find the corresponding
temperature rise as follows:
This yields Tj+1 in terms of Tj (computed in the previous upstream element) and the com-
puted enthalpy change from Equation 9.96. This linearization requires a sufficiently small Δx
to assume a locally constant value of cp within the element. When the fluid exists entirely as a
superheated vapor, the vapor specific heat at the mean temperature (between Tj and Tj+1)
can be used.
In the two-phase regime, if the enthalpy in Equation 9.96 is less than the saturated vapor
enthalpy, then the mass fraction of vapor (or quality), χg, in element j + 1 becomes:
hj+1 − hf
χ g,j+1 = (9.98)
hg − hf
where hg and hf refer to the enthalpy of saturated vapor and liquid, respectively. Using this
phase fraction, an updated estimate of the convection coefficient can be calculated based on
the flow regime and two-phase flow map associated with this vapor fraction (Chapter 5).
Heat Exchangers 411
Then an updated overall heat transfer coefficient is computed based on Equation 9.94,
−1
1 Di
Ui,j+1 = + (9.99)
hc,j+1 hh,j+1 Do
During the change of phase, the wall temperature between the two fluid streams is
assumed to remain constant within the discrete element.
The overall solution method requires a marching procedure whereby the problem vari-
ables are computed in the first element, j = 1, after which variables in successive elements
are calculated based on the previous element’s values. For example, after Ui,1 is calculated
from Equation 9.99, the enthalpy, temperature, and quality for the following element, j = 2,
can be calculated from Equations 9.96 through 9.98. Then the values are computed in the
following element, and so on, over the entire length of the tube, for j = 3, 4, … N. This sol-
ution procedure can be summarized as follows:
A similar procedure can be adopted for both condensation and boiling problems.
A numerical solution for power plant condensers based on this procedure was presented
by Zhang, Sousa, and Venart (1993). A two-phase flow map can also be used to identify
the flow regime based on the computed phase fraction. This two-phase flow map would dis-
tinguish between the various flow regimes, for example, wavy, annular, and slug flow
regimes.
For example, when saturated liquid enters a heated horizontal pipe, boiling occurs ini-
tially with bubbles forming and growing along the pipe surface. As the vapor fraction
increases, transition occurs to plug flow, followed by slug flow. Transition to annular
flow occurs at a vapor fraction of approximately χg = 0.04 and wavy flow at χg = 0.94. After
the mixture becomes saturated vapor, further heat transfer leads to a temperature increase of
the superheated vapor. Using Equation 9.92 over the entire tube, the total heat exchange
between the fluids can be calculated by the mass flow rate multiplied by the total enthalpy
difference of the fluid between the inlet and outlet.
Regular maintenance and safety are important practical operating requirements for con-
densers and boilers. Tubes should be periodically cleaned through removable components
or higher liquid velocities to reduce fouling on heat exchanger surfaces. In systems operating
at pressures different than atmospheric pressure, leakage can occur. Pressurizing a system
412 Advanced Heat Transfer
further and observing changes in pressure over time are useful indicators of the “tightness”
of the system, but not necessarily the location of leakage. Chemical leaks can be detected
individually. For example, sulfur dioxide can be detected by white smoke forming when
ammonia is brought into contact with the leakage point.
This section has discussed liquid–gas phase change in condensers and boilers. Similar
models and numerical methods have been developed for solid–liquid phase change such
as freezing of water in heat exchanger tubes. Nabity (2014) presented a model of a freezable
water-based heat exchangers with three modes of operation: (i) fully thawed mode which
rejects the full heat flow; (ii) partially frozen state rejecting an intermediate heat flux; and
(iii) completely frozen state (except for an insulated region) rejecting the minimum heat
flux. Further detailed aspects of heat exchanger design, operation, and safety are presented
in books by Fraas (1989), Hewitt (1998), Kakac, Liu, and Pramuanjaroenkij (2012), and
Penoncello (2015).
PROBLEMS
9.1 In a cross-flow heat exchanger, water flows over a copper pipe (1.9 cm inner diame-
ter; 2.4 cm outer diameter) with a convection coefficient of 210 W/m2K. Oil flows
through the pipe. Find the convection coefficient of an oil flow that provides an over-
all heat transfer coefficient of 140 W/m2K per unit length of pipe (based on the inner
tube area).
9.2 In a counterflow heat exchanger, water enters at 20◦ C at 3 kg/s and cools oil
(cp = 2.2 kJ/kgK) flowing at 2 kg/s with an inlet temperature of 160◦ C. The heat
exchanger area is 12 m2. What overall heat transfer coefficient is required to generate
a water outlet temperature of 50◦ C?
9.3 Water flows through a copper pipe (3.2 cm inner diameter; 3.5 cm outer diameter) in a
cross-flow heat exchanger. Air flows across the pipe. The convective heat transfer
coefficients for the air and water sides of the pipe are 120 and 2,400 W/m2K, respec-
tively. Up to what additional fouling resistance can be tolerated if its effect must not
reduce the overall heat transfer coefficient (based on the outside area without fouling)
by more than 5%?
9.4 A tubular heat exchanger operates in a counterflow arrangement. A design modifica-
tion requires a higher mass flow rate of fluid in the inner tube. What change of tube
length is required to maintain the same inlet and outlet temperatures of both fluids?
Express your answer in terms of the tube diameter, thermophysical properties, fluid
velocity, and convection coefficient of the external flow in the outer annulus. Assume
that the external convection coefficient remains approximately identical under both
flow conditions.
9.5 A concentric tube heat exchanger is used to condense steam in the annulus between
the inner tube (copper; 1.6 cm inner diameter, 0.2 cm wall thickness) and surface of
the outer tube (4 cm outer diameter). The water flows at 6 m/s with an average tem-
perature of 25◦ C through the inner pipe. The thermal resistances due to steam con-
vection and fouling on the outer surface of the inner tube are 2 × 10−5 and
10−6 K/W, respectively. The total length of the inner pipe is 80 m. Find the total
rate of heat transfer to the water and average inner wall temperature of the
inner pipe.
Heat Exchangers 413
9.6 A set of finned tubes is arranged uniformly in rows and placed normal to an
incoming airflow at 15◦ C with a flow rate of 300 m3/h into a cross-flow heat
exchanger. Hot water at 85◦ C enters the tubes (inner and outer diameters of
10 and 12 mm, respectively) with a mean velocity of 0.22 m/s. The heat
exchanger volume is 0.006 m3 and the Colburn and friction factors are j = 0.014
and f = 0.049, respectively. The estimated thermal resistance due to fouling is
10−5◦ C/W. What number of finned tubes is required to produce an air outlet
temperature of 64◦ C? The following additional parameters are specified in the
design of the heat exchanger: Afr = 0.1 m2 (frontal area); Aff/Afr = 0.3; Dh = 5 mm;
copper fin diameter and thickness of 18 and 0.2 mm, respectively; Af/A = 0.9; A/
V = 350 m2/m3 (ratio of the total surface area to the volume of the heat
exchanger) and L = 40 cm (flow length).
9.7 A cross-flow heat exchanger in a power plant is used for intercooling between two
compressor stages. Air enters the heat exchanger at 400◦ C with a flow rate of
20 kg/s and flows past finned tubes with a fin efficiency of 90%. The Colburn and
friction factors are 0.008 and 0.03, respectively. Cooling water enters the heat
exchanger at 10◦ C with a flow rate of 40 kg/s. Find the outlet temperature and pres-
sure drop of the air. The following additional parameters are specified in the heat
exchanger design: Ac,w = 0.4 m2; Dh,a = 2 cm = Dh,w; Ac,a = 2 m2; Af/A = 0.8; Aw/Aa =
0.2; A/V = 350 m2/m3; and Aff/Afr = 0.8. The dimensions of the heat exchanger are
2 m (length) × 0.75 m (height) × 0.8 m (width) and the thermal resistance due to
tube fouling is 10−5◦ C/W.
9.8 Derive the expression in Equation 9.45 for the effectiveness of a concentric tube (par-
allel flow) heat exchanger.
9.9 In a cross-flow heat exchanger, air is heated from 10◦ C to 30◦ C (unmixed stream) by
another airstream entering at 90◦ C (mixed stream). The mass flow rates of the cold
and hot airstreams are 1 and 2 kg/s, respectively. What percentage increase of the
overall heat transfer coefficient is required if the cold stream must be heated to
35◦ C at the same flow rates of air? Assume that the heat transfer area is maintained
equally in both cases, although its configuration may be altered to achieve higher heat
transfer coefficients.
9.10 In a shell-and-tube heat exchanger, the inner and outer diameters of each copper tube
are 1.5 and 1.9 cm, respectively. The convection coefficients for fluid flow inside and
over the tubes are 4,800 and 4,600 W/m2K, respectively. What change in the number
of tubes will increase the overall heat transfer coefficient (based on the outside area,
per unit length) by a factor of 30%? Assume that this change does not affect the shell-
side fluid velocity and the flow rate in the tubes remains constant for both cases.
9.11 Oil is cooled by water flowing through tubes in a shell-and-tube heat exchanger with
a single shell and two tube passes. The pipe’s outer diameter, wall thickness, thermal
conductivity, and length are 1.9 cm, 2 mm, 26 W/mK, and 4 m, respectively. Water
flows at 0.3 kg/s on the shell side with a convection coefficient of 4,950 W/m2K
and an inlet temperature of 280 K. Oil flows at 0.2 kg/s in the tubes with a convection
coefficient of 4,800 W/m2K (note: cp = 2.2 kJ/kgK for oil). If the oil inlet temperature is
360 K, determine its outlet temperature.
9.12 Water enters a shell-and-tube heat exchanger with a single-shell and four-tube pass at
30◦ C. It is heated by an oil flow (cp = 2.3 kJ/kgK) that enters the tube at 240◦ C and
leaves at 140◦ C. The mass flow rate on the tube side and heat exchanger area are
414 Advanced Heat Transfer
2 kg/s and 18 m2, respectively. What overall heat transfer coefficient is required to
yield a water outlet temperature of 70◦ C on the shell side?
9.13 Air is heated by water flowing through tubes at 2 kg/s in a cross-flow heat exchanger
(water unmixed, air mixed). The overall heat transfer coefficient is 500 W/m2K and
the length and diameter of each tube are 3 m and 1.9 cm, respectively. The water inlet
temperature is 80◦ C. Air enters the heat exchanger at 3 kg/s and 10◦ C. How many
tubes are required to produce an air outlet temperature of 60◦ C?
9.14 A shell-and-tube heat exchanger contains one shell pass and two tube passes. The
overall heat transfer coefficient is 480 W/m2K and the total surface area of the
tubes is 16 m2. Water enters the shell side at 40◦ C and leaves at 80◦ C. Find the
required mass flow rate of water to cool the fluid in the tubes from 260◦ C at
the tube inlet to 120◦ C at the tube outlet. The specific heat of the fluid in the
tube flow is 2.4 kJ/kgK.
9.15 Water is heated as it flows through a single-pass shell-and-tube heat exchanger
with N tubes internally. The outside surface of each tube (diameter D) is heated by
steam condensing at a temperature of Tsat. Find the number of tubes required to con-
dense a prescribed mass flow rate of steam, ṁs . Express your answer in terms of the
mass flow rate of water, ṁw , water inlet and outlet temperatures and thermophysical
properties of water.
9.16 Discuss the main design features of importance when selecting a shell-and-tube heat
exchanger for condensing refrigerant fluid.
9.17 What methods can be used for heat transfer enhancement in heat exchangers?
Explain the methods and their effects on the heat exchanger performance.
9.18 The compactness of a heat exchanger is calculated from the heat transfer area density,
or in other words, the total surface area of heat transfer divided by the heat exchanger
volume. Perform a review of the technical literature to assess the degree of compact-
ness of various types of heat exchangers such as automotive radiators and plate
heat exchangers.
9.19 A square-shaped array of 1-cm-diameter tubes is located within a condenser. Water
at 20◦ C enters at a velocity of 1 m/s into the tubes. The heat exchanger is required to
condense 4 kg/s of refrigerant 12 (hfg = 130 kJ/kg) at 45◦ C outside the tubes.
a. How many tubes are required to keep the temperature rise of the water below
3◦ C?
b. What total surface area is required to produce the required condensate mass
flow rate of refrigerant fluid?
9.20 A condenser consists of a heat exchanger with a single shell pass and internal copper
tubes (diameter of 1.9 cm, length of 1 m). Steam condenses at atmospheric pressure on
the outer surface of the tubes which are arranged in a square array. The outside wall
temperature of the tube is 92◦ C.
a. How many tubes (per row) are required to produce a rate of steam conden-
sation of 60 g/s?
9.21 What temperature rise of the cooling water occurs in this tube configuration? The
flow rate is 0.1 kg/s per tube.
Heat Exchangers 415
References
A. Bejan. 2013. Heat Transfer, 4th Edition, New York: John Wiley & Sons.
T.L. Bergman, A.S. Lavine, F.P. Incropera and D.P. DeWitt. 2011. Fundamentals of Heat and Mass
Transfer, 7th Edition, New York: John Wiley & Sons.
R.A. Bowman, A.C. Mueller and W.M. Nagle. 1940. “Mean Temperature Difference in Design,”
Transactions of ASME, 62: 283–294.
J. Buongiorno. 2006. “Convective Transport in Nanofluids,” ASME Journal of Heat Transfer, 128(3):
240–250.
A.P. Fraas. 1989. Heat Exchanger Design, 2nd Edition, New York: John Wiley & Sons.
G.F. Hewitt, Ed., 1998. Heat Exchanger Design Handbook, New York: Begell House.
S. Kakac, H. Liu and A. Pramuanjaroenkij. 2012. Heat Exchangers, 3rd Edition, Boca Raton: CRC Press/
Taylor & Francis.
W.M. Kays and A.L. London. 1984. Compact Heat Exchangers, 3rd Edition, New York: McGraw-Hill.
J.A. Nabity. 2014. “Modeling a Freezable Water-Based Heat Exchanger for Use in Spacecraft Thermal
Control,” AIAA Journal of Thermophysics and Heat Transfer, 28: 708–716.
G.F. Naterer and C.H. Lam. 2006. “Transient Response of Two-Phase Heat Exchanger with Varying
Convection Coefficients,” ASME Journal of Heat Transfer, 128: 953–962.
S.G. Penoncello. 2015. Thermal Energy Systems: Design and Analysis, Boca Raton: CRC Press/Taylor &
Francis.
J. Harrison, Ed., Standards of the Tubular Exchange Manufacturers Association, 8th Edition. 1999. Tarry-
town, NY: Tubular Exchange Manufacturers Association.
Y. Vermahmoudi, S.M. Peyghambarzadeh, M. Naraki and S.H. Hashemabadi. 2013. “Statistical Anal-
ysis of Nanofluid Heat Transfer in a Heat Exchange System,” AIAA Journal of Thermophysics and
Heat Transfer, 27: 320–325.
C. Zhang, A.C.M. Sousa and J.E.S. Venart. 1993. “Numerical and Experimental Study of a Power Plant
Condenser,” ASME Journal of Heat Transfer 115: 435–445.
10
Computational Heat Transfer
Numerical solution methods are commonly used to analyze thermal engineering systems
due to the rapidly increasing speed and capabilities of computers. A brief introduction to
numerical heat transfer will be presented in this chapter. Topics to be covered include the
fundamentals of numerical methods based on finite differences, weighted residuals, finite
elements, and finite volumes. The finite element method will be examined in detail
including two-dimensional formulations, triangular elements, quadrilateral elements,
and time-dependent problems. The SIMPLE and SIMPLEC methods are presented for
the pressure-velocity coupling in finite volume methods. The reader is also referred to
other more detailed books on computational heat transfer such as Ozisik et al. (2017)
and Reddy and Gartling (2010).
The finite difference equations can be obtained by a Taylor series method. Performing a Taylor
series expansion of temperatures about the nodal point (i, j), where i and j refer to the column
417
418 Advanced Heat Transfer
Qn n Control
Δy volume
w e
Qs s
FIGURE 10.1
Schematic of a finite difference grid layout.
∂T ∂2 T Δx2 ∂3 T Δx3
Ti−1,j = Ti,j − Δx + − + ··· (10.3)
∂x i ∂x2 i 2 ∂x3 i 6
where Δx refers to the grid spacing in the x-direction. A uniform grid spacing will
be assumed.
Adding Equations 10.2 and 10.3 and rearranging,
where O(Δx 2) refers to the higher order terms with an order of magnitude proportional to
Δx 2. Writing a similar expression for the y-direction derivative, combining with Equation
10.4, and assuming steady-state conditions
where higher order terms have been neglected. This result represents the finite difference
approximation of the steady-state heat equation. The approximation is second-order accurate
since the truncation errors correspond to terms truncated above the second-order from the
original Taylor series expansions.
For a uniform grid with Δx = Δy, Equation 10.5 becomes:
1
Ti,j = (Ti+1,j + Ti−1,j + Ti,j+1 + Ti,j−1 ) (10.6)
4
Computational Heat Transfer 419
As expected, steady-state heat conduction yields temperatures that are arithmetic averages
of their surrounding values.
Alternatively, another approach for deriving the finite difference equations is based on an
energy balance method. The grid is subdivided into finite volumes, where a nodal point is
located at the center of each volume (see Figure 10.1). For the control volume at point p,
or (i, j) in Figure 10.1, the steady-state energy balance is expressed as:
Q n + Qw + Qs + Qe = 0 (10.7)
where Qn, Qw, Qs, and Qe denote the heat flows across the north, west, south, and east edges,
respectively, of the control volume about point p. These edges are located halfway between
the adjacent nodal points. For example, point n is located halfway between (i, j) and (i, j + 1)
along column i.
Using a uniform grid spacing of Δx and Δy in the x- and y-directions, respectively, together
with Fourier’s law and a linear interpolation of temperature between the nodal points, Equa-
tion 10.7 becomes:
Ti,j − Ti,j+1 Ti,j − Ti−1,j Ti,j − Ti,j−1 Ti,j − Ti+1,j
−kΔx − kΔy − kΔx −kΔy =0 (10.8)
Δy Δx Δy Δx
per unit depth. For a uniform grid (Δx = Δy), it can be readily verified that this result
becomes the same as Equation 10.6. Both the Taylor series method and energy balance
method yield the same results (as expected).
Once the finite difference equations are obtained for each node, the full resulting set of
algebraic equations must be solved to yield the temperature values at each node in the
domain. The following example demonstrates the solution procedure of the finite difference
method.
At boundary node 4, the convective cooling condition is applied through Newton’s law
of cooling at the edge of the control volume,
For the energy balance at node 6, the boundary contribution to the heat flow is Qb = 0
since the right boundary is insulated, thereby yielding after simplification,
These three above algebraic equations for T4, T5, and T6 can be solved by standard solv-
ers, such as Gaussian elimination, iterative, or matrix inversion methods (Cheney and Kin-
caid 1985). The solution yields: T4 = 71.1◦ C, T5 = 148.9◦ C, and T6 = 164.4◦ C.
An overall heat balance may be used to validate these results. The total heat flow across
the top boundary of the domain, joined by nodes 1 (top left corner), 2, and 3 (top right cor-
ner), is given by Q123 = 813.3 W/m. Similarly, the heat flows across the bottom and left
boundaries are calculated as Q789 = 213.3 W/m and Q147 = –1,026.6 W/m, respectively.
Based on these results,
This result confirms that the finite difference results correctly yield individual and over-
all energy conservation under steady-state conditions.
If the nodes are located along a curved boundary, additional boundary node modeling is
required to establish the appropriate finite difference equations. The discretized equations
may become dependent on the boundary configuration, which compromises the generality
of the algorithm for various grid configurations. In an upcoming section, finite element
methods will be described for better geometric flexibility in spatial discretization of the
problem domain.
∂T ∂2 T
ρcp =k 2 (10.19)
∂t ∂x
Computational Heat Transfer 421
In this section, two types of transient formulations will be described, called explicit and
implicit methods. Explicit methods calculate the state of the system at the current time based
on values from the previous state, while implicit methods find a solution at the current time
using variables at both the current state and the previous time level.
Using a one-dimensional energy balance and backward difference in time to approximate
the time-dependent and spatial derivatives in Equation 10.19,
n+1 n n
Ti − Tin Ti+1 − Tin Ti − Ti−1
n
(AΔx)ρcp = kA − kA (10.20)
Δt Δx Δx
where the superscripts n + 1 and n refer to the current and previous time levels, respectively,
and Δt is the time step size. This formulation represents an explicit method because the flux
terms on the right side, Qe and Qw, are evaluated at the previous time level. An explicit
method requires less computational time than an implicit method because the matrix of
coefficients for the algebraic equations becomes diagonal. However, a more stringent
time step restriction is usually required to maintain numerical stability.
Rearranging Equation 10.20,
Tin+1 = (1 − 2Fo)Tin + Fo Ti−1 n + Ti−1 n (10.21)
where Fo = αΔt/Δx 2 is the Fourier modulus (nondimensional time) and α = k/ρcp is the ther-
mal diffusivity. Based on the first term on the right side of Equation 10.20, a time step stabil-
ity criterion can be derived for the interior nodes: Fo ≤ 1/2. This constraint is imposed to
maintain a positive leading coefficient of Tin . Otherwise, an increase of Tin may lead to a
decrease of Tin+1 and potentially numerical oscillations and resulting lack of solution
convergence.
Unconditional stability can be achieved by evaluating the diffusion terms in Equation
10.20 at the current time step (n + 1), rather than the previous time step (n). For an
implicit method which evaluates the diffusion flux terms at the current time level (n + 1)
in Equation 10.19,
n+1
Tin+1 − Tin Ti−1 n+1 − Tin+1 T − Ti−1 n+1
(AΔx)ρcp = kA − kA i (10.22)
Δt Δx Δx
This formulation is implicit because the transient and diffusion flux terms involve temper-
atures that are currently sought, rather than known from a previous time level. Rearranging
this equation in terms of the nondimensional Fourier number,
(1 + 2Fo)Tin+1 − Fo Ti−1 n+1 + Ti−1 n+1 = Tin (10.23)
In this case, there is not a stability limit involving the time step size, Δt. However, a set of
simultaneous algebraic equations needs to be solved at each time step with coefficients in
nondiagonal entries of the solution matrix. Thus, there is a trade-off between more time
steps (an explicit formulation) or more computational effort required at each time step
(implicit method). In most problems, implicit solutions generally have more attractive ben-
efits, robustness, and accuracy than explicit methods.
422 Advanced Heat Transfer
This function contains unknown coefficients (a1, a2, …, an) which are determined from the
numerical solution.
In general, the residual will not equal zero because it is unlikely that the initial appro-
ximate solution will precisely match the exact solution. As a result, the coefficients (a1,
a2, …, an) are determined by minimizing the residual over the domain or setting the inte-
grated and weighted value of the residual to zero. The weighted residual method poses n
constraints on the coefficients a1, …, an through the following formulation:
Wi L(ϕ̂) dV = Wi R dV = 0 (10.26)
V V
where V and Wi refer to the problem domain (volume in three dimensions) and weight
functions, respectively.
In the weighted residual method, n weight functions, W1, … , Wn, are selected to establish
n equations for the unknown coefficients a1, … , an. In practice, the unknown coefficients rep-
resent problem variables such as temperature (energy equation) or velocity (momentum
equations). The solution procedure yields n integral equations for these n coefficients, lead-
ing to an n × n matrix for the resulting set of linear algebraic equations.
The weighted residual formulation in Equation 10.26 distributes the solution error
throughout the domain. If the weight function becomes large, then the residual approaches
zero and the governing equation is satisfied exactly at the given point where the weight
Computational Heat Transfer 423
function is applied. Alternatively, if the weight function is nonzero, then the residual is also
nonzero and errors are distributed away from the grid points. If the weight functions are
selected to be equal to the basis functions (or shape functions) of the approximate solution,
then this method is called a Galerkin weighted residual method. A Galerkin-based finite ele-
ment method will be presented in the next section.
1. Discretize the problem domain by specifying the number of elements, shape of ele-
ments, and their spatial distribution.
2. Select a type of interpolation for the dependent variable(s), such as linear or
quadratic interpolation.
3. Determine the element property equations (or stiffness equations). For example, find the
temperature equations for a heat transfer analysis.
4. Assemble the elements by following a specified group of assembly rules.
5. Apply the boundary conditions.
6. Solve the discrete equation set and post-process the results.
The following example outlines how these steps are applied in a problem involving one-
dimensional heat conduction in a solid with internal heat generation.
1 2 3 4
·
P
e=1 e=2 e=3
N1 N2 N3 N4
1
T = 0°C
T = 0°C
x=L
x
x1 x2 x3 x4
FIGURE 10.2
Schematic of heat conduction in a solid with internal heat generation.
d2 T Ṗ
L(T) = + =0 (10.27)
dx2 k
The solution procedure involves the following six steps of the finite element method:
1. The one-dimensional domain is discretized into four nodes with three linear elements
located uniformly within the domain of length L. Further grid refinements may be
required to obtain a final converged solution.
2. Interpolation within each element is performed based on a locally linear approximate
solution. Using dimensionless coordinates, defined by x* = x/L, within each element,
x − xe1
x∗ = (10.28)
xe2 − xe1
where the subscript refers to the local node number. There are two local nodes for a lin-
ear one-dimensional element. The superscript e indicates a local evaluation within the
element. Also, based on a locally linear temperature profile within each element,
e
T̃ = ae1 + ae2 x∗ (10.29)
where the overhat notation (denoting an approximate solution) will be dropped in sub-
sequent equations for brevity.
If the temperatures at the local nodes are T1e (at x* = 0) and T2e (at x* = 1), then the
coefficients in Equation 10.29 can be readily evaluated to give:
T e = T1e + T2e − T1e x∗ (10.30)
Alternatively,
These results indicate that there are two nodal degrees of freedom in the interpola-
tion with linear elements.
Computational Heat Transfer 425
3. The third step involves the derivation of the element property equations. Recall that
Galerkin’s method selects the weight functions equal to the shape functions of the
approximate solution. Therefore, using Galerkin’s method within an element,
Wie Re dV e = Nie Re dV e (10.34)
Ve Ve
Based on Equation 10.34 for both weight functions within the element, the local finite
element equations (indicated by the superscript e) can be written in the following
matrix form:
[c]{Te } = {re } (10.36)
In the weighted residual method, the sum of residual terms becomes zero after the
assembly of all elements. As a result, the first weighted residual term on the right
side of the equation can be dropped from further consideration.
Two problems can be observed in the formulation in Equation 10.37: (i) a singular
matrix due to zero second-order derivatives arising from linear shape functions; and
(ii) no direct mechanism to readily invoke boundary conditions other than temper-
ature-specified conditions. A possible approach to address these difficulties is to
maintain more interelement continuity. The value of temperature and all of its
derivatives, up to and including one less than the highest order derivative in the
element property equations, should be continuous across element boundaries. For
example, piecewise cubic interpolation functions could be used.
Alternatively, integration by parts can be used to reduce the order of the highest
derivative in the element equations. This would prevent a singular matrix and allow
other types of boundary conditions, such as heat flux-specified conditions. Using
Galerkin’s formulation with Equation. 10.34 and 10.35, together with integration by
parts using local coordinates,
2 1
d T Ṗ 1 e dT 1 1 1 dNie dT ∗ Ṗ
Nie + dV e
= N − (Δxdx ) + Nie (Δxdx∗ ) = WRei
Ve dx2 k Δx i dx∗ 0 Δx2 0 dx∗ dx∗ 0 k
(10.39)
This result has three main terms between the equality signs. In the first term, the
shape function, Nie , may be evaluated at both local nodes, yielding,
1 e dT 1 e dT Qe2 Qe1
Ni ∗ − Ni ∗ = + (10.40)
Δx dx 1 Δx dx 0 kA kA
426 Advanced Heat Transfer
where A refers to the cross-sectional area of the one-dimensional element in the direc-
tion of heat flow.
A nodal value of Q (heat flow rate) is defined as positive into an element. Using local
node 1 as an example, it can be observed that a negative temperature gradient yields a
positive Qe1 , or heat flow into the element. Conversely, a positive temperature gradient
at node 1 yields a negative heat flow (out of the element). In both cases, the sign con-
vention is retained since both Q values are positive into the element.
For the second term in Equation 10.39,
1 1
1 dNie dT 1 dNie dN1e e dN2e e
∗ dx∗
dx∗ = ∗
T + T dx∗ (10.41)
Δx 0 dx Δx 0 dx dx∗ 1
dx∗ 2
Substituting i = 1, the right side becomes T1e − T2e /Δx. Also, i = 2 yields the same
result, except with a leading negative sign. Furthermore, for the third term in Equation
10.39,
1
Ṗ ∗ Δx Ṗ
Δx Nie dx = (10.42)
0 k 2 k
where i = 1, 2.
The finite element equations can now be obtained by adding all individual terms
into Equation 10.39. The stiffness matrix in Equation 10.37 is multiplied by kA
(conductivity × area). Then this matrix becomes kA/Δx along the diagonal and –kA/
Δx along the off-diagonal entries. Also, the right side in Equation 10.38 becomes,
1
r̂ = (kA)rm = Qem + ΔxAṖ − kAWRei (10.43)
2
4. In the fourth step, all elements are assembled within the mesh. The assembled WRei
terms, summed over all elements, become zero. The sum of Qei over all elements
becomes Qi (heat flow at node i). In the assembly process, the temperature at a specific
node is unique so a nodal map between local and global nodes is required. For example,
the temperature at global node 2, T2, is equivalent to T21 , as well as T12 (Figure 10.2),
where the subscript refers to the local node and the superscript refers to the element
number. The nodal mapping array is represented by ie(e, j), where e, j, and ie(e, j) refer
to the element, local node, and corresponding global node. Here ie(1, 1) = 1; ie(1, 2) = 2;
ie(2, 1) = 2; ie(2, 2) = 3 and so on.
Since all nodal Q values are positive into the element, their sum must be supplied at
the node, externally. Thus, the sum of elemental Q contributions from adjacent ele-
ments must coincide with the resulting nodal Q value, for example, Q2 = Q21 + Q12
at node 2. In the current problem, no other heat sources, other than heat generation rep-
resented by Ṗ, are experienced. As a result, assembling the element equations and using
the nodal mapping array,
⎡ ⎤⎧ ⎫ ⎧ ⎫
1 −1 0 0 ⎪⎪ T1 ⎪
⎪ ⎪
⎪ ΔxAṖ/2 ⎪
⎪
kA ⎢ ⎥⎨ ⎬ ⎨ ⎬
⎢ −1 2 −1 0 ⎥ T2 = ΔxAṖ
(10.44)
⎣
Δx 0 −1 2 −1 ⎪ ⎦
⎪ T3 ⎪
⎩ ⎪ ⎪ ΔxAṖ ⎪
⎭ ⎪
⎩ ⎪
⎭
0 0 −1 1 T4 ΔxAṖ/2
The coefficients in the left side of the matrix constitute the global stiffness matrix, while
the right side is called the global right-side vector.
Computational Heat Transfer 427
5. In the fifth step, boundary conditions are applied. In this problem, T1 = 0 = T4 at the left
and right boundaries, respectively. These boundary conditions replace the first and
fourth rows of the global stiffness matrix and right-side vector.
6. Finally, the discrete equation set is solved with a method such as Gaussian elimination.
The interior temperatures are obtained as:
Δx2 Ṗ
T2 = = T3 (10.45)
k
Thus internal heat generation leads to a temperature increase within the solid. The
accuracy of the finite element predictions can be improved by using more grid points
in the discretization of the problem domain.
The stiffness matrix in a Galerkin formulation has three significant properties or features.
First, for steady-state diffusion-type problems, the stiffness matrix is symmetric, so
cmn = cnm. Second, using the convention of positive Q values into an element, the diagonal
coefficients are positive, cmm . 0. Lastly, for steady-state and linear problems, the sum of
entries along a row or column of the stiffness matrix is zero before boundary conditions
are applied.
Recall that the superposition principle can be applied to linear systems such as steady-state
heat conduction (Chapter 2). For example, if Tn represents the solution of nodal temperature
values, then Tn plus any constant, C, is also a solution,
nnp
nnp
nnp
cmn (Tn + C) = cmn Tn + C cmn = 0 (10.46)
n=1 n=1 n=1
where nnp refers to the number of nodal points. The first term on the right side is a summa-
tion involving Tn which becomes zero since Tn is a solution itself, thereby requiring that
∑cmn = 0 (along a row or column). Therefore entries along a column or row must sum to
zero for linear problems.
The previous example outlined the main steps in the finite element method for one-dimen-
sional elements. The geometric preliminaries for two-dimensional finite element analysis
will be described in the next section.
ϕ = a1 + a1 x + a3 y (10.47)
428 Advanced Heat Transfer
Local node 1 2e
(x1, y1) t
ξ1 1e
(x2, y2)
3e
s
ξ2
(x3, y3) 4e
FIGURE 10.3
Triangular and quadrilateral finite elements.
where the unknown coefficients, a1, a2, and a3 can be determined based on substitution of
nodal values.
For example, at node 1, the position is (x1, y1) and the scalar value is φ1. Similarly, at local
nodes 2 and 3 and by inverting the resulting 3 × 3 system from Equation 10.47 in terms of the
unknown coefficients, it can be shown that,
1
a1 = [(x2 y3 − y2 x3 )ϕ1 + (x3 y1 − x1 y3 )ϕ2 + (x1 y2 − x2 y1 )ϕ3 ] (10.48)
2A
1
a2 = [(y2 − y3 )ϕ1 + (y3 − y1 )ϕ2 + (y1 − y2 )ϕ3 ] (10.49)
2A
1
a3 = [(x3 − x2 )ϕ1 + (y3 − y1 )ϕ2 + (x2 − x1 )ϕ3 ] (10.50)
2A
1
A = (x2 y3 − x3 y2 − x1 y3 + x1 y2 + y1 x3 − y1 x2 ) (10.51)
2
Rearranging terms, the interpolation in Equation 10.47 can be rewritten in terms of the
shape functions, N1, N2, and N3, as follows:
ϕ = N1 ϕ1 + N2 ϕ2 + N3 ϕ3 (10.52)
where,
1
N1 = (a1 + b1 x + c1 y) (10.53)
2A
1
N2 = (a2 + b2 x + c2 y) (10.54)
2A
1
N3 = (a3 + b3 x + c3 y) (10.55)
2A
Computational Heat Transfer 429
and,
a1 = x2 y3 − x3 y2 ; a2 = x3 y1 − x1 y3 ; a3 = x1 y2 − x2 y1 (10.56)
b1 = y2 − y3 ; b2 = y3 − y1 ; b3 = y1 − y2 (10.57)
c1 = x 3 − x 2 ; c2 = x1 − x3 ; c3 = x2 − x1 (10.58)
∂Nj bj ∂Nj cj
= ; = (10.61)
∂x 2A ∂y 2A
where j = 1, 2, 3.
Within a triangular element, global coordinates (x and y) can be expressed in terms of local
coordinates (ξ1, ξ2) in the following fashion:
These equations can be inverted or isolated in terms of the local coordinates, yielding,
a1 + b1 x + c1 y 1
ξ1 = = (a1 + b1 x + c1 y) ; N1 (10.64)
a1 + a2 + a3 2A
1
ξ2 = (a2 + b2 x + c2 y) ; N2 (10.65)
2A
Since there are three nodal points in a linear triangular element, three shape functions
are required. Consider a point within the element and lines connecting this point to the
nodes of the element. Then the local coordinates may be interpreted as ratios of each sub-
area to the total area of the element. For example, ξ1 = A1/A, where A1 refers to the sub-
area consisting of side 2–3 joined with the interior specified point. Similar expressions are
obtained for the other subareas and local coordinates, including ζ3, which is given by the
ratio of A3/A. Since the sum of individual subareas is A, dividing this summed equation
by A yields:
ξ 3 = 1 − ξ 1 − ξ2 ; N 3 (10.66)
430 Advanced Heat Transfer
Thus there are only two independent local coordinates (ξ1 and ξ2) for linear triangular
elements.
The local coordinates can be used for interpolation of the dependent scalar, φ, as well as
geometrical interpolation,
6
ϕ= Ni (ξ1 , ξ2 , ξ3 )Φi (10.68)
i=1
where,
Using higher order interpolation can increase the accuracy of interpolation within an ele-
ment, but generally at the expense of increased computational time and storage.
Local and global coordinates must be related with each another for differentiation of the
scalar variables involving triangular elements. Using the chain rule of calculus,
∂ϕ ∂ϕ ∂x ∂ϕ ∂y
= + (10.71)
∂ξ1 ∂x ∂ξ1 ∂y ∂ξ1
∂ϕ ∂ϕ ∂x ∂ϕ ∂y
= + (10.72)
∂ξ2 ∂x ∂ξ2 ∂y ∂ξ2
Rearranging and inverting these equations by writing the derivatives in global coordi-
nates with respect to the local derivatives,
⎧ ⎫ ⎛ ⎞⎧ ⎫
⎪ ∂ϕ ⎪ ∂y ∂y ⎪ ∂ϕ ⎪
⎪
⎨ ⎬ ⎪ − ⎪
⎨ ⎪
⎬
∂x 1⎜⎜ ∂ξ2 ∂ξ1 ⎟ ⎟ ∂ξ1
=
∂ϕ ⎪ |J| ⎝ ∂x ∂x ⎠⎪ (10.73)
⎪
⎪ ∂ϕ ⎪
⎩ ⎪ ⎭ − ⎪
⎩ ⎪
⎭
∂y ∂ξ2 ∂ξ1 ∂ξ2
Computational Heat Transfer 431
where,
∂x ∂y ∂y ∂x
|J| = − (10.74)
∂ξ1 ∂ξ2 ∂ξ1 ∂ξ2
dA = dx × dy (10.75)
where,
∂x ∂x
dx = dξ ξ̂ + dξ ξ̂ (10.76)
∂ξ1 1 1 ∂ξ2 2 2
∂y ∂y
dy = dξ ξ̂ + dξ ξ̂ (10.77)
∂ξ1 1 1 ∂ξ2 2 2
Here, the overhat notation designates the unit vector in the direction of interest.
Combining these above three equations,
∂x ∂y ∂x ∂y
dA = − dξ1 dξ2 ξ̂3 (10.78)
∂ξ1 ∂ξ2 ∂ξ2 ∂ξ1
where ξ3 is the unit vector normal to the ξ1 and ξ2 planes. The magnitude may be
written as:
For example, if an exponential heat source appears in the energy equation, then a spatial
integration of this heat generation term within the element can be calculated from the
above equation.
Some integrals can be calculated in a closed form, whereas for more complicated inte-
grands, a closed-form analytical integration may not be possible. In such cases, numerical
integration is required. A commonly used numerical integration technique is Gaussian
quadrature, which can be written as follows for triangular elements:
1 1−ξ1
n
f (ξ1 , ξ2 , ξ3 ) dξ2 dξ1 = Wi f ξi1 , ξi2 , ξi3 (10.81)
0 0 i=1
432 Advanced Heat Transfer
where the weights, Wi, and abscissae, ξij , are summarized for a few selected elements in
Table 10.1. The subgrid quadrature can have a significant impact on the numerical method’s
overall accuracy, particularly in transient flow problems.
The order of accuracy indicates the magnitude of the integration error in terms of a char-
acteristic length of the triangular element, such as a side length. For example, for a linear tri-
angle, the integration is second-order accurate since the error is approximately reduced by a
factor of 4 when the element size is reduced in half, whereas a quadratic element yields
third-order accuracy due to a reduction by a factor of 8, or proportional to h 3 rather than h 2.
1
N1 (s, t) = (1 + s)(1 + t) (10.85)
4
1
N2 (s, t) = (1 − s)(1 + t) (10.86)
4
1
N3 (s, t) = (1 − s)(1 − t) (10.87)
4
TABLE 10.1
Gaussian Quadrature for Triangular Elements
n i ξ1i ξ2i ξ3i Wi Order of Accuracy
For linear quadrilateral elements, spatial derivatives of φ can also be obtained based on the
shape functions as follows,
∂ϕ 4
∂Ni
= Φi (10.88)
∂x (s,t) i=1 ∂x (s,t)
∂ϕ 4
∂Ni
= Φi (10.89)
∂y (s,t) i=1
∂y (s,t)
Using the chain rule of calculus, the shape function derivatives can be expressed as:
⎧ ⎫ ⎛ ∂y ⎧ ⎫
⎪
⎪ ∂Ni ⎪ ⎪ ∂y ⎞⎪ ∂Ni ⎪
⎨ ⎬ − ⎨ ⎬
∂x 1⎜ ∂s ⎟
∂N = ⎝ ∂t ⎠ ∂s (10.90)
⎪
⎪ i ⎪
⎪ |J| ∂x ∂x ⎩ ∂Ni ⎪
⎪ ⎭
⎩ ⎭ −
∂y ∂t ∂s ∂t
∂x ∂y ∂y ∂x
|J| = − (10.91)
∂s ∂t ∂s ∂t
Also,
∂x 4
∂Ni ∂x 4
∂Ni
= xi ; = xi (10.92)
∂s i=1
∂s ∂t i=1
∂t
∂y 4
∂Ni ∂y 4
∂Ni
= yi ; = yi (10.93)
∂s i=1
∂s ∂t i=1
∂t
The shape function derivatives with respect to the local coordinates can be readily evalu-
ated from the expressions for Ni (s, t).
Functional expressions can be integrated based on the following area transformation:
1 1
f (x, y) dx dy = f (s, t)Det(J) ds dt (10.94)
A −1 −1
where (si, ti), Wi, and Wj correspond to the Gauss points and weights, respectively (see
Table 10.2), and ngp refers to the number of Gauss points.
If analytical expressions for integrated quantities over the finite element are difficult or
impractical to evaluate, then these quadrature rules can be used to find accurate approxima-
tions of the integrations.
434 Advanced Heat Transfer
TABLE 10.2
Gauss–Legendre Quadrature for Quadrilateral Elements
n Number of Gauss Points Gauss Points Weights Error
Recall that Galerkin’s method selects the weight functions equal to the basis functions
(shape functions) of the approximate solution. This selection is motivated by Fourier anal-
ysis (Chapter 2) which obtains the unknown coefficients of an infinite series for the tem-
perature distribution after multiplying by orthogonal functions and integrating over the
domain. Similarly in the Galerkin weighted residual method, shape functions are selected
as the basis functions and integrated over the domain to determine the unknown coeffi-
cients of an approximate solution. This section uses Galerkin’s method to develop the
two-dimensional equations of the finite element method. The solution method is demon-
strated through another example problem involving two-dimensional heat conduction in
a planar wall.
1. The first step requires discretization of the domain into eight uniformly distributed tri-
angular elements (see Figure 10.4). There are different possible methods of subdividing
y
Adiabatic wall
QIN 3 6 9
Element
(8)
number
T=0 (4)
x 1 4 7
FIGURE 10.4
Problem domain and finite element discretization of heat conduction in a planar wall.
Computational Heat Transfer 435
the domain into eight elements. For example, the discretization could involve four sub-
squares, each of which contains the same triangular alignment, or eight triangles all
joined by a common point at the center of the domain. Although both methods would
yield the same final solution, the former approach will be adopted. The latter method
with a common center point would contain a larger final matrix bandwidth since the
center point would contain contributions from all nodes.
2. Interpolation is handled with linear, triangular, isoparametric elements, as follows,
where Ni represent the shape functions. Although the shape function coefficients are
generally obtained separately for each element, these coefficients are identical in the
current problem due to the orientation of elements within the domain. As a result,
a1 = 0.01, a2 = 0 = a3 = b3 = c2, b1 = –0.1 = c1, and b2 = 0.1 = c3. Also, 2A = 0.01, and
therefore the temperature becomes:
T = 1 − 10x − 10y T1e + (10x)T2e + 10y T3e (10.97)
in terms of the global coordinates.
3. Using Galerkin’s weighted residual method for the steady-state heat conduction
equation,
∂ ∂T ∂ ∂T
Ni k + k dxe dye = WRei (10.98)
Ae ∂x ∂x ∂y ∂y
where WRei refers to the weighted residual and the superscript e refers to elemental, or
local within the element. Using integration by parts for the first part of the integral,
∂ ∂T ∂T ∂T ∂Ni ∂T e e
Ni k dxe dye = Ni k dye − Ni k dye − k dx dy
∂x ∂x ∂x y(x+ ) ∂x y(x− ) ∂x ∂x
Ae y(x+ ) y(x− ) ye xe
(10.99)
The first and second terms on the right side can be understood by reference to a
sample triangular element subdivided into differential segments of thickness dy
with endpoints of y(x −) (left side) and y(x +) (right side). Within a segment, dS and
θ are used to represent the segment length along the triangle edge and
angle between the end edge and the y (vertical) axis, respectively. Based on these def-
initions,
dy = |dS|cos(θ) (10.100)
which applies to either the left or right sides of the segment. In terms of the direction
cosine with respect to the x-axis, denoted by l1,
which applies to the right side at y(x +), where θ1 = θ. From geometrical consider-
ations of the component of the heat flux vector normal to the element’s edge, it
can be shown that,
∂T ∂T
k = k l1 (10.102)
∂x ∂n
436 Advanced Heat Transfer
∂ ∂T ∂T ∂Ni ∂T e e
Ni k dxe dye = Ni k l21 |dS|î − k dx dy (10.103)
Ae ∂x ∂x bndy ∂n bndy y x ∂x
e e ∂x
where î is the unit x direction vector and bndy refers to the boundary. A similar result is
obtained for the second term in Equation 10.98, except that l1 is replaced by l2 (direction
cosine with respect to the y-axis). Then, combining this result with the analogous y-
direction result into Equation 10.98,
∂T e ∂Ni ∂T ∂Ni ∂T
Ni k dS − k + k dxe dye = WRei (10.104)
bndy ∂n ye xe ∂x ∂x ∂y ∂y
where the direction cosines have summed to 1, thereby not appearing individually in
the equation.
Various boundary conditions can be incorporated through the boundary integral
term. A general form of the boundary equation can be expressed as:
∂T
k = −he T + Ce (10.105)
∂n
where he and Ce refer to the convection coefficient and a suitable constant describing the
boundary condition, respectively. Four types of common boundary conditions are
listed below:
a. Dirichlet: h → ∞ and C → hTs (where Ts refers to a specified surface
temperature);
b. Neumann: h → 0 and C → qs (where qs is a specified wall heat flux);
c. Adiabatic (insulated boundary): h → 0 and C → 0;
d. Robin (convection): h → h (convection coefficient) and C → hT∞.
If a boundary condition is not specified, then a default zero-flux (adiabatic) condition
is recognized by the numerical model.
Substituting Equations 10.105 into 10.104,
∂Ni ∂T ∂Ni ∂T
k + k dxe dye + Ni he T dSe = Ni Ce dSe − WRei (10.106)
ye xe ∂x ∂x ∂y ∂y Se Se
The first term represents contributions from all elements in the domain. The second
and third terms (surface integral terms) represent only the boundary element contribu-
tions, since all internal adjacent elements have common sides where the surface inte-
grals mutually self-cancel.
The temperature is approximated based on interpolation via the shape functions,
3
T= Nj Tje (10.107)
j=1
where,
!
∂Ni ∂Nj ∂Ni ∂Nj
kije = kde + kge = k + k dxe dye + Ni he Nj dSe (10.109)
ye xe ∂x ∂x ∂y ∂y Se
rei = {reg } + {rer } = Ni Ce dSe − WRei (10.110)
Se
In these equations, the subscripts d, g, and r refer to interior elements, boundary ele-
ments, and residual terms, respectively.
The values of the coefficients in the stiffness matrix are then evaluated for triangular
elements. Based on Equations 10.52 through 10.60, the interior coefficients are obtained
as follows:
bi bj ci cj bi bj ci cj
kij,d e = k + k dxe dye = k +k (10.111)
ye xe 2A 2A 2A 2A 4A 4A
For the specific domain discretization in this example problem (see Figure 10.4), the
coefficients can be calculated so that the stiffness matrix becomes:
⎡ ⎤
2 −1 −1
e 1 ⎢ ⎥
kd = ⎢ −1 1 0 ⎥ (10.112)
2⎣ ⎦
−1 0 1
It can be observed that the stiffness matrix is symmetric and the sum of entries along
a column or row is zero (as expected). Also, rer in Equation 10.110 is not included since it
becomes zero after summation over all elements, as expected by the weighted residual
method.
4. In this fourth step of the finite element method, all elements within the domain are
assembled together to construct the entire mesh. An element definition array, denoted
by ie(e, j) (where e is the element number and j denotes the local node number), is used
as a mapping between local and global nodes in the assembly process. The table below
outlines this mapping for the mesh discretization illustrated in Figure 10.4. The values
within the table represent the ie(e, j) array, where the top row refers to the local node
numbers ( j).
For example, the mapping array indicates that local node 2 in element 3 corresponds
to global node 5 in the finite element mesh. The mapping array, ie(e, j), identifies the
row and column in the global stiffness matrix that corresponds to the entry in the local
stiffness matrix.
438 Advanced Heat Transfer
where i = ie(e, iw) and j = ie(e, jw) are the row and column numbers in the global
matrix, respectively. Also, iw = 1, 2, 3 and jw = 1, 2, 3 are the row and column num-
bers of the local stiffness matrix, respectively. The updates in the loop involving
Equation 10.113 mean that the global stiffness matrix entry is based on the previous
global matrix value plus the current value in the local stiffness matrix.
The entries from the local stiffness matrix are placed in the row and column for which
the global node corresponds to the local node from the given element. For example, in
the local stiffness matrix for element 1, the second row is placed in the fourth row of the
global matrix since ie(1, 2) = 4. This proper placement is obtained when the ie(1, 2) entry
is used to identify index i in the calculation of the row number for a(i, j).
After all local stiffness matrices and right-side vectors are assembled into the global
matrix, the following result is obtained for the current problem:
⎡ ⎤⎧ ⎫ ⎧ ⎫
2 −1 0 −1 0 0 0 0 0 ⎪ ⎪ T1 ⎪⎪ ⎪
⎪ 0⎪⎪
⎢ −1 4 −1 0 −2 0 0 0 0 ⎥⎪⎪
⎪ T ⎪
⎪
⎪
⎪
⎪
⎪ 0⎪⎪
⎪
⎢ ⎥⎪ 2 ⎪ ⎪ ⎪
⎢ 0 −1 2 0 0 −1 0 0 0 ⎥⎪⎪
⎪ T ⎪
⎪
⎪
⎪
⎪
⎪ 0 ⎪
⎪
⎪
⎢ ⎥⎪⎪
3 ⎪
⎪ ⎪
⎪ ⎪
⎪
⎢ −1 0 0 4 −2 0 −1 0 0 ⎥⎪⎨ T 4⎬
⎪ ⎪
⎨ 0 ⎪
⎬
⎢
1⎢ ⎥
0 −2 0 −2 8 −2 0 −2 0 ⎥ T5 = 0 (10.114)
⎢ ⎥ ⎪
2⎢
0 −1 0 −2 4 0 −1 ⎥ ⎪ T6 ⎪⎪ ⎪
⎪ 0⎪
⎪
⎢ 0 0 ⎥⎪⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎢ 0 0 0 −1 0 0 ⎥
2 −1 0 ⎥⎪ ⎪ T7 ⎪⎪ ⎪
⎪ 0⎪⎪
⎢ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎣ 0 0 0 0 −2 0 −1 4 −1 ⎦⎪ ⎪ T ⎪ ⎪ ⎪
⎪0⎪ ⎪
⎪
⎩ 8⎪ ⎭ ⎪
⎩ ⎪ ⎭
0 0 0 0 0 −1 0 −1 2 T9 0
It can be observed that the sum of entries along a row or column is zero. Also, the
global stiffness matrix is symmetrical (as expected).
5. In this step, the boundary conditions for the physical problem are applied. The
boundary integral terms in Equation 10.104 must be added into the appropriate
places in the local and global stiffness equations. Each elemental boundary region
is approximated by a linear section between the two local nodes on the boundary.
For example, considering a counterclockwise ordering convention for an element
with local nodes 1 and 2 on the domain boundary, the following boundary inte-
gral is added to kije :
e
kij,g = Ni he Nj dSe (10.115)
Se
Similarly,
⎧ ⎫
" # Ce ΔSe ⎨ 1 ⎬
reg = 1 (10.117)
2 ⎩ ⎭
0
which is added to the right-side vector in the global system on the domain boundaries.
Computational Heat Transfer 439
Since the results were obtained for side 1–2, the results must be generalized for cases
where the boundary lies on a side other than side 1–2,
h2 ΔSe
e
kij,g =C (10.118)
6
where C = 2 if i = j and the boundary is side i–j; C = 1 if i ≠ j and the boundary is side i–j,
and C = 0 if the boundary is not on side i–j. Similarly, for the right side,
Ce ΔSe
rei,g = (10.119)
2
when local node i lies on the current boundary and rei,g = 0 otherwise, if node i does not
lie on the current boundary.
The implementation of boundary conditions also requires a mapping between local
and global nodes to place the boundary stiffness matrix components and right-side
entries correctly in the global system. For the current problem, the table below shows
the global nodes, N1 and N2, together with boundary condition coefficients, h and C,
for each surface, s, along the domain boundaries.
s N1 N2 h C
1 2 1 0 100
2 3 2 0 100
3 6 3 0 0
4 9 6 0 0
5 1 4 0 0
6 4 7 0 0
7 7 8 105 0
8 8 9 105 0
Using this boundary node mapping to add the boundary stiffness matrix, Equation
10.118, and right side, Equation 10.119, to the global system, the following final global
system of algebraic equations is obtained.
⎡ ⎤⎧ ⎫ ⎧ ⎫
2 −1 0 −1 0 0 0 0 0 ⎪
⎪ T1 ⎪⎪ ⎪
⎪ 5⎪⎪
⎢ −1 4 −1 0 −2 0 0 0 0 ⎥⎪⎪
⎪ T ⎪
⎪
⎪
⎪
⎪
⎪ 10 ⎪
⎪
⎪
⎢ ⎥⎪⎪
2 ⎪
⎪ ⎪
⎪ ⎪
⎢ 0 −1 2
⎢ 0 0 −1 0 0 0 ⎥
⎥⎪
⎪
⎪ T3 ⎪⎪
⎪
⎪
⎪
⎪ 5⎪⎪
⎪
⎪
⎢ −1 0 ⎥⎪⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
0 4 −2 0 −1 0 0 ⎥ ⎨ T ⎬ ⎨ 0 ⎬
1⎢
⎢ 0 −2 0 −2 8 −2 ⎥
4
0 −2 0 T
⎥⎪ ⎪ ⎪ ⎪ = 0 (10.120)
2⎢
5
⎢ 0 0 −1 0 −2 4 −1 ⎥ ⎪ T6 ⎪ ⎪ 0⎪
⎢ 0 0 ⎥⎪⎪ ⎪ ⎪ ⎪
⎪ ⎪
⎪
⎢ 0 0 0 −1 0 0 3, 335.3 4, 999 0 ⎥⎪⎪
⎪ T
⎪
⎪
⎪
⎪
⎪
⎪ 0⎪
⎪
⎪
⎢ ⎥⎪⎪ 7⎪
⎪ ⎪
⎪ ⎪
⎪
⎣ 0 0 0 0 −2 0 1, 665.7 3, 337.3 1, 665.7 ⎪⎦ ⎪ T ⎪
⎪ ⎪
⎪ 0 ⎪
⎪
⎪
⎩ ⎪ 8
⎭ ⎪ ⎩ ⎪ ⎭
0 0 0 0 0 −1 0 1, 665.7 3, 335.3 T9 0
6. In this last step, the system of algebraic equations is solved by a method such as Gauss-
ian elimination. The final solution yields T1 = 20◦ C = T2 = T3 along the left boundary;
T4 = 10◦ C = T5 = T6 along the vertical midplane; and T7 = 0 = T8 = T9 along the right
boundary. It can be readily verified by Fourier’s law that these temperature results
imply that the heat flow through the solid is 100 W, which correctly balances the spec-
ified heat inflow from the left boundary under steady-state conditions. Furthermore,
using the shape functions and nodal temperature values in elements along the right
boundary, it can also be verified that the heat flow rate there is also 100 W, as expected.
440 Advanced Heat Transfer
The results of a finite element solution should be grid independent. In other words, the
results should converge to the same final values, or within a specified small convergence tol-
erance, when finer grids are used in the domain discretization. In a convergence study, one
or more key parameters are selected and monitored during each subsequent grid refinement
to confirm the values approach a converged solution. Although it cannot always be ensured,
all grids within a convergence study should be contained within each subsequent refined
mesh. Also, all material within the domain should be fully contained by each mesh in the
convergence studies. In a grid refinement study, the type of interpolation of problem vari-
ables should not change from one mesh to another mesh.
where Γ, λ, and Q refer to the diffusion coefficient, capacitance, and source term, respec-
tively. For example, Γ = k (thermal conductivity) and λ = ρcp for heat transfer problems.
To accommodate the transient term in Galerkin’s weighted residual method, new shape
functions can be defined, such as Ni (x, y, t), including a time dependence. But this approach
generally leads to complicated resulting mathematical expressions in the weighted residual
integrations.
Alternatively, the transient derivative of φ can be approximated in the following manner:
nnpe
ϕ̇ = Njt ϕ̇j (10.122)
j=1
where the overdot denotes a time derivative, the superscript t refers to transient and nnpe is
the number of nodal points per element (three for triangular elements).
Following the procedure of Galerkin’s weighted residual method in the previous section,
it can be shown that the finite element formulation becomes:
nel $ %
∂Ni ∂Nj ∂Ni ∂Nj
kij = Γ + Γ dV e + Ni he Nj dSe (10.124)
e=1 V e ∂x ∂x ∂y ∂y Se
nel
ri = Ni Q dV +
e e e
Ni C dS e
(10.125)
e=1 Ve Se
nel
cij = Ni λNj dV e (10.126)
e=1 Ve
Computational Heat Transfer 441
Here, he anc Ce represent elemental coefficients for general Robin boundary conditions
along the external surfaces and nel refers to the number of elements in the domain. The latter
terms in Equations 10.124 and 10.125 are only evaluated along the external boundaries of
the domain.
In a consistent transient model, the transient term is assumed to vary spatially in the subele-
ment about the node in a manner consistent with the order of interpolation used in the other
diffusion terms. Substituting linear shape functions and performing the resulting integra-
tions in Equation 10.126, the entries in the capacitance matrix, cij, become λΔx/3 along the
diagonal (i = j) and λΔx/6 in the off-diagonal entries (i ≠ j) for one-dimensional elements
in a 2 × 2 elemental matrix. An analogous 3 × 3 matrix is obtained for triangular elements,
however, the entries are multiplied by a factor of 1/2 of the values for the one-
dimensional element.
Alternatively, in a lumped transient model, the transient integral is taken as constant over
the time step and subelement. A step function is used for the shape function in Equation
10.126, yielding a value of zero in the left half-element (x* ≤ 1/2 in one-dimensional ele-
ments) and 1 in the right half-element. Then after performing the resulting integrations in
Equation 10.126, the capacitance matrix entries become λΔx/2 along the diagonal (i = j)
and zero for off-diagonal entries (i ≠ j).
Then the transient derivative of φ in Equation 10.123 can be replaced by a backward dif-
ference in time, as follows,
ϕt+Δt
j − ϕtj
ϕ̇j = (10.127)
Δt
where t + Δt is the time level following time t. At a point in time between t and t + Δt, the
variable φj is extrapolated forward from time t and backward from t + Δt.
The other terms in Equation 10.123 can also be evaluated at any intermediate point in time,
for example, at t + βΔt,
& t+Δt '
t+βΔt
ϕj − ϕtj
[kij ]{ϕj } + [cij ] = {ri } (10.128)
Δt
ϕt+βΔt
j = (1 − β)ϕtj + βϕt+Δt
j (10.129)
The terms in round brackets represent the modified stiffness matrix and right-side vector
for transient problems.
The explicit and implicit methods are then obtained as special cases of Equation 10.130.
For example, β = 0 represents an explicit formulation, where future φj values are based on
calculations of diffusion and other terms at the previous time level. As discussed previously,
explicit time advance is usually subject to stringent time step constraints to ensure numerical
stability.
442 Advanced Heat Transfer
This formulation leads to a full elemental stiffness matrix with all nonzero entries. These
full nonzero entries preclude a simple matrix inversion so the expected benefits of reduced
computational effort in an explicit formulation are generally not realized. As a result, a con-
sistent formulation is usually not used in explicit schemes.
A lumped explicit formulation yields only diagonal entries in the local stiffness matrices.
This significantly reduces the computational effort to solve the resulting algebraic equations.
A time step stability criterion can be observed from the first term on the right side of Equa-
tion 10.131. For small time steps, if all conduction heat flows are directed into a control vol-
ume, the right side of Equation 10.131 indicates that φ in the control volume will increase due
to the net heat inflow. However, for large time steps, the right side of Equation 10.131 may
become negative, thereby causing φ to decrease in a control volume even with a net heat
inflow to the control volume. Therefore, a time step limitation is required to maintain
numerical stability in explicit formulations.
A positive diagonal matrix to ensure numerical stability requires positive right-side values
in Equation 10.131. For the first term on the right side of Equation 10.131 to stay positive in
two-dimensional problems,
Δx2 + Δy2
Δt ≤ λ (10.132)
2Γ min
where the subscript min refers to the minimum value in the domain. These time step restric-
tions are often stringent constraints that make explicit methods impractical. The constraints
become more pronounced when Δx → 0 as the grid is refined.
Alternatively, implicit formulations are more commonly used methods. Here the diffu-
sion and source terms are evaluated at the new time level. Substituting β = 1 in Equation
10.130,
1 1
[kij ] + [cij ] {ϕj }t+Δt = [cij ]{ϕj }t + {ri }t+Δt (10.133)
Δt Δt
Increased temporal accuracy can be achieved by setting β = 1/2 in Equation 10.130. This
corresponds to a Crack–Nicolson scheme, whereby diffusion and source terms are evaluated
at the intermediate time level (t + Δt/2). Substituting β = 1/2 into Equation 10.130,
1 1 t+Δt 1 1
[kij ] + [cij ] {ϕj } = − [kij ] + [cij ] {ϕj }t + {ri }t+Δt/2 (10.134)
2 Δt 2 Δt
This formulation can be shown to have second-order accuracy in time. Despite its
increased accuracy, this method has practical disadvantages in storing all values at an inter-
mediate time level, t + Δt/2. As a result, the fully implicit method (β = 1) is more commonly
adopted and will be used hereafter in the remainder of this chapter.
Computational Heat Transfer 443
2
∂v ∂v ∂p ∂ v ∂2 v
L2 (u, v, p) = ρuo + ρvo + − μ + =0 (10.136)
∂x ∂y ∂y ∂x2 ∂y2
∂u ∂v
L3 (u, v, p) = + =0 (10.137)
∂x ∂y
where the superscript o denotes an evaluation at the previous iteration as a lagged value.
The system of equations represents three coupled equations in three unknowns—two veloc-
ity components (u, v) and pressure ( p).
Since the nonlinear convection terms have been linearized, an iteration procedure is
required in the numerical solution. An initial estimate of the lagged velocity from the pre-
vious iteration is used in the convection terms. Then the Navier–Stokes equations are solved
and the new solution is compared with the initial estimate. The solution procedure is
repeated by using the new solution as the next linearized estimate. Convergence is achieved
when there is close agreement between the previous and current values of the velocity and
pressure fields.
Using the weighted residual method for Equations 10.135 through 10.137,
∂ũ ∂ũ ∂p̃ ∂ ∂ũ ∂ ∂ũ
Wiu ρuo + ρvo + − μ − μ dV = 0 (10.138)
V ∂x ∂y ∂x ∂x ∂x ∂y ∂y
∂ṽ o ∂ṽ ∂p̃ ∂ ∂ṽ ∂ ∂ṽ
Wiv ρuo
+ ρv + − μ − μ dV = 0 (10.139)
V ∂x ∂y ∂y ∂y ∂x ∂y ∂y
p ∂ũ ∂ṽ
Wi + dV = 0 (10.140)
V ∂x ∂y
where the superscripts u, v, and p refer to the x-momentum, y-momentum, and continuity
equations, respectively.
444 Advanced Heat Transfer
Define the approximate local velocities and pressure as a linear combination of nodal val-
ues weighted by the shape functions at their corresponding nodes,
nnpe
nnpe
nnpe
p
ũ = Niu ui ; ṽ = Niv vi ; p̃ = Ni pi (10.141)
i=1 i=1 i=1
where nnpe refers to the number of nodal points per element (e.g., nnpe = 2 for linear ele-
ments and nnpe = 4 for quadrilateral elements). Using Galerkin’s method, the weight func-
tions are selected as equal to the shape functions so that Wiu = Niu , Wiv = Niv , and
p p
Wi = Ni .
Assembling all finite elements and using integration by parts in the x-momentum equa-
tion, analogously to the method presented earlier for heat conduction, leads to:
nel
nel
∂ũ ∂ũ ∂p̃ ∂Ni ∂ũ ∂Ni ∂ũ ∂ũ
Niu ρuo + ρvo + + μ + μ dV e − Niu μ dSe = 0
e=1 Ve ∂x ∂y ∂x ∂x ∂x ∂y ∂y e=1 S
e ∂n Se
(10.142)
where the second summation represents the equivalent nodal force in the element due to the
shear stress, τ, in the x-direction. It is analogous to the surface heat flux in the heat
conduction equation.
Rewriting the equations in matrix form,
( )
[kuu ]{ui } + [kup ]{pi } = rui (10.143)
( )
[kvv ]{vi } + [kvp ]{pi } = rvi (10.144)
( p)
[kpu ]{ui } + [kpv ]{vi } = ri (10.145)
nel
∂Nju ∂Nju
u u
∂Ni ∂Nj ∂Niu ∂Nj
u
kijuu = Ni ρu
u o
+ ρvo
+ μ + μ dV e (10.146)
e=1 V e ∂x ∂y ∂x ∂x ∂y ∂y
nel
∂Nj
p
up
kij = Niu dV e (10.147)
e=1 Ve ∂x
nel
rui = Niu τeu dSe (10.148)
e=1 Se
nel
∂Njv ∂Njv
v v
∂Ni ∂Nj ∂Niv ∂Nj
v
kijvv = Niv ρuo + ρvo + μ + μ dV e (10.149)
e=1 V e ∂x ∂y ∂x ∂x ∂y ∂y
Computational Heat Transfer 445
nel
∂Nj
p
vp
kij = Niv dV e (10.150)
e=1 Ve ∂y
nel
rvi = Niv τev dSe (10.151)
e=1 Se
nel
∂Nju
pu p
kij = Ni dV e (10.152)
e=1 Ve ∂x
nel
∂Njv
pv p
kij = Ni dV e (10.153)
e=1 Ve ∂x
p
ri = 0 (10.154)
where τeu and τev refer the net elemental shear stress components in the x- and y-directions,
respectively. The surface integral terms, rui and rvi are evaluated only at external surfaces. The
above expressions may be written in a local form (locally within an element) by removal of
the summation over all elements. The local contribution of each element is assembled into
the global system on an element-by-element basis.
Boundary conditions must be specified for closure of the problem definition. Examples of
common types of CFD boundary conditions are listed below.
As discussed earlier, an iterative solution procedure is required for the nonlinear convec-
tion terms. The following steps can be used for an iterative solution.
1. Make initial estimates of the uo and vo values (e.g., a start-up value or the value from
a previous iteration). These values must satisfy conservation of mass.
2. Solve the finite element equations for the new (updated) u, v, and p values.
3. Update the estimate in the first step as follows: uo → uo + ω(u – uo) and vo → vo + ω
(v − vo). Over-relaxation (1 ≤ ω ≤ 2) is used if convergence is proceeding but slowly
446 Advanced Heat Transfer
The performance of the CFD solver is closely related to the structure of the matrix of coef-
ficients from the liner system of algebraic equations. The bandwidth of the matrix (maximum
number of columns of nonzero coefficients in any given row of the matrix) significantly
affects the solver performance. The placement of coefficients in the stiffness matrices and ori-
entation of elements are key factors which influence the bandwidth. Rather than ordering of
variables in the usual fashion (u followed by v and p), a rearrangement of the ordering
sequence can drastically reduce the overall bandwidth. For instance, at each node, the con-
tinuity and y-momentum equations can be interchanged, thereby ensuring nonzero entries
on the diagonal and reducing the matrix bandwidth.
∂(ρϕ)
+ ∇ · (ρvϕ) = ∇ · (Γϕ ∇ϕ) + Sϕ (10.156)
∂t
Using the Gauss divergence theorem of calculus to transform the volume integral to a sur-
face integral over the surface Si encompassing Vi, and taking the volume average of the last
term over the volume,
t+Δt t+Δt t+Δt t+Δt
∂(ρϕ)
dV dt + (ρvϕ) · ndS dt = (Γϕ ∇ϕ) · n dS dt + Sϕ dV dt
t Vi ∂t t Si t Si t Vi
*+++++++++++,-+++++++++++. *+++++++++++++,-+++++++++++++. *++++++++++++++,-++++++++++++++. *+++++++++,-+++++++++.
transient term convection fluxes diffusion fluxes source terms
(10.158)
where Si represents the total surface area of the control volume and n is a unit vector normal
to the surface and pointing outward.
Consider one-dimensional steady-state flow in Figure 10.5 without source terms but includ-
ing convection and diffusion. An approximation of φe at the integration point, e, between
nodes P and E, is needed to determine the convection flux in Equation 10.158. If the mass
flow rate is very low, it is reasonable to assume a linear profile, φe ∼ (φP + φE)/2. On the other
hand, if the mass flow rate is very high, from left to right, φe ∼ φP (upwinded value), or else
φe ∼ φE if the fluid moves from right to left. A general formulation of this interpolation can
be written as:
1 1
ϕe ≈ + sign(ṁe )γ e ϕP + − sign(ṁe )γ e ϕE (10.160)
2 2
where γe is a weighting function that depends on the mass flow rate. There are several options
for selecting the weight function as follows:
For diffusion dominated transport, a Central Difference Scheme (CDS) is commonly used.
Here γe = 0 represents a linear profile approximation of φ between nodes P and E. This approx-
imation is suitable in the limit of a low mass flow rate where diffusion is dominant. The grid
Peclet number, Pe = ρueΔx/Γ, can be used to assess the relative importance of diffusion and
convection, such as, Pe ≪ 1 (diffusion dominates) or Pe ≫ 1 (convection dominates).
An Upwind Difference Scheme (UDS) can be used for convection dominated flows. For
γe = 1/2 = γw, the resulting interpolation is weighted entirely in the upstream direction.
CDS is ideal if Pe ≪ 1 and UDS is ideal if Pe ≫ 1.
Alternatively, an Exponential Differencing Scheme (EDS) provides a smooth transition
between the above diffusion and convection limits. The weighting factor can be determined
based on an exact solution of the one-dimensional steady-state convection—diffusion equa-
tion as follows:
∂ϕ ∂2 ϕ
ρu =Γ 2 (10.161)
∂x ∂x
Solving this equation subject to boundary conditions, φ(0) = φP and φ(Δx) = φE, yields:
N Δx
vn n
Δy
w e
W uw P ue E
Control
volume
Vs s
FIGURE 10.5
Schematic of a mesh discretization in a finite volume method.
Computational Heat Transfer 449
1 exp (Pe/2) − 1
γe = − (10.163)
2 exp (Pe) − 1
At approximately Pe = 10, the weighting factor gives the UDS convection limit. An alter-
native approximation of the weighting factor over the entire range is given by:
Pe2
γe ≈ (10.164)
10 + 2Pe2
For the diffusion fluxes, consider the flux of φ across the east face of the control volume.
Again a linear variation of ∂φ/∂n can be assumed. With reference to Figure 10.5, the flux
can then be written as:
∂ϕ Γϕ A e Γϕ A e
(Γϕ ∇ϕ) · n dS ≈ Γϕ Ae = β ϕE − β ϕP (10.165)
Si ∂n e Δx Δx
where a diffusion weighting function, β, is introduced as a correction factor because the tem-
perature profile may not be linear in the presence of fluid motion. Using a similar derivation
as previously obtained for the convection weighting factor, it can be shown that:
Pe · ePe/2
β= (10.166)
ePe − 1
or alternatively by approximation,
1 + 0.005Pe2
β= (10.167)
1 + 0.05Pe2
piecewise constant approximation of the source term within the control volume, the inte-
grated source term becomes:
t+Δt
Sϕ dV dt ≈ Sϕ VP Δt (10.168)
t Vi
For the mass conservation equation, φ = 1. For two-dimensional flows and an orthogonal
grid, the equation is given by:
where the subscripts e, w, n, and s refer to the east, west, north, and south integration points
in Figure 10.5.
Substituting φ = u (x-momentum equation) and φ = v (y-momentum equation) in Equa-
tion 10.158, the general form of the conservation of momentum equations becomes:
ae ue = aunb unb + bu + (pP − pE )Ae (10.171)
ae vn = avnb vnb + bv + (pP − pN )An (10.172)
where nb, b, and A refer to neighboring nodal points, source term, and area of the cell
faces, respectively.
The iterative solution procedure proceeds in four steps as follows:
1. Guess the velocity fields, u and v, based on the previous time step or iteration.
2. Use these velocities to obtain the coefficients of the momentum equations.
3. Guess a pressure field, p*.
4. Solve the discretized momentum equations to obtain updated solutions of the
velocities, denoted by u* and v*.
The discretized equations for u*, v*, and p* obey the same momentum balances as above
for u, v, and p, and therefore:
ae u∗e = aunb u∗nb + bu + p∗P − p∗E Ae (10.173)
Computational Heat Transfer 451
an v∗n = avnb v∗nb + bv + p∗P − p∗N An (10.174)
To improve the updated velocities, u* and v*, such that they also satisfy the mass conser-
vation equation, a pressure correction term, p′ , and corresponding velocity correction terms,
u′ and v′ , are added to their current values. Thus the velocity and pressure fields are repre-
sented by:
u = u∗ + u′ ; v = v∗ + v′ (10.175)
p = p∗ + p′ (10.176)
Subtracting the u* and v* momentum equations from the u and v equations yields the fol-
lowing momentum equations for the velocity and pressure correction terms:
ae u′e = aunb u′ nb + p′ P − p′ E Ae (10.177)
an v′n = avnb v′ nb + p′ P − p′ N An (10.178)
This result indicates that the velocity corrections consist of two parts. The first term on
the right side refers to neighboring velocity corrections, while the second term is the pres-
sure correction difference between two adjacent points in the same direction as the
velocity field.
In the SIMPLE algorithm, the influence of nearby velocity corrections is neglected in Equa-
tions 10.177 and 10.178. The velocity correction equations are simplified as follows:
u′e = de p′ P − p′ E ; de = Ae /ae (10.179)
v′n = dn p′ P − p′ N ; dn = An /an (10.180)
The velocity corrections are determined and the updated velocities in Equation 10.175 are
then used as solutions at the next iteration level. The updated velocities are then substituted
into the mass conservation equation, Equation 10.170, yielding the following pressure cor-
rection equation,
aP p′P = anb p′ nb + bP (10.181)
where,
Under-relaxation may be required for p’ since the pressure correction may be relatively
large when the nearby velocity corrections are neglected in the SIMPLE algorithm.
The SIMPLE procedure can then be summarized by the following six steps.
452 Advanced Heat Transfer
In the SIMPLE algorithm, the nearby velocity corrections were neglected in Equations
In contrast, the SIMPLEC algorithm uses a “consistent” approximation
10.177 and 10.178. /
wherein the term anb u′nb is subtracted from both sides of Equation 10.177, rather than
neglecting the nearby velocity corrections. Therefore, instead of the simplified result of
Equation 10.179 in SIMPLE, the velocity correction equations become:
0 1
ae − anb u′e = anb u′ nb − u′ e + p′ P − p′ E Ae (10.185)
0 1
an − anb v′n = anb v′ nb − v′ n + p′ P − p′ N An (10.186)
In SIMPLEC, the first term on each right side is neglected, yielding the following expres-
sions for d:
Ae An
de = / ; dn = / (10.187)
ae − anb an − anb
This subtraction of the summation of coefficients of the neighboring velocities in the above
denominators reduces the influence of neglecting the adjacent velocity corrections in Equa-
tions 10.177 and 10.178. Also, no under-relaxation is needed for the pressure correction
equation in the SIMPLEC algorithm.
In SIMPLEC, the velocity correction equations in Equations 10.177 and 10.178 are
extended to neighboring velocity corrections by assuming:
This assumption extrapolates the expression of the velocity correction to adjacent loca-
tions so that the velocity correction equations become:
ae de Δp′e = anb dnb Δp′ nb + Ae Δp′ e (10.189)
Further comparisons between SIMPLE and its other variants for several test cases on var-
ious benchmark problems were presented by Zeng and Tao (2003).
the Navier–Stokes equations directly to the smallest scales in the flow (called the Kolmogorov
scales). Despite its excellent accuracy, the high computational costs and memory require-
ments for realistic industrial scale simulations are often prohibitive in DNS methods.
Another widely used approach involves large eddy simulations. This method applies mod-
els of the small scales in the flow field to resolve the large eddies. In Reynolds stress models,
separate stress transport equations are solved for individual turbulent stress and heat flux
terms. A variety of higher order turbulence quantities, such as the pressure strain (interac-
tion of fluctuating pressure with fluctuating strain rates) are introduced. The resulting com-
plexity again becomes prohibitive in many applications.
The most commonly used approach in industrial applications is called a k–ϵ model. This
method solves the coupled turbulent kinetic energy and dissipation rate equations (Chapter
3). The eddy viscosity hypothesis (Chapter 3), together with the k–ϵ model, can provide rea-
sonably accurate results for many flows. The procedure initially solves the mean flow equa-
tions, namely the mass, momentum, and energy equations. Then the equations for the
transport of turbulent kinetic energy (k) and dissipation rate (ϵ) are solved, based on the
mean flow variables. The eddy viscosity is determined based on k and ϵ. Then this eddy vis-
cosity is combined with the molecular viscosity in the mean momentum equations, yielding
the effective stresses due to molecular diffusion and turbulence transport. In a similar way,
the molecular diffusivity is replaced by the sum of molecular and turbulent heat diffusivities
in the energy equation.
In the k–ϵ model, two additional partial differential equations for k and ϵ must be solved.
Since the timescales associated with turbulent mixing are much smaller than corresponding
scales associated with the mean flow, careful attention is required in the iterative procedure
for the coupled mean flow and turbulence equations. A common approach is to first calcu-
late the mean flow quantities based on the eddy viscosity and diffusivity from a previous
iteration or time step. Then the k–ϵ equations are solved based on the computed mean
flow quantities. Nonlinear terms in the k–ϵ equations must be linearized in a suitable man-
ner. They can be linearized similarly to the procedure used in the mean flow equations. Iter-
ations between the mean flow and turbulence equations are performed until satisfactory
convergence between the equations is achieved.
In turbulent flow simulations, convergence may be difficult to achieve since numerical
instability can arise from various factors, leading to negative (nonphysical) values of k or
ϵ. As a result, under-relaxation is often used during iterations in these equations. Instead
of using the updated velocity in the next iteration, a combination of this value and the pre-
vious estimate are used. Also, the model must be analytically and computationally realiz-
able. This means that positive definite quantities, such as the turbulent kinetic energy,
approach zero in an asymptotic manner. Special numerical treatment of positive definite
quantities is required to ensure physically realistic and realizable results.
Although a finer grid should be used for the turbulence quantities (compared to the mean
flow) due to smaller turbulence length scales, this approach is usually not practical. Instead
an alternative upwinding approach in the convection terms for k and ϵ is used. For example,
local blending of a central difference scheme with a low-order upwind discretization is used.
This approach leads to certain discretization errors for k and ϵ but it may be necessary for
computations of different flow scales to be performed on a single grid.
Boundary equations are also required for the turbulence equations. At a wall, experimen-
tally obtained near-wall velocity and/or shear stress profiles, such as the law of the wall
(Chapter 3), are specified rather than highly refining the grid near the wall. At a high Rey-
nolds number, the viscous sublayer of the boundary layer is very thin so it is impractical to
use enough grid points to full resolve the flow behavior. Also, the turbulent kinetic energy, k,
454 Advanced Heat Transfer
and derivative of the dissipation rate, ϵ, perpendicular to the wall, are set to zero at the wall.
At an inflow boundary, k and ϵ are often unknown. Experimental measurements can be used
if available. Otherwise, k and ϵ can be selected based on a scale analysis of other relevant
length and velocity scales.
where the shape functions and their derivatives for quadrilateral elements were defined
earlier in Equations 10.84 through 10.87. The uppercase notation (Φ) refers to nodal
variables, whereas the lowercase notation (φ) will represent values internally within
an element.
For two-dimensional flows, the same form of governing transport equation and integrated
control volume equation are obtained as the previous section, Equations 10.155 and 10.158,
∂(ρϕ) ∂(ρuϕ) ∂(ρvϕ) ∂ ∂ϕ ∂ ∂ϕ
+ + = Γϕ + Γϕ + Sϕ (10.191)
∂t ∂x ∂y ∂x ∂x ∂y ∂y
t+Δt t+Δt t+Δt t+Δt
∂(ρϕ)
dV dt + ρvϕ · n dS dt = Γϕ ∇ϕ · n dS dt + Sϕ dV dt
t Vi ∂t t Si t Si t Vi
(10.192)
Computational Heat Transfer 455
t+Δt
∂(ρϕ)
dV dt = J1 ρΦt+Δt − ρΦt1 (10.193)
SCV1 ∂t
1
t
where Φ denotes a nodal value, the subscript 1 refers to local node 1, and J1 represents the
Jacobian determinant (area of SCV1) from Equation 10.91. Upon the assembly of all finite
elements, the SCV terms from other elements sharing the same node are added to this
transient term to form the full integrated transient term over the full control volume.
Methods such as UDS or EDS may be used to determine integration point values and
their convection fluxes. Alternatively, the method of integration point equations is a robust
method to determine the convection fluxes along the SCV boundaries of a control vol-
ume. Discretized approximations of the transport equation, Equation 10.191, are formed
at the integration points of each SCV and then solved locally within each element to
determine integration point values in terms of nodal quantities. Using these integration
point variables, the convection fluxes can then be determined. It should be noted that the
integration point equations are pointwise approximations that are discretized locally at
each integration point, unlike the control volume equations which are integrated over
finite volumes.
The pointwise discretization of Equation 10.191 at each integration point again involves
transient, convection, diffusion, and source terms. The transient term at an integration
point is approximated by a backward difference time, for example, at integration point 1
(subscript ip1),
∂ϕ ϕt+Δt
ip1 − ϕip1
t
= (10.194)
∂t ip1 Δt
SS1 Control
t t = 0 axis volume
SS2
s
SS3
s = 0 axis SS4
2e 1e Element
ip1 SCV2 SCV1
ip4
FIGURE 10.6
Schematic for a control volume-based finite element method.
456 Advanced Heat Transfer
a 0 a1
ϕu = Φ2 + 1 − Φ3 (10.196)
b b
where pi refers to the pressure at local node i. In a similar manner, the pressure gradient in
the y-direction is approximated in terms of bilinear interpolation using derivatives of the
shape functions in the y-direction.
The diffusion (Laplacian) operator is approximated by a central difference,
⎛ ⎞
2
∂ u ∂2 u 1 ⎝ 4
+ = Nj Φj − ϕip1 ⎠ (10.198)
∂x2 ∂y2 ip1 L2d j=1
where Ld is a diffusion length scale. For integration point 1, it can be shown that:
−1
2 8
L2d = + (10.199)
Δx2 3Δy2
Local source terms, such as thermal buoyancy using the Boussinesq approximation
(Chapter 3), can be evaluated by either direct substitution of corresponding integration
point values or interpolation of nodal values using the shape functions.
After all integration point operators are assembled back into Equations 10.191 at the inte-
gration point, a local matrix inversion can provide the integration point values explicitly in
terms of nodal variables alone. These inverted matrices are called influence coefficient matrices
since the individual coefficients express the relative contributions of convection, diffusion,
pressure, and source terms on the integration point scalar variable. The matrices provide
a direct coupling between integration point and nodal point values in order to calculate
Computational Heat Transfer 457
the diffusion fluxes for the control volume equations. Further details of the formulation of
influence coefficient matrices for the integration point equations were presented by
Schneider and Raw (1987).
The diffusion term in Equation 10.192 must be evaluated at both SS1 and SS4 surfaces
corresponding to SCV1. At SS1,
t+Δt
∂ϕ ∂ϕ
Γϕ ∇ϕ · n dS dt = Γϕ ΔtΔy1 − Γϕ ΔtΔx1 (10.200)
t Si ∂n ip1 ∂n ip1
where a midpoint approximation has been used. The diffusion flux is related to the scalar
variable, φ, through a suitable phenomenological law, such as Fourier’s law for the
energy equation.
For the source term in Equation 10.192, a lumped approximation is used,
t+Δt
Sϕ dV dt = J1 ΔtS|(1/2,1/2) (10.201)
t SCV1
where the (1/2, 1/2) subscript refers to the local coordinate position in the center of SCV1.
where Q refers to the rate of heat flow across the subsurface and the subscripts e1,1 and e2,1
refer to external elements that also contribute their heat flows to the energy balance for
SCV1. Source terms have been neglected.
The heat flows within an element, such as Q2,1 (from SCV2 into SCV1) at SS1, are com-
puted as follows,
1
∂T ∂T
QSS1 = ( − k∇T) · n dS = − k dy − dx dt (10.203)
SS1 0 ∂x ∂y s=0
Along the path of integration for SS1, the differentials dx and dy may be written in terms
of a single local coordinate, t. Using the chain rule, ds vanishes during the integration since
s = 0 along SS1. Substituting those expressions for the differentials and utilizing the shape
functions for the evaluation of the temperature derivatives,
& '
4 1
∂Ni ∂y ∂Ni ∂x
Q2,1 =− k −k dt Ti (10.204)
i=1 0 ∂x ∂t ∂y ∂t s=0
458 Advanced Heat Transfer
Similar expressions are obtained for the other heat conduction terms in the energy balance.
It should be noted that the remaining heat flows satisfy Qi,j = –Qj,i for i ≠ j since the heat
entering a given SCV is equivalent to the heat flow leaving the adjacent SCV.
Using a lumped capacitance approximation, the transient term is written as:
t+Δt
∂T T1 − T1t
ρcp = ρcp J1 (10.205)
∂t Δt
where the superscripts t + Δt and t refer to current and previous time levels, respectively.
Also, J1 refers to the Jacobian determinant of SCV1.
Then the energy balance for SCV1 can be written in the following form:
4
e
k1,j Tje = re1 (10.206)
j=1
where the local (elemental) stiffness matrix coefficients and right-side vector for SCV1 can be
written as:
1 1
∂Nj ∂y ∂Nj ∂x ∂Nj ∂y ∂Nj ∂x 1
e
k1,j = k −k dt − k −k ds + ρcp J1 (10.207)
0 ∂x ∂t ∂y ∂t s=0 0 ∂x ∂s ∂y ∂s t=0 Δt
1
re1 = ρcp J1 T1t (10.208)
Δt
Similar terms are constructed for the other SCVs such that the entire 4 × 4 stiffness coef-
ficient matrix can be assembled and completed. The previous equations have been con-
structed locally within the element so that the subscripts refer to local nodes. For
example, T1 refers to the temperature at local node 1 within the current element. Schneider
and Zedan (1982) analyzed this CVFEM formulation for heat conduction problems.
In a computer program, the local stiffness coefficient matrix can be programmed as a dou-
ble loop over i = 1, 2, 3, 4 and j = 1, 2, 3, 4. The first loop is indicated by the range of the sub-
script j in Equation 10.207. The second loop would cover the four SCVs within the entire
element. The heat conduction terms from external elements, Qe1,1 and Qe2,1 in Equation
10.202, are not required in the stiffness coefficient in Equation 10.207, because those terms
mutually self-cancel each other after all elements in the mesh are assembled. In other words,
the heat flow entering SCV1 from e1 balances the heat flow leaving the adjacent SCV from
the external element. Thus both terms self-cancel each other. An exception is elements
located along the external boundaries of the domain, where external flows or boundary tem-
peratures must be specified through appropriate boundary conditions.
Once the elemental stiffness matrix is completed, a standard finite element assembly pro-
cedure is performed for all finite elements in the mesh. The assembly process ensures that
energy conservation is obtained over the full control volume rather than individual SCVs
within an element. The full control volume consists of all SCVs associated with a particular
node in the mesh. Through this process, the geometric flexibility of the finite element
method is retained when the governing equations are formed locally within an element,
independent of the mesh configuration and layout.
Computational Heat Transfer 459
where the subscript AD refers to acceptor–donor. The interface orientation inside a cell is
important because it contains the free surface. The VOF method assumes that the boundary
can be approximated by a straight line passing through the cell. By determining the line’s
shape, it can be moved across the cell to a position such that it intersects the cell’s edges
to provide a known amount of fluid volume in the cell.
The free surface can be represented as a function Y(x) or X(y) after taking the weighted
average in terms of Fi,j of three adjacent cell boundary positions. If the magnitude of dY/dx
u u u
D A D A D A
Fj+1 Δyj+1
Fj Δyj
Interface orientation
within a discregte cell
Fj–1 Δyj–1
FIGURE 10.7
Schematic of the VOF method.
460 Advanced Heat Transfer
is larger than the magnitude of dX/dy, the surface is more horizontal than vertical. If
dX/dy , 0, fluid lies below the surface. With the slope and its sign, a line can be constructed
in the cell with a correct amount of fluid beneath it (see Figure 10.7). Then local surface ten-
sion forces that affect the dynamics of the free surface movement can be estimated based on
the Y(x) and X(y) profiles.
The motion of F (void movement) can be predicted based on the solution of an appropriate
transport equation. For two-dimensional flows, the following transient advection equation
can be used:
∂F ∂F ∂F
+u +v =0 (10.211)
∂t ∂x ∂y
This equation states that the transient accumulation of F in a control volume balances its
net inflow across the cell surface.
Appropriate boundary conditions must be applied along the free surface, such as matching
conditions of shear stress, temperature, and velocity. Also, surface tension forces act on a fluid–
gas interface. In free surface flows, the position of the moving interface is often unknown. As a
result, iterative or matching procedures from both sides of the free surface are required to track
the interface movement. The following example illustrates how the VOF method can be used to
find the movement of a free surface between donor and acceptor cells.
The cell spacing and normal velocity at the face are denoted by Δx and u, respectively.
Calculate the change of F advected across the cell face between donor and acceptor cells
over a time step of Δt for each of the three cases.
1. In the first case, the cells have FD = FAD = 1/4 and uΔt = xD/4. In this case, it is expected
that ΔF = xD/16 = uΔt/4, which is less than the maximum permissible amount that
could be convected into A. Equations 10.209 and 10.210 verify these trends,
If the donor cell is almost empty, then the Min condition in Equation 10.208 prevents
the advection of more fluid from the donor cell than it has to give. Similarly, the Max
feature of Equation 10.210 adds an additional flux, CF, which is available through
advection in the time step Δt.
2. In this case,
3
ΔF = FD ΔxD = ΔxD (10.214)
25
Computational Heat Transfer 461
Using the VOF method, the amount advected into the acceptor cell is computed as
follows:
$ %
19 3
CF = Max 1 − uΔt − 1 − ΔxD , 0 = 0 (10.215)
25 25
It can be observed that cell D advects all of its fluid into cell A.
3. In the third case,
22
ΔF = FD ΔxD = ΔxD (10.217)
25
In a way similar to the first case, the Max feature of Equation 10.210 adds the addi-
tional flux, CF, which is available due to advection over the time step of Δt.
The results showed a high accuracy and exponential convergence of results, even though a
relatively small number of nodes were used in the simulations.
In the boundary element method, the external boundary of the problem domain is subdi-
vided into a surface mesh. The method uses the boundary conditions to fit boundary values
into an integral equation rather than solving the differential equations within the domain. In
the post-processing stage, the integral equation can then be used to determine the solution at
any point in the interior of the domain. For more details, the reader is referred to a book of
Wrobel and Aliabadi (2002) and applications such as transient liquid sloshing in tanks
(Kolaei, Rakheha, and Richard 2015).
The boundary element method provides an exact solution of the differential equation(s) in
the domain. It is parameterized by a finite set of parameters along the boundary. It has sev-
eral advantages over other methods. Only the boundary of the domain needs to be discre-
tized, which allows simpler data input and storage. Also, the convergence rate is normally
high in the interior of the domain. However, the method also has disadvantages, as it
requires an explicit knowledge of a solution of the differential equation, which is generally
only available for linear partial differential equations with constant coefficients. Also, error
estimates are limited since there are different boundary integral equations for any given
problem and each has several approximation methods. Furthermore, if the boundary is
not smooth, but has corners or abruptly changing edges, then the boundary integral solution
has singularities at the boundary.
The Direct Simulation Monte Carlo method (or DSMC method) uses simulation molecules to
represent a large number of real molecules in order to solve the Boltzmann equation (an
equation describing the statistical behavior of a system not in a state of equilibrium). Mol-
ecules are moved throughout the physical domain in a manner that is coupled with the
appropriate time and length scales. Molecular movement and collisions are decoupled
over time periods less than a mean collision time. Collisions among the molecules and
between molecules and walls are calculated based on probabilistic models. Examples of
common collision models include the Hard Sphere Model (HSM), Variable Hard Sphere Model
(VHSM), and Variable Soft Sphere Model (VSSM). The DSMC method is commonly applied to
problems involving spacecraft reentry aerodynamics and micro- and nano-electromechan-
ical systems. It is a powerful method for the computation of complex, nonequilibrium gas
flows, such as hypersonic flows where nonequilibrium conditions occur at high altitude
and in regions of small length scales. Boyd (2015) presented a comprehensive overview of
the theoretical basis of the DSMC method.
For radiation problems, the method of discrete ordinates is often used to solve the radiation
equation of transfer (Chapter 4) by discretizing both the xyz-domain and angular variables
in the direction of radiation. A position and direction dependent radiation intensity function
is required in radiation problems. This intensity field can be solved by an integral–differen-
tial radiation equation of transfer. However, analytical solutions are usually not available
even in simple geometrical configurations. The method of discrete ordinates provides an
approximate solution by discretizing both the xyz-domain and angular velocities represent-
ing the direction, for example, Kamden (2015).
A high resolution method can be used for a numerical solution of differential equations
where high accuracy is required to resolve small-scale phenomena such as shock waves
and discontinuities. Second-order or higher accuracy is used to prevent spurious oscillations
in the solutions. These methods use the concept of flux limiters, or slope limiting functions
that have the effect of limiting the solution gradient near shock waves or discontinuities.
In high-resolution methods, the number of mesh points containing a discontinuity is usually
small compared with a first-order scheme of similar accuracy. Shu (2009) presented a review
Computational Heat Transfer 463
of high resolution schemes for convection dominated flows and applications in computa-
tional fluid dynamics.
PROBLEMS
10.1 Annular tapered fins with a thickness of c/r2 (where c is a constant) are designed
to enhance the rate of heat transfer from air-cooled tubes of a combustion
cylinder. One-dimensional, steady-state heat transfer occurs from the base (r = a)
to the tip (r = b) of each fin in the radial (r) direction. Ambient air flows past the
fins with a temperature and convection coefficient of T∞ and h, respectively, and
the temperature at the base of the fin is Tb.
a. Derive the finite difference equations for the interior and boundary nodes of
the domain. Assume a constant thermal conductivity, k, within the fin.
b. Explain how the numerical solution procedure in part (a) is modified to
include variations of thermal conductivity with temperature.
10.2 A plane wall of 4 cm width is cooled along its right boundary by a fluid at 20◦ C and a
convection coefficient of h = 800 W/m2 K. The left boundary is maintained at 160◦ C
and the initial wall temperature is 60◦ C. The thermal conductivity and diffusivity
are 50 W/m K and 10−5 m2/s, respectively. Using a five-node discretization, one-
dimensional domain, and explicit finite difference method, calculate the steady-state
temperatures at the midplane and right boundary. How would the trends change if
2Fo (1 + Bi) . 1, where Fo and Bi refer to the Fourier and Biot numbers, respectively?
10.3 The top edge of a silicon chip in an electronic assembly is convectively cooled by a
fluid at 20◦ C with a convection coefficient of 600 W/m2 K. A two-dimensional
explicit finite difference method (Δx = 6 mm, Δy = 5 mm) is used to predict the tran-
sient temperature variation within the chip (ρcp = 1,659 kJ/m3 K, k = 84 W/mK). The
rate of internal heat generation within the chip is 108 W/m3.
a. What minimum time step is required to ensure stable time advance for a
node on the top boundary? Determine the stability criterion by deriving
the relevant finite difference equation on the top boundary.
b. Would the time step constraint be more restrictive for a fixed heat flux boun-
dary condition? Explain your response.
10.4 Consider a turbulent flow and decay of turbulent kinetic energy as described by the
following k – ϵ equations.
∂k
= −ε
∂t
∂ε ε2
= −cε2
∂t k
Both the turbulent kinetic energy (k) and dissipation rate (ϵ) are positive definite
quantities. Develop an explicit finite difference formulation which ensures that solu-
tions of these equations remain positive.
10.5 A weighted residual method uses Wi = P at the nodal points (where P is a large num-
ber) and Wi = 0 away from the nodal points. Describe how this selection influences
the residual and error distributions throughout the problem domain.
464 Advanced Heat Transfer
10.6 State three significant advantages of the finite element method over other types of
numerical methods for structured grids.
10.7 Consider the interpolation of a scalar variable and its derivatives in a triangular
element. A triangular element with nodal points at (0.14, 0.01), (0.22, 0.05), and
(0.14, 0.15) has scalar values of φ = 160, 140, and 180 at local nodes 1, 2, and 3 in
the element, respectively. Calculate the value of φ and its derivatives within the
element and point (0.2, 0.05). At what point along the bottom side (1–2) of the ele-
ment does the φ = 150 contour intersect that side?
10.8 Are the shape functions linear along the edge of a four-node quadrilateral element?
Explain your response.
10.9 Describe the differences between a residual solution error and the order of accuracy
in the context of finite element analysis.
10.10 Can the elemental boundary integral of Ni h Nj over the elemental surface, S, be a full
(entirely nonzero) 3 × 3 matrix for triangular elements in linear and steady diffusion
problems? Explain your response.
10.11 Explain how over-relaxation and under-relaxation models affect the solution perfor-
mance and accuracy in finite element analysis of heat and fluid flow.
10.12 Explain how unstructured grid capabilities give a key advantage of finite element
methods in comparison to other conventional numerical methods for structured grids.
10.13 Under what circumstances should Gaussian quadrature with quadrilateral elements
be used in the weighted residual method?
10.14 A specified velocity profile is provided at the inlet of a diffuser. The outflow con-
ditions are unknown. What boundary location and conditions at the outlet would
be suitable for internal flow simulations?
10.15 Determine the requirement of the shape functions to ensure that a scalar, φ, is cons-
tant throughout a finite element.
10.16 Consider the following differential equation subject to two boundary conditions:
d2 u
+4=0
dx2
where i refers to the complex imaginary number. Find the discrete equations for
this one-dimensional heat conduction problem using weight functions of Wn = xn,
where n = 0, 1, and 2.
10.18 A composite wall consists of the following three different layers: 9 cm thick Section 1
(k1 = 5 W/mK), 6 cm thick Section 2 (k2 = 0.8 W/mK), and 7 cm thick Section 3 (k3 =
16 W/mK). The left and right boundaries are maintained at T = 400◦ C and T =-
100◦ C, respectively. Use a one-dimensional finite element method to determine
the interface temperatures and heat flux through the third (right) section.
10.19 Heat transfer in a circular fin is governed by the following fin equation:
d2 T
kA − hP(T − Tf ) = 0
dx2
where h = 1 W/cm2 K with a fluid flowing past the fin at a temperature of Tf = 12◦ C.
The thermal conductivity of the fin is 3 W/cmK. The fin diameter and length are
1 and 4 cm, respectively. The base temperature of the fin is maintained at 60◦ C.
Use a Galerkin finite element method with four equal one-dimensional elements
to find the nodal temperature distribution within the fin.
10.20 A nuclear fuel element of thickness 2L is covered with a steel cladding of
thickness a. Heat is generated within the nuclear fuel at a rate of q̇ and removed
from the left surface by a fluid at T∞ = 90◦ C with a convection coefficient of
h = 6,000 W/m2 K. The right surface is well insulated. The thermal conducti-
vities of the fuel and steel are kf = 60 W/mK and ks = 64 W/mK, respectively.
Use the finite element method with one-dimensional elements to find the tem-
peratures at the external boundaries, steel–fuel interface and fuel element
centerline.
10.21 The cross-section of a metal component is triangular. The left and lower boundaries
are insulated while the other surface is exposed to a convection condition with h =
5 W/m2 K. A uniform internal heat source generates 2 kW/m3 within the element.
Use a single triangular element to estimate the nodal temperature distribution in
terms of the positions of the nodal points.
10.22 Consider two adjacent linear quadrilateral elements. A local coordinate system (s, t)
is defined within each element in order to interpolate values of the problem variable
φ. Evaluate φ along the common edge from both elements. Determine whether there
is a unique single distribution for both cases.
10.23 Consider two adjacent linear triangular elements along the external boundary of a
problem domain. An insulated (zero gradient, ∂T/∂y = 0) boundary condition is
applied along the upper horizontal boundary.
a. Use isoparametric shape functions for triangular elements to find the
boundary temperature at node 3 that satisfies the insulated boundary
requirement. The temperatures at nodes 1, 2, and 4 are 170◦ C, 120◦ C, and
210◦ C, respectively.
b. Determine the x–y coordinates where the contour line for the temperature
at node 3 in part (a) intersects the element boundary alongside 1–2 in
element 1.
c. Use the linear shape functions to check that this intersection point lies on
side 1–2 in element 1.
466 Advanced Heat Transfer
d. Use the linear shape functions to check that the temperature at each
point along the contour line is constant and equal to the value obtained at
node 3.
10.24 A liquid layer of constant thickness flows under gravity down an inclined surface in
a chemical purification process. Assume no fluid velocity component perpendicular
to the plate and steady-state conditions. The film flow is governed by a balance
between frictional and gravitational forces, as follows,
d2 u
μ + ρgsin(α) = 0
dy2
Also assume that air resistance is negligible and the shear stress at the free surface is
zero, μ(du/dy) = 0 on the free surface.
a. Derive a finite element solution for the film velocity, u = u( y). Use three lin-
ear (one-dimensional) elements.
b. Find the exact solution and compare it with the above computed solution
from part (a).
c. How can closer agreement be achieved between the computed and exact
solutions?
10.25 Consider a thin wide plate with an externally applied heat flux at the top boundary
and an insulated bottom surface. The initial plate temperature is 0◦ C. Using an
explicit lumped method with a time step size of 1 s, determine whether the
resulting model is numerically stable. Explain whether numerical oscillations will
occur in the simulations.
10.26 The purpose of this question is to write four computer subroutines (STIFF, ASSMBL,
SHAPE, and BNDRY) to solve the following heat equation using the Galerkin
weighted residual method and triangular finite elements:
∂2 T ∂2 T
k + k =0
∂x2 ∂y2
Write a subroutine, SHAPE, to evaluate the shape functions and their derivatives
for a linear, quadrilateral, isoparametric finite element.
∙ Subroutine SHAPE (nnp, nnpe, nel, ie, e, x, y, s, t, dn, dx, dy, jac, dqdx, dqdy)
The shape functions should possess the following features: (i) Ni = 1 at local node i
and 0 otherwise; (ii) Ni = 0 on the sides opposite to node i. Use an example of a sam-
ple element to verify these features in the program. Mesh data should be imported
from a file (sample format shown below).
∙ 9, 8, 8 (number of nodal points, elements, and boundary surfaces)
∙ 5, 2, 1, 4 (counterclockwise listing of nodes in the first element, stored within
the ie global-local mapping array)
∙ … (Other elements)
∙ 0, 0 (x and y global coordinates of the first node, stored in the x, y arrays)
∙ … (Other nodes)
Write two subroutines (called STIFF and ASSMBL) that find the element pro-
perties for the two-dimensional field equation (stiffness matrix and right side).
Then perform the assembly of elemental stiffness matrices.
∙ Subroutine STIFF (e, nel, nnp, nnpe, x, y, lc, s, t, o, ie, dn, cond, flux, aq, rq)
For each element, e, an ASSMBL subroutine should call STIFF to obtain the entries
of the stiffness matrix. Each shape function is obtained from the SHAPE subroutine.
Consider an example problem of diffusion transport (G = 0 = Q) and the follow-
ing boundary conditions: (i) temperature-specified and (ii) flux-specified conditions.
468 Advanced Heat Transfer
∂ϕ
A + Bϕ = C
∂n
∙ Subroutine BNDRY (ie, o, nel, nnpe, nnp, nsrf, bgn, bel, x, y, bc, c, r)
The boundary condition data should be imported from a file (sample format
shown below).
∙ 2, 1 (pair of global nodes on the side of the first boundary element, stored in the
bgn array)
∙ … (Other boundary nodes and elements)
∙ 1 (element number corresponding to the first boundary element, stored in the
bel array)
∙ … (Other boundary elements)
∙ 0, 1, 10 (A, B, C values for the first node of the first element, stored in the bc
array)
∙ … (Other boundary nodes and elements)
Consider a sample problem of steady-state heat conduction in a rectangular
domain. The top and bottom boundaries are insulated. The left and right boundaries
have specified heat flux and temperature conditions, respectively. Find the internal
nodal temperature distribution. A banded solver can be used to solve the resulting
system of algebraic equations. A grid refinement should be performed to demon-
strate grid independence of the results. Problem parameters, such as conductivity,
should be imported from a separate project data file.
Also select another heat transfer problem of practical or industrial relevance and
prepare a numerical solution with the finite element program. The presentation
should include a brief description of the finite element model and a discussion of
results. Provide the source program, input files and numerical results for both val-
idation and application problems. For the application problem, provide additional
background information, problem parameters, boundary conditions, and a discus-
sion of solution errors.
Note: e = element number; nel = number of elements; nnp = number of nodal
points; nnpe = number of nodes per element; x, y(nnp) = nodal x and y values; ie
(nel, nnpe) = local–global node bookkeeping array; aq(nnpe, nnpe) = element stiff-
ness matrix; c(nnp, nnp) = global stiffness matrix; rq(nnpe) = element right side; r
(nnp) = global right side; dn = shape functions and derivatives; isemi = semi-band-
width; cond = conductivity; s, t = local coordinates; dx, dy = Δx and Δy; jac =
Jacobian; dqdx, dqdy = scalar derivatives; lc = integration point local coordinates;
o = orientation array for ip and SVCV numbers; flux = diffusion coefficients; nsrf =
number of boundary surface elements; bgn = boundary global nodes; bel =
boundary elements; bc = boundary condition coefficients (A, B, C).
10.28 Consider a project to develop and apply a two-dimensional CVFEM code to a
selected fluid flow problem. A general scalar quantity, φ, such as energy or concen-
tration of a pollutant in an airstream, is transported through a flow field by diffusion
and convection processes. The purpose of this project is to write two computer
Computational Heat Transfer 469
subroutines (IPOINT and STIFF) to solve the following governing differential and
finite volume equations with a CVFEM for two-dimensional steady flows.
∂(ρuϕ) ∂(ρvϕ) ∂ ∂ϕ ∂ ∂ϕ
+ = Γϕ + Γϕ + Sϕ
∂x ∂y ∂x ∂x ∂y ∂y
(ρvϕ) · n dS = (Γϕ ∇ϕ) · n dS + Sϕ dA
S S A
where v and n refer to the velocity and unit normal vector at the surface,
respectively.
To determine the integration point values of φ in terms of nodal quantities, Φi,
consider the one-dimensional convection–diffusion equation at an integration point,
i + 1/2, in between nodal points i and i + 1. Develop an algorithm to determine the
integration point value, φ i+1/2, in terms of nodal values, Φi and Φi+1, using the fol-
lowing schemes:
∙ CDS (Central Differencing Scheme): φi+1/2 = 0.5 Φi + 0.5 Φi+1.
∙ UDS (Upwind Differencing Scheme): φi+1/2 = Φi (flow from left to right).
∙ EDS (Exponential Differencing Scheme): φi+1/2 = ((1 + α)/2) Φi + ((1 −α)/2)
Φi+1, where α ∼ Pe 2/(5 + Pe 2) and Pe = ρuiΔxi/Γ (Peclet number).
∙ PINS (Physical INfluence Scheme): φi+1/2 dependence on Φi and Φi+1 obtained
by a local approximation of the integration point equation at i + 1//2.
The relative influences of upstream and downstream nodal values depend on
the Peclet number, Pe, and influence coefficients which premultiply each nodal
value. Write a computer subroutine (IPOINT) to determine the two-dimensional
influence coefficient array, called ic. Verify that the influence coefficients approach
CDS and UDS for low- and high-mass flow rates, respectively, and that the proper
influences in the x- and y-directions are achieved.
Apply the IPOINT and other subroutines developed in previous questions (STIFF,
ASSMBL, SHAPE, and BNDRY) to solve the problem of two-dimensional flow over
a step in a channel. The velocity components may be specified in the IPOINT sub-
routine as uniform throughout the flow field for this test case. Apply a specified φ
boundary condition at the inlet and verify that this upstream boundary condition
is propagated over time through the channel. For high Pe values, verify that convec-
tion influences are dominant in the results. Conversely, for low Pe values (including
the limit of zero flow), check that diffusion causes uniform transport of φ in both
upstream and downstream directions.
The presentation should include a brief description of the numerical model, prob-
lem parameters, boundary conditions, and a discussion of the results and solution
errors. Include the source program as well as input files and numerical results for
the step flow problem as well as validation against past reported data in the
archival literature.
10.29 For the numerical solution in the previous question, describe the sources of error
that arise due to conventional upwinding (such as UDS) for time-dependent con-
vection. Suggest alternatives to overcome these sources of errors. Explain the
response by reference to a specific example, such as a one-dimensional five-node
discretization with a step function of φ = 1 convected into the domain over time.
470 Advanced Heat Transfer
References
I.D. Boyd. 2015. “Computation of Hypersonic Flows Using the Direct Simulation Monte Carlo
Method,” AIAA Journal of Spacecraft and Rockets, 52, pp. 38–53
W. Cheney and D. Kincaid. 1985. Numerical Mathematics and Computing, 2nd Edition, Brooks/Cole
Publishing Company, Pacific Grove, CA.
C.W. Hirt and B.D. Nichols. 1981. “Volume of Fluid (VOF) Method for the Dynamics of Free Bound-
aries,” Journal of Computational Physics, 39: pp. 201–225.
H.T. Kamden. 2015. “Ray Effects Elimination in Discrete Ordinates and Finite Volume Methods,”
AIAA Journal of Thermophysics and Heat Transfer, 29: pp. 306–318.
A. Kolaei, S. Rakheja, and M.J. Richard. 2015. “A Coupled Multimodal and Boundary Element Method
for Analysis of Anti-slosh Effectiveness of Partial Baffles in a Partly-filled Container,” Computers
& Fluids, 107: pp. 43–58.
W.J. Minkowycz, E.M. Sparrow, and J.Y. Murthy. 1988. Handbook of Numerical Heat Transfer, New York:
John Wiley & Sons.
M.N. Ozisik, H.R. Orlande, M.J. Colaco, and R.M. Cotta. 2017. Finite Difference Methods in Heat Transfer,
2nd Edition, Boca Raton: CRC Press/Taylor & Francis.
S.V. Patankar. 1980. Numerical Heat Transfer, Washington, D.C.: Hemisphere.
J.N. Reddy and D.K. Gartling. 2010. Finite Element Method for Heat Transfer and Fluid Dynamics, 3rd Edi-
tion, Boca Raton: CRC Press/Taylor & Francis.
G.E. Schneider and M.J. Raw. 1987. “Control-Volume Finite Element Method for Heat Transfer and
Fluid Flow using Co-located Variables: 1. Computational Procedure,” Numerical Heat Transfer,
11: pp. 363–390.
G.E. Schneider and M. Zedan. 1982. “Control Volume Based Finite Element Formulation of the Heat
Conduction Equation,” AIAA Paper 82–0909, IAA/ASME 3rd Joint Thermophysics, Fluids,
Plasma and Heat Transfer Conference, St. Louis, MO.
C.W. Shu. 2009. “High Order Weighted Essentially Non-oscillatory Schemes for Convection Domi-
nated Problems,” SIAM Review, 51(1): pp. 82–126.
Y.S. Sun and B.W. Li. 2010. “Spectral Collocation Method for Transient Conduction–Radiation Heat
Transfer,” AIAA Journal of Thermophysics and Heat Transfer, 24(4): pp. 823–832.
J.P. Van Doormaal and G.D. Raithby. 1984. “Enhancements of the SIMPLE Method for Predicting
Incompressible Fluid Flows,” Numerical Heat Transfer, 7: pp. 147–163.
L.C. Wrobel and M.H. Aliabadi. 2002. The Boundary Element Method, New York: John Wiley & Sons.
M. Zeng and W.Q. Tao. 2003. “Comparison Study of the Convergence Characteristics and Robustness
for Four Variants of SIMPLE Family at Fine Grids,” Engineering Computations, 20: pp. 320–340.
Appendices
A:B = a11 b11 + a12 b12 + a21 b21 + a22 b22 (A.3)
Tensor notation (or indicial notation) provides a concise method of representing lengthy or
complicated equations in fluid mechanics, heat transfer, and other fields. Tensors represent
a generalized notation for scalars (a tensor of rank 0), vectors (rank of 1), matrices (rank of 2),
etc. A tensor is denoted by a variable with appropriate subscripts. For example, ai and Aij
represent the previous vector and matrix, respectively, where the range of subscripts is i = 1,
2 and j = 1, 2.
The summation convention of tensors requires that repetition of an index in a term denotes
a summation with respect to that index over its range. For example, the dot product of two
vectors is represented as:
ai bi = a1 b1 + a2 b2 (A.4)
The range of the index is the set of integer values over their range, such as i = 1, 2 in the
above dot product. A dummy index refers to an index that is summed, whereas a free index is
not summed. The rank of the tensor is increased for each index that is not repeated. For
example, aij contains two nonrepeating indices and therefore it is a tensor of rank 2 (matrix).
Further details describing the operations of vectors and tensors can be found in advanced
calculus or continuum mechanics textbooks.
471
472 Appendices
TABLE B.1
Conversion of Units and Constants
Conversion Factors SI Imperial
Acceleration 1 m=s 2
¼4.252 107 ft=hr2 ¼3.2808 ft=s2
Area 1 m2 ¼1,550.0 in.2 ¼10.764 ft2
Density 1 kg=m 3
¼0.06243 lbm=ft
Dynamic viscosity 1 kg=m · s ¼2,419.1 lbm=ft · h
Energy 1 kJ ¼0.9478 Btu ¼737.56 ft · lbf
Force 1N ¼0.22481 lbf
Heat flux 1 W=m2 ¼0.3171 Btu=h · ft2
Heat transfer coefficient 1 W=m K 2
¼0.1761 Btu=ft2 h F
Heat transfer rate 1W ¼3.4123 Btu=h ¼1.341 103 hp
Kinematic viscosity 1 m2=s ¼10.7636 ft2=s
Latent heat 1 kJ=kg ¼0.4299 Btu=lbm
Length 1m ¼39.37 in. ¼3.2808 ft
Mass 1 kg ¼2.2046 lbm ¼1.1023 103 U.S. tons
Mass diffusivity 1 m2=s ¼10.7636 ft2=s ¼3.875 104 ft2=h
Mass flow rate 1 kg=s ¼7,936.6 lbm=h
Mass transfer coefficient 1 m=s ¼1.181 104 ft=h
Pressure, stress 1 Pa ¼1.4504 104 lbf=in.2
Specific heat 1 kJ=kgK ¼0.2388 Btu=lbm F ¼0.2389 cal=g C
Temperature K ¼ C þ 273.15 ¼5=9( Fþ=459.67)
R ¼ F þ 459.67 ¼(9=5)( K)
Temperature difference 1K ¼1 C ¼(9=5) F
Thermal conductivity 1 W=mK ¼0.57782 Btu=h · ft · F
Thermal diffusivity 1 m2=s ¼10.7636 ft2=s ¼3.875 104 ft2=h
Thermal resistance 1 K=W ¼0.5275 F=h · Btu
Velocity 1 m=s ¼3.2808 ft=s ¼3.6 km=h
Volume 1m 3
¼264.17 gal (U.S.) ¼1,000 L
Volume flow rate 1 m3=s ¼1.585 104 gal=min ¼2,118.9 ft3=min
SI Unit Conversions
Prefix (Symbol) Multiplier
Tera (T) 1012
Giga (G) 109
Mega (M) 106
Kilo (k) 103
Milli (m) 103
Micro (μ) 106
Nano (n) 109
Pico (p) 1012
(Continued )
Appendices 473
qz + dz
T (x, y, z) qy + dy
dz
•
Eg
qx •
Est
qx + dx
z
y
x
qy dy
dx
qz
Mass Conservation
@ρ @ @ @
þ (ρu) þ (ρv) þ (ρw) ¼ 0 (C:1)
@t @x @y @z
qz + dz
rdθ
qr
qθ + dθ
dz
z
qθ
r qr + dr
T (r, θ, z)
dr
y
r
θ
qz
x
Schematic of cylindrical coordinates
Mass Conservation
@ρ 1 @ 1 @ @
þ (ρrvr ) þ (ρvθ ) þ (ρvz ) ¼ 0 (C:7)
@t r @r r @θ @z
(Continued )
Appendices 475
qφ + dφ
r sinφ dθ
qr qθ + dθ
z rdφ
φ qθ
T (r, θ, φ)
y dr qr + dr
θ r
x qφ
Mass Conservation
@ρ 1 @ 1 @ 1 @
þ (ρr2 vr ) þ (ρv sinϕ) þ (ρvθ ) ¼ 0 (C:13)
@t r2 @r r sin ϕ @ϕ ϕ r sin ϕ @θ
Conservation of Momentum (r, φ, and θ-directions)
2
!
Dvr vϕ v2θ @p 2vr 2 @vϕ 2vϕ cot ϕ 2 @vθ
ρ ¼ þ μ r2 vr 2 2 2 þ Fr (C:14)
Dt r r @r r r @ϕ r2 r sin ϕ @θ
(Continued )
476 Appendices
!
Dvϕ vr vϕ v2θ cotϕ 1 @p 2 @vr vϕ 2cosϕ @vθ
ρ þ ¼ þ μ r2 v ϕ þ 2 2 2 2 2 þ Fϕ (C:15)
Dt r r r @ϕ r @ϕ r sin ϕ r sin ϕ @θ
Dvθ vθ vr vϕ vθ cotϕ 1 @p v 2 @vr 2cosϕ @vϕ
ρ þ þ ¼ þ μ r2 vθ 2 θ 2 þ 2 þ 2 2 þ Fθ (C:16)
Dt r r rsinϕ @θ r sin ϕ r sinϕ @θ r sin ϕ @θ
where,
D @ @ vϕ @ v @
¼ þ vr þ þ θ (C:17)
Dt @t @r r @ϕ rsinϕ @θ
1 @ @ 1 @ @ 1 @2
r2 ¼ r2 þ 2 sin ϕ þ (C:18)
r @r
2 @r r sin ϕ @ϕ @ϕ r2 sin ϕ @θ2
2
Energy Equation
@T @T vϕ @T v @T
ρcp þ vr þ þ θ ¼
@t @r r @ϕ rsinϕ @θ
(C:19)
1 @ @T 1 @ @T 1 @2T
k 2 r2 þ 2 sinϕ þ 2 2 þ q_ þ μΦ
r @r @r r sinϕ @ϕ @ϕ r sin ϕ @θ2
" #
@vr 2 1 @vϕ vr 2 1 @vθ vr vϕ cotϕ 2 @ vϕ 1 @vr 2
Φ¼2 þ þ þ þ þ þ r þ
@r r @ϕ r rsinϕ @θ r r @r r r @ϕ
2 2
sinϕ @ vθ 1 @vθ 1 @vr @ vθ
þ þ þ þr (C:20)
r @ϕ rsinϕ rsinϕ @θ rsinϕ @θ @r r
TABLE D.1
Properties of Metals at STP (101 kPa, 25◦ C)
Boiling Thermal Specific Coefficient of Heat of
Melting Point Conductivity, Heat, cp Expansion, β Density, ρ Fusion, hsl
Metal Point ( C) ( C) k (W== mK) (kJ== kgK) (106= K) (kg== m3) (kJ== kg)
Aluminum 660 2,441 237.0 0.900 25 2,700 397.8
Antimony 630 1,440 18.5 0.209 9
Beryllium 1,285 2,475 218 1.825 12
Bismuth 271.4 1,660 8.4 0.126 13
Cadmium 321 767 93 0.230 30
Chromium 1,860 2,670 91 0.460 6 7,150 330.8
Cobalt 1,495 2,925 69 0.419 12 8,860 276.4
(Continued )
Appendices 477
TABLE D.2
Properties of Non-Metals and Phase Change Materials
Thermal Conductivity, Specific Heat,
Material Density, ρ (kg== m3) k (W== mK) cp (kJ== kgK)
Asbestos millboard 1,400 0.14 0.837
Asphalt 1,100 1.67
Brick, common 1,750 0.71 0.920
Brick, hard 2,000 1.3 1.00
Chalk 2,000 0.84 0.900
Charcoal, wood 400 0.088 1.00
(Continued )
478 Appendices
TABLE E.1
Properties of Air at Atmospheric Pressure
Temperature, Density, Specific Heat, Viscosity, μ Thermal Conductivity,
T (K) ρ (kg== m3) cp (kJ== kgK) (kg== ms) k (W== mK) Pr
6
150 2.367 1.010 10.28 10 0.014 0.758
200 1.769 1.006 13.28 106 0.018 0.739
6
250 1.413 1.005 15.99 10 0.022 0.722
260 1.359 1.005 16.50 106 0.023 0.719
270 1.308 1.006 17.00 106 0.024 0.716
275 1.285 1.006 17.26 106 0.024 0.715
6
280 1.261 1.006 17.50 10 0.025 0.713
290 1.218 1.006 17.98 106 0.025 0.710
300 1.177 1.006 18.46 106 0.026 0.708
310 1.139 1.007 18.93 106 0.027 0.705
6
320 1.103 1.007 19.39 10 0.028 0.703
330 1.070 1.008 19.85 106 0.029 0.701
340 1.038 1.008 20.30 106 0.029 0.699
350 1.008 1.009 20.75 106 0.030 0.697
6
400 0.882 1.014 22.86 10 0.034 0.689
450 0.784 1.021 24.85 106 0.037 0.684
500 0.706 1.030 26.70 106 0.040 0.680
550 0.642 1.040 28.48 106 0.044 0.680
6
600 0.588 1.051 30.17 10 0.047 0.680
700 0.504 1.075 33.32 106 0.052 0.684
800 0.441 1.099 36.24 106 0.058 0.689
900 0.392 1.121 38.97 106 0.063 0.696
6
1,000 0.353 1.142 41.53 10 0.068 0.702
Source: Data reprinted with permission from R.C. Weast, Ed., 1970. CRC Handbook of Tables for Applied
Engineering Science, Table 1-2, Boca Raton: CRC Press=Taylor & Francis.
480 Appendices
TABLE E.2
Properties of other gases at STP (110 kPa, 25oC)
Thermal Dynamic
Density, Specific Heat, Gas Constant Conductivity, Viscosity,
Gas ρ (kg== m3) cp (kJ== kgK) (J== kg C) k (W== mK) μ (kg== ms)
TABLE E.3
Properties of other Gases (Effects of Temperature)
Density, Specific Thermal
Temperature, ρ Heat, Conductivity, Dynamic Viscosity, μ
Gas T ( C) (kg== m3) cp (kJ== kgK) k (W== mK) (kg== ms)
TABLE E.4
Ideal Gas Properties of Selected Gases (Effects of Temperature)
Carbon Dioxide, CO2 Oxygen, O2 Nitrogen, N2
o ¼ 393,520 kJ== kmol
h o ¼ 0 kJ== kmol
h o ¼ 0 kJ== kmol
h
f f f
h, ,
u so ,
h, ,
u so ,
h, ,
u so ,
T(K) kJ== kmol kJ== kmol kJ== kmol K kJ== kmol kJ== kmol kJ== kmol K kJ== kmol kJ== kmol kJ== kmol K
0 0 0 0 0 0 0 0 0 0
250 7,627 5,548 207.337 7,275 5,197 199.885 7,266 5,188 186.370
300 9,431 6,939 213.915 8,736 6,242 205.213 8,723 6,229 191.682
350 11,351 8,439 219.831 10,213 7,303 209.765 10,180 7,270 196.173
400 13,372 10,046 225.225 11,711 8,384 213.765 11,640 8,314 200.071
(Continued )
484 Appendices
h, ,
u so ,
h, ,
u so ,
h, ,
u so ,
T(K) kJ== kmol kJ== kmol kJ== kmol K kJ== kmol kJ== kmol kJ== kmol K kJ== kmol kJ== kmol kJ== kmol K
450 15,483 11,742 230.194 13,228 9,487 217.342 13,105 9,363 203.523
500 17,678 13,521 234.814 14,770 10,614 220.589 14,581 10,423 206.630
550 19,945 15,372 239.135 16,338 11,765 223.576 16,064 11,492 209.461
600 22,280 17,291 243.199 17,929 12,940 226.346 17,563 12,574 212.066
650 24,674 19,270 247.032 19,544 14,140 228.932 19,075 13,671 214.489
700 27,125 21,305 250.663 21,184 15,364 231.358 20,604 14,784 216.756
750 29,629 23,393 254.117 22,844 16,607 233.649 22,149 15,913 218.889
800 32,179 25,527 257.408 24,523 17,872 235.810 23,714 17,061 220.907
850 34,773 27,706 260.551 26,218 19,150 237.864 25,292 18,224 222.822
900 37,405 29,922 263.559 27,928 20,445 239.823 26,890 19,407 224.647
950 40,070 32,171 266.444 29,652 21,754 241.689 28,501 20,603 226.389
1,000 42,769 34,455 269.215 31,389 23,075 243.471 30,129 21,815 228.057
1,100 48,258 39,112 274.445 34,889 25,753 246.818 33,426 24,280 231.199
1,200 53,848 43,871 279.307 38,447 28,469 249.906 36,777 26,799 234.115
1,300 59,522 48,713 283.847 42,033 31,224 252.776 40,170 29,361 236.831
1,400 65,271 63,631 288.106 45,648 34,008 255.454 43,605 31,964 239.375
1,500 71,078 58,606 292.114 49,202 36,821 257.965 47,073 34,601 241.768
1,600 76,944 63,741 295.901 52,961 39,658 260.333 50,571 37,268 244.028
1,700 82,856 68,721 299.428 56,652 42,517 262.571 54,099 39,965 246.166
1,800 88,806 73,840 302.884 60,371 45,405 264.701 57,651 42,685 248.195
1,900 94,793 78,996 306.122 64,116 48,319 266.722 61,220 45,423 250.128
2,000 100,804 84,185 309.210 67,881 51,253 268.655 64,810 48,181 251.969
2,100 106,864 89,404 312.160 71,668 54,208 270.504 68,417 50,957 253.726
2,200 112,939 94,648 314.988 75,484 57,192 272.278 72,040 53,749 255.412
2,300 119,035 99,912 317.695 79,316 60,193 273.981 75,676 56,553 257.027
2,400 125,152 105,197 320.302 83,174 63,219 275.625 79,320 59,366 258.580
2,500 131,290 110,504 322.808 87,057 66,271 277.207 82,981 62,195 260.073
2,600 137,449 115,832 325.222 90,956 69,339 278.738 86,650 65,033 261.512
Source: Adapted from D.R. Stull and H. Prophet, 1971. JANAF Thermochemical Tables, Tables CO2, O2 and N2, 2nd
Edition, U.S. National Bureau of Standards, Office of Standard Reference Data, Washington, DC.
TABLE E.5
Diffusion of Gases into Air (101 kPa)
Diffusion Coefficient, Diffusion Coefficient, Schmidt Number, Schmidt Number,
Substance D (m2= s) at 0 C D (m2= s) at 25 C Sc (ν== D) at 0 C Sc (ν== D) at 25 C
TABLE F.1
Properties of Liquids (300 K, 1 atm, 297 K)
Dynamic Thermal
Density, ρ Specific Heat, Viscosity, μ Conductivity, k Freezing
Liquid (kg== m3) cp (kJ== kgK) (kg== ms) (W== mK) Point (K)
TABLE F.2
Properties of Saturated Water
T ρf ρv hfg cρ,f μf · 106 kf σf
(oC) P (kPa) (kg== m3) (kg== m3) (kJ== kg) (kJ== kgK) (kg== ms) (W== mK) Prf (N== m)
0.01 0.612 999.8 0.005 2,501 4,229 1,791 0.561 13.50 0.0757
10 1.228 999.7 0.009 2,477 4,188 1,308 0.580 9.444 0.0742
20 2.339 998.2 0.017 2,453 4,182 1,003 0.598 7.010 0.0727
30 4.246 995.6 0.030 2,430 4,182 798 0.615 5.423 0.0712
40 7.381 992.2 0.051 2,406 4,183 653 0.631 4.332 0.0696
50 12.34 988.0 0.083 2,382 4,181 547.1 0.644 3.555 0.0680
60 19.93 983.2 0.130 2,358 4,183 466.8 0.654 2.984 0.0662
70 31.18 977.8 0.198 2,333 4,187 404.5 0.663 2.554 0.0645
80 47.37 971.8 0.293 2,308 4,196 355.0 0.670 2.223 0.0627
90 70.12 965.3 0.423 2,283 4,205 315.1 0.675 1.962 0.0608
100 101.3 958.4 0.597 2,257 4,217 282.3 0.679 1.753 0.0589
110 143.2 951.0 0.826 2,230 4,233 255.1 0.682 1.584 0.0570
120 198.5 943.2 1.121 2,202 4,249 232.2 0.683 1.444 0.0550
130 270.0 934.9 1.495 2,174 4,267 212.8 0.684 1.328 0.0529
140 361.2 926.2 1.965 2,145 4,288 196.3 0.683 1.232 0.0509
150 475.7 917.1 2.545 2,114 4,314 182.0 0.682 1.151 0.0488
160 617.7 907.5 3.256 2,082 4,338 169.6 0.680 1.082 0.0466
170 791.5 897.5 4.118 2,049 4,368 158.9 0.677 1.025 0.0444
180 1,001.9 887.1 5.154 2,015 4,404 149.4 0.673 0.977 0.0422
190 1,254.2 876.2 6.390 1,978 4,444 141.0 0.669 0.937 0.0400
200 1,553.7 864.7 7.854 1,940 4,489 133.6 0.663 0.904 0.0377
220 2,317.8 840.3 11.61 1,858 4,602 121.0 0.650 0.857 0.0331
240 3,344.7 813.5 16.74 1,766 4,759 110.5 0.632 0.832 0.0284
260 4,689.5 783.8 23.70 1,662 4,971 101.5 0.609 0.828 0.0237
280 6,413.2 750.5 33.15 1,543 5,279 93.4 0.581 0.848 0.0190
300 8,583.8 712.4 46.15 1,405 5,751 85.8 0.548 0.901 0.0144
320 11,279 667.4 64.6 1,239 6,536 78.4 0.509 1.006 0.0099
340 14,594 610.8 92.7 1,028 8,241 70.3 0.469 1.236 0.0056
360 18,655 528.1 143.7 721 14,686 60.2 0.428 2.068 0.0019
373 21,799 402.4 242.7 276 21,828 46.7 0.545 18.69 0.0001
Source: Data reprinted with permission from R.C. Weast, Ed., 1970. CRC Handbook of Tables for Applied Engineering
Science, Tables 1-7, 1-8, Boca Raton: CRC Press=Taylor & Francis.
Appendices 489
TABLE F.3
Properties of Liquid Metals (101 kPa)
Thermal Dynamic
Metal Melting T T Density, ρ Specific Heat, Conductivity, k Viscosity, μ
Point, ( C) ( F) ( C) (kg== m3) cv (kJ== kgK) (W== mK) (kg== ms)
Surface Material
White surfaces: paint, paper, plaster, plastics, fresh snow 0.1–0.3
Surface Material
Highly polished (white) metals, gold, yellow brass 0.02–0.08 0.1–0.4
Clean (dark) metals 0.1–0.35 0.3–0.6
Metallic-pigment paints 0.35–0.55 0.4–0.6
White, non-metal surfaces 0.7–0.9 0.1–0.35
Dark-colored non-metals 0.7–0.9 0.45–0.8
Black paint, asphalt, carbon, water 0.85–0.95 0.7–0.9
Diffuser or Enclosure
Thin quartz or silica 0.90
Clear glass or plastic (1=8 in.) 0.90
Ground or frosted glass 0.75
Opal-white glass 0.50
Heat-absorbing plate glass (1=4 in.) 0.60
Source: Data reprinted with permission from R.C. Weast, Ed., 1970. CRC Handbook of Tables for Applied Engineering
Science, Tables 2-8, 2-9, 2-10, 2-16, 2-17, Boca Raton: CRC Press=Taylor & Francis.
492 Appendices
TABLE G.2
Total Emissivity of Carbon Dioxide and Water Vapor
P L (ft atm)
(total pressure 500 R 1,000 R 1,500 R 2,000 R 2,500 R 3,000 R
1 atm) (4 C) (282 C) (560 C) (838 C) (1116 C) (1393 C)
CO2
0.01 0.03 0.03 0.03 0.03 0.02 0.01
0.02 0.04 0.04 0.04 0.04 0.03 0.02
0.05 0.06 0.06 0.06 0.06 0.05 0.04
0.10 0.08 0.07 0.07 0.07 0.06 0.05
0.20 0.10 0.09 0.09 0.09 0.08 0.07
0.50 0.13 0.12 0.12 0.12 0.11 0.09
1.0 0.15 0.14 0.14 0.14 0.13 0.12
2.0 0.17 0.16 0.16 0.16 0.16 0.15
5.0 0.20 0.19 0.19 0.19 0.19 0.18
H2O
0.02 0.06 0.04 0.02 0.02 0.01 –
0.05 0.10 0.07 0.06 0.04 0.03 0.02
0.10 0.15 0.11 0.08 0.06 0.05 0.04
0.20 0.20 0.16 0.12 0.10 0.08 0.06
0.50 0.30 0.24 0.20 0.16 0.14 0.12
1.0 0.37 0.32 0.27 0.23 0.19 0.16
2.0 0.44 0.39 0.35 0.30 0.26 0.22
5.0 0.55 0.50 0.47 0.41 0.36 0.31
Source: Data reprinted with permission from R.C. Weast, Ed., 1970. CRC Handbook of Tables for Applied Engineering
Science, Table 2-11, Boca Raton: CRC Press=Taylor & Francis.
TABLE G.3
Total Emissivities of Gases (Dependence on Partial Pressure)
PL PL PL PL PL
Gas T (R) T (K) (0.01 atm ft) (0.05 atm ft) (0.1 atm ft) (0.5 atm ft) (1.0 atm ft)
TABLE I.1
Selected Thermodynamic Properties at STP (101 kPa, 25◦ C)
Compound o (kJ== mol)
h gof (kJ== mol) so (J== K mol) cop (J== K mol)
f
TABLE I.2
Heats of Combustion
o
o
Δh c Δh c
Molecular Formula Name (J== K mol) Molecular Formula Name (J== K mol)
References
G.F. Hewitt, G.L. Shires, and Y.V. Polezhaev, Eds. 1997. International Encyclopedia of Heat and Mass
Transfer, Boca Raton: CRC Press/Taylor & Francis.
M. Kaufman. 2002. Principles of Thermodynamics, Boca Raton: CRC Press/Taylor & Francis.
M. Kenisarin and K. Mahkamov. 2016. “Salt Hydrates as Latent Heat Storage Materials: Thermophys-
ical Properties and Costs,” Solar Energy Materials and Solar Cells, Tables 1, 22, vol. 145, 255–286.
D.R. Stull and H. Prophet. 1971. JANAF Thermochemical Tables, 2nd edition., U.S. National Bureau of
Standards, Office of Standard Reference Data, Washington, DC.
R.C. Weast, Ed. 1970. CRC Handbook of Tables for Applied Engineering Science, Boca Raton: CRC Press/
Taylor & Francis.
Index
497
498 Index
boundary layer flow on vertical flat plate, chemically reacting flows, 347–384
113–118 combustion reaction, 358–360
boundary layer thickness, 13, 83, 91–93, 96–98, energy balance of reacting flows, 355–358
100, 114, 116–117, 122, 130, 291 fluidized beds, 373
boundary node mapping, 439 heat and mass transfer, 377–378
boundary value problems, 47 hydrodynamics, 373–377
Boussinesq approximation, 114, 456 noncatalytic gas–solid reaction model,
Brownian diffusion, 286–287 381–384
bubble diameter, 217, 376 reaction rate equations for solid
bubble flow, 205–206, 209, 213–217 conversion, 378–381
bubble rise velocity, 216, 376 gas–gas reactions, 371–372
Buckingham equation, 286 gas–liquid reactions, 368–371
Buckingham Pi theorem, 82–83, 198–199, 377 gas–solid reacting mixtures, 360–368
buffer volume, 237 energy balance and heat transfer, 366–368
bulk phase temperature, 363 progressive conversion model, 363–366
bulk velocity, 264 shrinking core model, 360–363
Buongiorno correlation, 290 material balance for chemical reactors, 352
buoyancy, 1, 6, 13, 17, 67–68, 71, 78–81, 111, batch reactor, 353
113–114, 118, 132, 195, 197–198, 217, continuous stirred tank reactor, 353
223, 236, 308, 449, 456 general mole balance equation, 352–353
force, 6, 67–68, 78–80, 113–114, 195, 217, 449 packed bed reactor, 354–355
“burnout” condition, 240 plug flow reactor, 354
mixture properties, 348–349
reaction rates, 349–352
C chemical potential, 302, 310
chemical reactions, 4, 68, 77, 133, 264,
C2H4 (ethylene), 2
347–348, 350–352, 355–368,
calcium carbonate (CaCO3), 364
371–372, 376–379, 449
calculus, 75, 133, 137, 261, 315, 430–431, 433, 447
chemical reactor, 199, 264, 349–350, 352–353
caloric, 1
Chilton–Colburn analogy, 88–89, 403
CaO (solid) crystals, nucleation of, 364
cap bubbles, 206 chlorofluorocarbons (CFCs), 153
capillary action, 68, 237–239 Churchill and Chu correlation, 116
capillary force, 237–238, 259 churn flow, 205–206, 214
capillary pump, 238 churn turbulent bubble flow, 206
carbon dioxide, 90, 152, 280–281, 357, 364–365 Clausius–Clapeyron equation, 207
carpenter–colburn correlation, 234 cloud cover, 179–181
carrier phase equations, 257, 260–263 clustered bubble flow, 206
Cartesian coordinates, 27, 37–44, 77, 104 CO2 (gas) formation, 364
systems, 39–42 coefficient of friction, 81
casting, 1, 292, 299, 306, 319–320, 343 coefficient of thermal expansion, 10
catalysts, 264, 347, 350 coefficient β determination, 337
catalytic reactions, 347, 383–384 coflowing droplets, 267
catalytic reactors, 347 cohesive force, 89, 259
cell spacing, 459–460 Colburn factor, 88, 403
central difference scheme (CDS), 448, 453 Colburn j factors, 81, 88
central standard time (CST), 177 cold fluid stream, 394–395, 397
centrifugal casting, 319 cold stream specific heat, 395
centrifugal fields, 71 collection efficiency, 269
ceramics, 2–3, 152 collector efficiency and heat losses, 182–185
CFD boundary conditions, examples, 445 collector efficiency factor, 187, 189, 193
CH4 (methane), 2 collector plate, 182, 184–186, 189, 259
chemical equilibrium constant, 363 absorptivity, 183, 189
chemical leaks, 412 collector surface area, 183
Index 499
higher nonlinearity, 409 interface velocity, 305, 308, 310, 317–319, 325, 328
high-speed compressible flows, 81 interfacial entropy balance, 311
high-velocity jets, 370 interfacial entropy production, 311
Hilpert’s correlation, 267 interfacial heat balance, 322
homogeneous flow, 256 interfacial mass flux, 261
regime, 284 interfacial roughness relationship, 211
homogeneous fluidization, 373 interfacial shear parameter, 230
homogeneous model, 215 intermolecular energy exchange, 18
homogeneous nucleation, 301 internal energy, 357
homogeneous systems, 347 equation, 76–78
homogeneous turbulence, 126 internal flow, 67, 103–113
horizontal radiation, 180 in tubes, 231–234
hour angle, 176–180 internal horizontal two-phase flows, 213–222
human eye, 153 internal resistance, 54
hydraulic diameter, 239 interphase effects, 449
hydraulic radius, 264 interphase interaction forces, 262
hydrocarbons, 2, 182, 234, 278, 347, 358–360 interpolation, 430, 435
hydrodynamics, 373–377 interstitial atoms
hydrostatic pressure component, 78 mass diffusion flow of, 18
hyperbolic relation, 71 net flux of, 18
hypereutectoid, 300 interstitial diffusion, 299
hypoeutectoid, 300 inverse annular flow, 209
inverse mapping, 50–51
inviscid fluid, 73
I inviscid freestream, 92
ionic bonds, 9
ice accretion, 2, 271, 276
irradiation, 151
ice density, 269
irreversible process, 6, 133
ice thickness, 269, 271, 327
isobaric (constant pressure) process, 7
ideal gas law, 77, 89, 357, 371–372
isogonal mapping, 47
ideally separated bubble flow, 206
isomerization, 349
impinging droplets, 267
isotherm gradient model, 309
on freezing surface, 269–271
isothermal (constant temperature) process, 7
implicit formulation, 449
isothermal conditions, 361
implicit method, 421, 441, 449
isothermal confocal ellipses, 52
incident radiation, 14, 151, 155–156, 161, 167,
isothermal object in flow stream, 78
172–173, 180, 182–183, 188, 282
isothermal wall boundary condition, 107
incident solar flux, 180–182
isotherms, 7, 24, 30–31, 52, 101, 107, 112–113, 122,
incident solar radiation, 21, 160, 179–182
166, 174, 256, 281, 309, 360–361, 363, 447
inclined solar collector, 178–179
isotropic turbulence, 126
incomplete combustion, 358
induction period, 275
influence coefficient matrices, 456–457 J
inlet conditions, 445
instantaneous collector efficiency, 183 Jacob number, 81, 208, 226, 406, 409
insulation layer, 28 Jacobian determinant, 41, 431, 433, 455, 458
integral analysis, 91–93 Joukowski mapping, 48
integral solution, 315–317
integration point, 446, 454
K
equations method, 455
interacting bubble flow, 206 Kandlikar correlation, coefficients, 222
interface structure, 303–305 Kelvin–Helmholtz instabilities, 214
interface tracking method, 299 kinematic component, 78
interface tracking of a free surface flow, 460–461 kinematic properties, 7–8
506 Index